Search for the Higgs boson in the H->WW(*)->lvlv decay channel in pp collisions at sqrt(s) = 7 TeV with the ATLAS detector

A search for the Higgs boson has been performed in the H->WW->lvlv channel (l=e/mu) with an integrated luminosity of 2.05/fb of pp collisions at sqrt(s) = 7 TeV collected with the ATLAS detector at the Large Hadron Collider. No significant excess of events over the expected background is observed and limits on the Higgs boson production cross section are derived for a Higgs boson mass in the range 110<mH<300 GeV. The observations exclude the presence of a Standard Model Higgs boson with a mass 145<mH<206 GeV at 95% confidence level.

by the ATLAS detector during the LHC run of spring and summer 2011. As described in detail below, the search examines events containing two leptons and up to one jet. The main backgrounds are suppressed by cuts on angular distributions, invariant masses, and b-jet tagging information. The background normalization and composition is estimated in situ using several control samples defined by relaxing or reversing selection cuts. Similar searches were performed by CMS and ATLAS in 36 pb −1 [8] and 35 pb −1 [9], respectively. The ATLAS experiment [10] is a multipurpose particle physics detector with forwardbackward symmetric cylindrical geometry allowing tracks within the pseudorapidity range |η| < 2.5 and energy deposits in calorimeters covering |η| < 4.9 to be recon-structed. It is modeled using GEANT4 [11] and simulated events are reconstructed using the same software that is used to perform the reconstruction on data. The effects of multiple pp interactions ("in-time" pile-up) and residual energy deposits from neighboring bunch crossings ("out-of-time" pile-up) are modeled in the Monte Carlo (MC) samples by superimposing a number of simulated minimum-bias events on the simulated signal and background events. MC samples with different numbers of pile-up interactions are re-weighted to match the conditions observed in the present data: about 6 interactions per bunch crossing, with a 50 ns bunch spacing. The data used in this analysis were recorded during periods when all ATLAS sub-detectors were operating under nominal conditions. The events were triggered [12] by requiring the presence of a high-p T electron or muon in the event.
Electron candidates are selected from clustered energy deposits in the electromagnetic (EM) calorimeter with an associated track reconstructed in the inner detector and are required to satisfy a stringent set of identification cuts [13] with an efficiency of 71% for electrons with transverse momentum E T > 20 GeV and |η| < 2.47. Muons are reconstructed by combining tracks in the inner detector and muon spectrometer. The efficiency of this reconstruction is 92% for muons with p T > 20 GeV and |η| < 2.4. Events are required to have a primary vertex with ≥ 3 tracks with p T > 0.4 GeV. For both electrons and muons, the track associated with the lepton candidate is required to be consistent with having been produced at the event's primary vertex. Leptons are required to be isolated, satisfying stringent cuts on tracks and calorimeter depositions inside a cone ∆R = ∆φ 2 + ∆η 2 < 0.2 around the lepton candidate, where ∆φ and ∆η are the transverse opening angle and pseudorapidity difference between the lepton and the track or energy deposit. The lepton reconstruction efficiencies are evaluated with tag-and-probe methods using Z → ℓℓ, J/ψ → ℓℓ, and W → ℓν events in data [14].
Jets are reconstructed from calibrated clusters using [18] is reconstructed from calibrated energy clusters in the calorimeters and the reconstructed momenta of the muons, which generally deposit only a small fraction of their energy in the calorimeters. The E miss T distribution in the presence of pile-up has been studied, and both E miss T as a function of the number of reconstructed primary vertices and E miss T as a function of the event's position in the bunch train are well-modeled by MC.
Exactly two opposite-sign lepton candidates (e or µ) with p T > 15 GeV for muons or E T > 20 GeV for electrons are required. The leading lepton must have transverse momentum > 25 GeV so the selected events have a high efficiency for the trigger selection.
After the selection of events with two leptons, the significant backgrounds are the Drell-Yan process, tt and single top (tW/tb/tqb), W W , other diboson processes (W Z/ZZ/W γ), and W +jets where a jet is misidentified as a lepton. In addition to data-driven validations of the background estimates discussed later, MC simulations of the signal and backgrounds are studied in detail. The gg → H and qq → qqH processes are modeled using Powheg, with Pythia to handle the parton shower [19], and the gg → H Higgs boson p T spectrum is reweighted to agree with the prediction of Ref. [20]. Pythia is used to model W H/ZH production. Signal MC is generated in steps of 5 GeV for m H below 200 GeV and in steps of 20 GeV for larger masses. Signal expectations for intermediate mass values are obtained by linear interpolation of the signal efficiency. The tt, s-channel single top (tb), and qq/qg → W W/W Z/ZZ processes are generated with MC@NLO, t-channel and W t single top with AcerMC (interfaced to the parton shower algorithm in Pythia), gg → W W with gg2ww interfaced to the parton shower algorithm in Herwig [21], W γ with MadGraph interfaced to Pythia, and W +jets and Z/γ * +jets with Alpgen interfaced to Pythia [22].
If the two leptons have different flavors, their invariant mass (m ℓℓ ) is required to be above 10 GeV. Otherwise, they must satisfy m ℓℓ > 15 GeV and they must lie outside the region with |m ℓℓ − m Z | < 15 GeV to suppress backgrounds from Υ and Z production, respectively.
The quantity E miss T,rel is defined as E miss T if the angle ∆φ between the missing transverse momentum and the transverse momentum of the nearest lepton or jet is greater than π/2, or E miss T sin(∆φ) otherwise. E miss T,rel is less sensitive to the mismeasurement of a single lepton or jet than E miss T . To suppress backgrounds from multijet events and Drell-Yan production, it is required that E miss T,rel > 40 GeV if the two leptons have the same flavor, or E miss T,rel > 25 GeV if they have different flavor. After these requirements, the data are separated into H + 0-jet and H + 1-jet [23] samples based on whether they have zero or exactly one jet. In the H + 0-jet channel, the dilepton system is required to have a large transverse boost, p ℓℓ T > 30 GeV, to suppress backgrounds from Z+jets and continuum W W production. To suppress background from top-quark production, events in the H + 1-jet channel are rejected if the jet is identified as the decay of a b-quark. These candidates are further required to have |p tot T | < 30 GeV, where p tot T is the vector sum of the transverse momenta of the jet, the two leptons, and the E miss T vector. This latter selection suppresses events with significant hadronic activity that recoils against the p tot T system but does not leave high p T jets in the detector. In the H + 1-jet channel, the event is required to pass the Z → τ τ rejection cut used in the H → W W analysis of Ref. [24].
Top and W W backgrounds are suppressed by an upper bound on m ℓℓ . Because the m ℓℓ distribution for the signal depends strongly on m H , the chosen upper bound depends on the Higgs boson mass hypothesis. For m H < 170 GeV, m ℓℓ < 50 GeV is required, while for 170 ≤ m H < 220 GeV, the cut is m ℓℓ < 65 GeV. For m H ≥ 220 GeV, the requirement is 50 < m ℓℓ < 180 GeV.
For m H < 220 GeV, an upper bound is imposed on the azimuthal angle between the two leptons to exploit differences in spin correlations between signal and background: ∆φ ℓℓ < 1.3 for m H < 170 GeV, or ∆φ ℓℓ < 1.8 for m H < 220 GeV. The final requirement uses the transverse mass m T [25] where the subscripts v and i denote the visible and invisible decay products The upper bound on this window reduces the W W and top backgrounds and excludes regions of phase space where interference effects between the signal and the gg → W W background are large [26]. Table I shows the expected and observed event yields after these cuts. As described below, the W +jets background is entirely determined from data, whereas for the other processes the expectations are based on simulation, with Z/γ * + jets, tt, and tW/tb/tqb corrected by scale factors derived from control samples. The uncertainties shown are the sum in quadrature of systematic uncertainties and statistical errors due to the finite number of MC events. Figure 1 shows the distributions of m ℓℓ and ∆φ ℓℓ before the final cut on m ℓℓ , and the distribution of m T after the cut on ∆φ ℓℓ .
The background from W +jets events where one jet is misidentified as a lepton is estimated from data using a control sample where one of the two leptons satisfies a loosened set of identification and isolation criteria but not the full set of criteria normally used. The extrapolation from this control sample to the signal region is extracted from dijet events [27].
The Drell-Yan background is corrected for mismodeling of the distribution of E miss T at high values based on the observed difference between the fraction of events passing the E miss T,rel > 40 GeV selection in data and MC simulation for events with m ℓℓ within 10 GeV of the Z boson mass. The correction factors are all found to be between 0.8 and 0.9, which indicates that the background in the signal region is about 15% less than the MC estimates.
The W W and top backgrounds are normalized by a simultaneous fit to the numbers of observed events in the signal region and several control samples. A sample enriched in W W background is defined by removing the selections on m T and ∆φ ℓℓ and changing the selection on m ℓℓ . For m H < 220 GeV, the cut is changed to m ℓℓ > 80 GeV, while for m H > 220 GeV, the control region is the union of the regions with 15 < m ℓℓ < 50 GeV and m ℓℓ > 180 GeV. This control sample is studied separately for the H + 0-jet channel and the H + 1-jet channel, and the observed yields are consistent with expectations in both cases. The yields in these control regions, shown in Table I, are propagated to the signal region using scale factors computed with MC.
In the H + 0-jet channel, the top-enriched control sample consists of the same preselected sample used in the rest of this analysis: events with two leptons and E miss T,rel . The scale factor used to propagate the tt yield from this sample to the signal region is estimated as the square of the efficiency for one top decay to survive the jet veto (estimated using another control sample, defined by the presence of an additional b-jet), with a correction computed using MC to account for the presence of single top [28]. A sample enriched in top background is defined for the H + 1-jet channel by reversing the b-jet veto and removing the cuts on ∆φ ℓℓ , m ℓℓ , and m T . The extrapolation to the signal region is done using a scale factor computed using MC. The control samples for top in the H + 0-jet and H + 1-jet channels also normalize the top contamination in the corresponding W W control regions. In both cases, the estimated top backgrounds are consistent with the expected yields in Table I. The signal significance and limits on Higgs boson production are derived from a likelihood function that is the product of the Poisson probabilities of each of the lepton flavor and jet multiplicity yields for the signal selections, the W W +0-jet and W W +1-jet control regions, and top control region for the H + 1-jet channel. The normalization of the signal, the W W cross sections for the H +0-jet and H + 1-jet channels, and the top cross section for the H + 1-jet channel are allowed to vary independently; the control regions included in the fit constrain all of these except the signal yield. All other components are normalized to their expectations scaled by nuisance parameters constrained by Gaussian terms that include the systematic uncertainties described below. The results from the control sample measurements for the top background in the H + 0-jet channel and for the W +jets and Drell-Yan backgrounds everywhere are used as the expected values for the corresponding backgrounds in the fit. Since these contributions are small, the control samples themselves are not explicitly modeled in the fit as they are for top in the H + 1-jet channel and for W W everywhere.
The systematic uncertainties include contributions from the 3.7% uncertainty in the luminosity [29], and from theoretical uncertainties, which are -8/+12% and ±8% from the QCD scale and 1% and 4% from the parton density functions, for gg → H and qq → qqH respectively. Additional theoretical uncertainties on the acceptance are assessed as described in Ref. [30]. In particular, the uncertainty in the assignment of events to jet multiplicity bins is included separately as an uncertainty on the cross section of each bin, calculated from the approximate 10% and 20% uncertainties of the inclusive 0-jet and 1-jet cross sections, respectively.
Several sources of measurement uncertainty are taken into account. The uncertainty on the jet energy scale is less than 10% on the global scale including flavor composition effects, with an additional uncertainty of up to 7% due to pile-up [16]. The electron and muon efficiencies are determined from samples of W and Z boson data with uncertainties of 2-5% and 0.3-1%, respectively, depending on |η| and p T . Uncertainties are < 1% and < 0.1%, respectively, on the lepton energy scale and < 0.6% and < 5% on the resolution [14]. The uncertainties on the b-tagging efficiency and mistag rate are 6-15% and up to 21%, respectively [17]. A 13% uncertainty is applied to the energy scale for low-p T depositions in the E miss T measurement. All these sources of detector uncertainty are propagated to the result by varying reconstructed quantities and observing the effect on the expected yields. For the W W background, the total (theoretical and experimental) uncertainty on the ratio of cross sections in the signal and control regions is 7.6% in the H +0-jet channel and 21% in the H +1-jet channel; for the top background in H + 1-jet the total for the extrapolation to the signal region is 38%, and 29% to the W W control region.
No significant excess of events is observed. The largest observed deviation from the expected background is 1.9σ. A 95% CL upper bound is set on the Higgs boson cross section as a function of m H using the CL s formalism [31]. Figure 2 shows the expected and observed limits. Discontinuities occur where the selection changes, since the signal regions there are less statistically correlated between adjacent masses. In the absence of a signal, one would expect to exclude a Standard Model Higgs boson in the range 134 < m H < 200 GeV at the 95% CL. The Higgs boson mass interval excluded by the measurements presented in this Letter, 145 < m H < 206 GeV, is consistent with that expectation. This measurement excludes, at 95% CL, a larger part of the mass range favored by the electroweak fits than previous limits [32].