Search for the Standard Model Higgs boson in the diphoton decay channel with 4.9 fb-1 of pp collision data at √s = 7 TeV with ATLAS

A search for the standard model Higgs boson is performed in the diphoton decay channel. The data used correspond to an integrated luminosity of 4 : 9 fb (cid:1) 1 collected with the ATLAS detector at the Large Hadron Collider in proton-proton collisions at a center-of-mass energy of ﬃﬃﬃ s p ¼ 7 TeV . In the diphoton mass range 110–150 GeV, the largest excess with respect to the background-only hypothesis is observed at 126.5 GeV, with a local signiﬁcance of 2.8 standard deviations. Taking the look-elsewhere effect into account in the range 110–150 GeV, this signiﬁcance becomes 1.5 standard deviations. The standard model Higgs boson is excluded at 95% conﬁdence level in the mass ranges of 113–115 GeVand

Search for the Standard Model Higgs Boson in the Diphoton Decay Channel with 4:9 fb À1 of pp Collision Data at ffiffi ffi s p ¼ 7 TeV with ATLAS G. Aad et al. * (ATLAS Collaboration) (Received 7 February 2012;published 13 March 2012) A search for the standard model Higgs boson is performed in the diphoton decay channel. The data used correspond to an integrated luminosity of 4:9 fb À1 collected with the ATLAS detector at the Large Hadron Collider in proton-proton collisions at a center-of-mass energy of ffiffi ffi s p ¼ 7 TeV. In the diphoton mass range 110-150 GeV, the largest excess with respect to the background-only hypothesis is observed at 126.5 GeV, with a local significance of 2.8 standard deviations. Taking the look-elsewhere effect into account in the range 110-150 GeV, this significance becomes 1.5 standard deviations. The standard model Higgs boson is excluded at 95% confidence level in the mass ranges of 113-115 GeV and 134.5-136 GeV. DOI: 10.1103/PhysRevLett.108.111803 PACS numbers: 14.80.Bn, 12.15.Ji, 14.70.Bh The Higgs mechanism [1] is one of the best-motivated processes to explain electroweak (EW) symmetry breaking. In the standard model (SM), this mechanism explains the generation of the W and Z boson masses and predicts the existence of the only elementary scalar in the SM, the hypothetical Higgs boson. Prior direct searches at LEP, Tevatron and LHC exclude the SM Higgs boson with a mass m H < 114:4 GeV and 145 < m H < 206 GeV at 95% confidence level (C.L.) [2][3][4]. The present search for H ! uses the full 2011 data sample collected by ATLAS at 7 TeV center-of-mass energy and updates prior results with 1:08 fb À1 [5].
The ATLAS detector [6] consists of an inner tracking detector surrounded by a superconducting solenoid providing a 2 T magnetic field, electromagnetic and hadron calorimeters, and a muon spectrometer. The main subdetectors relevant to the search presented here are the calorimeters, in particular, the electromagnetic section, and the inner tracking system. The inner detector provides tracking in the pseudorapidity region jj < 2:5 and consists of silicon pixel-and microstrip detectors inside a transition radiation tracker. The electromagnetic calorimeter, a lead liquid-argon sampling device, is divided in one barrel (jj < 1:475) and two end-cap (1:375 < jj < 3:2) sections. The barrel (jj < 0:8) and extended barrel (0:8 < jj < 1:7) hadron calorimeter sections consist of steel and scintillating tiles, while the end-cap sections (1:5 < jj < 3:2) are composed of copper and liquid argon.
The data were recorded using a diphoton trigger [7], each photon having a transverse energy, E T , of at least 20 GeV, seeded by a lower-level trigger that required two clusters in the electromagnetic calorimeter with E T > 12 or 14 GeV, depending on the data-taking period. The trigger efficiency for the signal events passing the final offline selection is 99%. After applying data quality requirements, the total integrated luminosity of the data set used in this analysis is 4:9 AE 0:2 fb À1 [8].
Events are required to contain at least one vertex with at least three associated tracks, where the transverse momentum, p T , of each track is required to be larger than 0.4 GeV, as well as two photon candidates each seeded by an energy cluster in the electromagnetic calorimeter with E T > 2:5 GeV. Photons that convert to electron-positron pairs in the inner detector leave one or two tracks that are reconstructed and matched to the clusters in the calorimeter. The photon energy is calibrated separately for converted and unconverted photon candidates using Monte Carlo (MC) simulations of the detector [9]. A correction, depending on pseudorapidity and typically of the order of AE1%, is applied to the calibrated photon energy as obtained from studies using Z ! ee decays in data [10]. Photons are reconstructed in the fiducial region jj < 2:37, excluding the calorimeter barrel-to-end-cap transition regions 1:37 < jj < 1:52. The photon candidates are ordered in E T and the leading (subleading) candidate is required to have E T > 40 GeV (25 GeV). Both candidates are required to pass further identification criteria based on shower shapes measured in the electromagnetic calorimeter and on the energy leakage into the hadron calorimeter [11]. The photon reconstruction and identification efficiency ranges typically from 65% to 95% for E T in the range 25 to 80 GeV. The two photon candidates are required to be isolated by having at most 5 GeV energy deposited in the calorimeters in a cone of ÁR ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi ðÁÞ 2 þ ðÁÞ 2 p ¼ 0:4 around the candidate, where is the azimuthal angle, after subtracting the energy assigned to the photon itself. The measured isolation [11] is corrected for lateral shower leakage and ambient energy from *Full author list given at the end of the article.
Published by the American Physical Society under the terms of the Creative Commons Attribution 3.0 License. Further distribution of this work must maintain attribution to the author(s) and the published article's title, journal citation, and DOI. multiple proton-proton interactions (pileup), following the method in Ref. [12]. The isolation cut retains $87% of Higgs boson signal events with m H ¼ 120 GeV while rejecting $44% of the selected data, which includes jets that can be misidentified as photons.
The opening angle of the two photons, used in the calculation of their invariant mass, is determined using the trajectories of the photons. For a converted photon with a well-measured conversion vertex, the trajectory is determined from the straight line between the barycenter of the associated energy deposits in the calorimeter and the conversion vertex. Otherwise, the trajectory is determined from the barycenters of the showers in the first and second layers of the calorimeter. The extrapolation of the trajectories as well as the average beam spot position are used to determine the origin of the photons along the beam axis, z. The resolution of the z vertex coordinate is $6 mm on average for two converted photons with reconstructed tracks, and $15 mm otherwise. The contribution of the resulting angular resolution to the mass resolution is negligible in comparison to that of the energy resolution.
In total, 22 489 events pass the selection in the diphoton mass range 100-160 GeV. To confirm the dominance of the diphoton processes () over backgrounds with one or two misidentified jets (j, jj), the composition of the selected sample is estimated using the data. A sideband technique [5] is used to estimate the numbers of , j, or jj events. The fraction of true diphoton events is estimated to be ð71 AE 5Þ%. The amount of Drell-Yan background is estimated by selecting Z ! ee decays in data where either one or both electrons pass the photon selection. The measured composition is summarized in Table I and is compatible with MC expectations. This decomposition is not directly used in the signal search; however, it is used to validate the parametrization of the background fit (see below).
The events are separated into nine mutually exclusive categories with different mass resolutions and signal-tobackground ratios, to increase the sensitivity to a possible Higgs boson signal. Categories are defined by the conversion status, of the selected photons, and p Tt [13], the component of the diphoton pT that is orthogonal to the thrust axis, as proposed in Ref. [14]. Events with two unconverted photons are separated into unconverted central (jj < 0:75 for both candidates) and unconverted rest (all other events). Events with at least one converted photon are separated into converted central (jj < 0:75 for both candidates), converted transition (at least one photon with 1:3 < jj < 1:75), and converted rest (all other events). Excepting the converted transition category, each category is further divided by a cut at p Tt ¼ 40 GeV into two categories, low p Tt and high p Tt . MC studies show that signal events, particularly those produced in vector-boson fusion (VBF) or in associated production (W=ZH and t " tH), have on average larger p Tt than background events. The number of data events in each category is given in Table II.
The distribution of the invariant mass of the diphoton events, m , summed over all categories, is shown in Fig. 1. The sum of the background-only fits (described below) to the invariant mass in each of the categories is superimposed. The signal expectation for a SM Higgs boson with m H ¼ 120 GeV is also shown. The presence of the Higgs boson will appear as a narrow resonance in the  ) and FWHM (both in GeV), expected number of signal events (N S ) for m H ¼ 120 GeV, and number of events in the data (N D ) in each category for 4:9 fb À1 , N S and N D are for the mass range 100-160 GeV. The signal-tobackground ratios (S=B) are given in a mass window containing 90% of the signal for m H ¼ 120 GeV. 111803-2 invariant mass of the selected photon pairs superimposed on a smoothly falling background. The residual of the data with respect to the total background as a function of m is also shown in Fig. 1.
Higgs boson production and decay are simulated with several MC samples that are passed through a full detector simulation [15] using GEANT4 [16]. Pileup effects are simulated by overlaying each MC event with a variable number of MC inelastic proton-proton collisions [17]. POWHEG [18], interfaced to PYTHIA [19] for showering and hadronization, is used for generation of gluon-fusion and VBF production. PYTHIA is used to generate the Higgs boson production in association with W=Z and t " t. The Higgs boson production cross sections are computed up to next-to-next-to-leading order (NNLO) [20] in QCD for the gluon-fusion process. In addition, QCD softgluon resummations up to next-to-next-to-leading log improve the NNLO calculation [21]. The next-to-leading order (NLO) EW corrections are applied [22]. These results are compiled in Refs. [23] assuming factorization between QCD and EW corrections. The cross sections for the VBF process are calculated with full NLO QCD and EW corrections [24], and approximate NNLO QCD corrections are available [25]. The W=ZH processes are calculated at NLO [26] and at NNLO [27], and NLO EW radiative corrections [28] are applied. The full NLO QCD corrections for t " tH are calculated [29]. The Higgs boson cross sections, branching ratios [30], and their uncertainties are compiled in Ref. [31].
The cross sections multiplied by the branching ratio into two photons are listed in Table III. The number of signal events produced by gluon fusion is rescaled to take into account the expected destructive interference between the gg ! continuum background and the gg ! H ! process [32], leading to a reduction of the production rate by 2-5% depending on m H and analysis category. The fractions of gluon fusion, VBF, WH, ZH, and t " tH production are approximately 87%, 7%, 3%, 2% and 1%, respectively, for m H ¼ 120 GeV.
The shower shape variables of the simulated samples are shifted to agree with the corresponding distributions in the data [11] and the photon energy resolution is broadened to account for differences observed between Z ! ee data and MC events. Events generated with POWHEG at NLO have been reweighted to match the Higgs boson p T distribution predicted by HQT [33]. The signal yields expected for 4:9 fb À1 and selection efficiencies are given in Table III.
The invariant mass shape of the signal in each category is modeled by the sum of a Crystal Ball function [34] describing the core of the distribution with a width CB , and a wide Gaussian with a small amplitude describing the tails of the mass distribution. In Fig. 2, the sum of all signal processes in all categories is shown for a Higgs boson with m H ¼ 120 GeV. The expected full-width-at-halfmaximum (FWHM) is 4.1 GeV and CB is 1.7 GeV. The resolution varies with category (see Table II). The signalto-background ratio (S=B), calculated in a mass window symmetric about the signal maximum and containing 90% of the signal, varies from 0.11 to 0.01 depending on the category and is also shown in Table II.
The background in each category is estimated from the data by fitting the diphoton mass spectrum in the range 100-160 GeV with an exponential function with free slope and normalization parameters. The background curve in Fig. 1 is the sum of these nine contributions. For each category, a single exponential fit satisfactorily describes the mass spectrum. This has been checked using large samples of diphoton events produced by the RESBOS [35] and DIPHOX [36] MC generators.  1 (color online). Invariant mass distribution for the selected data sample, overlaid with the total background (see text). The bottom inset displays the residual of the data with respect to the total background. The Higgs boson expectation for a mass hypothesis of 120 GeV corresponding to the SM cross section is also shown. The difference between the exponential function and the true background will contribute to an excess or a deficit of events over background expectations. In order to take this into account in a conservative way, a term is included in the likelihood function that allows for a signal-like component that is consistent with the background uncertainty. For each category this uncertainty is estimated from MC simulations by the difference between the mass distribution of diphoton events generated with RESBOS and the result of the exponential fit to this distribution. Photon reconstruction and identification efficiencies are taken into account. The MC events are scaled to correspond to 4:9 fb À1 of data. The uncertainty is then the maximal difference between the MC shape and the model integrated in a sliding mass window of 4 GeV, the approximate FWHM of the expected signal. The uncertainties obtained are AEð0:1 À 7:9Þ events depending on the category. Pseudoexperiments are used to check that the sum of , j, and jj events can also be described well by the exponential model. The background uncertainties are further validated by fitting the data with functions that have more degrees of freedom than the single exponential, and comparing the residuals to those obtained with the exponential fit.
The dominant experimental uncertainty on the signal yield is the photon reconstruction and identification efficiency ( AE 11%), which is estimated with data by using electrons from Z and W decays and photons selected from Z ! '' (' ¼ e, ) events. Pileup also affects the identification efficiency and contributes to the uncertainty (AE 4%). Further uncertainties on the signal yield are related to the trigger (AE1%), Higgs boson p T modeling (AE1%), isolation (AE5%), and luminosity (AE3:9%). Uncertainties on the predicted cross sections are due to uncertainties on the QCD renormalization and factorization scales ( þ12 À8 %) and on the parton density functions ( [37] and references therein) and s (AE8%). The total uncertainty on the signal yield is þ20 À17 %. The total uncertainty on the mass resolution is AE14%, dominated by the uncertainty on the energy resolution of the calorimeter, determined from Z ! ee events (AE12%). Further uncertainties on the mass resolution result from an imperfect knowledge of material in front of the calorimeter affecting the extrapolation from electron to photon calibration (AE6%), the impact of pileup (AE3%) estimated from events taken with random triggers, and the photon angle measurement (AE1%) estimated using Z ! ee events. The uncertainty on the knowledge of the material in front of the calorimeter is used to derive the amount of event migration between the converted and unconverted categories (AE 4:5%). Different parton density functions and scale variations in HQT calculations are used to derive possible event migration between high and low p Tt categories (AE 8%).
A modified frequentist approach (CL S ) [38] for setting limits and a frequentist approach to calculate the p 0 value are used [39]. The p 0 is the probability that the background fluctuates to the observed number of events or higher. The combined likelihood, which is a function of the ratio of the measured cross section relative to that of the SM prediction, is constructed from the unbinned likelihood functions of the nine categories. Systematic uncertainties are incorporated by introducing nuisance parameters with constraints. Asymptotic formulae [40] are used to derive the limits and p 0 values, which are refined with pseudoexperiments [41], as functions of the hypothetical Higgs boson mass.
The  background-only hypothesis in the mass range 110-150 GeV is observed at 126.5 GeV with a local significance of 2.9 standard deviations. The uncertainty on the mass position ( AE 0:7 GeV) due to the imperfect knowledge of the photon energy scale has a small effect on the significance. When this uncertainty is taken into account, the significance is 2.8 standard deviations; this becomes 1.5 standard deviations when the look-elsewhere effect [42] for the mass range 110-150 GeV is included. The median expected upper limits of the cross section in the absence of a true signal, at the 95% C.L., vary between 1.6 and 1.7 times the SM cross section in the mass range 115-130 GeV, and between 1.6 and 2.7 in the mass range 110-150 GeV. The observed 95% C.L. upper limit of the cross section relative to the SM cross section is between 0.83 and 3.6 over the full mass range. A SM Higgs boson is excluded at 95% C.L. in the mass ranges of 113-115 GeV and 134.5-136 GeV. These results are combined with SM Higgs searches in other decay channels in Ref. [41].
We thank CERN for the very successful operation of the LHC, as well as the support staff from our institutions without whom ATLAS could not be operated efficiently.