UvA-DARE (Digital Search for the Standard Model Higgs boson decay to μ+μ− with the ATLAS detector

A search is reported for Higgs boson decay to μ + μ − using data with an integrated luminosity of 24 . 8 fb − 1 collected with the ATLAS detector in pp collisions at √ s = 7 and 8 TeV at the CERN Large Hadron Collider. The observed dimuon invariant mass distribution is consistent with the Standard Model background-only hypothesis in the 120–150 GeV search range. For a Higgs boson with a mass of 125.5 GeV, the observed (expected) upper limit at the 95% conﬁdence level is 7.0 (7.2) times the Standard Model expectation. This corresponds to an upper limit on the branching ratio BR ( H → μ + μ − ) of 1 . 5 × 10 − 3 . (http://creativecommons.org/licenses/by/3.0/). by SCOAP 3 .


Introduction
The Standard Model (SM) describes a wide range of particle physics phenomena to a high degree of precision. In the SM, the Brout-Englert-Higgs (BEH) mechanism [1][2][3] spontaneously breaks the electroweak (EW) gauge symmetry and generates masses for the W and Z gauge bosons as well as for the charged fermions via Yukawa couplings [4][5][6]. In searches for the Higgs boson predicted by the BEH mechanism, the ATLAS and CMS Collaborations have discovered a new particle, via decays into gauge bosons [7,8], with a mass of approximately 125.5 GeV and measured properties consistent with those predicted by the SM [9][10][11][12].
Higgs boson decays to bb, τ + τ − and μ + μ − can be measured at the LHC with their SM branching ratios proportional to the squares of the fermion masses. The SM branching ratio for the H → μ + μ − decay is 21.9 × 10 −5 for a Higgs boson mass (m H ) of 125 GeV [13,14]. The H → μ + μ − decay has a clean final state signature that allows a measurement of the Higgs boson coupling to second-generation fermions. The dominant irreducible background is the Z /γ * → μ + μ − process, which has an approximately three orders of magnitude higher production rate compared to that of the expected signal.
In this Letter a search for the H → μ + μ − decay of the SM Higgs boson is presented. This search for the presence of a narrow H → μ + μ − resonance, with a signal width determined by experimental resolution, is performed by fitting the invariant mass distribution in the 110-160 GeV region. This range allows determining background shape and normalisation and setting a limit E-mail address: atlas.publications@cern.ch. on the dimuon decay of the SM Higgs boson with a mass of 125.5 GeV. Section 2 gives a description of the experimental setup and summarises the data sample and Monte Carlo (MC) simulation samples used to model the signal process and to develop an analytical model for the background processes. Sections 3 and 4 describe the event selections and categorisation. Analytical models used to describe invariant mass distributions for signal and background processes are discussed in Section 5, and systematic uncertainties are detailed in Section 6. The results are presented in Section 7.

Experimental setup, data and simulated samples
This search is performed on the data sample recorded in 2011 and 2012 by the ATLAS detector in pp collisions at the LHC at √ s = 7 and 8 TeV, respectively. ATLAS [15] is a general-purpose particle detector with a cylindrical geometry and consists of several subdetectors surrounding the interaction point and covering almost the full solid angle. 1 The trajectories and momenta of charged particles are measured within the pseudorapidity range of |η| < 2.5 by multi-layer silicon pixel and microstrip detectors as well as a transition radiation tracker. The tracking system is immersed in a 2 T magnetic field produced by a superconducting solenoid, and is surrounded by a high-granularity 1 ATLAS uses a right-handed coordinate system with its origin at the nominal T of tracks is considered to be the primary vertex. Muon candidates [53] are reconstructed by matching tracks in the inner detector to tracks reconstructed in the muon spectrometer. In addition to stringent track quality requirements imposed for muon identification, the muon tracks must be consistent with having originated from the primary vertex. All selected muon candidates are required to be within |η| < 2.5. Muon candidates must pass track and calorimeter isolation requirements that scale with the p T of the muon track. The isolation is calculated as the scalar sum of the p T of additional tracks or the E T of calorimeter energy deposits within cone of Jets are reconstructed from clusters of calorimeter cells using the anti-k t algorithm [54,55] with a radius parameter of 0.4.
The selected jets must satisfy E T > 25 GeV for |η| < 2.4 and E T > 30 GeV for 2.4 ≤ |η| < 4.5. Muon candidates overlapping with the selected jets within a cone of radius R = 0.4 are removed from the analysis. In the pseudorapidity range |η| < 2.5, jets originating from b-quarks are identified using a b-tagging algorithm [56,57] with an efficiency of approximately 80%, determined from tt MC events, and with a misidentification rate for selecting light-quark or gluon jets of less than 1%. The missing transverse momentum [58], E miss T , is the magnitude of the vector sum of the p T of muons, electrons, photons, jets and clusters of calorimeter cells with |η| < 4.9 not associated with these objects.
Corrections are applied to simulated MC samples in order to account for differences between data and MC simulation for the trigger and identification efficiency and for the muon momentum scale and resolution. The trigger and reconstruction efficiency corrections are measured using Z → μ + μ − events and are found to be within 2% of unity. The muon momentum corrections are determined by comparing the reconstructed invariant mass distribution of Z → μ + μ − events in data with that from simulated events; these corrections are within 0.1% of unity.    Table 1. The MC background yields are given to illustrate the expected background composition. The selection efficiency times acceptance for signal events with m H = 125 GeV after all selection criteria described thus far is approximately 55%.
The expected background processes produce smooth m μ + μ − distributions in the search window, allowing the total background normalisation and shape in each category to be derived from fitting the data as described in Section 5. The m μ + μ − distribution is examined in the range 110-160 GeV. This range is larger than the 120-150 GeV search window in order to account for signal resolution effects and to allow sufficient sidebands for background normalisation.

Event categorisation
To increase sensitivity to the Higgs boson signal, the selected events are separated into seven mutually exclusive categories with different signal-to-background ratios based on their muon pseudorapidity (η μ ), p μ + μ − T , and VBF dijet signature. Events produced in the VBF process are characterised by two forward jets with little hadronic activity between them. The VBF category is thus defined by requiring the events to have at least two jets with an invariant mass greater than 500 GeV, |η jet 1 − η jet 2 | > 3 and η jet 1 × η jet 2 < 0.
In events with more than two jets, those with the highest p T are used in the selection. Events with at least one jet identified as originating from a b-quark are excluded from the VBF category.
The events that are not selected for the VBF category are classified using p Signal events have on average larger values of p μ + μ − T than the Z /γ * background events. Therefore, the remaining    and high (> 50 GeV). To further improve the search sensitivity, each of these three categories is also subdivided into a central category with |η μ 1 | < 1 and |η μ 2 | < 1 and a non-central category containing all remaining events. This value for the η μ boundary has been chosen by scanning a range of η μ values and selecting a value with the highest signal sensitivity. The muon momentum measurement for the central muons is more precise, producing a narrower m μ + μ − distribution for signal events in the central category and thus resulting in a higher overall signal sensitivity. Table 2 shows the signal event yields, N S / √ N B ratios, approximate signal width and results of the fits to the data, described in Section 5, for all analysis categories.

Signal and background models
Analytical models are used to describe the m μ + μ − distributions for signal and background processes. The simulated samples detailed in Section 2 are used to develop background models which are designed to describe essential features of the background m μ + μ − distributions, dominated by Z /γ * , while having sufficient flexibility to describe different categories and to absorb potential differences between data and MC simulation.
The background model selected to describe the m μ + μ − distribution for the p μ + μ − T categories is the sum of a Breit-Wigner (BW) function convolved with a Gaussian function (GS), and an exponential function divided by x 3 : The background model for the VBF category is the product of a Breit-Wigner and an exponential function: For all categories, the BW parameters are fixed to M BW = 91.2 GeV and Γ BW = 2.49 GeV. The parameters f , A and the overall background normalisation are determined from fits to the data, as shown in Fig. 2 for the central medium p μ + μ − T category. Similar fit quality is observed for all other categories.
The signal model is obtained from simulated Higgs boson signal samples, where contributions from the ggF, VBF and VH Higgs boson production processes are added together. This model is the sum of a Crystal Ball (CB) 2 and a Gaussian function: where x represents m μ + μ − and f CB represents the fraction of the CB contribution when each individual component is normalised to unity. The parameters α and n define the power-law tail of the CB distribution. The parameters σ CB and σ S GS denote the widths of the CB and GS distributions, respectively. The parameters m, σ CB and σ S GS are determined from the fits to the simulated Higgs boson samples. In order to improve stability of the fits, the remaining parameters f CB , α and n are fixed to values determined from empirical tests where a range of possible values have been tested. Fig. 3 shows how the signal model reproduces the simulation for the medium p μ + μ − T category for the expected signal dimuon mass distributions. Similar fit quality is obtained for all other categories. The signal model parameters are linearly interpolated in steps of 1 GeV between the generated signal samples.
To derive the results presented in Section 7, a binned maximum likelihood fit to the observed m μ + μ − distributions in the range 110-160 GeV is performed using the sum of the signal and background model. The fit is done simultaneously in all seven categories with separate distributions for 7 TeV and 8 TeV data. Free fit parameters include the background model fit parameters described earlier and an overall background normalisation in each category. The signal model parameters are fixed in the fit to data except for the H → μ + μ − signal strength μ S defined such that μ S = 0 corresponds to the background-only hypothesis and μ S = 1 corresponds to the SM H → μ + μ − signal hypothesis. 2 A Gaussian function with a power-law tail.

Table 3
Main sources of experimental and theoretical uncertainty on the signal yield, excepting the error from mismodelling bias. "QCD scale" indicates the theoretical uncertainty on the Higgs boson production due to missing higher-order corrections estimated by varying the QCD renormalisation and factorisation scales, while "PDFs + α s " indicates uncertainty due to parton distribution functions, as described in Refs. [13,14]. The ranges for the uncertainties cover the variations among different categories and data-taking periods.

Systematic uncertainties
The main theoretical and experimental sources of uncertainty on the number of expected signal events are shown in Table 3.
The uncertainty on the integrated luminosity is ±1.8% for 7 TeV data [59] and ±2.8% for 8 TeV data; it is obtained following the same methodology as that detailed in Ref.
[59], from a preliminary calibration of the luminosity scale derived from beam-separation scans performed in November 2012.
Sources of experimental uncertainty include the efficiency of the muon trigger, reconstruction, identification, and isolation requirements, as well as the muon momentum scale and resolution. Uncertainties on the jet energy scale and resolution affect the selection of jets used in the VBF category definitions. Smaller uncertainties arise from pile-up and the primary vertex selection. The total experimental uncertainty on the predicted signal yield is a The theoretical uncertainties on the production and H → μ + μ − decay of a SM Higgs boson of mass m H = 125 GeV are taken from Refs. [13,14]. The uncertainty on the relative populations of the p μ + μ − T categories, due to the uncertainty on the description of the Higgs boson p T spectrum arising from missing higher-order corrections, is determined by varying the QCD renormalisation, factorisation and resummation scales used in the HRES program. To evaluate these uncertainties, the scales are independently varied up and down by a factor of two while keeping their ratio between 0.5 and 2.0. The ggF contribution to the VBF category has large uncertainties due to missing higher-order corrections; they are estimated using the method described in Ref. [26]. The uncertainties associated with the modelling of multi-parton interactions (MPI) are estimated by turning off the MPI modelling in the event generation, according to the recommendations in Ref. [26].
In addition to the samples described in Section 2, samples of the dominant Z /γ * background are generated with Powheg + Pythia8 and parameterised with a detector response measured using simulated MC events. These samples contain approximately 170 times more events than expected in the data and are used to validate the background models and to derive systematic uncertainties due to potential mismodelling bias. This bias is estimated by fitting the parametrised signal plus background model to the simulated m μ + μ − background distribution in the mass range 110-160 GeV where the signal strength μ S is a free parameter.
The bias is then defined as the root mean square of the signal yield obtained from the fit for Higgs boson masses in the range 120-150 GeV. This uncertainty varies from 3% to 20% of the statistical uncertainty on the signal strength μ S , depending on the selection category and data-taking period.

Results and conclusions
The statistical procedure used to interpret the data is summarised in Ref. [7]. The observed data is consistent with the expected backgrounds and no evidence for a signal is found. Upper limits are computed on the signal strength μ S using a modified frequentist CL s method [60,61] based on a Poisson log-likelihood ratio statistical test.
The observed and expected 95% confidence level (CL) limits on the H → μ + μ − signal strength are shown in Fig. 4.