Higgs Boson Searches at ATLAS

The present article gives an overview of the status of Higgs boson studies in the ATLAS experiment at the LHC, at the time of the DISCRETE 2012 Symposium. Accounts of the latest Higgs boson searches in the H → WW(*) → ℓνℓν, H → bb̄ and H → τ+τ− channels are given, as well as a summary of the investigation into the couplings of the new particle, and an indication of the sensitivity of future Higgs studies at the LHC.


Introduction
The discovery [1,2] of a new particle in the search for the Standard Model Higgs boson at the LHC represents a major milestone in the understanding of the mechanism responsible for electroweak symmetry breaking. After this discovery, it is now crucial to establish if the new particle is indeed the Higgs boson predicted by the Standard Model (SM) or whether it could provide evidence for physics beyond the SM. To do that, it is essential to make clear observations of the new particle in all expected channels and measure its properties accurately.
The new particle discovery in ATLAS [3] was based on datasets with integrated luminosities of 4.8 fb −1 and 5.8 fb −1 , produced at centre of mass energies of √ s = 7 TeV and √ s = 8 TeV, respectively [1]. At the LHC, the Higgs boson is produced predominantly through a top quark dominated loop in the gluon-fusion process (ggF ). Vector boson fusion (VBF), comes next in cross section, where the Higgs is produced by the fusion of two W or Z bosons, followed by the associated production of the Higgs boson with a single W or Z (V H), or with a pair of top quarks (ttH). Five decay channels were combined in the discovery of the new particle: H → ZZ ( * ) → 4ℓ, H → γγ, H → W W ( * ) → ℓνℓν, H → τ + τ − , and H → bb [1]. Clear observations of the new particle in a single channel have so far been made in analyses focusing on bosonic decays, such as H → ZZ ( * ) → 4ℓ and H → γγ. Fermionic decay channels such as H → τ + τ − and H → bb will play an important role in ascertaining the nature of the new particle. Our present knowledge of the new particle indicates that it is a neutral resonance with a mass in the region of 125 GeV (as observed in the H → γγ and H → ZZ ( * ) → 4ℓ channels). The observation in the H → γγ channel disfavours a spin 1 particle [4].
Our knowledge of the coupling properties of the new particle is discussed in section 2. This is followed by updates of the H → W W ( * ) → ℓνℓν, H → τ + τ − and H → bb ATLAS searches, in sections 3, 4, and 5. Finally, section 6 gives an impression of what results can be expected from ATLAS given 300 fb −1 and 3000 fb −1 of √ s = 14 TeV collision data.    [8]. Right: 68% and 95% CL contours for the fit to the fermion (κ F ) and vector (κ V ) coupling scale factors [5].

Coupling properties of the new particle
The same data, analyses, and statistical models used for the ATLAS discovery was examined to investigate the couplings of the new particle [5]. It is convenient to define a signal strength observable, µ, as the ratio of the observed cross section to the SM expected cross section, µ = σ/σ SM , in each investigated channel or the statistical combination of several channels. The background-only hypothesis corresponds to µ = 0 and the SM hypothesis including the Higgs boson corresponds to µ = 1. The couplings of the new particle are investigated with the following assumptions [6]: only modifications of coupling strengths are taken into account; the new particle is assumed to be a single, CP-even scalar, as expected of the SM Higgs boson; it is assumed to have a mass of 126 GeV with a negligible width. This allows the product of the cross section and branching ratio in each channel ii → H → f f to be decomposed as: where the κ factors multiply the Yukawa couplings of each initial-(i) or final-state (f ) particle flavour. This is a convenient and flexible framework in which to explore the couplings of the new particle. It should be noted, however, that new physics responsible for a deviation of the observed couplings with respect to the SM ones might also modify the kinematics of particles in the observed final states. This is not accounted for by this framework, which is however still assumed to be valid to identify small deviations in the observed couplings with respect to the SM expectations. The fit to the experimental data is done using a profile likelihood method [7].
The simplest model tested in this framework assumes all coupling parameters are modified by a single scaling factor κ, which corresponds to the square root of the global signal strength µ = 1.4 ± 0.3 [5]. The fit to the data yields κ = 1.19 ± 0.11(stat) ± 0.03(syst) ± 0.06(theory). The combined signal strength value has since been updated [8] and is shown in Figure 1 (left).
A different benchmark scenario can be explored by assuming a single scale factor κ F = κ t = κ b = κ τ for all fermion couplings, and a different one, κ V = κ W = κ Z , for vector bosons. In addition to the weak constraints provided by the H → bb and H → τ + τ − channels, sensitivity to the fermion coupling comes from the top-quark loop in the gluon-fusion diagram contributing to several channels. Sensitivity to the vector boson coupling parameter is provided by H → γγ, H → ZZ ( * ) → 4ℓ, and H → W W ( * ) → ℓνℓν decays. Assuming further that only SM particles contribute to the Higgs boson width, a fit to the coupling strength scale factors can be made. Figure 1 (right) shows the correlation between κ F and κ V scale factors. Only the relative sign between κ F and κ V is physical, so negative values of the scale factors are shown. Some sensitivity to this sign is found from the negative interference between the W and t loops in H → γγ decays.   Figure 2. Transverse mass distribution in the H + 0-jet category of the H → W W ( * ) → ℓνℓν analysis (left) and the H +1-jet category (centre) [9]. Right: probability for the background-only hypothesis, p 0 , for H → W W ( * ) → ℓνℓν combining all analysis categories [9]. The solid line corresponds to the observed p 0 , while the dashed line corresponds to the expectation for the signal+background hypothesis as a function of the mass of the hypothesised signal.
The fit shows a small preference for the local minimum near the SM expectation.
and updates the previously published results [1]. The high branching ratio of this channel (BR(H → W W ( * ) → ℓνℓν) ≈ 22% for m H around 125 GeV [10]) and the availability of leptonic final states which facilitate triggering and rejecting the main backgrounds make this a very sensitive channel. The updated analysis focuses on the H → W W ( * ) → eνµν channel. In addition to two isolated, opposite charge leptons of different flavour and high transverse momentum (p T above 15 or 25 GeV), events are required to have a large missing transverse momentum (E miss T ) relative to the direction of the direction of the leading lepton or the closest jet, E miss T,rel . Due to spin correlations in the W W ( * ) system, arising from the spin-0 nature of the SM Higgs boson and the V-A structure of W boson decay, the charged leptons tend to have similar directions for the SM Higgs signal and a flatter distribution for the dominant W W background. This feature is exploited by requiring the di-lepton invariant mass, m ℓℓ to be less than 50 GeV, and the azimuthal angle difference between the two leptons, ∆φ ℓℓ less than 1.8 radians.
The background composition depends on the jet multiplicity. Events are therefore classified into H + 0-jet and H + 1-jet categories, and different cuts are applied to improve the sensitivity in each one. In this and the following analyses, jets are reconstructed using the anti-k t algorithm [11] with distance parameter R=0.4. Events with higher jet multiplicity are dominated by top quark production and are not considered in the statistical analysis. An important feature of this and later analyses is the data-driven determination of the most important sources of background with the help of control regions enriched in each process. The main backgrounds are W W and top production (both tt and single-top), but W +jets and Z/γ * → τ τ are also determined from specific control samples. More details can be found in [9].
The presence of high transverse momentum neutrinos in the signal prevents the reconstruction of the Higgs boson mass. A transverse mass variable, m T [9], is thus used as the final discriminating variable in the statistical analysis. Figure 2 (left and middle) shows distributions of this variable in the H + 0-jet and H + 1-jet categories. The statistical analysis of the m T distribution in each category employs a binned likelihood, L(µ, θ), constructed as the product of Poisson probability terms. This depends on the signal strength, µ, and on a set of nuisance parameters, θ, which affect signal and background predictions [7]. The probability  of the background-only hypothesis, p 0 , is calculated in the statistical analysis, as well as the exclusion limits on the signal cross section, following the CL s method [13]. Figure 2 (right) shows the p 0 value observed in the H → W W ( * ) → ℓνℓν search, combining the various analysis categories. A broad minimum is observed, which is caused by the poor mass resolution of the m T discriminating variable. The observed p 0 value for m H = 125 GeV is 4×10 −3 , corresponding to 2.6 standard deviations from the background-only hypothesis.

H → ττ searches
Due to the relatively high mass of tau leptons, the H → τ + τ − decay corresponds to the dominant leptonic decay mode of the SM Higgs boson (BR(H → τ + τ − ) ≈ 6.3%). Both leptonic (τ lep ) and hadronic (τ had ) decays of the tau lepton were considered in this search [12], which considered the H → τ lep τ lep , H → τ lep τ had , and H → τ had τ had decay channels. Events were selected by requiring exactly two opposite-charge reconstructed τ candidates. The invariant mass of the τ + τ − system, m τ τ , was used as the discriminant variable in all three analyses, and was estimated with a m τ τ resolution of 13 -20% [14], depending on the event topology and kinematics. Events were further subdivided into different categories to target different topologies, the main ones being the VBF category, containing events with two high-p T jets widely separated in rapidity, and the Boosted category, which includes events with a high transverse momentum of the τ τ system. Other less significant categories were considered, and more details can be found in [12].
As can be seen in Figure 3 (left), the main source of background after the event selection is Z → τ τ , with other sources such as top (both tt and single top), multijet production, and backgrounds from τ mis-identification being important in some categories. To avoid uncertainties in the simulation of Z → τ τ production, this background was modelled by selecting a sample of Z → µµ in collision data and replacing the muon momenta with simulated tau leptons decays. Other backgrounds were estimated from collision data in control samples defined by inverting selection cuts specific to each analysis category. A detailed account can be found in [12].   A broad excess of the observed limit with respect to the expected limit, could indicate the presence of signal. The local p 0 value for this channel is shown in Figure 3 (right) and shows that this excess corresponds at most to a 1.5σ deviation from the background-only hypothesis.

H → bb searches
The Higgs boson decays predominantly to a pair of b quarks in the mass region of m H ≈ 125 GeV (BR(H → bb) ≃ 58% [10]). On the other hand, the experimental observation of the dominant ggF production mechanism is extremely challenging, due to the overwhelming background from multijet production. Instead, the ATLAS searches concentrate on channels where the Higgs boson is produced in association with gauge bosons, W H and ZH, or top quarks, ttH. These analyses relied on a b-tagging algorithm tuned to identify true b-initiated jets with an efficiency of 70%, and rejection factors of around 5 and 150 for c-quark jets and light-quark jets, respectively.

H → bb searches in associated production channels with a top quark pair
As the SM Higgs coupling to fermions is proportional to the fermion mass, the ttH channel provides experimental sensitivity to the largest Higgs coupling constant. It is also free from assumptions that affect other channels sensitive to this coupling through loops. A search [15] the ttH → bb bℓν bqq , decay channel was performed using 4.7 fb −1 of √ s = 7 TeV LHC data.
Events were required to have an electron (muon) with p T above 25 GeV (p T > 20 GeV) and at least four jets with p T above 25 GeV. Additional cuts were applied on E miss T and m T [15]. Selected events were classified into different categories depending on the jet multiplicity and on how many jets were b-tagged. The signal to background ratio is highest for events containing at least 5 jets, 3 of which are b-tagged. For each category, a discriminating variable was considered: either the scalar sum of jet transverse momenta (H had T ) or the invariant mass of the two jets most likely to originate from the Higgs boson decay (m bb ). In the analysis categories where m bb is used (with 6 jets and 3 or ≥ 4 jets), a kinematic likelihood fit was performed to reconstruct the tt system in each event. The remaining b-tagged jets were identified as Higgs boson decay products.
In addition to improving the analysis sensitivity, the analysis categories were used in the statistical analysis to constrain nuisance parameters corresponding to various sources of   Figure 4 (right) as a function of the Higgs boson mass. For a mass of 125 GeV, an exclusion limit of 13.1 × σ SM (10.5 × σ SM ) was observed (expected) at the 95% CL.

H → bb searches in associated V H production
The ATLAS search [16] for associated production of the Higgs boson with a W or a Z vector boson (V H) was made in three exclusive channels, corresponding to events containing 0, 1, or 2 charged leptons, in addition to two b-tagged jets. These target respectively the ZH → ννbb W H → ℓνbb and ZH → ℓℓbb decay channels, where ℓ refers to an electron or a muon. It used 4.7 fb −1 of √ s = 7 TeV LHC data and 13.0 fb −1 of √ s = 8 TeV data. This analysis updated a previously published analysis [17] by re-optimizing selection cuts and the analysis strategy.
In the 0-lepton channel, transverse energy triggers were used, with efficiencies above 50% for E miss T above 120 GeV, and essentially 100% for E miss T above 160 GeV. The trigger turn-on curve in simulated events was corrected to that obtained in data using W +jets events. In all categories, jets were selected with cuts of p T > 45 GeV and p T > 25 GeV for the leading and remaining jets, respectively. The main background in the 2-lepton channel is Z+jets production, whereas both top (single-top and tt) and W +jets are the main backgrounds affecting the 1-lepton analysis (see Figure 5). All three backgrounds are significant in the 0-lepton analysis. Cuts are applied to topological variables which were optimized for each channel and event category. In each channel, events were further classified into separate categories defined by the transverse momentum of the reconstructed vector boson, p V T . The p V T categories in the 1-and 2-lepton analyses were: to background, and drive the analysis sensitivity. Control samples rich in top events were also used in the statistical analysis. The invariant mass of the bb system, m bb , was the discriminating variable used in all categories. As in the previous analysis, the main sources of background, top, W + b jets, and Z + b jets, were constrained from the data during the statistical analysis. The production of W Z and ZZ with one Z boson decaying as Z → bb has a very similar signature as the V H signal, but with a cross section ≈ 5 times higher, and p V T amd m bb distributions peaking at lower values. As a validation of the V H search, a separate fit was made by fixing the Higgs cross section at its SM value and fitting the amount of diboson background. Figure 5 (centre) shows the resulting m bb distribution after subtracting all sources of background except for the diboson contribution. The measured diboson signal strength is µ D = 1.09 ± 0.20(stat) ± 0.22(syst), in agreement with the SM expectation, and corresponded to a significance of 4.0 σ. After fixing the diboson contribution to its SM value, a fit was performed to the Higgs boson signal strength to extract the 95% CL exclusion limits. Figure 5 (right) shows the result of this fit for the combination of all channels and analysis categories as a function of the Higgs boson mass. For a mass of m H = 125 GeV, a 95% CL exclusion limit of 1.8 × σ SM (1.9 × σ SM ) was observed (expected). way. In addition to existing analyses of H → γγ, H → ZZ ( * ) → 4ℓ, H → W W ( * ) → ℓνℓν and H → τ + τ − , this study has considered low rate Higgs boson channels: V H and ttH production with H → γγ decays, and H → µµ. The expected precision on the signal strength precision achieved in each individual channel is shown in Figure 6 (left) for a Higgs boson mass of 125 GeV and integrated luminosities of 300 or 3000 fb −1 collected at √ s = 14 TeV.

The long view
Individual measurements are used in a combined statistical analysis. For generality, no assumptions are made on the presence of physics beyond the SM. This means that no assumption can be made on the total width of the Higgs boson, Γ H , and only ratios of partial widths can be determined. For all production (i) and decay (f ) modes, the cross sections times branching ratios, σ i × BR f , are assumed to be proportional to the product of partial widths given by Γ i × Γ f /Γ H . In the notation of section 2 this means Γ X /Γ Y = κ 2 X /κ 2 Y = λ 2 XY , and the relative uncertainty on the ratio is given by ∆(Γ X /Γ Y ) Figure 6 (right) shows the expected relative uncertainties on the estimated partial width ratios. A minimal coupling benchmark scenario, already discussed in section 2, assumes no new physics and only two independent parameters for the vector (κ V ) and fermion (κ F ) couplings. In this case, experimental precisions of ≈2% on κ V and ≈3.5% on κ F are expected with 3000 fb −1 .
In order to completely determine the parameters of the SM and establish the Higgs mechanism as the origin of electroweak symmetry breaking, it is important to measure the self-couplings of the Higgs boson. The observation of Higgs boson pair production provides sensitivity to the Higgs trilinear coupling and may be accessible at the LHC. For a Higgs boson mass of around 125 GeV, a relative uncertainty of ≈30% on the Higgs trilinear coupling scale factor should be achievable from observations in the HH → bbγγ and HH → bbτ + τ − channels, assuming 3000 fb −1 of √ s = 14 TeV data.

Conclusions
The discovery of a new, Higgs-like boson raises crucial questions on the nature of the new particle, and determines a programme of measurements needed to measure its properties. This paper gives an overview of the latest Higgs studies in ATLAS. After describing the methods used, and results of a first exploration of the couplings of the new particle, it gives a brief description of new searches in the H → W W ( * ) → ℓνℓν, H → τ + τ − , and H → bb channels. Finally it gives an overview of the sensitivity achievable in future Higgs boson studies in ATLAS.