Exclusion of heavy, broad resonances from precise measurements of WZ and VH final states at the LHC

A novel search for heavy vector resonances in the H→bb¯\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$H\rightarrow b{\bar{b}}$$\end{document} and Z→bb¯\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Z\rightarrow b{\bar{b}}$$\end{document} final states in association with a leptonically decaying V (Z or W) and W-only respectively, is proposed. It is argued that excesses with respect to the Standard Model prediction should be observed in all final states (0, 1 or 2 leptons), with the 1-lepton final state being the strongest. Since the relative strengths of these excesses depend on branching ratios and efficiencies, this is a clear signal for the presence of heavy resonances or their low mass tails. A general vector-triplet model is used to explore the discovery potential as a function of the resonance mass and width. Recent Higgs to bb¯\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$b{\bar{b}}$$\end{document} observation data reported by the experiments ATLAS and CMS are used to test the model. Current limits are extended to resonance widths over mass as large as 9%.

Broader resonances (Γ /m > 4%) appear for larger values of the resonance coupling strength to weak bosons and the Higgs, g V , and depending on their mass, width, and production cross section, the LHC experiments can be sensitive to a large part of the resonance distribution. It is interesting to study to what extent the experiments can be sensitive to the full distribution or tails of such resonances and what the best observables are. In this work, we start addressing this quesa e-mail: stathes.paganis@cern.ch tion with a specific final state: a pair of b quarks produced in association with a weak gauge boson (Z or W ) that subsequently decays leptonically (where leptons are electrons and muons). The interest in this final state is because in the large coupling regime, g V > 3, the heavy boson branching fraction is dominated by decays to W Z and V H, while the decays to fermions are suppressed [24]. At the same time, both the Drell-Yan qq and the VBF production modes contribute, providing non-negligible cross sections. Here, we propose as an early indirect signal of the presence of vector resonances, the simultaneous excess in the 0, 1, and 2-lepton final states, for both W Z → νbb and V H → (ν)bb, in a measurement of the bb quark invariant mass M bb . The strength of these excesses in the different final states can be predicted by branching ratios and experimental efficiencies. We argue that the 0, 1, and 2-lepton final states should all be sensitive, albeit with very different sensitivity, to exotic V → V H decays. In particular, the 1-lepton state will be the most sensitive in Higgs excesses and in addition, it should also show a smaller level of excess in the Z -boson bb peak. The pattern of these excesses should be a clear sign of a heavy vector triplet (HVT), even before its observation as a (broad) mass resonance at the TeV scale. Using a general HVT framework that is also used by the experiments to model such heavy vector triplets [25], we derive the sensitivity of a single LHC experiment as a function of the resonance width.
In this paper, we first introduce the HVT framework and summarize the recent LHC results. In Sect. 3 we present an analysis used to search for heavy vector triplets at the LHC, and we report on the search potential for the bb + 0, 1, and 2-lepton final states, for a range of masses and widths. Finally, the compatibility of our predictions with the recently published 13 TeV H → bb data from ATLAS and CMS in terms of expected and observed limits, is discussed.

BSM heavy vector resonances
A general HVT phenomenological Lagrangian can be used for the modelling of resonances predicted by a wide range of Beyond the Standard Model (BSM) scenarios [25]. The Lagrangian describing the interactions of these resonances V a , a = 1, 2, 3 with quarks, leptons, vector bosons and the Higgs boson is shown below: where q k and k are the quark and lepton weak doublets, H is the Higgs doublet and σ a the three Pauli matrices. In this Lagrangian, the HVT triplet V a = (W + , W − , Z ) interacts with the Higgs doublet, i.e. the longitudinal degrees of freedom of the SM W and Z bosons and the SM Higgs, with a coupling strength g V . In order to allow for a broader class of models, this coupling strength can be varied by the parameter c H , so in the Lagrangian the full coupling to the SM weak and Higgs bosons is g V c H . The HVT resonances also couple to the SM fermions, again through their coupling to the SM weak and Higgs bosons, g 2 /g V , where g is the SM SU (2) L weak gauge coupling. This coupling between HVT resonances and fermions is also controlled by an additional parameter c F to allow for a broader range of models to be included, as follows: g 2 c F /g V . As discussed in [23], the experiments consider two Drell-Yan (DY) production scenarios: model A is a scenario that reproduces the phenomenology of weakly coupled models based on an extended gauge symmetry [9]. In this case, the couplings are g 2 c F g V = −0.55 and g V c H = −0.56, with the fermion coupling being universal. The second DY scenario, referred to as model B, implements a strongly coupled scenario as in composite Higgs models with g 2 c F g V = 0.14 and g V c H = −2.9. In model B, the V resonances are broader than in the weakly coupled scenario, model A, but for |g V c H | ≤ 3 they remain narrow relative to the experimental resolution. For |g V c H | > 3, the resonance intrinsic width becomes significant and cannot be neglected. In summary, the ATLAS and CMS benchmarks correspond for Model A to c H = − g 2 g 2 V , c F = − 1 3 and g V = 1, and for model B to c H = −1, c F = 1 and g V = 3. ATLAS and CMS experimental data from direct m V H and m V V searches, exclude part of the parameter phase space (g V , c H , c F ), for which the intrinsic width Γ of the new bosons is dominated by the experimental resolution (∼ 4% of the mass). This is also the case, for example, for |g V c H | ≤ 3, where negligible natural width is assumed (narrow width assumption). The goal of this work is to explore the part of the HVT parameter space with |g V c H | > 3, where the heavy resonances have a significant natural width. Following the model B benchmark, we fix the two constant factors to c H = −1, c F = 1 and allow g V to vary.
Results from recent experimental analyses that exclude part of the HVT phase space are shown in Fig. 1. The validity of these exclusions is actually limited due to the narrow width approximation. For larger, |g V c H | > 3, couplings, the width grows and becomes comparable to or larger than the experimental resolution, leaving a large part of the HVT phase space unexcluded: 3 < |g V c H | < 4π , where g V 4π is the perturbative limit. This is true even for resonance masses of order TeV.
In this work we argue that the presence of non-zero width resonances, could be observed as an excess of events in the V H → (ν)bb and W Z → νbb decays with 0, 1, or 2 leptons in the final state. Independent of the fact that due to the experimental resolution the close-by Z → bb and H → bb peaks have a partial overlap, in the 1-lepton case the excess must be present in both peaks, while in the 0 and 2-lepton final states it should appear only in the Higgs peak (Z does not couple to Z Z). In addition, the excess in the 1-lepton category should be more significant, since in this case both W + and W − contribute. A combined analysis of the three final states can quantify the excess and correlate it to events with bb pairs of higher p T than in the SM Higgs production, since these pairs originate from heavy TeV-scale objects.
ATLAS and CMS recently published results in the search for the SM H → bb in association with a gauge boson decaying leptonically [26,27]. This final state is identical with the one we propose here having sensitivity to HVT resonances. Broad heavy resonances mainly decaying to dibosons and pairs of fermions, occur for larger values of the g V coupling (g V > 3 for c H = −1 and c F = 1). The resonance width as a function of g V is shown in Fig. 4 and the branching ratios to a pair of fermions and diboson final states are shown in Fig. 5. In both figures the calculations shown were performed by implementing the HVT model in MadGraph5 [28]. Figure 4 shows that for g V > 3 the resonance natural width over the mass ratio exceeds ∼ 3%, and as it can be seen in Fig The magnitude of the impact of these resonances on SM measurements depends on their cross section which in turn, depends on the mass and the g V coupling. Heavy resonance Drell-Yan and Vector Boson Fusion (VBF) production cross sections for masses in the range 1-3 TeV are shown in Fig. 6. The DY mode is always dominant up to large values of g V close to the perturbative limit g V 4π , and the total cross section drops with g V . In Fig. 6 we can see that in the mass range 1 − 2 TeV, the cross sections for g V > 3 range from O(1) to O(100) fb, leading to non-vanishing contributions at present and future measurements. For parts of the parameter space corresponding to low resonance masses and high g V that are theoretically excluded (parts for which the input electroweak parameters α E W , G F and M Z , are not reproduced by the HVT model), the cross section is not provided in Fig. 6.

Resonance search potential at colliders
To evaluate the expected signal yields for a broad range of masses and widths, we rely on Monte Carlo samples generated with MadGraph5 [28], interfaced to PYTHIA [29].
To simulate the response of an LHC-like experiment, realistic resolution and reconstruction efficiencies for electrons, muons, photons and jets were applied with the Delphes framework [30]. The specific cut-based event selection was based on the recent V H with H → bb results from ATLAS and CMS [26,27]. As discussed later, the efficiencies were normalized to the experimental ones. The selected events were split in three categories according to the final state: two b jets with 0, 1 or 2 leptons. We follow more closely the ATLAS selection including the requirement of 2 or more jets out of which exactly two must be b jets. We will call this SM selection, the baseline selection, in order to differentiate from additional discriminants which enhance the BSM Higgs signal. The BSM Higgs signal decays to a bb pair with a transverse momentum, p ⊥ , significantly larger than the SM Higgs. For this reason, an additional requirement applied to the baseline analysis, is a cut on the transverse momentum of the bb system. As in the LHC analyses, we only consider resolved bb pairs although the search can be extended to a single, fat bb jet analysis. As a discriminant, the bb invariant mass M bb is used, which after full selection and subtraction of the bb continuum shows two peaks due to the presence of Z → bb and H → bb.
The characteristic of the BSM signal is that the non-SM produced H → bb has high p ⊥ . This can be seen in Fig. 7, where the exotic Higgs has typical p ⊥ > 150 GeV while the SM Higgs is softer. The expected signal and resonant SM background yields for the baseline selection are presented in Table 1 for the 0, 1 and 2-lepton final states, for a resonance of mass 1.5 TeV, g V = 5 and an integrated luminosity of 100 fb −1 . For a p ⊥ > 200 GeV cut, we observe a significant increase in the fraction of BSM signal over the SM resonant yields, that allows to estimate the discovery potential as a function of the HVT model parameters. In the calculations shown in Table 1, in order to realistically estimate the heavy resonance parameter space that can be probed at LHC as a function of luminosity, we have normalized the efficiencies to the published results by the LHC experiments. In this way, we maintain realistic non-resonant background levels and to some extent take systematic differences between the full simulation used by the experiments and the fast simulation used here partially into account. The signal efficiency × acceptance is at the level of few % and based on the signal cross-section expectations summarized in Fig. 6, observable excesses in M bb should be expected for broad resonance masses below 2 TeV, even for present LHC luminosities.
Example scenarios with parameters M Z = 1.5 TeV, g V = 4 and g V = 5 after baseline selection and a tight transverse momentum cut p ⊥ > 300 GeV, for which the bosonic decay branching ratio is dominant, are presented in Figs. 8 and 9 respectively. In these figures, the combined M bb distribution is shown for an integrated luminosity of 200 fb −1 . Assuming that the new physics does not modify significantly the SM Higgs couplings, the excess over the SM expectation will indicate the presence of heavy vector resonances decaying into V H and W Z final states. As we have already seen from Table 1, most of the Higgs excess comes from the 1lepton category, while all of the Z W excess comes from the 1-lepton category but it is rather small.
The resonance search potential for a single LHC experiment and an integrated luminosity of 200 fb −1 , is shown in Figs. 10, 11, and 12, for the 0-lepton, 1-lepton and 2-lepton  A p ⊥ > 300 GeV cut has been applied to the bb system categories, respectively [31]. As seen from these results, the search is sensitive to part of the phase space not currently probed by existing searches. For 200 fb −1 , the sensitivity extends to masses up to about 2 TeV, and resonance widths Γ /m 10%, although the higher the width the lower the resonance mass reach. For Γ /m 10% only a mass up to 1.6 TeV can be reached. The reason for the search not being sensitive to larger widths is not only due to the reduction of cross section with g V , but also due to a loss of efficiency because for V masses greater than 1.5 GeV, the majority of b quark pairs fall in the same jet. More precisely, for resonance masses of 2 TeV or higher, the fraction of resolved bb pairs is 5% or less, respectively. Here it should be stressed that an extension of the proposed search to include fat jets could push these limits at higher values.
In an attempt to test the prediction of the model against the ∼ 80 fb −1 data recently reported by ATLAS and CMS, the pattern of the excess with the HVT model. It should be pointed out that the signal cross-sections were calculated at leading order, thus an additional theory uncertainty due to missing higher order corrections should be considered. These theoretical uncertainties are expected to be small with respect to the rest of the systematic uncertainties of this analysis.

Summary and conclusions
In this work we explored the potential of a novel search of heavy vector resonances of non-zero natural width, decay-