ATLAS searches for resonances decaying to boson pairs

Many extensions to the Standard Model predict new particles decaying into two bosons (W, Z, photon, or Higgs bosons) making these important signatures in the search for new physics. Searches for such diboson resonances have been performed in final states with different numbers of leptons, photons, jets and b-jets where new jet substructure techniques to disentangle the hadronic decay products in highly boosted configuration are being used. This document summarizes recent ATLAS searches for resonances decaying to diboson final states, VV , VH and HH with LHC Run 2 data collected.


Introduction
The discovery of the Higgs boson [1] with a mass of approximately 125 GeV in 2012 finalizes the Standard Model (SM) in the description of particle interactions at energies up to a few hundred GeV.Despite the fact that a lot of precision measurements and searches with data from the Large Hadron Collider (LHC) show no significant deviations from the SM, some unanswered questions still exist.Therefore one of the main purposes of particle physics research at the LHC is to seek new intimations of physics Beyond the Standard Model (BSM).Many of BSM models, motivated by hierarchy and naturalness arguments [2], predict the existence of new heavy resonances decaying into diboson, such as composite Higgs models [3], warped extra dimensions [4], models with an extended Higgs sector [5] and Grand Unified Theories [6].
In order to evaluate the sensitivity, to optimise the event selections, and for comparison with the data, a number of models are used as benchmarks.Limits on cross-sections as a function of the particle mass are obtained in the framework of three representative benchmark models: an extended Higgs sector serves as benchmark model for spin-0 resonances, Heavy Vector Triplets (HVT) Model A (which has comparable branching ratios to fermions and gauge bosons) and Model B (which has enhanced couplings to gauge bosons and where the fermionic couplings are suppressed) for spin-1 in form of W and Z [7], and bulk Randall-Sundrum Gravitons (RSG) [8] for spin-2 resonances.
In this document, the results from recently performed searches for resonant production of VV, VH and HH are summarised.They are carried out using data taken with the ATLAS detector [9] during the 2015-2016 run of the LHC (Run 2).
The searches are performed by looking for localized excesses above the smoothly falling SM background.For this purpose, the mass of the diboson system is used as the final discriminant.This is the invariant mass unless Z boson decays to neutrinos (ν) are included, where only the transverse mass reconstruction is feasible.e-mail: Andrey.Ryzhov@ihep.ru

Analyses overview and results
This section summarizes the individual searches presented in this article.Subsection 3.1 gives the results for the VV, VH, HH searches in fully hadronic decay modes, followed by the results for VV resonances in semileptonic decay modes in Subsect.3.2.And finally a short summary is given in Table 1, which shows observed excluded resonance masses at the 95% confidence level (CL) for all analyses.The full description of each analysis can be found in the reference papers.These analyses are performed using data collected with the ATLAS detector at √ s = 13 TeV with an integrated luminosity that differs for each particular analysis and that will be indicated in each subsection.

VV , VH and HH resonances in fully hadronic decay modes
The VV → qqqq analysis [16] is performed only in the merged regime using 36.7 fb −1 of protonproton collision data at a centre-of-mass energy of √ s = 13 TeV recorded with the ATLAS detector at the Large Hadron Collider in 2015 and 2016.Events are required to contain at least two large-R jets with |η| < 2.0 (to guarantee a good overlap with the tracking acceptance) and mass m J > 50 GeV.The leading (highest p T ) large-R jet must have p T > 450 GeV and the subleading (second highest p T ) large-R jet must have p T > 200 GeV.The invariant mass of the dijet system formed by these two jets must be m JJ > 1.1 TeV to avoid inefficiencies due to the minimum jet-p T requirements.Boson-jet candidates are identified by applying a boson-tagging procedure.Further requirements are applied on the number of tracks in the two jets, their rapidity difference and their p T asymmetry.This selection strongly suppresses the large backgrounds due to SM multi-jet events.Other contributions to the background are small.The dijet invariant-mass spectrum of the background is estimated directly from data using a parametric form, and the search proceeds by seeking resonant structures in the falling spectrum of the background.The ability of the parametric shape to model distributions is tested with validation regions from collision data (see Figure 1(left)).No significant excess over the background is found in the signal regions, as shown for the WZ selection in Figure 1(right) and thus 95% CL exclusion limits are set on the production cross section times branching fraction in the HVT and bulk RS models as shown in Figure 2. The VH → qqbb analysis [17] is carried out with 36.1 fb −1 of Run 2 data.There is only merged regime which uses two large-R jets selection with p T > 450 (250) GeV for the leading (sub-leading) jet, |η| < 2 and an invariant mass larger than 50 GeV.The analysis is performed categorizing the signal in four regions, namely WH 1 b-tag, WH 2 b-tag, ZH 1 b-tag and ZH 2 b-tag.Signal and backgrounds from tt and W/Z+jets production are modelled with Monte Carlo (MC) simulation.While multijet MC events are used as a cross-check, the primary multijet background estimation is performed using data.This background shape is extracted from 0 b-tag "SR", which consists of ≈ 99% multijet events.The normalization of each distribution is obtained using the sideband regions.The data are in agreement with the Standard Model expectations (see Figure 3 = 13 TeV, 36.7 fb s qqqq → VV Observed 95% CL limit Expected 95% CL limit = 13 TeV, 36.7 fb s qqqq → VV Observed 95% CL limit Expected 95% CL limit Upper limits at the 95% CL on the cross section times branching ratio for (left) WW + WZ production as a function of V mass, (right) WW + ZZ production as a function of G KK mass.The dotted and solid red lines show the predicted cross section times branching ratio as a function of resonance mass for the HVT Models A and B with g V = 1 and g V = 3, respectively, or the bulk RS model with k/M Pl = 1.Both figures are taken from ref. [16] Events / 100 GeV The m JJ distributions in the VH signal regions for data (points) and background estimate (histograms) after the likelihood fit for events in the 1-tag categories.The pre-fit background expectation is given by the blue dashed line.The expected signal distributions (multiplied by 50) for a HVT benchmark Model B V boson with 2 TeV mass are also shown.Right: The observed and expected cross-section upper limits at the 95% confidence level on the cross section times branching ratio for HVT Model A and Model B in the ZH signal regions.The red and magenta curves show the predicted cross-sections as a function of resonance mass for the models considered.Both figures are taken from ref. [17] with p T > 450 (250) GeV for the leading (sub-leading) jet, |η| < 2 and an invariant mass larger than 50 GeV.Since high-mass resonances tend to produce jets that are more central than multijet background processes, the two large-R jets are required to have a separation |∆η| < 1.7.The resolved selection begins with the requirement that the event contains at least four b-tagged anti-k t R = 0.4 jets with p T > 30 GeV and |η| < 2.5.The leading jets are combined in pairs forming the Higgs candidates and requiring to pass ∆R and p T cuts which depend on invariant mass of HH system.For both the boosted and the resolved analyses the dominant background is QCD multijet which is estimated in a side band region and validated in a control region in the m H 1 , m H 2 plane as shown in Figure 4(left), where m H refers to the invariant mass of the Higgs candidate.Good agreement is observed between the data and the background prediction (see Figure 4(right)).

VV resonances in semileptonic decay modes
All semileptonic analysis is carried out with 36.1 fb −1 of LHC proton-proton collision data collected at √ s = 13 TeV with the ATLAS detector in 2015 and 2016.One boson candidate is required to be formed by a lν, ll or νν pair and the other boson by a qq pair or single jet, where both bosons should satisfy V mass window cuts.Events with two additional small-R jets (referred to as tag-jets) with large pseudorapidity separation and invariant mass are classified as vector-boson fusion (VBF) candidates, while absence of this topology is interpreted as gluon-gluon fusion/Drell-Yan (ggF/DY) production.
The ZV → llqq search [19] explores the VBF and ggF production of a heavy Higgs boson H, the VBF and DY production of an HVT W boson, and the ggF production of a bulk RS graviton G KK .It also utilises both the merged and resolved reconstruction for the V → qq decay.The search begins with the identification of the Z → ll decay (the leading lepton must satisfy E T (p T ) > 28 GeV and the dilepton invariant mass m ll is required to be consistent with the Z boson mass), followed by classifying events into the VBF (η tag 1 • η tag 2 < 0, |∆η tag | > 4.7, m tag > 770 GeV) or ggF/DY categories and finally the selection of either the ZV → llJ or ZV → ll j j final states.Merged events are further required to have p J T > 200 GeV and min(p ll , p J )/m llJ > 0.3 for the H → ZZ search and > 0.35 for the W → ZW and G KK → ZZ searches.Two signal regions are defined, high-purity (HP) and low-purity (LP) SR, that depend on boson tagging working point requirement.Events that have failed merged selection are required to have at least two small-R jets with p T > 60 GeV.The kinematic quantity (p ll T ) 2 + (p j j T ) 2 /m ll j j is required to be greater than 0. In order to extract the signal rate information, a simultaneous maximum-likelihood fit is performed to the observed distributions of the final discriminants in the signal regions, m llJ or m ll j j .The data are found to be consistent with the SM background predictions and no evidence of heavy resonance production is observed (see Figure 5).Limits are presented in Figure 6.→ p miss T .Due to the presence of neutrinos in the final state, it is not possible to fully reconstruct the invariant mass of the ννJ system, so the transverse mass is used as the final discriminant: where E T,J = m 2 J + p 2 T,J .The dominant backgrounds are Z+jets, W+jets and tt processes.Data control regions are defined to check the modelling of each contribution.As in the llqq search, the shapes of kinematic variables, including the final discriminant m T , are taken from MC simulations.Figure 7 shows that no significant excess over the background is found in the signal regions.Limits are presented in Figure 6.The WV → lνqq analysis [20] is very similar to the previous ZV → llqq, also with resolved and merged regime, but some additional requirements are used to eliminate the multijet contamination (E miss T > 100 GeV, p T (lν) > 200 GeV, E miss T /p T (lν) > 0.2).The analysis selects events that contain exactly one charged signal lepton and no additional veto electrons or muons, then events are categorized to VBF or DY categories and are assigned to merged and resolved regions.Studies using simulated events show that W+ jets and tt production are the dominant background sources.The shapes of the mass distributions for events from SM production of W+jets and tt are modeled using simulated events.The results are extracted by performing a simultaneous binned maximum-likelihood fit to the m(WV) distributions in the signal regions and the control regions for W+jets and tt.The m(WV) distributions are presented in Figure 8.The data are compatible with the Standard Model background hypothesis and no significant excess has been found.

Conclusion
A brief summary of the most recent searches for resonances decaying to diboson final states, VV, VH and HH using ATLAS data from the 2015 and 2016 run at the LHC have been presented.No discrepancies with respect to the Standard Model expectations are observed and thus 95% CL exclusion limits are set on the production cross section times branching ratios in a number of benchmark scenarios.Observed excluded resonance masses at the 95% confidence level for the HVT and bulk RS models are shown in the table 1.The utilization of unprecedented luminosity of ≈ 85 fb −1 available at the end of the whole LHC Run 2 should give significant improvements on the exclusion limits.

2 Figure 1 .
Figure 1.Left: Dijet mass distributions for data in the sideband validation regions.The shaded bands represent the uncertainty in the background expectation due to the maximum-likelihood fit's statistical uncertainty.The lower panels show the significance of the observed event yield relative to the background fits.Right: Dijet mass distributions for data in the WZ signal region.The red lines correspond to the result of the fit and the shaded bands represent the uncertainty in the background expectation.Expected signals are shown for the HVT model B and the bulk RS model.The predictions for G KK production are multiplied by a factor of 10.Both figures are taken from ref. [16] (left)), with the largest excess observed at m JJ ≈ 3.0 TeV in the ZH channel with a local significance of 3.3 σ.The global significance of this excess is 2.1 σ (see Figure 3(right)).The search for HH → bbbb final state[18]  is done in both boosted and resolved regimes, covering high and low energy regimes, respectively.The results presented here benefit from an integrated luminosity of 13.3 fb −1 .The merged analysis selects events with at least two anti-k t R = 1.0 jets EPJ Web of Conferences 158, 02003 (2017) DOI: 10.1051/epjconf/201715802003 QFTHEP 2017

Figure 3 .
Figure3.Left: The m JJ distributions in the VH signal regions for data (points) and background estimate (histograms) after the likelihood fit for events in the 1-tag categories.The pre-fit background expectation is given by the blue dashed line.The expected signal distributions (multiplied by 50) for a HVT benchmark Model B V boson with 2 TeV mass are also shown.Right: The observed and expected cross-section upper limits at the 95% confidence level on the cross section times branching ratio for HVT Model A and Model B in the ZH signal regions.The red and magenta curves show the predicted cross-sections as a function of resonance mass for the models considered.Both figures are taken from ref.[17]

Figure 4 .
Figure 4. Left: The m subl J vs. m lead J distribution for the boosted analysis background model.The signal region is the area surrounded by the dashed red contour line, centred on (m lead J = 124 GeV, m subl J= 115 GeV).The control region is the area between the signal region and the orange contour line.The sideband region is the area between the control region and the yellow contour line.Right: Distributions of m 2J in the signal regions of the boosted analysis for the two-tag-split sample compared to the predicted backgrounds.The grey hatched bands represent the combined statistical and systematic uncertainties in the total background estimates.The expected signal shape for a G KK resonance of mass 2.0 TeV is also shown.Both figures are taken from ref.[18] 4 for H → ZZ and 0.5 for W → ZW and G KK → ZZ.The dijet invariant mass must be in the window[70, 105]  GeV for Z → qq and in the window [62, 97] GeV for W → qq.Also the ZV → ll j j candidates are divided into two signal regions: events with two b-tagged jets and events with fewer than two b-tagged jets.Multiple signal regions are defined to enhance the sensitivity of the search.The dominant backgrounds are the Z+jets, top quark and diboson processes.Their contributions are estimated from a combination of MC and data-driven techniques.The shapes of kinematic variables are taken from MC simulations.

− 1 H− 1 H− 1 HFigure 5 .
Figure 5. Comparisons of the observed data and expected background distributions of the final discriminants of the ggF category for the H → ZZ → llqq search: m llJ of (top left) high-purity and (top right) low-purity signal regions; m ll j j of (bottom left) b-tagged and (bottom right) untagged signal regions.For illustration, expected distributions from the ggF production of a 1 TeV Higgs boson with σ × B(H → ZZ) = 20 fb are also shown.The middle panes show the ratios of the observed data to the background predictions.The uncertainty in the total background prediction, shown as bands, combines statistical and systematic contributions.The bottom panes show the ratios of the post-fit and pre-fit background predictions.All figures are taken from ref. [19]

Figure 6 .
Figure 6.Observed (black solid curve) and expected (black dashed curve) 95% CL upper limits: on σ × B(H → ZZ) for the (top left) ggF and (top right) VBF production of a heavy Higgs boson as a function of its mass; on σ × B(W → ZW) for the (middle left) DY and (middle right) VBF production of a W boson in the HVT model as a function of its mass; on σ × B(G KK → ZZ) for the production of a G KK in the bulk RS model with couplings of (bottom left) k/M Pl = 1 and (bottom right) k/M Pl = 0.5 as a function of the graviton mass, combining llqq and ννqq searches.Limits expected from individual searches (dashed curves in blue and magenta) are also shown for comparison.Limits are calculated in the asymptotic approximation below 2 TeV and are obtained from pseudo-experiments above that.All figures taken from ref. [19]

Figure 7 .
Figure 7. Comparisons of the observed data and expected background distributions of m T in the VBF category of the H → ZZ → ννqq search: (left) high-purity and (right) low-purity signal regions.For illustration, expected distributions from the VBF production of a 1.6 TeV Higgs boson with σ × B(H → ZZ) = 6 fb are also shown.The middle panes show the ratios of the observed data to the background predictions.The bottom panes show the ratios of the post-fit and pre-fit background predictions.Both figures taken from ref. [19]

Figure 8 .
Figure 8. Post-fit signal region m(WV) distributions in the DY category.Event failing the VBF selection are assigned to this category.The merged high-purity (HP) sample of WZ (left) events, the merged low-purity (LP) sample of WZ (right) events are presented.The background expectation is shown after the profile likelihood fit to the data and signal expectations are overlaid.The HVT Model A signal at 2000 GeV is presented.The band denotes the statistical and systematic uncertainty on the background after the fit to the data.The lower panels show the ratio of the observed data to the SM background estimation.In all regions, the overflow events are included in the last bin.Both figures are taken from ref.[20]

Table 1 .
Observed excluded resonance masses (at 95% CL) for the HVT and bulk RS models.