ATLAS High-mass MSSM H/A ττ search at 13 TeV LHCP 2016 (University of Oxford) on behalf of the ATLAS collaboration 1
Motivation MSSM Higgs sector: two Higgs doublets five Higgs Bosons Two charged H+/-, one neutral CP-odd A, two neutral CP-even, h and H Described by ma and tanβ (vev ratio of doublet) at tree level Search focuses on the neutral H/A decaying to a pair of τ-leptons Status after Run-I H/A ττ analysis sensitive to unique area of parameter space (grey) Channel sensitive to tanβ > 40 for ma ~1 TeV 2
Tau Particles In ATLAS Decay hadronically or leptonically QCD jets form a major background Boosted decision tree method used to identify τhad and reject QCD jets Discriminating variables include narrowness of tau jet, tau decay length, and number of charged tracks (1 or 3) Missing transverse energy associated with decay, unable to fully reconstruct parent particle mass ETMiss Hadronic Tau Decay Tau Decay Mode 3
Analysis Overview Analysis split by di-tau decay Gluon-gluon fusion and b-associated production mechanisms considered Fully-hadronic (Had-had) and lepton+hadron (Lep-had) decay b-associated production increasingly important for high tanβ Simulated signal with Higgs mass range 200 GeV to 1.2 TeV Use a likelihood function binned in mttot : mass discriminant (Lorenz invariant vector sum in transverse plane) Search for excess of events over SM prediction Interpret results in tanβ-ma space Example tree level production mechanisms 4
Had-had Signal Region Signal Region selection Trigger: pt > 125 GeV online τhad Veto events with e/μ lead τ pt > 135 GeV, ptsublead τ > 55 GeV Taus pass BDT identification Δφ (lead-τ, sublead-τ) > 2.7 Taus opposite charge Control Region: Same sign charge taus Signal region events as a function of the total transverse mass, 500 GeV signal in purple Dominant backgrounds Jet τhad fakes from multijet processes (blue) data-driven Z ττ (yellow) simulation Non-multijet jet τhad fake sources e.g. W+jet, top (red/green) simulation with data driven correction 5
Had-had Background Multijet Poor modelling of jet τhad fakes in simulation Data driven multijet background (blue) estimate via tag and probe method in QCD enriched CR Jet trigger, no selection on lead tag jet 1 or 3 charged tracks & ±1 charge on probe τ Calculate fake factor (FF), ratio of pass/fail τ-id on probe τ in Multijet CR Bin in pt and number of tracks Apply FF to data events failing sublead τid requirement for multijet contribution Same sign charge control region events as a function of the total transverse mass 6
Had-had Background Jet τ Fakes (non-multijet) Poor modelling of jet τ fakes in simulation Data-driven correction to simulation of fake taus passing the τ-id criteria for non-multijet backgrounds Calculate rate for jet passing τ-id criteria in W( μν)+jets dominated control region Require muon in event Fake rate (FR), % jets passing τ-id on probe τ in Multijet CR Apply fake-rate to non-truth matched taus in simulated backgrounds largest source is W+jets Fake trigger rate instead of trigger application Same sign charge control region events as a function of the total transverse mass 7
Lep-had Signal Region Signal Region object selection Fire electron or muon trigger # e/μ = 1, ptlep >30 GeV, isolated # taus 1, ptτ >20 GeV Tau passes BDT Identification Signal Region selection ΔΦ (lep,τ)> 2.4 Transverse mass window : mt(lep,etmiss) <40 GeV OR > 150 GeV e + τhad : mvis exclusion window around Z peak (80-110 GeV) H/A τe+τhad H/A τμ+τhad Signal region events as a function of the total transverse mass, 500 GeV signal in purple 8
Lep-had Background Overview True lepton + true τhad μ τhad fake Simulation Signal Region 9
Lep-had Background Overview Signal Region True lepton + e τhad fake Simulation with scale factor Z ee CR Mvis 80-110 GeV 10
Lep-had Background Overview W+jets CR Invert mt True lepton + fake τhad W+jets fake factor Fail τhad ID CR Jet τhad fakes Combined fake factor Multijet CR Invert lepton isolation Fake lepton + fake τhad Fake factor Signal Region 11
Limit on cross-section x BR Had-had (blue) drives sensitivity at high mass Lep-had (red) more important at low mass 12
Limit on cross-section x BR Run-1 limit shown in red (right) Improvement above 700 GeV mass reach extended to 1.2 TeV 13
Summary No significant excesses found Exclusion limits placed on cross-section x BR and tanβ-ma plane ATLAS-CONF-2015-061 Improved sensitivity compared to Run-1 above ma = 700 GeV Update to analysis in final stages Optimised selection Improved background estimation Large improvement in expected sensitivity Watch this space! 14
Backup 15
Systematics Overview Had-had τhadτhad Multijet FF: statistics in dijet CR + uncertainty from OS/SS difference 7% MC backgrounds with mis-id τhad: FR uncertainty given by statistics in W( μν) +jets CR 9% for W( τν)+jets MC-estimated samples detector-related syst: Lep-had τlepτhad trigger SF: <30%, low stats in SF measurement τ-id, e-veto, tau reconstruction, tau energy scale, high-pt systematics also significant (up to 15%) Multijet FF: Stats in CRs dominate effect. True Lepton contamination in CR also contributes W+jets FF: q/g fraction between SR-CR source of syst. Also CR contamination from multi-jet. Total 4-8% 1P, 5-30% 3P MC-estimated samples detector-related syst: Tau-ID, reco,e-veto, tau energy scale significant impact on Z ττ, top, signal (~10%) high pt tau systematic also significant e/μ : trigger, reco, isolation, identification, energy scale (2-5%) JES/JER <4%, MET syst < 4% 16
Final combined limits continued 17
Final combined limits continued 18
Final combined limits continued 19
Had-had Signal Region Signal Region selection recommended GRL applied overlap removal order μ>e>τ>jet Taus: # taus 2, 1 or 3 tracks, opposite (±1) charge η < 2.5 (not crack region) ptlead τ > 135 GeV ptsublead τ > 55 GeV ΔΦtau_0, tau_1 > 2.7 trigger: HLT_tau125_medium1_tracktwo matched to leading tau leading tau medium isolation* subleading pass loose isolation* Control Region: same charge taus *Isolation Definition if data or truth-matched tau MC: apply loose/medium τ-id and ID SF(MC only) non-truth matched tau MC: apply fake rate 20
Had-had Signal Region 21
Lep-had Signal Region Combination e + τhad & μ + τhad 22
Lep-had Combined Fake Factor Tau ID fails loose fails medium, BDT score >0.35 Anti-tau Region Isolated Lepton Isolation Antiisolated medium Signal Region QCD Control region Inverted mt cut W+jets Control region Aim: Using the known number of events in the QCD and W+jets control regions, extract the proportion of jets faking taus in the signal region from the Anti-tau region 23
Lep-had Combined Fake Factor Tau ID fails loose fails medium, BDT score >0.35 Anti-tau Region Isolated Lepton Isolation Antiisolated medium Signal Region QCD Control region Inverted mt cut W+jets Control region 1) Get the ratio jets faking taus in W+jets CR region - FF(W+jets) 2) Get the ratio jets faking taus in QCD CR region - FF(QCD) 3) Get the proportion of QCD jets faking taus in the anti-tau region - rqcd 24
Lep-had Combined Fake Factor Tau ID fails loose fails medium, BDT score >0.35 Isolated Lepton Isolation medium Signal Region Antiisolated Inverted mt cut FFW+jets denominator FFW+jets numerator Factor of jets in W+jets faking taus 25
Lep-had Combined Fake Factor Tau ID fails loose fails medium, BDT score >0.35 Isolated Lepton Isolation medium Signal Region Antiisolated FFQCD denominator FFQCD numerator Inverted mt cut Factor of QCD faking taus 26
Lep-had Combined Fake Factor Tau ID Lepton Isolation fails loose Isolated FFlep numerator Antiisolated FFlep denominator fails medium, BDT score >0.35 medium Signal Region Inverted mt cut Factor of QCD faking leptons 27
Lep-had Combined Fake Factor Tau ID Isolated Lepton Isolation fails loose fails medium, BDT score >0.35 Extract rqcd medium Signal Region Apply FFlep Antiisolated Inverted mt cut Factor of QCD faking leptons Proportion of Wjets in Anti-tau Proportion of QCD in Anti-tau region 28
Lep-had Combined Fake Factor Tau ID fails loose fails medium, BDT score >0.35 Anti-tau Region Isolated Lepton Isolation medium Signal Region Apply FFcomb Antiisolated Inverted mt cut 29
Lep-had Combined Fake Factor Tau ID Isolated Lepton Isolation Antiisolated Inverted mt cut fails loose fails medium, BDT score >0.35 FFlep numerator Extract rqcd FFlep denominator Apply FFlep Medium Signal Region Apply FFcomb FFQCD denominator FFQCD numerator FFW+jets denominator FFW+jets numerator 30
MC samples Had-had τhadτhad Lep-had τlepτhad Signal samples gluon-gluon fusion Powheg+Pythia8 bbh amc@nlo+pythia8 (AtlFast-II) 300-1200 GeV mass points Background Z+jets + Z ττ Powheg+Pythia8 in boson mass slices ttbar + top Powheg+Pythia6 Diboson samples Sherpa W+jets samples Sherpa W+jets samples Powheg+Pythia8 Lead tau pt slices Various corrective factor for mismodelling requried 31
Mass Reconstruction - Algorithms Various mass reconstruction algorithms considered tuned to high mass Missing Mass Calculator (MMC) Assume missing transverse momentum is due entirely to the neutrinos Scan over the angles between the neutrinos and the visible τ decay products Each solution is weighted according to probability density functions that are derived from simulated τ decays MAXW solution of maximum weight point of phase space MLM highest probability mass in calculation MLN3P point with most likely neutrino decay The Matrix-element Oriented SAmplIng Calculator (MOSAIC) similar technique to the MMC in using likelihood function and probability density function uses a matrix element based maximum likelihood. Total Transverse Mass MTOT Used in Run-I MTOT considered to be the optimal mass reconstruction technique for maximising signal-background separation 32
Derivations and Analysis Frameworks Had-had Running on HIGG4D4* derivations: 1 3 track tau candidates with: ptlead τ > 160 GeV && ptsublead τ > 45 GeV OR HIGG4D4 Derivation ptlead τ > 80 GeV && ptsublead τ > 50 GeV && jet tau looseidsublead τ Significant reduction in time to produce ntuples Two xaod based Frameworks: xtau Framework and Dresden based ELCore framework xtau produces ntuples from xaods ELCORE runs directly on derivations ELCORE xtau FW xtau FW Plotting codes for ntuples vary between analysers Cross checks and acceptance challenges in place * For EOYE QCD estimation in di-jet control region we use the looser SUSY11 Requires firing of single jet triggers no tau ID requirement * For jet tau fake rate calculation use HIGG4D2 derivations Local ntuple analysis Medium quality lepton (pte > 15 GeV ptμ > 12 GeV) + hadronic tau ptτ > 18 (1 3 track) 33