ATLAS Experiment Université de Genève Diboson resonance search in the all-hadronic final state On behalf of the ATLAS collaboration Moriond 16-23 March 2019 Sofia ADORNI BRACCESI CHIASSI
ATLAS Experiment Université de Genève VV ➜ JJ physics motivation ✸ Searching for heavy resonances decaying in WW, ZZ or WZ ➜ sensitive to BSM physics (spin-0/1/2 resonances) ✸ Why focus on fully hadronic decay products? High sensitivity in the high • mass regime (BR(W → qq) ≃ 3 x BR(W → l 𝛏 ), BR(Z → qq) ≃ 10 x BR(Z → ll)) Probe all decay modes in one • analysis Sensitive to the unexpected • Large-R (generic bump hunt) jets Sofia ADORNI BRACCESI CHIASSI 1
ATLAS Experiment Université de Genève Track-CaloClusters ✸ At high p T the hadronic decays of the W/Z are very collimated (reaching the granularity limits of the calorimeter) 0.7 resolution ATLAS Simulation Preliminary ✸ Track-CaloClusters (TCC) (reference) are a type of 0.6 s = 13 TeV anti k R=1.0, WZ qqqq → T particle flow designed for the high energy regime: jet jet 0.5 2 | |<2.0, p >200 GeV η jet D T • The calorimeter has better energy resolution but poor 0.4 Fractional spatial resolution 0.3 • The tracker has poorer transverse momentum (p T ) 0.2 resolution but good angular resolution LC Topo 0.1 ➜ use tracker angles and calorimeter energy scale TCCs 0 very roughly : TCC 4-vec = (p Tcalo , 𝛉 track , 𝛠 track , E calo ) 500 1000 1500 2000 2500 Generated jet p [GeV] T Sofia ADORNI BRACCESI CHIASSI 2
ATLAS Experiment Université de Genève New W/Z tagger ✸ Fully hadronic mode = large QCD dijet background ➜ use Jet SubStructure (JSS) to distinguish background and signal ( mass, D 2 and n trk ) ➜ JSS variables are more powerful with TCC jets ✸ New tagger optimised for best analysis significance ➜ novel tagging strategy replaces previous fixed efficiency tagger, non-optimal for analyses • 1000 Background rejection W-tagging efficiency 1.4 ATLAS Simulation Preliminary ATLAS Simulation Preliminary Mass efficiency cut 900 D2 efficiency cut s =13 TeV s=13 TeV nTrk efficiency cut 800 1.2 Total efficiency 700 1 600 0.8 500 0.6 400 300 0.4 200 0.2 100 0 0 0.5 1 1.5 2 2.5 3 3.5 4 0.5 1 1.5 2 2.5 3 3.5 4 Jet p Jet p [TeV] [TeV] [TeV] T T Sofia ADORNI BRACCESI CHIASSI 3
ATLAS Experiment Université de Genève ✸ No significant excess in any of the observed channels Results and ✸ WW + WZ: HVT model B (A) excluded up to 4.4 (4.1) TeV ✸ WW + WZ: Radion excluded up to 3.2 TeV exclusion limits ✸ WW + ZZ: Bulk RS excluded up to 2.8 TeV WW + WZ SR WW + ZZ SR 4 4 10 10 Events / 0.1 TeV Events / 0.1 TeV ATLAS Preliminary ATLAS Preliminary Data Data 3 3 10 10 Fit Fit s = 13 TeV, 1 39 fb -1 s = 13 TeV, 1 39 fb -1 Fit + HVT model A m=2.0 TeV Fit + Bulk RS m=1.5 TeV 2 2 10 10 Fit + HVT model A m=3.5 TeV Fit + Bulk RS m=2.6 TeV 10 10 1 1 1 1 10 − 10 − ZZ or WW SR WZ or WW SR 2 2 − − 10 10 2 /DOF = 3.1/3 2 /DOF = 6.0/4 χ χ 3 3 − − 10 10 Significance Significance 2 2 0 0 2 2 − − 1.5 2 2.5 3 3.5 4 4.5 5 1.5 2 2.5 3 3.5 4 4.5 5 m [TeV] m [TeV] JJ JJ Sofia ADORNI BRACCESI CHIASSI 4
ATLAS Experiment Université de Genève 3 Conclusions 10 WW+WZ) [fb] ATLAS Preliminary VV qqqq → s = 13 TeV ✸ We didn’t observe any significant excess -1 2 Phys. Lett. B 777 (2018) 91 (36.7 fb ) 10 ✸ Improvement in sensitivity equivalent -1 Phys. Lett. B 777 (2018) 91 (Scaled to 139 fb ) -1 Current Result (139 fb ) to redoing the 36.7 fb -1 study on entire → B(V’ HL-LHC dataset of ~3000 fb -1 ! HVT V’ → WW + WZ 10 ✸ This very large improvement is due to the × V’) combination of two major innovations x4 1 → • Use of Track-CaloClusters as inputs to (pp x2 jet reconstruction x4 σ • Use of new tagger (optimised for 1 − 10 significance + use of N trk variable) 1.5 2 2.5 3 3.5 4 4.5 5 m(V’) [TeV] ✸ The gain in sensitivity observed goes well beyond statistics : this analysis really shows the potential of new methods for reconstruction, tagging and statistics analysis Sofia ADORNI BRACCESI CHIASSI 5
ATLAS Experiment Université de Genève BACKUP Sofia ADORNI BRACCESI CHIASSI
ATLAS Experiment Université de Genève Event selection Trigger Trigger on lowest unprescaled large-R jet trigger (year-by-year) Quality GRL, DQ checks, jet cleaning Veto events with leptons of p T > 25 GeV and | 𝛉 | < 2.5 Leptons Leading p T > 500 GeV (for trigger), subleading p T > 200 GeV (for calibration), Jet Kinematics m > 50 GeV (for calibration), (TCC jets) | 𝛉 | < 2.0 (for tracks) m JJ > 1.3 TeV Trigger fully efficient at 1.3 TeV (for background) m JJ < 7.0 TeV Upper limit fixed by common range with other analyses for heavy resonance combination | 𝞔 y| < 1.2 Reducing t-channel QCD jet pair production Jet p T asym < 0.15 Signal is balanced W/Z selections defined as X < m J < Y, D 2 < Z, n trk < K Boson tagging Reduces QCD background ~5 orders of magnitude W/Z mass windows overlap ➜ signal regions are not orthogonal Sofia ADORNI BRACCESI CHIASSI
ATLAS Experiment Université de Genève Systematic uncertainties Background fit normalisation and shape (~25% (100%) at 3(5) TeV) Main uncertainties Boson-tagging signal efficiency (~25%) Jet p T scale (JP T S) ~5% ISR - FSR (3% for HVT, 5% for RSG) PDF (1%, up to 12% for HVT) Remaining Luminosity scale (2.1 %) uncertainties JP T R (<1%) Tagging and jet uncertainties approved by the JSS group Sofia ADORNI BRACCESI CHIASSI
ATLAS Experiment Université de Genève Boson tagging SF Events / 5 GeV ATLAS Preliminary Data -1 using V+jets s =13 TeV, 139 fb Fit V+jets control region 20000 Fit bkd. W/Z+jets W+jets ✸ The boson tagging efficiency is Z+jets evaluated in data enriched W/Z+jet events 10000 ✸ We tag one jet (leading or subleading) and anti-(D2)tag the other one Fitted W/Z+jet events: 17112 777 ± ✸ We fit the distribution we get using = 0.92 0.04 ± s Tag a signal+background function 0 ✸ We obtain the scale factor and the 60 80 100 120 140 160 180 200 m [GeV] uncertainty from differences between data and MC J • S Tag = 0.92 ± 0.04 (stat) ± 0.02 (closure) ± 0.03 (tt) ± 0.02 (fit) ± 0.05 (high pT) ± 0.1 (theory) = 0.92 ± 0.13 Sofia ADORNI BRACCESI CHIASSI
ATLAS Experiment Université de Genève Background modelling - ABCD method 4 10 Events / 0.1 TeV ATLAS Preliminary 3 ✸ The fit range is from 1.3 TeV Data 10 s = 13 TeV, 1 39 fb -1 2 to 8 TeV 10 Fit 10 ✸ Validation was done in a 1 dedicated control region 1 − 10 created with the ABCD WZ CR 2 − 2 /DOF = 3.9/5 10 χ method and parametrised 3 − 10 tagging efficiencies of QCD Significance 2 Fit was able to describe the expected • 0 m JJ spectra in all fit control region 2 − ✸ The behaviour of the fit at high masses 1.5 2 2.5 3 3.5 4 4.5 5 m [TeV] was checked and we are confident that JJ the extension of the fit range is valid Sofia ADORNI BRACCESI CHIASSI
ATLAS Experiment Université de Genève Background modelling ✸ To perform the bump hunt we first have to fit the background We use a parametric function: • m JJ dn with dx = p 1 (1 − x ) p 2 − ξ p 3 x − p 3 x = 13[ Tev ] ✸ The choice of the fit function could possibly have an impact on the analysis Comparison of two working fit functions on pseudo experiments showed significant effect only at high • mass (where limiting factor is the lack of statistics) ✸ The difference observed for this specific choice is significantly smaller than the uncertainty on the fit ➜ neglected in statistical treatment ✸ Why extend the range? From fit up to 6 TeV (limits up to 5 TeV) ➜ fit up to 8 TeV (limits up to 7 TeV) • ✸ We are preparing for the full Run2 “grand combination” Background is understood • Very large uncertainties: expectation is << 1 in this regime with 100% uncertainty • ✸ Comparison of mean and spread between data and MC gives us confidence that they are well modelled. Sofia ADORNI BRACCESI CHIASSI
Recommend
More recommend
