1 Search for high-mass dilepton resonances with the ATLAS detector Sarah Heim, Michigan State University Experimental Particle Physics Seminar University of Pennsylvania, 02.28.2012
2 Overview 1. Why are we looking for high-mass dilepton resonances? 2. 1fb -1 analysis ( Phys. Rev. Lett. 107, 272002 (2011)) - event selection - backgrounds - high energy electrons and muons - signal search and limit setting 3. 5 fb -1 analysis (in preparation) - updates Sarah Heim
3 Why are we looking for Physics beyond the SM? The Standard Model of Particle Physics is a very successful theory, but cannot be the end of the story... For example, it doesn't - have a dark matter candidate - explain, why gravity is so weak compared to the other fundamental forces Also: Before we find the Higgs, we cannot be sure of how electroweak symmetry is broken Sarah Heim
4 Search for dilepton resonances (ee/μμ) l + q/g ? q/g l - Sarah Heim
5 Search for dilepton resonances Dilepton resonances have been the window to a better understanding of elementary particles and forces before... Here be dragons Sarah Heim
6 Search for dilepton resonances Dilepton resonances could be a signature of - new heavy gauge boson in the E6 model (Grand Unified Theory model) ----> spin-1 - excited Kaluza-Klein mode of the Randall Sundrum graviton ----> spin-2 ...and many others (resonance search is fairly model independent) ---------------------------------------- Benchmark model: Sequential Standard Model Z' (same couplings as Z boson), not theoretically motivated Sarah Heim
7 Additional symmetries Dilepton resonances could be a signature of - new heavy gauge boson in the E6 model (Grand Unified Theory model) (Phys. Rev. D 34 (1986), arXiv:0801.1345v3) GUT theories: Unification of electroweak and strong forces at high energies ----> 1 overall symmetry, which breaks down at lower energies 2 additional U(1) groups lead to Z' Several motivated choices of θ E6 Sarah Heim
8 Randall Sundrum graviton Dilepton resonances could be a signature of - excited Kaluza-Klein mode of the Randall Sundrum graviton (arXiv:hep-ph/9905221v1) 2 parameters: 1 finite warped extra dimension, 2 branes - m 1 (first excitation) - k (curvature) Planck Standard brane brane G(x) 1 extra dimension 0.1 Only gravitons propagate to extra brane ----> wave functions are suppressed away from extra brane ----> gravity is weak! Finite extra dimension ----> excitation like in harmonic oscillator possible ----> Kaluza-Klein tower of massive graviton states Sarah Heim
9 1 fb -1 analysis
10 Event Selection – electrons - ATLAS data quality (stable beam, functioning subdetectors etc.) - pick two standard electrons, central (η < 2.47), E T > 25 GeV - require track and EM shower shape cuts (“Medium”) - require hit in innermost layer of detector - leading electron should be isolated (calorimeter based) ---> energy in a cone of 0.2 around the electron < 7 GeV - form invariant mass, require mass > 70 GeV Main background: Z/γ* Acceptance (Z', 1.5 TeV): 67% Sarah Heim
11 Event Selection – muons - ATLAS data quality (stable beam, functioning subdetectors etc.) - pick two standard combined muons, p T > 25 GeV - require 3 hits in 3 muon spectrometer layers, no overlap barrel-endcap, veto misaligned chambers - distance from primary vertex needs to be small - require tracks to be isolated - muons must have opposite charge - form invariant mass, require mass > 70 GeV Main background: Z/γ* Acceptance (Z', 1.5 TeV): 42%
12 Backgrounds Backgrounds with two prompt electrons/muons: - Drell Yan - WW,WZ,ZZ - ttbar (dileptonic decay) Backgrounds with QCD jets, which can fake prompt leptons - W+jets - QCD multijet production ----> what is the fake rate at high energies? Drell Yan is dominant background by far, all background except for QCD multijet taken from simulated samples Sarah Heim
13 QCD multijet background electrons Jets can fake electrons. How large is the fake rate? 1 . Baseline Method ►Reverse Identification - dijet shape from reverted electron identification cuts - extrapolation to high invariant masses by fitting with empirical function - normalization by 2-component template fit 2. Cross-check and systematic uncertainties ►Fake rate estimate ►Isolation fit method - measure probability for jet-like - use calorimeter isolation distributions objects to pass Z' selection (η, E T ) - fit signal/background templates from data for 1 st and 2 nd electron - apply fake rate on normalization - system of equations sample (Z' selection on leading, to avoid double counting jet selection on second electron)
14 QCD multijet background muons QCD multijet background much smaller for muons 1. Shape: Anti-track-isolated data (0.1 - 1.0) 2. Normalization: Ratio of isolated (0.00-0.05) /anti-isolated (0.1-1.0) dimuon events in QCD (heavy flavor) simulated samples QCD from simulation Sarah Heim
15 High p T leptons Looking for resonances at high invariant masses ---> need to understand properties of highly energetic objects in ATLAS Very small control sample, handles: calibration runs, cosmics, Tag-and-Probe around Z pole ---> extrapolation, simulation Sarah Heim
16 High p T leptons – resolution Electrons (Resolution 1.1 – 1.8% at 1 TeV): - energy measurement from electromagnetic calorimeter - resolution at high energies dominated by constant term - improvement at high energies of linearity and resolution shown in calibration runs Muons (Resolution > 15% at 1 TeV): - pt measurement from hits in inner detector and muon spectrometer - require stringent cuts on number of hits, veto misaligned areas - measured (as a function of pt) using cosmics, magnet off runs, overlap regions, inner detector vs. muon spectrometer comparisons, Z peak Sarah Heim
17 Efficiencies and scale factors Determine trigger, reconstruction and identification efficiencies in data with Tag-and-Probe l+ - allows to get relatively unbiased control sample by applying strict cuts on “Tag” and test efficiency on “Probe” (p.ex. Z ---> l+l-) l- Z - electrons: need to subtract QCD jet background - leptons from W/Z decay: no estimate above ~200 GeV - extrapolation by observing of trends, simulation Electrons: No decrease of selection efficiency expected at high energies (careful with isolation cut) Muons: highly energetic muons occasionally radiate so much bremsstrahlung, that their tracks might be too distorted for reconstruction Sarah Heim
18 Invariant mass distributions Electrons Muons The MC and QCD estimate are normalized to data in the mass range 70-110 GeV (Normalization factor: 99% for both electron and muon channel.) Sarah Heim
19 Systematic Uncertainties Uncertainties on yield at invariant mass of 1.5 TeV: - only mass-dependent uncertainties are considered - no theoretical uncertainties on signal (by convention) (except for 5% Z boson cross-section uncertainty, which replaces the luminosity uncertainty) - uncertainties below 2% negligible (Pileup, energy calibration, momentum/energy resolution, electron trigger, reconstruction, identification efficiency, QCD multijet estimate) Sarah Heim
20 Signal templates Two templates for every tested signal mass: Z' (spin-1) - limits will be set on Z' (SSM) and motivated Z' (E6) - use shape of Sequential Standard Model Z' - neglect interference with DY - reweight flat sample (with Breit-Wigner, Parton Luminosity) G* (spin-2): - limits will be set for different couplings (0.01-0.1) 1 - use shape with largest width (0.1) 0.1 - fully simulated (in 5fb-1 analysis: flat sample) Phys.Rev. D63 (2001) 075004 Sarah Heim
21 Statistical Method – Likelihoods - compare invariant mass distribution in data with SM background and signal templates (for invariant masses above 130 GeV) - binned Poisson likelihood (invariant mass bins k): - Systematic uncertainties are considered through nuisance parameters (for which we assume Gaussian probability functions) - Reduced likelihood: Integral over all nuisance parameters - Convert N sig to cross-section Sarah Heim
22 Search for a signal - use 2D maximum likelihood fit to find most probable M Z' and σ Z' - get p-value by comparing to background-only pseudo-experiments ---> what is the percentage of pseudo experiments, which show an excess at least as significant as the one seen in data - electrons: p = 54%, muons: p = 24% (evidence: 0.1%, discovery 0.00003%) ---> No significant excess found ---> Setting limits
23 Statistical Method – Setting limits - Bayes theorem: - no prior knowledge on cross-section: - 95% confidence level (C.L.) limits can be set by finding upper edge of integral, such that: ----> with 95% C.L. we say, that σ sig B is below (σ sig B) 95 Sarah Heim
24 Limits Spin 1 - Sequential Standard Model Z' as baseline model (same couplings as the Z boson) - different E 6 models Limits on SSM Z' [TeV]: Observed Expected ee 1.70 1.70 mumu 1.61 1.61 combined 1.83 1.83 Combined limits on E 6 models [TeV]: Z' (ψ) Z' (N) Z' (η) Z' (I) Z' (S) Z' (χ) 1.49 1.52 1.54 1.56 1.60 1.64 Previous limits on SSM Z' [TeV]: Tevatron (1.071), ATLAS (1.042), CMS (1.140) LEP (indirect, 1.79) CMS 1 fb -1 [TeV]: 1.94 Sarah Heim
Recommend
More recommend
