Search for Narrow Search for Narrow Resonance Decaying to Resonance Decaying to Muon Pairs in 2.3 fbMuon Pairs in 2.3 fb-1-1 Chris Hays1, Ashutosh Kotwal2, Ye
Li3, Oliver Stelzer-Chilton1
1 Oxford University2 Duke University
3 University of Wisconsin-Madison
APS April Meeting, St. Louis - 14 April 2008 2Ye Li
MotivationMotivation Theory Driven
Standard Model successful but incompleteStrong discovery potential in dimuon
channelNew models predict narrow neutral
resonance, e.g. • additional U(1) symmetry: Z’ • extra space-time dimension: Randall-Sundrum
gravitonThe present analysis focuses on Z’ →
channel
APS April Meeting, St. Louis - 14 April 2008 3Ye Li
MotivationMotivation Experiment Driven
Last CDF and DØ dimuon resonance searches performed with integrated luminosity 200 pb-1
→ Our search: L ≈ 2.3 fb-1 of CDF Run II dataSignificant increase of sensitivity to
dielectron and diphoton channelsExcellent tracking resolution (Central Outer
Tracker, Drift Chamber)
APS April Meeting, St. Louis - 14 April 2008 4Ye Li
MethodologyMethodology Model Drell-Yan background and
signal resonance with PYTHIA + fast simulation for W mass measurement
Use Z region for normalization Remove uncertainty on luminosityEasy accounting
Compare CDF fast simulation (FastSim) to full Geant simulation (CDFSim) and data for acceptance and efficiency study
APS April Meeting, St. Louis - 14 April 2008 5Ye Li
MethodologyMethodology Inverse Mass (1/m) Scan
Excellent angular resolution → negligibleTrack curvature (~1/PT) resolution constant
for high PT → constant 1/m resonance width
1/m ≈ 0.16/TeV
APS April Meeting, St. Louis - 14 April 2008 6Ye Li
MethodologyMethodology
Fit for NZ’ (number of Z’ candidate)Calculate Binned Poisson likelihood
L(NZ’;MZ’) for region 1/m < 10/TeVConstruct the narrowest possible interval in
NZ’ at 95% C.L. Scan 1/m spectrum for Z’
resonanceUse Monte-Carlo Pseudo-experiments to
determine the significance
APS April Meeting, St. Louis - 14 April 2008 7Ye Li
Dataset & SelectionDataset & Selection Dataset from high PT muon trigger The dimuon event selection
The muon identification requirement • EM energy cut
tuned for high efficiency of Z
• High identification efficiency ~ 95%
APS April Meeting, St. Louis - 14 April 2008 8Ye Li
EfficiencyEfficiency Mass dependence
Assume track and muon-hit cuts independent of mass
Momentum dependence Only consider P dependence, due to the
normalization of background expectationAssume no P dependence of trigger
efficiency for PT > 30 GeV Separate the sample into signal
and normalization (Z-pole) regions
APS April Meeting, St. Louis - 14 April 2008 9Ye Li
EfficiencyEfficiency EM and Hadronic Cut Efficiency
Signal region: constant ratio between FastSim and CDFSim (no inefficiency of Had cut for FastSim → 2% const. offset)
Z-pole region: ratio between FastSim and Data drops at low P (due to incomplete modeling)
• insufficient data for signal region• compute uncertainty from data-simulation difference
APS April Meeting, St. Louis - 14 April 2008 10Ye Li
AcceptanceAcceptance Implement detector Geometric
information on FastSimMap angular distribution of CDFSim to
FastSim; W → data and FastSim agree reasonably
Muon for 0.6 < || < 1.0 (CMX)
Muon for || < 0.6 (CMUP)
APS April Meeting, St. Louis - 14 April 2008 11Ye Li
AcceptanceAcceptance Mass-dependent Acceptance
Larger mass → Lower boost → More central events → Larger acceptance
Constant Ratio between FastSim and CDFSim
→ Validate acceptance calculation from FastSim
Uncertainty from the small slope of the ratio
APS April Meeting, St. Louis - 14 April 2008 12Ye Li
BackgroundBackground Drell-Yan */Z →
PYTHIA + FastSim WW and tt-bar
CDF Simulation (PYTHIA + CDFSim) Cosmic Rays
Identified Cosmic-ray samples QCD Jets and Decays-in-Flight
Data
APS April Meeting, St. Louis - 14 April 2008 13Ye Li
Drell-YanDrell-Yan Dominant source for background Mass spectrum affected by higher-
order correctionsCalculate up to next-to-next-to leading order
(NNLO) correction → k-factorDifferent Calculations give different k-
factorsAverage k-factor; Difference as uncertainty
APS April Meeting, St. Louis - 14 April 2008 14Ye Li
Drell-YanDrell-Yan
• The Stirling and Hamburg, van Neervan and Matsuura (HNM) calculations of the k-factor
• About 6% difference ( ~3% systematic uncertainty)
APS April Meeting, St. Louis - 14 April 2008 15Ye Li
WW tt & Cosmic RayWW tt & Cosmic Ray WW, tt → + missing ET : Simulate
PYTHIA samples using CDFSim to compute background
Cosmic Ray : Use timing information of Drift Chamber to estimateBackground fraction
~ 1.2 X 10-6
APS April Meeting, St. Louis - 14 April 2008 16Ye Li
QCD & DIFQCD & DIF Assumtions
QCD jets faking muons: same-sign dimuon (SS) and Opposite-sign
dimuon background (OS) distribution have similar shape, i.e. constant OS/SS ratio
Decay-in-flight muons:flat distribution of DIF muons at small curvature (high PT → small 1/m)
Track 2 cut reduces DIF events Same-sign samples contains both
jet fakes and decays-in-flight
APS April Meeting, St. Louis - 14 April 2008 17Ye Li
QCD & DIFQCD & DIF
• SS dimuon obtained from jet triggered data • SS dimuon obtained from signal dataset, with 2 cut removed• SS dimuon obtained from signal dataset, with 2 cut on
APS April Meeting, St. Louis - 14 April 2008 18Ye Li
Other IssuesOther Issues Momentum Scale & Resolution
Momentum scale measurement done by fitting Z peak using templates made with FastSim
Resolution tuned on the width of the Z peak Systematic Uncertainties
Dominant uncertainties:• Parton distribution functions• Mass-dependent of the NNLO k-factor
Other uncertainties:• Arise from PT-dependent acceptance and efficiency• Affect the signal and background prediction at high
mass
APS April Meeting, St. Louis - 14 April 2008 19Ye Li
Signal ScanSignal Scan Pseudo-experiment: Standard
Model process
APS April Meeting, St. Louis - 14 April 2008 20Ye Li
Signal ScanSignal Scan Pseudo-experiment: MZ’ = 250 GeV
APS April Meeting, St. Louis - 14 April 2008 21Ye Li
Signal ScanSignal Scan Expected limits on NZ’ from 1000
pseudo-experiments on 50 Z’ masses
Data: to be implemented …
APS April Meeting, St. Louis - 14 April 2008 22Ye Li
SummarySummary Use 1/m distribution for constant
resolution Fitter and Simulation in place to study
signal acceptance and identification efficiency
Analysis on different background fractions Systematic uncertainties to be determined Signal scan performed on pseudo-
experiments