B tag efficiency measurement

Download Report

Transcript B tag efficiency measurement

B Tagging Efficiency and
Mistag Rate Measurement in ATLAS
Thomas Delemontex (LPSC)
Jin Wang (Shandong University && LPSC)
Objective and Motivation
B Jet Identification
B Tagging Efficiency
B Tagging Efficiency
Light Jet Mistag Rate
Light Jet Mistag Rate
W+jets events
QCD events
W+jets events
QCD events
B-tagging Algorithm
• B-tagging algorithms are used to tag the b-jets originating from b quarks.
In first data, three algorithms are going to be used:
– TrackCounting: use tracks with large impact parameter.
– JetProb: calculate the probability of a track to be from light jet.
– Secondary Vertex Reconstruction (SV0): use the signed decay length
significance of the secondary vertex .
Definition
• Each tagger provides a tag weight w. On this w distribution of all
reconstructed jets, we cut at certain threshold w , and define :
cut
– B-tagging efficiency  b :
b 
number of b jets above wcut
total number of b jets
– Mistag rate  l :
l 
number of light jets above wcut
total number of light jets
– Light jet rejection :
rejection = 1 /  l
B tag efficiency measurement : pTRel method (1/2)
PTRel
Definition : for a muonic jet, it’s the transverse
momentum of the muon compared to the jet axis
CSC Note: CERN-OPEN-2008-020
Jet axis
Soft
muon PT
Form CSC note
θμ-jet
Primary
vertex
pTRel is sensitive to jet flavors
B tag efficiency measurement : pTRel method (2/2)
Simply a fit of an unknown pTRel
distribution with previous templates
Form CSC note
Jet sample without any btag
requirement
Adjustment of template normalizations
We find nb jets in this sample
Same sample with a btag
requirement
Form CSC note
We find ntaggedb jets in this sample
B tagging efficiency is
a simple ratio
b 
tagged
N bjet
N bjet
pTRel method : Results
Form CSC note
- pTrel method is OK up to 80 GeV and then becomes less efficient due to
similarities between light, c and b templates
- Soon dominated by systematic uncertainties (fragmentation function used,
muon pT measurement , misalignment …)
- Can be used with ~ 20pb-1
B tag efficiency measurement : System8 method
Use of 3 uncorrelated taggers :
J = b tagging algorithm we want to calibrate  B tagging efficiency in the system = εb(J)
M = muon tagger (pTrel), B = presence of an opposite b tagged jet (enhance bbar events)
• Q(X,Y) fractions of jet that
survive the X and Y tagger
requirement in data
• εX efficiency for the X
tagger (unknowns)
• Κ(X,Y) correlation between X
and Y taggers (found in MC)
8 equations and 8 unknowns (εX , fb and fcl) = system fully constraint
System8 method results
Advantages compared to the pTrel method :
- Reduce dependences in Monte Carlo
- Systematics clearly defined : error on correlation factors Κ(X,Y)
Problems :
- Need more statistics than the pTrel method
- Still some problems at high jet pT (use of soft muon tagger pTrel )
Form CSC note
Mistag Rate Measurement: Negative Tag Method
System
Test of Negative Tag Method in MC
Test Results of Negative Tag Method
Obtain Mistag Rate Matrix from Data
Test Results of Mistag Rate Matrix
Real Application in Background Estimation
Real Application in Background Estimation
Summary