Title: Standard model, Tracking, vertexing, btagging, taus
1Standard model,Tracking, vertexing, b-tagging,
taus
2Standard Model
- Focussed on early analysis of W and Z xsect
- W and Z xsect with electrons
- W and Z xsect with muons
- Many aspects of analysis are data-driven
- Triggering and reconstruction efficiencies from
tag and probe - Electron fake rate from QCD ET-miss spectrum
- Very tightly coupled to combined performance
groups - Will be application of standard performance
analysis eg efficiencies of isolated lepton
trigger - Some work on taus
- Discuss more
- No mention of pure QCD, this still wide open
3Interplay between groups
- Can not be exhaustive ? illustrate this
interplay through several inputs/outputs - Well operating detector (DAQoffline)
- Signal reconstruction
- Trigger (HLT pass-through at beginning)
- Alignment
- EM energy scale
- EM calo intercalibration
- Material in front of EM calo
- Trigger and identification efficiencies
- MC tuning (detector description, physics param.,
) - Background control with data
- Estimate of all systematic uncertainties
Detector (trigger)
egamma
SM group
- Not a step by step program !! Iterations needed
- Need good cooperation between communities
W/Z electron channel, F.Hubaut
4Z(?ee) extraction
- Fast and robust extraction of the signal in
early data taking phase - Trigger not discussed here (see dedicated
meeting) - Large part of initial bandwidth dedicated to
leptons, no isolation criteria
- Selection steps
- e10 trigger (single electron trigger to measure
efficiency from data, see next slide) - Kinematics 2 EM clusters pTgt15 GeV, hlt2.47,
exclude largely around crack (1.3lt h lt1.6) - Loose identification cuts robustness when
detector perf. not understood in detail.
Can even use simple criteria based on EM
calorimeter only ? unbiased tracker studies
- 24 800 200 signal events with 50 pb-1
CSC
- Large sample, stat. error lt1
- At 10 TeV, reduced by 1/3rd
- Data-driven background determination
- Fit exponential slope after kinematical cuts
? normalise on side-bands - 2300 400 events estimated for 50 pb-1
W/Z electron channel, F.Hubaut
5Z(ee) sample ? Efficiency determination with
data
More details in eg session
- Measure trigger/reconstruction/identification
efficiencies with Z(ee) data sample - Well known TagProbe method
- Single lepton trigger to allow unbiased probe
- Background contamination taken into account
Medium identification efficiency
- Reproduce differential structures
- 2 error on overall efficiency per electron with
50 pb-1
(Integrated over whole spectrum. Mainly limited
by Z sample statistics.)
W/Z electron channel, F.Hubaut
6Z(?ee) cross section measurement
s (Nsignal - Nbackground) / (A etot Lumi)
Acceptance uncertainty (mainly limited
knowledge of underlying physics ISR, PDFs, )
? determined with MC
Previous slides
- Overall uncertainty for 50 pb-1 0.8 (stat)
3.5 (syst) dL/L
- Systematic errors dominate, even with 50 pb-1
- Main systematics from electron selection
efficiency (except luminosity) - Comparable to muon channel
- Extrapolation to 1 fb-1
- estimated directly on data
- limited to 1.5 by acceptance uncertainties
(PDF, ISR, )
- use differential cross sections (vs h and pT)
W/Z electron channel, F.Hubaut
7W(?en) extraction
- Selection steps
- e20 trigger
- Electron ? 1 EM cluster with pTgt25 GeV, hlt2.47
exclude crack region - Medium identification
criterion - Missing ET ? gt25 GeV
1 electron only increase pT cut and tigthen
identification criteria
- Not discussed here, but need detailed detector
understanding
CSC
- 217 100 400 signal events with 50 pb-1
- Large sample, stat. error ltlt1
- At 10 TeV, reduced by 1/3rd
MTW (GeV)
- QCD background level and shape must be estimated
directly with data
W/Z electron channel, F.Hubaut
8Data driven background determination
- Dominant background jets
- Large uncertainties, difficult to simulate, poor
MC statistics - Must be measured directly on data
- Principle of the method
- QCD enriched sample (98) g trigger (g20)
and same kinematical EM cluster selection ?
missing ET shape parametrization - Normalise to side-band in electron sample (Zee
removed)
- Uncertainty on background contamination 4
(9200 events) with 50 pb-1
(limited by MC stat.)
W/Z electron channel, F.Hubaut
9W(?en) cross section measurement
s (Nsignal - Nbackground) / (A etot Lumi)
- Acceptance uncertainty
- only theoretical (ISR, PDFs, )
- impact of missing ET scale and resolution
uncertainties has to be quantified
- Overall uncertainty for 50 pb-1 0.2 (stat)
5 (syst) dL/L
- Systematic errors dominate largely with 50 pb-1
- main from background uncertainty (except
luminosity) - Luminosity uncertainty vanishes in s ratios,
e.g. sW/sZ - Comparable precision to muon channel (for which
background less important, Z?mm dominates) - Extrapolation to 1 fb-1
- estimated directly on data
- limited to 2.5 by acceptance uncertainties
(PDF, ISR, )
W/Z electron channel, F.Hubaut
10Hadronic taus means for identification
Tracking ? object with low track multiplicity
( 1 ?? or 3 ?? ) ? tracks more collimated
than for average jet, (invariant mass,
weighted width of tracks system) ? decay
length makes it possible to use impact parameter
and transverse flight path (three-prong)
? isolation cone from other tracks Calorimetry
? collimated deposition in EM (radius, width
in strips) use shower shape variables
? strong EM component for single prong
(50 energy by ?0) ? reconstruct ?0
subclusters ? isolation cone ? both EM
and HAD components
E.Richter-Was, UJ/IFJ-PAN ATLAS CP
Week, 9 June 2008
10
11Reconstruction
Track-seeded and calo-seeded algorithms
integrated for rel.14.2.0
A.Kaczmarska, S. Lai, N. Meyer, L. Janyst
- Track-seed and calo-seed
- - use good quality tracks (pTgt6 GeV) as initial
seed - - candidates with 1-8 quality tracks (pTgt1 GeV)
in ?Rlt0.2 from the seed - - the ?,f using pT weighting of tracks, check
charge consistency (Q 2) - find matching cone 0.4 TopoJets (gt10 GeV, DR lt
0.2) as calo-seed - ET (calorimetric) using H1-style calibration on
cells from calo-seed - ETeflow with energy-flow method (EM calo -
separating neutral/charged sources of energy) - - reconstruct p0 subclusters
- Calo seed only
- - use cone 0.4 TopoJets (gt10 GeV) as calo-seed,
matching seed not found from tracking - define the ?,f using calo-seed (h corrected for
z vertex), - looser tracks-quality selection, track pTgt 1
GeV - - ET (calorimetric) using H1-style calibration
on cells from calo-seed
Track-seed only small fraction (few ) of
total only
E.Richter-Was, UJ/IFJ-PAN ATLAS CP
Week, 9 June 2008
11
12Reconstruction
Only calo-seed
Both seeds
QCD J2 pThard35-70 GeV
Only track-seed
QCD J2 sample pT35-70 GeV
Z? tt
Overall purity in the sample Z-gttt 57 for
both seeds (yellow),
23 for only calo
seeds (red)
E.Richter-Was, UJ/IFJ-PAN ATLAS CP
Week, 9 June 2008
12
13(No Transcript)
14Tracking-1
- Many new changes in release 14
- Improved tuning of tracking following CSC
- New functionality BackTracking, TRTOnly,
ConversionFinder, V0Finder, LowPtTracking - Geometry updates
- Detector condition information introduced
- Tracking for startup single-beam, beam-halo
Tracking summary M.Elsing
15Tracking-2
- Handling real data
- Noisy SCT and TRT modules
- Condition service introduced, allows bad
channel/module masking - This links in with monitoring work from Saverio,
Dan, Aidan and Mary - Running in different configurations
PixelSCTTRT or SCTTRT - Need tuning
Tracking summary M.Elsing
16Tracking-3
- Mass resolution after alignment is a worry
- Z-mass is sensitive to weak eigenmodes
Tracking summary M.Elsing
17Vertexing
- Primary vertex and beamspot
- How to use beamspot to find PV and then PV find
BS - Pileup
- Vertex code can handle pileup
- But more testing required, eg identification of
correct PV for b-tagging, identification of
correct PV in high levels of pileup - Technical issues
- What to do?
- Study pileup in minbias Craig
- Study reconstruction of PV in pileup for
b-tagging --? - Maybe using a top sample?
Vertex summary A.Wildauer
18B-tagging at startup
- Avoid using PDFs
- Related to PV finding discussed earlier
b-tagging summary L.Vacavant
19Commissioning
- Once we have some understanding of
tracking/alignment (previous talks) - First taggers
- Track counting no calibration
- JetProb negative d0/s from data
- SV0 simple inclusive secondary vertex
- taggers relying on LR for b,u(,c) hypotheses next
- Simple baseline switch on progressively the
extra features - V0 rejection
- Dedicated treatment for shared tracks, other
categories - Samples
- min.bias, QCD resolution function for JetProb
- QCD, bbbar JetProb, SV0
- muonjet b-tagging efficiency measured in data
- ttbar
Monitor with jet events
b-tagging summary L.Vacavant
20Early jet taggers
- DCP with BS using jets
- Look at track jets
- Optimise tracks for use in this
- Introduce PV, remove V0s and conversions
- Secondary vertex taggers
- Introduce L3D/sigma, no pdf used
- Very strong links with tracking groups
- Optimise track selection, tracking performance in
jets, vertex reconstruction
b-tagging summary L.Vacavant
21What to do?
- Kenny
- Top in dileptons
- Start looking at robust early tagger jetprob
- Mary
- SV0 and then SV1 etc
- Saverio will talk to Richard Hawkings
- Tracking interface via Craig and Will
22TODO List for taus (I)
E.Richter-Was, UJ/IFJ-PAN ATLAS CP
Week, 9 June 2008
22
23TODO List for taus (II)
E.Richter-Was, UJ/IFJ-PAN ATLAS CP
Week, 9 June 2008
23
24TODO List for taus (III)
E.Richter-Was, UJ/IFJ-PAN ATLAS CP
Week, 9 June 2008
24