1. W/Z+Jets production studies in ATLAS
Ellie Dobson
(University of Oxford)
On behalf of ATLAS and the W/Z+Jets CSC note group
DIS 2008
Why study W/Z+jet events?
How do we see such events in ATLAS?
Summary of current MC studies
Ellie Dobson

2. Why study W/Z+jets? Meff Events Detector
10 fb1
Detector
Tests lepton reconstruction & MEt in multi jet environment
Precision tests of jet reconstruction algorithms and techniques
Standard model physics
Testing ground of pQCD
Testing LO/NLO predictions
Tests of BFKL logarithms
PDF studies
Beyond the standard model physics
Background to many BSM searches
Events must be well understood before we start looking for new physics!
Ellie Dobson

3. W/Z+jets production in ATLAS
Hard jets
Zooming in….. g
PDF1 u
Hard Scatter
Underlying event
(low pt hadronic recoil) d W
PDF2
Example of W+1jet production
W/Z
LHC will be a ‘W and Z factory’
Early running will produce ~1fb-1 of data:
~20 million Ws, ~2 million Zs that can be seen (electron decay)
~1/3 of these produced in association with jets….
Lepton
MET/second lepton
Health warning! Back of the envelope calculation
Ellie Dobson

4. W/Z+jets reconstruction in ATLAS
Muon trigger used to write Z→μμ and W→μν events to disk
Jets reconstructed in the calorimeters. Jets are built from calorimeter towers which are ‘H1 style’ calibrated to hadron level)
Muons reconstructed as a combination of a track in the inner detector and the muon spectrometer
Met and SumPt determined by summing over calorimeter cells.
More atlas specific
Electrons reconstructed as a combination of a track in the inner detector and the energy deposit in the EM calorimeter
Electron trigger system used to write Z→ee and W→eν events to disk
Ellie Dobson

5. Current activities
CSC (Computer Systems Commissioning) activity currently underway in ATLAS to produce public notes on performance and analysis
This talk summarises the activities of the W/Z+jets CSC note
The work in the W/Z inclusive CSC note is also relevant
Electronic and muonic decay channels have been studied
The focus at the moment is on understanding the detector: current activities include:
Put picture of first page of note here
Leptons (ID, efficiency, background)
Methods of background subtraction
Jets (reconstruction, algorithms, bjets)
Unfolding corrections
Evaluating uncertainties
Ellie Dobson

6. MC simulation and event selection
W/Z+jets sample (used for main analysis)
Matched Alpgen/Herwig samples (using LO PDF set CTEQ6LL), with generator level filter requiring 1 truth cone04 jet and 1(2) electrons/muons within fiducial requirements
Offline procedure (default)
Z EVENTS
Mass cuts
2 leptons
W EVENTS
1xMet
1 lepton
BG and comparison samples
Pythia and Alpgen W, Z, dijet and other background samples
Renormalisation scale of Mz2+Ptz2 used
Reference cross sections calculated at NLO (where possible) using MCFM (generated using CTEQ6.1 PDFs)
LEPTON
Pt > 25/20/15GeV
Fiducial η cuts
Track matching
Inner detector track
Isolation cut from jets
Deposit of appropriate shape in calorimeter (electrons)
Hits in muon spectrometer (muons)
JET
Pt>20 or 40 GeV
Fiducial rapidity cuts
Isolation from leptons
Seeded cone 04 algorithm used
MET
Met>25GeV (W events)
Ellie Dobson

7. Correcting from parton-hadron level
Data measurements compared to theory predictions at hadron level
→ need to correct theory with respect to fragmentation and underlying event
Determine corrections by comparing Pythia hadron level results with non perturbative effects switched on/off
Underlying event and fragmentation have the opposite effect
Precise behaviour depends on the jet algorithm used
Can check underlying event in data using minimum bias events
Pythia gives a better description of the underlying event at the Tevatron
In general the underlying event wins
Effects get smaller at higher energy because underlying event is essentially constant but Pt jet increases
Fragmentation effect becomes smaller because jet becomes more collimated
Fragmentation corrections for cone07 smaller than for cone04 whereas underlying event corrections are larger
Performance of kt06 comparable to cone04
Underlying event adds energy to the hadron level jet
Fragmentation reduces the amount of energy in the jet cone
Negligible effect for higher energy jets
Ellie Dobson

8. Trigger and reco efficiencies (Z→ee)
Single electron trigger (e25i) used to select Z → ee and W → eν events
e25i efficiency determined using a ‘tag and probe’ data driven method (although results shown are obviously ‘pseudo data’ for the time being….)
The numbers may be used to determine Z → ee or W → eν cross section
Sufficient in inclusive study to consider the efficiency as a function of η and pt only
In these more ‘jetty’ events is important to study the efficiency as a function of jet variables (distance to closest jet, hadronic activity, jet multiplicity….)
Global trigger efficiency ~1.5% lower than in the Pythia inclusive sample
Trigger efficiency
Reconstruction efficiency
Fall with higher hadronic
activity due to the implicit
isolation cut in the level 1 trigger
Ellie Dobson

9. Unfolding of detector effects (Z→ee)
Need to unfold data (jet and electron resolutions and reconstruction efficiencies) from detector level to hadron level
Will eventually determine jet resolution and efficiency from real data
Bias in reconstruction at low jet pt
For jet resolution used a global average correction. Effect is worse in tevatron experiments
The idea is to set up the machinery for making cross sections with data- test for perturbative QCD
Jet resolution and efficiency from real data possible- dijet matching in different planes
Make graphs by truth-reco matching in pt and dR (0.2 and )
At higher Pt deviations could suggest quark compositeness
BFKL logarithms can show up in these plots
Can probe these jets up to the multi TeV range
No global behaviour seen
Within errors, the
Pt distributions of truth and
corrected reconstructed jets agree
Ellie Dobson

10. BG estimation and subtraction (Z)
Z-ττ, ttbar and W → lν are, for now, estimated and subtracted using MC estimates
Weighting factors used for QCD due to the enormous cross section of this process
Fits may be used to estimate QCD backgrounds in real data
Can also derive this background by inverting selected electron ID cuts
Muons coming from background (particularly bbar) can be rejected using isolation requirements
QCD background dominates at a 1 jet signal. At higher jet multiplicity ttbar dominates.
Mass cuts imposed to reduce background
I use a JF17 sample (3Mio evts) where all W,Z,TTbar are already discarded in order to avoid double-counting.
In the signal and background plots I do the complete electron ID for all samples except for JF17 where I only require two AOD electron candidates. The events are then weighted by the additional rejection from the isEM medium and the trigger conditions.
This additional rejection is derived using JF17 events with single AOD-electron-candidates, which fulfil all kinematic requirements.
I derive a rejection separately for AOD-electron-candidates which don't match with a truth electron (actually the majority) and for AOD-electron candidates which match with a truth electron (probably mostly from B-decays).
> Do you have any plans to use data driven estimates for the other processes?
Yes, in particular for the TTbar. The most difficult part is to separate the Ttbar from the QCD because we expect a different PT distribution for TTbar and for QCD. A few ideas: Loosening the electron ID would increase the QCD but not the TTbar, requiring low Etmiss will decrease the TTbar but the QCD will stay the same, whereas requiring a minimum ETmiss will decrease the QCD but the Ttbar will stay the same.
Ellie Dobson

11. BG estimation and subtraction (W)
QCD
Zee
Weν
QCD largest background
ttbar dominates at high Njets
W→τν and Z →ee estimated using MC
Pythia underestimates QCD
QCD can be estimated from data using photon trigger and normalising to electron spectrum in sideband
QCD also can be estimated using fake rates
ttbar
W→τν
Ellie Dobson

12. Jet energy scale (JES) uncertainties
Dominant experimental systematic for W/Z+jets at the LHC
Miscalibration of ±1, 3, 5% assumed on the jet energy and effect on jet pt calculated
Main effect on the cross section results from the selection cut on the jet pt
Within early running (1fb-1) we hope to be within 3% JES uncertainty
→ systematic on the cross section of up to 10%
JES uncertainty of 1% (ultimate ATLAS goal) yields systematic of 0.5%
W events
Uncertainties increase with jet multiplicity
Ellie Dobson

13. PDF uncertainty studies
Must get these right or SM physics could be misinterpreted as new physics!
PDF uncertainties often the dominant theoretical uncertainty
PDF forms determined from combination of theoretical calculations and data
Free parameters of fit are given in a PDF set with their upper and lower errors
Overall PDF uncertainties calculated with PDF reweighting on NLO CTEQ6M set
PDF uncertainty ≤ JES uncertainty
Uncertainties always remain <10%
Hessian method used
PDF reweighting used to avoid generating two MC samples for each free parameter in the global fit.
Apply an event to the hard process- probability of picking up the same partons with the same momentum fractions
Shown results for W→eν.
Similar results seen in the muons and Z analysis
PDF errors increase at central η and at low electron Pt (no y dependence seen)
Ellie Dobson

14. Cross section results (Z→ee)
MCFM predictions corrected (for underlying event and fragmentation) to hadron level
Data unfolded to hadron level
PDF and JES uncertainties are included in the error determination
Cross sections quoted wrt inclusive to factor out luminosity uncertainty in early data
Can scale the MC samples globally with (inclusive cross section predicted with the MC sample - inclusive cross section measured in data)
- only interested if the XS fraction of the subsamples with >=N jets is reproduced correctly for detector validation
Pythia predicts a larger inclusive cross section for >=1jet
Results normalised to NLO inclusive (DrellYann) MCFM cross section
Pythia predicts a softer pt spectrum
Pythia discrepancies due to tuning of leading soft radiation in parton shower
Ellie Dobson

15. XS (Z →μμ) Data unfolded to hadron level as before
Choice of isolation cone size must be optimised so to balance muon reconstruction efficiency and background rejection
Cross sections normalised to NLO
MCFM predictions corrected and compared to unfolded reconstructed events
Unfolded distribution
- Count more jets in Pythia if you do soft cuts on the jet PT.
- For Cone 20- isolation there is +- no efficiency loss.
Pytha predicts more low pt jets
Pythia parton
shower predicts softer jet pt distribution
Alpgen predicts more high pt jets
Ellie Dobson

16. Z→μμ+bjets
A cross section measurement in this channel will provide an important test of pQCD
Will reduce the current uncertainty on the partonic heavy flavour content
This channel will be a lot easier at the LHC than at the Tevatron (higher production cross section, smaller background from Z+cjet)
btagging: cutting on the weight parameter (secondary vertex and impact parameter)
Z+cjet background depends on the sea quark distributions.
gluon distributions dominate at the LHC (gb→Zb will dominate over qqbar → Zbbar)
Thus the sea quark contributions are relatively suppressed.
At the Tevatron, the sea quark contributions can contribute more strongly
Contamination from light jets of the order 30%
Ellie Dobson

17. Conclusions
W/Z+jets used as a standard candle to understand the detector
Understanding necessary in SM and BSM sectors
Specific findings:
Efficiencies ~1% lower than the inclusive due to hadronic isolation
Cuts developed so that signals are larger than background sum
Tools for estimating the dominant backgrounds from data developed
Two main non perturbative effects neglible above a jet Pt of 40GeV
Comparisons made between theory (corrected to the hadron level) and data (unfolded from detector level) are in agreement
JES dominant systematic. Will be reduced to ~3% after 2 years
This is larger than dominant theoretical uncertainty (PDF uncertainties)
Ellie Dobson