Transcript Document

Commissioning ATLAS with top events
W. Verkerke
Wouter Verkerke, NIKHEF
Introduction to physics commissioning
• What are we going to do with the first month of data?
– Many detector-level checks (tracking, calorimetry etc)
– Try to see large cross section known physics signals
– But to ultimately get to interesting physics, also need to calibrate
many higher level reconstruction concepts such as jet energy scales,
b-tagging and missing energy
• Algorithms benefiting from early data for calibration include
– B-tagging
• Identify jets originating from b quarks from their topology
• Exploit relatively long lifetime of B decays  displaces vertex
– Jet energy scale calibration
• Relate energy of reconstructed jet to energy of parton
• Detector and physics calibration (some fraction of parton energy is
undetectable to due production of neutrinos, neutral hadrons etc…).
• Dependent of flavor of initial quark  need to measure separately for b jets
Wouter Verkerke, NIKHEF
Introduction to physics commissioning
• Jet energy scales (cont’d)
– Ultimate goal for JE calibration is 1%
– At startup calibration will be less known
– Important – Illustrated of effect on m(top) measurement
Uncertainty
on light jet scale:
1%
10%
Hadronic
 Mt < 0.7 GeV
 Mt = 3 GeV
Uncertainty
On b-jet scale:
1%
5%
10%
Hadronic
 Mt = 0.7 GeV
 Mt = 3.5 GeV
 Mt = 7.0 GeV
– Impacts many measurements, not just m(top)
• Need to start data to good use for calibration purposes as
quickly as possible
– Top physics ideal candidate to do the job
– Also candidate for clean physics channel for early cross section
measurement
Wouter Verkerke, NIKHEF
Top physics at LHC
• Large ttbar production cross section at LHC
– Effect of large s at LHC  threshold for ttbar production at lower x
sˆ  sx1 x2 ; x1 x2 ~ 103
ggtt
stt(tot) =
759±100 pb
Nevt ~ 700/hour
qqtt
– Production gluon dominated at LHC, quark dominated at Tevatron
– About 100 times larger than cross section at Tevatron (lumi
also much larger)
Wouter Verkerke, NIKHEF
Top physics at topology
• Decay products are 2 W bosons and two b quarks
– About 99.9% to Wb, ~0.1% decay to Ws and Wd each
t
t
• For commissioning studies focus on events where one W
decays hadronically and the other W decays semi-leptonically
– About 30% of total ttbar cross section
Wouter Verkerke, NIKHEF
What can we learn from ttbar production
• Abundant clean source of b jets
– 2 out of 4 jets in event are b jets
 O(50%) a priori purity
(need to be careful with ISR
and jet reconstruction)
– Remaining 2 jets can be kinematically
identified (should form W mass) 
possibility for further purification
t
t
Wouter Verkerke, NIKHEF
What can we learn from ttbar production
• Abundant source of W decays into light jets
– Invariant mass of jets should add
up to well known W mass
– Suitable for light jet energy scale
calibration (target prec. 1%)
• Caveat: should not use W mass in jet
assignment for calibration purpose
to avoid bias
– If (limited) b-tagging is available,
W jet assignment combinatorics
greatly reduced
t
t
Wouter Verkerke, NIKHEF
What can we learn from ttbar production
• Known amount of missing energy
– 4-momentum of single neutrino in each
event can be constrained from event
kinematics
• Inputs in calculation: m(top) from Tevatron,
b-jet energy scale and lepton energy scale
t
t
Wouter Verkerke, NIKHEF
What can we learn from ttbar production
• Two ways to reconstruct the top mass
– Initially mostly useful in event selection,
as energy scale calibrations must be
understood before quality measurement
can be made
– Ultimately determine m(top)
from kinematic fit to complete event
• Needs understanding of bias and resolutions
of all quantities
• Not a day 1 topic
t
t
Wouter Verkerke, NIKHEF
How to identify ttbar events
• Commissioning study  Want to restrict ourselves to
basic (robust) quantities
– Apply some simple cuts

– Hard pT cuts really clean up
sample (ISR).

– Possible because
of high production rate

Combined efficiency of requirements
is ~5%  still have ~10 evts/hour
4 hard jets
(PT >40 GeV)
1 hard lepton
(Pt >20 GeV)
Missing ET
(ET >20 GeV)

Wouter Verkerke, NIKHEF
Can this be done?
• Selecting ttbar with b-tagging expected to be easy: S/B
is O(100)
• But we would like to start without b-tagging
– Major worry: background. Can we see a signal?
– Does the idea hold with increasingly realistic detector simulation?
• Short history of study
– Freiburg 2004: Initial Fast Simulations studies by M. Cobal and S.
Bentvelsen demonstrate viability of idea
today
– Rome 2005: Repeat studies with Full simulation (I. van Vulpen &
W. Verkerke)
– Oct 2005 Physics week: Improve background estimates, add
effects of trigger efficiency
Wouter Verkerke, NIKHEF
Backgrounds that you worry about
W+4jets (largest bkg)
QCD multi-jet events
e-,p0
Wln
– Problematic if 3 jets line up m(t) and W +
remaining jet also line up to m(t)
– Cannot be simulated reliably
by Pythia or Herwig. Requires dedicated
event generator AlpGen
– Ultimately get rate from data Z+4 jets
rate and MC (Z+4j)/(W+4J) ratio
– Vast majority of events can be rejected
exploiting jet kinematics.
– Problematic if one jets goes down
beampipe (thus giving ETmiss) and
one jets mimics electron
– Cross section large and not well
unknown, but mostly killed by
lepton ID and ETmiss cuts.
– Rely on good lepton ID and ETmiss
to suppress
Wouter Verkerke, NIKHEF
‘Standard’ top analysis
• First apply selection cuts
Missing
ET > 20 GeV
1 lepton
PT > 20 GeV
Selection efficiency = 5.3%
4 jets(R=0.4) PT > 40 GeV
W CANDIDATE
• Assign jets to W, top decays
1 Hadronic top:
Three jets with highest
vector-sum pT as the decay
products of the top
TOP
CANDIDATE
2 W boson:
Two jets in hadronic top with
highest momentum in reconstructed
jjj C.M. frame.
Wouter Verkerke, NIKHEF
Samples for ‘Rome’ study
ttbar (signal)
• Generator: MC@NLO
• Includes all LO + NLO m.e.
W+jets (background)
• Dedicated Generator: AlpGen
• Includes all LO W + 4 parton m.e.
Hard
Process
CPU intensive!
Fragmentation,
Hadronization &
Underlying event
Atlas Detector
Simulation
Herwig (Jimmy) [ no pileup ]
ATLAS Full Simulation 10.0.2 (30 min/ev)
‘T1’ Sample
175K event = 300 pb-1
‘A7’ Sample
145K event = 61 pb-1
Wouter Verkerke, NIKHEF
Signal-only distributions (Full Simulation)
W CANDIDATE
TOP
CANDIDAT
E
•
Clear top, W mass peaks visible
•
Background due to mis-assignment of jets
–
•
Masses shifted somewhat low
–
mtop = 162.7±0.8 GeV
m(tophad)
Easier to get top assignment right than
to get W assignment right
Effect of (imperfect) energy calibration
MW = 78.1±0.8 GeV
m(Whad)
L=300 pb-1
Jet energy scale
calibration
possible from
shift in m(W)
(~1 week of running)
S
B
S/B = 1.20
S/B = 0.5
Wouter Verkerke, NIKHEF
Signal + Wjets background (Full Simulation)
• Plots now include W+jets background
W CANDIDATE
– Background level roughly triples
TOP
CANDIDAT
E
– Signal still well visible
– Caveat: bkg. cross section quite uncertain
m(tophad)
m(Whad)
Jet energy scale
calibration
possible from
shift in m(W)
S
L=300 pb-1
B
(~1 week of running)
S/B = 0.45
S/B = 0.27
Wouter Verkerke, NIKHEF
Signal + Wjets background (Full Simulation)
• Now also exploit correlation
between m(tophad) and m(Whad)
W CANDIDATE
TOP
CANDIDAT
E
– Show m(tophad) only for events with
|m(jj)-m(W)|<10 GeV
m(tophad)
m(tophad)
m(Whad)
S
L=300 pb-1
(~1 week of running)
S/B = 1.77
B
S/B = 0.45
Wouter Verkerke, NIKHEF
Signal + Wjets background (Full Simulation)
•
TOP
CANDIDATE
Can also clean up sample by with
requirement on m(jln) [semi-leptonic top]
–
SEMI LEPTONIC
TOP CANDIDATE
•
NB: There are two m(top) solutions for each
candidate due to ambiguity in reconstruction of pZ
of neutrino
Also clean signal quite a bit
–
m(W) cut not applied here
m(tophad)
m(tophad)
L=300 pb-1
S
(~1 week of running)
|m(jln)-mt|<30 GeV
B
S/B = 0.45
S/B = 1.11
Wouter Verkerke, NIKHEF
Effect if increasing realism
• Evolution of m(top) resolution, yield with improving realism
Effect of
detector
simulation
Effect of
increasing
Wjets bkg.
Effect of
mW cut
m(top) (GeV)
resolution (GeV)
s(N) stat
Truth jets
171.1 ± 0.4
7.0 ± 0.2
6.0%
Full simulation
162.7 ± 0.8
15.8 ± 0.8
6.3%
+50%
164.1 ± 1.0
17.0 ± 1.5
10%
+100%
165.9 ± 1.4
19.8 ± 2.8
17%
Hadronic MW=
80.4±10 GeV
160.0 ± 1.0
15.4 ± 1.2
8.3%
Wouter Verkerke, NIKHEF
Exploiting ttbar as b-jet sample (Full Simulation)
W CANDIDATE
TOP
CANDIDAT
E
• Simple demonstration use of ttbar
sample to provide b enriched jet
sample
– Cut on m(Whad) and m(tophad) masses
– Look at b-jet prob for 4th jet (must be b-jet
if all assignments are correct)
W+jets (background)
‘random jet’,
no b enhancement expected
ttbar (signal)
‘always b jet if all jet assignment are OK’
b enrichment expected and observed
AOD b-jet probability
AOD b-jet probability
Clear
enhancement
observed!
Wouter Verkerke, NIKHEF
Moving beyond Rome – Improving the analysis
• We know that we underestimate the level of background
– Only generating W + 4 partons now, but W + 3,5 partons may
also result in W + 4 jet final state due to splitting/merging
W + 3 partons
(80 pb*)
W + 4 partons
(32 pb*)
Wln
Wln
W + 5 partons
(15 pb*)
Wln
2 parton
reconstructed
as single jets
parton is
reconstructed
as 2 jets
* These are the cross sections with the analysis cuts on lepton andWouter
jet pT Verkerke,
applied at NIKHEF
the truth level
Moving beyond Rome – Improving the analysis
• Improving the W + 4 jets background estimate
– Need to simulate W + 3,5 parton matrix elements as well
– But not trivial to combine samples: additional parton showering in
Herwig/Jimmy leads to double counting if samples are naively
added
– But new tool available in AlpGen v2.03: MLM matching
prescription.
• Explicit elimination of double counting by reconstructing jets in event generator
and killing of ‘spillover’ events.
• Work in progress
– Expected for upcoming Oct Physics week
– To set upper bound: naïve combination of W + 3,4,5 parton
events would roughly double W+jets background.
Wouter Verkerke, NIKHEF
Moving beyond Rome – effect of trigger
• Look at Electron Trigger efficiency
– Event triggered on hard electron
• Triggering through 2E15i, E25i, E60 channels
– Preliminary trigger efficiency as function of lepton pT
• Efficiency = fraction of events passing all present analysis cuts that are triggered
• Analysis cuts on electron include requirements on isem flag and etcone40
• Includes effects of ‘untriggerable’ events due to cracks etc…
#triggered events / # events
• In cooperation with M. Wielers (work in progress)
73.5%
Nominal analysis cut
Electron pT (GeV)
Wouter Verkerke, NIKHEF
Summary
• Can reconstruct top and W signal after ~ one week of data
taking without using b tagging
– Can progressively clean up signal with use of b-tag, ET-miss, event topology
• Many useful spinoffs
– Hadronic W sample  light quark jet energy scale calibration
– Kinematically identified b jets  useful for b-tag calibration
• Continue to improve realism of study and quality of analysis
– Important improvement in W+jets estimate underway
– Incorporate and estimate trigger efficiency to few (%)
– Also continue to improve jet assignment algorithms
• Expect estimate of s(ttbar) with error < 20% in first running
period
– One of the first physics measurements of LHC?
Wouter Verkerke, NIKHEF