Top physics during ATLAS commissioning

1
Top physics during ATLAS commissioning
Ivo van VulpenWouter VerkerkeStan Bentvelsen

2
Top physics at the LHC in 1 slide
stt(tot) 759 pb
stt(semi-lept e,m) 30
Nevents 700 per hour
1 hard lepton (Pt gt20 GeV)
4 hard jets (PT gt40 GeV)
2 of which are b-jets
with b-tag S/B 80
Missing ET (ET gt20 GeV)
systematics dominated
3
What if b-tagging does not work perfectly during
first LHC runs ?
Can we see top quarks after 1 week of data ?
(no b-tag information)
Fit for stt(tot) and mtop
First light-jet energy scale calibration (W)
Select b-jet enriched jet sample (b-tag
commissioning)
4
Studies performed in ATLAS up to last year
Work initiated and performed by Marina Cobal and
Stan Bentvelsen
Top-peak
W-peak
2004 Freiburg meeting(150 pb-1)
Conclusion Can be done, but verify using FULL
simulation.
5
Monte Carlo samples

• For all plots in this talk we have used
• Full simulation 10.0.2 AOD files
• Monte Carlo Event samples
• MC_at_NLO ttbar 175K events ? 303 pb-1
• WJets 145K events ? 61 pb-1
• Monte Carlo equivalent statistics
• All plots are shown with statistics
  corresponding to 303 pb-1 ? a small week at
  low lumi LHC running

6
Standard top analysis
Selection cuts
Missing ET gt 20 GeV
1 lepton Pt gt 20 GeV
Selection efficiency 5.3
4 jets PT gt 40 GeV
Trigger efficiency not taken into account yet
7
Influence of Jet pT-min cut on number of selected
events
Note require 4 good jets, with Good jet PT gt
PT(min) and h lt 2.5
Nevents with exactly 3 good jets
Fraction of events
Nevents with exactly 4 good jets
12 of events has 4 reconstucted jets
Nevents with exactly 5 good jets
Minimum Jet pT-cut (GeV)
8
Standard Analysis on MC_at_NLO signal Monte Carlo
RooFitting the top and W mass
No cut on m(W)
No cut on m(t)
m(W)
m(t)
m(t)
Number of events / 5.1 GeV
Number of events / 5.1 GeV
Number of events / 5.1 GeV
Mtop 162.70.8 GeV
S/B 1.20
S
MW 78.10.8 GeV
B
S/B 0.5
W mass (GeV)
Top mass (GeV)
Use peak position for light jet energy calibration
9
Check behaviour combinatorics using MC_at_NLO signal
Monte Carlo
m(t)
m(W)
m(t)
m(t)
m(t)
Subset of events where chosen 3-jet combination
does not line up with top quark (using AOD truth
information)
Number of events / 5.1 GeV
Number of events / 5.1 GeV
Empirical background shape describes combinatoric
background well under peak
Top mass (GeV)
10
The top and W peaks when including W4jets
background
Can still see top peak! Wjets bckg is large (and
has large uncertainty)
m(t)
m(W)
S/B 0.45
S/B 0.27
S
Number of events / 5.1 GeV
Number of events / 5.1 GeV
B
Wjets and MC_at_NLO signal
Wjets and MC_at_NLO signal
W mass (GeV)
Top mass (GeV)
Signal Properly reconstructed ttbar
events Background Wjets events and
improperly reconstructed ttbar
events
in mass range130 lt M(jjj)lt 200 GeV
11
Improving sensitivity cut on hadronic W-mass (I)
Using the hadronic W
Ask for 70 lt M(jj) lt 90 GeV
So far looked at m(W) and m(top) independently
? Use m(W) information in m(top) results in
a clean signal even in presence of Wjets
background
12
Improving sensitivity cut on hadronic W-mass (II)
Mw within 10 GeV of nominal W mass
Standard analysis
m(t)
m(t)
S/B 0.45
S/B 1.77
Top mass (GeV)
Top mass (GeV)
Top peak clearly visible after 1 week of LHC data
13
Overview of fit results
Mtop (GeV) Resolution (GeV) s(N)
stat

• Signal only

• Adding Wjet background

• 100 background plus cut on m(W)

14
Systematics and further ideas and improvements
15
In progress various ways to clean up hadronic
top peak
Event topology
Using the leptonic top
Cut on mtop(lept) sstat(N) 16 ? 8.1

• Construct likelihood from several
  topological variables

S/B 1.11
ttbar
Wjets
140 lt M(lnj) lt 200 GeV (use ET-miss)
top likelihood
Excellent tool to calibrate/check ET-miss
16
The b-jet enriched jet sample
17
The 4th jet a b-jet ??
W4jets sample(control sample)
TOP CANDIDATE
ttbar sample(expect enrichment)
B-JET CANDIDATE
clear B-enrichment
b-tag likelihood
18
Reverse logic use b-tag of 4th jet to clean up
hadronic top
Standard analysis (for comparison)
Cut on b signal probability gt 0.90 on 4th jet
S/B 0.45
S/B 1.36
ttbar
ttbar
Wjets
Wjets
Closer to post-commissioning (normal) analysis
? see Etienvres talk
19
m(t)
1 week LHC data
CONCLUSIONS

• Can reconstruct top and W signal after one week
  of data taking (without using b tagging)
• Hadronic W sample? light quark jet energy scale
  calibration
• Kinematically identified b jets? usefull for b
  tag-calibration
• Use of b-tag, ET-miss, event topology ? clean up
  signal, move towards standard analysis

Thanks to Marina Cobal, Sascha Caron, Maaike
Limper, Marcello Barisonzi, Paul de Jong, Peter
Kluit, Ian Hinchliffe, Arthur Moraes, Stefano
Frixione and others
20
Backup slides
21
Can we do better with different jet algorithms?
m(t)
m(W)
Cone algoritm R0.7
worse
m(t)
m(W)
KT algoritm
worst