Summary of Commissioning Studies Top Physics Group

Summary of Commissioning Studies

Top Physics Group

M. Cobal, University of Udine

Top Working Group, CERNOctober 29th, 2003

Top Quark Event Yields

• NLO Xsect for t-tbar production = 833 pb8 million t-tbar pairs produced per 10 fb-1

• We reconstruct the top mass in the lepton+jets channel Clean sample (1 isolated lepton, high Etmiss).

Statistical Error

Period tt events

1 year 8x106

1 month 2x106

1 week 5x105

In the single lepton channel, where we plan to measure m(top) with the best precision:

Period evts Mtop(stat)

1 year 3x105 0.1 GeV

1 month

7.5x104 0.2 GeV

1 week 1.9x103 0.4 GeVL = 1x1033 cm-2s-1

Top mass precision

One top can be directly reconstructed

Reconstruct t Wb (jj)b

Selection cuts:

1 iso lep, Pt > 20 GeV, || < 2.5, Etmiss > 20At least 4 jets with Pt > 40 GeV and || < 2.5At least 2 b-tagged jets Selection effic. = 5% 126k events, with S/B = 65

Two methods:

Reconstruction of the hadronic part W from jet pair with the closest invariant mass to m(W) cut on |mjj-mW| < 20 GeV Association of W with a b-tagged-jet

Cut on |mjjb-<mjjb>| < 35 GeV

Kinematic fit

The leptonic part is reconstructed |mlb-<mjjb>| < 35 GeV -30k signal events-14k bkgnd events

Kinematic fit to ttbar, with m(top) and m(W) mass constraintsMain Background is the combinatorial one.

Systematics for the lepton + jet analyses

At the beginning the jet energy scale will be not known as well as 1%

Energy scale

From M. Bosman:

- Will start to calibrate calorimeter with weights from MC- Assume:

• EM scale correct to the percent level from the very beginning • fragmentation correctly described in MC• corrections for calorimeter non-compensation and dead material

correct calibration coefficients should be predicted

1) First check fragmentation function with the tracker, then dijet differential cross-section, distribution, check pT balancing across different detectors, etc.

2) Start lo look at in-situ calibration samples: At the very beginning, start with W->jj.

Taking TDR numbers:

1500 ttbar->bW(l)bW(jj) requiring 4 jets above 40 GeV/day at low L.

In 1 week: 10k W to jj decays In 1 month: 35k W to jj decays

Jets have a pT distribution: ~ 40 to 140 GeV with changing calibration. Consider pT bins of 10 GeV, and bins of 0.3. There are 150 "samples" to consider: After a week, about 70 W per "sample" or a statistical error on m(W) sigma(about 8 GeV with perfect calibration) divided by sqrt(70) This makes ~1% of statistical error

On top there is the systematic errors due to FSR and jet overlap...

Observed linearity dependence of the top mass shift on the b-jet absolute scale error for the inclusive sample.

Can scale correspondingly: Hadronic Kin fit 1% jet energy uncertainty M(top) = 0.7 0.7 GeV

5% jet energy uncertainty M(top) = 0.7*5 = 3.5 3.5 GeV

10% jet energy uncertainty M(top) = 0.7*10 = 7 7 GeV

b-jet scale

Here as well linear dependenceIf one performs constrained fit onW-mass, is less important than b-jet scale.

Can scale correspondingly: Hadronic 1% jet energy uncertainty M(top) < 0.7 GeV

10% jet energy uncertainty M(top) = 3 GeV

Light-jet scale

B-tagging

From S. Rozanov:

Main effects of initial layout:

2 pixel barrel layers rejection of light jets reduced by ~30%. Another important parameter is the efficiency of the pixel chips and modules (not predicted).

Effect of alignment precision:

Precise alignment of ID could be reached only after a FEW MONTHS work. (studies undergoing) Impact of misalignment much higher than effect of 2 or 3 layers. Can also compromise a jet energy calibration based on W from tt at startup: could be difficult to select W’s over background.

Estimates for initial (t-tbar) measurement

• Initial lum = 1x1033 cm-2 s-1 t-tbar production rate = 0.85 Hz

~ 500k t-tbar events produced per week

• With same analysis and detector performance as in Physics TDR, predict:– Selection of 8000 single lepton plus jets events, S/B =

65

– In ± 35 GeV window around m(top), would have:• 1900 signal events• 900 bkgnd events (dominated by “wrong

combinations” from t-tbar events)

stat error on (t-tbar) 2% after 1 week

• What happens with degraded initial detector performance?

– eg. Consider case where b-tagging is not available in early running:

– Drop b-tagging requirement: signal effic. increases from 5% to 20%, but bkgnd increases faster

– For one week, would select 32000 signal events, but with S/B = 6

– Biggest problem comes from large increase in combinatorial bkgnd when trying to reconstruct t Wb (jj)b with b-tagging

W jj t Wb (jj)b

– Fit of m(jjb) spectrum provides Xsect measurement with stat. error 7%

– Even with no b-tagging, can measure (t-tbar) to < 10% with two days of integrated luminosity at 1x1033

Results presented

An initial uncertainty of 5% on the b-jet energy scale, gives a top massuncertainty of 3.5 for the mass reconstuction.If we go to 10% , the uncertainty on the top mass is of ~7 GeV

An initial uncertainty of 10% on the light jet energy scale, gives a top mass uncertainty of 3 GeV for the mass reconstuction. Kinematic fit less sensitive to light jet energy scale. But can have very large combinatorial background in case of b-tagging not working

After 1 week of data taking we should be able to measure the cross-section with a 2% statistical error

Even without b-tagging, with two days of data taking, can measure at < 10% (stat. error)

In Athens:

In Prague:

First evaluation of Mtop, assuming no b-tagging at the startup (V. Kostiouchine)

Investigation of differences found in the combinatorial backgnd between TDR and DC1 (V. Kostiouchine)

Mtop reconstruction in ATLAS at startup

Work done by V. Kostioukhine

Assumptions:

• No jet energy calibration, no b-tagging.• Uniform calorimeter response • Good lepton identification.

TDR signal+backgrounds estimation

In case of no b-tag:

tt signal: ~500k evt ( 4 times reduction due to b-tag)W+jets: ~85k evt (50 times reduction due to b-

tag)

Signal selection without b-tag

Lepton+4jets exactly (R=0.4): signal ~76% with respect to

4jet W+jets ~83% with respect to

4jets

Select the 3-jet combination with maximal

Select among them 2 jets with maximal

3

1iiPP

2

1iiPP

jjj

jj

Having 3 jets from t-quark decay,there are 3 possible jet assignments for W(jj)b.

• A kinematical constraint fit can be used for a further selection: MW

1=MW

2 and Mt

1=Mt

2.

An approximate calibration is obtained with the W peak

• Select the combination with lowest 2 out of the 3 available. Event is accepted is this minimal 2 is less than a fixed value.

Big 2 events

Reconstructed Mtop

Signal selection: ( 4jets exactly+2 cut) ~40% (~200k evt)

W+jets selection: with the same cuts ~9% (~8k evt)

2 signal 2 W+jets3-jet mass W+jets

Preliminary results with full simulation

TDR top sample(same cuts as fast sim.)

Top mass

W mass

DC1 sample (same cuts as fast sim.)

Top mass

W mass

Conclusions on Mtop

1. A tt signal can be selected without b-tagging and precise jet energy calibration

2. Signal / backgnd ratio is ~20 in this case (~70 in the region Mjjb<200 GeV) . Here only W+jets events are considered as background.

3. Such a clean sample could be also used for jet energy calibration.

4. Results confirmed by full simulation

Combinatorial background in DC1 data

Work done by V. Kostioukhine

• Increase of the combinatorial background in DC1 samples with respect to the TDR ones

• Vadim checked better and.....

W(TDR) W (DC1)

TDR +jets sample

Selection: 1 lep with Pt>20 GeV, Pt miss >20 GeV, at least 4 jets with

Pt>40GeV, 2 b-jets (parton level). 2 non-b jets with min|Mjet-jet – MW|

taken as W decay products. b jet is selected so that Pt jet-jet-b -> max

t-quark peak after application of constraint fit

jj mass jjb mass

top

DC1 +jets sample

Same selection

DC1 sample

t-quark peak after application of constraint fit

DC1 sample with application of“TDR-like” generation level cuts

jj mass jjb mass

top top

jj mass jjb mass

DC1 e+jets sample Selection: the same

DC1 sample

t-quark peak after application of constraint fit

DC1 sample with application of“TDR-like” generation level cuts

jj mass jjb mass jjb massjj mass

toptop

DC1 summary e,+jets sample

Same selection DC1 sample with application of“TDR-like” generation level cuts

DC1 sample

t-quark peak after application of constraint fit agreement with TDR !!

toptop

jj mass jj massjjb mass jjb mass

Next Steps

More detailed MC study: W + jets background.

Study of background level dependence on b-tagging .

Measure the cross-section and top mass assuming different efficiency for the b-tagging (and no b-tagging at all) and looking at various channels. What is the minimal b-tagging needed?

……………

First look at data in 2007

Study of high pT isolated electrons and muons

Select a “standard” top sample, and a “golden” top sample with tighter cuts.

Try to reconstruct the two top masses (in single lepton events, one top decays hadronically, the other one leptonically)

Take top events: try a first measurement of the cross section, and of the mass in various channels (as a cross check, since systematic errors are different)

(tt) : initial measurement dominated by L and detector uncertainties 10-20%?

In addition, very pessimistic scenario considered : b-tag not yet available S increases by ~ 4 S/B decreases from 65 to 6 large combinatorial background

W jj t bjj

M (jj) M (bjj)

Still a top peak is visible Statistical error from fit: from 2.5% (perfect b-tag) to 7% (no b-tag) for ~ one weekWhat about B systematics ?

M (jj)

W jj

difference of distributionsfor events in the top peak andfor events in the side-bands

Feedback on detector performance:-- m (top) wrong jet scale ? -- golden-plated sample to commission b-tag

W jj t Wb (jj)b

– Fit of m(jjb) spectrum provides Xsect measurement with stat. error 7%

– Even with no b-tagging, can measure (t-tbar) to < 10% with two days of integrated luminosity at 1x1033

Conclusions

An initial uncertainty of 5% on the b-jet energy scale, gives a top massuncertainty of 3.5 for the mass reconstuction.If we go to 10% , the uncertainty on the top mass is of 7 GeV

An initial uncertainty of 10% on the light jet energy scale, gives a top mass uncertainty of 3 GeV for the mass reconstuction. Kinematic fit less sensitive to light jet energy scale. But can have very large combinatorial background in case of b-tagging not working

After 1 week of data taking we should be able to measure the cross-section with a 2% statistical error

Even without b-tagging, with two days of data taking, can measure at < 10% (stat. error)

Additional studies (e.g. di-lepton) undergoing