E. Soldatov

E. Soldatov 01.09.2011

Tight photon efficiency study using FSR photons Tight photon efficiency study using FSR photons fromfrom ZZ decay decay

E.Yu.Soldatov*

*National Research Nuclear University “MEPhI”

Outline:

1. MC study of the chosen sample

2. Background estimation study

3. Tight cut study

Photon ID efficiencies meeting

E. Soldatov

IntroductionIntroduction

№ 2

1. Mass of two leptons should have a missing part e.g. < Z bozon mass

For electron channel 60 < m(ee) < 83 GeV, and for muon channel 40 < m() < 82 GeV

A kinematic approach of the photon selection

2. And 3 body mass should correspond to a mass of Z bozon

For electron channel 80 < m(ee)< 94 GeV and for muon channel 81 < m() < 95 GeV

Kinematics approach is based on a simple criteria of the event selection.

Background-ISR-Brem-Jets

-FSRSignal

Three body mass spectrum (Z)

01.09.2011Photon ID efficiencies meeting

E. Soldatov

IntroductionIntroduction

A “minor” problem: a small production cross section.

What do we want?

The main idea is to obtain a photon sample with maximum purity with the method decoupled from the standard analysis

methods. Why do we want this?

Studies of the ATLAS detector performance with tagged photons and first of all study and optimization of the tight cut criteria.

Very important for analysis all photon containing processes.

The best source of such kind of photons is a production of

FSR photons from Z-lepton decay:

№ 301.09.2011Photon ID efficiencies meeting

Photon purity after all cuts appliedPhoton purity after all cuts applied

E. Soldatov

Purity of the sample after Et (in cone 0.2)<3 GeV and N=0 cuts

Figure. Purity of the sample as a function of photon energy

Purity=Signal/(Signal+Background)

Et cut [GeV]

Differential purity for each bin.


E. Soldatov

TightTight cut efficiency studycut efficiency study

Efficiency of robust tight cut on signal (lift) and on background (right) vs Et photon spectrum, Et (in cone 0.2)<3 GeV

ET [GeV]

- Sgn sample MC

- Data 2011

ET [GeV]

- Bkg sample MC

- Data 2011

Signal Background


Tight CUT efficiency for signal sample Tight CUT efficiency for background sample

E. Soldatov


Efficiency of robust tight cut on signal (lift) and on background (right) vs Et photon spectrum, Et (in cone 0.2)<3 GeV (barrel only)

Signal Background



ET [GeV]ET [GeV]

E. Soldatov



ET [GeV]

Signal Background



ET [GeV]

E. Soldatov


Multiplicity for data


E. Soldatov


For multiplicity<5


E. Soldatov


For multiplicity>6


E. Soldatov

Effect of background uncertainty on the tight cut efficiencyEffect of background uncertainty on the tight cut efficiency

Background estimation from MC

5-10 GeV 15-20 GeV

MC Purity=0.827 (slide 11)

S+B Efficiency=0.17 (sl.20)

Bkg Efficiency=0.06 (sl.20)




S Efficiency=(0.17-0.06*0.173)/0.827=0.193 S Efficiency=(0.53-0.32*0.027)/0.973=0.5358


Correction gives us uncertainty: (0.193-0.17)/0.17*100%=13.5%

Correction gives us uncertainty: (0.5358-0.53)/0.53*100%=1%

E. Soldatov

Data driven background estimationData driven background estimationWork of the background estimation method on real data

Cut Et (in cone 0.2)<3 GeVREAL DATA

ET()>5 GeV, ET()<10 GeV

N (tracks in cone20)

First bin: Data=2312; Bkg estimation=853 => Purity=0.631№ 1201.09.2011Photon ID efficiencies

meeting

E. Soldatov

Data driven background estimationData driven background estimationWork of the background estimation method on real data

Cut Et (in cone 0.2)<3 GeVREAL DATA

First bin: Data=774; Bkg estimation=66 => Purity=0.915

ET()>15 GeV, ET()<20 GeV



E. Soldatov

5-10 GeV 15-20 GeV

Estimated Purity=0.631 (sl.22)






S Efficiency=(0.17-0.06*0.369)/0.631=0.234

Correction gives us uncertainty:(0.234-0.17)/0.17*100%=38%

Uncertainty due to the method:

(0.234-0.193)/0.193*100%=21%

S Efficiency=(0.53-0.32*0.085)/0.915=0.5495

Correction gives us uncertainty:(0.5495-0.53)/0.53*100%=3.7%


(0.5495-0.5358)/0.5358*100%=2.5%

Effect of background uncertainty on the tight cut efficiencyEffect of background uncertainty on the tight cut efficiency

Background estimate – data driven


E. Soldatov

TightTight cut efficiency study (Z->eecut efficiency study (Z->ee))



ET [GeV]ET [GeV]

SignalBackground


electron channel

Photon purity after all cuts applied (Z->eePhoton purity after all cuts applied (Z->ee))

E. Soldatov

Purity of the sample after Et (in cone 0.2)<3 GeV and N=0 cuts


Et cut [GeV]

MC purity of the sample:

Estimated purity from here:

ET()>5 GeV

ET()>15 GeVData driven estimation of the background amount in the first bin

electron channel

E. Soldatov

5-10 GeV 15-20 GeV









S Eff (MC pur)=(0.15-0.06*0.407)/0.593=0.212

S Eff (Est pur)=(0.15-0.06*0.418)/0.582=0.215


(0.215-0.212)/0.212*100%=1.4%

S Eff (MC pur)=(0.55-0.26*0.04)/0.96=0.562

S Eff (MC est)=(0.55-0.26*0.091)/0.909=0.579


(0.579-0.562)/0.562*100%=3%

Effect of background uncertainty on the tight cut efficiency (Z->eeEffect of background uncertainty on the tight cut efficiency (Z->ee))

Background estimate – MC&data driven


electron channel

E. Soldatov

ConclusionsConclusions

The improvement of signal selection has been done. The purity became larger than 97% for 15 GeV photon sample.

Monte Carlo simulation has been compared with the 2011 Data (periods D-I with ~1.33 fb-1 of integral luminocity).

To improve the agreement between simulation and data background estimation method from data has been proposed and implemented.

Tight cut efficiency has been estimated using two methods of sample purity estimation. Efficiency from data has significant differences from MC.


E. Soldatov

Back-up slidesBack-up slides


Data driven background estimationData driven background estimation

E. Soldatov

Crosscheck on a larger statistics!

In order to obtain more background photons under in a signal sample less stringent kinematic conditions where taken:

40 < m() < 88 GeV75 < m() < 105 GeV

Excellent agreement!

ET()>5 GeV


Number of tracks in cone 0.2 around photon vector in ID.

Again the black line shows the background estimate using extrapolation method described above.


E. Soldatov

MC: Study of the pure background sample.MC: Study of the pure background sample.

Spectrum of background photons from the signal sample after 3 body

mass CUT applied

ET [GeV]

Spectrum of background photons from the background sample

Applying 3 body mass cutwe deform a bit the photon

spectrum.

Does it affect the background estimation?

Background photon spectrums

Cut Et (in cone 0.2)<3 GeV

№ 2128.07.2011Physics in ATLAS

MC study: Pt cone and N tracks (cone 0.2)MC study: Pt cone and N tracks (cone 0.2)

E. Soldatov

Background sample extraction will provide us exact distribution of ID track information for background photon candidates

ET()>10 GeV

PT(cone20) [GeV] N (tracks in cone20)

Sum of all momentums in cone 0.2 around photon vector in ID

Number of tracks in cone 0.2 around photon vector in ID

Background sample kinematic cuts are: for muon channel 89 < m() < 93 GeVCut Et (in cone 0.2)<3 GeV


MC study: Pt cone and N tracks (cone 0.2)MC study: Pt cone and N tracks (cone 0.2)Background sample extraction will provide us exact distribution

of ID track information for background photon candidates

We have almost the same picture for all energies.

But for high energies, we have a bit large fraction of

the signal.

ET()>15 GeV

Sum of all momentums in cone 0.2 around photon vector in ID

Number of tracks in cone 0.2 around photon vector in ID

E. Soldatov

PT(cone20) [GeV] N (tracks in cone20)

Background sample kinematic cuts are: for muon channel 89 < m() < 93 GeVCut Et (in cone 0.2)<3 GeV


E. Soldatov


Distribution of number of tracks in the cone 0.2 in Inner Detector around background photon (black line histogram) in the signal sample the three-body invariant mass cut applied, and for photon background sample (blue filled histogram)

ET()>10 GeV

Comparison of the track number distribution for different background photon samples

Cut Et (in cone 0.2)<3 GeV


E. Soldatov


Distribution of number of tracks in the cone 0.2 in Inner Detector around background photon (black line histogram) in the signal sample the three-body invariant mass cut applied, and for photon background sample (blue filled histogram)

Comparison of the track number distribution for different background photon samples

Cut Et (in cone 0.2)<3 GeVThe shapes of the

distribution are in a less good agreement,

than for small energies. However

due to big statistical errors, we can use this

shape from photon background sample

for a data driven background estimate

normalizing on number of events for

N>1. There is no signal in this area

ET()>15 GeV


Data driven background estimationData driven background estimation

E. Soldatov

Signal to background ratio is much higher for larger photon energies

Always condition Et in cone20 < 3 GeV is used

Relaxed conditions:40<m()<88 GeV75<m()<105 GeV(to gain more bkg statistics)


We use these bins for extrapolation


Number of tracks in cone 0.2 around photon vector in ID.

Again the black line shows the background estimate using extrapolation method described above.

We have a problem: signal here!

ET>10 GeV


E. Soldatov

TightTight cut effect on background samplecut effect on background sampleEfficiency of robust tight cut on background vs sum of Et in cone 0.2 in calorimeter around photon (left) and Et photon spectrum (right), Et(in cone20)<3 GeV (one photon candidate in event)

List all the conditions: all kinematic cuts, Et in cone20 < 3 GeV , N tracks = 0

ET(cone20) [GeV]

Distribution of Et photon spectrum for background sample before (left) and after (right) robust tight photon cut.

ET [GeV]


E. Soldatov

TightTight cut effect on background samplecut effect on background sample

N=Ns+Nb=17228after tight cut we haveNt=effs*Ns+effb*Nb=8414We also know purity=Ns/N (slide 9)

We can extract pure effs (we know effb from slide 25):Effs=(Nt-effb*Nb)/Ns=Nt/N*purity-effb*(1/purity-1)=0.38 (for 5 GeV)

List all the conditions: all kinematic cuts, Et in cone20 < 3 GeV , N tracks = 0


E. Soldatov

Documents

Transcript of E. Soldatov