E. Soldatov Pure photon sample selection 1

A photon sample selection from Zll decay

Alexey Antonov*, Evgeny Yu Soldatov*

*Moscow Engineering Physics Institute (MEPHI)

Outline:

•Motivation

•True level Signal Background

•Reconstruction level control plots

•Reconstruction Signal/Background optimization

•Results for Z -> ee•Results for Z -> •Conclusions

“Physics&Computing in ATLAS” – 22/09/2010

E. Soldatov Pure photon sample selection 2

Photon sample selection using Z->llprocess

Truth M(Z->ee)

M(ee) [GeV]

SignalBackground

Idea is to obtain pure photon sample from known physics process with distinctive kinematical feature - -M(ee) and dR(e) -M() and dR()

One of a problems is a small production cross section.

-ISR-Brem-Jets

Main goal of the analysis is to find the signal selection method - check Signal to Background ratio - and statistical yield

-FSR

E. Soldatov Pure photon sample selection 3

MC sample and preselection

Public FullSim MC samples - 5M events ~4.5 fb

Public Analysis code rel. 15.6.1

-1

Z -> mc08.106051.PythiaZmumu_1Lepton.merge.AOD.e347_a84_t53_tid061356

Z -> ee mc08.106050.PythiaZee_1Lepton.recon.AOD.e347_s462_r541_tid028253

Et() > 5 GeV;

Pt(e) > 15 GeV; Pt() > 15 GeV;

|| (e)| < 2.5; || ()| < 2.5;

E. Soldatov Pure photon sample selection 4

Z -> ee signal and signal and backgrounds (truth backgrounds (truth

level)level)

4

E. Soldatov Pure photon sample selection 55

R distribution for signal and all backgrounds (TRUTH)

Signal (FSR) ISR

Brem

-to-closest lepton angle

Jets treated as

22 R

E. Soldatov Pure photon sample selection 6

Remaining statistics & efficiency of photon reconstruction vs. R cut.

6

R>0.2always used

further

Remaining signal

statistics vs. R cut

FSR Efficiency

R

R cut selection

R>0.2

22 R

E. Soldatov Pure photon sample selection 7

TRUTH invM(eeee) for FSR, ISR and Jets (as photon)

FSR ISR Jets

JetsISRFSR

For bottom plots 80GeV < invM(eeee) < 100GeV

0.9K events 148K events52K events

M(eeee) [GeV]M(eeee) [GeV]

M(eeee) [GeV]M(eeee) [GeV]M(eeee) [GeV]

M(eeee) [GeV]

E. Soldatov Pure photon sample selection 8

Z -> ee Reconstruction levelReconstruction level

8

E. Soldatov Pure photon sample selection 9

Control plots for reconstructed FSR photons (signal)

- with tight cut

9

ET (), MeV

Efficiency vs.

()

Et() distribution

ET (), MeV() distribution

- no tight cut

Efficiency vs. Et ()

E. Soldatov Pure photon sample selection 10

Bkg.

all Z-> eeee events Signal/Background

M(eeee) [GeV]

Optimal M((eeee) window is 82<M(eeee)<95 GeV (see backup slides)

Reconstructed invM(eeee)

FSR – 26k

M(eeee) [GeV]

M(eeee) [GeV]

M(eeee) [GeV]

- Tight cut- No Tight cut

- Tight cut- No Tight cut

- Tight cut- No Tight cut- Tight cut

- No Tight cut

E. Soldatov Pure photon sample selection 11

eeee selection optimization

Photon tight cut

Mass window 82<M(eeee)<95 GeV window

Usual selection:

Mass window for invariant M(ee)ee);

Photon ET cut;

Upper R cut.

Extra cuts:

All these cuts will be applied sequentially to the signal with initial statistics 26139 events.

E. Soldatov Pure photon sample selection 12

eeee mass selection optimization

E. Soldatov Pure photon sample selection 13

Reconstructed invM M(eeee) - 82<M(eeee)<95 GeV window

Bkg.FSR

Signal/Background

Bkg.

Signal/Background

FSR

No tight CUT

Tight CUT

15.8k 3.9k events

11.4k events21.1k

Max s/b ~ 7

Max s/b ~ 3.3

M(eeee) [MeV]

M(eeee) [MeV]M(eeee) [MeV]

M(eeee) [MeV] M(eeee) [MeV]

M(eeee) [MeV]

E. Soldatov Pure photon sample selection 14

Reconstructed invM(eeee) @ 82<M(eeee)<95 GeV & 50<M(eeee)<82 GeV

NO Tight cut

Tight cut

M(eeee) [MeV]

- 96% of background rejected if M(eeee) and M(eeee) window cuts applied

FSR

12.7k events

Bkg.

Bkg.FSR

2.k events

16.7k events 6.1k events

<S/B>~6.2

<S/B>~2.75

- 50% signal reduction (with tight cut)

M(eeee) [MeV] M(eeee) [MeV]

M(eeee) [MeV]

E. Soldatov Pure photon sample selection 15

Upper R cut (Tight and Mee cuts applied))

S/B ratio

15

With the cut R<1.6 the S/B ratio reaches level of 7.47 with remaining statistics 39% of the initial one.

R R

Remaining statistics

Remaining signal statistics and averaged s/b ratio (integral from 0.2 to upper R cut) as a function of the upper R cut with Mee and tight cut.

E. Soldatov Pure photon sample selection 16

Remaining statistics

16

For ET>9 GeV averaged S/B=10.45 can be achieved and statistics is 31% from the initial one.

S/B ratio

Effect of the low Et cut for photons (all previous cuts applied)Effect of the low Et cut for photons (all previous cuts applied)

From ET distribution, it is clear, that low ET cut can give an

effect

ET, MeV

- Signal- Background

ET, MeV

ET, MeV

E. Soldatov Pure photon sample selection 17

Mee, MeV

S/B ratio

Differential distributions of s/b ratio as a function of InvMee for all previously presented cuts and with addition of the tight cut

17

For black line averaged S/B=7.18 and remaining statistics is 39% from the initial one. For green – it was given on the previous slide (10.45 and 31%

accordingly).

- all presented all presented cuts - extra tight cut

Ratio of S/B for selection w/o TightCut to selection with TightCut

Z->ee could be useful for the tight cut efficiency validation

E. Soldatov Pure photon sample selection 18

Z -> ee Selections Summary

18

SelectioSelectionn

S/B ratio

Statistics from initial

Number of signal events

No cuts 0.18 100% 26139

Inv Mee<95 GeV

1.40 89% 23221

+ Inv Mee>82

GeV

1.87 82% 21499

+ Tight cut 4.22 62% 16112

+ Inv Mee<82 GeV

5.58 53% 13770

+ Inv Mee>50 GeV

6.19 50% 13004

+ R<1.6 7.47 39% 10318

+ ET>9 GeV 10.45 31% 8187

E. Soldatov Pure photon sample selection 19

Z -> Reconstruction levelReconstruction level

19

E. Soldatov Pure photon sample selection 20

Reconstructed invM() @ 80<M()<96 GeV & 15<M()<80 GeV

- Tight cut has marginal effect on selection if M() and M() cuts applied

NO Tight cut

Tight cut

M() [MeV]

- 99% of background rejected if M() and M() cuts applied

FSR

15k.events

Bkg.

Bkg.FSR

1290 events

16k.events 1348 events

<S/B>~11.6

<S/B>~11.9

- 30% signal reduction

M() [MeV]M() [MeV]

M() [MeV]

Same R>0.2 used

E. Soldatov Pure photon sample selection 21

Differential Signal to Background Ratio as a function of

invM()

Tight cut has a marginal effect on selection if M() and M() cuts applied

NO Tight cut

Tight cut

<S/B>~11.9

<S/B>~11.6

80<M()<96 GeV & 15<M()<80 GeV

M() [MeV]M() [MeV]

No TightCut/TightCut Ratio

Z-> could be useful for the tight cut efficiency validation..

E. Soldatov Pure photon sample selection 22

Statistics reduction

Integral S / B

For Et() >10 GeV; => <S/B>~25;

Cost – 15% of signal events and Et() range

Low Et() selection @ 80<M()<96 GeV & 15<M()<80 GeV

-Signal (FSR)- All Background

ET () MeV

ET () MeV

ET () MeV

E. Soldatov Pure photon sample selection 23

Z -> selections Summary

Selection <S/B> ratio

Remaining events

Number of signal events

No Cuts 0.25 100% 33173

80<M()<96

GeV 2.4 86% 28523

+ Tight cut 6.7 65% 21516

15<M()<80 GeVNo Tight cut

11.9 48% 16055

15<M()<80 GeV+Tight cut

11.6 45% 14992

+Et() >10 25 41% 13600

Public FullSim MC sample of Z -> - 5M events ~4.5 fb-1

E. Soldatov Pure photon sample selection 24

About real data (I)

After all Z-boson note (ATL-COM-PHYS-2010-701) cuts for 310 nb-1 (runs 161520-161948) we have:

E. Soldatov Pure photon sample selection 25

About real data (II)

Probability of considering process is ~0.5% of Z decays to two leptons. (For previous slide picture – 0 such events).

Expecting integral luminosity this year: ~60 pb-1 ~20k of Z (~100 our photons);

Expecting luminosity next year: ~1 fb-1 ~1M of Z (~5000 our photons).

So in the end of this year there will be necessary statistics for start of the analysis such process. But good analysis and good photon sample to obtain (there will be losses of statistics due to cuts) will be possible in the next year.

E. Soldatov Pure photon sample selection 26

- All in all, about ~22k pure photons from ~4.5 fb can be selected using Z boson decays to electrons and muons using proposed method. Analysis of the real data statistics has been done.

-Selection of a pure photons in the process Z -> ee & Z -> has been studied using 5M (~4.5 fb ) events MC samples.

Conclusions

- Tight cut has a marginal effect on selection if M() and M() applied. Means that Tight cut can be studied using Z -> .

- Z -> : set of cuts 80<M(ee)<96 GeV & 15<M(ee)<82 GeV & Et() >10 GeV & 0.2<R provides ~13600 photons with purity ~96%

- Tight cut allows to identify 62% of photons with the signal-to-background ratio of 4.2 (ee ) and 6.7() in the invariant mass range 82-95 GeV.- Z -> ee : set of cuts 82<M(ee)<95 GeV & 50<M(ee)<82 GeV & Et() >9 GeV & 0.2<R<1.6 provides ~8200 photons with purity ~90%

-1

-1

E. Soldatov Pure photon sample selection 27

Backup slides

27

E. Soldatov Pure photon sample selection 28

Resolution of reconstructed invM(eeee), FSR only

trutheeZ

trutheeZ

recoee

M

MM

28

E. Soldatov Pure photon sample selection 29

EETT andand distributions for reconstructed ISR photons distributions for reconstructed ISR photons (background) matched with truth and reconstruction (background) matched with truth and reconstruction

efficiencyefficiencyET Efficiency

- without tight cut

- with tight cut

29

EfficiencyET, MeVET, MeV

E. Soldatov Pure photon sample selection 3030

With the cut Mee<95 GeV averaged S/B=1.4 and statistics is 89% of the initial one.

Mee, MeVMee, MeV

S/B ratioRemaining statistics

A choice of the mass window for A choice of the mass window for ZZeeeegg - decay study - decay study (I)(I)

Here presented the remaining signal statistics and averaged s/b ratio (integral from 0 to Mee limit) as a function of the upper Mee limit.

E. Soldatov Pure photon sample selection 3131

Further interval mass 82< Mee<95 GeV, with averaged S/B=1.87 will be used for analysis (statistics is 82% of the initial one).

Mee, MeVMee, MeV

S/B ratio

A choice of the mass window for A choice of the mass window for ZZeeeegg - decay study (II)- decay study (II)Here presented the remaining signal statistics and averaged s/b ratio Here presented the remaining signal statistics and averaged s/b ratio (integral from bottom M(integral from bottom Meeee limit limit to 95 GeV) as a function of bottom Mto 95 GeV) as a function of bottom Meeee

limit.limit.

Remaining statistics

E. Soldatov Pure photon sample selection 3232

- with tight cut

- without tight cut

S/B ratio

Differential distributions of s/b ratio as a function Differential distributions of s/b ratio as a function of InvMof InvMeeee for InvM for InvMeeee

With the tight cut averaged S/B ratio in the chosen mass window is 4.22. Remaining statistics is 62% of initial.

Mee, MeV

E. Soldatov Pure photon sample selection 3333

Mee, MeV Mee, MeV

Remaining statistics S/B ratio

Bottom cut for theBottom cut for the Invariant Mass of Invariant Mass of two two electrons. electrons. Upper cut is 82 GeV Upper cut is 82 GeV (additional to the tight cut)(additional to the tight cut)

Here presented the remaining signal statistics and averaged s/b ratio Here presented the remaining signal statistics and averaged s/b ratio (integral from bottom M(integral from bottom Meeee limit limit to 82 GeV) as a function of bottom Mto 82 GeV) as a function of bottom Meeee

limit.limit.

For the cut Mee>50 GeV averaged S/B=6.19 and remaining statistics is 50% from the initial one.

E. Soldatov Pure photon sample selection 34

- with 50<InvM50<InvMeeee<82 cut only<82 cut only

- with tight cut addition

S/B ratio

34

Mee, MeV

For black line averaged S/B=2.75 and remaining statistics is 65% from the initial one. For green – it was given on the previous slide (6.19 and 50% accordingly).

Differential distributions of s/b ratio as a function of Differential distributions of s/b ratio as a function of InvMInvMeeee for 50<InvM for 50<InvMeeee<82 cut only and with addition of the <82 cut only and with addition of the

tight cuttight cut

E. Soldatov Pure photon sample selection 35

Comparison of the s/b ratio for Comparison of the s/b ratio for tight cut tight cut onlyonly with the s/b ratio for with the s/b ratio for 50<InvM50<InvMeeee<82 <82

cut onlycut only

- Tight cut

InvMInvMee ee cut / Tight cutcut / Tight cutS/B ratio

Mee, MeVMee, MeV

- Mee cut

35

It could be useful for the tight cut efficiency validation.It could be useful for the tight cut efficiency validation.

E. Soldatov Pure photon sample selection 3636

S/B ratio

Differential distributions of s/b ratio as a function of InvMee for 50<InvMee<82 and DR<1.6 cuts and with addition of the tight

cut

Mee, MeV

For black line averaged S/B=3.98 and remaining statistics is 51% from the initial one. For green – it was given on the previous slide (7.47 and 39%

accordingly).

- with 50<InvM50<InvMeeee<82 <82 and DR<1.6 cuts - with tight cut addition

E. Soldatov Pure photon sample selection 37

Comparison of the s/b ratio for Comparison of the s/b ratio for tight cut onlytight cut only with the s/b with the s/b ratio for ratio for 50<InvM50<InvMeeee<82 and <82 and R<1.6 cuts onlyR<1.6 cuts only

- tight cut only

- InvMInvMeeee&&RRupup cuts only cuts only

InvMInvMeeee&&RRupup cuts / Tight cut cuts / Tight cutS/B ratio

Mee, MeVMee, MeV

37

It could be useful for the tight cut efficiency validation.It could be useful for the tight cut efficiency validation.

E. Soldatov Pure photon sample selection 38

EfficiencyTruth JetsReco JetsET, MeV ET, MeV

EETT and and distribution of the truth and reconstructed (in distribution of the truth and reconstructed (in

photon container) jets and efficiency of such reconstructionphoton container) jets and efficiency of such reconstruction

EfficiencyTruth JetsReco Jets

ET, MeV

38

E. Soldatov Pure photon sample selection 3939

Mee, MeV

Mee, MeV

Truth signal, background and ratio of s/b ratio as a function of Truth signal, background and ratio of s/b ratio as a function of invariant mass eeinvariant mass ee in full area and in window in full area and in window

FSR - Signal

Mee, MeV Mee, MeV

Mee, MeV Mee, MeV

Truth Jets Truth S/B ratio

E. Soldatov Pure photon sample selection 40

EETT and and distributions of all reconstructed photons (with distributions of all reconstructed photons (with and without tight cut)and without tight cut)

The peak in ET distributions around half Z-boson mass corresponds to an error in the reconstruction of the conversion photon from one of the Z-boson’s electron.

40

ET

With tight cut

With tight cut

ET, MeV

ET ET, MeV

E. Soldatov Pure photon sample selection 41

Backup slides Z ->

41

E. Soldatov Pure photon sample selection 42

RR distributions for signal and background (TRUTH)

Signal (FSR) ISR

Brem Jets treated as

R

R

R

R -to-closest lepton angle22 R

E. Soldatov Pure photon sample selection 43

Resolution of reconstructed M()

t r u t he eZ

t r u t he eZ

r e c oe e

M

MM

E. Soldatov Pure photon sample selection 44

Control plots for reconstruction - Z ->

FSR photons reconstruction

Efficiency vs Et()Et() distribution

Efficiency vs) ) distributionEt() [MeV] Et() [MeV]

) )

Tight Cut