Photon-jet reconstruction with the EEMC – Deuxième Partie Pibero Djawotho Indiana University Cyclotron Facility June 18, 2008 STAR.

Photon-jet reconstruction with the EEMC – Deuxième

PartiePibero Djawotho

Indiana University Cyclotron Facility

June 18, 2008

STASTARR

June 18, 2008 Pibero Djawotho – STAR – UC Davis 2

Dominant background to prompt γ production:π0(η)→γγ

• γ/π0≈1/40 at pT=5 GeV to 1/10 at pT=10 GeV

• dNγ/dpT~exp(-0.69pT) from Pythia 6.406

• Challenge: how low in pT can analysis be reasonably carried out while retaining high efficiency and purity

• Heavily rely on clever software algorithms for γ/π0 separation and specialized subdetectors: shower max and preshower

June 18, 2008 Pibero Djawotho – STAR – UC Davis 3

γ/π0 discrimination in Endcap SMD: Maximum Sided Residual

• Basic idea:– Look at transverse shower profile in the SMD

– γ and e transverse shower profile single peak narrow Gaussian+wide Gaussian with common centroid in each SMD plane (u and v)

– π0→γγ double peak structure: main peak and peaklet (asymmetric π0 decay)

– Fit main peak and compute residual=data-fit on each side of main peak pick maximum residual

– For given energy E, π0 should have more residual than γ

June 18, 2008 Pibero Djawotho – STAR – UC Davis 4

22

2

221

2

1

0 2

)(exp

2

2.0

2

)(exp

2

8.0)(

xx

Pxf

Functional form of fit function

12 5.3

Real data (run=7155062/ev=254105)

June 18, 2008 Pibero Djawotho – STAR – UC Davis 5

Single thrown γ and π0

• 10k γ/π0 each sample• STAR y2006 geometry• z-vertex at 0

• Flat in pT=10-30 GeV/c

• Flat in η=1.0-2.1

Quadratic y(x)=100+0.1x2

June 18, 2008 Pibero Djawotho – STAR – UC Davis 6

Background rejection vs. signal efficiency

75% eff @75% rejection

Use perp distancefrom quadratic toproject in 1D

Not quite the 80-80from original proposalbut this simulation hasmost up-to-date detectorconfiguration.

June 18, 2008 Pibero Djawotho – STAR – UC Davis 7

Background rejection vs. signal efficiency

We start to loseefficiency withthis method athigher γ energies.

June 18, 2008 Pibero Djawotho – STAR – UC Davis 8

Pythia 6.406 prompt γ production in pp collisions at √s=200 GeV

Pythia 6.406 prompt → production subprocesses:•q+qbar → q+γ (10% contribution)•f+fbar → γ+γ•q+g → q+γ (qg Compton scattering dominant subprocess)•g+g → γ+γ•g+g → g+γ

June 18, 2008 Pibero Djawotho – STAR – UC Davis 9

How realistic is simulation of SMD response?

• All shower shapes are normalized to unit area• MC photons are default GEANT+STAR simulation response• Will’s photons are selected from η-region of a π0→γγ finder on Run 6 data• Pibero’s photons are from simple η→γγ finder with soft isolation in SMD and no

EMC clustering on Run 6 data• Conclusion:

– Simulation does not accurately reproduce data– MC shower shapes and RMS are narrower

Comparison of Shower Shapes

0.001

0.010

0.100

1.000

0 5 10 15 20 25

Strip Number

He

igh

t Will's photons

MC photons

Pibero's photons

Comparison of RMS values

0.000

0.020

0.040

0.060

0.080

0.100

0.120

0.140

0 5 10 15 20 25

Strip Number

RM

S u

nc

ert

ain

ty

Will's photons

MC photons

Pibero's photons

June 18, 2008 Pibero Djawotho – STAR – UC Davis 10

How to make MC more realistic

• Compile library of shower shapes from data

• In MC, replace all γ shower shapes (25 strips) with shapes from library after proper energy scaling, translation in SMD plane and superposition on underlying event Data-driven MC

• Library shapes are binned by:– SMD plane (U and V)– Sector configuration (plane

ordering)– Photon energy– Preshower energy Consistency check: data-

driven MC agrees with data!!!

June 18, 2008 Pibero Djawotho – STAR – UC Davis 11

Photons from etas (η→γγ)• Use standard π0 finder with L2-

gamma trigger• 0.45<mγγ<0.65 GeV• pT(η)>6 GeV• Turn off splitting algorithm• 5 MeV seed threshold• No floors• No dead strips• Minimum 20-strip separation

between clusters• Energy sum of middle 5 strips

over 20 strips>70% soft SMD isolation cut

• Require 2 points/plane

S/B better than 1:1

June 18, 2008 Pibero Djawotho – STAR – UC Davis 12

Photons from γ-jets (See Ilya’s talk)

• Select dijets from Run 6

• Define neutral energy fraction REM=(ET(Endcap)+ ET(Barrel))/ET(total)

• REM(jet1)>0.9 and REM(jet2)<0.9

• Number of tracks(jet1)<2

• cos(φ1-φ2)<-0.9 “back-to-back” jets

• 0<number of Endcap towers<3

June 18, 2008 Pibero Djawotho – STAR – UC Davis 13

Shower Shapes

• All shower shapes normalized to unit area

• MC shower shape is narrower

• 3-Gaussian better describes the data (esp. tails)

• All data shower shapes are consistent (γ’s from η’s and γ’s from γ-jets)

June 18, 2008 Pibero Djawotho – STAR – UC Davis 14

Maximum sided residual revisited

• Generate prompt γ with Pythia• Generate QCD background with Pythia• Run through GEANT+STAR reconstruction chain• Replace all MC γ shower shapes with data shapes from library

in appropriate bins• Apply maximum sided residual cut background rejection vs.

signal efficiency

June 18, 2008 Pibero Djawotho – STAR – UC Davis 15

Conclusion and Outlook• γ-jets offer clean probe to ΔG at RHIC by

predominantly sampling qg-Compton channel• Very good agreement between MC and data with

preshower1=preshower2=0 Can achieve 1:1 signal-to-background ratio before any SMD cut

• Ongoing studies to understand discrepancies between MC and data shower shapes with preshower1>0 and preshower2>0

• Analysis of Run 8 data (SVT and support structures removed) once produced will provide crucial information on amount of material (conversion) before the calorimeter

EXTRA SLIDES

June 18, 2008 Pibero Djawotho – STAR – UC Davis 17

STAR Endcap Electromagnetic Calorimeter

• Coverage: 1.086<η<2.0, 0<φ<2π• 12 sectors×5 subsectors×η-bins=720 towers• 1 tower=24 layers:

– Layer 1=preshower-1– Layer 2=preshower-2– Layer-24=postshower

• SMD-u and –v plane at 5X0

• 288 SMD strips/plane/sector