13 July 2005 ACFA8

Gamma Finding procedurefor Realistic PFA

T.Fujikawa(Tohoku Univ.), M-C. Chang(Tohoku Univ.), K.Fujii(KEK), A.Miyamoto(KEK), S.Yamashita(ICEPP), T.Yoshioka(ICEPP)

On behalf of ACFA-SIM-J Group

13 July 2005 ACFA8

Motivation

To get high energy resolution in multi jet event, we use the tracker information for charged particles and calorimeter hit information for neutral particles. (tracker resolution is much better than calorimeter’s.)

Particle Flow Algorithm (PFA) is studied to achieve an event reconstruction. We studied part of PFA, “gamma finding method” for GLD detector, using full simulator (Jupiter).

13 July 2005 ACFA8

1.Small Clustering (collect fired cells with its neighbors) → Define thrust and broadening for each small cluster.2. Gamma Finding (Separate gamma/e from hadron) → Small Cluster based.3. Cluster-Track Matching (Separate charged to neutral) → Calorimeter Cell based.

Note: → Cheating method is used for charged hadrons in

the Endcap region. → Assuming the remaining clusters be neutral hadrons.

Flow

to make P

FO

bject

- Current Scheme of the PFA

Clustering flow for GLD detector

13 July 2005 ACFA8

Detector configuration

40 cm270 cm

210 cm

Full One Tower　 EM + HD

27 X0

6.1λ

Y.Yamaguchi (Tsukuba Univ.)

In current Jupiter geometry, calorimeter has 4cmx4cm/12cmx12cm scintillator tile size in ECAL/HCAL as a default setting. We used this configuration in our study.

13 July 2005 ACFA8

Why good gamma finding is necessary?

Expected jet-event in the future ILC, it has about 20% gamma contribution.

Also, if we can separate gamma (and also electron) well, it make the charged hadron clustering easier.

good gamma/charged hadron separation is necessary!

(Z->qqbar @ Ecm = 91.2 GeV)

K

n pLK

13 July 2005 ACFA8

Flow of the gamma finding

As a first step of the clustering, we collect hit cells with its neighbors (with ~50MeV energy deposit upper limit). We call it “small cluster”. Our gamma finding method is performed based on the small clusters.

1. Find “large” small cluster within ECAL region (we call it “mother small cluster”).

2. Apply “Gamma-Fitter” chi-square /ndf cut. (cut 1)

3. Apply “the distance from the nearest track” cut for mother small cluster. (cut 2)

4. Collect small clusters within a tube.

The flow of Gamma Finding

13 July 2005 ACFA8

Gamma finding method

1. Gamma Fitter (chi-square cut of fit)Find large small cluster within ECAL region and make a small tube along its thrust axis.

Take an average of energy deposit for each 3-layers within a tube and fit the total energy deposit by Gamma distribution function.

The result of chi-square/ndf is used to separate gamma/electron from hadrons.

13 July 2005 ACFA8

Energy deposit in three layersare merged to reduce fluctuation.

Choose large small cluster (mother small cluster) within ECAL region and make a small tube (R = 5cm) along to its thrust axis.

Fit the longitudinal shower profile of the energy deposition within a tube as an electromagnetic cascade (gamma distribution function).

Gamma Fitter: algorithm

HCAL projection axis

take an average energy deposit for each 3-layers

mother small cluster

ECAL

13 July 2005 ACFA8

Gamma Fitter: single particle test

5.0,/ where

)(

)exp()(0

Functionon Distributi Gamma

0

1

bXxt

a

btbtbE

dt

dE a

The result of fitting for a single particle case (gamma at 3GeV)

13 July 2005 ACFA8

3 GeV

10 GeV5 GeV

1 GeV Yellow: gamma

Red : pion

We can separate gamma/electron from pion using chi2/ndf cut.

Comparison of chi2/ndf (single particle)

13 July 2005 ACFA8

Blue : efficiency

Red : purity

Gamma Fitter: efficiency and purity

(Z->qqbar @ Ecm = 91.2 GeV)

The “Gamma fitter” works good for z->qqbar event, too. But it doesn’t have high efficiency for electron/gamma finding. (maximum ~ 90%)

13 July 2005 ACFA8

Chi2/ndf (Z->qqbar @ Ecm = 91.2 GeV)

Yellow: e/gamma

Red : all particles

cut position

We set a cut position at 1.0 for below gamma finding.

(~70% efficiency and ~70% purity)

13 July 2005 ACFA8

2. The distance from the nearest track

Calculate the distance between mother small cluster and its nearest track. The distance is used to separate gamma from charged particles.

Gamma finding method

HCAL

ECAL

gamma

track

distance

mother small cluster

13 July 2005 ACFA8

Distance from the nearest track

Blue : electron

Red : all particles

Yellow: Gamma

(Z->qqbar @ Ecm = 91.2 GeV)

This plot shows the distribution of the distance from the nearest track for all small clusters within ECAL region. (energy-weighted)

13 July 2005 ACFA8

Distance distribution (Z->qqbar @ Ecm = 91.2 GeV)

Blue : electron

Red : all particles

Yellow: Gamma

After Chi2/ndf < 1.0 cut

cut position

We set a lower limit at 13cm. It kills the electron as well, but can reduce large amount of charged hadron contamination. We gain the benefit anyway.

13 July 2005 ACFA8

The result of Gamma finding

1. efficiency

Finally, we got

efficiency ~ 60%

using “gamma fitter cut” and “the distance cut”.

13 July 2005 ACFA8

The result of Gamma finding

2. purity

w/o n,KL contamination

Purity ~ 93%

Purity ~ 99%

contamination is almost caused by n/KL

13 July 2005 ACFA8

The result of energy sum

E/%3.41~ E/%5.34~

Realistic gamma finder Cheated gamma finder

cheated etc.) (neutrinoenergy missing%,2.82~ ,%4.86 :matchingtrack purity

13 July 2005 ACFA8

Other idea (not used now)

1. Energy deposit comparison between ECAL/HCAL regionAssume a tube region and compare the total energy deposit between ECAL/HCAL region within a tube.

ECAL

HCAL

E1

E2

separate events using E1/E2

13 July 2005 ACFA8

2. Energy waited Broadening/Thrust ratio of the clusterCalculate thrust and broadening for the cluster, then calculate these ratio. (EM shower has lower value compare to HD shower’s.)

Other idea (not used now)

ECAL

HCAL

In general, (B/T)EM > (B/T)HD

cluster

cluster

13 July 2005 ACFA8

3. Calculate energy deposition ratio within a tube in ECALSet the tube and “Maximum layer” according to cluster energy, then calculate the energy ratio between upper/lower region of Maximum layer within a tube (region 1 and 2).

Other idea (not used now)

All of them are not effective “after” gamma fitter cut…

ECAL

region 1

region 2

region 1

region 2

gamma hadron

Maximum layer (energy

dependent)

Tube radius is also energy dependent

13 July 2005 ACFA8

Summary

Good gamma finding (clustering) method is necessary for future LC experiment.

We got ~60% efficiency and ~99% purity using Gamma Fitter (chi2/ndf) + the nearest track distance requirement.

Current gamma finding method has high purity but which doesn’t have enough efficiency.

13 July 2005 ACFA8

Next step

Combine other gamma finding method (or study another method other than “gamma fitter”) to get high efficiency/purity.

Do the cut value optimization.

Study other method which use transverse shower profile.

Study the “relationship” with clustering method for charged hadron.

13 July 2005 ACFA8

The result of other gamma finding method with gamma fitter chi-square cut

Backup

13 July 2005 ACFA8

HCAL/ECAL e-dep comparison

After Chisquare/ndf < 1.0 cut

Yellow: e/gamma

Red : all particles

The plot shows the energy deposit comparison between HCAL/ECAL regions after applying the chi-square cut for Z->qqbar event.

13 July 2005 ACFA8

Broadening/Thrust

Yellow: e/gamma

Red : all particlesAfter Chisquare/ndf < 1.0 cut

Broadening/Thrust ratio for the cluster after applying the chi-square cut for Z->qqbar event.