ef with simple multi-particle states

EF with simple multi-particle states

Vishnu V. ZutshiNIU/NICADD

Hadron Position Resolution

• Since Eflow invariably involves associating clusters near an extrapolated track …

• How to do this in a digital calorimeter ?• Study this using charged pions• Resolution is defined w.r.t. the MC

extrapolated position

10 GeV charged pionsD

ensi

ty w

eigh

ted

Ener

gy w

eigh

ted

unweighted

10 GeV charged pions

“Density”

• Need a hierarchy in the absence of an energy measurement

• Clumpiness of the surrounding • A simple-minded realization of this used here: di = (1/dRij) where dRij is the angular

distance between cell ‘i’ and cell ‘j’

10 GeV

unweighted

Density weighted

Mea

sure

d re

lativ

e to

the

ener

gy w

eigh

ted

reso

lutio

nsC

ell area at first layer=0.64cm2

10 GeV

Mea

sure

d re

lativ

e to

the

ener

gy w

eigh

ted

reso

lutio

nsC

ell area at first layer=4cm2

Density weighted

unweighted

10 GeV

Mea

sure

d re

lativ

e to

the

ener

gy w

eigh

ted

reso

lutio

nsC

ell area at first layer=6cm2

Density weighted

unweighted

10 GeV

Mea

sure

d re

lativ

e to

the

ener

gy w

eigh

ted

reso

lutio

nsC

ell area at first layer=9cm2

Density weighted

unweighted

10 GeV

Mea

sure

d re

lativ

e to

the

ener

gy w

eigh

ted

reso

lutio

nsC

ell area at first layer=12cm2Density weighted

unweighted

10 GeV

Mea

sure

d re

lativ

e to

the

ener

gy w

eigh

ted

reso

lutio

nsC

ell area at first layer=16cm2Density weighted

unweighted

2GeV Photons

maxima

maxima

maxim

a

Clustering

• Local ‘density’ maxima chosen as seed clusters

• Membership of each cell in the seed clusters decided with a distance function

• Calculate centroids• Iterate steps 2 and 3 till distortion is below

some threshold

Could be unique or shared

Parameters

• Cell thresholds• How many layers to lump together• Neighborhood for maxima search• Minimum no. of layers hit• Neighborhood for membership• Proto-cluster definition• Uniqueness of membership

10 GeV 0H

igh asymm

etryD

ensi

ty w

eigh

ted

10 GeV 0M

edium asym

metryD

ensi

ty w

eigh

ted

10 GeV 0Low

asymm

etryD

ensi

ty w

eigh

ted

10 GeV 0Low

asymm

etryD

ensi

ty w

eigh

ted

10 GeV 0

Recon. energy

Recon. mass

18%

10 GeV 0

Recon. energy

Recon. mass

Energy asymmetryA

= a

bs(E

–

E

)/(E

+ E

)

p

p

Den

sity

wei

ghte

d

p

p

Den

sity

wei

ghte

d

p

E/Egen

Eflow

Cal only

n

EMCal

HCal

n

n

EMCal

HCal

n

p

nre

cE (

)

genE ()

Not reliable due to noncom

pensation

n

• Get the e/pi for the SD detector • Scale the MC truth with that function• Take the ratio of the reconstructed pion

energy with the scaled MC truth• This should have a mean of 1.0 (with an

atrocious resolution) if things are working ok

n

Summary/Outlook

• A first pass clustering/track association algorithm exists applicable to EM/HAD, both analog/digital

• Encouraging results for multiparticle events• More detailed study to expand and enhance

(for instance particle id)• Move to jets• Feedback into calorimeter design