Документ взят из кэша поисковой машины. Адрес оригинального документа : http://zebu.uoregon.edu/~uochep/talks/talks02/iwasaki.pdf Дата изменения: Thu Mar 13 01:24:33 2003 Дата индексирования: Tue Oct 2 10:36:47 2012 Кодировка: Поисковые слова: п п п п п п п п п п п п п п п п п п п п п

Full-Simulation Calorimeter Study For Energy Flow Jet reconstruction

Jan. 8, 2002 M. Iwasaki University of Oregon


In the future linear collider experiments, many physics processes have multi jets in the final states ! Good Jet reconstruction is essential Jet Reconstruction ... Energy Flow method We use Tracker for charged particles Calorimeters neutral particles
For example, in e+e- !tt (6jets,4jets) events, Charged particles carry ~60% of event energy ! Tracker Photons 20% EM Cal Neutral hadrons 10% HAD Cal Neutrino 10% cannot detect
(ECM = 500GeV, Generator level)

At first, reconstruction in EM CAL is important!!


In this talk, 1. 2. 3.

I'll report Clustering algorithm GISMO vs GEANT4 Photon reconstruction study

In this study, we use e+e- ! tt (6jets, 4 jets) events Generated by Pandora-Pythia Through Full detector simulation (GISMO & GEANT4) Assuming SDMar01 detector parameter:
W-Si EM cal granularity 3.7 mrad (5mm segmentation) # layers 30 layer inner radius 127 cm B = 5.0 Tesla ECM = 500 GeV, Include ISR, Beamstrahlung, Parton showering


1) Clustering algorithm So far we used cheater algorithm to form clusters ! gather all CalHits associated to the same particle
Single particle generation ()

EMCalHit HADCalHit

Even there are scattering particles ... Make 1 cluster by gathering these CalHits


In real experiments, we cannot associate the hits from scattered particles to the original particle ! Cheater algorithm is not a realistic clustering method We introduce modified clustering algorithm: 1) Form a cluster by Cheater algorithm 2) Make cluster(s) from grouping CalHits
1. At 1st layer, gather the neighboring hits 2. Calculate energy-weight mean " reference position 3. Go to the next layer 4. Gathering hits within a cone from the reference position 5. Repeat 2 to 4 cone width: st layer 1 20mrad 30mrad (SD)


and between the near calhit in the same particles


Now we have more realistic clustering
( Below threshold )

Cluster (Starting position) EMCalHit HADCalHit

... This clustering will be available from LCDROOT V3.5


As we showed, there are one or more clusters from one particle

(Cluster Ecls>0.1GeV, |cos|<0.9)

Average # of clusters in a particle ... 1.5 Scattered particles carry ~15% of total cluster energy sum Scattered particle-original particle distance ... peaked around 5cm


2) GISMO vs GEANT4 (Now GEANT4 available : See Dr. Abe's talk) In this study, we use two simulators: GISMO & GEANT4 EMCAL hit energy distributions for EM particles

EGEANT4
SMO

(~10%) (Use same energy scale)


EMCAL hit energy distributions for hadrons and muons

GISMO ... less high energy tail


Max hit energy in a cluster (EM particles)


Max hit energy in a cluster (hadron, muon)

GISMO/GEANT4 quite different shape... need more study in future


3-1) selection by transverse information
To separate Charged/Neutral Clusters we see track-cluster matching
1) Extrapolate Charged tracks to the Cluster radius, 2) Associate the nearest track to the cluster

Apply a cut: Track-cluster distance > 2.5 cm ! selection 48% 98% (for Ecls>0.2GeV clusters: SDMar01)



selection by longitudinal information

We determine the longitudinal shape by fitting

Fitting function: p1 ( (p2x)

p3-1

exp(-p2x) )/ (p3) GISMO 0.299 0.599 2.220

x:distance (cm)

GEANT4 p1: 0.316 p2: 0.634 p3: 2.262

(SDMar01)





Summary of selection (SDMar01 detector)
GEANT4 Ecls>0.2GeV Tack-cluster cut Longitudinal cuts GISMO Ecls>0.2GeV Tack-cluster cut Longitudinal cuts 32552 31762 27586 33383 8000 2060 25506 24266 3258 36% 50% 84% 98% 85% 33974 33170 29148 Not 33471 8352 2316
scattered

purity 36% 48% 85%

efficiency

27999 26907 2760

98% 86%

·Transverse information is effective to reject charged particles ·Longitudinal information is effective to reject both charged and scattered clusters Overall selection performance ... ~85% purity, ~85% efficiency


Mass reconstruction
Using the energy flow object (track+neutral clusters), reconstruct W and Top-quark in e+e- ! tt events (6jets & 4jets) GEANT4 W mass error Top mass error 67.1±15.9 GeV (28%) 141.0±33.5 (24%) Track + Track + (true) 70.2±16.9 (24%) 147.0±31.7 (22%) Track + (true) + h0(true) 77.2±15.1 (20%) 159.7±30.7 (19%) GISMO Track + 72.2±18.6 Track + (true) 75.1±18.0 Track + (true) + h0(true) 84.4±16.4 (26%) 149.2±35.1 (24%) (24%) 156.5±35.9 (23%) (19%) 177.3±35.9 (20%)

No selection cut (Xjet cut for example) applied

True-/selected- difference ... 2~3 % ! We get very good photon-selection performance Adding the neutral hadron clusters at HADCAL ! can improve mass resolution 3~4%


Summary
1) We update the clustering algorithm for more realistic simulation studies 2) GISMO / GEANT4 comparison now GEANT4 is ready ... Many thanks to T.Abe significance difference in hadron and muon ! Need mode study 3) Photon reconstruction in EMCal Using Transverse+Longitudinal information, we get good reconstruction ~85% ~85% 4) HAD Cal clusters improve 3-4% jet-mass resolution neutral hadron reconstruction study ! Future plan


In the future linear collider experiments, many physics processes have multi jets in the final states ! Good Jet reconstruction is essential Jet reconstruction ... Energy Flow method
Tracker Charged hadron, electron Neutral hadron, photon Calorimeter

Average momentum of charged particle ~ 2.4 GeV @ECM = 500GeV .... P(Tracker) is better than E(Calorimeter)


Summary of selection (SDMar01 detector)
GEANT4 Ecls>0.2GeV Tack-cluster cut 2 for <5 Depth(>min-I)<2.5cm 0.5 < MaxE depth < 7cm GISMO Ecls>0.2GeV Tack-cluster cut 2 for <5 Depth(>min-I)<2.5cm 0.5 < MaxE depth < 7cm 33974 33170 30606 29785 29148 32552 31762 29084 28237 27586 Not scattered 33471 27999 8352 26907 2656 3727 2470 3176 2316 2760 Not scattered 33383 25506 8000 24266 2634 4396 2257 3702 2060 3258 36% 48% 83% 84% 85% 36% 50% 81% 83% 84% 98% 90% 88% 86% 98% 89% 87% 85%