Calorimeter PIDs Vanya Belyaev, Ioury Gouz, Vladimir Romanovsky, Grygory Rybkin

Vanya Belyaev Vanya Belyaev

CERN & ITEP/MoscowCERN & ITEP/Moscow

Calorimeter PIDsVanya Belyaev, Ioury Gouz, Vladimir Romanovsky, Grygory Rybkin

18 March'2k+318 March'2k+3 CaloSoft CaloSoft Vanya Belyaev CERN & ITEP/MoscoVanya Belyaev CERN & ITEP/Moscoww

22

Calorimeter Particle ID techniques

ee++/e/e-- PIDs PIDs

• EcalPIDeEcalPIDe• derived from derived from 22 match match

• PrsPIDe PrsPIDe • derived from energy derived from energy

deposition in Prsdeposition in Prs

• BremPIDe BremPIDe • derived from brem derived from brem 22

matchmatch

• HcalPIDeHcalPIDe• derived from energy derived from energy

deposition in Hcal along deposition in Hcal along the particle trajectorythe particle trajectory

++-- PIDs PIDs

• HcalPIDmuHcalPIDmu

• Derived from energy Derived from energy deposition in Hcal along deposition in Hcal along particle trajectoryparticle trajectory

• EcalPIDmuEcalPIDmu

• Derived from energy Derived from energy deposition in Ecal along deposition in Ecal along particle trajectoryparticle trajectory

//00 PIDs PIDs

• CaloMatchCaloMatch

• ShowerShapeShowerShape

• SpdPrs SpdPrs

• ClusterMassClusterMass


33

Technique

•All PIDs evaluate All PIDs evaluate LLLL

•difference in log-difference in log-likelihoods between the likelihoods between the particle hypothesis and particle hypothesis and the “background” the “background”

•The “measurable” value The “measurable” value (e.g. (e.g. 22 or or EEPrsPrs) is ) is converted into converted into LLLL by by comparison with comparison with “reference” distributions “reference” distributions for “Signal” and for “Signal” and “Background”“Background”

•Signal Signal • The associated MC The associated MC

particle has “correct” particle has “correct” typetype

•BackgroundBackground• The associated MC The associated MC

particle has “wrong” typeparticle has “wrong” type

•Background does not Background does not include “ghosts” include “ghosts”

•Momentum dependencies Momentum dependencies are taken into accountare taken into account


44

EcalPIDe - procedure

•Start from Start from 22

• ““ElectronMatch”ElectronMatch”

•Select minimal Select minimal 22 for for given given TrStoredTrackTrStoredTrack

•Produce 2D reference Produce 2D reference distributions for large distributions for large reference data sets reference data sets 800k for channels & b-800k for channels & b-inclusiveinclusive

tanh(tanh(22/100)/100)

versusversus

tanh(p/50 GeV/c)tanh(p/50 GeV/c)

•Normalize reference Normalize reference distributions to 1 in each distributions to 1 in each bin on “momentum” bin on “momentum”

•Make a ratio of normalized Make a ratio of normalized distributions for “signal” distributions for “signal” and “background”and “background”

•Take a (natural) logarithm Take a (natural) logarithm of this ratioof this ratio

•The resulting histogram is The resulting histogram is used for evaluation of used for evaluation of

LLLL: EcalPIDe: EcalPIDe


55

EcalPIDe

Raw distributions for 800k events Raw distributions for 800k events

tanh(p/50 GeV/c)tanh(p/50 GeV/c)tanh(tanh(22/100)/100)

ee++/e/e--

hh//ghostsghosts

ee++/e/e--

hh//ghostsghosts


66

EcalPIDe

Raw distributions for 800k events Raw distributions for 800k events

tanh(p/50 GeV/c)tanh(p/50 GeV/c) tanh(p/50 GeV/c)tanh(p/50 GeV/c)

tanh

(ta

nh(

22 /10

0)/1

00)

tanh

(ta

nh(

22 /10

0)/1

00)

ee++/e/e--


77

EcalPIDe

Normalized distributions Normalized distributions

ee++/e/e--


tanh

(ta

nh(

22 /10

0)/1

00)

tanh

(ta

nh(

22 /10

0)/1

00)



88

EcalPIDe

Results and Results and performanceperformance

LLLL


tanh

(ta

nh(

22 /10

0)/1

00)

LLLL

ee++/e/e--

hh//ghostsghosts


99

PrsPIDe


LLLL

LLLL


tanh

(Eta

nh(E

PrsPrs/1

00 M

eV)

/100

MeV

) ee++/e/e--

hh//ghostsghosts

LLLL


1010

BremPIDe

Results and Results and performanceperformanceLLLL

tanh

(ta

nh(

Bre

mB

rem

22 /20

0)/2

00)

tanh(p/50 GeV/c)tanh(p/50 GeV/c) LLLL

ee++/e/e--

hh//ghostsghosts


1111

HcalPIDe


tanh

(Eta

nh(E

Hca

lH

cal/5

GeV

)/5

GeV

)


LLLL ee++/e/e--

hh//ghostsghosts

LLLL


1212

ID

EEHcalHcal/E/EEcalEcal spectra spectra

tanh(Etanh(EHcalHcal/10 GeV)/10 GeV) tanh(Etanh(EEcalEcal/5 GeV)/5 GeV)

++//--

h/eh/eghostsghosts

++//--

h/eh/eghostsghosts


1313

HcalPID


LLLL

tanh

(Eta

nh(E

Hca

lH

cal/1

0 G

eV)

/10

GeV

)


LLLL

LLLL

++//--

h/eh/eghostsghosts

LLLL


1414

EcalPID


tanh

(Eta

nh(E

Eca

lE

cal/5

GeV

)/5

GeV

)


LLLL

LLLL

++//--

h/eh/eghostsghosts


1515

(intermediate) summary

Calorimeter PIDsCalorimeter PIDs

•New package New package Calo/CaloPIDs v1r0Calo/CaloPIDs v1r0

•6 different PID 6 different PID techniquestechniques•4 techniques for 4 techniques for ee++/e/e-- ID ID

•2 techniques for 2 techniques for ++//-- ID ID

•All information is All information is available in DaVinciavailable in DaVinci

•Reference histograms Reference histograms could be updatedcould be updated

To be doneTo be done

•Combine all information Combine all information to produce to produce• CaloPIDeCaloPIDe forfor e e++/e/e-- ID ID

• CaloPIDCaloPID forfor ID ID

Work onWork on //00 PID PID


1616

Sum CaloPIDs in a blind way:

Results and performance (no RICH, no Muon, only Calo)Results and performance (no RICH, no Muon, only Calo)

++//--

h/eh/eghostsghosts

ee++/e/e--

hh//ghostsghosts

LLLL(() combined) combinedLLLL(e(e++/e/e--)-combined)-combined

Calorimeter PIDs Vanya Belyaev, Ioury Gouz, Vladimir Romanovsky, Grygory Rybkin

Documents

e e id calopidm

e e id2 techniques

combineddlle e

calorimeter pidsvanya

heghostse ehmghostsdllm

particle hypothesis

d reference distributions

energy deposition