[PPT] - Hadronic Shower Reconstruction in an Imaging Calorimeter Marina PowerPoint Presentation

SLIDE 1

Hadronic Shower Reconstruction in an Imaging Calorimeter

Marina Chadeeva, ITEP, Moscow

for the CALICE Collaboration

Marina Chadeeva (ITEP) 2011 IEEE NSS, Valencia, Spain October 27, 2011 1 / 15

SLIDE 2

Imaging calorimeters for PFA-based reconstruction

Detectors for a future Linear Colliders

ILD event display

Goal for detectors at ILC:

σjet Ejet ∼ 3-4%

Classic calorimeter approach not sufficient Possible solution: Particle Flow Analysis ⇒ high granularity required CAlorimeters for LInear Collider Experiments CALICE calorimeter prototypes high transversal and longitudinal granularity test of detector concepts test of PFA concept test of MC models study of 3D shower profiles

Scintillator tiles assembled in one layer of hadronic calorimeter Marina Chadeeva (ITEP) 2011 IEEE NSS, Valencia, Spain October 27, 2011 2 / 15

SLIDE 3

Imaging calorimeters for PFA-based reconstruction

CALICE test beam setup

Test beam campaign: since 2006 up to now at DESY, CERN, FNAL Muons, electrons, hadrons in the energy range 1-180 GeV, different setup configurations CALICE test beam setup at CERN in 2007 ECAL: Si-W, ∼0.8λI (30 layers), 18x18cm2, ∼10000 cells: 1x1cm2 HCAL: Sc-Fe, ∼4.5λI (38 layers), ∼1x1m2, 7608 tiles: 3x3, 6x6, 12x12cm2 analogue with SiPM read-out TCMT: Sc-Fe, ∼5λI (16 layers), 90x90cm2, 5-cm strips with SiPM read-out Units of MIP (visible signal from Minimum Ionizing Particle) used to equalize cell-by-cell response

30 GeV pion shower in ECAL and HCAL Marina Chadeeva (ITEP) 2011 IEEE NSS, Valencia, Spain October 27, 2011 3 / 15

SLIDE 4

Imaging calorimeters for PFA-based reconstruction

Advantages of high granularity

Hadronic shower structure First inelastic interaction Identification of track segments and high density clusters Spatial energy density distribution 30 GeV pion event with track in ECAL and hits above 3.5 MIP shown in red Particle Flow Analysis Possibility to disentangle showers induced by charged and neutral particles Software compensation Improvement of the energy resolution by means of software compensation techniques based on the analysis of the detailed energy density spectra

Marina Chadeeva (ITEP) 2011 IEEE NSS, Valencia, Spain October 27, 2011 4 / 15

SLIDE 5

Software compensation Basic idea and techniques

Software compensation: basic idea and techniques

Hadronic shower comprises electromagnetic and hadronic components with significant event-by-event fluctuations of electromagnetic fraction fEM Non-compensating calorimeter: different response to electrons and hadrons ⇒ Hadron energy resolution is deteriorated w.r.t. electromagnetic one Software compensation: take into account fEM fluctuations to improve resolution Local Compensation technique (LC) weighting of signals of individual cells depending on the cell energy density

Energy density spectrum

Global Compensation technique (GC) applying one weight calculated from the energy density spectrum to energy sum

Correlation between global compensation factor and HCAL energy sum

Both methods: energy dependent weights, parameters of the energy dependence extracted from test beam data, do not require a prior knowledge of particle energy

Marina Chadeeva (ITEP) 2011 IEEE NSS, Valencia, Spain October 27, 2011 5 / 15

SLIDE 6

Software compensation Application to test beam data

Software compensation for π− and π+ test beam data

Hadron energy reconstruction Ereco = E track

ECAL + EHCAL + ETCMT

Selected events with track in ECAL EHCAL non-corrected or corrected Weights for software compensation: depend on total event energy calculated from uncorrected Ereco energy dependence parameters extracted from data

[GeV]

beam

E

10 20 30 40 50 60 70 80 90

beam

)/E

beam

E

reco

(E

0.05
0.04
0.03
0.02
0.01

0.01 0.02 0.03 0.04 0.05

π

Initial: + π Initial:

π

GC: + π GC:

π

LC: + π LC:

π

Systematics for initial

CALICE Preliminary Relative residuals to beam energy

Hadron energy resolution

[GeV]

beam

E

10 20 30 40 50 60 70 80 90

reco

/E

reco

σ

0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22

c/E ⊕ b ⊕ E Fit: a/ 0.3% c = 0.18 ± 0.4% b = 1.6 ± Initial: a = 57.6 0.2% c = 0.18 ± 0.3% b = 1.6 ± GC: a = 45.8 0.2% c = 0.18 ± 0.3% b = 1.6 ± LC: a = 44.9

π

+

π

CALICE Preliminary

Stochastic: from ∼58% ↓ to ∼45% Constant: 1.6% - not changed Noise: 0.18 GeV fixed for full setup Similar improvement of relative resolution for π− and π+

Marina Chadeeva (ITEP) 2011 IEEE NSS, Valencia, Spain October 27, 2011 6 / 15

SLIDE 7

Software compensation Application to test beam data

Software compensation: improvement of resolution

Energy distributions before and after compensation ( χ2

NDF < 2 for Gaussian fits)

Relative improvement of absolute resolution 12% < σSC/σinitial < 25% Similar improvement for π− and π+ Local approach gives 3% better improvement in the energy range 25-60 GeV than the global one Global uses twice as less parameters as the local

[GeV]

beam

E

10 20 30 40 50 60 70 80 90

initial

σ /

SC

σ

0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1

π

GC:

+

π GC:

π

LC:

+

π LC:

CALICE Preliminary

Reconstructed energy, GeV

2 4 6 8 10 12 14 16 18

Events / 0.5 GeV

1000 2000 3000 4000 5000 6000

10 GeV

π

Initial LC GC CALICE Preliminary (a) Reconstructed energy, GeV

25 30 35 40 45 50 55 60

Events / 1.0 GeV

500 1000 1500 2000 2500 3000 3500 4000 4500 5000

40 GeV

+

π Initial LC GC CALICE Preliminary (b) Reconstructed energy, GeV

60 65 70 75 80 85 90 95 100

Events / 1.0 GeV

1000 2000 3000 4000 5000 6000 7000

80 GeV

π

Initial LC GC CALICE Preliminary (c)

Marina Chadeeva (ITEP) 2011 IEEE NSS, Valencia, Spain October 27, 2011 7 / 15

SLIDE 8

Software compensation Comparison with MC

Software compensation: simulated samples and data

Local compensation

[GeV]

beam

E

10 20 30 40 50 60 70 80 90

reco

/E

reco

σ

0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22

c/E ⊕ b ⊕ E Fit: a/ 0.2% c = 0.18 ± 0.3% b = 1.6 ± Data LC: a = 44.9 0.1% c = 0.18 ± 0.2% b = 2.3 ± QGSP_BERT LC: a = 41.7 0.1% c = 0.18 ± 0.3% b = 3.2 ± FTF_BIC LC: a = 40.2

π

Data LC:

+

π Data LC:

π

QGSP_BERT LC:

+

π QGSP_BERT LC:

π

FTF_BIC LC:

+

π FTF_BIC LC:

CALICE Preliminary

[GeV]

beam

E

10 20 30 40 50 60 70 80 90

initial

σ /

SC

σ

0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1

π

Data LC:

+

π Data LC:

π

QGSP_BERT LC:

+

π QGSP_BERT LC:

π

FTF_BIC LC:

+

π FTF_BIC LC:

CALICE Preliminary

Global compensation

[GeV]

beam

E

10 20 30 40 50 60 70 80 90

reco

/E

reco

σ

0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22

c/E ⊕ b ⊕ E Fit: a/ 0.2% c = 0.18 ± 0.3% b = 1.6 ± Data GC: a = 45.8 0.3% c = 0.18 ± 0.1% b = 0.0 ± QGSP_BERT GC: a = 42.9 0.2% c = 0.18 ± 0.3% b = 1.8 ± FTF_BIC GC: a = 42.4

π

Data GC:

+

π Data GC:

π

QGSP_BERT GC:

+

π QGSP_BERT GC:

π

FTF_BIC GC:

+

π FTF_BIC GC:

CALICE Preliminary

[GeV]

beam

E

10 20 30 40 50 60 70 80 90

initial

σ /

SC

σ

0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1

π

Data GC:

+

π Data GC:

π

QGSP_BERT GC:

+

π QGSP_BERT GC:

π

FTF_BIC GC:

+

π FTF_BIC GC:

CALICE Preliminary

GEANT 4.9.4 QGSP BERT FTF BIC parameters for compensation extracted from data Local: MC follows data Global: MC predicts further improvement above 40 GeV

Marina Chadeeva (ITEP) 2011 IEEE NSS, Valencia, Spain October 27, 2011 8 / 15

SLIDE 9

PFA test using test beam data

10 GeV and 30 GeV pion showers at 12 cm 10 GeV and 30 GeV pion showers at 24 cm

Pairs of single particle events from the CALICE prototype are superimposed and mapped into ILD geometry PandoraPFA was used as a reconstruction tool Goal is to compare disentangling efficiency: test beam data vs. GEANT 4.9.2 simulations Estimated confusion term: RMS deviation of a neutral cluster reconstructed energy from its measured energy in the vicinity of a charged cluster Confusion term agrees for QGSP BERT and data

Marina Chadeeva (ITEP) 2011 IEEE NSS, Valencia, Spain October 27, 2011 9 / 15

SLIDE 10

PFA test using test beam data

Summary

Hadron energy resolution of the CALICE scintillator-steel analogue HCAL was estimated for π− and π+ test beam data samples in the range from 10 to 80 GeV intrinsic resolution:

57.5%

√

E/GeV ⊕ 1.6% ⊕ 0.18 E/GeV, linearity of response within ±2%

Local and global software compensation techniques were developed for the CALICE AHCAL and applied to test beam data contribution from stochastic term reduced down to ∼

45%

√

E/GeV

PFA performance was compared for CALICE test beam data and GEANT 4.9.2 simulated samples similar performance observed for QGSP BERT model and data ⇒ extrapolation to jets in the complete detector is reliable

Marina Chadeeva (ITEP) 2011 IEEE NSS, Valencia, Spain October 27, 2011 10 / 15

SLIDE 11

Backup slides

Marina Chadeeva (ITEP) 2011 IEEE NSS, Valencia, Spain October 27, 2011 11 / 15

SLIDE 12

Backup slides

Event selection and data samples

Sample cleaning

Muons: initial admixture 4-30% in the cleaned sample <0.5% Multiparticle: 1-2% Electrons from π−: ˇ Cerenkov counter Protons from π+: ˇ Cerenkov counter

Training and test subsamples

Samples from different runs of the same beam energy and particle charge are merged Merged samples are split into two subsamples (even and odd event numbers) Statistically independent samples are used to test software compensation approaches set of even subsamples is used to adjust software compensation factors adjusted compensation factors are applied to the set of odd subsamples

Marina Chadeeva (ITEP) 2011 IEEE NSS, Valencia, Spain October 27, 2011 12 / 15

SLIDE 13

Backup slides

Intrinsic AHCAL resolution for single hadrons

10 20 30 40 50 60 70 80 90

[GeV]

reco

E

10 20 30 40 50 60 70 80 90

CALICE Preliminary

π

Single runs:

+

π Single runs:

π

Merged Even:

+

π Merged Even:

π

Merged Odd:

+

π Merged Odd: [GeV]

beam

E 10 20 30 40 50 60 70 80 90

beam

)/E

beam

E

reco

(E

0.02

0.02

Systematic uncertainties δEreco = 0.9% (⊕δEbeam for residuals)

[GeV]

beam

E

10 20 30 40 50 60 70 80 90

reco

/E

reco

σ

0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22

c/E ⊕ b ⊕ E Fit: a/ 0.2% c = 0.18 ± 0.3% b = 1.5 ± Single runs: a = 57.3 0.3% c = 0.18 ± 0.4% b = 1.6 ± Merged Even: a = 57.6 0.4% c = 0.18 ± 0.4% b = 1.4 ± Merged Odd: a = 57.9

π

+

π

CALICE Preliminary

Fit results coincide within errors Stochastic term: ∼

57.5%

√

E/GeV

Constant term: ∼1.5% Noise fixed at 0.18 GeV for full setup Similar resolution for π− and π+

Marina Chadeeva (ITEP) 2011 IEEE NSS, Valencia, Spain October 27, 2011 13 / 15

SLIDE 14

Backup slides

Local Compensation technique (LC)

Energy density (ED) distribution is divided into energy density bins E SC

reco = EECAL + hit Ehit · ωhit + ETCMT

The weights depend on ED and initial reconstructed event energy E (p0, p1, p2 are energy dependent): ωhit = p0(E) · exp(p1(E) · ED) + p2(E) Shape of parameters p0, p1, p2 is found via an iterative procedure using beam energy.

energy density

20 40 60 80 100 120

weight value

0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

energy density

20 40 60 80 100 120

weight value

0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Weight parametrization:

2

* x) + p

1

* exp( p p Colors for beam energies 10GeV, 30GeV, 50GeV, 70GeV

beam energy [GeV]

10 20 30 40 50 60 70 80 90 100

p

1.2 1.4 1.6 1.8 2 2.2 2.4 2.6

beam energy [GeV]

10 20 30 40 50 60 70 80 90 100

1

p

0.7
0.6
0.5
0.4
0.3
0.2
0.1

beam energy [GeV]

10 20 30 40 50 60 70 80 90 100

2

p

0.4 0.5 0.6 0.7 0.8 0.9

Marina Chadeeva (ITEP) 2011 IEEE NSS, Valencia, Spain October 27, 2011 14 / 15

SLIDE 15

Backup slides

Global Compensation technique (GC)

Global compensation factor Cgl calculated on event-by-event basis:

number of shower hits Nav with ehit < e, e is a mean of shower hit energy spectrum number of shower hits Nlim with ehit < elim, elim = 5 MIP Cgl = Nlim

Nav

Mean of global compensation factor Cgl is energy dependent; coefficients a0, a1, a2 to describe this dependence are derived using beam energy Reconstructed energy: E GC

reco = EECAL + Esh · (a0 + a1Esh + a2E 2 sh),

where Esh = Cgl · (EHCAL + ETCMT)

Marina Chadeeva (ITEP) 2011 IEEE NSS, Valencia, Spain October 27, 2011 15 / 15