geant4 physics evaluation with testbeam data of the ......the atlas hadronic end-cap calorimeter a....

XIII International Conference on Calorimetry in High Energy PhysicsPavia, Italy, May 2008

GEANT4 Physics Evaluation with Testbeam Data of

the ATLAS Hadronic End-Cap Calorimeter

A. Kiryunin, H. Oberlack, D. Salihagic, P. Schacht, P. Strizenec

“Simulation” session

May 29, 2008

Calor 2008 May 29, 2008

The validation of GEANT4 physics models is based on the detailedcomparison of experimental data from beam tests of modules of theATLAS hadronic end-cap calorimeter with GEANT4 predictions

• ATLAS hadronic end-cap calorimeter and its testbeam

• GEANT4: simulation framework, parameters and features

• Results of the validation

– 1 –

Calor 2008 May 29, 2008

ATLAS Hadronic End-Cap Calorimeter

• ATLAS hadronic end-cap calorimeter(HEC) — a liquid argon (LAr)

sampling calorimeter with parallelcopper absorber plates

• Pseudorapidity coverage:1.5 < |η| < 3.2

• Two wheels per end-cap: front andrear

• Total thickness:∼1.8 m, ∼103 X0, ∼10 λ

• Wheel outer diameter: ∼4 m

• Each wheel contains 32 identicalmodules

• Granularity ∆η × ∆ϕ:

– 0.1×2π/64 for |η| < 2.5– 0.2×2π/32 for |η| > 2.5

• Four longitudinal layers

– 2 –

Calor 2008 May 29, 2008

Beam tests of HEC modules

• Beam tests of HEC serial

modules in 2000-2001

• H6 beam line of the CERN SPS

• Secondary and tertiary beams:

– charged pions of 10-200 GeV

– electrons of 6-150 GeV– muons of 120, 150 and 180 GeV

• ∼20,000 triggers per run

• Set-up with three front and

three rear HEC modules

– 3 –

Calor 2008 May 29, 2008

GEANT4 based simulations of the HEC testbeam

• Stand-alone code for HEC testbeam simulations

• Detailed description:

– calorimeter modules (sensitive LAr, copper plates, electrodes)

– cryostat and LAr excluder

– beam elements (scintillator counters, MWPCs)

• Studies for validation of GEANT4 physics:– different hadronic physics lists

– influence of the Birks’ law

– time structure of hadronic showers

• GEANT4 simulations:

– version 9.0, released in June 2007

– range cut = 30 µm

– 4 –

Calor 2008 May 29, 2008

Hadronic physics lists

• QGSP

– based on theory-driven models

– quark-gluon-string model for

interactions

– pre-equilibrium decay model for the

fragmentation

• QGSP-BERT

– Bertini cascade model for

particle-nuclear interactions

below ∼10 GeV

• QGSP-BERT-HP

– high precision data-driven modeling

for low energy neutrons

Birks’ law

• Saturation of response in LAr for

particles with large dE/dx

• Parametrization:

∆E′= ∆E

A

1 + cρ

∆E∆x

A = 1

c = 0.0045 g/(MeV cm2)

ρ = 1.396 g/cm3

– 5 –

Calor 2008 May 29, 2008

Time structure of hadronic showers

100 GeV charged pions (Birks’ law ON)

T [ns]0 20 40 60 80 100

T [ns]0 20 40 60 80 100

dE

/ d

T

[MeV

/ns]

-110

1

10

210

310

Visible energy VS TimeVisible energy VS Time

QGSP

QGSP-BERT

QGSP-BERT-HP

T [ns]0 20 40 60 80 100

T [ns]0 20 40 60 80 100

Cu

mu

lati

ve f

ract

ion

C

0

0.2

0.4

0.6

0.8

1

Cumulative Fraction CCumulative Fraction C

T [ns]0 20 40 60 80 100

T [ns]0 20 40 60 80 100

1 -

C

-210

-110

1

1 minus C1 minus C

• QGSP-BERT predicts slower showers than QGSP and QGSP-BERT-HP

• Late energy depositions (after a few tens of nanoseconds) are at the percent level

– 6 –

Calor 2008 May 29, 2008

Measurement of calorimeter signals

• Convolution of time profiles and

shaping functions

• Resulting amplitude — measured at

the position of the maximum of the

shaping function TMAX

• 51 types of HEC channels:

different shaping functions

• For the HEC testbeam:

50 < TMAX < 70 ns

• Measured signal — energy deposition

in a HEC channel integrated over a

few tens of nanoseconds

T [ns]0 200 400 600 800

T [ns]0 200 400 600 800

Sh

apin

g f

un

ctio

n

0

0.5

1

Shaping function for channel type 2Shaping function for channel type 2

• Two types of signal measurements are considered:

– after convolution with a shaping function– no time cut

– 7 –

Calor 2008 May 29, 2008

Simulation / Reconstruction / Analysis

• Simulated samples:– energy scans with negatively charged pions– energy scans with electrons

• 5000 events per beam particle type,

beam energy, physics list and Birks’

law switch

• Ratio of simulation times (for pions):– QGSP-BERT / QGSP = 1.7– QGSP-BERT-HP / QGSP = 4.9

• Energy reconstruction:– following experimental procedure– electromagnetic scale calibration

∗ defined by the sampling fraction∗ returns the total deposited energy for

electrons– cluster of the fix size (effective radius of

35-40 cm)

– Gaussian fit: E0 and σ– no electronic noise added to Monte Carlo

(MC) signals (spread of the noise wassubtracted from the resolution of the

experimental data)

• Analysed variables:– pion energy resolution σ/E0

– ratio e/π (ratio of energies in electronand pion clusters)

– fraction of energies in HEC longitudinallayers for charged pions

– 8 –

Calor 2008 May 29, 2008

Pion energy resolution

[GeV]BEAME0 50 100 150 200

[%

]0

/ E

σ

0

5

10

15

20

25 No time cutNo time cutQGSP-BERT-HP, Birks’ law ON

QGSP-BERT-HP, Birks’ law OFF

QGSP-BERT, Birks’ law ON

QGSP-BERT, Birks’ law OFF

QGSP, Birks’ law ON

QGSP, Birks’ law OFF

Exp

[GeV]BEAME0 50 100 150 200

[%

]0

/ E

σ0

5

10

15

20

25 After convolutionAfter convolution

Time cuts strongly influence the energy resolution

• QGSP-BERT: 10-30 % relative increase

• QGSP and QGSP-BERT-HP: relative increase by ∼5 %

– 9 –

Calor 2008 May 29, 2008

Pion energy resolution: Ratio to experiment

[GeV]BEAME0 50 100 150 200

[GeV]BEAME0 50 100 150 200

EX

P ) 0

/ E

σ /

( M

C ) 0

/ E

σ(

0.6

0.8

1

1.2

1.4

1.6No time cutNo time cut

QGSP-BERT-HP, Birks’ law ON

QGSP-BERT-HP, Birks’ law OFF

QGSP-BERT, Birks’ law ON

QGSP-BERT, Birks’ law OFF

QGSP, Birks’ law ON

QGSP, Birks’ law OFF

[GeV]BEAME0 50 100 150 200

[GeV]BEAME0 50 100 150 200

EX

P ) 0

/ E

σ /

( M

C ) 0

/ E

σ( 0.6

0.8

1

1.2

1.4

1.6After convolutionAfter convolution

QGSP describes well energy resolution below EBEAM ≃ 100 GeV.

– 10 –

Calor 2008 May 29, 2008

Pion energy resolution: Two-term parametrization

• σ/E0 = A/√

EBEAM ⊕ B

• Experimental values:

A = 69 ± 1 %√

GeV , B = 5.8± 0.1 %

• Simulation predictions:

Birks’ Physics After convolution

law list A[%√

GeV ] B [%]

QGSP 68.1 ± 0.8 6.2 ± 0.1OFF QGSP-BERT 57.1 ± 0.7 5.30 ± 0.09

QGSP-BERT-HP 60.2 ± 0.7 5.4 ± 0.1

QGSP 67.9 ± 0.8 6.9 ± 0.1ON QGSP-BERT 58.6 ± 0.7 5.83 ± 0.09

QGSP-BERT-HP 59.6 ± 0.7 6.13 ± 0.09

Birks’ law:• does not change the sampling term• increases the constant term

After convolution and with Birks’ lawswitched ON:• sampling term is described well by QGSP• constant term is predicted better by

QGSP-BERT and QGSP-BERT-HP

– 11 –

Calor 2008 May 29, 2008

Ratio e/π

[GeV]BEAME0 50 100 150

πe

/

1

1.1

1.2

1.3

1.4

1.5

1.6 No time cutNo time cutQGSP-BERT-HP, Birks’ law ON

QGSP-BERT-HP, Birks’ law OFF

QGSP-BERT, Birks’ law ON

QGSP-BERT, Birks’ law OFF

QGSP, Birks’ law ON

QGSP, Birks’ law OFF

Exp

[GeV]BEAME0 50 100 150

πe

/

1

1.1

1.2

1.3

1.4

1.5

1.6 After convolutionAfter convolution

Time cuts strongly influence e/π-ratio for QGSP-BERT: 4-8 % increase.

For QGSP and QGSP-BERT-HP increase is smaller: 1-2 %.

Ratio e/π is 2-3 % larger, when Birks’ law is switched ON for all physics lists.

– 12 –

Calor 2008 May 29, 2008

e/π: Ratio to experiment

[GeV]BEAME0 50 100 150

[GeV]BEAME0 50 100 150

EX

P )π

/ (

e /

MC

)π(

e /

0.8

0.9

1

1.1

1.2

1.3 No time cutNo time cutQGSP-BERT-HP, Birks’ law ON

QGSP-BERT-HP, Birks’ law OFF

QGSP-BERT, Birks’ law ON

QGSP-BERT, Birks’ law OFF

QGSP, Birks’ law ON

QGSP, Birks’ law OFF

[GeV]BEAME0 50 100 150

[GeV]BEAME0 50 100 150

EX

P )π

/ (

e /

MC

)π(

e /

0.8

0.9

1

1.1

1.2

1.3 After convolutionAfter convolution

After applying time cuts and with Birks’ law switched ON: all three physics lists describe e/π ratio well.

– 13 –

Calor 2008 May 29, 2008

Fraction of energy in HEC longitudinal layers

After convolution, Birks’ law ON

0 50 100 150 200

SU

M>

/ EL

AY

ER

F =

<E

0.1

0.2

0.3

0.4

0.5

0.6 Layer 1Layer 1

QGSP

QGSP-BERT

QGSP-BERT-HP

Exp

0 50 100 150 200

0.4

0.45

0.5

0.55

0.6 Layer 2Layer 2

0 50 100 150 200

0

0.05

0.1

0.15

0.2

0.25 Layer 3Layer 3

[GeV]BEAME0 50 100 150 200

0

0.02

0.04

0.06

Layer 4Layer 4

Four HEC longitudinal layers: 8/16/8/8 LAr gaps, 1.5/2.9/3.0/2.8 λ

F =< ELAY ER > /ESUM , where ESUM = Σ < ELAY ER >

– 14 –

Calor 2008 May 29, 2008

Fraction of energy in HEC longitudinal layers: Ratio to experiment

After convolution, Birks’ law ON

0 50 100 150 2000 50 100 150 200

EX

P /

F M

CF

0.7

0.8

0.9

1

1.1

1.2

1.3 Layer 1Layer 1

0 50 100 150 2000 50 100 150 200

0.7

0.8

0.9

1

1.1

1.2

1.3 Layer 2Layer 2

0 50 100 150 2000 50 100 150 200

0.8

1

1.2

1.4

1.6

1.8

2 Layer 3Layer 3

[GeV]BEAME0 50 100 150 200

[GeV]BEAME0 50 100 150 200

0

1

2

3

4 Layer 4Layer 4

QGSP

QGSP-BERT

QGSP-BERT-HP

• Fraction of energy in the second (main) layer is described within a few percent by all physics lists

• QGSP: hadronic showers start earlier and are more compact

• QGSP-BERT and QGSP-BERT-HP: good description of shower profiles (except lowest beam energy)

• Only small differences between “No time cut” and “After convolution” measurements

• No dependence on Birks’ law switch

– 15 –

Calor 2008 May 29, 2008

Conclusions

GEANT4 based simulations of the HEC testbeam were caried out with different physics lists, namely:

QGSP, QGSP-BERT and QGSP-BERT-HP. Influence of the Birks’ law and time cuts on the calorimeterperformance parameters was investigated. Comparison with experimental results, obtained during beam

tests of HEC modules, was done.

• Usage of the Birks’ law increases the e/π-ratio and the constant term of the energy resolution forcharged pions

• QGSP-BERT physics list predicts much slower hadronic showers than QGSP and QGSP-BERT-HP

• Applying of time cuts (following the experimental procedure of signal measurements in calorimetercells) has influence on the energy resolution and response for charged pions

• After applying time cuts and with Birks’ law switched ON: the better description of studied experi-mental parameters, in total, is given by QGSP-BERT and QGSP-BERT-HP

– good description of longitudinal profiles of hadronic showers

– agreement in the e/π-ratio– rather close predictions of the resolution at high beam energies (i.e. of the constant term)

• Questions addressed to GEANT4 experts:

– better description of the sampling term of the energy resolution for charged pions for BERT-basedphysics lists

– decrease of the simulation time for QGSP-BERT-HP or/and improvement of the neutron physics inQGSP-BERT

– 16 –