- Email: [email protected]
Comparison between Geant4, Fluka and the TileCal test-beam data
ARTICLE IN PRESS Nuclear Instruments and Methods in Physics Research A 617 (2010) 74–77
Contents lists available at ScienceDirect
Nuclear Instruments and Methods in Physics Research A journal homepage: www.elsevier.com/locate/nima
Comparison between Geant4, Fluka and the TileCal test-beam data M. Cascella a,M. Gallas b, W. Pokorski b, A. Ribon b a b
INFN and University of Pisa, Italy CERN, Switzerland
For the TileCal Collaboration a r t i c l e in f o
a b s t r a c t
Available online 7 October 2009
We present a study of the signal produced by charged pions of energies ranging between 20 and 350 GeV in modules of ATLAS tile calorimeter. The results from test beam data are compared to the predictions of different Monte Carlo simulations (Geant4 and Fluka). The goal is to assess in a quantitative way how well different Monte Carlo codes can reproduce the distribution of visible energy in the calorimeter and the details of the hadronic shower. & 2009 Elsevier B.V. All rights reserved.
Keywords: TileCal ATLAS Calorimeters Monte Carlo Geant4 Fluka Validation Hadronic showers
Monte Carlo simulation of detector hardware is increasingly important and many aspects of modern experiments demand a reliable software simulation. In ATLAS the calibration scheme envisioned to correct for the calorimeter response to jet depend on the ability of the software simulation to correctly reproduce the details of the hadronic showers. This work was developed to validate the simulation chosen by the ATLAS collaboration (Geant4) but also to compare its predictions to a completely different software simulation (the Fluka package).
1. Experimental setup 1.1. TileCal the ATLAS hadronic calorimeter The ATLAS hadron tile calorimeter (TileCal) is an ironscintillator sampling calorimeter [1]. TileCal is divided into one barrel and two extended barrel cylindrical sections. The sections are formed by 64 modules, each segmented in ‘‘pseudo-projective’’ towers (0:1 0:1 in DZ Df) and in three longitudinal sections. This segmentation is obtained by connecting groups of scintillating tiles of the same readout PMT.
Corresponding author.
E-mail address: [email protected] (M. Cascella). 0168-9002/$ - see front matter & 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2009.09.112
1.2. The TileCal test beam The TileCal stand alone test beam took place at CERN from 2001 to 2003 in the H8 beam of the CERN SPS. Beam energies ranged between 20 and 350 GeV. A stack of several modules was placed on the beam line. On the bottom there was the Module 0 reference prototype, the middle layer was a production barrel module and on the top there were two extended barrel modules. Further details on the experimental setup can be found in Ref. [2]. The beam was pointed along the Z ¼ 0:35 projective tower, hitting TileCal at the center of cell A4 with an approximate angle of y ¼ 203 . The H8 beam is a mixture of pions, protons, muons and electrons; on the beam line there were two wire chambers to monitor the beam, three trigger scintillators and a Cherenkov counter to assist particle identification. 1.3. Particle selection At low energy the Cherenkov counter is used to identify electrons. If we cannot rely on the Cherenkov counter to separate electrons from hadrons we use a combination of two adimensional variables rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi P P ðEac c Eac =Ncell Þ2 X c Ncell Ec P a : ; Ctot ¼ Clong ¼ ð E E beam c cÞ c A sample1;2
ARTICLE IN PRESS M. Cascella et al. / Nuclear Instruments and Methods in Physics Research A 617 (2010) 74–77
Clong is the fraction of energy release in the first two samples of TileCal and Ctot is related to the shower size. The Clong 2Ctot distribution is rotated and projected on C2 the axis of maximum separation [3]. The bias induced by this calorimetric selection is estimated by Monte Carlo simulations and a correction is applied on data. Muons are easily identified as they are minimum ionizing particles and release little energy in TileCal. Table 1 summarizes the beam composition at all the energy points. Electrons are rejected using the Cherenkov counter at 20 GeV while a calorimetric selection is used at 50 and 100 GeV. At 50 GeV protons are separated from pions using the Cherenkov detector. The 20 and 350 GeV beams have negative polarity and the anti-proton contamination is negligible.
2. The Monte Carlo simulations For Geant4 we tested the releases Geant4-07-patch-01 and Geant4-08-01-patch-01, we also used two different physics lists: QGSP and QGSP_BERT [4]. The version of Fluka used in the test is Fluka-2006.3 with the CALORIMEter configuration card [5,6]. In all cases the Birks law was enabled. The beam is simulated reproducing the characteristics of the real beam profile. The primary particle and the generated secondaries are tracked through the detector. In Geant4 the details of the processes simulated are described in the physics list. QGSP (that uses the quark-gluon string model
Table 1 Summary of the beam composition after selection at each energy point. Nominal energy (GeV) 20 50 100 180 350
e þ or e (%)
(Anti)proton (%)
371 0:28 7 0:10 0:07 7 0:05 o 0:1 o 0:1
o 0:1 572 64 7 10 74 7 10 o 0:1
75
for hadronic interactions) and QGSP_BERT (like QGSP, but uses the Bertini intra-nuclear cascade below 10 GeV). Fluka hadronic interactions are based on resonance production and decay below a few GeV, and on the dual parton model above. Both modules include a form of generalized intra-nuclear cascade. Geant4 and Fluka simulations share the same geometry description (through the FLUGG interface for Fluka), and the output of Fluka is produced with a similar format as for Geant4, so that the same digitization code is used for both simulations.
3. Validation results The electromagnetic scale for experimental data and for the Monte Carlo simulations is set using a sample of 20 GeV electrons. Fiducial volume is 70:15 in f (the three layers of the stack) and 7 0:35 in Z around the beam direction. The exact beam composition is reproduced in the Monte Carlo samples.
3.1. Distribution of visible energy Fig. 1 shows the mean value of the energy released in TileCal by a hadron beam for several energies. The mean is extracted with a two sigma Gaussian fit of the energy distribution.
3.2. Longitudinal shower shape In Fig. 2 we show the fraction of the total energy released in each longitudinal sample of the Tile calorimeter. The first section (sample A), of about 1.4 interaction lengths, samples the beginning of the shower development; the second (sample BC) is the longest section ð4lÞ and contains the bulk of the hadronic shower; the last sample (sample D, 2l) measures the energy in the tail of the shower. 3.3. Lateral shower shape The lateral profile of the hadronic shower in TileCal is shown in Fig. 3. Ecore is the energy released in the projective tower hit by the
Total Energy TileCal stand alone test−beam − 2002 setup 0.88
Etot / Ebeam
0.86 0.84 0.82 Data GEANT4 7.0p1 QGSP GEANT4 8.1p1 QGSP GEANT4 7.0p1 QGSP_BERT GEANT4 8.1p1 QGSP_BERT Fluka 2006.3 CALORIMEter
0.8 0.78 ATLAS Preliminary 0.76
0
50
100
150
200 Ebeam (GeV)
Fig. 1. Visible energy in TileCal.
250
300
350
ARTICLE IN PRESS 76
M. Cascella et al. / Nuclear Instruments and Methods in Physics Research A 617 (2010) 74–77
Longitudinal shower shape TileCal stand alone test−beam − 2002 setup
0.7
Sample BC
ESample / Etot
0.6
Data GEANT4 7.0p1 QGSP GEANT4 8.1p1 QGSP GEANT4 7.0p1 QGSP_BERT GEANT4 8.1p1 QGSP_BERT Fluka 2006.3 CALORIMEter
0.5 0.4 Sample A
0.3
ATLAS Preliminary 0.2 0.1 Sample D
0 0
50
100
150
200 Ebeam (GeV)
250
300
350
Fig. 2. Fraction of total energy released in each of the three TileCal samples.
Lateral shower shape TileCal stand alone test-beam - 2002 setup
0.98
ECore / E tot
0.96 0.94 0.92 0.9 0.88
ATLAS Preliminar y
0.86 0.84 0
50
100
150
200 Ebeam (GeV)
250
300
350
Data GEANT4 7.0p1 QGSP GEANT4 8.1p1 QGSP GEANT4 7.0p1 QGSP_BERT GEANT4 8.1p1 QGSP_BERT Fluka 2006.3 CALORIMEter
0.14
EHalo / E tot
0.12 0.1 0.08 0.06 0.04 0
50
100
150
200 E beam (GeV)
250
300
350
Fig. 3. Fraction of energy in the core and the halo of the hadronic shower.
pion beam (0:1 0:1 in DZ Df). The energy in the volume around the tower ðEhalo Þ measures the size of the shower halo.
4. Conclusions In this study we have found that the predictions for the average visible energy of Fluka and Geant4 with the QGSP_BERT
physics list are compatible with the data of the TileCal test beam while Geant4 with QGSP cannot reproduce them. Both Geant4 with the Bertini model and Fluka can reproduce the gross behavior of the hadronic shower (the energy in the middle section and in the core of the shower) but Geant4 showers are shorter and narrower than those observed in the data; Fluka also underestimates the signal in
ARTICLE IN PRESS M. Cascella et al. / Nuclear Instruments and Methods in Physics Research A 617 (2010) 74–77
the tails but is, in general, in better agreement with the experimental measurements.
Acknowledgments The authors wish to thank Alfredo Ferrari, Paola Sala, John Apostolakis and Vladimir Ivanchenko for help and suggestions received.
77
References [1] ATLAS Collaboration, ATLAS tile calorimeter technical design report, ATLAS TDR 3, CERN/LHCC/96-42, 1996, CERN, Geneva, Switzerland. [2] ATL-TILECAL-PUB-2009-002, CERN, 2009. [3] M. Cascella, et al., Comparison of Monte Carlo simulations based on Geant4 and Fluka to the TileCal pion test beam data, submitted for publication. [4] S. Agostinelli, et al., Nucl. Instr. and Meth. A 506 (2003) 250. et al., FLUKA: a multi-particle transport code, CERN-2005-10, INFN/ [5] A. Fasso, TC_05/11, SLAC-R-773, 2005. [6] G. Battistoni, et al., AIP Conf. Proc. 896 (2007) 31.
Contents lists available at ScienceDirect
Nuclear Instruments and Methods in Physics Research A journal homepage: www.elsevier.com/locate/nima
Comparison between Geant4, Fluka and the TileCal test-beam data M. Cascella a,M. Gallas b, W. Pokorski b, A. Ribon b a b
INFN and University of Pisa, Italy CERN, Switzerland
For the TileCal Collaboration a r t i c l e in f o
a b s t r a c t
Available online 7 October 2009
We present a study of the signal produced by charged pions of energies ranging between 20 and 350 GeV in modules of ATLAS tile calorimeter. The results from test beam data are compared to the predictions of different Monte Carlo simulations (Geant4 and Fluka). The goal is to assess in a quantitative way how well different Monte Carlo codes can reproduce the distribution of visible energy in the calorimeter and the details of the hadronic shower. & 2009 Elsevier B.V. All rights reserved.
Keywords: TileCal ATLAS Calorimeters Monte Carlo Geant4 Fluka Validation Hadronic showers
Monte Carlo simulation of detector hardware is increasingly important and many aspects of modern experiments demand a reliable software simulation. In ATLAS the calibration scheme envisioned to correct for the calorimeter response to jet depend on the ability of the software simulation to correctly reproduce the details of the hadronic showers. This work was developed to validate the simulation chosen by the ATLAS collaboration (Geant4) but also to compare its predictions to a completely different software simulation (the Fluka package).
1. Experimental setup 1.1. TileCal the ATLAS hadronic calorimeter The ATLAS hadron tile calorimeter (TileCal) is an ironscintillator sampling calorimeter [1]. TileCal is divided into one barrel and two extended barrel cylindrical sections. The sections are formed by 64 modules, each segmented in ‘‘pseudo-projective’’ towers (0:1 0:1 in DZ Df) and in three longitudinal sections. This segmentation is obtained by connecting groups of scintillating tiles of the same readout PMT.
Corresponding author.
E-mail address: [email protected] (M. Cascella). 0168-9002/$ - see front matter & 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2009.09.112
1.2. The TileCal test beam The TileCal stand alone test beam took place at CERN from 2001 to 2003 in the H8 beam of the CERN SPS. Beam energies ranged between 20 and 350 GeV. A stack of several modules was placed on the beam line. On the bottom there was the Module 0 reference prototype, the middle layer was a production barrel module and on the top there were two extended barrel modules. Further details on the experimental setup can be found in Ref. [2]. The beam was pointed along the Z ¼ 0:35 projective tower, hitting TileCal at the center of cell A4 with an approximate angle of y ¼ 203 . The H8 beam is a mixture of pions, protons, muons and electrons; on the beam line there were two wire chambers to monitor the beam, three trigger scintillators and a Cherenkov counter to assist particle identification. 1.3. Particle selection At low energy the Cherenkov counter is used to identify electrons. If we cannot rely on the Cherenkov counter to separate electrons from hadrons we use a combination of two adimensional variables rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi P P ðEac c Eac =Ncell Þ2 X c Ncell Ec P a : ; Ctot ¼ Clong ¼ ð E E beam c cÞ c A sample1;2
ARTICLE IN PRESS M. Cascella et al. / Nuclear Instruments and Methods in Physics Research A 617 (2010) 74–77
Clong is the fraction of energy release in the first two samples of TileCal and Ctot is related to the shower size. The Clong 2Ctot distribution is rotated and projected on C2 the axis of maximum separation [3]. The bias induced by this calorimetric selection is estimated by Monte Carlo simulations and a correction is applied on data. Muons are easily identified as they are minimum ionizing particles and release little energy in TileCal. Table 1 summarizes the beam composition at all the energy points. Electrons are rejected using the Cherenkov counter at 20 GeV while a calorimetric selection is used at 50 and 100 GeV. At 50 GeV protons are separated from pions using the Cherenkov detector. The 20 and 350 GeV beams have negative polarity and the anti-proton contamination is negligible.
2. The Monte Carlo simulations For Geant4 we tested the releases Geant4-07-patch-01 and Geant4-08-01-patch-01, we also used two different physics lists: QGSP and QGSP_BERT [4]. The version of Fluka used in the test is Fluka-2006.3 with the CALORIMEter configuration card [5,6]. In all cases the Birks law was enabled. The beam is simulated reproducing the characteristics of the real beam profile. The primary particle and the generated secondaries are tracked through the detector. In Geant4 the details of the processes simulated are described in the physics list. QGSP (that uses the quark-gluon string model
Table 1 Summary of the beam composition after selection at each energy point. Nominal energy (GeV) 20 50 100 180 350
e þ or e (%)
(Anti)proton (%)
371 0:28 7 0:10 0:07 7 0:05 o 0:1 o 0:1
o 0:1 572 64 7 10 74 7 10 o 0:1
75
for hadronic interactions) and QGSP_BERT (like QGSP, but uses the Bertini intra-nuclear cascade below 10 GeV). Fluka hadronic interactions are based on resonance production and decay below a few GeV, and on the dual parton model above. Both modules include a form of generalized intra-nuclear cascade. Geant4 and Fluka simulations share the same geometry description (through the FLUGG interface for Fluka), and the output of Fluka is produced with a similar format as for Geant4, so that the same digitization code is used for both simulations.
3. Validation results The electromagnetic scale for experimental data and for the Monte Carlo simulations is set using a sample of 20 GeV electrons. Fiducial volume is 70:15 in f (the three layers of the stack) and 7 0:35 in Z around the beam direction. The exact beam composition is reproduced in the Monte Carlo samples.
3.1. Distribution of visible energy Fig. 1 shows the mean value of the energy released in TileCal by a hadron beam for several energies. The mean is extracted with a two sigma Gaussian fit of the energy distribution.
3.2. Longitudinal shower shape In Fig. 2 we show the fraction of the total energy released in each longitudinal sample of the Tile calorimeter. The first section (sample A), of about 1.4 interaction lengths, samples the beginning of the shower development; the second (sample BC) is the longest section ð4lÞ and contains the bulk of the hadronic shower; the last sample (sample D, 2l) measures the energy in the tail of the shower. 3.3. Lateral shower shape The lateral profile of the hadronic shower in TileCal is shown in Fig. 3. Ecore is the energy released in the projective tower hit by the
Total Energy TileCal stand alone test−beam − 2002 setup 0.88
Etot / Ebeam
0.86 0.84 0.82 Data GEANT4 7.0p1 QGSP GEANT4 8.1p1 QGSP GEANT4 7.0p1 QGSP_BERT GEANT4 8.1p1 QGSP_BERT Fluka 2006.3 CALORIMEter
0.8 0.78 ATLAS Preliminary 0.76
0
50
100
150
200 Ebeam (GeV)
Fig. 1. Visible energy in TileCal.
250
300
350
ARTICLE IN PRESS 76
M. Cascella et al. / Nuclear Instruments and Methods in Physics Research A 617 (2010) 74–77
Longitudinal shower shape TileCal stand alone test−beam − 2002 setup
0.7
Sample BC
ESample / Etot
0.6
Data GEANT4 7.0p1 QGSP GEANT4 8.1p1 QGSP GEANT4 7.0p1 QGSP_BERT GEANT4 8.1p1 QGSP_BERT Fluka 2006.3 CALORIMEter
0.5 0.4 Sample A
0.3
ATLAS Preliminary 0.2 0.1 Sample D
0 0
50
100
150
200 Ebeam (GeV)
250
300
350
Fig. 2. Fraction of total energy released in each of the three TileCal samples.
Lateral shower shape TileCal stand alone test-beam - 2002 setup
0.98
ECore / E tot
0.96 0.94 0.92 0.9 0.88
ATLAS Preliminar y
0.86 0.84 0
50
100
150
200 Ebeam (GeV)
250
300
350
Data GEANT4 7.0p1 QGSP GEANT4 8.1p1 QGSP GEANT4 7.0p1 QGSP_BERT GEANT4 8.1p1 QGSP_BERT Fluka 2006.3 CALORIMEter
0.14
EHalo / E tot
0.12 0.1 0.08 0.06 0.04 0
50
100
150
200 E beam (GeV)
250
300
350
Fig. 3. Fraction of energy in the core and the halo of the hadronic shower.
pion beam (0:1 0:1 in DZ Df). The energy in the volume around the tower ðEhalo Þ measures the size of the shower halo.
4. Conclusions In this study we have found that the predictions for the average visible energy of Fluka and Geant4 with the QGSP_BERT
physics list are compatible with the data of the TileCal test beam while Geant4 with QGSP cannot reproduce them. Both Geant4 with the Bertini model and Fluka can reproduce the gross behavior of the hadronic shower (the energy in the middle section and in the core of the shower) but Geant4 showers are shorter and narrower than those observed in the data; Fluka also underestimates the signal in
ARTICLE IN PRESS M. Cascella et al. / Nuclear Instruments and Methods in Physics Research A 617 (2010) 74–77
the tails but is, in general, in better agreement with the experimental measurements.
Acknowledgments The authors wish to thank Alfredo Ferrari, Paola Sala, John Apostolakis and Vladimir Ivanchenko for help and suggestions received.
77
References [1] ATLAS Collaboration, ATLAS tile calorimeter technical design report, ATLAS TDR 3, CERN/LHCC/96-42, 1996, CERN, Geneva, Switzerland. [2] ATL-TILECAL-PUB-2009-002, CERN, 2009. [3] M. Cascella, et al., Comparison of Monte Carlo simulations based on Geant4 and Fluka to the TileCal pion test beam data, submitted for publication. [4] S. Agostinelli, et al., Nucl. Instr. and Meth. A 506 (2003) 250. et al., FLUKA: a multi-particle transport code, CERN-2005-10, INFN/ [5] A. Fasso, TC_05/11, SLAC-R-773, 2005. [6] G. Battistoni, et al., AIP Conf. Proc. 896 (2007) 31.