PANDA Electromagnetic Calorimeter (EMC)

FAIR/PANDA/Technical Design Report - EMC

Technical Design Report for:

PANDAElectromagnetic Calorimeter (EMC)

(AntiProton Annihilations at Darmstadt)

Strong Interaction Studies with Antiprotons

PANDA Collaboration

ii PANDA - Strong interaction studies with antiprotons

Cover: The figure shows the barrel (left) and forward endcap part (right) of the electromagnetic calorime-ter. The parts of the barrel calorimeter are cut to get a better view to the inside. Displayed are thePWO crystals with the carbon fiber alveole packs, the support feet connecting to the support beam, heldby the support rings at front and back. On the right side the forward endcap crystals inside the carbonfiber packs are shown along with the insulation (shown transparent) and the eight supporting structureson the outside.

FAIR/PANDA/Technical Design Report - EMC iii

The PANDA Collaboration

Universitat Basel, SwitzerlandW. Erni, I. Keshelashvili, B. Krusche, M. Steinacher

Institute of High Energy Physics, Chinese Academy of Sciences, Beijing, ChinaY. Heng, Z. Liu, H. Liu, X. Shen, O. Wang, H. Xu

Universitat Bochum, I. Institut fur Experimentalphysik, GermanyF.-H. Heinsius, T. Held, H. Koch, B. Kopf, M. Pelizaus, M. Steinke, U. Wiedner, J. Zhong

Universita di Brescia, Brescia, ItalyA. Bianconi

Institutul National de C&D pentru Fizica si Inginerie Nucleara ”Horia Hulubei”, Bukarest-Magurele,Romania

M. Bragadireanu, P. Dan, T. Preda, A. Tudorache

Dipartimento di Fisica e Astronomia dell’Universita di Catania and INFN, Sezione di Catania, ItalyM. De Napoli, F. Giacoppo, G. Raciti, E. Rapisarda

IFJ, Institute of Nuclear Physics PAN, Cracow, PolandE. Bialkowski, A. Budzanowski, B. Czech, S. Kliczewski, A. Kozela, P. Kulessa, K. Malgorzata,

K. Pysz, W. Schafer, R. Siudak, A. Szczurek

Instytut Fizyki, Uniwersytet Jagiellonski, Cracow, PolandW. Bardan, P. Brandys, T. Czyzewski, W. Czyzewski, M. Domagala, G. Filo, D. Gil, P. Hawranek,

B. Kamys, P. Kazmierczak, St. Kistryn, K. Korcyl, M. Krawczyk, W. Krzemien, E. Lisowski,A. Magiera, P. Moskal, J. Pietraszek, Z. Rudy, P. Salabura, J. Smyrski, L. Wojnar, A. Wronska

Gesellschaft fur Schwerionenforschung mbH, Darmstadt, GermanyM. Al-Turany, I. Augustin, H. Deppe, H. Flemming, J. Gerl, K. Gotzen, G. Hohler, D. Lehmann,

B. Lewandowski, J. Luhning, F. Maas, D. Mishra, H. Orth, K. Peters, T. Saito, G. Schepers,L. Schmitt, C. Schwarz, C. Sfienti, P. Wieczorek, A. Wilms

Technische Universitat Dresden, GermanyK.-T. Brinkmann, H. Freiesleben, R. Jakel, R. Kliemt, T. Wurschig, H.-G. Zaunick

Veksler-Baldin Laboratory of High Energies (VBLHE), Joint Institute for Nuclear Research. Dubna,Russia

V.M. Abazov, G. Alexeev, A. Arefiev, V.I. Astakhov, M.Yu. Barabanov, B.V. Batyunya, Yu.I. Davydov,V.Kh. Dodokhov, A.A. Efremov, A.G. Fedunov, A.A. Feshchenko, A.S. Galoyan, S. Grigoryan,

A. Karmokov, E.K. Koshurnikov, V.Ch. Kudaev, V.I. Lobanov, Yu.Yu. Lobanov, A.F. Makarov,L.V. Malinina, V.L. Malyshev, G.A. Mustafaev, A. Olshevski, M.A.. Pasyuk, E.A. Perevalova,

A.A. Piskun, T.A. Pocheptsov, G. Pontecorvo, V.K. Rodionov, Yu.N. Rogov, R.A. Salmin,A.G. Samartsev, M.G. Sapozhnikov, A. Shabratova, G.S. Shabratova, A.N. Skachkova, N.B. Skachkov,E.A. Strokovsky, M.K. Suleimanov, R.Sh. Teshev, V.V. Tokmenin, V.V. Uzhinsky A.S. Vodopianov,

S.A. Zaporozhets, N.I. Zhuravlev, A.G. Zorin

University of Edinburgh, United KingdomD. Branford, K. Fohl, D. Glazier, D. Watts, P. Woods

Friedrich Alexander Universitat Erlangen-Nurnberg, GermanyW. Eyrich, A. Lehmann, A. Teufel

Northwestern University, Evanston, U.S.A.S. Dobbs, Z. Metreveli, K. Seth, B. Tann, A. Tomaradze

Universita di Ferrara and INFN, Sezione di Ferrara, ItalyD. Bettoni, V. Carassiti, A. Cecchi, P. Dalpiaz, E. Fioravanti, M. Negrini, M. Savrie, G. Stancari

iv PANDA - Strong interaction studies with antiprotons

INFN-Laboratori Nazionali di Frascati, ItalyB. Dulach, P. Gianotti, C. Guaraldo, V. Lucherini, E. Pace

INFN, Sezione di Genova, ItalyA. Bersani, M. Macri, M. Marinelli, R.F. Parodi

Justus Liebig-Universitat Gießen, II. Physikalisches Institut, GermanyW. Doring, P. Drexler, M. Duren, Z. Gagyi-Palffy, A. Hayrapetyan, M. Kotulla, W. Kuhn, S. Lange,

M. Liu, V. Metag, M. Nanova, R. Novotny, C. Salz, J. Schneider, P. Schonmeier, R. Schubert,S. Spataro, H. Stenzel, C. Strackbein, M. Thiel, U. Thoring, S. Yang,

University of Glasgow, United KingdomT. Clarkson, E. Downie, M. Hoek, D. Ireland, R. Kaiser, J. Kellie, I. Lehmann, K. Livingston,

S. Lumsden, D. MacGregor, B. McKinnon, M. Murray, D. Protopopescu, G. Rosner, B. Seitz, G. Yang

Kernfysisch Versneller Instituut, University of Groningen, NetherlandsM. Babai, A.K. Biegun, A. Bubak, E. Guliyev, V.S. Jothi, M. Kavatsyuk, H. Lohner, J. Messchendorp,

H. Smit, J.C. van der Weele

Helsinki Institute of Physics, Helsinki, FinlandF. Garcia, D.-O. Riska

Forschungszentrum Julich, Institut fur Kernphysik, Julich, GermanyM. Buscher, R. Dosdall, A. Gillitzer, F. Goldenbaum, F. Hugging, M. Mertens, T. Randriamalala,

J. Ritman, S. Schadmand, A. Sokolov, T. Stockmanns, P. Wintz

University of Silesia, Katowice, PolandJ. Kisiel

Chinese Academy of Science, Institute of Modern Physics, Lanzhou, ChinaS. Li, Z. Li, Z. Sun, H. Xu

Lunds Universitet, Department of Physics, Lund, SwedenS. Fissum, K. Hansen, L. Isaksson, M. Lundin, B. Schroder

Johannes Gutenberg-Universitat, Institut fur Kernphysik, Mainz, GermanyP. Achenbach, M.C. Mora Espi, J. Pochodzalla, S. Sanchez, A. Sanchez-Lorente

Research Institute for Nuclear Problems, Belarus State University, Minsk, BelarusV.I. Dormenev, A.A. Fedorov, M.V. Korzhik, O.V. Missevitch

Institute for Theoretical and Experimental Physics, Moscow, RussiaV. Balanutsa, V. Chernetsky, A. Demekhin, A. Dolgolenko, P. Fedorets, A. Gerasimov, V. Goryachev

Moscow Power Engineering Institute, Moscow, RussiaA. Boukharov, O. Malyshev, I. Marishev, A. Semenov

Technische Universitat Munchen, GermanyC. Hoppner, B. Ketzer, I. Konorov, A. Mann, S. Neubert, S. Paul, Q. Weitzel

Westfalische Wilhelms-Universitat Munster, GermanyA. Khoukaz, T. Rausmann, A. Taschner, J. Wessels

IIT Bombay, Department of Physics, Mumbai, IndiaR. Varma

Budker Institute of Nuclear Physics, Novosibirsk, RussiaE. Baldin, K. Kotov, S. Peleganchuk, Yu. Tikhonov

Institut de Physique Nucleaire, Orsay, FranceJ. Boucher, T. Hennino, R. Kunne, S. Ong, J. Pouthas, B. Ramstein, P. Rosier, M. Sudol,

J. Van de Wiele, T. Zerguerras

Warsaw University of Technology, Institute of Atomic Energy, Otwock-Swierk, PolandK. Dmowski, R. Korzeniewski, D. Przemyslaw, B. Slowinski

Dipartimento di Fisica Nucleare e Teorica, Universita di Pavia, INFN, Sezione di Pavia, Pavia, ItalyG. Boca, A. Braghieri, S. Costanza, A. Fontana, P. Genova, L. Lavezzi, P. Montagna, A. Rotondi

FAIR/PANDA/Technical Design Report - EMC v

Institute for High Energy Physics, Protvino, RussiaN.I. Belikov, A.M. Davidenko, A.A. Derevschikov, Y.M. Goncharenko, V.N. Grishin, V.A. Kachanov,D.A. Konstantinov, V.A. Kormilitsin, V.I. Kravtsov, Y.A. Matulenko, Y.M. Melnik A.P. Meschanin,

N.G. Minaev, V.V. Mochalov, D.A. Morozov, L.V. Nogach, S.B. Nurushev, A.V. Ryazantsev,P.A. Semenov, L.F. Soloviev, A.V. Uzunian, A.N. Vasiliev, A.E. Yakutin

Kungliga Tekniska Hogskolan, Stockholm, SwedenT. Back, B. Cederwall

Stockholms Universitet, Stockholm, SwedenC. Bargholtz, L. Geren, P.E. Tegner

Petersburg Nuclear Physics Institute of Academy of Science, Gatchina, St. Petersburg, RussiaS. Belostotski, G. Gavrilov, A. Itzotov, A. Kisselev, P. Kravchenko, S. Manaenkov, O. Miklukho,

Y. Naryshkin, D. Veretennikov, V. Vikhrov, A. Zhadanov

Universita del Piemonte Orientale Alessandria and INFN, Sezione di Torino, Torino, ItalyL. Fava, D. Panzieri

Universita di Torino and INFN, Sezione di Torino, Torino, ItalyD. Alberto, A. Amoroso, M. Anselmino, E. Botta, T. Bressani, M.P. Bussa, L. Busso, F. De Mori,

L. Ferrero, A. Grasso, M. Greco, T. Kugathasan, M. Maggiora, S. Marcello, C. Mulatera,G.C. Serbanut, S. Sosio

INFN, Sezione di Torino, Torino, ItalyR. Bertini, D. Calvo, S. Coli, P. De Remigis, A. Feliciello, A. Filippi, G. Giraudo, G. Mazza, A. Rivetti,

K. Szymanska, F. Tosello, R. Wheadon

INAF-IFSI and INFN, Sezione di Torino, Torino, ItalyO. Morra

Politecnico di Torino and INFN, Sezione di Torino,Torino, ItalyM. Agnello, F. Iazzi, K. Szymanska

Universita di Trieste and INFN, Sezione di Trieste, Trieste, ItalyR. Birsa, F. Bradamante, A. Bressan, A. Martin

Universitat Tubingen, Tubingen, GermanyH. Clement

The Svedberg Laboratory, Uppsala, SwedenC. Ekstrom

Uppsala Universitet, Institutionen for Stralningsvetenskap, Uppsala, SwedenH. Calen, S. Grape, B. Hoistad, T. Johansson, A. Kupsc, P. Marciniewski, E. Thome, J. Zlomanczuk

Universitat de Valencia, Dpto. de Fısica Atomica, Molecular y Nuclear, Valencia, SpainJ. Dıaz, A. Ortiz

Soltan Institute for Nuclear Studies, Warsaw, PolandS. Borsuk, A. Chlopik, Z. Guzik, J. Kopec, T. Kozlowski, D. Melnychuk, M. Plominski, J. Szewinski,

K. Traczyk, B. Zwieglinski

Osterreichische Akademie der Wissenschaften, Stefan Meyer Institut fur Subatomare Physik, Wien,Austria

P. Buhler, A. Gruber, P. Kienle, J. Marton, E. Widmann, J. Zmeskal

vi PANDA - Strong interaction studies with antiprotons

Editors: Fritz-Herbert Heinsius Email: [email protected]

Bertram Kopf Email: [email protected]

Bernd Lewandowski Email: [email protected]

Herbert Lohner Email: [email protected]

Rainer Novotny Email: [email protected]

Klaus Peters Email: [email protected]

Philippe Rosier Email: [email protected]

Lars Schmitt Email: [email protected]

Alexander Vasiliev Email: [email protected]

Technical Coordinator: Lars Schmitt Email: [email protected]: Bernd Lewandowski Email: [email protected]

Spokesperson: Ulrich Wiedner Email: [email protected]: Paola Gianotti Email: [email protected]

FAIR/PANDA/Technical Design Report - EMC vii

Preface

This document presents the technical lay-out and the envisaged performance of theElectromagnetic Calorimeter (EMC) for thePANDA target spectrometer. The EMC hasbeen designed to meet the physics goals ofthe PANDA experiment. The performancefigures are based on extensive prototype testsand radiation hardness studies. The docu-ment shows that the EMC is ready for con-struction up to the front-end electronics in-terface.

viii PANDA - Strong interaction studies with antiprotons

The use of registered names, trademarks, etc. in this publicationdoes not imply, even in the absence of specific statement, thatsuch names are exempt from the relevant laws and regulationsand therefore free for general use.

ix

Contents

Preface vii

1 Executive Summary 1

1.1 The PANDA Experiment . . . . . . . 1

1.2 The PANDA ElectromagneticCalorimeter . . . . . . . . . . . . . . 3

1.2.1 PWO-II Scintillator Material . . . 3

1.2.2 APD and VPT Photo Detectors . 5

1.2.3 Electronics . . . . . . . . . . . . . 5

1.2.4 Mechanics and Integration . . . . 7

1.2.5 Calibration and Monitoring . . . . 8

1.2.6 Simulation . . . . . . . . . . . . . 8

1.2.7 Performance . . . . . . . . . . . . 10

1.3 Conclusion . . . . . . . . . . . . . . . 11

2 Overview of the PANDA Experiment 13

2.1 The Physics Case . . . . . . . . . . . 13

2.2 The High Energy Storage Ring . . . . 15

2.2.1 Overview of the HESR . . . . . . 15

2.2.2 Beam Cooling . . . . . . . . . . . 15

2.2.3 Luminosity Estimates . . . . . . . 15

2.3 The PANDA Detector . . . . . . . . . 17

2.3.1 Target Spectrometer . . . . . . . 18

2.3.2 Forward Spectrometer . . . . . . . 25

2.3.3 Luminosity monitor . . . . . . . . 26

2.3.4 Data Acquisition . . . . . . . . . . 27

2.3.5 Infrastructure . . . . . . . . . . . 28

References . . . . . . . . . . . . . . . . . . . 29

3 Design Considerations 31

3.1 Electromagnetic Particle Reconstruc-tion . . . . . . . . . . . . . . . . . . . 31

3.1.1 Coverage Requirements . . . . . . 31

3.1.2 Resolution Requirements . . . . . 33

3.2 Environment . . . . . . . . . . . . . . 34

3.2.1 Surrounding Detectors . . . . . . 34

3.2.2 Count-rate and Occupancy . . . . 35

3.2.3 Operational Aspects . . . . . . . . 37

3.3 Production and Assembly . . . . . . 37

3.3.1 Production Schemes . . . . . . . . 37

3.3.2 Assembly Schemes . . . . . . . . . 37

3.4 Conclusion . . . . . . . . . . . . . . . 37

4 Scintillator Material 39

4.1 Inorganic Scintillators . . . . . . . . . 39

4.1.1 Specific Requirements for the tar-get spectrometer EMC . . . . . . 40

4.1.2 Hit Rates and Absorbed EnergyDose in Single Crystals . . . . . . 40

4.2 Lead Tungstate PWO . . . . . . . . . 42

4.2.1 General Aspects . . . . . . . . . . 42

4.2.2 Basic Properties of PbWO4 andthe Scintillation Mechanism . . . 42

4.2.3 The Improved Properties ofPWO-II . . . . . . . . . . . . . . . 44

4.3 Radiation Induced Absorption inPWO-II Crystals . . . . . . . . . . . 46

4.3.1 The Irradiation Facility at IHEP,Protvino . . . . . . . . . . . . . . 46

4.3.2 The Irradiation Facility at theJustus-Liebig-University Giessen . 48

4.3.3 The Irradiation Facilities at INP(Minsk) and BTCP . . . . . . . . 49

4.3.4 Irradiation Studies with Neu-trons, Protons and High Energyγ-rays . . . . . . . . . . . . . . . . 50

4.4 The PWO-II Crystals for PANDA . . 50

4.4.1 Light Yield and Decay Kinetics . 51

4.4.2 Radiation Hardness of PWO-II . . 51

4.4.3 Radiation Hardness Required forPANDA . . . . . . . . . . . . . . . 54

4.4.4 Pre-Production Run of PWO-IICrystals . . . . . . . . . . . . . . 57

4.5 Quality Requirements and Control . . 59

4.5.1 Measurements of the Light Yieldand the Light Yield Non-uniformity 60

4.5.2 Inspection of the Optical Properties 61

4.5.3 Measurement of the GeometricalDimensions . . . . . . . . . . . . . 61

4.5.4 Analysis and Documentation . . . 61

x PANDA - Strong interaction studies with antiprotons

4.5.5 Control Measurements of Produc-tion Stability . . . . . . . . . . . . 62

4.5.6 Manufacturer for the Final Pro-duction . . . . . . . . . . . . . . . 62

References . . . . . . . . . . . . . . . . . . . 64

5 Photo Detectors 67

5.1 Avalanche Photodiodes (APD) . . . . 67

5.1.1 Introduction . . . . . . . . . . . . 67

5.1.2 Characteristics . . . . . . . . . . . 68

5.1.3 Radiation Damage due to Differ-ent Kinds of Radiation . . . . . . 72

5.1.4 APD Screening Procedure . . . . 78

5.1.5 Mounting Procedure . . . . . . . . 78

5.2 Vacuum Phototriodes (VPT) . . . . . 79

5.2.1 Introduction . . . . . . . . . . . . 79

5.2.2 Available Types . . . . . . . . . . 79

5.2.3 Characteristics and Requirements 80

5.2.4 Testing . . . . . . . . . . . . . . . 81

5.2.5 Screening Procedure . . . . . . . . 83

References . . . . . . . . . . . . . . . . . . . 83

6 Electronics 85

6.1 General EMC Readout Scheme . . . 85

6.2 Preamplifier and Shaper for barrelEMC APD-readout . . . . . . . . . . 86

6.2.1 Requirements and Specifications . 87

6.2.2 Integrated Circuit Development . 88

6.3 Preamplifier and Shaper for forwardendcap EMC VPT-readout . . . . . . 92

6.3.1 Requirements and Specifications . 92

6.3.2 Circuit Description . . . . . . . . 94

6.3.3 Performance Parameters . . . . . 95

6.3.4 SPICE Simulations . . . . . . . . 96

6.4 APD Timing Performance withFADC Readout . . . . . . . . . . . . 96

6.5 Digitizer Module . . . . . . . . . . . . 97

6.6 Data Multiplexer . . . . . . . . . . . 98

6.7 Signal Routing and Cabling . . . . . 99

6.7.1 Requirements . . . . . . . . . . . 99

6.7.2 Cable Performance and Specifica-tions for Proto60 Assembly . . . . 100

6.7.3 Cable Performance and Specifica-tions for barrel EMC and forwardendcap EMC . . . . . . . . . . . . 100

6.7.4 Circuit Description . . . . . . . . 100

6.7.5 Manufacturing, Operation, andSafety . . . . . . . . . . . . . . . . 100

6.7.6 Alternatives . . . . . . . . . . . . 100

6.8 Detector Control System . . . . . . . 101

6.8.1 Goals . . . . . . . . . . . . . . . . 101

6.8.2 Process Control . . . . . . . . . . 102

6.8.3 Detector Control and Monitoring 103

6.9 Production and Assembly . . . . . . 104

6.9.1 ASIC preamplifier . . . . . . . . . 104

6.9.2 Discrete preamplifier . . . . . . . 105

References . . . . . . . . . . . . . . . . . . . 106

7 Mechanics and Integration 107

7.1 The Barrel Calorimeter . . . . . . . . 107

7.1.1 The crystal geometry and housing 107

7.1.2 Mechanics around Crystals - SliceDefinition . . . . . . . . . . . . . . 112

7.1.3 Electronics Integration . . . . . . 112

7.1.4 Thermal Cooling . . . . . . . . . . 114

7.1.5 Integration in the PANDA TargetSpectrometer . . . . . . . . . . . . 119

7.1.6 Construction of the Slices and As-sembly of the Barrel . . . . . . . . 120

7.2 Forward Endcap . . . . . . . . . . . . 121

7.2.1 Requirements . . . . . . . . . . . 121

7.2.2 Crystal shape . . . . . . . . . . . 122

7.2.3 Subunit Structure . . . . . . . . . 122

7.3 Backward Endcap of the Calorimeter 128

References . . . . . . . . . . . . . . . . . . . 129

8 Calibration and Monitoring 131

8.1 Calibration . . . . . . . . . . . . . . . 131

8.1.1 Calibration with Physics Events . 131

8.1.2 Precalibration with CosmicMuons and at Test Beams . . . . 131

8.1.3 Online Calibration . . . . . . . . . 132

8.2 Monitoring . . . . . . . . . . . . . . . 132

8.2.1 Concept of a Light Source andLight Distribution System . . . . 133

FAIR/PANDA/Technical Design Report - EMC xi

8.2.2 Concept of the Light MonitoringSystem . . . . . . . . . . . . . . . 133

8.2.3 Light Pulser Prototype Studies . . 133

8.2.4 Light Monitoring System forPANDA Calorimeter . . . . . . . . 136

References . . . . . . . . . . . . . . . . . . . 139

9 Simulations 141

9.1 Offline Software . . . . . . . . . . . . 141

9.1.1 Photon Reconstruction . . . . . . 141

9.1.2 Electron Identification . . . . . . 144

9.2 Material Budget in front of the EMC 146

9.2.1 DIRC Preshower . . . . . . . . . . 147

9.3 Benchmark Studies . . . . . . . . . . 148

9.3.1 hc Detection with the hc → ηc γDecay and the Role of low Energyγ-ray Threshold . . . . . . . . . . 148

9.3.2 Y(4260) in Formation . . . . . . 150

9.3.3 Charmonium Hybrid in Production 151

9.3.4 Time-like Electromagnetic Form-Factors . . . . . . . . . . . . . . . 152

References . . . . . . . . . . . . . . . . . . . 154

10 Performance 157

10.1 Results from Prototype Tests . . . . 158

10.1.1 Energy Resolution of PWO arrays 158

10.1.2 Position Resolution . . . . . . . . 164

10.1.3 Particle Identification . . . . . . . 165

10.1.4 Construction and Basic Perfor-mance of the Barrel PrototypeComprising 60 Modules . . . . . . 165

10.2 Expected performance of the EMC . 167

References . . . . . . . . . . . . . . . . . . . 169

11 Organisation 171

11.1 Quality Control and Assembly . . . . 171

11.1.1 Production Logistics . . . . . . . 171

11.1.2 The PWO-II Crystals . . . . . . . 171

11.1.3 Detector Module Assembly . . . . 171

11.1.4 Final Assembly, Pre-calibrationand Implementation into PANDA 172

11.1.5 Other Calorimeter Components . 172

11.1.6 Integration in PANDA . . . . . . 172

11.2 Safety . . . . . . . . . . . . . . . . . . 172

11.2.1 Mechanics . . . . . . . . . . . . . 172

11.2.2 Electrical Equipment and Cooling 173

11.2.3 Radiation Aspects . . . . . . . . . 173

11.3 Schedule . . . . . . . . . . . . . . . . 173

Acknowledgments 177

List of Acronyms 179

List of Figures 181

List of Tables 187

xii PANDA - Strong interaction studies with antiprotons

1

1 Executive Summary

1.1 The PANDA Experiment

PANDA is a next generation hadron physics detec-tor planned to be operated at the future Facility forAntiproton and Ion Research (FAIR) at Darmstadt,Germany. It will use cooled antiproton beams witha momentum between 1.5 GeV/c and 15 GeV/c in-teracting with various internal targets.

At FAIR the antiprotons will be injected intoHESR, a slow ramping synchrotron and storage ringwith excellent beam energy definition by means ofstochastic and electron cooling. This allows to mea-sure masses and widths of hadronic resonances withan accuracy of 50–100 keV, which is 10 to 100 timesbetter than achieved in any e+e−-collider experi-ment. In addition states of all quantum numberscan be directly formed in antiproton-proton an-nihilations whereas in e+e−-collisions states withquantum numbers other than JPC = 1−− of a vir-tual photon can only be accessed by higher orderprocesses with corresponding lower cross section orproduction reactions with much worse mass resolu-tion. In the PANDA experiment, the antiprotonswill interact with an internal target, either a hydro-gen cluster jet or a high frequency frozen hydrogenpellet target, to reach a peak luminosity of up to2 · 1032cm−2s−1. For reactions with heavy nucleartargets thin wires or foils are inserted in the beamhalo.

The experiment is focusing on hadron spectroscopy,in particular the search for exotic states in thecharmonium mass region, on the interaction ofcharmed hadrons with the nuclear medium, ondouble-hypernuclei to investigate the nuclear poten-tial and hyperon-hyperon interactions as well as onelectromagnetic processes to study various aspectsof nucleon structure.

From these goals a list of physics benchmarks canbe derived defining the requirements for the PANDAdetector system. For precision spectroscopy of char-monium states and exotic hadrons in the charmo-nium region, full acceptance is required to performa proper partial wave analysis. Final states withmany photons can occur, leading to a low pho-ton threshold as a central requirement for the elec-tromagnetic calorimeters. To reconstruct charmedmesons, a vertex detector and the identification ofkaons is necessary.

Further requirements result from other signals of

interest: final states with muons occur in J/ψ de-cays, semi-leptonic charm meson decays and theDrell-Yan process. Measuring wide angle Comp-ton scattering requires detection of high energy pho-tons. Investigating hypernuclei requires the detec-tion of hyperon cascades. The measurement of pro-ton formfactors relies on an efficient e± identifica-tion and discrimination against pions.

For the reconstruction of invariant masses a goodmomentum resolution in the order of δp/p ∼ 1% isdesirable. Low cross-section processes and precisionmeasurements lead to a high rate operation at upto 20 million interactions per second. To performseveral measurements in parallel an efficient eventselection is needed.

General Setup. To achieve almost 4π accep-tance and good momentum resolution over a largerange, a solenoid magnet for high pT tracks (targetspectrometer) and a dipole magnet for the forward-going reaction products (forward spectrometer) areforeseen (Fig. 1.1). The solenoid magnet is super-conducting, provides a field strength of 2 T and hasa coil opening of 1.89 m and a coil length of 2.75 m.The target spectrometer is arranged in a barrel partfor angles larger than 22 and an endcap part forthe forward range down to 5 in the vertical and10 in the horizontal plane.

The dipole magnet has a field integral of up to 2 Tmwith a 1.4 m wide and 70 cm high opening. The for-ward spectrometer covers the very forward angles.

Both spectrometer parts are equipped with track-ing, charged particle identification, electromagneticcalorimetry and muon identification.

To operate the experiment at high rate and with dif-ferent parallel physical topologies a self-triggeringreadout scheme was adopted. The frontend elec-tronics continuously digitizes the detector data andautonomously finds valid hits. Physical signatureslike energy clusters, tracklets or ringlets are ex-tracted on the fly. Compute nodes make a fastfirst selection of interesting time slices which thenis refined with further data in subsequent levels.Data logging happens only after online reconstruc-tion. This allows full flexibility in applying selec-tion algorithms based on any physics signatures de-tectable by the spectrometer. In addition physicstopics with identical target and beam settings canbe treated in parallel.

2 PANDA - Strong interaction studies with antiprotons

Muon Detectors

Superconducting Solenoid

Central Tracker

Micro Vertex Detector

Electromagnetic Calorimeters

DIRC

Dipole Magnet

Drift Chambers

Forward RICH

Forward SpectrometerTarget Spectrometer

GEM

p-Beam

Muon/Hadron ID

Figure 1.1: Overview of the PANDA spectrometer.

Tracking detectors. The silicon vertex detectorconsists of two inner layers of hybrid pixel detec-tors and two outer layers with silicon strip detec-tors. The pixel layers consist of a silicon sensorcoupled to a custom-made self-triggering readoutASIC realized in 0.13µm CMOS with 100×100µm2

pixel size and cover 0.15 m2 with 13 million chan-nels. The strip part is based on double-sided siliconstrips read by a self-triggering 128-channel ASICwith discriminator and analog readout and covers0.5 m2 with 70 000 channels.

For the tracking in the solenoid field low-mass strawtubes arranged in straight and skewed configura-tion are foreseen. The straws have a diameter of 1cm and a length of 150 cm and are operated withAr/CO2 at 1 bar overpressure giving them rigiditywithout heavy support frames.

Alternatively an ungated time projection chamberbased on GEM foils as readout stage is being devel-oped. The GEM foils suppress the ion backflow intothe drift volume, minimizing the space charge build-up. As detector gas Ne/CO2 is used. The readoutplane consists of 100 000 pads of 2×2 mm2. With500 hits per track and more than 50µs drift time500 events are overlapping, leading to a very highdata rate which has to be handled online.

Tracks at small polar angles (5 < θ < 22) are

measured by large planar GEM detectors. Furtherdownstream, in the forward spectrometer, strawtube chambers with a tube diameter of 1 cm willbe employed.

Particle Identification. Charged particle iden-tification is required over a large momentum rangefrom 200 MeV/c up to almost 10 GeV/c. Differentphysical processes are employed.

The main part of charged particles is identified byvarious Cherenkov detectors. In the target spec-trometer two detectors based on the detection ofinternally reflected Cherenkov light (DIRC) are un-der design, one consisting of quartz rods for thebarrel region, the other one in shape of a disc forthe forward endcap. Novel readout techniques areunder study to achieve correction or compensationof dispersion in the radiator material. In the for-ward spectrometer a ring imaging Cherenkov detec-tor with aerogel and C4F10 as radiators is planned.

Time of flight can be partly exploited in PANDA.Although no dedicated start detector is available, ascintillator wall after the dipole magnet can mea-sure the relative timing of charged particles withvery good time resolution in the order of 100 ps.

The energy loss within the trackers will be employedas well for particle identification below 1 GeV/c

FAIR/PANDA/Technical Design Report - EMC 3

since the individual charge is obtained by analogreadout or time-over-threshold measurement. Herethe TPC option would provide best performance.

The detection system is complemented by muondetectors based on drift tubes located inside thesegmented magnet yoke, between the spectrometermagnets and at the end of the spectrometer. Muondetection is implemented as a range system with in-terleaved absorbing material and detectors to bestdistinguish muons from pions in the low momentumrange of PANDA.

Calorimetry. In the target spectrometer highprecision electromagnetic calorimetry is requiredover a large energy range from a few MeV up to sev-eral GeV. Lead-tungstate is chosen for the calorime-ters in the target spectrometer due to its good en-ergy resolution, fast response and high density, al-lowing a compact setup.

Good identification and reconstruction of multi-photon and lepton-pair channels are of utmost im-portance for the success of the PANDA experiment.Low energy thresholds and good energy and spatialresolution are important assets to achieve high yieldand good background rejection. Due to the highluminosity, fast response and radition hardness areadditional requirements which have to be fulfilled.Table 1.1 shows the detailed list of requirements forthe EMC.

To achieve the required very low energy threshold,the light yield has to be maximized. Therefore im-proved lead-tungstate crystals are employed with alight output twice as high as used in CMS. Thesecrystals are operated at -25 C which increases thelight output by another factor of four. In addi-tion, large area APDs are used for readout, pro-viding high quantum efficiency and an active areafour times larger than used in CMS.

The largest sub-detector is the barrel calorimeterwith 11360 crystals of 200 mm length. In the back-ward direction 592 crystals provide hermeticity atworse resolution due to the presence of readout andsupply lines of other detectors. The 3600 crystalsin the forward direction face a much higher range ofparticle rates across the acceptance of the calorime-ter in the forward endcap. A readout with vacuumphototriodes is foreseen to be able to safely oper-ate at the higher particle rates and correspondinglyhigher radiation load.

The crystal calorimeter is complemented in theforward spectrometer with a shashlik type sam-pling calorimeter consisting of 1404 modules of55× 55 mm2 cell size covering 2.97 × 1.43 m2.

This document presents the details of the technicaldesign of the electromagnetic crystal calorimeter ofthe PANDA target spectrometer.

1.2 The PANDAElectromagneticCalorimeter

1.2.1 PWO-II Scintillator Material

The concept of PANDA places the target spectrom-eter EMC inside the super-conducting coil of thesolenoid. Therefore, the basic requirements of theappropriate scintillator material are compactnessto minimize the radial thickness of the calorimeterlayer, fast response to cope with high interactionrates, sufficient energy resolution and efficiency overthe wide dynamic range of photon energies given bythe physics program, and finally an adequate radia-tion hardness. In order to fulfill these requirements,even a compact geometrical design must provide ahigh granularity leading to a large quantity of crys-tal elements. Their fabrication relies on existingproven technology for mass production to guaranteethe necessary homogeneity of the whole calorimeter.Presently and even in the near future there is no al-ternative material besides lead tungstate available.

The intrinsic parameters of PbWO4 (PWO) as de-veloped for CMS/ECAL are meeting all require-ments, except for the light output. Therefore, anextended R&D program was initiated to improvethe luminescence combined with an operation at lowtemperatures such as T=-25C. Several successfulsteps have been taken, which have lead to a signifi-cantly improved material labelled as PWO-II. Theresearch was aiming at an increase of the structuralperfection of the crystal and the optimized activa-tion with luminescent impurity centers, which havea large cross section to capture electrons from theconduction band, combined with a sufficiently shortdelay of radiative recombination. Beyond the im-provement of the crystal quality, the reduction ofthe thermal quenching of the luminescence processby cooling the crystals leads to an additional in-crease of the light yield without any considerabledistortion of the scintillation kinetics. Operatingat a temperature of T=-25C provides an overallgain factor of about 4 compared to T=+25C. Asa result, full size crystals with 200 mm length de-liver a light yield of 17-20 photoelectrons per MeV(phe/MeV) at 18C measured with a photomul-tiplier with bi-alkali photocathode (quantum effi-ciency ∼20 %).


Required performance valueCommon propertiesenergy resolution σE/E ≤ 1 %⊕ ≤2 %√

E/GeVenergy threshold (photons) Ethres 10 MeV (20 MeV tolerable)energy threshold (single crystal) Extl 3 MeVrms noise (energy equiv.) σE,noise 1 MeVangular coverage % 4π 99 %mean-time-between-failures tmtbf 2000 y(for individual channel)Subdetector specific properties backward barrel forward

(≥ 140) (≥ 22) (≥ 5)energy range from Ethres to 0.7 GeV 7.3 GeV 14.6 GeVangular equivalent of crystal size θ 4 1

spatial resolution σθ 0.5 0.3 0.1

maximum signal load fγ (Eγ > Extl) 60 kHz 500 kHz(pp-events) maximum signal load fγ (Eγ > Extl) 100 kHz 500 kHz(all events) shaping time ts 400 ns 100 nsradiation hardness 0.15 Gy 7 Gy 125 Gy(maximum annual dose pp-events)radiation hardness 10 Gy 125 Gy(maximum annual dose from all events)

Table 1.1: Main requirements for the PANDA EMC. Rates and doses are based on a luminosity of L = 2 ·1032cm−1s−1.

The radiation level will be well below typical LHCvalues by at least two orders of magnitude even atthe most forward direction and be much lower atlarger angles. The radiation hardness and the ef-fective loss of light yield as a function of dose rateand accumulated total dose are well known fromthe studies for CMS but investigated at room tem-perature. The effective deterioration of the opticaltransmission during the experiment is caused bythe interplay of damaging and recovering mecha-nisms. The latter are fast at room temperature andkeep the loss of light yield moderate. As shown bydetailed investigations, at temperatures well below0C the relaxation times of color centers become ex-tremely slow, reaching values well above 200 hours.As a consequence, there is a continuous and asymp-totic reduction of the scintillation output. However,the present quality of PWO-II has reached inducedabsorption values well below the limits of the CMScrystals. A saturation is reached after an integraldose of 30-50 Gy, leading to a maximum loss of lightyield of 30 %. To emphasize, these effects have tobe considered only in the very forward region of theforward endcap EMC. Taking the radiation-damageeffects in forward-angle detectors into account, theproposed operation of the calorimeter at T=-25Ccan be based on a net increase of the light yield by

a factor of 3 compared to room temperature oper-ation in addition to the improvement due to crys-tal quality and a more effective arrangement of thephotosensors. A detector being back at room tem-perature will recover from radiation damage within1-2 weeks.

The performance parameters are based on a largequantity of full size crystal samples and the expe-rience of a pre-production run of 710 crystals com-prising all geometrical shapes to complete one sliceof the barrel. Detailed specifications as well as aquality control program have been elaborated, tak-ing into account the experience of the CMS/ECALcollaboration.

The requested crystal quality can be provided at themoment by the Russian manufacturer BTCP. Theonly alternative producer SICCAS in China has notyet delivered full size samples of comparable quality.

Based on the ongoing developments of PWO-IIcrystals, a detailed program has been worked out forthe quality assurance of the crystals. This includesthe preparation of various facilities for irradiationstudies. The quality control is planned to be per-formed at CERN, taking advantage of the existinginfrastructure and experience developed for CMS.One of the two ACCOS machines, semi-automatic


robots, is presently getting modified for the differ-ent specification limits and geometrical dimensionsof the PANDA crystals.

1.2.2 APD and VPT PhotoDetectors

The low energy threshold of 10 MeV for the PANDAEMC requires the usage of excellent photo detec-tors. The magnetic field of about 2 T precludes theuse of conventional photomultipliers. On the otherhand the signal generated by ionization in a PINphotodiode by a traversing charged particle is toolarge for our applications. To solve these problems aphotosensor insensitive to magnetic fields and witha small response to ionizing radiation has to beused.

Since lead tungstate has a relatively low light yield,the photosensor is required to have an internal gainin addition. Due to the envisaged operation tem-perature of T = −25C for the EMC, not only thePWO-II crystals but also the used photo detectorshave to be radiation hard in this temperature re-gion, which implies detailed studies concerning pos-sibly occurring radiation damages.

For the barrel EMC an Avalanche Photodiode(APD) which has an internal signal amplification(gain) in the silicon structure is chosen as photodetector. To maximize the light yield the cover-age of the readout surface of the crystals has to beas large as possible, leading to the development ofLarge Area APDs (LAAPD) with an active areaof 10×10 mm2. Their characterisation has been fo-cused on the given requirements.

Several prototypes of LAAPDs have been studiedfor their characteristics, especially for their radia-tion hardness: These devices have been exposed toproton, photon and neutron radiation, while theirdark currents have been monitored to analyse thedamage of the surface and of the bulk structure.

The response to hadrons was investigated and willprovide complementary information to the particletracking or even identification of muons based onthe energy deposition or different cluster multiplic-ity of trespassing hadrons.

In addition to several screenings done usingmonochromatic light, certain properties of thediodes could only be determined by scanning thecomplete spectrum of visible light. The latter pro-cedure is not only used for studying the quantumefficiency but also to detect effects on the structureof an LAAPD after irradiation. These studies en-able us to identify the best structure of the APDs

fulfilling the requirements of the PANDA EMC.

The operation of an LAAPD requires the correctknowledge about its gain factor. Since this condi-tion changes with the operation temperature, thescreening of this photo detector is being done atroom temperature as well as at the operating tem-perature of T = −25C to study its gain-voltage de-pendence. During the screening process the knowl-edge about this voltage dependence will be used forgrouping LAAPDs in the detector to enable the us-age of one HV line for one block of sensors. Againwe take advantage of the CMS experience for thescreening of the photo sensors.

During the development of the readout system forthe EMC the typical size of the crystal readoutsurface of 27×27 mm2 lead to the development ofan LAAPD with a rectangular shape (active area:14×6.8 mm2) to cover a maximum of this space withtwo neighboring photo detectors. The new geome-try allows to fit two sensors on each crystal irre-spectively of its individual shape.

In contrast to the barrel EMC, the forward end-cap EMC has to deal with rates up to 500 kHz percrystal and magnetic fields up to 1.2 T. Thereforevacuum phototriodes (VPTs) have been chosen forthe photo detection in this EMC part due to thefollowing reasons: rate capability, radiation hard-ness, absence of nuclear counter effect and absenceof temperature dependence. Standard photomulti-pliers are excluded due to the magnetic field envi-ronment. Different to the barrel region, the mag-netic field is oriented in the axial direction of theVPTs and thus makes it feasible to use them for theendcap readout. Vacuum phototriodes are essen-tially a photomultiplier tube with only one dynodeand weak field dependence. Tests and the parame-ters presented are based on VPTs produced for theCMS experiment. A new VPT with a significantlyhigher quantum efficiency combined with a largerinternal gain is under development and will be pro-duced by the company Photonis and will be testedas soon as it becomes available to us.

The operation of these photo detectors at differentangles relative to the orientation of a magnetic fieldas well as the high rate capability are being studied.These studies are being done both at room temper-ature as well as at the envisaged EMC operationtemperature of T = −25C.

1.2.3 Electronics

The PANDA Electromagnetic Calorimeter (EMC)will provide an almost full coverage of the final


state phase space for photons and electrons. Thelow-energy photon threshold will be around 10 MeV,while the threshold for individual detectors can beas low as 3 MeV with correspondingly low noiselevels of 1 MeV. Such settings allow the precisionspectroscopy of charmonium states and transitions.With a dynamic range of 12000 for the readout elec-tronics, a maximum photon energy deposition of12 GeV per crystal can be detected which allowsthe study of neutral decays of charmed mesons atthe maximum beam energy of the HESR. Typicalevent rates of 10 kHz and maximum 100 kHz areexpected for the barrel part of the calorimeter andup to 500 kHz in the forward endcap. Two differentphoto detectors, LAAPD and vacuum photo triodes(VPTs) will be used. The photo sensors are directlyattached to the end faces of the individual crystalsand the preamplifier is placed as close as possibleinside the calorimeter volume for optimum perfor-mance and minimum space requirements. Since ev-ery barrel EMC crystal is equipped with two APDs,not only redundancy is achieved but also a signif-icantly (max.

√2) improved signal to noise ratio

and a lower effective threshold level.

The readout of small and compact subarrays ofcrystals requires very small preamplifier geometries.The low-temperature environment of the EMC willimprove the noise performance of the analogue cir-cuits. For efficient cooling and stable temperaturebehavior, low power consumption electronics hasbeen developed in combination with extremely low-noise performance.

Two complementary low-noise and low-power(LNP) charge-sensitive preamplifier-shaper (LNP-P) circuits have been developed. The LNP designbased on discrete components utilizes a low-noiseJ-FET transistor. The circuit achieves a very goodnoise performance using signal shaping with a peak-ing time of 650 ns. This preamplifier will be used forthe readout of the forward endcap EMC for whichwe expect a single-channel rate up to 500 kHz. Suchan approach minimizes the overall power consump-tion and keeps the probability for pileup events ata moderate level well below 1 %. The noise floor ofthe LNP-P at -25C, loaded with an input capaci-tance of 22 pF, has a typical equivalent noise charge(ENC) of 235 e− (rms) using signal shaping with apeaking-time of 650 ns. Because the VPT has al-most no dark current, the noise is not increased dueto the leakage current of that photo detector. Byapplying the quantum efficiency and the internalgain of the VPT, the ENC of 235 e−(rms) corre-sponds to an energy noise level of 0.8 MeV (rms).

The second circuit for the LAAPD readout is

a state-of-the-art CMOS ASIC (APFEL), whichachieves a similar noise performance with a shorterpeaking time of 250 ns. The advantage of theCMOS ASIC is the very low power consumption.The noise floor of the APFEL ASIC at -20C,loaded with an input capacitance of 270 pF, hasa typical equivalent noise charge (ENC) of 4150e− (rms). Based on measurements of PWO-IIlight production and the amplification characteris-tics of the photon sensor, this corresponds to anenergy noise level of 0.9 MeV (rms). This is aboutthe same level as achieved in the forward end-cap EMC with the VPT readout. Both systemswill allow an energy threshold of 3 MeV and thusfulfill the requirements. The ASIC was designedin a 350 nm CMOS technology. The power con-sumption at -20C is 52 mW per channel. Proto-types have been produced on Multi Project Wafer(MPW) runs of the EUROPRACTICE IC proto-typing program. For the instrumentation of theelectromagnetic calorimeter about 23000 pieces ofthe preamplifier are needed for the barrel EMC.These amounts of ASICs can no longer be producedcost effectively with MPW runs, thus a chip produc-tion campaign has to be started in due time.

The readout of the electromagnetic calorimeter isbased on the digitization of the amplified signal-shape response of LAAPD and VPT photo sensorsto the light output of PWO-II crystals. The digi-tizer modules are located at a distance of 20–30 cmand 90–100 cm for the barrel EMC and the forwardendcap EMC, respectively, away from the analoguecircuits and outside the cold volume. Signal transferfrom the front-end over short distances is achievedby flat cables with low thermal budget. The digi-tizers consist of high-frequency, low-power pipelinedADC chips, which continuously sample the ampli-fied and shaped signals. With an 80 MHz sam-pling ADC a time resolution better than 1 ns hasbeen achieved at energy deposits above 60 MeV and150 ps at energies above 500 MeV. At the lower en-ergies the time resolution is limited by APD- andpreamplifier-noise. This time resolution is sufficientfor maintaining a good event correlation and reject-ing background hits or random noise. The sam-pling is followed by the digital logic, which pro-cesses time-discrete digital values, detects hits andforwards hit-related information to the multiplexermodule via optical fibers. The multiplexer moduleswill be located in the DAQ hut and they perform ad-vanced signal processing to extract amplitude andsignal-time information.

The front-end electronics of the barrel EMC is lo-cated inside the solenoid magnet where any access


for maintenance or repair is limited to shutdown pe-riods of the HESR, expected to occur once a year.Redundancy in the system architecture is achievedby arranging the digitization of the two APDs ofevery crystal in two blocks, respectively, and pro-viding interconnections at the level of FPGA usinghigh-speed links. A prototype ADC module hasbeen developed containing 32 channels of 12 bit 65MSPS ADCs. The total power consumption of themodule is 15 W.

The EMC and its subcomponents will be embed-ded in the general Detector Control System (DCS)structure of the complete PANDA detector. Theaim of a DCS is to ensure the correct and stableoperation of an experiment, so that the data takenby the detector are of high quality. The scope of theDCS is therefore very wide and includes all subsys-tems and other individual elements involved in thecontrol and monitoring of the detector.

1.2.4 Mechanics and Integration

The calorimeter design in the target region is in fullaccordance with the constraints imposed by a fixedtarget experiment with the strong focusing of themomenta in forward direction. A nearly full cov-erage of ∼ 99 % solid angle in the center-of-masssystem is guaranteed in combination with the for-ward electromagnetic calorimeter, which is locateddownstream beyond the dipole magnet. The gran-ularity is adapted to the tolerable maximum countrate of the individual modules and the optimumshower distribution for energy and position recon-struction by minimizing energy losses due to deadmaterial. The front sizes of the crystal elementscover a nearly identical laboratory solid angle andhave absolute cross sections close to the Moliere ra-dius.

The mechanical design is composed of three parts:the barrel part for a length of 2.5 m and 0.57 m ofinner radius; the forward endcap for a diameter of2 m located at 2.1 m downstream from the target;and the backward endcap of 0.8 m in diameter lo-cated at 1 m upstream from the target. The con-ceptual design of these elements is equivalent. Thebasic PWO crystal shape is a truncated pyramidof 200 mm length and this principle is based on the“flat-pack” configuration used in the CMS calorime-ter. Right angle corners are introduced in order tosimplify the CAD design and the mechanical man-ufacturing process to reduce machining costs. Thepresented geometry foresees that the crystals arenot pointing toward the target position. A tilt of afew degrees is added on the focal axis to reduce the

dead zone effect and to ensure that particles enter-ing between crystals will always cross a significantpart of a crystal. The crystals are wrapped in a foilof a high reflectivity (98 %) and inserted in carbonfiber alveoles to hold them by the back. This helpsalso to avoid any piling-up stress and any heavy ma-terial in front of the crystals. The nominal distancebetween crystals is around 600µm taking into ac-count the thickness of the reflector (2x65µm), car-bon fiber alveoles (2x200µm) and mechanical freegaps.

All crystals have a common length of 200 mm corre-sponding to 22 radiation lengths, which allows op-timum shower containment up to 15 GeV photonenergy and limits the nuclear counter effect in thesubsequent photo sensor to a tolerable level.

The necessary thermal shield for the low-temperature operation of the EMC is made ofpanels using either thin components with highthermal resistivity and low material budget (interms of X0) in front of the crystals or thicker stan-dard foam in the other areas. The cooling circuit iscomposed of carbon fiber or copper thermal screensdepending on the position. Silicone oil, SylthermXLT has been chosen as cooling liquid for its lowviscosity and high thermal efficiency. To avoid anymoisture or ice, dry nitrogen gas is flowing throughthe detector which therefore has to be madeairtight. All these elements, crystals, front-endelectronics and thermal screens are sustained bya metallic support structure at room temperatureconnected to the mainframe of the PANDA magnetor yoke, for the barrel or the endcaps, respectively.On these structures boards and various servicesare implemented as the digitizing, optical fibersfor calibration and data transfer, or cables forthe supply of electrical power and sensors for thedetector control system. The electronics itselfconsists of a low-noise and low-power consumptioncharge sensitive preamplifier connected to thedigitizer part by flat cables to reduce the heattransfer.

The mechanical designs of the barrel and the end-caps differ in some details. The barrel is divided in16 slices of 710 crystals each and 11 different shapesof crystals are necessary to fill the volume. Theaverage crystal has approximatively a square frontface of 21.3 mm and a square rear face (readout) of27.3 mm for an average mass of 0.98 kg. Each sliceis divided in 6 modules for an easier constructionand assembled to a stainless steel support beam.Two rings connect all these slices to make a com-pact cylindrical calorimeter inserted into the mag-net by special tools. All the barrel services are going


through the yoke on its backward side.

The endcaps are designed as a wall structure withquadrant symmetry. Each quarter is composed ofsubunit structures composed of 16 crystals each in-serted in carbon fiber alveoles and screwed throughan interface insert to a thick aluminum back plate.Contrary to the barrel, only one shape of crystal isnecessary for the endcaps and the average dimen-sions are for the front face 24.4 mm and for the rearface 26 mm.

The feasibility of the construction of such deviceshas been validated by the construction of severalprototypes. The last one and the most representa-tive is a 60-crystal prototype which uses some crys-tals from the barrel and integrates all the differentelements described above. It permits to check themechanical design and to focus on the result of thetemperature stability which shows a promising re-sult of ±0.05C. In a next iteration, a mechanicalprototype with 200 crystals is presently under con-struction. This device is primarily meant for study-ing stable operation and cooling by simulating twoadjacent detector slices of the barrel section of theEMC.

1.2.5 Calibration and Monitoring

The ultimate performance of the EMC can only bereached with a precise calibration of the individualcrystal channels. The energy resolution of less than2% at energies above 1 GeV demands a precisionat the sub-percent level. To reach this goal, threemethods are combined: A pre calibration with cos-mic muons, in situ calibration with physics eventsand continuous monitoring with a light-pulser sys-tem.

Before the start up of the experiment all crystalswill be calibrated in situ with cosmic muons at alevel of 10% accuracy. The correspondence betweenthe light output of a muon and a photon will bedetermined with test beams. The position recon-struction algorithm will also be optimized at testbeams.

The final calibration will be performed with physicsevents during data taking. Events with three to fourπ0 or η mesons in the final state serve as an inputto apply the calibration by constraining the energymeasurements to the invariant mass of the meson.A dedicated software trigger will select the eventsbased on total energy and multiplicity at a rate ofabout 4 kHz. An integrated luminosity of 8 pb−1,accumulated in less than a day will be enough toperform a full calibration. With the hardware-

trigger free concept of PANDA the full event selec-tion relies on the online trigger system. An efficientoperation requires a quick calibration of the EMC.This will be achieved by combining the calibrationconstants from the previous day with the informa-tion from the light-pulser system.

The amount of collected light per MeV may changeslightly due to radiation damage. Radiation dam-age affects only the optical transmission and canthus be corrected for by measuring the effect withthe light-pulser system.With a light-pulser systemoperating at several wavelengths (455 nm, 530 nmand 660 nm) the effects from radiation damage canbe disentangled from other effects such as changes ofgain in the photodetectors and preamplifiers. Thewavelength at 455 nm is close to the emission wave-length at 420 nm. A dominant defect center due toMolybdenum is at 530 nm and far in the red spec-tral region one does not expect any radiation dam-age so that one can control separately the readoutchain including the photo sensor.

Ultra-bright LEDs driven by electronic circuits re-producing the time dependence of the PWO scintil-lation light will serve as the light source. The lightwill be distributed to the crystal back face by radia-tion hard silica/silica fibers. The normalizations ofthe pulses are done with temperature stabilized SiPIN photodiodes. A test system showed stabilitiesof 0.1 to 0.2% over a day, which is enough to followthe variations of the light output up to the next fullcalibration.

A first prototype of the light-pulser system is al-ready implemented into the PROTO60 array andoperating.

1.2.6 Simulation

The simulation studies are focused on the expectedperformance of the planned EMC with respect tothe energy and spatial resolution of reconstructedphotons, on the capability of an electron hadronseparation, and also on the feasibility of the plannedphysics program of PANDA.

The software follows an object oriented approach,and most of the code is written in C++. Sev-eral proven software tools and packages from otherHEP experiments have been adapted to the PANDAneeds and are in use. It contains event generatorswith proper decay models for all particles and reso-nances involved in the individual physics channels,particle tracking through the complete PANDA de-tector by using the GEANT4 transport code, a dig-itization which models the signals and the signal


processing in the front-end electronics of the indi-vidual detectors, the reconstruction of charged andneutral particles as well as user friendly high-levelanalysis tools.

The digitization of the EMC has been realized withrealistic properties of PWO crystals at the opera-tional temperature at -25C. A Gaussian distribu-tion of σ = 1 MeV has been used for the constantelectronics noise. The statistical fluctuations wereestimated by 80 phe/MeV produced in the LAAPDwith an excess noise factor of 1.38. This resultsin a photo statistic noise term of 0.41% /

√E. A

comparison with a 3x3 crystal array test measure-ment demonstrates that this digitization gives suffi-ciently realistic results. The simulated line shape atdiscrete photon energies as well as the energy res-olution as a function of the incident photon energyare in good agreement with the measurements.

Different EMC detector scenarios have been inves-tigated in terms of energy and spatial resolutions.While the variation of the crystal length (15 cm,17 cm and 20 cm) show nearly the same resultsfor photons below 300 MeV, the performance getssignificantly better for higher energetic photons forlonger crystals. The 20 cm setup yields an energyresolution of 1.5% for 1 GeV photons, and even <1% for photons above 3 GeV. Another importantaspect is the choice of the individual crystal recon-struction threshold, which is driven by the electron-ics noise term. Comparisons of the achievable res-olution for the most realistic scenario with a noiseterm of σ = 1 MeV and a individual crystal recon-struction threshold of Extl = 3 MeV and a worsecase (σ = 3 MeV, Extl = 9 MeV) show that thedegradation increases by more than a factor of 2for the lowest photon energies. This result demon-strates clearly that the single crystal threshold hasa strong influence on the energy resolution.

The high granularity of the planned EMC providesan excellent position resolution for photons. Basedon a standard cluster finding and bump splitting al-gorithm a σ-resolution of less than 3 mm can be ob-tained for energies above 1 GeV. This correspondsto roughly 10% of the crystal size.

Apart from accurate measurements of photons, theEMC is also the most powerful detector for the iden-tification of electrons. Suitable properties for thedistinction to hadrons and muons are the ratio ofthe energy deposit in the calorimeter to the recon-structed track momentum (E/p) and shower shapevariables derived from the cluster. Good electronidentification can be achieved with a Multilayer Per-ceptron. For momenta above 1 GeV/c an electron ef-ficiency of greater than 98% can be obtained while

the contamination by other particles is substantiallyless than 1%.

The reconstruction efficiency of the EMC is affectedby the interaction of particles with material in frontof the calorimeter. The largest contribution to thematerial budget comes from the Cherenkov detec-tors, which consist of quartz radiators of 1-2 cm.This corresponds to a radiation length between 17%and 50%, depending on the polar angle.

Four different benchmark channels relevant to theEMC have been investigated in order to demon-strate the feasibility of the planned physics programof PANDA. The studies cover charmonium spec-troscopy, the search of charmed hybrids as well asthe measurement of the time-like electromagneticformfactors of the proton. All channels have incommon that the particle detection with the elec-tromagnetic calorimeter plays an essential role.

The first charmonium channel is the production ofhc in the formation mode. One of the main decaychannels of this singlet P wave state (11P1) is theelectromagnetic transition to the ground state ηc(pp → hc → ηc γ). The analysis is based on theηc → φφ decay mode with φ → K+K−. Thethree background channels pp→ K+K−K+K−π0,pp→ φK+K−π0 and pp→ φφπ0 are considered asthe main contributors, having a few orders of mag-nitude higher cross sections. With one γ-ray froma π0 decay left undetected, these reactions have thesame list of decay products as the studied hc de-cay, and thus the γ reconstruction threshold playsan essential role for the background suppression. Asignal to background ratio of better than 3 can beobtained with a reasonable photon reconstructionthreshold of 10 MeV. With a 30 MeV threshold in-stead the signal to background ratio decreases by20% to 40%.

The second charmonium channel is the formationof the recently discovered vector-state Y (4260) inthe reaction pp → Y (4260) → J/ψπ0π0. Thechallenge of this channel is to achieve an efficientand clean electron identification for the reconstruc-tion of J/ψ → e+e−, and also an accurate mea-surement of the final state photons originating fromthe π0 decays. After an event selection by apply-ing kinematical and vertex fits, the reconstructionefficiency is found to be 14%, whereas the suppres-sion rate for the background channels pp → J/ψ η ηand pp → J/ψ η π0 is better than 104. Anothersource of background which has been investigatedare non-resonant pp → π+ π− π0 π0 events. Dueto the expected high cross section of this channela suppression of at least 107 is required. With thecurrently available amount of MC events a suppres-


sion better than 107 is obtained.

Closely connected with charmonium spectroscopyis the search for charmonium hybrids. The groundstate ψg is generally expected to be a spin-exoticJPC = 1−+ state within the mass range of4.1 GeV/c2 and 4.4 GeV/c2. In pp annihilations thisstate can be produced only in association with oneor more recoil particles. In the study, the decay ofthe charmonium hybrid to χc1π

0π0 with the sub-sequent radiative χc1 → J/ψγ decay is considered.The recoiling meson is reconstructed from the de-cay η → γγ. The reconstruction efficiency afterall selection criteria is 4% and 6% for events withJ/ψ → e+e− and J/ψ → µ+µ− decays, respec-tively. One major background source are eventswith hidden charm, in particular events includ-ing a J/ψ meson. The suppression is found to be7 · 103 for the channel pp → χc0 π

0 π0 η, 3 · 104 forpp → χc1 π

0 η η and 1 ·105 for pp → χc1 π0 π0 π0 η

and thus low contamination of the ψg signal fromthese background reactions is expected for the fore-seen PANDA EMC.

The 4th channel is related to the measurementof the time-like electromagnetic formfactors of theproton. The electric (GE) and magnetic (GM ) formfactors can be described by analytic functions of thefour momentum transfer q2 ranging from q2 = −∞to q2 = +∞. The pp annihilation process allowsto access positive q2 (time like) starting from thethreshold of q2 = 4m2

p. In this region GE andGM become complex functions and their determi-nation for low to intermediate momentum transfersis an open question. At PANDA the region between5 (GeV/c)2 and 22 (GeV/c)2 can be accessed. Thefactors |GE | and |GM | can be derived from the an-gular distribution of pp→ e+e− events. For a pre-cise determination of the form factors a suppressionbetter than 108 of the dominant background frompp → π+ π− is required. With the currently avail-able amount of MC events the background rejectionof pp→ π+π− is found to be better than 108. Fur-thermore it was shown that the angular distribu-tions can be obtained with a sufficient accuracy forthe determination of GE and GM .

1.2.7 Performance

To prove the concept for the EMC, test experi-ments have concentrated on the response to pho-tons and charged particles at energies below 1 GeVsince those results are dominated by the photonstatistics of the scintillator, the sensitivity and effi-ciency of the photo sensor and the noise contribu-tions of the front-end electronics. The conclusions

are drawn based on crystal arrays comprising up to25 modules. The individual crystals have a length of200 mm and a rectangular shape with a cross sectionof 20× 20 mm2. Only the most recently assembledarray PROTO60 consists of 60 crystals in PANDAgeometry. The tapered shape will improve the lightcollection due to the focusing effect of the geometryas known from detailed simulations at CMS/ECAL.

The performance tests completed up to now havebeen aiming at two complementary aspects. On onehand, the quality of full size PWO-II crystals has tobe verified in in-beam measurements with energy-tagged photons covering the most critical energyrange up to 1 GeV. Therefore, the scintillator mod-ules were read out with standard photomultipliertubes (Philips XP1911) with a bi-alkali photocath-ode, which covers ∼35 % of the crystal endface witha typical quantum efficiency QE=18 %. The noisecontribution of the sensor can be neglected and thefast response allows an estimate of the time re-sponse. The achieved resolution deduced at variousoperating temperatures can be considered as bench-mark limits for further studies including simulationsand electronics development. The achieved resolu-tion represent excellent lower limits of the perfor-mance to be expected. Operation at T=-25C usinga photomultiplier readout delivers an energy resolu-tion of σ/E = 0.95 %/

√E/GeV + 0.91 % for a 3×3

sub-array accompanied with time resolutions belowσ=130 ps.

From the experimental tests one can extrapolateand confirm an energy resolution at an operatingtemperature of T= −25C, which will be well be-low 2.5 % at 1 GeV photon energy. The resolutionof 13 % at the lowest investigated shower energyof 20 MeV reflects the excellent statistical term. Athreshold of 3 MeV is expected for an individual de-tector channel in the final setup. Relevant for the ef-ficient detection and reconstruction of multi-photonevents, an effective energy threshold of 10 MeV canbe considered for the whole calorimeter as a startingvalue for cluster identification, which will enable usto disentangle from the measured data even physicschannels with extremely low cross section.

The second R&D activity intended to come close tothe final readout concept with large area avalanchephotodiodes (LAAPD), which are mandatory forthe operation within the magnetic field. All thereported results are obtained by collecting and con-verting the scintillation light only with a singlequadratic LAAPD of 10× 10 mm2 active area witha quantum efficiency above 60 % using newly devel-oped low-noise preamplifiers but commercial elec-tronics for the digitization. Not even taking ad-


vantage of an operation at the lowest temperature,an energy resolution of σ/E = 1.86 %/

√E/GeV +

0.65 % has been achieved at T=0C for a 3×3 sub-array, which would come very close to a measure-ment using photomultiplier readout under similarconditions. In addition timing information can beexpected with an accuracy well below 1 ns for en-ergy depositions in a single crystal above 100 MeV.The present data document experimentally only thelower limit of the envisaged performance.

As mentioned above, radiation damage might re-duce asymptotically the light output at the mostforward angles, since relaxation processes becomevery slow at low temperatures. However, due to thefurther improved radiation hardness of the crystals,the loss of light yield will stay below 30 %. Com-bining all improvements including the foreseen sen-sor concept the target spectrometer EMC can relyon a factor of 15-20 higher light yield compared toCMS/ECAL.

In order to study the operation of large arrays, themechanical support structures, cooling and temper-ature stabilization concepts and long term stabili-ties, a large prototype comprising 60 tapered crys-tals in PANDA-geometry has been designed andbrought into operation. First in-beam tests arescheduled for summer 2008.

1.3 Conclusion

The intrinsic performance parameters of the presentquality of PWO-II, such as luminescence yield, de-cay kinetics and radiation hardness and the addi-tional gain in light yield due to cooling down toT=-25C fulfill the basic requirements for the elec-tromagnetic calorimeter. The general applicabilityof PWO for calorimetry in High-Energy Physics hasbeen promoted and finally proven by the success-ful realization of the CMS/ECAL detector as wellas the photon spectrometer ALICE/PHOS, both in-stalled at LHC. The necessary mass production ofhigh-quality crystals has been achieved at least atBTCP in Russia.

The operation at low temperatures imposes a tech-nological challenge on the mechanical design, thecooling concept and thermal insulation under theconstraints of a minimum material budget of deadmaterial. Detailed simulations and prototypinghave confirmed the concept and high accuracy hasbeen achieved in temperature stabilization takinginto account also realistic scenarios of the powerconsumption of the front-end electronics.

Concepts for the photon sensors and the readout

electronics have been tested. As the most sensitiveelement, a prototype of a custom designed ASICimplementing pre-amplification and shaping stageshas been successfully brought into operation andwill provide a large dynamic range of 12,000 with atypical noise level corresponding to ∼ 1 MeV.

The overall performance of the calorimeter will becontrolled by injecting light from LED-sources dis-tributed via optical fibers at the rear face of thecrystal.

Based on the ongoing developments of PWO-IIcrystals and LAAPDs, respectively, a detailed pro-gram has been elaborated for quality assurance ofthe crystals and screening of the photo sensors.

The general layout of the mechanical structure iscompleted including first estimates of the integra-tion into the PANDA detector. The concepts forsignal- and HV-cables, cooling, slow control, moni-toring as well as the stepwise assembly are workedout to guarantee that the crystal geometries couldbe finalized. Prototypes of the individual crys-tal containers, based on carbon fiber alveoles, havebeen fabricated and tested and are already imple-mented in the PROTO60 device.

The experimental data together with the elaboratedesign concepts and simulations show that the am-bitious physics program of PANDA can be fully ex-plored based on the measurement of electromag-netic probes, such as photons, electrons/positronsor the reconstruction of the invariant mass of neu-tral mesons.


13

2 Overview of the PANDA Experiment

PANDA represents the efforts of the world-widehadron physics community to attack the most fun-damental and burning problems in the strong inter-action. Given the complexity of the strong inter-action in the non-perturbative regime this requiresstudies of rather diverse topics. The PANDA exper-iment should complement the new antiproton accel-erator complex at the FAIR facility in Darmstadt.The high-energy storage ring HESR will deliver an-tiproton beams of unprecedented precision and in-tensity. It is a definite challenge for experimental-ists to build an experiment which on one side coversan “as broad as possible” range of physics topicsand on the other side has to go beyond the pre-cision reached in the past with specialized setups.The proposed detector clearly has to be as modernand versatile as possible to fulfill the physics needswithout jeopardizing quality.

The concept of PANDA is based on the experiencefrom previous experiments in the field like the Crys-tal Barrel and OBELIX detectors at LEAR, theE835 experiment at Fermilab, the running FIN-UDA experiment at Frascati, and takes into accountthe concepts, which have been developed and im-plemented for the most modern LHC experiments.PANDA is a hermetic detector for charged and neu-tral particles in the energy range of 10 MeV up to10 GeV. Precise micro-vertex tracking is mandatoryas is good particle identification. The possibility tocombine all these elements of information simulta-neously for each given event gives the PANDA detec-tor its unprecedented power, both in the ability toestablish (rather than just suggest) new unexpectedphenomena and in the redundant identification ofinteresting processes predicted by present models.Since the ratio of interesting events to backgroundsis often rather small we shall need all the capabilitythat we can provide.

Clearly the design choices for a detector represent awell-thought balance between physics needs and theavailable resources. The hadron physics communitywill not have a second detector besides PANDA forthis physics available and hence the detector hasto be sufficiently robust, redundant and resistantto radiation damage for an operation over manyyears. Superb calibration and monitoring capabil-ities must be present for all subsystems. The datarate of 2·107 antiproton annihilations per secondposes not only a challenge for the individual detec-tors but also for the data acquisition system and

the online data selection.

In the cost/performance optimization one has todistinguish between items that at a later stagecould be improved by upgrades if the need arisesand items where scope reductions remain forever.To the latter belong in our opinion the choice ofmagnets and expensive items like calorimeters andthe overall performance of a central tracker. Eventhough savings could be achieved by reducing eventheir parameters, we believe that going below thelevels proposed in this document would lead tounacceptable technical and performance degrada-tion. The current report should justify the param-eter choices in view of the physics that should beachieved.

2.1 The Physics Case

One of the most challenging and fascinating goalsof modern physics is the achievement of a fullyquantitative understanding of the strong interac-tion, which is the subject of hadron physics. Sig-nificant progress has been achieved over the pastfew years thanks to considerable advances in exper-iment and theory. New experimental results havestimulated a very intense theoretical activity and arefinement of the theoretical tools.

Still there are many fundamental questions whichremain basically unanswered. Phenomena such asthe confinement of quarks, the existence of glueballsand hybrids, the origin of the masses of hadrons inthe context of the breaking of chiral symmetry arelong-standing puzzles and represent the intellectualchallenge in our attempt to understand the natureof the strong interaction and of hadronic matter.

Experimentally, studies of hadron structure can beperformed with different probes such as electrons,pions, kaons, protons or antiprotons. In antiproton-proton annihilation particles with gluonic degreesof freedom as well as particle-antiparticle pairs arecopiously produced, allowing spectroscopic studieswith very high statistics and precision. Therefore,antiprotons are an excellent tool to address the openproblems.

The recently approved FAIR facility (Facility forAntiproton and Ion Research), which will be builtas a major upgrade of the existing GSI laboratoryin Germany, will provide antiproton beams of the


highest quality in terms of intensity and resolution,which will provide an excellent tool to answer theaforementioned fundamental questions.

The PANDA experiment (Pbar ANnihilations atDArmstadt) will use the antiproton beam from theHigh-Energy Storage Ring (HESR) colliding withan internal proton target at a CMS energy between2.2 GeV and 5.5 GeV within a general purpose spec-trometer to carry out a rich and diversified hadronphysics program.

The experiment is being designed to fully exploitthe extraordinary physics potential arising fromthe availability of high-intensity, cooled antiprotonbeams. The aim of the rich experimental programis to improve our knowledge of the strong interac-tion and of hadron structure. Significant progressbeyond the present understanding of the field isexpected thanks to improvements in statistics andprecision of the data.

Many measurements are foreseen in PANDA, partof which can be carried out in parallel. In the follow-ing the main topics of the PANDA physics programare outlined.

Charmonium spectroscopy. The cc spectrumcan be computed within the framework of non-relativistic potential models and, more recently, inLattice QCD. A precise measurement of all statesbelow and above open charm threshold is of fun-damental importance for a better understanding ofQCD. All charmonium states can be formed directlyin pp annihilation.

At full luminosity PANDA will be able to collectseveral thousand cc states per day. By means offine scans it will be possible to measure masses withaccuracies of the order of 100 keV and widths to10% or better. The entire energy region below andabove open charm threshold will be explored.

Search for gluonic excitations (hybrids andglueballs). One of the main challenges of hadronphysics is the search for gluonic excitations, i.e.hadrons in which the gluons can act as principalcomponents. These gluonic hadrons fall into twomain categories: glueballs, i.e. states of pure glue,and hybrids, which consist of a qq pair and excitedglue. The additional degrees of freedom carried bygluons allow these hybrids and glueballs to haveJPC exotic quantum numbers: in this case mix-ing effects with nearby qq states are excluded andthis makes their experimental identification easier.The properties of glueballs and hybrids are deter-mined by the long-distance features of QCD and

their study will yield fundamental insight into thestructure of the QCD vacuum. Antiproton-protonannihilations provide a very favourable environmentto search for gluonic hadrons.

Study of hadrons in nuclear matter. Thestudy of medium modifications of hadrons embed-ded in hadronic matter is aimed at understand-ing the origin of hadron masses in the context ofspontaneous chiral symmetry breaking in QCD andits partial restoration in a hadronic environment.So far experiments have been focused on the lightquark sector. The high-intensity p beam of up to15 GeV/c will allow an extension of this programto the charm sector both for hadrons with hiddenand open charm. The in-medium masses of thesestates are expected to be affected primarily by thegluon condensate.

Another study which can be carried out in PANDAis the measurement of J/ψ and D meson productioncross sections in p annihilation on a series of nucleartargets. The comparison of the resonant J/ψ yieldobtained from p annihilation on protons and differ-ent nuclear targets allows to deduce the J/ψ-nucleusdissociation cross section, a fundamental parameterto understand J/ψ suppression in relativistic heavyion collisions interpreted as a signal for quark-gluonplasma formation.

Open charm spectroscopy. The HESR runningat full luminosity and at p momenta larger than 6.4GeV/c would produce a large number of D mesonpairs. The high yield (e.g. 100 charm pairs per sec-ond around the ψ(4040)) and the well defined pro-duction kinematics of D meson pairs would allow tocarry out a significant charmed meson spectroscopyprogram which would include, for example, the richD and Ds meson spectra.

Hypernuclear physics. Hypernuclei are sys-tems in which up or down quarks are replaced bystrange quarks. In this way a new quantum num-ber, strangeness, is introduced into the nucleus. Al-though single and double Λ-hypernuclei were dis-covered many decades ago, only 6 events of doubleΛ-hypernuclei were observed up to now. The avail-ability of p beams at FAIR will allow efficient pro-duction of hypernuclei with more than one strangehadron, making PANDA competitive with planneddedicated facilities. This will open new perspectivesfor nuclear structure spectroscopy and for studyingthe forces between hyperons and nucleons.


Electromagnetic Processes. In addition to thespectroscopic studies described above, PANDA willbe able to investigate the structure of the nucleonusing electromagnetic processes, such as Wide An-gle Compton Scattering (WACS) and the processpp → e+e−, which will allow the determination ofthe electromagnetic form factors of the proton inthe timelike region over an extended q2 region.

2.2 The High Energy StorageRing

The FAIR facility at Darmstadt, Germany will pro-vide anti-proton beams with very high quality. A 30GeV/c proton beam is used to produce anti-protonswhich subsequently are accumulated and cooled at amomentum of 3.7 GeV/c. A derandomized bunch ofanti-protons is then fed into the High Energy Stor-age Ring (HESR) which serves as slow synchrotronto bring the anti-protons to the desired energy andthen as storage ring for internal target experiments.HESR will have both stochastic and electron cool-ing to deliver high quality beams.

2.2.1 Overview of the HESR

An important feature of the new anti-proton facil-ity is the combination of phase-space cooled beamsand dense internal targets, comprising challengingbeam parameter in two operation modes: high-luminosity mode (HL) with beam intensities up to1011, and high-resolution mode (HR) with a mo-mentum spread down to a few times 10−5, respec-tively. Powerful electron and stochastic cooling sys-tems are necessary to meet the experimental re-quirements.

The HESR lattice is designed as a racetrack shapedring, consisting of two 180 arc sections connectedby two long straight sections. One straight sectionwill mainly be occupied by the electron cooler. Theother section will host the experimental installationwith internal H2 pellet target, RF cavities, injectionkickers and septa (see Fig. 2.1). For stochastic cool-ing pickup and kicker tanks are also located in thestraight sections, opposite to each other. To im-prove longitudinal stochastic cooling a third pickuplocation in the arc is presently being investigated.Special requirements for the lattice are dispersionfree straight sections and small betatron amplitudesin the range of a few meters at the internal inter-action point. In addition the betatron amplitudesat the electron cooler are adjustable within a largerange.

Table 2.2.1 summarizes the specified injection pa-rameters, experimental requirements and operationmodes.

2.2.2 Beam Cooling

Beam equilibrium is of major concern for the high-resolution mode. Calculations of beam equilibriafor beam cooling, intra-beam scattering and beam-target interaction are being performed utilizing dif-ferent simulation codes like BETACOOL (JINR,Dubna), MOCAC (ITEP, Moscow), and PTAR-GET (GSI, Darmstadt). Cooled beam equilibriacalculations including special features of pellet tar-gets have been carried out with a simulation codebased on PTARGET.

An electron cooler is realized by an electron beamwith up to 1 A current, accelerated in special ac-celerator columns to energies in the range of 0.4 to4.5 MeV for the HESR. The 22 m long solenoidalfield in the cooler section has a field range from 0.2to 0.5 T with a magnetic field straightness in theorder of 10−5 [1]. This arrangement allows beamcooling for beam momenta between 1.5 GeV/c and8.9 GeV/c.

In the HR mode RMS relative momentum spreadsof less than 4 · 10−5 can be achieved with electroncooling.

The main stochastic cooling parameters were deter-mined for a cooling system utilizing quarter-waveloop pickups and kickers with a band-width of 2to 4 GHz. Stochastic cooling is presently specifiedabove 3.8 GeV/c [2]. Applying stochastic coolingone can achieve an RMS relative momentum spreadof 3 to 4 · 10−5 for the HR mode. In the HL modean RMS relative momentum spread slightly below10−4 can be expected. Transverse stochastic coolingcan be adjusted independently to ensure sufficientbeam-target overlap.

2.2.3 Luminosity Estimates

Beam losses are the main restriction for high lu-minosities, since the antiproton production rate islimited. Three dominating contributions of beam-target interaction have been identified: Hadronicinteraction, single Coulomb scattering and energystraggling of the circulating beam in the target. Inaddition, single intra-beam scattering due to theTouschek effect has also to be considered for beamlifetime estimates. Beam losses due to residualgas scattering can be neglected compared to beam-target interaction, if the vacuum is better than


Figure 2.1: Schematic view of the HESR. Tentative positions for injection, cooling devices and experimentalinstallations are indicated.

Injection Parameters

Transverse emittance 1 mm ·mrad (normalized, RMS) for 3.5 · 1010 particles,

scaling with number of accumulated particles: ε⊥ ∼ N4/5

Relative momentum spread 1 · 10−3 (normalized, RMS) for 3.5 · 1010 particles,

scaling with number of accumulated particles: σp/p ∼ N2/5

Bunch length Below 200 m

Injection Momentum 3.8 GeV/c

Injection Kicker injection using multi-harmonic RF cavities

Experimental Requirements

Ion species Antiprotons

p production rate 2 · 107 /s (1.2 · 1010 per 10 min)

Momentum / Kinetic energy range 1.5 to 15 GeV/c / 0.83 to 14.1 GeV

Number of particles stored in HESR 1010 to 1011

Target thickness 4 · 1015 atoms/cm2

Transverse emittance < 1 mm ·mrad

Betatron amplitude E-Cooler 25–200 m

Betatron amplitude at IP 1–15 m

Operation Modes

High resolution (HR) Luminosity of 2 · 1031 cm−2s−1 for 1010 p

RMS momentum spread σp/p ≤ 4 · 10−5,

1.5 to 9 GeV/c, electron cooling up to 9 GeV/c

High luminosity (HL) Luminosity of 2 · 1032 cm−2s−1 for 1011 p

RMS momentum spread σp/p ∼ 10−4,

1.5 to 15 GeV/c, stochastic cooling above 3.8 GeV/c

Table 2.1: Injection parameters, experimental requirements and operation modes.


10−9 mbar. A detailed analysis of all beam lossprocesses can be found in [3, 4].

0

2 1031

4 1031

6 1031

8 1031

1 1032

0 5000 10000 15000 20000

Avera

ge L

um

ino

sit

y /

cm

-2s

-1

Cycle Time / s

tprep

= 120 s

ττττ = 1540 s

nt = 4*10

15 cm

-2

f0 = 0.443 MHz

2*107 pbar / s

1*107 pbar / s

0

5 1031

1 1032

1.5 1032

2 1032

2.5 1032

3 1032

0 5000 10000 15000 20000

Avera

ge L

um

ino

sit

y /

cm

-2s

-1

Cycle Time / s

2*107 pbar / s

1*107 pbar / s

nt = 4*10

15 cm

-2

f0 = 0.521MHz

tprep

= 290 s

ττττ = 7100 s

Figure 2.2: Average luminosity vs. cycle time at 1.5(top) and 15 GeV/c (bottom). The maximum numberof particles is limited to 1011 (solid line), and unlimited(dashed lines).

The maximum luminosity depends on the antipro-ton production rate dNp/dt = 2 · 107 /s and lossrate

Lmax =dNp/dt

σtot(2.1)

and ranges from 0.8 ·1032 cm−2s−1 at 1.5 GeV/c to3.9 · 1032 cm−2s−1 at 15 GeV/c.

To calculate the average luminosity, machine cyclesand beam preparation times have to be specified.After injection, the beam is pre-cooled to equilib-rium (with target off) at 3.8 GeV/c. The beam isthen ac-/decelerated to the desired beam momen-tum. A maximum ramp rate for the superconduct-ing dipole magnets of 25 mT/s is specified. Afterreaching the final momentum beam steering and fo-cusing in the target and beam cooler region takes

place. Total beam preparation time tprep rangesfrom 120 s for 1.5 GeV/c to 290 s for 15 GeV/c.

In the high-luminosity mode, particles should bere-used in the next cycle. Therefore the used beamis transferred back to the injection momentum andmerged with the newly injected beam. A bucketscheme utilizing broad-band cavities is foreseen forbeam injection and the refill procedure. Duringacceleration 1% and during deceleration 5% beamlosses are assumed. The average luminosity reads

L = f0Ni,0ntτ[1− e−texp/τ

]texp + tprep

(2.2)

where τ is the 1/e beam lifetime, texp the experi-mental time (beam on target time), and tcycle thetotal time of the cycle, with tcycle = texp + tprep.The dependence of the average luminosity on thecycle time is shown for different antiproton produc-tion rates in Fig. 2.2.

With limited number of antiprotons of 1011, as spec-ified for the high-luminosity mode, average lumi-nosities of up to 1.6 · 1032 cm−2s−1 can be achievedat 15 GeV/c for cycle times of less than one beamlifetime. If one does not restrict the number ofavailable particles, cycle times should be longerto reach maximum average luminosities close to3 · 1032 cm−2s−1. This is a theoretical upper limit,since the larger momentum spread of the injectedbeam would lead to higher beam losses during injec-tion due to the limited longitudinal ring acceptance.For the lowest momentum, more than 1011 particlescan not be provided in average, due to very shortbeam lifetimes. As expected, average luminositiesare below 1032 cm−2s−1.

In the operation cycle of HESR the time for re-injection and acceleration of a bunch of anti-protonscan be used by the experiment to do pulser calibra-tions and tests required to monitor noise and signalquality.

2.3 The PANDA Detector

The main objectives of the design of the PANDAexperiment pictured in Fig. 2.3 are to achieve 4π ac-ceptance, high resolution for tracking, particle iden-tification and calorimetry, high rate capabilities anda versatile readout and event selection. To obtain agood momentum resolution the detector is split intoa target spectrometer based on a superconductingsolenoid magnet surrounding the interaction pointand measuring at high angles and a forward spec-trometer based on a dipole magnet for small angle


Figure 2.3: Artistic view of the PANDA Detector.

tracks. A silicon vertex detector surrounds the in-teraction point. In both spectrometer parts track-ing, charged particle identification, electromagneticcalorimetry and muon identification are available toallow to detect the complete spectrum of final statesrelevant for the PANDA physics objectives.

In the following paragraphs the components of alldetector subsystems are briefly explained.

2.3.1 Target Spectrometer

The target spectrometer surrounds the interactionpoint and measures charged tracks in a solenoidalfield of 2 T. In the manner of a collider detectorit contains detectors in an onion shell like config-uration. Pipes for the injection of target materialhave to cross the spectrometer perpendicular to thebeam pipe.

The target spectrometer is arranged in a barrel partfor angles larger than 22 and an endcap part forthe forward range down to 5 in the vertical and 10

in the horizontal plane. The target spectrometer isgiven in a side view in Fig. 2.4.

A main design requirement is compactness to avoida too large and a too costly magnet and crystal

calorimeter.

2.3.1.1 Target

The compact geometry of the detector layers nestedinside the solenoidal magnetic field combined withthe request of minimal distance from the interactionpoint to the vertex tracker leaves very restrictedspace for the target installations. The situation isdisplayed in Fig. 2.5, showing the intersection be-tween the antiproton beam pipe and the target pipebeing gauged to the available space. In order toreach the design luminosity of 2 · 1032 s−1cm−2 atarget thickness of about 4·1015 hydrogen atoms percm2 is required assuming 1011 stored anti-protonsin the HESR ring.

These are conditions posing a real challenge for aninternal target inside a storage ring. At present, twodifferent, complementary techniques for the internaltarget are developed further: the cluster-jet targetand the pellet target. Both techniques are capableof providing sufficient densities for hydrogen at theinteraction point, but exhibit different propertiesconcerning their effect on the beam quality and thedefinition of the interaction point. In addition, in-ternal targets also of heavier gases, like deuterium,


Figure 2.4: Side view of the target spectrometer.

nitrogen or argon can be made available.

For non-gaseous nuclear targets the situation is dif-ferent in particular in case of the planned hyper-nuclear experiment. In these studies the whole up-stream end cap and part of the inner detector ge-ometry will be modified.

Cluster-Jet Target The expansion of pressur-ized cold hydrogen gas into vacuum through aLaval-type nozzle leads to a condensation of hy-drogen molecules forming a narrow jet of hydrogenclusters. The cluster size varies from ·103 to ·106 hy-drogen molecules tending to become larger at higherinlet pressure and lower nozzle temperatures. Sucha cluster-jet with density of ·1015 atoms/cm3 actsas a very diluted target since it may be seen as alocalized and homogeneous monolayer of hydrogenatoms being passed by the antiprotons once per rev-olution.

Fulfilling the luminosity demand for PANDA still re-

quires a density increase compared to current appli-cations. Additionally, due to detector constraints,the distance between the cluster-jet nozzle and thetarget will be larger. The size of the target re-gion will be given by the lateral spread of hydro-gen clusters. This width should stay smaller than10 mm when optimized with skimmers and colli-mators both for maximum cluster flux as well asfor minimum gas load in the adjacent beam pipes.The great advantage of cluster targets is the homo-geneous density profile and the possibility to focusthe antiproton beam at highest phase space den-sity. Hence, the interaction point is defined trans-versely but has to be reconstructed longitudinallyin beam direction. In addition the low β-functionof the antiproton beam keeps the transverse beamtarget heating effects at the minimum. The possi-bility of adjusting the target density along with thegradual consumption of antiprotons for running atconstant luminosity will be an important feature.


Figure 2.5: Schematic of the target and beam pipe setup with pumps.

Pellet Target The pellet target features a streamof frozen hydrogen micro-spheres, called pellets,traversing the antiproton beam perpendicularly. Apellet target presently is in use at the Wasa atCOSY experiment. Typical parameters for pelletsat the interaction point are the rate of 1.0 -1.5 ·104

s−1, the pellet size of 25 - 40 µm, and the velocityof about 60 m/s. At the interaction point the pel-let train has a lateral spread of σ ≈ 1 mm and aninterspacing of pellets that varies between 0.5 to 5mm. With proper adjustment of the β-function ofthe coasting antiproton beam at the target position,the design luminosity for PANDA can be reachedin time average. The present R&D is concentrat-ing on minimizing the luminosity variations suchthat the instantaneous interaction rate does not ex-ceed the acceptance of the detector systems. Sincea single pellet becomes the vertex for more thanhundred nuclear interactions with antiprotons dur-ing the time a pellet traverses the beam, it will bepossible to determine the position of individual pel-lets with the resolution of the micro-vertex detectoraveraged over many events. R&D is going on to de-vise an optical pellet tracking system. Such a devicecould determine the vertex position to about 50 µmprecision for each individual event independently ofthe detector. It remains to be seen if this devicecan later be implemented in PANDA.

The production of deuterium pellets is also well es-tablished, the use of other gases as pellet targetmaterial does not pose problems.

Other Targets are under consideration for thehyper-nuclear studies where a separate target sta-tion upstream will comprise primary and secondary

target and detectors. Moreover, current R&D isundertaken for the development of a liquid heliumtarget and a polarized 3He target. A wire targetmay be employed to study antiproton-nucleus in-teractions.

2.3.1.2 Solenoid Magnet

The magnetic field in the target spectrometer isprovided by a superconducting solenoid coil withan inner radius of 90 cm and a length of 2.8 m. Themaximum magnetic field is 2 T. The field homogene-ity is foreseen to be better than 2 % over the vol-ume of the vertex detector and central tracker. Inaddition the transverse component of the solenoidfield should be as small as possible, in order toallow a uniform drift of charges in the time pro-jection chamber. This is expressed by a limit of∫Br/Bzdz < 2 mm for the normalized integral of

the radial field component.

In order to minimize the amount of material infront of the electromagnetic calorimeter, the latteris placed inside the magnetic coil. The trackingdevices in the solenoid cover angles down to 5/10

where momentum resolution is still acceptable. Thedipole magnet with a gap height of 1.4 m providesa continuation of the angular coverage to smallerpolar angles.

The cryostat for the solenoid coils has two warmbores of 100 mm diameter, one above and one belowthe target position, to allow for insertion of internaltargets.


Figure 2.6: The Microvertex detector of PANDA

2.3.1.3 Microvertex Detector

The design of the micro-vertex detector (MVD) forthe target spectrometer is optimized for the detec-tion of secondary vertices from D and hyperon de-cays and maximum acceptance close to the interac-tion point. It will also strongly improve the trans-verse momentum resolution. The setup is depictedin Fig. 2.6.

The concept of the MVD is based on radiation hardsilicon pixel detectors with fast individual pixelreadout circuits and silicon strip detectors. Thelayout foresees a four layer barrel detector with aninner radius of 2.5 cm and an outer radius of 13cm. The two innermost layers will consist of pixeldetectors while the outer two layers are consideredto consist of double sided silicon strip detectors.

Eight detector wheels arranged perpendicular to thebeam will achieve the best acceptance for the for-ward part of the particle spectrum. Here again, theinner two layers are made entirely of pixel detectors,the following four are a combination of strip detec-tors on the outer radius and pixel detectors closerto the beam pipe. Finally the last two wheels, madeentirely of silicon strip detectors, are placed furtherdownstream to achieve a better acceptance of hy-peron cascades.

The present design of the pixel detectors comprisesdetector wafers which are 200 µm thick (0.25%X0).

The readout via bump-bonded wafers with ASICsas it is used in ATLAS and CMS [5, 6] is foreseenas the default solution. It is highly parallelized andallows zero suppression as well as the transfer ofanalog information at the same time. The read-out wafer has a thickness of 300 µm (0.37%X0). Apixel readout chip based on a 0.13 µm CMOS tech-nology is under development for PANDA. This chipallows smaller pixels, lower power consumption anda continuously sampling readout without externaltrigger.

Another important R&D activity concerns the min-imisation of the material budget. Here strategieslike the thinning of silicon wafers and the use ofultra-light materials for the construction are inves-tigated.

2.3.1.4 Central Tracker

The charged particle tracking devices must handlethe high particle fluxes that are anticipated for aluminosity of up to several 1032 cm−2s−1. The mo-mentum resolution δp/p has to be on the percentlevel. The detectors should have good detection ef-ficiency for secondary vertices which can occur out-side the inner vertex detector (e.g. K0

S or Λ). Thisis achieved by the combination of the silicon ver-tex detectors close to the interaction point (MVD)with two outer systems. One system is covering a


large area and is designed as a barrel around theMVD. This will be either a stack of straw tubes(STT) or a time-projection chamber (TPC). Theforward angles will be covered using three sets ofGEM trackers similar to those developed for theCOMPASS experiment at CERN. The two optionsfor the central tracker are explained briefly in thefollowing.

Straw Tube Tracker (STT) This detector con-sists of aluminized mylar tubes called straws, whichare self supporting by the operation at 1 bar over-pressure. The straws are arranged in planar layerswhich are mounted in a hexagonal shape aroundthe MVD as shown in Fig. 2.7. In total there are24 layers of which the 8 central ones are tilted toachieve an acceptable resolution of 3 mm also in z(parallel to the beam). The gap to the surroundingdetectors is filled with further individual straws. Intotal there are 4200 straws around the beam pipe atradial distances between 15 cm and 42 cm with anoverall length of 150 cm. All straws have a diameterof 10 mm. A thin and light space frame will holdthe straws in place, the force of the wire howeveris kept solely by the straw itself. The mylar foil is30 µm thick, the wire is made of 20 µm thick goldplated tungsten. This design results in a materialbudget of 1.3 % of a radiation length.

The gas mixture used will be Argon based with CO2

as quencher. It is foreseen to have a gas gain nogreater than 105 in order to warrant long term op-eration. With these parameters, a resolution in xand y coordinates of about 150 µm is expected.

Time Projection Chamber (TPC) A chal-lenging but advantageous alternative to the STT isa TPC, which would combine superior track resolu-tion with a low material budget and additional par-ticle identification capabilities through energy lossmeasurements.

The TPC depicted in a schematic view in Fig. 2.8consists of two large gas-filled half-cylinders enclos-ing the target and beam pipe and surrounding theMVD. An electric field along the cylinder axis sepa-rates positive gas ions from electrons created by ion-izing particles traversing the gas volume. The elec-trons drift with constant velocity towards the anodeat the upstream end face and create an avalanchedetected by a pad readout plane yielding informa-tion on two coordinates. The third coordinate of thetrack comes from the measurement of the drift timeof each primary electron cluster. In common TPCsthe amplification stage typically occurs in multi-wire proportional chambers. These are gated by

an external trigger to avoid a continuous backflowof ions in the drift volume which would distort theelectric drift field and jeopardize the principle ofoperation.

In PANDA the interaction rate is too high and thereis no fast external trigger to allow such an oper-ation. Therefore a novel readout scheme is em-ployed which is based on GEM foils as amplificationstage. These foils have a strong suppression of ionbackflow, since the ions produced in the avalancheswithin the holes are mostly caught on the backsideof the foil. Nevertheless about two ions per pri-mary electron are drifting back into the ionisationvolume even at moderate gains. The deformationof the drift field can be measured by a laser calibra-tion system and the resulting drift can be correctedaccordingly. In addition a very good homogeneityof the solenoid field with a low radial component isrequired.

A further challenge is the large number of tracks ac-cumulating in the drift volume because of the highrate and slow drift. While the TPC is capable ofstoring a lot of tracks at the same time, their as-signment to specific interactions has to be done bytime correlations with other detectors in the targetspectrometer. To achieve this, first a tracklet re-construction has to take place. The tracklets arethen matched against other detector signals or arepointed to the interaction. This requires either highcomputing power close to the readout electronics ora very high bandwidth at the full interaction rate.

Forward GEM Detectors Particles emitted atangles below 22 which are not covered fully by theStraw Tube Tracker or TPC will be tracked by threestations of GEM detectors placed 1.1 m, 1.4 m and1.9 m downstream of the target. The chambers haveto sustain a high counting rate of particles peakedat the most forward angles due to the relativisticboost of the reaction products as well as due to thesmall angle pp elastic scattering. With the envis-aged luminosity, the expected particle flux in thefirst chamber in the vicinity of the 5 cm diameterbeam pipe is about 3·104 cm−2s−1. In addition itis required that the chambers work in the 2 T mag-netic field produced by the solenoid. Drift cham-bers cannot fulfill the requirements here since theywould suffer from aging and the occupancy wouldbe too high. Therefore gaseous micropattern detec-tors based on GEM foils as amplification stages arechosen. These detectors have rate capabilities threeorders of magnitude higher than drift chambers.

In the current layout there are three double planeswith two projections per plane. The readout plane


Figure 2.7: Straw Tube Tracker in the Target Spectrometer.

Figure 2.8: GEM Time Projection Chamber in the Target Spectrometer.

is subdivided in an outer ring with longer and aninner ring with shorter strips. The strips are ar-ranged in two orthogonal projections per readoutplane. Owing to the charge sharing between striplayers a strong correlation between the orthogonalstrips can be found giving an almost 2D informationrather than just two projections.

The readout is performed by the same frontendchips as are used for the silicon microstrips. Thefirst chamber has a diameter of 90 cm, the last oneof 150 cm. The readout boards carrying the ASICsare placed at the outer rim of the detectors.

2.3.1.5 Cherenkov Detectors andTime-of-Flight

Charged particle identification of hadrons and lep-tons over a large range of angles and momenta isan essential requirement for meeting the physics

objectives of PANDA. There will be several dedi-cated systems which, complementary to the otherdetectors, will provide means to identify particles.The main part of the momentum spectrum above 1GeV/c will be covered by Cherenkov detectors. Be-low the Cherenkov threshold of kaons several otherprocesses have to be employed for particle identifi-cation: The tracking detectors are able to provideenergy loss measurements. Here in particular theTPC with its large number of measurements alongeach track excels. In addition a time-of-flight barrelcan identify slow particles.

Barrel DIRC Charged particles in a mediumwith index of refraction n, propagating with veloc-ity βc < 1/n, emit radiation at an angle ΘC =arccos(1/nβ). Thus, the mass of the detected par-ticle can be determined by combining the velocityinformation determined from ΘC with the momen-tum information from the tracking detectors.


A very good choice as radiator material for thesedetectors is fused silica (i.e. artificial quartz) witha refractive index of 1.47. This provides pion-kaon-separation from rather low momenta of 800 MeV/cup to about 5 GeV/c and fits well to the compactdesign of the target spectrometer. In this way theloss of photons converting in the radiator materialcan be reduced by placing the conversion point asclose as possible to the electromagnetic calorimeter.

At polar angles between 22 and 140, particle iden-tification will be performed by the detection of in-ternally reflected Cherenkov (DIRC) light as re-alized in the BaBar detector [7]. It will consistof 1.7 cm thick quartz slabs surrounding the beamline at a radial distance of 45 - 54 cm. At BaBarthe light was imaged across a large stand-off vol-ume filled with water onto 11 000 photomultipliertubes. At PANDA, it is intended to focus the im-ages by lenses onto micro-channel plate photomul-tiplier tubes (MCP PMTs) which are insensitive tomagnet fields. This fast light detector type allows amore compact design and the readout of two spatialcoordinates. In addition MCP PMTs provide goodtime resolution to measure the time of light prop-agation for dispersion correction and backgroundsuppression.

The DIRC design with its compact radiatormounted close to the EMC minimises the conver-sions. Part of these conversions can be recoveredwith information from the DIRC detector, as wasshown by BaBar[8].

Forward Endcap DIRC A similar concept canbe employed in the forward direction for particlesbetween 5 and 22. The same radiator, fused sil-ica, is to be employed however in shape of a disk.At the rim around the disk focusing will be doneby mirroring quartz elements reflecting onto MCPPMTs. Once again two spatial coordinates plus thepropagation time for corrections will be read. Thedisk will be 2 cm thick and will have a radius of110 cm. It will be placed directly upstream of theforward endcap calorimeter.

Barrel Time-of-Flight For slow particles atlarge polar angles particle identification shall beprovided by a time-of-flight detector. In the tar-get spectrometer the flight path is only in the orderof 50 – 100 cm. Therefore the detector must have avery good time resolution between 50 and 100 ps.

Implementing an additional start detector would in-troduce too much material close to the interactionpoint deteriorating considerably the resolution of

the electromagnetic crystal calorimeter. In the ab-sence of a start-detector relative timing of a mini-mum of two particles has to be employed.

As detector candidates scintillator bars and stripsor pads of multi-gap resistive plate chambers areconsidered. In both cases a compromise betweentime resolution and material budget has to befound. The detectors will cover angles between 22

and 140 using a barrel arrangement around theSTT/TPC at 42 - 45 cm radial distance.

2.3.1.6 Electromagnetic Calorimeters

Expected high count rates and a geometrically com-pact design of the target spectrometer require a fastscintillator material with a short radiation lengthand Moliere radius for the construction of the elec-tromagnetic calorimeter (EMC). Lead tungstate(PbWO4) is a high density inorganic scintillatorwith sufficient energy and time resolution for pho-ton, electron, and hadron detection even at interme-diate energies [9, 10, 11]. For high energy physicsPbWO4 has been chosen by the CMS and ALICEcollaborations at CERN [12, 13] and optimized forlarge scale production. Apart from a short decaytime of less than 10 ns good radiation hardness hasbeen achieved [14]. Recent developments indicatea significant increase of light yield due to crystalperfection and appropriate doping to enable pho-ton detection down to a few MeV with sufficientresolution. The light yield can be increased by afactor of about 4 compared to room temperatureby cooling the crystals down to -25C.

The crystals will be 20 cm long, i.e. approximately22X0, in order to achieve an energy resolution be-low 2 % at 1 GeV [9, 10, 11] at a tolerable energy lossdue to longitudinal leakage of the shower. Taperedcrystals with a front size of 2.1 × 2.1 cm2 will bemounted with an inner radius of 57 cm. This implies11360 crystals for the barrel part of the calorime-ter. The forward endcap calorimeter will have 3600tapered crystals, the backward endcap calorimeter592. The readout of the crystals will be accom-plished by large area avalanche photo diodes in thebarrel and vacuum phototriodes in the forward andbackward endcaps.

The EMC allows to achieve an e/π ratio of 103

for momenta above 0.5 GeV/c. Therefore, e-π-separation does not require an additional gasCherenkov detector in favor of a very compact ge-ometry of the EMC.


2.3.1.7 Muon Detectors

Muons are an important probe for, among others,J/ψ decays, semi-leptonic D-meson decays and theDrell-Yan process. The strongest background arepions and their decay daughter muons. However atthe low momenta of PANDA the signature is lessclean than in high energy physics experiments. Toallow nevertheless a proper separation of primarymuons from pions and decay muons a range track-ing system will be implemented in the yoke of thesolenoid magnet. Here a fine segmentation of theyoke as absorber with interleaved tracking detec-tors allows the distinction of energy loss processes ofmuons and pions and kinks from pion decays. Onlyin this way a high separation of primary muons fromthe background can be achieved.

In the barrel region the yoke is segmented in a firstlayer of 6 cm iron followed by 12 layers of 3 cm thick-ness. The gaps for the detectors are 3 cm wide. Thisis enough material for the absorption of pions inthe momentum range in PANDA at these angles. Inthe forward endcap more material is needed. Sincethe downstream door of the return yoke has to ful-fill constraints for space and accessibility, the muonsystem is split in several layers. Six detection lay-ers are placed around five iron layers of 6 cm eachwithin the door, and a removable muon filter withadditional five layers of 6 cm iron is located in thespace between the solenoid and the dipole. Thisfilter has to provide cut-outs for forward detectorsand pump lines and has to be built in a way that itcan be removed with few crane operations to alloweasy access to these parts.

As detector within the absorber layers rectangularaluminum drift tubes are used as they were con-structed for the COMPASS muon detection system.They are essentially drift tubes with additional ca-pacitively coupled strips read out on both ends toobtain the longitudinal coordinate.

2.3.1.8 Hypernuclear Detector

The hypernuclei study will make use of the mod-ular structure of PANDA. Removing the backwardendcap calorimeter will allow to add a dedicatednuclear target station and the required additionaldetectors for γ spectroscopy close to the entrance ofPANDA. While the detection of anti-hyperons andlow momentum K+ can be ensured by the univer-sal detector and its PID system, a specific targetsystem and a γ-detector are additional componentsrequired for the hypernuclear studies.

Active Secondary Target The production ofhypernuclei proceeds as a two-stage process. Firsthyperons, in particular ΞΞ, are produced on a nu-clear target. In some cases the Ξ will be slow enoughto be captured in a secondary target, where it reactsin a nucleus to form a double hypernucleus.

The geometry of this secondary target is determinedby the short mean life of the Ξ− of only 0.164 ns.This limits the required thickness of the active sec-ondary target to about 25–30 mm. It will consistof a compact sandwich structure of silicon microstrip detectors and absorbing material. In this waythe weak decay cascade of the hypernucleus can bedetected in the sandwich structure.

Germanium Array An existing germanium-array with refurbished readout will be used for theγ-spectroscopy of the nuclear decay cascades of hy-pernuclei. The main limitation will be the load dueto neutral or charged particles traversing the ger-manium detectors. Therefore, readout schemes andtracking algorithms are presently being developedwhich will enable high resolution γ-spectroscopy inan environment of high particle flux.

2.3.2 Forward Spectrometer

2.3.2.1 Dipole Magnet

A dipole magnet with a window frame, a 1 m gap,and more than 2 m aperture will be used for themomentum analysis of charged particles in the for-ward spectrometer. In the current planning, themagnet yoke will occupy about 2.5 m in beam di-rection starting from 3.5 m downstream of the tar-get. Thus, it covers the entire angular acceptanceof the target spectrometer of ±10 and ±5 in thehorizontal and in the vertical direction, respectively.The maximum bending power of the magnet will be2 Tm and the resulting deflection of the antiprotonbeam at the maximum momentum of 15 GeV/c willbe 2.2. The design acceptance for charged parti-cles covers a dynamic range of a factor 15 with thedetectors downstream of the magnet. For particleswith lower momenta, detectors will be placed insidethe yoke opening. The beam deflection will be com-pensated by two correcting dipole magnets, placedaround the PANDA detection system.

2.3.2.2 Forward Trackers

The deflection of particle trajectories in the field ofthe dipole magnet will be measured with a set ofwire chambers (either small cell size drift chambers


or straw tubes), two placed in front, two within andtwo behind the dipole magnet. This will allow totrack particles with highest momenta as well as verylow momentum particles where tracks will curl upinside the magnetic field.

The chambers will contain drift cells of 1 cm width.Each chamber will contain three pairs of detectionplanes, one pair with vertical wires and two pairswith wires inclined by +10 and -10. This configu-ration will allow to reconstruct tracks in each cham-ber separately, also in case of multi-track events.The beam pipe will pass through central holes inthe chambers. The most central wires will be sepa-rately mounted on insulating rings surrounding thebeam pipe. The expected momentum resolution ofthe system for 3 GeV/c protons is δp/p = 0.2 %and is limited by the small angle scattering on thechamber wires and gas.

2.3.2.3 Forward Particle Identification

RICH Detector To enable the π/K and K/pseparation also at the very highest momenta aRICH detector is proposed. The favored design is adual radiator RICH detector similar to the one usedat Hermes [15]. Using two radiators, silica aero-gel and C4F10 gas, provides π/K/p separation ina broad momentum range from 2–15 GeV/c. Thetwo different indices of refraction are 1.0304 and1.00137, respectively. The total thickness of the de-tector is reduced to the freon gas radiator (5%X0),the aerogel radiator (2.8%X0), and the aluminumwindow (3%X0) by using a lightweight mirror fo-cusing the Cherenkov light on an array of photo-tubes placed outside the active volume. It has beenstudied to reuse components of the HERMES RICH.

Time-of-Flight Wall A wall of slabs made ofplastic scintillator and read out on both ends by fastphototubes will serve as time-of-flight stop counterplaced at about 7 m from the target. In addition,similar detectors will be placed inside the dipolemagnet opening, to detect low momentum particleswhich do not exit the dipole magnet. The relativetime of flight between two charged tracks reachingany of the time-of-flight detectors in the experi-ment will be measured. The wall in front of theforward spectrometer EMC will consist of verticalstrips varying in width from 5 to 10 cm to accountfor the differences in count rate. With the expectedtime resolution of σ = 50 ps π-K and K/p separa-tion on a 3σ level will be possible up to momentaof 2.8 GeV/c and 4.7 GeV/c, respectively.

2.3.2.4 Forward ElectromagneticCalorimeter

For the detection of photons and electrons aShashlyk-type calorimeter with high resolution andefficiency will be employed. The detection is basedon lead-scintillator sandwiches read out with wave-length shifting fibers passing through the block andcoupled to photomultipliers. The technique has al-ready been successfully used in the E865 experi-ment [16]. It has been adopted for various otherexperiments [17, 18, 19, 20, 21, 22]. An energy res-olution of 4%/

√E [20] has been achieved. To cover

the forward acceptance, 26 rows and 54 columns arerequired with a cell size of 55 mm, i.e. 1404 mod-ules in total, which will be placed at a distance of7–8 m from the target.

2.3.2.5 Forward Muon Detectors

For the very forward part of the muon spectruma further range tracking system consisting of inter-leaved absorber layers and rectangular aluminiumdrift-tubes is being designed, similar to the muonsystem of the target spectrometer, but laid out forhigher momenta. The system allows discriminationof pions from muons, detection of pion decays and,with moderate resolution, also the energy determi-nation of neutrons and anti-neutrons.

2.3.3 Luminosity monitor

In order to determine the cross section for physicalprocesses, it is essential to determine the time in-tegrated luminosity L for reactions at the PANDAinteraction point that was available while collect-ing a given data sample. Typically the precisionfor a relative measurement is higher than for anabsolute measurement. For many observables con-nected to narrow resonance scans a relative mea-surement might be sufficient for PANDA, but forother observables an absolute determination of Lis required. The absolute cross section can be de-termined from the measured count rate of a spe-cific process with known cross section. In the fol-lowing we concentrate on elastic antiproton-protonscattering as the reference channel. For most otherhadronic processes that will be measured concur-rently in PANDA the precision with which the crosssection is known is poor.

The optical theorem connects the forward elasticscattering amplitude to the total cross section. Thetotal reaction rate and the differential elastic reac-


tion rate as a function of the 4-momentum transfert can be used to determine the total cross section.

The differential cross section dσel/dt becomes dom-inated by Coulomb scattering at very low valuesof t. Since the electromagnetic amplitude can beprecisely calculated, Coulomb elastic scattering al-lows both the luminosity and total cross section tobe determined without measuring the inelastic rate[23].

Due to the 2 T solenoid field and the existenceof the MVD it appears most feasible to mea-sure the forward going antiproton in PANDA. TheCoulomb-nuclear interference region corresponds to4-momentum transfers of −t ≈ 0.001 GeV2 at thebeam momentum range of interest to PANDA. At abeam momentum of 6 GeV/c this momentum trans-fer corresponds to a scattering angle of the antipro-ton of about 5 mrad.

The basic concept of the luminosity monitor is to re-construct the angle (and thus t) of the scattered an-tiprotons in the polar angle range of 3-8 mrad withrespect to the beam axis. Due to the large trans-verse dimensions of the interaction region when us-ing the pellet target, there is only a weak correlationof the position of the antiproton at e.g. z=+10.0 mto the recoil angle. Therefore, it is necessary toreconstruct the angle of the antiproton at the lumi-nosity monitor. As a result the luminosity monitorwill consist of a sequence of four planes of double-sided silicon strip detectors located as far down-stream and as close to the beam axis as possible.The planes are separated by 20 cm along the beamdirection. Each plane consists of 4 wafers (e.g. 2cm × 5 cm × 200 µm, with 50 µm pitch) arrangedradially to the beam axis. Four planes are requiredfor sufficient redundancy and background suppres-sion. The use of 4 wafers (up, down, right, left) ineach plane allows systematic errors to be stronglysuppressed.

The silicon wafers are located inside a vacuumchamber to minimize scattering of the antiprotonsbefore traversing the 4 tracking planes. The ac-ceptance for the antiproton beam in the HESR is±3 mrad, corresponding to the 89 mm inner di-ameter of the beam pipe at the quadrupoles lo-cated at about 15 m downstream of the interactionpoint. The luminosity monitor can be located in thespace between the downstream side of the forwardspectrometer hadronic calorimeter and the HESRdipole needed to redirect the antiproton beam outof the PANDA chicane back into the direction ofthe HESR straight stretch (i.e. between z=+10.0m and z=+12.0 m downstream of the target). Atthis distance from the target the luminosity moni-

tor needs to measure particles at a radial distanceof between 3 and 8 cm from the beam axis.

As pilot simulations show, at a beam momen-tum of 6.2 GeV/c the proposed detector mea-sures antiprotons elastically scattered in the range0.0006(GeV)2 < −t < 0.0035 (GeV)2, which spansthe Coulomb-nuclear interference region. Basedupon the granularity of the readout the resolutionof t could reach σt ≈ 0.0001 (GeV)2. In realitythis value is expected to degrade to σt ≈ 0.0005(GeV)2 when taking small-angle scattering into ac-count. At the nominal PANDA interaction rate of2 · 107/s there will be an average of 10 kHz/cm2 inthe sensors. In comparison with other experimentsan absolute precision of about 3% is considered fea-sible for this detector concept at PANDA, which willbe verified by more detailed simulations.

2.3.4 Data Acquisition

In many contemporary experiments the trigger anddata acquisition (DAQ) system is based on a twolayer hierarchical approach. A subset of speciallyinstrumented detectors is used to evaluate a firstlevel trigger condition. For the accepted events,the full information of all detectors is then trans-ported to the next higher trigger level or to stor-age. The available time for the first level decision isusually limited by the buffering capabilities of thefront-end electronics. Furthermore, the hard-wireddetector connectivity severely constrains both thecomplexity and the flexibility of the possible trig-ger schemes.

In PANDA, a data acquisition concept is being de-veloped which is better matched to the high datarates, to the complexity of the experiment and thediversity of physics objectives and the rate capabil-ity of at least 2 · 107 events/s.

In our approach, every sub-detector system is aself-triggering entity. Signals are detected au-tonomously by the sub-systems and are prepro-cessed. Only the physically relevant informationis extracted and transmitted. This requires hit-detection, noise-suppression and clusterisation atthe readout level. The data related to a particlehit, with a substantially reduced rate in the prepro-cessing step, is marked by a precise time stamp andbuffered for further processing. The trigger selec-tion finally occurs in computing nodes which accessthe buffers via a high-bandwidth network fabric.The new concept provides a high degree of flexi-bility in the choice of trigger algorithms. It makestrigger conditions available which are outside thecapabilities of the standard approach. One obvious


example is displaced vertex triggering.

In this scheme, sub-detectors can contribute to thetrigger decision on the same footing without re-strictions due to hard-wired connectivity. Differ-ent physics can be accessed either in parallel or viasoftware reconfiguration of the system.

High speed serial (10 Gb/s per link and beyond)and high-density FPGA (field programmable gatearrays) with large numbers of programmable gatesas well as more advanced embedded features arekey technologies to be exploited within the DAQframework.

The basic building blocks of the hardware infras-tructure which can be combined in a flexible way tocope with varying demands, are the following:

• Intelligent front-end modules capable of au-tonomous hit detection and data preprocessing(e.g. clustering, hit time reconstruction, andpattern recognition) are needed.

• A precise time distribution system is manda-tory to provide a clock norm from which alltime stamps can be derived. Without this,data from subsystems cannot be correlated.

• Data concentrators provide point-to-pointcommunication, typically via optical links,buffering and online data manipulation.

• Compute nodes aggregate large amounts ofcomputing power in a specialized architec-ture rather than through commodity PC hard-ware. They may employ fast FPGAs (fast pro-grammable gate arrays), DSPs (digital signalprocessors), or other computing units. Thenodes have to deal with feature extraction, as-sociation of data fragments to events, and, fi-nally, event selection.

A major component providing the link for all build-ing blocks is the network fabric. Here, special em-phasis is put on embedded switches which can becascaded and reconfigured to reroute traffic for dif-ferent physics selection topologies. Alternatively,with an even higher aggregate bandwidth of thenetwork, which according to projections of networkspeed evolution will be available by the time the ex-periment will start, a flat network topology whereall data is transferred directly to processing nodesmay be feasible as well. This requires a higher totalbandwidth but would have a simpler architectureand allow event selection in a single environment.The bandwidth required in this case would be atleast 200 GB/s. After event selection in the orderof 100-200 MB/s will be saved to mass storage.

An important requirement for this scheme is that alldetectors perform a continuous online calibrationwith data. The normal data taking is interleavedwith special calibration runs. For the monitoring ofthe quality of data, calibration constants and eventselection a small fraction of unfiltered raw data istransmitted to mass storage.

To facilitate the association of data fragments toevents the beam structure of the accelerator is ex-ploited: Every 1.8 µs there is a gap of about 400ns needed for the compensation of energy loss witha bucket barrier cavity. This gap provides a cleandivision between consistent data blocks which canbe processed coherently by one processing unit.

2.3.5 Infrastructure

The target for antiproton physics is located in thestraight section at the east side of the HESR. At thislocation an experimental hall of 43 m × 29 m floorspace and 14.5 m height is planned (see Fig. 2.9). Aconcrete radiation shield of 2 m thickness on bothsides along the beam line is covered by concrete barsof 1 m thickness to suppress the neutron sky shine.Within the elongated concrete cave the PANDAdetector together with auxiliary equipment, beamsteering, and focusing elements will be housed. Theroof of the cave can be opened and heavy compo-nents hoisted by crane.

The shielded beam line area for the PANDA ex-periment including dipoles and focusing elementsis foreseen to have 37 m × 9.4 m floor space and aheight of 8.5 m with the beam line at a height of3.5 m. The general floor level of the HESR is 2 mhigher. This level will be kept for a length of 4 m inthe north of the hall (right part in Fig. 2.9), to facil-itate transport of heavy equipment into the HESRtunnel.

The target spectrometer with electronics and sup-plies will be mounted on rails which makes it re-tractable to a parking position outside the HESRbeam line (i.e. into the lower part of the hall inFig. 2.9). The experimental hall provides additionalspace for delivery of components and assembly ofthe detector parts. In the south corner of the hall, acounting house complex with five floors is foreseen.The lowest floor will contain various supplies forpower, high voltage, cooling water, gases etc. Thenext level is planned for readout electronics includ-ing data concentrators. The third level will housethe online computing farm. The fourth floor is atlevel with the surrounding ground and will housethe control room, a meeting room and social roomsfor the shift crew. Above this floor, hall electricity


Figure 2.9: Top view of the experimental area indicating the location of PANDA in the HESR beam line. Thetarget center is at the center of HESR and is indicated as vertical dash-dotted line. North is to the right and thebeam comes in from the left. The roll-out position of the detector will be on the east side of the Hall.

supplies and ventilation is placed. A crane (15 t)spans the whole area with a hook at a height ofabout 10 m. Sufficient (300 kW) electric power willbe available.

Liquid helium coolant may come from the maincryogenic liquefier for the SIS rings. Alternatively, aseparate small liquefier (50 W cooling power at 4 K)would be mounted. The temperature of the build-ing will be moderately controlled. The more strin-gent requirements with respect to temperature andhumidity for the detectors have to be maintainedlocally. To facilitate cooling and avoid condensa-tion the target spectrometer will be kept in a tentwith dry air at a controlled temperature.

References

[1] Baseline Technical Report, subproject HESR,Technical report, Gesellschaft furSchwerionenforschung (GSI), Darmstadt,2006.

[2] H. Stockhorst et al., Stochastic Cooling forthe HESR at the GSI-FAIR Complex, inProc. of the European Accelerator ConferenceEPAC, Edinburgh, 2006.

[3] A. Lehrach, O. Boine-Frankenheim,F. Hinterberger, R. Maier, and D. Prasuhn,Nucl. Instrum. Meth. A561, 289 (2006).

[4] F. Hinterberger, Monte carlo simulations ofThin Internal Target Scattering in Ceslius, inBeam-Target Interaction and Intra-beamScattering in the HESR Ring: Emittance,Momentum Resolution and Luminosity,Bericht des Forschungszentrum Julich, 2006,Jul-Report No. 4206.


[5] Technical report, ATLAS Technical DesignReport 11, CERN/LHCC 98-13.

[6] Technical report, CMS Technical DesignReport 5, CERN/LHCC 98-6.

[7] H. Staengle et al., Nucl. Instrum. Meth.A397, 261 (1997).

[8] A. Adametz, ”Preshower Measurement withthe Cherenkov Detector of the BABARExperiment Aleksandra Adametz”, Diplomathesis, Master’s thesis, University Heidelberg,2005.

[9] K. Mengel et al., IEEE Trans. Nucl. Sci. 45,681 (1998).

[10] R. Novotny et al., IEEE Trans. Nucl. Sci. 47,1499 (2000).

[11] M. Hoek et al., Nucl. Instrum. Meth. A486,136 (2002).

[12] Technical Proposal, CERN/LHC 9.71.

[13] Technical Proposal, 1994, CERN/LHCC94-38, LHCC/P1.

[14] E.Auffray et al., Moscow, 1999.

[15] N. Akopov et al., Nucl. Instrum. Meth.A479, 511 (2002).

[16] G. S. Atoyan et al., Nucl. Instrum. Meth.A320, 144 (1992).

[17] G. David et al., Performance of the PHENIXEM calorimeter, Technical report, PHENIXTech. Note 236, 1996.

[18] A. Golutvin, (1994), HERA-B Tech. Note94-073.

[19] LHCb Technical Proposal CERN LHCC 98-4,LHCC/P4, 1998.

[20] I.-H. Chiang et al., (1999), KOPIO Proposal.

[21] H. Morii, (2004), Talk at NP04 Workshop atJ-PARC.

[22] G. Atoyan et al., Test beam study of theKOPIO Shashlyk calorimeter prototype, inProceedings of “CALOR 2004”, 2004.

[23] T. A. Armstrong et al., Phys. Lett. B385,479 (1996).

31

3 Design Considerations

The PANDA experiment aims at various physicstopics related to the very nature of large distancestrong binding. Although the details and observ-ables turn out to be different, most channels shareone important feature - many photons and/or elec-trons/positrons in the final state. Examples arehidden charm decays of charmonium hybrids withneutral recoils and low-mass isoscalar S-waves (ap-pearing in π0π0), radiative charm decays and thenucleon structure physics. This puts special empha-sis on the electromagnetic calorimeter, and its basicperformance parameters have to be tuned to accom-plish the effective detection of these channels in or-der to succeed in the basic programme of PANDA.

The basic function of an electromagnetic calorime-ter is the efficient reconstruction of electrons,positrons and photons with high efficiency and lowbackground. This is performed by measuring thedeposited energy (E) and the direction via the pointof impact. (θ and φ). High resolution is mandatoryfor a sufficient resolving power for final states withmultiple electrons, positrons and photons.

Photons in the final state can originate from var-ious sources. The most abundant sources are π0

and η mesons. Important probes are radiative char-monium decays (like χc1 → J/ψγ), which are sup-pressed by the charm production yield or directphotons from rare electromagnetic processes. Todistinguish radiatively decaying charmonium anddirect photons from background with undetectedphotons (from π0 and η rich states) it is of utmostimportantance to identify very efficiently π0 and ηby reducing the number of undetected photons dueto solid angle or energy threshold.

The EMC may also provide timing information.This is needed to accomplish a proper distinctionamong different events. The annihilation rate goesup to several ·107/s leading to σt ≈ 10 ns. Thus afast scintillator is required for operation.

PANDA will not have a threshold Cherenkov de-tector to discriminate pions from electrons andpositrons. Therefore, the EMC has to add com-plementary information to the basic E/p informa-tion. Lateral shower shape information is neededto discriminate e± from background. These infor-mations are deduced from the difference of lateralshower shapes. Hadronic showers (KL, n, chargedhadrons) in an electromagnetic calorimeter differsignificantly due to the difference in energy loss per

interaction and the elementary statistics of theseprocesses. The quality of this discrimination doesnot (to first order) depend on the actual choice ofcrystal geometry and readout, as long as the frontface size of the crystal is matched to the Moliereradius. Therefore, this does not place a strong re-quirement on the calorimeter design. Nevertheless,the final design process must incorporate an opti-mization of the electron-pion separation power.

One basic aspect of the PANDA EMC is the require-ment on compactness to reduce cost. The price ofthe scintillator and the surrounding magnet scaleswith the cube of their dimension thus, e.g., leadingto a 50 % increase in price for ≈ 15 % increase inradius.

3.1 Electromagnetic ParticleReconstruction

3.1.1 Coverage Requirements

3.1.1.1 Energy Threshold

Apart from energy resolution the minimum photonenergy Ethres being accessible with the EMC is animportant issue since it determines the very accep-tance of low energy photons.

Figure 3.1: Percentage of π0 loss as a function of en-ergy threshold.


It is easily shown that the bare photon and π0 lossrate is not very large even for energy thresholdshigher than 20 MeV as long as a limit below 50 MeVis maintained (e.g. π0 loss in Fig. 3.1). Also thenumber of events dropped due to the energy thresh-old within the limits just mentioned is not dramatic.What is really driving the limit is the fact, that thephysics being performed with PANDA requires aneffective background rejection to distinguish radia-tive charmonium decays (as a tag for exotic charmo-nia) and other electromagnetic probes from back-ground events with at least one undetected photon.So even small losses result in an unacceptable dropof the signal-to-background ratio.

How the energy threshold affects the sensitivity forphysics with PANDA can be exemplified by the hy-brid production channel pp → ηc1η with the char-monium hybrid ηc1 decaying to χc1π0π0 leading tothe final state J/ψγπ0π0η. The production ratio be-tween potential background and signal is expectedto have the same order of magnitude. We consideronly background with cc content - e.g. J/ψ3π0η,since generic light quark background disappears toan undetectable level after the electron-id and char-monium cuts. However the decay branching ratiossuppress the signal by one or two orders of magni-tude compared to the background. Therefore, everyeffect of background leaking into the selection hasto be minimized. Simulations show that the signal-to-background ratio depends almost quadraticallyon the minimum photon energy.

This is due to the fact that, if the energies of theundetected photons are small enough, the residualparticles may be recognized as an exclusive eventand may contribute to the background of a chan-nel with one photon less. Due to imperfect energyresolution, there is a certain probability, that the re-duced set of particles of the background event fulfillall selection criteria of the signal channel and evensurvive a kinematic fit with high probability.

From this discussion it is clear, that the lowestachievable value of Ethres is necessary to get theoptimum in terms of photon detection. AlthoughEthres = 10 MeV would be ideal in that context,technical limitations like noise or a reasonable cov-erage of a low energy shower may increase thisvalue, but at least Ethres ≤ 20 MeV should beachieved to reach the physics goals of PANDA. Moredetails are given in the simulation part of the report(Sec. 9).

3.1.1.2 Geometrical Coverage

The acceptance due to geometrical cuts is to 1st or-der proportional to (Ω/4π)n (n being the number ofe±, γ. This is illustrated by the example of 6 pho-tons and 90 % solid angle coverage where the geo-metrical acceptance drops to 1/2. Since final stateswith many electrons, positrons and/or photons areone of the prime signals, these put, therefore, strongrequirements on the angular coverage. As demon-strated in Sec. 3.1.1.1, undetected photons are animportant source of background effects and the lossdue to the solid angle coverage should be minimizedto the mechanical limit. In the backward regionthe beampipe is the limiting factor, but a maximumopening of≈ 10−15 should be reached. In the for-ward direction a dipole bends all charged particles.In particular the p-beam is inclined by 2, whichallows for 0 calorimetry to maximize performance.Additional holes for mechanics, support, pipes andcables have to be considered. Nevertheless, the an-gular coverage should be maximized and the aim is99 % 4π coverage in the center-of-mass system. Inaddition to the target spectrometer EMC the for-ward part down to 0 will be covered by a shashlykdetector. The actual partioning between forwardendcap EMC and forward shashlyk is optimized toallow high momentum tracks to enter the spectrom-eter dipole.

3.1.1.3 Dynamical Energy Range

Fig. 3.2 shows the range of energies from the DPMgenerator for two different momentum settings forthe antiproton beam. The highest energies are inforward direction, while backward particles are rel-atively low in energy. Since low energy capabilitiesare mandatory for all regions of the calorimeter thedynamic range is mainly driven by the highest en-ergy possible. The dynamic range for the variousdetector parts should at least cover in

• backward endcap EMC: 10(20) MeV- 0.7 GeV,

• barrel EMC: 10(20) MeV- 7.3 GeV, and

• forward endcap EMC: 10(20) MeV- 14.6 GeV.

3.1.1.4 Vertex Distribution

The primary vertex distribution is dictated by theoverlap of beam and target stream. The worst caseappears for a cluster-jet target with a spread in theorder of a cm. In order to ensure that no photonescapes in the dead area between neighbouring crys-tals, a non-pointing geometry is needed. Ideally, the


Figure 3.2: Photon energy distribution vs. lab. angle for two momentum settings.

distance of closest approach of the pointer normalto the frontface of the crystal to the primary ver-tex should be at least 4 cm. Thus the focus of theendcap is off the average vertex position in z by atleast 10 cm. In the barrel part a tilt of 4 is neededto fulfill the same requirement.

3.1.2 Resolution Requirements

3.1.2.1 Energy Resolution

Apart from obtaining the best resolution to ensurethe exclusiveness of events, the choice of the ap-propriate energy resolution has various additionalaspects:

• Precise measurement of electron and positronenergies for

– very accurate E/p determination, and

– optimum J/ψ mass resolution

• Efficient recognition of light mesons (e.g. π0

and η) to reduce potential background.

Precise E/p measurement is an important asset topositively identify electrons and positrons againstpions. This is achieved if the error on the electronenergy is negligible compared to the momentum er-ror from the tracking detectors (≈ 1 %). This puts alimit on the resolution σE

E ≤ 1.0 % at high energies.

Another effect of bad energy resolution is the badmass determination of π0 and η mesons. At lowenergies this is due to the 1/

√E dependence of

the energy resolution, while at high energies thisis dominated by the constant term. Experimentslike Crystal Barrel, CLEO, BaBar and BES, withsimilarities in the topology and composition of fi-nal states have proven, that a π0 width of less than8 MeV and η width of less than 30 MeV is necessaryfor reasonable final state decomposition. Assumingan energy dependence of the energy resolution ofthe form

σEE

= a⊕ b√E/GeV

(3.1)

leads to the requirement a ≤ 1 % and b ≤ 2 %.This balance of values also ensures a J/ψ resolutionwhich is well matched with the resolution of thetypical light recoil mesons (like η and ω).

3.1.2.2 Single Crystal Threshold

Energy threshold and energy resolution place a re-quirement on the required minimum single crystalenergy Extl.

As a consequence this threshold puts a limit on thesingle crystal noise, since the single crystal cut (sev-eral MeV) should be high enough to exclude a ran-dom assignment of photons. This requirement canbe relaxed by demanding a higher single-crystal en-ergy to identify a bump, i.e. a local maximum inenergy deposition (e.g. 10 MeV). Starting withthose seeds, additional crystals are only collectedin the vicinity of this central crystal. With typ-ically 10 neighbours and not more than 10 par-ticles the probability of less or equal of one ran-dom crystal hit per event for a single crystal cut of


Extl = 3σnoise. Fig. 3.3 shows that a single crystalthreshold of Extl = 3 MeV is needed to obtain therequired energy resolution. Also the containment oflow energy photons in more than the central crystalcan only be achieved for Extl ≤ 3 MeV. From thisconsideration we deduce a limit for the total noiseof σE,noise = 1 MeV.

Figure 3.3: Comparison of the energy resolutions forthree different single crystal reconstruction thresholds.The most realistic scenario with a noise term of σ =1 MeV and a single crystal threshold of Extl = 3 MeVis illustrated by triangles, a worse case (σ = 3MeV,Extl = 9 MeV) by circles and the better case (σ =0.5 MeV, Extl = 1.5 MeV) by rectangles.

3.1.2.3 Spatial Resolution

The spatial resolution is mainly governed by thegranularity. The reconstruction of the point of im-pact is achieved by weighted averaging of hits in ad-jacent crystals. In addition, to identify overlappingphotons (e.g. due to π0 with small opening angles)it is mandatory to efficiently split crystal clustersinto individual photons. This requires, that the cen-tral hits of the involved photons are separated by atleast two crystal widths to assure two local maximain energy deposition. If this can not be achieved,a cluster moment analysis has to be performed toidentify π0 without identifying individual photons.This is possible down to a spatial separation of onecrystal.

Fig. 3.4 shows the energy distribution of π0 andη for a cocktail of events with many photons forthe three detector regions. The average (maxi-mum) π0/η momenta for the three detector partsare < 1 GeV (< 1 GeV), ≈ 2 GeV (≈ 7 GeV) and≈ 5 GeV (≈ 14 GeV) for backward endcap EMC,barrel EMC and forward endcap EMC, respectively,for the highest incident antiproton momentum of

15 GeV/c. Fig. 3.5 shows the minimum opening an-gle for various π0 momenta. The angular coverageof a crystal should be tuned to the smallest π0 open-ing angle possible for this subdetector and shouldnot exceed 10, 2 and 0.5, respectively, to fullyresolve the photons. These angles may be a factor2 larger when taking cluster moment analysis intoaccount.

The required spatial resolution is mainly governedby the required width of the π0 invariant mass peakin order to assure proper final-state decomposition.Fig. 3.6 shows the effect on the π0 mass resolu-tion as a function of momentum for various spa-tial resolution values. Taking into account the av-erage π0 energies a narrow π0 width below 8 MeVis maintained for a cos θ dependent spatial resolu-tion of ≤0.5, ≤0.3 and ≤0.1 for backward endcapEMC, barrel EMC and forward endcap EMC, re-spectively. These resolutions are in accordance withthe granularity requirement. They can be achievedby an asymmetric geometry where compactness isstill maintained for the radial component, while itextends more to the forward direction.

3.2 Environment

3.2.1 Surrounding Detectors

3.2.1.1 Magnet System

The EMC will be operated in a high solenoidal mag-netic field (2 T). Therefore the sensors perpendicu-lar (barrel part) to the field have to be field insen-sitive. In the endcaps the requirements due to themagnetic field are relaxed since the sensors wouldbe differently oriented.

Backscattering in the magnet may take place.These effects are accounted for in the detector sim-ulations in subsequent chapters.

3.2.1.2 DIRC

Detectors of Internally Reflected Cherenkov Light(DIRC) are placed in the barrel part and in the for-ward endcap for particle identification. They arelocated near the front face of the scintillator andadd substantially to the material budget before theEMC. It has been shown, that the preshower infor-mation of the Cherenkov light of the shower elec-trons of such a detector can be used to repair thedistorted calorimeter information.


Figure 3.4: π0 and η energy spectrum (scale in GeV) for pp = 15 GeV/c for the forward endcap EMC (left),

barrel EMC (middle) and backward endcap EMC (right), for π0 (top) and η (bottom).

3.2.1.3 Other Systems

Various different target systems are foreseen for thePANDA experiment. Gas-Jet and Pellet targets re-quire a target pipe perpendicular to the beam pipe.Since pumps have to be placed outside of the de-tector, the target pipes extend to the end of themagnet, thus crossing the EMC. Mechanical cut-outs for these pipes have to be foreseen.

3.2.2 Count-rate and Occupancy

3.2.2.1 Signal Load

Fig. 3.7 and 3.8 show the hit rates (Eγ >1 MeV,DPM background generator) for a geometricalsetup which fulfills the spatial requirements alreadymentioned. The rates are for the barrel part andthe forward endcap, respectively. The detector hasto be able to digest the maximum hit rate whichhappens for the highest beam momenta. The max-imum rates per crystal are ≈60 kHz and ≈500 kHzfor barrel EMC and forward endcap EMC respec-tively. For heavy targets the single crystal rate forthe barrel EMC goes up to ≈100 kHz due to the


Figure 3.5: Minimum π0 opening angle vs. beam mo-mentum.

[GeV/c]0π

p0 2 4 6 8 10 12 14

]2 [G

eV/c

)0 π

m(

σ

0.005

0.01

0.015

0.02

0.025

0.03 E/E = 0.017 / Eσ with )0πm(σ

° = 0.02ασ

° = 0.06ασ

° = 0.10ασ

° = 0.28ασ

° = 0.50ασ

Figure 3.6: π0 mass resolution for various spatial res-olution values vs. beam momentum.

reduced boost compared to pp events.

3.2.2.2 Response and Shaping Time

The response time has to be short enough to allowevent identification. Since the p beam will have atime structure, PANDA has to be prepared to ac-cept and instantaneous rate of up to 50 MHz onshort timescales. This condition leads to a requiredtime resolution for the relevant hit time t0 of 3 ns orless. The shaping time of the preamplification stageshould be longer than the actual decay time of thescintillation mechanism to collect the whole signal.The shaping time should be maximised to improvethe noise level. If the shaping time becomes toolong pile-up occurs, since new hits are delivered on

Figure 3.7: Hit rate in the barrel part from the DPMbackground generator at pp=14 GeV/c.

Figure 3.8: Hit rate in the forward part from the DPMbackground generator at pp=14 GeV/c.

top of a preceding hit. As mentioned previously, un-detected photons are a severe cause of background.Therefore a maximum effective pile-up rate of lessthan 1% is required. In particular in the forward re-gion this would lead to extremely short (and there-fore unrealistic) shaping times. To reduce photonloss, the pile-up hits should be recovered. This canbe achieved with a FADC system with appropri-ate sample frequency and proper hit detection (inhardware or software). In case of a FADC readouta pile-up rate up to 10% can be tolerated (includ-ing variations of the instantaneous rate) leading toshaping times of not more than ≈ 100 ns for theforward endcap EMC and ≈ 400 ns for barrel EMCand backward endcap EMC.


Using the hit rates from Sec. 3.2.2.1 and requiringless than 1 % loss due to pile-up the shaping timeswould drop.

3.2.2.3 Radiation Hardness

Potential detector setups which fulfill the require-ments already mentioned have been used to checkfor radiation doses. These simulations show thatfor the highest beam momenta the innermost crys-tals (setup as in Sec. 3.2.2.1) accumulate an energydose of 25 mJ/h for pp generic background (DPMbackground generator). The maximum dose doesnot increase with heavier targets. Due to the muchreduced boost of the overall system, the hit ratein the innermost crystals is reduced dramatically(even with the same luminosity) and the dose in thiscase is distributed more evenly among crystal rings.For a crystal weight of 1 kg (in case of e.g. PWOor BGO) and a typical annual operation of notmore than 5 kh the maximum annual dose would be125 Gy (or 12.5 kRad/year). This strong constraintapplies only to the forward crystals. For backwardendcap EMC and barrel EMC the maximum energydoses (for pp events) are 30µJ/h (0.15 Gy annually)and 1.4 mJ/h (7 Gy annually), respectively. Dueto the reduced boost, part of the barrel EMC andin particular the whole backward endcap EMC willhave a much higher rate than with pp events. Tobe on the safe side, all non-forward crystals shouldbe radiation hard to an annual limit of 10 Gy.

3.2.3 Operational Aspects

3.2.3.1 Long- and Medium-term Stability

PANDA will be operated in a factory mode, thusrunning for several months (up to 9 months) andwill be suspended for the rest of the year for main-tenance, repair and upgrade. This requires a longMean-Time-between-Failure to the level of not morethan 10 channels lost per year.

3.2.3.2 Maintenance of Sensors,Electronics and Environment

PANDA is planned to operate for at least onedecade. Therefore it is mandatory to allow accessto all crystals, sensors and electronics for mainte-nance during down-time periods. It is envisagedto dismount the detector annually on the detectormodule level for service work. Therefore the overallmounting procedure has be reversible and detectormodule oriented.

3.2.3.3 Calibration and MonitoringPrerequisites

The energy calibration has to be performed at a pre-cision much better than the energy resolution, thusat the sub percent level. The PANDA DAQ relieson online trigger decisions performed on computenodes. The use of the electromagnetic calorimeterfor the trigger decisions requires the availability ofcalibration constants with sufficient precision in realtime.

The monitoring system should detect any changesin the readout chain starting from the light trans-mission in the crystals to the digitizing modules.

3.3 Production and Assembly

3.3.1 Production Schemes

The fabrication of calorimeters has been exercisedmany times in the past. Many production and as-sembly procedures have been used in the past forother large scale projects. Whenever possible exist-ing production schemes, equipment and personnelshould be assimilated to reduce overall cost. De-cent quality assurance processes are needed for allmass-produced parts.

3.3.2 Assembly Schemes

Modern large-scale detectors require a large numberof channels and many mass-produced parts whichhave to be screened and stored for further assem-bly. Since not all parts can be stored and assem-bled at a single laboratory, complex logistics hasto be arranged to ensure a smooth supply of partsthroughout the production process.

3.4 Conclusion

The full list of requirements is compiled in Ta-ble 3.1.

In this technical design report of the EMC, we willdemonstrate that a compact lead tungstate crys-tal calorimeter being operated at low temperaturesand read out with LAAPDs and VPTs will fulfillthe listed requirements on energy resolution, spatialresolution, energy threshold, timing and radiationhardness.

Test equipment of CMS, employed in the past forlarge-scale production, can be reused for PANDA.


Required performance valueCommon propertiesenergy resolution σE/E ≤ 1 %⊕ ≤2 %√

E/GeVenergy threshold (photons) Ethres 10 MeV (20 MeV tolerable)energy threshold (single crystal) Extl 3 MeVrms noise (energy equiv.) σE,noise 1 MeVangular coverage % 4π 99 %mean-time-between-failures tmtbf 2000 y(for individual channel)Subdetector specific properties backward barrel forward

(≥ 140) (≥ 22) (≥ 5)energy range from Ethres to 0.7 GeV 7.3 GeV 14.6 GeVangular equivalent of crystal size θ 4 1

spatial resolution σθ 0.5 0.3 0.1

maximum signal load fγ (Eγ > Extl) 60 kHz 500 kHz(pp-events) maximum signal load fγ (Eγ > Extl) 100 kHz 500 kHz(all events) shaping time ts 400 ns 100 nsradiation hardness 0.15 Gy 7 Gy 125 Gy(maximum annual dose pp-events)radiation hardness 10 Gy 125 Gy(maximum annual dose from all events)

Table 3.1: Main requirements for the PANDA EMC. Rates and doses are based on a luminosity of L = 2 ·1032cm−1s−1.

Moreover, the expertise of technicians operatingthese devices can be advantageously exploited.

39

4 Scintillator Material

4.1 Inorganic Scintillators

The concept of PANDA places the target spectrom-eter EMC inside the super-conducting coil of thesolenoid. Therefore, the basic requirement of anyappropriate scintillator material has to be its com-pactness to minimize the radial thickness of thecalorimeter layer [1, 2]. Of similar importance,high interaction rates, the ambitiously large dy-namic range of photon energies, sufficient energyresolution and efficiency and finally a moderate ra-diation hardness rule out most of the well knownscintillator materials. Finally, even a compact ge-ometrical design requires due to a minimum gran-ularity a large quantity of crystal elements, whichrely on existing technology for mass production toguarantee the necessary homogeneity of the wholecalorimeter. Presently and even in the near future,no alternative materials besides lead tungstate willbecome available.

Of the materials listed in the table below BismuthGermanate (Bi4Ge3O12, BGO) is a well known scin-tillator material since many decades [3]. It has beenapplied in several experiments, like L3 at CERNor GRAAL at Grenoble, providing well performingelectromagnetic calorimeters [4, 5, 6]. However,presently it is only used in large quantities but smallsamples for medical applications such as Positron-Electron-Tomography (PET) scanners [7]. Theproperties of BGO, listed in Table 4.1, allow forthe design of a very compact calorimeter. The lightyield is comparable to cooled PWO-II crystals. Theemission wavelength is well suited for an efficientreadout with photo- or avalanche diodes. However,the moderate decay time imposes a strong limita-tion on the necessary high rate capability for theenvisaged interaction rate of 107/s of PANDA. Inaddition, it excludes in general any option for thegeneration of a fast timing information on the levelof ≤1–2 ns or even below for higher energies. Con-cerning the radiation hardness, which has not beena major issue for the former applications, contro-versial results are reported in the literature. Someauthors report recovery times in the order of daysor online recovery by exposure to intense light fromfluorescent lamps at short wavelengths [8, 9, 10, 11].Since most of these studies are rather old, an updatewith new studies using full size scintillation crystalsproduced recently would become necessary.

CeF3 has been invented already in 1989 by D. F.

Anderson [12] and identified as a fast scintillatorwith maximum emission wavelength between 310and 340 nm. The luminescence process is domi-nated by radiative transitions of Ce3+ ions, whichexplains the fast decay time of ∼30 ns and the in-sensitivity to temperature changes. The lumines-cence yield is comparable or even higher than thefast component of BaF2, corresponding to approx-imately 5 % of NaI(Tl) as a reference. Since thecrystal matrix is extremely radiation hard, it wasconsidered for the CMS calorimeter during a longperiod of R&D, supported by the short radiationlength and Moliere radius, respectively. However,homogeneous crystal samples beyond 10X0 length,grown by the Bridgeman method, have never beenproduced with adequate quality. Detailed studiesof the response function [13] to low energy protonsand high energy photons have documented excel-lent time and energy resolutions, which are limitedonly in case of the reconstruction of the electro-magnetic shower energy by the inhomogeneity ofthe available crystals. Further improvement relieson a significant optimization of the manufacturingprocess. In case of PANDA, the implementation ofCeF3 has not been considered. It would on onehand double the radial thickness of the calorimetershell. On the other hand, the effective light yieldwould be degraded due to the mismatch of the emis-sion wavelength in the near UV with the optimumquantum efficiency of LAAPDs.

In the last decade cerium doped silicate based heavycrystal scintillators have been developed for medi-cal applications, which require high light output forlow energy γ-ray detection and high density to allowfor small crystal units. Mass production of smallcrystals of Gd2SiO5 (GSO), Lu2SiO5 (LSO) andLu2(1−X)Y2XSiO5 (LYSO) has been established inthe meantime. Because of the high stopping powerand fast and bright luminescence, the latter mate-rial has also attracted interest in calorimetry. How-ever, the need for large homogeneous samples andthe presently high costs, partly justified by the highmelting point, are retarding the R&D. LYSO, whichhas almost identical physical and scintillation prop-erties as LSO and shows identical emission, exci-tation and optical transmittance spectra, has beenrecently produced in samples up to 20 cm length[14, 15, 16, 17]. The light output of small sam-ples can reach values 8 times of BGO but witha decay time shorter by one order of magnitude,


Parameter CeF3 LSO/LYSO:Ce BGO PWO PWO-IIρ g/cm3 6.16 7.40 7.13 8.28X0 cm 1.77 1.14 1.12 0.89RM cm 2.60 2.30 2.30 2.00τdecay ns 30 40 300 6.5λmax nm 330 420 480 420n at λmax 1.63 1.82 2.15 2.24/2.17relative LY % (LY NaI) 5 75 9 0.3 at RT 0.6 at RT

0.8 at -25C 2.5 at -25Chygroscopic no no no nodLY/dT %/C 0.1 0 -1.6 - 2.7 at RT -3.0 at RTdE/dx (MIP) MeV/cm 6.2 9.6 9.0 10.2

Table 4.1: Relevant properties of CeF3, LSO/LYSO:Ce, BGO, PWO taken from the updated table of the ParticleData Group 2007. The specific parameters of PWO-II are deduced from the presented test measurements.

determined by the Ce-activation. Detailed stud-ies even within the PANDA-collaboration have beenperformed and response functions have been ob-tained even for high-energy photons up to 500 MeV[16]. Aiming at applications for the next generationof homogeneous calorimeters, detailed studies of fullsize crystals with respect to homogeneity, scintilla-tion processes and radiation hardness are part of aresearch program promoted by a group at Caltech[17].

4.1.1 Specific Requirements for thetarget spectrometer EMC

As pointed out in the previous chapter in selectedfigures based on simulations assuming already tosome extent parameters of PWO-II, the target spec-trometer EMC requires very ambitious specifica-tions to be adapted to the physics program. Themain features are:

• high rate capability, which requires a fast scin-tillation kinetics, appropriate granularity tominimize pile-up as well as guarantee optimumreconstruction of the center of the electromag-netic shower;

• sufficient luminescence yield to achieve goodenergy resolution in particular at the lowestphoton energies in the MeV range, which goesin parallel with a minimum energy threshold ofthe individual crystal;

• timing information, primarily to reduce back-ground and to provide an efficient correlationwith other detector components for particleidentification;

• adapted geometrical dimensions to contain themajor part of the electromagnetic shower andto minimize the impact of leakage fluctuationsand direct ionization in the photo sensor;

• radiation hardness to limit the loss in opticaltransparency to a tolerable level.

Based on the its intrinsic parameters, PWO, asdeveloped for CMS/ECAL, most of the require-ments could already be met except the light output.Therefore, an extended R&D program was initiatedto improve the luminescence combined with an op-eration at low temperatures such as T=-25C. Thedifferent steps to achieve the quality of PWO-II,as described later, have shown that radiation hard-ness becomes a very sensitive parameter, when theoperating temperature of the calorimeter is belowT=0C. Therefore, the estimated radiation envi-ronment is discussed here in more detail.

4.1.2 Hit Rates and AbsorbedEnergy Dose in SingleCrystals

The single crystal rates under the proposed runningconditions have been simulated using the Dual Par-ton Model (DPM) as generator for primary events.The maximum beam momentum of 15 GeV/c hasbeen used to evaluate the highest load for pp→ Xinteractions. A full tracking of the primary gen-erated events through the whole PANDA detectorusing GEANT3/4 was simulated. The results arescaled to the canonical interaction rate of 107 pri-mary events/s. Fig. 4.1 and Fig. 4.2 show the in-tegral rate for the forward end cap and the barrelpart for an energy threshold E>3 MeV.


Figure 4.1: Integrated single crystal rate for the barrelsection for an energy threshold of E>3 MeV using DPMat 15 GeV/c incident beam momentum.

Figure 4.2: Integrated single crystal rate for the for-ward endcap for an energy threshold of E>3 MeV usingDPM at 15 GeV/c incident beam momentum.

The energy spectrum of a single crystal dependsstrongly on the polar angle and is represented inFig. 4.3 for four different polar angles (5, 25,90 and 135). The summation of the givenenergy distributions yields total absorbed energyrates of 27 mGy/h, 1.5 mGy/h, 0.16 mGy/h and0.03 mGy/h, respectively, for each angle. The en-ergy absorption is homogeneous in lateral directionof the crystals, but in radial depth the situation

changes. The different interaction mechanisms ofthe electromagnetic and hadronic probes with thescintillator material lead to a strong radial depen-dence of the energy absorption as shown in Fig 4.4under similar conditions. The strong variation canbring the stated average numbers up to peak values,which can be higher by factors of 2-3 within the firstfew cm of the crystals. The centers of the electro-magnetic shower of photons up to 10 GeV incidentenergy will be concentrated within the first third ofthe crystal length, which has to be considered forthe optimization of a homogeneous collection of thescintillation light.

Figure 4.3: Single crystal energy differential rate spec-trum for polar angles of 5 (black), 25 (blue), 90 (red)and 135 (green) using DPM at 15 GeV/c incident beammomentum.

Figure 4.4: Relative energy dose normalized to 1 asfunction of the radial depth in a single crystal usingDPM at 15 GeV/c incident beam momentum.


Summarizing, the individual scintillation crystalshave to be capable to handle average hit rates abovea threshold of 3 MeV of 100 kHz in the barrel EMCand increasing up to 500 kHz in the most forwardparts of the forward endcap EMC. However, thesevalues might be even exceeded due to fluctuationsin the time structure of the beam. Consideringthe complete cocktail of impinging probes, absorbeddose rates up to 20-30 mGy/h have to be expectedat most forward regions. The values in most partsof the barrel EMC are two orders of magnitudelower. Therefore, the absolute values are well belowthe environment expected for LHC experiments.Due to the significantly slower recovery processesat T=-25C, as outlined in chapter 3.4.1.1, radi-ation induced changes of the optical transmittancewill reach a final level asymptotically on a time scaleof a typical run period. However, these limits willonly be reached at the most forward angles of theforward endcap EMC.

4.2 Lead Tungstate PWO

4.2.1 General Aspects

Lead tungstate, PbWO4 (PWO), crystals meet ingeneral the requirements to represent an extremelyfast, compact, and radiation hard scintillator ma-terial. However, a significant improvement of thelight output was mandatory for the applicationin PANDA. Table 4.1 considers already the finallyachieved quality described as PWO-II.

The specifications, which were developed andobtained for the application in experiments atLHC at CERN such as the ElectromagneticCALorimeter (ECAL) of CMS [18] and thePHOton Spectrometer (PHOS) of ALICE [19]served as a starting point for the further optimiza-tion of lead tungstate crystals [20]. Radiation hard-ness, fast scintillation kinetics, large-scale produc-tion and full size crystals of 23 cm length (28X0)with a light output of 9–11 phe/MeV (measuredwith a bi-alkali photocathode at room temperature)became standard.

Looking backwards, in the R&D phase, performedin 1998–2002, the light output was improved by ad-ditional doping [21, 22, 23]. Co-doping with molyb-denum (Mo) and lanthanum (La) at concentrations<100 ppm delivered optimum results. However,an additional strong slow scintillation component(τg = 1 − 4µs) appeared limiting applications athigh count rates [24]. Two additional concepts werepursued:

1. increase of the structural perfection of the crys-tal, and

2. activation of the crystal with luminescent im-purity centers. These have a large cross sectionto capture electrons from the conduction band,combined with a sufficiently short delay of ra-diative recombination.

Beside crystal activation with impurity centers, au-thors of Ref. [25] investigated the possibility to re-distribute the electronic density of states near thebottom of the conduction band by the change of lig-ands contained in the crystal, which unfortunatelyintroduced slow components in the scintillation ki-netics. An obvious possibility to reduce the con-centration of point structure defects is doping withyttrium (Y), lanthanum (La) or lutetium (Lu) ions,which suppress oxygen and cation vacancies in thecrystal matrix. An activator concentration at alevel of 100 ppm is needed. This concept has beenfollowed up successfully by the CMS collaboration,and led to mass production technologies of radiationhard crystals at the Bogoroditsk Technical Chemi-cal Plant [20]. A further increase of the scintillationyield by 30–50 % can be achieved by a reductionof the La-concentration to ∼40 ppm or less, whichcan compensate vacancies only if point defects arefurther suppressed. That was realized by an im-proved control of the stoichiometric composition ofthe melt.

4.2.2 Basic Properties of PbWO4

and the ScintillationMechanism

The world’s largest PWO crystal producer Bo-goroditsk Technical Chemical Plant (BTCP) fromRussia produces lead tungstate crystals with highyield by the Czochralski method, using standard”Crystal 3M” or ”Lazurit” equipment. Crystalsare grown from raw material with a purity levelclose to 6N specification in Pt crucibles. In or-der to grow high quality crystals an additional pre-crystallization is required. Ingots of 250 mm lengthand slightly elliptical cross-sections with a large axisbetween 36 and up to 45 mm in diameter becameavailable. Crystals can be pulled with identicalquality in the directions close to the a or c axis.Nevertheless, the production along the direction ofthe a axis appears to be the best for producing longsamples. The process of pulling is carried out inan isolated chamber filled with nitrogen. Pullingrate and rotation speed can be varied over a wide


range adapted to the required ingot diameter. Sev-eral crystal ingots to produce modules for the for-ward endcap EMC are shown in Fig. 4.5.

Figure 4.5: Selected PWO-II ingots, which have beenused to machine crystals in PANDA geometry.

Lead tungstate crystals occur in nature as tetrag-onal stolzite [26], scheelite type or monoclinicraspite [27]. The structure of the synthetic leadtungstate crystal is determined by X-ray diffrac-tion as scheelite-type of tetragonal symmetry witha space group of I41/a and unit-cell parameters a =b = 5.456(2), c = 12.020(2) A. The density of thesynthetic crystal can vary by a few percent depend-ing on the technology.

The electronic structures of the conducting and va-lence band for pure and defective lead tungstatecrystals are described in detail in [28]. According tothe electron band structure, luminescence appearsin PWO crystals because of charge-transfer transi-tions in a regular (WO4)2− anionic molecular com-plex. Since the (WO4)2− complex has a Td pointsymmetry of the crystalline field, the final config-urations of the energy terms are found to be 3T1,3T2 and 1T1, 1T2 with the 1A1 ground state. Thespin and parity forbidden radiative transition 1T1,1T2 →1A1 in the anionic complex is responsible forthe intrinsic blue luminescence of the lead tungstatecrystal. The luminescence has a large Stokes shiftof 0.44 eV and therefore is characterized by a strongtemperature quenching.

Crystalline material of lead tungstate is transparentin the visible spectral range and fully colorless. Atypical transmission spectrum measured along thecrystallographic axis a of a 200 mm long sample andthe radio-luminescence spectrum measured at roomtemperature are compared in Fig. 4.6. There is anoverlap of the distribution of luminescence with thecut-off of the transmission spectrum. Therefore, op-

timizing the light yield relies significantly on thecrystal transmittance in the near UV region.

Figure 4.6: The optical transmission and radio-luminescence spectra at room temperature measured fora 200 mm long PWO crystal.

Due to strong temperature quenching of the lumi-nescence in PbWO4, the light yield shows a strongtemperature dependence. As a consequence, thescintillation kinetics is fast at room temperatureand is dominated by an exponential decay constantof τ=6.5ns. The temperature dependence of thescintillation light yield in the range +20C to -25Cwas found to be close to linear with a gradient of-2 %/C. The typical kinetics of the scintillationprocesses in PWO is shown in Fig. 4.7.

Figure 4.7: The scintillation kinetics of PWO mea-sured at room temperature.

Both scintillation and optical properties of the cur-rently produced lead tungstate crystals are ratheruniform along the axis of crystal growth. The vari-ation with respect to the wavelength of the opti-


cal transmission value of T=50 % does not exceed∆λ = 3 nm for a 200 mm long crystal.

4.2.3 The Improved Properties ofPWO-II

Up to date the CMS Collaboration has the largestcomprehensive experience with the exploration andproduction of lead tungstate [29, 30]. RecentlyBTCP has completed the production of 77,000 scin-tillation crystals for the CMS/ECAL using theCzochralski method. A smaller percentage was de-livered by the Shanghai Institute of Ceramics (SIC-CAS) using a modified Bridgeman method [31]. Asan outcome of seven years of mass production thefollowing quality parameters of the production tech-nology were obtained [32], summarized in Table 4.2.

One can state that both crystal producers reached ahigh quality of the crystals. Moreover, the distribu-tions of crystal parameters are significantly superiorto the limits set by the CMS specifications. As aconsequence, only a small fraction at the level of1.5 % was rejected by the certification of scintilla-tion elements repeated by the customer at CERN.

Experimental applications for photon detection inthe MeV range below 200 MeV are predominantlylimited by the photon statistics [33, 34]. Obviously,further improvement of the crystal light yield be-came mandatory. Detailed studies of the differentdopants in the crystal did not lead to a spectacu-lar increase of the light yield [21], as illustrated inFig. 4.8.

PWO is very sensitive to the conditions at the syn-thesis. During the crystal growth by the Czochral-ski method from stoichiometric raw material a dom-inant leakage of lead from the melt takes place lead-ing to the creation of cation vacancies Vc on thelead site in the host. This holds also for crystalsobtained by the modified Bridgeman method [31].The systematic lead deficiency introduces a super-structure created by cation vacancies as identifiedby combined X-ray and neutron diffraction mea-surements [35]. In fact, cation vacancies Vc arecompensated by oxygen vacancies VO or Frenkel-type defects and their ordering is compensated bythe distortion of the tungstate anionic polyhedra.During the development, point structure defects,such as electron capturing centers were identifiedby Thermo Stimulated Luminescence (TSL) andElectron Paramagnetic Resonance (EPR) measure-ments [36, 37, 38, 39]. As a result, the shallowestcenter occurs in all crystals as an intrinsic defect:an additional electron auto-localized at an anionic

(WO4)2− complex via a Jahn-Teller distortion cre-ating a (WO4)3− polaronic center [40, 41, 42]. Thisshallow trap is emptied near a temperature of 50 Kwith an activation energy of 0.05 eV. Part of the re-leased electrons recombine radiatively or are caughtby deeper traps. The second one is a Pb1+-V0

center, which is stable in the crystal up to 175 K[43, 44]. It is not excluded that instead of Pb another ion may create such a center near an anionvacancy. However, it counts that the electron istrapped by a hetero-valent cation in the vicinity ofan oxygen vacancy. That can be proved by a rel-atively large deviation of the g-factor of all suchmagnetically non-equivalent species from ge. Thiscenter is not clearly correlated with the detectedTSL band, however it is photo-ionized by IR withthe threshold 0.9 eV. The third one is a (WO4)3−

electron center, which is created on a base of aregular tungstate anionic complex disturbed by anearby rare earth (RE) trivalent impurity ion likeLa, Lu or Y [41]. It decays near 97 K and has anactivation energy 0.2 eV. Careful partial anneal-ing experiments showed that the rate of decreaseof the EPR intensity of (WO4)3−-La coincides withthe TSL emission in this temperature region [43].In addition, the doping of the crystal with stabletrivalent RE ions such as La, Lu, Gd and Y at con-centrations of some tens of ppm redistributes elec-tron trapping centers and reducing the number ofdeep ones [43, 45, 46]. Such ions localized at Pbsites introduce in the crystal an additional positiveuncompensated charge and thus will compete withthe creation of vacancies. In addition, the dopantsintroduced in the crystal above the stoichiometricratio occupy empty lead sites in the lattice and sup-press the superstructure fraction with its distortedtetrahedra in the crystal.

Another trap center with an activation energy of0.13 eV appears in the crystal with RE doping.This RE-(WO4)4− center is not paramagnetic. Thishypothesis is in agreement with the fact that RE-doped PWO crystals show higher TSL intensity inthe vicinity of 100 K temperature. An electronrelease from the 0.13 eV traps causes simultane-ous production of the lowest electron centers by re-trapping as well as by contributing to an increaseof TSL intensity in that region via the creation ofRE-(WO4)3−.

The deepest electron trap centers are double va-cancy centers having an activation energy in therange of 0.4-0.5 eV and Frenkel-type defect with anactivation energy 0.7 eV, which occurs in the crystalat the shift of the oxygen ion into the inter-site posi-tion. Double vacancies cause induced absorption in


number parameter mean value standardof samples deviation

40607 length/mm 229.95 0.0261267 LY/phe·MeV−1 10.2 1.261267 LY non-uniformity / %X0 -0.13 0.0261267 opt. transm. at 420nm/ % 69.5 2.261267 rad. ind. absorption at 420nm/m−1 1.04 0.4

Table 4.2: Distribution of the crystal properties of accepted PWO crystals for the CMS/ECAL project.

Figure 4.8: Integral dependence of the radioluminescence of lead tungstate crystals doped with different ions asa function of the activator concentration. The measurements have been performed at an absolute temperature ofT=300 K.

the near IR spectral region in the crystal. Frenkel-type defects cause an optical absorption band ata wavelength of 360 nm, which is converted underionizing radiation in the absorption band with amaximum near 410 nm.

We attribute a molybdenum (Mo) based centerto the characteristic luminescence centers in leadtungstate. It is related to a (MoO4)2−-anion com-plex impurity, which is a stable electron trap cen-ter. Although the raw material is purified before

crystal growth, especially of Mo, the molybdenumion is chemically very close to the tungsten ion andit is rather hard to separate both at the produc-tion quality level of the raw material. The effectivemolybdenum impurity in PWO-II is typically at thelevel of <1 ppm. The properties of the (MoO4)3−-center and its impact on the scintillation parametersof PWO are described in detail in [24].

Following the disposition of electronic centers nearthe bottom of the conducting band it appears to be


possible to describe the influence of each kind of de-fects on the crystal scintillation properties even atdifferent temperatures. On one hand shallow pola-ronic and RE-distorted regular centers contribute tothe scintillation via relatively fast electron exchangewith the conducting zone. A reduction occurs dueto loss of electrons into the deep traps. The increaseof the concentration of (WO4)3−-RE centers in thecrystal leads to a drop of the light yield. There-fore, an increase of the light yield is in conflict withoptimization of radiation hardness by a higher con-centration of activating ions such as La. To fulfillboth, perfection of the crystal is the only way out.

Decreasing the operating temperature of thecalorimeter down to T=-25C slows down the timeconstants of the spontaneous recovery of the elec-tron centers by thermal activation. Table 4.3 showsthe parameters of selected electron capturing cen-ters in PWO crystals [38] obtained by the TSLmethod, where ET , S, γ are the activation energy,the frequency factor and the order of the kinet-ics, respectively. The calculated spontaneous re-covery constants for two temperatures are shownin addition. When the temperature is decreasing toT=-25C the deepest electron center becomes meta-stable and increases effectively the level of dynamicsaturation of the radiation induced optical absorp-tion in the crystal. This effect has to be consideredand its consequences have to be investigated underthe experimental conditions for PANDA.

Structural perfection of the crystal has been reachedby an improved control of the stoichiometric compo-sition of the melt. Since the full-size crystal sampleshave a length of 200 mm, crystal activation was pur-sued at different stages of the raw material prepa-ration by Y and La ions with distribution coeffi-cients of 0.95 and 1.47, respectively. EPR stud-ies have confirmed that only single RE-(WO4)3−

(RE=La,Y) centers are formed, if the concentra-tion of the dopants remains below 20 ppm for eachion species. These crystals contain finally at leasttwo times less Frenkel type defects leading to an in-crease of light by 80 % compared to CMS quality.The kinetics of the enhanced material remains fastand allows to integrate 97 % of the light within a100 ns wide time gate at room temperature.

4.3 Radiation InducedAbsorption in PWO-IICrystals

The stability of the parameters of the scintillationmaterial in a radiation environment is one of themost important properties required to build a pre-cise electromagnetic calorimeter. As common formost of the scintillators, the scintillation mecha-nism in PWO itself is not damaged when grownunder optimized conditions [29]. Irradiation affectsonly the optical transmission due to the creationof color centers with absorption bands in the visi-ble spectral region. The induced absorption evolveswith time, depending on the dose rate and the ori-gin of the participating centers and their mecha-nism of annihilation and relaxation. The absorptionis quantified by the absorption coefficient k givenin m−1 assuming a pure exponential change of thetransmission loss. Fig. 4.9 shows typical spectra ofthe induced absorption loss measured for PWO-IIsamples at room temperature after a total absorbeddose of 0.2 kGy using a 60Co source expressed bythe difference of the coefficients before and after ir-radiation at given wavelengths.

The obtained values stay below the considered qual-ity limit ∆k=1m−1 in the relevant spectral rangeof the luminescence. Nevertheless, enhanced ab-sorption is visible near the maximum of scintilla-tion emission at 420 nm or at a broad structure nearλ=550 nm.

In contrast to previous applications, the calorimeterwill be operated at T=-25C. Therefore, a detailedresearch program had to be started to investigatepossibly new and unexpected effects due to irradi-ation at crystal temperatures below 0C. Besidesthe facilities performing quality control during thephases of development and pre-production at themanufacturer locations BTCP or INP (Minsk, Be-larus) two additional and complementary facilitiesat IHEP (Protvino, Russia) and JLU (Giessen, Ger-many) have been adapted for that purpose.

4.3.1 The Irradiation Facility atIHEP, Protvino

An irradiation facility, dedicated to the investiga-tion of the radiation damage of cooled and temper-ature stabilized scintillation crystals, was designedand installed at IHEP (Protvino, Russia). A 137Csradioactive source, emitting γ-rays of 662 keV en-ergy with an activity of 5·1012Bq, is used for the ir-radiation measurements. The dose rate can be con-


ET S γ τ rec at T=+25C τ rec at T=-25oCmeV s s0.05 4.00E+03 1.1 1.80E-03 2.70E-030.135 1.0E+08 1.6 2.80E-06 8.36E-060.19 2.90E+08 1.35 6.40E-06 2.98E-050.23 6.7E+09 2.2 2.2E-06 1.42E-050.27 9.0E+10 2.2 7.3E-07 6.51E-060.40 2.0E+10 1.1 3.1E-04 7.93E-030.50 9.0E+11 1.4 5.0E-04 2.87E-020.49 2.0E+09 1,0 9.0E-02 4.770.58 1.1E+07 1.0 4.80E+02 5.28E+040.7 5.5E+7 1.0 1.00E+04 2.91E+06

Table 4.3: Selected parameters of the electron capturing centers in PWO crystals.

Figure 4.9: The measured change of the induced absorption coefficient of various PWO-II samples after irradi-ation with a dose of 0.2 kGy (60Co) over the relevant range of wavelength.

tinuously adjusted by changing the distance to theexposed crystals. The dose rate in air is measuredclose to the crystals with a commercial dosimeter(DKS-AT1123) with a typical accuracy of <20 %[47]. Fig. 4.10 shows the schematic layout of thefacility and Fig. 4.11 the major components of the

installed setup. All major components are remotecontrolled and the relevant parameters are recordedfor the off-line analysis. The up to five full size crys-tals are mounted in an insulated container, whichcan be cooled and temperature stabilized by a cryo-thermostat (LAUDA RC6CP) with an absolute ac-


curacy of 0.1C even during a test cycle of severaldays. The crystals are attached via air light-guidesto photomultiplier tubes (PMT). A gas flow of drynitrogen helps to equalize the temperature and toavoid condensation of moisture. Several thermo-sensors measure and control the temperature. Anychange of the optical transmittance of the crystalsdue to irradiation as well as possible gain instabil-ities of the PMTs are controlled by a LED-basedmonitoring system distributed by optical quartzfibers. Two spectral regions near the wavelengthsof 450 nm and 640 nm, respectively, are inspectedusing blue and red LEDs. The highly stabilizedmonitor system is similar to the one described inreference [48, 49].

Figure 4.10: Schematic layout of the irradiation facil-ity at IHEP, Protvino.

Figure 4.11: Photograph of the major components ofthe irradiation setup at IHEP, Protvino.

The DC current of the PMT is directly digitized viaa 10 kΩ resistor. That signal, which is proportionalto the product of light yield and light collection ef-ficiency in the crystal, provides general information

on any change of the light output due to irradiationor temperature change. Complementary, the sig-nal triggered by the LED pulses is monitoring anyvariation of the light transmittance of the crystal.

The facility will be used during final production toinvestigate selected crystal samples with respect tothe radiation damage of cooled crystals, comple-mentary to the standard procedures at the man-ufacturer location performed at room temperature.The measurements will confirm the linear relationbetween the radiation hardness at both tempera-ture regions. Finally, tests will be performed for allcrystal geometries in order to determine the corre-lation between the loss in transmission and the re-duction of the integrated scintillation light collectedafter multiple reflections inside the crystal.

4.3.2 The Irradiation Facility at theJustus-Liebig-UniversityGiessen

The irradiation facility at Giessen at the so calledStrahlenzentrum is based on a set of five 60Cosources, which can be used in any combination.The maximum dose rate amounts to∼30 Gy/h mea-sured at a distance of 1m. Complementary to thepossibilities at Protvino an experimental setup in-cluding a double beam photospectrometer has beenadapted to irradiate cooled crystals and measureimmediately after irradiation or hours and dayslater any changes of the spectral distribution of theoptical transmission. That arrangement allows todetermine at various temperatures the induced ab-sorption as a function of integral dose as well as therecovery as a function of time. Present experimentswere performed with a typical dose rate of 12 Gy/hat a distance of 1m (see Fig. 4.12).

For the future quality control, the crystal samplescan be either exposed in a radiation protected lab-oratory room or placed into a sufficiently large con-tainer, which can be loaded outside and moved infront of the array of sources. The equipment in-cludes a Varian photospectrometer to measure af-terwards the radiation induced absorption via thelongitudinal optical transmission. All measure-ments are meant to be performed at room temper-ature and will be the final acceptance test for asignificant percentage of all the crystals as a crosscheck of the measurements performed at the man-ufacturer.


Figure 4.12: The experimental setup arranged on topof a photon spectrometer for irradiation studies at JLU,Giessen (left). The right part of the figure shows themovable crystal container with the open region for themeasurement of the transmission.

4.3.3 The Irradiation Facilities atINP (Minsk) and BTCP

For the mass production for the CMS detector, anirradiation facility was installed at BTCP with tech-nical assistance of INP (Minsk). A 60Co source ofround geometry is placed in a well for radiation pro-tection, see Fig. 4.13. Up to four crystals, placedin cylindrical light-tight containers, are irradiatedlaterally from all sides at a dose rate of 1200 Gy/h.In addition, an inner container equipped with opti-cal windows allows to determine simultaneously thelongitudinal induced absorption at three character-istic wavelengths in an optical spectrometer withfour independent channels. The procedure of mea-surement is illustrated in Fig. 4.14.

The change of longitudinal transmission after irra-diation is inspected at three wavelengths: 650 nm(RED), 570 nm (GREEN) and 420 nm (BLUE) [50].According to these three colors the spectrometricsetup is called RGB System. Three light emittingdiodes emit alternately short (tens of nanoseconds),stable light pulses. The light emitted from eachLED is mixed by four diffusion cavities and illumi-nates the samples in longitudinal direction as illus-trated in Fig. 4.15. After traversing the crystal andthe diffusion of the light on a receiving reflector, itimpinges on PIN-photodiodes. The amplified sig-nals are shaped and sent to a PC-based ADC card.

Some fraction of the light from the LEDs is directedto a reference channel based on a thermo-stabilizedPIN-diode to ensure the long-term stability at thelevel of 10−4. This concept of the automated con-

Figure 4.13: The irradiation facility operating atBTCP in Russia.

Figure 4.14: Schematic layout of the spectrometricsetup at BTCP.

trol of the light pulse intensity is similar to one de-scribed in details in [51]. The measurement cycle offour PWO crystals includes 2.5 minutes of irradia-tion and optical transmission measurements beforeand after irradiation (taking about 2 minutes each).Due to short irradiation time and short time of thecrystal transmission measurement after irradiation,unwanted fast crystal recovery components can alsobe measured along with induced optical absorption.


Figure 4.15: Schematic layout of a single spectromet-ric channel of the setup at BTCP.

Per working day at least 100 crystals can be char-acterized with respect to their radiation hardness.This equipment was adopted to measure PANDAbarrel type crystals of 200 mm length. The radia-tion hardness of the crystals of the pre-productionlot has been characterized already with the adaptedequipment.

The irradiation facility at INP Minsk comprises awell shaped 60Co source with a constant dose rateof 2300 Gy/h. Variations of the crystal transmissionare measured 60 min after irradiation by a VarianCary1E photospectrometer in the wavelength rangeof 300-900 nm. The control of other scintillationparameters can also be carried out at INP.

BTCP and INP irradiation facilities are cross cali-brated two to three times per year on the occasionof change of the crystal dimensions. It is agreedthat a set of reference crystals including all typesof PANDA scintillation elements will be producedbefore crystal mass production. It will be used forexamination of the equipment before each measure-ment session and to verify the cross calibration ofthe irradiation facilities within the project.

4.3.4 Irradiation Studies withNeutrons, Protons and HighEnergy γ-rays

It is foreseen to perform selective irradiation studieswith hadrons and electromagnetic probes. Withinthe collaboration there are several accessible facili-ties to perform such studies. The following optionsare available and have been already used for thecharacterization presented in the previous section.

• The irradiation facility at IHEP, Protvino pro-vides a Pu-α-Be source with a fluence of 3·104

neutrons/(cm2·s). The emitted neutrons havea mean kinetic energy of En=3.5 MeV. How-ever, there is an accompanying background ofγ-rays with a mean energy of Eγ=300 keV cor-responding to a dose of 7 mGy/h at the locationof irradiation of samples.

• The cyclotron at the KVI at Groningen, (TheNetherlands) provides a dedicated experimen-tal area for irradiation studies with protons upto an energy of 150 MeV. The facility allowsto irradiate locally as well as homogeneouslyover a typical surface of 15 × 15 cm2 with anaccurate control over the overall dose rate andintegral dose.

• Besides the opportunity to measure responsefunctions to high-energy photons over therange up to 3.5 GeV, both tagging facilities atthe electron accelerators MAMI (Mainz, Ger-many) and ELSA (Bonn, Germany) allow tostudy the radiation damage in a controlled way.The absolute flux of the energy marked pho-tons is known with high accuracy and can beused to test pure crystals as well as assembleddetectors or arrays. High-energetic electronscan be simultaneously used with well knownenergies and rates directly behind the focalplane of the magnetic dipole

4.4 The PWO-II Crystals forPANDA

The achievements in the understanding of the scin-tillation mechanism, the defect structures and theimplications for the production technology haveprovided the quality of lead tungstate fully ade-quate for the PANDA application. PWO-II, whenoperated at T=-25C becomes a sufficiently brightscintillator, in particular when nearly 50 % ofthe crystal endface are covered and readout with


LAAPDs of high quantum efficiency. The specifictemperature keeps the decay time fast to cope withhigh count rates. However, the processes recover-ing the irradiation damage become extremely slowat low temperatures and cause an asymptotic re-duction of the light yield due to the changing op-tical transparency. High quality crystals with aminimized concentration of defects limit the loss toa value below 30 %. Therefore, for the operatingscenario of the target spectrometer EMC a net in-crease of the light yield of a factor 3 compared toroom temperature can be considered even for themost forward located modules in the forward end-cap EMC. The next chapters are describing in detailthe overall parameters of the scintillator materialfor PANDA.

4.4.1 Light Yield and DecayKinetics

As a result of the significant improvement of thetechnology of crystal production, full size crystalswith 200 mm length deliver today a light yield of17-20 phe/MeV at 18C measured with a photo-multiplier with bi-alkali photocathode (quantum ef-ficiency ∼20 %. Documenting the breakthroughyears ago, Fig. 4.16 shows for only 150 mm longPWO-II samples the light yield in comparison toCMS Endcap type crystals of 230 mm length. Thelight yield is shown as a function of the optical ab-sorption at 360 nm wavelength, which is sensitiveto the concentration of Frenkel-type defects in thecrystal, in order to take in to account the differentcrystal dimensions.

Reducing the thermal quenching of the lumines-cence process by cooling the crystals leads to theexpected increase of light yield. Obviously, the lowconcentration of deep traps in PWO-II does notdistort the scintillation kinetics at least down toa temperature of T=-25C. Fig. 4.17 a) and b)show typical pulse height spectra, which were mea-sured for a full size crystal (20×20×200mm3) attwo operating temperatures of T=±25C and twodifferent integration gates of 75 ns and 4000 ns, re-spectively, using a photomultiplier as sensor (Hama-matsu R2059). The overall gain factor of the lightyield at T=-25C compared to T=+25C amountsto more than 4 for both time gates.

Compared to the standard material, PWO-II crys-tals follow a different temperature dependence ofthe scintillation yield, as shown in Fig. 4.17 in therelevant temperature region between T=+25C andT=-25C, respectively. The average temperaturegradient of the light yield amounts to ∼ −3 %/C,

which is 50 % larger than for standard PWO. Theobviously steeper increase of the light yield withlower temperature can be addressed to a lower con-centration of capturing centers.

Fig. 4.19 shows for different temperatures the meanvalue of the scintillation yield per MeV depositedenergy including standard deviations obtained forlow energy γ-rays. The signals obtained with photo-multiplier tubes equipped with bi-alkali photocath-odes (Hamamatsu R2059) were integrated over agate width between 75 ns up to 1µs, respectively.The figure confirms a mean gain factor up to 4 be-tween the two extreme temperatures and a suffi-ciently fast decay time, which allows to collect 95 %of the scintillation light within a time window of300− 400 ns.

Mass production requires sensitive criteria for thecertification of crystals. As shown in Fig. 4.20,a strong correlation has been verified with a co-efficient above 4 for the light yield observed atT=+25C and the corresponding value at T=-25C. Experimental data are shown in part a) andb) for ten PWO-II crystals after integrating the re-sponse within 100 ns and 1000 ns, respectively. Con-sequently, the linear correlation allows quality con-trol to be performed in the more convenient envi-ronment at room temperature.

The overall light collection depends on the opticalquality, the geometrical shape, surface treatmentand reflector material. The distribution of the emis-sion of the scintillation light and the variation of thelight path cause a position dependence of the lightyield collected at the photo sensor located at therear end of the crystal. These effects cause a typi-cal non-linearity, which can have an impact on thefinal energy resolution, explicitly the constant term.The effect has to be corrected in particular, whenthe maximum of energy deposition within the crys-tal is spread over a large region, which is not thecase as illustrated in Fig. 4.4. Nevertheless, investi-gations have been started to optimize the additionalsurface treatment by partly de-polishing or selectionof various reflector materials. Fig. 4.21 illustratesthe change of light output for different geometriesand reflector materials when a collimated source ismoved relative to the sensor position along the crys-tal axis.

4.4.2 Radiation Hardness ofPWO-II

The radiation hardness and the effective loss of lightyield as a functions of dose rate and accumulated


Figure 4.16: Light yield of full size polished scintillation crystals (28×28×230mm3) used for CMS endcapconstruction (right distribution) and of PANDA prototypes of slightly smaller size (20×20×150mm3) shown as afunction of the absorbance to compensate for different geometries.

dose are well known from the studies for CMS butinvestigated at room temperature. The deteriora-tion of the optical transmission in the experiment iscaused by the interplay of damaging and recoveringmechanisms. The latter are fast at room tempera-ture and keep the loss of light yield moderate. Firstexploratory experiments at IHEP (Protvino) haveshown a completely different scenario if the irra-diated crystals are cooled down to T=-25C andare kept cool even after the irradiation has beenstopped. Fig. 4.22 and Fig. 4.23 illustrate for twocrystal samples the typical behavior. At a start-ing dose rate of 20 mGy/h the scintillation responsedrops continuously and does not even reach satura-tion after 1000 hours. In a similar way, almost norecovery can be noticed after irradiation for the twosamples kept at T=-25C.

Therefore, more specific studies became necessaryin order to characterize the nature and propertiesof defects with slow decay times. The experimentalinstallation at Giessen allows to measure in a wave-

length specific manner the deterioration of the op-tical transmission expressed as the change of the in-duced absorption coefficient for crystals, while theyare kept cool after irradiation for a longer time.Investigations of full size crystals were performedat temperatures of T=-3.5C, T=-10.0C and T=-22.5C. Fig. 4.24 shows the representative resultobtained for the absolute loss in transmission atT=-3.5C right after irradiation as a function ofphoton energy. The contributions of two specificabsorption bands are marked for further analysis.The shape of the spectrum is similar to the resultsat the two lower temperatures. Fig. 4.25 repeatsthe above data obtained at T=-3.5C but re-scaledas induced absorption as a function of wavelength.The previous measurement at room temperature(see Fig. 4.9) shows in a similar manner the in-duced absorption primarily in the same wavelengthrange between 400-600 nm and two broad absorp-tion bands located at wavelengths of 528 nm and420 nm, corresponding to photon energies of 2.35 eV


Figure 4.17: Pulse height spectra measured for aPWO-II crystal (20×20×200mm3) for γ-rays of a 60Co-source (Eγ ∼1.2MeV). The scintillation light has beenconverted with a photomultiplier tube (HamamatsuR2059) at two temperatures and integrated over timewindows of 75 ns (part a) or 4000 ns (part b), respec-tively.

Figure 4.18: Temperature dependence of the yield ofscintillation light of PWO-II in the temperature rangebetween T=+25C and T=-25C, respectively.

and 2.95 eV, respectively.

Performing the irradiations of cooled crystals withsignificantly higher dose rates a saturation of the

damage can be observed at an integral dose of 30-60Gy. This value does not depend on the temperatureand might only reflect the individual concentrationof defects in the crystal. In the final PANDA ex-periment such an integral dose will be accumulatedonly in the most forward part of the forward end-cap EMC at maximum beam intensity after 3000hours of operation and might be never reached inthe barrel region over a typical experimental cyclelasting 6 months.

The recovery of the optical transmission has beenstudied in a time range up to 5 days and at differ-ent temperatures. From the measurements similarto Fig. 4.24 but performed at different times afterthe end of irradiation an estimate of the lifetimesof the color centers can be deduced. Fig. 4.26 andFig. 4.27 illustrate the kinetics of the two selectedabsorption regions. Besides a fast component therecovery times are in the order of hundreds of hoursin both cases. At all investigated temperatures be-low 0C the relaxation of the color centers remainsslow and does not compensate even partially thedamage in parallel.

Detailed studies of the impact of Mo impurityon the scintillation performance of the crystal al-lowed also to clarify its contribution to the radi-ation damage. It is well known that a contam-ination with MoO4-complexes gives rise to greenluminescence and extremely slow decay times inPWO crystals [24]. Therefore, the present gener-ation of PWO-II crystals maintains the impurityon the level of several ppm. Our recent studieshave identified the presence of (MoO4)3− centersexploiting the EPR technique. Spectral distribu-tions of the induced absorption of various samplesmeasured at room temperature gave first indica-tions based on the broad absorption band peakednear 535 nm as already shown in Fig. 4.9. System-atic studies of paramagnetic resonance and thermo-luminescence of PbWO4/PbMoO4 mixed crystals[36] have investigated the properties of (MoO4)3−

centers with respect to the molybdenum concentra-tion. The temperature of maximum thermal de-cay depends strongly on the Mo concentration andreaches values near 250 K at concentrations below1 %. TL spectra of mixed crystals after 50 keV X-ray excitation show a pronounced glow peak slightlyabove 250 K for mixed crystals with low Mo con-centration. Measurements of the concentration of(MoO4)3− and (WO4)3− paramagnetic centers inCaWO4:Pb after X-ray irradiation and subsequentstepwise heating allow deducing the radiation in-duced optical absorption due to (MoO4)3− centers[37]. The spectral distribution indicates a broad


Figure 4.19: Dependence of the average light yield on the temperature and the integration gate for 10 PWO-IIcrystals (20×20×200mm3) produced at BTCP. The variations are given as error bars.

distribution peaked at a photon energy of 2.35 eV(FWHM=0.8 eV) corresponding to a wavelength of528 nm. At T=-25C, which is below the activa-tion temperature of (MoO4)3−, one observes as aconsequence of the thermo-activation a slow decaytime. Fig. 4.28 shows the EPR signal identifyingthe (MoO4)3− center in the PWO-II crystal, whichwas stored directly after the irradiation in liquidnitrogen. The drop of the signal intensity after arecovery of 96 hours was estimated to 25 %. Sim-ilar measurements at different temperatures are inpreparation.

Therefore, even a small concentration of Mo impu-rities can contribute to the reduced transmittancewithin that range of wavelengths. In order to ensuresufficient radiation hardness one has to put even astronger limit on the Mo contamination in the crys-tal, even at a level below 1 ppm. This limit of Moconcentration has been so far specified for the crys-tal mass production.

4.4.3 Radiation Hardness Requiredfor PANDA

The radiation damage reduces the output of thescintillation light and consequently the number ofphotoelectrons per unit of deposited energy. There-fore, the expectable light output has to be corre-lated with the quantified radiation resistivity. Thelatter is specified by the radiation induced absorp-tion coefficient ∆k at the wavelength of maximumscintillation emission at λ=420 nm, which can beeasily measured and cross-checked, if it can be de-termined at the relevant facilities at a temperatureof T=+20C. The correlation depends on the indi-vidual crystal shape, light collection and efficiencyof the photo sensor and has not been experimen-tally determined for the target spectrometer EMCup to date. We intend to follow the approach usedby CMS [52]. The information is obtained at smalldose rates, which are similar to the experimentalconditions of PANDA. The procedure determinesthe ratio of the changes of the scintillator and mon-


Figure 4.20: Correlation between the light yieldinduced by low energy γ-rays at temperatures ofT=-25C and T=+25C for 10 PWO-II crystals(20×20×200mm3). The photomultiplier response hasbeen integrated over a time gate of 100 ns (part a) and1000 ns (part b).

itoring response, respectively, and models the re-lationship by the expression S/S0=(R/R0)α, whereS/S0 and R/R0 are the relative variations of the re-sponse to the scintillation and monitoring signal, re-spectively. Since the geometrical shape of the crys-tals of CMS and PANDA are comparable, a typicalvalue of 1.5 can be assumed for the parameter α.

The measurement at the irradiation facility atIHEP (Protvino) estimates the change of opticaltransmission, deduced from the response to thetransmitted light of two LEDs of different color,and the reduction of the scintillation response viathe luminescence initiated by the absorbed γ-raysemitted from the 137Cs source. This measurementis not fully compatible with the response to electro-magnetic probes hitting the crystal from the frontface, but can give a first hint. Fig. 4.29 shows thedropping DC-current of the photomultiplier tube asa function of the irradiation time. The five inves-tigated PWO-II samples were characterized beforeby the value of the induced absorption at λ=420 nm(∆k(420 nm)) determined at T=+20C at BTCP

Figure 4.21: Variation of the light output measuredfor different crystal geometries and reflector materialswhen a collimated γ-source is moved relative to the sen-sor position along the crystal axis.

Figure 4.22: Relative change of the light yield of twoPWO-II samples with irradiation time for two differ-ent dose rates and two temperatures. The curves arenormalized to the initial light yield at T=+20C.

and afterwards annealed at T=200C for 2 hours.Table 4.4 summarizes the experimental data includ-ing the relative loss of scintillation response mea-sured at T=-20C.

Both specification parameters are correlated inFig. 4.30. In spite of the lack of sufficient statistics,a linear correlation can be proposed, which supportsthat the quality limits defined for control measure-ments at room temperature are sufficiently selectivefor the final operation at low temperatures. Thedata are consistent with the assumption of α=1.5


crystal ID ∆k/m−1 R/R0 estimated S/S0

T=+20C T=-20C T=-20Cb4 0.402 0.905 0.86b16 0.420 0.910 0.87b17 0.289 0.915 0.88b24 0.405 0.920 0.88b30 0.753 0.890 0.84

Table 4.4: Comparison of the induced absorption ∆k(420 nm) measured at BTCP, the value R/R0 at λ=450nmmeasured at IHEP (Protvino) and the estimated change of scintillation response.

Figure 4.23: Relative change of the monitoring signalprovided by the blue emitting LED illustrating the re-covery of two PWO-II crystals after γ-irradiation at twodifferent temperatures.

Figure 4.24: The absolute loss in transmission dueto the irradiation of a PWO-II crystal with a dose of30.3 Gy at a temperature of T=-3.5C shown as a func-tion of photon energy. The contributions of two absorp-tion bands at photon energies of 2.35 eV and 2.95 eV,respectively, are marked for detailed analysis.

as discussed above. Therefore, due to the slow re-

Figure 4.25: The induced absorption due to the ir-radiation of a PWO-II crystal with a dose of 30.3 Gyat a temperature of T=-3.5C given as a function ofwavelength.

Figure 4.26: The decay kinetics of an absorption re-gion near a photon energy of 2.35 eV (λ=528 nm) af-ter the irradiation of a PWO-II crystal with a dose of30.3 Gy at a temperature of T=-3.5C.

laxation of color centers in cooled crystals one hasto cope with a typical loss in scintillation responsebetween 20 and 30 % as an asymptotic value aftera deposited dose of 30-50 Gy for typical PWO-II


Figure 4.27: The decay kinetics of an absorption re-gion near a photon energy of 2.95 eV (λ=420 nm) af-ter the irradiation of a PWO-II crystal with a dose of30.3 Gy at a temperature of T=-3.5C.

Figure 4.28: EPR signal of the (MoO4)3− center ina PWO-II crystal measured at T=-25C immediatelyafter irradiation (upper part) and 96 hours later. Bothprobes were kept continuously at T=-25C.

crystals.

These results have been further confirmed by a di-rect measurement of the pulse height spectrum ofthe energy deposition of minimum ionizing cosmicmuons measured with a cooled crystal detector as-sembly. Fig. 4.31 compares the spectra measuredat T=+24C and T=-26C, before and after irra-diation at the lower temperature with an integraldose of 53Gy. In spite of a reduction of the lightoutput by 31 % the photon statistics remains a fac-tor 3 above the value at room temperature. In spiteof the significant asymptotic degradation of the re-sponse, the slow change can be very well monitoredwith high precision and stability.

Figure 4.29: Change of the scintillation responsedue to a continuous irradiation with low-energy γ-rays(137Cs) for 5 samples of the pre-production run de-duced from the DC-signal of the photomultiplier re-sponse. The measurement has been performed at IHEP,Protvino.

Figure 4.30: Correlation between the maximumchange of the scintillation response, as shown inFig. 4.29, and the induced absorption coefficient ∆k atλ=420 nm measured at BTCP at room temperature.

4.4.4 Pre-Production Run ofPWO-II Crystals

A subgroup of scintillator modules of the barrelEMC (40 crystals of each type: 8,9,10) has beenproduced by BTCP in December 2007/February2008. Fig. 4.32 shows a few optically polished bar-rel type crystals produced at BTCP. The remain-ing crystals to complete one slice of the barrel EMChave been delivered in May 2008 and are presentlyinspected for a detailed quality control and to ad-just and calibrate the ACCOS machine and test in-stallations at CERN and JLU, respectively. The


Figure 4.31: Change of the scintillation response ex-pressed by the measured energy deposition of cosmicmuons in an assembled PWO-II detector. The figureshows the measurement at T=+24C and T=-26C, re-spectively. The spectra at the lower temperature are ob-tained before and after irradiation with an integral doseof 53Gy. The measurement was carried out at Giessen.

pre-production run was aiming for:

1. check of the reproducibility of the productiontechnology

2. confirmation of the correspondence of the crys-tal quality to the specification

3. production of several reference crystals to beused for adaptation of the certification tools toPANDA type crystals

4. estimation of the technological yields of thecrystals at different modes of operation of theproduction.

The performed tests of the first 120 units have con-firmed, that all mechanical dimensions are withinthe specified tolerances. Some of the relevant qual-ity parameters are summarized in the next threefigures.

Fig. 4.33 shows the range of the achieved light yieldmeasured with a standard photomultiplier with bi-

Figure 4.32: Machined and optically polished barreltype PWO-II crystals for the PANDA electromagneticcalorimeter.

alkali photocathode at room temperature. The nexttwo figures document the variation of the inducedabsorption. Fig. 4.34 shows for 1 cm thick samplescut from the ingots of a pre-production lot the in-duced absorption as a function of wavelength. A60Co source was used to deposit a total dose of 1kGy at a rate of 2 kGy/h. Fig. 4.35 projects thedistribution of the induced absorption value at thewavelength of 420 nm. These measurements wereperformed at the irradiation facility at BTCP.

Figure 4.33: Distribution of the light yield of barreltype PWO-II crystals of the pre-production lot mea-sured at room temperature using a photomultiplier tubewith bialkali photocathode.


Figure 4.34: Radiation induced absorption spectra of1cm thick PWO-II samples cut from the ingots of a pre-production lot measured at room temperature. A 60Cosource has deposited a total dose of 1 kGy at a dose rateof 2 kGy/h.

Figure 4.35: Distribution of the induced absorptionmeasured at λ=420nm after an absorbed dose of 50 Gyat room temperature for PWO-II crystals of the pre-production run.

4.5 Quality Requirementsand Control

The electromagnetic calorimeter of PANDA requiresthe production of nearly 16,000 scintillating crys-tals. The envisaged high performance of the detec-tor imposes strict requirements on the crystal pa-rameters, such as sufficient and uniform light yield,short decay time free of slow components, very lim-ited light yield loss under irradiation and finallyprecise geometrical dimensions. To ensure a highand efficient production rate, the certifying proce-dure for such a large scale project has been elabo-rated for both the producer and the customer site

taking advantage of the many years of experiencefor CMS/ECAL [53]. The equipment for automaticmeasurement of the optical and scintillation prop-erties, the associated software tools as well as themethods and facilities for the radiation hardnesstests are described in the upcoming sections.

The target spectrometer EMC requires the produc-tion of PbWO4 crystal elements of 200 mm length ofone shape for both endcaps and 11 different shapesfor the barrel. The latter ones are used in two sym-metric versions. Each crystal has to pass two certi-fying procedures, one before delivery at the manu-facturer site and a final one performed by the cus-tomer. Presently, the specification limits, which arelisted below, are more stringent mainly due to therequested higher light yield and the operation atlow temperature down to T=-25oC.

1. General Properties

• Crystal dimensions have to be conform tothe drawings given in Fig. 7.3, Fig. 7.24,and Fig. 7.35.

• Crystals have to be oriented with theirfront face to the seed part of the ingot.

• The appearances of crystals with respectto color, visible spatial and surface defectshave to conform to the coordinated refer-ence crystal.

• The polished surfaces are polished witha roughness of Ra ≤0.02µm, the lappedface has a roughness Ra = 0.40±0.05µm.This parameter may be subject of modi-fication under mutual agreement with theproducer.

• The surface finish of chamfers should bemade at a roughness of not more than0.5µm (lapping).

• No cracks prolonging deeper than 0.5 mminto the crystal are allowed on the cham-fers.

2. Optical Properties: These properties of allcrystals should be measured with cross-checkedequipment.

(a) Longitudinal transmission (absolute val-ues)≥35 % at λ=360nm≥60 % at λ=420nm≥70 % at λ=620nm

(b) The non-uniformity of the transversaltransmission δλ at the transmission valueof T=50 % has to be δλ≤=3 nm, for 6


measurements every 4 cm; the first one isat 1.5 cm from the front face.

(c) Light yield ≥15 phe/MeV at T=18C (forcrystals with one lateral side de-polished).

(d) Light yield ≥16 phe/MeV at T=18C (forcrystals with all sides polished).

(e) Decay time: LY(100 ns)/LY(1000ns)>90 % at T=18C.

3. Radiation Hardness

(a) Radiation hardness is evaluated by themeasurement of the induced optical ab-sorption in the crystal along its axis. Thecrystals shall be kept at all times at a tem-perature T=20± 5C between the irradi-ation and the completion of the measure-ments. The value of the induced absorp-tion calculated from the optical transmis-sion of the crystal caused by the irradi-ation is limited to ∆k≤1m−1 at λ=420nm due to lateral 60Co irradiation. Thetotally accumulated dose shall be 30 Gyat a rate of 50-500 Gy/h. The meanvalue of the ∆k distribution defined foreach lot and based on the measurementsof not less that 50% of the lot shall be< ∆k >≥ 0.75 m−1.

The conventional laboratory methods used for crys-tal measurements cannot provide the necessarythroughput. Therefore, a high-productivity indus-trial system - Automatic Crystal Control System(ACCOS) has been constructed [54] and installed atboth the crystal production facility and at CERNduring implementation of the CMS project (seeFig. 4.36). There is an agreement with the CMSmanagement on the use of this equipment and thedirectly related infrastructure. The equipment ishighly automated in order to reduce the influenceof the human factor and to provide an output ofat least 30 crystals per day. The PWO crystal ismechanically fragile and can easily be damaged bycontact with hard equipment parts. Therefore, un-necessary frequent handling of crystals should bestrictly avoided. The adopted solution is to keepcrystals lying on special supports during all mea-surements and to move the spectrometers alongthem. Consequently, special devices of a very com-pact layout were designed. The overall setup in-cluding the associated 3D-machine is installed ina temperature-controlled and light-tight room. Abar-code reader identifies each crystal from the bar-coded label glued on the small end-face at the pro-duction stage. Two identical ACCOS machines are

installed both at the production site and at CERNto provide non-stop operation even in case of mal-functioning.

Figure 4.36: The ACCOS machine installed at theCMS regional center at CERN to perform the qualityinspection of PWO-II crystals in a semi-automatic pro-cedure.

4.5.1 Measurements of the LightYield and the Light YieldNon-uniformity

A start-stop or delayed coincidence method is usedboth for the measurement of the scintillation de-cay time and light yield (LY). The proportionalitybetween the count rate of the stop signals and thelight yield of the scintillator for small quantities oflight is used to determine the LY by integration ofthe decay time spectrum. It means that both pa-rameters can be measured simultaneously. In thiscase, neither optical coupling between scintillatorand PM nor scintillator wrapping are needed incontrast to the very widely used method of pulse-height spectrum measurement by using radioactiveγ-sources (most commonly 60Co) with consequenttotal absorption peak position determination. Thestart-stop method puts two constraints on the eventrates, which limits the productivity. One is thelimitation on the annihilation source activity. Itshould not exceed 105 Bq. At higher activity theprobability of random coincidences of two γ-quantafrom different decays becomes too high and leadsto background. The second is the limitation on theprobability of detecting scintillation photons in thestop channel per single excitation. A conventionalTDC requires that the average number of detectedphotons to be not more than 0.1 per single excita-tion. Otherwise, the measured scintillation decay


curve will be distorted, especially in the slow com-ponent region. The use of a multi-hit TDC wouldavoid this problem. The dead time of the stop-channel detector can also cause distortion, in par-ticular in the region of the fast component, whenthe fastest component in scintillation is less than orcomparable to the stop channel dead time, whichis the case for PWO. A single-hit TDC (or TDCwith multi-stop rejection [55]) was implemented toovercome the distortion problem. The most com-prehensive analysis of TDC with multi-stop rejec-tion is made in [56]. The start-channel detectorconsists of a Hamamatsu UV-extended photomulti-plier (R5900) coupled to a 20 × 20 × 20mm3 BaF2

scintillator. The 22Na annihilation source with anactivity of 200 kBq is mounted as close as possi-ble to the scintillator and delivers about 2 · 104 /sstarts. The start photomultiplier with scintillatorand source is mounted on a moving platform to-gether with the transversal transmission photospec-trometer for scanning along the crystal. The stop-channel is based on another Hamamatsu R5900 se-lected with a low count rate of dark pulses. Theaverage number of stop signals (detected scintilla-tion photons) amounts to ∼0.3 per single excitationfor the most luminous PWO crystals. The overallcount rate of good events (scintillation photons pro-duced by scintillation decay in PWO and collectedin the decay time spectrum) is typically 1 kHz, and aspectrum with acceptable statistics can be acquiredwithin one minute.

4.5.2 Inspection of the OpticalProperties

The longitudinal transmission can be reconstructedusing transmission data taken in transverse direc-tion. However, it is impossible to obtain the re-quired precision in case of PWO scintillation ele-ments because of uncertainties in the Fresnel re-flection level from crystal to crystal. These un-certainties are caused by the deviation of the crys-tal growth axis from the crystallographic axis a by± 5. Therefore, both types of transmission mea-surements should be performed. Since PWO is abirefringent crystal and the shape of the scintil-lation elements varies, a wide-aperture photo de-tector is required. This prevents from the use ofspectrometers, which are based on photodiode ar-rays with a light dispersing prism. On the otherhand, the PWO transmission spectrum has no nar-row absorption bands, so it is not necessary to mea-sure the spectrum at many wavelengths. It can bemeasured only at several well-chosen wavelengths

with subsequent reconstruction of the total spec-trum to speed up the measurements. For this pur-pose, a set of 11 interference filters was used. Twodedicated compact photospectrometers for opticaltransmission measurement have been designed andbuilt. The optical part of each photospectrometerincludes a 20 W halogen lamp, a four-lens collima-tor, three objective lenses, a rotating changer withinterference filters, collecting mirrors, a 20×20 mm2

UV-extended photodiode (Hamamatsu S6337) andreadout electronics. The spectrometer dimensionsare 70 × 80 × 220 mm3. The compact design al-lows the fast movement of the spectrometer fromthe measurement zone to the calibration position(air measurement). During operation, the rotatingwheel with interference filters crosses the light beamsequentially. Such a system provides a fast switch-ing (0.5 s) of the wavelength. The time necessary tomeasure a transmission spectrum in the range from330 nm to 700 nm takes less than 10 s for one spatialpoint.

4.5.3 Measurement of theGeometrical Dimensions

All dimension measurements are carried out withthe standard three-dimensional machine TOPAZ 7supplied by Johansson AB (Sweden), which pro-vides three coordinates of several points on eachcrystal face with a precision of ±5µm and subse-quent calculation of the crystal dimensions and theplanarity of all faces.

4.5.4 Analysis and Documentation

Software is provided for a user interface, acquisi-tion, data processing, results storage and presen-tation. It also supports communication with theC.R.I.S.T.A.L. database software [57] and the 3D-machine. Software components are implementedin LabViewTM graphical programming system andMicrosoftTM Visual C++ and runs on an IBM-compatible PC under MicrosoftTM Windows oper-ating system. A special algorithm was developedfor the processing of the scintillation kinetics data[58]. Instead of a non-linear iterative fitting pro-cedure usually used for multi-exponential functionsand giving sometimes unstable results due to inher-ent convergence problems, a new approach is basedon non-iterative 3-step integrating/subtraction pro-cedure working without user intervention.


4.5.5 Control Measurements ofProduction Stability

To provide the stability of the technology and min-imize the number of rejections the following set ofcontrol measurements should be carried out in ad-dition to the certification measurements along thetotal manufacturing process.

• Quality check of the raw material providedby pilot crystal growth and measurements (1per 100 crystals) - the only acceptable rate interms of time and expenses;

• Technology discipline check:analysis of average crystals parameters per lot;analysis of average crystals parameters duringsome period for each growing machine;analysis of rejected crystals

• Radiation hardness characterization onthe sampling basis:measurement of radiation induced absorptionin top parts of crystals (5-10 top parts per 100crystals);measurement of radiation induced absorptionin full-size crystals devoted to the specific ovenin case a previous check gives marginal results

Pilot crystals along with top parts of regular crys-tals will be analyzed in detail at the INP (Minsk)using the same methods and laboratory equipmentas were exploited during the development phase ofPWO crystals. All crystals rejected by the certi-fication measurements have to be investigated byindependent laboratory methods with full confirma-tion of the results. Even accepted crystals, whichhave passed the certification procedure, will bechecked by laboratory methods on the sampling ba-sis. All required measurements during PWO mass-production can be summarized in a 3-level schemeto make decisions. On the first level, the manu-facturer and INP provide pilot crystal growth andtests to prove the specific raw material batch to beacceptable for crystal production. On the secondlevel, the produced crystals pass through the certi-fication procedure on the ACCOS system and ra-diation hardness checks at the manufacturer plant.The data analysis is performed at INP including thedecision on the delivery of the crystals to the cus-tomer. On the third level, the customer carries outthe visual inspection, the certification on the AC-COS machine and radiation hardness tests. If all re-quirements are fulfilled the crystals will be acceptedfor the assembly into the calorimeter. The first twosteps of the final acceptance will be performed at

the CERN facilities. Final radiation tests will bedone at the facility at Giessen. Depending on theoverall productivity and stability of the manufac-turing process some tests at the customer facilitiesmight be done either on all crystals or on selectedsamples.

4.5.6 Manufacturer for the FinalProduction

All presently installed or operating calorime-ters, such as ECAL (CMS), PHOS (ALICE)and HYCAL of the PrimEx experiment atJLAB [59] are composed of PWO crystalsmanufactured by ShanghaiInstituteofCeramics(SICCAS, China) or the two Russian producersBogoroditskTechnicalChemicalPlant or NorthCrystals.The latter one, the major producer for PHOS, hasmoved out of the scintillator business. As outlinedin the previous chapters, the specification limits forthe calorimeter crystals have to be identical to thequality of PWO-II, which was developed in closecollaboration with BTCP and the CMS/ECALgroup. Up to now a first pre-production runperformed at BTCP was completed in May 2008comprising more than 710 crystals of all 23 differentshapes with an overall length of 200 mm. Thefirst lot allows to complete one out of the 16 slicesof the barrel EMC and a first subarray of theforward endcap EMC. Some of the achieved qualityparameters were shown in the previous chapter.

There exists a long collaboration with the ShanghaiInstitute of Ceramics and many state-of-the-art testcrystals of various sizes were investigated. Finally, afull matrix comprising 25 modules of 200 mm lengthwith identical cross section of 20×20 mm2 were de-livered in 2005 for comparison.

These crystals are grown along the c-axis by a modi-fied Bridgeman method [61] and consequently differin the longitudinal and transverse optical transmit-tance. Fig. 4.37 shows a comparison of measuredtransverse and longitudinal transmission. The mea-surements performed for a sample from both manu-facturers are compared to theoretical limits assum-ing no internal absorption [62, 60]. The chosen crys-tal from BTCP corresponds to CMS quality. Thethree different calculations are using the refractiveindex of ordinary light, which propagates along thec-axis and has a polarization perpendicular to the c-axis. Extraordinary light propagates perpendicularto the c-axis and has a polarization along the c-axis. Finally, unpolarized light is assumed to prop-agate perpendicular to the c-axis. Because of thebirefringence, the theoretical limit of the extraordi-


Figure 4.37: Comparison of the optical transmission of two crystal samples from the manufacturers BTCP andSICCAS. The experimental data are compared to theoretical calculations described in the text. The figures aretaken from Ref. [60].

nary light is about 3 % higher than that of the or-dinary light. Since the crystals produced in Chinaare grown along the c-axis, the experimental lon-gitudinal transmission should be compared to thetheoretical limit for propagation along the c-axis in-dependent of polarization. Consistent with the the-oretical calculations, the samples from BTCP showa better longitudinal transmission. However, themeasurement of the absolute yield of the detectedscintillation light is slightly superior for the Chineseproducts, which are doped exclusively with yttriumions. Compared to CMS standards, these crystalsshow approximately 30–40 % more light.

Small crystal samples (16×16×30 mm3), whichwere readout with photomultiplier tubes, confirmthe good photon statistics expressed by the excel-lent energy resolution of σE/E=14.5 % obtained for662 keV γ-rays at a temperature of T=-26C. The25 full size crystals at room temperature deliver amean light yield of ∼18.0 phe/MeV and are compa-rable to PWO-II. Both values are determined withphotomultiplier tubes. Fig. 4.38 summarizes for oneof the brightest samples the change of light yieldat reduced temperatures. The values are obtainedwith a calibrated photomultiplier tube (HamamatsuR2059, quantum efficiency QE(420 nm)=20.5 %) byintegrating the response over time gates rangingbetween 75 ns and 1µs, respectively. The figure

Figure 4.38: Absolute light yield of a 200 mm longrectangular crystal manufactured by SICCAS. The scin-tillation light has been measured with a photomultipliertube with bialkali photo cathode. The response is shownat different temperatures as a function of integrationgates varying between 75 ns and 1µs, respectively.

shows a high and nearly constant light yield atT=+31C. There is a strong increase of the emit-ted light when the temperature is decreased down toT=-25C. However, in strong contrast to crystalsof similar size from BTCP, there is a significant andrising contribution of slow scintillation components.


Therefore, in order to collect ≥90 % of the light out-put, integration times between 0.5 and 1.0µs be-come mandatory leading to a severe impact on thecount rate capabilities of a calorimeter [63].

Figure 4.39: Longitudinal optical transmission of two200 mm long crystals manufactured by SICCAS. Theslightly colored sample shows a strong absorption bandpeaking at λ∼420 nm.

In addition, detailed investigations have observedsignificant inhomogeneities of the crystals. In par-ticular, about half of the delivered samples show ayellowish color, which leads in the optical transmis-sion spectrum to a strong absorption region peakedat λ∼420nm as shown in Fig. 4.39. A complemen-tary study of the transversal optical transmission,determined at several positions along the crystalaxis, documents an insufficient homogeneity of sev-eral crystals. Fig. 4.40 shows strong deviations ofthe transmission values appearing typically at bothends of the crystal when measured at the criticalwavelength of λ∼420 nm.

As a consequence, the manufacturer SICCAS hasstarted further developments in order to adapt thescintillation kinetics at lower temperatures to thePANDA needs. Some of the samples with a volumeof a few cm3 show faster decay kinetics but therehas been up to date not yet the delivery of full sizescintillation crystals for a realistic comparison.

Figure 4.40: Inhomogeneity in the transversal opti-cal transmission of several 200 mm long crystals manu-factured by SICCAS and measured at the wavelengthλ∼420 nm. The cross section of the rectangular crystalsis 20×20 mm2. Most of the samples indicate deviationseven at both end sections of the crystal.

References

[1] R. Novotny, Nucl. Instrum. Meth. A537, 1(2005).

[2] R. Novotny, High-resolution electromagneticcalorimetry in medium and high energyphysics based on inroganic scintillators, inProceedings SCINT 2005, edited byA. Gektinj and B. Grinyov, page 310, ISBN9666-02-3884-3, 2005.

[3] E. Lorenz, Nucl. Instrum. Meth. A225, 500(1984).

[4] R. Sumner, Nucl. Instrum. Meth. A265, 252(1988).

[5] J. A. Bakken et al., Nucl. Instrum. Meth.A343, 456 (1994).

[6] O. Adriani et al., Nucl. Instrum. Meth.A302, 53 (1991).

[7] W. W. Moses et al., Nucl. Instrum. Meth.A353, 189 (1994).

[8] G. J. Bobbink, A. Engler, R. W. Kraemer,J. Nash, and R. B. Sutton, Nucl. Instrum.Meth. A227, 470 (1984).

[9] C. Laviron and P. Lecoq, Nucl. Instrum.Meth. A227, 45 (1984).

[10] S. K. Sahu et al., Nucl. Instrum. Meth.A388, 144 (1997).


[11] K. C. Peng et al., Nucl. Instrum. Meth.A427, 524 (1999).

[12] D. F. Anderson, IEEE Trans. on Nucl. Sci.36, 137 (1989).

[13] R. Novotny, Nucl. Instrum. Meth. A486, 131(2002).

[14] J. Chen et al., Trans. on Nucl. Sci. 52, 3133(2005).

[15] R. W. Novotny et al., IEEE NSS 2006,Conference Record , 1118.

[16] M. Thiel et al., Trans. on Nucl. Sci. 55, 101(2008).

[17] J. Chen et al., Trans. on Nucl. Sci. 54, 1319(2007).

[18] CERN-LHCC-97-33.

[19] CERN-LHCC-95-71.

[20] A. A. Annenkov, M. V. Korzhik, andP. Lecoq, Nucl. Instrum. Meth. A490, 30(2002).

[21] A. Annenkov et al., Nucl. Instrum. Meth.A450, 71 (2000).

[22] M. Kobayashi et al., volume A537, page 312,2005.

[23] M. Korzhik, Talk at FEMC03, FZ Julich,Germany, March 10-11, 1003.

[24] M. Bohm et al., phys. stat. sol. A. 167, 243(1998).

[25] X.Liu, G.Hu, X.Feng, Y.Huang, and Y.Zhang,Phys. Stat. Sol. (a) 190, R1 (2002).

[26] P. Richter, G. Kruger, and C. Pistorius, ActaCrystall B32, 928 (1976).

[27] T. Fujita, I. Kawada, and K. Kato, ActaCrystall B33, 162 (1977).

[28] R. Williams, Y. Zhang, Y. Abraham, andN. Holzwarth, 1999.

[29] P. Lecoq et al., Nucl. Instrum. Meth. A365,291 (1995).

[30] R. Y. Zhu et al., Nucl. Instrum. Meth. A376,319 (1996).

[31] B. Han, X. Feng, G. Hu, Y. Zhang, andZ. Yin, J. Appl. Phys. 86, 3571 (1999).

[32] E. Auffray, IEEE Trans. on Nucl. Sci. 55, 57(2008).

[33] R. N. K. Mengel et al., IEEE Trans. on Nucl.Sci. 45, 681 (1998).

[34] R. Novotny et al., IEEE Trans. on Nucl. Sci.47, 1499 (2000).

[35] J. Moreau, R. Gladyshevski, P. Galez,J. Peigneux, and M. Korzhik, J. of Alloys andCompound 284, 104 (1999).

[36] A. Hofstaetter et al., phys. stat. sol. B. 89,375 (1978).

[37] M. Bohm et al., Journal de Physique C 6, 508(1980).

[38] M. Bohm et al., Proceedings of SCINT99,Moscow, Russia, Aug. 16-29, 1999 , 619(2000).

[39] S. Baccaro et al., Proc. Intern. Workshop onTungstate Crystals, Rome, Italy, Oct 12-14(1998).

[40] M. Bohm et al., Proc. Intern. Workshop onTungstate Crystals, Rome, Italy, Oct 12-14 ,139 (1998).

[41] M. Bohm et al., Rad. Effects and Defects inSolids 150, 413 (1999).

[42] V. Laguta, J. Rosa, M. Zaritski, M. Nikl, andY. Usuki, J. Phys. Condensed Matter” 10”,7293” (1998”).

[43] A. Hofstaetter et al., Proceedings ofSCINT99, Moscow, Russia, Aug. 16-29, 1999 ,128 (2000).

[44] A. Hofstaetter et al., RadiationMeasurements 33, 533 (2001).

[45] M. Nikl et al., Applied Phys. Lett. 71, 3755(1997).

[46] S. Baccaro et al., Phys. Stat. Sol. (a) 160, R5(1997).

[47] http://www.atomtex.com.

[48] V. A. Batarin et al., Nucl. Instrum. Meth.A550, 543 (2005).

[49] P. A. Semenov et al., Nucl. Instrum. Meth.A582, 575 (2007).

[50] G. Drobychev et al., IEEE Trans. on Nucl.Sci. 48, 1177 (2001).


[51] A. Fedorov et al., Nucl. Instrum. Meth.A413, 352 (1998).

[52] P. Adzic et al., Eur. Phys. J. C44, 1 (2006).

[53] E. Auffray et al., CMS Note (1998/038).

[54] O. Missevitch et al., Nucl. Instrum. Meth.A456, 325 (2001).

[55] W. W. Moses, Nucl. Instrum. Meth. A336,253 (1993).

[56] A. F. Cherniavskii et al., Atomizdat (inRussian) , 352 (1974).

[57] J.-M. L. Goff et al., CMS Note (1996/003).

[58] O. V. Missevitch et al., Instruments andExperimental Techniques 45, 34 (2002).

[59] A. Gasparian, Proceedings of 11th Int. Conf.on Calorimetry in Particle Physics, Perugia,Italy, 2004 , 109 (2005).

[60] R. Y. Zhu, private communication.

[61] X. Qu et al., Nucl. Instrum. Meth. A480, 470(2002).

[62] R. Mao et al., IEEE Trans. on Nucl. Sci. 51,1777 (2004).

[63] F. Hjelm et al., Proceedings of SCINT05,Alushta, Ukraine, 2005 , 381 (2006).

67

5 Photo Detectors

The detection of scintillation light of lead tungstateunder the given conditions for the PANDA EMCrequires excellent photo detectors.

The magnetic field of about 2 T precludes the use ofconventional photomultipliers. On the other handthe signal generated by ionization in a PIN photodi-ode by a traversing charged particle is too large forour applications. To solve these problems a pho-tosensor insensitive to magnetic fields and with asmall response to ionizing radiation has to be used.

Since lead tungstate (PbWO4) has a relatively lowlight yield, the photosensor is required to have inter-nal gain in addition. Due to the increase of the crys-tal light yield accomplished by cooling the scintilla-tor down to a temperature of T = −25C, the usedphoto detectors have to be radiation hard in thistemperature regime, which implies detailed studiesconcerning possibly occurring radiation damages.

For the barrel EMC an Avalanche Photodiode(APD) which has an internal signal amplificationin the silicon structure is chosen as photo detec-tor. The low-energy photon threshold in the orderof 10 MeV requires to maximise the coverage of thereadout surface of the crystals, leading to the de-velopment of Large Area APDs (LAAPD) with anactive area of 10×10 mm2, which have been testedfor the given requirements. During the developmentof the readout system for the EMC the typical sizeof the crystal readout surface of 27 x 27 mm2 wasleading to the development of an LAAPD with arectangular shape to cover a maximum of this spacewith two neighboring photo detectors.

The photon detection in the forward endcap EMChas to deal with rates up to 500 kHz and magneticfields up to 1.2 T. Therefore we have chosen vac-uum phototriodes (VPT) with a diameter of 22 mmas photon detectors. The main reasons for thischoice are rate capability, radiation hardness, ab-sence of nuclear counter effect and absence of tem-perature dependence. Standard photomultipliersare excluded due to the magnetic field environment.In contrast to the barrel region, the magnetic fieldis oriented in the axial direction of the VPTs andthus makes it feasible to use VPTs for the endcap.Vacuum phototriodes are essentially a photomul-tiplier tube with only one dynode and weak fielddependence.

5.1 Avalanche Photodiodes(APD)

5.1.1 Introduction

The requirements to detect the scintillation light oflead tungstate in the barrel part of the EMC can besatisfied by silicon avalanche photodiodes (APDs).Avalanche photodiodes are reverse biased diodeswith an internal electric field used for avalanchemultiplication of charge carriers. The CMS collab-oration has developed APDs in collaboration withHamamatsu Photonics optimized to detect the scin-tillation light from lead tungstate crystals. ThoseAPDs have several advantages: compactness withan overall thickness of ≈ 200µm (Fig. 5.1), highquantum efficiency of 70–80 % at the wavelength ofmaximum emission intensity of lead tungstate, in-sensitivity to magnetic fields and low cost in massproduction. A disadvantage of this photosensor isits relatively small active area of 5 × 5 mm2 com-pared to the area of the crystal end faces. Thereforewe started the R&D of large area APDs (LAAPDs)with an active area of (10 × 10) mm2 in collabo-ration with Hamamatsu Photonics. Our R&D ofLAAPDs is based on the same internal structurethat has already been tested by the CMS collab-oration (see Fig. 5.1): Light enters the APD viathe p++ layer and is absorbed in the following p+

layer, where electron-hole pairs are generated. Theelectrons drift in the electric field towards the p-njunction where they are amplified by impact ioniza-

Figure 5.1: Schematic view of an APD with reversestructure [1].


tion, yielding avalanche gain, and drift through then-material to the n++ electrode where the chargecollection takes place. In front of the p++ layera passivation layer made of silicon nitride Si3N4 isused which reduces the decrease of quantum effi-ciency caused by reflection losses from the surfaceof the Si wafer.Due to the large capacitance of the (10 × 10) mm2

LAAPDs (see Table 5.1) an additional prototypeof same dimensions but lower capacitance has alsobeen tested. As shown in the Technical ProgressReport [2] the first prototype of the (10× 10) mm2

APDs showed a nonsatisfying radiation hardnessconcerning proton irradiation. Therefore the APDinternal structure was modified by adding a grooveto reduce the occuring surface current and in thepresent status of R&D work mainly two differentAPD types are under investigation. Those areLAAPDs with an active area of (10 × 10) mm2

with different capacitance values developed in co-operation with Hamamatsu Photonics: a so called’normal C’ type LAAPD (C = 270 pF ) and a’low C’ type LAAPD with a capacitance of C =180 pF . The foreseen rear side dimensions of thelead tungstate crystals (typ. (21.4× 21.4) mm2)make it impossible to use two LAAPDs withquadratic active area of our dimensions (packagesize: (14.5× 13.7) mm2). Therefore LAAPDs withrectangular package/active area have to be used tocover the main part of the crystal readout area.Those diodes with rectangular shape are presentlyunder development and first prototypes will beavailable mid of 2008. A preliminary drawing ofthe LAAPD dimensions of the rectangular LAAPDtype is shown in Fig. 5.2. This avalanche photodi-ode will have an active area of (7× 14) mm2. Twoof these rectangular LAAPDs will be used to detectthe scintillation light of one crystal in the barrelpart of the electromagnetic calorimeter.

5.1.2 Characteristics

To ensure a stable operation of an avalanche photo-diode several device properties have to be measuredduring screening or rather recorded during opera-tion. The main parameter to be reported is thetemperature of the APD during operation, becausethe dark current as well as the internal gain arestrongly depending on temperature. Therefore thetemperature has to be held stable down to an uncer-tainty of ∆T = ±0.1 C. Another essential param-eter is the value of the applied bias voltage, whichhas to be kept constant down to ∆UR = ±0.1V toensure a proper determination of the internal gainas well as a proper measurement of the device dark

current depending on bias voltage. Special atten-tion will be devoted to the main APD parametersin the following paragraphs.

5.1.2.1 Dark Current and Gain

Due to interactions of charge carriers with phononsthe gain of an APD decreases with increasing tem-perature. On the other hand the free mean path andthe band gap of semiconductor devices are temper-ature dependent. In our case these two parameterscan be neglected because of the envisaged operationtemperature of T = −25C. The degree of sensitiv-ity to temperature changes can be determined bythe proximity of the applied bias/reverse voltage(UR) to the value of breakdown voltage (UBr) andby the device characteristics as a result of internalstructure:A decrease of capacitance of the APD results in anincreasing sensitivity to temperature changes. Thetemperature dependence of gain has a value of -2.2 %/C at an internal gain of M = 50 in case ofa CMS-APD [3].The dark current Id of an APD can be devidedinto the bulk current Ib and the surface current Is.While the surface current is independent from theapplied gain M the value of the bulk current in-creases with increasing gain. The contributions ofthose two can be calculated separately by using thefollowing relation if the gain M and the overall darkcurrent Id have been determined:

Id = Ib ·M + Is. (5.1)

The internal gain M of the APDs at a fixed tem-perature is measured by using the following method:The dark current (Id) and the current under con-tinous illumination (Iill) at a fixed wavelength ofλ = 420 nm, according to the maximum emissionwavelength of PbWO4, is recorded for several biasvoltages up to breakdown. From the ratio of photocurrents (Iill−Id) of high bias voltages and the volt-age value where no amplification occurs (equivalentM = 1) the gain M is calculated by:

M =Iill(UR)− Id(UR)

Iill(M = 1)− Id(M = 1). (5.2)

The results of these measurements done at roomtemperature are shown for a ’normal C’ type APDin Fig. 5.3 and for a ’low C’ type version in Fig. 5.4.For the evaluation of the maximum gain of the de-

vice up to which a stable operation of the diode canbe ensured, the dark current dependence of gain hasto be regarded in parallel. For the investigated ’nor-mal C’ LAAPD gains up to a value of M = 3300


Figure 5.2: Technical drawing of the developed rectangular APD type.

could be reached before the breakdown voltage ofthe diode was reached. A closer look on the darkcurrent reduces this value down to M = 500, be-cause at higher gains the linearity of the dark cur-rent given by Eq. 5.1 is no longer guaranteed (seeFig. 5.3). Similar behaviour can be observed in caseof the ’low C’ version, where the maximum gain de-creases from M = 460 to a value of M = 160 due tothe observed dark current unlinearity. The discrep-ancy in the maximum usable gain of the two APDtypes may originate from their different dark cur-rent behaviour depending on the internal APD gainand is caused by their different internal structure.This circumstance could be clarified by consideringthe ratio Id/M depending on gain which is shownin Fig. 5.5. As it can be seen the two APD typesbehave completely different for gain values largerthan M = 20: In case of the ’normal C’ LAAPDthe ratio is decreasing with increasing gain in con-trast to the Id/M behaviour of the ’low C’ versiondiode where the ratio is slightly increasing.The values of the gain variation depending on biasvoltage changes 1/M ·dM/dV were also determined.At gainM = 50 a value of 1/M · dM/dV = 3.4 %/Vhas been evaluated for the ’normal C’ versionand a similar value for the ’low C’ version of1/M · dM/dV = 3.2 %/V. These values are in goodagreement with the value given for CMS-APDs in[3].

5.1.2.2 Excess Noise Factor

The APD internal charge carrier multiplication viaavalanche is a statistical process. A characteriza-

tion of these statistical fluctuations of the APD gainσM is given by the excess noise factor F and has itsorigin in inhomogeneities in the avalanche regionand in hole multiplication. The value of the excessnoise factor is determined by the internal structureof the APD and is related to the amplification ofelectrons and holes at a given value of the internalgain M :

F ≈ k ×M +(

2− 1M

)× (1− k), (5.3)

with k defined as the ratio of the ionization coeffi-cients for electrons to holes. Its value is determinedby the shape of the electric field near the p-n junc-tion. Its r.m.s. broadening of a signal from Npephotoelectrons is given by

√F/Npe. If Nγ photons

per MeV are emitted by the crystal the number ofphotoelectrons created inside the conversion layer ofthe diode with quantum efficiency QE can be cal-culated by Npe = Nγ · QE (assuming full coverageof the crystal rear side by the APD). Therefore ashower of energy E(MeV) creates ENpe photoelec-trons.The processes mentioned above result in an addi-tional contribution to the energy resolution of [4]

σEE

=1√ENpe

·√M2 + σM 2

M

=1√E·

√F

Npe. (5.4)

From Eq. 5.4 it can be clearly seen that the ex-cess noise factor should be as small as possible tomaintain an excellent energy resolution. The value


[V]RU0 100 200 300 400 500

M

0

500

1000

1500

2000

2500

3000

3500

gain M0 100 200 300 400 500

[n

A]

dI

0

20

40

60

80

100

120

Figure 5.3: Determined gain M of the ’normal C’ ver-sion LAAPD (above) and the corresponding dark cur-rent dependence of M in the range of linear dark currentbehaviour (below).

of the excess noise factor at an internal gain ofM = 50 of the CMS-APDs is F = 2.0 (Fig. 5.6),at gain M = 100 the excess noise factor is increas-ing to F = 2.33. The excess noise factor of the firstLAAPD ’normal C’ prototype was measured at theAPD laboratory of the CMS collaboration at CERNin April 2004. The result of this measurement isshown in Fig. 5.7. At an internal gain of M = 50the excess noise factor of these LAAPDs has a valueof F = 1.38, at gain M = 100 its value is F = 1.57.

5.1.2.3 Nuclear Counter Effect (NCE)

The Nuclear Counter Effect (NCE) describes an ex-tra amount of charge produced inside a photodiode

[V]RU0 50 100 150 200 250 300 350 400

M

0

100

200

300

400

500

gain M0 20 40 60 80 100 120 140 160 180

[n

A]

dI

0

2

4

6

8

10

12

14

Figure 5.4: Determined gain M of the ’low C’ ver-sion LAAPD (above) and the corresponding dark cur-rent dependence of M in the range of linear dark currentbehaviour (below).

by a charged particle directly hitting it, in additionto the charge produced by the scintillation light ofthe used crystal [6]. Therefore the NCE can causea decrease of resolution in the energy measurementof an electromagnetic shower due to the leakage(secondary produced electron/hole pairs) of chargedparticles from the crystal.Charged particles crossing a silicon layer of giventhickness create a number of electron/hole pairs cal-culated by:

dn

dx=dE

dx· % · 1

Ee/h≈ 100 e/h pairs / µm. (5.5)

In an APD only the photoelectrons created infront of the avalanche region, inside the conversion


gain M0 100 200 300 400 500

/ M

[n

A]

dI

0

0.2

0.4

0.6

0.8

1

1.2

1.4

Figure 5.5: Id/M depending on M of the ’normal C’version (blue) and the corresponding contribution forthe ’low C’ type APD (red).

Figure 5.6: Measured excess noise factor F as a func-tion of the internal gain of a CMS APD [5].

layer, always experience full amplification. Are theelectrons created inside the avalanche region theirreachable amplification depends on their place ofcreation (see details Sec. 5.1.3). To give a quantita-tive measurement of this effect, an effective thick-ness deff of the APD can be defined. Its value canbe determined by exposing the APD directly to a90Sr source, emitting beta electrons up to an en-ergy of 2 MeV. The charge Q collected in the APD,operating at gain M , will be compared to the valuemeasured with a PIN diode of well-known thicknessdPIN which results in the following relation [4]:

deff =dPIN

Q(PIN)· Q(APD)

M. (5.6)

The calculation of the NCE includes the determina-tion of the number of photoelectrons Npe detectedby the crystal-APD system. Its value depends on

Figure 5.7: Measured excess noise factor F as a func-tion of the internal gain of the first prototype of theLAAPD S8664-1010SPL with ’normal C’ manufacturedby Hamamatsu.

the light yield (LY) of the crystal, the APD quan-tum efficiency QE and the fraction of the crystalrear area covered by the APD (f). The asump-tion of a 22 % coverage using a (10×10) mm2 largeAPD mounted on a (2.14×2.14) cm2 crystal endface(f = 0.22) results in:

Npe = LY ·QE · f ≈ 500× 0.7× 0.22

= 77.0pe

MeV(5.7)

at an operation temperature of T = −25C. Basedon these values one MIP creates a signal inside theAPD (operating at T = −25C) equivalent to thelight signal released in the crystal by a photon ofenergy:

EMIP =dn

dx× deffNpe

≈ 1.3 MeV× deff (µm). (5.8)

From Eq. 5.8 it can clearly be seen, that the effec-tive thickness of the diode should be as small aspossible to minimize the influence of the NuclearCounter Effect (assuming deff = 5.6µm (CMS)EMIP would have a value of 7.28 MeV). On theother hand a reduction of deff increases the capac-itance of the APD. Therefore a compromise has tobe found and in fact as shown in Fig. 5.8 one of thetested APD types appears to fulfill these require-ments. While the exposure with electrons seems tohave no measureable effect on the ’normal C’ ver-sion LAAPD (Fig. 5.8 b), the ’low C’ version showsa clear signal created by the electrons directly hit-ting the APD (Fig. 5.8 c). The determination of


ADC channel0 2000 4000 6000 8000 10000 12000 14000 16000

cou

nts

0

100

200

300

400

500

600

700

800

(a)

ADC channel0 2000 4000 6000 8000 10000 12000 14000 16000

cou

nts

0

100

200

300

400

500

600

700

800

(b)

Figure 5.8: Measured Nuclear Counter Effect of a PINdiode of 300µm thickness (green), of the ’normal C’version LAAPD (red) and of the ’low C’ version LAAPD(blue). In part a) the comparison between PIN diodeand ’normal C’ version APD is shown and b) shows thecorresponding comparison between PIN and the ’low C’version LAAPD.

the effective thickness of the used LAAPDs is hin-dered compared to the measurement done by CMSbecause of the much larger device capacitance. Thisvalue gives a major impact on the diodes noise be-haviour and has to be known to conceive the layoutof a low noise preamplifier which will be a majorpart of the APD readout chain. For PANDA themeasurement of deff requires the minimization ofany possible noise source and is currently under de-velopment.

5.1.2.4 Electrical and Optical Propertiesof the Tested APD Types

The optical and electrical properties of the testedtwo LAAPDs mentioned above are summarized inTable 5.1 and compared to the APDs to be usedin the electromagnetic calorimeter ECAL of the

Figure 5.9: QE of a PIN diode (squares/pink) com-pared to the QE measured for a LAAPD (stars/green).

CMS experiment. Most of the measured parame-ters of the LAAPDs are in good agreement withthe properties of the APDs developed by the CMSgroup in collaboration with Hamamatsu Photonics.The measured quantum efficiency of the PANDA-LAAPDs is shown in Fig. 5.9 compared to valuesmeasured with a PIN diode. Due to the four timeslarger active area of the LAAPD prototypes theircapacitance is much larger than those of the tested(5× 5) mm2 APD type of CMS.

5.1.3 Radiation Damage due toDifferent Kinds of Radiation

To analyze the influence of radiation on a semi-conductor device the knowledge of the internal de-vice structure has to be assumed. Therefore severalmeasurements to gain inside on the internal APDstructure are inevitable. The possibility of calcu-lating the effective thickness deff of the tested de-vices was already mentioned in Sec. 5.1.2.3. An-other important parameter of the diode which wasnot discussed in detail so far and has also influ-ence on the NCE is the conversion layer thicknessdconv. In Sec. 5.1.2.3 it was already mentioned thatthe longitudinal position x where the electron holecreation inside the diode takes place has influenceon the reachable amplification of the photoelectronspassing the device structure: For light absorbed be-fore reaching the avalanche region the gain is con-stant. The absorption of light inside the avalancheregion leads to a decrease of the internal gain andtherefore to additional gain fluctuations. Assuming


Property Condition CMS APD S8664-1010SPL S8664-1010SPL’normal C’ ’low C’

Active area [mm2] 5× 5 10× 10 10× 10Quantum efficiency QE [ % ] M = 1,λ = 420 nm 70 70 70Breakdown voltage Ubr [V ] Id = 100µA 400 400 500-600Dark current Id [nA] M = 50 5 10-40 20-50Capacitance C [pF ] M = 50 80 270 180Excess noise factor F M = 50 2.0 1.38 not measuredExcess noise factor F M = 100 2.33 1.57 not measured

Table 5.1: Electrical and optical properties of the tested APD types.

a constant electric field in the avalanche region ofwidth W , the gain M could be calculated by usingthe relation

M = eαi·(W−x) (5.9)

in which αi defines the ionization coefficient. Ad-ditionally the light absorption inside the avalancheregion follows the exponential law

N(x) = N0 · e−(α(λ,T )·x). (5.10)

The absorption coeffcient α(λ, T ) respectivelyα(Eγ , T ) of silicon is shown in Fig. 5.10 depend-ing on the incident photon energy. The conversionlayer thickness of the APD could be determined byapplying the definition of the optical mean depth ofpenetration [8]:

dconv =1

α(λ, T ). (5.11)

By normalizing the measured gain values depend-ing on wavelength, as shown in the lower plot ofFig. 5.11 for the ’normal C’ type APD (Fig. 5.12

Figure 5.10: Absorption coefficient α(λ, T ) of Si de-pending on the photon energy for different temperatures[7].

wavelength α(λ) dconvAPD version λcut [nm] [1/cm] [µm]’low C’ type 500 1.1 · 104 0.9’normal C’ type 540 7.05 · 103 1.4

Table 5.2: Determined conversion layer thicknessesdconv of the tested APD types.

for the ’low C’ type respectively), it is possible todefine a ’cut-off’ wavelength λcut above which lightis absorbed inside the avalanche region [9]. Theevaluated conversion layer thicknesses of the twodifferent APD types and their corresponding ’cut-off’ wavelenghts are assorted in Table 5.2 using theα(λ)-values given for a temperature of 300K. Itis quite obvious that any kind of irradiation of theAPD has an impact on the M(λ) behaviour of thedevice. Therefore measurements as e.g. shown inFig. 5.11 will lead to a more consolidated under-standing of the caused radiation damages and their

[nm]λ300 400 500 600 700 800

gain

M

0

100

200

300

400

500

600

[nm]λ300 400 500 600 700 800

norm

aliz

ed g

ain

M

40

50

60

70

80

90

100

110

540 nm≈ cutλ

Figure 5.11: M(λ) for the ’normal C’ version APD fordifferent UR values (above). Determination of λcut bynormalizing the data (below).


[nm]λ300 400 500 600 700 800

gain

M

50

100

150

200

250

[nm]λ300 400 500 600 700 800

norm

aliz

ed g

ain

M

30

40

50

60

70

80

90

100

110

120

500 nm≈ cutλ

Figure 5.12: M(λ) for the ’low C’ version APD fordifferent bias voltages (above). Determination of λcutby normalizing the data (below).

locations inside the APD structure.

The results of different irradiation experiments willbe reported in the following section. All irradia-tions have been done at T = −25 C to account forthe real operation conditions of the PANDA electro-magnetic calorimeter.

5.1.3.1 Proton Irradiation

Irradiation of an APD with protons causes two dif-ferent kinds of radiation damage:

• ionization effects at the surface reflected in thesurface current Is, and

• atom displacements inside the silicon bulk in-creasing the bulk current Ib.

The proton irradiation of the diodes has been doneat the KVI Groningen (the Netherlands) using a90 MeV proton beam providing a homogenous beamprofile. The diodes were irradiated until an inte-grated fluence of 1.1 · 1013 p was reached which iscomparable with the dose expected for 10 years ofPANDA operation. Using the NIEL theory this pro-ton fluence corresponds to an equivalent neutronfluence of 1.42 · 1013 n of an energy of 1 MeV [10].

The first step in evaluating the influence of irradi-ation on the APD characteristics is the remeasure-ment of the internal gain of the diode. The typicalchange of M(UR) due to proton irradiation is shownfor instance in Fig. 5.13 for the ’normal C’ version.It can be clearly seen that the gain of this diodetype has dramatically decreased. Taking into ac-count the assumed linearity of the dark current de-pending on M (see Fig. 5.14), the usable maximum

[V]RU0 100 200 300 400 500

M

0

200

400

600

800

1000

Figure 5.13: Determined gain values M of the’normal C’ type APD before (blue) and after (red) ex-posure with 1.1 · 1013 p of Ep = 90 MeV.

gain M0 100 200 300 400 500

[n

A]

dI

0

20

40

60

80

100

120

gain M0 20 40 60 80 100 120 140 160 180 200 220

[n

A]

dI

0

10000

20000

30000

40000

50000

60000

70000

80000

90000

Figure 5.14: Comparison of the measured dark currentvalues of the ’normal C’ type APD before (left/blue)and after (right/red) proton irradiation.

gain decrease of the ’normal C’ type APD after pro-ton exposure is in the order of 60%. The decreasein case of the ’low C’ version diode (see x-axes ofFig. 5.15) has a similar value about 50-60%. Theinfluence of proton irradiation on the internal APDstructure and the associated characteristic changescould be concretized by having a closer look on thedark current behaviour. For the two tested APDversions the dark current was measured in depen-dence of the internal gain before and after protonexposure. The results are shown in Fig. 5.14 andFig. 5.15. As already mentioned above this kindof measurement makes it possible to evaluate thedifferent rates of the surface current and the bulkcurrent to the overall dark current. The values ofthe single dark current contributions shown in Ta-ble 5.3 were evaluated by fitting Eq. 5.1 to the dat-


APD version Ib1 [nA] Ib2 [nA] Is1 [nA] Is2 [nA] Ib2/Ib1 Is2/Is1’low C’ type 0.07 268 0.2 ≈ 1554 ≈ 3829 ≈ 7770’normal C’ type 0.24 389 3.47 ≈ 1770 ≈ 1621 ≈ 510

Table 5.3: Influence of proton irradiation on the APD dark current. The values Is1 and Ib1 have been measuredbefore irradiation, Is2 and Ib2 afterwards.

gain M0 20 40 60 80 100 120 140 160 180

[n

A]

dI

0

2

4

6

8

10

12

14

gain M0 10 20 30 40 50 60 70 80 90

[n

A]

dI

0

5000

10000

15000

20000

25000

Figure 5.15: Comparison of the measured dark currentvalues of the ’low C’ type APD before (left/blue) andafter (right/red) proton irradiation.

[nm]λ350 400 450 500 550 600 650 700 750 800

no

rma

lize

d g

ain

M

40

45

50

55

60

65

70

75

Figure 5.16: M(λ) of ’normal C’ type after protonexposure with 1.1 · 1013 p at KVI Groningen.

apoints in Fig. 5.14 and Fig. 5.15 respectively. Theenormous increase of the ’low C’ bulk current com-pared to the increase factor Ib2/Ib1 measured forthe ’normal C’ type APD has its origin in the in-creased layer thickness of this diode type used forthe required reduction of the device capacitance.To get an impression where these observed pro-ton induced damages are mainly located the mea-surement of the gain dependence on wavelengthhas to be reconsidered after proton exposure forthe two different kinds of tested avalanche photo-diodes. The results of this measurement are shownin Fig. 5.16 and Fig. 5.17. Obviously in case of

[nm]λ400 500 600 700 800

no

rma

lize

d g

ain

M

45

50

55

60

65

70

75

80

Figure 5.17: M(λ) of ’low C’ type after proton expo-sure with 1.1 · 1013 p at KVI Groningen.

the ’normal C’ type APD the proton induced dam-ages are mainly located inside the conversion layerand at the end of the avalanche region. To clar-ify the corresponding situation in case of the ’lowC’ version APD more APDs of that type have tobe irradiated. Nevertheless Fig. 5.17 provides anindication of the locations of the damages similarto those found in case of the normal capacitanceAPDs.

5.1.3.2 Photon Irradiation

The γ-irradiation of the diodes took place at theStrahlenzentrum Giessen, Germany. The APDs wereirradiated using 60Co sources of different dose ratesto reach the envisaged integrated dose of 1012 γs perAPD. To determine the influence of γ exposure onthe APD characteristics basically the same proce-dure was used as described in the former paragraphin case of proton exposure. The irradiation has beenaccomplished for the ’normal C’ version only, dueto the fact that only 5 pieces of the ’low C’ typeAPDs were available at the specific date of mea-surement. Similar to the proton irradiation doneat KVI the photon exposure of the diodes has beendone at T = −25 C. The remeasurements of theAPD characteristics have been done after a storageof the diodes at room temperature in the order ofone month. Similar to the results presented for theproton irradiation mesurement the gain values afterγ exposure are compared to those measured usingan unirradiated diode of the same internal structure


[V]RU0 100 200 300 400 500

M

0

500

1000

1500

2000

2500

3000

3500

Figure 5.18: Determined gain values M of the’normal C’ type APD before (blue) and after (red) ex-posure with a 60Co source.

(Fig. 5.18). Taking into account the determinationof the maximum usable gain of the ’normal C’ typeAPD no obvious gain decrease could be observed(see x-axes of Fig. 5.19), which is in contrast tothe dramatical gain decrease of the diode in termsof proton exposure. Comparing the measured darkcurrent after γ irradiation and one month of stor-age with the values determined for an unirradiateddiode (Fig. 5.19) a nearly complete annealing of thedark current could be observed (see y-axes values).Looking at the gain range between 180 and 300 thedark current behaviour is dramatically changing sothat the required linearity of the dark current de-pending on the internal gain seems to be no longerensured. Detailed studies have to be done in thenear future to determine the influence of surfacedefects caused by photon irradiation in this gainregion. On closer examination of Fig. 5.20 no indi-cation for any kind of bulk damage induced by theaccomplished photon exposure could be observed.This result suggests some kind of surface damageto be cause of the atypical behaviour of the darkcurrent shown in Fig. 5.19.

5.1.3.3 Neutron Irradiation

The neutron irradiation of the quadratic shapedAPDs which has been done until now was exe-cuted at the APDlab Frankfurt using a 241Am −α−Be source. This source generates approximately0.65 · 10−4 neutrons per emitted α particle of thealpha decay of americium via the reaction

4He+9 Be→12 C + n. (5.12)

Taking into account the activity of the employedsource (1.1GBq) a neutron rate of 7.15 · 104n/s

Figure 5.19: Dark current dependence on gain ofthe ’normal C’ type APD before (blue/left) and after(red/right) exposure with 1012 γs.

[nm]λ300 400 500 600 700 800

no

rma

lize

d g

ain

M

30

40

50

60

70

80

90

100

Figure 5.20: M(λ) of ’normal C’ type after photonexposure with an integrated dose of 1012 γs at Strahlen-zentrum Giessen.

could be reached.In contrast to the exposure procedures describedin the former paragraphs the irradiation results de-scribed in this section are based on measurementsdone at T = 20C. The measuring equipment pro-viding a save handling of the neutron source andwhich has to be completely filled with boron sili-cate due to radiation protection reasons is shown inFig. 5.21. As already discussed in cases of protonand photon exposure of the diodes the gain of thetwo tested APD types has decreased after irradia-tion. The curve progression of the measured gaindependence on the applied bias voltage is shownexemplarily for the ’normal C’ type in Fig. 5.22 incomparison to the behaviour observed before expo-sure. The dark current dependence on the deter-mined gain values is shown for both APD types inFig. 5.23 and in Fig. 5.24 respectively. In case ofthe ’normal C’ type the dark current shows nearlyno increase and its linearity depending on M isstill ensured. The converse situation could be ob-


Figure 5.21: Schematic of the apparatus used for neu-tron irradiation of the avalanche diodes including APDhousing und holder for the used neutron source.

[V]RU0 100 200 300 400 500

M

0

500

1000

1500

2000

2500

3000

3500

Figure 5.22: Determined gain values M of the’normal C’ type APD before (blue) and after (red) ex-posure using a 241AmBe source.

served in case of the ’low C’ type APD where themaximum useable gain has dramatically decreasedto a value 50% below the determined one for theunirradiated device. Having a closer look on thewavelength dependence of the gain measured afterneutron exposure lead to the situations shown inFig. 5.25 and Fig. 5.26. For both APD types,’normal C’ and ’low C’, no indication of the exis-tence of any kind of bulk damages could be foundusing the 241Am− α−Be source described above.

To complete the radiation hardness studies of theAPDs an irradiation of the devices using a neutrongenerator located at the FZ Rossendorf providingneutrons with an energy of 14 MeV is foreseen. Thisbeamtime will take place during first half of 2008and will clarify open questions concerning the de-

gain M0 100 200 300 400 500

[n

A]

dI

0

20

40

60

80

100

120

gain M0 50 100 150 200 250 300 350 400 450

[n

A]

dI

0

10

20

30

40

50

60

70

80

Figure 5.23: Dark current dependence on gain ofthe ’normal C’ type APD before (blue/left) and after(red/right) exposure using a 241AmBe source (linearfits are added).

gain M0 20 40 60 80 100 120 140 160 180

[n

A]

dI

0

2

4

6

8

10

12

14

gain M0 20 40 60 80 100 120 140 160 180

[n

A]

dI

0

5

10

15

20

25

30

Figure 5.24: Dark current dependence on gain ofthe ’low C’ type APD before (blue/left) and after(red/right) exposure using a 241AmBe source.

[nm]λ300 400 500 600 700 800

no

rma

lize

d g

ain

M

30

40

50

60

70

80

90

100

Figure 5.25: M(λ) of ’normal C’ type after neutronexposure using a 241AmBe source.

vice radiation hardness due to neutron exposure.


[nm]λ300 400 500 600 700 800

no

rma

lize

d g

ain

M

30

40

50

60

70

80

90

100

Figure 5.26: M(λ) of ’low C’ type after neutron ex-posure using a 241AmBe source.

5.1.4 APD Screening Procedure

To ensure an optimum operation of the calorime-ter all components have to be tested before theirmounting as a part of the EMC. In case of thereadout devices of the EMC barrel several stationsof the APD testing have to be passed before theAPDs will be definitively glued on the rear side ofthe scintillator crystals. The screening procedurefor the LAAPDs described in this section containsthe timeline from delivery of the diodes until theirmounting to the crystals.Due to the stability of one APD production runprovided by Hamamatsu Photonics some propertiesof the APDs will only be measured for control sam-ples and are not foreseen as a part of the APD massscreening:

• the quantum efficiency QE,

• the Excess Noise Factor at an internal gain ofM = 50 and M = 100 at room temperatureand at T = −25 C,

• the capacitance of the APD depending on biasvoltage/gain and

• the gain uniformity of the APD active area sur-face.

To ensure a stable operation of the large amount ofLAAPDs the stability of several parameters duringoperation have to be guaranteed in addition. There-fore a complex screening procedure for the diodesused for the EMC crystal readout has to be con-ceived. The measurements of the APD parametershave to be done for each single diode at room tem-perature (to be comparable to the parameter valuesgiven by Hamamatsu) and at the envisaged opera-tion temperature of T = −25 C. For each APDthe gain-bias voltage characteristic has to be mea-sured as well as the dark current dependence of UR.

From these measurements several APD parameters,which will be stored in a database, will be extracted:

• Bias voltage values for the corresponding envis-aged operation gain (in case of CMS: M = 50)and other gain values,

• Dark current Id for different fixed gain values(e.g. M = 50 and M = 100) and

• the gain variation with varying bias voltageUR: 1/M · dM/dV [ %/V] for different gainvalues.

To accommodate the influence of radiation on allthese parameters an irradiation of all APDs usinga high dose 60Co source will be part of the screen-ing process. After irradiation all APDs have to beannealed (under bias voltage) using an oven oper-ating at a temperature of T = 80 C until the mon-itored dark current of the APDs will reach an equi-librium. After this procedure all APD parameterslisted above have to be measured again and have tobe compared to the values determined before irra-diation.

Due to the application of one preamplifier board(including two channels) per crystal one additionalstep, as part of the screening procedure, has to bedone: Based on the results of the remeasured APDcharacteristics two of them have to be assorted interms of similar dark current and bias voltage valuesat a given gain value for the readout of one leadtungstate crystal of the EMC barrel part.

5.1.5 Mounting Procedure

After all APDs have passed the screening procedureexplained above they have to be mounted on therear side of the lead tungstate crystals. The mount-ing technique has to fulfill three main requirements:

1. The thermal contact between the APD and thecrystal has to be as good as possible to guar-antee an overall temperature stability of thecrystal-APD system down to ∆T = ±0.1 C atT = −25 C.

2. The optical coupling between crystal and APDhas to be done as proper as possible to avoidany kind of light/signal loss in this region. Ad-ditional emphasize should lie on the opticaltemperature dependent properties of the usedcoupling material especially in view of the en-visaged operation temperature of the calorime-ter.


Figure 5.27: PEEK capsule designed to carry two(10× 10) mm2 LAAPDs.

3. The readout diodes have to be as good as pos-sible electrically insulated to avoid the appear-ance of additional noise sources (increase ofdark current, etc.· · · ).

Most of these specifications could be fulfilled by us-ing a special kind of plastic (PEEK) produced byinjection-moulding as carrier for the APDs. Dur-ing the ongoing R&D work a so called ’capsule’built out of PEEK was designed and manufacturedfor the housing of 2 APDs of quadratic shape. Apicture of one of the first prototypes is shown inFig. 5.27. In addition to its function as APD hous-ing a hole for one fibre of the envisaged light pulsersystem (see Sec. 8.2) used for radiation damagemonitoring during the PANDA operation was im-plemented. On the rear side of the capsule spacefor the ASIC preamplifier and for a possibly neededtemperature sensor is allocated (see right side ofFig. 5.27). To avoid scintillation light losses due toa low reflectivity of the capsule material the admix-ture of titanium oxide of different granularity to themoulding process is under investigation. The as-sembled capsule (including light fibre, preamplifier,thermistor and cables) offers a proper adjustment ofthe APDs on each crystal of the EMC barrel partand will be coupled to the scintillator using a spe-cial kind of optical glue between the APD entrancewindow and the rear side of the crystal.

5.2 Vacuum Phototriodes(VPT)

5.2.1 Introduction

For the photon detection in the forward end-cap EMC we have chosen vacuum photo triodes(VPT). A schematic view of a VPT is shown inFig. 5.28. Typical high voltage settings are: cath-ode at −1000 V, the anode on ground and the dyn-ode at −200 V. Photoelectrons emitted from thephotocathode are accelerated towards the anodeconsisting of a fine metal mesh. The electrons pass-ing the anode strike the solid metal dynode, locatedbehind the anode. The secondary electrons ejectedfrom the dynode are accelerated back towards theanode and are collected there. The VPTs have ahigh rate capability and can be produced radiationhard, thus they are matching the requirements ofthe forward endcap EMC.

Mesh AnodePhotocathode

Dynode

Figure 5.28: Schematic diagram of a 22 mm diameterVPT. Typical distances between the photocathode andthe mesh anode are 4.5 mm. The distance between meshanode and dynode is 4.5 mm. The total length is around46 mm and is needed for connections in and outside thetube and for production purposes.

Vacuum phototriodes were previously in operationat the electromagnetic calorimeter endcap of theOPAL [11], DELPHI [12] and CMD-2 [13] exper-iments. Recently radiation hard vacuum phototri-odes were developed for the CMS endcap electro-magnetic calorimeter for operation in a magneticfield of up to 4 T.

5.2.2 Available Types

Vacuum phototriodes produced by three differentcompanies are investigated for PANDA. Samples of


the R2148 from Hamamatsu1 provide a typical gainof 10 and were tested in the lab. The ResearchInstitute Electron2 produced 15000 radiation hardVPTs for the CMS endcap. Presently Photonis3

is developing phototriodes with high quantum effi-ciency photocathodes and high gain (>40 at 1000V). The latter VPTs may produce only about a fac-tor of two less charge per deposited energy in PWOas 1 cm2 sized APDs. Currently we assume onlythe availability of VPTs with a standard quantumefficiency and gain of about 10 as already availablefrom RIE and Hamamatsu.

5.2.3 Characteristics andRequirements

For the forward endcap EMC 4000 phototriodes areneeded. To match the size of crystals a maximumdiameter of 22 mm and to stay within the space as-signed to the detector an overall length of 46 mm isavailable. The specifications are listed in Table 5.4.Parameters of the RIE FEU-189 VPT are listed inTable 5.5. It matches most of the specifications,though further development is needed.

In Fig. 5.29 and Fig. 5.30 the gain is shown as afunction of anode voltage for Hamamatsu R2148and CMS FEU-188 VPT, respectively. Comparedto APDs they exhibit only a small dependence ofthe gain on the high voltage. Typical values areless than 0.1 % per volt. New developments in thematerial design of the dynodes allow the produc-tion of VPTs with gain above 40. Prototypes fromPhotonis are currently investigated. The high gainis favourable since it reduces the noise contributionfor low energy calorimeter signals.

The spectral response of the photocathodes shouldcover a region from 350 nm to 650 nm. Typicalquantum efficiencies of bialkali photocathodes areabove 15 %, much less than the quantum efficien-cies of APDs, which are above 65%. In Fig. 5.31the quantum efficiency of VPTs are compared tothe PWO emission spectrum. An increase of thequantum efficiency would be desirable to improvethe signal over noise ratio. Recently photocathodeswith quantum efficiencies above 30% were producedby Photonis and Hamamatsu [15, 16]. These wouldbe favourable. However, in the current design andsimulations we assume only the availability of stan-dard photocathodes.

The endcap is located close to the door of thesolenoid magnet. At the position of the photode-tectors the magnetic field varies between 0.5 T and1.2 T and has a direction between 0 and 17 degrees

Figure 5.29: Variation of VPT gain as a function ofcathode voltage at a dynode voltage of -200 V for R2148VPT.

Figure 5.30: Variation of VPT gain as a function ofcathode voltage at a dynode voltage of -200 V for CMSVPT.

with respect to the axial direction of the VPTs. Thegain variation under a magnetic field depends on themesh size of the anode. In Fig. 5.32 the variationof the anode response is shown as a function of an-gle to the axial field for the VPTs produced for theCMS experiment.

The dependence of the VPT quantum efficiency andgain on the temperature has been measured to besmall and can be completely neglected compared

1. Hamamatsu Photonics, Electron Tube Division ExportSales Dept. 314-5, Shimokanzo, Iwata City, Shizuoka Pref.,438-0193, Japan

2. National Research Institute Electron (RIE), AveM. Toreza 68, 194223 St. Petersburg, Russia

3. PHOTONIS France SAS, Avenue Roger Roncier, 19100Brive La Gaillarde, France


Parameter ValueExternal diameter max. 22 mmOverall length about 46 mmDynode number 1 or 2Gain 10 to 30 or moreRegion of max. spectral response 420 nm (PbWO4)Magnetic field max 1.2 T

in 0-17 degrees in axial direction of VPTPhotocathode useful diameter 16-20 mmQuantum efficiency at 430 nm 20% or higherRadiation hardness 10 Gy per annumOperational temperature range -30C to 35CRate capability above 500 kHz

Table 5.4: Specifications for the PANDA VPTs.

Parameter ValueExternal diameter 21 mmPhotocathode useful diameter 15 mmOverall length 41 mmOperating bias voltage: Vu, Vd(Vc = 0V) 1000 V, 800 VDark current 1 - 10 nA(dM/dV )/M < 0.1%/V(dM/dT )/M < 0.1%/CQuantum efficiency at 430 nm > 15%Range of spectral response 300 – 620 nmEffective gain (B=0 T) 12Effective gain (B=4 T, Θ = 0) 6Effective gain (B=4 T) Θ = 20) 8Anode pulse rise time 1.5 nsExcess noise factor F at B = 0 2.0 – 2.5Excess noise factor F at B =1 – 4 T, Θ = 20 2.2 – 2.6

Table 5.5: Parameters of the RIE FEU-189 VPT [14].

to the temperature dependence of the light yield ofPWO. VPTs can be operated like photomultipliersin temperature ranges down to −30 C.

Unlike PIN diodes, where energetic charged parti-cles may produce electron-hole pairs in the silicon(nuclear counter effect), photomultipliers and VPTsare not susceptible to charged particles, due to thethinness of the entrance window and photocathode.

The capacitance of the VPT is small compared tothe cables connecting to the preamplifiers, such thatlow noise values can be reached. Typical values areabout 22 pF or less. Therefore the cables to thepreamplifiers must be kept as short as possible. Inthe design it is foreseen to place the preamplifierswithin less than 10 cm to the VPTs (see Chap. 6).Also the dark current of 1 nA is very small compared

to the maximum accepted 50 nA of the LAAPD.

5.2.4 Testing

5.2.4.1 Radiation Hardness

Radiation may change the response of the anodeand the excess noise factor. Damages resulting insevere modifications of the response would deteri-orate the energy resolution. Extensive studies ofthe radiation hardness were performed by the CMSexperiment for the FEU-188 VPTs [18]. It wasshown that under gamma irradiation the VPT an-ode response is fully determined by the loss of theVPT faceplate transmittance (Fig. 5.33). The CMSVPTs use the UV glass type US-49A. The VPTs


Figure 5.31: Quantum efficiencies of a caesium an-timony photocathode in a Hamamatsu R5189 tetrodeand a RIE triode compared with the emission spectrumof PWO [1].

Figure 5.32: Variation of anode response for constantpulsed LED illumination as a function of the VPT angleto the axial field of 1.8 T for the CMS type VPT [17].

were irradiated in Russia with 20 kGy at a dose rateof 0.24 kGy/h. The decrease of the VPT anode sig-nal did not exceed 4% at 20 kGy (Fig. 5.34) [18]. AtPANDA we do not expect gamma dose rates above0.2 kGy.

Radiation hardness tests with reactor neutrons havea high level of gamma background and activate theVPT material by a thermal neutron flux. Thereforethey are difficult to evaluate. To circumvent thisproblem the FEU-188 VPTs were irradiated at aneutron generator providing En = 14 MeV. Theanode response of VPTs with cerium-doped glasswas found independent of the neutron fluence upto 2.4 · 1015 n/cm2 within the experimental error of±5 % (Fig. 5.35).

Figure 5.33: Light transmittance spectra of differentUV glasses produced in Russia before and after 20 kGy60Co gamma irradiation and the emission spectrum ofPWO. Dose rate is 0.24],kGy/h [18].

Figure 5.34: Relative US-49C (squares) and US-49A(circles) faceplate light transmission in the range of thePWO emission spectrum and relative anode response ofthe VPT as a function of the gamma dose at B = 0 T(crosses) [18].

5.2.4.2 Rate Studies

The particle rate in the forward endcap EMC canbe as high as 500 kHz. Signals of the VPT are asshort as the PWO scintillation pulses. To checkthe rate capabilities of the VPTs a LED pulser pro-ducing 445 nm pulses not shorter than the PWOscintillation pulses was built. The VPTs of the dif-ferent producers will be tested with the light pulserat varying rates up to 1 MHz.


Figure 5.35: Relative anode response of VPT FEU-188 with faceplates from cerium-doped glass C1-96 ver-sus the neutron fluence (En = 14 MeV) [18].

5.2.5 Screening Procedure

After delivery of the VPTs they will undergo ascreening procedure to check relevant parameters.The results of the test of each of the 4000 VPTswill be stored in a database. Basic dimensions ofthe VPTs and the connectors will be checked firstafter the arrival. Following a burn-in procedure theanode leakage current and the gain times the quan-tum efficiency will be tested with an LED pulsersystem at wavelengths of 455 nm and 470 nm. Thestability of the gain will be checked by varying thefrequency of the LED pulser system between 500kHz and 0 kHz and back to 500 kHz. Finally thevariation of gain between 0 T and 1 T axial magneticfield and the excess noise factor will be measured.

References

[1] CMS: The electromagnetic calorimeter.Technical design report, Technical report,1997, CERN-LHCC-97-33.

[2] Panda Technical Progress Report, Technicalreport, 2005.

[3] D. Renker, Status of the avalanchephotodiodes for the CMS ElectromagneticCalorimeter, Annecy, 2000, Prepared for 9thInternational Conference on Calorimetry inParticle Physics (Calor 2000).

[4] CMS Conference Report: Progress onAvalanche Photodiodes as photon detectorsfor PbWO4 crystals in the CMS experiment,Technical report, 1997.

[5] B. Patel et al., (1999), Prepared for 5thWorkshop on Electronics for the LHCExperiments (LEB 99), Snowmass.

[6] K. Ueno et al., arXiv:physics/9704013v1(1997).

[7] S. M. Sze, Physics of Semiconductor Devices,1981.

[8] M. Kneifel and D. Rabe,Halbleiterbauelemente, 1993.

[9] T. Kirn et al., Nucl. Instrum. Meth. A387,202 (1997).

[10] M. Moll, Radiation Damage in SiliconDetectors, 1999, DESY-THESIS-1999-040.

[11] M. Akrawy et al., Nucl. Instrum. Meth.A290, 76 (1990).

[12] M. Bonesini et al., Nucl. Instrum. Meth.A387, 60 (1997).

[13] P. M. Bes’chastnov et al., Nucl. Instrum.Meth. A342, 477 (1994).

[14] N. A. Bajanov et al., Fine-MeshPhotodetectors for CMS EndacapElectromagnetic calorimeter, Technicalreport, CMS NOTE 1998/080, 1998.

[15] R. Mirzoyan, M. Laatiaoui, and M. Teshima,Nucl. Instrum. Meth. A567, 230 (2006).

[16] J.-C. Lefort, UBA/SBA PMTs -Hamamatsu’s “Bialkali Climbing Party” hasnow reached 43% QE, Technical report,Hamamatsu News 2008, Vol. 1, 2008.


[17] K. W. Bell et al., Nucl. Instrum. Meth.A504, 255 (2003).

[18] Y. I. Gusev et al., Nucl. Instrum. Meth.A535, 511 (2004).

85

6 Electronics

The PANDA Electromagnetic Calorimeter (EMC)will consist of PbWO4 (PWO-II) crystals arrangedin the cylindrical barrel EMC with 11360 crystals,the forward endcap EMC with 3600 crystals andthe backward endcap EMC with 592 crystals. Thepurpose of this detector is an almost full coverage,as far as the acceptance of the forward spectrometerallows, of the final state phase space for photons andelectrons. Since one of the physics goals is e.g. pre-cision spectroscopy of the charmonium spectrum,the low-energy photon threshold should be around10 MeV, which requires the threshold for individualcrystals to be about 3 MeV and correspondingly lownoise levels of 1 MeV. Neutral decays of charmedmesons require the detection of a maximum pho-ton energy deposition of 12 GeV per crystal at thegiven maximum beam energy of the HESR. Theserequirements dictate a dynamic range of 12000 forthe readout electronics.

The placement of the calorimeter inside the 2 Tsolenoidal magnetic field requires photo sensorswhich provide a stable gain in strong magneticfields. Therefore a Large Area Avalanche PhotoDi-ode (LAAPD) has been developed for the PANDAEMC and will be employed in the barrel EMC partwhere typical event rates of 10 kHz and maximum100 kHz are expected (see Sec. 4.1.2). Because ofhigher rates (up to 500 kHz) vacuum photo triodes(VPTs) have been chosen for the forward endcapEMC. The photo sensors are directly attached tothe end faces of the individual crystals and thepreamplifier has to be placed as close as possibleinside the calorimeter volume for optimum perfor-mance and minimum space requirements. The read-out of small and compact subarrays of crystals re-quires very small preamplifier geometries. In orderto gain maximum light output from the PWO-IIcrystals, the calorimeter volume will be cooled to-25C. Efficient cooling thus requires low powerconsumption electronics to be employed in combi-nation with extremely low-noise performance.

To minimize the input capacitance and pickupnoise, the analogue front-end electronics is placednear the APD and kept at the same temperatureas envisaged for the PWO-II crystals, namely at -25C. The low-temperature environment improvesthe noise performance of the analogue circuits and,at the same time, constrains the power consump-tion of the analogue front-end electronics. ThePANDA collaboration has developed two comple-

mentary low-noise and low-power (LNP) charge-sensitive preamplifier-shaper (LNP-P) circuits.

First, a LNP design was developed based on discretecomponents, utilizing a low-noise J-FET transistor.The circuit achieves a very good noise performanceusing signal shaping with a peaking time of 650 ns.Second, a state-of-the-art CMOS ASIC was devel-oped, which achieves a similar noise performancewith a shorter peaking time of 250 ns. The advan-tage of the CMOS ASIC is the very low power con-sumption. Both designs are complementary sincethe preamplifiers, based on discrete components,will be used for the readout of the forward end-cap EMC for which we expect a maximum rate percrystal of 500 kHz. Such an approach minimizes theoverall power consumption and keeps the probabil-ity for pileup events at a moderate level well below1 %.

The subsequent digitization stage will be placed asclose as possible to the calorimeter volume but out-side the low-temperature area. This allows signaltransfer from the front-end over short distances byflat cables with low thermal budget. Optical linkswill be employed to transfer digitized data via amultiplexing stage to the compute node outside theexperimental setup.

In the following paragraphs the requirements andperformance of the various stages of the readoutchain will be discussed: the general readout scheme,the preamplifiers, the digitizer modules, the multi-plexer stage and, finally, the detector control systemsupervising the performance of the whole detectionchain.

6.1 General EMC ReadoutScheme

The readout of the electromagnetic calorimeter isbased on the fast, continuous digitization of the am-plified signal-shape response of LAAPD and VPTphoto sensors to the light output of PWO-II crys-tals. The readout chain will consist of extremelylow-noise front-end electronics, digitizer modulesand data multiplexer (see Fig. 6.1 ).

The digitizer modules are located at a distance of20–30 cm and 90–100 cm for the barrel EMC andthe forward endcap EMC, respectively, away fromthe analogue circuits and outside the cold volume.


Optical LinksCSA/Shaper

DigitizerCSA/Shaper

‐25 ˚CInside Calorimeter Volume

DATAMultiplexer

DigitizermoduleDigitizer

module

CSA/Shaper

DigitizermoduleAPD

CNAPD

CSA/Shaper

Digitizermodule

CSA/Shaper

DigitizerCSA/Shaper

Di iti DATAMultiplexer

moduleDigitizermoduleAPD

APD

1

Figure 6.1: The readout chain of the Electromagnetic Calorimeter.

The digitizers consist of high-frequency, low-powerpipelined ADC chips, which continuously samplethe amplified and shaped signals. The sampling isfollowed by the digital logic, which processes time-discrete digital values, detects hits and forwards hit-related information to the multiplexer module viaoptical fibers. At this step the detection of clustersof energy deposition can be efficiently implemented.A cluster seed requires an energy deposition of typ-ically at least 10 MeV with a number of neighbor-ing crystals surpassing the single-crystal thresholdof typically 3 MeV. The multiplexer modules willbe located in the DAQ hut and they perform ad-vanced signal processing to extract amplitude andsignal-time information.

The front-end electronics of the barrel EMC is lo-cated inside the solenoid magnet where any accessfor maintenance or repair is limited to shutdown pe-riods of the HESR, expected to occur once a year.This condition requires to implement a redundancyin the system architecture. One of the most im-portant decisions, that has been taken by the col-laboration, is to equip every EMC crystal with two

APDs. Apart from redundancy, the system withtwo independent readout channels offers a signifi-cantly (max.

√2) improved signal to noise ratio

and a lower effective threshold level.

6.2 Preamplifier and Shaperfor barrel EMCAPD-readout

A low-noise and low-power charge preamplifierASIC (APFEL) was designed and developed for thereadout of the LAAPD for the PANDA EMC. TwoLAAPDs with an active area of 7 × 14 mm2 eachare attached to the end face of the lead tungstatescintillating crystals (PWO-II) which have a typi-cal geometry of (200 × 27 × 27) mm3. Contrary toa photomultiplier, the LAAPDs can also be oper-ated in a strong magnetic field. In the barrel EMCthe LAAPDs act as photo detectors converting thescintillating light to an electrical charge signal. Thepreamplifier linearly converts the charge signal from


the LAAPD to a voltage pulse which is transmittedto the subsequent electronics.

6.2.1 Requirements andSpecifications

6.2.1.1 Power Consumption

Since the complete barrel EMC, together with theAPDs and the preamplifiers, will be cooled to lowtemperatures (-25C) to increase the light-yield ofthe PWO-II crystals, the power dissipation of thepreamplifier has to be minimized. Low power dis-sipation leads to a smaller cooling unit and thin-ner cooling tubes; it also helps to achieve a uniformtemperature distribution over the length of the crys-tals.

6.2.1.2 Noise

To reach the required low detection threshold ofabout 1 MeV, the noise performance of the pream-plifier is crucial. The newly developed rectangu-lar LAAPD from Hamamatsu (Type S8664-1010)has an active area of 14 × 7 mm2 resulting in aquite high detector capacitance which requires alow-noise charge preamplifier. The total outputnoise is a combination of the preamplifier noise andthe noise generated by the dark current flowingthrough the APD. By cooling the APD to -25Cthe dark current is reduced by a factor of aboutten, with respect to room temperature. Using alow-leakage LAAPD at low temperature, the chargepreamplifier is the dominating noise source due tothe relatively high detector capacitance of around270 pF. The noise floor of the APFEL ASIC at-20C, loaded with an input capacitance of 270 pF,has a typical equivalent noise charge (ENC) of 4150e− (rms) (see Sec. 6.2.2.2).

Investigations of PWO-II light production (seeFig. 4.19) yield on average 90 photoelectrons perMeV, measured at -25C with a photomultipliertube of 18% quantum efficiency and an integrationgate width of 300 ns. This value results in 500 pho-tons/MeV at the end face of the cooled (-25C)PWO-II crystal. With the average back face crosssection of barrel crystals of 745 mm2 (see Fig. 7.3)we obtain 66 photons/MeV on the active area of7 × 14 mm2 of a single rectangular LAAPD. Thequantum-efficiency of the LAAPD is around 70%for the scintillating light of the PWO-II crystals andthe voltage biased LAAPD will be operated at aninternal gain M = 100. Applying these numbers, aprimary photon with the energy of 1 MeV induces

an input charge of 0.74 fC (4620 e−) to the pream-plifier. Thus, an ENC of 4150 e− (rms) correspondsto an energy noise level of about 0.9 MeV (rms).

6.2.1.3 Event Rate

To cope with the expected event rates in the barrelEMC of maximum 100 kHz per crystal, a feedbacktime constant has to be chosen which is a trade-off between noise performance and pile-up problem-atic. The preamplifier is designed for an event rateup to 350 kHz.

6.2.1.4 Bias Voltage

At the maximum event rate of 100 kHz with themaximum expected photon energy deposition percrystal of 12 GeV (8.9 pC from the LAAPD) a meancurrent of 890 nA is flowing through the LAAPD.Under these extreme conditions, the voltage dropover the low-pass filter for the APD bias voltagegets relevant, since the internal gain (M = 100) ofthe LAAPD varies by about 3%/V. This means,that the measured energy will be dependent on theevent rate. To minimize this effect one can de-sign the low-pass filter with low series resistanceresulting in a degraded noise performance of thepreamplifier and the need for larger filter capaci-tors. Eventually, the various regions of the barrelEMC with different event rates could be equippedwith adapted low-pass filters to obtain an optimumnoise performance combined with low rate/energydependence. Another solution is that the APD bias-voltage supply sources a voltage which is correctedon the output current (the more current the morevoltage). This means, that the output resistanceof the bias-voltage supply is negative and there-fore compensates the series resistance of the low-pass filter. The advantage is a better noise perfor-mance of the preamplifier due to the highly resis-tive APD bias-voltage filter. This solution woulddemand a more complex bias-voltage supply sys-tem with a sophisticated overvoltage and overloadcontrol. Each LAAPD has its own measured biasvoltage where it reaches the nominal internal gainof M = 100. To reduce the number of APD bias-voltage channels, it is foreseen to group LAAPDswith similar bias-voltages. Since this grouping willbe very local, in regions where similar event ratesare expected, the solution with the negative resis-tance bias-supply could still be feasible.


6.2.1.5 FADC Readout

Readout with a flash ADC (FADC) does not need adedicated timing amplifier. Only a good anti alias-ing low-pass filter/amplifier has to be implementedin front of the FADC. The low-pass filter/amplifiercan be safely installed at distances below 70 cm fromthe preamplifier directly at the FADC in the room-temperature environment. The cut-off frequency ofthat filter is given by the sampling frequency of theFADC divided by 2.5; this prevents aliasing effectsdue to the sampling. To determine the energy witha high signal to noise ratio, the digitized values areprocessed by a digital shaping filter followed by adigital peak determination. It could be possibleto choose different digital energy shaping filters forthe different regions of the barrel EMC: A shortpeaking-time for the small angles in forward direc-tion and the region orthogonal to the target posi-tion, where high event rates are expected. For otherregions of the detector, where lower event rates areestimated, the peaking-time could be larger andtherefore result in a better energy resolution. Evena rate-dependent automatic adaptation of the dig-ital energy shaping filter could be imagined. Thetiming information is extracted by processing thedigitized values similarly to the traditional signalchain with a fast digital timing filter followed bya digitally implemented CFD. Measurements haveshown, that a timing resolution well below 1 ns canbe reached with that regime. One has to point out,that the processing for the energy- and timing infor-mation extraction must be performed in real time.This can be implemented by using programmabledigital signal processors (DSPs) or completely inhardware by using field programmable gate arrays(FPGAs). The needed signal throughput at thehigh sampling frequency (80 MHz) combined withthe complex algorithm will result in a remarkablepower dissipation.

6.2.2 Integrated CircuitDevelopment

The high compactness of the electromagneticcalorimeter in connection with the required tem-perature homogeneity at low operation tempera-ture puts high demands on space and power con-sumption. It appears almost mandatory to inte-grate the preamplifier and shaper on a single chipto fulfill these demands. For this reason an Appli-cation Specific Integrated Circuit (ASIC ”APFEL”)for the readout of Large Area APDs as foreseen atthe PANDA EMC was developed.

Figure 6.2: Schematic diagram of the folded cascodefront end.

6.2.2.1 Circuit Description

For the front-end amplifier design a single endedfolded cascode architecture was chosen as it is fre-quently described in literature [1, 2, 3] and shownin Fig. 6.2. This architecture combines a high openloop gain with a large output swing.

A signal at the input transistor T1 effects a currentchange of I1. As the current I = I1 + I2 is keptconstant by the current sink T4 also I2 changes bythe same value. This variation creates a voltagedrop at the output node of T2. The cascodetransistor T3 separates the input transistor fromthe output node. To get the best gain and stabilitythe ratio of the currents I1 and I2 in Fig. 6.2 shouldbe in the order of I1/I2 = 9.

The signal charge is integrated on a capacitance Cwhich has to be discharged by a resistor R to pre-vent the preamplifier from saturation. The integra-tion capacitance is not across the feedback resistoras usual but between the input and the dominantpole node realizing a Miller compensation. This re-sults in a reduction of wideband output noise [1]and minimizes the sensitivity to variations in de-tector capacitance [4].

For the preamplifier design the noise considerationplayed the leading part. The parts in equivalent in-put noise of transistors T2 and T4 scale with theirtransconductances gm2/gm1 and gm4/gm1, respec-tively, so they can be minimized by a dedicatedchoice of parameters [5]. The main noise contrib-utor is the input transistor T1. From noise theoryone can see, that the drain source current IDS andthe transistor width W are the free parameters tocontrol the transistor noise which decreases with in-creasing IDS and W . Since at W ≈ 104 µm the


Figure 6.3: Principle diagram of the self biasing feed-back network.

noise reaches a minimum, W = 12.8 mm was cho-sen. A tradeoff between noise performance and thepower consumption led to a current of IDS = 2 mA.

Another tradeoff is the choice of the feedback resis-tor to balance low parallel noise (large R) and thehit rate the preamplifier has to cope with (small R).The capacitance C is given by the maximum of theinput charge the preamplifier has to deal with.

To realize the mandatory high resistivity for thefeedback resistor R1, a transistor operating in thesubthreshold region is used. Concerning the tem-perature and process independence, a self biasingtechnology as described by O’Connor et al. in [3] isrealized. As shown in Fig. 6.3 a MOS transistor T3in diode connection is used together with a currentsink to generate the gate source voltage VGS of thefeedback transistor T1. The source potential of thisMOS diode is fixed by a downscaled version A3 ofthe preamplifier circuit A1.

The pole which is introduced by the output resis-tance roT1 of T1 and the capacitance C1 is compen-sated by transistor T2 in parallel connection to thedifferentiation capacitance C2. This way a zero isintroduced into the transfer function. By choosingthe time constants τ2 = τ1 with τ1 = C1 · roT1 andτ2 = C2 · roT2 any undershoot in the pulse shape iseliminated.

The gate of T2 is connected to the same potential

as the gate of T1. To ensure that the drain andsource potentials of T1, T2 and T3 are equal, theamplifiers A2 and A3 are downscaled versions of theinput amplifier A1.

The amplifier A2 is used as the first stage of a3rd order integrator. With a capacitive feedbacka 1st order low pass filter with a time constant ofτ = 90 ns is realized. At this point the signal is splitinto two paths. To add two more poles to the trans-fer function, a 2nd order integrator stage based ona fully differential operational amplifier follows oneach path.

On one of these paths an amplification factor of 16 isrealised so this signal path is optimised to measureat the low energy part of the dynamic range witha minimized influence of pick up noise on the con-nection between preamplifier and ADC. The otherpath has no additional amplification so it covers thewhole upper part of the dynamic range.

After the shaper build up by the differentiator stageand the integrators, detector pulses have a semiGaussian pulse shape with a peaking time of 250 ns.The shapers are followed by output drivers with adriving capability of 10 pF parallel to 20 kΩ. Forshaper operation two reference voltages are needed.To guarantee the full dynamic range over a temper-ature range from −30C to +30C these referencevoltages have to be adjusted. An adjustable volt-age reference based on two 10 bit digital to analogueconverters is implemented on the ASIC to providethis functionality.

A charge injection unit for each channel is imple-mented on the chip which gives the possibility toinject a defined amount of charge into the pream-plifier input for readout electronics monitoring. Theamount of charge can be programmed in four steps.For programming of the adjustable voltage refer-ences as well as for programming and triggering ofthe charge injection a serial interface to a two-wireserial bus is implemented on the integrated circuit.Data transfer has to follow a specific bus protocolwhich was defined for this circuit. Each transferconsists of 20 bits. After a start signature the firsteight bits are used as an address to select a readoutchip. The next two bits are used to select one offour internal data registers into which the following10 data bits are written.

To avoid introduction of additional noise by sub-strate coupling from a running digital logic, thereis no continuous clock signal for the digital logic.Receiving data and latching into the internal regis-ters is triggered by the external serial clock signal.After data transmission the clock line stays on high


Figure 6.4: Left side: Overall block diagram of the preamplifier and shaper ASIC. Right side: Photograph ofthe prototype preamplifier ASIC.

level and the on-chip digital logic is inactive. Soit neither produces any additional noise nor it con-sumes power.

An overall block diagram of the integrated pream-plifier circuit is shown in Fig. 6.4 on the left side.The choice of technology was driven by the fact thatlarger feature size technologies have the advantageof higher core voltages, which affects directly thedynamic range. On the other hand, the noise per-formance benefits from a smaller feature size onlyon a minor level. Therefore, for the preamplifier de-sign a 350 nm technology from AMS1 was chosen,which provides a sufficient high core voltage andwhich is widely used e.g. in automotive industry.Thus, we may expect that this technology is avail-able with a long term perspective. The prototypeshown on the right side of Fig. 6.4 was producedin 2007. Since summer 2007 several measurementshave been done to specify the ASIC prototype.

6.2.2.2 Prototype Performance

For specifying the preamplifier prototype over atemperature range from −20C to +20C theASIC was mounted on a printed circuit board (seeFig. 6.5) which can be cooled by a Peltier element.To avoid condensation of air humidity the setupis placed in an evacuated chamber. For measure-ments a voltage step generated by an AWG 510on a well defined capacitor is used as charge in-

jection. A second well defined capacitor connectedparallel to the amplifier input simulates the detec-tor capacitance. The output signals are monitoredwith a digital oscilloscope type DPO 7254. Serialprogramming is done with a data timing generatorDTG 5078. For the temperatures −20C, −10C,+10C and +20C dynamic range, gain and outputnoise voltage were measured so that the equivalentinput noise charge could be calculated. The resultsfor four different detector capacitances are plottedin Fig. 6.6. As expected, the noise increases linearlywith the detector capacitance. Also the tempera-ture dependence is clearly visible.

A linear fit indicates that the slope SENC = (5.73±0.5) e−/pF of the equivalent noise charge is almostindependent of temperature. The constant term in-creases linearly with temperature with a tempera-ture coefficient of ST = 26.15 e−/C.

For an operating temperature of −20C the mea-sured noise of the preamplifier prototype is

ENC = [2610± 103 + (5.69± 0.3) · CDet] e−

so with a detector capacitance of CDet = 270 pFa noise of ENC = (4146 ± 131) e− (rms) can beexpected. This value corresponds (see Sec. 6.2.1.2)to an energy of 0.9± 0.02 MeV.

Fig. 6.7 shows the measured characteristics of theintegrated preamplifier for the high and the low am-

1. Austria Mikrosysteme AG


Figure 6.5: Test PCB with glued and bonded pream-plifier ASIC.

Figure 6.6: Measured noise values of the preamplifierprototype.

plification path. The measurement covers a rangefrom 10 fC to 10 pC input charge. Both paths showan excellent linear behavior with an overlappingrange up to 200 fC. The -1 dB compression pointof the low amplification path can be determined at7.84 pC. Together with the measured noise of 0.66fC this leads to a covered dynamic range of 11900.

In Fig. 6.8 the dependence of power consumptionon the temperature is shown. For a temperature ofT = +20C the power consumption for one channelamounts to 10 mW for the charge sensitive ampli-fier, to 15 mW for the shaper stages and to 17 mWfor the buffers. In addition there are about 3 mW

0.01

0.1

1

10

0.001 0.01 0.1 1 10 100

Out

put v

olta

ge [V

]

Input charge [pC]

Preamplifier Dynamic Range

"messung001_1-20_low4_4pF_gain.txt" using ($1/9.3*4.4):3"messung001_1-20_high48pF_gain.txt" using ($1/9.3*48):3

Figure 6.7: Amplification characteristics of the pream-plifier at −20C.

Figure 6.8: Power consumption for one chip with twochannels in dependence of temperature.

for the bias circuit and the digital part. So theoverall power consumption for one chip with twochannels is 90 mW at +20C. At the temperatureof −20C the power consumption slightly increasesto 104 mW for one chip with two channels. Neithersimulations nor measurements revealed any signifi-cant dependence of the power consumption on theevent rate for an event rate up to 350 kHz.

A compilation of the measured results is given inTable 6.1.


Parameter

ENC (-20C, 270 pF) 4146 ± 131 e−

Max. input charge 7.84 pC

Dynamic Range 11900

Max. Event rate 350 kHz

Peaking Time 250 ns

Power Consumption 104 mW

(2 channel, -20C)

Table 6.1: Measured preamplifier parameters.

6.3 Preamplifier and Shaperfor forward endcap EMCVPT-readout

6.3.1 Requirements andSpecifications

A discrete charge preamplifier, the Low Noise /Low Power Charge Preamplifier (LNP-P) has beendeveloped in first instance for the LAAPD read-out of the barrel EMC and was implemented inthe barrel EMC prototype detector. It has an ex-cellent noise performance in combination with lowpower consumption. This preamplifier has been fur-ther developed and adapted for the readout of Vac-uum Photo Triodes (VPTs). VPTs will be used asphoto detectors in the forward endcap EMC. TheVPT is attached to the end face (26 × 26 mm2) ofthe lead tungstate scintillating crystals (PWO-II)which have a length of 200 mm. Due to the higherevent rates and large photon energies in the for-ward endcap EMC with respect to the barrel EMC,the APDs would suffer from radiation damage (in-creased noise) and from the nuclear counter effect.Therefore, APDs are not suitable at that positionof the EMC. The VPT is a single-stage photomul-tiplier with only one dynode which can also be op-erated in a strong magnetic field without loosingsignificantly in gain. The VPT translates the scin-tillating light of the PWO-II crystals into an electri-cal charge which is linearly converted by the LNP-P to a positive voltage pulse; this output pulse isthen transmitted via a 50 Ω line to the subsequentelectronics. The low capacitance of the VPT fa-vors the development of a faster low-noise pream-plifier with discrete components which is suitablefor the high-rate environment. For the following de-sign considerations, we adopt the parameters of theVPT type RIE-FEU-188 used in the CMS ECAL. Anew VPT with a significantly higher quantum effi-ciency (> 30%) combined with a larger internal gain(> 40) is under development and will be produced

by the company Photonis. By using this new VPT,the energy noise level will be reduced remarkably.

6.3.1.1 Power Consumption

Since the complete forward endcap EMC, includ-ing VPTs and preamplifiers, will be cooled to lowtemperatures (-25C) to increase the light-yield ofthe PWO-II crystals, the power dissipation of thepreamplifier has to be minimized. Low power dis-sipation leads to a smaller cooling unit and thin-ner cooling tubes; it also helps to achieve a uniformtemperature distribution over the length of the crys-tals. The LNP-P has a quiescent power consump-tion of 45 mW. The power dissipation is dependenton the event rate and the photon energy; Fig. 6.9shows the measured power dissipation on the LNP-P versus count-rate in the worst case of maximumoutput amplitude. Since the high rates are pre-dominantly occurring at lower energies, a reason-able maximum power consumption of 90 mW canbe presumed.

6.3.1.2 Noise

To reach the required low detection threshold ofonly several MeV, the noise performance of thepreamplifier is crucial. The VPT has an outsidediameter of 22 mm and a minimum photocathodediameter of 16 mm (see Table 5.4), resulting inan active area of ca. 200 mm2. This area is thesame as the combined active area of two rectangu-lar 7 × 14 mm2 LAAPD. The VPT anode capaci-tance is around 22 pF which is more than 10 timeslower than the capacitance of the LAAPD; this re-sults in a much lower noise from the LNP-P. Thus,the shielded cable between the VPT and the LNP-P has a significant impact on the total detectorcapacitance; it must be kept as short as possible.The dark current of the VPT (1 nA) is significantlylower than the one from the LAAPD (50 nA), bothmeasured at room temperature. The quantum ef-ficiency of the standard available VPT type RIE-FEU-188 is about 20%, compared to 70% of theLAAPD. Further, the internal gain of the VPT isaround ten, which is five times lower than that ofthe LAAPD. The noise floor of the LNP-P at -25Cloaded with an input capacitance of 22 pF, has atypical equivalent noise charge (ENC) of 235 e−

(rms). This is measured with an ORTEC450 shap-ing filter/amplifier with a peaking-time of 650 ns.Because the VPT has almost no dark current, thenoise is not increased due to the leakage current ofthat photo detector.


Figure 6.9: Power dissipation as function of continuous event rate for the LNP preamplifier for the VPT readout.

As already discussed in Sec. 6.2.1.2, measurementsof PWO light production yield 500 photons/MeVat the end face (26 × 26 mm2 for the forward end-cap EMC) of the cooled (-25C) PWO-II crystal.This results in 150 photons/MeV on the active areaof the VPT. By applying the quantum efficiencyand the internal gain of the VPT, a primary photonwith the energy of 1 MeV induces an input chargeof 48 aC (300 e−) to the preamplifier. So an ENCof 235 e−(rms) corresponds to an energy noise levelof 0.78 MeV (rms). This is about the same levelas achieved under the same conditions in the barrelEMC with the LAAPD readout (0.9 MeV (rms)).Therefore, the signal to noise is in the same orderwhen using a VPT or an LAAPD for the readoutof a PWO-II crystal. The noise level is increasedas the shaping time is decreased. Shorter shapingtimes are mandatory to cope with the expected highevent rates in the forward endcap EMC. By decreas-ing the peaking-time from 650 ns (reference values)to 200 ns, the noise level is raised by around 25%.So, the noise floor with the more realistic shapingusing a peaking-time of 200 ns corresponds to about1 MeV (rms) for the presently available VPT. As al-

ready mentioned in Sec. 6.3.1, a new VPT with asignificantly higher quantum efficiency (ca. 30%)combined with a larger internal gain (ca. 40) isunder development and will be produced by Photo-nis. By applying these values for quantum efficiencyand the internal gain of the VPT, a primary photonwith the energy of 1 MeV induces an input charge of290 aC (1800 e−) to the preamplifier. In this casethe ENC of 235 e−(rms) corresponds to a signifi-cantly reduced energy noise level of 160 keV (rms)for a peaking-time of 200 ns.

6.3.1.3 Event Rate

The expected event rate in the forward endcapEMC is maximum 500 kHz per crystal. The LNP-Phas a feedback time constant of 25 µs. This feed-back time constant is a trade-off between noise per-formance and pile-up problematic. Reducing thefeedback time constant by a factor of two will in-crease the noise by about 10%. For a single pulse (orvery low rates) the LNP-P accepts an input chargeof up to 4 pC; for a continuous event rate of 500 kHzan input charge of up to 8 pC is allowed. This dis-


crepancy is due to the following reason: A singleoutput pulse starts from zero output voltage andis limited by the positive supply voltage (+6 V) ofthe LNP-P. At high continuous event rates the out-put pulses will swing between the negative (-6 V)and the positive (+6 V) supply voltage; thereforethe maximum input charge is doubled. If a 500 kHzevent rate is applied abruptly (burst) to the LNP-Pit takes around one second until a continuous inputcharge of up to 8 pC is allowed. During that transi-tion period, a maximum input charge of 0.3 pC canbe handled. With this charge restriction, the outputvoltage of the preamplifier stays always in the linearrange and is never limited by the power supply volt-ages. Nevertheless, the electronics after the pream-plifier has to perform a good base-line correction,because at higher rates it is likely that one pulse sitson the trailing edge of the previous one. If a chargeof 48 aC/MeV (290 aC/MeV) is assumed from theVPT (see Sec. 6.3.1.2) the maximum expected pho-ton energy deposition of 12 GeV per crystal resultsin an input charge of 0.58 pC (3.5 pC). Under theexpected operational conditions with high rates pre-dominantly occurring at low energy, the LNP-P willnot be restricted by pile-up even with the relativelylong feedback time constant of 25µs.

6.3.1.4 Bias Voltage

Because the anode of the VPT will be referencedto ground by the LNP-P, the photocathode (PC)and the dynode (DY) must be biased with negativehigh voltages (HV). To have the input of the pream-plifier referenced to ground has the advantage thatany noise on the HV supply is not directly coupledinto the charge sensitive input. The typical biasvoltages for the VPT type RIE-FEU-188 are: Pho-tocathode: VPC = -1000 V; Dynode: VDY = -250V. Even if these bias voltages do not directly coupleto the charge input of the preamplifier, they haveto be cleaned from external noise by an efficient lowpass (LP) filter before they are wired to the VPT.Also, if all the VPT are biased with the same twohigh voltages, each VPT must have its own LP fil-ter to prevent crosstalk. These LP filters for thetwo negative bias voltages (VPC , VDY ) are not in-tegrated on the preamplifier printed circuit board(PCB). A separate LP filter board has to be used;to minimize the noise level it is important that theground of this LP filter board is tightly connectedto the ground of the LNP-P. During the prototypingphase it is reasonable that both bias voltages of theVPT can be adjusted independently. In the finalrealization a passive voltage divider can eventuallybe used to generate the two bias voltages. At the

maximum event rate of 500 kHz with the maximumexpected photon energy deposition per crystal of12 GeV (0.58 pC from the VPT) a mean current of290 nA is flowing through the VPT. This current ismainly drawn from the dynode bias supply. Sincethe internal gain of the VPT varies only by about0.1%/V, the voltage drop over the LP filter for theVPT dynode bias is not critical.

6.3.1.5 Dynamic Range

As explained in the Sec. 6.3.1.3, the LNP-P is designed for a single pulse charge inputof maximum 4 pC. With an input charge of48 aC/MeV (290 aC/MeV) coming from the VPT(see Sec. 6.3.1.2) this corresponds to a maximumphoton energy of 83 GeV (14 GeV). Therefore thedynamic range of the LNP-P is restricted by thenoise floor only. Thus, the specification of a dy-namic range is strongly dependent on the appliedshaping filter. In principle, the energy range of theLNP-P spans from the noise floor of 1 MeV (rms)(presently available VPT, peaking-time of 200 ns,see Sec. 6.3.1.2) up to the maximum input chargecorresponding to an energy of 83 GeV; this corre-sponds to a theoretical dynamic range of 83000.In practice, the typical energy range will start at2 MeV (2·σnoise) and end at 12 GeV which corre-sponds to an effective dynamic range of 6000.

6.3.1.6 FADC Readout

For the readout with a flash ADC (FADC) the samearguments apply as discussed in Sec. 6.2.1.5. Thedistance from the preamplifier in the cold volume tothe digitizing electronics in the warm environmentis maximally 110 cm which can be bridged with flatcables and differential signal lines. Therefore also inthis case the anti-aliasing low-pass filter/amplifierwill be placed right in front of the FADC.

6.3.2 Circuit Description

The LNP-P (Version SP 883a01) for the VPT is afurther development of the charge preamplifier de-scribed in [6]. Some modifications on the circuitare made and a couple of components are changedto SMD types. The circuit diagram of the LNP-Pis shown in Fig. 6.10. The AC-coupled input stageconsists of a low-noise J-FET of the type BF862from the company NXP Semiconductors (formerPhilips). This industrial standard J-FET is oftenused in preamplifiers of car radio receivers. It isspecified with a typical input voltage noise density


of 0.8 nV/√

Hz at 100 kHz and at room temperature.The J-FET input capacitance is 10 pF and the for-ward transductance is typically 30 mS at a drain-source current (IDS) of 5 mA. Along with the 470 ΩAC-dominant drain resistor this transductance re-sults in a typical AC-voltage gain of 14 for the J-FET input stage. The gate of the J-FET is pro-tected against over-voltages by two low-leakage sili-con diodes of the type BAS45AL. The input stage isfollowed by a broadband (300 MHz), fast (2000 V/s)and low power (1 mA) current feedback operationalamplifier of the type AD8011AR from the com-pany Analog Devices. With its typical input volt-age noise density of only 2 nV/

√Hz at 10 kHz, this

amplifier suits well for such a low noise design. Theproper frequency compensation is performed by thecapacitor C13 (100 pF), in combination with R2(10 Ω); this leads to high-frequency feedback to theinverting input of the operational amplifier. Over-shoot and ringing can be efficiently suppressed andthis compensation also prevents from oscillationswhen no VPT is connected.

The output of the operational amplifier is DC-coupled via the feedback network (1 pF || 25 MΩ)to gate of the J-FET. In parallel the output is AC-coupled via a 1 µF capacitor and a 47 Ω series re-sistor to the output of the LNP-P. Therefore, theoutput voltage is divided by a factor of two if it isterminated with 50 Ω.

With a symmetrical supply voltage of ±6 V the out-put voltage can swing symmetrically between thepositive and negative supply when high continu-ous event rates at high energies occur. The LNP-Pdraws a typical quiescent current of 6.3 mA fromthe +6 V supply and 1.2 mA from the -6 V sup-ply; this leads to a total power consumption of only45 mW.

To set the 5 mA operating point of drain-source cur-rent through the J-FET, a gate voltage in a range of-0.2 V to -0.6 V (typically -0.3 V, depending on theDC characteristics of the individual J-FET) has tobe applied. This negative DC voltage is fed from theoutput of the operational amplifier via the 25 MΩresistor to the gate of the J-FET. The operatingpoint (IDS = 5 mA) is fixed by the well filtered DCvoltage applied to the inverting input of the opera-tional amplifier. This set-point voltage is obtainedby subtracting 2.5 V from the positive supply volt-age (+6 V) by using a 2.5 V reference diode. So, thesame voltage drop of 2.5 V must also be present overthe total drain resistor of 503 Ω (470 Ω + 33 Ω); thisresults in a stabilized DC drain current of 5 mA.

As shown in Fig. 6.10 the anode of the VPT is ref-erenced to ground by a 20 MΩ resistor and the gate

input of the J-FET is decoupled by a 4.7 nF highvoltage capacitor.

As already discussed in the Sec. 6.3.1.4, the voltagedrop over the LP filter for the VPT bias voltagehas to be proven. At high rates in combinationwith high energies, a maximum current of 240 nAis flowing through the VPT, mainly drawn from thedynode bias voltage supply. The planned series re-sistance of the LP filter is 40 MΩ, resulting in amaximum voltage drop of about 10 V. By usingthe typical gain sensitivity of 0.1%/V of the VPT,this voltage drop corresponds to a maximum en-ergy/rate error of 1%. By reducing the series resis-tance of the LP filter, this energy/rate error can bekept at an acceptable level.

6.3.3 Performance Parameters

A summary of the LNP-P (Version SP 883a01) per-formance and specifications is given below:

• J-FET (BF862, NXP Semiconductors) in com-bination with a low power, high-speed current-feedback operational amplifier (AD8011AR,Analog Devices)

• Supply +6 V at 6.3 mA, -6 V at 1.2 mA leadingto 45 mW quiescent power consumption

• Rise-time 13 ns at Cd = 22 pF

• Feedback time-constant 25µs

• Gain: 0.5 V/pC at 50 Ω termination

• Maximum single pulse input charge: 4 pC

• Maximum 500 kHz burst input charge: 0.3 pC

• Maximum continuous 500 kHz input charge:8 pC

• Single channel LNP-P version: PCB size 48 ×18 mm2

• Typical noise performance at -25 C and Cd =22 pF (see also Fig. 6.11)

– ENC = 235 e− (rms) (shaping with apeaking-time of 650 ns)

– ENC = 300 e− (rms) (shaping with apeaking-time of 200 ns)

The single-ended output of the LNP-P is designedto drive a 50 Ω transmission line. The charge sensi-tivity is 0.5 V/pC and so the maximum input chargeof 4 pC corresponds to a positive output pulse with


Figure 6.10: Circuit diagram of the LNP-P prototype for the VPT readout. The flexibility of the discretedesign allows easy modifications in the future development process. The HV filter indicated in the top left is notintegrated on the preamplifier PCB. This is the revision 1 of the LNP-P and it has the identification number SP883a01.

a peak voltage of 2 V at 50 Ω. As an incident pho-ton energy of 100 MeV corresponds to a pulse peakof only 2.4 mV, the subsequent electronics has alsoto be designed with low-noise performance. If thefollowing electronics is located at distances of sev-eral 10 cm from the preamplifier, it may be neces-sary to add an additional amplifier onto the LNP-Pprinted circuit board. Advantageously an ampli-fier with a differential output driver should be inte-grated because differential signals are less sensitiveto noise-pickup caused by an improper ground sys-tem. The differential driver AD8137YR from thecompany Analog Devices seems to be applicable foran extra gain of 5, while capable to drive a 120 Ωterminated differential line. It has a typical volt-age noise density of only 9 nV/

√Hz at 10 kHz. By

using such an additional amplifier/driver, the quies-cent power consumption of the preamplifier wouldincrease to around 85 mW and a larger rise time ofabout 20 ns is expected at a detector capacitanceof 22 pF. Also more space on the printed circuitboard of the LNP-P would be needed for such anadditional amplifier.

The LNP-P (Version SP 883a01) can handle detec-tor capacitances in a range from 0 pF to 250 pF.To reach an optimal rise time, the frequency com-pensation of the amplifier can be tuned by a capac-itor. For different ranges of detector capacitancesthe frequency compensation must be matched. The

actual frequency compensation is suitable for de-tector capacitances in a range of 0 pF to 100 pF. Itresults in a short rise-time of only 13 ns at a detec-tor capacitance of 22 pF; this allows precise timingmeasurements. The anode of the VPT is connectedto the LNP-P via a short and shielded cable.

6.3.4 SPICE Simulations

A precise SPICE model of the LNP-P including theshaping filter (peaking-time 650 ns) has been de-veloped. The LNP-P circuit is based on the SPICEmodels of the BF862 (March 2007, NXP Semicon-ductors) and the model of the AD8011 (Rev. A1997, Analog Devices). The shaping filter is mod-eled noiseless by using the Laplace block from theanalog behavioral modeling (ABM) library. Allsimulations are made with PSPICE version 16.0from the company Orcad/Cadence. An example ofthe good agreement between simulation and mea-surement is given in Fig. 6.12.

6.4 APD Timing Performancewith FADC Readout

The EMC readout electronics is being designedto provide the best possible energy resolution and


Figure 6.11: The measured noise performance of theLNP-P versus the detector capacitance (Cd) at roomtemperature and at -25C. Measurements are per-formed by using an ORTEC450 Research Amplifier withTint = 250 ns and Tdiff = 2 µs which corresponds toa peaking-time of 650 ns. One can notice the strongdecrease of the noise if the detector capacitance dropsfrom 270 pF for a LAAPD, (1250 e− (rms)) to 22 pF forthe VPT (235 e− (rms)). At high event rates, a moreadequate shaping filter with a peaking-time of 200 nsmust be used; in that case the noise for a detector ca-pacitance of 22 pF is increased by 25%, which results inan ENC of around 300 e− (rms).

Figure 6.12: PSPICE simulation of ENC versus thedetector capacitance (dashed red) together with themeasured ENC (blue line), both at -25C. The simula-tion and the measurement are in very good agreementover the entire capacitance range.

highest dynamic range. However, a time resolutionin the order of at least 1 ns is desirable to rejectbackground hits or random noise. To investigatethe timing performance of the APD readout a se-ries of test measurements was performed. The ex-perimental setup consisted of two Hamamatsu APDS8664-1010 mounted on opposite sides of a 150 mmlong PWO-II crystal. The crystal was mounted in

Figure 6.13: Time resolution as function of the corre-sponding pulse energy.

an alcohol-cooled aluminium case placed in a dry-nitrogen flooded dark box. Light pulses of 3 ns risetime and variable intensity were supplied by a LEDpulser using quartz fibers, coupled perpendicularlyto either crystal end face. The APD signals afterthe LNP preamplifier were shaped using a newlydeveloped two-channel two-stage shaper unit andwere digitized by 10 bit 80 MHz sampling ADC. Thefull pulse shape was digitized in typically 60 timesamples with 7 time samples in the leading edge ofthe pulse. Time information for each pulser eventwas determined using the method of constant frac-tion timing. Fig. 6.13 shows the time resolution asfunction of the collected charge. The measurementswere performed at room temperature and at -25C.An arrow indicates the position of the cosmic-raypeak with an energy deposit of roughly 22 MeV,which was used to establish the charge-energy cal-ibration. At the lower energies the time resolutionis limited by APD- and preamplifier-noise. Thesemeasurements show that it is possible to achievea time resolution better than 1 ns at energy de-posits above 60 MeV and 150 ps at energies above500 MeV using the proposed sampling ADC read-out scheme of the APD signals.

6.5 Digitizer Module

A functional diagram of the digitizer module isshown in Fig. 6.14. The module employs commer-cial multi-channel 12 bit ADC chips. One modulehouses up to 120 ADC channels, FPGAs and twofiber optic links.

There will be two versions of the digitizer modulewith Low and High digitization frequency for theASIC and for the LNP preamplifier, respectively.In both cases the digitization frequency is a factorof three higher than the frequency of the highest


Figure 6.14: The functional diagram of the digitizermodule.

harmonic. The digitization frequency range will bearranged between 40 MHz for Low and 80 MHz forHigh frequencies. These values will be implementedfor test experiments with the Proto60 prototype ofthe barrel EMC equipped with a prototype versionof the ADC module and the LNP preamplifier.

The ADC chips have a resolution of 12 bit sothat with two overlapping amplification ranges (seeFig. 6.7) digitized in two ADC channels the full dy-namic range of 12000 can be covered. A specialrange selection circuit, suitable for operation in con-junction with the LNP preamplifier, is introducedin front of the ADC chip. The circuit multiplexesdirect or attenuated signals depending on the signalamplitude. The range selection circuit consists of acomparator, an attenuator and an analogue switch.The switch is synchronized with the ADC clock.Since the switching time is less than one clock pe-riod, one sample can be distorted during switching.The two endcaps, equipped with VPT and LNPpreamplifier, thus require in total 4192 ADC chan-nels. The APFEL preamplifier ASIC provides twooutputs with different gain and does not require therange selection circuit. For the independent readoutof two LAAPD per crystal we thus require 4×11360ADC channels for the barrel part.

The FPGAs perform the following tasks:

• time adjustment and distribution of the globalclock signal;

• noise calibration;

• common mode noise suppression;

• pedestal subtraction;

• autonomous hit detection;

• conversion of ADC data and linearization ofthe full data range;

• transporting the hit information together withthe time stamp to the data multiplexer;

• slow control.

The architecture of the digitizer module preservesthe redundancy policy, introduced by equipping ev-ery crystal with two APDs. The digitizer consists oftwo blocks of 60 channels each. The blocks have in-terconnections at the level of FPGAs but may func-tion independently. The EMC channels are mappedin a way that the first APD of the crystal is con-nected to block one and the second APD to blocktwo. In case of failure of any component at most 60channels out of 120 will not provide data. However,during normal operation the data of two APDs ofthe crystal are merged inside one of the FPGAs byusing high-speed links between FPGAs.

An important parameter for the construction of thedetector system is the power consumption and thechannel density of the readout system. If one wouldstart the development of the digitizer module today,using presently available commercial components,the power consumption of the digitizer would bebelow 400 mW per channel and the channel densitywould be about 3 cm2/channel.

A prototype ADC module, shown in Fig. 6.15 [7],has a size of 70×130 mm2 and contains 32 channelsof 12 bit 65 MSPS ADCs. The total power con-sumption of the module is 15 W. The parametersof this prototype module are used as a reference forthe design of the final readout system. The mod-ule is part of the setup for experiments with theProto60 prototype of the barrel EMC.

6.6 Data Multiplexer

The data multiplexer provides the interfaces be-tween the user program and the front ends, andbetween the front-end and the DAQ system. Theforeseen data multiplexer module will comply tothe new Advanced Telecommunication Architecturestandard. The following Physical Interfaces will beincluded:

• 1 bidirectional 1 Gbit/s optical link to/fromthe Time Distribution System;

• 10 bidirectional 1 Gbit/s optical links to/fromthe front-end electronics (the digitizer mod-ules);

• 2 bidirectional 2 Gbit/s copper links to/fromthe backplane to the neighboring multiplexersfor


Figure 6.15: The layout of the prototype ADC module.

• 2 bidirectional 2 Gbit/s links to the DAQ sys-tem;

• 1 Ethernet link to a general purpose networkfor configuration and slow control.

The data multiplexer performs advanced data pro-cessing by extracting the signal amplitude and time,combining single hits into clusters, and sorting theclusters in a time-ordered sequence. A data flowthrough the module of up to 200 MBytes/s seemsfeasible.

For transmission to the data acquisition system amaximum data rate of 4 GByte/s is estimated. Thisestimate is based on an event rate of 2·107 antipro-ton annihilations per second, a multiplicity of 5detected clusters with a typical cluster size of 10crystals, 8 Bytes per hit crystal using two indepen-dent readout channels per crystal (as foreseen inthe barrel). Further is assumed, that a clusteriza-tion process is applied before sending data to thedata acquisition system. If clusterization would notbe included, the data rate would increase by a fac-tor 2 to 3. For 4 GByte/s a number of 40 opticallinks is required, assuming an optical link capacityof 100 MByte/s, and an equivalent number of datamultiplexers.

6.7 Signal Routing andCabling

6.7.1 Requirements

The detector system cooled to -25C requires a min-imum amount of heat conducting copper into thesystem. However, to provide shielded, controlledimpedance of 50 Ω signal lines and high voltageinsulation, we can not use standard solutions. InProto60 we designed a rigid multilayer-back-PCB(Fig. 6.16). The drawback was the reduction of flex-ibility in mechanical design. Possible solutions are,apart from round cables, flat ribbon cable, Flat FoilCable (FFC), Flat Laminated Cable (FLC), exFCand Flexible Printed Circuit (FPC). Because of theheterogeneous types of conductors it is planned toimplement flexible flat multilayer cables (FPC) in-stead of conventional bulky flexible solutions withcoax cables. The signal, ±6 V power supply andthe bias voltage (HV) is designed in one 4-Layercable per channel. These flexible cables are usuallycustom made in different lengths (for PANDA EMCabout 350 mm for the barrel EMC and 1000 mmfor the forward endcap EMC) and not availableas a standard product. Long cables are increas-ing the noise, especially between APD/VPT andthe preamplifier, but also between the preamplifierand the Shaper/ADC. Therefore, good shielding isessential. The temperature sensors (0.2 mm ther-mocouples) and the light guides for calibration arenot discussed here.


6.7.2 Cable Performance andSpecifications for Proto60Assembly

The distance between the preamplifier and the APDacts as a thermal decoupling to the APD andthe crystals but must not be too long because ofnoise induced into the cable. A four fold devicereduces the amount of cables from the warm tothe cooled zone and saves space. Serviceabilityis provided through the use of a removable mul-tilayer backplane-PCB (Fig. 6.16, which is a cheapand technically suitable solution to distribute sup-ply voltages to the preamplifiers with a minimumamount of copper and to break out to ambient withmaximum signal integrity through impedance con-trolled signal lines. Further connections are thenmade through Lemo00-connectors and RG174 coaxcables. At the moment, the disadvantage of this so-lution is the limitations imposed on the mechanicalconstruction in the rigid PCB version.

From APD to the preamplifier a 70 mm longshielded twisted pair cable (Krophon Liff2Y-DY2 × 0.073 mm2) with an outer diameter of 2.2 mmand specified operating voltage of 500 VDC. Theflexibility of the cable is sufficient to mount theparts using 2.54 mm connectors.

6.7.3 Cable Performance andSpecifications for barrel EMCand forward endcap EMC

6.7.3.1 Barrel

The cabling between preamplifier and ADC for eachchannel is planned to be provided by a 350 mm ×ca. 15 mm Flex-PCB flat cable (see scheme inFig. 6.17). The bias voltage is maximum 500 VDCfor the barrel EMC.

6.7.3.2 Forward Endcap

A new version of the single-channel LNP pream-plifier (SP883a01) was designed for use in the for-ward endcap EMC with improved stability alsofor the small capacitance of the VPT’s. Thispreamplifier will be implemented in a ”Proto16”VPT-subunit. The cabling between preamplifierand ADC for each channel is planned to be pro-vided by 1000 mm× ca. 15 mm Flex-PCB flat cable.The preamplifier works also with a 50 Ω line over1000 mm to the ADC or one of the eight patch pan-els from where RG178 (diameter 1.8 mm) or other

coax cables can be connected. The maximum biasvoltage is 1000 VDC and two high voltages mightbe supplied to the forward endcap EMC VPTs.

From the HV-Filter to the VPT and from the VPTto the preamplifier the distance must be as short aspossible. The signal must be shielded and hold the1000 VDC bias voltage.

The distance between the preamplifier and thePatchpanel/Shaper/ADC is about 1 m. The VPTgives about 10 times lower pulse height (with VPTgain of 10) than the APD at the moment. A newVPT with gain up to 40 is under development.

6.7.4 Circuit Description

The AC coupled signal is transmitted over a 50 Ωline. A suitable conductor width for an impedancecontrolled microstrip design is for example 0.2 mmon an isolation layer of 0.1 mm to ground, but de-pends on the material specifications. Crosstalk isminimized by separate shielding of every channel.

6.7.5 Manufacturing, Operation,and Safety

The materials are polyimide (e.g. Kapton), copper,tin-lead and are not flammable. Connectors haveonly local relevance but care has to be taken also(UL94-V0 approval). Polyimide (e.g. Kapton) iswidely used in accelerator environments. Solder-type Sn60Pb40 is used to avoid ”tin pest” of lead-free solder points at long time temperature exposureof < 13C (β-Sn to α-Sn transformation).

6.7.5.1 Manufacturing and Connectors

Economic automatic processing of connectors with-out soldering can be provided with the use of crimp-ing contacts (e.g. with Schleuniger HFC) or/andpress-fit connectors.

6.7.6 Alternatives

There are alternatives to custom made cables withseparated signal line, power supply and HV-cables.This is probably cheaper but needs also morespace and has a significantly higher thermal impact,which also leads to higher costs. An additional dif-ferential amplifier stage on the preamplifier can re-duce the sensitivity of the signal lines but causeshigher power consumption in the cooled stage. Thesolution with the ADC directly mounted behind the


Figure 6.16: The layout of the Layer Stack Back-PCB ”SP903” used in the barrel EMC prototype Proto60

Figure 6.17: Scheme of the layer stack of flat cables.

preamplifier is not recommended because of highercooling power required at that point. A resistivevoltage divider near the VPT consumes roughly1 mW (1000 V × 1µA) but can save a separate biasvoltage line.

6.8 Detector Control System

6.8.1 Goals

The aim of a Detector Control System (DCS) is toensure the correct and stable operation of an exper-iment, so that the data taken by the detector are ofhigh quality. The scope of the DCS is therefore verywide and includes all subsystems and other individ-ual elements involved in the control and monitoringof the detector. The EMC and its subcomponentswill therefore be embedded in the more general DCSstructure of the complete PANDA detector. Here wewill focus primarily on DCS aspects relevant for de-sign and operation of the electromagnetic calorime-ter.

The DCS extends from the active elements of thecomplete setup of the experiment, the electronics atthe detector and in the control room, to the com-munications with the accelerator. The DCS alsoplays a major role in the protection of the exper-iment from any adverse occurrences. Many of thefunctions provided by the DCS are needed at alltimes, and as a result some parts of the DCS must

function continuously, on a 24-hour basis, duringthe entire year. The primary function of the DCSwill be the overall control of the detector status andits environment. In addition, the DCS has to com-municate with external entities, in particular withthe run control and monitoring system, which is incharge of the overall control and monitoring of thedata-taking process and of the accelerator. Systemwide we require the DCS to be:

• reliable, with respect to safe power, as well asredundant and reliable hardware in numerousplaces;

• modular and hierarchical;

• partitionable in order to allow independentcontrol of individual subdetectors or parts ofthem;

• automated to speed up the execution of com-monly requested actions and also to avoid hu-man errors in highly repetitive actions;

• easily operated such that a few non-experts areable to control the routine operations of theexperiment;

• scalable, such that new subdetectors or subde-tector components can be integrated;

• generic: it must provide generic interfaces toother systems, e.g. the accelerator or the runcontrol and monitoring system;


Figure 6.18: The various functions of the DetectorControl System.

• easily maintainable;

• homogeneous, which will greatly facilitate itsintegration, maintenance and possible up-grades, and displays a uniform ’look and feel’throughout all of its parts.

The DCS has to fulfill the following functions (seeFig. 6.18):

• process control

• detector control

• detector monitoring

• ambient circumstances control

• trigger control

• data monitoring

• calibrations

• archiving

6.8.2 Process Control

The Electromagnetic Calorimeter (EMC) is onepart of a complex detector system and it is calleda subdetector. The subsystem must be compati-ble with the requirements imposed on the DetectorControl System as stated above. One important re-quirement is the process control, realized by the Su-pervisory Control and Data Acquisition (SCADA)software. The process control supervises the follow-ing items:

• system status

• data flow

Figure 6.19: The DCS network structure.

• run start/stop status

• critical conditions

The definite knowledge of the system status is theheart of each control system. It enables a high levelof abstraction and a simplified representation of de-tector control systems. A finite set of well-definedstates is introduced, in which each of its subsystemscan be, and rules are defined, that govern transi-tions between these states. The system status ofeach subsystem depends on the current state of theunderlying hardware. At the same time, the systemstatus enables a logical grouping of DCS subsys-tems into a hierarchical tree-like structure, where”parent” states are uniquely determined by statesof its children and system-specific logic. Each par-ent in such system status tree can issue an actioncommand to its children. Action commands at thelowest level imply appropriate commands issued tothe controlled hardware. The PANDA EMC con-trol software will be implemented in this way. Thesoftware granularity is driven by the EMC sub-system structure. The High Voltage (HV), LowVoltage (LV), cooling, temperature, humidity andsafety systems are controlled by independent appli-cations. On top of these applications there is theEMC supervisory application that implements anhierarchical structuring of the whole EMC controlsoftware. In addition the EMC DCS applicationsinclude numerous other functionalities, such as e.g.full parametrization and visualization of each sub-system, loading from and storing to the PANDAconfiguration database the start-up and operationalparameters for EMC DCS subsystems. Among thecharacteristics of the SCADA system we can distin-guish two of them as most important: its capability


to collect data from any kind of installation and itsability to control these facilities by sending (lim-ited) control instructions. The standard SCADAfunctionality can be summarized as follows:

• data acquisition

• data logging and archiving

• alarm and alert handling

• access control mechanism

• human-machine interface, including manystandard features such as alarm display.

Two basic layers can be distinguish in a SCADAsystem: the ’client layer’, which caters for the man-machine interaction, and the ’data-server layer’,which handles most of the process data control ac-tivities (see Fig. 6.19). The data servers communi-cate with devices in the field through process con-trollers. The latter, e.g. programmable logic con-trollers (PLCs), are connected to the data serverseither directly or by networks or fieldbuses. Dataservers are connected to each other and to clientstations via an Ethernet local area network (LAN).There should be also a detector database fully inte-grated into the SCADA framework. This databaseshould be accessible from the SCADA so that theSCADA features (e.g. alarming or logging) can befully exploited. This database can incorporate:

• firmware

• readout settings (thresholds etc.)

• hardware settings (e.g. HV, LV)

• alignment values

• calibration constants.

6.8.3 Detector Control andMonitoring

The EMC Detector Control System should providethe monitoring of the detector conditions of the on-detector electronics as well as of all EMC subsys-tems (HV, LV, cooling system, gas system, status oflaser monitoring system). All these monitored datashould be recorded and archived as part of the com-mon PANDA’conditions database’. The DCS alsohas to provide early warnings about abnormal con-ditions, issue alarms, execute control actions andtrigger hardwired interlocks to protect the detectorand its electronics from severe damage. Regarding

control functions, the DCS will switch on/off andramp up/down the HV and LV, as well as set uptheir operational parameters. Overall PANDA willhave a hierarchical DCS tree. This tree is a soft-ware layer built on top of the experiment controls.Every detector integrates its controls in this tree-like structure. The PANDA DCS supervisor, whichis connected to the PANDA run control, will sit ontop of the tree. In this way the EMC DCS will bedirectly controlled by the PANDA supervisor. How-ever, when EMC runs separately (i.e. commission-ing) its DCS is under control of an EMC run control.The EMC DCS also has connections to the PANDAdetector safety system. All of the DCS functional-ities must be constantly available during the EMC(PANDA) run time and some functionality practi-cally non-stop (24h/365d) during the whole PANDAdetector life time. Parts of the DCS functionalitiesare going to be implemented through software ap-plications running on dedicated DCS computers, asis the case with the HV, LV, cooling and laser mon-itoring systems. These applications communicateto hardware or to embedded computers using stan-dard network or field-bus protocols. The other partof the functionalities will be implemented via dedi-cated DCS applications whose readout systems arecompletely independent of the EMC DAQ. Theseare the EMC monitoring system for temperatureand humidity, and the monitoring system for the airtemperature of the EMC electronics environment,the water leakage sensors, the proper functioning ofthe EMC cooling and LV cooling systems, and thecontrol system to automatically perform predefinedsafety actions and generate interlocks in case of anyalarm situation. The EMC DCS must have an in-terface to the data acquisition system (DAQ). Acommunication mechanism between DAQ and theDCS is needed for several purposes. During normalphysics data taking, the DCS will act as a slaveto the run control and monitoring system and willtherefore have to receive commands and send backstatus information. Partitions will also have to besynchronized between the DAQ and the DCS: therun control and monitoring system will instruct theDCS to set up specific partitions, which will be putunder the control of the run control and monitoringsystem. Furthermore, alarms will be sent from onesystem to the others.


6.9 Production and Assembly

6.9.1 ASIC preamplifier

Since the beginning of the integrated preamplifierdevelopment in 2005 two prototypes have been de-signed. These prototypes have been produced onMulti Project Wafer (MPW) runs of the EURO-PRACTICE IC prototyping program. At least onemore prototype iteration is expected to be neededbefore a final ASIC design is reached for production.

Typically the designer gets some 20 chips by eachMPW prototyping run which is sufficient for testingand device characterisation but may be not suffi-cient for a detector test setup with a medium scaledetector array like the Proto60. For such cases theEUROPRACTICE program offers the possibility toprocess additional wafers on a MPW run. Each ad-ditional wafer yields about 50 pieces. If the deci-sion will be made to adapt the existing integratedpreamplifier for the VPT readout of both endcapsof the EMC as well, additional prototypes have tobe designed, produced and tested for this versions.Also these prototypes can be produced very costeffective by MPW runs.

For instrumentation of the electromagneticcalorimeter about 23000 pieces of the preamplifierare needed for the barrel EMC part (2 ASICs forone crystal with 2 LAAPD) and about 5000 piecesfor both endcaps (1 ASIC for one crystal with VPTreadout). These amounts of ASICs can no longerbe produced cost effectively with MPW runs, soone has to start a chip production campaign. Sucha campaign starts with the production of a set ofphotomasks for the lithography steps during chipproduction. With this mask set an engineering runis started to optimize the production parameter forthis design. During this engineering run 6 wafersare produced but only 2 wafers are guaranteed tobe within the specifications.

With a die size of 10 mm2 and a wafer diameter of 8inches one will get about 3000 pieces on each wafer,see Fig. 6.20, so two wafers do not suffice to getenough preamplifier ASICs for the whole detector.That means, after testing the dies produced on theengineering run, one has to order the production of1 wafer lot in addition which, in case of a productionrun at Austria Mikrosysteme, is 25 wafers.

After wafer production an intensive test phase hasto follow. For wafer tests so called needle cards areused to connect the circuits on the wafer temporary.These needle cards are special printed circuit boardswith very fine needles which are placed to fit the

Figure 6.20: Microchip wafer.

Figure 6.21: An integrated circuit connected by nee-dles during a wafertest.

bonding pads of the integrated circuit. On a waferprober this needle card is lowered on the wafer so asingle circuit on the wafer can be tested electrically.After the test is finished the needle card is left offand the wafer has to be moved by one chip to startthe next test. In Fig. 6.21 one can see 6 needlesconnecting an integrated circuit.

For mass tests this procedure has to be done witha semiautomatic prober which can do the lowering,lifting and wafer stepping in an automatic manner.


Figure 6.22: The top- and the bottom side of the single-channel LNP-P prototype for the VPT readout. Ithas a PCB size of 48× 18 mm2 and four holes (connected to ground) with a diameter of 2.3 mm are foreseen formounting. The connector for the VPT is on the left side and the supply voltage (±6 V) is connected via thewhite socket on the right. For the testing phase a Lemo-00 connector is equipped at the signal output. On thebottom side the two 10 MΩ HV resistors and the HV gate decoupling capacitor are located.

Nevertheless, wafer changing and frequently clean-ing of the needles has to be done manually. As allof these tests have to be done 3000 times for eachwafer and 81000 times in total, test-algorithms haveto be very efficient and fast. After testing, the chipassembly has to be started by an external assem-bly company. Assembly for Chip On Board (COB)technology consists of several steps:

• Sawing wafers into single dies

• Picking up good dies controlled by a wafermapwhich was compiled from wafer test results

• Placing and gluing dies on a PCB

• Wire bonding the dies

• Globtop the dies for mechanical protection

• Placing additional components

• Soldering discrete components

The preamplifier printed circuit board modules pro-duced this way have to be tested once again beforethey are ready to be mounted on the detector.

6.9.2 Discrete preamplifier

The LNP-P is a simple, robust and low-cost combi-nation of a standard J-FET with a fast integratedoperational amplifier. The single-channel version ofthe LNP-P prototype for the VPT readout is im-plemented on a small-size double layer printed cir-cuit board (PCB) with the mechanical dimensionsof 48× 18 mm2 (see Fig. 6.22).

All components, except the connectors, are surface-mount devices (SMD). Therefore the LNP-P is wellsuited for automated mass production. Due to the

discrete design of the LNP-P for the VPT read-out, adaptations and modifications can be smoothlymade in the future. For example, the power con-sumption can be easily reduced by changing onlythe value of two resistors. Of course, lowering thepower consumption would also increase the noiselevel of the preamplifier.


References

[1] D. M. Binkley et al., IEEE Trans. Nucl. Sci.39 No. 4, 747 (1992).

[2] T. Suharli, J. vd Spiegel, and H. H. Williams,IEEE Journal of Solid-State Circuits 30 No.2, 110 (1995).

[3] G. Gramegna, P. O’Connor, P. Rehak, andS. Hart, Nucl. Instrum. Meth. A390, 241(1997).

[4] R. G. Meyer and R. A. Blauschild, IEEEJournal of Solid-State Circuits 21 No. 4, 530(1986).

[5] K. R. Laker and W. M. C. Sansen, Design ofAnalog integrated circuits and systems,McGraw Hill, 1994.

[6] Panda Technical Progress Report, section8.5.2, page 204ff, 2005.

[7] I. Konorov, A. Mann, and S. Paul, A VersatileSampling ADC System for On-detectorApplications and the Advanced TCA CrateStandard, 15th IEEE-NPSS Real-TimeConference, 2007.

107

7 Mechanics and Integration

The electromagnetic calorimeter of PANDA com-prises two main parts: the central target calorime-ter covers in cylindrical geometry almost completelythe target area. A second planar arrangement lo-cated further downstream behind the dipole magnetserves the most forward range up to an azimuthalangle of 5 with respect to the beam axis. Thetarget calorimeter is illustrated in Fig. 7.1 and com-prises three major parts as summarized in Table 7.1.

The present design concept is based on a homoge-neous electromagnetic calorimeter composed of fastand compact scintillator crystals as active absorbermaterial to be operated within the solenoidal mag-netic field of maximum 2T field strength. Largearea avalanche photodiodes (LAAPD) are consid-ered as photosensors in the barrel part. To achievethe envisaged large dynamic range in energy reach-ing from 10 GeV down to 10 MeV, the proposedscintillator PbWO4 (PWO) has to be operated atlow temperatures down to -25C to guarantee suf-ficient luminescence yield. The limited size of thephotosensor and, consequently, the restricted cover-age of the crystal endface can be compensated by asignificantly higher quantum efficiency of the sensorcompared to a standard bialkali photocathode andan improved scintillator performance.

The operation conditions impose additional, butstill feasible, requirements on the mechanical con-

Parts Barrel Forward Backwarddownstream upstream

Crystals 11360 3600 592Axial depth 2.5 m - -Distance - 2.05 m 0.55 mfrom targetInner radius 0.57 m 0.18 m 0.1 mOuter radius 0.94 m 0.92 m 0.3 mInner angle 22 5 vert. 169.7

10 horiz.Outer angle 140 23.6 151.4

Solid angle 84.7 3.2 5.5(%4π)

Table 7.1: Geometrical parameters of the EMC refer-ring to the front face of the crystal arrangement of 200mm long crystals.

struction, the insulation and the temperature sta-bility in order to control the strong temperaturedependence of the luminescence yield and keep theLAAPD gain at a tolerable level.

This report presents the principles of the calorime-ter design focusing on:

• the definition of crystal geometry

• the mechanical housing and support structure

• the thermal aspects with respect to cooling(-25C) and its fine regulation (±0.1C)

• the integration with respect to the other de-tector components and the overall geometry asshown in Fig. 7.2.

The calculations are based on PWO as detector ma-terial. The proposed crystal depth is chosen in or-der to obtain 22 radiation length (X0). The over-all granularity of the calorimeter is related to theMoliere radius, which describes the radial showerprofile. The granularity has to guarantee the re-construction of the electromagnetic showers withan adequate energy and position resolution and tolimit the occupancy even for the highest event mul-tiplicities.

The presented concept is based on experiencewithin the collaboration and on similar calorime-ter concepts for BaBar [1], CMS [2], ALICE [3] orCLAS-DVCS [4].

This design relies on the construction of prototypesprimarily to study the technology of extremely lightcrystal containers as well as the cooling and tem-perature control. The prototype results are alsopresented in this report.

7.1 The Barrel Calorimeter

The barrel calorimeter including the mechanicalstructure covers the polar angular region between22 and 140 with an inner radius of 570mm andan outer radius of 950 mm.

7.1.1 The crystal geometry andhousing

The basic crystal shape is a tapered parallelepiped,shown in Fig. 7.3, and is kept fixed for all calorime-


Figure 7.1: The components of the PANDA electromagnetic calorimeter and their acronyms as used in thePandaRoot simulation framework: barrel EMC (Barrel), forward endcap EMC (FwEndCap) and backward endcapEMC (BwEndCap), with the beam going from left to right.

Figure 7.2: Overall view of the integration of the elec-tromagnetic calorimeter into PANDA.

ter elements. It is based on the “flat-pack” config-uration used in the CMS calorimeter. Right anglecorners are introduced in order to simplify the CADdesign and the mechanical manufacturing processto reduce machining costs. The average mass ofone crystal is 0.98 kg (from 0.88 to 1.05 kg). Allgiven dimensions are nominal and the tolerances

of the crystals are 0/-100µm (the achievable tol-erance is based on the present delivery for CMSfrom the Bogoroditsk plant as well as for the DVCScalorimeter at JLAB). The dimensions of the indi-vidual crystals are related to the global shape andto the discretization of the calorimeter, defined cir-cumferentially and longitudinally in Sec. 7.1.1.1 andSec. 7.1.1.2, respectively.

7.1.1.1 Crystal Arrangements along theCircumference

Fig. 7.4 shows the crystal arrangement on the ringbased on the gap dimensions defined in Sec. 7.1.1.5.Choosing the front size of an individual crystal closeto 20 mm (21.28 mm exactly) at a radius of 570 mm,which corresponds to the Moliere radius, the ring isdivided into 160 crystals. The crystals are groupedinto packs of 4 × 10 (one alveole pack) leading to 16sectors of 22.5 coverage, which are termed slices.

The presented geometry foresees that the crystalsare not pointing towards the target position. A tiltof 4 is added on the focal axis of the slice to re-duce the dead zone effect. This means, that tracksoriginating at the target never pass through gapsbetween crystals, but always cross a significant partof a crystal.


Figure 7.3: The shape of the scintillator crystal and the definition of geometrical parameters. The averagecrystal corresponds to squares of 21.3 mm for the front face and 27.3 mm for the back face and its mass is 0.98 kg.

Figure 7.4: The segmentation of the calorimeter alongthe circumference of the barrel part. The 160 crystalsare grouped into 16 subunits named slices.

7.1.1.2 Longitudinal Crystal Positioning

Along the length of the barrel (parallel to the beamaxis) the crystal positions and individual geome-tries are shown in Fig. 7.5. The mirror symmetrywith respect to the vertical axis reduces the num-ber of different crystal shapes in the arrangementfrom 18 to 11. The lateral sizes of the rear (read-out) faces vary between 24.35 mm and 29.04 mmand the average area is equivalent to a square of27.3 mm (±15 % area variation between the 11 typesof crystals). For the front face, the lateral sizes varybetween 21.18 mm and 22.02 mm and the averagearea is equivalent to a square with lateral size of21.3 mm. In total 71 crystals are aligned at the ra-dius of 570 mm. A tilt angle of 4 is introduced toreduce the dead zone effect and this angle corre-sponds to a shift of the focus by ≈ 37 mm down-stream. In one slice of 710 crystals, 3 or 5 alveolepacks are grouped together into 6 modules: 4 mod-ules of 120 crystals each, 1 module with 70 and 1with 160 crystals. The entire barrel contains 11360crystals for a total length of 2466 mm and a volume


Figure 7.5: Geometrical arrangement of the crystals of the barrel in a cut along the beam axis. The definitionof subgroups by pack of 4 and by module is indicated. The use of the mirror symmetry decreases from 18 to 11different types of shapes according to the definition in Fig. 7.3.

Figure 7.6: View of the total barrel volume composedof 11360 crystals and, separated, a single slice of 710crystals covering 1/16 of the barrel volume.

of 1.3 m3. Fig. 7.6 highlights one slice out of thetotal barrel volume.

7.1.1.3 Crystal Light Collection

The crystals are wrapped with a reflective mate-rial in order to optimize light collection as well asto reduce optical cross talk. Considered materialis Radiant Mirror Film ESR from 3M, commonlycalled ”VM2000” in the past, accounting for a thick-ness of 63.5µm. This wrapping material, a non-

Figure 7.7: Deformation test of carbon alveoles in ver-tical position. Here the measure is around 10µm com-pared to the 80µm in the horizontal position.

metallic multilayer polymer, was employed for thecrystals of the DVCS calorimeter at JLAB) and theGLAST calorimeter. In order to foresee a small airgap between the reflector and the crystal face, thefoil must be shaped in a mold heated at 80C tosharpen the corners. In order to compensate for thelongitudinal dependence of the light collection, thestructure of the crystal surface (optically polished,roughed, lapped) can be modified or an inhomoge-


Figure 7.8: Summary of the expected dead space between calorimeter elements.

neous reflector can be selected. This is still understudy.

7.1.1.4 Carbon Fiber Alveoles

In the present design, 4 crystals will be containedin one carbon fiber alveole in order to avoid anyload transfer to the fragile PWO while the crystalsare held in place by their rear end. The expectedwall thickness of the alveoles is 200µm and theyare grouped to compose an alveole pack of 40 crys-tals. Each alveole is epoxy-glued to an aluminuminsert which is the interface with the support ele-ments. Temperature cycling tests of the gluing be-tween -40C and 30C are performed to check thereliability of this support stressed by the differentialthermal expansions. In front of the alveoles, a car-bon plate is added to avoid the movement of crys-tals. Epoxy pre-impregnated carbon plain weavefabric is precisely moulded in complex tools to fab-ricate the alveoles. Each type of crystal correspondsto one mold composed, for technical reasons, of 2alveoles to overlap the 2 wrapping joints. Real sizealveole prototypes have been produced to check thefeasibility, to optimize the final thickness and toperform mechanical tests. Fig. 7.7 shows one of the

deformation tests, here in the vertical position. Theresults 80µm horizontally and 10µm vertically aretolerable and are in agreement with the analyticalcalculation.

7.1.1.5 Distances between Crystals

The distance between two crystals is calculatedfrom the thickness of materials, structure deforma-tion and mechanical tolerances. Fig. 7.8 presentsdrawings of the different gaps which are explainedin detail below. A conservative concept has beenchosen, which can be considered as an upper limitof the expected dead space.

The basic distance between crystals inside a pack isdefined to 0.68 mm and represents the sum of:

• 400µm, the double thickness of the carbonalveoles;

• 130µm, the double thickness of the wrappingmaterial;

• 100µm, the free distance left for the alveoledeformation;

• 50µm, the approximate manufacturing toler-ance.


This calculation is based on the nominal dimensionof the crystal. There might appear an additionaldistance of < 0.2 mm between two adjacent crys-tals due to polishing tolerances. Gaps between theidentical shapes of 4 crystals in one alveole packmight introduce an additional lateral spacing up to350µm.

Between alveole packs, the thermal expansion of themounting plate amounts to about 120µm and themounting tolerance is assumed 100µm. Togetherwith the basic distance between crystals of 0.68 mmthis adds up to a total value of 0.9 mm.

Between modules, a distance varying between 2.4and 3.3 mm has to be considered to take into ac-count the mounting feasibility and tolerances in themechanical assembly, and in the structure deforma-tions.

Between adjacent slices, a first gap up to 2.62 mmdue to mechanical mounting tolerances caused bythe deformation of the whole structure and a secondgap of 1.5 mm have to be assumed. The distancebetween crystals in this area amounts to < 4.8 mm.

7.1.2 Mechanics around Crystals -Slice Definition

Fig. 7.9 presents the design principle of one slicecoping with all constraints: thermal, mechanicaland electronical integration.

7.1.2.1 Insert and Module Definition

The inserts, glued to the carbon alveoles, make theprecise connection to the back module plates andalso hold the preamplifiers and the optical fiberswhich are shown in Fig. 7.10. The shape of theseinserts are all different and complex to fabricate dueto the tapered slopes of crystals and due to the face-tized shape of the barrel calorimeter. Correspond-ing to the 6 modules defined in the Sec. 7.1.1.2,the back module plates have a thickness of 14 mmmachined with a good flatness in order to stay a ref-erence plane even under the weight of the crystals.Each plate is linked to a support beam through 6support feet designed for low thermal transfer andlow deformation. In addition, these feet have mo-tional freedom in particular directions in order to al-low translations due to the thermal expansion. Theback module plates are cooled down by the upperthermal screen. Further thermal details are givenin Sec. 7.1.4. Fig. 7.11 shows an exploded view ofone module with all individual components.

7.1.2.2 Support Beam Definition

The 710 crystals of one slice are supported by astainless steel support beam whose shape is a rect-angular tube of 2.7 m length. The bending of thisbeam is calculated to be between 0.1 and 0.4 mmfor the horizontal and the vertical position of theslice, respectively, as shown in Fig. 7.12. Thesevalues show its rigid behavior but the position ofthe modules will have to be corrected in order toalign all the crystals. The magnetic field of thesolenoid requires to verify that the applied stainlesssteel material is indeed of sufficient non-magneticquality. The magnetic quality may have been al-tered by the machining or the welding process ofthis part and a magnet-relieving anneal is foreseenin an oven. The internal part of this tube is usedfor storing all the services as electronics boards andpower supply cables. The details about the inte-gration of services are provided in Sec. 7.1.3. Theseservices are located at room temperature and ac-cess must be possible simply by opening the topexternal cover when the barrel is in maintenanceposition. The support beam is fixed on 2 supportrings at its both extremities, where a possibility toadjust the alignment of the slices is foreseen. Theserings (shown in Fig. 7.13), rest on support pointsadded to the inner vessel of the coil cryostat.

7.1.2.3 Adapted Design for the Target

The target system is passing through the calorime-ter barrel on the vertical axis. It is foreseen toplace two slices, an upper and a lower one, spe-cially designed with a central hole. Some crystalsare removed, the mechanics and the thermal shieldsare modified and a hollow cylinder of insulation isadded. In any case, the target tube will not be incontact with the cold area and will be let free tomove.

7.1.3 Electronics Integration

7.1.3.1 Photosensor, Preamplifier, FlatFlexible Cables

Due to the magnetic field, Large Area AvalanchePhoto Diodes (LAAPD) are employed. TwoLAAPD of 7 × 14 mm2 each are used and gluedon the back face of the crystal. Each LAAPD isconnected to a charge preamplifier with a 40 mmlong twisted wire. The length is discussed in theSec. 7.1.4.5 as it plays a role in the heat trans-fer to the LAAPD. The preamplifier has a power


Figure 7.9: Schematic view of the concept and of the major components of a slice.

Figure 7.10: CAD integration of the insert with thepreamplifier and optical fiber at the back of crystals.First design uses single square LAAPD but 2 rectangu-lar LAAPDs fit as well.

consumption of 52 mW/channel and is fixed on theinserts. It is connected to the read-out electron-ics through a 350 mm long flat flexible cable whichdrives: the 8 signals, 1 high and 2 low voltages,

stacked between 2 shielding layers. The sectionsare resumed in Fig. 7.14. The lack of space requireshigh density connectors. The length of the flat flex-ible cables is calculated in order to be able to groupthem and thus reduce the number of holes in thebottom face of the support beam. Besides, a in-creased length and a smaller cross section decreasethe conductive heat transfer.

7.1.3.2 Read-out Electronics

Sec. 7.1.2.2 indicated that the read-out electronics isintegrated into the support beam. This electronicscomprises the digitizing ADC boards. In additionto this equipment, the support beam also containsservices as the high and low voltage power supply,the ADC read-out optical fibers, the ADC powersupply and the water cooling tubes for the roomtemperature regulation (discussed in Sec. 7.1.4.9).Fig. 7.14 illustrates this arrangement and resumesthe sections of the services. The size of the boards is300× 150× 10 mm3. Using high-density connectorsand a stacking in pairs, 2 boards can fit even in thesmall central region of the beam and could performthe readout of up to 240 channels, equivalent to amodule equipped with 2 LAAPD per crystal.


Figure 7.11: Exploded view of one slice showing all the individual components.

Figure 7.12: Deformation of 0.4 mm of the supportbeam in the upper position (horizontal beam).

7.1.3.3 Laser Calibration System

The LAAPD is enclosed in a light tight plastic boxinto which an optical fiber for light injection is in-serted in one corner. This LED- or LASER-lightis injected from the rear side of the crystals due tothe limited space in front of the calorimeter. Thelight pulser system is primarily intended for stabil-ity control of the complete readout chain includingthe LAAPD. A first prototype is under construc-tion based on a LED light source with optical fiber

distribution (see Sec. 8.2.1). The routing of the op-tical fibers is taken care of in the early stage of thedesign in order to respect the minimum bendingradius and to integrate special guiding tubes insidethe mechanical construction. The fiber installationmust be finished before the upper thermal insula-tion is closed. The use of a non-rigid insulator, asvermiculite granulate mentioned in the Sec. 7.1.4.3,is a reasonable solution due to the high number offibers in random positions.

7.1.4 Thermal Cooling

7.1.4.1 Requirements and Method

The crystals are to be cooled down to a nominaloperating temperature of -25C. The temperaturegradients of the LAAPD gain and of the crystallight yield of -2.2 %/C and -1.9 %/C, respectively,require a stable temperature with peak to peak vari-


Figure 7.13: Barrel supported on the magnet.

Figure 7.14: Integration of the services in the support beam.

ation over time of at most ±0.1C in order to keepthe initial calibration. The read-out electronics isstabilized at room temperature and this regulation

is described in the Sec. 7.1.4.9.


Parameter PWOρ 8.28 g/cm3

Specific heat 262 J/kg.CConductivity 3.22 W/m.C

Table 7.2: The relevant properties of PbWO4 (PWO).

7.1.4.2 Thermal Properties of the PbWO4

Based on information from the CMS/ECAL Collab-oration, the thermal properties of PWO are listedin Table 7.2 and are used in the thermal analysisdiscussed below.

7.1.4.3 Thermal Shields

The crystals are surrounded by thermal shields ba-sically made of panels, cooled by serpentines filledwith a coolant, and of foam to insulate from the am-bient air. As shown in Fig. 7.9, two thermal shieldsare introduced:

• Above the crystals, the module plate is cooledvia a thin copper plate and serpentine tubesbrazed on it. These tubes have a square crosssection of 10× 20 mm2 to limit the height andkeep enough insulating foam. This upper area,containing lots of services like flexible read-outcables and optical fibers, is filled with 50 mm ofvermiculite granulate. The module plate andthe insert are made of aluminum because ofits good thermal conductivity and thus keep agood thermal transfer for the preamplifier andfor the area up to the rear end of the crystal.

• For the front side of the crystals, a specialthin thermal screen was developed, in order toreduce the distance to the DIRC for achiev-ing a better resolution. Carbon fiber materialis used for its low radiation length, in orderto improve the transparency for particles, andfor its negligible thermal expansion coefficient.This shield has a thickness of only 25 mm dis-tributed in 4 mm of carbon coolant channelsand in 21 mm of a vacuum super-insulatingpanel, respectively. Super-insulation is usedin the cryogenic technology and is in our case2.5 times more efficient in terms of thermalconductivity versus total thickness. The vac-uum (0.03 Torr), however, is absolutely neces-sary and is contained between two skins linkedwith Rohacell blocks, a structural and vacuumtight foam. The warm side is aluminum, in or-der to homogenize the temperature and avoid

Figure 7.15: Front thermal screen.

local cold points in front of each Rohacell block,while the cold side is a skin of carbon material.Its low thermal expansion reduces the differ-ential constraints between the two faces of thesandwich, which thus keeps a good flatness inany case. Fig. 7.15 shows the first sandwichstructure constructed for the thermal and me-chanical tests. Installed on prototype, it showson its external face a satisfactory temperaturevariation of 3C below room temperature.

The temperature on the sides of the slice is assumedto be constant at -25C, because of the adiabaticboundary condition with the neighboring slice atthe same temperature. For reasons of protection, a1.5 mm thick aluminum plate is inserted betweentwo slices, which additionally improves the ther-mal cooling homogeneity. In fact, the insulationbetween slices is achieved at the level of the ther-mal shields.

The heat transfer through the thermal shields isthe first external heat source of the cooling system.The definition of their thickness minimizes the heattransfer and the limit is to have their external facesat approximatively room temperature (above thedew point). This criterium ensures that the internaltemperature stability is nearly independent of theambient air temperature variation. A ratio of 25has been found from measurements performed with


Figure 7.16: Equivalent chain for the analytical com-putation of the LAAPD temperature.

crystals in the prototype detector Proto60, whichmeans that for an ambient temperature change of1C the internal temperature could vary by as muchas 0.04C without chiller regulation.

7.1.4.4 Thermal Bridges

The second external heat source is introduced bythe heat transfer through the mechanical supportsand the metallic conductors of cables for the read-out. In this conductive process, the section and thenumber of these thermal bridges have to be min-imized. Design studies and simulations have beenperformed on the support feet in order to reduce thethermal transfer. Flexible long cables are preferred.

7.1.4.5 Internal Heating of LAAPD andCrystal

The preamplifiers introduce an internal heat sourcein the calorimeter. Their total power consump-tion is 50 mW/channel, and due to the thermalimpedance of their components (200C/ W maxi-mum), the hottest point is +4C higher. This heatproduced is partially transferred by direct contactto the metallic support through its fixing screws orthrough the conductive silicone interface (Bergquistgap pad) put on top of the printed circuit board ofthe preamplifiers. But the heat propagates over-all to the LAAPD connector, and therefore to theLAAPD itself through the twisted pair wire. TheLAAPD temperature is calculated and settles at anequilibrium under the influence of the bottom ther-mal screen in front of the crystals. The analyti-cal formula is given in Eq. 7.1 and is deduced fromthe crystal/LAAPD/preamplifier model illustratedin Fig. 7.16.

Tapd = Tpcb+Rthwire∗(Tfront − Tpcb)

Rthair +Rthcrystal +Rthapd +Rthwire(7.1)

From this equation, the calculated length of thetwisted pair wire must be at least 150 mm in order

Figure 7.17: Thermal simulation with a preamplifierof 50 mW linked to the LAAPD with a 40 mm wire.

to have less than 0.1C of temperature variation onthe LAAPD. This length, however, is not compat-ible with the available space on top of the crystalsand with the good electronics functioning where thepreamplifiers have to be placed as close as possible.In the present design, the length is fixed to 40 mm,and the LAAPD temperature rises up to 0.7C. Un-fortunately, the temperature inside the barrel willnot be uniform and the LAAPD stability will bedependent of the preamplifier power consumptionwhich has to be stable in the order of 10%. The lowthermal conductivity of PWO (see Table 7.2) atten-uates any crystal temperature changes on a shorttimescale (≈ 10 sec). The crystals are affected lin-early, however, by their longitudinal temperaturenon-uniformity which, in fact as for the LAAPD,is corrected during the pre-calibration. Fig. 7.17shows a simulation of the design model.

7.1.4.6 Definition of the Cooling System

Three types of heat sources have been defined previ-ously and are resumed in Fig. 7.18. The total powerconsumption for 16 slices is ≈ 2 600 W. The externalcooling circuit is subject to heat transfer, too, esti-mated to be ≈ 700 W. Finally, the cooling machinemust have a minimum effective cooling capacity of3 300 W.

The coolant is SYLTHERM XLT from Dow Corn-


ing. This liquid is a high performance silicone poly-mer designed for use at low temperature. Comparedwith water/alcohol or with hydrofluoroether (usedin ALICE/PHOS) fluids, it offers the best ratio basedon the heat transfer versus pumpability due to itshigh specific heat and low viscosity. It has essen-tially no odor, and is not corrosive for long termuse.

The design of the cooling circuit is taking in ac-count the pump capacities, flow rate and pressurein order to optimize the flow and thus minimize thetemperature variation along the longitudinal axisof the crystal. Inside a slice, the nominal flow isaround 15 liters/min which gives a non-uniformityof 1.1C between the inlet and the outlet. Twocooling machines, one for each half-barrel, will sup-ply the barrel circuit, basically split into 4 parallelsectors.

The stability of the coolant temperature at the en-trance of the barrel must be much better than therequired stability in the barrel. A starting point forthe specifications in the machine design is ±0.05Cand a time reactivity of the order of tens of sec-onds. This value depends mainly of the quality ofthe cooling machine. Further studies will be per-formed and industrial solutions are foreseen to getthe optimal reliability versus price ratio. The lowthermal conductivity of the 20 tons of PWO mate-rial creates a very long time constant for the barrel.The expected time to reach the final temperature isseveral tens of hours.

Figure 7.18: Cooling power summary.

7.1.4.7 Read-out of the Temperatures

A slice is equipped with up to 50 thermal sensorsplaced in representative positions: along the lengthof the crystals, in the center or in the extremitiesof the slice, near the LAAPD, close to the thermalshields and in contact with inlet and outlet coolingtubes. Sensors are also placed externally, in the sup-port beam to check the water regulation. In fact,the temperature measurement controls the stabilityof the calibration but also can return informationabout possible problems in the cooling system. Allthe thermal sensors will be cross-calibrated at thenominal temperature. Two types of thermal sensorsare used: a) type T thermocouples working at lowtemperatures which give correct results for relativemeasurements and are thin enough to be insertedinto carbon fiber alveoles, or b) a few Pt100 sen-sors installed in parallel to give a better absolutevalue. The data acquisition is performed reliablyand cheap with a commercial system at a frequencyof one readout per several minutes.

7.1.4.8 Air-tightness Studies and RiskAnalysis for Low-temperatureRunning

The condition for a feasible operation at low tem-perature is to avoid any ice generation as well insideas outside of the calorimeter. If this would not beachieved, the ice could introduce water at reheat-ing of the detector or could break the crystals byice pressure between the crystals. The method isto remove the humidity of the air by circulatinga dry gas like nitrogen inside an air-tight envelopsurrounding all the calorimeter. The continuity ofthis sealing is kept at the boundaries with a feed-through like those used for electrical or supply con-nectors. This envelop is mainly made with plasticcovers wrapped with a special industrial thick ad-hesive developed for long term domestic gas seal-ing. For the outside part, the method is to avoidany temperature lower than the dew point (around12C) on the structure like on the external coolingtubes or on the mechanical supports for instance.Thermal studies are done on every critical part andmassive parts at room temperature are used in or-der to equilibrate the low temperature transferedby conduction. From a risk analysis the following 2difficulties arise:

• Danger for the calorimeter in case of ice andmoisture: if the dry gas is not completely wellcirculating or if humidity remains after anyswitch off of the dry gas circulation. Humidity


Figure 7.19: Dimensions and volume of one slice.

sensors can provide some information on thequality of the atmosphere. A break of the air-tightness envelop can lead to ice, too, and evenalter the quality of the thermal shields.

• Danger for the other detectors if the thermalshields break. The cold can reach the outsidesides and put moisture and water in the targetspectrometer.

The Proto60 crystals have shown the feasibility ofthis air-tightness envelop. Thermal studies andrisk analysis will continue and be performed withthe next prototype of 480 crystals to improve thereliability and define the emergency procedures.Fig. 7.20 introduces the CAD design of this set-upwhich will be available in summer 2008. It inte-grates all the components of a slice in their finalshape (e.g. support feet, front thermal shield). Itis equipped with stainless steel dummy crystals forcost reasons but should exhibit the same thermalbehavior.

7.1.4.9 Room Temperature Stabilizationfor the Electronics

The electronics is dissipating 560 W/slice and ismounted on copper plates brazed to cooling tubes(see Fig. 7.14). Water flows inside the tubes to reg-ulate the support beam at approximatively roomtemperature (±2C). The machine must have aminimum effective cooling capacity of 9 000 W, and

Figure 7.20: Next thermal prototype 480.

a total flow of 65 liters/min. This level of require-ments can be found in standard industrial systems.

7.1.5 Integration in the PANDATarget Spectrometer

Fig. 7.19 defines the complete volume of the barrel.The overall dimensions are required for the integra-tion of the neighboring detectors. The total mass is≈ 20 tons composed of 11 tons of crystals and 7 tonsof support structure. The services are going out


Figure 7.21: Services going outside the barrel into the corners of the octagonal yoke.

Service Detail (mm) Qty.Bundle 1 High/Low volt. 50 16

ADC supplyoptical fibers

Bundle 2 Gas, humidity/ 40 16temperatur sensorscalibr. opt. fibers

Tube 1 Water 20C 50 16Tube 2 Coolant -25C 105 8Vacuum 50 4

Table 7.3: List of services in the backward area.

of the slices in the backward area and afterwardsthrough the corners of the octagonal shape of theyoke as shown in Fig. 7.21. A list of services andrespective details is given in Table 7.3.

7.1.6 Construction of the Slices andAssembly of the Barrel

This report presents the design of a slice. A prelim-inary mounting sequence is proposed here:

1. Gluing of LAAPD onto crystals (after controland reference measurements).

2. Wrapping of crystals.

3. Insertion into alveoles. Installation of thermalsensors between crystals. Gluing of the insertafter having controlled the good functioning ofthe LAAPD with twisted pair wire.

4. Assembly of the alveole pack on the moduleplate. Installation of the guiding tube for theoptical fiber. Fixing the preamplifier. Controlof the alignment.

5. Mounting of the modules plates on the main-frame tool (as in CMS). Adding top and bot-tom cooling circuit. Adding calibration op-tical fibers. Checking of the keep in volumeand alignment. Mounting the support feet andspreading the insulation.


Figure 7.22: Barrel final mounting.

6. Put the support beam. Connecting flexible ca-bles to the boards and adding all electronics.First-level air sealing.

7. Transfer the slice to its support for storage andshipment and for the support with insulationsides to cool it down. This prepares the testwith cosmic muons.

8. Mount the rings on the rolling mounting tool.Then mount the slices one by one.

To perform all these tasks, special tools need tobe designed and manufactured. The constructionof the first prototype of a slice is foreseen and willvalidate this sequence.

The construction of a single slice and the assem-bly of the complete barrel require a large amountof manpower, installation space and time for test-ing. The production time is of the order of 3 or4 years and will need a good coordination betweenthe different laboratories.

7.1.6.1 Final Assembly of the Barrel

Each slice is installed one by one on the two supportrings and a special rolling system (as in CMS) canbe used. Fig. 7.22 presents a sketch of the mountingsequence of the barrel. Once the barrel is complete,tested and air-tight insulated, it can slide to its finalposition on a stable central beam going through thePANDA detector.

7.2 Forward Endcap

7.2.1 Requirements

The envisaged physics program of PANDA requiresmeasurements of photons and charged particleswith good position and timing resolution over awide dynamic range from a few MeV up to sev-eral GeV energy. The electromagnetic calorimeterof PANDA comprises the barrel EMC, the forwardendcap EMC, and the backward endcap EMC, seeFig. 7.1.

The forward endcap EMC is designed as a wallstructure with off-pointing projective geometry, i.e.the crystals are oriented to a point on the beamaxis which is located at a certain distance (in thiscase 890 mm) away from the target. This arrange-ment guarantees that particles originating from thetarget will never pass exactly along the boundariesbetween two neighbouring crystals where they couldremain undetected. Since at the same time quad-rant symmetry is required, this condition can notbe maintained for the boundaries between the fourquadrants of the forward endcap EMC which willbe oriented along the lines (x,y=0) and (x=0, y).

In order to catch electromagnetic showers com-pletely at the boundary between the Barrel and theforward endcap EMC, the acceptance of the forwardendcap EMC must foresee one full crystal overlapwith the Barrel. The acceptance is limited at themost forward angles by the space required for theforward magnetic spectrometer. This defines theouter opening angle to be < 23.6 and the inneropening angle to be > 5 in vertical and > 10 inhorizontal direction.

Simulations performed at 15 GeV (see Sec. 4.1.2)indicate that the particle rates per calorimeter cellwith a detection threshold of 3 MeV reach 500 kHzat the smallest angles and still amount to 100 kHz atthe largest angles. Because of these high rates andthe increased risk of radiation damage the large-area avalanche photodiodes (LAAPD), foreseen forthe Barrel, will be replaced by vacuum phototriode(VPT) photosensors in the forward endcap EMC.From two production sites VPT’s are available withan outer diameter of 22 mm. This size can be ac-commodated on a crystal with a rear cross sectionof 26×26 mm2, while still maintaining the conditionfor optimum position resolution, namely an averagecrystal width of about one Moliere radius (20 mmfor PWO). The forward endcap EMC will containPWO crystals (PWO-II) of 200 mm length, which isequivalent to 22 radiation length and sufficient for95% containment of the maximum expected photon


energy of 15 GeV.

7.2.2 Crystal shape

In the forward endcap EMC the crystals are closelypacked in ”off-pointing” geometry, i.e. oriented to-wards a point on the beam axis 950 mm farther thanthe target and 3000 mm away from the front face ofthe crystal plane. The off-pointing geometry is il-lustrated in Fig. 7.23. The ratio of target distanceto off-pointing distance has been chosen such thatthe angle of incidence of particles on the crystalfront face is minimally 1.6 with respect to nor-mal incidence. This arrangement guarantees thatparticles originating from the target will never passmore than 15% of the crystal length in the gap be-tween two neighbouring crystals. Including photosensors, front-end electronics, cooling and insula-tion, the overall depth of the forward endcap EMCwill amount to 430 mm. The off-pointing geometryand the VPT diameter determine that each crystalhas a front-face of 24.4×24.4 mm2, see Fig. 7.24. Inorder to save costs for cutting and polishing crys-tals, each crystal will be shaped with two taperedand two right-angled sides. A cluster of 4 crys-tals, touching at the right-angled sides, thus forms amini-unit of trapezoidal cross section, mounted in asingle carbon-fiber alveole with 0.18 mm wall thick-ness. The crystals should be manufactured with tol-erance of +0/-0.1 mm in all transversal and longitu-dinal dimensions. An angular precision of < 0.01

will be requested while a planarity within 0.02 mmshould be maintained for all faces with chamfers be-tween 0.5 and 0.7 mm. All surfaces will be polishedwith roughness Ra < 0.2 mm.

7.2.3 Subunit Structure

Since PWO crystals are very fragile, they must notbe exposed to bending- or shear forces. There-fore crystals are mounted in frames made fromcomposite material (carbon fiber alveoles) whichare designed to absorb tolerances in crystal dimen-sions and accommodate the thickness of the light-reflecting foils (100µm). Four mini-units of crys-tals will be combined to form a 16-crystal sub-unit of ca. 19 kg that can be attached individu-ally to the 30 mm thick aluminum mounting plate.Fig. 7.25 shows the alveole wall thickness and thespace foreseen between crystals. The front face ofthe alveole will be covered with 1 mm thick com-posite C-fiber/epoxy material. The company Fiber-Worx B.V. (Netherlands) has designed, engineeredand prototyped a C-fiber alveole for 16 crystals.

Loaded with the weight of the crystals the alve-ole material is requested to stretch less than 1.5 %.The proposed and prototyped design results in amaximum stretch of 0.04 %, which results in a com-fortable safety factor of 30. Fig. 7.26 shows threeC-fiber alveoles produced according to the abovegiven specifications.

The alveoles carrying 16 crystals will be mountedindividually to the aluminum backplane withouttouching the neighboring alveole. For this purposealuminum inserts will be glued into the rear partof the alveole downstream of the crystal, thus al-lowing a good thermal contact, providing a rigidsupport for the VPT photosensor, and creating amounting structure for the backplane. Fig. 7.27shows the explosion view of a subunit housing 16crystals, VPT and inserts. In Fig. 7.28 is demon-strated how the subunits will be attached to themounting plate using angled interface-pieces whichallow a precise arrangement. In total 3520 crystalswill be mounted in this way in 4× 4 subunits, and80 crystals in smaller 2 × 2 subunits in order toapproach as much as possible a homogeneous cov-erage between minimum and maximum acceptanceangle of the forward endcap EMC. Optionally, foreven larger coverage at the expense of more specificalveole development, an additional number of ca.40 subunits housing only 1 or 2 crystals is beingconsidered.

Preliminary thermal calculations indicate a temper-ature gradient in the crystal of ca. 4C if only themounting plate will be cooled. Presently a proto-type setup for 16 crystals is being constructed inorder to test the mechanical accuracy, the thermalproperties and photon response in photon-beam ex-periments.

7.2.3.1 Mounting Structure andImplementation in the Solenoid

One quadrant of the forward endcap EMC houses900 crystals in subunits of mostly 16 and occasion-ally 4 crystals. The arrangement as seen from thetarget is shown in Fig. 7.29. The backplane is con-structed from 30 mm thick aluminum with ellip-tic holes at the center of every alveole in order tofeed the cables from the front-end electronics to thedownstream part of the mounting plate and fromthere to the circumference of the forward endcapEMC structure.

The total weight of the forward endcap EMC willamount to ca. 5100 kg. For the design of the mount-ing plate the stiffness is the most important cri-terium. The loads acting on the mounting plate are


Figure 7.23: The position of the forward endcap EMC with respect to the target.

Figure 7.24: The geometry of a single forward endcap EMC crystal.

the gravity and the moment caused by the center ofgravity of the crystals attached in front of the plate.In its final vertical position the maximal Von Misesstress is ca. 20 MPa which results in a maximumdeflection of 0.23 mm.

The endcap mounting plate will shrink by 0.16 mmper row or column of subunits at the operation tem-perature of -25 C. With 18 rows and columns ofsubunits, there are 17 gaps in between in horizontaland vertical direction. Per gap a space has to be re-served of approximately 0.16 mm in order to absorbshrinking or expansion. In addition we have to takeinto account that the magnetic field of the solenoidwill cause deflections of the mounting structure. Itis foreseen that the forward endcap EMC will be

completely constructed outside the PANDA area,inserted into the downstream part of the solenoidmagnet yoke, and held in place by 8 holding armsinside an octagonal frame structure. This structuremust be able to absorb a shrinking or expansion ofmaximally 3 mm.

7.2.3.2 Insulation and Cooling

For the thermal insulation of the crystal area at -25C from room temperature, a space of 30 mm isreserved for a layer thermal-insulation foam. De-tailed thermal calculations are needed and havebeen started in order to determine the requiredcooling power and shielding material. Preliminary


Figure 7.25: The geometry of a C-fiber alveole for 16 crystals.

Figure 7.26: Photograph of three produced C-fiber alveoles for 16 crystals each.

calculations and experience with the Barrel proto-type indicate, that the crystals also need cooling

from the front side in order to avoid a temperaturegradient in the crystal. However, maintaining a sta-


Figure 7.27: The explosion view of a 16-crystal subunit.

Figure 7.28: The attachement of the C-fiber alveoles in front of the mounting plate using angled interface pieces.

tionary temperature gradient will be investigated,since the corresponding gradient in light produc-tion could compensate inhomogeneities due to lightcollection in a tapered crystal. This method wouldavoid an expensive and time consuming ”depolish-ing” treatment of the crystal surface. Cooling at

the front side of the crystal wall could be achievedby cooling a thin (ca. 2 mm) carbon plate in frontof the crystals. A flow of cooled dry N2 gas in-side the insulated volume would support an homo-geneous cooling of the crystal volume. To preventice forming around the crystals, dry N2 gas will be


Figure 7.29: One quadrant of the PANDA forward endcap calorimeter.

supplied to the individual alveoles by gas pipes in-serted through the cable-feedthrough holes in themounting plate. Fig. 7.30 shows part of the ther-mal insulation cover and the holding structures tofacilitate the mounting inside the solenoid.

It is foreseen to embed four separate cooling circuitsinto grooves in the downstream part of the mount-ing plate. The total cooling circuit will consist ofa meander of 10 mm diameter cooling pipes with alength of 56 m.

7.2.3.3 Sensors and Front-End Electronics

The crystals of the forward endcap EMC will beequipped with VPT’s as photosensors for which again of 30 to 50 is expected. Development of ahighly sensitive VPT with super-photocathode andquantum efficiency above 40 % is in progress. ALow Noise / Low Power (LNP) Charge Preamplifierhas been developed on basis of the equivalent LNPpreamplifier for LAAPD readout. The output pulseis transmitted via a 50 Ω line to the subsequent dig-itizer module electronics at the circumference of theforward endcap EMC. The LNP-Preamp has a qui-

escent power consumption of 45 mW and is suitedfor operation in the cooled volume. It is foreseen toattach the preamplifier directly to the VPT on theupstream side of the mounting plate.

7.2.3.4 Cabling and Supplies forMonitoring

For signal transmission from the VPT preampli-fier to the digitizer electronics a multi-pin flexi-ble laminate cable of .15 mm thickness is fore-seen. Q.P.I. BV Netherlands is able to producea double-sided, copper-clad all-polyimide compos-ite (PyraluxAP). This is a polyimide film bondedto copper foil. Fig. 7.31 gives the relation betweenimpedance and conductor width and allows to makea suitable choice for 50 Ω signal transmission.

7.2.3.5 Arrangement of ReadoutElectronics

The digitizing electronics will use sampling ADC’sfor signal-shape analysis and timing determinationand will be located near the detector in the warm


Figure 7.30: The complete endcap with thermal insulation cover and holding structures.

Figure 7.31: Impedance of laminate cable as function of conductor width.

volume inside the solenoid magnet. From there sig-nals can be transmitted via optical link to multi-

plexer units and compute nodes outside the PANDAexperimental area. With commercial components


Figure 7.32: Arrangement of Digitizer modules at thecircumference of the forward endcap EMC.

an ADC channel density of about 300 mm2/channelcan be reached. This means that per octant of theforward endcap EMC an area of 3200 cm2 must beavailable. Between the circumference of the forwardendcap EMC mounting plate (radius = 1050 mm)and the inner boundary of the solenoid (radius =1450 mm) a surface area of maximally 2000 cm2 isavailable. Thus with a sandwich of two ADC boardsthere is sufficient space available for accommodat-ing the digitizing electronics. Fig. 7.32 shows theposition of the endcap inside the solenoid structureand the space available at the forward endcap EMCcircumference for arranging the electronics boardsin a sandwich layer. In addition, the 8-fold mount-ing structure is indicated and the routing of cablesand cooling pipes.

7.3 Backward Endcap of theCalorimeter

The backward endcap EMC is required in order tocomplete the hermetic coverage of the target spec-trometer EMC for electromagnetic energy, with ex-ception of the beam entrance and the acceptance ofthe forward spectrometer. As shown in Fig. 4.3,at polar angles above 135 the differential ratesper crystal are about 1 kHz per 20 MeV at ener-gies of 50 MeV and the maxium energy depositionis about 200 MeV. Since both endcaps require analmost planar arrangement of crystals, the basic de-sign concept of the forward endcap EMC has beenchosen also for the backward endcap EMC. This

Figure 7.33: Acceptance of the backward endcapEMC.

approach simplifies the mechanical construction ofcrystals and submodules and creates synergy in theapplication of photosensors and readout electron-ics. In addition, the readout with VPT is superiorin timing performance at the low energies expectedin the backward region which allows efficient reduc-tion of background. Since design and developmentwork on other detectors in the surrounding of thebackward endcap EMC is still in progress, the me-chanical constraints are not yet defined wellenough,so that the integration in PANDA could not yet bespecified in detail. In Fig. 7.33 the angular accep-tance is presented which is allowed by the radialspace available inside the target spectrometer EMCand the distance of 550 mm of the crystal front faceto the target (see Fig. 7.2). The maximum openingdiameter is 920 mm. Allowing 20 mm for toler-ances and mounting space and 60 mm for thermalinsulation, we arrive at 840 mm maximum diame-ter of the mounting plate to which the readout endof the crystals is attached. The inner hole of 200mm diameter of the backward endcap EMC is de-termined by the beam pipe and a safe distance toprevent disturbance from beam halo. This definesthe maximum polar angle of 169.7. The tilting ofthe crystals due to the off-pointing projective geom-etry, oriented towards a point 200 mm farther thanthe target, determines the minimum polar angle of151.4. The off-pointing geometry is illustrated in


Fig. 7.34. The same ratio of target distance to off-pointing distance has been chosen as for the for-ward endcap EMC. Including photo sensors, front-end electronics and insulation, the overall depth ofthe backward endcap EMC will amount to 430 mm.

The geometry of a single crystal is shown inFig. 7.35. Given the cross section of 26 × 26 mm2

of the readout face to accommodate the VPT, thefront face of each individual crystal will result underthe given geometrical conditions in 20.5×20.5 mm2.The overall dimensions are close to the average di-mensions of the barrel crystals. The arrangementof individual crystals in one quadrant of the back-ward endcap EMC is shown in Fig. 7.36. As forthe forward endcap EMC, the size of the carbonfiber alveoles, the tolerances and the configurationin subunits of 4×4 or 2×2 crystals has been chosen.The optimum coverage of the available geometricalacceptance is achieved with 7 subunits of 16 crys-tals and 9 subunits of 4 crystals. This results in148 crystals per quadrant and 592 crystals for thewhole backward endcap EMC. The provisions fornitrogen gas flow and cooling will be arranged asin the forward endcap EMC. Due to space restric-tions, the Digitizer modules can not be attached tothe mounting plate, but will be positioned furtherupstream between the upstream barrel part and thesolenoid yoke.

References

[1] B. Aubert. et al., Nucl. Instrum. Meth. A479,1 (2002).

[2] CERN-LHCC-97-33.

[3] CERN-LHCC-99-4.

[4] P. Rosier, R&D Detection - IPN Orsay,France, internal report RDD 2004-02, 2004.


Figure 7.34: The position of the backward endcap EMC with respect to the target.

Figure 7.35: The dimensions of a single crystal of thebackward endcap EMC.

Figure 7.36: The geometry of one quadrant of thebackward endcap EMC.

131

8 Calibration and Monitoring

To achieve the required energy and spatial resolu-tion of the electromagnetic calorimeter it is essentialto precisely calibrate the individual crystal chan-nels. Time dependent variations of the calibrationfactors are expected due to several effects: change inlight transmission, change in the coupling betweenthe crystal and the photodetectors and variations inthe photodetectors itself and the following electron-ics. Both the scintillation of the crystals and theamplification of the APDs depend on the tempera-ture. Therefore stable cooling at the level of 0.1Cis required. Long term variations in the tempera-ture need to be tracked by the calibration. Givenan energy resolution between 1% and 2% at ener-gies above 1 GeV, the precision of the calibrationneeds to be at the sub-percent level.

For the calibration several methods performed instages are planned:

• Precalibration at test beams and with cosmicmuons at the level of 10%

• In situ calibration with physics events (neutralmesons and electrons)

• Continuous monitoring with a light-pulser sys-tem

They are discussed in the following subsections.

8.1 Calibration

8.1.1 Calibration with PhysicsEvents

Based on the present experience with the temper-ature sensitivity of the complete detector elementsincluding the crystal and the photosensor, a finalin-beam calibration of the whole calorimeter doesnot appear appropriate. All calibrations have to beperformed using the complete setup operating atthe final temperature.

Therefore energy calibration of each crystal channelwill be performed in situ with physics events. Lowmultiplicity events with π0 and η decaying into twophotons will be selected. The constraint on the in-variant mass m2

meson = Eγ1Eγ2(1 − cos θ12) givescorrections to the reconstructed photon energy inthe crystal channels of the well separated clusters

γ1 and γ2. After a few iterations all calibration con-stants are available with a sufficient precision. Theprecalibration of the crystals with cosmic muonswill provide the initial seed to start the procedurefor the first time. The method was applied success-fully at the Crystal Barrel experiment. There 1.2million pp events with up to 8 photons in the fi-nal state (mainly pp→ 3π0) were used to calibrate1380 crystals [1, 2].

For a full calibration of the PANDA electromag-netic calorimeter a total number of about 5 · 107

events are required. It is essential to select low oc-cupancy channels to avoid the overlap of clustersin the electromagnetic calorimeter. The channelspp→ π0π0π0 and pp→ π0π0η have a cross sectionof about 30µb, thus producing 4200 events per sec-ond at a luminosity of 1032 cm−2s−1 (η → γγ only).Having a dedicated software trigger, calorimetercalibration data will be produced at a rate of about4 kHz. The selection of the events with completelyneutral final states makes the calibration indepen-dent of any other detectors and thus quickly to per-form.

With the available statistics it is expected that acalibration can be performed once a day.

Variations of the calibration for shorter timescalesthan required for the calibration will be tracked bythe monitoring system.

8.1.2 Precalibration with CosmicMuons and at Test Beams

The absolute geometrical position of the individ-ual calorimeter elements has to be deduced fromthe mechanical design. The point of impact of thephoton relies on the reconstruction of the electro-magnetic shower and primarily on the appropriatemathematical algorithm, which will be optimizedfor the prototype arrays. The optimal position re-construction algorithm depends on the point of im-pact and will be evaluated with full size prototypesat test beams.

The precise calibration of the calorimeter by usingthe constraint on the π0 and η mass relies on a pre-calibration at a 10% level. Before the final assemblyof the calorimeters all submodules will undergo acheck and precalibration with cosmic muons. Partof the submodules will be subject to beam teststo verify the precalibration and the position recon-


struction algorithms. The final precalibration of allcrystal channels will be performed in situ with cos-mic muons. This will be done before the startupwith physics beam and at the final and stable tem-perature. Depending on the position relative to theearth surface different energies are deposited by theminimum ionizing particles. Even for upward point-ing crystals enough statistics will be reached withinone day of dedicated running. Due to the hardware-trigger-free concept of the PANDA DAQ system it iseasy to set up algorithms to store all relevant eventsby requiring certain cuts on energy depositions inneighboring crystals. The precision of the calibra-tion will be verified at test beams by comparing thecalibration obtained with cosmic muons to the cal-ibration with defined photon energies. The CMSexperiment has obtained a precision of 2.5% withthe calibration of PWO with muons. It is expectedthat the resulting precision at PANDA will be bet-ter than 10%, enough for the final calibration withphysics events.

8.1.3 Online Calibration

The hardware-trigger-free concept of the PANDADAQ system requires the availability of calibrationdata at real time. The trigger decision is takenby software operated on compute nodes where fullevent information is available. For efficient trigger-ing with the EMC it is therefore essential to havereasonable calibration constants available, to per-form a quick analysis of the event.

The calibration constants can be determined oncea day by the algorithm described in Sec. 8.1.1.The compute nodes filter the events where onlycalorimeter information and no charged track isseen. In addition a cut on 5 to 8 clusters and aminimum cut on the total energy are performed.These filtered events are stored for the calibration.Variations on the timescale of minutes and hours aredetected by the light monitoring system (Sec. 8.2).Each crystal is flashed 100 times per minute by thelight pulser. The data is evaluated on the computenodes online and variations in the light transmis-sion are measured every minute. These are appliedto the calibration constants determined before.

8.2 Monitoring

The monitoring system provides a reference to iden-tify and record any changes of the coefficients forthe energy calibration of all calorimeter cells. Theresponse can shift due to changes of the lumines-

cence process and the optical quality of the PWOcrystals, the quantum efficiency and gain of theAPD, and of the preamplifier/ADC conversion gain,for example. These effects can originate from tem-porary and permanent radiation damages and/ortemperature changes. Due to extensive research onthe radiation damage of PWO carried out duringthe last decade it was established that the lumines-cence yield of PWO crystals is not affected by irra-diation, only the optical transmission. Any changeof transparency can be monitored by the use oflight injection from a constant and stabilized lightsource. The system will be based on light sources atmultiple wavelengths. The monitoring of the radia-tion damages will be performed at low wavelengths(UV/blue to green, 455 nm and 530 nm). With lightin the red region at 660 nm, a wavelength where ra-diation damages play no role, the chain from thelight coupling to the photosensitive detector and theamplifier digitization can be monitored. The wave-lengths, which are used for monitoring purposes, areindicated in Fig. 8.1 together with the induced ab-sorption of radiation damaged PWO crystals. Thevalue of 455 nm corresponds to the closest avail-able LED with respect to the peak of the PWOemission. The concept to be used depends on thetime scale when these deteriorations appear. Asoutlined in Sec. 3.2.2.3 the expected dose rate willstay well below the values expected for CMS oper-ation. Therefore, one can assume that the possibledegradation of the optical performance will evolvegradually and very slowly in time. In that case,trends and corrections can be deduced from kine-matic parameters such as invariant masses calcu-lated off-line. To avoid photon conversion in frontof the calorimeter, the overall thickness of dead ma-terial has to be minimized. Mechanical structuresfor support or cooling should be installed as close aspossible to the crystal front face. A system of op-tical fibers for light injection from the front side isexcluded due to the space needed to cope with thelarge bending radius of fibers. Injecting light fromthe rear with reflection at the front side back to thephotosensor is possible. However, the effects due toenhanced multiple scattering must be studied. Thetemperature gradient of the luminescence yield ofPWO, which varies due to temperature quenchingbetween 2 and 3 %/K requires a temperature sta-bilization of the whole system with a precision of∼ 0.1 including a finely distributed temperaturemeasurement. In addition the intrinsic gain andthe noise level of the APD photosensors show simi-lar temperature sensitivity. The thermal contact tothe crystals should be sufficiently reliable. Checksof the linearity of the electronics at the sub-percent


level require to cover a dynamic range of 10 000.This will be achieved by neutral density filters.

Figure 8.1: Induced absorption of various PWO sam-ples after irradiation with a dose of 20 krad (60Co)shown with the indication of the available wavelengthsof LEDs, which are used for monitoring purposes.

8.2.1 Concept of a Light Source andLight Distribution System

To maintain exact correspondence between ob-served changes of the measured amplitude of thelight monitoring signal and the detector signal ini-tiated by high energy photons, which is influencedby the variation of the optical transmission of PWOcrystals, it is necessary to meet two requirements:the optical emission spectrum of the light sourceshould be similar to the radio-luminescence spec-trum of PWO; the effective optical path length formonitoring light in the crystal should be identicalto the average path length of the scintillation lightcreated by an electromagnetic shower in the crys-tal. Further technical requirements for the moni-toring light source are: the number of photons perpulse should generate the output signal equivalentto about 2 GeV in each cell; the duration of thepulse should be similar to the decay time of the scin-tillation of PWO; a long-term pulse stability betterthan 0.1 % has to be achieved; minimal pulse heightvariation from pulse to pulse (determined only byphoton statistics without any additional line broad-ening to keep the required number of monitoringevents at the minimum). Nowadays available ultra-bright LEDs (brightness > 10 cd) emitting at var-ious wavelengths give the opportunity to create asystem to meet most of the above requirements.In case of blue-violet LEDs the emission spectrum(centered at 420–430 nm) as well as the spectral

width (approx. 70 nm FWHM) correspond exactlyto those for PWO radio-luminescence. The num-ber of photons per 20 ns pulse can be as large as109 and can be further increased by more than oneorder of magnitude by combining optically manyLEDs in one emitting block. Moreover, combin-ing LEDs of different color allows fine-tuning to theentire scintillator emission spectrum. Even takinginto account the light losses in the distribution sys-tem one might illuminate 1 000–1 200 calorimetercells with one large LED block. Consequently 10 to15 such stabilized blocks can provide the monitor-ing of the whole electromagnetic calorimeter. Sucha concept allows the selective monitoring of differ-ent parts of the calorimeter by electronic triggeringof the appropriate LED blocks without the neces-sity of optical switching. The distribution systemshould provide the light transfer to each cell withthe option to fire selectively only ∼ 10 % of all cellssimultaneously to avoid interference problems in thefront-end electronics and an overload of the data ac-quisition system. This can be implemented with op-tical fibers grouped into 10 bunches attached to theoutputs of multi LED blocks. Each fiber bunch willcontain additional fibers controlling possible lossesin fiber transparency.

8.2.2 Concept of the LightMonitoring System

The concept of the system should provide flexiblecontrol and redundancy in terms of measured pa-rameters to distinguish between different sources ofinstability. Each emitting block will be equippedwith an optical feedback and a thermo-stabilizedPIN-photodiode as reference. Such a device hasbeen successfully implemented for CMS ECAL [3].A central sequencer will trigger all blocks with max-imum frequency up to a few kHz. A separate unitis foreseen to measure the light amplitude of thereference fiber in each bundle. The system does notrequire regular maintenance. Fig. 8.2 and 8.3 showthe schematic layout and the major components ofa first functioning prototype.

8.2.3 Light Pulser PrototypeStudies

8.2.3.1 Experience at Test Beam

A LED-based light pulser system with four LEDs ofdifferent wavelengths was made for the test beamstudies at Protvino to monitor gain variations ofthe PMTs and transmission variations of the PWO


Figure 8.2: Schematic layout of the stabilized light pulser system.

Figure 8.3: Major components of a first prototype ofthe monitoring system.

crystals. The LEDs emit at red (660 nm), yellow(580 nm), green (530 nm), and blue (470 nm) wave-lengths. For the actual analysis of data, the red andthe blue LEDs were most useful.

Between two accelerator spills, 10 light pulses ofone color were sent to the crystals. Then in thefollowing interval, light pulses of another color wereinjected in the crystals. This way, four spills wereneeded to collect data for all four colors. The lightfrom all the LEDs was fed into the same set of opti-cal fibers and they delivered the light to individualcrystals.

In the test beam setup, the LED light was injectedat the front end of the crystals. So the typical pathlength of LED light in the crystal approximately

equals the length of the crystal. Since the lightcomes out of the optical fiber with a characteristicfull angle spread of 25, and this angle is reduced to11 as the light enters the crystal from air, the pathlength of light in the crystal should be increased by1/ cos 11. As for the scintillation light from inci-dent particles, half of the light travels directly tothe PMT while the other half will travel towardsthe front of the crystal and gets reflected before itis detected by the PMT. Averaging these two cases,the mean path length of scintillation light to thePMT also equals the crystal length in 0th approx-imation. In order to estimate the 1st order correc-tion, we need to know how much the light zigzags onits way to the PMT. The maximum angle that thelight makes with respect to the crystal axis is deter-mined by the reflection angle on the side surfacesdue to the total internal reflection angle, which isabout 64. This leads to many more zigzag pathsthan the paths for LED light. Taking into accountthat the scintillation light is emitted isotropically,the average < 1/ cos θ > factor arising from thezigzag paths is about 1.4.

The LED system monitors the transparency of thecrystal at a specific wavelength (in our case, 470nm was chosen partially due to the availability ofblue LEDs) and thus does not sample the entirespectrum of scintillation light. The radiation dam-age effect is less severe at 470 nm than at 430 nm,the center of the PWO scintillation emission peak.From these considerations, we expect that the ratio,


R, of the light loss factors for the LED signal andthe particle signal is about 1/1.4 = 0.7 to 1/1.6 =0.6.

One of our goals in the test beam studies of the cal-ibration system is to measure this ratio, R, experi-mentally, and to observe how it varies from crystalto crystal. Naively, since this ratio only depends onthe geometrical lengths of light paths for the LEDand scintillation light, it should not vary from onecrystal to the next. If there are variations in theshape of the absorption as a function of wavelengthamong crystals, the ratio, R, may vary among crys-tals. In addition, since the crystals will not be pol-ished to optical flatness, actual reflections of lightby the side surfaces do not follow the simple law ofgeometrical light reflection. This may also lead tovariations of the ratio, R, among crystals. Thus wefeel that it is very important to measure the varia-tion of R values experimentally.

Since it is not practical to measure this ratio for allproduced crystals at a test beam facility (it wouldtake too much time), we need to know if the varia-tion, if there is any, is small enough so that we willnot spoil the resolution even if we assume and usean average value of the ratio for all crystals.

8.2.3.2 Monitoring Systems for the LightPulser System and Stability ofLight Pulser

We built two monitoring systems to check the sta-bility of the magnitude of the light pulses. (i.e.monitoring systems of the monitoring system.) Onewas based on a PIN photodiode, which is consideredvery stable even when the temperature varies. Ac-cording to the literature, the temperature variationof PIN photodiodes is less than 0.01%/C. How-ever, since we needed an amplifier to detect the PINphotodiode signal, the amplifier gain needed to bestabilized by housing it in the crystal box where thetemperature was stable to ± 0.1C.

The second system used a PMT, a scintillation crys-tal and a radioactive source. The PMT (Hama-matsu R5900) monitored the LED pulser while thePMT was monitored using the stable scintillationlight produced when a Y AlO3 : Ce crystal was ir-radiated with an 238Pu alpha source (YAP) [4]. Theα energy spectrum measured by the PMT and thepeak position of this spectrum as a function of timeis presented in Fig. 8.4. The width of the peak is2.3% r.m.s. as determined by a fit to a Gaussian.The peak position was stable over 85 hours to betterthan 0.2%.

Figure 8.4: (a) α energy spectrum accumulated over1.5 hours. (b) α spectrum peak position as a functionof time over 85 hours. Each point corresponds to a 15-minute measurement duration.

The measured variations (drifts) of the magnitudesof light pulses (averaged over 120 pulses) over dif-ferent time periods were measured for the periodsof test beam studies. They were:

• 0.1 to 0.2% over a day;

• 0.5% over a week;

• 1% over a few months.

Temperature variations were the main cause for thevariations in the size of pulses. When correctionsbased on the temperature were made in the pulse-height analysis, the long term variation significantlydecreased to 0.4% over a few months and down to0.3% over a week. No LED ageing effects were ob-served after 3000 hours of operation.

The stability of this system was better than weneeded to monitor the gain variations of the PMTsand the transparency variations of the crystals overthe relevant time periods. For example, we wereable to track the crystal transparency change withan accuracy of better than 1% over a week whenwe measured how much radiation damage the crys-tals suffered. Additionally, we were able to trackthe change in crystal transparency with an accu-racy of better than 1% over a few months whenwe measured the recovery process of the radiation-damaged crystals. Finally, we were able to trackthe PMT gain variations over a day well enough sothat it did not contribute appreciably to the energyresolution. This last accomplishment implies thatwe already have a good enough system for PANDAexcept that we need to have a much larger system,and temperature stabilization must be considered.


Figure 8.5: Linear fit coefficients calculated from thecorrelation plots of relative changes for the blue LEDvs. electron signal under (a) pion and (b) electron irra-diation.

8.2.3.3 Crystal Light Output Monitoring

As was shown by radiation hardness studies, PWOcrystals behave in a similar way in radiation envi-ronments of different nature; clear correlations be-tween electron and LED signal changes were ob-served. The dedicated study has been carried outto confirm that these correlations are not depen-dent on the type of irradiation using a particularoptical monitoring scheme. To be more specific,the same crystals were calibrated with a low inten-sity electron beam first, then they were exposed tothe highly intense electron radiation. The irradia-tion of crystals continued with a pion beam. Bothelectron and pion irradiations alternated with cal-ibration runs using a low intensity electron beam.Changes in the crystal transparency were monitoredcontinuously with the use of the LED monitoringsystem. A linear fit of the distributions of signalchange of the relative blue LED signal vs. the elec-tron signal was calculated for both the electron andpion irradiations. Coefficients of the linear fit arepresented in Fig. 8.5

As it was expected, on average the measured coeffi-cients are not different and are in a good agreementwith the calculations for the current optical scheme.The fact that the linear approximation works well,significantly simplifies the procedure of the EMC in-tercalibration that will be performed using an LED-based light monitoring system over the time inter-vals between two subsequent in-situ calibrations.

8.2.4 Light Monitoring System forPANDA Calorimeter

The light monitoring system is designed to injectlight pulses into each PWO crystal in order to mea-sure optical transmission near the scintillation spec-trum peak (430 nm). The red light pulses are usedto monitor a photodetector gain stability. The sys-tem includes both blue and red LEDs, their drivercircuits, and optical fibers to deliver light pulses toeach of the PWO crystals.

We plan to use very powerful LEDs, assisted by areflector and a light mixer so that each light pulsingsystem produces enough light for ∼3000 crystals,each receiving light pulses equivalent to scintillationlight from 2 GeV photons.

The distribution of light among the 3000 fibersshould be very uniform. This is accomplished bydesigning a good light mixer which will distributelight uniformly across an area of 38 ×38 mm2. Eachbunch of fibers (containing about 3000 fibers, out ofwhich 400 are spares) and two reference PIN siliconphotodiodes will be contained in this area. Severallight pulsers will serve the whole PANDA calorime-ter.

The time dependence of pulse heights from thepulser is monitored by the two reference PIN photo-diodes. One of the pulser systems will be activatedat any given time to limit the power requirementsof the light source, the size of data transfers, as wellas high and low voltages current demands.

The principal goal of the system is to monitor short-term variation in the photodetector gains and thelight transmission of the crystals. The system willalso be used to check out the entire crystal-readoutchain during the assembly of the calorimeter. Itwill also permit a rapid survey of the full calorime-ter during the installation or after long shutdowns.Furthermore, the light monitoring system can beused to measure the response linearity of the PWOcrystal’s photodetector and its readout chain. Thisshould complement measurements with electroniccharge injection at the preamplifier level which doesnot test the photodetector.

Some results obtained with a prototype system arepresented in Sec. 8.2.4.3 and in Fig. 8.6. In moredetail, the monitoring system prototype design andperformance are discussed in [5].

A picture of the whole prototype system is pre-sented in Fig. 8.7. A LED driver is presented inFig. 8.8.


Figure 8.6: Stability of the LED pulser prototype: (a),(b) - behaviour in time of blue and red LED signalscorrespondingly detected by one of the photodiodes over one week of measurements. Each entry is a mean valueof amplitude distribution collected over 20 min; (c),(d) - normalized projections on the vertical axis of the diagrams(a) and (b). R.m.s. characterizes instability of the system over the period of measurements.

8.2.4.1 Monitoring System Components

The monitoring system will be located directly atthe outer radius of the calorimeter support struc-ture and an optical-fiber light distribution systemconnects the pulser to the crystals. Since the lightpulser is located in a low radiation zone, its com-ponents and electronics are not required to be ra-diation hard. In contrast, many fibers are routedthrough a high radiation zone and they must bemade of radiation-hard materials.

The characteristics of the LEDs we plan to usein the PANDA calorimeter monitoring system aregiven in Table 8.2.4.1.

Besides the exceptional luminous fluxes, we findthat two additional features of the Luxeon technol-ogy are very important for our monitoring system:very long operating lifetime (up to 100,000 hoursin DC mode), and small temperature dependenceof the light output (≈ 0.1%/C). The producer isLumileds Lighting, USA. The reflector which wasmade at IHEP has a trapezoidal shape and is made


Property blue (royal blue) LED red LEDBrand Luxeon 5-W emitter Luxeon 1-W emitter

Typical Luminous flux 30 lm (@700 mA) 45 lm (@350 mA)Radiation Pattern Lambertian Lambertian

Viewing Angle 150 140

Size of Light Emission Surface 5 × 5 mm2 1.5×1.5 mm2

Peak Wavelength 470 nm (455 nm) 627 nmSpectral Half-width 25 nm (20 nm) 20 nm

Average Forward Current 700 mA 350 mA

Table 8.1: Properties of LEDs.

Figure 8.7: View of the prototype monitoring system.Inside the box there is a LED driver, blue and red LEDs,a light mixer, a temperature stabilization system, anda referenced PIN-diode system. In use we have a fiberbunch coming out the far side of the box instead of thecables which are pictured; the cables are for tests onlyand will not go to the calorimeter. The size of the boxis 370 mm x 70 mm x 60 mm.

Figure 8.8: LED driver of the Prototype of the moni-toring system. It will be inside the box (see the previousFigure) near the side opposite to the one with a bunchof fibers.

of aluminum plated Mylar or Tyvek.

The optical fibers we plan to use are produced byPolymicro Technologies, USA. Their properties are:

• Silica / Silica optical fiber

• High - OH Core

• Aluminum Buffer

• Core Diameter 270 µm

• Outer Diameter 400 µm

• Numerical Aperture 0.22

• Full Acceptance Cone 25.4

This fiber has very good radiation hardness. Ac-cording to the tests made by the CMS ECAL group,this fiber has shown no signal degradation undergamma irradiation with an absorbed dose of up to12 Mrad.

8.2.4.2 Reference PIN Silicon Photodiodes

An essential element of the light monitoring sys-tem is a stable reference photodetector with goodsensitivity at short wavelengths. PIN silicon pho-todiodes with a sensitive area of about 6 mm2 arewell suited for this task. In particular, such lowleakage currents are achieved with PIN diodes, dueto their very narrow depletion zone resulting fromheavy (p and n) doping, which is less sensitive tothe type inversion than typical PIN diodes. Therather large sensitive area of this photodiode allowsus to work without preamplifiers and improve thestability of the reference system itself. A PIN sili-con photodiode S1226-5BQ (Hamamatsu) was usedin our test measurements. It has an active area of2.4×2.4 mm2 and a dark current less than 50 pA(at 5 V reverse-bias voltage).


Figure 8.9: Schematic view of the light pulser proto-type.

Figure 8.10: Distribution of pulse heights (in mV)measured over an area of 34 × 34 mm2 in 2 mm steps.

8.2.4.3 Tests of the Light Pulser Prototype

A schematic view of the light pulser prototype isshown in Fig. 8.9. The light distribution unifor-mity was measured with a single fiber scanner. Allthe measurements were made with a scope and amanual scan with a step size of 2 mm. The scanarea was 34×34 mm2. The results are shown inFig. 8.10. The FWHM of this pulse height distri-bution is 2%, and the full width is 8%. The energyequivalent is 20 GeV for the whole scan area. Theaverage forward current in the tests was 20 mA. Themaximum forward current is 700 mA. So we havea large safety factor for the amount of light. Thislight pulser can illuminate more than 3000 fibers.

The short-term stability of light uniformity over thearea of 34×34 mm2 for a day has been measured tobe 0.05% and a long-term stability was 0.1% over20 days. In spite of these encouraging results, ther-mostabilisation of the light pulser by means of thePeltier cell is foreseen in the design of the wholesystem.

References

[1] I. Augustin, Search for Scalar Gluonium inAntiproton-Proton Annihilations at Rest, PhDthesis, 1992.

[2] E. Aker et al., Nucl. Instrum. Meth. A321, 69(1992).

[3] A. Fyodorov, M. Korzhik, A. Lopatik, andO. Missevitch, Nucl. Instrum. Meth. A413,352 (1998).

[4] V. A. Kachanov et al., Nucl. Instrum. Meth.A314, 215 (1992).

[5] V. A. Batarin et al., Nucl. Instrum. Meth.A556, 94 (2006).


141

9 Simulations

The simulation of the EMC and the facilitated soft-ware is described in this chapter. The goal of thestudies described in the following is the expectedperformance of the planned EMC. We focus here onthe energy and spatial resolution of reconstructedphotons, the capability of an electron hadron sepa-ration, and also the feasibility of the PANDA physicsprogram. A couple of benchmark studies will bepresented. These accurate simulations highlight thenecessity of the planned EMC. Due to the fact thatmost of the physics channels have very low crosssections - typically between pb and nb - , a back-ground rejection power up to 109 has to be achieved.This requires an electromagnetic calorimeter whichallows an accurate photon reconstruction within theenergy range between 10 MeV and 15 GeV, and aneffective and an almost clean electron hadron sepa-ration.

9.1 Offline Software

The offline software has been devised for detaileddesign studies of the PANDA detector and for thepreparation of the PANDA Physics Book, which willbe published by September 2008. It follows an ob-ject oriented approach, and most of the code is writ-ten in C++. Several proofed software tools andpackages have been adapted from other HEP exper-iments to the PANDA needs. The software contains

• event generators with proper decay models forall particles and resonances involved in the in-dividual physics channels as well as in the rel-evant background channels,

• particle tracking through the complete PANDAdetector by using the GEANT4 transport code[1, 2],

• a digitization which models the signals and thesignal processing in the front-end-electronics ofthe individual detectors,

• the reconstruction of charged and neutral par-ticles, comprising a particle identification thatprovides lists of particle candidates for the finalphysics analysis, and

• user friendly high level analysis tools with thepurpose to make use of vertex and kinematicalfits and to reconstruct even complicate decaytrees in an easy way.

The simulations have been done with the com-plete setup which was already described in detailin Sec. 2.3. The still not finally established TimeOf Flight and the Forward RICH detectors havenot been considered and the Straw Tube option hasbeen used for the central tracker device.

9.1.1 Photon Reconstruction

9.1.1.1 Reconstruction Algorithm

A photon entering one crystal of the EMC devel-ops an electromagnetic shower which, in general,extends over several crystals. A contiguous area ofsuch crystals is called a cluster.

The energy deposits and the positions of all crys-tals hit in a cluster allow a determination of thefour vector of the initial photon. Most of the EMCreconstruction code used in the offline software isbased on the cluster finding and bump splitting al-gorithms which were developed and successfully ap-plied by the BaBar experiment [3, 4].

The first step of the cluster reconstruction is thefinding of a contiguous area of crystals with en-ergy deposit. The algorithm starts at the crystalexhibiting the largest energy deposit. Its neigh-bors are then added to the list of crystals if theenergy deposit is above a certain threshold Extl.The same procedure is continued on the neighborsof newly added crystals until no more crystal ful-fills the threshold criterion. Finally a cluster getsaccepted if the total energy deposit in the contigu-ous area is above a second threshold Ecl.

A cluster can be formed by more than one particleif the angular distances of the particles are small.In this case the cluster has to be subdivided intoregions which can be associated with the individ-ual particles. This procedure is called the bumpsplitting. A bump is defined by a local maximuminside the cluster: The energy deposit of one crystalELocalMax must be above Emax, while all neigh-bor crystals have smaller energies. In addition thehighest energy ENMax of any of the N neighboringcrystals must fulfill the following requirement:

0.5 (N − 2.5) > ENMax /ELocalMax (9.1)

The total cluster energy is then shared between thebumps, taking into account the shower shape of thecluster. For this step an iterative algorithm is used,


which assigns a weight wi to each crystal, so thatthe bump energy is defined as Ebump =

∑i wiEi.

Ei represents the energy deposit in the ith crystaland the sum runs over all crystals within the cluster.The crystal weight for each bump is calculated by

wi =Ei exp(−2.5 ri / rm)∑j Ej exp(−2.5 rj / rm)

(9.2)

, with

• rm = Moliere radius of the crystal material,

• ri,rj = distance of the ith and jth crystal tothe center of the bump and

• index j runs over all crystals.

The procedure is iterated until convergence. Thecenter position is always determined from theweights of the previous iteration and convergence isreached when the bump center stays stable withina tolerance of 1 mm.

The spatial position of a bump is calculated via acenter-of-gravity method. Due to the fact that theradial energy distribution originating from a pho-ton decreases mainly exponentially, a logarithmicweighting with

Wi = max(0, A(Ebump) + ln (Ei/Ebump)) (9.3)

was chosen, where only crystals with positiveweights are used. The energy dependent factorA(Ebump) varies between 2.1 for the lowest and3.6 for the highest photon energies.

9.1.1.2 Digitization of the Crystal Readout

Fast FADC will digitize the analog response of thefirst amplification and shaping stage. Therefore sig-nal waveforms are important to be considered in thesimulation. For performance reasons a simplifiedmethod has been applied in the simulation stud-ies. It is based on an effective smearing of the ex-tracted Monte Carlo energy deposits. Reasonableproperties of PbWO4 at the operational tempera-ture of -25C have been considered. A Gaussiandistribution of σ = 1 MeV has been used for theconstant electronics noise. The statistical fluctua-tions were estimated by 80 phe/MeV produced inthe LAAPD. An excess noise factor of 1.38 has beenused which corresponds to the measurements withthe first LAAPD ’normal C’ prototype at an inter-nal gain of M = 50 (see Sec. 5.1.2.2). This resultsin a photo statistic noise term of 0.41% /

√E(GeV).

9.1.1.2.1 Comparison with 3×3 Crystal Ar-ray Measurements In order to demonstratethat the digitization is sufficiently described by thesimplified method, the simulation was comparedwith the results of a measurement with a 3×3 crys-tal array at an operational temperature of 0C (seeSec. 10.1.1.2). Fig. 10.16 and Fig. 10.17 show theobtained line shapes, and Fig. 10.19 the measuredenergy resolution as a function of incident photonenergy. The simulations have been done for a 3×3array of crystals of 2 x 2 x 20 cm3. Photons with dis-crete energies between 40.9 MeV and 1 GeV wereshot to the center of the innermost crystal. To con-sider the crystal temperature of 00C the digitiza-tion has been done with just 40 phe/MeV instead of80 phe/MeV produced in the LAAPD. The result-ing line shape at the photon energy of 674.5 MeVis illustrated in Fig. 9.1. A fit with a Novosibirskfunction defined by

f(E) = AS exp(−0.5ln2[1 + Λτ · (E − E0)]/τ2 + τ2)

,with

• Λ = sinh( τ√

ln 4)/(σ τ√

ln 4),

• E0 = peak position,

• σ= width, and

• τ = tail parameter

yields to a resoultion of σ/E = 2.63 %, whichis in good agreement with the measurements (seeFig. 10.17). Moreover the energy resolution as afunction of the photon energy can be reproducedvery well. A comparison between the measurementsand the simulations are shown in Fig. 9.2, and theresults are consistent within a tolerance of 10%.The resolutions obtained with the simulation aresystematically better, which may be caused by theuncertainties of the estimated electronic noise termand the number of phe/MeV, as well as by inho-mogeneities of the light yield of the crystals, whichwere not taken into account.

9.1.1.3 Reconstruction Thresholds

In order to detect low energetic photons and toachieve a good energy resolution, the three pho-ton reconstruction thresholds should be set as lowas possible. On the other hand the thresholds mustbe sufficiently high to suppress the misleading re-construction of photons from the noise of the crys-tal readout and from statistical fluctuations of theelectromagnetic showers. Based on the propertiesof the PANDA EMC (see Sec. 9.1.1.2) the followingthresholds were chosen:


Entries 10000

A 144.1

mean 0.6265 σ 0.01649

tail -0.2287

[GeV]depE0.5 0.55 0.6 0.65 0.7

coun

ts

0

20

40

60

80

100

120

140

160Entries 10000

A 144.1

mean 0.6265 σ 0.01649

tail -0.2287

=674.5MeVγE

/E=2.63%σ

Figure 9.1: Simulated line shape of the 3×3 crystalmatrix at the photon energy of 674.5 MeV. The fitwith a Novosibirsk function gives an energy resolutionof σ/E = 2.63 %

Figure 9.2: Energy resolution as a function of incidentphoton energy between 40.9 MeV and 1 GeV for a 3×3crystal array. The measurements are represented byblack circles and the simulation results are illustratedwith red rectangles.

• Extl = 3 MeV

• Ecl = 10 MeV

• Emax = 20 MeV .

As already discussed in Sec. 3.1.1.1 the ability toidentify photons down to approximately 10 MeV isextremely important for PANDA. A lot of channels– especially in the charmonium sector (exotic andconventional) like pp → ηc → γ γ, pp → hc →ηc γ or pp → J/ψ γ - require an accurate and cleanreconstruction of isolated photons. The main back-ground channels here have the same final stateswith just the isolated photon being replaced by aπ0. If one low energetic photon from a π0 decay

gets lost, the background event will be misidenti-fied. The cross sections for the background channelsare expected to be orders of magnitudes higher thanfor the channels of interest. Therefore an efficientidentification of π0 is mandatory for this importantpart of the physics program of PANDA. Fig. 3.1 inSec. 3.1.1.1 shows the upper limit for the identifica-tion of π0’s with respect to different photon thresh-olds. While roughly 1% of the π0’s get lost for athreshold of 10 MeV, the misidentification increasesby one order of magnitude for the scenario wherephotons below 30 MeV can not be detected.

9.1.1.4 Leakage Corrections

The sum of the energy deposited in the crystalsin general is a few percent less than the energyof the incident photon. This is caused by showerleakages particularly in the crystal gaps. The re-constructed energy of the photon is expressed as aproduct of the total measured energy deposit anda correction function which depends logarithmicallyon the energy and – due to the layout of the PANDAEMC – also on the polar angle. Single photonMonte Carlo simulations have been carried out todetermine the parameters of the correction functionEγ,cor = E ∗ f(lnE,Θ) with

f(lnE,Θ) = exp(a0 + a1 lnE + a2 ln2E + a3 ln

3E

+a4 cos(Θ) + a5 cos2(Θ) + a6 cos

3(Θ)+a7 cos

4(Θ) + a8 cos5(Θ)

+a9 lnE cos(Θ))

Fig. 9.3 shows the result for the barrel part in theΘ range between 22 and 90.

9.1.1.5 Energy and Spatial Resolution

The energy resolution of the EMC strongly dependson the length of the crystals. A sufficient contain-ment of the electromagnetic shower within the de-tector and marginal fluctuations of the shower leak-ages must be guaranteed for the detection of pho-tons with energies up to 15 GeV. Fig. 9.4 comparesthe simulation results for the foreseen PANDA EMCequipped with crystals of 15 cm, 17 cm and 20 cmlength. While the resolution for low energetic pho-tons – below 300 MeV – is nearly the same for allthree scenarios, the performance becomes signifi-cantly better for higher energetic photons with anincreasing crystal length. The 20 cm setup yieldsin an energy resolution of 1.5% for 1 GeV photons,and of below 1% for photons above 3 GeV. Anotheraspect which favors the 20 cm scenario is the toler-


Figure 9.3: Leakage correction function for the barrelEMC in the Θ range between 22 and 90.

able level of the nuclear counter effect in the photosensors (Sec. 2.3).

Figure 9.4: Comparison of the photon energy resolu-tion for three different crystal lengths. The resolutionfor 15 cm crystals is illustrated by circles, for 17 cm byrectangles, and the 20 cm scenario is shown with trian-gles.

As already described in 9.1.1.2 and 9.1.1.3 thechoice of the single crystal threshold Extl, which isdriven by the electronics noise term, strongly effectsthe resolution. Three different scenarios have beeninvestigated: Fig. 9.5 compares the achievable res-olution for the most realistic scenario with a noiseterm of σ = 1 MeV and a single crystal reconstruc-tion threshold of Extl = 3 MeV with a worse case(σ = 3 MeV, Extl = 9 MeV) and a better case(σ = 0.5 MeV, Extl = 1.5 MeV). While the max-imal improvement for the better case is just 20%for the lowest photon energies, the degradation in

the worse case increases by more than a factor of2. This result demonstrates clearly that the sin-gle crystal threshold has a strong influence on theenergy resolution.

Figure 9.5: Comparison of the energy resolutions forthree different single crystal reconstruction thresholds.The most realistic scenario with a noise term of σ =1 MeV and a single crystal threshold of Extl = 3 MeVis illustrated by triangles, a worse case (σ = 3MeV,Extl = 9 MeV) by circles and the better case (σ =0.5 MeV, Extl = 1.5 MeV) by rectangles.

The high granularity of the planned EMC providesan excellent position reconstruction of the detectedphotons. The accuracy of the spatial coordinatesis mainly determined by the dimension of the crys-tal size with respect to the Moliere radius. Fig. 9.6shows the resolution in x direction for photons upto 3 GeV. A σx-resolution of less than 0.3 cm canbe obtained for energies above 1 GeV. This cor-responds to roughly 10% of the crystal size. Forlower energies the position becomes worse due tothe fact that the electromagnetic shower is con-tained in just a few crystals. In the worst case ofonly one contributing crystal the x-position can bereconstructed within an imprecision of 0.5 - 0.6 cm.

9.1.2 Electron Identification

Electron identification will play an essential role formost of the physics program of PANDA. An accu-rate and clean measurement of the J/ψ decay ine+ e− is needed for many channels in the charmo-nium sector as well as for the study of the p anni-hilation in nuclear matter like the reaction pA →J/ψX. In addition the determination of electromag-netic form factors of the proton via pp → e+ e− re-quires a suppression of the main background chan-nel pp → π+ π− in the order of 108.


Figure 9.6: Position resolution in x-direction for pho-tons below 3 GeV.

The EMC is designed for the detection of photons.Nevertheless it is also the most powerful detectorfor an efficient and clean identification of electrons.The character of an electromagnetic shower is dis-tinctive for electrons, muons and hadrons. Themost suitable property is the deposited energy inthe calorimeter. While muons and hadrons in gen-eral loose only a certain fraction of their kineticenergy by ionization processes, electrons deposittheir complete energy in an electromagnetic shower.The ratio of the measured energy deposit in thecalorimeter to the reconstructed track momentum(E/p) will be approximately unity. Due to the factthat hadronic interactions within the crystals cantake place, hadrons can also have a higher E/p ra-tio than expected from ionization. Figure Fig. 9.7shows the reconstructed E/p fraction for electronsand pions as a function of momentum.

Figure 9.7: E/p versus track momentum for electrons(green) and pions (black) in the momentum range be-tween 0.3 GeV/c and 5 GeV/c.

Furthermore the shower shape of a cluster is help-ful to distinguish between electrons, muons andhadrons. Since the chosen size of the crystals cor-responds to the Moliere radius of lead tungstate,the largest fraction of an electromagnetic showeroriginating from an electron is contained in just afew crystals. Instead, an hadronic shower with asimilar energy deposit is less concentrated. Thesedifferences are reflected in the shower shape of thecluster, which can be characterized by the followingproperties:

• E1/E9 which is the ratio of the energy de-posited in the central crystal and in the 3×3crystal array containing the central crystal andthe first innermost ring. Also the ratio of E9

and the energy deposit in the 5×5 crystal arrayE25 is useful for electron identification.

• The lateral moment of the cluster defined bymomLAT =

Pni=3 Eir

2iPn

i=3 Eir2i +E1r20+E2r20with

– n: number of crystals associated to theshower.

– Ei: deposited energy in the i-th crystalwith E1 ≥ E2 ≥ ... ≥ En.

– ri: lateral distance between the centraland the i-th crystal.

– r0: the average distance between two crys-tals.

• A set of zernike moments [5] which describethe energy distribution within a cluster by ra-dial and angular dependent polynomials. Anexample is given in Fig. 9.8, where the zernikemoment z31 is illustrated for all particle types.

Figure 9.8: Zernike moment z31 for electrons, muonsand hadrons.


Due to the fact that a lot of partially correlatedEMC properties are suitable for electron identifica-tion, a Multilayer Perceptron (MLP) has been ap-plied. The advantage of a neural network is thatit can provide a correlation between a set of in-put variables and one or several output variableswithout any knowledge of how the output formallydepends on the input. Such techniques are also suc-cessfully used by other HEP experiments [6, 7].

The training of the MLP has been done with a dataset of 850.000 single tracks for each particle species(e, µ, π, K and p) in the momentum range between200 MeV/c and 10 GeV/c in such a way that the out-put values are constrained to be 1 for electrons and-1 for all other particle types. 10 input variablesin total have been used, namely E/p, p, the po-lar angle Θ of the cluster, and 7 shower shape pa-rameters (E1/E9, E9/E25, the lateral moment ofthe shower and 4 zernike moments). The responseof the trained network to a test data set of singleparticles in the momentum rage between 300 MeV/cand 5 GeV/c is illustrated in Fig. 9.9. The loga-rithmically scaled histogram shows that an almostclean electron recognition with a quite small con-tamination of muons and hadrons can be obtainedby applying a cut on the network output.

Figure 9.9: MLP output for electrons and theother particle species in the momentum range between300 MeV/c and 5 GeV/c.

The global PID, which combines the PID informa-tions of the individual sub-detectors, has been re-alized with the standard likelihood method. Eachsub-detector provides probabilities for the differentparticle species, and thus a correlation between thenetwork output and the PID likelihood of the EMChas been calculated. Fig. 9.10 shows the electronefficiency and contamination rate as a function ofmomentum achieved by requiring an electron like-

lihood fraction of the EMC of more than 95%. Formomenta above 1 GeV/c one can see that the elec-tron efficiency is greater than 98% while the con-tamination by other particles is substantially lessthan 1%. For momenta below 1 GeV/c instead theelectron identification obtained with just the EMCis quite bad and not sufficient.

Figure 9.10: Electron efficiency and contaminationrate for muons, pions, kaons and protons in differentmomentum ranges by using the EMC information.

A good electron-ID efficiency and small contamina-tion rates can be achieved by taking into consid-eration additional sub-detectors (MVD, Cherenkovdetectors and muon counters) (Fig. 9.11). By ap-plying a 99.5% cut on the combined likelihood frac-tion, the electron efficiency becomes better than98% while the pion misidentification is just on the10−3 level over the whole momentum range. Thecontamination rates for muons, kaons and protonsare negligible.

9.2 Material Budget in frontof the EMC

The reconstruction efficiency as well as the energyand spatial resolution of the EMC are affected bythe interaction of particles with material in frontof the calorimeter. While the dominant interactionprocess for photons in the energy range of interest ise+ e− pair production, electrons lose energy mainlyvia Bremsstrahlung (e → e γ). Fig. 9.12 illustratesthe material budget in front of the EMC originatingfrom the individual sub-detectors in units of radia-tion lengths as a function of Θ. The largest contri-bution comes from the Cherenkov detectors, whichconsist of quartz radiators of 1-2 cm thickness. This


Figure 9.11: Electron efficiency and contaminationrate for muons, pions, kaons and protons in differentmomentum ranges by using the combined PID informa-tions.

corresponds to 17% to 50% of a radiation length,depending on the polar angle.

Figure 9.12: Material in front of the EMC in units ofa radiation length X0 as a function of the polar angleΘ.

9.2.1 DIRC Preshower

As Fig. 9.12 shows, the DIRC detector contributesmost to the material budget in front of the EMC.Therefore detailed Monte Carlo studies have beendone to investigate the impact of this device on theγ reconstruction efficiency and energy resolution.Fig. 9.13 represents the γ conversion probabilityin the DIRC for 1 GeV photons generated homo-geneously in the Θ range between 22 and 145.While 15% of the photons convert at the polar an-gle of 90, the DIRC preshower probability increasesup to 27% at 22.

Figure 9.13: γ conversion probability in the DIRC asa function of Θ.

DIRC preshowers mainly lead to a degradation ofthe energy resolution. A comparison of the re-constructed energies for 1 GeV photons with andwithout DIRC preshowers is shown in Fig. 9.14.While the energy distribution for non-preshowerphotons can be well described by a Gaussianfunction with σ(∆E/E) < 2%, the distributionfor DIRC preshower clusters becomes significantlyworse. Due to the fact that a fraction of the photonenergy is deposited in the quartz bars, a broad andasymmetric distribution with a huge low energy tailshows up.

Figure 9.14: Reconstructed energy of 1 GeV photonswithout (black circles) and with (red triangles) DIRCpreshowers. For a better comparison the plot withDIRC preshowers is scaled by a factor of 4.87.

9.2.1.1 Outlook: Preshower Recognitionand Energy Correction.

If the Cherenkov light originating from the pro-duced e+ e− pairs gets measured, the number of


detected Cherenkov photons provides a measure forthe energy loss, and thus an energy correction ofsuch clusters could be feasible. A DIRC preshowerrecognition with an additional energy correctionwould yield in a better performance of the photonreconstruction.

Based upon recent investigations for the BaBarexperiment, it is expected to achieve a DIRCpreshower detection efficiency of better than 50%and an improvement for the photon energy resolu-tion of more than 1% [8].

9.3 Benchmark Studies

As discussed in the introduction of this document,the PANDA physics program covers many fields andthe addressed topics range from light and charmhadron spectroscopy over the study of charm in nu-clear matter and the measurement of the protonform-factors. To fulfill the requirements defined bythe different fields, the EMC is essential in manycases. To demonstrate that the proposed PANDAdetector and in particular the EMC is able to reachthe defined goals, detailed Monte Carlo studies havebeen performed. In the following the results ofthese benchmark studies will be presented, wherethe EMC plays a major role. These studies covercharmonium spectroscopy as well as the measure-ment of the time-like electromagnetic form-factorsof the proton. In the field of charmonium spec-troscopy recent observations of new states [9] fallinginto the mass range of the charmonium system,underline the importance of the ability to detectthese states in as many decay modes as possible tocompare the observed decay patterns with theoret-ical expectations, and thus understand the natureof these states. For the envisaged comprehensivestudy of the charmonium system at PANDA thisimplies the detection of hadronic, leptonic and ra-diative decay modes, where the final states can con-sist out of charged and neutral particles. Thereforephotons have to be detected with precise energy andspatial resolution over a wide energy range and anexcellent coverage of the solid angle. For the re-construction of leptonic decays to e+e− (i.e., J/ψand ψ(2S) decays) the EMC supplements the PIDcapabilities of the other detector components andprovides together with these a clean identificationof electrons. This is also crucial for the measure-ment of the electromagnetic proton form factors inthe time-like region via the process pp → e+e−,where the rejection of the dominant pp → π+ π−

background requires an excellent separation of elec-trons and pions.

9.3.1 hc Detection with thehc → ηc γ Decay and the Roleof low Energy γ-ray Threshold

9.3.1.1 Description of the Studied Channel

hc is a singlet state of P wave charmonium (11P1)with a mass of M=3526 MeV/c2. One of its maindecay modes is the electromagnetic transition to theground state, ηc, with the emission of a γ with anenergy of Eγ = 503 MeV. This decay mode waspreviously observed by E835 [10] and CLEO [11].hc can be observed exclusively in many decay modesof ηc, neutral (ηc → γ γ) or hadronic. The givenanalysis is based on the decay modes ηc → φφwith a branching fraction of BR = 2.6 · 10−3, andφ → K+K−, BR = 0.49.

9.3.1.2 Background Consideration

The previous observation of hc was done in theneutral decay mode of ηc in pp annihilations [10],or in hadronic decay modes of ηc from e+ e− →Ψ(2S) → π0 hc, hc → ηc γ [11]. In PANDA, wherehc is produced in pp annihilations and the detectoris capable to detect hadronic final states, the mostimportant aspect of the analysis is the evaluation ofsignal to background ratios, because the productionof the hadronic final states is much more enhancedin pp annihilation in comparison to e+e−. For theexclusive decay mode considered in this study:

pp → hc → ηc γ → φφγ → K+K−K+K−γ

the following 3 reactions are considered as the mainbackground contributors:

1. pp→ K+K−K+K−π0,

2. pp→ φK+K−π0,

3. pp→ φφπ0.

With one γ-ray from the π0 decay left undetected,these reactions have the same signature of decayproducts as the studied hc decay.In Fig. 9.15 the energy distributions of γs are pre-sented for the signal and one of the backgroundchannels.

γs from hc decays cover an energy range of[0.15; 2.0] GeV. The corresponding distribution forthe background also covers all this range, but in-creases for small energies. If we want to recover π0sto separate signal from background, the low energy


Figure 9.15: Energy distribution of γs from pp →hc → ηc γ and pp → φφπ0.

γ reconstruction threshold has be as low as possi-ble, as it was already discussed in section 9.1.1.3.Moreover, if one γ from the π0 is low energetic, themomenta of the other γ and the charged hadronsmove closer to the total momentum of the initialpp system. Such events pass cuts on the fit prob-ability of the 4C-fit, and this makes a low energythreshold mandatory for applying anti-cuts on noπ0 in the event, and correspondingly to suppressthe background.

There are no measurements, to our best knowledge,of the cross-sections of the studied background re-actions. The only way to estimate these cross-sections we found was to use the DPM (Dual Par-ton Model) event generator [12]. 107 events weregenerated with DPM at a beam momentum ofpz = 5.609 GeV/c, which corresponds to the studiedhc resonance. The corresponding numbers of eventsare 60 and 6 for the first two background channels.No event from the pp → φφπ0 reaction was ob-served. With a total pp cross-section at this beammomentum of 60 mb, the cross-sections for the cor-responding background channels are estimated to360 nb, 60 nb and below 6 nb, respectively.

9.3.1.3 Event Selection

The following selection criteria were applied to se-lect the signal:

1. constraint on a common vertex of K+ and K−

with φ, pre-fit selection ηc invariant mass win-dow of [2.6;3.2] GeV/c2 and φ mass window of[0.8; 1.2] GeV/c2,

2. 4C-fit to beam momentum, CL > 0.05,

3. ηc invariant mass in [2.9;3.06] GeV/c2,

Figure 9.16: The number of reconstructed EMC clus-ters for the pp → hc → ηc γ and pp → φφπ0 reac-tions.

4. Eγ within [0.4;0.6] GeV,

5. φ invariant mass in [0.97;1.07] GeV/c2,

6. no π0 candidates in the event, i.e. no 2 γ invari-ant mass in the range [0.115;0.15] GeV/c2 with2 different low energy γ threshold options: 30MeV and 10 MeV.

Fig. 9.16 presents the multiplicity distribution of re-constructed EMC clusters for the signal and one ofthe background channels. One may note that themean numbers of clusters, i.e. neutral candidates,significantly exceed one, expected for the signal, ortwo expected for the background from π0 decay.This is caused by hadronic split-offs and preventsthe use of the number of clusters as a selection crite-ria. This observation emphasizes the importance ofother selection criteria, in particular of the anti-cuton no π0 in an event. The latter strongly dependson the assumed low energy γ-ray threshold.

9.3.1.4 Signal to Background Ratio versuslow Energy γ-ray Threshold

The efficiencies of the chosen selection criteria aregiven in Table 9.1 for the signal and the consideredbackground channels.

Assuming an hc production cross-section of 40 nbyields in the signal to background ratios given inTable 9.2.

For the pp → φφπ0 background channel a re-duction of the γ-ray threshold from 30 MeV to 10MeV gives a 20 % improvement in the signal tobackground ratio, for pp → φK+K−π0 the cor-responding improvement is 40 %.


Selection criteria signal 4Kπ0 φK+K−π0 φφπ0

pre-selection 0.51 9.8 · 10−3 1.3 · 10−2 4.9 · 10−2

CL > 0.05 0.36 1.6 · 10−3 2.0 · 10−3 6.8 · 10−3

m(ηc), Eγ 0.34 4.2 · 10−4 5.2 · 10−4 1.7 · 10−3

m(φ) 0.33 1.0 · 10−5 1.2 · 10−4 1.7 · 10−3

no π0(30MeV ) 0.27 1.0 · 10−5 4.5 · 10−5 8.6 · 10−4

no π0(10MeV ) 0.25 5.0 · 10−6 3.0 · 10−5 7.0 · 10−4

Table 9.1: Efficiency of different event selection criteria.

channel signal/backgr. ratiopp→ K+K−K+K−π0 3.5pp→ φK+K−π0 10pp→ φφπ0 > 6

Table 9.2: Signal to background ratio for different hcbackground channels.

For the final signal selection efficiency of 25 % andthe assumed luminosity of L = 1032cm−2s−1 weexpect a signal event rate of 55 events/day.

9.3.2 Y(4260) in Formation

The recently discovered vector-state Y (4260) hav-ing a mass and width of (4259 ± 8+2

−6) MeV/c2 and(88 ± 23+6

−4) MeV [13], is observed in radiative re-turn events from e+e− collisions [14] and direct for-mation in e+e− annihilation [15]. Evidence is alsoreported in B → J/ψ π+ π−K decays [16]. TheY (4260) is found in the J/ψππ and J/ψK+K− de-cay modes. Here the reaction Y (4260) → J/ψ π0 π0

is studied in pp-formation and the decay chain isexclusively reconstructed via J/ψ → e+e− andπ0 → γ γ.

Electron candidates are identified by the global like-lihood selection algorithm described in Sec. 9.1.2.For the reconstruction of J/ψ → e+e− decays atleast one of the lepton candidates must have a like-lihood value L > 0.99 while the other candidateshould fulfill L > 0.85. The corresponding tracks ofthe e+e− candidates are fitted to a common vertexand an accepted candidate should have a confidencelevel of CL > 0.1%. Photon candidates are formedfrom the clusters found in the EMC applying thereconstruction algorithm disussed in Sec. 9.1.1.1,must have an energy > 20 MeV and are combinedto π0 → γ γ candidates.

Accepted J/ψ and π0 candidates are combinedto J/ψπ0π0 candidates, where the π0 candidatesshould not share the same photon candidate. The

final state lepton and photon candidates are kine-matically fitted by constraining the sums of theirspatial momentum components and energies tothe corresponding values of the initial pp system.An accepted J/ψπ0π0 candidate must have a con-fidence level of CL > 0.1% and the invariantmass of the e+e− and γγ subsystems should bewithin [3.07; 3.12] GeV/c2 and [120; 150] MeV/c2, re-spectively. A FWHM of ≈ 14 MeV/c2 (≈ 7 MeV/c2)is obtained for the J/ψ (π0) signal after the kine-matic fit (Fig. 9.17,Fig. 9.17). For the final eventselection J/ψπ0π0 candidates are refitted kinemat-ically by constraining the invariant e+e− mass tothe J/ψ mass and the invariant γγ mass to the π0

mass on top of the constraints applied to the ini-tial pp four-vector. Only events where exactly onevalid J/ψπ0π0 candidate with CL > 0.1% is foundare accepted for further analysis.

To reject background from the reactions pp →J/ψ η η (pp → J/ψ η π0) with η → γ γ the eventsare reconstructed and kinematically fitted similarto the fit with the J/ψ π0 π0 signal hypothesis de-scribed above but assuming the J/ψ η η (J/ψηπ0) hy-pothesis, where η candidates should have an invari-ant mass within [500; 600] MeV/c2. Events whereat least one valid J/ψηπ0 or J/ψ η η candidate withCL > 0.01% is found are rejected.

The reconstruction efficiency after all selection cri-teria is found to be 13.8%. For pp → J/ψ η ηand pp → J/ψ η π0 events the suppression rateis better than 104 and the contamination of theJ/ψπ0π0 signal with events from these reactionsis expected to be negligible unless the cross sec-tions for pp → J/ψ η η and pp → J/ψ η π0 are en-hanced by more than an order of magnitude com-pared to the signal cross section. Another source ofbackground which has been investigated are non-resonant pp → π+ π− π0 π0 events. For this reac-tion an upper limit of 132 µb at 90% CL can be de-rived from the measurement at

√s = 4.351 GeV/c2

[17], 92 MeV/c2 above the Y (4260) resonance. Thecross section for the signal reaction is estimatedfrom Ref. [18] to be in the order of 50 pb. This


Figure 9.17: Invariant a) J/ψ → e+e− and b) π0 → γ γ mass after the kinematic fit with constraints onthe initial pp system as described in the text. The vertical lines indicate the mass windows chosen for furtherselection.

requires a suppression for pp → π+ π− π0 π0 ofat least 107, whereas with the currently availableamount of MC events a suppression better than 108

is obtained.

In summary the reconstruction of the Y (4260) →J/ψ π0 π0 decay mode with one e+e− pair and fourphotons in the final state yields a good detectionefficiency and suppression of the dominant pp →π+ π− π0 π0 background.

9.3.3 Charmonium Hybrid inProduction

Closely connected with charmonium spectroscopy isthe search for predicted exotic hadrons with hiddencharm, such as charmonium molecule, tetra-quarkor hybrid states. For the latter the ground stateis generally expected to be a spin-exotic JPC =1−+ state and lattice QCD calculations predict itsmass in the range between 4100 and 4400 MeV/c2

[19, 20, 21]. In pp annihilation this state can be pro-duced only in association with one or more recoilingparticles. Here the results of a study assuming theproduction of a state having a mass 4290 MeV/c2

and a width of 20 MeV together with a η meson arereported.

Flux-tube model calculations predict for a hybridstate of this mass suppressed decays to open charmwith respect to hidden charm decays [22]. An OZI-allowed decay to hidden charm would be the tran-

sition to χc1 with emission of light hadrons, prefer-able scalar particles [23]. The lightest scalar systemis composed out of two neutral pions in a relatives-wave.

In the study the decay of the charmonium hybrid(labeled as ψg in the following) to χc1 π

0 π0 withthe subsequent radiative χc1 → J/ψ γ decay withJ/ψ decaying to a lepton pair e+e− is considered.The recoiling meson is reconstructed from the de-cay η → γ γ. The total branching fraction for thesubsequent ψg and η decays is given by B(ψg →χc1 π

0 π0)×0.81%, where B(ψg → χc1 π0 π0) is the

unknown branching fraction of the ψg decay.

Photon candidates are selected from the clustersfound in the EMC with the reconstruction algo-rithm explained in Sec. 9.1.1.1. Two photon candi-dates are combined and accepted as π0 and η can-didates if their invariant mass is within the inter-val [115;150] MeV/c2 and [470;610] MeV/c2, respec-tively.

For the reconstruction of J/ψ decays two candidatesof opposite charge, both identified as electrons ap-plying the likelihood selection algorithm describedin Sec. 9.1.2, where one of the candidates shouldhave a likelihood value L > 0.2 and the other avalue L > 0.85, are combined and accepted as J/ψcandidates if their invariant mass is within the in-terval [2.98; 3.16] GeV/c2. The corresponding tracksof the two lepton candidates are kinematically andgeometrically fitted to a common vertex and their


invariant mass is constrained to the nominal J/ψmass.

Afterwards χc1 → J/ψ γ candidates are formedby combining accepted J/ψ and photon candi-dates, whose invariant mass is within the range 3.3-3.7 GeV/c2. From these χc1 π0 π0 η candidates arecreated, where the same photon candidate does notoccur more than once in the final state. The cor-responding tracks and photon candidates of the fi-nal state are kinematically fitted by constrainingtheir momentum and energy sum to the initial ppsystem and the invariant lepton candidates massto the J/ψ mass. Accepted candidates must havea confidence level of CL > 0.1% and the invari-ant mass of the J/ψ γ subsystem should be withinthe range [3.49; 3.53] GeV/c2, whereas the invariantmass of the η candidates must be within the inter-val [530; 565] MeV/c2. A FWHM of ≈ 13MeV/c2 isobserved for the χc1 and η signals (Fig. 9.18) afterthe kinematic fit.

For the final event selection the same kinematic fitis repeated with additionally constraining the in-variant χc1, π0 and η invariant mass to the corre-sponding nominal mass values. Candidates havinga confidence level less than 0.1% are rejected. Ifmore than one candidate is found in an event, thecandidate with the biggest confidence level is chosenfor further analysis.

The reconstruction efficiency after all selection cri-teria is 3.1%. The invariant χc1 π

0 π0 mass isshown in Fig. 9.19. The ψg signal has a FWHMof 27 MeV/c2.

The final state with 7 photons and an e+e− lep-ton pair originating from J/ψ decays has a distinc-tive signature and the separation from light hadronbackground should be feasible. Another source ofbackground are events with hidden charm, in par-ticular events including a J/ψ meson. This type ofbackground has been studied by analyzing pp →χc0 π

0 π0 η, pp → χc1 π0 η η, pp → χc1 π

0 π0 π0 ηand pp → J/ψ π0 π0 π0 η. The hypothetical hy-brid state is absent in these reactions, but the χc0and χc1 mesons decay via the same decay path asfor signal. Therefore these events have a similartopology as signal events and could potentially pol-lute the ψg signal. The suppression after applica-tion of all selection criteria is found to be 4 · 103

(pp → χc0 π0 π0 η), 2 · 104 (pp → χc1 π

0 η η),> 1 · 105 (pp → χc1 π

0 π0 π0 η) and 9 · 104 (pp →J/ψ π0 π0 π0 η). Therefore only low contaminationof the ψg signal from this background reactions isexpected.

9.3.4 Time-like ElectromagneticForm-Factors

The electric (GE) and magnetic (GM ) form factorsof the proton parameterize the hadronic current inthe matrix element for elastic electron scatteringe+e− → pp and in its crossed process pp → e+e−

annihilation. The form-factors are analytic func-tions of the four momentum transfer q2 rangingfrom q2 = −∞ to q2 = +∞. The annihilationprocess allows to access the formfactors in the time-like region (q2 > 0) starting from the threshold ofq2 = 4m2

pc4. Measurements are concentrated to

the region near threshold and few data points inthe q2 range between 8.8 (GeV/c)2 and 14.4 (GeV/c)2

[24]. Thus the determination for low to intermedi-ate momentum transfer remains an open question.This region can be accessed at PANDA betweenq2 = 5 (GeV/c)2 and q2 = 22 (GeV/c)2. The dif-ferential cross section for the reaction pp → e+e−

is given by [25]

dσd cos θ

=πα2(~c)2

8m2p

√τ (τ − 1)

(|GM |2(1 + cos2 θ

)+|GE |2

τ

(1− cos2 θ

)) (9.4)

with τ = s/4m2pc

4, where θ is the angle betweenthe electon and the antiproton beam in the cen-ter of mass system. The factors |GE | and |GM |can be determined from the angular distribution ofpp → e+e− events in dependence of the beam mo-mentum. A measurement of the luminosity will pro-vide a measurement of both factors |GE | and |GM |,otherwise the absolute height of the cross sectionremains unknown and only the ratio |GE |/|GM | isaccessible. However, for a precise determination ofthe form factors a suppression better than 108 of thedominant background from pp → π+π−, having across section about 106 times higher than the signalreaction is mandatory. For the simulations, a modellagrangian for the process pp → π+π− has beencreated and fitted to the data [26, 27, 28, 29, 30]in order to get cross section predictions at angularregions, where no data exist.

To achieve the required background suppression agood separation of electrons and pions over a widemomentum range up to ≈ 15 GeV/c is mandatory,where the PID information from the individual de-tector components has to be exploited. In particu-lar information from the EMC is important in thehigh momentum range where PID measurementsfrom the tracking, Cherenkov and muon detectorsdo not allow to distinguish between the two species.


Figure 9.18: Invariant a) χc1 → J/ψ γ and b) η → γ γ mass after the kinematic fit with constraints on the initialpp system as described in the text. The vertical lines indicate the mass windows chosen for further selection.

Figure 9.19: Invariant χc1 π0 π0 mass after the kine-

matic fit with constraints on the initial pp system andresonances as described in the text.

For the reconstruction of pp → e+e− two candi-dates identified as electrons (see Sec. 9.1.2) havinga likelihood value > 0.998 are combined. The twotracks are kinematically fitted by constraining thesum of the four-momenta to the four-momentum ofthe initial pp system and are accepted as a e+e−

candidate if the fit converges. The same fit is re-peated but assuming pion hypothesis for the twotracks. For events where more than one valid e+e−

candidate is found, only the candidate with thebiggest confidence level is considered for further

analysis.

For the measurement of |GE | and |GM | the angle θbetween the e+ candidate’s direction of flight andthe beam axis is computed in the center of masssystem. The obtained cos(θ) distribution is cor-rected by the cos(θ)-dependent reconstruction effi-ciency. In lack of an absolute luminosity measure-ment for MC data only the ratio |GE |/|GM | can bedetermined from a fit to the corrected cos(θ) distri-bution. In Fig. 9.20 the corrected distribution for√s = 3.3 GeV/c is shown exemplarily, whereas a

ratio |GE |/|GM | of 0, 1 and 3 has been assumed inthe simulation. The values −0.14±0.07, 1.04±0.02and 2.98 ±0.03 derived from a fit to these distribu-tions are in good agreement with the input valuesof the simulation.

Currently, 108 MC events are available for the back-ground process pp → π+π− at a beam momen-tum of 3.3 GeV/c and additionally the same amountfor the same process at a beam momentum of7.9 GeV/c. Figure Fig. 9.21 shows the rejection ofpp → π+ π− when we apply different cuts based onPID and kinemtical fits. We find 2 out of 108 eventsmisinterpreted as electron positron pairs at a beammomentum of 3.3 GeV/c based on PID only and 0events out of 108 events misinterpreted if we addkinematical constraints. The respective numbersfor a beam momentum of 7.9 GeV/c are 8 events mis-interpreted out of 108 from PID only, and 0 eventsout of 108 when adding kinematical constraints.


Figure 9.20: Angular distribution for reconstructedpp → e+e− events at an incident beam momentum of3.3 GeV/c after efficiency correction. For the ratio of theform-factors |GE |/|GM | a value of 0 (black), 1 (red) and3 (blue) has been assumed in the simulation.

Figure 9.21: Angular distribution for reconstructedpp → π+π− events at an incident beam momentumof 3.3 GeV/c. Different colors represent different cuts.When requiring a probability to be an electron greaterthan 99.8 % (from a combined PID analysis of EMC,DIRC, MVD and STT), we have only two out of 108 sim-ulated and reconstructed pp → π+π− events which aremisinterpreted as an electron positron pair. If we addthe constraints from the kinematical fit, there are nomisidentified pp → π+π− events left over out of the 108

simulated pp → π+π− events. In the case of 7.9 GeV/cbeam momentum 8 events out of 108 pp → π+π−

events are left over after PID cuts. Those 8 events arealso rejected by adding the constraints from the kine-matical fit.

References

[1] J. Allison et al., IEEE Transactions onNuclear Science 53, 270 (2006).

[2] S. Agostinelli et al., Nucl. Instrum. Meth.A506, 250 (2003).

[3] B. Aubert et al., Nucl. Instrum. Meth. A479,1 (2002).

[4] P. Strother, Design and application of thereconstruction software for the BaBarcalorimeter, PhD thesis, 1998, University ofLondon and Imperial College, UK.

[5] F. Zernike, Physica 1 , 689 (1934).

[6] H. Abramowicz, A. Caldwell, and R. Sinkus,Nucl. Instrum. Meth. A365, 508 (1995).

[7] V. Breton et al., Nucl. Instrum. Meth. A362,478 (1995).

[8] A. Adametz, Preshower Measurement withthe Cherenkov Detector of the BaBarExperiment, 2005, Diploma Thesis,University of Heidelberg, Germany.

[9] E. S. Swanson, Phys. Rept. 429, 243 (2006),and references therein.

[10] M. Andreotti et al., Phys. Rev. D72 (2005).

[11] J. Rosner et al., Phys. Rev. Lett. 95 (2005).

[12] A. Capella et al., Phys. Rept. 236 (1994).

[13] W.-M. Yao et al., J. Phys. G33, 1 (2006).

[14] B. Aubert et al., Phys. Rev. Lett. 95, 142001(2005).

[15] T. E. Coan et al., Phys. Rev. Lett. 96,162003 (2006).

[16] B. Aubert et al., Phys. Rev. D73, 011101(2006).

[17] E. J. Flaminio, D. J. Hansen, D. R. O.Morrison, and N. Tovey, Compilation ofCross Section, 1970, CERN/HERA 70-5.

[18] M. Negrini, Measurement of the branchingratios ψ′ → J/ψX in the experiment E835 atFNAL, PhD thesis, 2003,FERMILAB-THESIS-2003-12.

[19] C. W. Bernard et al., Phys. Rev. D56, 7039(1997).

[20] F. E. Close et al., Phys. Rev. D57, 5653(1998).

[21] P. R. Page, Acta Phys. Polon. B29, 3387(1998).


[22] P. R. Page, Phys. Lett. B402, 183 (1997).

[23] C. Michael, Beyond the quark model ofhadrons from lattice QCD, 2002,International Conference On Quark NuclearPhysics.

[24] M. Ambrogiani et al., Phys. Rev. D60,032002 (1999), and references therein.

[25] A. Zichichi et al., Nuovo Cim. 24, 170 (1962).

[26] E. Eisenhandler et al., Nucl. Phys. B 96, 109(1975).

[27] T. Buran et al., Nucl. Phys. B 116, 51 (1976).

[28] T. Berglund et al., Nucl. Phys. B 137, 276(1978).

[29] R. Dulude et al., Phys. Lett 79B, 329 (1978).

[30] T. Armstrong et al., Phys. Rev. D 56, 2509(1997).


157

10 Performance

As documented by the intrinsic performance of in-dividual crystals of the present quality of PWO-II,such as luminescence yield, decay kinetics and ra-diation hardness and the additional gain in lightyield due to cooling down to T=-25C, there ispresently no alternative scintillator material for theelectromagnetic calorimeter of PANDA besides leadtungstate. The general applicability of PWO forcalorimetry in High-Energy Physics has been pro-moted and finally proven by the successful realiza-tion of the CMS/ECAL detector as well as the pho-ton spectrometer ALICE/PHOS, both installed atLHC.

The necessary mass production of high-quality crys-tals has been achieved primarily at BTCP andNorth Crystals in Russia and SICCAS in China.However, driven by the much more stringent re-quirements to perform the Physics program ofPANDA, from the beginning it became mandatoryto search for a significantly higher scintillation yieldin order to extend photon detection down to theMeV energy range. Beyond the realization of anearly perfect crystal the operation at temperatureswell below room temperature will reduce the effectof the thermal quenching of the scintillation pro-cess and improve the photon statistics. The spec-trometer PHOS is following the same line. Our re-quirements are even more stringent, since the targetspectrometer EMC of PANDA has to cover nearlythe full solid angle and stresses even more the con-cept of thermal insulation and cooling.

The test experiments to prove the concept and ap-plicability of PWO have concentrated on the re-sponse to photons and charged particles at energiesbelow 1 GeV, since those results are limited by thephoton statistics of the scintillator, the sensitivityand efficiency of the photo sensor and the noise con-tributions of the front-end electronics. The conclu-sions are drawn based on crystal arrays comprisingup to 25 modules. The individual crystals have a fi-nal length of 200 mm and a rectangular cross sectionof 20× 20 mm2. Only the most recently assembledarray PROTO60 consists of 60 crystals in PANDAgeometry. The tapered shape will improve the lightcollection due to the focusing effect of the geometryas known from detailed simulations at CMS/ECAL.

The performance tests completed up to now havebeen aiming at two complementary aspects. On onehand, the quality of full size PWO-II crystals hasto be verified by in-beam measurements with en-

ergy marked photons covering the most critical en-ergy range up to 1GeV. Therefore, the scintillatormodules were readout with standard photomulti-plier tubes (Philips XP1911) with a bi-alkali photo-cathode, which covers ∼35 % of the crystal endfacewith a typical quantum efficiency of QE=18 %. Thenoise contribution of the sensor can be neglectedand the fast response allows an estimate of the timeresponse. All achieved resolutions, which have beendeduced at various operating temperatures, can beconsidered as benchmark limits for further studiesincluding simulations and electronics development.

The second R&D activity was aiming to comeclose to the final readout concept with largearea avalanche photodiodes (LAAPD), which aremandatory for the operation within the magneticfield. The new developments of the sensor as wellas the parallel design of an extremely low-noiseand low-power preamplifier, based on either discretecomponents or customized ASIC technology, are de-scribed in detail in separate chapters. All the re-ported results are performed by collecting and con-verting the scintillation light with a single quadraticLAAPD of 10 × 10 mm2 active area with a quan-tum efficiency above 60 %. The final readout con-siders two LAAPDs of identical surface but rect-angular shape to fit on the crystal endface, whichis presently prevented by the dead space of the ce-ramic support. The reported experiments are usingin addition individual low-noise preamplifiers andcommercial electronics for the digitization. Again,the present data deliver a lower limit of the finalperformance, which will be achieved with twice thesensor surface.

In order to simulate the operation of large arrays,the mechanical support structures, cooling and tem-perature stabilization concepts and long term sta-bilities, a large prototype comprising 60 taperedcrystals in PANDA-geometry has been designed andbrought into operation. First in-beam tests arescheduled in summer 2008. In a next iteration,a mechanical prototype for housing 200 crystals ispresently under construction. This device is primar-ily meant for studying stable operation and coolingsimulating two adjacent detector slices of the barrelsection of the EMC.

In spite of some compromises compared to the fi-nal concept, the achieved resolutions represent ex-cellent lower limits of the performance to be ex-pected. Operation at T=-25C using a photomulti-


plier readout delivers an energy resolution of σ/E =0.95 %/

√E + 0.91 % (E given in GeV) for a 3×3

sub-array accompanied with time resolutions belowσ=130 ps. The complementary test using a singleLAAPD for readout reaches even at T=0C a fullysufficient resolution of σ/E = 1.86 %/

√E + 0.65 %

(E given in GeV). Timing information can be ex-pected with an accuracy well below 1 ns for energydepositions above 100 MeV. Summarizing, accord-ing to the detailed simulations and the selected testresults, operation of the calorimeter at T=-25Cwill allow to perform the very ambitious researchprogram, even if radiation damage at most for-ward directions might reduce asymptotically thelight output by up to 30 %.

10.1 Results from PrototypeTests

10.1.1 Energy Resolution of PWOarrays

Almost in parallel to the research project forCMS/ECAL, investigations have been started toexplore the applicability of PWO in the low andmedium energy regime. Beside the response to lowenergy γ-rays of radioactive sources the initial pilotexperiments used electrons as well as energy markedphotons at the tagging facility of MAMI at Mainz[1] between 50 and 850MeV energy [2, 3]. The crys-tals were rectangular of 150 mm length and with aquadratic cross section of 20×20mm2 with all sidesoptically polished. The quality was similar to CMSsamples of the pre-production runs and containedslow decay components on the percent level due toMo-contamination. The readout of the scintilla-tion light was achieved with photomultiplier tubesand commercial electronics of NIM and CAMACstandard. In most of the cases the anode signalswere carried via RG58 coaxial cables over a dis-tance of 30− 50 m outside the experimental area tothe DAQ system. For practical reasons, energy andtime information were deduced and digitized afterlong passive delays, causing a significant signal at-tenuation and loss of the high-frequency response.The compact detector arrays were operated at astabilized temperature, always above T=0C butslightly below room temperature. Keeping the crys-tals at a fixed temperature had to guarantee stableoperation, not with the intention to increase theluminescence yield.

The response to monoenergetic photons showedfor the first time an excellent energy resolution

even for photons well below the expected rangefor CMS/ECAL. Resolution values of σ/E =1.54 %/

√E+0.30 % (E given in GeV) were achieved

for a matrix of 5×5 elements and in addition timeresolutions of σt≤130ps for photon energies above25 MeV [4].

Besides the studies using low and high-energy pho-tons exploratory experiments at KVI, Groningen,and COSY, FZ Julich, have delivered results forlow and high energy protons and pions [5]. How-ever, in all these cases PWO of standard qual-ity according to the CMS specifications was used.For protons, which are completely stopped in acrystal of 150 mm length, an energy resolution ofσ/E = 1.44 %/

√E + 1.97 % (E given in GeV) has

been extracted as an upper limit, since the protonenergy had to be deduced from the time of flightmeasured using a fast plastic start counter. For90 MeV protons typical resolutions between 4 % and5 % were obtained.

In the following chapters, the obtained experimen-tal results are based on scintillator crystals, whichcorrespond to the performance parameters of PWO-II [6].

10.1.1.1 Energy Response Measured withPhotomultiplier Readout

As outlined in the previous chapters on PWO(Sec. 4.1) the specifications of the calorimeter canbe further optimized by the reduction of the ther-mal quenching of the relevant scintillation processby cooling the crystals down to temperatures as lowas T=-25C, which leads to a significant increase ofthe light yield by a factor 4 compared to T=+25C.For a typical PWO-II crystal 500 photons can becollected at the endface of a 200 mm long crystalfor an energy deposition of 1 MeV at 100 % quan-tum efficiency.

The reported experiments [7] are based on large sizecrystals of 200 mm length grouped in arrays com-posed of 3×3 up to 5×5 crystals manufactured atBTCP or comparable in quality grown at SICCAS.All crystals were rectangular parallelepipeds to al-low in addition a direct comparison to initial pilotexperiments with limited quality as stated in theprevious chapter.

The temperature stabilization and the operation atsignificantly lower temperatures in an atmospherefree of moisture required the installation of a newexperimental setup, which comprises an insulatedcontainer of large volume to house the different de-tector arrays with high flexibility. A computer con-


Figure 10.1: The experimental setup at the taggedphoton facility at MAMI comprising the cooling unit(right), the connection pipe to circulate cold dry airand the insulated container for the test detectors (left).

Figure 10.2: The insulated container to house the var-ious test arrays. The cold air is circulated via the twoblack pipes.

trolled cooling machine circulates cold and dry airof constant temperature via thermally insulated 5 mlong pipes of large diameter to the detector con-tainer. Fig. 10.1, Fig. 10.2 and Fig. 10.3 show theexperimental installation as well as details of a typi-cal detector matrix mounted in the A2 hall at Mainzbehind the CB/TAPS experiment.

The setup shown particularly in Fig. 10.3 wasused to test the inner 3×3 section of a 5×5 ma-trix of 200 mm long crystals from BTCP, individ-ually readout via photomultiplier tubes (PhilipsXP 1911). The measurement was performed atT=+10C and T=-25C, respectively, to illustratethe effect of cooling. The advantage of reduced tem-peratures becomes immediately obvious in Fig. 10.4in the overlay of the response of the central de-

Figure 10.3: A typical 5×5 PWO matrix shown in-side the insulated box including photomultiplier read-out, signal and high voltage cables.

Figure 10.4: Lineshape of the central detector mod-ule measured at eight different photon energies and twooperating temperatures. The change of the light yielddue to the difference in the thermal quenching of thescintillation light can be deduced directly.

tector to eight different photon energies between64 MeV and 520 MeV, respectively, measured at thetwo operating temperatures. The light-output wasmeasured under identical experimental conditionsof the photomultiplier bias and electronics settingsand documents a gain factor of 2.6 of the overallluminescence yield.

Figure 10.5 and Fig. 10.6 show for the low temper-ature of T=-25C with reduced statistics the line-shape of the central detector, the total energy de-position in the surrounding ring of eight crystalsand the integral of the full array for the two ex-treme photon energies. In particular at the high-est energy, the persisting low energy tail indicatesthe significant lateral shower leakage, which lim-


Figure 10.5: Response function of the 3×3 matrix tophotons of 63.8 MeV energy measured at T=-25C.

Figure 10.6: Response function of the 3×3 matrix tophotons of 520.8 MeV energy measured at T=-25C.

its the obtainable resolution. To parameterize theachieved energy resolution the reconstructed line-shapes of the electromagnetic showers have been fit-ted with an appropriate function to determine theFWHM. Finally, an excellent energy resolution ofσ/E = 0.95 %/

√E + 0.91 % (E given in GeV) was

deduced, which represents the best resolution evermeasured for PWO (see Fig. 10.7). A value wellbelow 2 % extrapolating to an incident photon en-ergy of 1 GeV is very similar to the performance ofseveral operating EM calorimeters, which are basedon well known bright scintillator materials such asCsI(Tl), NaI(Tl) or BGO, respectively.

In comparison, the reduced energy resolution ofσ/E = 1.74 %/

√E + 0.70 % (E given in GeV) de-

duced at T=+10C is consistent with the lower lightoutput due to thermal quenching and is quantita-tively expressed by the higher statistical term in the

Figure 10.7: Comparison of the energy resolution of a3×3 PWO-II matrix of 200 mm long crystals measuredat two different operating temperatures of T=-25C andT=+10C, respectively. The crystal responses werereadout individually with photomultiplier tubes.

parametrization of the resolution [8].

In order to extend and quantify experimentallythe response function to even lower energies, testswere performed at the MAX-Lab laboratory atLund, Sweden. The tagging facility deliverslow-energy photons in the energy range between19 MeV and 50 MeV with an intrinsic photon res-olution far below 1 MeV. The rectangular crys-tals of 20×20×200mm3 were individually wrappedin Teflon and coupled to a photomultiplier tube(Philips XP 1911). The detector array was mountedin a climate chamber and cooled down to a temper-ature of T=-25C. The assembled detector matrixas well as the cooling unit accommodating the setupare shown in Fig. 10.8.

The individual modules have been calibrated usingthe energy loss of minimum ionizing muons in orderto reconstruct the electromagnetic shower withinthe matrix. Fig. 10.9 illustrates the energy dis-tribution within the array for an incident photonenergy of 35 MeV. Nearly Gaussian-like lineshapeshave been deduced for energies as low as 20 MeV asillustrated in Fig. 10.10.

Fig. 10.11 summarizes the obtained energy resolu-tions over the investigated range of low incidentphoton energies and documents the excellent perfor-mance even at the lowest energies to be studied withthe target spectrometer EMC. One should considerthat the readout with a standard photomultiplierrepresents only a lower limit compared to the fu-ture readout with two LAAPDs. They will converta significantly higher percentage of the scintillationlight, which should have a strong impact on photon


Figure 10.8: The 3×3 PWO-II matrix (left) and theclimate chamber (right) for detector operation at lowtemperatures as installed for response measurements atMAX-Lab.

Figure 10.9: Energy deposition of a 35 MeV incidentphoton in a 3×3 PWO-II matrix measured at a temper-ature of T=-25C at MAX-Lab.

detection in particular at very low energies.

10.1.1.2 Energy Response Measured withLAAPD Readout

The operation of the calorimeter inside the solenoidwill require a readout with LAAPD or equivalentsensors, which are not affected by strong magneticfields. As outlined in more detail in the relevantchapters, a new low-noise and compact charge sen-sitive preamplifier was developed and used in pro-totype tests. As a test of principle, with low energyγ-sources and small PWO-II crystals, cooled downto T=-25C, a clean identification of the photopeakof γ-rays as low as 511 keV was demonstrated, thusindicating the low level of noise contributions.

Figure 10.10: Response of a 3×3 PWO-II matrix toincident photons of 20 MeV energy measured at a tem-perature of T=-25C at MAX-Lab. A Gaussian-likelineshape is shown for comparison.

Figure 10.11: Experimental energy resolution of a3×3 PWO-II matrix in the energy range between19 MeV and 50 MeV, respectively, measured at a tem-perature of T=-25C at MAX-Lab.

In a first in-beam test [9] in 2004 at the taggedphoton facility of MAMI at Mainz, a 3×3 matrix of150 mm long crystals of PWO-II quality was op-erated at T=-25C. All modules were read outwith 10×10mm2 LAAPD manufactured by Hama-matsu [10]. The charge output was collected inthe new charge sensitive preamplifier and passivelysplit afterwards to integrate the signals in a spec-troscopy amplifier as well as to deduce a timing in-formation from a constant fraction discriminator af-ter shaping in a timing filter amplifier (INT=50 ns,DIFF=50 ns). Both signals were transferred via50 m coaxial cables to the DAQ system for digitiza-tion.

In spite of the long transfer lines and the miss-


Figure 10.12: Response function of the 3×3 PWO-IImatrix to photons of 63.8 MeV and 520.8 MeV energymeasured at T=-25C. The 150 mm long crystals arereadout with LAAPDs.

Figure 10.13: Energy resolution of a 3×3 PWO-IImatrix of 150 mm long crystals measured at T=-25Creadout with LAAPDs.

ing option to optimize the adjustments of theelectronics an excellent result has been achieved.Fig. 10.12 shows the obtained lineshape at 64 MeVand 520 MeV incident photon energy, respectively,after reconstruction of the electromagnetic showerdeposited within the array. It should be mentionedthat even one of the detector modules was not func-tioning at all. The deduced energy resolution issummarized in Fig. 10.13 and can be parameter-ized as σ/E = 0.90 %/

√E ⊕ 2.13 % (E given in

GeV). The low value of the statistical factor reflectsthe high electron/hole statistics due to the signifi-cantly larger quantum efficiency of the avalanchediode compared to a standard bialkali photocath-ode. The large constant term can be related to themissing detector, reduced signal/noise ratio due to

Figure 10.14: Time resolution of a 3×3 PWO-II ma-trix of 150 mm long crystals measured at T=-25C read-out with LAAPDs.

passive splitting and the not yet optimized adjust-ments of the used commercial electronics.

The used readout represents a first attempt to de-duce as well timing information. In order to identifyoff-line the selected photon energies the trigger sig-nal of the responding scintillators of the tagger lad-der had to be in coincidence. The recorded time in-formation serves as a reference with an intrinsic res-olution of approximately 1.2 ns (FWHM). In spiteof the by far not optimized timing measurement, aresolution of σt1 ns can be reached already abovethe typical energy deposition in the central crys-tal for an incident photon energy of ≥200 MeV.The result emphasizes the excellent timing capa-bilities of PWO even for a readout via LAAPDs.The achieved time resolutions are summarized inFig. 10.14.

In order to investigate a larger range of operationaltemperature a 3×3 matrix of 200 mm long PWO-II crystals, produced in 2007 under more strin-gent specification limits, was tested at a temper-ature of T=0C [11]. In addition, the rectangu-lar crystals were covered for the first time withone layer of a thermally molded reflector foil con-sisting of the 3M radiant mirror film VM2000TM

and mounted in a container made of carbon fibre,similar to the alveole-structure to be foreseen forthe final assembly of the EMC. Detailed tests withrespect to the properties of VM2000 show an in-crease of the collected light of 10–15 % compared toa wrapping with several layers of Teflon. The ma-trix was mounted in an insulated container, whichcould be cooled down to a low temperature wellstabilized (∆T<0.1C) by circulating cold and dryair similar to the setup as shown in Fig. 10.1 and


Fig. 10.2. A thin plastic scintillator covering the to-tal front of the prototype served as a charged parti-cle veto to reject electrons and/or positrons due toconverted photons upstream of the detector. Thecrystal matrix could be moved remote controlledin two dimensions perpendicular to the axis of thecollimated photon beam by stepping-motors to per-form a relative calibration of each detector elementunder beam conditions. Each of the optically pol-ished crystals was coupled to a single LAAPD usingoptical grease (BAYSILONE 300.000). The usedprototype-2 version (S8664-1010SPL) has an activearea of 10×10mm2, a quantum efficiency of 70 %(at 420nm), a dark current of 10 nA at a gain of 50and a capacitance of 270 pF at a gain of 50 and ata frequency of 100 kHz. The charge signal of theLAAPD was amplified with a low-noise and low-power charge sensitive preamplifier with a sensitiv-ity of 0.5 V/pC and a small feedback time constantof 25µs.

The output signal of the preamplifier was fedinto a 16-fold multi-functional NIM module(MESYTEC, MSCF-16), which provided a remotecontrollable multi-channel spectroscopic amplifier(1µs Gaussian shaping), timing filter amplifier(INT=DIFF=50 ns), and constant fraction discrim-inator to serve both the analog and logic circuit inparallel. The energy response was digitized in apeak-sensing ADC (CAEN V785N). Time informa-tion of the individual modules was deduced relativeto the central module. The coincidence timing ofthe central crystal with one of the timing signals ofthe chosen tagger channels selected the event type.The energy and time information of each moduletogether with the timing response of the relevanttagger channels were recorded event-by-event foran off-line analysis. The response function of thecrystal matrix was measured for the first time atan intermediate temperature of T=0C.

The distribution of the electromagnetic showerwithin the 3×3 crystal matrix has been obtainedafter a carefully prepared relative calibration of theindividual modules. The readout concept of a com-mon integration gate did not impose any hardwareenergy threshold. The absolute calibration was de-duced from the simulation of the total energy de-position in the array based on a GEANT4 simula-tion, taking into account all instrumental details.Fig. 10.15 illustrates the measured energy distribu-tion within the crystal matrix at an incident photonenergy of 674.5 MeV. Since the cross section of theused crystals is comparable to the Moliere radius,approximately 70 % of the shower energy are con-tained in the central module.

Figure 10.15: Distribution of the electromagneticshower into the 3×3 PWO-II matrix at an incident pho-ton energy of 674.5 MeV readout with single LAAPDsas photo sensors at a temperature of T=0C.

Figure 10.16: Experimental line shape of the 3×3PWO-II matrix measured at the lowest photon energyof 40.9 MeV at a temperature of T=0C. The figureshows the individual energy depositions into the cen-tral module and the surrounding ring. Each crystal wasreadout with a single LAAPD.

The line shape and the resulting energy resolutionof the sub-array are obtained by summing event-wise the responding modules. All contributionsabove the pedestal were taken into account corre-sponding to a minimum energy threshold of approx-imately 1 MeV. Fig. 10.16 shows the excellent res-olution of σ/E=9.5 % for the lowest photon energymeasured in such a configuration. The correspond-ing distributions for the highest investigated inci-dent energy are reproduced in Fig. 10.17.

The response of the matrix over the entire energyrange follows a linear relation, both for the centralmodule as well as the whole detector block. The


Figure 10.17: Experimental line shape of a 3×3 PWO-II matrix measured at the highest photon energy of674.5 MeV at a temperature of T=0C. The figureshows the individual energy depositions into the cen-tral module and the surrounding ring. The crystal wasreadout with a single LAAPD.

Figure 10.18: The linearity of the response of a 3×3matrix over the entire range of photon energies mea-sured at a temperature of T=0C. The most probableenergy deposition into the central module is shown sep-arately.

correlation is shown in Fig. 10.18. The overall per-formance is summarized in Fig. 10.19. From theexperimental data one can extrapolate an energyresolution of 2.5 % at 1 GeV photon energy in spiteof the fact, that the lateral dimensions of the de-tector array are not sufficient at that energy. Thefigure contains for comparison the expected resolu-tion based on the Monte-Carlo simulation using thecode GEANT4 including the concept of digitizationas used in the overall simulation of the PANDA de-tector. It confirms that the simulations discussedin chapter 9 are close to the present experimental

Figure 10.19: Experimental energy resolution as afunction of incident photon energy measured for a 3×3PWO-II matrix readout with a LAAPDs at a tempera-ture of T=0C. The result of a GEANT4 simulation isshown for comparison.

performance.

The results show the excellent performance evenat operating conditions well above the envisagedtemperature of T=-25C and with a limited col-lection efficiency of the LAAPD compared to thefinal coverage with two sensors, which will lead toan improvement in resolution at least in the orderof ≥1.4 and will be significant at lower photon ener-gies, when the resolution is dominated by the pho-ton statistics. However, the final gain factor willstrongly depend on the read-out concept and noiseimposed by the front-end electronics [12].

10.1.2 Position Resolution

The point of impact of the impinging photon canbe reconstructed from the center of gravity of theelectromagnetic shower distributed over the clusterof responding detector modules. Such an analysiscan be performed in a straight forward fashion us-ing data from previously described measurements,independent of the used readout of the scintillationresponse. However, reconstructions and optimiza-tions adapted to the crystal geometry of the targetspectrometer EMC have been performed so far onlybased on Monte-Carlo simulations.

Experimental studies in the past with detector ar-rays of identical geometry but inferior crystal qual-ity have delivered a position resolution of σ<2 mm,slightly depending on the reconstruction algorithmbut typical for crystal cross sections close to theMoliere radius.


10.1.3 Particle Identification

The capabilities of particle identification, such asthe different cluster size of detector modules re-sponding to hadron or photon/electron inducedevents, have not been investigated so far experimen-tally but will be a major task using the PROTO60device within the next months. In previous exper-iments with 2.1 GeV proton induced reactions atCOSY [5, 13] a clean separation of hadrons and pho-tons was possible even in a sub-array of 5×5 units.More specific aspects are studied in simulations dis-cussed in the next section on simulations.

The high quality of the timing information of thecalorimeter elements was documented in several ex-periments and will be further exploited strongly de-pending on the final concept of the front-end elec-tronics and final digitization. Exploratory measure-ments were previously discussed in the context ofthe electronics, see Sec. 6.4. Time resolutions evenbelow 1ns will allow selective and precise correla-tions between different inner detector components,in particular the tracking as well as the particleidentification in the barrel DIRC. In addition, thecalorimeter allows the extension of muon detectionand identification down to low energies. Finally,the expected fast response can become a selectivetool to discard random coincidences and events ini-tiated by background such as due to residual gasesin the beamline near the target region or secondaryinteractions at beamline components up- and down-stream from the central target.

10.1.4 Construction and BasicPerformance of the BarrelPrototype Comprising 60Modules

Several prototypes have been constructed in orderto validate the concepts of performance with re-spect to physics, mechanics, thermics and integra-tion. The last one and the most complete is the real-size PROTO60, composed of 60 crystals of PWO-II.The setup is representative for the center part of aslice, assembled of type 6 crystals in their realisticpositions. The design principle is similar to the finalone except that the crystals, LAAPDs and pream-plifiers are easily accessible for reasons of mainte-nance or controls as needed for a prototype. Thecrystals are resting on a thick support plate madeof aluminum and the inserts into the carbon fiberalveoles are not permanently glued. The insulationbox has a removable back cover.

Fig. 10.20 shows a group of 2 pairs of left and

Figure 10.20: Pack of 4 crystals wrapped with reflec-tor and coupled to a LAAPD as part of the PROTO60.

Figure 10.21: Back view of PROTO60 before in-stalling the fiber system for monitoring.

Figure 10.22: Back view of the fully assembledPROTO60 including the optical fibers.

right crystals, wrapped with the ESR reflector. The


Figure 10.23: Temperature stability of the PROTO60.The temperature variation of the inlet coolant, the crys-tals and the ambient air stay within ±0.01C, ±0.05C,and ±0.6C, respectively.

Figure 10.24: Measured response of a single module ofthe PROTO60 to the energy deposition of penetratingminimum ionizing cosmic muons.

large LAAPDs (hidden behind their black capsules)are coupled with optical grease for easy removaland connected to the preamplifiers by twisted andshielded wires. They are arranged in groups of 4precisely positioned with a thin mylar tape beforebeing inserted into the alveoles. The Fig. 10.21shows a back view with all alveoles being alreadyfilled, with some inserts and one printed circuitboard in place for signal readout. This multilayerboard connects the preamplifiers and transfers thesignals to the data acquisition across the thermalshield. The next Fig. 10.22 shows the back view be-fore mounting the cover plate. Besides the printedboards one can notice the distribution of the op-tical fibers, which inject light pulses from a LED-pulser individually into each crystal. The fibers arejust contacting the rear surface without grease forbetter optical coupling. The light is provided by apulsed LED with a variable frequency up to severalkHz and distributed and homogenized via an op-tical system to serve the 60 optical fibers made offused silica.

The insulation is achieved by thermal screens com-posed of copper plates with serpentines, confined inan insulated plastic box filled with dry nitrogen toavoid ice formation. In front of the calorimeter, aprototype of the vacuum panel with coolant carbonchannels is successfully running with a temperaturevariation of up to 3C between both crystal endfaces. It means that the front side is well abovethe dew point and avoids any risk of moisture closeto the DIRC. The set of crystals is operating at atemperature of T=-25C stabilized to ±0.1C witha Julabo chiller recirculating Syltherm coolant at aflow rate of 3.5 liters/minute. A set of 13 thermo-couples measures the temperatures at different lo-cations and is recorded with an Agilent data acqui-sition. The Fig. 10.23 shows one cooling cycle overa period of 30 hours. The temperature variationof the inlet coolant, selected crystals and the am-bient air inside the prototype stay within ±0.01C,±0.05C, and ±0.6C, respectively, and verify therequired thermal stability.

In order to prepare the in-beam tests, the function-ality of the calorimeter prototype has been inves-tigated using the response to cosmic muons. Fromthe known high-voltage settings of the LAAPDs fora gain of 50, given by the manufacturer for a tem-perature at T=+20C, the corresponding lower biasvoltages at the final working temperature of T=-25C have been adjusted exploiting the light pulseras a reference. The fine-tuning of the gain includingthe amplification settings in the spectroscopy am-plifiers have been performed using the energy depo-sition of cosmic muons into the crystals. The indi-vidual spectra can be used as a relative calibrationof the calorimeter elements for later reconstructionof an electromagnetic shower. The linearity of theassembled module including the electronics chain ischecked with the LED light, which can be dimin-ished in intensity by a set of gray-filters. Presentlythe PROTO60 is ready for beam tests in the sum-mer of 2008 including a DAQ system based on com-mercial VME-modules for event-wise recording ofthe responding detectors. Fig. 10.24 shows for asingle module a typical spectrum initiated by pen-etrating muons. The prominent peak correspondsto the most probable path-length of minimum ion-izing muons, equivalent to an energy deposition of24 MeV. The prototype has been running for weeksunder stable conditions.


10.2 Expected performance ofthe EMC

Summarizing, the experimental data as well as thedetailed design concepts and simulations, respec-tively, show that the ambitious physics programcan be fully explored with respect to the measure-ment of the electromagnetic probes, such as pho-tons, electrons/positrons or the reconstruction ofthe invariant mass of neutral mesons.

The calorimeter design in the target region is in fullaccordance with the constraints imposed by a fixedtarget experiment with the strong focusing of themomenta in forward direction. It is composed of abarrel and two endcaps comprising in total 15,552tapered crystal modules of 11 + 2 different shapes.A nearly full coverage of ∼ 99 % solid angle in thecenter-of-mass system is guaranteed in combina-tion with the forward electromagnetic calorimeter,which is not part of the present TDR but is basedon well known technology of a sampling calorime-ter and is located downstream beyond the dipolemagnet. The granularity is adapted to the tolera-ble maximum count rate of the individual modulesand the optimum shower distribution for energy andposition reconstruction by minimizing energy lossesdue to dead material. The front faces of the crys-tal elements cover a nearly identical solid angle andhave lateral dimensions close to the Moliere radius.All crystals are pointing off the target center.

PWO has been chosen as the most appropriate scin-tillator material, which allows a compact designwithin the solenoidal magnet, a fast response andhigh-rate capability due to the short decay time ofthe dominating scintillation process concentratedon the blue emission. All crystals have a com-mon length of 200 mm corresponding to 22 radi-ation lengths, which allows optimum shower con-tainment up to 15 GeV photon energy and limitsthe nuclear counter effect in the subsequent photosensor to a tolerable level. The luminescence yield,which was sufficient for the high-energy regime atLHC, has been significantly improved by nearly afactor of two due to the optimization of the pro-duction technology, such as raw material selection,pre-fabrication processing and optimized co-dopingwith a minimum concentration of La- and Y-ions.To cope with the need for low energy photon detec-tion in the tens of MeV range, the calorimeter willbe operated at T=-25C in order to gain anotherfactor of 4 due to the lower thermal quenching ofthe scintillation light.

The radiation level will be well below typical LHC

values by at least two orders of magnitude at themost forward direction and even further below atlarger angles. However, the interplay between ra-diation damage and recovery processes is almostnegligible at the envisaged temperatures [14, 15].Therefore, the change of the optical transparencyof the crystals is degrading asymptotically to avalue, which is given by the absolute concentra-tion of defect centers in the crystal. As a conse-quence, the present quality of PWO, characterizedas PWO-II, provides an induced absorption coeffi-cient, which is almost a factor two below the qual-ity limits of CMS/ECAL. The light loss, asymp-totically reached in a typical running period of 6-8months, stays below 30 %, which reduces the opti-mum gain factor to a value of 3 compared to anoperation at T=+25C. It should be noted, thatthese effects have to be considered only in the veryforward region of the forward endcap EMC. Whenthe calorimeter is brought up to room temperatureafter a typical experimental period of 6-8 months,there will be a full recovery of the optical transmis-sion within a few weeks. In Y-, La-doped crystalsshorter the dominant relaxation times are on thelevel 20 to 25 hours. There are only very small con-tributions due to tunneling processes, with recoveryconstants near 75 hours.

The operation at low temperatures imposes a tech-nological challenge on the mechanical design, thecooling concept and thermal insulation under theconstraints of a minimum material budget of deadmaterial in particular in front and in between sub-structures. Detailed simulations and prototypinghave confirmed the concept and high accuracy hasbeen achieved in temperature stabilization takinginto account also realistic scenarios of the powerconsumption of the front-end electronics.

The direct readout of the scintillation crystals hasbeen significantly improved by the developmentof a large size avalanche photodiode, similar tothe intrinsic structure of the CMS/ECAL version,in collaboration with Hamamatsu. In total, twoLAAPDs of 100 mm2 active surface each will beused to convert the scintillation light with high ef-ficiency, since almost 30 % of the crystal endfaceare covered. Irradiation studies with electromag-netic and hadronic probes have confirmed the radi-ation hardness. The operation at low temperaturesis even in favor and leads to a significantly slower in-crease of the dark current due to radiation damage.In order to cope with a substantially higher countrate of the individual crystal modules in the forwardendcap EMC, vacuum photo triodes are foreseen asphoto sensors for the forward endcap. Besides the


design used at CMS, the development and testing ofprototypes continues in parallel at both the manu-facturers Hamamatsu and Photonis, which focus inaddition on the implementation of photocathodeswith higher quantum efficiency compared to stan-dard bialkali versions.

The concept of the readout electronics has beenelaborated. As the most sensitive element, a pro-totype of a custom designed ASIC implementingpreamplification and shaping stages has been suc-cessfully brought into operation and will provide alarge dynamic range of 12,000 with a typical noiselevel corresponding to ∼ 1 MeV. Again, the opera-tion at low temperatures reduces further any noisecontributions. Presently most of the experimen-tal tests were based on low-noise preamplifiers as-sembled of discrete components. A similar versionmight be used for the VPTs after some modifica-tions.

The expected overall performance of the calorime-ter is presently based on a series of test experimentswith detector arrays comprising up to 60 full sizecrystals. The basic quality parameters of PWO-IIare confirmed in many tests including a first pre-production run of 710 crystals which are sufficientin shape and quantity to complete one out of the 16slices of the barrel part. The results establish thereadiness for mass production, at least at the man-ufacturer plant BTCP in Russia. Parallel develop-ments of full size samples from SICCAS, China, areexpected soon for comparison.

The experimental tests, mostly performed at thetagged photon facility at MAMI at Mainz andMAXLab at LUND have delivered values for energyand time resolutions, which should be considered assafety limits. They have been performed at varioustemperatures and were read out via photomultipliertubes or just with a single LAAPD [11]. Therefore,only a significantly lower percentage of the scintil-lation light was converted into an analogue signal.From these results one can extrapolate and confirmat an operating temperature of T=-25C an energyresolution, which will be well below 2.5 % at 1 GeVphoton energy. At the lowest investigated showerenergy of 20 MeV a resolution of 13 % has evenbeen obtained, which reflects the excellent statisti-cal term due to the overall improved and enhancedlight yield. Relevant for the efficient detection andreconstruction of multi-photon events, an effectiveenergy threshold of 10 MeV can be considered forthe whole calorimeter as a starting value for clusteridentification along with a single crystal thresholdof 3 MeV for adjacent channels which will enable usto disentangle physics channels with extremely low

cross section.

The fast timing performance will allow a time reso-lution close to 1 ns even based on the readout withLAAPDs and provide a very efficient tool to corre-late the calorimeter with other devices for trackingand particle identification or an efficient reductionof background caused by residual gas in the beampipe or interactions at accelerator components act-ing as secondary targets. In a series of tests thedetection of hadronic probes such as low energy pro-tons or minimum ionizing particles has been provenand will provide complementary information to theparticle tracking or even identification of muonsbased on the energy deposition or different clustermultiplicity of trespassing hadrons.

The overall performance of the calorimeter will becarefully controlled by injecting light from LED-sources distributed via optical fibers via the rearface of the crystal. The change of optical trans-mission will be monitored near the emission wave-length at 420 nm, at a dominant defect center due toMolybdenum at 530 nm, and far in the red spectralregion to control independently the readout chainincluding the photo sensor. Radiation damage isnot expected to appear in the red. A first prototypeis already implemented into the PROTO60 arrayand operating.

Based on the ongoing developments of PWO-IIcrystals and LAAPDs, a detailed program has beenelaborated for quality assurance of the crystals andscreening of the photo sensors. The crystal studiesinclude the preparation of various facilities for irra-diation studies at Protvino and Giessen. In addi-tion, the quality control is planned to be performedat CERN, taking advantage of the existing infras-tructure and experience developed for CMS. One ofthe two ACCOS machines, semi-automatic robots,is presently getting modified for the different spec-ification limits and geometrical dimensions of thePANDA crystals.

Similar infrastructure has been developed for thefinal certification of the LAAPDs in rectangularshape. The intrinsic layout is identical to thequadratic version, which has been tested in greatdetail. The new geometry allows to fit two sensorson each crystal irrespective of its individual shape.

The general layout of the mechanical structure iscompleted including first estimates of the integra-tion into the PANDA detector. The concepts forsignal- and HV-cables, cooling, slow control, moni-toring as well as the stepwise assembly are workedout to guarantee that the crystal geometries couldbe finalized. Prototypes of the individual crys-


tal containers, based on carbon fiber alveoles, havebeen fabricated and tested and are already imple-mented in the PROTO60 device.

References

[1] I. Anthony et al., Nucl. Instrum. Meth.A301, 230 (1991).


[3] K. Mengel et al., IEEE Trans. on Nucl. Sci.45, 681 (1998).


[5] M. Hoek et al., Nucl. Instrum. Meth. A486,136 (2002).

[6] A. Fedorov et al., Proceedings of SCINT05,Alushta, Ukraine, 2005 , 389 (2006).

[7] K. Makonyi et al., Proceedings of SCINT05,Alushta, Ukraine, 2005 , 338 (2006).


[9] M. Thiel et al., Proceedings of SCINT05,Alushta, Ukraine, 2005 , 342 (2006).

[10] B. Lewandowski et al., Proceedings ofSCINT05, Alushta, Ukraine, 2005 , 391(2006).

[11] R. W. Novotny et al., IEEE Trans. on Nucl.Sci. 55, 1207 (2008).

[12] R. Novotny, IEEE NSS2007, ConferenceRecord , 161 (2007).

[13] M. Hoek et al., IEEE Trans. on Nucl. Sci. 49,946 (2002).

[14] R. W. Novotny et al., IEEE Trans. on Nucl.Sci. 55, 401 (2008).

[15] P. A. Semenov et al., Nucl. Instrum. Meth.A582, 575 (2007).


171

11 Organisation

11.1 Quality Control andAssembly

11.1.1 Production Logistics

The realization of the two major components of thetarget spectrometer EMC, the barrel part and thetwo endcaps of different size, is split into severalparts, which also require significantly different lo-gistics. The mechanical parts, such as the basicdetector modules including their housing as well asthe support structure, the thermal insulation andcooling as well as the container for the readout elec-tronics, are presently under development and firstdesign studies have been realized in prototype de-velopment. The detector components comprise ba-sically the nearly 16,000 PWO-II crystals and thetwo types of photo sensors, LAAPDs and vacuumtriodes. Both require a sophisticated program ofproduction by manufacturers, the quality controland assurance and screening and pre-radiation inparticular of the avalanche photo diodes. The con-cept and the procedures of the latter componentsare outlined in great detail in the chapter aboutphoto sensors.

The major steps in realization are defined on onehand by the optimum logistics but also by the ca-pabilities and interests of the participating institu-tions as well as the supporting funding profile. Thepresent layout assumes the sufficient funding and atiming profile to be available when necessary. Thelogistics is based on a practicable concept to achievean optimum realization and to guarantee the envis-aged performance of the calorimeter and is schemat-ically illustrated in Fig. 11.1.

11.1.2 The PWO-II Crystals

Presently, the production of high-quality PWOcrystals delivering the specifications of PWO-II canbe expected to be performed only by the two man-ufacturers BTCP and SICCAS, respectively, whichhave been successfully involved in the realizationof the CMS EM calorimeter. There is foreseen tohave prepared at both companies an infrastructureto measure the specifications primarily to have afast response to instabilities of the production lineand to deliver crystals according to the very selec-tive specification parameters. There will be a major

Figure 11.1: Schematic layout for the realization ofthe target spectrometer EMC of PANDA.

center for the crystal inspection and assembly of thedetector modules. As an option, it could be locatedat Giessen, which is in addition advantageous dueto the very close distance to FAIR. Already con-firmed by an official offer of the CMS collaboration,the quality tests will be performed at CERN us-ing the semi-automatic ACCOS machine. It guar-antees a well accepted and adapted procedure andallows to exploit more than 10 years of experience.Preparations have been initiated already, to adaptthe installations to the geometry and more selectivespecifications of the PANDA crystals. Afterwards,the crystals will be shipped back to control for eachcrystal its radiation hardness and the wavelengthdependence, which is very crucial for the operationat low temperatures and has to be performed for avery high percentage of the crystals. These mea-surements could be handled at the irradiation facil-ity at Giessen, as characterized in chapter 4.

11.1.3 Detector Module Assembly

In a second line, to be performed at an appropriateregional center, the crystals will be equipped witha reflector foil, photo and temperature sensors, andthe front-end electronics. All latter components aredelivered tested and screened from the responsiblelaboratories at GSI (LAAPD, front-end electronics)and Bochum/KVI (VPT), most probably.


11.1.4 Final Assembly,Pre-calibration andImplementation into PANDA

The final assembly of the barrel slices or the twoendcaps will be completed by the implementationof the optical fibers for the monitoring system,the electronics for digitization, the cables for sig-nal transfer and low- and high voltage support, thesensors and lines for slow control as well as the in-stallations for cooling. This task including the in-termediate storage have to be performed at or atleast close to FAIR in a dedicated laboratory space.The basic functionality of the assembled units willbe tested using on one hand the monitoring systemto check the electronics chain including the photosensors. A first hint of the expected sensitivity ofeach detector module is given by the measured lightyield of the crystal, the absolute value of opticaltransmission at the absorption edge and the knowngain of the photo sensor. These parameters will al-low a first step towards a pre-calibration with anaccuracy on the order of 5 %. On the other hand,a more sensitive relative and absolute calibrationof the calorimeter modules will be obtained by us-ing minimum ionizing cosmic muons. The submod-ules have to be mounted on a mechanical supportstructure, which allows to rotate most of the detec-tor modules into vertical positions. Requiring anti-coincidence with all neighboring modules selects thefull path length of 200 mm of the particle, whichcorresponds to an energy deposition of ∼240 MeV.This procedure will deliver calibration parameterswith an intended accuracy of 2.5 %, values whichbase on the experience of CMS. In order to obtaina more accurate calibration, a cross check with thecalibration steps explained above and to determinefinally the response function a substantial part ofthe units will be exposed to a direct beam of en-ergy marked photons, which can be provided at thetwo tagging facilities at MAMI (Mainz) and ELSA(Bonn), respectively, covering the complete energyrange up to ∼3.2 GeV. At both places experimentalset-ups are under consideration.

11.1.5 Other CalorimeterComponents

Besides the basic calorimeter components, crystals,photo sensors and pre-amplifier/shaper, all otherparts including mechanical structure, cooling, ther-mal insulation, electronics and DAQ are under de-velopment. In case of the most critical tasks proto-types have been designed and realized and confirm

the proposed concept. The responsibility for thefinal design and realization are distributed amongthose groups within the collaboration, which havethe infrastructure and experience with large scaleexperimental installations.

11.1.6 Integration in PANDA

The design of the target spectrometer EMC is infull agreement to the overall layout of the wholePANDA detector. The components including ser-vice connections and mounting spaces fit withinthe fiducical volume defined by the magnet and theother detector components. It is foreseen, that thefully assembled and tested submodules are stored inan air-conditioned area with stabilized temperatureand low humidity. The sealed components shouldbe kept in a nitrogen atmosphere. The necessarysteps for mounting and integrating the calorimeterinto PANDA have been described already in greatdetail in the chapter on mechanics for both majorparts and will be the final task performed by thewhole collaboration.

11.2 Safety

The design details and construction of the calorime-ter including the infrastructure for operation will bedone according to the safety requirements of FAIRand the European and German safety regulations.Design aspects for which the CMS calorimeters aretaken as examples take into consideration the de-tailed guidelines given by CERN, which are mostappropriate to the case of large scientific installa-tions.

11.2.1 Mechanics

The strength of the EMC support structures hasbeen computed with physical models in the courseof the design process. Details of finite element anal-ysis are shown in Sec. 7.1.2. Each mechanical com-ponent will undergo a quality acceptance examina-tion including stress and loading tests for weightbearing parts. Spare samples may also be testedup to the breaking point. A detailed material mapof the entire apparatus showing location and abun-dance of all materials used in the construction willbe created. For structural components radiationresistance levels will be taken into account in theselection process and quoted in the material map.For the calorimeter crystals themselves precautionswill be taken to rule out exposure of persons to toxic


material due to exhaustion in case of fire or due tomechanical operations. In normal conditions leadtungstate is stable, non-hygroscopic, and hardly sol-uble in water, oils, acids and basis.

11.2.2 Electrical Equipment andCooling

All electrical equipment in PANDA will complyto the legally required safety code and concur tostandards for large scientific installations followingguidelines worked out at CERN to ensure the pro-tection of all personnel working at or close to thecomponents of the PANDA system. Power sup-plies will have safe mountings independent of largemechanical loads. Hazardous voltage supplies andlines will be marked visibly and protected fromdamage by near-by forces, like pulling or squeez-ing. All supplies will be protected against overcur-rent and overvoltage and have appropriate safetycircuits and fuses against shorts. All cabling andoptical fibre connections will be executed with non-flammable halogen-free materials according to up-to-date standards and will be dimensioned withproper safety margins to prevent overheating. Asafe ground scheme will be employed throughoutall electrical installations of the experiment. Smokedetectors will be mounted in all appropriate loca-tions.

Lasers or high output LEDs will be employed inthe monitoring system and their light is distributedthroughout the calorimeter systems. For these de-vices all necessary precautions like safe housings,color coded protection pipes, interlocks, properwarnings and instructions as well as training of thepersonnel with access to these components will betaken.

The operation of the PWO EMC at temperaturedown to -25C requires a powerful and stable cool-ing system for the crystals and sensors. It consistsof a system of cooling pipes, pumps and heat ex-changers and employs a liquid silicone polymer ascoolant. This material is highly inert and non-toxic.Its flow will be in the order of 4 l/s for the fullEMC. A control system is employed to detect anymalfunctioning of the cooling system and includeinterlocks for electronics and high-voltage supplies.In the case of coolant loss or in case of abnormaltemperature gradients between coolant input andoutput the control system enacts the appropriatesafety procedures. A highly redundant system oftemperature sensors allows the continuous monitor-ing of the effectiveness of the system.

11.2.3 Radiation Aspects

The PWO crystals can be activated by low energyprotons and neutrons leading to low-energy radioac-tivity of the activated nuclei. However, simulationsbased on the computer code MARS deliver in caseof the maximum luminosity of 2×1032/s and for anoperation period of 30 days a dose rate of 20µSv/hat the crystal surface of the forward endcap EMCclosest to the beam axis, where the rate is expectedto be the highest in the entire EMC by several or-ders of magnitude. The estimated rate is one or-der of magnitude lower than in case of CMS atLHC. Shielding, operation and maintenance will beplanned according to European and German safetyregulations to ensure the proper protection of allpersonnel.

11.3 Schedule

Fig. 11.2 shows in some detail the estimated time-lines including the main tasks for the design phase,prototyping, production, pre-calibration and imple-mentation of the target spectrometer EMC into thePANDA detector. The schedule is based on the ex-perience gathered during the R&D and prototypingphase, the close collaboration with optional manu-facturers and the fruitful exchange of experience inparticular the CMS/ECAL project at CERN. Thetimelines assume sufficient funding, which shouldbecome available when necessary.

After an initial phase of research and developmentof the target spectrometer EMC concentrating onthe basic performance of the essential components,i.e. the optimal scintillator material and the photosensors of choice, the collaboration has completedthe design phase. The integration of the mechani-cal components into the PANDA detector has beenclarified, prototype crystals, sensors, discrete andASIC preamplifiers, and digitization modules are athand, and a prototype detector has been set up andis ready for beam experiments. A number of insti-tutions, who have gained specific expertise in pastand ongoing research programme at large-scale ex-periments such as BaBar, TAPS/Crystal Ball/CrystalBarrel at several accelerator facilities, are engagedin the remaining tasks leading to the completion ofthe target spectrometer EMC. The responsibilitiesfor the various work packages and tasks are listedin Fig. 11.3 and show the participating institutions.The coordination of the work packages is performedby those groups marked in bold-face. However,the responsibility is not automatically linked to theavailability of future funding.


ID Task Name Duration

1 Barrel mechanics 1312 days

2 design 32 mons

3 C-fiber alveole production 34 mons

4 support structure 33 mons

5 Barrel detector modules 1256 days

6 crystal pre-production 8 mons

7 crystal production & quality control 31 mons

8 LAAPD pre-production 8 mons

9 LAAPD production & screening 27 mons

10 ASIC final iteration 9 mons

11 ASIC production & screening 26 mons

12 HV supply 13 mons

13 detector module assembly 24 mons

14 subunit & slice assembly 28 mons

15 Fw Endcap mechanics 1032 days

16 design 24 mons



19 Bw Endcap mechanics 858 days

20 design 24 mons



23 Fw & Bw Endcap detector modules 894 days

24 crystal production & quality control 12 mons

25 VPT pre-production 9 mons

26 VPT production 21 mons

27 LNP production 26 mons

28 HV supply 10 mons

29 subunit & endcaps assembly 24 mons

30 common tasks 2075 days

31 digitizer modules 13 mons

32 multiplexer modules 25 mons

33 cooling design 30 mons

34 cooling implementation 50 mons

35 pre-calibration 33 mons

36 monitoring, DCS 40 mons

37 final assembly & integration in PANDA 32 mons

38 ready for commissioning 0 days 1/1

H2 H1 H2 H1 H2 H1 H2 H1 H2 H1 H2 H1 H2 H1 H2 H1 H2 H1 H2 H16 2007 2008 2009 2010 2011 2012 2013 2014 2015 20

Figure 11.2: Timeline for the realization of the target spectrometer EMC.


work package task participating (coordinating) institute

design Giessen, Orsay, Protvino

C-fiber alveole production Orsay, Protvino

Barrel mechanics

support structure Orsay, Protvino

crystal pre-production Giessen, Protvino

crystal production & quality

control

Giessen, GSI, Minsk, Protvino

LAAPD pre-production GSI

LAAPD production & screening GSI

ASIC final iteration & production GSI, KVI

ASIC screening GSI, KVI

HV supply Basel, Giessen

detector module assembly Giessen, GSI, Orsay, Protvino,

Valencia, Warsaw

Barrel detector modules

subunit & slice assembly Giessen, Orsay, Protvino

design Bochum, KVI, Orsay

C-fiber alveole production Bochum, KVI, Lund, Stockholm,

Uppsala

Fw endcap mechanics

support structure KVI

design GSI, KVI, Mainz

C-fiber alveole production KVI

Bw endcap mechanics

support structure GSI, KVI, Mainz

crystal production & quality

control

Bochum, Giessen, GSI, Lund, Minsk,

Stockholm, Uppsala

VPT pre-production Bochum, KVI, Stockholm, Uppsala

VPT production Bochum, KVI , Lund, Stockholm,

Uppsala

LNP production Basel, Giessen

HV supply Basel, Uppsala

Endcap detector modules

subunit & endcaps assembly Bochum, Giessen, GSI, KVI, Mainz,

Stockholm, Uppsala

digitizer modules GSI, KVI, Munich, Uppsala

multiplexer modules KVI, Munich

cooling design & implementation Bochum, KVI, Orsay, Protvino

pre-calibration Basel, Bochum, Giessen, GSI, KVI,

Lund, Mainz, Protvino, Stockholm,

Uppsala, Valencia, Warsaw

monitoring, DCS Bochum, KVI, Protvino, Warsaw

Common tasks

Final assembly & integration in

PANDA

Basel, Bochum, Giessen, GSI, KVI,

Lund, Mainz, Minsk, Orsay, Protvino,

Stockholm, Uppsala, Valencia,

Warsaw

Figure 11.3: Tasks and responsibilities for the realization of the target spectrometer EMC.

Since the table documents as well the capabilitiesof the collaboration members, the different insti-tutions are shortly described and characterized by

their expertise gathered in previous or ongoing re-search activities.

• Basel (B. Krusche, M. Steinacher): The group


is conducting experiments at tagged photon fa-cilities and has expertise in analogue electron-ics.

• Bochum(F.-H. Heinsius, U. Wiedner): TheBochum group is engaged in the design andoperation of calorimeters for BaBar (SLAC),Crystal Barrel (CERN and Bonn/ELSA) andthe WASA detector (Uppsala and Julich).

• Giessen (V. Metag, R. Novotny): The Giessengroup has a longstanding experience in crystalcalorimetry. It has been the leading institutionin building the TAPS detector including elec-tronics. The group has played a leading role inthe development of the PWO-II.

• GSI (B. Lewandowski, F. Mass, K. Peters, A.Wilms): The GSI group has long experiencein crystal calorimetry (BaBar (SLAC), CrystalBarrel (CERN), Mami A4 (Mainz)) and ASICdesign. Particular fields of expertise includedesign and construction of sensors and elec-tronics as well as calibration, monitoring andfront-end software. The APD-LAB branch isa renowned R&D center for LAAPDs channel-ing the developments for PANDA and R3B. Inaddition technical and infrastructure coordina-tion for PANDA is performed by GSI.

• KVI (H. Lohner, J. Messchendorp): TheGroningen group participated in the construc-tion and exploitation of the TAPS calorime-ter. They have major experience in applica-tions of scintillation detectors, photo sensors,light pulser systems for monitoring crystals, de-veloping FPGA electronics and DSP networks.

• Lund (B. Schroder): The Lund group per-forms research at the MAX-Laboratory withmonoenergetic photons in the 15 to 200 MeVrange with an energy resolution of 250 keV atbest.

• Mainz (F. Maas): The group has a long stand-ing experience in fast EM calorimetry, triggersystems and operation of a facility for energymarked photons up to 1.5 GeV energy.

• Minsk (M. Korzhik, O. Missevitch): Thegroup has long year expertise in solid statephysics of scintillation materials and has beenthe coordinator for PWO production for CMSat CERN.

• Munchen (I. Konorov, S. Paul): The grouphas expertise in design and development ofdigital electronics for large scale experiments

(COMPASS) and has developed ADC systemsfor physics and medical applications.

• Orsay (T. Hennino, P. Rosier): Long yearexperience in mechanical and electronics engi-neering and construction of large scale detectorsystems.

• Protvino (V. A. Kachanov, A. N. Vasiliev):The group has been strongly involved in thedesign and construction of the endcap ofCMS/ECAL and R&D for the proposed BTeVexperiment.

• Stockholm (P. E. Tegner): The StockholmUniversity group has a major experience in op-erating detectors in high magnetic fields at theformer CELSIUS machine in Uppsala.

• Uppsala (P. Marciniewski, T. Johansson):The Uppsala group has been involved in exper-iments at CERN, Uppsala and Julich. Theirexperience covers many aspects of calorime-ters (WASA), electronic signal digitization andreadout systems.

• Valencia (J. Diaz): The group is involved inscintillator systems of large scale detectors suchas TAPS and HADES.

• Warsaw (B. Zwieglinski): The SINS Warsawgroup has experience in response studies ofscintillators exploiting proton accelerators.


Acknowledgments

We acknowledge financial support fromthe Bundesministerium fur Bildung undForschung (bmbf), the Deutsche Forschungs-gemeinschaft (DFG), the University ofGroningen, Netherlands, the Gesellschaftfur Schwerionenforschung mbH (GSI),Darmstadt, the Helmholtz-GemeinschaftDeutscher Forschungszentren (HGF), theSchweizerischer Nationalfonds zur Forderungder wissenschaftlichen Forschung (SNF), theRussian funding agency “State Corpora-tion for Atomic Energy Rosatom”, theCNRS/IN2P3 and the Universite Paris-sud,the British funding agency “Science andTechnology Facilities Council” (STFC),the Instituto Nazionale di Fisica Nucleare(INFN), the Swedish Research Council, thePolish Ministry of Science and Higher Edu-cation, the European Community FP6 FAIRDesign Study: DIRACsecondary-Beams,contract number 515873, the EuropeanCommunity FP6 Integrated InfrastructureInitiative: HadronPhysics, contract numberRII3-CT-2004-506078, the INTAS, and theDeutscher Akademischer Austauschdienst(DAAD).We also would like to thank the CMS Col-laboration at CERN for their support.



ADC Analog to Digital Converter

ALICE A Large Ion Collider Experiment atCERN LHC

ALICE/PHOS ALICE Photon Spectrometer

APD Avalanche Photo Diode

APFEL ASIC for Panda Front End ELectronics

ASIC Application Specific Integrated Circuit

AWG Arbitrary waveform generator

BGO Bismuth Germanate

BTCP Bogoroditsk Techno-Chemical Plant

C Capacitance

CAD Computer Aided Design

CERN Conseil European pour la RechercheNucleaire

CLAS CEBAF Large Acceptance Spectrometer

CMOS Complementary Metal OxideSemiconductor

CMS Compact Muon Solenoid

CMS/ECAL CMS electromagnetic calorimeter

COMPASS Common Muon Proton Apparatus forStructure and Spectroscopy

COSY Cooler Synchrotron

DAQ Data Acquisition

DCS Detector Control System

DIRC Detector for Internally Reflected CherenkovLight

DPM Dual Parton Model

DSP Digital Signal Processor

DVCS Deeply Virtual Compton Scattering

EMC Electromagnetic Calorimeter

ENC Equivalent Noise Charge

EPR Electron Paramagnetic Resonance

FADC Flash ADC

FAIR Facility for Antiproton and Ion Research

FINUDA Fisica Nucleare a DAFNE

FPGA Field Programmable Gate Array

GEM Gas Electron Multiplier

GLAST Gamma ray Large Area Space Telescope

GSI Gesellschaft fur Schwerionenforschnung

HEP High Energy Physics

HESR High Energy Storage Ring

HV High Voltage

IHEP Institute for High Energy Physics

J-FET Junction Field-Effect Transistor

KVI Kernfysisch Versneller Instituut

LAAPD Large Area APD

LEAR Low Energy Antiproton Ring

LED Light Emitting Diode

LHC Large Hadron Collider

LNP Low Noise Preamplifier

LOI Letter of Intent

LY Light Yield

M Gain of APD

MAMI Mainz Microtron

MIP Minimum Ionizing Particle

MLP Multi Layer Perceptron

MVD Micro Vertex Detector

NCE Nuclear counter effect

NIEL Non-Ionizing Energy Loss

PANDA Pbar ANihilation at Darmstadt

PCB Printed Circuit Board

PEEK Polyetheretherketone

phe photo electrons

PID Particle Identification

PMT Photomultiplier

PSI Paul Scherrer Institute

PWO Lead Tungstate

QCD Quantum Chromo Dynamics

QE Quantum efficiency

RICH Ring Imaging Cherenkov Counter


RT Room Temperature

SCADA Supervisory Control And DataAcquisition

SICCAS Shanghai Institute of Ceramics, ChineseAcademy of Sciences

SIS Heavy ion synchrotron

SMD Surface Mount Device

STT Straw Tube Tracker

TDC Time to Digital Converter

TSL The Svedberg Laboratory

TSL Thermo stimulated luminescence

UrQMD Ultra-relativistic Quantum MolecularDynamic

VME Versa Module Eurocard

VPT Vacuum Photo Triode

181

List of Figures

1.1 Overview of the PANDA spectrometer. 2

2.1 Schematic view of the HESR. . . . . 16

2.2 Average luminosity vs. cycle time. . 17

2.3 Artistic view of the PANDA Detector. 18

2.4 Side view of the target spectrometer. 19

2.5 Schematic of the target and beampipe setup with pumps. . . . . . . . 20

2.6 The Microvertex detector of PANDA 21

2.7 Straw Tube Tracker in the TargetSpectrometer . . . . . . . . . . . . . 23

2.8 GEM Time Projection Chamber inthe Target Spectrometer . . . . . . . 23

2.9 Top view of the experimental area. . 29

3.1 π0 loss rate as a function of energythreshold. . . . . . . . . . . . . . . . 31

3.2 Photon energy distribution vs. lab.angle. . . . . . . . . . . . . . . . . . 33

3.3 Energy resolutions for different sin-gle crystal reconstruction thresholds. 34

3.4 π0 and η energy spectrum. . . . . . . 35

3.5 Minimum π0 opening angle. . . . . . 36

3.6 Mass resolution for π0 (spatial reso-lution). . . . . . . . . . . . . . . . . 36

3.7 Hit rate in the barrel part. . . . . . . 36

3.8 Hit rate in the forward part. . . . . . 36

4.1 Integrated single crystal rate for thebarrel section. . . . . . . . . . . . . . 41

4.2 Integrated single crystal rate for theforward endcap. . . . . . . . . . . . . 41

4.3 Single crystal energy differential ratespectrum. . . . . . . . . . . . . . . . 41

4.4 Relative energy dose as function ofthe radial depth. . . . . . . . . . . . 41

4.5 Selected PWO-II ingots. . . . . . . . 43

4.6 The optical transmission and radio-luminescence spectra. . . . . . . . . 43

4.7 The scintillation kinetics of PWOmeasured at room temperature. . . . 43

4.8 Integral dependence of the radiolu-minescence. . . . . . . . . . . . . . . 45

4.9 The measured induced absorption ofvarious PWO-II samples. . . . . . . 47

4.10 Schematic layout of the irradiationfacility at IHEP. . . . . . . . . . . . 48

4.11 Photograph of the major compo-nents of the irradiation setup at IHEP. 48

4.12 The experimental setup for irradia-tion studies at Giessen. . . . . . . . . 49

4.13 The irradiation facility operating atBTCP in Russia. . . . . . . . . . . . 49

4.14 Schematic layout of the spectromet-ric setup at BTCP. . . . . . . . . . . 49

4.15 Schematic layout of a single spectro-metric channel of the setup at BTCP. 50

4.16 Light yield of full size polished scin-tillation crystals. . . . . . . . . . . . 52

4.17 Pulse height spectra measured for aPWO-II crystal. . . . . . . . . . . . . 53

4.18 Temperature dependence of the yieldof PWO-II scintillation light. . . . . 53

4.19 Average PWO-II light yield vs. inte-gration gate. . . . . . . . . . . . . . 54

4.20 Correlation of light yield at low andhigh temperature. . . . . . . . . . . 55

4.21 Variation of the light output mea-sured for different crystal geometriesand reflector materials when a colli-mated γ-source is moved relative tothe sensor position along the crystalaxis. . . . . . . . . . . . . . . . . . . 55

4.22 Relative change of the light yield oftwo PWO-II samples. . . . . . . . . 55

4.23 Relative change of the monitoringsignal. . . . . . . . . . . . . . . . . . 56

4.24 The absolute loss in transmission dueto irradiation. . . . . . . . . . . . . . 56

4.25 The induced absorption due to irra-diation. . . . . . . . . . . . . . . . . 56

4.26 The decay kinetics of an absorptionregion after irradiation. . . . . . . . 56

4.27 The decay kinetics of an absorptionregion after irradiation. . . . . . . . 57


4.28 EPR signal measured after irradiation. 57

4.29 Change of the scintillation responsedue to a continuous irradiation. . . . 57

4.30 Maximum change of the scintillationresponse vs. induced absorption. . . 57

4.31 Change of the scintillation responsebefore and after irradiation. . . . . . 58

4.32 Machined and optically polished bar-rel type PWO-II crystals. . . . . . . 58

4.33 Distribution of the light yield of bar-rel type PWO-II crystals. . . . . . . 58

4.34 Radiation induced absorption spectra. 59

4.35 Distribution of the induced absorp-tion after an absorbed dose of 50 Gy. 59

4.36 The ACCOS machine installed at theCMS regional center at CERN. . . . 60

4.37 Optical transmission of crystal sam-ples from BTCP and SICCAS. . . . 63

4.38 Absolute light yield. . . . . . . . . . 63

4.39 Longitudinal optical transmission oftwo 200 mm long crystals. . . . . . . 64

4.40 Inhomogeneity in the transversal op-tical transmission. . . . . . . . . . . 64

5.1 APD reverse structure. . . . . . . . . 67

5.2 Technical drawing of the rectangularAPD type. . . . . . . . . . . . . . . . 69

5.3 Gain and dark current of ’normal C’version LAAPD. . . . . . . . . . . . 70

5.4 Gain and dark current of ’low C’ ver-sion LAAPD. . . . . . . . . . . . . . 70

5.5 Ratio Id/M for different APD types. 71

5.6 Excess noise factor of CMS-APDs. . 71

5.7 Excess noise factor of ’normal C’ ver-sion LAAPD from Hamamatsu. . . . 71

5.8 NCE of the investigated LAAPD ver-sions. . . . . . . . . . . . . . . . . . . 72

5.9 QE of a PANDA-LAAPD. . . . . . . 72

5.10 Absorption coefficient α(λ, T ) of Si. . 73

5.11 Normalized gain of ’normal C’ typeAPD. . . . . . . . . . . . . . . . . . 73

5.12 Normalized gain of ’low C’ type APD. 74

5.13 Gain M of ’normal C’ type after andbefore protonirradiation. . . . . . . . 74

5.14 Dark current rise of ’normal C’diodecaused by proton irradiation. . . . . 74

5.15 Dark current increase of ’low C’diodecaused by proton irradiation. . . . . 75

5.16 M(λ) of ’normal C’ type after protonexposure. . . . . . . . . . . . . . . . 75

5.17 M(λ) of ’low C’ type after proton ex-posure. . . . . . . . . . . . . . . . . . 75

5.18 Gain M of ’normal C’ type after andbefore photon irradiation. . . . . . . 76

5.19 Dark current Id of ’normal C’ typeafter and before photon irradiation. . 76

5.20 M(λ) of ’normal C’ type after γ ex-posure. . . . . . . . . . . . . . . . . . 76

5.21 Apparatus used for neutron irradia-tion. . . . . . . . . . . . . . . . . . . 77

5.22 Gain M of ’normal C’ type after andbefore neutron irradiation. . . . . . . 77

5.23 Dark current Id of ’normal C’ typeafter and before neutron irradiation. 77

5.24 Dark current Id of ’low C’ type afterand before neutron irradiation. . . . 77

5.25 M(λ) of ’normal C’ type after neu-tron exposure. . . . . . . . . . . . . . 77

5.26 M(λ) of ’low C’ type after neutronexposure. . . . . . . . . . . . . . . . 78

5.27 PEEK APD capsule. . . . . . . . . . 79

5.28 Schematic of VPT. . . . . . . . . . . 79

5.29 VPT gain vs. cathode HV for R2148. 80

5.30 VPT gain vs. cathode HV for FEU-188. . . . . . . . . . . . . . . . . . . 80

5.31 Quantum efficiency VPT. . . . . . . 82

5.32 VPT response to magnetic field vs.angle. . . . . . . . . . . . . . . . . . 82

5.33 Light transmission of VPT windowafter gamma irradiation. . . . . . . . 82

5.34 Light transmission of VPT windowas function of gamma irradiation. . . 82

5.35 Relative anode response of VPT ver-sus the neutron fluence. . . . . . . . 83

6.1 The readout chain of the Electro-magnetic Calorimeter. . . . . . . . . 86

6.2 Schematic diagram of the folded cas-code front end. . . . . . . . . . . . . 88

6.3 Principle diagram of the self biasingfeedback network. . . . . . . . . . . . 89

6.4 Block diagram and photograph ofthe preamplifier and shaper ASIC. . 90


6.5 Test PCB with glued and bondedpreamplifier ASIC. . . . . . . . . . . 91

6.6 Measured noise values of the pream-plifier prototype. . . . . . . . . . . . 91

6.7 Amplification characteristics of thepreamplifier at −20C. . . . . . . . . 91

6.8 Power consumption for one chip independence of temperature. . . . . . 91

6.9 Power dissipation as function of con-tinuous event rate. . . . . . . . . . . 93

6.10 Circuit diagram of the LNP-P proto-type for the VPT readout. . . . . . . 96

6.11 The measured noise performance ofthe LNP-P. . . . . . . . . . . . . . . 97

6.12 PSPICE simulation of ENC versusthe detector capacitance. . . . . . . . 97

6.13 Time resolution as function of thecorresponding pulse energy. . . . . . 97

6.14 The functional diagram of the digi-tizer module. . . . . . . . . . . . . . 98

6.15 The layout of the prototype ADCmodule. . . . . . . . . . . . . . . . . 99

6.16 The layout of the Layer Stack Back-PCB. . . . . . . . . . . . . . . . . . . 101

6.17 Scheme of the layer stack of flat cables.101

6.18 The various functions of the DetectorControl System. . . . . . . . . . . . . 102

6.19 The DCS network structure. . . . . . 102

6.20 Microchip wafer. . . . . . . . . . . . 104

6.21 An integrated circuit connected byneedles during a wafertest. . . . . . . 104

6.22 The top- and the bottom side of thesingle-channel LNP-P prototype. . . 105

7.1 The components of the PANDA elec-tromagnetic calorimeter. . . . . . . . 108

7.2 Overall view of the target spectrom-eter. . . . . . . . . . . . . . . . . . . 108

7.3 The shape of the scintillator crystal. 109

7.4 Segmentation of the calorimeteralong the circumference of the barrel. 109

7.5 Geometrical arrangement of thecrystals of the barrel. . . . . . . . . . 110

7.6 View of the barrel of 11360 crystalswith a separated slice. . . . . . . . . 110

7.7 Deformation test of carbon alveoles. 110

7.8 Summary of the expected dead spacebetween calorimeter elements. . . . . 111

7.9 Schematic view of the concept and ofthe major components of a slice. . . 113

7.10 CAD integration of the insert withthe preamplifier and optical fiber. . . 113

7.11 Exploded view of a slice. . . . . . . . 114

7.12 Deformation of the support beam. . 114

7.13 Barrel supported on the magnet. . . 115

7.14 Integration of the services in the sup-port beam. . . . . . . . . . . . . . . 115

7.15 Front thermal screen. . . . . . . . . . 116

7.16 Equivalent chain for the analyticalanalysis of the LAAPD temperature. 117

7.17 Thermal simulation. . . . . . . . . . 117

7.18 Cooling power summary. . . . . . . . 118

7.19 Dimensions of a slice. . . . . . . . . . 119

7.20 Next thermal prototype 480. . . . . 119

7.21 Services out of barrel. . . . . . . . . 120

7.22 Barrel final mounting. . . . . . . . . 121

7.23 The position of the forward endcapEMC with respect to the target. . . 123

7.24 The geometry of a single forwardendcap EMC crystal. . . . . . . . . . 123

7.25 The geometry of a C-fiber alveole for16 crystals. . . . . . . . . . . . . . . 124

7.26 Photograph of three produced C-fiber alveoles for 16 crystals each. . . 124

7.27 The explosion view of a 16-crystalsubunit. . . . . . . . . . . . . . . . . 125

7.28 The attachement of the C-fiber alve-oles in front of the mounting plate. . 125

7.29 One quadrant of the PANDA forwardendcap calorimeter. . . . . . . . . . . 126

7.30 The complete endcap. . . . . . . . . 127

7.31 Impedance of laminate cable. . . . . 127

7.32 Arrangement of digitizer modules forthe forward endcap EMC. . . . . . . 128

7.33 Acceptance of the backward endcapEMC . . . . . . . . . . . . . . . . . . 128

7.34 The position of the backward endcapEMC with respect to the target. . . 130

7.35 The dimensions of a single crystal ofthe backward endcap EMC. . . . . . 130


7.36 The geometry of one quadrant of thebackward endcap EMC. . . . . . . . 130

8.1 Induced absorption of various PWOsamples after irradiation. . . . . . . 133

8.2 Schematic layout of the stabilizedlight pulser system. . . . . . . . . . . 134

8.3 Major components of a first proto-type of the monitoring system. . . . 134

8.4 α signal stability. . . . . . . . . . . . 135

8.5 Linear fit coefficients for pion andelectron irradiation. . . . . . . . . . 136

8.6 Stability of the LED pulser prototype. 137

8.7 Monitoring system prototype. . . . . 138

8.8 Prototype LED driver. . . . . . . . . 138

8.9 Schematic view of the light pulserprototype. . . . . . . . . . . . . . . . 139

8.10 Light distribution uniformity. . . . . 139

9.1 Simulated line shape at the photonenergy of 674.5 MeV. . . . . . . . . . 143

9.2 Measured and simulated energy res-olution vs. incident photon energy. . 143

9.3 Leakage correction function for thebarrel EMC. . . . . . . . . . . . . . . 144

9.4 Energy resolution for three differentcrystal length scenarios. . . . . . . . 144

9.5 Energy resolutions for different sin-gle crystal reconstruction thresholds. 144

9.6 Position resolution in x-direction forphotons below 3 GeV. . . . . . . . . 145

9.7 E/p versus track momentum for elec-trons and pions. . . . . . . . . . . . . 145

9.8 Zernike moment z31 for electrons,muons and hadrons. . . . . . . . . . 145

9.9 MLP output for electrons and otherparticle species. . . . . . . . . . . . . 146

9.10 Electron efficiency and contamina-tion rate using EMC information. . . 146

9.11 Electron efficiency and contamina-tion rate using combined PID infor-mation. . . . . . . . . . . . . . . . . 147

9.12 Amount of material in front of theEMC. . . . . . . . . . . . . . . . . . 147

9.13 γ conversion probability in the DIRCas a function of Θ. . . . . . . . . . . 147

9.14 Reconstructed photon energywith/without DIRC preshowers. . . . 147

9.15 Energy distributions of γs frompp → hc → ηc γ and pp → φφπ0. 149

9.16 The number of reconstructed EMCclusters. . . . . . . . . . . . . . . . . 149

9.17 Invariant mass for J/ψ → e+e− andπ0 → γ γ. . . . . . . . . . . . . . . . 151

9.18 Invariant χc1 → J/ψ γ and η → γ γmass for the process pp → ψg η. . . 153

9.19 Invariant χc1 π0 π0 mass for the re-action pp → ψg η. . . . . . . . . . . 153

9.20 Angular distribution of recon-structed pp → e+e− events. . . . . . 154

9.21 Angular distribution of recon-structed pp → π+π− events. . . . . 154

10.1 The experimental setup at thetagged photon facility at MAMI. . . 159

10.2 The insulated container to house thevarious test arrays. . . . . . . . . . . 159

10.3 A typical 5×5 PWO matrix. . . . . . 159

10.4 Lineshape of the central detectormodule. . . . . . . . . . . . . . . . . 159

10.5 Response function of the 3×3 matrixto photons. . . . . . . . . . . . . . . 160

10.6 Response function of the 3×3 matrixto photons. . . . . . . . . . . . . . . 160

10.7 Comparison of the energy resolution. 160

10.8 Detector operation at low tempera-tures at MAX-Lab. . . . . . . . . . . 161

10.9 Energy deposition of a 35 MeV inci-dent photon. . . . . . . . . . . . . . 161

10.10Response of a 3×3 PWO-II matrixto incident photons. . . . . . . . . . 161

10.11Experimental energy resolution of a3×3 PWO-II matrix. . . . . . . . . . 161

10.12Response function of the 3×3 PWO-II matrix to photons. . . . . . . . . . 162

10.13Energy resolution of a 3×3 PWO-IImatrix. . . . . . . . . . . . . . . . . . 162

10.14Time resolution of a 3×3 PWO-IImatrix. . . . . . . . . . . . . . . . . . 162

10.15Distribution of the electromagneticshower into the 3×3 PWO-II matrix. 163

10.16Experimental line shape of the 3×3PWO-II matrix. . . . . . . . . . . . . 163


10.17Experimental line shape of a 3×3PWO-II matrix. . . . . . . . . . . . . 164

10.18The linearity of the response of a 3×3matrix. . . . . . . . . . . . . . . . . . 164

10.19Experimental energy resolution as afunction of incident photon energy. . 164

10.20Pack of 4 crystals wrapped with re-flector. . . . . . . . . . . . . . . . . . 165

10.21Back view of PROTO60. . . . . . . . 165

10.22Back view of the fully assembledPROTO60. . . . . . . . . . . . . . . 165

10.23Temperature stability of thePROTO60. . . . . . . . . . . . . . . 166

10.24Measured response of a single mod-ule of the PROTO60. . . . . . . . . . 166

11.1 Schematic layout for the realizationof the target spectrometer EMC ofPANDA. . . . . . . . . . . . . . . . . 171

11.2 Timeline for the realization of thetarget spectrometer EMC. . . . . . . 174

11.3 Tasks and responsibilities for the re-alization of the target spectrometerEMC. . . . . . . . . . . . . . . . . . 175


187

List of Tables

1.1 Main requirements for the PANDAEMC. . . . . . . . . . . . . . . . . . 4

2.1 Injection parameters, experimentalrequirements and operation modes. . 16

3.1 Main requirements for the PANDAEMC. . . . . . . . . . . . . . . . . . 38

4.1 Properties of various scintillationcrystals. . . . . . . . . . . . . . . . . 40

4.2 Distribution of the crystal propertiesof accepted PWO crystals. . . . . . . 45

4.3 Selected parameters of the electroncapturing centers in PWO crystals. . 47

4.4 Comparison of the induced absorption. 56

5.1 Properties of tested APDs. . . . . . 73

5.2 Conversion layer thicknesses of thetested APD types. . . . . . . . . . . 73

5.3 Dark current parameters after pro-ton irradiation of the tested APDs. . 75

5.4 Specification for PANDA VPT. . . . 81

5.5 Parameters of the RIE FEU-189 VPT. 81

6.1 Measured preamplifier parameters. . 92

7.1 Geometrical parameters of the EMC. 107

7.2 Thermal properties of PWO. . . . . 116

7.3 List of services in the backward area. 120

8.1 Properties of LEDs. . . . . . . . . . 138

9.1 Efficiency of different event selectioncriteria. . . . . . . . . . . . . . . . . 150

9.2 Signal to background ratio for differ-ent hc background channels. . . . . . 150

Documents

PANDA Electromagnetic Calorimeter (EMC)