Calorimetry in Particle Physics: Proceedings of the Tenth International Conference, California, USA 25-29 March 2002

Proceedingsof the Tenth InternationalConference on ALORIMETRY IN PARTICLE PHYSICS This page intentionally left blank...

Author: Ren-Yuan Zhu

20 downloads 888 Views 52MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Proceedingsof the Tenth InternationalConference on

ALORIMETRY IN PARTICLE PHYSICS

This page intentionally left blank

Proceedings of the Tenth International Conference on

ALORIMETRY IN PARTICLE

PHYSI

Pasadena, California, USA

25 - 29 March 2002

Editor

Ren=YuanZhu California Institute of Technology, USA

we World Scientific b

New Jersey London Singapore Hong Kong

Published by World Scientific Publishing Co. Re. Ltd. 5 Toh Tuck Link, Singapore 596224 USA office: Suite 202,1060 Main Street, River Edge, NJ 07661 UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

British Library Cataloguing-in-PublicationData A catalogue record for this book is available from the British Library.

CALORIMETRY IN PARTICLE PHYSICS Proceedings of the Tenth International Conference Copyright 0 2002 by World Scientific Publishing Co. Re. Ltd. All rights reserved. This book, or parts thereof; may not be reproduced in anyform or by any means, electronic or mechanical, includingphotocopying, recording or any informationstorage and retrieval system now known or to be invented, without wrinen permissionfrom the Publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the publisher.

ISBN 981-238-157-0

This book is printed on acid-free paper.

Printed in Singapore by Uto-Print

International Advisory Committee G. Bellettini J. Colas A. Ereditato F. L. Fabbri H. A. Gordon D. Green P. Jenni T. Kobayashi A. Maio A. Menzione H. Oberlack A. Para K. Pretzl J. Rutherfoord R. Wigmans R. Yoshida R-Y. Zhu

INFN /Pisa LAPP / Annecy INFN/Napoli INFN/Frascati BNL/Brookhaven FNAL /Bat avia CERN/Geneva ICEPP/Tokyo FCUL & LIP/Lisbon INFN/Pisa MPI/Munich FNAL/Batavia Bern U/Bern U. Arizona/Tucson Texas Tech/Lubbock ANL / Ar gonne Caltech/ Pasadena

Local Organization Committee J. Adams B. Barish B. Choudhary M. Gataullin D. Hitlin H. Newman F. Porter L. Xia L. Zhang R-Y. Zhu

MSFC/NASA Caltech Caltech Caltech Caltech Caltech Caltech Caltech Caltech Caltech

vii

viii

Artistry Liyuan Zhang

Caltech

Computing Support Juan Barayoga Frank Porter

Caltech Caltech

Conference Web Lei Xia Kejun Zhu

Caltech Caltech

Logistics Brajes h Choudhar y Marat Gataullin

Caltech Caltech

Photography Bob Paz Qing Wei Liyuan Zhang

Caltech Caltech Caltech

Scientific Secretariat Marat Gataullin Lei Xia

Caltech Caltech

Secretariat Debbie Kingston Virginia Licon Marc Rincon Betty Smith Meiske Van der Eb

Caltech Caltech Caltech Caltech Caltech

Special thanks to the following sponsors: Division of Physics, Mathematics and Astronomy, Caltech Hamamatsu Photonics K.K. Kuraray Chemical Co. Shanghai Institute of Ceramics, Chinese Academy of Sciences

PREFACE The “loth International Conference on Calorimetry in Particle Physics’’ (Calor2002) was held from March 25th to 2gth, 2002, at the Beckman Institute Auditorium in California Institute of Technology (Caltech) and at the elegant Huntington Library, Pasadena, California, USA. The nine previous International Conferences on Calorimetry in High Energy Physics were held in Annecy (2000), Lisbon (1999), Tucson (1997), Frascati (1996), Brookhaven (1994), Elba (1993), Corpus Christi (1992), Capri (1991) and Fermilab (1990). Starting from this edition, the name of this series conference is changed to engage broad participants in particle physics community. The Calor2002 scientific program consists of three parts: 6 introductory talks, 11 sessions and 3 perspective talks. The conference is featured in all presentations given in plenary sessions, which provides an excellent opportunity for young physicists to be introduced into the community. It is also the first time in this series of conference a session is dedicated to the “Calorimetry in astrophysics”. The conference was opened by T. Tombrello, Division Chair of Physics, Mathematics and Astronomy, and D. Hitlin, Director of Caltech High Energy Physics. The program started with the introductory session chaired by R.Y. Zhu and A. Ereditado. The presentations in this session, by K. Pretzl (history), D. Fournier (LHC), S. Swordy (astrophysics), W. Wisniewski (super B factory), R. Frey (linear collider) and V. Morgunov (energy flow), covered historical, current and new trends of calorimetry development. The conference also saw a lively discussion on the issue of the “energy flow”. The 11 sessions were organized around various calorimetry techniques and applications by distinguished conveners. The calorimetry in astrophysics session, convened by T. Parnell, covered a broad range of calorimetry technologies implemented in astrophysics experiments: AMANDA, AMS, ATIC, BLAZARS, CREAM, EUSO, GLAST, OWL, PAMELA, P A 0 and VERITAS. In the crystal calorimetry session, convened by W. Wisniewski, calorimeters in BaBar, BELLE, CMS, COSY and PRIMEX were presented. The medical application session, convened by C. Woody, covered PET detector and simulation. W. Moses’s enlightening perspective talk comparing calorimetry with PET was given in this session. In the silicon calorimetry session, convened by D. Strom, experiences from OPAL and ZEUS as well as on-going design studies by linear collider communities in Europe and US were presented. The simulation session, convened by C. Seez, covered physics studies and detailed calorimeter simulations by ATLAS, CDF and CMS, where implementation of

IX

X

the “energy flow” was discussed in details. In the calibration session, convened by M. Gataullin, a broad range of calibration and monitoring techniques implemented in AMANDA, BELLE, CDF, CMS, D 0 , KLOE, MINOS, PHENIX, PRIMEX, SNO and ZEUS was presented. The Cerenkov calorimetry session, convened by S. White, covered calorimetry in CMS and RICE, where an interesting new idea of radio Cerenkov detection was discussed. In the scintillation calorimetry session, convened by M. Cavalli-Sforza,calorimeters from ATLAS, Borexino, CDF, CMS MINOS and ZEUS as well as on-going design studies for a TESLA detector were presented. The electronics session, convened by J. Elias, covered electronics development in ATLAS, BaBar, CDF, CMS, H1 and LHCb. In the ionization session, convened by P. Schacht, calorimeters from ATLAS, DO and El58 were presented. Finally, the jet measurement session, convened by J. Terron, covered experiences from ALEPH, CDF, DELPHI, DO, HI, OPAL and ZEUS as well as on-going design studies by linear collider communities in Europe. The Calor2002 scientific program was concluded by the perspective session chaired by R. Wigmans, where perspective presentations by D. Green (High Energy Physics) and J. Krizmanic (Astrophysics) promised continuous interest of the community and pointed out several key issues for the calorimeter development in the future. The Calor2002 conference was attended by more than 150 physicists from different fields and countries. More than hundred presentations were given in the conference, most of which are published in these proceedings. Including convener’s reports, these proceedings contain a total of 114 papers. Also included in these proceedings are two papers which were not presented in the conference: a quartz fiber paper in the Cerenkov calorimetry session and an interesting study on jet energy measurement in the jet measurement session. All conference information, including presentations and proceedings papers, can be found on the Web: http://3w.hep.caltech.edu/calor2002. The credit of the scientific program goes to the members of the IAC and conveners who organized the session and recruited speakers. The success of the conference should also be attributed to the members of the LOC, scientific staff and secretaries who devoted much of their time to make the conference smooth and enjoyable. Finally, credit should also be given to all participants and speakers, both for their efforts in giving presentations and submitting their papers on the due time. The conference is also benefitted by the sponsorship of the Caltech PMA division and three industrial sponsors: Hamamatsu, Kuraray and SICCAS. Ren-yuan Zhu

CONTENTS

Preface

ix

Introduction Historical Review of Calorimeter Developments K. Pretzl

3

Overview and Status of Calorimetry at LHC D. Fournier

17

Calorimetry in Astrophysics S. P. Swordy

43

Considerations for Calorimetry at a Super B Factory W. Wisniewski (contribution not received)

-

Calorimeter Considerations for a Linear Collider Detector R. E. Frey

54

Calorimetry Design with Energy-Flow Concept (Imaging Detector 70

for High-Energy Physics)

V. L. Morgunov

Calorimetry in Astrophysics Covener’s Report T. Parnell

87

ATIC, a Balloon Borne Calorimeter for Cosmic Ray Measurements J. Isbert et al.

89

ATIC Backscatter Study Using Monte Carlo Methods in FLUKA & ROOT T. Wilson et al.

95

A Silicon-Tungsten Calorimeter for Cosmic-Ray Physics V. Bonvicini et al.

101

Electromagnetic Calorimeter for the AMS-02 Experiment R. Kossakowsti et al.

108

Performances of the AMS-02 Electromagnetic Calorimeter P. Maestro et al.

114

xi

xii

The Status of GLAST CsI Calorimeter A . Chekhtman

121

Performance of GLAST Calorimeter R. Terrier et al.

127

Cosmic Ray Energetics And Mass (CREAM): Calibrating a Cosmic Ray Calorimeter 0. Ganel et al.

133

VERITAS: A Next Generation Atmospheric Cherenkov Detector and Calorimeter for Gamma-Ray Astronomy F. Krennrich

139

Pierre Auger Observatory: The World’s Largest Calorimeter A . K. %pathi EUSO and OWL: Atmospheric Cosmic Ray Calorimetry from Space K. Arisaka (contribution not received)

151 -

Calorimetry (GeV-EeV) in AMANDA and IceCube Neutrino Telescopes J. Lamoureux (contribution not received)

Crystal Calorimetry Performance of a Small Angle BGO Calorimeter at BELLE H. - C. Huang

161

Performance and Calibration of the Crystal Calorimeter of the B a B a r Detector M. Kocian

167

A Systematic Study of Radiation Damage to Large Crystals of CsI(T1) in the B a B a r Detector T. Hryn’ova

175

Performance and Upgrade Plans of the BELLE Calorimeter B. A . Shwartz

182

Development of Yttrium Doped Lead Tungstate Crystal for Physics Applications Q. Deng et al.

190

Performance of PWO Crystal Detectors for a High Resolution Hybrid Electromagnetic Calorimeter at Jefferson Lab A . Gasparian

208

xiii

The PHOTON BALL at COSY R. Novotny et al.

215

Overview of the CMS Electromagnetic Calorimeter P. Lecomte (contribution not received) Performance of the PWO Crystals of the CMS Electromagnetic Calorimeter F. Cavallari

223

Avalanche Photodiodes for the CMS Lead Tungstate Calorimeter R. Rusaclc et al.

231

CMS/ECAL Barrel Construction and Quality Control E. Aufiay

240

Medical Applications Covener’s Report C. Woody

249

Synergies Between Electromagnetic Calorimetry and PET W. W. Moses

251

LSO - From Discovery to Commercial Development C. L. Melcher (contribution not received) New Scintillating Crystals for PET Scanners P. Lecoq A Simulation Framework for Positron Emission Tomography Based on GEANT4 J, Collot et al.

262

274

Silicon Calorimetry Covener’s Report D. Strom

285

Performance of the OPAL Si-W Luminometer at LEP 1-11 G. Abbiendi et al.

287

The ZEUS Hadron Electron Separator, Performance and Experience P. Gottlicher

296

xiv

Design Considerations for a Silicon-Tungsten Electromagnetic Calorimeter for a Linear Collider Detector R. Frey et al.

A Si-W Calorimeter for Linear Collider Physics H. Videau and J.-C. Brient

304 309

Simulation Covener’s Report C. Seez

323

QCD Jet Simulation with CMS at LHC and Background Studies to H -+ yy Process S. Shevchenko et al.

325

Comparisons of Electron and Muon Signals in the ATLAS Liquid Argon Calorimeters with GEANT4 Simulations P. Loch et al.

331

Data Volume Reduction Strategies in the CMS Electromagnetic Calorimeter P. Paganini

339

Performance of CDF Calorimeter Simulation for Tevatron Run I1 C. A . Cumat Simulation of Hadronic Showers in the ATLAS Liquid Argon Calorimeters A . Kiryunin et al. MC Simulation of the ATLAS Hadronic Calorimeter Performance M. J. Varanda Simulation Studies of the Jet and Missing Transverse Energy Performance of the ATLAS Calorimeters M. Wielers Jet Energy Reconstruction with the CMS Detector S. Kunori

345

354 361

367 375

Calibration & Monitoring Covener’s Report M. Gataullin

385

Calibration of the KLOE Electromagnetic Calorimeter C. Gatti et al.

388

xv

Monitoring and Calibration of the BELLE Electromagnetic Calorimeter K. Miyabayashi

394

Calibration and Monitoring of the ZEUS Uranium Scintillator Calorimeter at HERA M.Barbi

401

Calibration of the PHENIX Lead Scintillator Calorimeter

409

H. Torii DO Calorimeter Calibration U.Bassler

413

Calibrating a Longitudinally Segmented Calorimeter 0. Lobban

421

The Calibration of the MINOS Detectors R. Nacho1 et al.

428

First Results from the MINOS Calibration Detector P. Vahle et al.

436

Calibrating the SNO Detector Response A . Hamer

442

Time Calibration of AMANDA: Three Variations of a Theme of To K. D. Hanson

452

Absolute Calibration of Electromagnetic Calorimeter at LHC with Physics Processes L. Xia, T. Hu and R.-Y. Zhu

459

Monitoring Light Source for CMS Lead Tungstate Crystal Calorimeter at LHC L. Zhang et al.

469

LED based Light Monitoring System for the PRIMEX Experiment at Jefferson Lab S. Danagoulian

479

Cerenkov Calorimetry Covener’s Report S. White Influence of Phase Transition on the Optical Transparency of Lead Fluoride Crystals Q. Deng et al.

489

491

xvi

Explicitly Radiation Hard Fast Gas Cerenkov Calorimeter 0. Atramentov

497

Present Status of CMS HF Quartz Fiber Calorimetry Y. One1

504

Calorimetry of the RICE Detector S. Razzaque

515

Radio Cherenkov Detection of High Energy Particles D. Saltzherg (contribution not received) Radiation Hardness Studies of High OH- Quartz Fibres for a Hadronic Forward Calorimeter of the Compact Muon Solenoid Experiment at the Large Hadron Collider I. Dumanoglu et al.

-

521

Scintillation Calorimetry Covener's Report M. Cavalli-Sforza

531

Status of the ATLAS Tile Hadronic Calorimeter Production A. Henriques

532

Studies of the ATLAS Tile Hadron Calorimeter Performance S. Ne'meEek and I. Korolkov

538

An Overview of CMS Central Hadron Calorimeter S. Katta

544

Performance and Calibration of the Forward Plug Calorimeter at ZEUS A. Benen

549

Plug Shower Maximum Detector for CDF Run I1 A. Attal

557

CDF I1 Integrated Calorimetry Environment S. Dell'Agnello

563

Borexino: A Real Time Liquid Scintillator Detector for Low Energy Solar Neutrino Study L. Maramonti

570

The MINOS Far Detector Construction and Quality Assurance Testing L. Mualem

578

xvii

A. New Hermetic Electromagnetic Calorimeter Design for Future Collider Experiments E. Kistenev et al.

584

The Tile HCAL Calorimeter for the TESLA Detector, a Status Report V. Korbel

591

Electronics Covener’s Report J. Elias

605

The ATLAS Tile Calorimeter Front End Electronics F. Martin

607

Overview of Liquid Argon Front End Electronics E. Ferrer Ribas

613

First Results with the QIE8 ASIC s. Los (contribution not received)

-

Front-End Electronics for the CMS Preshower Detector A . Go et al.

621

The Front-end Electronics for the LHCb Calorimeters D. Breton

627

Front-end Electronics Upgrade for the CDF Calorimeters C. A . Nelson and T. M. Shaw

644

The Electronics of the New H1 Luminosity System V. Boudry et al.

652

The BaBar Electromagnetic Calorimeter in its Third Year of Operation I. G. Eschrich

H1 Calorimeter DAQ Upgrade for HERA-I1 D. Hoffmann, P.-Y. Duval and C. Vallee

658 665

Ionization Calorimetry Covener’s Report P. Schacht

677

Argon Purity Measurement of the DO Calorimeter A . Besson and G. Sajot

679

xviii

The Run I1 DO Calorimeter: Electronics Upgrade and its Performance N. Parua

687

ATLAS LAr EM Calorimeter: Construction and Uniformity of Response S. Rodier

695

Performance of ATLAS EM Modules in Test Beam D. Zerwas

703

The ATLAS Hadronic Endcap Calorimeter M. Fincke-Keeler

712

Performance of the ATLAS Hadronic End-Cap Calorimeter in Beam Tests A . E. Kiryunin

720

ATLAS Forward Calorimeter (FCAL)

K. K.

300

(contribution not received) A High Resolution Luminosity Monitor for SLAC Experiment El58 G. M. Jones

728

Jet Measurement Simulations and Prototyping Studies for a Digital Hadron Calorimeter V. Zutshi (contribution not received) Jet Measurements with the Aleph Detector at LEP2 M. N. Minard

739

Calorimetry Optimised for Jets H. Videau and J.-C. Brient

747

The Jet Calibration in the H1 Liquid Argon Calorimeter C. Schwanenberger

76 1

Setting the Jet Energy Scale for the ZEUS Calorimeter M. Wing

767

Jet Reconstruction at CDF M. Tonnesmann

773

Suppression of Pileup Noise in a Jet Cone A . Savine

781

xix

D0’s Recent Results and Experiences with the kT and Cone Jet Algorithms J. Krane

786

Jet Measurement in OPAL I. Nalcamura

793

Developments on Jet Reconstruction by DELPHI A. Kiiskinen

798

E-Flow Optimization of the Hadron Calorimeter for Future Detectors S. Magill et al.

806

On the Energy Measurement of Hadron Jets R. Wigmans, 0. Lobban and A . Sriharan

814

Perspective The Future of Calorimetry in High Energy Physics D. Green

837

Future Experiments in Astrophysics J . F. Krizmanic

867

Conference Pictures

881

Author Index

889

List of Participants

895

This page intentionally left blank

Overview and Introduction Chairpersons: R. Y. Zhu and A . Ereditato

K. Pretzl

Historical Review of Calorimeter Developments

D. Fournier

Overview and Status of Calorimetry at LHC

S. Swordy

Calorimetry in Astrophysic

*W. Wisniewski

Considerations for Calorimetry at a Super B Factory

R. E. Frey

Calorimeter Considerations for a Linear Collider Detector

V. L. Morgunov

Calorimetry Design with Energy-Flow Concept (Imaging Detector for High-Energy Physics)

*Written contribution not received 1

This page intentionally left blank

HISTORICAL REVIEW OF CALORIMETER DEVELOPMENTS

KLAUS PRETZL Laboratory for High Energy Physics, University of Bern, 3012 Bern, Switzerland E-mail: pretzlOlhep.unibe.ch Calorimeters are widely used in astro- and particle physics experiments. They provide the means t o explore new physics in an energy range from several eV to more than 1020 eV. An attempt is made to discuss the historical developments of these devices. This report is far from being complete and the author apologizes for possible omissions and misquotations.

1. Introduction Calorimeters belong t o the most important instruments t o measure the energy of neutral and charged particles produced with cosmic rays or with particle accelerators. Their development was very much driven by the quest for new frontiers in astro- and particle physics. Several types of calorimeters have been developed. There are the so-called true calorimeters, which operate at very low temperatures and are used as thermal sensors. Cryogenic calorimeters are the most sensitive devices in the energy region eV to keV. They are frequently used in direct dark matter detection and double beta decay experiments. Then there are the sampling calorimeters, which are based on measuring the energy loss of secondary particles in an active material, which is sandwiched between the absorbing material of the calorimeter. Ionization or scintillation detectors are commonly used as active layers. Sampling calorimeters are among the most frequently employed calorimeters in high energy physics experiments because they are relatively cheap and they can cover a large energy range, typically from GeV to TeV. Homogeneous total absorption calorimeters, like crystals or liquids (Ar, Kr, Xe) have mostly been developed for electromagnetic shower detection. They yield outstanding energy resolutions but they are rather expensive. Cerenkov calorimeters are based on the detection of Cerenkov radiation of relativistic particles in transparent materials. An early example is the lead glass calorimeter for electromagnetic shower detection. Cerenkov calorimeters with enormous dimensions using large underground water tanks, sea water, polar ice and the atmosphere of the earth have recently

3

4

been developed to measure solar, atmospheric and cosmic neutrinos as well as ultra high energy particles (UHEP) in cosmic rays. Large atmospheric calorimeters also make use of the fluorescence scintillation light emitted when particles pass through the atmosphere. The observation of the fluorescence light of air showers in the atmosphere from a space station is planned for the near future. Properly instrumented, the atmosphere would be the largest calorimeter ever put in operation and would enable us to study UHEP with energies in excess of lo2' eV. As of today calorimeters allow us to explore new physics in the energy range from a few eV to more than lo2' eV. It took the ideas and the devotion of many people and hard work to get there. I apologize to all those people whose contributions I omitted or misquoted in my attempt to give a historical review of the developments in this field. 2. Early Developments

One of the early pioneers in developing the art of calorimetry was W. Orthmann, a close collaborator of W. Nernst (Nobel prize winner for chemistry in 1920). Orthmann' developed a differential calorimeter with which he could measure heat transfers of the order of pWatt. Using this true calorimetric technique, he and L. Meitner2 were able to determine the mean energy of the continuous @-spectrum in 210Bi to be E = 0.337 MeV &6%. This value agrees very well with the mean kinetic energy of E = 0.33 MeV of the emitted @-particles. Their measurements contributed to the notion of a continuous ,8-spectrum leading to Pauli's neutrino hypothesis in 1930. Applying this technique to high energy particles would fail, since the temperature increase A T = caused by the energy loss AE of a high energy particle is unmeasurably small ( A T 10-'K for A E lo2' eV) due t o the heat capacity of the large absorber mass (- 10' g Fe) necessary t o contain the shower. Nevertheless this technique was revitalized in the 1980s by the development of calorimeters for dark matter detection (see chapter 9). In 1954 N.L. Grigorov3 put forward the idea of sampling calorimeters using ionisation chambers (arrays of proportional counters) interspersed between thick iron absorber sheets to measure cosmic ray particles with energies E > 1014 eV. The visible energy of the particle Evisible = Jn(x)dx is then determined from the number of secondary particles in the shower n(x) and their in the ionisation detectors. Because of the invisible energy loss energy loss in the absorber plates and in the detector sheets, sampling calorimeters need to be calibrated in particle beams with known energies in order t o obtain absolute energy measurements. In 1957 Grigorov and his collaborators4 constructed a sampling calorime-

3

-

-

5

ter in the Pamir mountains, at an altitude of 3860 m above sea level. Their calorimeter also employed 10m2of emulsion sheets, which they placed between 3 lead sheets to study details of the primary interaction. In order to identify corresponding events in the emulsions and the calorimeter, two layers of emulsion sheets on top of each other were used: one fixed in space, the other moved by a certain amount in a certain time interval with respect to the other one. The two emulsion film images were then matched and thus the time of the shower passage determined. In the 1960s high energy accelerators in the USA and Europe became the main facilities to study high energy phenomena. While charged particles were measured with high accuracy in magnetic spectrometers, the energy of neutral particles like photons, 7ro and neutrons could only be measured by calorimetric means. First successful attempts to measure photon and electron energies with compact electromagnetic sampling calorimeters were made at CALTECH by C. Heusch and C. Prescott5. They studied the electromagnetic shower development of electrons and photons in the energy region of 100 MeV to 5 GeV. For that purpose they built two sandwich counters, one out of plastic scintillator and one of lucite material with lead inserts as absorbers. With this they were able to exploit and compare the performance of calorimeters based on the ionization loss of shower particles in scintillators and those based on Cerenkov radiation in plastic materials. They also studied sampling fluctutations and shower containment by changing the thickness of the lead inserts. A similar study for a hadron calorimeter was done by the Karlsruhe group under the leadership of H. Schopper motivated by the idea of measuring n-p elastic scattering at C E R N ~ . With the vision that calorimeters will play a role not only in accelerator experiments but also in experiments on board space missions, R. Hofstadter and his collaborators7 developed large homogeneous N a I (Tl) and CSI total absorption calorimeters. Because of their robustness, CSI crystals were particularly suited for space-born gamma ray astronomy. The persuasively good energy resolutions obtained with these calorimeters also made them powerful tools for studying inclusive 'r7 production in hadron collisions as well as electron and gamma final states in e+e- collisions at SPEAR. Hofstadter's original ideas were later further developed and realised in the very successful CRYSTAL BALL calorimeter at SLAC, with which many of the charmonium states have been discovered.

3. Segmented Calorimeters Large charged particle spectrometers were disigned for exploring the physics in an energy domain which was made accessible by the new large proton accelera-

6

tors at Fermilab and at CERN (SPS) in the early 1970s. At that time the usefulness of calorimetric energy measurements to complement the measurements in magnetic spectrometers was not yet widely appreciated. However, the parton picture of hadrons, strongly supported by the SLAC single-arm spectrometer experiment, demanded a search for parton jets, which would for example manifest their presence by a large transverse energy flow ET in deep inelastic hadron-hadron collisions. For this search, segmented calorimeters covering a large solid angle were developed. Although jets could not be unambiguously identified at Fermilab, ISR and SPS energies, they were clearly detected a.t higher energies in the UA2 and UA1 experiments at the p-p collider at CERN. The extraordinarily successful use of segmented calorimeters in the UA1 and UA2 experiments in discovering not only jets but also the intermediate vector bosons in their decay-channels Z -+ 2 leptons and W* -+ l e p t o n + m i s s i n g E T ( EFiss)made them indispensable tools for future collider experiments searching for the top quark (t + j e t s leptons), the Higgs ( H -+ yy) and SUSY particles (EFiss). In 1973 neutral currents were discovered by GARGAMELLE at CERN. This important step was decisive for future neutrino experiments. To measure neutral current reactions in neutrino detectors required calorimeters of multiton mass with very large dimensions. One of the first upgrades in this direction was initiated by B. Barish and collaborators' of CALTECH. They improved the calorimetric shower detection capability of their neutrino experiment at Fermilab by introducing a wavelength shifting (WLS) read-out of their very large-sized scintillator tiles. The loss of photoelectrons due t o this WLS technique, typically a factor of 10, could be compensated somewhat by using thicker scintillators. However, the main advantage of the WLS technique over the conventional light-guides is its simplicity and the smaller number of readout channels necessary, which is reflected in the lower costs. By reading the light from at least two opposite sides of a scintillator tile, it was even possible to locate the center of gravity of the shower to within a few cm accuracy.

+

4. Segmentation Using Wavelength Shifter Read-out The WLS principle, which was first introduced by W. Shurcliffg, further discussed by R.L. Garwin" and later developed by G. Keil" provided the capability to construct tower-structured scintillation counter sampling calorimeters. Tower-structured calorimeters were an essential ingredient for the jet search, since they allowed to trigger on events with a large transverse energy ET and to determine the jet energy as well as the jet size. Two types of WLS systems were developed: WLS sheets and WLS rods (later replaced by fibers).

7

W. Selove12 used WLS sheets for his jet search experiment at Fermilab and D. Wegener13 developed a fine sampling tower-structured electromagnetic calorimeter for the ARGUS experiment at the DORIS e+e- collider. A novel read-out system using WLS rods was introduced by the author14 for the jet search experiment NA5 at the SPS at CERN. By using a different WLS color (yellow) in the electromagnetic part of the NA5 calorimeter15 than in the hadronic part (BBQ green), it was possible to guide the light of both colors in one rod to the two photomultipliers which, by using appropriate filters, were sensitive to either yellow or green light. This way the energy deposited in the electromagnetic and in the hadronic part of the calorimeter could be measured separately. The NA5 calorimeter was the first tower-structured scintillator sampling calorimeter in operation at CERN. It is still in use in the NA49 experiment in the North Area of the SPS. On the basis of the first positive experiences with WLS read-outs, UAl and UA2 at CERN and CDF at Fermilab developed their calorimeters using this principle. The next step was to introduce WLS-doped optical fibers and scintillation optical fibers for the next generation scintillator sampling calorimeters. However, no company in the fiber-producing industry was interested in developing these special fibers with only slim expectations for a large market. In 1982 D. Treiller, P. Sonderegger and the author persuaded J. Thevenin at SACLAY to develop scintillating fibers and WLS fibers for calorimetry. Thanks to J. Thevenin, very soon the first tower-structured electromagnetic calorimeter16 using WLS fibers doped with K27 (BBQ was not successful, since it cracked during extrusion) could be successfully tested. It was later given the name SHASHLIK. Due to its good energy resolution and its relatively low construction costs, SHASHLIK is being used in DELPHI, HERA-B, PHENIX (RHIC) and LHCB experiments. WLS fiber read-out in hadronic tile calorimeters was further developed by the CDF collaboration a t Fermilab. 0. Gildemeister proposed a novel structure for the ATLAS hadron calorimeter, the so-called TILECAL. In this structure, the scintillating tiles and the read-out fibers run parallel to the particles. This design turned out to be cheap and allowed a hermetic construction, projective geometry as well as depth sampling (3 times). Equally promising were the first tests of J. Thevenin’s scintillating fibers used in a prototype Calorimeter. This technique was further developed, leading to the SPACAL17 calorimeter - a t the time a most promising candidate for use in LHC experiments. SPACAL was a compensating, homogeneous, high resolution calorimeter with properties very close to homogeneous calorimeters. However, depth segmentation turned out to be difficult and its production was more expensive than other calorimeters with similar performances. Therefore it finally did not get used in LHC experiments, but its great performance

8

was fully appreciated in smaller-sized experiments like CHORUS, H1 backcalo, AMS-2 and KLOE. Mainly for heavy ion physics at the CERN SPS there was a demand for a fast and radiation-hard calorimeter to measure forward energy flow, which was used for the determination of the centrality of the collision. P. Gorodetzkyl* pioneered the development of a quartz fiber/lead calorimeter with a very fast response (6 ns pulse to pulse separation) due to Cerenkov radiation. The radiation resistance of quartz fibers is considerably higher than that of scintillating fibers. However, because of its moderate energy resolution this type of calorimeter is limited in its applications. Quartz fiber calorimeters are employed in the N50, NA52, ALICE (CASTOR), and CMS experiments. 5. Liquid Ionization Chambers

One of the limiting factors of a sampling calorimeter are the sampling fluctuations and the uniformity of response. To reduce both these factors and still keep the sampling calorimeter compact with minimal dead space between towers, W. Willis and V. Radekalg, in the early 1970s, introduced liquid argon (LAr) as active medium. This technique has proven to be highly successful and has been used in many fixed target and collider experiments (R807/ISR, MARK2, CELLO, NA31, SLD, HELIOS, DO, HERA, ATLAS). The development of the LAr ionization chamber technique was also pioneered by the group of J. Engler” at Karlsruhe. Although fine sampling LAr calorimeters surpass the performance of their scintillator counterparts in many respects, they have the disadvantage that they require cryogenics (boiling temperature of LAr is T = 87 K) and high purity of the liquid. In order to overcome the cooling problem, J. Engler and H. KeimZ1 developed liquid ionization chambers which operate at room temperature. They used tetramethylsilane Si(CH2)4 (TMS), which has a relatively high electron mobility. TMS and tetramethylpentane (TMP) have the double advantage of operating at room temperature and being hydrogeneous, providing the necessary presupposition for a compensating calorimeter. A disadvantage is the high degree of purification needed. Warm liquid calorimeters have been proposed for the UA1 upgrade (TMP) and are being used in the CASCADE experiment. In 1990 D. FournierZ2introduced a novel design for a LAr calorimeter, the so-called ”accordeon”,which has practically no dead space between towers and provides better uniformity of response, less cabling (signals can be extracted from the front and back face of the calorimeter) and fast signal extraction due to low capacitance. The ”accordeon” was adopted as electromagnetic calorimeter for the ATLAS detector. In the evaluation of all possible types of calorimeters, CMS chose PbW04 crystals for their electromagnetic calorimeter

9

with the argument of obtaining the best energy resolution, while for ATLAS the better uniformity of response and the better angular resolution with the ” accordeon” were decisive. In the mid-1980s C. Rubbia started the development of a LAr total absorption calorimeter for solar and atmospheric neutrino experiments at the Gran Sasso Laboratory. A 600 ton calorimeter has been realized and is going to be installed in the Gran Sasso Laboratory. The ICARUS collaboration plans t o complete the calorimeter to 3 kton to perform a long baseline neutrino oscillation experiment using the CNGS beam from CERN. A LKr total absorption calorimeter has been developed by the NA48 collaboration and is in operation to study CP violation in Kg,L + I T O I T O and Kg,L -+IT+IT- channels. At the Paul Schemer Institute in Villigen a LXe calorimeter is in construction to search for forbidden p + e y decays which, if found, would indicate physics beyond the Standard Model.

+

6. Compensating Calorimeters

Even with minimal sampling fluctuations the energy resolution of hadron calorimeters is mainly limited by large shower fluctuations, which are due to fluctuations in the electromagnetic component (IT’, etc.) of the shower and the invisible energy due to nuclear excitations, muons and neutrinos. It was suggested by C. Fabjan and W. W i l l i ~that ~ ~ some of this invisible energy can be recuperated using depleted 238U plates as absorber. The energy loss will be compensated by the emission of soft neutrons and gammas in fission processes of the uranium. Measurements supported these ideas and the first compensating 238 U/scintillator calorimeter was employed in the Axial Field Spectrometer at the ISR24. Compensation turned out to be essential for a correct jet energy determination and therefore was a main consideration for the design of calorimeters for high energy collider experiments. However, there are also other ways t o achieve compensation, namely by proper weighting of the electromagnetic and hadronic components, as shown by H. Abramowicz et a125for the WA1 calorimeter and W. Braunschweig et a126for the H1 calorimeter. In 1987, mainly due to the detailed work of R. wig man^^^ supplemented by the work of Brueckmann et a128,the compensation effects were fully understood. Wigmans could show that compensation can be achieved with absorber materials other than uranium by tuning the influencing parameters like choosing the appropriate sampling fractions. For example, compensation can be achieved by choosing the thickness of the absorber and of the scintillator as follows: U/sci = 1/1, Pb/sci = 4/1, Fe/sci = 15/1. The crucial role of hydrogen in the active material was clearly demonstrated by measurements of the

10

L3 collaboration. LAr calorimeters do not compensate, due to the missing hydrogen. However, compensation can be achieved by prolonging the charge collection time so that the gammas from neutron capture reactions can be detected or by using the above mentioned weighing procedure. In order to achieve the best jet energy measurements large compensating calorimeters were built at HERA (ZEUS: 238U/sci, H1: Pb, steel/LAr) and at the Fermilab collider (DO: 238 U/LAr). However, compensation degrades the energy resolution for electromagnetic showers considerably. Since the search for the Higgs, with its prominent decay channel H + yy, is one of the prime goals at the LHC, the ATLAS and CMS experiments have chosen non-compensating calorimeters with the highest energy resolution capabilities for electromagnetic showers.

7. Other Sampling Calorimeters The high segmentation capability and the relatively low cost were the main arguments for the development of gas sampling calorimeters. All four LEP experiments were equipped with such devices. The modest obtainable energy resolution (- 20%/&?) for electromagnetic showers with such calorimeters is mainly due to Landau and pathlength fluctuations. In addition, gas sampling calorimeters have to cope with stability problems due to temperature and pressure changes. The disadvantages outweighed the advantages such that they were not further developed. In 1983 P.G. Rancoita and A. Seidman” introduced silicon detectors for electromagnetic calorimetry. These calorimeters were further developed by the SICAPO c o l l a b ~ r a t i o n ~Large-sized ~. silicon detectors employing relatively low-resistivity (less expensive) material are appropriate for calorimeters used in experiments requiring compact geometry, fast signal response and operation in strong magnetic fields. Disadvantages are the high cost and the poor radiation resistance of silicon. Nevertheless, they are used in the ZEUS experiment as electron-hadron separator, in the OPAL experiment at LEP as luminosity monitor, and in the PAMELA experiment as imaging calorimeter. Si/W calorimeters are being seriously considered for e+e- linear collider experiments of the next generation.

8. Crystal Calorimeters Following earlier developments using NaI (Tl) crystals for physics at e+e- colliders, the CRYSTAL BALL c ~ l l a b o r a t i o nat ~ ~SLAC demonstrated very impressively the discovery potential of a tower-structured high-resolution crystal

11

calorimeter. It made major contributions to the discovery and the better understanding of the charmonium states. Crystal calorimeters are mainly used for electromagnetic shower detection. In later developments, where the CRYSTAL CLEAR collaboration played a leading role, NaI (Tl) crystals have been replaced by denser and non-hygroscopic materials. Among the most popular are: CsI, BGO and PbW04. The production as well as the physical properties of many of these crystals were already studied during World War 11, since fluorescent and scintillating crystals were developed to mark firing tanks or guns for airplane attacks at night. The typical energy resolutions of (1.5% - 3.5%) / obtained with crystal calorimeters have not been reached by sampling devices. Because of their energy resolution and compactness crystal calorimeters have been widely used in collider experiments: L3, CUSB, CLEO 11, KTEV, GLAST, BABAR, BELLE and CMS. One of the notable by-products is the use of CsI (Tl) crystals in medical tomography.

9. Cryogenic Calorimeters In 1935 F. Simon32 suggested measuring the energy deposited by radioactivity with low temperature calorimeters. He claimed that, with a calorimeter consisting of lcm3 tungsten in a liquid helium bath at 1.3 K, one could measure lo-’ cal/sec, which is about 1000 times more sensitive than the calorimeter of W. Orthmann. The argument is that at low temperatures the heat capacity C of a calorimeter is low and a small energy loss A E of a particle in the calorimeter can lead to an appreciable temperature increase A T = AE/C. More recently, the development of cryogenic calorimeters was motivated by the quest for the dark matter in the universe, the missing neutrinos from the sun and the neutrinoless double beta decay. First ideas and experimental attempts were discussed at a first workshop on low temperature detectors (LTD1)33, which was held in 1987 at Ringberg Castle in southern Bavaria. More workshops followed every other year, either in Europe or in the USA. Most calorimeters used in high energy physics measure the energy loss of a particle in form of scintillation light or ionization. In contrast, cryogenic calorimeters are able to measure the total deposited energy in form of ionization and heat. This feature makes them very effective in detecting very small energy deposits (order of eV) with very high accuracy. The use of superconductors as cryogenic particle detectors was motivated by the small binding energy lmeV of the Cooper pairs. Thus, compared to conventional detectors, several orders of magnitude more free charges are produced, leading to a much higher intrinsic energy resolution. The main advantage of cryogenic devices for the direct detection of dark matter particles, so-called WIMPS (weakly interacting massive particles), is

-

12

their effectiveness in measuring very low energy recoils and the possiblity of using a large variety of detector materials. WIMPS can be detected by measuring the nuclear recoil energies in coherent elastic WIMP-nucleus scattering. Depending on the mass of the WIMP and the mass of the detector nucleus, the average recoil varies between eV and keV. The most commonly used WIMP detectors are bolometers. E. Fiorini and T. N i i n i k ~ s k ipioneered ~~ the development of bolometers for measuring neutrinoless double beta decay. But also superheated superconducting granules (SSG) for dark matter detection have been developed. This technique was first introduced by H. Bernas et a135for beta radiation detection and later proposed for dark matter and solar neutrino detection by L. Stodolsky and A. D r ~ k i e r The ~ ~ . first generation WIMP experiments with cryogenic bolometers with absorber masses of about 1 kg (CDMS, EDELWEISS, CRESST and ROSEBUD) and with SSG of 0.5 kg (ORPHEUS) are in operation. Larger detector masses of more than 10 kg are planned for the future. One of the big advantages of cryogenic calorimeters over conventional WIMP detectors is their capability of active background recognition, which allows to discriminate between background recoils due t o electron scattering and genuine nuclear recoils by a simultaneous but separate measurement of phonons and ionization in each event. This is possible since, for a given deposited energy, the ionization generated by nuclear recoils is smaller than that generated by electrons. This feature increases the sensitivity for WIMP detection considerably. Mainly for applications other than WIMP detection, calorimeters on the basis of superconducting tunnel junctions (STJ) have also been developed. Cryogenic cameras37 consisting of STJ pixel arrays provide the astronomers with a powerful tool to observe very faint and distant objects and to determine their distance via red shift. Other research areas have also benefited from these cryogenic detector developments, such as x-ray spectroscopy in astrophysics, mass spectrometry of large molecules (DNA sequencing) as well as x-ray microanalysis for industrial applications. 10. Detection of Extraterrestrial Neutrinos

In 1968 R. Davis38 pioneered the detection of extraterrestrial neutrinos and started a new field of neutrino astronomy. In his chlorine detector buried in the Homestake mine in South Dakota he was the first to detect neutrinos from the sun. The neutrino flux, however, turned out to be lower than expected from the standard solar model. His findings were later confirmed by many experiments, such as BAKSAN, GALLEX, KAMIOKANDE, SUPERKAMIOKANDE. They remained a puzzle until very recently, when the SNO experiment confirmed a long-presumed hypothesis, namely that neutri-

13

nos have a mass and, as a result of this, they change flavour on their way from the sun to the earth. Atmospheric neutrino oscillations have previously been found by the SUPERKAMIOKANDE water Cerenkov detector. In order to look for extraterrestrial neutrinos, K. Lande39 and collaborators installed a water Cerenkov detector in the Homestake mine adjacent to the R. Davies experiment. The detector consisted of 7 water tanks, each viewed by 4 PMs on opposite sides of the tank. On January 4, 1974 the detector signalized an event which the experimenters interpreted as a possible antineutrino burst. Unfortunately, such events do not happen very often and a confirmation is practically impossible unless they are also witnessed by other experiments. If for nothing else, these findings wakened the curiosity of researchers to look for more of these events with larger and better detectors. Thus it was around this time that A. Roberts from Fermilab, V. Peterson from the University of Hawaii, R. March of Madison and others started to think about a deep sea water neutrino telescope near the coast of Hawaii. Their ideas later became a project called DUMAND, which was unfortunately discontinued. But it paved the way for the next generation of cosmic neutrino Cerenkov detectors: NESTOR and ANTARES in the Mediterranean sea as well as AMANDA and ICECUBE in the Antarctic ice. Massive calorimeters like NUSSEX in the Mont Blanc tunnel, BAKSAN in Russia, IMB in the Morton-Thiokol salt mine and KAMIOKANDE in Japan, originally designed for proton decay experiments, were, on 17th February 1987, witness to an extraterrestrial neutrino burst, which originated from a nearby supernova explosion (SN1987A). This was the first time neutrinos from a supernova explosion were detected with terrestrial calorimeters. This happy event encouraged all those who had already made plans to set new trends in neutrino astronomy. An interesting new method has been presented by D. Saltzberg at this conference. It is based on the detection of the coherent emission of Cerenkov radiation in radio- and microwaves. Large volumes of natural materials transparent to radio waves may be employed in future for the detection of ultra high energy (UHE) astrophysical neutrinos. 11. The Atmosphere as Calorimeter At sea level our atmosphere measures 1032 g/cm2. Very high energy particles from outer space trying to traverse our atmosphere would find material in front of them which corresponds to 28 radiation lengths and 16.6 collision lengths. Particles from cosmic origin like hadrons, photons or neutrinos interact with air nuclei, producing secondaries that in turn collide with air atoms, leading to extensive air showers (EAS). The most numerous particles in EAS are elec-

14

trons. Electrons traversing the atmosphere produce Cerenkov light, which is directed along their path. On their way, they can also excite metastable energy levels in atmospheric molecules which, after a short relaxation time, emit a characteristic fluorescence light, which peaks at wavelenghts from 330 to 450 nm. In contrast to the Cerenkov light the emitted fluorescent light is isotropic. The emitted Cerenkov and fluorescence light are proportional t o the EAS energy. Thus properly instrumented, the atmosphere can be utilized as a huge calorimeter. It was P. Blackett4’ who, in 1948, suggested to detect the Cerenkov light in the atmosphere caused by penetrating cosmic particles. A. Chudakov and collaborators4’ were the first to apply this idea in 1962 to detect celestial 7rays. Atmospheric Cerenkov light detection was also pioneered by the Whipple telescope on Mount Hopkin, Arizona, in 1987. It consisted of a l0m-diameter mirror dish and a pixel array of photon detectors42. This technique provided information on the direction, shape and energy of the shower as well as on the type of primary particle (cosmic hadrons produce 2 times less Cerenkov light than gammas of the same energy). The Whipple telescope was one of the most powerful early instruments which made major contributions to the study of high energy gamma rays in the energy range of several hundred GeV to several TeV. However, this technique was limited to a relatively small fiducial area yielding low event rates for high energy showers as well as too poor energy resolutions. In a further step the Fly’s Eye detector43 was built to overcome these deficiencies. The Fly’s Eye detector consists of an array of spherical mirrors with a cluster of PMTs mounted in the focal plane of each mirror. It records the fluorescence light, which is caused by ultra high energy cosmic ray showers in the atmosphere. The Fly’s Eye observatory in Utah consists of two detector stations (Fly’s Eye I and 11) situated 3.3 km distant from each other. Fly’s Eye I is equipped with 67 detector arrays and Fly’s Eye I1 with 8 detector units with 120 eyes. Events more than 20 km away from the observatory could be detected, giving rise to a very large fiducial area of about 100 km2 sr. The simultaneous observation from both Fly’s Eye stations allows a stereoscopic reconstruction of an event. From the observed shower profiles the total energy can be derived. However, the knowledge of the energy scale is still a central issue for atmospheric calorimeters. The original Fly’s Eye detector has now been replaced by a new facility called HIRES, which is sensitive in the energy domain of the Greisen-Zatsepin-Kuzmin (GZK) cut-off. Presently under construction is the P. Auger observatory in Malague in Argentina. It will consist of 1600 water Cerenkov detectors and 4 fluorescence ”eyes” spread over an area of 3000 km2. The P. Auger observatory will measure the energy

15

and arrival direction of UHE cosmic rays with energies in excess of 10’’ eV. For the future, an exploratory mission probing the extremes of the universe using the highest energy cosmic rays and neutrinos is planned. It is called EUSO/OWL (EUSO for Extreme Universe Space Observatory) and will orbit the earth in 500 km height and observe an area of about 3 . 1O5km2sr of the earth’s atmosphere, being able to detect several thousand air showers above lo2’ eV per year (see S. Swordy as well as K. Arisaka in these proceedings). With this sensitivity the experimenters aim to systematically study the energy spectrum around the GZK cut-off. They also hope to be able to detect relic Big Bang neutrinos through the 20-resonance absorption of cosmic neutrinos with energies > 1021 eV. 12. Conclusions The development of novel detectors is essential for the exploration of new domains in physics. Calorimeters are a very good example for this. Major discoveries, like neutral currents (by GARGAMELLE), quark and gluon jets (by UA2, UA1 and PETRA), W,Z bosons (by UA1 and UA2), top quark (by CDF and DO), neutrinos from the supernova explosion SN1987A (by NUSSEX, IMB, KAMIOKANDE and BAKSAN), atmospheric neutrino oscillations (by SUPERKAMIOKANDE) and solar neutrino oscillations (by SNO) were made with detectors employing calorimeters. The future will provide enough challenges for young people with imagination.

Acknowledgments I am very grateful to P. Grieder, U. Moser, P. Jenni, Ch. Fabjan, E. Lorenz and R. Wigmans for useful discussions as well as for providing me with material for this presentation. I would also like to thank R.Y. Zhu and his collaborators for the perfect organization of this conference and the pleasant atmosphere. References 1. W. Orthmann, Zeitschrift f. Physik Bd 6 0 , 10 (1930). 2. L. Meitner and W. Orthmann, Zeitschrift f. Physik Bd 60, 143 (1930). 3. N.L. Grigorov et al, Zh. Eksp. Teor. Fiz. 34, 506 (1954). 4. V.S. Murzin, Prog. Elemt. Part. Cosmic Ray Phys. 9, 247 (1967). 5. C. Heusch and C. Prescott, IEEE NS12, 213 (1965). 6. J. Engler et al, NIM 106, 189 (1973). 7. E.B. Hughes et al, IEEE NS17, 14 (1970) and NS19, 126 (1972). 8. B. Barish et al, IEEE NS25/1, 532 (1978). 9. W.A. Shurcliff, J. Optical SOC.Vol 41/3, 209 (1951). 10. R.L. Garwin, Rev. Sci. In&. 31, 1010 (1960).

16

G. Keil, NIM 83, 145 (1970). W. Selove et al, NIM 161, 233 (1979). W. Hofmann et al, NIM 195, 475 (1982). V. Eckardt et al, NIM 155, 389 (1978). C. De Marzo et al, NIM 217, 405 (1983). H. Fessler et al, N I M A 2 4 0 , 284 (1985). D. Acosta et al, NIM A294, 193 (1990) and NIM A308, 481 (1991). P. Gorodetzky, Rad. Phys. and Chem. 41, 253 (1993) and P. Gorodetzky et all NIM A361, 161 (1995). 19. W. Willis and V. Radeka, NIM 120, 221 (1974). 20. J. Engler et al, NIM 120, 157 (1974). 21. J. Engler and H. Keim, NIM 223, 47 (1984) and 3. Engler et al, N I M A 2 5 2 , 29 (1986). 22. B. Aubert et al, R D 3 coll. CERN/DRDC/90-31 and NIM A309, 438 (1991). 23. C. Fabjan et al, NIM 141, 61 (1977). 24. H. Gordon et al, NIM 196, 303 (1982). 25. H. Abramovicz et al, NIM 180, 429 (1981). 26. W. Braunschweig et al, NIM A265, 419 (1988). 27. R. Wigmans, NIM A259, 389 (1987). 28. H. Brueckmann et al, NIM A263, 136 (1988). 29. P.G. Rancoita and A. Seidman, NIM 266, 369 (1984). 30. G. Barbiellini et al, NIM A235, 55 (1985) and E. Borchi et al, CERN-EP/89-28 (1989). 31. M. Oreglia et al, Phys. Rev. D25, 2295 (1982). 32. F. Simon, Nature 135, 763 (1935). 33. Low Temperature Detectors for Neutrinos and Dark Matter, ed. K. Pretzl, N. Schmitz, L. Stodolsky, Springer, Berlin (1987); see also K. Pretzl, Proceedings of "Calorimetry in High Energy Physics" World Scientific, 32 (Lisbon 1999). 34. E. Fiorin i and T.O. Niinikoski, NIM 224, 83 (1984). 35. H. Bernas et al, Phys. Lett. A24, 721 (1967). 36. A. Drukier, L. Stodolsky, Phys. Rev. D30/11, 2295 (1984). 37. T. Peacock et al, Astron. Astrophys. Suppl. Ser. 127, 497 (1998). 38. R. Davies et al, Phys. Rev. Lett. 20, 1205 (1968). 39. K. Lande et al, Nature 251, (Oct.11.1974). 40. P. Blackett, Rep. Gassiot Committee 34, UK (1948). 41. A. Chudakov et al, J. Phys. Soc. Japan 17/AIII, 106 (1962). 42. M. Cawley et al, Experimental Astronomy 1, 173 (1990). 43. R.M. Baltrusaitis et al, NIM A240, 410-428 (1985). 11. 12. 13. 14. 15. 16. 17. 18.

OVERVIEW AND STATUS OF CALORIMETRY AT LHC

D. FOURNIER Laboratoire de 1’Acce‘le‘rateurLine‘aire - Centre Scientzfique d ’Orsay - B.P.34 - 91898 ORSAY CEDEX (FRANCE) E-mail: fournierOlal.in2p3.fr

1. Introduction Calorimeters play a central role in “general purpose detectors’’ for high energy proton-proton colliders, like ATLAS and CMS at the LHC: they allow to trigger, to identify and to measure electrons, photons and jets, as well as escaping neutrinos or other non-interacting particles appearing as missing transverse energy. They also complement muon detection both at the trigger level, and in providing an estimate of the energy lost in traversing them. Indeed calorimeters are only one part of the experimental set-up, and the overall optimisation of the experiment, i.e. reaching optimal combined performance with the tracker, the magnets and the muon systems was one of the most challenging tasks of the designers. Today, both ATLAS and CMS ca!orimeters are well in the construction phase. This is a time when one realises that some choices were indeed wise, when the detector will meet or sometimes exceed the specifications, without having caused too much trouble in the engineering phase. But there are also cases for which compromises had to be made, in order to cope with technical difficulties, budgetary constraints, and schedule. Taking in turn, electromagnetic, hadronic and forward calorimetry, the talk addresses, for ATLAS and CMS: - design evolutions since the Technical Design Reports (TDR)

- update of performances (test beam results, simulations) - systems aspects

- status of construction At the end, the particular cases of LHCb and ALICE are briefly considered.

17

18

2. EM Calorimetry

2.1. ATLAS EM Calorimetry 2.1.1. Main features of the design The ATLAS EM calorimeter uses the Liquid Argon technique, for its excellent stability and radiation resistance, with an “accordion geometry” to allow for hermeticity, speed and high granularity’. The barrel part which covers up to r] = 1.4 is separated in two cylindrical “half-barrels”, housed in a single cryostat which shares its isolation vacuum with the solenoid (figure 1). Each half-barrel is made of 16 modules and is preceded by a thin presampler layer. In the end-caps (figure 2), each EM wheel (made of 8 modules and covering 1.4 < r] < 3.2) is preceded by a thin presampler for 1.5 < r] < 1.8. The design energy resolution is l O % / a , with a constant term of 0.7%. The granularity and noise. (which was measured in beam tests with ATLAS-like electronics) are as in table 1. Thanks to the separation in 3 layers in depth, the front strips and central cells allow to measure the direction of photons, independently of the knowledge of the interaction vertex, with an accuracy of 50 m r a d l a . The front-end electronics scheme, unchanged since the TDR, features 3 gains (1, 10, 100) shaping amplifiers, analog buffering in SCAs (switch capacitor arrays) every 25 ns, and digitization of a given number of samples (typically 5) upon LVLl request.

WADOWMAR0

COID-lU-MRU CARIR P A 1 3 PAPNEL

’-

\

\\~I‘AHY

\

PRESAYPlbU

VWCL

COIJ VkXSEL

Figure 1. Sketch of ATLAS E M barrel liquid argon calorimeter (only one half is shown).

In this scheme (figure 3), the largest non-saturating gain is first chosen by digitising, in medium gain, the sample closest to the expected signal maximum. Then all samples are digitised with the same gain, allowing then to use digital filtering for combined optimisation of noise and pile-up as a function of

19 Table 1. Granularity and noise of ATLAS EM calorimeter.

Figure 2.

Sketch of ATLAS end-cap liquid argon calorimeters (one side).

luminosity.

Figure 3.

Simplified readout diagram of liquid argon calorimeters.

20

2.1.2. Difficulties with engineering and fabrication

Essentially three difficulties were met: 0 Fabrication of large electrodes: In the early prototypes3, 3-layer, flexible, copper-polyimide electrodes were built with the standard laminates and processing equipment, limited to 60 cm width. Sticking to this approach would have resulted in 7 electrodes to match one absorber, between q = 0 and q = 1.4, with the drawback of several thin cracks. It was thus decided, in 1996, to move to l m width laminates and processing equipment, in order to cover the same area with two electrodes only. It took several years, in collaboration with firms, to reach a satisfactory quality and rate of production. The tuning of non-standard equipment was made more acute due t o the presence, on each electrode, of almost 1000 serigraphied resistive pads, necessary t o distribute high voltage. The resistances proved rather fragile, in particular when bending electrodes to the desired accordion shape (figure 4). At the time of the conference, more than 75% of the 6000 electrodes of the calorimeter were available.

Figure 4.

View of one central barrel electrode (7= 0.0 - 0.8).

HV trouble shooting: Each electrode is kept in its nominal position, in the middle of the gap between two absorbers, using honeycomb bands linked by a set of wires in a sort of net matching the accordion waves (figure 5). The total area of the material used is about 20000m2. The nominal high voltage of lkV/mm in liquid argon is also used as test voltage in air. In the early phase of the stacking, done in clean rooms with controlled temperature and humidity, unexpected spurious sparkings, and even some shorts were experienced. A suitable mode of operation could be found only after that a thorough cleaning of honeycomb nets, followed by a specific HV training of each of them was installed. 0 Rad hard electronics: The front-end electronics resides in front-end crates, at the high q end of each cryostat, and rather large radius (- 2.5m), where the level of radiations is reduced n/cm2 and 30Gy each high luminosity

21

year) and accessibility is possible. All electronics on these boards was revisited in the last two years such that only radiation-hard processes be used (mostly DMILL). The SCA controller, originally planned in DMILL appeared to have a too low yield. It was redesigned in a 0.25 micron process and is now becoming available.

Figure 5. Close view of the transition between large and small wheel in an E M end module, showing the honeycomb nets.

A set of 30 boards to fill a full front-end crate will be available as of fall 2002, and submitted to detailed tests before series production of boards starts. Most custom designed chips (preamplifiers, shaping amplifiers, SCAs) are already mass produced and being subjected to acceptance tests. 2.1.3. Test beam results After thorough tests of “module 0s” in year 1999 and 20004, which revealed a few problematic points, later corrected (see below), two series modules of both the barrel and the end-caps were tested in beam in summer 2001 (see dedicated talks to this conference). The main problem encountered was to understand in detail the behaviour of fast signals (peaking time of 50 ns), in complex structures: from the electrode cell sensing the current signal in the module, to the preamplifier on the front-end board, are: the traces routing the signal (and ground !) out on the

22

electrode, the summing board, the mother board, cold cables, feedthrough and crate baseplane. Inductances and impedance discontinuities are affecting the physics signal (and the calibration signal injected at the mother board level) propagation, making the electronics calibration to the anticipated level (channel to channel dispersion of 0.25% rms) a more difficult task than anticipated. Problems were best seen from the scan of a module on constant q or constant q5 lines: the energy response should be flat, and was not. The results for series modules are now quite good, but one had to:

- improve the ground return on the electrodes, - re-route the traces on mother boards and summing boards,

- use rather complex formalism to analyse the pulse shapes. Combining in the same histogram (figure 6 ) all electrons from the scan of a production module 600 middle cells) with 245 GeV electrons gives a total resolution of 1.14% which splits more or less equally between the sampling term (0.7%) and the constant term (0.9%). Some corrections, like for the transition between electrodes at 7) = 0.8, or correcting some sick electronics channels, are still to be done. (W

2.2. CMS EM calorimetry 2.2.1. Main features of the design and evolution since TDR In CMS a PbW04 crystal calorimeter5 was chosen for its excellent sampling term of the energy resolution, and for its compactness (the radiation length of PbW04 is 0.9 cm). R&D concentrated on purification and controlled doping of the base material, and crystal growth conditions, in order to avoid as much as possible radiation dependent effects. The rather low light yield called for active light converters (APDs, VPTs) which also needed a specific development. The CMS EM calorimeter consists of two half-barrels, each made of 18 identical modules, and two end-caps each made of two Dees (figure 7). An evolution from the initial proposal, is that now only the endcaps are preceded by a Lead-Silicon preshower . The sampling term of the energy resolution is thus about 3 % / a for the barrel and 5 . 5 % / a for the end-caps. 2.2.2. Readout and noise The scintillation light from the crystals which has a broad spectrum around 420 nm, is converted -in the barrel- to an electrical signal by two APDs (25

23 iss

Figure 6 . Combined spectrum of 245 GeV electrons taken over not yet adjusted).

Figure 7.

N

600 cells (absolute scale

Geometry of the CMS crystal calorimeter.

mm2 each) glued on the back face of each crystal. In the end-caps, the solenoid magnetic field, almost parallel to the crystal direction, allows the use of 3-stages

24 Table 2. Granularity and channel count of the CMS crystal calorimeter.

I

A17XW

Barrel I 0 . 0 1 7 5 ~0.0175 17 < 1.48 End-cap Variable 1.48 < 17 < 3.0 End-cap preshower

I

Cell size (mm)

I

I

213x213

I

I

I

Depth (Xo) 25.8

I

Number of

I

61200

I channels

29.6 x 29.6

23

15632

63 x 1.9

3

N

I

130000

photo-multipliers (VPTs), less sensitive to radiations than APDs. A tight Quality Control is applied t o the production (see dedicated presentations to this conference). One of the criteria is a combined crystal-APD efficiency of more than 6 photo-electrons/MeV, measured with a radioactive source (Co60). The electronics noise depends both on this efficiency, on the APDs ”excess noise factor” ( w 2.2), and on the preamplifier characteristics. The current estimate is around 35 MeV/crystal (this figure is expected to approximately double after 10 years of high luminosity running, because of the shot noise associated t o the APDs’ leakage current). In the end-caps, read out by Vacuum Photo-Triodes, it is about 50 MeV. The preamplifier (figure 8) is followed by a 4 gain shaping amplifier (single RC-CR filter with a 43 ns time constant, with gains in the ratio 1, 5, 9, 33). The sample and hold system selects at 40 MHz the largest non saturating gain for digitisation in the fly. This means that the 5 or 7 samples used for a given pulse are, for large signals, taken with different gains. This system is called ”floating point preamplifier” since it provides to the downstream stage (ADC) a sampled analogue level, and two bits for the gain. Evaluations are still ongoing in the collaboration to decide if all ADC samples are continuously transmitted at 40 MHz (one 1.3 GHz fibre per crystal) for treatment outside of the detector, or if LVLl primitives (5 x 5 crystals) are formed locally, allowing to output only events (stored locally in a memory) selected by LVL1. The understanding after dedicated talks at the Conference was that the second way is more likely to be chosen, in particular because of financial constraints. 2.2.3. Short term follow-up of light output Exposed to low radiation levels (one Gray per hour or less), like in standard LHC conditions, PbW04 crystals show a small drop of light output (mostly due to absorption by colour centres), which saturates after a few hours of exposure, and is followed, when irradiation stops, by a recovery with a time constant of

'25

Figure 8. Simplified diagram of the CMS calorimeter front-end readout.

.;!

:.!irradiation ., i i

;

,

.j

i

j

.

:

111

B i

I,

I

Monitoring light Figure 9.

111

Ill

(11

11,

Ill

W,.,l,,,lh

Ill

IJI

(,I

Geometrical distribution of light, and frequency spectrum.

typically hours as well (figure 10). In order to limit the consequences of this dependence, the following actions were taken: Production crystals are requested to vary by less than 6% under exposures like above. In order to follow the crystal behaviour at the required pace (hours) during LHC data taking, an optical switching system will send in sequence laser light pulses (at 440 nm) to all crystals of each module. Because the laser pulses do not have a spectrum identical to the scintilla-

26 c

1.01

$

1 0.99

I? Green Cusor r?mitor;ny 0 Blue laser rnonitA 1 2 0 GeV skctronr

0.98 0.97

0.90

0.95 0.94 0.93 0.92

0

1

2

3

4

6

7

time Idayl

Figure 10. Simulated behaviour of crystal response to electrons and light pulses (green, blue) as a function of time.

tion light, nor the same geometrical distribution in the crystal, the correction cannot be perfect. It was estimated, and confirmed by exposing pre-production modules to test beam6, that the constant term associated to the residuals of the laser monitoring will not be larger than 0.4%.

2.3. System aspects While the behaviour of the basic ”bricks” of both ATLAS and CMS electromagnetic calorimeters is now well understood, the performance of EM calorimeters as “systems” still critically depends on many integration issues. A selection of the most relevant ones is considered below. 2.3.1. Effect of tracker material and magnetic field The strong requirements (granularity, speed) on tracking systems has lead to central tracking detectors which are unfortunately rather massive, as illustrated by the CMS case in figure 2.3.17 (ATLAS is similar)s: Electrons and photons are thus subject to early showering in the tracker (plus solenoid and cryostat in the case of ATLAS) before reaching the active calorimeter (or presampler) medium. The low energy calorimeter tail subsequently generated is particularly visible in CMS due to its better energy resolution, and larger magnetic field (4T for CMS, 2T for ATLAS). Figure 2.3.1 shows the energy spectrum of electrons after that an isolation criterium is applied at the trigger level (adding a 6% inefficiency), and reconstructing the shower with an ”hybrid cluster” (i.e. extended to sub-clusters in the azimuthal direction). In the same conditions, for 35 GeV ET, the position resolution, converted to angular resolution assuming a fixed vertex is about 1.0 mrad in 17 and 2.1 mrad

27

Figure 11. (a): Material in the CMS tracker; (b)Simulated electron energy spectrum (normalised to true energy) in the CMS detector.

in 4. For comparison, with the middle sampling of ATLAS the corresponding figures are 1.0 and 1.5 mrad respectively. Once more the worse resolution in azimuth is to be associated with the effect of bremsstrahlung . Photons which do not interact before a radius of -90 cm are much less affected than electrons. In CMS, for a 110 GeV Higgs decaying into yy, 78% of the reconstructed masses without conversion fall in a 1.9 GeV mass bin. However, the efficiency of the ”non-converting” photon cut alone is about 70% per leg. In ATLAS using both interacting and non-interacting photons the 80% acceptance mass bin is 3.1 GeV wide. The comparative performances of the two experiments for H + yy detection depend on the resolution that CMS will reach with the converted photons, and on several other aspects (background rejection, pointing - see below). 2.3.2. Calibration in situ

In ATLAS it is assumed (and now confirmed by beam test of series modules) that the detector is uniform ”by construction” to better than 0.4% rms in areas of Aq x Aq5 = 0.2 x 0.4 or larger. Taking this as “minimal hypothesis”, there would be 440 areas (0.2x0.4) to be intercalibrated in situ. It was shown by simulationsg that imposing the Z mass constraint to 2 + e f e - decays measured in the calorimeter only (i.e. without reference to tracking information) an adequate inter-calibration (0.3% rms cell to cell dispersion) can be obtained with data recorded during 48 hours at low luminosity (where the rate of 2 -+ e+e- is about 1 Hz). The same constraint gives indeed an extremely precise absolute energy scale calibration. In CMS the crystals are first calibrated in the laboratory with laser pulses

28

and/or radioactive sources. Experience shows that the correlation of this calibration with the response to high energy electrons from beam tests is not better than 6% rms. In the likely c a e that all crystals cannot be calibrated in test beam, the in-situ inter-calibration will use lab results as starting point. The currently developed strategy is based on using electrons from W decays, measured with the tracker. This is better adapted than the Z mass constraint, due to the higher rate. Preliminary studies" indicate that two months of low luminosity data are necessary to intercalibrate to 0.5% rms in this way. When this is done, Z decays give the absolute energy scale. Considerations on azimuthal uniformity of response to minimum bias events may help in this inter-calibration task.

2.3.3. Constant term Besides the residual of the crystals or cells inter-calibration, and of short term variations, like the crystal output light depending on recent irradiation history, there are several contributions to the constant term. One of them is associated with temperature dependence of signal response: the temperature dependence of the liquid argon signals was measured t o be -2%/OK (see ref 1, chapII, p.33). The liquid argon bath is subject t o free convection, which can be turbulent in some places despite the small gradients (the viscosity of argon is very low), and is therefore difficult to simulate. The total heat influx is about 2.5 kW per cryostat. Extensive finite elements simulations (preliminary) indicate an overall temperature dispersion within the barrel sensitive volume of f 0.15 degrees, hopefully small enough to avoid the need for corrections. In case of larger gradients, corrections will be possible offline, thanks to continuous recording of -300 precision temperature probes located at the surface of modules. 0 In CMS the temperature dependence of "crystal+APD" is about -4.3%/OK. Removing the heat dissipated by front-end electronics (2 W per channel, i.e. 160 kW in total) right in the back of the crystals, while keeping an excellent temperature uniformity, is a challenge. A thermal analysis of the crystal calorimeter was made, based on the use of pressurised water circulating in a dense pipe network, separated in three layers, each layer being isolated from the more external one by thermal shields. As a result, it is estimated that a temperature dispersion of f 0.05'K within the crystal volume can be reached. Overall, ATLAS claims a constant term in the energy resolution of 0.7%, and CMS of 0.55%. At this stage it might be useful to stress that those figures will only be reached if processing key physics channels (W+ ev, Z+ ee) is easy, with a fast turn-over, as of LHC start-up.

29

2.3.4. Use of granularity Background rejection: In the search for the Higgs in two photons (up to 150 GeV) it is necessary to bring the background from jet-jet and y-jet events well below the irreducible yy background. After standard shape cuts the remaining background is dominated by jets fragmenting to single r0s. Using the high granularity of the calorimeters the performances reported in table 3 were obtained by simulation, at 50 GeV ET. Somewhat inferior in the barrel, CMS is equivalent or slightly better in the end-caps,thanks to the high granularity of its lead-silicon preshower. Table 3.

rejection in ATLAS and

ATLAS(al1 y) CMS (unconverted)

0-0.45 3.6 3

0.45-1. 3.1 2.2

1.-1.5 2.6 2

-1.7 3.1 3.5

-2.4 2.8 2.5

Pointing: The measurement of the direction of non-converted photons is a unique feature of ATLAS (indeed measuring the direction of converted photons is easy in both detectors). Such direction measurements are important in several cases, in particular for Higgs decaying in two photons, where the direction error contributes t o the mass resolution (at high luminosity, the interaction vertex is often ambiguous for yy events). Gauge Mediated Susy Breaking models, in which the neutralino photon-gravitino i decay may have a long enough life time to produce “non-pointing photons”, which can be used t o sign the process, is another example where pointing is useful.

2.3.5. Linearity Excellent linearity is required for precision physics. One topic particularly demanding is an improved measurement of the W mass using the W -+ ev channel. Illustrated below is the linearity observed in one of the best calorimeters built so far, namely the homogeneous krypton Calorimeter of the NA48 CP Violation experiment at CERNll (figure 12). Over the energy range of interest for this experiment, the systematic uncertainty was estimated to be 6 rms. What would be needed to meet the W mass measurement requirement (25 MeV error, overall) is gaining a factor of 3, however on a restricted range (Mz/2 to Mw/2). This would probably be possible with the NA48 calorimeter, ... but ATLAS and CMS calorimeters are more complex devices, and not much was done so far to assess them, in this respect, at the required level.

30

1.002

1

.

(E+45MeVf/pjrom ,K

t.mi 1

0.969

0.ws 0.997

0.96'6

Energy (GeV) Figure 12. Ratio of electron energy, measured with the krypton calorimeter, and momentum from the magnetic spectrometer, in NA48.

2.3.6. Data reduction Large number of cells, large dynamic range, and high trigger rates produce very large amount of data (-1.0 Mbyte/event for the calorimeter itself, in both experiments). The associated cost for storing and processing these data was recently put under scrutiny12, which triggered some further dedicated work in the collaborations. One route for data reduction being explored by both ATLAS and CMS, is "zero suppression". While at first sight cutting cells containing "only noise" is rather tempting, this may generate subtle adverse effects on precision physics, difficult to resolve later on. Detailed Monte Carlo evaluations (and control samples) will be needed before such a data reduction can be implemented at the "on line" level, but prospects are interesting13. 3. Hadronic Calorimetry

All the electromagnetic component of jets, plus a fraction of the charged hadronic part, is deposited in EM calorimeters representing 1 to 1.5 interaction length A. EM calorimeters being calibrated with electrons, the energy measured for jets has to be converted to "hadronic scale" when, as it is the

31

case for ATLAS and CMS the electron to hadron ratio (e/h) is larger than 1. To catch the remaining part of jets, more massive devices are needed, the “hadronic calorimeters”, up to a total thickness of 9X or more. Full coverage in pseudo rapidity is mandatory, up to q N 5. The instrumentation of the most forward region (3 < q < 5) raised specific problems. This part is considered in the next section. N

3.1. CMS hadronic Calorimeter CMS uses scintillator sampling calorimetry both in the barrel and in the endcaps (up to q = 3.0)14. The choice to place the full calorimeter inside the coil imposed a non magnetic absorber (brass, in plates of 5 cm). The constraint of the coil size (inner radius 2.95 m, outer radius 3.80 m) in practice limited the total thickness at q =O to 7 A. In order to improve on this figure, it was found necessary to add a ”tail catcher” behind the solenoid coil, for a total of 9.4 A, compromising to some extend-in a limited pseudo rapidity range-the original design criterium. Scintillator tiles (3.7 mm thick) inserted in between the brass plates, are readout by wave length shifting fibres, fitted in grooves. The former are glued to clear fibres which bring light out of the solenoid, up t o Hybrid-PhotoDetectors (proximity focused single stage photomultipliers with ” pixelised” silicon diode target) able to work in the fringe magnetic field. The light yield is about 10 pe/GeV. The granularity is 0.087 x 0.087 (5 x 5 EM towers) with 3 samplings in depth, the first one being rather thin in order t o sample the particles coming out from the ECAL. Close to the upper ( q = 3) boundary, the cumulated radiations (3 lo4 Gy integrated over 10 years) will start to affect the collected light (mostly because of a reduction of the WLS fibre absorption length). This will be monitored by a moving source, and using LHC data itself. 3.2. ATLAS Tile Calorimeter ATLAS uses scintillator sampling calorimetry in the barrel and in the extended barrel, up to q = 1.715. The iron plates (5mm thick, grouped by 3) and scintillator plates (3mm thick), are perpendicular to the beam axis. The thickness up to the last active layer is 9 X at q=O. The scintillation light is collected by WLS fibres running on either side of scintillator plates. This allows a rather easy bundling of fibres to form towers, read out by photomultipliers located in the “girders” (outer radius 4.23 m). At this place the solenoid and toroid fringe fields are low, and further reduced by shielding with high permeability metal sheets.The light yield is about 40 pe/GeV. The granularity is 0.1 x 0.1 (4 x 4 EM towers) with 3 samplings in

32

depth (the last segment has a granularity of 0 . 2 ~ 0.1). Towers are pointing in azimuth, and pseudo pointing in 77. 3.3. A T L A S Hadronic End-Cap

Beyond 77 = 1.5, ATLAS uses the Copper-Liquid Argon technique’. Among other aspects this alleviates radiation resistance problems. The detector consists of two wheels-HEC1 and HEC2- in the same cryostat as the Electromagnetic end-cap and Forward calorimeters (see figure 2). The copper plates (25mm thick in HEC1, 50 mm in HEC2) are perpendicular to the beam axis and interleaved with 4-uple liquid argon gaps in a configuration of electrostatic transformer. Pairs of adjacent cells in depth are connected to a preamplifier located in the liquid, at the periphery of the wheels. Summing of signals to form readout towers is done downstream of the preamplifiers, in the cold as well. This scheme gives adequate speed of response, with signal rise-time of N 50ns. The granularity is 0.1 x 0.1 (4 x 4 EM towers) up to 77 = 2.5 and 0 . 2 ~0.2 up to 77 = 3.2, with 4 samplings in depth (at the time of the TDR, the two segments of HEC2 were ganged together).

3.4. Resolution, linearity Prototypes of the hadronic calorimeters briefly described above have been tested in high energy pion beams, both in ”standalone” mode, and with their EM counterpart in front. In all cases the electron to hadron response ratio (e/h)is greater than 1, both in the electromagnetic and hadronic compartments, which affects both the linearity and the constant term. The example of the CMS calorimeter16 is summarised in figure 13. With a suitable weighting, depending on signal height on an event by event basis, the combination crystal calorimeter-hadronic calorimeter can be made nearly as good as the hadronic part in stand alone. Results from the ATLAS combined LAr-Tile prototypes17 show a similar behaviour, as given in figure 14. Cell by cell weighting (a la H1) improves by 15% the resolution of the “benchmark” method, with a single weight per compartment. Performances have also been extrapolated to jets at LHC. A larger cone size improves the sampling and constant terms of the energy resolution, but the fluctuations of the noise and pile-up soon become dominant. Cone sizes down to 0.4 are used for the high luminosity case. Some relevant parameters are summarised in table 4 (the pile-up is the expected value a t nominal high luminosity in a cone AR=0.7). Attempts are being made in CMS to improve the jet resolution by using

-

33

s

0 Combined 96 (benchmarks)

0.3

*

n interocting in ECAl or HCAL

0

Combined 96 (HI weighting)

no weighting passive weighting

0.2

dynamic weighting

0.1

n U I "

0

'

" " '

SO

IW

'

, ' . I , . ' . .

IS0

'

"

,'

"

1

2W ZSO JW IS0 JCU pion beam momentum (GeV)

Figure 13. Pion resolution with prototypes of the CMS calorimeter.

0

0. I

0.2

0.3

11.1 Ekiu,, (GeV-") Figure 14. Pion resolution with prototypes of the ATLAS calorimeter.

Table 4. Performances of ATLAS and CMS hadronic calorimeters

the tracker information'', on the basis of what was done, with some success, in some LEP experiments and elsewhere. The so-called "energy flow" approach consists mainly in replacing the energy of each somewhat isolated charged hadron measured in the calorimeter, and linked to a track, by the corresponding momentum measured in the tracker. As an example, the energy resolution of 100 GeV jets is improved from 12 to 8 GeV (simulation done in a low luminosity case). While this looks promising, systematic effects associated to the method need to be evaluated case by case, depending on the physics analysis undertaken.

3 .5 . In situ Calibration 0 Z+jet: Given its cleanliness and high cross-section, this channel is particularly attractive to calibrate jets to the lepton scale of selected Z decays, by transverse momentum balance. ATLAS simulations showed that, with a veto on extra jets and/or an alignment cut in the transverse plane (A4 cut), the energy bias can be kept at the few % levelg (see figure 15).

34 N4.08

3

-

-

0

a” d 0.06 a”

Full simulation

v

0.04

-

0.02

: Q

-

0

-

0

-

-

-0.02 -1

-0.5

0

0.5

I 1

1 1 1 1

1 1 1 1

1 1 1 1

1 1 1 1

L

I

Figure 15. Jet calibration with 2 + j e t s events. Left: relative p l difference between 2 and jet, for p l >40GeV.Right: evolution of the average fractional imbalance with p l jet.

The method is also suitable t o calibrate b-jets (tagged with a displaced vertex), and forward jets by selecting events with the jet in the forward direction, while the leptons of the Z are kept central. The other main tool consists in imposing the W mass constraint to W -+ jet-jet decays suitably selected. This method is particularly well adapted to top mass measurements, using the event sample itself. 3.6. Some illustrations

To illustrate the methods briefly described above, a selected sample of results obtained in ATLAS by s i m ~ l a t i o n ’are ~ shown below: 0 Measurement of the top mass: It is presently estimated that, with the statistics of one year at low luminosity (10 fb-’) the systematic error on the top mass measurement, using the 3-jets decay mode, can be brought down to 1% (figure 16). 0 Higgs associate production: The discovery of a Higgs boson above the LEP exclusion limit, up to ~ 1 3 0 G e Vwill be difficult at LHC and requires several modes. The associated t-t H, H -+b b mode can usefully complement H-+ yy if the two b-jet mass resolution is good enough, and the background well understood. A simulated spectrum, with signal (120 GeV) and background is shown in figure 17.

35

t

2000

c 0

ll 100

200

300

400

mjjb(GeV)

Figure 16. Invariant jjb mass from top decays (loft-’ integrated luminosity). The background (shaded) is dominated by ”wrong combinations”.

Figure 17. Invariant mass of tagged b-jets pairs, above background, for associate t fH production and 1OOfb-1 integrated luminosity.

4. Forward Calorimetry 4.1. A harsh region

The first ”raison d’i5tre” of forward calorimetry is to avoid tails due to particles escaping around the beam pipe, in the measurement of missing ET. For practical reasons, it is difficult to instrument beyond q =5. This is in general adequate, although some channels where precise measurement of ET is necessary would benefit from an extended coverage (H+ TT at low luminosity is such an example). At shower max and q = 4, in ATLAS, the neutron flux is about 3 1015n/cm2/year, a factor 1000 above the barrel EM situation. The main constraint in designing these devices was thus their survival to 10 years of high luminosity. Enough granularity at high pseudorapidity is necessary in order that the relative precision on jet direction do not spoil the relative energy resolution. A summary of specifications and general parameters, for ATLAS and CMS, is given in table 5 below: 4.2. The CMS Hadronic Forward Calorimeter

CMS chose to recess the front face of their hadronic Forward Calorimeter (HF) at l l m from the collision point and to base the detection on Cerenkov light produced in quartz fibres embedded in a metal matrix (figure 18), with a pitch of 2.5 x 2.5 rnm2O.

36 table 5. Parameters fo ATLAS and CMS Forward Calorimeters.

ATLAS Technique Tungsten-Liquid Argon 3.1 < 71 < 4.9 Geom acceptance Layers in depth 3 Number of channels 1792fside Sampling term 9O%/JE Constant term 8% Noise pile-up * 6 GeV ET in R = 0.4 at high luminosity

+

-

CMS iron-quartz fibres 3.0 < 7 < 5.0 2

2096fside 200%/dE -10% 6 GeV ET

-

Figure 18. Central part of a CMS HF module, prior to fibre insertion.

This approach samples mostly the neutral component of the shower, and thus features a large e/h ratio. The detector sensitivity is 5 1 pe/GeV. Despite the high radiation tolerance of quartz it is expected that the light from the most central part of the device will be reduced by -30% after 10 years at high luminosity, this being mostly due to an effect on the fibre cladding (figure 19).

4.3. The ATLAS FCAL The choice of ATLAS was to integrate the forward calorimeter in the same cryostat as the EMEC and HEC calorimeters. The structure chosen is a metal

37

Figure 19. Simulation of radiation damage on light yield from fibres of the CMS HF.

matrix with holes parallel to the beam axis, in which tubes with rods are inserted (figure 20).

Figure 20.

Structure of the ATLAS Forward Calorimeter.

To avoid spill-out of showers, a dense calorimeter was mandatory, which dictated the use of Tungsten for the hadronic part (average density is 14 g/cm3). Radiation resistance of liquid argon made the device possible, even with its front face at 4.7 m from the collision point. The other adverse effect due to the high rates is space charge build-up. The tolerable flux goes like V2/d4p ( p is the ion mobility) and calls for extremely thin gaps. ATLAS chose d=250 microns in the first (EM) section and 375 microns downstream. Several prototypes demonstrated the soundness and expected level of performance of this concept (see ref 9, section 5.1.4). However the price to pay is an extreme cleanliness in assembling the device, in order t o prevent being plagued by high voltage problems (here high voltage is only 300 volts !).

38 4.4. Forward jet tagging

Jet tagging in the forward direction opens up the possibility t o select samples of final states (Higgs boson) produced by WW or ZZ fusion, for which the signal to background ratio is more favorable than in inclusive reactions. The main acceptance region is between 7 = 2 and 77 = 4 (see figure 21), which underlines the necessity to optimize the transition near 77=3 between the “end-cap” part and the “forward” part .

”!.

PT Figure 21. Pseudorapidity distribution of forward jets,in the production of a 300 GeV Higgs boson by WW fusion.

Figure 22. Effect of pile-up on the double-tag efficiency

As an example qqH +qqWW + lvlv jet-jet was studied by simulation in ATLAS. It was shown that requiring a double tag reduces the effect of fake tags to manageable effects (-10% relative loss on double tag efficiency -El> 15GeV- due to a cell cut at 1 GeV ET), even at the nominal high luminosity. 5 . Status of Construction

5.1. Construction advancement The construction of all parts of ATLAS and CMS calorimetry is now going “full speed”. There is however a rather large spread in the present degree of completeness, as a result of several factors, including: - the amount of

R&D which had to be carried out

39 - the difficulties in transferring techniques from labs to industry - the profile of available money and man-power - the overall integration and installation plan of the experiment.

In particular the CMS cavern shall be available later than the ATLAS one, and thus installation in the pit is taking place later (in 05 and 06), with large elements being assembled in advance in the dedicated large surface hall. As an illustration the state of advance of ATLAS production is given below (table 6), for a global detector commissioning as of oct 06 . Table 6. Status of production and integration of ATLAS Calorimeters in spring 02.

ATLAS

EM-barrel EM-EC Tile HEC FCAL

Main Components procured 80% 80% 90% 90%

Module assembled 30% 25% 75% 70%

70%

40%

Integration on surface ends Apr 04 ECC July 04

Installation in pit July 04 Nov 04

ECA

March 05

Nov 04

5 . 2 . Staging plans

Being so critical for the experiment performance, and central in the installation sequence, calorimeters have been so far protected against staging or descoping plans, which may however at some stage affect some back-end electronics.

6. Calorimetry in LHCb 6.1. General strategy LHCb detection is concentrated in a 250 mrad half-angle forward cone around the beam axis (77 >2), which optimises the ratio of event rate t o coverage for B physicsz2. The aim is to have at the collision point a luminosity which maximises the number of bunch crossings with a single proton-proton inelastic interaction. In order to go fully in the direction of "clean events", the calorimeter pulse shaping/clipping is made such that signals generated by collisions in the preceding or the following bunch crossings, are negligible at the time of the main bunch crossing. The experiment aims also at using B decay modes with n-' and qo + yy in the final state. 6 . 2 . LHCb layout The LHCb calorimeter system consists of

40

- A scintillator pad detector (6000 pads) - A preshower detector (scintillator pads after 3x0 of lead)

- A "shashlik" EM calorimeter (6000 towers) - An ATLAS like "Tile" hadronic calorimeter (1500 towers)

6.3. Some performances

The energy range of the calorimeters is limited to E 300 GeV and can be covered by a single electronics gain of 12 bits dynamic. EM calorimeter and pad/preshower cells have a transverse size of 4x4 cm in the central part, 8x8 cm in the middle part, and 16x16 cm in the outer part, giving a total transverse size of about 6.5x6.5m. All devices are read out by photomultipliers (multi anodes for the pad and preshower detectors). The pion/electron rejection provided by the pad-preshower system is around 12 between 10 and 50 GeV. This complements the rejection from the combined magnetic and calorimetric energy measurements (E/p). Expected energy resolutions are l O % O / a and 8 0 % / a for the EM and hadronic part respectively. 6.4. A

WOTTY:

scintillator damage by radiations

The central part of the calorimeter will see doses, integrated over several years of running, in the Mrad range (several lo4 Gy), which will induce light losses. It was estimated that residuals in the correction of this effect will lead to an increase of the constant term of the EM energy resolution from 0.5% to 1.5%. If things go to the worse, the central part could be replaced after some running time.

7. Calorimetry in ALICE The Alice experiment at the LHC is dedicated to ion-ion and proton-ion collisions23. 7.1. Photons at large angle

The EM PHOS ~ a l o r i m e t e ris~optimised ~ to measure direct photons, 7ro and qo between 0.5 and 10 GeV/c. Since only single spectra and correlations are looked for, a complete coverage is not mandatory. The chosen detector covers looo in azimuth and f 0.12 in rapidity. It is located at 4.6m from the beam axis. Alice chose PbW04 crystals, used in a very similar way to CMS. Crystals have a transverse size of 22x22mm and are read out by APDs. In total, there are 17280 crystals in the detector.

41

7.2. Z e n , degree calorimeters Zero degree calorimeters play the special role, in heavy ion collisions, of detecting neutrons and protons from nuclear break-up of the incident ions. In a symmetric machine like LHC, the neutron spot has a 1 cmz size, in between the two rings, 120 m away from the interaction point. Protons are bent away by the beam elements and fall on either side of the beam pipes. In these conditions, compactness and radiation resistance are the prime requirements. Alice has chosen quartz fibres embedded in a Tantalum matrix for neutrons, and in a brass matrix for protonsz5. These detectors are extrapolation of what was used in the heavy ion experiment NA50 at the CERN SPS where an energy resolution of 5.4% was observed when sending 33 TeV lead ions in a similar calorimeter. 8 . Summary

0

0

In many-if not all-cases ATLAS and CMS made different calorimetry choices: this is a safety for the LHC physics programme. Monte Carlo tools (not discussed) are essential in the design phase. While EM simulations give satisfactory results, hadronic packages (GEANT4) still need improvements before being usable for LHC physics. Electromagnetic calorimeters are detectors difficult to build: high precision, high granularity has a technical cost. Located downstream of tracking devices, calorimeters suffer from the associated material: attempts being made to combine information from both detectors (converted photons, energy flow,..) need to be pursued. With the construction now in full swing, calorimeter teams are making every effort t o stay on schedule. Still, they have to stay prepared for unexpected problems, both technical and financial.. .

Acknowledgements All what was presented comes from the hard work of several hundreds of physicists, engineers and technical staff engaged since years in LHC calorimeters development and construction. They deserve warm thanks. To prepare the talk I benefited of discussions and information from many colleagues, and in particular Ph. Bloch, F. Gianotti, D. Green, P. Loch, C. Seez and L. Serin. Finally I would like to thank the organisers of CALOR2002 for a well balanced program and many extremely interesting presentations. Without the help of C. Bourge and C. Drouet, writing these proceedings (almost) in time would not have been possible: many thanks.

42

References 1. ATLAS Liquid Argon Calorimeter Technical Design Report CERN/LHCC/96-41 dec 1996 2. ATLAS Calorimeter Performance Technical Design Report CERN/LHCC/96-40 dec 1996 3. Performance of a large scale prototype of the ATLAS accordion electromagnetic calorimeter. D.M. Gingrich et al., NIM A364(1995)290. 4. Performance of the ATLAS Electromagnetic Calorimeter barrel module 0. To be submitted to NIM. Performance of the ATLAS Electromagnetic Calorimeter end-cap module 0. To be submitted t o NIM. 5. CMS electromagnetic Calorimeter Technical Design Report CERN/LHCC/97-33 dec 1997 6. The lead Tungstate Electromagnetic Calorimeter for CMS. R.Brown in Proceedings of CALOR 2000. Frascati Physics series number 21, 2001. 7. CMS Note 2001/034 and C. Seez, private communication. 8. ATLAS Inner Detector Technical Design Report CERN/LHCC/97-16 apr 1997, and updates. 9. ATLAS Detector and Physics Performance Technical Design Report CERN/LHCC/99-14 may 1999, Vol I. 10. CMS ECAL calibration Strategy. Imperial College workshop, Jan 2002. 11. G.Unal for the NA48 Collaboration, in Proceedings of the 9th International Conference on Calorimetry, CALOR2000, Frascati Physics series, Vol 21,2001. 12. LHCC - March 2002 13. P. Paganini. Zero suppression in CMS. Presentation to this Conference. 14. CMS hadron Calorimeter Technical Design Report CERN/LHCC/97-31 june 1997 15. ATLAS Tile Calorimeter Technical Design Report CERN/LHCC/96-42 dec 1996 16. Studies of the response of the prototype CMS hadron Calorimeter. V.V. Abramonov et al, NIMA 457, 475 (2001) 17. Results of a new combined test of an EM Liquid argon calorimeter with a hadronic scintillating tile calorimeter.ATLAS Collaboration.S.Akhmadalievet a1 NIMA 449, 461 (2000) 18. D.Green Energy Flow in CMS (Sept 2001) and private communication. 19. ATLAS Detector and Physics Performance Technical Design Report CERN/LHCC/99-14 may 1999, Vol I1 20. Test beam of CMS quartz fibre prototype. N. Akchurin et a1 NIM A409 (1998) 593-597. Status of CMS HF Quartz Fiber Calorimetry. Y. Onel. Presentation to this conference. 21. Forward Tagging and Jet Veto studies for Higgs events produced via Vector Boson Fusion. V. Cavasinni et al. ATLAS-PHYS-2002-008 22. LHCb Technical Proposal CERN/LHCC/98-4 23. ALICE Technical Proposal CERN/LHCC/95-71 24. ALICE PHOS Technical Design Report CERN/LHCC/99-4 25. ALICE ZDC Technical Design Report CERN/LHCC/99-5

CALORIMETRY IN ASTROPHYSICS

SIMON P. SWORDY Enrico Fermi Institute, Dept. of Physics, Dept. of Astronomy and Astrophysics University of Chicago, Chicago I L 60637, USA E-mai1:[email protected]

Astrophysical environments produce charged particles and photons over an enormous range of energies. Single particles have been detected with energies in excess of 1OZ0eV. Studying these particles and photons requires a range of calorimetry techniques which are matched to the intensities and shower properties at various energies. At lower energies where the particle intensity is large, straight-forward techniques can be used on spacecraft or high altitude balloons. At the highest energies, the atmosphere itself is used as a n interaction medium because of the huge effective collecting area required. Weakly interacting particles require low background detectors with even larger mass. Experiments are underway t o use deep water volumes, the Antarctic ice, and even the Moon for neutrino detection.

1. Introduction

This review is intended to examine some of the key ideas and design considerations in making measurements of particles and photons from space. It is limited by necessity (and author expertise!) to the kinds of calorimetry which are familiar in the world of high energy physics. There is therefore no discussion here of the astrophysical techniques and detectors which probably more deserve the label ‘calorimetry’, as defined in any undergraduate text in the chapter(s) on thermodynamics. These other, more traditional, calorimetric techniques which have provided exquisite measurements of the microwave background and produced eV level energy resolution for single x-ray photons are not discussed here. A detector for high energy particles or photons placed above the atmosphere will see a constant bombardment of near isotropic radiation from space. Most particles are protons with an intensity 10-100 times that of cosmic ray muons at the Earth’s surface. Essentially all particles and photons above several GeV have their origins outside the solar system and thus truly deserve the name promoted by Compton of ‘Cosmic Radiation’. These cosmic rays consist of all stable particles and nuclei with abundances varying in a similar way to the overall abundance of matter in our galaxy. The differences between cosmic ray

43

44 abundances and those observed in our galaxy by other methods have provided some of the crucial insights into cosmic ray origin. Reliable measurements of cosmic ray elemental adundances at high energy remains a central scientific goal in astrophysics. Photons and electrons are rare in cosmic rays of the same energy. However they are extremely important: In the case of photons, they are not deflected by the magnetic fields of the galaxy and therefore can be traced back to their sources. This has lead to the discovery that many photons above lGeV originate outside our galaxy - in the case of gamma-ray bursts from distant objects, apparently at large redshiftsl . The dominant feature of arriving cosmic rays is shown in Figure 12. Measurements of overall particle ‘fluxes’ are given as a function of total particle energy. These decline as a power law in energy, E-“ with an index CY 2.7 - 3.0. The only features in this spectrum are the slight steepening or ‘knee’ near 1015eV and the flattening or ‘ankle’ near 101’eV. In principle a calorimeter could be built to cover a large part of this energy range, since the depth of hadronic shower maximum increases logarithmically with energy. As shown in Figure 1 this might be expected to get up to 611 at the highest energies. The real issue is how large an acceptance aperture is needed for devices at high energy. The steeply falling energy spectrum means something in the detection scheme has to get 50-100 times larger for every decade increase in particle energy to achieve the same count rate. The extremely low fluxes at high energy can only be sampled using huge, naturally occurring, media such as the atmosphere. While most measurements below 1014eV are made above the atmosphere, at higher energies detection techniques are based on observations of air showers. The situation with photons is similar: most sources exhibit power law flux spectra. Here the energy ‘break’ for techniques is at lower energy but is more pronounced. Current/past satellite experiments have little sensitivity above 30GeV and the atmospheric techniques begin to have sensitivity above 75GeV; there is no real overlap region.

-

-

-

-

-

2. Direct Measurements In the region below the ‘knee’ direct measurements are made with thin hadronic calorimeters. These are devices which because of weight constraints do not fully contain the hadronic shower. Instead they focus on collecting the energy from r 0 s created in the first interaction. Typically these instruments will contain a primary ‘target’ layer of 0.511 of low atomic number material followed by an electromagnetic calorimeter layer consisting of 15-20 r.1. of high atomic number material. A typical overall depth for these instruments is 211 with an rms energy resolution of 50%. The JACEE detectors with this design, flown on high altitude balloons, use a passive detector system consisting of

-

-

45

1 4

4

(1 particle per m2-second) tmax 2A

-

Q

0 9

‘9

-

*O *O .O ‘ ,o

Knee (I particle per m2-year) tmax-3 5A

,O ‘%

Direct <- Meas.

-

%\b

?%

\%

Indirect Meas. ->

-

1

-

-

-

Ankle (1 particle per km2-year) tmax 6A

-

-

(Swordy -U Chicago)

Figure 1. The overall energy spectrum of cosmic rays. The approximate intensity rates and locations of hadronic shower max. in an iron calorimeter are shown for some representative energies.

emulsion layers and x-ray films which are analyzed after the flight3. A general issue which arises with this kind of instrument is finding an adequate system to identify the nuclear charge before the interaction. As mentioned previously, a crucial aspect of cosmic ray science is the reliable identification of the elemental abundances. In principle this is a straightfor-

46

ward measurement since the particle energy loss before the first interaction is c( Z 2 ,where Z is the nuclear charge. However there can be a large background produced in the ionization detector caused by albedo particles from the subsequent calorimeter shower. This could make, for example, a proton appear to be a helium nucleus, or worse. Two methods have been developed to avoid this problem in electronic detectors. A highly pixellated ionization loss detector can reject background from all regions except very close to the particle path. This method is implemented in the silicon detector scheme of ATIC4, which had its first high altitude balloon flight around Antarctica in Winter 2000/2001. ATIC uses BGO crystals for the e/m calorimeter layers to measure particle energy. Another scheme, which uses timing in fast scintillators is being implemented in the CREAM5 experiment. Here nanosecond level timing is used to separate the incoming particle arrival from subsequent albedo particles. This experiment is planned to be flown around Antarctica in Winter 2003/2004 and uses scintillating fiber/tungsten sandwiches for the e/m calorimeter section. A key issue with these experiments is the calculation of instrument response. There are no calibration beams available which can cover the range of incoming particle energies and nuclear masses so high energy response is calculated using simulations which fit with lower energy data. The direct detection of gamma-rays above the atmosphere has produced a wealth of information about the non-thermal parts of our universe. At high energies these activities were revolutionized by EGRET' on CGRO which used tracking chambers to detect the initial electron-positron pair produced by the incoming gamma-ray. This was combined with a NaI crystal calorimeter, 8 radiation lengths deep, to measure the gamma-ray energies. The next step in direct observations at high energies will be GLAST7, which uses silicon strip trackers and a CsI calorimeter to produce a much wider field-of-view instrument.

3. Effects of Energy Resolution The measurement of cosmic ray or photon spectra is clearly directly affected by the calibration of the energy scale of these devices. What is sometimes not so well appreciated is that the energy resolution can have dramatic effects on the measured fluxes. Figure 2 shows schematically the effect of instrumental resolution on measurements of the overall cosmic ray spectrum as dashed lines. The absolute size of these effects is greatly exaggerated in these examples. The effect of constant energy resolution in a / E has a direct impact on the absolute flux, this is the leftmost example in the picture. The overall measured flux is larger

47

-$ "A*

\

Finite energy resolution in o/E

'0 \

'

\

9

\

\

\

/f

\

,0

Relative energy resolution improves with increasing E

'% ,.",

\

,~,\

\

x

\%... ,

Relative energy resolution degrades with increasing E

~nmul I ~ t u w lI U I ~ i a dI I ~ I I I I I ~1 1 a i i d lo9 l o i o l o 1 ' 1 0 ' ~ 1013 1014 l o f 5

I t111m1 I

I B I ~ I I ~

10''

I tiiwd

lo1* l o f 9 lozo Energy (eV)

Figure 2. The effect of energy resolution on measurements of cosmic rays. The dashed lines shows examples of measured spectra under conditions described in the text

-

than the actual flux at some energy by a factor 2 x a l E because of the finite resolution of the detector. More problematic are energy resolution functions which vary with energy. As shown in the center example, an energy resolution ( a / E )which improves with energy will produce a measured spectrum which is

48

steeper than the original. The rightmost example shows how a degradation of resolution with energy will flatten the measured spectrum. Understanding the energy resolution function is a crucial part of any cosmic ray experiment, since it is needed to transform these measured spectra back to the input spectrum of particles. In particular, devices with resolution ”tails” to the high energy side can prove to be useless because these resolution effects cannot be reliably corrected. 4. Atmospheric Calorimetry

For the highest energy particles and photons a large collecting area is required. The atmosphere has been used extensively since the 1950s for this purpose. The vertical depth of the atmosphere near sea level is 1000g/cm2, which is around 27 radiation lengths or 1111. Curiously this is a pretty good match to the highest energies shown in Figure 1, since a cosmic ray at 1OZ0eV produces a vertical shower which maximizes at -750g/cm2. There are several techniques which are used widely for atmospheric calorimetry:

-

-

Detection of particles at ground level. Detection of Cherenkov light from the particles in the shower. Detection of air fluorescence emission from the shower particles.

4.1. Particle Sampling The most straightforward technique for the detection of air showers involves distributing an array of particle counters at ground level to make multiple single-point samplings of the particle distributions in the showers. The electrons and muons produced in these showers range over distances of hundreds of meters to many kilometers depending on the incident cosmic ray energy and the atmospheric depth of observation. The shower particles in the shower arrive very closely in time (< 100ns) producing a strong shower ‘front’. The number of particles detected in each counter can be used to fit the shower arrival direction parameters and size. The details of the lateral distribution of particles can be used to estimate the ‘age’ of the shower or where the likely shower maximum occurred in the atmosphere. This is important because both the location of shower maximum and the shower size are affected by the initial cosmic ray energy and the nuclear mass. This can be broadly understood in a simple superposition model where the energy is spread evenly between the nucleons. The overall shower can be thought of as the superposition of A independent showers produced by the nucleons each carrying 1/A of the total energy. The overall shower size is similar for nuclei of the same energy but

49

the location of shower maximum is higher in the atmosphere for heavier nuclei since these showers look like the sum of A smaller showers. More recent air shower arrays of this type take many samples of the particles at ground level for each shower8. The KASCADE array even makes an additional measurement of the energetic particles in the shower core using a hadronic calorimeter located on the groundg. 4.2. Cherenkov Detection

During dark moonless nights air-showers are visible from the Cherenkov radiation emitted by high energy electrons in the showers. This radiation is beamed in the direction of the shower providing a fairly bright light signal from only a few thousand shower particles. This beaming makes a low energy detection threshold feasible. The pool of light on the ground has a typical radius of 100-200m which is produced by the intrinsic Cherenkov angle and multiple scattering contributions to the particle directions in the showers. The Cherenkov radiation from a shower can be fit to density distributions in much the same ways as particle distributions to get the locations of shower maximum and shower size. In some sense Cherenkov observations are more truly 'calorimetric' since they contain contributions from the entire shower not just a single plane at ground level. The overall size signal from Cherenkov observations has generally smaller fluctuations at a given energy than particle sampling. But setting an absolute energy scale is considerably more difficult for Cherenkov observations where optical collection efficiencies and atmospheric transmissions need to be well known. The most spectacular examples of Cherenkov detection are in high energy gamma-rays. Here imaging telescopes routinely detect gamma-rays at energies down to NlOOGeV from galactic and extragalactic sources. Imaging atmospheric gamma-ray telescopes, such the Whipple" 10m at Mt. Hopkins in Arizona, collect images of air showers in an -fl telescope with a few degrees field of view. A key method for providing signal to noise stems from the fact that the gamma-rays are expected from a point source whereas the background of cosmic ray events is isotropic. The gamma-ray showers that apparently come from the source have an identifiable asymmetry in the focal plane and can be separated from the more numerous cosmic ray induced air showers which have arbitrary directions. 4.3. Air Fluorescence

Fortunately nature provides scintillation light from air at wavelengths near 325nm. This radiation is emitted isotropically and therefore showers can be

50

viewed from essentially any geometry with sensitive enough cameras. In a simple model of this emission the excited states of nitrogen, caused by particle ionization losses, can either decay on timescales of -100ns or lose energy via collisional quenching. Since the quenching rate increases with pressure this decrease in light yield approximately compensates for the increased ionization loss with pressure. As a result the amount of light output per unit ionizing particle pathlength in the atmosphere is nearly constant. (This is an oversimplification since the effects of temperature have not been discussed - but the net result is very similar t o this simple picture) This means that a shower viewed in air fluorescence will have an absolute luminosity which only depends on the number of particles in the shower at some depth and not on the absolute atmospheric pressure. In principle the longitudinal shower profile can be directly measured. However, the light yield is low and this technique has only been useful in the region above 1017eV. The HiRes/Fly’s Eye group in Utah have built wide field-of-view cameras with mirrors and clusters of photomultipliers for observations of air fluorescence of the highest energy cosmic rays1’. More recent observations have produced stereo views of the showers from cameras widely separated on the ground, allowing for excellent geometrical reconstruction of the events. There are also efforts to detect these events with orbiting satellite cameras which can view very large regions of the atmosphere. While the air-fluorescence technique can in principle observe huge volumes of the atmosphere, great care needs to be taken with calibration and the corrections for atmospheric attenuation. N

4.4. A i r Shower Results

Some results from atmospheric calorimetry for cosmic rays are shown in Figures 3 and 4. In the first of these (taken from PDG section on cosmic rays”) the highest energy part of the spectrum is shown multiplied by a scaling factor E3 to emphasize differences between measurements. Here the filled squares at high energies come from the ground based array AGASA and the filled circles from the HiRes/Fly’s Eye group. There are many detected particles with energies in excess of 1OZ0eV.Figure 4 shows the mean location of shower maximum in the atmosphere as a function of particle energy from a variety of experiments (taken from the PhD thesis of J. Fowler13). At lower energies these come mostly from Cherenkov experiments and above 1017eV from air fluorescence. The two model lines derived from QGSjet14 suggest the mean nuclear mass lies between protons and iron nuclei over most of this energy range.

51

t I

I

I I I I l l /

I

I

I I I I l l

I

I

I I I Ill

Figure 3. Measurements of the fluxes of the highest energy cosmic rays using the atmosphere as a calorimetric volume. Taken from PDG handbook where full references appear. The experiments associated with these symbols are discussed in the text.

5. Other Techniques

The search for high energy neutrinos from astrophysical objects has produced a class of detectors which use other natural materials as the primary detection medium. Since the fluxes of high energy neutrinos from astrophysical objects are unknown the motto here is essentially ‘bigger is better’. The AMANDA15 detector has been developed using Antarctic ice as a Cherenkov radiator to provide a large mass detector. The events are viewed by strings of photomultipliers sunk into the ice. It is hoped to ultimately build a detector of size lkm3 with a billion tons of ice as the target material16. There are also groups working on deep water Cherenkov detectors17. Again, viewing Cherenkov light from showers with many photomultipliers submerged in long strings. The technical challenges for these experiments are enormous, ranging from providing photomultiplier enclosures, trigger and acquisition electronics which can survive these rough environments to understanding how to avoid luminous fish and other organisms. A crucial feature needed is the ability to separate locally produced neutrinos from cosmic ray interactions from those of astrophysical origin. This generally involves providing a method to select events which come ‘up’ through the

52

QGSJET

350;0i1

'"""1

' ' 10 l5

"""' 10 l6' ' """'10 l7' ' " " " 1 10

' '

"""'10 l9

Energy (eV)

Figure 4. The variation of mean location of shower maximum observed in the atmosphere by various experiments compiled by Joe Fowler. The lines show the expected values for P or Fe nuclei predicted by the QGSjet interaction model.

detector - using the mass of the Earth as a screen for cosmic rays. Interestingly a major mode of operation of these detectors is expected to be the detection of upwardly moving energetic muons produced by neutrinos in the rock mass below the detectors. So this is really moving towards using the planet as a calorimeter! There is also an effort underway to try to detect neutrinos interacting in the moon through their radio emissions'*. A large shower in a solid material produces radio waves through the Askaryan effectlg, where the shower becomes predominantly negative in solid media and the emission from the shower front is 'in-phase' which greatly enhances the radio signal for large showers. This effect has been recently detected in an experiment at SLAC2O. The idea is to view the moon through earth-based radio telescopes separated by a long baseline and look for coincidences. This is expected to be sensitive to neutrinos of energies 1020eV. N

53

6. Summary The range of calorimetry in astrophysics is as broad as the phenomena being observed. Ideas about calorimeters range from small detectors for gammarays at -100MeV to schemes which start to use significant bits and pieces of our planet and even its moon. Perhaps the strongest challenge facing these huge detectors is that of energy scale calibration. Certainly the bulk of the high energy calorimetry techniques discussed here rely on extrapolations of shower physics into energy regions where no measurements exist. Progress can undoubtedly be made by combining different techniques at the same site since each can be expected to have different types of systematic errors. For example the Augerz1 experiment for ultra-high energy cosmic rays is using both ground sampling and air fluorescence measurements at the site in Argentina. This seems to be a good principle for future experiments to follow. Measurements which at first might seem redundant may be able to prcvide confidence for measuring particles in energy regions which are beyond the reach of present day accelerators. References 1. D. Reichart A p . J. 554,2001,643. 2. Compilation available at http://astroparticle.uchicago.edu/archives.htm 3. T. H. Burnett et al. A p . J. 349 (1990)25. 4. http://atic.phys.lsu.edu/aticweb/ 5. http://cosmicray.umd.edu/cream/cream.html 6. http://lheawww.gsfc.nasa.gov/docs/gamcosray/EGRET/egret.html 7. http://glast.gsfc.naa.gov/ 8. see e.g. CASA at http://hep.uchicago.edu/ covault/casa.html 9. see KASCADE at http://iklaul.fzk.de/KASCADEhome.html 10. http://egret.sao.arizona.edu/ 11. http://hires.physics.utah.edu/ 12. http://pdg.lbl.gov/ 13. J . Fowler, PhD thesis, University of Chicago, 2001. 14. N.N. Kalmykov, S.S. Ostapchenko, Yad. Fiz. 56 (1993) 105. 15. http://amanda.berkeley.edu 16. http://icecube.wisc.edu 17. see e.g. ANTARES at http://antares.in2p3.fr/ and NESTOR

at http://www.nestor .org.gr/programme/scinet-index. htm 18. P. Gorham et al. in proceedings RADHEP2000 conference, AIP 579, 2001, 177. 19. G.Askaryan, Sov. Phys. J E T P 14, 441 (1962);21, 658(1965). 20. D. Saltzberg et al. Phys. Rev. Lett. 86 (2001)2802. 21. http://www.auger.org/

CALORIMETER CONSIDERATIONS FOR A LINEAR COLLIDER DETECTOR

RAYMOND E. FREY Physics Department and Oregon Center for High Energy Physics University of Oregon, Eugene, OR 97403 E-mail: myfreyOcosmic.uoregon.edu

Current trends in the consideration of calorimeters for a future linear e+e- collider detector are discussed. The physics requirements and LC environment are briefly reviewed. The paradigm that excellent jet reconstruction can best be realized when the charged and neutral jet components are separated in the calorimeter is discussed. Design ideas are given, citing specific examples now under consideration in Europe, Asia, and N. America.

1. Introduction

A concensus has emerged internationally in the last two years that an e+elinear collider (LC) in the energy range 0.5 to 1 TeV is the highest priority future project in elementary particle physics. Accelerator designs and test facilities in Europe, N. America, and Asia have made tremendous progress. This activity has inspired increased attention to LC detector design. Here, the prevalent ideas for calorimeter design are discussed. We start by highlighting the physics prospects, then discuss the LC detector environment, followed by design considerations and specific examples. There is perhaps some feeling in the hadron collider community that e+edetectors are straightforward, given the relatively benign LC environment, and perhaps do not merit significant R&D effort. However, I argue that this environment in fact allows one to think of detectors which are significantly better than the previous generation of excellent detectors at LEP and SLC. Hence, the LC provides an opportunity to achieve improved levels of physics measurement, and we should strive to take advantage of this. Calorimeters are perhaps undergoing the most extensive evolution in design to meet the LC challenge. An important element of the design philosophy is that the goal is not to build the best possible calorimeter, but rather the best possible detector. This highlights the fundamental interdependency of the calorimeter and other sub-detectors in making a physics measurement. Our experience with previous

54

55

detectors is pointing the way. But significant R&D will be required to take this next, significant step. And it is now underway. 2. Physics Requirements

In general, the main physics goals of the LC are the same as the LHC, namely to uncover the new physics responsible for electroweak symmetry breaking and to explore other phenomena at the TeV energy scale. Detectors must be prepared to study Higgs physics, supersymmetry (SUSY) if present, top physics, new and old gauge bosons, and so forth. While the LHC will likely be the discovery facility, the LC will be required to fully explore the physics. Perhaps one would expect scenarios similar to that of the W and 2 bosons, where discovery at hadron colliders was followed by a full exploration of their properties at LEP/SLC. So the LC is expected to make measurements which are difficult or impossible at the LHC, but are essential for the elucidation of of the underlying physics. Perhaps the most important example, and the one most pertinent to calorimeter design, is the capability to measure electroweak processes which decay hadronically. In fact, multi-jet final states are common signatures of most new physics processes, many of which involve W and 2 as intermediate states. Typical examples include the separation of the hadronic decays of WW from 22, or 22 from Z H . Some important final states, such as H H Z to determine the Higgs self-coupling, have small cross sections and hence require the reconstruction of all final states, including those with 6 or more jets. Assuming that the detectors have this capability, the LC can provide these measurements. In addition to jet final states, the LC physics also requires that leptons are well measured. Tau identification and measurement becomes very important at the LC, and is mentioned further below. SUSY final states require that the calorimeter coverage extend to small scattering angles, with no cracks. There is no clear physics case at the LC for excellent photon energy resolution, such as that for H + yy at the LHC. On the other hand, some SUSY models predict secondary vertices where a photon is the only visible particle. Thus, one would like to identify photons which do not originate from the IP.

3. The LC Environment The LC designs call for a maximum fi in the range 500 to 1000 GeV at a luminosity of a 2 - 4 ~ 1 0cm-2s-1. ~~ The ease of low energy (Mz) running is design dependent and a subject of current debate. A lovely feature of the LC is that the collision IP is indeed a point. Along with the small beam radius, this

56

allows excellent vertexing capability. The main environmental issues which drive calorimeter design are IP radiation and the accelerator bunch timing structure. 3.1. The IP and IP Radiation The high charge density of the beams at the LC IP gives rise to photon radiation (beamstrahlung) and production of low energy e+e- pairs. Roughly lo5 pairs are produced per beam crossing, as shown in Figure 1. Fortunately, the pairs have small, limited transverse momentum. Therefore, the detector solenoid prevents them from entering the detector proper. To allow the vertex detector to be 1 cm from the IP, the field strength needs to be about 3 T or greater. N

Figure 1. Pair production simulation from a LC bunch crossing. Note that the x and y scales are very different - the conical masks are about 2' from the beamline.

3.2. Bunch Timing The TESLA and NLC/JLC designs have quite different timing structures. The differences are intrinsic to the linac RF technologies employed. For TESLA, bunch trains are supplied at 5 Hz. Each train has a length of 0.95 ms and the bunches cross every 337 ns. For NLC/JLC, the corresponding numbers are 150 Hz, 269 ns, and 1.4 ns. The physics event rate is small, but it is not yet clear if timing within a bunch train is required to avoid pile-up from 2-photon

57

events and backgrounds. Individual bunch timing would clearly be a challenge at NLC, but should be readily achievable for the TESLA calorimeter. One notes that there are lengthy intervals between bunch trains. This implies that power cycling of calorimeter electronics could provide large reductions of the heat load. The NLC (duty cycle 5 x lop5) provides an advantage in this case compared to TESLA (5 x Finally, a small bunch crossing time interval requires a finite beam crossing angle to avoid additional unwanted bunch crossings. An angle of 20 (8) mrad is chosen for NLC (JLC). While a non-zero angle is not in principle necessary for TESLA, the zero-angle crossing design is technically difficult. A crossing angle has obvious implications for detectors placed at very small angle, but these detectors inside the masks are not discussed further here. 4. Making the Most of the Tracker: The Energy Flow Method

As discussed above, excellent jet reconstruction and measurement is the outstanding challenge for LC calorimeters. Two basic facts drive the approach to this measurement. First, jets are composed primarily of charged particles. For example, for 22 --t jets, the visible energy is 64% charged particles, 25% photons, and 11% neutral hadrons. (These numbers change very little for other hadronic final states.) Second, jet particles do not have large momenta, and the energy resolution for charged particles is vastly better in the tracker than the calorimeter. This last point is illustrated in Figure 2 which gives a typical momentum distribution, and in Figure 3 which compares tracker and calorimeter resolution for charged pions.

Trucker, SO 10-3

10-4

u 20

10

80

80

100

Momentum I G e V l c I

Figure 2. Typical charged particle momentum distribution for a multi-jet process, in this case e+e- + 22 +jets.

Figure 3. Comparison of typical singleparticle energy resolution for charged pions for calorimetry and for an LC tracker. The tracker is the SD design with 0 M 90°.

58

Clearly, one would like to take full advantage of the tracker for jet physics. In fact, this idea has been in use in e+e- detectors for ages, in one way or another. It is sometimes called “energy flow”, although there is no standard use of the term, as it is often used to describe the more far-reaching methodology discussed below. In any case, at the LC we are in position t o try t o push the energy flow idea t o a new level. Various approaches on how to achieve this are discussed in the next section. Finally, it is noted that energy flow will fail for sufficiently high fi due to higher track density and the eventual convergence of the single-particle resolution curves. Simulation studies indicate that this is well above 1 TeV. 4.1. Segmentation Requirement

Which strategy will best allow one to take advantage of the tracker measurement of the charged pions in jets? First, the photons will be measured in the electromagnetic calorimeter (ECal). So if one could make the ECal transparent to hadrons, then neglecting neutral hadrons, one would have excellent jet

4 HCal

G

I

Figure 4. Illustration of the interaction of a 7ro and a .rr+ from a jet in a cartoon calorimeter. The .rr+ does not interact in the ECal in (a), but does in (b). The ECal has no longitudinal segmentation and the transverse segmentation is indicated.

59

measurements - all the energy in the ECal would be from photons. This is illustrated in Figure 4(a). (Hence, it is helpful for the ratio of radiation length to interaction length, Xo/X, to be relatively small for the ECal.) Unfortunately, a real ECal will have X 1, so our cartoon will often look like that in Figure 4(b). Without sufficient segmentation in the ECal, we cannot separate the photonic contribution(s) from that due to the pion(s). And in general, one would like to have segmentation in 3 dimensions to do this efficiently. Similarly, one needs to separate charged and neutral hadrons in the hadron calorimeter (HCal) in order to take advantage of this method. The highly segmented approach to the implementation of energy flow is the one adopted by the TESLA and the American SD designs, discussed below. Two other approaches can be considered. One might emphasize the ECal energy resolution by using crystals. Here, one has the problem with lack of segmentation illustrated in Figure 4, but would presumably try to correct using average energy depositions. CMS has shown that this is helpful'. One would presumably include small Xo/X as a design criterion. Finally, one can emphasize the hadronic response of the calorimeter by choosing one which provides hardware compensation. Charged/neutral sepa-

-

0

2

4

6

8

10 3.2

14

16

18

20

0

2

4

6

S

10

l2 14

16

18

20

Tmck€hskr distance (an ) Figure 5. Example of ECal transverse segmentation requirement for e+e- -+ tf -+ jets for the SD detector. The left-hand figure gives the transverse separation between the charged particle position determined in the ECal and the centroid of all photon showers. The righthand figure is the transverse separation between charged particles as determined by the ECal and the tracker. Figure from M. Iwasaki.

60

ration would still be necessary, and it is difficult to find a technology which combines compensation with excellent segmentation. This is the approach of the JLC and American LD detector designs. 4.2. Requirements for the Electromagnetic Calorimeter

In considering ECal segmentation, one should first examine what is implied from LC jet physics and the detector induced charged particle sagitta, which is proportional to BR2. A typical example is given in Figure 5 for the SD detector. We see that the physics requires a separation at the level of x 1 cm. Hence, we strive to use a small MoliBre radius, tungsten being an obvious choice with R, = 0.9 cm. A figure of merit is BR2/R,. Finally, the segmentation need to be comparable to R,, smaller if possible. The TESLA and SD designs implement this segmentation using silicon detectors throughout the ECal, amounting to N lo3 m2 of silicon.

/

\ ECal

\

-104

/

-80 -78

-76 -74

-72

-70

X

Figure 6. Photon “tracking” in the SD detector Si/W ECal. The small squares are hit cells. The shower profile is fit and extrapolated to the IP, shown as the dark line. The front face of the ECal is at the upper right. Based on a GEANT4 simulation. Figure from T. Abe.

61

A dense, highly segmented ECal provides other important assets. It allows tracking of MIPs to extend into the calorimeter. As we point out below, this is important for the HCal, too. It also provides photon tracking, which opens the possibility to find photons not originating from the IP, potentially a critical signature for new physics. Figure 6 shows a display from a GEANT4 study of photon tracking in the SD detector. The resolution on the extrapolation to the IP was found to be 3.5 cm in both r-4 and z for 10 GeV photons. The same study applied to charged particles found a 1 cm error. Numerous physics studies have pointed out the importance of identifying 7's. Also, because of the benefits of using the tau as a polarimeter, one might demand that the calorimeter have the ability to identify some specific decay modes. Figure 7 shows that with sufficient granularity, one of these decay modes can be identified even at very high boost.

Figure 7. Simulation of a 300 GeV T with the decay mode Si/W ECal and digital HCal. Figure from H. Videau.

T

+ pv -+

7 r T + ~ O in u

the TESLA

4.3. Requirements for the Hadmnic Calorimeter

While hadronic showers are large and diffuse, one still expects a highly segented HCal to be very important, primarily because tracking MIPs through the HCal is seen to be a key element in energy flow pattern recognition. Here is an outline of how this information might be used: One would do tracking of charged particles in the HCal. If they do not interact, they are probably muons (to be verified with a muon detection system). If a track terminates at a shower(s), begin pulling in shower energy based on a x2 shape criteria. A constraint on E / p matching would help terminate the process. Whatever is left is due to neutral hadrons (subject to reasonable consistency constraints). And one gets reasonable muon identification as part of the process.

62

-

The “digital” HCal, discussed further below, pushes the transverse segmentation to 1 cm. This approach is being pursued as an option for TESLA and with the SD detector. At this time, sharp criteria for HCal segmentation are still being developed. So it may be the case that a practical compensating technology like Pb and scintillating tiles can also provide adequate segmentation. This path is being pursued with the JLC and LD detectors. We note that all designs to date have been able to put both ECal and HCal inside the solenoid coil, even with the large fields being pursued at the LC. Of course, this is beneficial to calorimeter resolution if it can be achieved. This is aided by the fact that the HCal at the LC does not have to be nearly as deep as those at a hadron collider, perhaps even more so for highly segemented calorimeters which allow tracking of MIPS.

4.4. Limits to Jet Resolution

It is interesting to evaluate jet measurement using the energy flow teechnique in the case where the reconstruction algorithms are capable of correctly matching all energy depositions with the corresponding particle, i. e. perfect pattern recognition. Using the process e+e- -+ qq as a benchmark, Figure 8 gives a sample jet-jet mass distribution. And Figure 9 gives the jet energy resolution as a function of Ejet. The width of the peak in the mass distribution is dominated by the ECal resolution for photons ( x 0.15/&?), while the tails result from fluctuations of neutral hadrons and neutrinos.

0.5 l

Figure 8. Jet-jet mass distribution for e+e- + qq at 200 GeV in the SD detector in the limit of perfect pattern recognition.

100 .

E,.,

IGeVI 200 1

300

1

Figure 9. Jet energy resolution for e+e- -i qq as a function of Ejet in the SD detector in the limit of perfect pattern recognition.

63

The fit in Figure 9 gives a resolution of 0.15/-. This agrees with the result derived2 using a formulaic approach. The agreement demonstrates that fluctuations in particle types are unimportant, the resolution is limited by detector resolution, and QCD effects are relatively unimportant. This is in dramatic contrast to the case for hadron colliders, where detector resolution is Table 1. Parameters of calorimeter designs currently under consideration. Note that many parameters will change as designs evolve. The labels T and D refer to two TESLA design options.

Tracker type

TESLA5

SD6

TPC

Silicon

L D ~ TPC

JLC7 Jet-cell drift

ECal Rmin barrel (m) Type Sampling Active Gap (mm) Long. readouts

1.68

1.27

2.00

1.60

Si pad/W

Si pad/W

scint. tile/Pb

scint . tiIe/Pb

30 X 0.4Xo +10 x 1.2Xo

30 X 0.71Xo

2.5 (0.5 Si)

2.5 (0.3 Si)

1 (scint.)

1 (scint.)

40

X

0.71Xo

38

X

0.71Xo

40

30

10

3

Trans. seg. (cm)

X 1

0.5

5.2

4

Channels ( x lo3)

32000

50000

135

5

zmin endcap (m)

2.8

1.7

3.0

1.9

1.91

1.43

2.50

2.0

TJSrpe

T: scint. tile/SS D: digital/SS

digital

scint. tile/Pb

scint. tile/Pb

Sampling

38 x 0.12X (B), 53 x 0.12X (EC)

34 x 0.12X

120 x 0.047X

130 x 0.047X

Active Gap (mm)

T: 6.5 (5 scint.) D: 6.5 (TBD)

1 (TBD)

2 (scint.)

2 (scint.)

Longitudinal readouts

T: 9(B), 12(EC) D: 38(B), 53(EC)

34

3

4

T: 5-25 D: 1

1

19

14

5"

2O

2O

8'

3.0

2.5

3.7

3.7

4

5

3

3

option: Si pad shower max. det.

scint. strip shower max. det.(2 layer)

HCal Rmin (m) barrel

Transverse segment. (cm) Omin

endcap

coil Rmin

( 4

B (TI Comment

Shashlik ECal option in TDR discontinued

64

in fact nearly negligible3 compared to QCD and underlying event effects. The ideal result should also be compared with simulations from the TESLA group, which give jet energy resolution of a E / E = 0.30/& for e+e- + qq with realistic simulations, as shown at the previous meeting4 in this series. So presently there is a factor two between the ideal case and an existing reconstruction. Presumably, some of this difference can be reduced by the development and improvement of reconstruction algorithms. To summarize, these studies allow one to make two general points: (1) LC jet resolution will be limited by detector resolution, not by QCD or other effects not under the control of the experimenter. (Nevertheless, it is still important to quantify QCD effects, which are likely to become more important for final states with many jets.) (2) The jet resolution will be in the range (0.15 to 0.30)/& (for 2-jet final states) for the TESLA or SD type of detector (see next section). It is important that proponents of various calorimeter designs be able to compare resolutions using equivalent full simulations, eventually reinforced by test beam results. 5 . Design Ideas

Table 1 summarizes some of the major calorimeter parameters for current detector designs. It should be noted that in some cases these parameters are rapidly evolving, and are likely to soon be obsolete, if they are not already. The TESLA parameters are based on the TDR from 2001, except that the Shashlik ECal option is no longer included. In the discussion below, we note some recent changes and highlight some areas of R&D activity. The JLC parameters are based on the recent (July 2002) 3T design in the ACFA report. 5.1. TESLA

The calorimeter for TESLA consists of a highly-segmented (imaging), silicontungsten (Si/W) ECal. There are two options for the HCal, one using scintillating tile detectors and the other a “digital” HCal discussed below. Both HCal options use the same mechanical layout. The barrel ECal consists of eight modules, a few of which can be seen in the figure. In the design from the TDR5, the ECal front end electronics is at the edges of each module, positioned to be out sight of the IP. The readout end of the detector slab is shown in Figure 11. Recently, an alternative design has been developed which inserts the front end electronics within the slab. This has several advantages’, but requires more attention to thermal management. Cost is always a concern for silicon detectors. In the TESLA TDR, the silicon detectors are about 70% of the ECal cost While the TDR design included 40 Si layers, an alternative design

65

Figure 10. Simulated event in the TESLA detector with the digital HCal option.

with coarser sampling (20 layers) would reduce expense with a corresponding increase in photon resolution (0.11 to 0 . 1 4 / a ) . A conservative estimate of the cost in 2005-10 for these relatively simple silicon detectors is $2/cm2. This puts the ECal cost estimate at roughly 70 MEuros for the 20-layer design. The tile HCal option was discussed in detail by Korbelg at this meeting. The digital HCal option calls for enough segmentation so that a one bit readout suffices. This offers the possibilty of a greatly simplified readout which can offset the cost of the increased segmentation. The current design calls for 1 cm transverse segmentation. Figure 12 compares single particle resolution for the usual analog sum of HCal cells t o the “digital” sum, that is, the multiplicity of hit cells. We see that the digital resolution is actually slightly better than that for the analog sum. An R&D effort is underway to choose an appropriate detector for this option. RPC detectors, if made to operate reliably, might be a good choice. A glass RPC, being considered for TESLA, is shown in Figure 13.

66

m ax45 m m l;lontmd ekbmjcs Figure 11. The end of the Si/W detector slab from the TESLA TDR design.

qo5

w a

0A

03

02

01 0

2

4

6

8 1 0 1 2

lxdmnenergy GeV Figure 12. Comparison of single hadron energy resolution for a sum of the full energy of hit cells (triangles) and the digital sum (squares). The circles are after additional pattern recognition is applied to the digital hits. Figure from H. Videau.

67

Figure 13. A glass RPC being considered for the TESLA digital HCal option. Figure from V. Amassov.

5.2. JLC Detector

The ACFA/ JLC detector is based on a hardware compensating configuration of P b and scintillating tile layers in the ratio 4:l. This structure has already undergone extensive test beam evaluation, an example of which is given in Figure 14. One challenge for this design is whether the segmentation, both

O p t k a l F h ~ ~ ~ f UEM X iC

(O.7mm

resolution vs. Pb thickness for n I

h

K v W 7

.

I

'

I

'

I

x144)

'

fit function J(u:+dXu:)

60

2

U.(Z) u,(Z rnm-'") 25.lf0.9 1 l . l f 0.2 25.6f 1.3 8.5f 0.4

!-4GeV 1 GeV

/

4

50

tphotnn

40

..

3c

1 CI"

I ZOSV

Jc." ...

0 '00"

2c

I

1

.

,

.

,

.

3 4 6 b thickness J d (rnrn"

Figure 14. Test beam energy resolution for pions for the JLC Pb/scintillator configuration.

Figure 15. Pixelated photon detector being developed to readout the fibers from the JLC calorimeter scintillators. Figure from Y. Fujii.

68

transverse and longitudinal, is sufficient, especially for the ECal. To aid spatial separation, a shower maximum detector is foreseen, consisting of one x and one y layer of scintillating strips. Another challenge is the readout of the light from the scintillator, which is only 1 mm thick in the ECal. Hence, the photon detectors must have high gain, but be formatted to read out thousands of channels. Hence, the R&D for these devices is critical. Figure 15 shows a CCD-based device under investigation. 5.3. SD and LD

The SD detector being considered in N. America is a larger version of a detector first proposed at Snowmass 19961°. The calorimeter for the current SD detector is similar to the TESLA design. It has a Si/W ECal and a digital HCal. The ECal R&D is currently focussed" on the issue of how to integrate detectors and electronics. A single readout chip mounted on a silicon wafer of pads would effectively reduce the readout channel count by a factor lo3. This is sketched in Figure 16. A simple cooling scheme may be possible if one uses power cycling to reduce the heat load by a factor lo3.

-

-

n n

Figure 16. Center of a silicon detector wafer with 1000-channel readout chip bump bonded to the detector array. The detector pixels are hexagonal cells 5 mm across. A few representative signal traces are indicated.

Figure 17. A schematic of a section of a gas electron multiplier (GEM) being considered for the SD digital HCal. Figure from A. White.

The SD digital HCal effort is investigating three different detector options: RPCs, GEMS, and extruded scintillator tiles. The goal for each is to provide 1 cm segmentation at reasonable cost. The simple electronics possible with these technologies makes them compatible with the digital scheme. A GEM

69 schematic is shown in Figure 17. The LD calorimeter closely resembles the JLC detector. It exists as a configuration file for simulation studies, but currently no attempt has been made to provide a realistic design.

6. Prospects Calorimeters are being designed to take full advantage of the wonderful detector environment at an LC. The most important criteria are based on detectorwide measurements, such as jet reconstruction. This is an interesting time in this development: There are new ideas to implement, test, and compare with alternatives. The energy flow methods are not simple to test, since to work well they require a full set of pattern recognition and reconstruction tools. Test beam measurements will be required to compare with full simulation results. R&D is required to answer both basic and detailed questions. We do not know now what the new physics will be at the LC, but hopefully the choices being made now will improve the discovery reach!

Acknowledgements This work is supported in part by the US Department of Energy under award DE-FG02-96ER40969.

References 1. 2. 3. 4. 5. 6.

7. 8. 9. 10. 11.

S. Kunori, these proceedings. V. Morgunov, these proceedings. D. Green, these proceedings. H. Videau, Proc. Calor 2000, http://wwwlapp.in2p3.fr/Calor2OOO/. The TESLA TDR, http://tesla.desy.de/new-pages/TDR-CD/start.html, DESY, March 2001. The American Linear Collider Working Group, “Resource Book for Snowmass 2001, Part IV,” hep-ex/0106058, 2001. ACFA LC Working Group Report, http://acfahep.kek.jp/acfareport/. H. Videau, these proceedings. V. Korbel, these proceedings. C. Damerell, et al., pg. 431, and J. Brau, et al., pg. 437, Proc. Snowmass 1996, www.slac.stanford.edu. R. Frey, these proceedings.

CALORIMETRY DESIGN WITH ENERGY-FLOW CONCEPT (IMAGING DETECTOR FOR HIGH-ENERGY PHYSICS)

V. L. MORGUNOV DESY, TESLA-FLC, Notkestrasse 85, 0-22603Hamburg, Germany E-mail: [email protected] Permanent address: ITEP, B. Cheremushkinskaya, 25, Moscow 11 7218, Russia

The Energy-Flow method is based on the idea to replace, for the charged fraction of the event, the energy and angle information as derived by the calorimeter by the much more accurate particle momentum as measured in the powerful tracker device. The optimal application of this method requires establishment of sufficient spatial and energy resolution for all individual tracks to allow efficient shower separation in the boosted jets. Thus the calorimeter has to have good energy resolution and it has to be highly granular. The imaging detector capability, along with the use of the Combined Energy-Flow technique, allows the reconstruction of almost all individual particles in an event. Reconstruction for di-jets from 2’ decay using the Combined Energy-Flow method within TESLA detector results in a final mass resolution of better then 3 GeV.

1. Introduction

The Energy-Flow notation originally used to describe the energy flux inside the developing QCD processes in the physical chain: Partons ( e + e - + qq or W+W- or Zoh or so ...) + Fragmentation (hard processes =+ partons + strings) + Hadronization (soft processes + . . . sea quarks) + Particles at the interaction point ( B , D, ..., Ao, ..., no,n*,p , ... ). The backward problems for this chain have been more or less successfully solved by jet finder algorithms (Cone, JADE, Durham, Cambridge, kt, ...), by means of heavy flavour tagging and with the modern methods of event analysis, assuming that the four-momenta of all particles at the IP are known, and all vertex positions have been reconstructed. This concept is not the subject of this article. The LEP experiments introduced so called Energy-Flow techniques (methods) in 1994-1995ll2 to get better jet energy resolution by taking into account tracker information, partially replacing the calorimeter information. This resulted in reconstruction of “pseudo-particles” (E-flow objects). The recon-

+

70

71

struction of individual particles (charged or neutral) in LEP detectors was very difficult because of coarse calorimeter granularity , lack of or small magnetic field, lack of longitudinal segmentation, and additional dead material in front of or inside the calorimeter. The Energy-Flow technique has been improved by modern experiments such as CDF3, H I 4, ZEUS5, and it will also be used in CMS‘ and ATLAS7. The conventional calorimeters was tuned to get the best energy resolution for single particle and it does not work well with the number of particles (jet) that impact with such a calorimeter. The jet energy resolution of any existing detector is not better then 9 0 % / ~ 1 9 ~ 6if ~it 1is 0not used the Energy-Flow method; and it is near by 50-60%/&11J2J3>7 if used. The current evolution of this concept leads us to a new definition of EnergyFlow, different from the original concept. The next generation of HEP detectors and calorimeters can be built in such a way that it will be possible to reconstruct four-vectors of almost all particles (charged and neutral) in an event by means of a new Combined Energy-Flow method. An example of a new HEP detector14 is the imaging detector for the next generation of accelerators, that hopefully will be the e+e- Linear Collider (LC) at an energy region of ECM = 90 - 800 GeV. The TESLA detector has a typical collider detector architecture. Close to the beampipe a high precision vertex detector is installed. It is surrounded by tracking system consisting of Si detectors and large volume TPC. The electromagnetic and hadronic calorimeters sit outside the trackers. The complete assembly is immersed into a solenoidal magnetic field of 3-4 Tesla. The LC has “clean” events: there is no event overlapping - low event rate, clean environment - small background, possible constraint to the beam energy. The detector at an LC will not have an event trigger at all. So, the detector will work similar to a bubble chamber. A long time ago it was understood that the di-jets mass resolution is not a simple problem. The reason for this is the overlapping showers of particles in the jet. “Clearly how to handle the jet-jet resolution problem is one of the lessons being learned right now at LEP2 and is more dificult than most people imagined. This is a reason for the emphasis o n better energy-flow tool for LC detector. ” /Conceptual Design Report, 1997, ECFA-DESY/. Many physical tasks for the LC exist that require the best di-jet mass resolution, such as the mass reconstruction and separation for hadronic decay modes. A partial list of reference reactions for the World-Wide LC Study15 is shown here: 0 Measurement of MW from jets, without using the beam-energy con-

72

Tracker and Calorimeter Resolution in Absolute Scale

10

1

10

10

-1

-2

1

10

lo2

Energy (GeV) andhlomentum (Ge V/c). Figure 1. Th e main reason t o use tracker for event/jet energy measurement together with calorimeter is the very good momentum resolution - that is much better then calorimeter energy resolution in the wide energy region. T h e hadron calorimeter resolution of 50 32 %, and electromagnetic calorimeter of 10 % resolution are shown, together with two TPC momentum resolutions.

straint; 0 W / Z separation by mass in hadronic decay modes. 0 Measurement of Mt from jets, without using the beam-energy constraint; 4- and 6-jet reconstruction of e+e- + tf events. 0 Measurement in e+e- + Z h of branching ratios for h decay to bb, cC, T+T-, gg, (or WW and 22). Measurement of Higgs self coupling in e+e- + Zhh. 0 Measurement of t f h coupling cross section for e+e- -+ tfh. All of these reactions require di-jet mass and jet energy resolution of about 30% / fior better. Such a difficult requirements lead t o new reconstruction methods and in turn to the new requirements for the detector properties and to new detector design. The task of this article is to introduce the new Combined Energy-Flow concept and to explain a new path to detector design that takes into account new imaging reconstruction methods from the very beginning.

73

Figure 2. The Combined Energy-Flow technique for the imaging detector uses a priori knowledge for the showers separation and substitution.

2. Combined Energy-Flow, Main Idea

The main idea and resulting benefit of Combined Energy-Flow is based on two rather general and well known statements: 0 The well-measured particle momentum substitutes the nearly random energy spread in the imaging calorimeter volume (see Figure 1). This leads t o a decrease of the energy fluctuations in general. 0 Vector subtraction of overlapped showers is more effective in comparison with scalar subtraction due to the utilization of the momentum vector. The additional constraints decrease any fluctuations. Figure 2, which shows a sketch of two charged particles overlapping with one neutral, can be used to illustrate this statement. If the momenta of charged particles are known one can collect the detector hits along the predicted direction and substitute the collected cluster energy with the well measured momentum of its origin. When doing this, the vector character and the bending of the charged particles in the magnetic field is taken into account. The rest of the detector hits are attributed to and measure the energy of neutral particle - treated by the usual cluster algorithm. Such a technique requires an imaging reconstruction methods. This implies that the

74

calorimeter must be constructed as a high-granularity imaging calorimeter. The difference between an old LEP-Energy-Flow method and the new one is that the Combined Energy-Flow is not trying to reconstruct the jets but it aspires for the separation and substitution of all charged particle showers, using a priori knowledge of their momentum. The fine-granularity imaging calorimeter helps to reconstruct also the neutral part of the event carefully. Finally this allows to reconstruct every particle at the interaction point. After such a procedure any of the existing jet finders can be applied to analyze the event. The details and strict mathematical definition of the reconstruction problem can be found in the appendix of Snowmass2002 proceedings16. Now let us look at the promise of this concept.

3. Combined Energy-Flow, Energy Resolution Limit for the jet/event which has negligible showers overlapping: If every particle in the event can be distinguished separately: 60 % of all energy will be measured with the tracker precision; 30 % of the energy will be measured with the rather good resolution of the electromagnetic part of the imaging calorimeter; only about of 10% of the energy will be measured with the hadron calorimeter with relatively poor resolution. The numbers illustrating this are shown in Table 1. One can estimate the combined resolution as:

Table 1. Energy resolution Energy fraction for e+e- + hadrons ECM = 90 - 800 Gel/

Measured Tracker ECAL HCAL

Charged particles Gammas Neutral hadrons

Ech.hadr M

Energy resolution

0.6 ' E

ffch.hadr

ET x 0.3. E Eneut.hadr

ffcomb = Jff%.hadr

ff7

0.1 ' E

un.hadr

10-4E2h.pa~t

x 0.11-

o . 4 G

+ ff; + ffieut.hadr

After substitution one has: ffcomb =

J(0.6)4

X

+

(10-4)2E4 0.3

X

+

(0.11)2E 0.1

X

(0.4)2E

with the lower limit of the energy resolution as: !%Z@ x

E

J ( E x 3.6 x 10-5)2 + (0.14/&)

2

+

0.14/&

75

A more general formula for the energy resolution, taking into account the constant terms U y = 0.12 0.005 E y ; Uneut.hadr = 0.4 0.05 Eneut.hadr leads to the same limit numerically:

JC

+

z)2(

+(

ucomb

(3.6 10-5 x E ) ~

E

+

T)2 0.038

+ (0.005)2 + 0.14/&

Thus, the lowest limit of energy resolution for a jet or whole event is about 14% so far unreachable by any calorimeter.

a,

35

...............................................

30

......,.............................................

25

......,.

..........

I , , , I

20 15

...... .........

10

......,. ........

I

-

.

.-

5 0

8..

...... - - - - - - - -,- -

0

:

a?,

20

40

60

80 100 I20 ZO mass (GeV)

Figure 3. Full simulation result for TESLA detector with SNARK reconstruction program. (2' mass width for this sample was set t o 100 MeV, to show the detector resolution only).

Figure 3 shows the mass resolution for ZO at its resonant energy; in this case the mass and energy of the event is the same parameter. The distribution in Figure 3 is not the scalar energy sum; but it is the sum of reconstructed particle four-vectors at the IP. The shape of the distribution is not Gaussian due t o the mixing of many different event geometries, with very different degree of shower overlap. In some events all particle showers were reconstructed separately and the energy resolution approaches the theoretical limit. That is visible as the narrow peak over the Gaussian distribution with an effective sigma of about 30 % over

a.

76

4. Combined Energy-Flow, Reconstruction Algorithm

Let us look in a bit more detail, but briefly a , at the reconstruction algorithm for a better understanding of the requirements for the imaging detector design that follows from the Combined Energy-Flow concept. The cluster search procedure for charged tracks in the imaging calorimeter starts from the predicted point and direction of the track at the imaging calorimeter “face” (see Figure 4).

HCAL

Figure 4.

Combined Energy-Flow algorithm in the reconstruction program SNARK.

For example, one builds a helix from the impact point at the imaging calorimeter face, and collect the hits along this curve - track core. Then one can build the first hypothesis for the collected cluster. If all hit’s amplitudes are near the MIP value so one can define the track as a muon. In this case one has to stop the collection, calculate properties and probability and remove hits from further analysis. If it is not a muon, one can check the next hypothesis - an electron. Again as an example, for the particular ECAL structure of a Si-W imaging calorimeter with 1 by 1 cm cell sizes, half of an input energy is collected in the cylinder of one cell size radius around the helix prediction. This can be shown in the separate simulation of such electromagnetic showers. Then one can take the next shell around the core - and so on and so on, up to the track energy and taking into account the ECAL energy resolution. At each step the 3-d ~

aThe algorithm for hypothesis building includes 35 logical branches for now.

77

electromagnetic hypothesis is built and compared with the collected cluster shape. The collection stops when the collected energy and cluster shape are agree with the prediction. If the collected cluster is not valid for the muon or electron hypotheses (both of them are much better defined in comparison with the hadron one) it is classified as a hadron. The fitting and collecting procedure for the hadron cluster is close to the electromagnetic one, however the expected shower shape and fluctuations are taken from a simulation of appropriate hadronic shower. All collected hits with well defined hypothesis are excluded from further examination. The cases of the overlapped showers treated at the separate algorithm branches with some additional conditions and bit collection procedures that include hit weighting depends on the input particle energy. One hit after such a procedure can be member of different clusters, or even it can go to the residual of the hits, that means it will belongs to some neutral cluster partially. The good quality building of the hypothesis for the collected cluster requires the high granularity mostly for the the electromagnetic part of imaging calorimeter. The rest of the hits after the charged track procedure is finished may go to the more or less usual clustering procedure that should resolve the overlapped gamma-showers. Again this leads to the fine granularity of the the electromagnetic part of imaging calorimeter. 5. From IP to Calorimeter in detail We will start from the requirements for the tracker, as it plays a significant role in the Combined Energy-Flow technique. Particles are propagated from the IP to the imaging calorimeter face through the relatively small amount of matter in a large magnetic field with possible decay, nuclear interaction, or gamma conversion. Charged tracks with Pt < 0.76 GeV/c in a 3 Tesla magnetic field will never reach the imaging calorimeter barrel surface. The big amount of low energy curling tracks in event leads to additional requirements for the tracker reconstruction procedure. The bending of charged tracks in the large magnetic field (?~p # ?fa,,) leads to a significant shift of the position of the jet center of gravity in the calorimeter volume in comparison with zero or small magnetic field. It leads in turn to the deterioration of jet energy and di-jet mass resolution because of the angular dependences. Any tracker reconstruction error leads t o the bad reconstruction of the particular shower and rather often to the hits doublecounting.

78

+ t€

& = 500 GeV, Process e'e-

................... . . .

....................................

0 2 4 6 8 101214161820

-+y + + y + q. a.

Process e+ e'

ZO

I00 80 60 40 20 0

.

.. I

,

.

,

,

I

.

,

,

. ,

I . , I......... . ,..... ................... .. .. .. .. ..

,

,

,

.

.

,

0 2 4 6 8 101214161820

Distance at the face (cm) Figure 5. Distribution of distances between particle (charged and neutral) impact points at the imaging calorimeter surface depends on the physical process - width of the jet cone.

Requirements for tracker and reconstruction:

*

The tracker should reconstruct the point where the particle impacts the imaging calorimeter volume with an accuracy better then electromagnetic calorimeter cell size in order to get the best matching with reconstructed shower image. The tracker should treat low energy curling tracks carefully. The tracker reconstruction procedure should include backward propagation of all tracks to the IP including particle decays and gamma conversion vertex treatment t o get the correct particle four-momentum and composition at the interaction point.

+

*

Requirements for absorber material and sampling structure: Electromagnetic part:

The distribution of distances between impact points at the imaging calorimeter face is shown in Figure 5. The distances are of order centimeters for the

79

boosted jets. + The ECAL sampling structure should have the smallest Moliere radius one can reach to make compact separated electromagnetic showers in the electromagnetic part of the imaging calorimeter. + Large longitudinal segmentation is needed to allow the application of the Combined Energy-Flow method to build high quality three-dimensional shower hypotheses for different types of particles.

Hadronic part: These showers are considerably broader than electromagnetic ones, and they have big fluctuation in shape and in the energy deposited in the imaging calorimeter volume, including fundamentally undetectable energy losses due to the nuclear binding energy. The typical size of a hadronic shower is 2-3 hint.which is of order tens of centimeters. It is impossible to prevent the hadron showers from overlapping in the HCAL volume for boosted jets. The distance between particles in such jets (see Figure 5) is much less than the calorimeter interaction length for any absorber material. The requirements will come later.

Requirements on the electronic cell sizes The shower develops in the imaging calorimeter volume as a 3-d space object, so we will not separate it into transverse or longitudinal projections. The readout cell does not necessarily corresponds to an individual physical cell of the detector. One electronic cell can collect information from one or several physical volumes. Any answer for the question of cell size optimization requires a huge amount of computer simulation and reconstruction investigation for each particular variant or imaging detector option. The cell sizes are dictated by physical processes in the calorimeter volume. They will be quite different for the electromagnetic and hadronic parts of the imaging detector due t o different physics in both of them. The building of the particle type hypothesis for each cluster is important for the definition of cell size. The electromagnetic showers in the calorimeter volume are significantly different in shape. Nevertheless it allows one to apply a fit to the shapes and to get a good estimation of the probability to distinguish these showers between muon and hadron hypotheses.

Electromagnetic part: Two shower topologies make appearance most often in the electromagnetic

80

part of the calorimeter. They influence to the definition of cell sizes. 1. Two close or overlapped gamma-showers from T O decay. Such gammashowers are quasi-parallel due to the large distance from IP, so the separation has to be done along all shower length. Two electromagnetic showers that do not overlap more than half of their width (a1 x 0 ) can be reconstructed separately if the transverse shower shape can be divided into more than four subdivisions. To fit the electromagnetic shower in 3-d space one needs more then 10 points in the longitudinal direction and 4-6 points in transverse at every layer. + ECAL granularity should be close to the 1 X O value for the particular sampling structure density. 2. Overlapping hadron or muon tracks with a gamma-shower. The track should be resolved in the shower background. The direction of the track can be quit different from the shower direction due to the bending in the big magnetic field. + This increases the significance of the longitudinal segmentation in comparison with that in the first case. Hadronic part:

The hadron shower spatial variations are much larger than the electromagnetic ones because more processes are involved in their development. Such a behaviour does not allow the application of any kind of hypothesis to that shower (at the level of different hadron type). For example, to distinguish a neutral hadron from a charged one. The charged hadrons easily run 1-2-3 A0 though the calorimeter before the first nuclear interaction - that is a rather big distance. The track bends at such a distance in a big magnetic field and the cluster that is created after the first hadron interaction can be found along this helix prediction with reasonable accuracy.b The hadron transverse shower shape density has three components (see Figure 6 ) . 0 The first is the primary ionization track. 0 The second is the narrow core that is mostly made of the electromagnetic part of the hadron cascade. 0 The third is the pure hadron tail around the axis with an exponential behaviour and M Ai,t. slope. ‘ bThe same effect of bending shifts the cluster center if one use the “old” reconstruction method of cluster searching based on angular space clustering or cone algorithm. cWe will not discuss the neutron component of the shower because the picture was made for the Fe-Sc structure that has a negligible neutron component.

81

+ The cell size for hadron part of the imaging calorimeter should be close to the size of the transverse hadron shower core for the first few interaction lengths to collect the shower core along the predicted trajectory. The cell sizes should be close to the 1XOvalue for the particular sampling structure to follow this requirement. + The HCAL sampling structure should have the smallest XOvalue t o make better shower separation in the hadronic part of imaging calorimeter. . ,

.

,

. .. . . . ..

.

.

..

...... ~ ~ ..... . ~ . .~

.

..

..

..

..

.

.

.

.

.

.

.

..

.

.

.

.. .. , .. . , . , .. ... .. .

.. .. .. ., . . ... ,. ......... , . . . .. ... .. . . .. .. .. . . ,

.

..

..

..

.

.

0 10 20 30 40 50 60 70 80 Distance from shower axes (cm) Figure 6. The radial distribution of hadronic shower energy density for particles that have MIP in the electromagnetic part of the imaging calorimeter.

.

.

.

n 0 5 10 1.7 20 25 30 3.7 40 45 SO

P momentum (GeV) Figure 7. T he integral energy fraction that carried by the particles with momentum P (in percent/0.5GeV) for e+e- =$ tf events at 500 GeV E C M .

6. Compensation and Combined Energy-Flow Requirements on the imaging calorimeter energy resolution: The first order of significance for the Combined Energy-Flow method is the shower separation in space and reconstruction of an individual shower correctly. The energy collected along the predicted shower direction is used in the substitution procedure. So, the energy resolution plays an important role during the reconstruction, but not the first one. The usage of the tracker and ECAL for measurement of about 90% of the energy allows one t o relax the requirements on the hadron part of the imaging calorimeter energy resolution. =+ The energy resolution of the ECAL mostly depends on the mass resolution for the pure electromagnetic processes at the parton level (e+e- + Z o H o + qQ^Iy) that we do not discuss here. It should not be worse then 12-13%/@ (see TDR14).

82

The dependence of the integral energy fraction carried by particles with momentum P is shown at the Figure 7. From this picture one can see that the biggest part of the energy in an event is carried by low-energy particles (80 % of energy by particles with momentum less then 30 GeV/c). It leads to less significance of the constant term and more significance of the sampling fluctuations. The hadron calorimeter energy resolution should be as good as sampling fluctuations allow in intermediate energy region. When we are talking about sampling fluctuations then we should keep in mind that both the HCAL and ECAL sampling structures should keep the smallest value of X O , as shown above.

+

Compensation:

The imaging calorimeter energy resolution is a second order parameter in the Combined Energy-Flow reconstruction technique. If this is so, calorimeter compensation and the constant term of the energy resolution come in as third order of significance in the reconstruction procedure, and in its influence on the final result - that is the jet energy and di-jet mass resolution. To prove this statement let us look at the ZEUS detector as an example: ZEUS has a very good quality, well compensated calorimeter with maybe the best hadron energy resolution, but the jet energy resolution in the ZEUS detector is bad in comparison with hadron energy resolution". In the reconstruction procedure ZEUS applied the Energy-Flow technique but, a rather small magnetic field and coarse granulated electromagnetic and hadron calorimeter do not allow this technique to achieve a good jet energy resolution.d The same situation is true for the all LEP detectors mostly due to bad calorimeter segmentation. The imaging detector (if we follow the Combined Energy-Flow requirements) will have two separated parts, electromagnetic and hadronic, with different absorbers and sampling structures, so it will be a non-compensated calorimeter by default. An intermediate atomic number absorber material (Fe) has less neutron yield of the hadronic cascade, leading to smaller overlap of showers in the case -

d The volume of hadronic part of the calorimeter, where compensation of the binding fluctuation occurred (by neutrons signal registration) significantly larger ( w one order of magnitude) than the main part of the hadron shower energy deposition, so this properly leads to the deterioration of the energy resolution f o r the overlapped hadron showers. This is might be one of the reasons why compensation did not work in the case of jet energy measurement in old detectors.

84

4.

5.

6.

7.

8. 9. 10. 11. 12.

13. 14. 15. 16.

A. Bhatti et al., (CDF Collaboration), Review of Jet Clustering a t Tevatron, talk at CALOR2000 Conference, Annecy (2000). S. Aid et al., (H1 Collaboration), A Measurement and QCD Analysis of the Proton Structure Function &(2, Q 2 )a t HERA, Nucl. Phys. B470, 3 (1996). C. Adloff et al., (H1 Collaboration), Diffraction Dissociation in Photoproduction a t HERA, Z. Phys. C74, 221 (1997). C. Adloff et al., (H1 Collaboration), Measurement ofNeutral and Charged Current Cross-Section in Positron-Proton Collisions a t Large Momentum Transfer, Eur. Phys. J. C13, 609-639 (2000). C. Issever et al., (H1 Collaboration), The calibration of the HI liquid Argon calorimeter, Proceedings of the IX CALOR2000 Conference, Annecy , 603-608 (2000). G. Briskin, Diffractive Dissociation in ep DIS, Ph.D. Thesis (1998). J. Breitweg et al., (ZEUS Collaboration), Eur. J. Phys. C1, 81 (1998). S. Chekanov and S. Magill (ZEUS Collaboration), Jet Energy Corrections with the ZEUS Barrel PREshower Detector, talk at CALOR2000 Conference, Annecy (2000). M. Wing (ZEUS Collaboration), the ZEUS Detector, talk at CALOR2000 Conference, Annecy (2000). J. Damgov, L. Litov, Application of neural networks for energy resolution, NIM A 482, 776-778 (2002). S. Kunori (CMS Collaboration), Jet Energy Reconstruction with the CMS Detector, talk at CALOR2002 Conference, Pasadena, (2002). M. Wielers (ATLAS Collaboration), Performance of Jets and missing ET in ATLAS, talk at CALOR2002 Conference, Pasadena, (2002). N. Isamu (OPAL Collaboration), Jet Measurement a t OPAL, talk at CALOR2002 Conference, Pasadena, (2002). A. Kiiskien (DELPHI Collaboration), Development on Jet Reconstruction by DELPHI, talk at CALOR2002 Conference, Pasadena, (2002). D. Foumier, Overview of Calorimetry at LHC, talk at CALOR2002 Conference, Pasadena, (2002). ZEUS Collaboration, Search for Resonances Decaying to e+ - j e t in e+p Interactions a t HERA, preprint, DESY 00-023, (2000). M. Minard (ALEPH Collaboration), Jet energy Measurement with the ALEPH detector a t LEP2, talk at CALOR2002 Conference, Pasadena, (2002). S. Dell’Agnello (CDF2 Collaboration), CDF2 Integrated Calorimetry Environment, talk at CALOR2002 Conference, Pasadena, (2002). TESLA Technical Design Report, DESY March 2001, http://tesla.desy.de/new-pages/TDR-CD/start .html F. R. K. h j i i , M. Peskin, Reference Reactions for the World-Wide LC Study, (2001), http://www.slac.stanford.edu/ mpeskin/LC/refrxns.html. V. Morgunov, Energy-ffow method for multy-jet effective mass reconstruction in the highly granulated TESLA calorimeter, Snowmass2001 proceedings (2001), http://www.slac.stanford.edu/econf/COlO63O/forweb/E3l5~m0rgunov.pdf.

Calorimetry in Astrophysics Covener: T. Parnell

T. Parnell

Covener’s Report

J. Isbert

ATIC, a Balloon Borne Calorimeter for Cosmic Ray Measurements

T . Wilson

ATIC Backscatter Study Using Monte Carlo Methods in FLUKA & ROOT

V. Bonvicini

A Silicon-Tungsten Calorimeter for Cosmic-Ray Physics

R. Kossakowski

Electromagnetic Calorimeter for the AMS-02 Experiment

P. Maestro

Performances of the AMS-02 Electromagnetic Calorimeter

A. Chekhtman

The Status of GLAST CsI Calorimeter

R. Terrier

Performance of GLAST Calorimeter

0. Ganel

Cosmic Ray Energetics And Mass (CREAM): Calibrating a Cosmic Ray Calorimeter

F. Krennrich

VERITAS: A Next Generation Atmospheric Cherenkov Detector and Calorimeter for Gamma-Ray Astronomy

A. K. Tripathi

Pierre Auger Observatory: The World’s Largest Calorimeter

*K. Arisaka

EUSO and OWL: Atmospheric Cosmic Ray Calorimetry from Space

*J. Lamoureux

Calorimetry (GeV-EeV) in AMANDA and IceCube Neutrino Telescopes

*Written contribution not received

85

This page intentionally left blank

ASTROPHYSICS

T. PARNELL U A H and Marshall Space Flight Center, N A S A E-mail: [email protected] (Convener’s Report)

The applications of calorimetry in astrophysics cover an enormous energy range from cryogenic bolometers for 1 keV X-rays (not covered in this session) to methods that instrument large masses of Earth’s atmosphere in order t o perform measurement of cosmic particles above 10l8 eV. Energy ranges, low and often isotropic fluxes, kinds of particles to be measured and economic considerations constrain the calorimetry methods that can be employed. The papers in this session provide a sample of scientific objectives, calorimetry techniques and instrument development challenges in astrophysics. The first set of papers in this session discuss calorimetry applications for direct measurements on cosmic gamma rays, electrons and nuclei. They use more traditional particle calorimetry techniques to measure the energy along with other detectors to directly identify the primary particle. These detector systems are designed for exposure on large balloons or on spacecraft. The Advanced Thin Ionization Calorimeter (ATIC) , with a bismuth germanate calorimeter has been flown around the South Pole on a balloon t o measure the energy spectra of cosmic ray nuclei (up to Fe) over the range of 0.1 to 100 TeV. Next a silicon and tungsten calorimeter for the satellite-borne PAMELA experiment is described. It is intended to measure the spectra of electrons, positrons and light nuclei up to 0.7 TeV. Another paper describes the lead-scintillating fiber electromagnetic calorimeter to be employed in the Alpha Magnet Spectrometer (AMS). It will be flown on the International Space Station for several years t o measure the spectra of cosmic ray hadrons and electrons and search for anti-nuclei (He and heavier). A CsI calorimeter for the Gamma Ray Large Area Space Telescope (GLAST) is also described. It will extend the energy range of an electron pair telescope (which employs silicon strips and tungsten foils) t o several hundred GeV. GLAST will continue the study of high energy discrete sources such as BLAZARS (active galaxies with jets). This set of “direct” instruments concludes with the calorimeter for the Cosmic Ray Energy and

87

88

Mass (CREAM) instrument to be flown on an “ultra-long duration” balloon flight. Along with a transition radiation detector, the tungsten-scintillating fiber calorimeter will be used to measure the spectra of cosmic ray nuclei to above 1014 eV. All these calorimeters are very thin (about two proton mfp), because the low fluxes demand a large area, but the flight vehicles limit the mass. Thus they cannot compete in energy resolution with accelerator experiment calorimeters. The second sample of calorimeters presented in this session use Earth’s atmosphere or the Antarctic polar ice cap for the calorimeter’s mass. The Whipple Observatory employs the atmospheric Cerenkov imaging technique for studying the gamma ray spectra of discrete objects such as pulsars and BLAZARS above 100 GeV. The instrument uses large mirrors with arrays of photomultipliers at the focus to track the candidate sources across the night sky. The tracking and characteristics of the imaged Cerenkov light from the atmospheric showers allow discrimination against the showers produced by nuclei. The Very Energetic Radiation Imaging Telescope Array System (VERITAS) will be more sensitive with better background rejection, and will extend down to 10 GeV, allowing comparison with future GLAST observations. The Pierre Auger Observatory (PAO) described next will provide a significant advance in cosmic ray air shower experiments which use detectors on the ground. Its large area of 3,000 square kilometers will extend the energy range above lo1’ eV and the use of 1,600 water Cerenkov counters and four “fly’s eye” atmospheric fluorescence detectors will provide important data to identify the primary particles. Not published here, but useful in describing the range of astrophysics calorimetry techniques are the AMANDA/Ice Cube and the EUSO/OWL experiments which were presented. AMANDA uses strings of photomultiplier tubes frozen deep in the Antarctic ice cap to provide a sensitive detector for high energy neutrinos which might be produced in hadronic processes in sources of cosmic rays or y-ray bursts. The Extreme Universe Space Observatory (EUSO) and the Orbiting Wide-angle Light-collector (OWL) are proposed instruments to observe about a million square kilometers of the atmosphere from space, and to measure the moving discs of nitrogen fluorescence produced by large air showers. The shower profiles would be used to identify the primary particles and to measure the cosmic ray spectra above lo2’ eV. Also, in the Cerenkov calorimetry session of this conference are two papers concerning the measurement of radio Cerenkov radiation that have potential application in astrophysics.

ATIC, A BALLOON B O R N E CALORIMETER FOR COSMIC RAY MEASUREMENTS

J . ISBERT, G. CASE, D. GRANGER, T.G. GUZIK, B. PRICE, M. STEWART, J.P. WEFEL Louisiana State University, Dept. of Physics & Astronomy Baton Rouge LA 70803-4001, USA E-mail: isbert%phunds.phys.lsu.edu

J.H. ADAMS, M. CHRISTL NASA/Marshall Space Flight Center, Huntsville, A L

H.S. AHN, 0. GANEL, K.C. KIM, S.A. NAQVI, E.S. SEO, R. SINA, J.Z. WANG, J. WU University of Maryland, College Park, MD, USA

A.R. FAZELY, R. GUNASINGHA Southern University, Baton Rouge, LA, USA

Y.J. HAN,H.J. KIM, S.K. KIM Seoul National University, Seoul, Korea

G. BASHINDZHAGYAN, E. KOUZNETSOV, M. PANASYUK, A. PANOV, G. SAMSONOV, N. SOKOLSKAYA, A. VORONIN, V. ZATSEPIN Lomonosov Moscow State University, Moscow, Russia

J. CHANG, W.K.H. SCHMIDT Max Planck-Institute fuer Aeronomie, Katlenburg-Landau, Germany

ATIC (Advanced Thin Ionization Calorimeter) is a balloon borne experiment designed to measure the Cosmic Ray composition for elements from hydrogen to nickel and their energy spectra from 50 GeV to near 100 TeV. It consists of a Simatrix detector to determine the charge of a CR particle, a scintillator hodoscope for tracking, carbon interaction targets and a fully active BGO calorimeter. ATIC had its first flight from Mcmurdo, Antarctica from 28/12/2000 to 13/01/2001, local time, recording over 360 hours of data. The constraints, the design and the operation of this balloon borne instrument are described.

89

90

1. The ATIC design ATIC is designed primarily to measure cosmic ray spectra for elements from hydrogen t o nickel. Balloon born instruments face severe limitations in terms of size (volume, shape), weight, power and communication. 1.1. Limitations The most severe for complex instruments utilizing calorimeters are weight and communication. The maximum weight limit is determined by the size of the available balloon. For the ATIC instrument this limit was set by a 28 million cubic feet helium balloon to 32001bs. In addition to lifting the science instrument the balloon also has to carry the balloon "craft" consisting of communication electronics, a parachute and termination packages. One termination package severs the balloon from the instrument at the end of the flight, the second termination package severs the parachute from the instrument after landing. The limits for communications is set by its route and location. The fastest link at about 330kbaud is direct line-of-sight VHF, which is limited t o about 600 miles. For distances outside this area satellites are used at a rate of about 4kbaud. ATIC generates data at about the rate of the line-of-sight link. 1.2. The ATIC instrument The most efficient way to determine energy utilizing a calorimeter for a given weight is ionization calorimetry. This technique utilizes the interactions of cosmic ray nuclei in a low Z target. The resulting secondary pions decay into high energy photons which then start an EM shower in the calorimeter. This is used as the energy measurement for the incident nucleus. In order to determine the energy spectra for each cosmic ray species the following quantities need to be measured:

(1) The charge of the cosmic ray particle, (2) Its energy and (3) The abundance of each species. The ATIC balloon instrument is composed of two major subsystems: the target module and the calorimeter. The target module has 4 functions:

(1) Provide the low Z target for the cosmic ray nuclei to interact (2) Determine the charge of the cosmic ray nuclei, (3) Provide a trigger for the instrument

91

(4) Provide tracking in combination with the calorimeter. The calorimeter has 2 functions: (1) Determine the energy of the cosmic ray, (2) Provide tracking in combination with the target module. 1.3. The Target Module

Figure 1 shows a cross section of the ATIC instrument. The target module consists of (from top to bottom): a silicon matrix detector array, three plastic scintillator XY planes, the first (Sl) placed directly below the silicon matrix detector, followed by lOcm of carbon target, the 2nd scintillator XY plane (S2) and 20cm of carbon target and then the 3Td scintillator XY plane (S3).

\

1

Silicon Detector Si

3

-

s1

-s2

Figure 1. Cross section of the ATIC Instrument.

A potential problem for charge determination in the presence of calorimeters are particles back scattered from the shower into the detectors above once the energy of the cosmic ray exceeds a few TeV. Simulations of high energy protons in the ATIC experiment indicate that, indeed, as the proton energy increases the number of "back-splash" particles per unit area increases in all three

92

scintillator planes as well as the silicon matrix detector, potentially adding to the charge signal and degrading the ability to distinguish between Protons and Helium. To combat this effect both charge detectors are segmented, thereby decreasing the probability of back scattered particles passing through the same detector element as the cosmic ray. The Silicon matrix is the primary charge detector, supplemented by the topmost scintillator plane. The Silicon matrix detector consists of 4480 silicon pixels, 1.5 x 2.0 cm in size (Adamsl), covering an area slightly larger than the aperture defined by the trigger scintillators. The individual pixels are read out with a specially designed low power application specific integrated circuit (ASIC), the CR1.4 consuming only 6.6 mW per channel. A chip consists of 16 channels. The analog output of this chip is digitized with a 16 bit ADC, thereby covering the charge range from protons to nickel in a single ADC range. Each of the three scintillator detectors consists of two planes composed of individual strips of plastic scintillator rotated 90" to each other. The individual strips are viewed by photomultiplier tubes at each end. Each photomultiplier is read out by two channels of a custom designed low power ASIC, consisting of 16 channels of a preamplifier, ADC and two discriminator outputs for the trigger logic. Each channel consumes only 14mW of power. The front end electronics containing these ASICs and their support electronics is mounted directly behind the photomultipliers. defining the active aperture of the instrument

1.4. The Calorimeter The calorimeter module at the bottom of the instrument consists of a "package" of 320 BGO crystals, each 2.5 cm by 2.5 cm by 25 cm in size and placed into 8 trays of 40 crystals each, covering an active area of 51 x 51 cm2. Alternating layers are rotated 90" relative to each other to form 4 X and 4 Y layers. Each BGO crystal is wrapped in teflon tape and covered with aluminum coated mylar foil for light tightness and viewed by a single photomultiplier tube, a Hamamatsu R5611-01. A light attenuator is placed on the front face of each photomultiplier to keep the readout linear over the entire signal range. The PMTs and the required electronics is mounted in a box directly to two opposing sides of each tray. For calibration and liveliness tests a light emitting diode is mounted onto the side of each PMT. To measure particle energies up to 100 TeV the readout device has to cover energy deposits from about 5 MeV to over 10 TeV in each individual crystal. To readout the high dynamic range of the signals in a single crystal of 2x106 the readout of the photomultiplier tube is split into three gain ranges. This

93

is accomplished by utilizing three dynode pickoffs each read out by its own preamplifier channel. The digitization is accomplished utilizing the same ASIC used for the scintillators. The absolute energy calibration of the calorimeter is done by first calibrating the highest gain range of every BGO crystal with cosmic ray muons. Then the higher energy ranges (=lower gain) are calibrated using the overlap between ranges. The slope in the linear region determines the ratio of low E range to medium E range. The ratio of the high E range to the medium range is determined using the overlap region of these ranges. Utilizing this bootstrapping calibration mechanism, it is possible to determine the deposited energy of particle showers in the fully active BGO calorimeter very accurately. 2. The ATIC Trigger

The ATIC experiment trigger needs to fulfill two requirements:

(1) It must trigger on potential events which pass through the active aperture of ATIC, and (2) It has to set an energy threshold Requirement 1 is fulfilled by deriving a trigger from the top (Sl) and bottom (S3) scintillator of the target module. Requirement 2 is fulfilled by utilizing discriminators on the BGO crystals and form an energy dependant trigger signal. These two triggers are formed separately due to their different timing characteristics. The plastic scintillator hodoscope form a pre-trigger derived from the topmost scintillator plane (Sl) in coincidence with the scintillator plane below the carbon target (S3). This pre-trigger (PT) also locks the preamplifier outputs for potential readout. The energy dependant trigger, the master trigger (MT) is formed from the discriminator outputs of the BGO crystal readout preamplifiers. If both triggers fire for the same particle the pulse heights are digitized and read out.

3. Read out and control The data are stored on a hard disk of one of the computers mounted on the instrument as well as transmitted through the line-of-sight link. A subset of the data as well as housekeeping data such as temperatures, pressures, position, altitude and remaining disk capacity is transmitted through all links to for monitoring during the entire length of the flight. The instrument is controlled via commands from the ground through the same links.

94

4. Summary

The ATIC instrument had a very successful long duration balloon flight from Mcmurdo, Antarctica. All detectors and their support systems worked well for the entire flight. ATIC is currently (June 2002) prepared for a 2nd flight from Mcmurdo, Antarctica in December 2002 / January 2003. References 1. Adams, J. H. et al., the ATIC Collaboration, Proc. 26th Int. Cosmic Ray Conf. (Salt Lake City), 5, 69 and 76, 1999. 2. Ganel, 0. et al., The ATIC Collaboration, Proc. 26th Int. Cosmic Ray Conf.(Salt Lake City), 5, 453, 1999. 3. Ganel, 0. et al., The ATIC Collaboration, Adv. I n Space Research, in press, 2001. 4. Guzik, T. G. et al., the ATIC Collaboration, SPIE International Symposium on Optical Science, Engineering, and Instrumentation, Denver, CO, 2806, 122, 1996.

5. Guzik, T. G. et al., the ATIC Collaboration, Proc. 26th Int. Cosmic Ray Conf. (Salt Lake City), 5, 6,1999. 6. Seo, E. S. et al., the ATIC Collaboration, SPIE International Symposium on Optical Science, Engineering, and Instrumentation, Denver, CO, 2806, 134, 1996. 7. Seo, E. S. et al., the ATIC Collaboration, Advances in Space Research, 19, No. 5, 711, 1997.

ATIC BACKSCATTER STUDY USING MONTE CARL0 METHODS IN FLUKA & ROOT

T. WILSON NASA, Johnson Space Center, Houston, Texas 77058, USA E-mail: twilsonOems.jsc.nasa.gov

L. PINSKY, A. EMPL, K. LEE, V. ANDERSEN University of Houston, Department of Physics, Houston, Texas 77204, USA

J. ISBERT, J. WEFEL Louisiana State University, Department of Physics and Astronomy Baton Rouge, Louisiana 70803, USA

F. CARMINATI, A. FASSO, A. FERRARI", P. SALA*, E. FUTO CERN, 1211 Geneva, Switzerland

J. RANFT Universiat Siegen, Fuchbereich Physik, 0-57068, Siegen, Germany

A Monte Carlo analysis, based upon FLUKA, of neutron backscatter albedoes is presented using the ATIC balloon experiment as a study case. Preparation of the FLUKA input geometry has been accomplished by means of a new semi-automatic procedure for converting GEANT3 simulations. Resultant particle fluences (neutrons, photons, and charged particles) produced by incident Carbon nuclei striking ATIC with energies up to 1 TeV/A are discussed. The analysis is part of a broad goal of simulating space radiation transport in materials science by means of the FLUKA code in conjunction with a ROOT-based interface.

1. Introduction The present study is the outcome of an ongoing investigation into nextgeneration Monte Carlo techniques for analyzing radiation transport in materials science. This is motivated by the need for a comprehensive understanding of mitigation measures for radiation protection in nuclear physics, particle * Permanent address: INFN and Dipartimento di Fisica dell' Universita, 20133 Milano, Italy.

95

96

accelerator physics, and space exploration. The tools involved derive from sophisticated physics transport codes and prove to be invaluable in the design of particle detectors as well as cosmic-ray payloads in high-energy astrophysics, an application that will be the focus of our discussion here. The particular name for this investigation is FLEUR (Fluka Executing Under Root, website: http://fleur.cern.ch/-project/links.html). The Monte Carlo of choice is FLUKA’ which is used to simulate heavy-ion interactions at energies above a few GeV/A utilizing the newly interfaced DPMJET 11.5 event generator (although no heavy-ion interaction model is present at low energies even though transport and energy loss are simulated). FLUKA also simulates all other known particle interactions. The graphical interface will eventually be the object-oriented (00)software called ROOT2 developed at CERN, a combination which is an adaptation of the AliROOT off-line system architecture shown in Figure 1.

Figure 1. Illustration of the virtual Monte Carlo concept in ALICE’S AliROOT off-line system design, showing the transport engines FLUKA and GEANT with differing geometry databases.

2. Benchmarking and Testing FLUKA

FLUKA (http://www.fluka.org/) is a fully integrated particle physics Monte Carlo simulation package that is particularly suited for backscatter problems.

97

It has many other applications in high-energy experimental physics and engineering, shielding, detector and telescope design, cosmic-ray studies, dosimetry, medical physics, and radiobiology. In order to test and validate the predictive power of FLUKA, recent studies include an analysis of the particle backscatter albedo of the Earth’s atmosphere as measured by AMS3 (Alpha Magnetic Spectrometer) , and the Earth’s atmospheric neutrino flux4. Additional benchmark presentations and references are available at the FLUKA website. 3. The Geometry Conversion Technique

A roadblock in Figure 1 has been the inability of the GEANTS community to convert their rich heritage of G3 detector geometries directly into a FLUKA input file, except via G4. Such a technique would save considerable time, and the present study illustrates how to do this. Conversion of a complex geometry from GEANT3 to FLUKA is not a trivial step because of the fundamentally different input method adopted by the two codes, a feature that seems to be one reason GEANT3 users avoid using FLUKA. A procedure has been designed to extract the geometry and related features from an existing GEANT3 simulation in double precision. The extracted information is then transformed into an equivalent FLUKA input by a program called flex4 (FLUKA for ex-GEANT3 users). The GEANT3 concept of detector sets is retained and can be applied to the (standard) FLUKA output or used in FLUKA user routines. Only a subset of the GEANTS shapes is implemented at this point. The detailed conversion technique will be published elsewhere5 and a tutorial will be provided at the FLEUR website. 4. Brief Description of the Instrument Geometry Studied

For a realistic assessment of the converter, we selected the ATIC6 balloon payload geometry. This experiment provides another opportunity for examining a cosmic-ray instrument design using the current version of the DPMJETinterfaced FLUKA. Although this paper is not benchmarking the code against ATIC flight data, it does highlight the possibilities and the advantages of having a physics model that fully understands the intrument’s performance. ATIC is a cosmic-ray astrophysics collaboration led by Louisiana State University, involving the University of Maryland, Marshall Space Flight Center, Southern University, and a number of international partners. ATIC is designed to look at the cosmic-ray composition from 100 GeV/A to 10 TeV/A. It consists of a telescope of silicon strip detectors on top of a set of carbon interaction targets followed by a BGO (Bismuth-Germanium-Oxygen, BidGe3012) calorimeter, with triggering scintillators interspersed at various points. The

98

experiment was flown near the top of the atmosphere for an extended time in Antarctica during January 2001 and additional flights are planned.

5. Resultant Backscatter Albedoes A total of 1600 Carbon nuclei events at 1 TeV/A normally incident but slightly offset from the central Z-axis (1.24 cm in X and Y) of the ATIC instrument were simulated. 1000 similar events at 100 GeV/A were also conducted, for side-by-side comparison of the effect of a 10-fold increase in energy. A histogram of the energy deposited in the top, first layer of the ATIC silicon detector is given in Figure 2, represented by the shaded spectrum. The peak at 5.29 MeV is the energy deposited by the primary C nucleus in the central Si pixel. A mzp deposits 147 keV in Si, times Z2 with Z=6 for Carbon, and the peak follows as it should.

counts 1o4

FLUKA - ATIC simulation raw energy spectrum silicon detector, first layer

1o3 1o2

10

1 0

2

4

6

8

10

12

14

16

18

20

MeV Figure 2. A raw energy spectrum in total counts from Carbon incident at 1 ATeV along the central axis of ATIC. A similar spectrum at 100 AGeV was generated for comparison.

The unshaded spectrum corresponds to the backscatter albedo. This is the total energy spectrum of the secondary radiation produced by the impact of the primary C, being the sum or integral of all charged plus neutral particles. One would expect, for example, that thermal and evaporation neutrons are principal contributors to the unshaded spectrum below 4 MeV. The fluence plots in Figure 3 represent a decomposition of the energy spectrum in Figure 2 into its component contributions from charged and neu-

100

plots as an expanding gas or cloud of particles produced by the excitation or perturbation from the impacting C nucleus. This helps visualize why the flux is moving backwards (to the right in Figure 3) with respect to the calorimeter’s center of mass at (X,Y,Z)=(O,O,-15), constituting backscatter. The intent of our study has not been to re-design the ATIC instrument but rather to benchmark changes in the FLEUR adaptation of FLUKA. In several respects, the fluence plots in Figure 3 raise interesting questions. The neutron balance plot, for example, indicates the presence of neutrons in the very top Si layer (Sl) of the instrument (Z=45 cm). As space-borne cosmic-ray detector technology exceeds 1 TeV/A one can use Monte Carlos such as FLUKA to evaluate the physics of ever-increasing neutron backscatter contamination for triggering schemes as well as the interpretation of certain data. The photon fluence in Figure 3 is equally interesting. Supposing one might want to place a transition radiation detector (TRD) above the calorimeter (Z > 0 cm), the backscatter photons can affect TRD triggering schemes. Thermal and evaporation neutrons can be attenuated by a layer of H-rich material such as plastic due to elastic scattering of neutrons by protons. Other materials can alter both the photon and neutron backscatter. In any case, the utility of FLUKA for assessing the optimal absorption layer at the top of the calorimeter (Z=O cm) for mitigating backscatter is obvious. 6. Conclusions

We have developed a semi-automatic procedure5 to facilitate the conversion of simulation problems involving GEANTS designs into FLUKA. Successfully applied to the ATIC instrument, a preliminary study of the backscatter albedo has been performed using FLUKA with its newly implemented DPMJET event generator module. A simplified conversion of numerous G3 particle detector designs using FLUKA seems feasible now as next-generation Monte Carlos continue t o evolve. References 1. A. Fass6, A. Ferrari, J. Ranft, and P. Sala, Proc. Monte Carlo 2000, 159 and 955

(Springer-Verlag, Berlin, 2001.). 2. R. Brun, et al., Proc. of Computing in High-Energy and Nuclear Physics (CHEP) (Elsevier, Berlin, 1997). 3. P. Zuccon, et al., ICRC-2001 (Copernicus Gesellschaft, Hamburg, 2001). 4. G. Battistoni, et al., ICRC-2001 (Copernicus Gesellschaft, Hamburg, 2001). 5. A. Empl, et al., Comp. Phys. Comm., to be submitted. 6. Isbert, J . , et al., these proceedings.

A SILICON-TUNGSTEN CALORIMETER FOR COSMIC-RAY PHYSICS

V. BONVICINI, M. BOEZIO, P. SCHIAVON, G. SCIAN, A. VACCHI, G. ZAMPA AND N. ZAMPA University of lkieste and INFN Sezione di lkieste, via A . Valerio 2, 1-3137 'Iheste, Italy E-mail: bonviciniOts.infn.it

E. MOCCHIUTTI Royal Institute of Technology, AlbaNova University Center (SCFAB), S-10691 Stockholm, Sweden A silicon-tungsten imaging calorimeter has been designed and built for the PAMELA satellite-borne experiment. The main physics task is the measurement of the flux of antiprotons, positrons and light nuclei in the cosmic radiation. The calorimeter is made by 22 layers of tungsten (each 0.74 XO thick) interleaved with X-Y silicon sensor planes. The signals are read out by a dedicated custom VLSI front-end chip, the CR1.4P, with a dynamic range of 7.14 pC or 1400 MIPS (Minimum Ionizing Particle) and self-trigger capability. We report on the calorimeter design details, the expected performance in PAMELA, the experimental results obtained in test beams and their comparison with simulations.

1. Introduction: the PAMELA experiment The PAMELA mission is part of the Wizard-RIM (Russian-Italian Mission) program, which foresees several space experiments with different scientific objectives. Two missions of the program (RIM-0 and RIM-1) have already been carried o ~ t ~ 1 ~ 1 ~ . The PAMELA4 experiment (RIM-2 mission) has the scientific goal of measuring the cosmic radiation over a wide energy range with unprecedented accuracy. The PAMELA apparatus will be installed on-board the Russian satellite Resurs-DK1, which will be launched in early 2003 with a Soyuz launcher. Its sun-synchronous, nearly polar orbit at 600 km of altitude will allow the measurement of the low-energy component of cosmic rays when the instrument is near the poles. The main objectives of the experiment are the precise measurement of the positron flux from 50 MeV to 270 GeV and of the antiproton flux from 80 MeV to 190 GeV, as well as the search for anti-helium with a sensitivity of in the E / H e ratio. For further details see Bonvicini et al.4.

101

102

The structure of PAMELA is similar to the one used by the Wizard Collaboration in its stratospheric balloons experiments5. The PAMELA apparatus is composed by the following detectors: 0

0

0

0 0

A system of plastic scintillators that includes: a time-of-flight counter (TOF), providing dE/dx measurement, timing information and the trigger for data acquisition, and an anti- coincidence system (ANTI), which identifies those particles that enter the spectrometer from outside its geometrical acceptance; A Transition Radiation Detector (TRD), made by 1024 Xe/Co2- filled straw tubes; A magnetic spectrometer (SPE), formed by a permanent magnet (providing a field of 0.4 T) and a tracking system realised with 6 layers of double-sided silicon microstrip detector; A Si-W electromagnetic calorimeter (CAL, see next sections); A plastic scintillator counter (S4) placed under the calorimeter for triggering of high energy (> 100 GeV) electrons.

Furthermore, a neutron counter is foreseen to be installed in the payload along with the PAMELA apparatus, just below S4. Its purpose is to work together with the calorimeter for measuring very-high energy electrons (see section 5). PAMELA has been designed taking into account the strict mass (430 kg) and power budget (350 W) limitations imposed by the satellite, as well as the vibration and shock loads occurring during the launch phase. 2. Design characteristics of the PAMELA Imaging Calorimeter The PAMELA calorimeter is a sampling calorimeter made of silicon sensor planes interleaved with plates of tungsten absorber. The instrument was designed aiming to an excellent granularity, both in the longitudinal (Z) and in the transversal (X and Y) directions. Longitudinally, the granularity is given by the thickness of the absorber layers, which is 0.74 XO (0.26 cm). Since there are 22 tungsten layers, the total depth is 16.3 XO (i.e. about 0.6 interaction lengths). This depth is not sufficient to fully contain high-energy electromagnetic showers. However, the fine granularity, along with the energy resolution of the silicon detectors, allows an accurate topological reconstruction of the showers, thus making the calorimeter a powerful particle identifier, as confirmed by simulations and experimental results (see next sections). Each tungsten plate is sandwiched between two printed circuit boards (called “frontend boards”), which house the silicon detectors and the front-end and ADC

103

electronics. The silicon detectors developed for the calorimeter are large area devices (8x8 cm2), each of them is segmented into 32 large strips with a pitch of 2.4 mm and has a thickness of 380 pm. In each front-end board the detectors are arranged in a square matrix made by 3 x 3 devices and each of the 32 strips of a detector is wire-bonded to the corresponding one of the other two detectors in the same row (or column), thus forming 24 cm-long strips. The orientation of the strips of two consecutive layers (conventionally referred to as X and Y “views”) is shifted by 90”, so to have 2-dimensional spatial informations. The whole calorimeter structure is modular. The basic unit, called a “detection plane”, is formed by a tungsten plate and its two front-end boards, fully equipped with detectors and electronics. Two such detection planes are assembled together forming a “detection module”. In a module, the two detection planes are kept together by special aluminium frames to which they are bolted at the edge of the absorber plates. Figure 1 shows a picture of a

Figure 1. Photograph of a detection module of the PAMELA Imaging Calorimeter.

detection module. The 11 modules of the calorimeter can be inserted into the mechanical main structure of the calorimeter by sliding them through guides which are precisely machined inside the structure itself. The front-end electronics is based on a VLSI analog processor specifically designed for the PAMELA calorimeter, the CR1.4P6i7. For more details about the calorimeter front-end and read-out electronics, see Boezio et al.7. 3. Simulated performance.

Similar silicon-tungsten calorimeters, differing in layout, number of layers and read-out electronics, were developed and extensively studied by our group

104

through simulations, beam testss and balloon flights9JoJ1J2. Throughout this activity, we had the opportunity to continuously upgrade and tune the simulation routines by comparing the simulated data with experimental results. The agreement between simulations and experimental data is excellent8J1J2. From these simulations, a Monte Carlo program based on the CERN GEANT/FLUKA- 3.21 code13 was developed to study the performance of the PAMELA calorimeter concerning the primary scientific goals of PAMELA, in particular the energy reconstruction for electrons and the identification of positrons and antiprotons in the background of protons and electrons, respectively. The simulated energy resolution of the calorimeter as a function of the energy of the electrons has a (17/fi)% dependence and the “constant term” is about 5% from 20 GeV to about 200 GeV. At higher energies the resolution starts to worsen mainly due to longitudinal losses. It is important to notice that this simulation did not consider only perpendicular tracks but, like in the real experimental situation, tracks coming at any angle inside the geometrical acceptance of the PAMELA trigger system. Therefore, the energy resolution takes into account also effects due to variations of the effective thickness of the materials and to the presence of non-active silicon volumes at the edge of the detectors. As mentioned in the previous sections, one of the main tasks of the calorimeter is to act as a powerful particle identifier to select positrons and antiprotons. In fact, in a cosmic-ray experiment like PAMELA, protons and electrons constitute most of the positive and negative components, respectively, of the cosmic radiation. The longitudinal and transversal segmentation of the calorimeter, combined with the measurement of the particle energy loss in each silicon strip, results in high identification power for electromagnetic showers. Therefore, in the electron and positron analysis, the calorimeter is used to identify electromagnetic showers whereas in the antiproton analysis it is used to reject them. Selection criteria were developed based on the experience gained using silicontungsten calorimeters in previous balloon-borne experiments”. The efficiency and contamination of the selections were studied simulating a large number of electrons, antiprotons and protons. The resulting values for different momenta spanning the entire range of interest for PAMELA are collected in Table 1.

4. Beam test results In July 2000 a prototype version of the calorimeter was tested on a particle beam at the CERN SPS, employing muons, electrons and pions up to 100 GeV/c. The calorimeter under test was equipped with all the tungsten

105 Table 1. Simulated performances: efficiencies in antiproton and electron detection versus electron and proton contamination, respectively. -

Momentum (GeV/c) 1

P efficiency 0.9192 f 0.0009

5

0.9588 f 0.0005

( 4 : ; )

20

0.9767 f 0.0004

< 6.2 X lop5 < 1.4 x 1 0 - ~ < 1.5 x 1 0 - ~

100

0.963f 0.001

200

0.954f 0.002

econtamination (2.5f0.2)X X

lop5

e+ efficiency 0.899 f 0.001

P contamination (1.9f 0.4) x 1 0 - ~

0.9533 f 0.0009

(i.4?;:;) x 10-5

0.970 f0.001

(3+;) x 10-5

0.944 & 0.002

< 3.3 x < 1.2 x

0.955 f 0.002

10-5 10-4

plates, but only 5 layers of silicon detectors (out of a total of 44) were installed. The main purpose of the test was to evaluate the performance of the CR1.4P front-end chip in real experimental conditions, to test the front-end electronics up to the ADC and to check the general behaviour of the calorimeter for electrons and hadrons, even with a few active layers. Figure 2 shows the signal distributions on one strip produced by minimum ionising particles (muons at 100 GeV/c). The signal-to-noise ratio for MIPS is

Signal (ADC values) Figure 2. Signal distribution on one strip for minimum ionizing particles (test beam data).

better than 9:1, thus confirming the expected performance of the CR1.4P

106

Figure 3 is a plot obtained by considering data from all runs of electrons Momentum 100 GeV/c

Total detected energy [MIPI

Figure 3. Total number of hit strips versus total detected energy for 100 GeV/c pions and electrons (test beam data).

and pions at a momentum of 100 GeV/c. On the vertical axis is reported the total number of hit strips whether on the horizontal axis is reported the total detected energy in units of MIP. Notwithstanding the few silicon detector views installed in the calorimeter, its power of separating electrons from hadrons by means of the different shower topologies is striking. 5 . Self-triggering operation of the calorimeter

The CR1.4P chip was designed to work not only with an external trigger signal (‘‘normal” operation mode) but it was also provided with a “self-trigger” capability, i.e. it generates a logic signal when the sum of the signals of all 16 preamplifiers exceeds a certain threshold, which can be externally regulated. The self-trigger option in the CR1.4P was developed t o enhance the calorimeter’s capability to measure very-high energy (from 300 GeV to more than 1 GeV) electrons in the cosmic radiation. At present, very few measurements have covered this energy range14. Since these events are quite rare in comparison with the “normal” event rate of PAMELA, it is important to have a N

107

large geometric factor in order to increase the statistics during the expected three-year lifetime of the mission. By using the calorimeter alone and requiring that the particles enter from one of the first four planes and cross at least 10 radiation lengths, the geometric factor becomes about 600 cm2sr, i.e. a factor of 30 larger than the normal PAMELA acceptance. The behaviour of the calorimeter in self-trigger mode was carefully studied by means of simulations7. The simulated energy resolution of the calorimeter in self-trigger mode is fairly constant (z12%) up to about 800 GeV. At higher energies the resolution decreases because of increasing longitudinal leakage and saturation of the signal from the strips (about 1100 MIP, this limit being set by the ADC dynamics rather than by the CR1.4P chip). Compared to the energy resolution of the calorimeter in “normal” PAMELA acquisition mode (see Figure 3), the resolution in self-trigger mode is worse (12% instead of 5% at 200 GeV), the worsening being due to the different acceptance conditions. 6. Conclusion

PAMELA will be launched in early 2003 and will allow measurement of antiparticle spectra in the cosmic radiation over a wide energy range with unprecedented accuracy. The Flight Model of the Si-W imaging calorimeter of PAMELA is presently being completed. Simulations and beam tests with the Engineering Model have shown that the calorimeter can fulfil all scientific requirements for PAMELA. References A. Bakaldin et al., Astropart. Phys. 8,109 (1997). V. Bidoli et al., Advances i n Space Research 25,2075 (2000). V. Bidoli et al., Astrophys. J. Suppl. 132,365 (2001). V. Bonvicini et al., Nucl. Instr. and Meth. A461, 262 (2001). M. Ambriola et al., Nucl. Phys. (Proc. Suppl.) B78 32 (1999). J. H. Adams et al., Proc. 26th Int. Cosmic Ray Conf., Salt Lake City, 5 69 (1999). M. Boezio et al., “A High Granularity Imaging Calorimeter for Cosmic-Ray Physics”, t o appear on Nucl. Instr. and Meth. A (2002). 8. M. Bocciolini et al., Nucl. Instr. and Meth. A333, 560 (1993). 9. R. Golden et al., Astrophys. J . 457,L103 (1996). 10. M. Boezio et al., Astrophys. J . 487,415 (1997). 11. M. Boezio, Ph. D. Thesis, Royal Institute of Technology, Stockholm (1998) (available at http://www.particle.kth.se/group_docs/admin/theses.html#phd) . 12. M. Boezio et al., Astrophys. J. 532,653 (2000). 13. R. Brun et al., Detector Description and Simulation Tool, CERN program library. 14. T. Kobayashi et al., Proc. 26th Int. Cosmic Ray Conf., Salt Lake City, 3 61 (1999). 1. 2. 3. 4. 5. 6. 7.

ELECTROMAGNETIC CALORIMETER FOR THE AMS-02 EXPERIMENT

R. KOSSAKOWSKI, F. CADOUX, V. CHAMBERT-HERMEL, G. COIGNET, J. M. DUBOIS, D. FOUGERON, N. FOUQUE, L. GIRARD, C. GOY, R. HERMEL, B. LIEUNARD, S. ROSIER-LEES, J. P. VIALLE LAPP

- BP

110, 74941 Annecy -1e-Vieux Cedex, France

G. CHEN, H.CHEN, Z. LIU, Y. LU, Z. YU, H. ZHUANG IHEP - Chinese Academy of Science, 100039 Beijing, China

F. CERVELLI, S. DI FALCO, T. LOMTADZE, G. VENANZONI INFN - Sezione di Pisa, Via Liuornese 1291, 56010 S. Piero a Grado, Italy

E. FALCHINI, P. MAESTRO, P. S. MARROCCHESI, R. PAOLETTI, F. PILO, N. TURINI, G. VALLE Gruppo Collegato INFN - Siena Physics Dept.,55 u. Banchi di Sotto, 53100 Siena, Italy

(presented b y R . Kossakowski at CA LOR 2002, Email : kossakowskiOlapp.in2p3.fr)

The electromagnetic imaging calorimeter made of Lead and scintillating fibers will identify the high energy leptons and y rays in AMS-02 experiment on the International Space Station. Physics requirements and space qualification constraints lead t o severe optimizations of the detector design, the mechanics and the electronics for this 16,5Xo calorimeter sampled by 1296 electronic channels.

1. Introduction

The AMS-02l experiment aims at measuring the cosmic ray spectra in the range of energy from GeV to TeV for three years on the International Space Station ISS. The experimental set-up consists of a superconducting magnet, a silicon tracker and a number of additional detectors, designed to measure the energy and to identify the nature of cosmic rays. These detectors are: the transition radiation detector (TRD), the time of flight (TOF) which also provides a standard AMS trigger, the ring imaging Cerenkov counter (RICH) and the electromagnetic calorimeter (ECAL). The detailed description of the AMS-02 detector can be found elsewhere2.

108

109

The electromagnetic calorimeter, which is being constructed by an Annecy (France), Beijing (China) and Pisa-Siena (Italy) collaboration, will be a major instrument to identify electrons, positrons and y-rays and to measure their energy in particular in the high energy part of the spectrum. The ECAL is an imaging calorimeter consisting of 9 modules made of layers of Lead and scintillating fibers (Figure 1). Each module has a 648x648 mm2 section and 18 mm depth, which corresponds to 1.8 radiation lengths. In two successive modules the fibers are rotated by 90 and follow on X or Y direction. The fibers of a module are read only at one end by the photomultiplier R7600-00-M4 from Hamamatsu3, placed alternatively on each side. One photomultiplier consists of 4 independent pixels. In this way the elementary cell of the calorimeter has the dimension of 648x9 mm2 (or 9x648 mm2)in X-Y directions and 9 mm in the Z direction. It corresponds roughly to a 0,5 Molire radius for the transverse dimension of the electromagnetic shower and to a 0,9 radiation length in the longitudinal direction. A particle impinging vertically on the ECAL crosses about 16,5 radiation lengths and the longitudinal profile of the electromagnetic shower is sampled by 18 independent measurements (Figure 1).

Figure 1. ECAL for the A M S - 0 2 experiment inside the supporting structure. 324 photomultipliers will be housed in the lateral panels. The electromagnetic shower is sampled in X and Y directions by scintillating fibers glued to the inside of the grooved lead foils.

The major challenge for the photomultiplier and for its front-end electronics is related to the very large dynamic range of light pulses created in optical fibers by cosmic rays. The signal in the photomultiplier ranges from a few photoelectrons for minimum ionizing particles (MIP) to about lo5 photoelectrons for electromagnetic showers corresponding to very high energy particles (for instance an electron of 1 TeV en erg^)^. Power consumption for all ECAL electronics (including PMT bleeders) is limited to 100 W.

110

Another challenge comes for mechanics of the ECAL. The total weight is limited to 630 kg (the weight of the Lead the fibers is 512 kg), first resonance frequency of the structure must be higher than 50 Hz and it has to support vibrations and accelerations up to 27g during the space qualification tests. All elements have to be designed to support 30 000 thermal cycles during 5 years of orbiting. The ECAL will be dipped into the stray magnetic field of the superconducting magnet, ranging up to 300 Gauss. The weight budget allowed for the detector and the space available between photomultipliers are very limited, which implies a fine optimization between the response of the photomultipliers on the magnetic field and the design of the magnetic shielding. In the following sections we will present the performances of the photomultiplier and its front end electronics and the design of the magnetic shielding and light collection system incorporated into the mechanical structure.

+

2. Photomultiplier and its front end electronics The total of 324 photomultipliers R7600-00-M4 from Hamamatsu will be used. The resistance for vibrations, resistance for magnetic field, square form, compactness and low weight were the major factors leading to the choice of this space qualified photomultiplier. The properties of R7600-00-M4 were extensively studied in order to optimize its dynamic range4. As mentioned in the introduction, the expected light signal from the calorimeter ranges from a few photoelectrons to lo5 photoelectrons. Several types of bleeders were tested and the dynamic range for each bleeder were determined using the LED light pulses. The deviation from the linearity was detected by comparing the PMT response for the signal of two LEDs flashing simultaneously with the sum of responses to LEDs flashing individually. As can be seen in Figure 2, for a given bleeder and given high voltage, the saturation occurs at the same output charge for all four pixels of the PMT. Figure 3 shows the relation between the number of photoelectrons at the saturation point and the gain of the photomultiplier for two different bleeders. One can observe that the saturation point differs by a factor of 5 between these two bleeders. Choosing the saturation at the level of lo5 photoelectrons and B type bleeders(see reference5 for detail) sets the gain at lo5. It was checked that in these conditions the signal corresponding to about 9 photoelectrons (expected for MIPS) can be clearly separated from the background (Figure 4 ). As a result, the design of the front end electronics (dedicated ASIC chip) was made to accept signals from 30 fC to 2 nC6. In this front end chip signals from PMT anodes are separated to low and high amplitude parts (ratio of

111

Figure 2. T h e saturation of the response of the photomultiplier as a function of the collected charge (B type HV bleeder and voltage supply of -6OOV - see text).

f

a

10'

z

Figure 3. Relation between the number of photoelectrons at saturation point and the gain of the photomultiplier for two different high voltage bleeders.

-

?!.

m

E

4

35

I

450

Noturol photoslectron fluctuotion for Npa

YII

5511

MI

~1

71x1

750

-

8

SUI

UXI

~*YI

High Voltage (V)

Figure 4. T h e separation of the signal and the pedestal as a function of the applied high voltage. T h e natural fluctuation of the signal corresponding t o 8.5 photoelectrons corresponds t o RMS = 2,9.

about 30), than shaped and integrated with 2 pus time constant and finally treated by track and hold logic. The signal from last dynode is also proceed

112

by low gain channel. Finally the multiplexer incorporated in the chip sends these 9 signals to the ADC. The whole system (including ADC) is placed on the small electronic board directly coupled to the bleeder board. The linearity tests performed on the first version of the chip are presented in Figure 5 . The right linearity was obtained in the whole expected range and the level of noise is compatible with the detection of MIPS.

High gain output (G-33

Figure 5. The output signal of the dedicated front end ASIC chip as a function of the signal delivered by the photomultiplier. One can observe a good linearity in the whole required dynamic range and the noise at the level of 10% of the MIP signal.

3. Magnetic shielding and light collection system As it was mentioned in the introduction, the stray magnetic field from the AMS-02 superconducting magnet is of the order of 200 - 300 Gauss in the region where the calorimeter is placed. The limit of the weight and the small space available between photomultipliers makes it necessary to optimize the thickness of the material used for magnetic shielding. Finite element calculations corresponding to this optimization can be found in reference7. In these calculations it was shown that when taking into account the global configuration of magnetic materials (those for both RICH and the calorimeter), the magnetic field can increase by a factor of 2 in some particular regions, where the density of surrounding magnetic materials is important. This is due to the attraction of magnetic field lines by the structure. Taking into account this effect, the local configuration of the magnetic shielding was calculated. It was shown that the shielding of photomultipliers by lmm thick, square soft iron tubes (section of 2.4 cm by 2.4 cm and length of 7 cm) lowers the magnetic field in the tube down to 10 Gauss in the central part. The measurements performed on some particular configurations confirmed the validity of calculations. It was also checked, that the response of the photomultiplier is affected

113

by less than 10% by such a field4. Finally, the light collection system between the fibers of the calorimeter and the photomultiplier was optimized: leaving the fiber of the calorimeter, the light crosses successively a RTV joint, 30mm Plexiglas conical light guide, a second RTV joint and enter the photomultiplier. The collection efficiency of this set up was measured to be about 70% . 4. Conclusions

The optimization of the processing of the light signal from the electromagnetic calorimeter of the AMS-02 experiment is described. The extensive study of the R7600-00-M4 photomultiplier from Hamamatsu was done leading t o following conclusions: (1) the high voltage bleeder of the type B from Hamamatsu allows the collection of signals in the whole required dynamic range (from 3 c to 2 C , corresponding to MIP and to electromagnetic showers of 1 TeV electron respectively) with the gain of lo5. (2) the magnetic field is required to be less than 10 Gauss to limit the signal reductionto 10% .

The dedicated front-end ASIC chip was designed and the linearity of the chip was tested in the whole required dynamic range. The noise level was measured at the level of 10% of the MIP signal. The magnetic shielding of the photomultipliers was optimized by finite element calculations. It was shown that l m m thick soft iron tubes lower the magnetic field from 200-300 Gauss down to about 10 Gauss for the whole geometry of the calorimeter. Finally, the light collection system between the calorimeter and the photomultiplier was designed (RTV joints and conical light guides made with Plexiglas).

References 1. http://ams.cern.ch/AMS and references quoted there. 2. B.Alpat - AMS on ISS - talk given on the Conference Frontiers Detectors for Frontier Physics - 8th Pisa Meeting on Advanced Detectors, May 2000. 3. Hamamatsu data sheet, March 1997. 4. R.Kossakowski et al, LAPP-EXP-2002-02 / AMS Note 2002-01-03. 5. Monte Car10 simulations with Geant 4 - AMS collaboration 6. V.Herme1 - EMC electronics status - presentation on AMS Technical Interchange Meeting, CERN, Juin 2001. 7. F.Cadoux, R.Kossakowski, J.P.Vialle - AMS Note 2001-05-01.

PERFORMANCES OF THE AMS-02 ELECTROMAGNETIC CALORIMETER

F. CERVELLI, S. DI FALCO, T. LOMTADZE, G. VENANZONI Istituto Nazionale d i Fisica Nucleare INFN, via Vecchia Livornese 1291, 56010 Pisa, Italy

E. FALCHINI, P. MAESTRO, P. S. MARROCCHESI, R. PAOLETTI, F. PILO, N. TURINI, G. VALLE University of Siena-INFN Gruppo Collegato, via Banchi d i Sotto 55, 53100 Siena, Italy

G. COIGNET, L. GIRARD, C. GOY, R. KOSSAKOWSKI, S. LEES-ROSIER] J. P. VIALLE Laboratoire d’Annecy-le- Vieux de Physique des Particules-LAPP Chemin d e Bellevue, 74941 Annecy-le- Vieux, France

G. CHEN, H. CHEN, Z. LIU, Y. LU, Z. YU, H. ZHUANG Institute of High Energy Physics-IHEP, Chinese Academy of Sciences, 19 Yuquan Road Shijing Shan District, 100039 Beijing, P.R. China

A full-scale prototype of the e m . calorimeter for the AMS-02 experiment was tested at Cern in October 2001 using 100 GeV pions and electrons beams with energy ranging from 3 t o 100 GeV. The detector, a lead-scintillating fibers sampling calorimeter about 17 radiation lengths deep, is read out by a n array of multi-anode photomultipliers. T h e calorimeter’s high granularity allows t o image the longitudinal and lateral showers development, a key issue t o provide high electron/hadron discrimination. From the test beam data, linearity and energy resolution were measured as well as the effective sampling thickness. T h e latter was extracted from the data by fitting the longitudinal e m . showers profiles at different energies.

1. Introduction

AMS 02 is a large acceptance spectrometer designed to operate on the International Space Station (ISS) for three years1. The main goals of the experiment include: (1) search for nuclear antimatter with an expected sensitivity of lo-’ for He and lop7 for

c2

114

115

(2) search for SUSY dark matter with high statistics precision measurements of the p , e+ and y spectra3 (3) astrophysical studies with high statistics precision measurements of the light isotopes (3He, B, C, 9Be, 1°Be) spectra in CR, and observations in VHE y-rays astronomy4. The accurate measurement of the e+ and y spectra at energies greater than 5 GeV requires a high rejection power (- lo4) against the CR background, mainly consisting of protons. This requirement can be achieved by a fine grained sampling e.m. calorimeter that can image the shower development in 3D, allowing for the discrimination between hadronic and e.m. cascades and the reconstruction of the shower direction5.

2. The electromagnetic calorimeter for AMS-02 The e.m. calorimeter (ECAL) of the AMS-02 experiment is a lead-scintillating fibers device6 with an active area of 648x648 mm2 and a thickness of 166.5 mm. The detector is subdivided into 9 superlayers, each consisting of 11 grooved lead foils (1 mm thick) interleaved with layers of scintillating fibers (1 mm diameter) glued by means of an epoxy resin; the superlayers are alternatively oriented along orthogonal directions. The light signal coming out from the fibers is collected by photomultipliers (Hamamatsu R7600 00-M4) through plexiglass light guides. The light guide geometry fits the active area of each of the four cathodes of the PMTs. The region delimited by one of the four PMT cathodes is called a cell (9x9 mm2); the calorimeter is subdivided into 1296 cells, corresponding to 324 photomultipliers. Further details about the detector can be found in reference?.

3. Test beam setup

A full scale prototype of the calorimeter was tested during October 2001 in CERN at the SPS X7 beam line, using 100 GeV pions and electrons with energy ranging from 3 to 100 GeV. The calorimeter was equipped with a total of 36 PMTs, therefore the effective active area was limited to 72x72 mm2 (8 instrumented cells per layer). The high voltage supply for each P MT was set to fix the working point at a gain of about lo6. The detector was read-out feeding the cathode signals into CAMAC gated charge integrating (12 bit) ADCs. No front-end sampling electronics was used during the test. Beam particles were triggered on by means of a 3-fold coincidence of scintillators aligned along the beam line.

116

4. Test beam data analysis

The ECAL cells were first equalized to take into account the differences among the 36 PMTs. Two different equalization methods were used, based respectively on the cells response t o electrons and mzp's'; the first procedure was found to be the most accurate one. After equalization specific analyses were devoted to: the X O measurement and the e.m. shower profile reconstruction the study of energy linearity and leakage corrections the study of energy resolution

4.1. Measurement of the effective sampling thickness The longitudinal profiles of e.m. showers for beam energy in the range 3-100 GeV were reconstructed and fitted (Fig. 1) using the functiong

where t is the layer index and the maximum of the shower occurs at t,,, Plotting t,,,

=f.

as function of beam energy (Fig. 2) and using the relation

t,,,

= XO . log E

+ const

(2)

to fit the data, a value of X O = (9.6 f 0.3) mm was extracted, which implies an ECAL total thickness of (17.3 f 0.5) X,.

Layer

Figure 1. Average longitudinal shower profile at 50 GeV beam energy. T h e superimposed histogram is the expected average profile from the Monte Carlo.

117

1 Figure 2.

10 Shower maximum t,,,

vs. beam energy.

4.2. Energy linearity

Electrons runs at different energies were used to calibrate the energy scale of the calorimeter. After the equalization, good linearity was found up to 30 GeV beam energy (where longitudinal leakage is negligible), with deviation of the order of 15% at 70 GeV and 20% at 100 GeV (Fig. 3). A leakage correction was applied at higher energies deriving the EOparameter from the average longitudinal profile fit (Fig. 1). The leakage corrected linearity shows residual small deviations (about 3% at 70 GeV and 4.5% at 100 GeV) caused by the incomplete coverage of the lateral development of the shower (due to the limited number of instrumented cells) and to dead channels. 4.3. Energy resolution

The energy dependence of the resolution of the calorimeter is shown in figure 4 where the fractional uncertainty on the energy measurement is plotted as a function of the nominal beam energy E. From the fit of data we found the energy resolution curve:

118

E

w80

60

40

20

0 0

Ebeam (GeV)

Figure 3.

20

60

40

80

104

Ebeam (Ge\

Energy linearity curve before (left) and after (right) the leakage correction.

0

20

40

60

80

100

Energy (GeV)

Figure 4. Energy resolution vs. beam energy.

5. 3D shower imaging

Taking advantage of its fine granularity and geometry with alternate planes of fibers oriented along orthogonal directions, the calorimeter can image the

119

longitudinal and lateral development of e.m. and hadronic showers, providing two orthogonal views for each event. In figures 5 and 6 two typical events are displayed, respectively, for an electromagnetic and a hadronic interaction. The calorimeter is capable to image showers so resolving the different topology of e m . versus hadronic events; this is essential to achieve a high e l h discrimination. A fully quantitative study of the hadron rejection power was not

X-L

vlew RUN 186 Event 252

fiview RUN 186 Event 252

Figure 5 . Image (2 orthogonal views) of an e.m. shower generated by a 100 GeV electron.

1x4 vtew RUN 171 Event 18

iy-zview RUN 171 Event 1s

t

Figure 6.

Image (2 orthogonal views) of a hadronic shower generated by a 100 GeV proton.

120

possible due to the small statistics collected with hadron beams and to the limited lateral detector coverage. Rejection power will be measured in a future planned test beam. 6. Conclusions

A prototype of the e.m. calorimeter for AMS-02 was successfully tested at CERN and the measurement of its properties showed a good agreement with its expected performances. The imaging capability of the calorimeter allowed a 3D visualization of the shower development, an essential tool to discriminate e.m. showers from showers originating by the hadronic interactions of protons and pions.

References B. Alpat, Nucl. Instr. and Meth. A461,272-274 (2001). A. Dolgov, Phys. Rep. 222, 311 (1992). G . Jungman et al., Phys. Rep. 267,195-373 (1996). R. Battiston, Nucl. Instr. and Meth. A409,458-463 (1998). 5. E. Choumilov et al., Nucl. Instr. and Meth. A426,625-632 (1999). 6. M. Antonelli et al. Nucl. Phys. B 54, 14-19 (1997). 7. F. Cervelli et al. “The AMS Electromagnetic calorimeter”, CALOR2002 proc. 8. F. Cervelli et al. “Analysis of Ecal 2001 test beam data”, AMS-02 Note. 9. E. Longo and I. Sestili Nucl. Instr. and Meth. A128,283 (1975). 1. 2. 3. 4.

THE STATUS OF GLAST CSI CALORIMETER

A.CHEKHTMAN, REPRESENTING GLAST COLLABORATION Space Science Division, Naval Research Laboratory, 4555 Overlook Avenue Washington DC 20375, USA E-mail: [email protected]

GLAST is a gamma-ray observatory for celestial sources in the energy range from 20 MeV to 300 GeV. This is NASA project with launch anticipated in 2006. The principal instrument of the GLAST mission is the Large Area Telescope (LAT), consisting of an Anti Coincidence Detector (ACD), a silicon-strip detector Tracker (TKR) and a hodoscopic CsI Calorimeter (CAL). It consists of 16 identical modules arranged in a 4 x 4 array. Each module has horizontal dimensions 38 x 38cm2 and active thickness 8.5 radiation length. It contains 96 CsI (Tl) crystals arranged in 8 layers with 12 crystals per layer. The scintillation light is measured by PIN photodiodes mounted on both ends of each crystal. The sum of signals at the two ends of the crystal provides the energy measurement. The difference in these signals provides the position measurement along the crystal. The calorimeter was designed to meet the goals of good energy resolution (better than 10% for photon energies 100 MeV - 100 GeV), position resolution of l m m for photon energies > l G e V , and a rejection factor of > 100 for charged cosmic rays, under limitations on calorimeter weight (95 kg per module) and power consumption (6 W per module). The Monte Carlo simulation and prototype beam test results confirm that proposed design meets the requirements. Calorimeter production is planned t o start in 2003. N

1. Introduction

GLAST is a next generation high-energy gamma-ray observatory designed for making observations of celestial gamma-ray sources in the energy band extending from 20 MeV to more than 300 GeV. The principal instrument of the GLAST mission is the Large Area Telescope (LAT) that is being developed jointly by NASA and the US Dept. of Energy (DOE) and is supported by an international collaboration of 26 institutions lead by Stanford University. The GLAST LAT' is a high-energy pair conversion telescope. It consists of an Anti Coincidence Detector (ACD), a silicon-strip detector Tracker (TKR), a hodoscopic CsI Calorimeter (CAL), and a Trigger and Data Flow system (T&DF). The design is modular with a 4 x 4 array of identical tracker and calorimeter modules. The modules are 38 x 38cm2. Figure 1 shows the LAT instrument concept.

-

121

122

Figure 1. View of the LAT Science Instrument with one Tracker tower module and one Calorimeter module pulled away from the Grid. GLAST is a 4 x 4 array of identical Tracker and Calorimeter modules.

2. LAT technical description The principal purpose of the LAT is to measure the incidence direction, energy and time of cosmic gamma rays while rejecting background from charged cosmic rays and atmospheric albedo gamma rays and particles. The data, filtered by onboard software triggers, are streamed to the spacecraft for data storage and subsequent transmittal to ground-based analysis centers. The Tracker provides the principal trigger for the LAT, converts the gamma rays into electron-positron pairs, and measures the direction of the incident gamma ray from the charged-particle tracks. The primary tasks of the GLAST calorimeter2 are to provide an accurate measure of the energy of the shower resulting from pair conversion of incident gamma rays in the tracker, and to assist with cosmic-ray background rejection through correlation of tracks in the silicon tracker with the position of energy deposition in the calorimeter. The calorimeter also provides triggers to the LAT, particularly for very large energy depositions. 3. Calorimeter design overview

The calorimeter is comprised of a segmented thallium-doped cesium iodide, CsI(Tl), scintillation crystal array.

123

To achieve the required energy coverage and resolution, the calorimeter is 8.5 radiation lengths (8.5Xo)deep. An additional depth of 1.5Xo resides in the tracker. To assist in track correlation for background rejection and to improve the energy measurement by shower profile fitting, the calorimeter is segmented into discrete detector elements and ,arranged into a hodoscopic or imaging configuration and read out using PIN photodiodes. The design of a single calorimeter module is shown on Figure 2.

CsI Detectors 3seout

Carb

Elect

Mounting Baseplate

A1 EM1 Shield

Figure 2. Exploded view of a single Calorimeter module. Eight layers of 12 CsI Crystals are readout by P I N photodiodes and electonics on the four module sides

Each CAL module contains 96 crystals of size 26.7 x 19.9 x 326mm3. The crystals are individually wrapped for improved light collection and optical isolation, and are arranged horizontally in 8 layers of 12 crystals each. Each layer is aligned 90 degrees with respect to its neighbors, forming an x-y array. The spectral response of the PIN photodiodes is well matched with the scintillation spectrum of CsI(Tl), which provides for a large primary si5nal ( N 5000 electrons collected in 1.5cm2 diode per MeV deposited), with correspondingly small statistical fluctuations and thereby good intrinsic spectral resolution. The PIN photodiodes are mounted on both ends of a crystal and measure the scintillation light at each end of a crystal from an energy deposition in the crystal. This provides a redundancy in the energy measurement. However,

124

the difference in light levels seen at the two ends of the crystal also provides a determination of the position of the energy deposition along the CsI crystal. The position resolution of this imaging method ranges from a few millimeters for low energy depositions (- 10MeV) to a fraction of a millimeter for large energy depositions (> 1GeV). The size of the CsI crystals has been chosen as a compromise between electronic channel count and desired segmentation within the calorimeter. The indicated size is comparable to the CsI radiation length (1.86 cm) and Moliere radius (3.8 cm) for electromagnetic showers. The hodoscopic array of CsI crystals is installed in a carbon composite cell structure. Aluminum side panels hold the CsI crystals in the cells, provide mounting space for the readout electronics printed circuit cards, and provide EM1 shielding. A baseplate provides for mounting of the calorimeter module to the LAT GRID structure and is integral to the strength of the GRID. As shown in Figure 2, the readout electronics for the calorimeter are mounted on the four sides of the module where they attach t o the PIN photodiodes. The major design challenges for the calorimeter electronics were 0 0

dynamic range of 5 x lo5 reduced power consumption per CsI crystal

The large dynamic range is supported by using two independent signal chains. A custom dual PIN photodiode assembly is used at each end of the crystals. The active areas of the two diodes have a ratio of 6 to 1. The larger area diode covers the low energy band (2 MeV - 1.6 GeV), while the smaller diode covers the higher energy band (- 15MeV to 100GeV). The significant overlap between the two ranges permits cross-calibration of the electronics. Each diode has dedicated preamp and shaping amplifiers that are part of a custom application specific integrated circuit (ASIC). The power for the readout electronics has been reduced by the development of analog and digital CMOS ASICs that are optimized to the performance requirements of the calorimeter. The mechanical structure is designed to have the structural stiffness to withstand environmental loads without requiring any contribution from the crystals. The honeycomb geometry of the structure, combined with light, high strength material ensures the required mechanical properties, while minimizing the amount of passive material between the CsI logs. The thickness of the wall within a layer is less than 0.4 mm and from layer to layer less than 0.8 mm. The outer walls are thicker since metallic inserts are embedded in the composite material to provides attachment point for the aluminum parts. A wrapped CsI crystal with bonded photodiodes is called Crystal Detector Element (CDE). The wrapping material - non-metalic reflector film VM2000

125

from 3M - was chosen to maximize the light yield. The wide qualification temperature range (from -3OC to +50C) and the significant mismatch of thermal expansion coefficients of CsI and photodiode carrier require a chareful choice of adhesive t o bond photodiodes to crystals. After a substantial test program, we have selected a Dow Corning silicone elastomer (DC93-500) and primer (DC92-023). These materials have excellent optical and mechanical properties and provide bonds that readily survive the mechanical stresses. The CDEs are mounted independently inside the composite cells and access is granted t o each of them until the close out plates are assembled. A clearance of 0.3 to 0.5 mm allow their integration inside the cells. A silicone elastomeric cord is placed between each the corners of the cells and the chamfers of the crystals to provide a support distributed along the full length of the logs and center the CDE in the cell. The cords are stretched to reduced their diameter and allow the insertion of the log. Tkansverse vibrations of the CDE are damped by the elastomeric cords. Longitudinal motion is damped by elastomeric pads in the cell closeout. Table 1 shows the sharing of the responsibilities of collaboration countries in calorimeter manufacturing. Table 1. The manufacturing responsibilities of collaborating countries. Sweden

- Acceptance and verification of crystals from vendor -

France

Acceptance, verification

- Assemble Crystal Detector Elements (CDEs) - Manufacture mechanical structure

USA

- Manufacture front-end electronics - Integrate CDEs with structure a.nd electronics - Test and calibration

4. Calorimeter status The CsI crystal production contract with Amcrys (Kharkov, Ukraine) is in place for more than 2000 prototype and flight crystals. 240 crystals have been received in Sweden. Minor adjustments have been made in crystal length, chamfer size, and tolerance since the original specification. 650 custom prototype Dual Photodiodes (DPDs) have been received from the vendor, Hamamatsu. The photodiodes have excellent optical and electronic characteristics. Radiation testing in France indicates no problem with

126

the GLAST environment. Thermal cycling tests show small cracks within the optical epoxy, with no degradation of optical or electrical performance. We expect that this problem will be readily solved by the vendor before procurement of flight DPDs. The bonding material and processes have been tested over 90 sample bonds of PIN diodes to CsI crystals. Bond strength is measured t o be the same (250N) before and after thermal cycling, significantly exceeding the tensile and shear strength requirements. Light yield tests on sample CDEs manufactured by the proposed bonding process indicate an expected yield of 7500 e/MeV for the final dimensions, which exceeds the requirement by 25The first two copies of bonding tools have been fabricated, and the first CDEs have been bonded. The prototype mechanical structure has successfully undergone vibration and thermal cycling. The calorimeter analog front-end ASIC design have evolved through 6 fabrication iterations. Essentially all required performance parameters have been demonstrated. Minor remaining issues will be tested in parts delivered in August. The first version of digital readout controller ASIC was received in March 2002 and demonstrated full functionality. Minor improvements and adjustments have been incorporated in parts to be received in August 2002. Two prototype versions of front end printed circuit board have been fabricated for testing of calorimeter readout components. The final version of PCB is currently in layout design. An Engineering Model calorimeter module is planned t o be assembled and tested in late 2002 and early 2003. The modified production schedule includes following milestones: CsI crystal production: Oct 2002 - Oct 2003 Calorimeter modules integration and tests: May 2002 - June 2004 Instrument integration and test: June 2004 - Sep 2005 Spacecraft integration and test: Sep 2005 - Sep 2006 Launch: Nov 2006 References 1. P. E. Michelson, "GLAST: A detector for high-energy gamma rays" Proc.

SPIE Conf. Gamma-Rays and Cosmic-Ray Detectors, Techniques and Missions 2806,B.D.Ramsey and T.A.Parnel1, Eds., Denver,CO, pp.31-40 (Aug,1996). 2. W.N.Johnson, J.E.Grove, B.F.Phlips,J.Ampe, SSingh, E.Ponslet, "The construction and performance of the CsI hodoscopic calorimeter for the GLAST beam test engineering module", IEEE IPrans. on NUC.Sci. 48, 1182 (2001).

PERFORMANCE OF GLAST CALORIMETER

R. TERRIER, M. JOHN, A. DJANNATI-ATAI Collbge de France, Paris, France E-mail: terraerOcdf.in2ppJ.fr

A. CHEKHTMAN, J.E. GROVE, W.N. JOHNSON Naval Research Laboratory, Washington DC, USA

The GLAST Large Area Telescope to be launched in 2006 is dedicated to gammaray astronomy from 20 MeV to 300 GeV. Its calorimeter consists of 16 modules of 8 layers of 12 CsI(T1) crystals arranged in an hodoscopic array. Each module is placed under a silicon tracker using tungsten converters. The calorimeter is only 8.5Xo thick . Therefore, depending on the energy regime, the shower containment is rather poor and corrections need to be applied. We present here the correction algorithms as well its the performances of GLAST calorimeter in terms of energy, position and direction, based on detailed simulations of the instrument and beam tests results.

1. Introduction

The GLAST calorimeter consists in 16 modules of 8 layers of 12 CsI(T1) crystals in an hodoscoping arrangement, this is to say alternatively oriented in X and Y directions, t o provide shower imaging capability. It is designed to measure energies from 30 MeV t o 300 GeV, and even up to 1 TeV. The energy measurement performance on such a broad energy range is limited due t o the presence of a 1.3 X o thick tracker, and due to a thickness of only 8.5 X O which limits the shower containement for high energy events. In these regimes, a correction must be applied in order to restore good linearity and energy resolution. The calibration and its impact on energy measurement is presented first. Then we will give details about the corrections applied in the low and high energy regime, and derive the energy measurement performance of the instrument. We will then briefly review the principles and performance of position estimation in the hodoscopic calorimeter. The results exposed here are based on Monte Carlo simulations of the GLAST instrument based on the GISMO toolkit'.

127

128

2. Calibration

Since the various characterisics of the calorimeter are expected to change strongly after the launch and during the lifetime of the instrument, an inflight calibration is necessary. The galactic cosmic-rays nuclei(GCR) energy deposits will be used to calibrate the gains and response asymetry of crystals. On axis, protons deposit 11 MeV, carbon ions 390 MeV and iron around 7.4 GeV; this provides several absolute gains determination along the dynamic range5. There are several steps in the calibration process:

0

extract multi MIP events and select likely GCR fit the tracks identify charges identify mass and charge changing interactions and reject them fit dE/dx

In order to check the possibility to separate the various species of nuclei, a prototype of one module of the calorimeter5 has been tested at GSI Darmstadt beam facility. To produce a beam of secondaries, a 700 MeV/A Ni beam with a polystyrene target upstream was used. By comparing the energy deposits in the first two layers, we can easily distinguish the various nuclei produced from nickel in spite of the spread of the ion energies. It is important to note that because of the calibration, the calorimeter measures only a deposited energy. To recover an incident energy (especially in the high energy regime), we have to apply different gains offline.

3. Energy Measurement 3.1. Low energy regime

The tracker consist of silicon strips and tungsten converter layers. The upper part is made of 12 3% radiation length thick converters, and the lower part of 4 18% radiation length thick ones. At low energies (under a few hundred of MeV), the tungsten scatter the particles and absorb a large fraction of the incident energy. The tracker must be used as a sampling calorimeter. Once the vertex and direction of the gamma have been fitted, we compute all the hits in a cone of 5 times the multiple scaterring angle aperture around the incident direction and correct the energy with an angle-dependent sampling fraction. Taking the number of hits and the sampling fraction in the thin tracker part (nl, a ) and in the thick layers (122, p), the corrected energy for an incidence

129

angle 8 is:

t Figure 1. Energy resolution in the low energy regime for normal incident photons. Since the energy dispersion function has large tails we give here the gaussian resolution (lower curve) and 68% resolution

The corrected energy response is linear and its resolution is ranging from 17% at 50 MeV to 7% at 500 MeV for normal incident photons. The resolution worsens with inceasing angle and in any case low-energy tails remain. This is summarized on figure 1 3.2. High energy regime

Around the GeV, a significant amount of energy starts to escape through the back of the calorimeter. Using the longitudinal segmentation, we can compute gains to be applied to the measured energy in each crystal in order to obtain the incident instead of the deposited energy. We have shown that these gains are all approximately equal to 1, except for the last layer. This is not surprising since the last layer carries the most important information concerning the leaking energy: the total number of particles escaping through the back should be nearly proportional to the energy deposited in the last layer. Using simulations we fit the energy and angle dependency of the correlation coefficient and obtain the following estimator for the incident energy. Here Ecal and Elast are the energies deposited in the whole calorimeter and in the last layer: Ecorr

= Ecal

+ a(Eca1,0 ) E l a s t

130

This gives good linearity and energy resolution up to a few tens of GeV (depending on the incidence) where the shower maximum begins t o leak out of the calorimeter. To give a correct estimation, even in the highest energy ranges, we fit a mean shower profile to the partially observed profile, using the usual gamma distribution parametrization4. The shower energy and starting point are taken as free parameters, whereas the two parameters T and X (respectively shower maximum position and shower length) are taken at their mean value for the estimated incident energy. This restores a good linearity over the whole energy range and provides a good energy measurement even when the shower maximum is not contained. The energy resolution obtained at 300 GeV is 15% and 19% at 1 TeV for normal incidence photons7. [Energy resolution]

t

0.12 0.14

0.041 0.021

3

4 5678910

30 40 Energy CeV

20

Figure 2. Energy resolution in the high energy regime for normal incident photons. Since the energy dispersion function have large tails we give here the gaussian resolution (lower curve) and 68% resolution. Beyond 30 GeV the shower maximum begins to leak out of the calorimeter

Both methods have been tested during the SLAC 1999-2000 beam test. The resolutions obtained for 20 GeV electrons were 4% and 5% for the shower profile fitting. For more details, see Do Couto e Silva et al.3 4. Position measurement

The hodoscopic arrangement provides a position measurement in both x and y direction for each crystal in order to provide useful information on the shower shape to help rejecting the hadronic background, but also t o improve the di-

131

rection estimation of high energy-events converted in the last tracker layers. We will detail here the longitudinal and transverse position reconstruction and performance. 4.1. Longitudinal position reconstruction

Using the asymmetry of light collection5, one can determine the position of the barycenter of the energy deposited in a crystal by:

L-R x=AL+R where L and R are respectively the left and right response of a crystal, and A the asymetry slope which is mostly dependent on surface treatment and wrapping of the crystals. Inside a layer the crystal with the highest energy deposition yields the best position estimate. The precision is limited by the intrinsic shower fluctuations, the electronic noise, and the uncertainty on A due to non-linearities, calibration etc. The feasability of the method has been demonstrated during the 1997 test beam2 where the position error in a crystal was found to be as low as 0.5 mm for a 5 GeV energy deposit. However the barycenter position is different from the incident direction intersection with the crystal because of the lateral extension of the shower. The radial and longitudinal profiles are mixed up when the incidence angle is large. Therefore we have a systematic bias at non-zero incidence angles. It depends on incidence angle, depth in the calorimeter and energy of the shower. 4.2. Transverse position

Because of the segmentation of the layer, the position given by the energy weighted mean is systematically shifted towards the center of a crystal, leading to the classical S-shape bias of the barycenter6. One can correct for this effect using the usual poatan(p1x) - p 2 z function. Once deconvolved in each layer, the position dispersion obtained is around a factor of 2 worse than in the longitudinal case. We can see on figure 3, the position dispersion obtained for 70 GeV normal incident photons for longitudinal (lower curve) and transverse(upper curve) positions. The precision is better by a factor of two using longitudinal position. The longitudinal error here is mostly limited by electronic noise, it should be noted though that an uncertainty of the asymmetry slope coming from calibration would reduce significantly this precision. The global barycenter dispersion varies from 1.5 mm at 10 GeV to 0.7mm

132

at 100 GeV for normal incidence photons, once again neglecting systematic errors. ~orltfonerror 70 GeVi

Figure 3. Position error for 70 GeV photons in each layer. Upper is deconvolved transverse position, lower is longitudinal position

5 . Conclusion

GLAST calorimeter, though its relative thinness, is able to provide a satisfactory energy measurement. The expected energy resolutions for normally incident photons range from 15% at 100 MeV to 7% at 10 GeV, and 15% at 300 GeV. Besides, the hodoscopic arrangement gives acces to a good position determination allowing an efficient background rejection and an improved direction estimation for high energy events. References 1. W. B. Atwood and T. H. Burnett, SLAC-REPRINT-1992-037 Prepared for 10th International Conference on Computing in High-energy Physics (CHEP 92), Annecy, France, 21-25 Sept 1992. 2. W. B. Atwood et al., Nucl. Instrum. Meth. A 446, 444 (2000) 3. E. do Couto e Silva et al., Nucl. Instrum. Meth. A 474, 19 (2001). 4. Grindhammer,G., Peters, S., 1993, Int. Conf. on Monte Car10 Simulation in High Energy and Nuclear Physics, Tallahassee, Florida 5. Johnson, W. N., Grove, J. E., Phlips, B. F., Ampe, J., Singh, S., & Ponslet, E. 2000, AAS/High Energy Astrophysics Division, 32, 6. R. Y. Zhu, G. Gratta and H. Newman, Nucl. Phys. Proc. Suppl. 44, 88 (1995). 7. Terrier, R. et al. 2001, Gamma 2001, Baltimore, Maryland

COSMIC RAY ENERGETICS AND MASS (CREAM): CALIBRATING A COSMIC RAY CALORIMETER

0. GANEL, E. S. SEO, H. S. AHN, R. ALFORD, K. C. KIM, M. H. LEE, L. LIU, L. LUTZ, A. MALININE, E. SCHINDHELM, J. Z. WANG AND J. W U Institute for Physical Science and Technology, University of Maryland College Park, MD 20742, USA E-mail: opherOcosmicmy.umd.edu

J. J. BEATTY, S. COUTU, S. A. MINNICK AND S. NUTTER Department of Physics, Penn State University University Park, PA 16802, USA

M. A. DUVERNOIS School of Physics and Astronomy, University of Minnesota Minneapolis, MN 55455, USA

M. J. CHOI, H. J. KIM, S. K. KIM AND I. H. PARK Department of Physics, Seoul National University Seoul, 151-742, Korea

S. SWORDY Enrico Fermi Institute and Department of Physics, University of Chicago Chicago, IL 60637, USA

CREAM is slated to fly as the first NASA Ultra Long Duration Balloon (ULDB) payload in late 2003. On this 60-plus-day flight CREAM is expected to collect more direct high-energy cosmic ray events than the current world total. With three such flights CREAM is expected to have a proton energy reach above 5x1Ol4 eV, probing near 100 TeV for the predicted kink in the cosmic-ray proton spectrum. With a Transition Radiation Detector (TRD) above a sampling tungsten/scintillator calorimeter, an in-flight cross-calibration of the absolute energy scale becomes possible with heavy ions. We report on results from a 2001 beam test of the calorimeter in an SPS beam at the European High Energy Physics lab (CERN) and on the planned in-flight calibration.

133

f 34

1. Introduction CREAM Science and measurement objectives

For over half a century it has been known that charged particles from outer space constantly hit the Earth’s atmosphere. The currently accepted Supernova Remnant shock acceleration model predicts the spectra of these cosmic-ray nuclei cuts off at different energies with Ecutoff 0: 2 x 1014 eV’ . This should lead to sudden changes in elemental spectra, causing a gradual change in the elemental composition of cosmic rays between l O I 4 eV and 1015 eV. CREAM (Figure 1) is an experiment comprised of a calorimeter, a TRD and a charge detector, intended to identify incident cosmic-ray nuclei and measure their energy2. From these measurements we will extract the spectra of nuclei from Hydrogen to Iron from 10l2 eV to 1015 eV, and search for a kink in the proton spectrum near 1014 eV. With the TRD and charge detector CREAM will measure ratios of primary to secondary cosmic rays, testing theoretical models of cosmic-ray passage through the interstellar medium. To enable these science objectives, the CREAM Timing-based Charge Detector (TCD) must measure particle charge with a resolution of 0.2e, the TRD must achieve an energy resolution of 15% for various nuclei (Carbon to Iron) at energies of 10s to 100s of TeV, covering a range of lo3 < y < lo5, and the calorimeter must achieve an energy resolution of < 50% with no high-end non-Gaussian tails3.

Figure 1. Schematic drawings of CREAM with and without its support structure.

2. The CREAM Detector

The TCD4 is comprised of 8 scintillator paddles in 2 layers, covering an area of 120x120 cm2. Each paddle is read out through two adiabatic light-guides by fast photo-multiplier tubes (PMT). Each PMT is read out by an array of fast TDCs that digitize the time of several threshold levels being exceeded by the leading edge of the scintillation pulse. These times are used to reconstruct the slew-rate of the pulse, proportional to the square of the particle charge.

135

The PMTs are also read out via ADCs that digitize the peak of the pulse, for more accurate charge measurements of heavier nuclei. The TCD measurement is completed within 3 ns, avoiding noise from back-scattered secondaries. Transition radiation can be used to measure the Lorentz factor of relativistic heavy nuclei5, and with particle ID, their energy. The CREAM TRD is made up of two 35 cm thick modules. A foam matrix in each module functions as both radiator and mechanical support for 6 layers of thin aluminized Mylar tubes. Each 2 cm diameter tube contains a sense wire and a Xenonlmethane gas mixture. The T R signal is read out via Amplex-chip based circuitry. The radiator is optimized for lo3 < y < lo5. Target layers

Hodoscoper Calorimeter

Figure 2.

Schematic side-view of CREAM calorimeter module.

The calorimeter module6 (Figure 2) is comprised of a 20 radiation length (X,) sampling tungsten/scintillating-fiber calorimeter, preceded by a 0.5 Xint graphite target to induce incident particle interaction with a minimal weight penalty. Twenty 1 XO absorber plates are each followed by a layer of 0.5 mm fibers in 1 cm ribbons. Each ribbon is aluminized on one end and glued into a light-mixer on the other. Light is transferred t o the readout via a bundle of thin clear fibers, split into low-, mid-, and high-energy ranges. Neutral density filters with different transmission for the mid- and high-energy ranges (with none for the low-energy range) increase the ratio of signals between ranges further, bringing the dynamic range to the required 1:200,000. The signal is read by multi-pixel hybrid photo-diodes (HPD) with a dynamic range of up to 1:1,000,000. Half-way through the target, a scintillating fiber hodoscope provides tracking information to augment that available from the calorimeter. Above the target, two additional hodoscopes provide both more tracking information and charge information for those particles not in the TCD geometry.

3. Calibration Calibrating a hadron calorimeter system is at best not a trivial task7. When the system is intended to measure energies up to several PeV, and must be sensitive

136

to low signals (5 MeV in CREAM) and high signals (1 TeV in CREAM), the problem becomes even more challenging. When attempting to measure shower energies not only of protons, but also of projectile nuclei up to Iron, an additional level of complexity is added. Finally, for flight calorimeters the problem is exacerbated yet further by requirements of low weight, low power, limited volume, no access in-flight, and no guarantee of recovery for post-flight recalibration. To address all the above one would ideally use testbeams of separate nuclei up to the highest expected energy, over a fine grid (1 cm pitch for CREAM), and at angles up to the highest acceptance (> 70 degrees for CREAM). In addition, one would like to have a test-beam with the same composition, angular distribution and energy spectra as those of cosmic rays, with each incident particle energy, charge and particle trajectory known accurately independent of the calorimetric measurement. Such a calibration facility, unfortunately, is not available. One is thus forced to combine different aspects of the above requirements from different calibration techniques. The CREAM calibration is divided into several phases and modes. Preflight, the calorimeter is tested in high-energy electron beams to provide an accurate assessment of detector performance and allow inter-calibration of ranges for each ribbon and between different ribbons. Proton runs are taken to validate the low energy portion of the CREAM hadronic Monte Carlo simulation. Heavy ions incident on a target produce a mixture of nuclear fragments to validate simulations of showers induced by projectiles with Z> 1. Simulation is then used to extrapolate to the highest expected energies. In-flight, shower data averaged over many events are used to maintain the inter-range calibration, and the inter-calibration between ribbons. In addition, LED flasher events monitor the HPD average gain ( e g . to help account for temperature gradients between different HPDs reading out different parts of the calorimeter). Charge injection events allow monitoring of the readout electronic chains, separate from the optical portion of the readout string. Finally, an in-flight cross-calibration is planned between the TRD/TCD system and the calorimeter. The TCD is expected to measure the charge of incident particles with good resolution. With the charge known, the energy of the incident nucleus can be reconstructed from the TRD Lorentz factor measurement. Since these are all electromagnetic phenomena, they can be modeled accurately. For a sub-sample of such events, the nuclei will interact and shower in the calorimeter. The reconstructed calorimeter-measured energy can be compared to the energy obtained from the TRD/TCD for an absolute energy calibration. After collecting data for 4-6 hours we expect to have a sufficiently large sample of events to provide a cross-calibration accuracy of 10%.

137

4. Beam Test Results

In September 2001, a prototype CREAM calorimeter module was placed in the H2 beam-line of the Super Proton Synchrotron (SPS) (Figure 3). The objectives of this test were to validate the design of the calorimeter, t o measure the light yield from showers, and to validate the calorimeter simulation in the energy range available at the SPS. After calibration, the resulting longitudinal profile for 250 GeV electron showers showed remarkably good agreement with the average simulated showers of such electrons (Figure 4). The showers were well contained, with shower maximum occurring as expected, at a depth of 8 XO.Lateral profiles showed the core of the shower is nearly contained in a single fiber ribbon (1 cm width) with negligible tails beyond a 3 cm width.

Figure 3.

CREAM prototype calorimeter at the H2 beam-line.

After calibrating the readout gain, the signal from the highest-signal ribbon in layer 8 was translated to photoelectrons (p.e.). Simulations were used to estimate the energy deposit in the fiber ribbon, expressed in terms of minimum ionizing particle (MIP) energy deposit. The light yield measured by the HPD was corrected for the measured efficiency of light transmission from the ribbon through the light-mixer and clear fibers, indicating a light yield of 3.9 p.e./MIP at the end of the ribbon. A Gaussian fit to measurements with a lo6 Ru source above a fiber ribbon coupled directly to a P M T provided a value of 3.2 p.e./MIP, within 25% of the beam test measurement.

5. Conclusions Beam test results confirm the CREAM calorimeter works as expected based on high energy electron shower shapes and light yield. The CREAM calorimeter

138

Beam test result : 250 GeV electron 0

experiment

- simulation

Layer number Figure 4. Experimental vs. simulated longitudinal shower profiles for 250 GeV electrons.

will be subjected to more detailed beam testing pre-flight to achieve full calibration. Calibration will be maintained through 60 or more day/night cycles by use of periodic in-flight stand-alone and cross-calibrations. Stand-alone calibrations will consist of LED events, charge-injection events, and shower data. Cross-calibration will utilize heavy nuclei crossing the TCD and TRD, interacting and showering in the calorimeter to obtain two independent measurements of incident energy, allowing absolute energy calibration.

Acknowledgments This work was supported by NASA grant NAG5-5249. The CREAM collaboration thanks John Mitchell of NASA/GSFC for coordinating the beam tests, and CERN for providing excellent beams and support for the reported testing.

References 1. P. 0. Lagage and C. J. Cesarsky, Astron. and Astroph., 118,223 (1983). 2. E. S. Seo et al. (CREAM Collaboration), Adv. in S p . Res. in press (2001). 3. H.S. Ahn et al. (CREAM Collaboration), Proc. 27th Int. Cosm. Ray Conf. (Hamburg), 6,2159, (2001). 4. J. J. Beatty et al. (CREAM Collaboration), Proc. 2sth Int. Cosm. Ray Conf. (Salt Lake City), 5, 61, (1999). 5. S. Swordy et al., Phys. Rev. D42,3197, (1990). 6. 0. Ganel et al. (CREAM Collaboration), Proc. 27th Int. Cosm. Ray Conf. (Hamburg), 6,2163, (2001). 7. 0. Ganel and R. Wigmans, Nucl. Instr. and Meth. A409, 621, (1998).

VERITAS: A NEXT GENERATION ATMOSPHERIC CHERENKOV DETECTOR AND CALORIMETER FOR GAMMA-RAY ASTRONOMY

F. KRENNRICH Physics d Astronomy Department, Iowa State University, Building 12 Ames IA 50011, USA E-mail: krennrichOiastate.edu The Very Energetic Radiation Imaging Telescope Array System (VERITAS) is a wide energy range (50 GeV - 50 TeV) imaging atmospheric Cherenkov detector that will provide a high sensitivity and good energy resolution for astrophysical y-ray sources. Recent discoveries of y-ray blazars have opened the possibility of prqbing the intergalactic IR fields by analyzing the shape of TeV y-ray spectra. Also, the search for the origin of cosmic rays using secondary y-rays requires accurate energy spectral measurements. The technical concept and design of VERITAS and its capabilities for calorimetric measurements are discussed.

1. Introduction

The objective of this paper is to give a short overview of the VERITAS instrument and discuss its capabilities to make energy spectral measurements. The third EGRET catalogue (Hartman et al. 1999) with > 270 y-ray sources has established the field of y-ray astronomy as a new discipline providing a wealth of information about supernova remnants, pulsars, cosmic rays, active galaxies and y-ray bursters. The extension of y-ray observations to TeV energies by atmospheric Cherenkov telescopes has resulted in a total of more than ten TeV sources (Weekes 2001), indicating the tip of the iceberg of the high energy end points of astrophysical photon spectra. Probing the physical emission processes of astrophysical y-ray sources, e.g., inverse Compton scattering of soft photons to high energy photons by relativistic electrons, often involves the measurement of energy spectra over several orders of magnitude in energy. This requires y-ray telescopes with a large dynamic range covering sub-GeV t o multi-TeV energies. Typically, these spect r a are approximated by power laws with curvature and exponential cutoffs, providing information about the physical conditions at the y-ray source and external absorption (pair production) phenomena. Energy reconstruction with sufficient resolution, accurate calibration and well understood systematic un-

139

140

certainties over a large energy range is key to understanding the physics of y-ray emission processes and absorption features as seen in some active galaxies (Aharonian et al. 1999; Krennrich et al. 2001). Hence, energy reconstruction and calibration play an important role for the next generation of space-based (GLAST; see Gehrels & Michelson 1999) and ground-based y-ray telescopes VERITAS (Weekes et al. 2002), HESS (Hofmann et al. 1999), MAGIC (Lorenz et al. 1999) and CANGAROO I11 (Matsubara et al. 1999). Together, GLAST and the next generation imaging atmospheric Cherenkov telescopes (IACTs) cover energies between 20 MeV to 50 TeV, providing the energy range necessary to constrain astrophysical yray emission mechanisms, making the combined spectra extremely valuable. Space-based pair conversion telescopes can be well calibrated to a few percent accuracy using laboratory beams. Calibration of IACTs requires, due to the lack of a TeV y-ray test beam, extensive modeling of air showers and the earth's atmosphere. However, the next generation of satellite instrument GLAST and IACTs have substantial overlap in energy. This will allow a cross-calibration using an astrophysical standard candle, the most prominent one is the Crab nebula. By measuring the Crab Nebula spectrum the systematic uncertainties for the combined GeV and TeV spectra can be substantially reduced. In the following the VERITAS design and its implications for energy reconstruction, dynamic range and calibration will be discussed. 2. The VERITAS concept The VERITAS proposal is to built an array of seven IACTs of 10 m aperture. The location of the array will be at the Whipple Observatory in southern Arizona. The individual telescopes will be placed on a hexagonal grid with 80 m spacing (see Fig. 1). Individual telescopes will be based on the proven design of its predecessor, the Whipple 10 m reflector, which has been the pioneering instrument for the field of ground-based y-ray astronomy (Weekes et al. 1989). It has an energy range of 200 GeV - 20 TeV and a sensitivity of 7 0 ,/for a flux level of 1 Crab. The VERITAS detector is primarily aimed at a substantially improved flux sensitivity and reduced energy threshold, but will also provide a better energy resolution and angular resolution over existing instruments. In particular, the dominant design goal has been maximum sensitivity in the energy range of 100 GeV - 10 TeV. Minimum detectable fluxes (5 sigma in 50 hours) will be 0.5% of the Crab Nebula at 200 GeV, a factor of 20 better than the Whipple observatory 10 m telescope. VERITAS provides an unprecedented angular resolution of 0.05' (0.03') at 300 GeV (1 TeV), substantially better than any existing y-ray telescope on the ground or in space.

141

Figure 1. The layout of the stereoscopic system of telescopes as at would be located at Montosa Canyon near Tucson Arizona.

The individual telescopes will be substantially improved over the Whipple 10 m telescope, providing adequate optical resolution, a wide FOV and fast and low noise electronics, in order to maximize the performance of VERITAS. The major improvements of the design are shortly described in the following.

2.1. Optics The design of the optical reflector of the VERITAS telescopes provide much better resolution to resolve the intrinsic characteristics of 7-ray Cherenkov images. A pixelsize of M 0.03" - 0.1", across a 3.5' field of view (FOV) would be desirablea. For practical considerations, cost of phototubes and readout, "Shower fluctuations limit the inherent accuracy of resolving the Cherenkov images, t o M 0.03' (300 GeV) as shown by Hillas (1989).

142

the upper end of 0.15" is a reasonable compromise. The optical reflectors will be a Davies-Cotton design (Davies & Cotton 1957) consisting of facet mirrors, each with a 24 m radius of curvature. The Davies-Cotton (for details see Weekes et al. 2002) design has off-axis aberrations smaller than a parabolic reflector, showing good image quality out to a few degrees from the optic axis (Lewis 1990). For example, 100% of the light from a point source is concentrated in a 0.12" diameter circle out t o 1.0" from the optical axis. The Davies-Cotton reflector design is not isochronous, however, the time spread has a full width of 3 - 4 ns, comparable to the intrinsic Cherenkov pulse width.

2.2. Camem

Fig. 2 shows the layout of the focal plane detector for VERITAS. It consists of 499 photomultipliers with 0.15" spacing, corresponding to a FOV of 3.5". The pixelation and the FOV are two competing factors given a limited number of phototubes. The choice of pixel spacing is driven by the structure of y-ray shower images. In order to trigger efficiently on low energy y-ray events (E x 100 GeV), it is necessary that the image width is approximately matched by the pixel size - a pixelation larger than the image width would accept a large noise contamination from night sky background light, reducing the signal to noise ratio. In Fig. 2 the Cherenkov light image of a 100 GeV y-ray shower is superimposed on a VERITAS camera. Images of sub-TeV showers have a RMS width and length (depending on energy) of 0.10" - 0.15" and 0.20" - 0.30", respectively. The camera pixelation (0.15") and excellent optical quality of the reflector will provide sufficient resolution to trigger efficiently and resolve useful shower structure on this scale. The Cherenkov light from air showers is dominantly emitted from shower maximum (at 8-10 km atmospheric height) and the images are off-set from the arrival direction by 0.6" - 1.2", depending on the impact distance of the shower from the telescope. A 2.5" FOV is required to capture most of the Cherenkov image. However, a larger FOV is necessary for resolving extended sources, the detection of multi-TeV y-rays (especially for E > 10 TeV), and stereoscopic operation, making 3.5" a better choice. A large FOV also improves the energy resolution by imaging the shower over the entire longitudinal development by capturing y-ray images that are fully contained within the camera.

143 -2.5

I

3 t?b o 2 2, Q)

.

,.

.

.

1.5

1

0.5

0

-0.5

-1

-1.5

-2

100 GeV p h o t o n -2.5

_

-2.5

.

-2

.

I

. . . .

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

X (degrees) Figure 2. The 499-pixel photomultiplier matrix for VERITAS. The pixel spacing i s 0.15' and the FOV is 3.5'. A 100 GeV 7-ray shower image is projected into the focal plane.

2.3. Electronics In order to produce a digital image of the Cherenkov light flash with minimal noise contamination, the electronics of the recording system has to preserve the short (6-10 ns) pulse until it is digitized. To minimize the noise in the electronics, the photomultiplier (PMT) signals are amplified using fast preamps. Preliminary tests with a prototype of 30 channels showed an electronic noise level of 0.2 mV r.m.s. at 500 MHz sampling frequency. This allows also an in situ calibration of the gain of the electronics system using single photoelectron (p.e., hereafter) pulses (1 p.e. x 2 mV). Preamps also enable to operate the PMTs at a lower gain allowing observations during moon light, hence in-

144

creasing the duty cycle of VERITAS by a factor of two in comparison to the Whipple 10 m telescope. The electronics also has to form a trigger decision based on the characteristics of a typical ?-ray image in an array of IACTs. The atmospheric Cherenkov technique is ultimately limited by the fluctuations from the night sky background. To estimate the lowest possible energy threshold for the VERITAS array, a careful analysis of the expected trigger rates from accidentals due to the night sky is requiredb. The trigger threshold, given in p.e., determines the energy threshold of the IACT array. The trigger of VERITAS is formed in a sequence, based on the different levels in the electronic chain. The level 1 trigger is simply a constant fraction discriminator. In order t o enable efficient triggering on compact y-ray images, a pattern trigger (level 2) can be programmed to select patterns of 2 N adjacent pixel (N = 2, 3 or 4) with a coincidence time window of 14 ns. This ensures that the accidental trigger rate from random night sky fluctuations is at a minimum level. A trigger condition of 2 3 adjacent pixels reduces the rate to 100 kHz. These local level 2 triggers from each telescope are transmitted by digital optical fiber cable to the central station. The level 2 trigger are sent through individual digital delays to account for the orientation of the shower front (delay range: 0-500 ns). The array trigger is required to be flexible for various operation modes: a single telescope trigger, using three and four telescopes independently, a trigger requiring 3 out of seven telescopes. At a threshold of 5 p.e., the array trigger (3 out of 7) produces a negligible background rate from random night sky fluctuations, at 4.2 p.e. the accidental rate is M 300 Hz. The digitization of the signals can begin, once a trigger decision has been formed. The trigger decision based on an array trigger of 3 telescopes can take up to 1 . 2 ~ assuming s operation at large zenith angles. A Flash ADC system (Buckley et al. 1999) at each telescope allows the digitized output samples of each channel to be written into a circulating memory (depth > 8ps), while waiting for the trigger decision. If a trigger arrives, the writing stops and the memory contents are examined for a signal in the corresponding time bin. 2.4. Performance of VERITAS: Simulations

Monte Carlo simulations have been used to determine the optimum configuration and to characterize the performance of VERITAS (Vassiliev et al. 1999). The design optimization was performed by the means of a full Monte Carlo bFor practical consideration, the array must trigger at a rate < 1 kHz, a higher rate would introduce significant dead time for the data acquisition system.

145

simulation of air showers and the telescopes. Subsequent event reconstruction is based on established methods developed for single telescope data (Fegan 1997) and extended through several new algorithms, allowing a lower reconstruction energy threshold. Realistic operating conditions, e.g., a realistic night sky background, are given in form of input parameters such as FADC integration time (8 ns), pixel coincidence gate (14 ns), the level l trigger threshold (4.2 pe) and level 3 (telescope array) trigger coincidence gate width (40 ns).

Table 1. Number of telescopes Telescope spacing Reflector aperture/area Focal length Field of View (FOV) Number of pixels Pixel Spacing/Photocathode Size Array Trigger Telescope Triggers

7 (hexagonal layout) 80 m 10 m / 78.6 m2 12 m 3.5'

499 0.148' / 0.119' 3 telescopes out of 7 2, 3 pixels (adjacent)

The VERITAS design is optimized for maximum point source sensitivity between 100 GeV - 10 TeV. The outcome of these studies has resulted in the so-called baseline design for VERITAS given in Table 1. For further details of the optimization process see Weekes et al. (2002). The ability to detect y-ray sources can be described by the flux sensitivity of the detector. We define the minimum detectable flux of y-rays requiring a 5u excess above background (or at least 10 photons) for 50 hours of observation assuming a source spectrum given by dN/dE 0; E-2.5. The y-ray flux sensitivity of the VERITAS baseline design as a function of energy is shown in Fig. 3. The flux sensitivity is limited by different effects depending on energy. The region above M 2 TeV is limited by the collection area and therefore by photon statistics. At lower energies (200 GeV - 900 GeV) the cosmic ray electrons are the major source of background. The flux sensitivity below 200 GeV is limited by night sky background and cosmic-ray protonsc. In the lower curve, a low night sky background is assumed, the less sensitive curve is for a bright region in the sky, e.g., the galactic plane. Further performance details are given in Table 2and Weekes et al. (2002).

'Note, that a cosmic-ray muon can trigger the VERITAS array, but is easily rejected by the presence of a hadronic shower halo in at least one of the cascade images or by parallax.

146

Figure 3.

The sensitivity of VERITAS for a point-source in 50 hours.

3. Energy Reconstruction and Calibration The IACT technique provides a good energy resolution for the measurement of y-ray induced air showers. The atmosphere is a fully active calorimeter with Cherenkov light emitted at all stages of the shower development. For yray induced showers above 20 GeV the Cherenkov light yield is approximately proportional to its primary energy. If the shower geometry can be measured with its location of the shower axis, impact point on the ground, height of shower maximum, and the total light density at the ground, the y-ray primary energy can be estimated with good accuracy. Simulations for VERITAS indicate that an energy resolution for individual y-ray showers of 10% - 20% can be reached. The energy resolution AE/E

147 Table 2. Characteristic Energy threshold” Flux sensitivityb

Angular resolution

Effective area Crab Nebula rate

VERITAS performance E

>lo0 GeV >300 GeV >1 TeV 50 GeV 100 GeV 1TeV 50 GeV (100 GeV) 300GeV (1TeV) >lo0 GeV

Value 75 GeV 9 . 1 10-12cm-2s-1/15 ~ mCrab 8 . 0 ~ 1 O - ~ ~ c r n - ~ smCrab -~/5 1 . 310-13cm-2s-1/7 ~ mCrab 0.14‘ 0.09O 0.03’ l.0x103m2 (l.0x104m2) 4.0x104m2 (l.0x105m2) 50/minute

aEnergy at which the rate of photons per unit energy interval from the Crab Nebula is highest. ‘Minimum integral flux for detecting a 5a excess (or a minimum of 10 events) in 50 hours of observations of a source with a Crab-like spectrum.

depends somewhat on the primary energy and ranges from M 20% at 100 GeV to 10% at 10 TeV, when using an algorithm that is based on shower core location and total light density. This can likely be improved using more advanced reconstruction algorithms taking into account a measurement of the height of shower maximum. This resolution is adequate to measure energy spectra that are power laws with curvature and/or exponential cutoffs. It is also a good match to the accuracy of spectral measurements at GeV y-rays (GLAST at 100 GeV: 10%) and X-rays (few %). For the physical interpretation of spectral cutoffs and the estimate of yray fluxes an absolute energy calibration of 10% or better would be desirable. Again, space-based telescopes achieve that by laboratory calibrations with a test beam. In contrast, the absolute calibration of the energy scale for IACTs is more difficult with a number of inherent systematic uncertainties: 1. atmospheric transparency, 2. response of the telescope, 3. variations in the instrument and atmosphere. The Cherenkov light transmission through the atmosphere is typically modeled using the “U.S. Standard Atmosphere, 1976” (U.S. Standard Atmosphere 1976). Cherenkov light traversing the earth’s atmosphere is attenuated by Rayleigh scattering, Mie scattering (Aerosols), and 0 2 and 0 3 . Pressure variations, changes in the Aerosol and Ozon concentration and their effects on the light throughput of the atmosphere and subsequent impact on energy reconstruction have been studied by Lewis (1997) and Krennrich et al. (1999). The uncertainty on the absolute energy was found t o be typically at the 5% - 7% level. An overall systematic uncertainty of the atmospheric model is estimated to 10%. The response of a Cherenkov telescope can be calibrated t o 5% - 10%

148

accuracy including mirror reflectivity, light concentrator efficiency, quantum efficiency of the photodetectors, cable losses etc. Variations in the relative light throughput of the atmosphere and drifts in the instrument gain can be monitored using the cosmic-ray rates and spectrum (see also Le Bohec & Holder 2002, in preparation). The cosmic-ray rate in the Whipple telescope is typially 20 Hz allowing a 5% accurate relative calibration for 30 minutes of data taking. For VERITAS this could be done on shorter time scales with better accuracy. Together, these effects could result in a 25% uncertainty (sum of errors) for the absolute energy estimate of IACTs, reason enough to develop more accurate calibration procedures. An alternative method could be to capitalize on the accurate calibration of space-based y-ray detectors. The substantially increased collection area of GLAST between 10 GeV - 1 TeV and the lower energy thershold of IACTs (VERITAS 50 GeV, MAGIC 20 GeV) might provide sufficient overlap in energy between the space-based and ground-based technique. The y-ray emission from the Crab Nebula is a strong and constant y-ray source that has its peak emission between 50 GeV - 100 GeV (Fig. 4) and is well suited as a test beam for a cross calibration between satellite and atmospheric telescopes. A calibration data set with sufficient statistics to use the overlap between VERITAS and GLAST (50 GeV - 300 GeV), requires M 40 days of exposure for the GLAST telescope, providing M 160 photons above 10 GeV. This estimate is based on the energy spectral fit shown in Fig. 4 of Hillas et al. (1998) with a differential spectrum of dN/dE 0: E-2.44-0.1510gloE and collection areas given in the GLAST proposal. We require, that at least lO(2) photons are detected above lOO(300) GeV, providing ample overlap with VERITAS between 50 GeV - 300 GeV. The ground-based instruments, due to their large collection areas (10, 000m2 for VERITAS at 100 GeV) are capable of collecting sufficient statistics on much shorter time scale. A calibration between GLAST and VERITAS could be achieved by normalizing the absolute fluxes to the well calibrated GLAST spectrum by allowing the spectrum of the IACT to shift in energy.d. With the assumed statistics for the GLAST spectrum, the flux normalization would carry a M 15% statistical uncertainty, translating in a 10% accuracy for the absolute energy calibration. The exercise shown here serves exemplary character and a more precise estimate needs to be done in the future. A curved spectrum (see Fig. 4) could also be used to perform a direct energy calibration, independent from the collection area, by measuring the y-ray emission peak for both instruments and determine the offset. The peak emisdThis uses the fact that the collection areas of IACTs can be well calculated using simulations.

149

Figure 4. The Whipple and EGRET observations of the Cmb unpulsed ?-myspectrum are shown above (f.om Hillas et al. 1998). The full-line curves are predicted inverse Compton fluxes for three different assumed magnetic fields in the region where the TeV T-mys are produced.

sion for the Crab Nebula seems to occur between 50 GeV - 100 GeV which is well in the range of overlap between GLAST, MAGIC and VERITAS. After a few years of GLAST operation the calibration range could be extended to 1 TeV. This of course needs substantially more statistics and could be considered at a later stage of the GLAST and VERITAS operation. 4. Conclusions

A next generation IACT with a flux sensitivity of a few mCrab and an energy range covering three orders of magnitude (50 GeV -50 TeV) is under development in the northern hemisphere. VERITAS will provide a highly sen-

150

sitive view of the high energy universe beyond GeV y-ray energies. VERITAS will complement GLAST, the next generation high energy y-ray instrument in space. GLAST with a wide FOV but small effective area, together with VERITAS with a smaller FOV but large effective area will provide for the first time a continuous coverage for y-ray observations between 20 MeV and 50 TeV. A good absolute energy calibration between northern hemisphere IACTs and GLAST could be achieved using the Crab Nebula spectrum as a test beam. This would enable accurate energy spectral measurements over from MeV TeV energies for astrophysical studies.

Acknowledgments If would like to thank David Carter-Lewis for reading the manuscript. This research is supported by grants from the U.S. Department of Energy. References F.A. Aharonian et al., A&A 349, 11 (1999). J.H. Buckley, et al., 26th ICRC (Salt Lake City) 5 , 267 (1999). J.M. Davies, & E.S. Cotton, J . Solar Energy Sci. and Eng. 1, 16 (1957). D.J. Fegan, J. Phys. G 23, 1013 (1997). 5. N. Gehrels, & P. Michelson, in TeV Astrophysics of extragalactic Sources, Astroparticle Physics 11,277 (1999). 6. R.C. Hartmann et al., ApJS 123, 79 (1999) 7. A.M. Hillas et al., ApJ 503, 744 (1998). 8. A.M. Hillas, Very High Energy Gamma Ray Astronomy, eds. A . A . Stephanian, D.J. Fegan & M.F. Cawley, 134 (1989). 9. W. Hofmann et al., in GeV-TeV Astrophysics: Towards a Major Atmospheric Cherenkov Detector VI (Snowbird), 500 (1999). 10. F. Krennrich et al., ApJ 511, 149 (1999). 11. F. Krennrich et al., ApJL 560, L45 (2001). 12. D.A. Lewis, Exp. Astron. 1, 213 (1990). 13. D.A. Lewis, internal report of Whipple collaboration, (1997). 14. E. Lorenz et al., i n GeV-TeV Astrophysics: Towards a Major Atmospheric Cherenkov Detector V I (Snowbird), 510 (1999). 15. Y. Matsubara et al., in GeV-TeV Astrophysics: Towards a Major Atmospheric Cherenkov Detector V I (Snowbird), 447 (1999). 16. United States Committee on Extension to the Standard Atmosphere US. Standard Atmosphere 1976 in Supt. of Docs., U.S. Govt. Print. Off. (1976). 17. V.V. Vassiliev, et al., 26th ICRC (Salt Lake City) 5,299 (1999). 18. T.C. Weekes, A I P Proc, of the International Symposium on High Energy Gamma-Ray Astronomy (Heidelberg), eds. Aharonian, F.A., Volk, H., 558, 15 1. 2. 3. 4.

(2001). 19. T.C. Weekes et al., Astroparticle Physics 17,221 (2002). 20. T.C. Weekes, ApJ 342, 379 (1989).

PIERRE AUGER OBSERVATORY: THE WORLD’S LARGEST CALORIMETER

A. K. TRIPATHI Department of Physics and Astronomy, University of California Los Angeles, Califronia, USA E-mail: arunOphysics. ucla.edu

The Pierre Auger Observatory is designed to observe and study a high statistics sample of ultra high energy cosmic rays (UHECR) with energy greater than 10’’ eV. The hybrid nature of the observatories, using both air-fluorescence and ground water Cherenkov array, will enable Auger to measure the energy of the cosmic ray showers with an improved accuracy compared to previous experiments. With a surface area of 3000 Km’, and an effective aperture of at least 7350 Km’Sr for each observatory, the P A 0 will be the world’s largest calorimeter to date, and will be able to conclusively determine whether the cosmic ray spectrum extends beyond the GZK cutoff. The Auger detector components, and the status of the observatory are described in this report. N

1. Introduction Observation of Ultra high energy cosmic rays (UHECR) with energies in excess of lo2’ eV poses as yet unanswered fundamental questions about their origin, acceleration mechanism, and propagation through space. The existing experimental data does not conclusively establish the existence or absence of GZK cutoff. The cosmic ray flux measured by AGASAl is inconsistent with those measured by the Fly’s Eye2 and HirRes3 experiments for energies greater than lo2’ eV. One possible source of this discrepancy between the two experiments could be a systematic difference in their energy calibration. AGASA is a ground array, measuring the energy of the cosmic ray showers by sampling the charged particle density at the ground, whereas HiRes is an air fluorescence experiment , measuring the shower energy and longitudinal development by detecting the nitrogen fluorescence light. The sources of systematic errors inherent in these two techniques are entirely different. Only a large hybrid detector can conclusive answer the question of the GZK cutoff. The large size will provide a sufficiently high statistics within a reasonable time period, and the hybrid technique will provides a greater degree of accuracy in the shower energy and composition estimates.

151

152

The Pierre Auger Observatory is such a hybrid detector, employing both ground array and air-fluorescence techniques simultaneously. A subset (-J 10%) of the events observed by the PA0 would be detected both by the ground array and the fluorescence detectors, thus allowing for important cross-calibration of the shower energy. This information then can be used to reconstruct the high statistics data sample obtained from the the surface detectors (because of their 100%duty cycle), enabling Auger to conclusively answer whether cosmic ray flux extends beyond the GZK cutoff. The details of the P A 0 detectors, and the status of the observatory are described below. 2. The Surface Detectors The ground array in the PA0 observatory consists of -1600 water Cherenkov detectors, arranged on a hexagonal grid with 1.5 Km spacing. Figure 1 shows the layout of the southern Auger observatory, currently under construction in Argentina. A schematic of the surface detectors is shown in Figure 2. Each surface detector is a cylindrical plastic water tank, with 3.6 m diameter. The interior of the tanks is lined with diffuse white laminated Tyvek layer. The tanks are filled with purified water to a height of 1.2 m, which corresponds to 3.3 radiation lengths. As a result, electrons and photons deposit most of their energy in the detector. Three PMTs are symmetrically mounted on the top of the tanks, with the photocathode facing vertically down. This arrangement of PMTs gives a uniform response to particles entering the detector at different locations. The PMTs used in the surface detectors are 9-inch diameter Photonis XP1805/D1, a custom made for Auger requirements5 in order to satisfy the requirement of good linearity over the needed large dynamic range of (- 5050000) photoelectrons/25 ns at low operating gain of 2 x lo5. These PMTs are not magnetically shielded in the field, but are oriented in a manner that minimizes the effect of earth’s magnetic field6 on the collection efficiency of photoelectrons in order to achieve good energy resolution. Each surface detector is a fully independent unit, and includes a local solar power supply, a computer, two flash ADCs, a GPS unit, a time tagging board, and a radio7. The PMT signals first go through a 20 MHz filter, and then

-

aThese PMTs will be used in the production version of surface detectors, which make up nearly all of the surface array. The 40 tank engineering array uses three different kinds of PMTs: Hamamatsu R5912, ETL 9353KB, and Photonis XP1802/FLB. These PMTs were already available in the market before the PMTs custom designed for Auger were manufactured.

153

Figure 1. The layout of the southern Auger observatory in Argentina. The small dots indicate the location of the surface detectors, and the four big dots indicate the location of the four fluorescence detectors according to the original design. After a recent review, it was decided to move the central FD to the northern edge of the array (near the top of the figure).

are digitized with a 40 MHz, 10 bit flash ADC (FADC). In order to achieve the needed dynamic range, the PMT base provides two signals: one from the anode (the low-gain channel) and the other from the last dynode, which is amplified using an AD8011 amplifier (the high-gain channel). The two signals are then digitized separately by two 10 bit FADCs, thus providing 15 bits of effective dynamic range. 3. The Air Fluorescence Detectors

The Auger surface detector array will be viewed by four fluorescence detectors (FD)8*9located along the periphery of the array. Each FD consists of 6 telescopes, each with a 30' x 30' field of view. A schematic of an Auger FD telescope is shown in Figure 3. Each telescope has a mirror system to collect and focus the fluorescence light, and a pixelized camera to detect the light emitted from different regions of the shower path. Auger used Schmidt optics, with a spherical mirror of 3.4 m radius. The

154

?-

t

Figure 2.

3.6 m

A schematic view of the Auger surface detectors.

fluorescence light emitted from a shower first passes through a UV transmitting filter which reduces the dark sky noise by a factor of 8, while allowing most of the nitrogen fluorescence light to pass through. A circular diaphragm of 1.7 m diameter, centered on the center of curvature of the spherical mirror, defines the aperture of the telescope. The shape of the camera is a section of a sphere with a radius of 1.743 m, located concentric to the spherical mirror. The light reaching the camera is detected by a hexagonal array of 440 PMTs (Photonis XP3062), which provide 1.5' x 1.5' angular resolution. The details of the FD electronics can be found in Gemmeke et al.''. 4. Communication, DAQ and Triggering

The 1600 surface detectors are separated by 1.5 Km and are spread over distances of up to 50 Km. similarly, the distance between any two fluorescence detectors is also quite large. This makes it prohibitively expensive to connect all the detectors with cables. As a result, Auger relies on radio communication to communicate between the detectors and the central data acquisition building. A detailed description of the the radio communication network is available elsewherell. The details of the Auger trigger and DAQ are also available in references be lo^^^^^^' ,14. 5. Aperture and Event Rates

-

Each observatory, when complete, will cover an area of -3000 Km2, providing an effective aperture of 7350 Km2Sr for cosmic rays with zenith angle less

155

UV Filter, corrector ring Figure 3.

A drawing showing the components of a fluorescence detetcor.

than 60'. If events with zenith angle greater than 60' accepted, the aperture increases by 50%. With the inter-detector spacing of 1.5 Km, the array will be fully efficient for cosmic rays with energy greater than 10'' eV. The fluorescence detectors can only be operated during clear, moon-less nights.This limits their duty factor to 10%. As a result, most of the data will come from the surface detectors, since they can be operated round the clock. Table 1 shows the expected event rates for different energies. Note that the event rates beyond 5 x 1019 eV are estimated assuming AGASA' spectrum.

-

6. Calibration and Energy Resolution The calibration of the surface detectors15 is carried out by through going muons. The calibration of fluorescence detectors, however, must be carried out with an absolutely calibrated light source''. A lidar system17J8 using the laser backscattering method will provide information about the atmospheric attenuation length. The energy resolution in Auger has the following components:

156 Table 1. Expected event rates from the southern Auger observatory. Event rates above 5 x 1019 eV are estimated using AGASA spectrum Energy (eV)

23x 21x 22x 25x 21x 22x

10'8

Surface Detectors

Fluorescence Detectors

(Number of Events)/Year

(Number of Events)/Year

15000

4700

1019

5150

515

1019

1590

159

1019

490

49

1020

103

10

1015

32

3

(1) the MC error due to the uncertainty in the physics interaction models at the energies involved, (2) In the case of surface detectors, shower-to-shower fluctuations, and (4)In the case of fluorescence detectors, contamination from Cherenkov light (both direct and scattered) and (5) any systematic errors in the detector calibration. MC s t ~ d i e s ' ~show ~ ~ ' that the expected RMS energy resolution from surface detectors alone is 12% averaged over all energies, assuming the primary cosmic rays consist of equal numbers of proton and iron. The angular resolution is better than 1.1' for all energies, falling to 0.6O for showers with energy above 1020 eV. In the hybrid mode, the expected energy resolution is approximately 10% and the angular resolution is 0.5% for showers with energies above 1020 eV.

7. Status of Auger The first Auger observatory is under construction near Malargue ( latitude = -35.2', longitude -69.2') in Mendoza Province of Argentina. This site is located at a mean altitude of approximately 1400 m above sea level. The construction of an engineering array of 40 surface detectors and two fluorescence telescopes was completed in the December of 2002. Figure 4 shows the picture of a fully instrumented and functional tank, and Figure 5 shows the prototype FD detectors used in the engineering run. The purpose of this array was to test, debug and optimize all the detector components, communications, and the data acquisition system. This array successfully ran in hybrid mode until the end of March 2002, observing several hybrid showers. The FD telescopes were then dismantled in order t o install the production version, but the surface detectors have been in operation continuously. At the the time of writing of this report, the deployment of new surface and fluorescence detectors is already

157

Figure 4. A picture of a fully instrumented and functional surface detector in the engineering array, currently under operation in Malargue, Argentina. The solar panel, the battery box and the communications antenna are visible.

Figure 5 . Pictures of the prototype FD detector system used in the engineering run. The aperture system and the camera can be seen on the left, and the mirror on the right.

in progress, and the southern observatory is expected to be complete by the end of 2004.

8. Summary Pierre Auger Observatory will be the world’s largest calorimeter to date, utilizing the atmosphere as a giant calorimeter. The hybrid nature of the ob-

158

servatory will allow Auger to measure t h e energy as well as composition of t h e UHECR with a n accuracy not possible before. This, combined with the large aperture, will allow Auger to conclusively settle t h e question of the GZK cutoff, a n d help solve the mystery still surrounding UHECR.

References 1. M. Takeda et al., Phys. Rev. Lett. 81, 1163 (1998). 2. D. J. Bird et al., Phys. Rev. Lett. 71,3401 (1993). 3. C. H. Jui, for the HiRes collaboration, in Proceedings of the 27th International Cosmic Ray Conference, Hamburg, (2001); F. Halzen and D. Hooper, submitted

to Rep. Prog. in Phys. (astro-ph/0204527) .

4. M. T. Dova, in Proceedings of the 27th International Cosmic Ray Conference, Hamburg, (2001). 5. A. Tripathi et al., “Tests of New Extended Dynamic Range PMTs for Auger Project”, Pierre Auger Project Technical Note GAP-2001-049, (2001). 6. A. Tripathi, K. Arisaka, T. Ohnuki, and P. Ranin, “Effect of Earth’s Magnetic

Field on Production Photonis PMTs”, Pierre Auger Project Technical Note GAP2002-013, (2002). 7. T. Suomijarvi, in Proceedings of the 27th International Cosmic Ray Conference, Hamburg, (2001). 8. G. Matthiae, in Proceedings of the 27th International Cosmic Ray Conference, Hamburg, (2001). 9. R. Cester, in Proceedings of the 27th International Cosmic Ray Conference, Hamburg, (2001). 10. H. Gemmeke, in Proceedings of the 27th International Cosmic Ray Conference, Hamburg, (2001). 11. P. D. J. Clark and D. Nitz, in Proceedings of the 27th International Cosmic Ray Conference, Hamburg, (2001). 12. D. Nitz, in Proceedings of the 27th International Cosmic Ray Conference, Hamburg, (2001). 13. R. Meyhandan, J. Matthews, D. Nitz, in Proceedings of the 27th International Cosmic Ray Conference, Hamburg, (2001). 14. H. Gemmeke et al., in Proceedings of the 27th International Cosmic Ray Conference, Hamburg, (2001). 15. H. Salazar, L. Nellen, L. Villasenor, in Proceedings of the 27th International Cosmic Ray Conference, Hamburg, (2001). 16. H. 0. Klages, in Proceedings of the 27th International Cosmic Ray Conference, Hamburg, (2001). 17. A. Filipcic et al., in Proceedings of the 27th International Cosmic Ray Conference, Hamburg, (2001). 18. J. Matthews and Roger Clay, in Proceedings of the 27th International Cosmic Ray Conference, Hamburg, (2001). 19. M. Ave, J. Lloyd-Evans and A. A. Watson, in Proceedings of the 27th International Cosmic Ray Conference, Hamburg, (2001). 20. B. Dawson and P. Sommers, in Proceedings of the 27th International Cosmic Ray Conference, Hamburg, (2001).

Crystal Calorimetry Covener: W. Wisniewski

H.-C. Huang

Performance of a Small Angle BGO Calorimeter at BELLE

M. Kocian

Performance and Calibration of the Crystal Calorimeter of the BaBar Detector

T . Hryn’ova

A Systematic Study of Radiation Damage t o Large Crystals of CsI(T1) in the BaBar

B. A. Shwartz

Performance and Upgrade Plans of the BELLE Calorimeter

Q. Deng

Development of Yttrium Doped Lead Tungstate Crystal for Physics Applications

A. Gasparian

Performance of PWO Crystal Detectors for a High Resolution Hybrid Electromagnetic Calorimeter at Jefferson Lab

R. Novotny

The PHOTON BALL at COSY

*P. Lecomte

Overview of the CMS Electromagnetic Calorimeter

F. Cavallari

Performance of the PWO Crystals of the CMS Electromagnetic Calorimeter

R. Rusack

Avalanche Photodiodes for the CMS Lead Tungstate Calorimeter

E. Auffray

CMS/ECAL Barrel Construction and Quality Control

*Written contribution not received

159

This page intentionally left blank

PERFORMANCE OF A SMALL ANGLE BGO CALORIMETER AT BELLE

H.-C. HUANG National Taiwan University, Taipei, Taiwan 106, R. 0.C. E-mail: [email protected] (For the Belle Collaboration)

The Extreme Forward Calorimeter (EFC) at Belle has been operating since the spring of 1999 on KEKB. It consists of 320 radiation-hard BGO crystals surrounding the beam pipe to cover the small angle region in both forward and backward directions. The main function of EFC is to provide on-line luminosity information using Bhabha scattering and it also acts as a tag for two-photon production events. The calorimeter has been performing well and running stably under high luminosity and beam currents. Its performance and status are presented.

1. Introduction

A small angle calorimeter, the Extreme Forward Calorimeter (EFC)' , has been installed last April in the Belle detector1 at the KEKB electron positron collider2. The main functions of EFC are providing the online luminosity information to the detector and accelerator, and acting as a tag for twophoton events. The electromagnetic shower medium is Bismuth Germanate (Bi4Ges012), commonly known as BGO. Since the radiation hardness is an important issue for this device, we choose the radiation-hard BGO crystals3 produced by the Institute of Inorganic Chemistry, Novosibirsk, Russia. Its radiation hardness has been checked to be good after receiving 100 Mrad dose4. EFC consists of two parts, forward and backward, which are mounted on the front surfaces of the cryostats of the compensating solenoids of KEKB. The angular coverage is from 6.4" to 11.5" and from 163.3" to 171.2" for forward and backward, respectively. Each part consists of 160 BGO crystals with 5 segments in 8 and 32 segments in 4.Due to space limitation, the lengths are only 12 and 11 radiation lengths for the forward and the backward BGO crystals.

161

163

The encoded time signals are then fed into a fan-out modules via 60m long twisted pair cables and split into a dual DAQ system, which consists of two almost identical sets of Fastbus and VME crates, called global and local DAQ, separately. The local DAQ is specifically for taking data in parallel with Belle global DAQ without any interference. This is important for monitoring and calibration purpose. The digitization is done with 16-bit multi-hit Fastbus TDCs operated in common-stop mode. The maximum data taking rate for the local DAQ is about 500 Hz. 3. Trigger System

The basic EFC trigger unit is a trigger cell. Two neighboring 4 segments (each segment consists of 5 crystals) of the detector have three trigger cells according to different 8 angles. In the innermost ring, a trigger cell consists of two crystals and its signal is formed by the analog sum of these two channels. The other two trigger cells consist of four crystals. The trigger output of each cell is generated by the CFD circuit in the receiver module. Currently the 2 GeV for the forward and 1 GeV for the trigger thresholds are set to backward. These trigger output are then grouped into four sectors in I$ with a logical OR in the forward and the backward, respectively. A typical trigger rate of cm-2s-1. one sector is around 50 Hz at instantaneous luminosity of The Bhabha trigger is defined by a back-to-back coincidence of energetic electromagnetic showers. Besides being used for the global DAQ and local DAQ, the Bhabha trigger signal is also sent to a CAMAC scalar for on-line luminosity measurement. To make a reliable measurement of luminosity, one has to subtract the accidental rate of the Bhabha trigger. A logic is made by geometrically wrong combinations (not back-to-back) of trigger sectors to estimate such fake rate. The typical fake rate is less than O.1Hz under normal beam condition. The EFC-Tag trigger is defined as a logical OR of all the forward trigger sectors. Currently, two kinds of tagged two-photon triggers are implemented in Belle data taking. One is defined as EFC-Tag A N D “at least one full charged track”; the other is EFC-Tag A N D “at least one electromagnetic cluster”. Each one gives a trigger rate of around 2 Hz at instantaneous luminosity of 1033~m-2s-1. N

-

4. Calibration

Bhabha events are used to calibrate the EFC detector. Considering the boost effect of asymmetric collision at KEKB, only inner 2 layers of forward EFC and

164

outer 3 layers of backward EFC can make coincidence to form the back-to-back Bhabha trigger and to accumulate Bhabha events. However, based on Monte Carlo (MC) study, and the rates from Bhabha trigger and ALL-OR trigger, the Bhabha scattering contribute 73% (83%)of the forward (backward) ALL-OR trigger rate. Other processes like Coulomb scattering, bremsstrahlung,twophoton and beam gas interaction are less important. Therefore, we took EFC “forward All-OR” and “backward ALL-OR” two different data sets by local DAQ to calibrate EFC. The calibration is done by minimizing the difference between the expected shower energy and the measured energy by EFC. The x2 is determined by

”

i= 1

where Eezpis 6 GeV (3 GeV) for the forward (backward) EFC Bhabha clusters, N is the total number of events, g j is the calibration constant, and the Ej is the measured energy in the j-th crystal in the 3 x 3 matrix cluster. Figure 2 shows the energy distribution of Bhabha data collected by EFC. We use a Bifurcated Gaussian function to fit the energy distribution and the tail at the left side of the peak is due to the energy leak. The Bifurcated Gaussian resolutions of Bhabha events are 14.8% (left a ) and 8.6% (righ at) for the forward EFC and 16.1% (lef a t ) and 10.4% (right a ) for backward EFC, as shown in Fig. 2.

Energy Spectrum of EFC Bhabha 350 300

250 200 150 100

50 0

0

2

4

6

E FEFC

8

10

0

1

2

3

4

5

E BEFC

Figure 2. The energy spectra of Bhabha events detected by EFC in the forward (left) and the backward (right).

165 5. Luminosity Measurement

In the e+e- collider, the luminosity can be determined by measuring the known QED process such as e+e- -+ e+e-(y) (Bhabha scattering). EFC is able to provide the instantaneous luminosity measurement by counting the event rate of Bhabha scattering. As described in Sec. 3, we use the fake rate subtracted back-to-back coincidence rate multiplies a conversion factor to obtain the luminosity, C = factor x (Bhabha rate - fake rate). The conversion factor is in the unit of cm-2s-1 and obtained from the MC according to the trigger geometry and energy threshold. The typical Bhabha cm-2s-1 luminosity. rate in the experiment is 140 Hz with 5 . 1 . IP Dependence

The largest systematics of luminosity measurement comes from the movement of interaction point (IP) of KEKB. Due to the small distance between EFC and IP, the IP position change on z axis affects the acceptance of Bhabha scattering. The MC shows the IP-dependent acceptance of EFC Bhabha measurement, as shown in Fig. 3. A linear correlation is found between the acceptance and IP position in z axis. We also see the same relation when we compare the ratio of EFC over the central electromagnetic calorimeter (ECL) luminosity measurement, as shown in Fig. 4. The luminosity ratio shows fluctuation during the experiment runs. After applying the corrections obtained by MC, the ratio of EFC over ECL remains stable during the experiment. 6. Two-photon Physics

The EFC-tag trigger described in Sec. 3 records the two-photon processes in which one of the re-scattered electrons hits EFC. The momentum transfer, q, of the corresponding photon can be measured by the energy deposit in EFC of that re-scattered electron and its position. From the EFC energy and angular resolution, we can estimate the Q2 resolution of the forward EFC for tagged two photon events. The Q2 of the forward EFC tagged data can be extracted by Q2 = -q2 = 2EE'(1 - cosB'), where the E' is the tagged electron energy, and 6'' is the tagged electron angle. The Q 2 range of the forward EFC tagged two photon process is between 0.2 GeV to 2.5 GeV. R o m the EFC angular and energy resolution, the expected Q2 resolution is 9.5%.

7. Summary EFC has been operated since 1999 and providing stable luminosity measurement for KEKB. The performance of energy resolution and stability of lumi-

166

-

Exp 13 Run 1 1500

-

MC IPI effect on EFC Bhabha

g

,

1 j:. .

:

. L: .....

i

.

. I

:

j

.

L.......

i

:

:

J:

j

: :

1

:

^I

1

1

IP in z (mm)

Figure 3. The EFC Bhabha acceptance versus IP-z positions. A linear correlation is observed in MC.

Figure 4. The luminosity ratio of EFC over ECL, IP position in z-axis, and corrected luminosity ratio.

nosity are reported. EFC can also give correct Q2 estimation for triggered two-photon events and make certain QCD studies feasible.

Acknowledgments I would like to thank the CALOR2002 organization committee t o host such a wonderful conference. I would also like to thank Rong-Shyang Lu and ChinChi Wang who help me operate and maintain EFC at KEK. This work is supported by the National Science Council and the Ministry of Education of Taiwan. References 1. Belle Collaboration, A. Abashian et al., Nucl. Inst. and Meth. A479, 117 (2002). 2. E. Kikutani ed., KEK Preprint 2001-157, t o appear in Nucl. Inst. and Meth. A . 3. Ya. V. Vasiliev et al., Nucl. Inst. and Meth. A379, 533 (1996). 4. K.C. Peng et al., Nucl. Inst. and Meth. A 4 2 7 , 524 (1999). 5. K. Ueno et al., Nucl. Inst. and Meth. A396, 103 (1997).

PERFORMANCE AND CALIBRATION OF THE CRYSTAL CALORIMETER OF THE BABAR DETECTOR

M. KOCIAN SLAC, M S 61, 2575 Sand Hill Rd.,Menlo Park, CA 94025, USA E-mail: kocianOslac.stanford.edu (for the B A B A R calorimeter group) The BABAR detector at the B-factory at SLAC is equipped with a calorimeter consisting of 6580 CsI(T1) crystals. This allows for the measurement of the energies of photons and neutral pions and the identification of electrons with high precision, needed in the reconstruction of B-meson decays. The detailed performance of the calorimeter will he presented. As the B-factory operates at high luminosity the calorimeter is exposed to substantial background and radiation damage. The calorimeter is calibrated regularly at different energies in order to meet the precision goals. The calibration methods include the use of a radioactive source, Bhabha events, radiative Bhahha events, 7ro-Mesons, and a light pulser system. This article largely follows reference'.

1. Introduction

BABAR is the detector at the PEP-I1 B Factory at SLAC. PEP-I1 is an asymmetric e+e--collider, currently operating at a luminosity of 4.5 x cm-2s-1 at a center-of-mass energy of 10.58 GeV, the mass of the Y(4S) resonance. The Y(4S) decays exclusively into BOBo and B+B- pairs. The main physics goal of BABAR is the study of CP-violating asymmetries in the decay of neutral B-mesons. Secondary goals are precision measurements of decays of bottom and charm mesons and of T leptons, and searches for rare processes that become accessible with the high luminosity of the PEP-I1 B Factory. The BABAR detector consists of 6 subdetectors. F'rom the inside to the outside, there is a Silicon Vertex Detector, a Drift Chamber, a DIRC (Cherenkov- Detector), an Electromagnetic Calorimeter, and an Instrumented Flux Return (fig. 1). 2. Calorimetry goals

The very small branching ratios of B mesons to C P eigenstates and the need for full reconstruction of final states with several r o s place stringent requirements

167

168

IFR Barrel

0

I

' scale ' 4m BABAR Coordinate System

8583A51

Figure 1. The components of the B A B A R detector

on the electromagnetic calorimeter: a large and uniform acceptance down to small polar angles relative t o the boost direction excellent reconstruction efficiency for photons down to 20 MeV energy resolution of order 1- 2 % and excellent angular resolution for the detection of photons from AO and decays in the range from 20 MeV to 4 GeV efficient electron identification with low misidentification probabilities for hadrons.

For these reasons the choice for BABAR was a CsI(T1) crystal calorimeter. The energy resolution of a homogeneous crystal calorimeter can be described empirically in terms of a sum of two terms added in quadrature (1)

where E and UE refer to the energy of a photon and its rms error, measured in GeV. The energy dependent term a arises from fluctuations in photon statistics,

169

electronics noise and beam background noise. The constant term b is dominant at higher energies and arises from non-uniformity in light collection, shower leakage or absorption in the material between and in front of the crystals, and uncertainties in calibration. The angular resolution is determined from the transverse crystal size and the distance from the interaction point. It is parametrized as

The actual resolution for the BABAR calorimeter will be discussed in section 5.2.

3. Layout and Assembly 3.1. General Overview

Interaction Point Figure 2.

1979

8572A03

A longitudinal cross-section of the Calorimeter

The calorimeter consists of 6580 CsI(T1) crystals. Its angular coverage is 126' in polar angle and 360" in azimuthal angle (see fig. 2). It is subdivided into a barrel and a forward endcap. The barrel consists of 5760 crystals arranged in 48 rings of 120 crystals, while the endcap contains 8 rings. The innermost 2 rings contain 80 crystals, the next three rings 100, and the three outer rings 120 crystals. The calorimeter's geometry is projective in 4, while in 0 there is a non-projectivity of 14 mrad, except in the transition region between barrel and endcap where it reaches 45 mrad. The non-projectivity minimizes the

170

energy losses through the spaces between the crystals. There is a gap of about 2 mm between the barrel and the endcap which is fully covered by the higher non-projectivity in this region.

3.2. Mechanical Assembly The individual crystals are assembled in carbon fiber modules. In the barrel, the modules contain 7 x 3 crystals (except for the most backward module which only contains 6 x 3 crystals). The entire barrel consists of 280 carbon fiber modules. In the endcap, each carbon fiber module contains 41 crystals. The carbon fiber housings are glued onto Aluminum strongbacks. The barrel modules are inserted into an Aluminum cylinder. The endcap modules are held in place by two semicircular structures. 3.3. Crystal Assembly The crystals are trapezoidal in shape. The typical area of the front face is 4.7 x 4.7 cm’, while the back face area is typically 6.1 x 6.0 cm’. The length of the crystals varies between 16 radiation lengths in the backward part and 17.5 radiation lengths in the forward part. The polished crystals are wrapped in two layers of Tyvek (2 x 165 pm) for reflection and tuning. The next layers are Aluminum foil (25 pm) and Mylar (13 pm) for electrical insulation (fig. 3). The outside layer is a 300 pm thick carbon fiber housing which is attached to the housings of the neighbour crystals. The crystals are read out through 2 silicon PIN photodiodes of 2 x 1 cm’ area each. The diodes are glued onto a polystyrene coupling plate which itself is glued onto the crystal with transparent epoxy. The area surrounding the diodes is covered with white, painted plastic plates to reflect light. Each diode is read out by a preamplifier that sits in a housing above the crystal. For the details of the electronics readout see reference’. 4. Calibration

4.1. Overview The energy calibration of the calorimeter proceeds in two steps. First, the measured pulse height in each crystal has to be transformed into the deposited energy. Second, the deposited energy in a shower has to be related to the energy of the incident photon or electron by correcting for energy loss mostly due to leakage and absorption in material between and in front of the crystals. Table 1 shows a summary of the calibrations used for the BABAR calorimeter. The electronics and light pulser calibrations are discussed in reference2.

171

Figure 3.

A schematic of the wrapped crystal

Table 1. Properties of the different calibrations for the calorimeter. “Absolute” refers to the ability to tie the measurement t o an absolute energy scale as opposed t o measuring the relative changes and differences in signal height.

Source

Duration

Energy Scale

Single Crystal

Absolute

20 min

0.00613 GeV

Yes

Yes

3-9 GeV

No

Yes

4h

0.03 - 3 GeV

No

Yes

Yes

No

Yes

No

Bhabha

0 - 13 GeV

I Light Pulser I

3 min

0 - 13 GeV

i I ~

172

4.2. Individual Crystal Calibration

In spite of the careful selection and tuning of the crystals, their light yield varies significantly and is non-uniform along the crystal axis. It also changes with time under the impact of beam-generated radiation. The absorbed dose is largest at the front of the crystal and results in increased attenuation of the transmitted scintillation light. The light yield must therefore be calibrated at different energies, corresponding to different average shower penetration, to account for the effects of the radiation damage3. The calibration of the deposited energies is performed at two energies at opposite ends of the dynamic range, and these two measurements are combined by a logarithmic interpolation. A 6.13 MeV radioactive photon source provides an absolute calibration at low energy, while at higher energies the relation between polar angle and energy in Bhabha events is used for calibration.

4.2.1. Radioactive Source Calibration The radioactive source calibration uses 6.13 MeV photons produced in the reaction

+ n +16

N

+ a,16N

+16

O* + p,16 O* +16

o + y.

(3)

16N has a lifetime of 7 seconds. A fluid of polychlorotrifluoro-ethylene, activated by neutrons from a generator, circulates through a system of tubes in front of the crystals. All crystals in the calorimeter are calibrated with this method. The average resolution for the constants is 0.33 %.

4.2.2. The Bhabha Calibration At high energies, single crystal calibration is performed with a pure sample of Bhabha events. As a function of the polar angle of the e*, the deposited cluster energy is constrained to equal the prediction of a GEANT based Monte Carlo simulation. For a large number of energy clusters, a set of simulates linear equations relates the measured to the expected energy and thus permits the determination of a constant for each crystal. 200 e* per crystal result in constants with a statistical error of 0.35 %.

4.3. Shower Energy Correction The correction for energy loss due to shower leakage and absorption is performed as a function of shower energy and polar angle. For low energies it is derived from 7ro decays, while for high energies corrections derived from single photon Monte Carlo or from radiative Bhabha events can be used.

173

4.3.1.

TO

Calibration

For the no calibration the energy range is subdivided into 16 bins in ln(E), while the angular range is divided into 9 bins in cos(8). For each bin a twophoton-mass plot is generated. By constraining the peaks to the nominal mass, a correction polynomial of third order in ln(E) and a correction polynomial of second order in cos(8) are determined in an iterative procedure. Typical corrections are of order 6 f 1 %. 5 . Performance

5.1.

d' and

77 mass and width

Figure 4a shows the two-photon invariant mass for hadronic events around the T O mass in data from 2001. Photons are required to exceed 30 MeV, while 7ro exceed 300 MeV. The reconstructed mass is measured t o be 134.9 MeV/c2. The width is 6.5 MeV/c2. The two-photon-invariant mass for symmetric 77s for Ev > 1 GeV is shown in figure 4b. The reconstructed mass is 547 MeV/c2, the width is 15.5 MeV/c2.

Figure 4. data.

Invariant mass of two photons in hadronic events. The solid lines are fits t o the

5 . 2 . Resolution

Figure 5a shows the energy resolution derived from a variety of processes: radioactive source, symmetric T O and 17 decays, xcl + J/y!ry, and Bhabha events. As the resolution of T O and 17 depends on the angular resolution also, a simultaneous fit to energy and angular resolution was done for those cases, assuming an asymmetry of the photon energy distribution derived from Monte Carlo. There is good agreement between the measured resolution and Monte

A SYSTEMATIC STUDY OF RADIATION DAMAGE TO LARGE CRYSTALS OF CSI(TL) IN THE BABAR DETECTOR

T. HRYN'OVA SLAG, MS 61, 2575 Sand Hill Rd., Menlo Park, C A , USA E-mail: [email protected]

(For the BaBar EMC Group) We study the impact of radiation damage on large CsI(T1) crystals in the BABAR electromagnetic calorimeter. Average radiation exposure of up to 400 Rad to date, originating primarily from beam backgrounds, has been measured by RadFETs located at the front face of crystals.

1. Introduction

The BABAR Electromagnetic Calorimeter' (EMC) consists of 6580 CsI(T1) crystals ranging between 16 and 17.5 radiation lengths. CsI(T1) was chosen for its good mechanical properties, high light output, convenient emission wavelength for use with Si-photodiodes and reasonable signal response time. The crystals were produced2 from a melt of CsI salt doped with 0.1%thalium using either Kyropoulos (Kharkov, Crismatec, Hilger) or Bridgman (Shanghai, Beijing) growth techniques. As sensitivity to radiation damage is generally found to be smaller for higher purity crystal, the quality of the salt and the recycled material was strictly controlled. In order to decrease the contributions to systematic errors on energy resolution it is important to understand the effect of radiation on CsI(T1) crystals. 2. Sources of Radiation Damage

Radiation damage in the BABAR EMC is believed3 t o be almost entirely caused by 'non-physics' events, or so called 'beam backgrounds' in the EMC. There are two distinct types of this background in the BABAR experiment: single beam background and colliding beam background. The single beam background is mainly caused by fixed dipole magnets which are situated near the interaction point. They tend to sweep off-energy primary beam particles into machine elements near the detector, resulting in a low-energy shower (<

175

176

10MeVl) which enters the EMC. For colliding beams, there is also a major contribution of photons from small-angle radiative Bhabhas in which an e* strikes a machine element. In both cases the occupancy increases significantly at smaller polar angles (in the endcap and backward barrel), while single beam backgrounds also peak in the horizontal plane.

3. Dose Monitoring 3.1. RadFET Monitoring

The dose received by the front of the EMC is measured by 116 RadFETs4 placed in front of the barrel and endcap crystals. RadFETs are red-time integrating dosimeters based on solid-state MOS technology. The dose increases approximately linearly with the integrated luminosity. The dose map obtained by the RadFETs reproduces the beam background angular distribution. The highest dose accumulated to date, 700Rad, is observed in the innermost ring of the endcap (EC) while both backward (BB) and forward (FB) barrels have similar doses of about 250 Rad (Figure la) on average.

Days since 1 Jan 1999

Days since 2 June 1999

Figure 1. Average dose in the EMC measured by (a) the RadFETs, (b) the leakage currents.

3.2. Leakage Currents An alternative way to calculate the dose accumulated by the crystals is using the leakage currents5. The dose then is proportional to the integral of beam-

177

correlated photodiode current ( I ) :

Dose =

~

EcTystal

M C T ystal

=

1

Ibeams on - Ibeams off McTystal .C

dt,

(1)

where McTystal is the mass of crystals in the section of the detector and C is the light output of 3900 photoelectrons/MeV obtained using the EMC readout (a light output of 7300 photoelectons/MeV was measured using a preamplifier with 2 ps shaping time). There are 10 independent bias voltage supplies for the EMC (four in the BB, four in the FB and two in the EC). Using the formula above, one can obtain the average dose in each sector. The RadFETs measure the dose seen at the front face of the crystal. The leakage currents average the above dose over the whole crystal volume. They give similar results within a scaling factor of approximately 3 (Figure lb), which corresponds to the fraction of the crystal volume exposed to the radiation since the electro-magnetic showers deposit energy preferentially towards the front of the crystals. The observed integrated dose induces damage to the crystals, which may be exhibited in two ways: a drop in the total light output and a change in the uniformity of the light output along the length of the crystal. We measure the change in the total light output using the standard BABAR calibration procedures6: radioactive source (6.13 MeV photons) and Bhabha events (3 8.75 GeV electrons). 4. Light Yield (LY) Monitoring 4.1. Source Measurements

We use 6.13 MeV photons from neutron-activated Fluorinert' circulating through a system of thin tubes in front of all crystals. These measurements are taken every 2 weeks and reach a precision of 0.33% for single crystals. The dependence of the LY drop (averaged over EC, BB and FB) on the dose is presented in Figure 2a. The value of the degradation is currently 9% in the EC, 6% in the FB and 3% in the BB. The LY decreases as a function of dose as expected, but the drop in LY differs for FB and BB, although they received similar doses as measured by the RadFETs. To address this effect the LY change was studied separately for each crystal vendor (Figure 3). Among crystals from the same vendor the values of the light yield degradation in the FB and the BB are similar. We are currently investigating the different rates of change of the LY in the barrel and the endcap. This may be explained by a significant portion of the EC crystals being irradiated both from the sides and from the front face, whereas the majority of the barrel crystals are irradiated from the front face only.

178

- b2 d

-0.02 -0.03 \

-0.1I

I

I

10 Dose, Rad

”: 00

AAFB s9 ooEC * 1/01/01

-0.04700 800 900 1000 1100 Days since 1/01/99

Figure 2. Average change in the light yield in the EMC measured with (a) source (August 1999-December 2001), (b) source(s) and Bhabhas(B) (January-December 2001).

4.2. Bhabha Measurements

Bhabha events allow the calibration of the calorimeter at high energies. In a 12-hour run at a luminosity of 3 x cm-’ s-l we reach a precision of 0.35% per crystal. For the source measurements nearly all of the energy is deposited in the front part of the crystal whereas in Bhabha events a large fraction of the electro-magnetic shower is contained in the back part of the crystal which consequently has less radiation damage. Currently we do not see any difference between the Bhabha and the source LY change measurements (Figure 2b), thus there is as yet no evidence of non-uniformity. Our previous studies have shown that it might become an issue around 1kRad4. To maintain a reasonable energy resolution the non-uniformity contribution to OE/Emust be less than 0.5%.

5. Crystal Scanner Experiment To study the light output uniformity behavior under irradiation wea have built an apparatus which allows in situ exposure and measurement of the longitudinal changes in the LY of large CsI(T1) crystals. This experiment will help us to develop a correction function to be used in Monte Carlo simulation of detector performance which incorporates the effect of the radiation damage of the crystals. We will study the spare full-size CsI(T1) crystals from different vendors. The diagram of the apparatus is shown in Figure 4. An assembly consists of &T.Hryn’ova, P. Kim, M. Perl, K. Phillips, H. Rogers, R. Schindler, W. Wisniewski

179

0

b

Crismatec $

A0 A

-0.025 -

n

*d -0.025-

%wx

-0.05 -

-0.075-

-0.1-

A

FB

-0.075-

BB

-0.1 -

Hilger

m

bp p

%'

-0.05 -

A

A

0

A

*A A%'

EC FB

*e>

I

Kharkov

%tnfi

'3

-0.05 -

-0.075-

-0.1-

A

'i

BB

-

10 Dose,Rad Figure 3.

0 ,

,

'e

-0.05 -

-0.075-

A

-0.1ul

I

Shanghai

* -0.0250- h t d

A

FB

-

,b $

d

EC

J

10 Dose,Rad

10 A Dose,Rad

2* -0.025 d

'.r

\

I

0

4

EC FB

'e

% .,'

\s

BB I

I

I

I

. 10 Dose,Rad

#,"I

Average change in the LY in the EMC by vendor.

2 x 8 crystals each with Kamamatsu R2154-06 photomultiplier tubes (PMT) and 4 stepper motors moving two collimators with 88Y sources in vertical and horizontal planes. The assembly is irradiated at a rate of 2 Rad/hour by photons from a 6oCo source which is located 1m from the assembly. The front faces of all crystals are irradiated uniformly. A small CsI(T1) crystal with PMT and 88Y, 22Na and 228Thsources, located behind lOcm of lead, is used as a standard reference to compensate for the drift of electronics. We use 10 AD592ss for the temperature monitoring. A light pulser system with light fibers connected to the face of each crystal is used to monitor the electronics. Radiation monitoring is done with 2 GM tubes with a computer readout for the current dose monitoring and 55 thermo-luminescent dosimeters for the total dose monitoring. Data is read out through the CAMAC crate/SCSI cardg to

180

Irradiating Source

Crystals & Scanner

Figure 4.

Readout

A diagram of Crystal Scanner Experiment.

a PC. The apparatus is specifically designed to measure sixteen crystals simultaneously and to minimize the systematic errors in these measurements by performing all the logitudinal scans completely in situ, interleaved with short 6oCoexposures. Data points are planned to be taken every 2 - 3 cm along the crystal length doubling the dose until it reaches 5 kRad. A typical spectrum is presented on Figure 5 . This experiment had been assembled and is ready to begin collecting data. 6. Conclusion

The 6580 crystals in the BABAR EMC along with extensive dosimetry allow us to study the impact of radiation damage on CsI(T1) crystals with high precision. Effects of radiation damage in the detector are visible but not yet problematic. Additional studies such as the Crystal Scanner test will help us to improve our understanding of the changes in the detector, decreasing systematic uncertainties in measurements which rely on the calorimeter.

Acknowledgments

I wish to thank all the members of the EMC group and the PEP crew for making these measurements possible. This work is supported by DOE and NSF (USA), CAS (China), BMBF (Germany), NFR (Norway) and PPARC (UK). Individuals have received support from Alexander von Humboldt Foundation.

181

88Y, 1.8 MeV

88Y,2.7 MeV 228Th,2.6 MeV

I

h0

l

l

800

1000

1200 1400 1600 1800 2000

ADC Counts

Figure 5. A typical spectra from the Crystal Scanner Experiment.

References 1. BABAR Collaboration, Nucl.Inst.Meth., A479, 1 (2002). 2. Amcrys-H, Kharkov, Ukraine; Crismatec, Nemours, France; Hilger Analytical, Margate, Kent, UK; Shanghai Institute of Ceramics, Shanghai, P.R. China; Beijing Glass Research Institute, Beijing, P.R.China. 3. S. H. Robertson, private communications. 4. B. Camanzi et.al., Nucl.Instr.Meth. A457,476(2001). 5. I. Eschrich, these proceedings. 6. M. Kocian, these proceedings. 7. Fluorinet((CPFSCI]) is manufactured by 3M Corporation, St. Paul, MN, USA. 8. AD592 is a two terminal monolithic integrated circuit temperature transducer produced by Analog Devices, USA. 9. Jorway 73A CAMAC crate controller and CAMAC controller software (Fermitools).

PERFORMANCE AND UPGRADE PLANS OF THE BELLE CALORIMETER

B. A. SHWARTZ Budker Institute of Nuclear Physics, 630090 Novosibirsk, RUSSIA, E-mail: shwartzOinp.nsk.su

(For the BELLE Electromagnetic Calorimeter Group)

The electromagnetic calorimeter of the BELLE detector built for experiments on B-meson physics is described. The calorimeter consisting of 8736 CsI(T1) crystals demonstrates its good performance in the experiment while its parameters are close to the project ones. The luminosity monitoring and tolerance to high background conditions are briefly described as well.

1. Introduction

The BELLE detector for experiments at the KEKB, an energy asymmetric B-Factory with high luminosity, has been constructed at KEK, Japan to study CP violation in B meson decays. The detailed description of the detector can be found in reference’. The collaboration comprises about 300 researchers from 55 universities and institutes of 14 countries. The detector consists of a three-layer vertex detector (SVD), a 50-layer drift chamber (CDC), an array of aerogel Cherenkov counters (ACC), time-of-flight scintillation counters (TOF), an electromagnetic calorimeter based on CsI(T1) crystals (ECL) and 14 layers of 4.7 cm thick iron plates interleaved with a system of resistive plate counters (KLM). All subdetectors besides the KLM are located inside a 3.4 m diameter superconducting solenoid which provides a 1.5 T magnetic field. Since one-third of B-decay products are re's and other neutral particles providing photons in a wide range from 20 MeV to 4 GeV, a high resolution calorimeter is a very important part of the detector. The CsI(T1) scintillation crystals were chosen as a material for the calorimeter due to its high light output, short radiation length, good mechanical properties and moderate price. The main tasks of the calorimeter are: detection of y-quanta with high efficiency,

182

183

0 0 0 0

precise determination of the photon energy and coordinates, electron/hadron separation, generation of the proper signal for trigger, on-line and off-line luminosity measurement.

2. Calorimeter structure and main features

The electromagnetic calorimeter (ECL) consists of a barrel section of 3.0 m in length with the inner radius 1.25 m and the annular endcaps at z=2.0 m (Forward part) and z=-1.0 m (Backward part) from the interaction point. The calorimeter covers the polar angle region of 12.4" < 8 < 155.1' except two gaps N 1' wide between the barrel and endcaps. The barrel part has a tower structure projected to the vicinity of interaction point. It contains 6624 CsI(T1) elements of 29 types. Each crystal is a truncated pyramid of the average size about 6 x 6cm2 in cross section and 30 cm (16.2X0) in length. The end caps contain altogether 2112 CsI crystals of 69 types. The total number of the crystals is 8736 with total mass of about 43 tons. Each crystal is wrapped with a layer of 200 pm thick Gore-Tex porous teflon and covered by the 50 p m thick aluminized polyethylene. For light read out two 10 x 20mm2 Hamamatsu S2744-08 photodiodes are glued to the rear surface of the crystal via an intervening 1 mm thick acrylite plate. The LED attached to the plate can inject the light pulses to the crystal volume to control the optical condition stability. Two preamplifiers are attached to the photodiodes. For the electronics channel control the test pulses are fed to the inputs of the preamplifier. The average light output of the crystals measured by calibration with cosmic rays is about 5000 photoelectrons/MeV while a noise level is equal to about 200 keV. The barrel crystals were installed in a honeycomb-like structure formed by 0.5 mm thick aluminum septum walls stretched between the inner and outer cylinders. Eight crystals, four rows in 13and two columns in (p, were inserted in each cell and fixed in position by support jigs at the back. The overall support structure is gas tight and flushed with dry air to provide a low humidity (5%) environment for crystals. The preamplifier generated heat (about 3 kW in total) is absorbed by the liquid cooling system. The end cap support structure is similar to the barrel one. The layout of the electronics for readout and trigger is presented in Fig. 1. The preamplifier attached to the crystal is followed with the shaper boards placed in the crates around the detector as well as digitizing and trigger modules placed at the electronics hut. The shaper board contains C R - (RC)4

184

Figure 1. Calorimeter electronics lay out.

active filter with r = lps shaping time and MQT300A chip which converts the input charge integrated over certain gate time to three time intervals which are measured by the multi-hit TDC 18778. The corresponding ranges are: 0.06 MeV/bin, 0.5 MeV/bin and 4 MeV/bin. Auto-range selection option provides readout of only one range with optimal sensitivity. In addition to the spectrometric channel the shaper board contains the fast shaper (T = 200ns) which generates the signal for trigger and timing. The calorimeter is under operation since June 1999. All counters work and demonstrate good performance. 3. Calorimeter performance

To reconstruct photons in the calorimeter clusters of hitted crystals are detected. The cluster is defined as an array of connected counters having the energy deposition over the threshold 0.5 MeV. Then the photon energy is determined as a total energy of the cluster: cluster

cluster

where TDCi is TDC counts, Pi is a pedestal value and ci - calibration coefficients. The calibration procedure is considered in details in the report2

185

at this conference. Photon energy distribution measured for e+eE-, < 8GeV) is presented in Fig. 2.

-+yy process (3.5GeV <

d 102

I000

0.6

0.8

i

1.2

0.6

0.8

I

1.2 E/E.,.

zoo00

10000

0.8

0.6

I

1.2

0.6

0.8

I

Figure 2. Photon energy spectra for e+esolid line is the fit results.

+ yy

1.2 WEI

WE.(

process. Histogram is a data while the

The energy deposition distribution can be well approximated by the logarithmic gaussian form : 2s; qdE dW = exp( - ln2(1- q ( E - Ep)/u)- -)2s; 2 JZ-;uso where Ep is peak energy; D - FWHM/2.35; q - asymmetry parameter; be expressed as: SO

2 a.n -1 (175 =T),

5

< = 2.35

SO

can

(3)

The energy resolution averaged over all calorimeter is a ~ / E = 1 . 7 % . The difference of uE/E between barrel and endcaps is connected with different amount of material in front of the counters as well as the photon energy angular dependence which varies from 3.5 GeV in the backward direction up to 8 GeV in the forward one. The obtained results are in a good agreement with Monte Car10 simulation.

186

Photon angles are measured as a corrected center of gravity of the energy deposition over the cluster:

where Ei, 8i, p i - are respectively the energy deposited in i-th crystal and it's angular coordinates. Correction functions (F) can be written depending from only one angular variable. The width of the experimental acollinearity distribution for e+e- + yy process was found to be UA,+, = 0.23" that is in good agreement with simulation. The yy invariant mass distribution presented in Fig. 3 shows clear peaks of TO and mesons with mass resolution of 4.8 MeV/c2 and 12 M e V / c 2 respectively. These results are in a good with M C simulation.

n' mass resolution

-q niass resolution

10'.

I2W

I

It

I

il

i

*

I

sow r

L

M,

Figure 3. yy invariant mass distributions. Histogram is a data while the solid line is the fit results.

The crucial function of the calorimeter is electron identification. The ratio of the energy measured by ECL to the particle momentum determined by CDC and the information about the shower lateral shape are used for that in combination with data for this particle from dE/dx, T O F and ACC systems. Obtained values of electron identification efficiency as well as the fake rate of misidentified pions are presented in the Fig. 4.

4. Tolerance to the high background environment Since KEKB collider is a B-factory, the machine with circulating beams higher than 1 A, the obvious concern was a possible degradation of the crystal pa-

187

Figure 4. Electron identification efficiency. Upper points correspond to the electron efficiency while lower points stand for misidentified pions rate.

rameters due to high radiation dose. To monitor the dose the direct current through photodiodes are measured and averaged over group of counters. 16 average current values were controlled to watch the dose dependence from counter position. The total dose absorbed during time interval t can be expressed by formulae:

1

l t Dose = KLme 'Idt7

K = 6.24 x 101'MeV/kg,

(5)

where L - light output (photoelectrons/MeV), m - mass of the crystal (kg), e electron charge, AI = I - I d d a r k . Then for t = 107sec Dose = 100rad/2.5nA. The measured integrated dose up to now is about 10 rad for barrel part and about 40 rad averaged over each endcap. This caused light output deterioration of about 2% in the barrel part and 3-4% in the region closest to the beam pipe. More details are given in the report2. These results are in a good agreement with previous measurements of the crystal radiation hardness3. Since these studies shows the loss of light output less or about 20% at 1 krad we do not anticipate big problems with signal decreasing at the present luminosity. More troubles are produced by so called pile up noise caused by the soft background photons with average energy of about 1 MeV. The fluctuation of the number of these photons coming during the integration time contribute to the total noise level. The ratio of the noise level, 0 ,with both beams in ~

188

the collider ( I + = I - = 360mA) to that without beams is shown in Fig. 5. It can be seen that even at the moderate beam currents the noise increases substantially, in the end caps it becames 4-5 times higher approaching to the level of 1 MeV.

Figure 5. Th e ratio of the noise level, u, with both beams in the collider ( I + = I- = 360mA) t o that without beams. T h e lower points correspond t o the net pile-up effect.

High energy photon background provides the random occupancy of the calorimeter as well as additional clusters. The average number of this clusters per event was measured to be 3 at 700 mA of the beam currents and threshold of 20 MeV. At present the upgrade of KEKB collider with 1035~m-2s-1 luminosity is proposed ( S U ~ ~ ~ K E KThe B)~ beam . currents in this collider are planned to be: I+ = 10A, I - = 4.4A while vacuum should be of (1 - 10) x 10-7Pa. It’s clear that the background effects mentioned above will increase by more than one order of magnitude. To accept these conditions the calorimeter should be upgraded as well. Now R&D works are going on. These are mostly connected with two items: 0

0

Calorimeter electronics will be changed from present (QT+TDC) chain to pipeline readout scheme. The option of pure CsI usage in the end caps is under study. Vacuum phototriodes and photothetrodes are considered as photodetectors instead PIN photodiodes.

1 89

5. Luminosity monitoring

On-line and off-line luminosity measurements is a very important task of the calorimeter. The e+e- elastic scattering events are used for both of these purposes. The on-line luminosity monitor is based on the counting of the rate of the events with electron and positron hitted the opposite quadrants of both end caps. The energy threshold is set at present to 1.5 GeV. The corresponding fast sum signals are provided by the trigger electronics. The counting rate is about 220 Hz at 5 x 1033cm-2s-1. The exploitation of the on-line luminosity monitor demonstrated its reliability and the accuracy better than 3%. For off-line integrated luminosity determination we use the sample of e+eevents with both, electron and positron, provided good tracks in CDC and clusters in the barrel part of ECL with the sum energy exceeding 4 GeV. The precision of the off-line luminosity measurement is better than 2%. It's limited at a moment by the Monte Carlo simulation with taking into account the radiative corrections. Detail description of the luminosity measurement at Belle can be found in reference5. 6. Conclusion 0

0

0 0

Belle calorimeter demonstrates good performance while its parameters are close to the project ones. Achieved yy invariant mass resolution is 4.8MeV/c2 for 7ro and 12.1MeV/c2 for q. The energy resolution for y(Ey > 3.5GeV): a E / E = 1.7%. luminosity are in R&D works to match the calorimeter to progress.

References 1. A.Abashian et al. (Belle collaboration), Nucl. Inst. bMeth. A479 (2002) 117. 2. K.Miyabayashi, "Monitoring and calibration of the Belle electromagnetic calorimeter", talk at this conference. 3. K.Kazui et al., Nucl. Inst. bMeth. A394 (1977) 46. B.Shwartz, Nucl. Inst. &Meth. A453 (2000) 205. 4. http://belle.kek.jp/workshops/HLO2/ 5. V.Zhilich, "Luminosity measurement at BELLE", 8-th Int. Conf. on Instrumentation for Colliding Beam Physiscs, Febr. 28 - March 6, 2002, Novosibirsk.

DEVELOPMENT OF YTTRIUM DOPED LEAD TUNGSTATE CRYSTAL FOR PHYSICS APPLICATIONS

Q. DENG, J.Y. LIAO, D.Z. SHEN, D.S. YAN, Z.W. YIN Shanghai Institute of Ceramics, 1295 Dingxi Road, Shanghai 200050, P . R . China

R.H. MAO, X.D. QUt, L.Y. ZHANG AND R.Y. ZHU California Institute of Technology, Pasadena, C A 91125, USA

In this paper we present results of the development of yttrium doped lead tungstate crystal at Shanghai Institute of Ceramics. The crystal growth by modified Bridgman method is described. The segregation coefficient of yttrium ions in lead tungstate crystals was determined. The scintillation emission and transmittance spectra, light output, decay kinetics, light response uniformity and radiation induced color centers were measured. It is found that yttrium doping suppresses the slow scintillation component and improve crystal’s radiation resistance.

1. Introduction

Because of its high density, small radiation length and Molihre radius and fast decay time, lead tungstate (PbW04) crystal has attracted wide attention in high energy and nuclear physics community. The Compact Muon Solenoid (CMS) experiment will use 11.2 m3 large size (25 X,) PbW04 crystals for its precision electromagnetic calorimeter (ECAL) at the Large Hadronic Collider (LHC)l. The Alice experiments at CERN, the BTeV experiment at Fermilab and the CLAS and PrimEx experiments2 at CEBAF will also use PbW04 crystals. Among all features, a good radiation resistance is required in many of the above applications. An effort has been made at Shanghai Institute of Ceramics (SIC) in the last six years to develop radiation hard lead tungstate crystal with fast decay time for physics applications. Our previous studies have shown that the radiation damage in PbW04 is caused by the host structure defects, such as oxygen vacancies3, which introduce local charge imbalance, trap electrons or holes and consequently form color centers under irradiation. One approach to reduce the density of the host structure defects is the optimization of the t Now at Imperial College, University of London, London, United Kingdom.

190

191

stoichiometric ratio between two oxides and control it during the crystal growing process. This approach was attempted by SIC as well as other group^^>^. It is found this optimization alone was not sufficient. Various approaches t o suppress or compensate the remaining defects were taken. Oxygen compensation, referring to post growth annealing at high temperature in an oxygen rich atmosphere, was found to be effective6, indicating the main mechanism of the radiation damage in PbW04 is oxygen vacancies, i.e. caused by electron centers. Doping during crystal growth is another common approach to either artificially introduce local charge imbalance and thus compensate structure defects or function as scavenger to further eliminate unwanted impurities. In development of BGO for the L3 experiment, Eu doping was used at SIC t o improve its radiation hardness7. In development of CsI(T1) for the BaBar and BELLE experiments, a special scavenger was used at SIC to remove oxygen contamination6. Pentavalent (niobium) doping in PbW04 was first reported by Lecoq e t al. to be effective in improving transmittance at 100 ppm level8. Trivalent (La) doping was first reported by Kobayashi et al. t o be effective in improving both transmittanceg and radiation hardnesslO. Consequent studies on doping with various ions, such as La, Lu, Gd, Y and Nb, at optimized level were reported to be effective in improving in transmittance as well as radiation resistancel1J2. Along the same direction, doping was extensively studied at SIC13914>15. This paper presents growth of yttrium doped PbW04 crystals by modified Bridgman method at SIC and discusses their optical properties and radiation hardness measured at Caltech as well as SIC. 2. Crystal Growth

PbW04 single crystals were grown by Modified Bridgman technique at SIC. Raw materials of high purity, PbO (5N) and W 0 3 (4N), are produced in Shanghai, and are mixed in precise stoichiometric proportion in an agate mortar. The mixture is first melted in a platinum crucible in air for a period of time to ensure complete homogeneity. After heated to high temperature this melt is sintered into platinum crucible to form polycrystalline PbW04 grogs for crystal growth. Figure 1 is a schematic showing typical structure of a modified Bridgman crucible used for the growth of PbW04 crystals at SIC. The detailed layout of the furnace and its temperature profile can be found in reference16. firnaces of this kind were early developed at SIC to grow BGO for the L3 experiment, and were later successfully adapted to grow CsI(T1) crystals for two B factories experiments BaBar and BELLE.

192

Crystal were grown along the c axis at SIC. Twenty eight crucibles of rectangular shape are constructed in every furnaces for PbW04 crystal growth. The shape of crucible makes efficient use of raw materials since only a small fraction of ingot is needed to be cut off to make final dimension. As-grown PbW04 crystals are transparent, colorless without visible defects, such as cracking, inclusions, scattering centers and growth striation.

A

%$holder

4%powder Pt ctucible

Molten PbW04

Grown crystal

Thenalcouple

Seed ciystal

4%holder

Metal support

Figure 1. A schematic of the modified Bridgman-Stockbarger crucible used for PbW04 growth a t SIC.

Figure 2. A linear fit of yttrium concentration in PbW04 crystals.

Table 1. Results of GDMS analysis (ppmw) for Y doped PbW04 ingots Element Na Si K Ca cu AS

Yttrium Mo Nb Sb La

Seed/Tail 1 0.2/0.8 0.5/0.2 0.3/1.8 0.9/<0.05 0.04/0.2 0.15/0.35 40/45 0.3/0.55 <.05/< .05 <.05/<.05 <.01/<.01

Seed/Tail 2 0.2/1.0 0.7/1.3 0.4/1.4 0.6/0.08 0.04/0.4 0.1/0.6 40/50 0.3/0.9 <.05/< .05 <.05/<.05 <.01/<.01

Seed/Tail 3 0.4/0.8 0.5/1.2 0.7/1.2 0.12/0.15 0.3/0.35 0.5/0.5 30/35 0.6/0.8 <.05/< .05 <.05/<.05 <.01/<.01

Seed/Tail 4 0.2/1.8 0.5/0.1 0.5/2.0 0.6/0.2 0.08/0.54 0.2/0.6 40/60 0.2/0.8 <.05/< .05 <.05/<.05 <.01/<.01

Tail 5 0.8 0.05 1.3 0.15 0.23 0.54 50 1.0 <.05/< .05 <.05 c.01

Table 1 lists impurities (ppmw) found by Glow Discharge Mass Spec-

193

troscopy (GDMS) analysis in several pairs of yttrium doped PbW04 samples cut from the seed and tail ends of the same ingot. It shows that impurities Na, K, Cu, As and Mo migrated to the tail end of ingots during growth, indicating that their segregation coefficient in PbWO4 crystals is less than one. The Ca, as well as La 13, are concentrated at the seed end with segregation coefficient of larger than one. The distribution of the yttrium, however, is rather uniform and is slightly less than 150 ppm mol of Y 2 0 3 in melt, indicating that its segregation coefficient in PbW04 is less but close to 1. Yttrium concentration (ppmw) in PbW04 samples

Table 2.

ID Melt (pprnw)

(4 yttrium (pprnw)

699 29.3 5.5 39

699 29.3 29.0 29

723 29.3 3.0 38

723 29.3 26.5 30

727 29.3 4.0 30

727 29.3 28.0 24

743 29.3 4.5 31

743 29.3 28.0 20

Table 2 lists the yttrium concentration (ppmw), obtained by GDMS analysis in four pairs of yttrium doped samples taken at positions Z (cm) from the top (the tail end) of the ingot. Also listed is the initial yttrium concentration (ppmw) in the melt (Melt). The data in Table 2 was used to extract the segregation coefficient of yttrium in PbW04. The segregation coefficient k,, defined as the ratio of the dopant concentration in the bulk crystal ( C c r y s t a l ) to that in the melt ( C m e l t ) , describes the ability of the dopant to be incorporated into the solid phase,

k, = -Ccr . y stal

(1)

Cmelt

Assuming a slow, steady state growth process, the distribution of a dopant concentration in a crystal can be expressed as

where COis the initial dopant concentration in the melt, g is the relative solidification coefficient, defined as the ratio of the volume of solidification part of the ingot to the whole volume of the melt. The solution of Equation 2 is Ccrystal

= keCo(1-

g)"-l.

(3)

Taking logarithm, Equation 3 can be written as a linear equation: In

Ccrystal ~

co

= Ink,

+ (k, - l ) l n ( l - 9 ) .

(4)

194

Figure 2 shows a linear fit to the GDMS data. The numerical result of the effective segregation coefficient of the yttrium ions in PbW04 crystal is found to be 0.91f0.04. This rather uniform distribution of the yttrium concentration in PbW04 crystals makes it a good dopant.

3. Crystal Properties Optical properties of yttrium doped PbW04 crystals were characterized at Caltech and SIC. All samples are full size of more than 20 radiation length, typically 2.6 cm x 2.6 cm at the large end, tapering to 2.2 cm x 2.2 cm at the small end, and 23 cm long. 3.1. Emission Photo luminescence was measured by using a Hitachi F-4500 fluorescence spectrophotometer. A schematic of the measurement setup is shown in Figure 3, where UV excitation light was shot to a bare surface of the sample and photo luminescence, without passing through sample, was measured by a photo multiplier tube (PMT) through a monochromator. The setup used to measure radio luminescence is shown in Figure 4, where whole body of a wrapped sample was irradiated by 6oCoy-ray at a dose rate of about 1,000 radfh, and the radio luminescence, passing through sample, was focused, passed through a monochromator and measured by a PMT. Yttrium doped PbW04 samples have a broad emission spectrum with a peak at about 420 nm, and the radio luminescence is 15 to 25 nm red shifted as compared to that of the photo luminescence as shown in Figure 5. This red shift is explained by the internal absorption since PbW04 transmittance at shorter wavelengths is poorer than that at longer wavelengths. A cross

Excitalion Monochromalor

I

Figure 3. A schematic of the setup used t o measure photo luminescence.

Figure 4. A schematic of the setup used to measure radio luminescence.

195

t

As-received

im

EX:330Nn

SIC-S301

v

5 .-

O:

AimriAdiatA ' '

950

Wavelength (nm)

Figure 5. A comparison of photo (solid lines) and radio (dashed lines) luminescence spectra for sample S301.

300

350' '

'

'4W

' '

4';

'

'

&i '

'

Wavelength (nrn) Figure 6. Excitation and photo luminescence spectra are shown before and after 7ray irradiations at 9 krad/h for sample S762.

check of photo luminescence with luminescence light passed through the sample showed similar red shift, and thus confirmed this explanation. Since the photo luminescence spectrum shown in Figure 5 is not affected by the internal absorption, it can be seen as the intrinsic emission spectrum, while the radio luminescence spectrum is a convolution of the intrinsic emission and the internal absorption with later depends on the light path15. It is interesting to note that both the excitation and emission spectra of PbW04 crystals is not affected by y-ray irradiation. Figures 6 shows the excitation and photo luminescence spectra before and after irradiations at 9 krad/h for sample S762. Within the measurement errors the shape of these spectra are identical. This observation consists with our previous conclusion that the scintillation mechanism in PbW04 crystals is not damaged by the y-ray irradiation17y3. 3.2. Transmittance and Birefringence

PbW04 crystal has a birefrigent scheelite crystallographic structure. While its a and b axes are equivalent, its c axis is the symmetry axis. The transmittance measured with light propagating along the c axis (ordinary component only) is different from that measured with light propagating perpendicular to the c axis, which may have a mix of ordinary and extraordinary polarization components. Figure 7 shows transverse and longitudinal transmittance for sample U517. Also shown in these figures are the theoretical limits of the transmit-

196

tance calculated by using the refractive index of the ordinary light, propagating along the c axis, extraordinary polarized light and unpolarized light, propagating perpendicular t o the c axis, assuming no internal a b ~ o r p t i o n l ~ >At ~ *the . emission peak, 420 nm, the transmittance of the extraordinary polarized light is about 3% higher than that of the ordinary light. The PbW04 specification on the transmittance, aiming at setting a limit on the internal absorption, thus is crystal orientation dependent, i.e. the transmittance specification along the c axis should be lower than that along the direction perpendicular to the c axis to allow the same density of absorption centers. Since PbW04 crystals are grown along the c axis at SIC, the difference between the longitudinal and transverse transmittance shown in Figures 7 reflects the internal absorption integrated over the light path as well as the birefringence. It is also interesting to note that the transverse transmittance curves measured in two directions perpendicular t o the c axis at both the seed and tail ends are almost identical, indicating that (1) PbW04 crystals grown along the c axis is isotropic transversely and (2) this sample has very good longitudinal uniformity. Figure 8 shows the longitudinal transmittance as a function of wavelength for three PbWO4 samples produced at different time. Also shown in the figure is the theoretical limit of transmittance along the c axis. As seen from the figure that recent grown crystals approach theoretical limit, indicating very low residual absorption. Early samples have a low transmittance at short wavelength caused by scattering centers, but no direct correlations between

80

1

SIC-U517 80

il.

*

-

t

.

e 6 0 -

.-Seed end transverse (x.y) :.--. Tail end transverse (x,y) - Longitudinal (z)

8 .. K m

-

.-s

iii

K

+E

E40;

II

I

Calculated transmittance 0 /I c axis, 0

0

d0""'

'do'

'

5&'

I c axis, unpolarized11 ht Ic axis, 6-polarized Ii&t

'

'

'6h

'

'

'7h' '

'

@ ' !lo

Wavelength (nm) Figure 7. Transverse and longitudinal transmittance of sample U517.

o Theoretical c-axis limit From top to bottom

U517,12/2001 S762,12/1999 s643,11/1999

20 -

u

900

400

500

600

700

t

0

Wavelength (nm) Figure 8. Progress of longitudinal trans&tame of three PbW04 samples.

197 40

J 1999

Transmission at 360 nm (%)

Transmission at 360 nm (%)

Transmission at 360 nm (%)

Transmission at 360 nm (%)

Figure 9.

Progress of longitudinal transmittance at 360 nm.

transmittance and the radiation hardness were observed. A fine tuning of growth parameters improved transmittance at short wavelength, as shown in Figure 8 and 9, where statistical distribution of longitudinal transmittance at 360 nm is presented for crystals grown in 1999 to 2002.

3.3. Light Output and Decay Kinetics

Full size (25 radiation lengths) yttrium doped PbW04 crystals produces typically 10 p.e. per MeV energy deposition. It has also a -2%/"C temperature dependence. Stringent control of temperature and its correction are thus necessary to achieve 1% precision in light output measurement15. Figures 10 shows light output as a function of integrated time for three yttrium doped PbW04 samples. Data in these plots are in units of number of photo electrons per MeV of energy deposition (p.e./MeV). The ratio between light outputs integrated to 100 and 1,000 ns is about 95%, as compared to 90% and 85% for Sb doped and undoped PbW04 crystals re~pectively'~.

198 11

SIC-S347 lo-

SIC-S392

c

,

L 0 = 8 4 pelMeV

(mn6.20 0%)

+ ++

t + + + +

++

t+

+ tttt

dose rate (,a&): 5

Gale(ns) Omdh

0

"

"

50 89

A '

lm

IWO

XK)

100

98 ~

'

2oM)

"

2MK)

'

15

105

104 '

'

'

'

4ow

Time (ns) Figure 10. Light output is shown as a function of integration time for three samples.

0.7

'

,

,

+

loo

+Em

,I M O ,

,

1W

200

3w

400

Time (hours) Figure 11. Normalized light output is shown as a function of time under irradiations for sample S392.

3.4. Light Output Degradation under Irradiation

The light output, however, degrades under irradiation. It is known that radiation induced color centers are created in PbW04 crystals by irradiation, and may annihilate in room temperature. During irradiation, both annihilation and creation processes coexist, the color center density reaches an equilibrium at a level depending on the dose rate applied6. Figures 11 shows light output normalized to that before irradiation (solid dots with error bars) as a function of time under irradiation for sample S392. The light output was defined as an average of nine measurements with a collimated 13'Cs source shooting at evenly distributed positions along the sample to properly evaluate the degradation of light output and reduce systematic uncertainty. Measurements were done step by step for different dose rates: 15, 100, 500 and 1,000 rad/h, as shown in these figures. The degradation of the light output shows a clear dose rate dependence, as described in reference6. Table 3 summarizes the numerical results of the light output before irradiation and the normalized light output (%) in equilibrium under certain dose rates. The light output (L.O.), in units of number of photoelectrons per MeV energy deposit (p.e./MeV), is defined as the average of nine measurements with an integration time of 200 ns at 2O.O0C. By using emission weighted quantum efficiencies of the R2059 P M T (14.8%), the measured light output was converted to the light yield in units of photons/MeV.

199 Table 3. Sample ID S301 S347 S392 S412 S643 S762 606 678 679 L411 U517 LS614 LS615 U685 U686 U688 U689 U690 U691 U692

Summary of light output measurements

L.O. (l/MeV) P.e. 9.4 9.9 8.4 8.3 8.9 10.6 10.4 10.4 10.8 11.7 10.3 8.6 9.9 11.4 10.8 11.1 10.4 12.3 10.8 10.5

Y 63.5 66.9 56.8 56.1 60.1 71.6 70.3 70.3 73.0 79.1 69.6 58.2 66.9 81.4 77.1 79.3 74.3 87.9 77.1 75.0

L.O.(%)@R(rad/h) 100 500 1000

Fraction(%) 50ns _ _ loons lps

lps

15

92.0 91.3 92.0 94.6 88.8 85.6 88.3 85.2 85.0 71.8 70.8 87.2 82.8 91.3 92.1 90.5 92.6 90.1 91.8 92.0

96.6 97.8 97.3 98.6 98.9 94.2 98.4 93.5 94.7 94.0 93.2 97.6 96.9 97.9 97.5 95.5 96.3 96.5 97.7 98.5

96.6 95.1 98.2 98.2 88.3 91.5 91.7 94.2 93.5 88.3 87.0 98.1 89.7 90.4 98.5 91.9 95.5 92.7 94.7 93.2

87.3 88.6 91.3 91.2 79.8 84.2 79.3 76.0 73.5 79.2 71.6 89.5 80.9

79.5 82.1 83.6 85.9

74.3 78.0 80.2 85.3

-

-

81.4

-

-

59.6 57.3 72.3 62.7 84.6 76.1

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

3.5. Damage Recovery

The radiation damage in PbW04 crystal recovers at room temperature. Since the light output of PbW04 crystals varies during both damage and recovery, a precision calibration is the key for maintaining the precision of crystal calorimetry in s i k . For CMS PbW04 calorimeter the calibration is achieved by using various physics while variations of the calibration caused by radiation damage and recovery are traced by a light monitoring system, which measures transmittance to 0.1% at the peak wavelength of PbW04 emission2'. Since a complete monitoring run for entire calorimeter takes about 37 minutes, the variation speed for the light output, either during damage or during recovery, is preferred to be less than l%/h. Figure 12 shows the damage (dots with error bars, left scale), corresponding exponential fit (solid lines) and damage speed (dashed lines, right scale) of normalized light output under irradiation at 15 rad/h for sample S392. Figure 13 shows the recovery (dots with error bars, left scale), corresponding exponential fit (solid line) and recovery speed (dashed lines, right scale) of normalized light output after an irradiation at 400 rad/h for sample S392. While the damage data were fit to

200 11

0 1

SICS392 Dose rate 15 rad%

l . m

0.8

SIC-S392 Alle~400 ram in Equilibrium

.

L Y I L Y . I l r n - O O l ( 1 -e'"m,

075';.

"

'

&'

'

'

'1L'

'

'

1';

'

'

'&

'

'

&'a2

Tirne(hours)

Time (hours)

Figure 12. Light output damage (dots with error bars, left scale) and damage speed (dashed lines, right scale) for sample S392.

Figure 13. Light output recovery (dots with error bars, left scale) and recovery speed (dashed lines, right scale) for sample S392.

Table 4. (Wh) Sample ID 5301 S347 S392 S412 S643 S762 606 678 679

Summary of PbW04 damage and recovery speed Damage - under (rad/h) . . . 15 100 400 1,000 -0.21 -2.3 -4.9 -0.35 -3.9 -0.28 -0.67 -0.54 -0.23 -1.3 -1.6 -0.04 -0.81 -0.10 -0.82 -1.7 -0.85 -1.4 -1.0 -3.6 -3.0 -1.9 -0.31 -0.53 -3.1 -9.0 -5.3 -1.1 -5.8 -0.21 -4.8

Recovery from 400 rad/h 0.37 0.27 0.31 0.19 0.41 0.04 0.48 0.59 0.41

the recovery data were fit to

LY/LYo = 1 - a - beFt/' = (1 - a - b) + b( 1 - e - t / T ) ,

(6)

where a and b are two constants representing the amplitude of the slow and fast components respectively, and r is the time constant of the fast component. Table 4 lists the maximum speed (in units of %/h) of light output variation for the damage process under different dose rates and for the recovery process after reaching equilibrium at 400 rad/h. As shown in the table, some samples may have too fast damage speed under irradiation, indicating possible less accurate calibrations for a short time at the beginning.

201

3.6. Light Response Uniformity

While variations of the amplitude of the light output can be inter calibrated, the loss of the energy resolution, caused by the degradation of light response uniformity is not recoverable3. To preserve crystal's intrinsic energy resolution the light response uniformity thus must be kept within tolerance. The light response uniformity was measured by moving a collimated y-ray source along the longitudinal axis of the sample at nine points evenly distributed along the crystal and the response (y) was fit to a linear function,

JL = 1 + 6 ( X / X m i *

- l),

(7)

Ymid

where ymid represents the light response at the middle of the crystal, 6 represents the deviation of the light response uniformity, and x is the distance from the small (front) end of a tapered crystal. Figures 14 shows the light response uniformity as a function of accumulated dose for samples S412, where 6 is a measure of the light response uniformity, as defined in equation 7. These figures show clearly that the slope ( 6 ) and the shape of the uniformity does not change. This good uniformity can be attributed to the fact that the light attenuation length at the emission peak (420 nm) is long enough even after irradiation, so provides an adequate compensation between the attenuation and the focusing effect caused by crystal's tapered shape3. SIC-S412 I

-

08

Q

c

O ' l

E

08

0,

J

u 1

8

2

O8

zb 1 08 1

08

0

23

46

69

92

115 138

181

184 207

230

Distance From the Front End of Cfystal (mm)

Figure 14. The light response uniformities are shown as a function of integrated dose for sample S412.

1.5

2

2.5

3

3.5

4

Photon Energy (eV)

Figure 15. Radiation induced color center density for sample S392.

202

3.7. Radiation Induced Color Centers The longitudinal transmittance data can be used to calculate radiation induced color center density. Figures 15 shows radiation induced color center density as function of photon energy measured for sample S392 in equilibrium at different dose rates. The stars in these figures represent corresponding radio luminescence spectra weighted with quantum efficiency of the PMT. The points with error bars in these figures are radiation induced color center density (D), or absorption coefficient, measured in equilibrium under dose rates specified. They were calculated according to an equation

D

= 1/LALeguilibrizLrn - 1 / L A L b e f o r e .

(8)

where LAL is light attenuation length calculated by using longitudinal transmittance according to Equation 1 of reference17, and the subscript “equilibrium” and “before” refer to “in equilibrium” and “before irradiation” respectively. The radiation induced color center density was decomposed to a sum (solid line) of two color centers with Gaussian shape in photon energy (dashed lines) :

i= 1

where Ei, C T ~and Ai denote the energy, width and amplitude of the color center i, and E is photon energy. As seen from these figures, the two center Gaussian fit provides a rather good description of the radiation induced color center data with good X2/DoF.

Table 5.

Summary of radiation induced color centers

A! AT A$ Ez/uz m-’ m-1 eV/eV 5301 0.04 0.07 0.11 3.07/0.76 S347 0.00 0.03 0.07 3.07/0.76 5392 0.04 0.06 0.10 3.07/0.76 0.03 0.04 0.06 3.07/0.76 5412 0.05 0.08 3.07/0.76 S643 0.00 0.00 3.07/0.76 S762 0.07 3.07/0.76 0.14 606 678 0.11 0.23 3.07/0.76 679 0.14 0.26 3.07/0.76 L411 0.08 0.10 3.07/0.76 U517 0.09 0.10 3.07/0.76 LS615 0.07 0.07 3.07/0.76 a , b , c , d represent 15, 100, 500, 1000 rad/h respectively. ID

Ei/u1 eV/eV 2.30/0.19 2.30/0.19 2.30/0.19 2.30/0.19 2.30/0.19 2.30/0.19 2.30/0.19 2.30/0.19 2.30/0.19 2.30/0.19 2.30’/0.19 2.30/0.19

A:

rn-l 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.03 0.05 0.01 0.01 0.02

~

A; m-l 0.10 0.10 0.10 0.10 0.14 0.11 0.10 0.10 0.10 0.13 0.17 0.13

A;

AC, m-’

0.22 0.13 0.18 0.15 0.22 0.21 0.26 0.38 0.39 0.43 0.47 0.27

0.35 0.14 0.29 0.19 0.25 0.24 0.54 0.78 0.77 0.54 0.48 0.29

A$ m-l 0.42 0.38 0.37 0.24 -

203

Table 5 lists numerical results of fits for yttrium doped PbW04 samples. Consistent fit with two common radiation induced color centers at the same energy and with the same width was found for all samples. One broad center is at wavelength of 400 nm (3.07 eV) with a width of 0.76 eV, and other narrow center is at a longer wavelength of 540 nm (2.30 eV) with a width of 0.19 eV. This observation is similar to what observed in doped BGO samples21, where three common radiation induced absorption bands were observed for BGO samples doped with different dopants, indicating that the radiation induced color centers are presumably not impurity specific rather lattice structure related. This observation also consists with our previous a s ~ u r n p t i o nthat ~~ the oxygen vacancies, which are lattice structure defects, are responsible for radiation damage in PbW04 crystals. 3.8. Stability

One peculiar property was found in yttrium doped PbW04 samples in early development stage. The light output of some yttrium doped samples was found to increase under irradiations. Figure 16 shows the variation of the light output of an early sample after various irradiation and annealing processes. As shown in the figure, its light output was increased by 12% under irradiation a t 15 rad/h, and was increased further after a thermal annealing at 100°C. It was decreased by 10% after a thermal annealing at 200°C. It was also observed that the variations of the light output of these samples correspond to the variations 12

PWOY L 0-

4 :I 51.1

t

-

5 6 p elMeV (200m. 20 0°C)

*3 4 4

6

E l

.-0

'

4

4

c

4 6

4

'

cmw u):Ho1.1.20

A

o

Dpnr V I I 5 Y . l m n

11 11

O 5

s

4 10

$8

Times

B C

D E

FO

H I

J

I

Figure 17. Samples investigated in hunting Figure 16. Normalized light output after for the origin of the "instability". various irradiation and annealing processes for an early sample.

204

of their transmittance. While showing good light output under irradiations, crystals of this type are not conventional since it is difficult to define their nominal light output. This peculiarity was understood as the effect of pre-existing color center in early crystals shown in Figure 8, and corresponding self optical bleaching by scintillation light, similar to the BaF2 crystal case discussed in reference23. Scintillation light was produced when sample is under irradiation by 6oCo yrays. If the annihilation speed of a pre-existing color center, caused by the bleaching effect of the scintillation light, is larger than the creation speed of a radiation-induced color center under low dose irradiation, the total density of color centers would decrease, not increase, under irradiation6. High temperature annealing also creates such pre-existing centers. A study of optical bleaching by using light of different wavelengths confirmed that the optical bleaching is indeed effective in removing pre-existing color centers. Although an initial thermal annealing at a lower temperature would bypass this problem, R&D was carried out to hunt for the origin of these pre-existing color centers. In this investigation, full size PbW04 ingots were cut into short pieces. As shown in Figure 17, five 2.5 x 2.5 x 4.4 cm samples were cut from ingot 4-1-20, and an 18 and a 5 cm long samples were cut from another full 1.2L

AB

L.Y.,.-,=

*

h

2

0.8il.

I

1 , '

,

1.8 radh

17.9p.e.lMeV

I:.. '

1

. '

" 1

' '

1

'

'

1

.. ' '

1

'

' I '

: CD L.Y.,.,,=17.9p.e./MeV

0 0 (u 1 ' '

" 1 ,

1.1

"

1.8 radh

I

Q

1

I

.

.

Dose rate: 4 0 radh

m

E

b

11 105

3

- IJ

= 12.9 p.e./MeV

L.Y.,,

1'800

0

0.8

' " ' I ' ' ' I

0

20

1.8 radh

11.

+ +

+ +

0

40

0

0

" ' " ' ' ~ " ' ~ " ' ~ ' " ~ " ' ~ ' "

Bo

80

100

120

140

180

0.95 0.9 0

25

50

75

100

180

Time (hours) Figure 18. Normalized light output under irradiation at 1.8 rad/h for five 4.4 cm samples cut from ingot 4-1-20.

125

150

1 5

Time (hours) Figure 19. Normalized light output under irradiation at 4 rad/h for 18 and 5 cm samples cut from ingot B13.

205

size tapered sample B13. All these samples went through standard radiation test at Caltech. It was found that only the sample at the tail end have this problem. Figures 18 and 19 show normalized light output of these short samples under irradiations at low dose rate. No “instability” was found for samples cut from the seed side of ingots. As shown in Table 1 in Section 2, the GDMS analysis revealed that the impurities of Na, K , Cu, As and Mo were concentrated at the tail end15. Following this investigation, several ionic impurities were deliberately added to PbW04 melt to study the consequence. The result shows that monovalant impurities, such as K+ and Na+, enhance pre-existing color center at 420 nm, as shown in Figure 20.

350

400

150

500

550

600

650

700

750

Wavelength (nrn)

Figure 20. Radiation induced absorption of yttrium, potassium and sodium doped PbW04 samples.

Figure 21. Normalized light output loss after irradiation at 35 rad/h.

As a consequence of this study, limitations t o the monovalant impurities were added to the the raw material specification and parameters of crystal growth were modified accordingly. The yttrium doped samples grown at SIC since then are free from this “instability”. Figure21 shows a statistical distribution of the light output loss after irradiation at 35 rad/h measured at SIC. No increase of light output was observed. This observation was also confirmed by measurement at Caltech. Correspondingly, the transmittance at 420 nm has also improved, as shown in Figure 22. 4. Summary

Yttrium doping is found to shift the emission to blue, and yttrium doped PbW04 crystal has a broad spectrum peaked at 420 nm. It is also effective in reducing slow scintillation component, and leading to PbW04 crystals with adequate radiation hardness for the LHC environment. With yttrium doping

206 21

15

5

20

1999

15

E

10

a

0

10

5

0 52

&

%

58

60

62

M

66

68

f0

0 50

52

54

56

56

60

62

€4

Transmission at 420 nm (%)

Transmission at 420 nm (%)

Transmission at420 nm (%)

Transmission at 420 nm (%)

Figure 22.

66

66

70

Progress of longitudinal transmittance at 420 nm.

it is not necessary to implement oxygen compensation. The concentration of yttrium ions in PbW04 crystals is rather uniform with a segregation coefficient of 0.91 f 0.04. The radiation induced absorption in all samples can be decomposed t o two color centers with two common centers peaked at wavelength of 400 nm (3.07 eV) and 540 nm (2.30 eV) with widths of 0.76 and 0.19 eV respectively. These centers are not as deep as that in Sb doped ~arnples’~. This explains a relatively large dose rate dependence of radiation damage and relatively large damage speed in yttrium doped PbW04 crystals. Light output of early yttrium doped PbW04 samples is sensitive to the temperature of thermal annealing. Pre-existing color centers, which are bleachable by scintillation light, led to the “instability”, i.e. the light output increase under irradiations. These pre-existing color centers were found to be caused by impurities at the tail end of ingots, and were formed by contamination of monovalant impurities, such as Na+ and K+. With refined growth parameters and stringent requirement to the purity of raw material these pre-existing color centers were eliminated, so that yttrium doped samples grown recently at SIC are free from the “instability”.

207

Acknowledgments This work is supported in part by U.S. Department of Energy Grant No. DEFG03-92-ER40701.

References 1. Compact Muon Solenoid Technical Proposal, CERN/LHCC LHCC/Pl (1994). 2. A. Gasparian, in these Proceedings. 3. R.Y. Zhu, Nucl. Instr. and Meth. A413 (1998) 297. 4. A.N.Annenkov et al., CMS NOTE 1997/055. 5. M.Nikl et al., J.Appl.Phys.82 (1997) 5758. 6. R.Y. Zhu, IEEE Trans. Nucl. Sci. NS-44 (1997) 468. 7. Z.Y. Wei et al., Nucl. Instr. and Meth. A297 (1990) 163. 8. P. Lecoq et al., Nucl. Instr. and Meth. A365 (1995) 291. 9. M. Kobayashi et al., Nucl. Instr. and Meth. A399 (1997) 261. 10. M. Kobayashi et al., Nucl. Instr. and Meth. A404 (1998) 149. 11. S.Baccaro et al., Phys. Stat. Sol. A164 (1997) R9. 12. E. Auffray et al., Nucl. Instr. and Meth. A402 (1998) 75. 13. Q. Deng et al., Nucl. Instr. and Meth. A438 (1999) 415. 14. X.D. Qu et al., A469 (2001) 193. 15. X.D. Q u et al., A480 (2002) 470. 16. Z.W. Yin et al., Proc. SCINT99, Ed. V. Mikhailin, (1999) 206. 17. R.Y. Zhu et al., Nucl. Instr. and Meth. A376 (1996) 319. 18. G. Bakhshiva and A. Morozov, Sov. J. Opt. Techno1 44 (1977) 9. 19. T Hu et al., in these proceedings. 20. L.Y. Zhang et al., in these proceedings. 21. R.Y. Zhu et al., Nucl. Instr. and Meth. A302 (1991) 69. 22. R.Y. Zhu et al., IEEE Trans. Nucl. Sci. NS-45 (1998) 686. 23. D.A. Ma et al., Nucl. Instr. and Meth. A356 (1995) 309.

94-38,

PERFORMANCE OF PWO CRYSTAL DETECTORS FOR A HIGH RESOLUTION HYBRID ELECTROMAGNETIC CALORIMETER AT JEFFERSON LAB

A. GASPARIAN Department of Physics, NCA&T State University, Greensboro, NC 27411, USA E-mail: [email protected] (On behalf of the PrimEx Collaboration)

The PrimEx collaboration at Jefferson Lab is planning to perform a precision measurement of the neutral pion lifetime via the Primakoff effect. This will be a state-of-the-art experimental determination of the lifetime with a precision of less than 1.5%. Such a measurement requires an electromagnetic calorimeter with high resolution and high efficiency for detecting the photons from pion decay. A new electromagnetic hybrid calorimeter is under construction at Jefferson Lab consisting of 1200 lead tungstate ( P b W 0 4 ) crystal detectors and 600 lead glass Cerenkov modules. Recent beam tests were performed with few GeV electrons on the P b W 0 4 crystals obtained from two different manufacturers (Bogoroditsk, Russia and Shanghai, China). Results from energy and position resolution studies, and the dependence of detector response on radiation rate are presented.

1. Introduction

The PrimEx Collaboration at Jefferson Lab (JLAB) is preparing t o perform a high precision (1.4%) measurement of the neutral pion lifetime using the small angle coherent photoproduction of neutral pions in the Coulomb field of a nucleus, ie., the Primakoff effect'. This measurement will provide a fundamental test of the predictions of the axial anomaly in quantum chromodynamics. Photons from the Hall B photon tagging facility at JLAB will be used to produce neutral pions in the Coulomb field of a nucleus. At the incident photon energies of this experiment ( E y = 4.6 - 5.7 GeV), the Primakoff cross section peaks at extremely small angles ( O p e a k N 0.02'). In order to identify and extract the Primakoff amplitude, the experimental setup must have sufficient angular resolution for detecting forward produced pions. The pions will be identified by detecting the decay photons ( T O -+ yy) in the multi-channel electromagnetic hybrid calorimeter (HYCAL). Therefore, the angular resolution of the detected pions will depend strongly on both the position and energy measurement ac-

208

209

curacies of the calorimeter. In addition, the kinematical constraints imposed by the knowledge of the initial photon energy provided by the tagging system results in an improvement of the angular resolution by about 30%l. Currently, we are constructing the HYCAL calorimeter (116 x 116 cm2 area) consisting of two types of shower detectors: 600 lead glass Cerenkov modules located on the periphery of the calorimeter, and 1200 lead tungstate scintillating crystals in the central region with a hole (4 x 4 cm2) in the middle for passage of the incident photon beam. In the past decade, PbW04 has became a popular inorganic scintillator material for precision compact electromagnetic calorimetry in high and medium energy physics experiments. The performance characteristics of the PbW04 crystals are well known mostly for high energies ( > l o GeV)2 and at energies below one GeV3. In this report, we are presenting results with few GeV electrons for the energy and position resolutions, as well as the dependence of the detector response on radiation rate. These measurements were done with crystals obtained from two different manufacturers: Bogoroditsk (BTCP), Russia and Shanghai (SIC), China. 2. Experimental Setup

In order to check the performance of the crystals and to select the manufacturer, we have done beam tests with a prototype detector. A 6 x 6 PbW04 crystal array (single crystal dimensions: 2.05 x 2.05 x 18.0 cm3) was assembled in a light-tight aluminum box maintained at a stable temperature of T = 5°C. A front-view of the prototype calorimeter is shown in Figure 1. The upper 3 x 6 section of the array was assembled from crystals made by BTCP while the bottom section consisted of the ones made by SIC. The scintillation light from the electromagnetic shower in the crystals was detected with Hamamatsu R4125HA photomultiplier tubes (PMT) coupled at the back end with optical grease. Each crystal was wrapped with lOOpm millipore paper which served both as a light reflector and as an optical isolator between the blocks. The anode signal from each PMT was digitized by means of a 14-bit charge-sensitive ADC (LeCroy 1881M, integration width=200 ns). The light yield of the crystal is highly temperature dependent (- 2%/"C). In order t o keep the detector array at a very stable temperature, the crystal assembly was surrounded by thick copper plates with circulating coolants. Temperature stability at the level of AT = fO.l"C was achieved during the entire period of data collection. Secondary electrons ( E , 4 GeV) pair-produced by tagged photons in a thin radiator was incident on the crystal array. Signal from the PrimEx/Hall B pair-spectrometer system was used as the trigger for the data acquisition. For finer definition of the impact coordinates of the electrons on the crystal array, a pair of X - Y array of scintillating fibers with a fiber-width of 0.2 cm was used.

-

210

Figure 1. The prototype lead tungstate detector.

3. Energy Resolution The energy calibration of the calorimeter was performed with 4 GeV electrons irradiating the centers of each crystal module. To measure the energy resolution, the centers of both the SIC and BTCP crystal arrays were irradiated and sufficient statistics was obtained. We found a slightly better energy resolution for the SIC crystals compared to those from BTCP. We attribute this to the relatively higher (- 20%) light yield of the SIC crystals. For the final energy and position resolution, the central part of the prototype detector was irradiated with 2 and 4 GeV electrons. The reconstructed energy distribution for the 4 GeV electrons is shown in Figure 2 for three different calibrated ADC sums: the central module; the inner section comprising 3 x 3 array; and the total array of 6 x 6 crystals. The central module already contains 75% of the total energy deposition. An excellent energy resolution of ( T E / E= 1.3% has been achieved by using a Gaussian fit of the line-shape obtained from the 6 x 6 array. After subtraction of the beam energy spread due to the finite size of the scintillating fibers, as well as multiple scattering effects in vacuum windows and in air, a level of 1.2% energy resolution was reached for 4 GeV electrons.

21 1

S225

6x6

d

us 2 0 0 1 175 1

1x1 a,/E=4% 3x3 aE/E= 1.6% 6x6 aE/E= 1.3% 3x3

150 1

125 1 100

+

75 1 50 L

25 1

Figure 2. Energy response of a PbW04 crystal array to 4.25 GeV electrons. Left peak: single crystal; center peak: 3 x 3 array; right peak: 6 x 6 array.

4. Position Resolution

The impact coordinates of the electrons incident on the crystal array were determined from the energy deposition of the electromagnetic shower in several neighboring counters. In the case of the PbW04 crystals, the transverse size of the shower is about two times smaller than that in lead glass. As a result, the position resolution in the PbW04 detector with an optimal cell size should be about twice smaller than that of lead glass detectors. To maximize the position resolution, we have optimized the crystals' transverse dimensions, and have selected it to be 2.05 x 2.05 cm2. This size is comparable to the Molihre radius (2.2 cm) of the crystal material. The distribution of the reconstructed coordinates for 4 GeV electrons hitting a crystal cell boundary is shown in Figure 3. The linear dependence of the reconstructed coordinates obtained from a logarithmically weighted average of the cell signals uersus the impact positions determined by the fiber scintillator detectors is shown in Figure 4. As is well known, there is a rather strong

212 c m,

E

5 50

-1.483

40

30 : 20

Reproduced Position (mm) Figure 3. Distribution of reconstructed positions at the boundary between two lead tungstate crystal detectors.

correlation between the position resolution (oz)and the point at which the incoming electrons (or photons) hit the detector face. The bottom plot of the figure shows this dependence for the PbW04 crystals. The (T, is smaller (1.28 mm) near the edge of the cell and increases to 2.1 mm at the cell center. These tests showed no measurable difference in the coordinate resolution for the two types of crystals produced by SIC and BTCP. 5 . Detector Response versus Radiation Dose Rate

It is well known that all types of crystal scintillators are sensitive to integrated radiation dosage. In case of PbW04, one of the more important characteristics is the change in crystal response due to radiation dose rate. For crystals that were developed a decade ago, it was observed experimentally that even for a small dose rate ( 5 1 Rad/hour), the gain changes were at the few per cent level. It was understood that this effect is not a result of the change of scintillation mechanism of the material, but the loss of light output due only to absorption by radiation-induced color centers4. A dramatic improvement

213

'3 30 20 3, 10 N' 0 -10 -20 3

h

E a W

bx

3 2.5 1.; 1

1

25

1 ,I 30

35

40 45 50 X Fibre Number

Figure 4. Reconstructed wersus actual position (top) and position resolution (bottom) across the face of PbW04 crystal array.

of this effect has been achieved in recent years by both commercial producers of the crystals. We have done experimental studies of this effect for crystals produced in the past two years by SIC and BTCP. For these tests, several crystal centers were irradiated with 4 GeV electrons at different rates by changing the incident photon beam intensity and the radiator thickness. The duration of each irradiation was kept to about 30 minutes. The information from all channels of the 6 x 6 array prototype was recorded continuously during the irradiation process. Preliminary data of the normalized pulse-heights versus the relative beam rate for three typical crystals from each of the manufacturers is plotted in Figure 5. It was observed that above the dose rate equivalent to 4 GeV electrons at 50 kHz, the SIC and BTCP crystals show opposite behavior in detector response. This behavior could be caused by three different effects: (1) change of scintillation mechanism in the crystals; (2) change in the light transmission in the crystals; and (3) change in PMT gain due t o rate variations. Currently, N

214

we are in the process of measuring performance of the PMT's at similar rates which will allow us to extract this contribution from the experimental data.

-----

SIC Shanghai BTCP Bogoroditsk

7 Figure 5. Relative pulse-height (preliminary) versus rate for SIC and BTCP PbW04 crystals. The location of the arrow indicates the dosage equivalent for 4 GeV electrons at 5 kHz obtained from GEANT simulations.

Acknowledgments This project is supported by the National Science Foundation under a Major Research Instrumentation (MRI) grant (PHY-0079840). JxeIerences 1. PrimEx Conceptual Design Report, 2000 (http://www.jlab.org/primex/). 2. Compact Muon Solenoid Technical Proposal, CERN/LHCC 94-38, LHCC/Pl (1994). 3. K. Mengel et al. IEEE Trans. Nucl. Sci. 45, 681-685 (1998). 4. R.Y. Zhu, et al. IEEE Trans. Nucl. Sci. 45, 686-691 (1998).

THE PHOTON BALL AT COSY

R.NOVOTNY, W.DORING AND M.HOEK II. Physics Institute, University Giessen, Heinrich- Buff-Ring 16 0-35392 Giessen, Germany E-mail: r.novotny4expZ.physik.uni-giessen.de

M.BUSCHER, V.HEJNY, H.R.KOCH, H.MACHNER, H.SEYFARTH AND HSTROHER Nuclear Physics Institute, Research Center Julich, D-52425 Julich, Germany

J.BACELAR AND H.LOHNER Kernfysich Versneller Instituut, Groningen, The Netherlands

A.WRONSKA Institute of Physics, Cmcow, Poland A compact photon detector with nearly 4a-coverage in solid angle has been proposed t o be implemented into the magnetic spectrometer ANKE at COSY, Julich. Based on the physics program and the experimental requirements the instrumental concept will be presented. As part of the R&D program, the response functions of PbW04 to photons and charged particles have been measured at different accelerator facilities up to energies of a few GeV and are compared to fast scintillator materials such as CeF3 and BaF2. The applicability of fine-mesh photomultiplier tubes (Hamamatsu R5505) has been tested successfully for the read-out of first prototype detectors.

1. Physics Motivation One of the internal target experiments at the cooler synchrotron COSY at the FZ Julich, Germany, is the magnetic spectrometer ANKE (Apparatus to study Nucleonic and Kaon Ejecticles), which consists of three C-shaped dipoles. They are designed to guide the beam to the target, serve as a spectrometer magnet and inflect the beam back to the original orbit. Equipped with a complex set of tracking chambers, Cherenkov detectors and plastic scintillators, charged reaction products of both signs can be detected and identified by their momentum, energy-loss and time-of-flight. The spectrometer covers a large solid angle in the forward hemisphere'.

215

216

Strangeness production in elementary reactions at medium energies has been the primary physics motivation of ANKE,which requires the detection of meson production in nucleon-nucleon and nucleon-nucleus collisions. However, many of the exit channels are accompanied or even dominated by the production of neutral mesons as well, which can only be reconstructed via a missing mass analysis, if the measured kinematical parameters are over-determined. Therefore, a dedicated nearly 4.rr photon detector is intended to be implemented into the ANKE environment to identify directly neutral mesons via their dominant decay channels into photons and to provide even for multi meson events a complete set of physical variables. 2. The Detector Concept 2.1. Technical Constraints

A nearly complete coverage of the solid angle around the target is mandatory for the clean detection of multiple photon events. Due to the fixed geometry of the magnetic spectrometer, the vacuum system and the installation of various target devices the available space is limited to a spherical volume of 75cm in diameter. During operation the spectrometer dipoles are ramped synchronously with the COSY magnets up to a magnetic field strength of 1.6T causing strayfields up to 0.2T pointing in various directions. Therefore, the new detector to measure directly photons up to an energy of 2GeV has to accomplish the following constraints: 0

0

0

0

0

0

large acceptance for detection of multi-photon events. the invariant mass analysis requires good energy resolution (2 12x0) and high granularity. excellent timing and high count-rate capability require as well a fast scintillator as a fast photosensor operating in magnetic environment. the compact design can be accomplished only with a fast and dense scintillator material. flexible geometry to integrate different target systems and to enable fast installation. on-line rejection of charged particles via a separate veto detector system.

2 . 2 . The Detector Design

The generally limited space as well as the fixed beamline components allow the installation of a spherical detector ball of 5 75cm diameter, covering the azimuthal angle range between 25°<0<1600. Initiated by a detailed study of

217

fast and compact scintillator materials for medium energy photon detection and to accomplish the severe technical constraints PbW04 (PWO) has been selected as the most appropriate materia12i3~4. Based on MC-simulations, taking into account the geometrical development of the electromagnetic shower, the expected photon and particle multiplicity and the acceptable count-rate of single detectors, a geometrical arrangement comprising 876 tapered detector modules of 120mm length (F 13.5Xo) with 11 different shapes has been chosen (see Figure 1). The crystals will be mounted in rings around the beam axis.

Figure 1. Schematic view of the PHOTON BALL in front of the dipole magnet D2 of ANKE.

The individual scintillator of maximum trapezoidal diameter of 43mm at the rear surface will be read-out via a 15mm quartz light-guide and a fine-mesh photomultiplier tube (Hamamatsu R5505) shielded in addition by a soft iron cylinder of 3mm wall thickness. On-line charged particle identification will be achieved by a polyhydral array of 72 hexagonal and pentagonal plastic scintillators (thickness 5mm) placed inside the hollow detector sphere. The read-out is foreseen via embedded green WLS-fibers and multi-anode phototubes (Hamamatsu H6568), a design similar to the charged particle veto system of the TAPS spectrometer5. For the fast read-out of the energy and time information and trigger decisions new highly compact 8-fold VME-modules are under consideration to be developed based on the currently produced new electronics for the BaF2calorimeter TAPS6.

218

3. Research and Development 3.1. The Performance of PbW04

3.1.1. The Response to High Energy Photons Pure and Nb/La-doped PWO crystals, manufactured and preselected by Bogoroditsk Techno-Chemical Plant (Russia) and RI&NC (Minsk, Belarus), respectively, have been tested in a series of experiments. The optically polished crystals of 150mm length (-17 X,) with a quadratic front face (20.5x20.5mm2) are rectangular or slightly tapered, wrapped in PTFE-foil and coupled with optical grease to fast photomultiplier tubes (Hamamatsu R4125, covering 35% of the crystal endface). An array of typically 5x5 elements is assembled into a light-tight box, which can be temperature stabilized at T=6f0.1°C. The detector block can be moved by remote control t o illuminate directly each crystal with the collimated test beam. The detector signals have been transfered in most of the experiments via long coaxial cables to the data acquisition system to deduce energy and time information by means of commercial CAMAC electronics. The photon response has been measured with monoenergetic photons between 45 and 770MeV at the tagging facility of MAMI at the university of Mainz, Germany. The intrinsic beam width varies between 2.4MeV and 1.2MeV over the investigated energy range. After the reconstruction of the electromagnetic shower the overall dependence of the energy resolution can be parametrized by o / E = 1 . 4 1 % / ~ ~ + 0 . 9 0 %In. particular at E7=45.4MeV, an excellent energy resolution of o/E=7.4% has been obtained. In parallel, typical time resolutions of a
3.1.2. The Response to Charged Particles The same PWO-array has been exposed to charged reaction products which were produced by an external proton beam of 1.2GeV energy at the synchrotron COSY hitting a 5mm thick aluminum target. A small plastic scintillator delivered a start for a time-of-flight measurement over a distance of 2.9m for particle identification. The intercalibration of the modules was based on the energy loss of cosmic muons and the absolute flight-time was given by particles arriving with the speed of light. Well distinct kinematical curves can be addressed to hydrogen isotopes, a few kaons, pions and minimum ionizing particles. The absolute energy scale has been determined for protons by adjusting the kinematical curve to the energy loss of minimum ionizing protons and the maximum energy stopped in 150mm PWO, taken from GEANT3 simulations. Hydrogen isotopes, which are completely stopped in the array, can

219

be selected in addition by a low multiplicity (M52) of responding detectors. Even for hadrons up to the highest energies the mean multiplicity of M=5.9 remains well below the values measured for electromagnetic showers. All hydrogen isotopes can be reproduced by the same energy calibration parameters. The energy response to protons at fixed energy values has been deduced by appropriate time-of-flight cuts. Taking into account the width of the selected energy bins, the overall energy resolution of g / E = 0 . 9 7 % / d m + 3 . 3 3 % is very similar to the previous values obtained for electromagnetic showers. However, the detected charged pions significantly deviate from the expected energy versus time-of-flight correlation. Comparing the kinetic pion energy, calculated independently from the measured flight time, to the scintillation response allows to estimate the luminescence yield for pions. Assuming a linear relation, the proton signal in PWO appears to be quenched significantly by a factor QF=3.4. Investigations by Kamenskikh7 have reported on a significant shift of the emission spectrum of PWO scintillators towards longer wavelengths and longer decay times, when the ionization density is drastically increased. In the present experiment, the reduced quantum efficiency of the photosensor for green scintillation light as well as the applied short integration gate used for the anode signal, could explain the observed effect.

20

40

60

80

100

proton energy (MeV) Figure 2. Energy spectra of scattered protons emitted from the reaction 85MeV p+CH2 at a scattering angle of 28.9O detected with different scintillator materials.

In an experiment performed at the KVI at Groningen, the Netherlands,

220

+

scattered protons from the reaction 85MeV p CHz/C have been detected with different scintillation detectors. The energy calibration is based on the kinematically known energies of well identified exit channels in the recorded energy spectra as shown in Figure 2. The two prominent peaks in each of the spectra correspond to protons elastically scattered on C and H nuclei. Similar spectra obained for large BaF2- and CeF3 crystals as well as a small Mo-doped PWO sample are shown for comparison. Figure 3 summarizes the deduced energy resolutions for protons at energies below lOOMeV and underlines the applicability of PWO crystals in photon and particle spectroscopy down to very low energies.

20

.

I

I

0

I

I

-+-PWO -+-PWO

2x2xi5cm3 3x3xi2cm

+PWO:Mo +CeF,

1.5x1.5x1.5cm 2x2x14crn3

+CeF3 --b-BaFz

3 x 3 x 13.2 cm TAPS 25 crn

'

PWO 4-

0

.

I

I

1

I

40

60

80

100

energy (MeV) Figure 3. Relative energy resolutions obtained for protons between 40 and 9OMeV for three fast scintillator materials and different crystal geometries.

3.2. The Photosensor In order to fully exploit the fast timing properties of PWO it is mandatory to use photomultipliers rather than slow standard or avalanche photo diodes. Fine-mesh phototubes are basically insensitive to magnetic field components parallel t o the multiplier axis ( B L l T ) . Therefore, only the perpendicular component has to be shielded. The fine-mesh tube R5505 (Hamamatsu, 25.6mm diameter, 15 stages) has been tested extensively in the magnetic field and with respect to linearity, resolution and gain in a first test experiment using a 3x3 matrix of prototype crystals.

221

3.2.1. Read-out in the Magnetic Field A fine-mesh tube illuminated with a blue LED-pulser has been operated in a static magnetic field up to 120mT at different orientations. Whereas the gain of an unshielded tube drops drastically above a perpendicular field >15mT, additional shielding by a soft iron tube with 3mm wall thickness will allow full operation in the expected strayfield up to 0.2T. The shielding of the extremely sensitive cathode region can be further improved by moving the photomultiplier window well inside the cylinder by coupling in between a quartz light-guide of at least 15mm thickness. Operating the tube in the ANKE environment there has not been observed any instability during the data taking cycle caused by the ramping procedure. 3.2.2. Read-out of P WO Crystals In a first test measurement with energy marked photons up to 740MeV energy two 3x3 arrays of PWO crystals with larger diameter but shorter in length (32x32x120mm3), manufactured at Bogoroditsk, have been individually readout with fine-mesh tubes (R5505) operating at bias voltages <2000V to cover an energy range of 800MeV corresponding to a maximum charge of 500pC in the QAC. One of the arrays was equipped with additional quartz lightguides of 10mm thickness. The reconstruction of the electromagnetic shower shows a perfect linearity of the crystal/photosensor response and a good energy resolution of a / E = 2 . 4 0 % / , / m ' + 1 . 3 3 % in spite of the fact that only 25% of the crystal end face are covered by the photocathode (see Figure 4). No difference in the performance has been observed performing the readout via a light-guide. A value of 05200ps can be given as upper limit for the achievable time resolution. Additional measurements with a calibrated blue LED-pulser allow to determine the gain of the tubes showing a variation between 5.105 and 1.106 at 2kV bias voltage. Assuming photons up to 2.5GeV energy and single detector count rates 5 lOOkHz for typical average photon energies of GOOMeV, the extrapolated anode current will remain well below the allowed limit as given by the manufacturer to maintain linearity and long term stability. 4. Conclusion and Outlook

The proposal for the PHOTON BALL, an upgrade of the ANKE experiment at COSY has been accepted by the PAC. The spherical calorimeter can be the first detector system for hadronic experiments at medium energies, which fully exploits the advantageous properties of PWO in spite of its moderate

222 -

I

-

,

-

,

.

,

.

3x3 PWO array with quartz light guide

0

I

1 I

2 00

400

.

,

I

600

.

I

. I

800

1000

photon energy (MeV) Figure 4. Energy resolution for high energy photons measured with a 3x3 matrix of PWO prototype crystals of 120mm length, read-out for the first time with fine-mesh phototubes (R5505).

luminescence yield. The presented research and development, studying the complete response function of PWO to electromagnetic and hadronics probes and the achieved concept for a fast photon read-out in a strong magnetic environment represent the first important milestones in the realization of the project. The development of the read-out electronics, trigger-system and dataacquisition is exptected to run in parallel to the design of the mechanical structure and delivery of crystals and phototubes. The installation at ANKE at the end of 2004 would allow first experiments of rl- and w-production in ppand pn-collisions in the first half of the year 2005.

Acknowledgments This work has been supported by the BMBF and FZ-Julich, Germany. References 1. 2. 3. 4. 5. 6: 7.

S.Barsov et al., Nucl. Instr. and Meth. in Phys. Res. A 462, 364 (2001). R.Novotny et al., IEEE R u n s . Nucl. Sci. 44, 477 (1997). K.Menge1 et al., IEEE Truns. Nucl. Sci. 45, 681 (1998). R.Novotny et al., IEEE Trans. Nucl. Sci. 47, 1499 (2000). S.Janssen et al., IEEE Runs. Nucl. Sci. 47, 798 (2000). P.Drexler, Actu Physicu Polonica. B33, 919 (2002). 1.A.Kamenskikh et al., Proc. of SCINT97, Shanghai, P.R.China, September 22-25, 1997, edt. by Yin Zhiwen et al., 195 (1997).

PERFORMANCE OF THE PWO CRYSTALS OF THE CMS ELECTROMAGNETIC CALORIMETER

F. CAVALLARI INFN, Setione di Roma, Roma, Italy E-mail: Francesca.CauallariQromal.infn.it

The CMS experiment at the LHC has decided to install a homogeneous electromagnetic calorimeter made of about 80000 PWO scintillating crystals. The LHC is a very hard environment for both the radiation levels in the detectors and the high beam crossing rate, so very demanding requirements are set on the characteristics of the detector. After a long R&D the PWO crystals are now well adapted to the experiment. We will review the crystal properties and their implications on the calorimeter performance. In particular we will discuss the light collection crystal uniformity and its effect on the constant term of the calorimeter resolution, measured in a high energy electron beam.

1. Introduction

The LHC at CERN will have as main goal the search for the Higgs boson. Indications from the LEP experiments show that the low mass range is favourite for this particle. In this case the search goes via the H -+ yy channel. Thus a very good energy and position resolution is required for electromagnetic showers of high energy. The CMS experiment at the LHC has decided to install a homogeneous electromagnetic calorimeter made of about 80000 PWO scintillating crystals'. The LHC is a very demanding environment for the detectors and the PWO is well suited for this application, in fact: 0

0

0

the LHC bunch separation is 25 ns but 80% of the PWO light can be collected in 25 ns; the LHC design luminosity is cm-2 s and this gives a high fluence inside the detector, but with the R&D, which has been made, the PWO is considered radiation tolerant and acceptable for the LHC; granularity and compactness are required and the PWO has XO =0.89 cm and RM =2.2 cm; the magnetic field inside the detector is 4 T so the photon sensors must not be sensitive to the field. On the other hand, the PWO light yield is

-'

223

224

rather low compared to other crystals (LY-10 pe/MeV with a PMT 2262B and tyvek wrapping at 18"C), so photon sensors with some internal amplification are required. Avalanche Photodiodes (APDs) and Vacuum Phototriodes (VPTs) have been chosen respectively for the barrel and endcap calorimeter. The PWO temperature coefficient of the LY is -2%/"C at 18"C, this implies that a good temperature stabilization is required.

1.1. The CMS P W O crystals

In 1992 there was the first CMS expression of interest for the PWO as scintillating crystal for the calorimeter and in 1994 the choice was made. From 1994 to 1998 there was an R&D phase2 where the properties of the crystals were understood especially from the point of view of the doping, the transparency and its correlation with the radiation hardness, a research essentially oriented towards the study of the crystal defects. In 1998 a contract was placed for the pre-production of 6000 crystals from the BTCP Institute in Bogoroditsk, which were received in 14 batches. The pre-production phase has led to several very important achievements: an increase of the production yield, an increase in the production rate, an improvement of the quality of the crystals from all the points of view and a very good homogeneity of the crystal parameters within the batch. In 2001 the real production phase started. Since the preproduction phase the crystals are produced in the final shape and dimensions for the CMS calorimeter: a troncopiramidal shape with the small face of about 2.2x2.2 cm2, the large face of 2.5x2.5 cm2 and 25 X, in length, while endcap crystals are larger. The crystals are grown with the Czochralski method and the barrel crystals are cut from ingots of 3.2 cm diameter. During 1999 endcap crystals were successfully obtained from ingots of 4.4 cm. In the course of the year 2000 BTCP was able to produce ingots of 6.5 cm where 2 barrel crystals can be obtained and recent developments show that 4 barrel crystals or 2 endcap crystals can be obtained from ingots of 8.5 cm.

225 2. Crystal Properties

The resolution for homogeneous electromagnetic calorimeters can be parametrized as follows: a(E) E

a

-=-@-@c

@

b

E

where a is the stochastic term, b is the noise term and c is the constant term. The stochastic term in the CMS ECAL barrel can be parametrized as:

a=

J"WAPD

(2)

where F is the APD excess noise factor (N 2) and LYAPD is the light yield of the crystals measured with the APDs. Another contribution to the stochastic term' can be due t o the incomplete radial containment of the shower (estimated to 2% when the shower energy is summed over a 3 x 3 matrix of crystals around the central one). The second parameter is due to the electronic noise. The third parameter is dominated by instabilities in the calibration, fluctuations in the longitudinal shower containment and nonhomogeneities of the crystals. All the crystal properties are measured and checked in the two regional centers with some automatic machines: INFN Rome Center at ENEA Casaccia and CERN. 2.1. Light Yield and Light Yield Uniformity The LY is measured for all crystals to check their quality and t o guarantee the resolution of the calorimeter. In fact this parameter is directly connected with the stochastic term of the detector resolution (see Eq. 2). The acceptance cut is at 8 pe/MeV, which guarantees about 5 pe/MeV in the APDs. In electromagnetic showers the depth of the first interaction inside the crystal can have large fluctuations. This is why any inhomogeneity in the longitudinal crystal response can cause an effect on the constant term and can spoil the resolution for high energy showers. Monte Carlo simulations3 have shown that if the light uniformity curve is uniform t o better than 0.35 %/Xo in the front of the crystals, the effect on the constant term is limited to 0.3 %. CMS checks the so-called Front-Non-Uniformity (FNUF) of the crystals, defined as the slope of the light yield from 3.5 to 11.5 cm from the crystal front, to be less than 0.35 %/Xo in absolute value.

226 0 all polished m Ra = 0 . 3 4 ~ A Ra=0.24y

2

16.5

I

$

Non-uniformity vs Roughness

............................ ... ... ... ... ... . . .

16

. . . .

15.5

.

I

. .

15

14.5 14 13.5

13

1

12.5

0 0

2.5

5

7.5

10

12.5 15

17.5

IIIIIIII 0.1

02

i,li4!,,,l,i, 0.3

0.4

05

20 22.5 2.

Dist. from PMT (cm)

Figure 1. Light yield uniformity curves for a crystal when still polished, after depolishing and after a further depolishing.

Figure 2. Roughness of the depolished face as a function of the FNUF (Ron-NonUniformity of the crystal light collection).

The CMS crystals have a tapered shape and this leads to a focusing effect: the light produced in the crystal front is better focused towards the photon sensors and gives a higher signal with respect to the light produced in the rear. So the crystals are naturally non uniform. A long R&D has led to the understanding of this effect and to the actual solution: all crystals are depolished on one of the side faces up to a given roughness and this is enough to cure the problem4. The value of the roughness of the depolished face has been tuned to render the curve flat. Figure 1 shows the uniformity curves for a crystal when still polished, after depolishing and after a further depolishing. Figure 2 shows how the roughness of the depolished face has been tuned to its actual value. The crystals are now received from the factory with one depolished face and they are just measured and checked. If found outside tolerances they are further treated. Figure 3 shows the resolution calculated on one single crystal for two test-beam matrices made respectively of russian crystals from BTCP depolished as explained, and a second set from SIC China, which had not yet been depolished. The higher value for the constant term of the Chinese crystals can be explained by the higher FNUF.

227

0.004

o.iw 0

0

-0.4

-0.2

0

russian proto 2000 cliiiiese proto 2000 0.2

0.4

0.6

0.K

I

1.2

I3NllF

Figure 3. FNUF of several crystals versus the energy resolution for each single crystal in a beam of 120 GeV electrons. The Figure shows two sets of crystals: a Chinese and a russian set. The Chinese crystals were not uniformized before the test-beam, so their FNUF is higher. Figure 4. Improvement of the LTO (420 nm) during the pre-production phase.

2.2. l'hnsmission

Before assembly the Longitudinal Transmission (LTO) and Transversal Transmission (TTO) are measured for each crystal. These measurements guarantee that the crystal has good general quality: 0

0

0

the LTO(600 nm) guarantees that the crystal has no macroscopic defects; the LTO(420 nm), that is at the emission peak of the PWO, guarantees a good LY; the LTO(360 nm), that is around the radiation-induced absorption band, togeteher with the slope at the inflection point of the curve, guarantee a good performance from the point of view of the radiation resistance. The dispersion of the TTO curves at a transmission of 50% guarantees an homogeneous distribution of the dopant.

Figure 4 shows the improvement in the LTO(420 nm) during the preproduction both from the point of view of quality and homogeneity of the distribution.

228 80

Front irrad., 1.5Gy, O.l5Gy/h

70

-

LYloss=(LY,-LYi,,)/Lyo

M)

(%)

50

E40

30

---after

20

irradiation4

93

-

10 0 3x7

350

400

450

500

5M

6w

650

700

wavelength (nm)

(a) Figure 5. (a) LTO curve for a barrel crystal before and after irradiation and (b) LY loss for several crystals.

2.3. Radiation Hardness

Monte Carlo simulations show that the total dose after 10 years of running at the LHC will be 2x103 Gy in the Barrel. The dose-rate at high Luminosity in the barrel is 0.15-0.3 Gy/h and in the endcaps 0.3-15 Gy/h, depending on q. Figures 5(a) and (b) show the LTO curve for a barrel crystal before and after irradiation and the LY loss for several crystals. These two figures show that the scintillation mechanism is not directly affected by the radiation but there is a significant transparency loss in the crystal, which induces a loss of light. The transparency loss reaches a saturation level after few Gy and it depends on the dose-rate. The acceptance cut for CMS is at 6%. The LY loss saturation level has been shown to be correlated with the LTO properties at the band edge and in particular with the slope at the inflection point and the LTO(360 nm). Figure 6 shows the distribution of the LY loss for several crystals, where the few crystals outside tolerance had already been rejected for their LTO curve.

3. Crystal measurements Among the important results of the R&D on the PWO crystals there is the fact that now all the relevant measurements have been defined and tolerances have been established: all the checks which are relevant for the performance of the calorimeter can now be done in the laboratory before assembly. In fact all these characteristics are measured and checked in the two regional centers before assembly with two automatic machines, called ACCOS machine5i6. These

229

l o 'Statistics on 368 crystals

Mean : 2.45 % StDev : 1.06%

a 6

4

2

-

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

0 0 . 1 1 . 2 2 . 3 3 . 4 4 . 55. 66. 7 7 . 8 8 . 9 9 . 1 LY loss

Figure 6. Saturation level in the LY loss for 368 crystals. T he two crystals outside the 6% tolerance had already been rejected for the Longitudinal Transmission at 360 nm or for the Slope.

laboratory measurements are so good that they can predict the intercalibration of the crystals at the level of about 5% as shown in Figure 7 where the actual calibration found in the test-beam is compared with the prediction based on the lab. measurements of the LY, the APD gains, and the electronics calibration7.

-

1.4

3

1.3

/

/

r

3

5

;

4

3.5

1

3 1.5

I5 I 0.5

0.6

0.8 0.9 1 1.1 1.1 1.3 1.4 LY ACCOS plrdklim in APD / LY chmnrl I

0.7

0 -0.3 -0.2

-0.1

0.1 0.1 0.3 (Rcdictcd Cali. - coliYcali

0

Figure 7. Crystal inter-calibration found in the test-beam compared with the prediction based on the lab. measurements of the LY, the APD gains, and the electronics calibration.

230

4. Test-beam performance The careful control of crystal parameters shows that a very good resolution can be obtained. Figure 8 shows the response of a matrix of crystals t o a very high energy electron beam (280 GeV). The resolution is extremely good ( o ( E ) / E21 0.4%). In fact this is a proof of principle of the detector performance. We are now waiting to test on the beam a larger set of crystals with the final electronics, but on the basis of our laboratory tests we are confident about the result.

Figure 8. Energy resolution of a PWO crystal matrix in a high energy electron beam (280 GeV).

References 1. CMS Technical Design Report 2. G . Yu. Drobychev et al., Results of PWO radiation hardness optimization, CMS NOTE 1999/062. 3. D.Graham, C. Seez, Simulation of Longitudinal Light Collection Uniformity in P W O Crystals, CMS NOTE 1996/002. 4. CMS NOTE 2001/004 G. J. Davies et al., Development of a Uniformisation Procedure for the PbW04 Crystals of the CMS Electromagnetic Calorimeter, CMS NOTE 2001/004. 5. E. Auffray Hillemanns et al., Performance of ACCOS, an automatic crystal quality control system for P W O , CMS NOTE 1999/067. 6. S. Baccaro et al., NIM A 459 (2001) 278-284. 7. E. Auffray et al., Test-beam results on the performance of two matrices of P W O crystals for the CMS E C A L and comparison with Laboratory measurements, CMS IN 2001/033.

AVALANCHE PHOTODIODES FOR THE CMS LEAD TUNGSTATE CALORIMETER.

J. GRAHL, I. KRONQVIST, R. RUSACK, A. SINGOVSKI University of Minnesota, Minneapolis, MN, USA

A. KUZNETSOV, Y. MUSIENKO, S. REUCROFT, J. SWAIN Northeastern Univerity, Boston, M A , USA

K. DEITERS, Q. INGRAM, D. RENKER, T. SAKHELASHVILI Paul Scherrer Institute, Villigen, Switzerland

Avalanche Photodidoes (APD’s) will be used to readout the CMS barrel electromagnetic calorimeter. After several years of detailed prototyping studies the final design of the CMS APD is now fixed and it is now in production. The design of the calorimeter will make access to the APD’s extremely difficult once they are installed. In view of this the APD group has devised a system of tests to ensure that the APD’s will have a high-probability of surviving for ten years in the harsh environment in the CMS detector. In this paper we describe the methods that we have adopted to select the APD’s for the CMS detector.

1. Introduction The CMS electromagnetic calorimeter will consist of a barrel and two endcap calorimeters made with lead tungstate crystals’. This choice is motivated by the desire to have the highest feasible electromagnetic energy resolution at LHC energies in order to be able to study the production of electrons and gammas by such processes as H + yy. The barrel portion of the calorimeter will consist of 61,200 lead tungstate crystals arranged symmetrically around the interaction region. The two endcap calorimeters will each be constructed from two Dee-shaped crystal arrays mounted on either side of the beam pipe. As the whole calorimeter will be placed inside a 4 T magnetic field, with the barrel crystals projecting towards the interaction region, the scintillation light from the crystals in the barrel cannot be readout with a conventional vacuum photodetector and instead a semiconductor photodetector with a small electron trajectory is required. Conversely, for the endcap calorimeters the situation is different, as the main axis

231

232

of the crystals is nearly parallel to the magnetic field and for these calorimeters we will readout the crystals with vacuum phototriodes2. The light yield from the lead tungstate crystals that we will use in CMS is approximately 50 photons per MeV of deposited energy, or approximately 1.5% of the light output from NaI(T1) crystals. This relatively low light yield precludes the use of PIN photodiodes since the signal from a minimum ionizing particle traversing the diode - -1,600 photoelectrons for a diode 200 p thick - would be the same magnitude as the signal from a 1 GeV electron in the crystal calorimeter. To circumvent this problem CMS has chosen to use an avalanche photodiode manufactured by Hamamatsu Photonics to readout the barrel calorimeter. For the construction of the barrel, the crystals are grouped into 36 ‘supermodules’ each containing 1,700 crystals. Every crystal is readout with two APD’s glued to the end. The output signal of the APD pair is combined and amplified by a custom preamplifier that is a 40 MHz multi-ranging track-andhold5. The output of this circuit is digitized by a 12-bit 40 MHz ADC, one per crystal, and the data is then sent to the data collection system. The eighteen supermodules will be mounted at each end in a ring that covers the full azimuth and the pseudo-rapidity range 1771 < 1.48 inside the magnet, supported by the hadron calorimeter. Once installed a supermodule can only be removed with great difficulty and accordingly we are selecting the supermodules’ components with sufficient reliability that it should never be necessary to remove a supermodule. This requirement on the component reliability is further complicated by the presence of ionizing radiation and neutrons during the operation of the LHC. The readout of the crystal calorimeter is at the hadronic shower maximum, where a total integrated dose of 2 kGy and 2xlOI3 neutrons/cm2 is expected in regions of high 1771. Since there is no prior experience with operating a large number of APD’s in a high radiation environment, either in a high energy physics experiment or elsewhere, we have had to define an extensive set of tests to ensure that the very stringent reliability requirement is met. 2. APD properties

The properties and operation avalanche photodiodes have been described extensively elsewhere3s4. In summary they are silicon diodes operated with a high internal electric field sufficient to cause impact ionization by the electrons as they propagate through the lattice. At low electric fields the hole amplification is small, but it becomes more significant as the electric field increases, introducing noise into the current amplification process. A schematic of their operation is shown in Figure 1.

233

Electric Field

Figure 1. First 3 normalized frequencies versus release location for clamped simply supported beam with internal slide release.

The CMS APD’s are manufactured by Hamamatsu Photonics. They are made by epitaxial growth on n++ silicon combined with ion impantation and diffusion. The junction is at a depth of approximately 5pm so that the blue light from the crystal will be absorbed, while the electron-hole pairs formed by ionizing particles that are subject to amplification is minimized. Behind the junction there is an additional n-type layer to increase the depletion depth and thus reduce the capacitance. The active area is 25 mm2 and they are mounted in a 1 cm2 package. The spectrum of light from the lead tungstate crystals is shown in Figure 2 The quantum efficiency of the CMS APD’s is shown in Figure 3 is well matched to this spectrum. The Gain - QE product turns down for wavelengths above 500 nm due to incomplete absorption of light in front of the junction. The gain ( M ) of a typical CMS APD as a function of voltage is shown in Figure 4. The operating voltage at gain 50 (VR)and the breakdown voltage (VB)are shown in the figure. Though gains in excess of 1000 can be achieved with these APD’s, the excess noise factor ( F ) also increases with the gain. Due

234

Figure 2. Relative output spectrum of the scintillation lead tungstate.

Figure 3. T he quantum efficiency at gain one as a function of wavelength.

t o the increase in the contribution from the holes t o the amplification process. Consequently there is an optimum operating value for the gain where the signal-to-noise is maximized. The actual value is determined by the properties of both the APD and the preamplifier that is used. For operation of the APD with the CMS preamplifier, which has a 43 nsec shaping time, the optimum gain is in a broad minimum between 50 and 100. We have chosen to operate them at a gain of 50. All APD's have a temperature dependence of the gain. For the CMS APD, this temperature variation, x is -2.2%0/"C. This temperature depen-

% &,

10000 1000

j

I

I

I

I

I

I

I

I

1

I

0

50

100

I

I

I

I

I

I

I

250

300

~r

I

J

I

I

I

1

0.1 150

200

Voltage [ V l

350

t 408 I

VR

450

1

vE3

Figure 4. Gain versus applied bias. T h e operating voltage (VR) is around 360 V while the breakdown voltage (V,) is approximately 40 V higher.

235 dence is the same magnitude and, unfortunately, has the same sign as the variation in light output of the crystal with temperature and hence the design of the calorimeter has very strict requirements on the temperature stability. It is our design goal to stabilize the temperature of the APD’s and crystals t o 0.1”C. The properties of the CMS APD’s are summarized in Table 1. Table 1. The CMS APD properties. 5 x 5mm2

Active Area Quantum Efficiency at 420 nm

75%

Light Collection in 20 nsec

99 f 1%

Operating Voltage

N

Operating Gain

50

Capacitance

80pF

Series Resistance

30

Dark Current before irradaition Voltage Sensitivity Temperature Sensitivity

380V

x

&

< 50nA 3.2%/V

x

&

Excess Noiose Factor at Gain 50 Thickness sensitive to ionizing particles

2.2%/0C 2.1 5Pm

3. APD Production The CMS APD’s are manufactured at Hamamatsu Photonics in Hamamatsu, Japan and, after packaging and testing, they are shipped to PSI in Switzerland. At PSI, and later at CERN, they undergo a rigorous set of tests so that any APD which shows an indication that it might fail during the detector operation is removed from the production stream. The flow of the testing begins before the APD’s leave the manufacturer, where the gain, dark current and capacitance are measured as a function of voltage and the breakdown voltage is determined. At PSI they are irradiated with a 6oCo source to 5 kGy, at a dose rate of 2.5 kGy/hr. After allowing twenty-four hours of relaxation time, the APD dark current at a gain of fifty and the breakdown voltage are re-measured to determine if the APD has been damaged by the irradiation. The APD’s are then shipped to CERN where the noise power is measured at frequencies up to 1 MHz and at gains up to 300. Then they are baked-out in an oven at 80°C under bias for one month, which corresponds t o three years of

236 50

0 -

7-

4

50

4

A

>

a

-200 -250

-300 1 5 0 >

-400

I -*J"

I

0

100

200

300

400

APD # Figure 5 . Change in the breakdown voltage with 6oCo irradiation. Th e APD's which show a significant change are not used in the CMS detector.

continuous operation. After these steps, the breakdown voltage and the dark current as a function of applied bias are re-measured to test for any changes. The results from these measurements are used to select the APD's to be installed into the calorimeter. Figure 5 shows the difference between the preand post-irradiation measurements for a group of 450 of APD's. As can be seen in the figure there are a small number APD's that show a significant change in the breakdown voltage under irradiation and these are rejected. Likewise in Figure 6 it can be seen that for some APD's there is a change in the dark current that is significantly larger then that of the other APD's from the same manufacturing lot. Those APD's which show these type of significant changes are rejected. Another criteria which we have developed as a way of identifying APD's that are more likely t o fail is by selecting on the shape of the plot of the ratio $'$ as a function of gain. Generally, for an APD that is operating correctly without any internal or surface breakdown this ratio should tend towards a stable value as the gain increases. This is because the bulk generation of electron-hole pairs in the p-type region in front of the junction will be amplified and surface currents will not. Thus at sufficiently high gains $'$ will be dominated by the amplified bulk current and the contribution from the surface current will become negligible. Cases where there is break-down on the surface or some other contribution to the cuurent occuring at higher gains, can be found by

237 1500

4

c

1200

Y

cr

$

900

22141-

k k ?

U

x

k

6oo

a

n 300

, 20500

21000

21500

22000

22500

APD number Figure 6.

Dark current at gain 50 after 6oCo irradiation.

examining these plots. Figure 7 shows this ratio for many APD’s and the cwes where there is anomolous behavior can be clearly seen. There is a similar selection procedure, that is well correlated with the I D/ M selection, made on the noise power measured after the 6oCoirradiation. Again, any APD’s which show anomalous behaviour rejected. In addition to these tests made on individual APD’s there is another type of selection procedure that is based on an analysis of the APD’s selected. If it is found that a particular wafer position has a high probability of producing a bad APD, then all the APD’s from that position on other are rejected, even if they have passed our selection tests. At an early stage during the production we were able to identify bad wafer positions which, after consultation with the manufacturer, were traced to a mask defect. The overall rejection rate after each step is given in Table 2. Those APD’s which pass all these tests are then sorted into bins according to their operating voltage and grouped into pairs prior to shipment to Lyon where the pairs are mounted in capsules. These capsules are then remeasured for gain and noise to ensure that there has been no damage during transit and assembly and then they are glued onto the crystals. While these selection procedures are intended to remove APD’s which show any sign of a tendency to failure. It is difficult to extract from these tests the failure rate that can be expected to occur during operation. However, a

238

Figure 7. The ratio of the dark current ( I o / M ) as a function of the APD gain. Table 2.

The APD rejection rates.

Rejection criteria

% Rejected

Change in VB after ""Co irrdiation

1.8%

Change in I D after 6oCo irrdiation

0.3%

Anomalous

2

Noise Power

0.4% 2.4%

Change in VB after bake-out

0.1%

Change in I D after bake-out

0.3%

2 after bake-out

0.1%

Change in

Bad Position analysis

0.6%

self-consistency check can be performed to ensure that the procedure is indeed selecting out weak APD's by repeating all the tests on a statistically significant sample. This test was performed on 475 APD's, which had already passed our screening test, it was found that only 1 one of them failed. While not conclusive, it does indicate that the procedure does effectively select good APD's. 4. Sample Testing

In parallel with the production testing, where every APD is tested, we also conduct more detailed tests on a sample, to date about 5% of the APD's de-

239

livered. The purpose of these tests is to ensure that critical parameters of the APD remain stable during production and that the CMS specifications continue to be met. The parameters that are measured are the quantum efficiency, the capacitance as a function of bias voltage and the temperature dependence of the gain. LFrom this sample of APD’s a subset are subjected to neutron radiation at the Minnesota 252Cf irradiation facility6 with a flux of 2 x 1013 1 MeV neutrons/cm2, the maximum neutron flux expected at the LHC in the region where the APD’s will be placed. In these tests the APD’s are irradiated at gain of 50. In all 1200 APD’s have so far been tested in this way and not one has failed, behaving as expected with a steady increase in the dark current during the irradiation and having only a small changes in the breakdown voltage after the irradiation. 5 . Conclusions

The APD that will be used to readout the lead tungstate crystals of the CMS electromagnetic calorimeter have been developed in collaboration with Hamamatsu Photonics. As they will be inaccessible during the operation of the detector they are required to be very robust. To select out the APD’s which are likely to fail we have devised a series of tests which include irradiation with a 6oCosource and a month-long bake-out at 80°C. These tests combined with a destructive neutron irradiation suggest that they will meet the strict requirement of a failure of rate less that one in a thousand. Acknowledgments

This work is support in part by the US National Science Foundation’ the US Department of Energy, including the Californium Industrial Loan Program adminsitered through the ORNL. References P. Lecompte, these proceedings. K.W. Bell et al. Nucl. Instr. and Meth. A469,29 (2001). P.P. Webb, R.J. McIntyre and J. Conradi. R C A Review 35,234 (1974). K. Deiters, et al. Nucl. Instr. and Meth. A442,193 (2000). J.P. Walder, H. Mathez, P. Pangaud, J.M. Bussat, P. Denes I E E E Trans.Nuc1. sci. 6 . To be published.

1. 2. 3. 4. 5.

CMS/ECAL BARREL CONSTRUCTION AND QUALITY CONTROL

E. AUFFRAY CERN, EP-CMA, Geneva, Switzerland E-mail: [email protected]

(On behalf of CMS-ECAL) The CMS electromagnetic calorimeter is now in its construction phase. The barrel part of this calorimeter consists of 36 super-modules containing in total 61200 PbW04 crystals equipped with 2 avalanche photodiodes (APDs). Each supermodule contains 1700 crystals assembled in 4 modules. The construction of these modules is shared between two regional centres, CERN and INFN/ENEA Rome. Prior to the construction, all the ECAL components follow a strict quality control procedure. In this paper, a presentation of the organisation of the ECAL barrel construction, a status of the construction progress as well as the results of the quality control of some components like crystals and APDs will be given.

1. Introduction In 2007, the new Large Hadron Collider (LHC) at CERN will allow to study new physics and in particular to search for the Higgs boson. In the interesting mass region of 100 to 150 GeV, where a light Higgs has a good probability to exist, only its two-photon decay stands a good chance to be observed. For this, a good electromagnetic calorimeter with an excellent mass resolution well below 1%is mandatory. Therefore CMS, which is one of the 4 LHC experiments, will be equipped with a homogeneous crystal calorimeter, composed of about 80000 scintillating lead tungstate (PWO) crystals. The calorimeter is subdivided into a barrel and 2 end-cap regions1. To provide best quality and homogeneity over a long construction and operation period of several years, the production and assembly of such a calorimeter require an industrial organisation and management. The production, the quality control and the assembly of thousands of individual components have to be thoroughly organised and supervised in order to avoid any impact on cost, planning and performance of the detector. In this paper, after a short overview on the barrel structure, the current production status with emphasis on the quality control will be reported and discussed.

240

24 1

2. Barrel construction scheme The barrel part of the electromagnetic calorimeter is made of 36 so-called supermodules, each of them containing 1700 PWO crystals. Each super-module has a modular structure as drawn in Fig. 1.

Supermodule assembly at CERN

Figure 1. Assembly sequence of CMS-ECAL barrel.

The basic part of a super-module is a so-called sub-unit, composed of a crystal and 2 APDs mounted inside a capsule glued on the back of the crystal. The crystals together with the capsules are mounted into a 2 by 5 alveolar structure made of glass fibre and specially treated aluminium foil t o form a submodule. Starting from v=O, 17 different sub-module geometries are required t o provide optimal coverage of the barrel region up t o v=1.5. The sub-modules are then mounted on a grid structure to form a module. 4 types of modules exist depending of the 7 region. The first module at v=O contains ,5 x 10 sub-modules in total, the 3 other ones 4 x 10 sub-modules. The assembly of these 4 modules results in a super-module. In the next step, the monitoring system will be installed on the front part of the super-module by plugging an optical fibre on each crystal. This step is followed by the installation of the cooling system at the back. In the last step,

242

electronic boards containing the front-end electronics will be mounted at the back of the super-module. 3. Component production overview

The different components of the ECAL are produced at different places around the world. For instance, the APD production is done by Hamamatsu in Japan. Their reception and quality control is done at PSI and at CERN under the coresponsibility of the Minnesota university, The North-Eastern university and the Paul Scherrer Institute2. They are then sent to IPN Lyon where the capsules are produced and characterised. The crystal production is done at the Bogoroditsk Techno-Chemical Plant (BTCP) in Russia, where an infrastructure for the production of the required amount of crystals for the whole ECAL has been installed. Two regional centres are in charge of the construction of the different modules. One is located near Rome and operated by INFN/ENEA, the second one is at CERN under the responsibility of the CERN CMS-ECAL group. The main tasks of those centres are:

0 0

0

Part reception and storage Crystal quality control Capsule gluing Sub-module and module assembly

Under full production conditions, 1 super-module per month will be assembled, thus resulting in a total amount of 50 crystals to treat per day in both regional centres. This, firstly, requires an effective and reliable quality control of each part of the calorimeter at each step of the assembly in order t o avoid nonworking channels and therefore to ensure a fully operational detector. Secondly, a high level of automatisation in the quality control in each step of the production and assembly chain is mandatory. The current state of the quality control as well as the production will be described in the following section. 4. Quality Control Quality control of the detector elements comes in at different levels of the production chain. 4.1. Crystal quality control

After reception and a fast visual inspection, the crystals are characterized at CERN and at Rome using the so-called ACCOS machine (Automatic Crystal

243

Control System), an automatic control bench for measuring crucial crystal parameters like dimension, light yield, transmission and decay time3t4. Figure 2 shows the ACCOR and ACCOS systems in operation. Every 6 hours and half, 30 crystals are fully characterized in each regional centre. If one of the parameters is out of the defined specifications5, the crystal is rejected. The rejection rate is currently between 1 and 2%6.

Figure 2. ACCOS machines: Left: ACCOR at INFN/ENEA and Right: ACCOCE at CERN regional centres.

4.2. APDs and Capsules quality control

After their delivery at PSI, the APDs have to follow a very strict quality control concerning their intrinsic properties (gain, noise, dark current) and their evolution under irradiation. This quality control is described elsewhere2. They are then sent to Lyon where they are assembled by pair inside a plastic white square piece to form the so-called capsule (see the left side of Fig. 3). A kapton cable is then soldered. 23 types of capsules exist depending on the Kapton length and on the presence or not of a thermal sensor. Once the capsules are made, they are tested in terms of gain curve, dark current and noise behaviour on an automatic bench called Cupucine, allowing test of a large number of capsules in parallel (200 per day), as shown in the right side of Fig. 3. This control allows to fully characterize the capsules, and to reject the non-working ones before to send them to the regional centre.

244

Figure 3. Left: Different types of Capsules. Right: Automatic control test bench of Capsules at IPN Lyon: Capucine.

4.3. Capsule gluing

In the two regional centres, the capsules are glued on the crystals, at a rhythm of 50 per day, to form sub-units. The quality of the gluing is tested using an optical bench for bubble control. Furthermore, electronic functionality tests on those subunits are performed on a bench similar to Cupucine: Cupucinette, in order to check that the gluing operation did not damage the capsule. 4.4. Sub-module assembly

In this step, 10 sub-units are assembled in a sub-module. As a quality check, test is performed on the correct positioning of the crystals inside the alveole and correct cabling of the capsules. The photodetectors are illuminated with a standard LED, and the signal is checked. 4.5. Module assembly

In the next step, each sub-module is fixed on the grid. As for the sub-modules, the correct cabling of the capsules is checked in case of accidental unplugging by mechanical operations. 4.6. Data tracking

During the complete assembly sequence described so far all production and quality check data are acquired using the CRISTAL database system7>*.The CRISTAL system, which has been developed in parallel to the construction setting-up, provides a dynamic workflow execution in a highly distributed environment at each level of the production: Each operator at his “working place” has to follow the instructions and the workflow provided by CRISTAL.

245

Each operation done during the assembly of one part has to be validated. For the composite part, the exact position of each component is stored. A query facility allows for navigating the construction database at every stage of the construction and during the operation of the experiment and for retrieving the critical parameters of the crystals and capsules needed for the calibration of the detector. 5. Current production and assembly status The pre-production of the crystals started in September 1998 until December 2000, when 6000 crystals had been produced. The good quality of the preproduction crystals allowed starting the mass production of the barrel crystal in Bogoroditsk techno-chemical plant in Russia, in year 20009. In parallel to the crystal production, the two regional centres progressively set-up the module construction facility. The CERN regional centre is on production since spring 2001 and Rome since beginning of 2002. After a “warming up” period, which has allowed to test and validate all the assembly procedure and tooling in both regional centres, the following number of parts of the detector have been received, characterised and assembled using the assembly and quality check procedures described in section 4:

Element Crystals APDs Capsules Sub-units Sub-modules Modules

Number of elements 8700 30000 4000 3500 350 8

The 4 modules (see two of them in Fig. 4) for the first super-module are now ready and its assembly is currently under-work. After the full validation of the first super-module and its components, their mass production will be launched and the routine super-module assembly will start beginning of 2003 at an interval of 1 super-module per month. 6. Conclusion

The complexity of the CMS electromagnetic calorimeter and the huge amount of its components require a sophisticated and efficient organisation in terms of

246

Figure 4. 2 modules (SMO/Ml & SMO/M4) of the first super-module.

management and quality control. Each production centre and particularly the two centres responsible for the module assembly have succeeded in setting-up such a good organisation that should allow building a performing detector in time.

Acknowledgments The author wants to thank all the members of the CMS-ECAL group and particularly her colleagues of the production centres.

References 1. ECAL Technical Design report, CERN/LHCC97-33, CMS TDR4 (Decl5, 1994). 2. R. Rusack, paper presented in these proceedings. 3. E. Auffray et al., Proceedings of the 1998 IEEE Nuclear Science Symposium Conference in Toronto, Canada, Vol. 1, p. 508-13 (Nov. 1998). E. Auffray et al, Nucl. Instr. and Meth., A 456, 325-341 (2001). 4. S.Baccaro et al., Nucl. Instr. and Meth., A 459 278-284 (2001). 1.Dafinei et al., Proceedings of the SCINT99 conference i n Moscow, Russia (August 16-20, 1999). 5. E. Auffray et al., CMS Note 1998-038. 6. F. Cavallari, paper presented in these proceedings. 7. N. Baker et al., Computer Physics Communication, Vol. 110, No 1-3, pp170-176, (1998). 8. A. Bazan et al., ZEEE n u n s on Nuclear Sci., Vol. 46, No 3, pp392-400, (1999). 9. E. Auffray et al., Proceedings of the SCINT2001 conference in Chamonix, fiance, (September 16-21, 2001), to be published in NIM.

Medical Applications Covener: C. Woody

C. Woody

Covener’s Report

t W. W. Moses

Synergies Between Electromagnetic Calorimetry and P E T

*C. L. Melcher

LSO - from Discovery to Commercial Development

P. Lecoq

New Scintillating Crystals for PET Scanners

J. Collot

A Simulation Framework for Positron Emission Tomography Based on GEANT4

+The perspective talk *Written contribution not received

247

This page intentionally left blank

MEDICAL APPLICATIONS

CRAIG WOODY Physics Department, Brookhaven National Lab, Upton, N Y 11973, USA E-mail: [email protected] (Convener’s Report)

There are a number of areas where the techniques used in calorimetry for high energy and nuclear physics are similar to those used in medical applications. Perhaps first and foremost of these involves the use of scintillating crystals and photodetectors to detect gamma rays in medical imaging and nuclear medicine. While these gamma rays may be considered low energy from the perspective of high energy calorimetry, the types of materials and the techniques used for measuring them are very similar in both cases. In addition, the need for large solid angle coverage for tomographic imaging using highly pixilated detectors is also common to both applications. More recently, there have also been many similarities in the areas of developing fast front end electronics, data acquisition, Monte Carlo techniques and image reconstruction. W. Moses gave an excellent overview on the methods and techniques used in medical imaging, particularly in Positron Emission Tomography (PET),including the use of scintillating crystals, such as BGO and LSO, and various types of readout devices, such as position sensitive photomultiplier tubes and avalanche photodiodes. These systems typically use crystals which have dimensions on the order of a few mm2 cross section by 1-2 cm in length. Human scanners used for imaging the brain have apertures with dimensions on the order of 30 - 40 cm in radius, and can contain more than 100,000 readout channels. With its fast front end electronics and high rate data acquisition systems, a P E T scanner can be seen as miniaturized but equally complex calorimeter system. C. Melcher from CTI, Inc., which is one of the world’s largest producers of P E T systems, described the development of LSO crystals for use in medical imaging. This was a particularly interesting and relevant description of what is required to produce a new crystal scintillator for use in tomography, since the same process applies for developing new crystals for calorimetry. The process begins with a detailed understanding of the basic scintillation properties of

249

250

the material, followed by thorough study of the parameters and conditions needed for growing the crystals. In the case of LSO, this required a significant investment of time, money and resources to set up a large production facility at CTI in Knoxville, which is now successfully producing crystals for PET scanners that are sold commercially. The message from this was that the same process invariably applies to the development of any new crystal, as has occurred for BGO, BaFz, CsI, CeF3, PbW04 and others which have been used in calorimetry. P. LeCoq described the current R&D efforts by CERN and the Crystal Clear Collaboration to develop new crystal scintillators for medical imaging. This study is presently focused on the development of LuAP (LuA103), which is a new bright scintillator based on lutetium that has a density of 8.3 g/cm3 and a decay time of 17 ns. A considerable amount of effort has gone into developing LuAP by CERN in cooperation with the Bogoroditsk chemical plant in Russia, which is now producing more than 100 tons of lead tungstate for the CMS experiment. There have now been many samples of LuAP produced which show excellent light output properties, and the results look encouraging for developing this new scintillator for use in medical imaging. Several other new scintillators were also described, including those based on hafnium and barium, as well as new possibilities for increasing the light output of PbW04 that would greatly improve its resolution at low energies. This latter possibility would be very attractive given the enormous effort that has gone into developing PbW04 for high energy physics applications, and could provide a very low cost material for medical imaging. J. Collot described a Monte Carlo program based on GEANT for simulating PET scanners. While there are other simulation programs available for this purpose, the use of the GEANT program is an excellent example of the use of a tool that was developed for simulating high energy detectors being used for medical applications. This required tuning various GEANT parameters for an accurate description of the physics processes at lower energies (down to a few hundred keV) and testing the results with experimental data. The program was used to help design a microPET scanner based on liquid xenon. In summary, there are many areas where the techniques used in calorimetry for high energy and nuclear physics are common to those used in a variety of medical applications. Several examples of this were given in this session which included the development of new scintillating crystals and the use of standard simulation programs to design detectors. It is hoped and expected that these developments of mutual benefit to both fields will continue in the future.

SYNERGIES BETWEEN ELECTROMAGNETIC CALORIMETRY AND PET*

W. W. MOSES Lawrence Berkeley National Laboratory Mailstop 55-121, 1 Cyclotron Road Berkeley, C A 94709, USA E-mail: wwmoses0lbl.gov

The instrumentation used for the nuclear medical imaging technique of Positron Emission Tomography (PET) shares many features with the instrumentation used for electromagnetic calorimetry. Both fields can certainly benefit from technical advances in many common areas, and this paper discusses both the commonalties and the differences between the instrumentation needs for the two fields. The overall aim is to identify where synergistic development opportunities exist. While such opportunities exist in inorganic scintillators, photodetectors, amplification and readout electronics, and high-speed computing, it is important to recognize that while the requirements of the two fields are similar, they are not identical, and so it is unlikely that advances specific to one field can be transferred without modification to the other.

1. Introduction

There are many similarities between the instrumentation for electromagnetic calorimetry and PET. Incident gamma rays are detected (and their energy and arrival time measured) with inorganic scintillators coupled to photodetectors. Thousands of these channels are read out in parallel using custom analog integrated circuits. However, each field has its own unique requirements, which often force different optimizations to be made. This paper explores the similarities and differences between the fields. It assumes that the reader is familiar with high energy physics and electromagnetic calorimetry, but relatively unfamiliar with PET.

* This work was supported in part by the U S . Department of Energy under contract No. DE-AC03-76SF00098, and in part by Public Health Service Grants Nos Pol-HL25840 and R01-CA67911.

25 1

252

2. The PET World Picture

PET is a nuclear medical imaging technique whereby the patient is injected with positron-labeled r a d i o p h a r m a ~ e u t i c a l ~This ~ ~ ~ drug ~ ~ ~ ~localizes ~. within the patient according to its physiologic properties and the radioisotope decays, emitting a positron. The positron annihilates with an electron (from the patient’s tissue) to form back-to-back 511 keV annihilation photons. As shown in Figure 1, a planar ring of 511 keV photon detectors surround the patient - if two photons are detected simultaneously (within 10 ns) then the radioisotope is assumed to lie somewhere along the line joining the two detector elements. The mathematical technique of computed tomography6, shown in Figure 2, is used to reconstruct the two-dimensional activity distribution within the plane defined by that detector ring - images from multiple detector rings are used to create a three-dimensional, volumetric image of the radiopharmaceutical distribution. PET’S strength is that it images the physiological properties of the pharmaceutical and so provides physiologic information, whereas many other medical imaging techniques provide mainly anatomical information7. For example, x-ray techniques image electron density and MRI (magnetic resonance imaging) mainly images water density. Therefore PET is predominantly used for diseases that are metabolically based (such as cancer or neurological disease). Despite the many similarities between PET and electromagnetic calorimetry, there are significant differences. One of the main constraints in PET is the low energy of the photons that must be measured with high accuracy - 511 keV is many orders of magnitude lower than the energies commonly measured with calorimeters, implying greater signal-to-noise challenges in PET. This also implies that there are no electromagnetic showers involved in PET - the only relevant interaction processes are photoelectric absorption and Compton scatter, and Compton scatter is the dominant process (at 511 keV, the photoelectric fraction is 45% in BGO scintillator and 0.02% for tissue). The high Compton fraction and relatively short interaction length (10 cm) in tissue also imply that in only a small percentage (
253

\ S cntir8br

Septum

Lead S h k U

Figure 1. PET Camera Schematic. Positron annihilations yield back t o back 511 keV photons, which are individually detected in a ring of photon detectors, shown on the top. Pairs are identified by time coincidence. Multiple rings are stacked up, as shown on the bottom, t o create a 3-dimensional image.

The detectors in most P E T cameras consist of block of BGO scintillator crystal that is partially sawn through to make a group of quasi-independent crystals that are optically coupled to four photomultiplier tubes, as shown in Figure 3. When a gamma ray interacts in the crystal, the resulting scintillation photons are emitted isotropically but the saw cuts limit (but do not entirely prevent) their lateral dispersion as they travel toward the photomultiplier tubes. The position of the gamma ray interaction within the scintillator block is then determined by the analog ratio of the photomultiplier tube output signals, and the gamma ray energy is determined and a timing pulse generated by the sum of these signals'. Based on current PET cameras, the gamma ray detector requirements for P E T are, in order of decreasing importance, (1)

254

li)in ensbnal Ho?&on@l Pmjx~n

Ve-1

Figure 2. Computed Tomography. T h e 1-dimensional horizontal and vertical projections (the line integral of the density along parallel horizontal and vertical lines) are shown adjacent to the 2-dimensional object that they were taken of. Computed tomography is the process of reconstructing the 2-dimensional object given its 1-dimensional projections from all angles.

>85% detection efficiency (to minimize statistical fluctuations), (2) <5 mm fwhm position resolution (to obtain good spatial resolution) , (3) <$100/cm2 parts cost (PET cameras are widely available commercially), (4) <1 p s cm2 dead time product (single event dead time times detector area affected - high counting rates are often encountered), (5) <2 ns timing resolution (to identify coincident pairs), and ( 6 )
3. Comparison of Calorimetry and PET 3.1. Scintillators

For both electromagnetic calorimetry and PET, the highest performance instruments use inorganic scintillator crystals coupled to photodetectors t o detect gamma rays with high efficiency, energy resolution, spatial resolution, and

255

Four 1 Square Photomultiplier Tubes

t

2"

+--

BGO Crystal Block, sawed into 64 segments, each 6 mm square

m Figure 3. PET Detector Module. Scintillation light from gamma ray interactions is detected by multiple photomultiplier tubes. The interaction position is determined by the ratio of the analog signals, and the energy by the analog sum of the signals.

timing resolution. The last decade has seen remarkably parallel developments of a new scintillator for electromagnetic calorimetry (PbW04)10i11 and a new scintillator for PET (LS0)l2. Both were discovered in approximately 1992, both are just now entering the full-scale production phase, and the development of both have required the efforts not only of the end users, but also of luminescence scientists, spectroscopists, defects scientists, materials scientists, and crystal growers. While an ideal scintillator would satisfy the requirements for both disciplines, the ideal scintillator has not yet been developed and compromises must be made. As Table 1 indicates, many properties of these two newly developed scintillators are very similar, such as the density, attenuation length, decay time, and emission wavelength. However, there are some important differences (which are highlighted in bold face type). The light output of LSO is much higher than PbW04 in order to (partially) compensate for the relatively low energy of the photons measured by PET. The cost per unit volume of PbW04 scintillator must be significantly lower than scintillators used for PET - even though the budget per channel is almost two orders of magnitude higher for calorimetry than it is for PET, the volume of scintillator required for PET is roughly three orders of magnitude higher. Finally, PbW04 must be very

256

radiation hard in order to withstand the background radiation present at high luminosity hadron colliders. Although LSO happens to be fairly radiation hard (10 Mrads), it could be several orders of magnitude less radiation hard without impairing its performance for PET. Table 1. Scintillator properties for PbW04 and LSO scintillator. PbW04

LSO

Density (g/cc)

8.3

7.4

Attenuation Length (cm)

0.9

1.2

Light Output (photons/MeV)

200

25,000

10

40

Emission Wavelength (nm)

420

420

Radiation Hardness (Mrad)

>10

10

Y , Nb

Ce

$1.6

>$25

Decay Time (ns)

Dopants Cost per cc

3.2. Avalanche Photodiodes

Avalanche photodiode (APD) arrays promise substantial improvements for PET’. They promise a pixel size that is much better matched to that of the scintillator crystals (3-5 mm) with minimal dead area between pixels, which leads to higher spatial resolution. APD arrays are not without challenges though. For PET, minimal dead area around the perimeter of the packaged device is essential, and large (or even moderate) scale manufacturing techniques have not yet matured to where the cost and reliability cease t o be issues. Finally, while their high quantum efficiency and gain promise to give acceptable system energy resolution and timing resolution, their signal t o noise ratio is likely t o be inferior to that of PMT-based systems. Many of these same features have made APD arrays attractive to the electromagnetic calorimeter community, which has worked extensively t o develop these devices. As with the scintillator material, the optimization of tradeoffs is different for high energy physics than it is for PET, and is summarized in Table 2. While both applications desire high gain and high quantum efficiency (especially at the shorter wavelength region of the visible spectrum), high energy physics experiments have a high flux of charged particles traversing the APDs. This requires that the APDs be radiation hard as well as have a relatively thin (typically 100 pm) depletion depth to minimize the nuclear counter

257

effect. APDs used for PET do not have these restrictions, but have tighter signal-to-noise requirements. In order to obtain low electronic noise at high frequencies (to obtain an accurate timing signal), the depletion region for PET APDs should be as large as possible in order to minimize the detector capacitance. Leakage current must also be minimized for both applications, but the requirements are more stringent for PET. Finally, the pixels in PET are roughly an order of magnitude smaller than they are for calorimetry. While neither application can tolerate significant gaps between pixels ( 2 . e., high packing fraction is needed for both), the smaller size of the pixels in PET implies that the size of the inactive area around the perimeter of the array must be proportionally smaller for PET. Table 2. Avalanche photodiode requirements for electromagnetic calorimetry and PET. ~~~

~

~

Calorimetry

PET

Yes

Yes

Yes

Yes

Radiation Hardness?

High

Low

Reduced Nuclear Counter Effect?

Yes

No

Timing Signal (low C)?

No

Yes

Small Dead Area at Perimeter?

No

Yes

-

Yes

High Gain? High QE

/

Blue Sensitivity?

Sensitive to Leakage Current?

3.3. Electronics The electronics for electromagnetic calorimetry and PET share many features. Large numbers of analog amplifiers with very low noise and low power consumption are required, and are usually implemented in mixed-mode (analog and digital) custom ASICs. Reasonably complex data correction is performed on-the-fly. The data acquisition systems support extremely high data rates, generally by using highly parallel architectures. However, there are also many differences in the electronics, and these are summarized in Table 3. As with the scintillator and APDs, radiation hardness is critical for calorimetry but irrelevant for PET. The analog electronics for calorimetry must accurately measure input signals that span a large dynamic range, but the dynamic range for PET is low, as the input signals consist mainly of monochromatic, 511 keV photons (the remainder are 511 keV photons that have undergone Compton scatter). The clocking / synchronization

258 Table 3. Electronics requirements for electromagnetic calorimetry and P E T . Calorimetry

PET

Low Noise Analog Amplifier?

Yes

Yes

Low Power Consumption?

Yes

Yes

Mixed-Mode Custom ASICs?

Yes

Yes

Real-Time Data Correction?

Yes

Yes

Highly Parallel readout?

Yes

Yes

High Data Rate?

Yes

Yes

Radiation Hardness?

Yes

No

Analog Dynamic Range?

High

Low

Self-Generated Timing Signal?

No

Yes

Asynchronous Inputs?

No

Yes

Event Size

/

Complexity?

High

Low

Multiple Trigger Levels?

No

Yes

“Good” Event Rate?

kHz

MHz

of the systems are also quite different. For high energy physics, events can only be generated when the accelerator bunches cross, and the accelerator provides the readout electronics a timing signal that is synchronized to this bunch crossing. Thus, the electronics for high energy physics are externally clocked with a constant clock period. With PET, events are produced when radioisotopes undergo radioactive decay, which is an excellent physical example of a random process. Finally, there are major differences in the event data that must be collected. With high energy physics, many different event topologies are possible, implying that complicated, multi-level event triggering schemes are required and that the size of the event (number of bits that are transferred) is highly variable. However, the overall event rate is relatively low. With PET, only a single event topology is possible (a pair of detected 511 keV gamma rays), leading to simple triggering schemes and a fixed event size. However, the overall event rate is comparatively high. 3.4. Computing

P E T and electromagnetic calorimetry share many computing requirements, which are summarized in Table 4. In both cases a significant amount of computation is needed, both before and after the apparatus is constructed. Extensive Monte Carlo simulations are performed at the design stage, modeling

259

both the interactions of gamma radiation in the detectors and the underlying physics that will be measured. Both are reasonably large software projects, necessitating many programmers working in a well-coordinated way. However, there are also many differences. The size of data set for a high energy physics experiment is enormous, possibly ranging from terabytes to petabytes. In contrast, the data set size for a single P E T experiment is measured in megabytes to gigabytes. The complexity of the analysis for high energy physics is quite high, with many different experimental signatures that must be recognized. Although the same data is mined many times, the analysis for each “experiment” (ie. an analysis of the data set resulting in the measurement of a single physics result, often synonymous with “Ph.D. thesis”) is different and custom code must be developed for each of these experiments. Although the images produced by P E T may be complex, the data from each “experiment” (in this case, one patient study) is reconstructed the same way using the same code. However, the time available to analyze the data from P E T is only a few minutes, as imaging centers need to verify that good quality images have been obtained before the patient can be released. High energy physics analyses, on the other hand, can often take years to complete. Finally, software bugs in P E T can lead to incorrect diagnosis with potentially fatal results! Because of this, a significantly higher level of scrutiny and approval by the Food and Drug Administration (FDA) is required for P E T software. Table 4. Computing requirements for electromagnetic calorimetry and PET.

I

I

Calorimetry

I

PET

Significant Computation?

Yes

Yes

Monte Carlo Simulation?

Yes

Yes

Large Programming Project?

Yes

Yes

TB-PB

MB-GB

Complexity of Analysis?

High

Low

Time to Finish Analysis?

Years

Minutes

No

Yes

Data Set Size?

FDA Certification Required?

4. Discussion

It is clear that P E T and electromagnetic calorimetry have much in common. At the most basic level, the requirements (such as highly efficient detection of gamma rays) and core technologies are identical. However, the previous

260

section makes it evident that the two disciplines have different requirements in a number of areas, some of which are contradictory. These differences imply that no matter how attractive it may be, it is naive to expect that technologies or devices developed and optimized for electromagnetic calorimetry can be transferred unchanged to P E T . Does this suggest that synergies between calorimetry and P E T are impossible? Far from it! It merely means that effort is needed to translate technologies optimized for one discipline to the other discipline. The high energy physics community has traditionally been at the forefront of developing advanced tools and technologies. The nuclear medical imaging community often adapts the tools and technologies most relevant t o its needs, optimizing them to satisfy the unique requirements and tradeoffs required for P E T imaging. This transfer is beneficial to both communities and should continue to be encouraged. As is common in such situations, the most effective element in nurturing this synergy is probably communication between the communities, most especially in understanding the requirements and tradeoffs involved in the application-specific optimization.

5. Conclusion

Electromagnetic calorimetry and P E T both rely on highly efficient detection of gamma rays, hence the technologies used by these two disciplines are extremely similar and many of the tools and technologies currently under development could be applied to either discipline. Examples of these tools and technologies include scintillators, photodetectors (especially APD arrays), electronics, and high-performance computation. The resulting devices are extremely similar the differences lie mostly in the details. However, the details are important and often preclude having a device that is developed and optimized for one application from being used without modification by the other. Nevertheless, strong synergies exist and should be encouraged, as the effort required t o make a modification is often less than that necessary for the initial creation.

Acknowledgments This work was supported in part by the Director, Office of Science, Office of Biological and Environmental Research, Medical Science Division of the U.S. Department of Energy under Contract No. DE-AC03-76SF00098, and in part by the National Institutes of Health, National Cancer Institute under grant No. R01-CA67911, and National Institutes of Health, National Heart, Lung, and Blood Institute under grant No. Pol-HL25840.

261

References 1. Sandler, M.P., Coleman, R.E., Wackers, F.J.T., et al., Diagnostic Nuclear Medicine, Baltimore, MD: Williams & Wilkins (1996). 2. Hendee, W.R. and Ritenour, R., Medical Imaging Physics, St. Louis, MO: Mosby Year Book (1992) 3. Macovski, A., Medical Imaging Systems, Englewood Cliffs, NJ: Prentice Hall (1983). 4. Webb, S., The Physics of Medical Imaging, Bristol: Institute of Physics Publishing (1993). 5. Cherry, S.R. and Phelps, M.E., Positron Emission Tomography: Methods and Instrumentation, In Diagnostic Nuclear Medicine, (Edited by Sandler, M.P., Coleman, R.E., Wackers, F.J.T., Patton, J.A., Gottschalk, A. and Hoffer, P.B.), Baltimore, MD: Williams & Wilkins, 139 (1996). 6. Cormack, A.M. J . Appl. Phys. 34, 2722-2727 (1963). 7. J. Nucl. Med. 32, 561-748 (1991). 8. Cherry, S.R., Tornai, M.P., Levin, C.S., et al. ZEEE 'Prans. Nucl. Sci. NS-42, 1064 (1995). 9. Moses, W.W., Derenzo, S.E. and Budinger, T.F. Nucl. Znstr. Meth. A-353, 189 (1994). 10. P. Lecoq and M. Korzhik, ZEEE Trans. Nucl. Sci. NS-47, 1311 (2000). 11. Qu, X., Zhang, L., Zhu, R.Y., et al. Nucl. Znstr. Meth. A-480, 470 (2002). 12. C.L. Melcher and J.S. Schweitzer, ZEEE Trans. Nucl. Sci. NS-39, 502 (1992).

NEW SCINTILLATING CRYSTALS FOR PET SCANNERS

PAUL LECOQ CERN, Geneva, Switzerland Email: Paul.LecoqOcern. ch

Systematic R&D on basic mechanism in inorganic scintillators, initiated by the Crystal Clear Collaboration at CERN 10 years ago, has contributed not to a small amount, to the development of new materials for a new generation of medical imaging devices with increased resolution and sensitivity. The first important requirement for a scintillator to be used in medical imaging devices is the stopping power for the given energy range of X and 7 rays to be considered, and more precisely the conversion efficiency. A high light yield is also mandatory to improve the energy resolution, which is essentially limited by the photostatistics and the electronic noise at these energies. A short scintillation decay time allows to reduce the dead time and therefore to increase the limiting counting rate. When all these requirements are fulfilled the sensitivity and image contrast are increased for a given patient dose, or the dose can be reduced. Examples of new materials under development by the Crystal Clear Collaboration will be given with an emphasis on the major breakthrough they can bring in medical imaging, as compared to present equipments.

1. Introduction Positron Emission Tomographs (PET scanners) are more and more recognized as very powerful tools for basic research in cognitive sciences, clinical oncology and kinetic pharmaceutical studies. Their working principle is based on the reconstruction of the product decay (2 y rays) of a /3+ labelled tracer injected to the patient. Detection of the two 511 KeV gamma rays produced in the electron-positron annihilation allows the in vivo reconstruction of the three dimensional distribution of the isotope in the body. The detection of the two y rays in cohincidence requires the use of scintillation detectors. Scintillators used in PET must be dense to optimise detection efficiency, fast to limit number of random coincidences, and have sufficient energy resolution to reject scattered coincidences. State-of-the-art commercial PET scanners are usually based on BGO detector blocks which have a good detection efficiency, but are quite slow (scintillation decay constant 300 ns). Consequently, these scanners operate at a sensitivity of about 1000 kcps/pCi/ml with a coincidence time window of about 10 ns and a scatter fraction above 30 to 45% . Next generation PET scanners

262

263

need faster scintillators as well as depth-of-interaction encoding, which can be provided by the combination of scintillators with similar density and light yield but different decay time or emission wavelength. Besides LSO which is now produced at an industrial scale, LuAP is a particularly interesting scintillator material for application in P E T . It is dense (8.34 g/cm3), fast (18 ns), and has potentially a high light yield. It is therefore a factor 2 faster than LSO with a density about 10% higher and is then a very attractive candidate for PET applications, to be used alone or in combination with LSO in a phoswich configuration for depth of interaction determination. For these reasons the Crystal Clear Collaboration' decided in the beginning of the year 2000 t o start an ambitious program of technology development of LuAP:Ce crystals in collaboration with the Bogoroditsk Technochemical Plant in Tula (Russia), with the strong support of CERN and ISTC (International Science and Technology Centre). Several prototype small animal P E T scanners using such crystals are presently under construction for evaluation by different medical and biologist groups. In our study we have also looked at the possibility to reduce the amount of Lutetium in an attempt to reduce the production cost of the crystal. One obvious candidate is Y3+, as the scintillation properties of YAP:Ce are well known. Much effort has been spent and is still on the way to define the best balance between Lutetium, Yttrium and Cerium for an optimization of the light yield, the photofraction and the growth condition of LuYAP crystals2i3. We have also analysed the role of a second cation in the complex oxide compounds doped with Ce. Anionic complexes on the base of two cations in the host lattice play a major role in the creation of excited matrix states and in the structure of its luminescent properties. We paid attention to the rare earth, Ba and Hf compounds which, being doped with Ce, show luminescence in the blue-green region. Both the effective charge increase of the host matrix and the decrease of the amount or even the complete suppression of the relatively expensive Lu ion in the crystal motivated us for this research. In our report we discuss the spectroscopic properties of several new heavy compounds such as, LazHf207 and BasLu409 doped with Ce. Another motivation of this study was to clarify the existence of the trivalent Ce interconfiguration 5d+4f luminescence in the complex structure oxide compounds where the second heavy cation of the matrix host has an electronic shell 5p. It is well known that room temperature high light yield of 5d-+4f luminescence of Ce3+ is observed in oxides where the second cation is B3+(ls2);Si4+,A13+(2p6);S~3+(3p6)2i4 but not for the case of heavier cations like Nb5+(4p6)or Ta5+(5p6)based compounds. In our case we could observe a strong Ce3+ luminescence in Hf4+(5p6) and Ba2+(5p6) oxyde based materials.

264

Another approach is aiming at the production of very cost effective scintillators for the construction of full body machines to be used for systematic clinical investigations. We expect here to benefit as a direct spin-off from the results of the large efforts which have been invested for High Energy Physics detectors in the last decade. We have demonstrated that a potential improvement of the light yield by a factor between 5 to 10 exists in this crysta14i5, which would make it suitable for large full body PET scanners to be mass produced at a moderate cost for systematic clinical investigations. In particular the possibility t o use crystals about 2 times shorter than BGO will maintain a very good spatial resolution over a large field of view (FOV), without having to integrate complex and expensive depth of interaction (DOI) systems. 2. LuAP developments

The initial work has concentrated on a systematic study of the perovskite phase stability in the phase diagram, and on an analysis of the thermal conductivity of the melt and of the perovskite and garnet phase. The main difficulty for the growth of large size and high quality crystals relies on the fact that we are in a situation of non congruent growth, where the perovskite phase competes with the garnet and monoclinic phases in a very narrow region of the phase diagram (Fig. 1). Not only a very high precision in the stoechiometry of the starting raw materials is needed, but a strong requirement on the precision of the heating system must result in a control of the temperature in any part of the melt in a range smaller than f3OC. Seeding the crystal is a particularly delicate part of the process as it locally introduces temperature variations which easily exceed the f3OC of the perovskite phase stability. On the other hand it was shown that addition of some quantity of Yttrium helps stabilizing the perovskite phase. This quantity must of course remain small enough to keep the photoelectric absorption coefficient high enough for an efficient detection of 511 KeV gamma rays. Our work with the Bogoroditsk Technochemical Plant has concentrated in particular on a very good control of the thermal leaks and a well designed geometry of the crucible in order to achieve this goal for Cerium doped Lu,Yl-,AP crystals where z is at least 70%. As a result of this, ingots of about 2.5 cm in diameter and 15 cm in length are now consistently grown with a very good phase stability and high transparency (Fig. 2). When cut in 1 cm long pixels such crystals have a photo-efficiency of 19%, to be compared t o 23% for LSO and 26% for LuAP (100% Lutetium) crystals of similar size. The combination in a phoswich configuration of one L u o . ~ o Y o . ~ o and A P one LSO pixel, each 1 cm long would give a photoefficiency of about 40%, which is an excellent result for a P E T scanner.

265

Figure 1. Phase diagram of the L u z 0 3 - A l z 0 3 system.

The light yield of pixels 2x2x10mm has been measured on both Lu0.70Y0.30AP and LSO crystals. The pixels were installed in a vertical position with Teflon wrapping on a PM XP2020Q with a silicon oil coupling agent. Another measurement with the crystals horizontally coupled to the photomultiplier window allowed us to measure the crystal self absorption at the emission wavelength of the crystal (365 nm for LuYAP and 420 nm for LSO). We found an average of (4374f400) ph/MeV in the vertical position and (8100f400) ph/MeV in the horizontal position, to be compared for LSO to 16920 ph/MeV and 22800 ph/MeV respectively. The ratio of the light yield measured in the vertical and the horizontal position is then 54% for LuAP and 60% for LSO. The slightly worse result for LuAP is related t o the fact that the emission peak is much closer to the fundamental absorption edge than for LSO. Presence of even small quantities of Ce4+ instead of Ce3+ as an activator produces a tail of the absorption edge which leads to a higher self absorption.

266

Figure 2.

Lu0.70Y0.30AP ingot

and pixels grown in Bogoroditsk.

This problem was well known in the early stage of development of the YAP crystals, for which technical solutions have been found meanwhile. It is interesting to mention that in spite of a light yield typically about 25% of the light yield from LSO, the LuAP crystals have an excellent energy resolution, as can be seen on Fig. 3 where the Lutetium escape peak is clearly resolved13. Indeed on a set of 30 crystals an average energy resolution FWHM of 7.7% (horizontal) and 14% (vertical) was measured. In similar conditions we found 8.6% and 15% for LSO. The reason for this excellent result is a very good linearity of the light response of LuAP crystals as a function of energy. As can be seen in Fig. 4 the light yield per unit of deposited energy is constant down to 150 KeV13. Considering that the energy resolution is the quadratic sum of an intrinsic term and a photostatistic term6, we find for the Lu0.70Y0.3OAP crystals an intrinsic resolution of 2.7%, to be compared to 1.3%for YAP, 4.1% for CsI(Tl), 5.7% for NaI(T1) and 7.6% for LSO. The pulse shape of the Lu0.70Y0.30APcrystals has been measured using the standard delayed photon counting method. A fast component of 23 ns has been measured, but contrary to LuAP (100% Lutetium) it represents only about half of the total light and a slow component of about 320 ns appears. LSO shows only one component of 40ns. This slow component is directly correlated to the amount of Yttrium in the crystal. A large effort is presently underway in

267

A.

d

L .u escape peak

Lu,,Y,$P horkontal Na-22 source

I1 energy resolubon

9 2% FWHM

1275 k s v p e a k from Ns-22

0

200

400

600

Eon

imo

1200

1400

energy [keV]

Figure 3.

Typical Lu0.70Y0.30AP light spectrum with a 22Na source.

120 110

m 100 r

0"

90

2

5

70

Q

z

00

order to progressively increase the Lutetium fraction up to at least 90% and to significantly reduce the amount of slow components.

268

3. Study of new scintillators Several polycrystalline compounds have been investigated. The most interesting are listed in Table 3. Undoped and Ce doped samples of the five listed materials have been obtained by solid phase synthesis from a stoechiometric mixture of oxides at 1400°C in air. One atomic percent of Ce dioxide has been blended to each compound and they have been annealed in a covered Pt crucible. They have then been annealed in two steps for 6 hours each, excluding LazHf207. The crystalline structure has been controlled by X-ray diffraction method. The fraction of the desired phase in the mixtures was 95, 70, 50, 50, 30% respectively. The other fraction was a composition of unreacted oxides. In spite of a relatively high crystallization temperature, at least two of these compounds, La2€If207and Ba3Lu40g, which exist in single crystalline form, may be competitive in density and effective charge to the well known Lu2SiOs:Ce and LuA103:Ce. Moreover, La2Hf207 does not contain the expensive lutetium whereas the Lutetium fraction in Ba3Lu409 is reasonably small. Table 1. Some physical properties of the studied compounds. Compound

LazHf2O.r

Structural Type

p

and space group

(g/cm3)

pyrochlor Ed3m

7.84

Z,ff

Melting point ("C)

64

2285

Yz03-HfO2

Fluorite Fm3m

6.8

62

2400

Lu203- HfO2

Fluorite Fm3m

8.8

69

2510

BaLaz04

Rhombic Pnam

6.34

54

1845

Ba3Lu409

Hexagonal R3m

8

65

2210

In addition, two tungstate compounds: Y2W3O12 and LuzW3012 have been studied. They have a rather low density of 4.85 and 5.34 g/cm3, however their Z,ff is 65 and 69 respectively. The Y based compound has been smelted at 1200°C and the Lu one has been prepared according to the above mentioned procedure. All undoped Hf and Ba based compounds did not show any luminescence under 122 keV X-ray (57C0, 15mCi) or UV excitation of Xenon or Hydrogen lamp at room temperature. The room temperature emission and excitation spectra of three of four compounds doped with Ce are shown in Fig. 5(a-c). La2Hf207:Ce shows an intense luminescence band peaked near 465 nm and excitation maxima near 268, 310 and 371 nm. Lu203-HfO2 (1:l) shows a luminescence band peaked at 480 nm and excitation maxima at 268, 310 and 365 nm. Doped BaLa204 did not show any luminescence at room temperature.

269

On the contrary, Ba3Lu409:Ce shows a strong luminescence band peaked at 510 nm with excitation maxima at 315, 369 and very weak at 260 nm. a,

16

d 14

a

250

350

450

550

650

550

650

h. nm 30 25

t-

=

20

0

15

0 250

350

450

h, nm

Figure 5 . Room temperature luminescence excitation (1) and luminescence (2) spectra of LazHfiO7:Ce (a), Lu203- HfO2:Ce (b), Ba3Lu40g:Ce (c). Luminescence band peak intensities (a&) decreased 5 and 10 times respectively.

All samples show a weak additional green luminescence band with maximum at 525 nm. This luminescence has an excitation maximum at 375 nm. The same luminescence and excitation have been observed for pure 4N cerium dioxide in powdered form. We do not exclude that the long wavelength excitation band in all spectra results from the superposition of the Ce3+ excitation luminescence in compounds with the one from unreacted cerium. The common feature of the excitation spectra is the presence of three bands, the

270

position of which vary from compound t o compound. These observed bands strongly suggest Ce3+ interconfiguration luminescence in oxide compounds7 with relatively high Stokes shifts. The observation of Ce3+ luminescence in hafnium compounds demonstrates that the desactivation of the excited Ce3+ ions through ionization’ is weak in this material. This mechanism causes on the other hand a strong quenching of the Ce3+ luminescence in tantalatesg for which scintillation has never been observed at room temperature. To clarify the scintillation properties of these compounds another synthesis at temperaturs close to melting point is required. Our study of tungstate compounds Y2W3012 and Lu2W3012 shows a wide and strong luminescence band in the green region under UV excitation. Luminescence and excitation spectra of undoped compounds are shown in Fig. 6. Doping with Ce does not change the luminescence properties of these compounds. This indicates that the lowest Ce3+ 5d level is in the conduction band. Thus a fast electron delocalization occurs under excitation and the Ce3+ luminescence is not observed. As for many tungstates, their spectroscopic properties are mainly caused by optical transitions in oxyanionic tyngstate complexes WOi-. Both compounds have a scintillation yield under y-ray excitation of about 200 ph/MeV, similar to the yield of lead tungstate crystals at room temperature.

200

300

400 h, nm

500

600

Figure 6. Luminescence excitation spectrum of Y2W3O12 (1) and luminescence spectra for 300 nm excitation of Y2W3O12 (2) and LuzW3012 (3) taken at room temperature. Peak intensity of Y2W3O12 is increased 2 times.

4.

L e a d Tungstate light yield improvement

It is generally recognized that the high effective Z (75.6), short decay time (15 ns) and low production cost of Lead Tungstate crystals could make this

27 1

crystal very attractive for low cost full body P E T scanners for cancer screening if its light yield could be increased at a level of 100 pe/MeV or more. It must be said that the present light yield of 15 to 20 pe/MeV(for small pieces) results from an optimisation of this crystal for high energy physics applications on high luminosity hadron colliders where the radiation hardness is more important than a high light yield. It is well known that the WOi- luminescence group is strongly quenched at room temperature in PWO by energy migration processes between regular groups. A decrease of the temperature gives a rise of the luminescence yield but at the expense of a dramatic increase of the decay time of the luminescence. A further improvement of the PWO crystal light yield is possible by introducing additional luminescence centres with specific properties into the crystal. As was determined, the energy of the first excited and radiating state of such centres must be less than 3 eV, which is lower than the energy of the lowest zero-phonon transition of the regular anionic oxi-complex (the estimated value is 3.35 eV). Moreover, a large e- capture cross-section by the high electronic energy levels of the doping ion is important. Because the concentration of the doping is usually small and direct excitation of the doping ion by irradiation is negligible, the increase of light yield is based on the redistribution of the non-radiative losses from the first excited state of the regular tungstate oxicomplexes. Figure 7 shows the integral light yield dependence of PWO crystals doped by different luminescent centres on the activator concentration at y-ray steady-state excitation. It was found that only Mo and T b satisfy the requirements and stimulate a visible rise of the scintillation yield. Recently it was confirmed that a combination of Mo, La, T b and Y5710,11~12 impurities increases the total scintillation yield in the crystal at room temperature. Scintillation in Mo-doped crystals is due to simultaneous contribution of the regular WOi- and doping MOO;- oxicomplexes, and the maximum of the resulting luminescence band is 500 nm. Mo impurity quenches fast scintillation of the regular tungstate group^^>^; however, it produces its own relatively fast scintillation owing to the radiating electronic 3T1 + I A l transition as well as slow components owing to thermo-activation of the electron centres created by Mo in lead tungstate. La doping in the crystal creates two shallow electron centres with thermo-activation energies 130 and 200 meV. Both centres are released through the conducting zone and have energy depths above the radiating level of the Mo oxicomplex. Thus, being co-doped with Mo they are an additional source of electrons to be captured by Mo oxycomplexes. For this reason the total light yield of Mo-La co-doped crystals is higher. The fraction of the slow component in scintillation strongly increases with the increase in Mo concentration for single doped crystals but this effect is much limited and

272

0

0

403

w

OM

om

0.1

0.u

ammiat a o n a m tat. XI

Figure 7. Light yield improvement of PWO crystals as a function of concentration of different doping ions ( l p s gate).

easily controlled in the case of co-doping with La. We have measured a light yield of 80pe/MeV in Mo-La doped PWO with an integration gate of 1 ps and 213 of the light collected in 100 ns. A superimposed luminescence of regular tungstate groups and Tb3+ ions is observed in Tb-doped samples. At a Tb concentration around 100 ppm, a wide luminescence band with a maximum at 420 nm overlapped with narrow Tb3+ emission bands with maxima at 348,382,410,440,490,540 and 587 nm. Besides the two regular radiating states 5D3, 5D4 of Tb3+ ions which have an energy less than 3.35 eV, upper f-levels also contribute to the luminescence at 348 nm at low activator concentration. The luminescence from the 5D3 level is strongly quenched by interaction with regular oxicomplexes so that it provides only fast scintillation components. Such additional luminescence almost doubles the scintillation light yield. Scintillation is extremely fast in Tb-doped crystals at small activator concentration. 5.

Conclusions

The large effort from the Crystal Clear Collaboration for the development of LuAP has led to the stabilization of a production line of Luo.~oYo.~oAP crystals to be used in combination with LSO crystals in several small animal PET scanners presently under construction by the collaboration. R&D is continuing to increase the Lutetium fraction up to at least 90%. The investigation of new oxycomplexes on the base of two heavy cations in the host lattice shows promising results for materials with 5p6 outer shell

273

cations like Ba and Hf which have a bright and fast luminescence in the green region. Finally a significant light yield improvement has been obtained in Lead Tungstate crystals for different doping conditions. Particularly in the case of Mo-La codoping a light yield of 80 pe/MeV has been measured, which is close to the limit where this crystal could be seriously considered for large, low cost P E T scanners for systematic cancer screening.

Acknowledgement The author wants to associate M. Korzhik from INP Minsk who took a very active part in all the work mentioned in this paper. He is also very thankful to all his colleagues from the Crystal Clear Collaboration and other colleagues who spent much time in developing these crystals. He is also extremely grateful to CERN and the International Science and Technology Centre, Moscow, Russia, for their financial support of PWO and LuAP research and technology development within Projects 354, 1489 and 2039.

References 1. Study of new fast and radiation hard scintillators f o r calorimetry at L H C , Crystal Clear Collaboration, RD 18, CERN/DRDC/P27/91-15. 2. A.G. Petrosyan et al, Journal of Crystal Growth 198/199 (1999) 492-496. 3. J. Chval et al, NIM A 443 (2000) 331-341. 4. A. Annenkov, A. Borisevich, A. Hofstaetter, M. Korzhik, V. Ligun, P. Lecoq, 0. Missevitch, R. Novotny, J.P. Peigneux, NIM A 450 (2000) 71. 5. P. Lecoq, M. Korzhik, IEEE Trans. Nucl. Sci. 47 (2000) 1311. 6. P. Dorenbos et al., IEEE Trans. Nucl. Sci. Vol 42, N6, p2190-2202, Dec 1995. 7. P.Dorenbos, T h e 4f-5d transitions of the3 trivalent lanthanides in halogenides and chalcogenides, Journ. of Luminescence 91 (2000) 155. 8. William M.Yen, Photoconductivity and delocalization in rare earth activated insulators, Journ.of Luminescence 83-84 (1999) 399. 9. L.I.Kazakova, A.B.Dubovski, G.V.Semenkovich, O.A.Ivanova, Luminescence of Y T a O 4 single crystals, Radiation measurements, 24 (1995) 359. 10. M.Nikl, P.Bohacek, A.Vedda, M.Martini, G.P.Pazzi, P.Fabeni, M.Kobayashi, E f i c i e n t m e d i u m speed Pb WO4:Mo, Y scintillator (to be published in phys. status solidi). 11. M.Kobayashi et al., Proc. SCINT2001, Chamonix, France, sept. 2001 M.Kobayashi et al., NIMA 434 (1999) 412-423. 12. R.Y. Zhu et al., in “Proceedings of the 9th International Conference on Calorimetry in High Energy Physics”, ed. B. Aubert et al., Franscati Physics Series, Vol XXI (2000) 709 R.Y. Zhu et al., Proceedings IEEE NSS/MIC 2000, Lyon. 13. C. Kuntner, P. Lecoq, to be published in NIM.

A SIMULATION FRAMEWORK FOR POSITRON EMISSION TOMOGRAPHY BASED ON GEANT4

J. COLLOT, S. JAN, M.L. GALLIN-MARTEL, P. MARTIN, E. TOURNEFIER ISN Grenoble, 53 avenue des Martyrs, 38026 Grenoble Cedex , France E-mail: collotOin2p3. fr, sjanOin2p3.fr

GePEToS is a simulation framework which we have developed over the last few years for assessing the instrumental performance of P E T scanners under development. It is based on Geant4, written in 00 C++ and runs on Linux platforms. The validity of GePEToS was tested on the well-known Siemens ECAT EXACT HR+ camera. We also present the results of two application examples : the configuration optimization of a liquid Xe pPET camera dedicated to small animal imaging ; the evaluation of the effect of a strong axial magnetic field on the image resolution of a Concorde P4 pPET camera.

1. Introduction

Over the last decade, the performance of Positron Emission Tomography (PET) scanners has considerably improved. For instance, commercial cameras dedicated to small animals now feature a space resolution below 2 mm as long as a sensitivity which exceeds 1%l. No matter how beneficial this performance gain has been to users, for instrument designers, it has somehow hardened the challenge of finding new solutions which would go beyond the present instrumental limits at an affordable cost. This is why the complete simulation of new T E P configurations under study has now become even more justified than in the past, and calls for the development of a dedicated simulation framework, sufficiently versatile to allow fast and very detailed approaches with the best-existing emulation of all the underlying physical processes. Since its first public release in 1998, the stability, the validity and hence the popularity of Geant42- the Object-Oriented particle tracking and transport simulation framework developed by the High Energy Physics community - have noticeably progressed. In our opinion, it has become the best tool box from which any common and public TEP simulation framework should be developed for the next decade. GePEToS : Geant4 Positron Emission Tomography Simulation, is a first attempt that we have made over the last few years to meet this goal .

274

275

2. Adequacy for TEP scanners of GeantCsimulated physical

processes As we aimed at providing the possibility t o totally simulate all the processes which take place in the short life of a positron and then in the transport and interaction of its two 511 keV annihilation photons, we have made some effort to check the validity of Geant4 and, when the need arose, t o correct or complete it. Hence, we have added as part of GePEToS the possibility t o generate the positrons according to their respective ,B+ spectra (for 18F, 1 5 0 , "C). The positrons are fully tracked down until they annihilate. As Geant4 did not correctly reproduced the accolinearity of the two annihilation photons ((OYY - 180') = 0.5 degrees FWHM) which affects the image resolution, we have modified the native Geant4 algorithm to obtain a correct annihilation behavior in water where this phenomenon takes place in PET. Finally, we have tested two Geant4 electromagnetic (EM) process packages : the standard one, and a second one which provides a better description of the EM physical processes at low energy. We have achieved some comparisons of the total attenuation coefficients as obtained from Geant4 for photons, either using the standard EM processes or the low energy ones, to NIST experimental data3. As can be seen on figure 1, the low energy processes provide results which come closer t o the experimental data. The difference is explained by the absence of Rayleigh scattering in the standard EM package. Indeed, these are now used in GePEToS even though they slightly increase the code CPU consumption. 3. Framework description

GePEToS as Geant4 , is fully written in Object-Oriented C++. It runs on Linux platforms (tested on RedHat 6.2). It utilizes a simple mechanism to define the geometry and the material composition of most of the P E T cameras presently commercialized or under development (multi-ring and multi-crystal block devices). This is achieved by preparing an ASCII configuration file in which the users can select the desired ,B+ isotope, the number of active rings, the dimensions and the segmentation of the crystal blocks, the nature of the crystal (NaI, LSO, BGO), the phantom type, the acquisition mode (2D or 3D). In addition, users have t o provide the energy resolution measured or estimated at 511 keV which is then normally scaled according to the energy deposited in the crystals. For standard configurations, neither any code modifications nor any recompilation are necessary. For every event (positron) and each of its two annihilation photons,

276

I

-Bra

.E

E

5

0 .

:

:.

LII

-5

1

.

.

../

. .: . . . . . . . .. .. . ... .. . . . . . . . : . . . . . . . . . . . . .. . . .: :. i

.

I

10 keV

1

6

i

i

.

.

1 1 1 1 1 1

1

I

I

I

loz

Figure 1. Deviation of photon total attenuation coefficients as computed by Geant4 from NIST data for two Geant4 EM process packages

GePEToS computes the deposited energy and an energy-weighted position in the transverse plane of the crystals which is then used to determine which crystal was hit. The depth of interaction (DOI) in crystals is also computed and stored if this readout mode is selected by users. The hits information is written in ROOT4 files. The sinograms are separately prepared by using a ROOT application and finally processed by IDL5 to reconstruct the tomographic images. More complex geometries, which depart from the standard multi-ring crystal block model can be handled with a bit more effort by modifying the code and rebuilding the application.

4. Validation test

We have achieved an exhaustive validation test of GePEToS on one of the most common P E T cameras used in medical examination centers : the Siemens ECAT EXACT HR+ PET scanner6. The ECAT EXACT HR+ P E T scanner consists of 32 rings, featuring an internal diameter of 82.7 cm and spanning 15.2 cm in the axial field of view. It is made of blocks of BGO crystals. Each crystal has a transverse cross-section of 4 x 4.1 mm2 and is 30 mm long. This device, as modeled in GePEToS, is presented in figure 2, prepared for a 2D acquisition, for which the lead septa have been slid in front of the crystal rings. Also shown on the picture is

277

one of the typical water phantoms (@=20 cm, L=20 cm) that can be used in GePEToS to assess the performance of the cameras.

Figure 2. Graphical view of the ECAT EXACT HR+ PET camera as described in GePEToS in 2D acquisition mode, with the lead septa slid in front of the crystal rings - The rays exiting the water phantoms represent a few simulated annihilation photons

All the comparisons of the simulated performance, that we have carried out against the available experimental data7 have shown an excellent agreement. To illustrate this statement, we show in figure 3 our results for the fraction of scattered coincidences as measured in the NEMA experimental protocol in 2D and 3D acquisition modes with 18F. Also presented in figure 4 is the radial resolution determined with IsF-loaded capillaries which again shows a nice agreement between the simulation and the experimental data.

5. Application examples In this section, we briefly describe two application examples of GePEToS which show that this simulation framework although it is still in its infant stage, can be used to investigate a wide variety of PET problems . 5.1. A liquid xenon @ P E T camera

For several years, liquid xenon has been considered by two groups to build PET cameras8. We have used GePEToS to optimize the configuration of a small animal pPET camera which would exclusively use the scintillation light of LXe.

278

6 F '

280

"

'

loo

160

'

UD

'

"

Unln-bV

Figure 3. Simulated fraction of scattered coincidences compared to experimental data obtained following the NEMA protocol in 2D and 3D acquisition modes

0

2

4

6

8

SO

12

r-mm

14

16

18

a0

2

Figure 4. Radial resolution of the ECAT EXACT HR+ P E T camera obtained by GePEToS and compared to experimental data

The active part of the camera is a ring which features an internal diameter of 10 cm and has a radial extension of approximately 25 mm. It is filled with liquid xenon and placed in a cryostat which is composed of thin aluminum walls (especially around the imaging port). 16 identical modules of the same type as the one which is shown in figure 5, are immersed in this ring. Each module has a cross-section in the trans-axial plane of the camera of 2 x 2 cm2. The axial field of view spans 5 cm. A module is optically subdivided by 100 2 x 2 mm2 MgF2-coated aluminum UV light guides. The UV light is collected on both sides of a module by two position sensitive photo-tubes. The positions measured by the photo-tubes determine which light guides have been fired : we then finely measure the trans-axial Depth Of Interaction (DOI) of the photons. For each module, the ratio L-R/L+R of the photo-tube signals provides the axial coordinate. The performance of this device has been totally evaluated using GePEToS

279

Light guide

,

Liquid xenon module

1

Figure 5 . Sketch of an elementary module of the LXe pPET camera : the z-axis is along the axial direction of the wPET

plus a light collection program written by us in C++. It assumes a quantum efficiency of the photo-tubes of 15% and a UV reflection coefficient of 90% for the light guides. The simulated 18F sensitivity of this device evaluated on a water cylinder of 4 cm in diameter and 4 cm in length is 0.6% for an energy threshold of 250 keV. Its image resolution after filtering is 1.6 mm (FWHM) throughout the view field , thanks to the DO1 capability of the device. We show on figure 6 the reconstructed image of point-like 18F sources placed in the z=O trans-axial plane of a 4 cm diameter water cylinder. After filtering these are clearly resolved.

Figure 6. Reconstructed images of point-like "F sources placed in the z=O trans-axial plane of a 4 cm diameter water cylinder . Left : unfiltered ; Right : filtered

280

5.2. Eflect of an axial magnetic field on image resolution

As Geant4 features the capability to transport and track charged particles in strong magnetic fields, we have used GePEToS to evaluate the image resolution gain of a P4 Concorde pPET camera1 which would be operated in the strong axial field of a MRI scanner. We found no improvement for 18F and a marginal gain for l'C. However for 150and provided the device is operated in a 15 T solenoidal field, figure 7 shows that the field helps to resolve two point-like sources separated by 4 mm. The physical explanation of this effect is very comparable to what happens in a TPC operated in a magnetic field. The axial magnetic field blocks the trans-axial diffusion of electrons and confines them within a spiral around their creation points. Our results are in good agreement with what was found in previous studiesg. 15 T MRI scanners are now becoming available for small animals, but the operation of photo-sensors in such a strong magnetic field remains a very difficult challenge for the future.

Figure 7. Reconstructed image of two point-like I5O sources placed in the z=O trans-axial plane of a 4 cm diameter water cylinder imaged by a Concorde P4 pPET. The sources are 4 mm apart .

281

6. Conclusion and perspectives

We have established the basis of a PET simulation framework, called GePEToS, which utilizes Geant4 as a transport and tracking engine. GePEToS has been validated against the available data of a Siemens ECAT EXACT HR+ PET scanner. We have used GePEToS to guide our development effort toward a LXe PET camera dedicated to the imaging of small animals. We would now like to start preparing the public distribution of this software : documentation , uses examples. For the future, we would also like to add the time dimension (event time and detector time response) to GePEToS.

Acknowledgments This work was made possible thanks to the financial grants allocated by the Rh6ne-Alpes region through its ”Emergence” science program and by CNRS/INSERM via its IPA program dedicated to the imaging of small animals. We are also deeply indebted to Jean-Franqois Le Bas and Daniel Fagret of the Medical department of the Joseph Fourier University of Grenoble for the support and motivation they brought to this project.

References 1. Chatziioannou A.F.et al., Performance evaluation of microPET : a high-resolution oxyorthosilicate P E T scanner for animal imaging:, Journal of Nuclear Medicine, 40 (1999)p. 1164 2. Geant4 web page : http://wwwinfo.cern.ch/asd/geant4/geant4.html 3. NIST web page : http://physics.nist.gov/PhysRefData/ 4. ROOT web page : http://root.cern.ch 5. IDL web page : http://www.rsinc.com/idl/index.asp 6. Brix G. et a1 , Performance Evaluation of a whole-body P E T scanner using the NEMA protocol, Journal of Nuclear Medicine, 38 (1997) 7. Private communication of Service Hospitalier Frederic Joliot - CEA - DSV - Orsay France 8. V. Chepel et al. , ”performance study of liquid xenon detector for PET” Nucl. Instr. and Meth. A392 (1997) 427 J. Collot, S. Jan and E. Tournefier, ” A liquid xenon P E T camera for neuroscience”, IX Int. Conf. On Calorimetry in Part. Phys. - Annecy 2000 - F’rascati Physics Series Vol. XXI (2001)305 9. R. Raylnan et al. , IEEE Transactions on Nuclear Science, 43 (1996) 2406 B.Hammer et al. , IEEE Transactions on Nuclear Science, 42 (1995) 1371

This page intentionally left blank

Silicon Calorimetry Covener: D. Strorn

D. Strom

Covener’s Report

G. Abbiendi

Performance of the OPAL Si-W Luminometer at LEP 1-11

P. Goettlicher

The ZEUS Hadron Electron Separator, Performance and Experience

R. Fkey

Design Considerations for a Silicon/Tungsten Electromagnetic Calorimeter for a Linear Collider Detector

H. Videau

A Si-W Calorimeter for Linear Collider Physics

283

This page intentionally left blank

SILICON CALORIMETRY

D. STROM Dept. of Physics, University of Oregon, Eugene, OR, 97403-1274, USA E-mail: [email protected] (Convener's Report)

Electromagnetic calorimeters based on silicon sensors are under consideration for detectors at future linear colliders. This session reviews some existing special purpose devices as well as plans for calorimeters at linear colliders.

1. Introduction

Silicon is an ideal medium for highly granular calorimeters because it can be very finely segmented. Given the importance of this segmentation for the energy flow algorithms at future linear colliders (see below and elsewhere in these proceedings) these detectors have been proposed for use as electromagnetic calorimeters at these facilities'. We heard reports from both the European2 and North American3 groups on their plans for silicon tungsten calorimeters. The first practical devices to be used in large experiments were the luminosity monitors from SLD4, ALEPH5 and OPAL6. These groups made use of both the fine granularity and the precision geometry (micron level) of these detectors. These devices were successfully operated for many years in the high background small angle region of the LEP and SLC colliders. We had a contribution from OPAL7 on one of their detector. Another interesting use of silicon is the Zeus Hadron Electron Separator (HES)9 which consists of one layer of silicon located inside the Zeus Uranium-Scintillator calorimeter. The experience gained by the ZEUS HES group indicates that thermal management and cooling are likely t o be an important issues for linear collider detectors. 2. Silicon Detectors and Energy Flow

At LEP it was found that jet energy and direction could be most precisely measured if charged particles were measured using the tracking detectors. Neutral hadrons and gammas are then measured using electromagnetic and hadronic calorimeters. If the granularity of the calorimeters is not sufficiently fine, clus-

285

286

ters from charged and neutral particles will overlap. The correct energy can be recovered, but only after subtracting the expected deposition of energy from the charged hadrons. This procedure degrades the energy resolution because of the large fluctuations in hadronic energy deposition. As a silicon-tungsten calorimeter provides the highest granularity possible, it is a natural candidate for calorimetry at linear colliders. This advantage must be weighed against its modest energy resolution for electromagnetic showers ( > 15%/@). 3. A Look Ahead

F'uture work on silicon calorimetry is likely to concentrate on two issues. First, studies of processes such as Higgs boson production in WW fusion (e+e- -+ vPho), where kinematic constraints can not be applied, will tell us how important precise jet reconstruction is to the physics program of linear colliders. Second, much detailed design work needs to be done to demonstrate that compact and hermetic silicon sampling calorimeters can be constructed within reasonable financial and thermal constraints.

Acknowledgments This work was supported by DOE grant DE-FG03-96ER40969.

References 1. J. E. Brau, A. A. Arodzero and D. M. Strom, Published i n New Directions for High-energy Physics: Proceedings. Edited b y D. G. Cassel, L. "kindle Gennari, R.H. Siemann. Stanford, C A , Stanford Linear Accelerator Center, 1997. pp. 437-

444

2. 3. 4.

5. 6. 7. 8. 9.

T. . Behnke, S. . Bertolucci, R. D. Heuer and R. . Settles, detector for TESLA," DESY-01-011.; T. Abe et al., in Proc. of the APS/DPF/DPB Summer Study on the Future of Particle Physics (Snowmass 2001) ed. R. Davidson and C. Quigg, SLAC-R-570 Resource book for Snowmass 2001, 30 Jun - 21 Jul 2001, Snowmass, Colorado. H. Videau, these proceedings. R. Frey, these proceedings. S. Berridge et al., IEEE "kans. Nucl. Sci. 37,1191 (1990). D. Bederede et al., Nucl. Instrum. Meth. A 365,117 (1995). G. Abbiendi et al. [OPAL Collaboration], Eur. Phys. J. C14, 373 (2000). G. Abbiendi, these proceedings. G. Lindstrom et al., DESY-90-109. P. Gottlicher, these proceedings.

PERFORMANCE OF THE OPAL SI-W LUMINOMETER AT LEP 1-11

G. ABBIENDI Dipartimento di Fisica dell'Universitd di Bologna and INFN V i d e Berti Pichat 6/2, 40127 Bologna, Italy E-mail: Giovanni.A b biendi Obo.infn.it

R. G. KELLOGG Department of Physics, University of Maryland, College Park, M D 20742, USA E-mail: Richard.KelloggOcern.ch

D. STROM University of Oregon, Department of Physics, Eugene O R 97403, USA E-mail: stromObovine.uoregon.edu

A pair of compact Silicon-Tungsten calorimeters was operated in the OPAL experiment at LEP to measure the integrated luminosity from detection of Bhabha ek scattered at small angles from the beam line. The performance of the detector at both LEP-I and LEP-I1 is reviewed.

1. Introduction

The LEP e+e- collider at CERN operated for more than a decade: in 198995 at center of mass energies close to the Z peak (LEP-I); in 1996-2000 at higher energies, up to 209 GeV (LEP-11). In the first phase a large number of Zo events were collected, of the order of 5 x lo6 events per experiment. To match the inherent precision of this data sample, the error on the integrated luminosity had t o be better than lop3. At LEP the relevant process for the luminosity measurement is Bhabha scattering at small angle, which delivers a counting rate higher than the Zo event rate at resonance. The Bhabha angular spectrum falls like 1/03, implying a high sensitivity to the definition of the minimum polar angle of the acceptance. For example an uncertainty 60 = 10 prad (which in our configuration is equivalent to 25 pm in radius at the face of the detector) would give an unacceptable systematic error of lop3. Precision luminosity measurement was thus a demanding task, dictated by interest in measuring absolute cross sections at the 2' peak. In particular,

287

288

cross sections were needed to determine the Invisible Ratio Rinv = l ? i f l u / F l l , the ratio of the Z decay width to invisible particles and to charged lepton pairs. From Rinv the LEP experiments determined the number of light neutrinos to be 3 and limited possible contributions from extra new physics like cold dark matter. a 2. Detector

The OPAL Si-W luminometer consists of 2 identical cylindrical calorimeters, encircling the beam pipe simmetrically at f 2 . 5 m from the interaction point. A detailed description can be found in the OPAL paper'. Each calorimeter is a stack of 19 silicon layers interleaved with 18 tungsten plates, with a sensitive depth of 14 cm, representing 22 X O . The first 14 tungsten plates are each 1 X O thick, while the last 4 are each 2 X O thick. The sensitive area fully covers radii between 6.2 and 14.2 cm from the beam axis, giving a total area of silicon of 1.0 m2 per calorimeter. Each silicon layer is divided into 16 overlapping wedges. Even and odd layers are staggered by an azimuthal rotation of half a wedge. Water cooling pipes run as close as possible to the readout chips to remove the 340 W dissipated in each calorimeter. The distribution of material upstream of the calorimeters is kept at a minimum especially in the crucial region of the inner acceptance cut where it amounts to 0.25 X O . In the middle of the acceptance this material increases to about 2 X O due to cables and support structures of the beam pipe. The effects of the degraded energy resolution are important and are corrected for. Each detector wedge is a thick-film ceramic hybrid carrying a 64-pad silicon wafer diode plus the readout electronics. The pad layout of the silicon diodes is shown in Figure 1. The pads are arranged in a R - 4 geometry, with a radial pitch of 2.5 mm. Readout is done with 4 DC-coupled AMPLEX chips (each one reading 16 channels in a given 4 column). The diodes have an average depletion voltage of 62 V and are operated at 80 V bias voltage. The complete luminometer has in total 608 wedges with a total of 38,912 readout channels. For a typical LEP-I Bhabha electron with E, = 45 GeV the charge deposited on a single detector layer at shower maximum is 300 - 400 mips ( M 1.0 - 1.3 pC) which is typically spread over a few pads. The AMPLEX chip has a full scale limit of more than 1000 mips for each pad, thus providing a sufficient dynamic range. The equivalent noise for each channel remained at a level of 1500 to 2000 electrons for a typical detector capacitance of 20 pF, giving better than 10 : 1 signal to noise for mips. aOPAL results are N , = 2.984 & 0.013 and:::r

< 3.7 MeV at 95% confidence level.

289

Figure 1. Layout of the pad geometry of one wedge.

The calibration has been studied with electrical pulses generated both on the AMPLEX chips themselves and on the hybrids, as well as with ionization signals generated in the Si using test beams and laboratory sources. The overall channel-to-channel uniformity in gain was 1 % but gain variations among the 16 channels of each AMPLEX were 5 0.25%. This allowed optimum resolution for trigger thresholds and eliminated the need for a database of calibration constants for off-line energy reconstruction. Cross talk among channels in each AMPLEX was at the level of 2%/channel (coherent 30 %/AMPLEX) and was subtracted. Any residual gain variations depending on the channel position within each AMPLEX were cancelled by inverting the channel radial ordering between the two 4 columns of each wedge. The calorimeters were exposed to substantial radiation from occasional catastrophic beam losses. To limit this damage a protection system monitored the bias currents and induced a fast beam dump if the absorbed energy was greater than 3 x lo8 GeV within 1s. The leakage current at 22°C was uniformly 1 nA/cm2 when the detector was installed in 1993. Radiation damage during eight years of operation at LEP increased it to 12 nA/cm2 on average, although at shower maximum the typical values are 5 times higher (the AMPLEX bias current limit is x 200 nA/pad). From such increase of the leakage current we have estimated an effective absorbed dose of about 4 Krad, or a total absorbed energy of x 5 x 10l2 GeV, using measurements from J. Lauber et a1.2. At the end of LEP running only 0.6% of the Si-W Luminometer was not functional.

290

3. Lateral shower profile

The lateral profile of electromagnetic showers in the dense medium of the SiW calorimeters is characterized by a sharp central peak (FWHM < 1 pad = 2.5 mm) and broad tails extending to almost 10 pads, as shown in Fig. 2.

im 50

0

Pad Nurnbar

Pad Nvrnber

Figure 2. Average radial shower profile at 6 XO for E = 45.5 GeV electrons in linear (left) and logarithmic (right) scale.

Peak finding is based on the second spatial derivative of the pad charge, so that a sufficiently pronounced shoulder can be identified as a secondary cluster. Radiative Bhabha events with one or more photons contained within the acceptance can produce such configurations. The two cluster resolution efficiency has been determined from such radiative data events with a well separated secondary cluster with E > 5 GeV. The pad signals belonging to the secondary cluster are rotated about the beam axis until they have the same azimuth as the primary cluster and added to the signals actually observed on the local pads. The standard reconstruction is then applied and the separation efficiency as a function of the radial distance between the two clusters is obtained, as shown in Fig. 3. It is greater than 50% for cluster separation greater than 1.0 cm, equivalent to 4 pad widths. The overall inefficiency of primary cluster finding is less than OPAL

p , ...............

f

r

+.......................................................

0.4

Fadialduster separalian (mm)

Figure 3. Efficiency of reconstructing a secondary cluster as a function of the radial separation with the primary one when they have equal azimuth.

291

4. Position measurement The detector segmentation is very different in R and 4, owing to their different impact on the luminosity measurement. Here we will be interested only in the precise radial position measurement. The radial coordinate is first determined in each layer by interpolating a coordinate within the pad displaying the maximum signal in that layer. Then all the good layer coordinates from 2 XOto 10 X Oin depth are projected onto a reference layer chosen at 7 X o , and averaged there. The reference layer lies near the average shower maximum to minimize systematic effects. The resolution of the layer coordinate varies strongly across a pad, from about 300 pm at pad boundaries to 750 pm at pad center. This variation is reflected even in the average R coordinate, where a periodical structure following the radial pitch is apparent. To remove such oscillation, as the last step, a smoothing algorithm is applied, subjected to boundary conditions at the pad boundaries. A key issue for the luminosity measurement is knowledge of the absolute radial dimensions of the calorimeters. Very accurate positioning and monitoring of detector wedges in each layer using microscopes and micro-manipulators have achieved an RMS scatter of 1.3 pm of the radius of each wedge with respect to the best-fit circle of each half-layer. Taking into account deviations of each half-layer with respect to its ideal position in the calorimeter stack, mechanical deformations, temperature effects and measurement errors, the final precision on the absolute average radius is 4.4 pm. The final position resolution of the average smoothed radial coordinate has been determined to be 130 pm at pad boundaries and 170 pm at pad centers, from test beam measurements. The test beam used a 45 GeV electron beam alternated with a 100 GeV muon beam. Alignment of the calorimeter with respect to a high resolution Si-strip beam telescope was carried out with the muon beam. Sensitivity of the Si-W electronics to mips was essential for this purpose. The effect of upstream material was studied using a 0.84 X O plate which could be inserted in front of the detector. The reconstruction method respects the symmetry condition that a shower which deposits equal energies on two adjacent pads in the reference layer at 7 X O has to be reconstructed in the mean exactly at the boundary between the pads. In reality due to the R - 4 geometry of the pads, the true position of such showers is at a smaller radius than the pad boundary. This is the so called p a d boundary baas, which depends on the lateral shower spread and has been measured in the test beam. As the radial position of the incoming particles is scanned across a radial pad boundary in a single layer, the probability for observing the largest pad signal above or below this boundary

292 OPAL

.............................

Figure 4.

0

50

m

O-250

-200 -150

-1M)

-50

O-250

-200 -150

-100

50 0 50 1w 150 200 150 Dilance m i m l Imm pad boundav (microns)

100

150

250

Pad boundary images for test beam electrons and muons at 7 X O .

shifts rapidly, giving an image of the pad boundary as shown in Fig 4. The pad boundary images are modelled with an error function, where the gaussian width w is called the pad boundary transition width and R , f f is the radial offset between the apparent and the nominal pad boundary. The difference in R , f f obtained by changing from electron to muon beam is the measured pad boundary bias, which is shown in Fig. 5 as a function of w. The reconstructed radial coordinate is sensitive to the distribution and type of material in front of the detector as well as to the incidence angle of the particles. The test beam configuration could not reproduce the exact features of the OPAL running, so an indirect approach has been followed, called anchoring. Details of the method are fully explained in the cited paper'. The procedure is applied separately on individual data samples, each one characterized by different beam parameters, and obtains net systematic corrections on the radius of the acceptance cuts. The inner acceptance cut is corrected by 5-10 pm with an uncertainty of 3.5 pm, while the outer acceptance cut is corrected by 10-20 pm with an uncertainty of 6 pm. These radial corrections are then easily turned into acceptance corrections which are applied to data. We have also studied the energy dependence of the pad boundary transition width using data from OPAL running, as there was no test beam data at LEPI1 energies. In Fig. 6 w is plotted as a function of depth into the calorimeter for LEP-I and LEP-I1 Bhabha electrons. There is a sizeable shrinkage of the shower core with increasing energy. As w is related to the position resolution

293 OPAL

~"""""""""'1

Figure 5. The pad boundary bias as a function of the pad boundary transition width 20. The points refer to different depths in the bare calorimeter (solid circles) or after an optional preshowering layer (open boxes).

-2,,

800

8 600

.51

g 500

E-96-101GeV r E =45.5 GeV

400

300

200

,w

v 11

. ... v

Figure6. Pad boundaryransition t width as a functionof depth in thecalorimeter at a fixed radius for LEP-1 and LeP-II energies. near the pad boundaries, this indicates that the radial resolution inherently improves a t energies higher than LEP-I.

5. Energy measurement The distribution of the summed energy in the left and right calorimeters (after all other cuts) is shown in Fig. 7 for a typical OPAL run. The bulk of selected Bhabha events have back-to-back electrons and positrons with energies close to the beam energy. The large accidental background is visible at small en-

294

g 1.4

OPAL

9 1.2 -=I

w

1

5 1.4

9 1.2 1w

1

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0

0

Figure 7. Distribution of E L vs E R after all selection cuts except for the energy cuts and before the acollinearity cut ( a ) or after it (6). The lines show the cuts applied to the energies.

ergies and is reduced to negligible levels by applying tight energy cuts, which also eliminate a small fraction of real Bhabha events. The visible radiative tails extending from the full energy spot originate from events which have lost energy due to a single initial state photon emitted along the beam axis. For these events transverse momentum conservation implies: ER/EL = RL/RR. A useful quantity to improve our understanding of the Bhabha events failing the energy cut is the radius difference or acollinearity A R = RR - RL. A cut at AR < 10 mrad reduces both the background and the impact of uncertainties in the low energy tail of the detector response function, as can be seen from Fig. 7/a-b. The systematic error due to the energy measurement is reduced by almost a factor 3 with the A R cut. By cutting on the acollinearity one can also effectively limit or constrain the energy lost to initial state radiation. Therefore it is also useful to provide clean samples of beam energy electrons for studying the energy response of the calorimeters. Also samples with a selected lower energy can be isolated, though with lower statistics. The energy resolution has stayed almost constant during all the LEP running. At LEP-I ( E x 45 GeV) AEIE = 3.8 - 4.5% (for right - left calorimeter); at LEP-I1 ( E 5 104 GeV) AEIE = 5.0% (for both right and left calorimeter). Differences between the two calorimeters as well as from LEP-I and LEP-I1 are due to different amounts of preshowering material. 6. Final error on luminosity

The main experimental systematic errors on the OPAL luminosity measurement at LEP-I' are summarized in table 1. After all the effort on radial recon-

295 Table 1. T h e most important systematic errors in the final luminosity measurement for LEP-I.

Energy Inner Anchor Radial Metrology Total Experimental

I

Total Theoretical

I

5.4

I

struction, the dominant systematic error is related to the energy measurement , mostly due to uncertainties in the tail of the energy response function and the nonlinearity. The final experimental systematic error successfully matches the desired level of precision, well below lop3, and even surpasses the present theoretical precision of the calculated Bhabha cross section, which is one of the most deeply studied QED processes.

7. Conclusions The OPAL Si-W luminometer has reliably operated at LEP for 8 years (19932000), with high efficiency and negligible losses of Si detectors and readout electronics in a non-trivial background environment. Its performance can be summarized by these figures: (1) Energy resolution M 4% almost constant from E = 45 GeV t o E = 100 GeV; (2) Good efficiency to resolve close lying clusters: E 2 50% for A R 2 1.0 cm; (3) Good S/N ratio for mips: lO/l; (4) Position resolution on the radial coordinate of 130-170 pm with a residual bias less than 7 pm.

In particular the very small residual bias on the position of the acceptance cut was crucial to achieve the extraordinary experimental systematic error of only 3.4 x 10-4. References 1. OPAL Collaboration, G. Abbiendi et al., Eur. Phys. J. C14,373 (2000). 2. J. A. Lauber, S. Gascon-Shotkin, R. G. Kellogg, G. R. Martinez, Nucl. Instrum. Meth. A396, 165 (1997).

THE ZEUS HADRON ELECTRON SEPARATOR, PERFORMANCE AND EXPERIENCE

P. GOTTLICHER Deutsches Elecktronen-Synchrotron DESY, Hamburg, Germany E-mail: Peter. [email protected] (For the ZEUS-HES Group) The hadron electron separator (HES), a component of the ZEUS experiment is designed to improve the identification of electrons generally and, in particular, within jets. It consists of 20518 silicon diodes with 20m2 active area. The diodes are installed after 3-5 X O of the electromagnetic uranium-calorimeter, where the maximum intensity of the shower is expected. With an analog readout of each channel the deposited energy is measured. The HES improves the electron identification by a energy dependent factor 2.5 to 5 and the granularity by a factor 10. The attained position resolution is 5.4mm.

1. Introduction

In HERA at DESY protons (820GeV/920GeV) collide with 27.5GeV electrons or positrons. The scattered or produced e* and y’s have energies from very low values to a few 100GeV. For example in DIS and heavy flavor physics it is important to study electrons with a few GeV. These electrons are identified with the tracking detectors and the calorimeter. A cross-section of the ZEUS detector is shown in figure 1’. The Uranium calorimeter contains forward (FCAL), barrel (BCAL) and rear (RCAL) elements which are longitudinally segmented into one electromagnetic and two hadronic sections readout separately as towers. Hadronic towers have 20 x 20cm2 and electromagnetic towers 5 x 20cm2 (FCAL,BCAL) and 10 x 20cm2 (RCAL) planar sizes. This is much broader than the radius of the electromagnetic shower R ~ ~ l =i 2cm. h ~ The ~ measurement of the showers is improved by a highly segmented planar detector at a depth close to the shower maximum. This component called hadron electron separator (HES) consists of two circular planes of diodes, RHES and FHES, in the RCAL and FCAL with a radius of 1.9771. The HES uses the different showering behavior to distinguish e* and 7’s from hadrons. Typically electromagnetic showers develop early into narrow cones while most hadronic particle showers start only deep inside the calorime-

296

297

-

JERA

proton

820l920GeV

Gal

Rear: RHES Figure 1. A cut through the inner part of the ZEUS detector along the beam line showing the tracking detector, the magnet and the calorimeter split into forward (FCAL), barrel (BCAL) and rear part (RCAL). The HES is the cross-hatched bar from the top to the bottom of the electromagnetic RCAL and FCAL.

ter, This leads to the idea of measuring the number of particles crossing a plane in the early stage of a shower. The position of the plane is chosen as 3 to 5x0, where the maximum intensity of the electromagnetic shower is expected. Comparing this position to the 25x0 or 1 X (nuclear interaction length) longitudinal segmentation of the calorimeter allows HES to identify electromagnetic showers

Figure 2. A view of the front face of the FCAL together with a blow up where the cell sizes of the electromagnetic and hadronic calorimeter (EMC,HAC) and the HES are indicated.

298

:Ocm

E

N

m

Figure 3. A module of the ZEUS forward calorimeter with the Uranium and scintillator plates, the optical readout and the HES.

more easily. The finer segmentation of the HES diodes, 10cm2 in comparison with the 5 x 20cm2 or 10 x 20cm2 of the FCAL and RCAL, allows the HES to search for electromagnetic showers within jets.

2. Constraints on the design and the construction of the HES How the HES is embedded into the calorimeter is sketched in figure 3l. The calorimeter consists of Uranium plates with a height of 4.6m . These alternate with the scintillator plates. On the sides of the module wavelength shifters and light guides are mounted. The HES is constructed to have a low impact on the energy measurement. Therefore the 5m long gap after the forth scintillator plate has a depth of only 1.5cm. In the gap only a small amount of material has been installed, because this is passive for the energy measurement and the position close to the shower

299

maximum is most sensitive. The Uranium, scintillator and wavelength shifters surround the gap in the horizontal plane completely and constrain the geometry. The only access to the gap is at the top of the calorimeter through an opening of 16.3 x 1.5cm2. Through this HES modules, with lengths equal to the 4.6m height of the active calorimeter, has to be installed including the infrastructure for 672 readout channels - cooling, power and signal lines. As the gap is close to the tracking chambers the magnetic field of the solenoid at the HES-gap is close to the nominal 1.42’. The frontend electronics is housed in cabinets of the iron support structure on top of the calorimeter. 3. Experimental Setup

Silicon diodes forming the active medium have the advantage of producing a high electrical signal in a small volume and of being insensitive to the magnetic field. In the 400pm thick silicon minimum ionizing particles (MIP) produce 33000 electron-hole pairs. A compromise between number of channels and shower size R ~ ~ l + ~ ~ = leads 2 c r nto a size 3.47 x 3.07cm’ of a diode with 3.32 x 2.96cm2 being the active area’. This improves the granularity by a factor 10 compared t o the cell size of the FCAL. Figure 4a illustrates, how 2 diodes and their preamplifiers are combined on one ceramic plate. The HES detector contains 20m2 of silicon or 20518 channels. Since the available space is small and additional material worsens the energy resolution, central parts of the construction have multiple functionality. The multilayer board is used as a cable for 112 signals from 56 ceramic plates, supply and calibration lines. It also gives the HES-module the mechanical strength. Since the signals have to be transported from the bottom of the CAL to the c) Cross section of a HES module

Multilayer boards yater tubes// AI-frame

/

Detector card with diodes Detector card, with pceamplifier

Figure 4. Construction of a HES module: a) a ceramic plate with 2 diodes and preamplifiers; b) equipment of 2 long multilayer boards. The full structure has 56 cards per multilayer board and a total length with connectors of 4.6m c ) cross section through a closed module.

300

electronics on the top, these 18-layer boards have an unusual length of 4.6m. Because of the special production techniques for such a length, no electrical connections between the layers are possible. The inner layers are contacted by cutouts in the layers above at the solder pads for the detector plates. Adverse effects on electrical signals is limited by keeping the capacitance of the signal lines below 1nF. Nevertheless the rise time of the signal grows over the full length from 50ns to loons, which is compatible to the bunch crossing time of HERA. Each multilayer board is equipped with up to 56 ceramic plates (figure 4b). Along the multilayer board the active areas of the diodes alternate with the preamplifiers and only half of the area is sensitive to particles. To get full coverage two of these structures are shifted against each another by the length of a diode and then folded over each another as depicted in figure 4c. Each diode faces a preamplifier of the opposite multilayer board. Together with a few additional mechanical parts a HES-module is formed containing two multilayer boards and 224 diodes. The robust parts covers the more sensitive diodes and electronics. The power consumption of 9OmWlchannel has to be water cooled, because the heat conductivity of the surrounding sandwich calorimeter is low and the gap is 4.6m long. Three modules with each 224 diodes are installed into each calorimeter module and covers, sitting side by side, the area between the wavelength shifters of the calorimeter with 672 diodes. HES adds 0.lXc.1of material to the calorimeter. The active area of the HES diodes covers 94% of the area incident to particles. The non-sensitive area breaks down as follows: the field stop in the diodes before their mechanical edge: 2.5%, the mechanical construction of the modules: 2.5% and the limited size of the diodes: 0.5%, because of economical reasons four diodes are produced from a single 4-inch-wafer. In addition the wavelength shifters of the calorimeter pass the HES plane making 9.5% of the area inaccessible. Figure 5 shows an overview of the further readout electronics. Inside the cabinets of the iron structure shaping amplifiers slow the rise time of the pulses to 180ns, which has the advantage of removing the position dependent effects within the multilayer board. The overlay of consecutive bunch crossings can be tolerated, since the interaction rate at HERA is much lower then the bunch crossing rate. With an integrated circuit3 developed for the ZEUS-calorimeter the signal height is sampled each 96ns and stored in capacitors of a circular analog pipeline. This is overwritten until the ZEUS trigger decides to keep the data. Then 24 channels and four samples for each channel around the triggered

301

time [bunch crossings]

Figure 5 . The electronic chain of the HES

bunch crossing are multiplexed to a single signal line. This reduces the multiplicity of cables to the electronic hut. After digitizing a DSP calculates the energy and the time of the diode signal and suppresses signals below 0.4MIP. Typically per trigger data from 5% of the channels are sent out to the ZEUS data acquisition chain. 4. Performance and Experience

The calibration of the HES is performed by charge injection into the preamplifiers. Since drifts are very small this essentially identifies faults. The energy scale is checked with muons. At HERA halo-p are produced far upstream of the ZEUS experiment and propagate parallel to the proton beam crossing the HES-plane perpendicularly. The energy deposition spectrum (figure 6) follows a peak fitted by a convolution of a Landau-distribution for the ionization with a Gaussian for the detector resolution4. The energy deposition of a MIP is 120keV. For electrons the energy of the shower is distributed to a few diodes. A typical cluster is shown in figure 7. The clustering algorithm used looks for a diode with a local maximum of energy deposition and associates all 8 neighbors to it. This algorithm is verified by analysing, how much more energy is measured in an enlarged area of 5 x 5 diodes5. On average the cluster contains 96% of the energy, which is sufficient for electron identification. The incident position of the e* and 7’s can be reconstructed by a weighted mean of the diode positions. The weighting coefficients are functions of the pulse heights. Away from

302

the edges (85% of the HES area) the position resolution is measured with test beams to be 5.4mm for 25GeV e* (figure 7)6. The Monto-Carlo simulation describes the data for DIS electrons ( z 25GeV) well and shows a resolution of 5mm inside the ZEUS experiment7. The improvement for the electronlhadron separation was measured with clean samples at test beams. Hadrons and electrons show distinguishable spectrums with a relatively low misidentification probability (figure 6)'. A cut with 90% electron identification efficiency has been defined. In table 1 the misidentification rates for calorimeter only, HES only and combined measurements are shown. For the calorimeter data the analysis is restricted to a tower with a front surface of 20 x 20cm2, to adapt the analysis to the real situation at ZEUS with jets. Using HES improves the misidentification by a factor between 2.5 and 5 depending on energy. The HES was installed into the ZEUS experiment as a second stage upgrade. The plane in the rear calorimeter (RHES) was installed during the years 1992 to 1994 and the plane in the front calorimeter (FHES) from 1996 to 1998. FHES was operated since then until 2000 with a bad channel rate of 2 to 3% and RHES until 1999 with a bad channel rate of 3 to 6 %. The reason for these failure rate are mainly connectors. To keep this rate low continuous repairs on

a) Muons at ZEUS

'

'

'

Table 1.

'

'

'

'

'

'

-

Rate of misidentified hadrons in [%] for 90% e - efficiency.

Used detector

2GeV

Energy 3GeV 5 G eV

Calorimeter

7.79%

4.03%

0.65%

0.37%

HES

3.8%

3.4%

3.5%

3.6%

HES and calorimeter combined

1.47%

0.86%

0.22%

0.15%

Factor of improvement by HES

5.3

4.7

3.0

2.5

9 G eV

303

c

P

'03

3 102 L

10 1

1 1.5 2 ratio: E(SxS)/E(3~3-~luster) reconstructed - beam position [mm]

Figure 7. Cluster for electrons: a) a typical cluster for a 21GeV e - , each tower corresponds to a diode; b) spectrum of the energy ratio in an increased area ( 5 x 5 diodes) to the 3x3cluster; c) Position resolution for 25GeV e- away from the edges of the HES area.

detector access days are made. The number of broken and repaired channels of 100/month is mainly due to the fact, that a single faults effects large groups of channels. In the RHES water leakages developed during the year 1999. The copper tubes used inside the HES-gap of the calorimeter has a wall thickness of only 0.3mm to keep space and material low (figure 4). These tubes corroded from the inside. Later water analysis showed traces of sulfate-ions and organic molecules, both are known to speed up corrosion. The extensive repairs were possible during the long shutdown for the HERA luminosity upgrade and the complete HES is prepared for the HERA-I1 runs. Acknowledgements

The HES is an international project. I thank all contributors - physicists, engineers and technicians - from Germany, Israel, Japan, Korea , Russia, Spain and the USA for their combined effort. References 1. The ZEUS-Detector, Status-Report 1993, ed. U.Holm, DESY 1993 2. K.H. Barth, Messungen zur Homogenitaet des ZEUS Hadron-Elektron-

Separators, Diploma-Thesis, University Hamburg (1991) 3. W. Buttler et al., Design and Performance of a lOMHz CMOS Analog Pipeline, Nucl. Instr. Meth. A277, 217 (1989) 4. T.Kuhn-Sander, Untersuchung von Strahl-Halo Myonen im ZEUS Detektor, Internal Note DESY F35D-96-15, DESY (1996) 5. J.I. Fleck, K. Ohrenberg, Electron Identification in the HES and a new way to determine the efficiency of electron finders, Internal ZEUS-Note 95-9, DESY (1995) 6. U. Wollmer, Studium der Ortsaufloesung im Hadron-Elektron-Separator des ZEUS-Experimentes, Internal Note DESY F35D-95-09, DESY (1995) 7. F.Goebe1, Measurement of the Diffractive Contribution to the DIS Cross Section Using the ZEUS Forward Plug Calorimeter, DESY-Thesis-2001-049, DESY (2001)

DESIGN CONSIDERATIONS FOR A SILICON/TUNGSTEN ELECTROMAGNETIC CALORIMETER FOR A LINEAR COLLIDER DETECTOR

R. FREY Physics Department and Center for High Energy Physics University Of Oregon, Eugene, OR 97403 E-mail: rayfreyOcosmic.uoregon.edu

M. BREIDENBACH, D. FREYTAG, G. HALLER, M. HUFFER, J.J. RUSSELL Stanford Linear Accelerator Center, Stanford, CA 94309

We discuss some issues relevant for a highly granular silicon-tungsten electromagnetic calorimeter, such as those currently being designed for a future linear collider detector. An important issue is the interplay between the silicon pixels and readout electronics. Here, we propose an integrated solution.

1. Introduction

Accurate jet reconstruction and excellent jet energy resolution are widely considered to be prerequisites for calorimeters at a future linear collider (LC) detector. Current designs' strive to push jet reconstruction t o new levels of performance. An essential function of the electromagnetic calorimeter (ECal) in such designs is the separation of jet energy depositions due to photons from those due to charged particles. This calls for a dense, highly granular (in 3-d) ECal. A natural technology is alternating tungsten and silicon layers. Tungsten has a small Moliitre radius (9 mm) and the silicon can provide practically any transverse segmentation. Hence, one has the possibility of providing a high-quality image of the energy depositions. Such a silicon-tungsten system would have lo8 silicon channels (pixels). Clearly, this calls for a compact, economical readout. In this brief report, we introduce a scheme for an integrated front end readout of a silicon-tungsten ECal. It is discussed in the context of the SD detector concept for the NLC, although it may in part be applicable elsewhere.

-

304

305 2. Silicon and Readout Configuration Since we are still far from a final technical detector design, the current SD design concept2 should be considered fluid, the parameters being educated guesses based on existing simulation studies. Both barrel and endcap of the SD ECal consist of 30 longitudinal layers each of tungsten and silicon. The tungsten layers have thickness t w = 2.5 mm (z 0.7Xo). The interleaved readout layers are also set to t, = 2.5 mm thickness, comprised of 0.3 mm of silicon, 2 mm of (G10) motherboard, and a small air gap. The transverse segmentation of the silicon is 5 mm x 5 mm. With this segmentation, the number of detector pixels to be read out is about 50 million. We may expect the cost of these relatively simple silicon detectors (compared to strip trackers) to be less than $2 per cm2 by the time they are purchased in quantity for a LC detector, making a silicon-tungsten ECal feasible. But we also have to provide a simple, rational electro-mechanical detector and readout configuration. So to control cost and complexity, we seek an integrated configuration, where one channel of electronics serves a large number of pixels, and the electronics is compatible with the detector mechanical structure. To reduce cost, we start with the largest silicon wafer readily available, which for now we assume to be of 6 inch diameter. Figure 1 depicts the central region of a possible detector on such a wafer. The hexagonal silicon pixels are 5 mm across, giving about 1000 pixels per 6 inch wafer. Metallization on the wafer connects the pixels to an array of bump bonding pads. A single 1000 channel readout chip is bump bonded to this array. Figure 2 provides a cross-sectional view of the connections between the wafer metallization, the readout chip, and a G10 motherboard. Only 10 lines are required on the motherboard per chip - power, silicon bias, a multiplexed digital output, and various control lines. We assume these connections are wire bonds. An important design consideration is the minimization of the readout gap, since the effective Molihre radius of the ECal is R E ( 1 ( t , / t w ) ) ,where RE M 9 mm is for tungsten only. The bump bonding procedure has become rather standard in recent years for pixellized detectors in high energy physics. The relatively large pixel size for our application allows for a relaxed bump bond pitch of 250 pm, roughly an order of magnitude from current limits. In our case, the pitch is driven more by the area required by the readout chip, assuming 0.25 pm technology. Thermal management of the detector is critically important because the electronics are embedded in the calorimeter in this approach. The best arrangement (for minimizing dead spaces and gaps) would be a fixed temperature heat sink at one edge of the ECal structure, and conduction cooling through

-

+

N

306

Figure 1. T h e center of one 1000-pixel silicon wafer showing the bump bond array a t the center for the single readout chip. A few representative traces from pixels t o the bump bond array are shown.

\

S i l h wafer

0 Figure 2.

I

I

I

I

1

2

8

4

5 mm

Cross-sectional view in the vicinity of the readout chip.

each layer to this heat sink. Thus the maximum thermal path would be about 1 meter. The conduction layer might be a thick copper ground plane in the G10 motherboard, or even a separate copper layer approximately 1 mm thick. This approach requires very aggressive control of the average power of the read out chips. The small accelerator duty cycle at a linear collider plays an important role here, allowing a large average power reduction if power pulsing of

307

electronics is employed. For NLC, if the turn on transients can be managed, then power pulsing should reduce the front end power by a factor of 1000. For the TESLA LC design, the duty cycle is an order of magnitude larger, possibly requiring more elaborate cooling3. We use the charge amplifier of the GLAST tracker5 as an existence proof of an applicable low noise, low power front end design. Assuming the back end power can be made small compared to the preamps, a full chip at NLC would average about 2 mW. We estimate that this power would cause AT M 1°C using a 6 oz copper ground plane in the G10. This should be fine, as would another factor of 2 or 3 in the power. However, achieving these levels for the average power while keeping the rather high performance standards will be a fundamental engineering challenge. 3. Dynamic Range and Resolution Requirements

Each ECal pixel must be sensitive to a large dynamic range, from MIPS to Bhabha electrons at the full beam energy. We have employed an EGS simulation which incorporates* the subtleties in simulating thin silicon sampling layers to evaluate this. We find that the ratio of 500 GeV Bhabha to MIP is at most 2000 for pixels near shower maximum. Because of the exponential transverse falloff of shower energy with distance, using smaller pixels does not decrease the dynamic range significantly. This is illustrated in Figure 3. Again using the GLAST chips as an example, the noise would be x 300 rms electrons. 1.0 - - - . +Mean m D

0.8

-

--

-

1

- OMbxlmUm -

a

;

l " " l " " I " " 1 " " l " " --------------

+

-

0

L

al

r

m C

0

+

0.6 4

-

0

C

+

2 0.4

-

0.2

-*

-

0

LL

D W a

w

0

.

0

6 X, Depth 100 GeV e'

~

"

"

~

"

"

~

"

"

~

"

"

~

"

"

~

"

"

1 -

-

-

Figure 3. Energy deposition in the central pixel of a silicon layer at shower maximum, Expressed as a fraction of the deposited energy in the entire layer, as a function of the pixel size (in mm). Based on an EGS4 simulation with 100 GeV incident electrons.

308

The requirements for an excellent S/N for MIPS and the large dynamic range leads to a readout element with a two-gain analog stage followed by a multiplexer and 12-bit ADC, effectively providing two overlapping 12-bit ranges. It may be possible to go to fewer than 12 bits with further study. A schematic of a front end electronics channel is shown in Figure 4.

MUX - A D C - - O

Figure 4 . Schematic front end electronics channel. We anticipate C M 10 pF and G1/G2 M 15 for a 12-bit ADC. An additional capacitor M C will be required for calibration.

4. Plans

We wish to develop this design in staged prototypes, starting with single chips and wafers to demonstrate the integration approach, and leading to construction of a full-depth module for testing in a beam of both electrons/photons and hadrons.

Acknowledgements This work is supported in part by the US Department of Energy under award DE-FG02-96ER40969 (Oregon) and contract DE-AC03-76SF00098 (SLAC).

References 1. R. F’rey, see “Calorimeter Considerations for a Future Linear Collider Detector,” and references therein, these proceedings. 2. The American Linear Collider Working Group, “Resource Book for Snowmass 2001, Part IV,” hep-ex/0106058, 2001. 3. H. Videau, these proceedings. 4. G. Lindstrom, et al., Proc. Workshop on Calorimetry for the Supercollider (1989) 215, World Scientific. 5. GLAST document LAT-22-00169, 2002, available from http://www-glast .slac.stanford.edu/documents/LATDocSort .asp.

A SI-W CALORIMETER FOR LINEAR COLLIDER PHYSICS

HENRI VIDEAU AND JEAN-CLAUDE BRIENT Labomtoire Lepn’nce-Ringuet Ecole polytechnique - F-91128 Palaiseau, France E-mail: Henri. VadeauOin2p3.fr

In the framework of the ECFA- DESY study for an electron linear collider we have studied a calorimetry intended to be well suited for jet measurement. As a result we contend that the best solution for the electromagnetic part is a silicon- tungsten sandwich. We describe here the current status of the study going on inside the CALICE collaboration: http://polywww.in2p3.fr/tesla/calice.html After a short reminder on the physics to be expected at the linear collider, we state what may be the requirements for the electromagnetic calorimetry and a possible ”optimum” design. As an example of implementation the ECFA- DESY TDR design is briefly sketched. We then focus on the current R&D developments for electromagnetic calorimetry in the CALICE collaboration and its objectives, the current design of prototypes and the perspectives.

1. Requirements 1.1. The physics The physics expected as being of interest at an electron positron collider in the energy range 90 to 1000 GeV is going to be dominated by the production of tops, W’s, Z’s and Higgs’s. Some obvious reactions to study are, e+e- + tt, e+e- + Z H , e+e- + W + W - , e+e- + WWvV or e+e- + Z Z Y V . Experimentally this means a physics of many jets (or better dijets) and leptons. Under that assumption, the first requirement for the calorimetry is to optimise the jet reconstruction and the lepton identification. It can be tempting to rely mostly on calorimetry and measure a flow of energy as recorded by the calorimeter, but the resolution of hadron calorimeters is known to be mediocre, the strong magnetic field disconnects the actual direction of the charged hadrons seen in the calorimeter from the direction at the interaction point, the calorimeter does not provide a measurement of the muons energy. Another approach largely used at LEP under the slightly misleading name of ”energy flow” is better suited. To better identify the method we will call it ” analytic energy flow” (AEL). This method was adequate at LEP but the presence of the coil in front or in the middle of the calorimeter did not allow it to bring its full efficiency.

309

310

It is still perfectly adequate for our purpose even though it is not a universal panacea. The principle is , remembering that p j e t = X P c h a r g e d p a r t i c l e s P n e z l t r a l h a d r o n s , that the charged particles will be measured in the py tracker, the photons in the front part of the calorimeter, the neutral hadrons mostly in the rear. To summarise the argument, observing, first, that the charged particles make about 60% of the energy, the photons 30% and the neutral hadrons 10% with large fluctuations, second, that being of rather low energy the charged particles are much better measured by the tracker, third, that the photon energy is properly measured in an electromagnetic calorimeter, it appears that the main purpose is to separate well the 10% neutral hadron energy and t o measure it at best. It is worth noticing that, if using the tracker is very effective, the number of lost or spurious tracks has to be very well controlled and it is essential to identify the secondary vertices. The behaviour of the detector with respect to jets is then the following, the large magnetic field sweeps the charged particles away from the neutrals and the separation between photons and neutral hadrons is obtained by choosing a calorimeter with a first part with a large interaction to radiation length ratio. Remember that the presence of a charged lepton is the sign of the presence of a neutrino, and it is important to be able to separate jets with and without neutrinos.

+

+

1.2. The electromagnetic calorimeter The requirements for the electromagnetic calorimetry can then be drawn: Identify efficiently the photons even at low energy without collecting fake ones from hadronic debris, identify electrons even in dense jets, measure photons and electrons in jets as accurately as possible. An important feature is to be able to identify in the full domain of energy the electrons even in jets. The first reason is that, due to bremsstrahlung the energy of an electron is best estimated by the calorimeter, the second is that identifying prompt charged leptons signs the presence of neutrinos, which is essential in comparing jet energies. The calorimeter is used as soon as the electrons reach the calorimeter, below that energy some extra identification, like dE/dx is mandatory. Note that it is essential to distinguish the prompt electrons from conversions, this is a requirement on the tracker. Once photons or electrons separated from the charged particles, a good measurement comes from the calorimeter sampling, it is important to recognise that the uncertainty coming from the collection of the shower cells may be dominant over the intrinsic resolution coming from the sampling.

312

An example of an actual design can be found in the TESLA Technical Design Report. The absence of projective cracks was a strong motivation for the design described in the TDR, it can be seen on figure 2. We chose an eightfold symmetry which gives the possibility of getting the signal out behind the electromagnetic part without creating any azimuthal crack, we obtain that way a full phi coverage. To solve the mechanical problem of a structure without holes we went to a structure made of an epoxy-carbon fibre composite. There are many very thin walls which are not projective since the choice of a very small grain read out makes the projectivity of the read out of no real interest. The impact of these walls is negligible. By using 40 layers of sampling, 30 thin ones (0.4 Xo) followed by 10 coarser (1.2 Xo) the resolution is close to O . l / f l at low energies. In fact, as the effective resolution in jets is largely related to the efficiency in collecting the right cells contributing to the shower, a coarser sampling could be enough, 20 to 30 layers. The TDR choice generates 32 million channels for a 1 cm2 pad size.

Figure 2. View of the barrel structure on the left and of an end-cap on the right, the electromagnetic calorimeter is in grey and the hadronic in bordeaux.

In a given module the structure is flat, the signals are taken out along the layers toward a reserved area behind the ECAL carbon fibre/epoxy structure embedding tungsten slabs, as shown in figure 3. The mechanical structure is more clearly described in the figure 4, it looks like a bookshelf where the detecting elements can be slit in slots called "alveoli". One layer of tungsten every second layer is part of the structure, the other tungsten layer is part of the detecting element. The tungsten is cut in long

313

Figure 3. Detail of the calorimeter structure showing the way the signal is extracted behind the electromagnetic calorimeter without generating a crack.

Ecole Pol technique L.P.~W.E 2010412000

Tungsten slab

I/

Alveolus\

\P a P t i t i o n

Figure 4. T h e mechanical structure showing the alveoli and the detector slabs sliding in.

314

slabs, the structural and the detector slabs are orthogonal.

Figure 5.

Detail of the detector slabs like envisaged in the TDR.

The drawing 5 shows the structure of the detecting elements like envisaged in the TDR. On each side of the tungsten, silicon wafers are glued to a G10 board, the connection to the front-end electronics situated at the end of the module is made by layers of kapton or special wires. This design is currently being revised. 4. Current developments in the CALICE collaboration, the

objectives

A large R&D effort has started after the TDR publication in a collaboration named CALICE formed around a proposal sent to the DESY PRC. This large collaboration counts now more than 130 people from eight countries, Czech Republic, France, Germany, Great Britain, Korea, Russia, United States and Uzbekistan, a detailed list of the people involved can be found at the URL given on the title page. Its aim is to prove the existence of a design, hardware and software, fulfilling the requirements for a linear collider experiment. To reach that goal, specific technological developments take place up to a physics

315

prototype of the complete calorimeter to be tested in a beam. That way, not only would a design exist but also a collaboration ready to build it. The organisation is not monolithic and leaves room for different solutions to coexist up to the time when the choice becomes clear. Currently one version of electromagnetic calorimeter is being followed, namely the Si-W sandwich, but two versions of hadron calorimeter, one putting the emphasis on resolution, the second on granularity. Currently the main areas of work, concerning the electromagnetic part , cover the price of the silicon, the wafer design to keep the space at a minimum, the channel number and the space needed for the read out, the way to get the signals out and to master the coherent noise, the making of the mechanical structure. The main points to be tested in beam are the dead zones and the accuracy with which they can be modelled, and the hadron behaviour in this calorimeter, the validity of the simulation, here Geant4, not being fully proven today.

Moore's Law for Silicon Detectors

Figure 6.

Evolution of the area of the silicon detectors area.

The figure 6 shows the evolution of the area of silicon being used in experiments with time. By the time of the linear collider, 3000 square meters is the area to be expected and is not far from what is envisaged in the TDR. The data are from H.F-W. Sadrozinski, UC-Santa Cruz. An adequate guess of the price derived from the evolution with years of

316

Moore's Law for Silicon Detectors

f

15 DATA From H.F-W. Sadrozinski, UC-Santa Cruz Figure 7.

Evolution of the price of silicon for microstrip detectors.

the prices is more difficult. As can be seen on the picture 7, the error in the extrapolation is larger. Note that the reduction of the price has been related t,n

t,hp urnwinu

nf t h e wafers

9 i 7 ~

Tt

i s tn

he e m n h a c i c d t h a t thn

fimiirnc

correspond to microstrip detectors; in the case discussed here, the number of masks is reduced and the yield may be much better since the detector, due to the huge number of channels, is not very sensitive to the fraction of dead channels, a fraction of few percent being tolerable if the dead channels are randomly distributed. Under these assumptions the price of two dollars per cm2 looks adequate and the price of 130ME taken from the TDR is likely t o be around 96 without considering any saving on the sampling or on the size of the calorimeter. The figure 8 illustrates the current wafer design for the prototype. A very important point is to show that using only one guardring around the full wafer instead of around each pad is tolerable, which means that a problem on one pad does not destroy the full wafer. It is envisaged to couple the pads to

317

the electronics capacitively, the resistor and the capacitance being obtained by depositing few microns of amorphous silicon and silicon nitride respectively on large parts of the pads. In the case of the prototype, the pad matrix is a square of 6.2 cm side cut out of a 4” wafer.

Figure 8. The matrix of pads and the guard ring design.

One more challenge is to get the many signals out of this dense calorimeter. Where should the front-end electronics be put? In the TDR it was at the end of the modules, this was for making it accessible and easy to cool since the heat generated is considered to be around 100 kW. The drawbacks are that the density of lines to transfer the signal to that point and the making of the connections arc out of current knowledge. New studies lead to try to put it inside the calorimeter itself. The number of connections at the edge of the module will be much less numerous and the accessibility can be better because the detectors elements can be, due to this reduced number of connections, more easily removed from their alveoli. Remains the cooling. The mechanical structure does not help to release the heat and an active cooling is mandatory. This can be done without loosing space as shown in the figure 9 but the pipes are tiny and the water, if used, generates quite commonly problems. This is

318

actively looked at.

Figure 9.

Cut through the detector slab showing the front-end chip and the cooling.

The figure shows a scheme for installing the front-end electronics inside the detector. The PCB connects to the pads by glue, the area of the pads makes it quite straightforward, and to the chip. It contains the power lines, the command line and the lines to get the signal out. Recalling the small occupancy of the channels, a zero suppression in the chip makes a serial output easy. With about 3mW/ch, a value not that easy to reach, cooling is mandatory but is not so demanding for the tube thickness. The ongoing studies with prototyping concern the cooling, the pick-up noise control, the front-end electronics, the building of the detector slabs, the mechanical structure. These three last items concern the final design but also the physics prototype which may require slightly different solutions. For example the front-end chip for the prototype is directly derived from a chip done for the Opera experiment. The structure will also be quite different in particular in view of the use of 6x6 cm2 wafers in the prototype when a larger size is considered for the final detector. The general aspect of the complete prototype of calorimeter t o be put in a beam is shown in the figure 10. On the left is a general sketch showing along the beam: the hodoscope, currently being built, then the ECAL made of three modules as shown on the right, and finally the HCAL. The complete system is

319

Movable tabie

Figure 10. Artist view of the Ecal and HCAL prototypes put in a beam

;o 5 6 5 mm!

- .I>? ,, 11 simcture ' mou.dng o: tunc;sten and e m o n epoxy conpos te easy .ntegrat,on of PCBIS. I

1:'composi;e Layer to ismmi

- 0 7 mm th.cktungs:en pates

-Tungsten ,l 4 mm

i x i 4 or 3 x 1 4

nu1

not n d e d w y paces pioclubgon -

PCB ~O.ut,on

ted1mo:ogy sk ... s

3rd compos,te ayer t o i s nun) froinMarc Anduze LLR CAL:CE Co. aborwon

Figure 11.

Detail of the curent mechanical design for the detector slab.

320

slightly tilted in order to disalign the structural elements with the beam in a way similar to the final calorimeter. The HCAL prototype will be built in such a way that the detecting medium can be either scintillator or gas chambers. The figure 11 presents the new design for the detector slab, made of a core of tungsten embedded in epoxy-fibre in the form of a H surrounded by the detecting elements, silicon with its PCB and cooling, surrounded by a grounded shield. 5. Perspectives and conclusions Even though the electromagnetic calorimetry only was presented, it does not make much sense to design independently electromagnetic and hadronic parts of the calorimeter. See the paper on jet calorimetry. It appears that a calorimeter adequate to the physics of the linear collider below 1 TeV is both feasible and affordable. The CALICE collaboration should be able to test with beam a complete prototype in 2004. Nevertheless a lot of effort on hardware technologies and also on software is needed. People having interest in that field, please join CALICE in a world-wide effort.

Sirnulation Covener: C. Seex

c. seez

Covener’s Report

S. Shevchenko

QCD Jet Simulation with CMS at LHC and Background Studies to H+ yy Process

P. Loch

Comparisons of Electron and Muon Signals in the ATLAS Liquid Argon Calorimeters with GEANT4 Simulations

P. Paganini

Data Volume Reduction Strategies in the CMS Electromagnetic Calorimeter

C. Currrat

Performance of CDF Calorimeter Simulation for Tevatron Run I1

A. Kiryunin

Simulation of Hadronic Showers in the ATLAS Liquid Argon Calorimeters

M.J. Varanda

MC Simulation of the ATLAS Hadronic Calorimeter Performance

M. Wielers

Simulation Studies of the Jet and Missing Transverse Energy Performance of the ATLAS Calorimeters

S. Kunori

Jet Energy Reconstruction with the CMS Detector

321

This page intentionally left blank

SIMULATION SESSION: CONVENERS REPORT

C. SEEZ Imperial College of Science, Technology and Medicine, London, U K E-mail: [email protected] (Convener’s Report)

Continuous use of, and interaction with simulation is evident throughout the design, development, construction and use of today’s calorimeters. h e quently this uses the full machinery of detailed shower simulation. We saw numerous examples of the use of simulation throughout the conference - not only in the simulation session. In particular, there were many talks touching on the investigation of ’energy flow’ algorithms, which attempt to exploit information from charged particle tracking in front the calorimetry, together with a fine spatial granularity of calorimeter readout, t o obtain improved jet resolution. Simulation is used t o extrapolate from single particle response, observable in a test beam, to jets from an event simulator, in a strong magnetic field. In the simulation session Shuichi Kunori presented results of detailed studies of the ’energy flow’ technique applied to the CMS detector which show that some improvement of the jet energy resolution can be obtained. There were two talks, starting and ending the session, describing studies in which simulation had been used t o extrapolate from calorimeter response to physics performance: in the CMS detector (Sergey Shevchenko), and in the ATLAS detector (Monica Wielers). There were three talks on comparison of test beam data t o shower simulation, with particular emphasis on the ongoing transition from GEANT3 to GEANT4. These talks were all from the ATLAS collaboration, where the Calorimeter geometry is often both challenging and important to simulate precisely ( e g . the accordion geometry). Peter Loch showed detailed comparisons for electron and muon signals in the liquid argon calorimeters. In general, and even bearing in mind the very high degree of precision expected for simulation of electro-magnetic processes, the agreement is excellent. In some cases GEANT4 does better than GEANT3, although there are parameters and situations in which GEANT4 does less well. Results on hadronic shower simulation were presented by Maria Varanda and Andrei Kiryunin. Considerable work

323

324

tuning, is required to get good agreement between simulated hadronic showers and test beam data. Working in close collaboration with the GEANT4 team reasonable agreement has been achieved over quite a wide range of measurements and shower parameters. Pascal Paganini talked about simulation of the CMS ECAL front end electronics to arrive at digitised time samples. Using such simulations, into which are fed shower simulations of simulated events, data reduction schemes are being studied. It will probably remain true that for many uses in physics data analysis, full detailed shower simulation has difficulty delivering enough events fast enough. So 'fast simulation' techniques remain of considerable interest. Charles Currat described the adaption and tuning of the Gflash package for CDF calorimeter simulation, which achieves a speedup of O(100) over full, GEANT, shower simulation.

Acknowledgments I would like to thank the speakers in the simulation session for their well prepared talks, and the organisers of the conference for its impressively smooth running.

QCD JET SIMULATION WITH CMS AT LHC AND BACKGROUND STUDIES TO H+ 77 PROCESS

V. LITVIN, H. NEWMAN, S. SHEVCHENKO, N. WISNIEWSKI Lauritsen Laboratory, California Institute of Technology, Pasadena, USA

We have simulat.ed and reconstructed one million of QCD jet events. This study was done with CMS full detector simulation, based on GEANT3 package, and object-oriented CMS C++ reconstruction program. T h e understanding of QCD jet background is important for the Higgs search in two-photon decay mode. Th e comparison with other types of backgrounds was also done. It was shown that the isolation tools were important ones t o isolate the signal process from the huge background one. Using the isolation criteria based on the information from P b W 0 4 electromagnetic calorimeter and the tracker we were able to reduce the QCD jet background to 15% of the total one.

1. Introduction

The process H+ yy provides one of the most promising signals to search for the Higgs boson in the mass region between 90 and 150 GeV't'. We present the results of the full detector simulation of different types of backgrounds to H+ yy process. The irreducible background such as prompt di-photon and prompt photon plus isolated bremstrahlung photon productions, and reducible background such as prompt photon plus jet and QCD jet productions, where one or two jets are misidentified as photons, are investigated. The QCD jet-jet background cross section is huge (- 1O'O pb). Therefore previous s t u d i e ~ l were > ~ ~done ~ at the generator level or QCD jet-jet background have obtained estimates for the rate, rj,t, at which a jet would be misidentified as a photon, and due to limited computational power, to estimate the QCD jet-jet background, the rates were simply multiplied together (factor r;,t). The problem is that the correlations within an event are not taken into account. Also the simulation was done with simplified geometry, so non-Gaussian tails in the resolution have not been adequately simulated. A special goal of this study is to simulate for the first time a large enough sample of QCD jet background to directly estimate the di-photon misidentification and compare it with contributions from other types of backgrounds. The isolation tools are important ones to isolate the signal process from

325

326

the background one. The aim of this note is to use the isolation algorithms, which are based on information from the PbW04 electromagnetic calorimeter and the tracker, to select isolated photons from Higgs decay, while rejecting the huge jet background below the level of irreducible one. 2. Different types of backgrounds

The search for H+ yy signal at LHC has treat three types of backgrounds: 0

0

0

Promp di-photon production from the quark annihilation and gluon fusion diagrams, which provides an intrinsic or ‘irreducible’ background (see Figure 1). Prompt di-photon production from significant higher-order diagrams - primarily bremsstrahlung from the outgoing quark line. The background from QCD jets, where an electromagnetic energy deposit results from the decay of neutral hadrons (especially isolated xos) in a jet and from 1 jet 1 prompt photon.

-+-

q

Y

Figure 1. Diagrams of irreducible background: (left) quark annihilation, (right) gluon fusion.

Table 1 shows the cross sections of different types of backgrounds. The threshold of the transverse momentum of the parton in the hard interaction process, P k T d ,which was used for generation4. 3. Simulation tools and Monte Carlo data samples

For the Higgs background study we used the following data samples:

327

Table 1. cross sections of different types of backgrounds. Background process

Pkrd(GeV)

Cross section (pb)

QCD jets 1 prompt photon Brem and 1 jet Gluon fusion Quark annihilation

35 25 25 25

8x107 8x104 27 45

+

Signal: -

50K H - + yy events at mH=llO GeV

Background: - 1 million of preselected QCD jet events -

+

500K of preselected y jets events 50K of quark annihilation events 50K of gluon fusion events

All events were generated with PYTHIA 6.1524 at f i = 1 4 TeV. The CTEQ 4L parton density structure functions were used in generation. The simulation of tracking in the detector was done using CMSIM version 121 (CMSIM is a full detector simulation program based on GEANT3 p a ~ k a g e ) The ~ . simulated events were digitized using the reconstruction and analysis program ORCA‘. The digitization was done at a luminosity of 2x 1033/cm2/s. Due to the huge QCD jet cross section (0 N lo8 - lo9 pb) a strong preselection at the generator level is needed to simulate a sufficient number of background events in a reasonable time. The QCD cross section strongly depends on the transverse momentum of the parton in the hard interaction process, Therefore a proper choice of this parameter is needed to generate QCD jet events. We have chosen PkTd to be equal to 35 GeV. Further generator level preselection was performed as follows: we looked for events with at least one pair of any of the particles, y, T O , e, Q, Q’, p , w , which could deposit a significant part of their energy in electromagnetic calorimeter. The transverse energies of these two particles should be greater than 37.5 GeV and 22.5 GeV (note that proposed offline CMS cuts are 40 GeV and 25 GeV, respectively’), and invariant mass of these two particles should be in the range 80-160 GeV, which is the range where we are going to search for the Higgs in two photon decay mode. The obtained rejection factor at generator level is -3000, what means that 3000 events should be generated to produce one event to be passed through full detector simulation. Therefore one million of simulated and re-

PkTd.

328

constructed QCD events corresponds to 3 x lo9 generated events. A check was done that we do not lose the events, which could fake a Higgs signal, due to the above preselection. A much looser preselection was used to simulate y jets event sample: we looked for events with at least one particle of any of the particles, y,no,e, 77, p, w , which has a transverse energy greater than 37.5 GeV.

+

4,

4 . Isolation 4 . 1 . Isolation based on electromagnetic calorimeter

Energy deposits from electromagnetic showers in the electromagnetic calorimeter are constructed into basic clusters using the Island clustering algorithm7. These basic clusters are in turn clustered into ‘super-clusters’ to recover the energy radiated from photon conversions, which fall outside the seed shower cluster7. The reconstructed supercluster is associated with electron or photon candidate. The isolation criteria is based on the sum of transverse energies of basic clusters in some cone R ( R = , / m ) around electron or photon candidates. The basic clusters, which belong to electron (photon) candidate supercluster, do not count. The algorithm contains two parameters: 0

the size of the cone R around electron(photon) candidate, where the transverse energies of basic clusters are summed. The transvere energy sum threshold, Epresh. If the sum of transverse energies is below this threshold, the photon candidate is considered as isolated, otherwise - non-isolated.

The detailed results are presented in elsewhere’. There is no strong dependence of jet rejection factor on the cone size R and the slightly better rejection factor is obtained for the cone sizes R = 0.30 - 0.35. 4.2. Isolation based on the tracker

The isolation criteria is based on the number of charged tracks, with PT greater than some PT threshold, p p T e s h ,calculated in some cone R ( R = , / m ) around photon candidate. The algorithm contains three parameters:

0

the size of the cone R around photon candidate, where the number of charged tracks is calculated. the PT threshold, p p r e s h . Only the charged tracks with p~ greater than p p T e s hare accepted for isolation.

329

0

the ‘number of track’ threshold, Nthresh. If the number of charged tracks with PT greater than choosen ppresh, calculated in cone R, is greater than Nthresh, the photon candidate is considered as nonisolated, otherwise - isolated.

The detailed results are presented in elsewhereg. The jet rejection factor is very sensetive t o the ‘number of track’ threshold, Nthresh. By going from N t h r e S h = 0 to N t h r e S h = 1, one can increase the Higgs signal efficiency by 6-lo%, while the jet rejection factor drops by a factor of 2. Therefore we decided to fix parameter Nthresh to be equal to zero. The pFresh threshold was choosen to be 1.5 GeV.

-

5. Results

Before study the photon isolation, the 12 GeV Level 1 double isolated electron trigger was applied both for the signal and background data samples. The transverse energies of the two photon candidate should be greater than 40 GeV and 25 GeV, respectively, and invariant mass should be in the range 80 160 GeV. The fiducial volume in rapidity was restricted to 1771 < 2.5 for both photon candidates. The cross section of different backgrounds are shown in the second column of Table 2 after the above cuts were applied. One can see that the QCD jet cross section is 25 times higher than the cross section of irreducible background.

Table 2. Background cross sections, do/dmrr, in fbb/GeV. Background process

All cuts before isolation cuts

Track isolation

ECAL isolation

QCD jets 1 prompt photon Brem and 1 jet Gluon fusion Quark annihilation

2500 1200 42.3 49.7

100.0 190 35.7 40.9

30.2 112.4 33.5 39.1

+

Figure 2 shows the invariant mass distribution of different backgrounds. As expected the behaviour is smooth and decreases with increasing of the invariant mass. The last two columns of Table 2 shows the effect of isolation. One can see that the isolation is a very effective tool to reduce the huge QCD background. 15% of the total After isolation was applied the QCD background is only

-

330 20

7

I

Gluon fusion

16 12

700 420

8

280

4

140

0 20

7

560

0

80

,

100

120

140

160 1

80 600 480

l&

I20

140

1 i0 1

j

Bremsstrahlung and 1 jet + 1 prompt photon

360

240 120 0

80

100

120

140

160

Figure 2. Invariant mass distribution of different backgrounds.

one. The overall Higgs efficiency is 33.5% after all cuts, including isolation cuts, were applied.

6. Acknowledgements We would like to thank J. Pino for the help in writing the analysis program and C. Seez for useful discussions.

References 1. C. Seez and J. Virdee, Detection of a n intermediate mass Higgs boson at L H C via its two photon decay mode, CMS TN/92-56 (1992) 2. CMS Collaboration, " T h e Electromagnetic Calorimeter Project - Technical Design Report", CERN/LHCC 97-33 (1997)

3. V. Tisserand, T h e Higgs to two photon decay in the A T L A S Detector, ATLAS Internal Note, PHYS-NO-90 (1996). 4. http://www.thep.lu.se/torbjorn/Pythia.html 5 . http://cmsdoc.cern.ch/cmsim/cmsim.html 6. http://cmsdoc.cern.ch/orca 7. E. Meschi et al., "Electron Reconstruction in the CMS Electromagnetic Calorimeter, CMS Note 2001/034 (2001) 8. V. Litvin, H. Newman, S. Shevchenko, Isolation studies f o r electrons and photons based o n ECAL, CMS IN 2002/023 (2002) 9. V. Litvin, H. Newman, S. Shevchenko, N. Wisniewski, Isolation studies f o r photons f r o m Higgs decay based o n the 'Tracker, CMS IN 2002/034, (2002)

COMPARISONS OF ELECTRON AND MUON SIGNALS IN THE ATLAS LIQUID ARGON CALORIMETERS WITH GEANT4 SIMULATIONS

D. BENCHEKROUN Universite' Hassan 11, Casablanca-Maarif, Morocco

G. KARPETIAN, R. MAZINI Universite' de Montre'al, Montre'al, Que'bec, Canada

A. KIRYUNIN', D. SALIHAGIC, P. STRIZENECt Max Planck Institut fur Physik, Werner Heisenberg Institut, Munich, Germany

J. KISH Institute of Experimental Physics, Kosice, Slovakia

K. KORDAS~,G. PARROUR Labomtoire de 1 'Acce'le'rateurLine'aire, Orsay, France

M.LELTCHOUK, S. NEGRONI, W. SELIGMAN Neuis Laboratories, Columbia University, Irvington, New York, USA

P. LOCH University of Arizona, Tucson, Arizona, USA

A. SOUKHAREV Budker Institute of Nuclear Physics, Novosibirsk, Russia

Signals from electrons and muons taken at testbeams with different modules of the ATLAS Liquid Argon Calorimeter have been compared to corresponding simulations using the GEANT4 toolkit. These simulations have also been compared in some detail with GEANT3 based predictions. Results for signal linearity, energy resolution, and shower shapes all generally indicate a good agreement between experiment and the two simulation packages, typically at the level of a few percent.

331

332 1. Introduction

The ATLAS Liquid Argon GEANT4 Comparison working group has been set up to evaluate the GEANT4 toolkit' for physics simulation. This includes comparisons of relevant calorimeter performance parameters like the average response, energy resolution and shower shapes, to testbeam data and GEANT32 simulations. Results from these comparisons for electron and muon signals in various calorimeter modules are summarized in this note. First comparisons of hadron signals in ATLAS liquid argon calorimeters can also be found elsewhere in these proceedings3. This note starts out with a brief description of the relevant ATLAS liquid argon calorimeters used for the comparison, followed by a discussion of the testbeam setups and their description in the simulation. Results for muons are presented in section 4, and results for electrons can be found in section 5. 2. The ATLAS Liquid Argon Calorimeters The ATLAS detector at the future Large Hadron Collider (LHC) features a liquid argon calorimeter system housed in three cryostats, one barrel and two endcaps4. The ElectroMagnetic Barrel (EMB) calorimeter is a liquid argon/lead device featuring accordion shaped absorbers. It is about 24 radiation lengths ( X O )deep, and covers the pseudorapidity range 1771 < 1.475. The most important performance characteristics of the EMB is excellent linearity and energy resolution with stochastic terms significantly below 10% and constant terms of the order of 0.3%, all well within the requirements for physics reconstruction in ATLAS. More details on the calorimeter have been presented5 at this conference. Each of the endcap cryostats contains an Electromagnetic EndCap (EMEC), a Hadronic EndCap (HEC) calorimeter, and one electromagnetic and two hadronic Forward Calorimeter (FCal) modules. The EMEC is an accordion type liquid argon/lead calorimeter, while the HEC features a classical parallel plate liquid argon/copper structure6. The HEC is about 103 X O deep for electrons, and about 10 absorption lengths (A) for hadrons. The electromagnetic and hadronic endcap calorimeters cover 1.3 < 1171 < 3.2, approximately, while the FCal closes the gap to the beam line (3.1 < Iql < 4.9). The electromagnetic Forward Calorimeter is a liquid argon/copper device featuring tubular electrodes with thin (250 pm) liquid argon gaps for charge collection. It is about 28 Xo deep. The hadronic modules have the same * On leave of absence from Institute for High Energy Physics, Protvino, Russia t On leave of absence from Institute of Experimental Physics, Kosice, Slovakia Now at University of Toronto, Toronto, Ontario, Canada

333

electrode structure, with slightly larger gaps (350 - 500 pm) and tungsten as absorber7. The around 10000 electrodes in each FCal module are oriented parallel to the beam at LHC, and arranged in a hexagonal pattern, which introduces interesting response features for this particular calorimeter. 3. Testbeam Setups in the Simulation

Test beam experiments conducted with production modules for the EMB and HEC, and special engineering prototypes for the FCal, have been described in the GEANT3 framework and using GEANT4 toolkit. To allow detailed comparisons not only to experimental signals, but also between GEANT3 and GEANT4, detailed and identical descriptions of the testbeam module and the environment are needed. 3.1. Geometry Description

Emphasis was put not only on a precise description of the calorimeter modules, but also relevant elements of the testbeam environment, like beam counters and other detectors used for trigger and particle selection, and upstream inactive materials. Figure 1 shows a part of the EMB testbeam module setup in the GEANT4 description.

lead absorber readout board

Figure 1. Th e ATLAS Electromagnetic Barrel Calorimeter in GEANT4. Th e setup includes the cryostat walls, the pre-sampler, cables and other components of the readout chain, all located at the inner (front) face of the module.

334

The EMB modules used in the testbeam are actual half-barrel production modules (27r/16 azimuthal segments). The accordion structure, as seen in Figure 1, is described in GEANT4 using a combination of boxes, trapezoids and tube segments, similar to the description in GEANTS. An alternative description with a new volume shape containing one accordion absorber is available in GEANT4 only. Both geometry descriptions produce basically identical signals. The HEC testbeam setup consists of three production front and rear modules, each of which is a 27r/32 segment of the HEC wheel. The description in both simulations includes details of the readout cell, as well as virtual lateral and longitudinal leakage detectors around the actual modules allowing to estimate the leakage energies for the testbeam setup. For the FCal, engineering prototypes were used in testbeam experiments. These so-called Module 0’s are full depth, quarter segments of the full FCal modules, and have only been built for the electromagnetic and first hadronic module. Each of the about 2300 - 2500 electrodes is individually described in GEANTS and GEANT4. They are positioned using the hardware cabling and machining databases. 3.2. Simulation Conditions

GEANTS and GEANT4 use slightly different parameter spaces to control the tracking and production of particles. In GEANT4, all particles are tracked to zero kinetic energy, while the production of secondaries is controlled by a minimum range requirement. This can of course be translated into a minimum energy requirement in a given material. In GEANTS, both tracking and production energy thresholds are implemented. The tracking thresholds for e* and y in the various studied setups are 100 keV for EMB and HEC, and 10 keV for the FCal. The secondary production thresholds are 100 keV for e* in EMB, 1 MeV for e* and y in the HEC, and 10 keV for e* in the FCal, and 10 keV for y in the EMB and FCal. In GEANT4, only one range threshold at a time is specified for all secondary particles and all materials in a given setup. In the EMB, 30 pm is used, which corresponds to an energy threshold of 112(41) keV for e* and lS(1.1) keV for y in lead(argon). The default minimum range in the HEC is 700 pm, which corresponds to an energy threshold of about 1 MeV for e* in copper. Most of the GEANT4 FCal electron signals have been simulated with 500 p m range threshold, which corresponds to about 17(4.4) keV in copper(argon). In addition, the dependence of the electron signal on the range threshold has been studied for the HEC and FCal setup, in the range of 1 . ..2000 pm. Figure 2 shows the expected rise in the signals in both calorimeters with decreasing range cut, reflecting the increasing number of particles produced in

335

26.6 26.4

-

26.2 26

25.8

""""

' """" 10

1 o2

'

1

""""

' " 1 o3 1o4 G4 ronae cut Ivml

Rmgc lllunl

Figure 2. The dependence of the simulated electron signal current in the HEC on the range cut in GEANT4 (left). The upper plot on the right shows the same dependence for the FCal, here measured by the electron sampling fraction. Also shown are the total deposited energy (right, center) and the relative energy resolution (right, bottom) in the FCal as function of the range cut. The gray bars indicate the corresponding GEANT3 predictions and their errors.

the copper and reaching the liquid argon, thus contributing to the signal. The unexpected drop in signal around 20 pm can be observed in both setups, and is presently under study. 4. Muon Signals in the EMB and HEC Modules

Signals from muons have been studied in the EMB and HEC. The signal spectrum for 100 GeV/c muons in the EMB is shown in Figure 3. Major differences especially in the tails of the distributions can be observed for GEANT3 and GEANT4. In general the later can describe the shape to about 1%over the whole signal range, which indicates a good description of the signal contribution from &electron production around the muon track. GEANT3 does not reproduce the tails to better than 3.5%. The simulated signals from both GEANT3 and GEANT4 have been smeared by 41 MeV Gaussian electronic noise, as determined in the experiment. Both GEANTS and GEANT4 can describe the muon signal spectrum in the HEC quite well. A Kolmogorov-Smirnoff test yields the same likelihood for both simulated spectra to be identical to the experimental one. The dependence of the tails in the spectrum on the range cut in GEANT4 has been studied as well, clearly indicating a rise in the signal contribution from 6electrons when lowering the range cut from 4 to 0.2 mm. The lower value then gives the best description of the experimental data.

336

..............

0.1

Energy in LAr (scaled) tGeVI

Energy bin (10 MeV bins, I* bin at -0.5 GeV)

0.2

0.5

0.4

0.5

0.6

0.1

0.8

0.9

1

Energy in LAr (scaled) IGeVl

c)

Energy bin (10 MeV bins, I* bin at -0.5 GeV)

Figure 3. The signal spectrum of 100 GeV/c muons in the ATLAS Electromagnetic Barrel calorimeter, as observed in a testbeam and simulated by GEANT3 (upper left) and GEANT4 (upper right plot). The corresponding cumulative differences between the simulated and measured distributions are shown in the lower plots.

5. Comparison of Electron Signals and Shower Parameters The electron response of the EMB, the HEC and the FCal has been studied extensively in testbeam experiments and with corresponding simulations. Relevant parameters used to estimate the quality of the Monte Carlo predictions are the energy dependence of the average signal, mostly introduced by inactive upstream material, the relative energy resolution, and the lateral and longitudinal shower spread. The electron signal in the EMB testbeam module, which is discussed in more detail in these proceedings', shows the expected linearity with energy to better than 1%. This is also well reproduced by the GEANT3 and GEANT4 simulations. The relative energy resolution of the testbeam module has a constant term well below 0.3%,and a stochastic term of about 9.2%. GEANT3 and GEANT4 simulations reproduce the stochastic term within the statistical errors, but only GEANT4 can predict the experimental constant term, again within errors. GEANT3 simulations show a significantly larger constant term of about 0.45%. An interesting observation in the EMB is that GEANT4 electromagnetic showers start later and are shorter (more compact) than the GEANT3 ones. This discrepancy increases with energy, and is measured by about 3% less signal in the first sampling ( w 4.3 X O )for 20 GeV/c electrons in GEANT4, and about 14% less for 245 GeV/c beam momentum. The signal in the second sampling (- 16 X o ) is between 1.5% and 2.5% higher in GEANT4, rising with

337

energy, and about 5% lower in the third sampling (- 2 X o ) , rather independent of the energy. The GEANT4 electron signal in the HEC is systematically about 3% smaller than the GEANT3 signal, if a 700 pm range cut is used. Lowering this cut t o 20 pm reduces this difference to slightly less than 1%) see Figure 4. As can be seen in the same figure, the energy deposited in the absorber is correspondingly larger, so that the total deposited energy is again very comparable. The experimental electron energy resolution is very well reproduced by both Monte Carlos. A.

4.4

3

;

4.3

w V

4.2

4.1

K

96

, 3

95.5

4 ...p e P...-0

A

v

.........

95

............ ,....... . .. ~. .. ... .....; ............. '........

.... ....

0 ....(;.............. ;.... ......... I;............... ..* ~

94.5

4t 94

K

0

.. . ....... .. ... ..

.. ...

...

140

; '

.i'

i.

.

.

.... .

20

40

60

80

100

120

.

.

100

G4 $700

. '

..

. ..

q)

n c.4 7",,rn

AA

t *

0

.

(

..

:

..

..

. ....

.. ....

.. ....

.

20

40

60

80

.

.

.

...

...

100

120

. ... 140 L.GeV

Summed response

Figure 4. The energy deposited by electrons in the liquid argon of the HEC calorimeter (left,top), the energy deposited in the absorber (left,center) and the total deposited energy (left,bottom), as function of the incident beam energy. T h e plots on the upper right show the relative difference between the GEANT3 (circles) and GEANT4 (triangles) electron signal and experimental data as function of the cell noise cut level, for 20 and 60 GeV/c electrons. The FCal cell signal significance spectrum for 60 GeV/c electrons is shown on bottom right.

The difference between GEANTS and GEANT4 simulated signals and experimental electron data has been studied in the FCal for various cell level

338

noise cuts. In general GEANT4 can describe the signal dependence on this analysis cut within &l%, while GEANTS shows up to 4% mismatch with the experimental data, see Figure 4. For this study, experimental noise has been added event by event, and channel by channel, to both simulations. The compactness of the electromagnetic shower in the electromagnetic FCal, which has only one longitudinal compartment, can be measured by the cell signal significance spectrum, also shown in Figure 4. This quantity is given by the signal to electronic noise ratio in each cell. The spectrum clearly shows that both Monte Carlos have less often highly significant signals in one cell, indicating their shower composition is softer. 6. Conclusions and Outlook

The results available so far for GEANT4 electron and muon simulations in the ATLAS liquid argon calorimeters suggest that the available electromagnetic shower and muon energy loss models in this toolkit are mature enough to understand signals in complex detector readout geometries and environments at about the same level as is possible with GEANTS. Major remaining discrepancies like the signal density are addressed and under discussion with relevant experts from the GEANT4 collaboration.

Acknowledgments We gratefully acknowledge the help and support from the ATLAS testbeam communities, the CERN staff, and the involved funding agencies. In addition, we especially like to thank M. Maire (LAPP Annecy) and J. Apostolakis (CERN), both from the GEANT4 collaboration, for their help.

References 1. see http://geant4.veb.cern.ch/geant4/geant4.html. 2. R. Brun and F. Carminati, G E A N T Detector Description and Simulation Tool, CERN Programming Library Long Writeup W5013 (1993). 3. A. Kiryunin et al., these proceedings. 4. ATLAS Liquid Argon Coll., Liquid Argon Calorimeter Technical Design Report, CERN/LHCC/96-41 (1996); ATLAS Coll., A T L A S Calorimeter Performance Technical Design Report, CERN/LHCC/96-40 (1996); ATLAS Coll., ATLAS Detector and Physics Performance Technical Design Report, CERN/LHCC/9914/15 (1999). 5. S. Rodier et al., these proceedings. 6. M. Fincke-Keeler et al., these proceedings. 7. K.K. Joo et al., these proceedings. 8. D. Zerwas et al., these proceedings.

DATA VOLUME REDUCTION STRATEGIES IN THE CMS ELECTROMAGNETIC CALORIMETER

P. PAGANINI LLR, Ecole Polytechnique, Palaiseau, France E-mail: paganiniOin2p3.fr The electromagnetic calorimeter of CMS consists of a barrel and two endcap calorimeters containing a sum of over 80000 lead tungstate crystals. If all the crystals were to be read-out in a triggered event, the total amount of ECAL data would excess by a factor 20 the CMS data acquisition system limits allowed for ECAL. This paper presents the strategies developed by CMS in order to reduce the ECAL data volume to the required level.

1. Introduction The ECAL electromagnetic calorimeter of CMSl consists of 75848 readout channels (barrel plus endcap). The data are a set of samples, each one being coded after digitization, on 2 bytes. With the present ECAL raw data format, the event size is2:

Event size (Kbytes) = 37.75 +

24

X Ncrystals

1024 where N c r y s t a l s corresponds to the number of crystals read-out for the whole ECAL in the event. In this formula, 10 samples per channel are assumed. However, preliminary studies3 showed that 7 samples are already enough to achieve a reasonable energy resolution and bunch-crossing identification. Using N c r y s t a l s = 75848, one gets an event size of 1.8 Mbytes. This value exceeds the total event size of 1Mbytes assumed in the DAQ design. Moreover, considering the 100 kHz level 1 trigger rate, a total bandwidth of 1440 Gbits/s would be needed to transfer these 1.8 Mbytes which is higher than the allowed value for the full CMS event builder: 500 Gbits/s. So, only 100 Kbytes per event are allocated to the ECAL data. To match this constraint, only 2656 crystals must be selected: 3.5% of the whole calorimeter. Two complementary approaches are usually considered to reach that reduction. The first one, called ”Zero suppression” is based on the rejection of informations on a channel by channel basis: only channels with energy above a certain threshold are kept. This approach is described in section 3.1. The

339

340

second approach, called ” Regional selective readout” defines regions of interest (event by event) of the calorimeter where all channels are read-out. In the other regions, the data are either removed or a severe ”zero suppression” may be applied. The details are given in section 3.2. The next section of this paper is dedicated to the simulation of the ECAL data used for this study and their reconstruction. 2. Simulation and reconstruction of ECAL data

The CMS ECAL calorimeter consists of lead-tungstate crystals. The electromagnetic shower in this volume is simulated by a software based on GEANT 3 (in a near future GEANT 4 will be used) producing crystal hits characterized by their energy and time. The next steps of the simulation (and reconstruction) are performed by ORCA4 (Object Oriented Reconstruction for CMS Analysis) software. The front-end electronic response results from the convolution of the Avalanche Photo Diode and the multislope amplifiers responses. In the simulation, the front-end response is factorized by an electronic signal shape. Quantization noise (coming from the multislope amplifiers and ADC) are taken into account in ORCA but have been neglected in the present study. The last stage corresponds to a 40 MHz sampling of the signal shape ’. The electronic noise is taken into account by adding to each sample a typical dispersion of 40 MeV in the barrel and 150 MeV in the endcaps. Correlations among the samples have been neglected in this paper. To simulate the pile-up contribution, several events per bunch crossing are superimposed. High luminosity cm-2s ) corresponds to about 20 minimum bias events. The crystal hits resulting from these events are mixed together and their signal shape are superimposed. The resulting signal shape is then sampled. The black curve of figure 1 shows the samples energy dispersion (in GeV) as function of the crystal pseudorapidity (the RMS being used as an estimator). The transition between the barrel and the endcap (at 7 = 1.48) is clearly seen. One sees that pile-up contributes mainly in the endcap at large pseudorapidity. The crystal energy measurement is directly related t o the amplitude of the signal. The amplitude measurement has to take into account the presence of electronic noise, pile-up events and possibility of time jitter fluctuations. However, since data volume reduction will be done on-line, the algorithm must be fast enough. The method used here is based on a linear filter: the signal amplitude corresponds to a weighted sum over the samples, the weights being determined by filter theory or x2 minimization and depend on signal shape,

-’

a40 MHz is the LHC beam crossing frequency.

34 1

Eta Figure 1. RMS of the samples energy distribution (in black) and RMS of signal amplitude distribution in GeV (light-grey: using basic filter, mid-grey: using optimized filter for pileup) as function of the pseudorapidity at LHC high luminosity. The value of the electronic noise used in the barrel is here 50 MeV and 150 MeV in the endcaps.

electronic noise and pile-up properties. When a filter optimized for electronic noise3 is used b , the RMS of the signal amplitude is given by the light-gray curve of figure 1, which follows closely the black one. To minimize pile-up effects, we have taken into account the correlations among samples introduced by pile-up (given by the autocorrelation of the signal shape itself3). An improvement of the RMS of the signal amplitude is shown by the mid-grey curve of figure 1. It is negligible in the barrel but can reach a few hundred MeV in the endcap. 3. Ecal data volume reduction

As already mentioned, our goal is to select only 3.5% of calorimeter channels without degrading physics performances. In CMS, we have studied 2 level of data reduction: the so-called ”zero suppression” and the ”regional selective readout”. that case, the weights depend only on the electronic noise dispersion and the signal pulse shape

342

3.1. Zero Suppression

With the ”zero suppression” approach, the selection criterion is quite simple: crystals are kept if their energy is larger than a threshold. On figure 2, the fraction of kept crystals (both in barrel and endcap) is shown as function of the threshold. The light-grey curve is obtained using filter optimized for electronic noise only3, and the black one is obtained when optimal weights are used to reconstruct the signal amplitude. The second method is more efficient (see figure 2): to keep less than 3 . 5 % of channels at high luminosity, the threshold has to be set to 110 MeV in the barrel (2.20electronic noise), and to 470 MeV in the endcap (Q.10electronic n o i s e ) .

0

0.2

0.4

0.6

0.8

1

Threshold (GeV) Figure 2. Fraction of calorimeter read-out channels as function of the threshold, using basic filter (in light-grey) or optimized filter for pile-up (in black).

But, applying a threshold could introduce several biases in physics quantities. For example, non-linearity effects are present in electromagnetic shower energy reconstruction as shown on figure 3 . This figure shows that for unconverted photons, the ratio between the reconstructed photon energy (obtained by a simple 5 x 5 crystals energy summation) and its initial simulated energy (called Kine on figure 3) is non linear when the energy rises. Applying a 30electronic noise threshold can introduce up to 1% non-linearity effect for photons energy between 40 MeV and 200 MeV.

343

i

Zero suppression 2 Sigma

i

i

A

Zero suppression 3 sigma

7

Zero suppression 4 Sigma

/

0.96

-

d; A

I , ,

) , , , I

,

I , . , I , . . , , . . l

, , I .

.

I

I

L

0 Energy, GeV Figure 3. Ratio between reconstructed photon energy and its true energy as function of the energy for different level of zero suppression.

3.2. Regional Selective Readout

The readout is based on trigger tower geometry. Two thresholds are considered. Towers are classified in the following way: when the transverse energy is greater than the high threshold, towers are tagged as center towers. The towers surrounding a center tower are neighbour towers whatever their energy is. Neighbour window size can be 3 x 3 or 5 x 5 towers; when the energy of a tower is between the high and low thresholds, the tower is classified as single tower. The other towers, called suppressed tower, not yet classified, are candidate for suppression. Two schemes have been studied in CMS. In the first one2, the neighbour window size is 3 x 3 towers, and the suppressed towers are not read-out at all. Moreover, a minimal zero suppression is applied everywhere: only reconstructed positive energies are kept (negative energy can occur because of noise fluctuation). Different threshold values have been tested as shown on figure 4. The best compromise is obtained when the high threshold is set to 2 GeV, and the low one to 0.6 GeV. With these parameters, the average event size is 95 K byt es . In the second scheme5, the two thresholds are higher: 5 and 2.5 GeV and

344 120 110

- 100

i-

8 80

VI

70 60 50

40 0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

LSH (CSV)

Figure 4. Event size (Kbytes) as function of the low threshold (GeV) for different zero suppression level (expressed as a number of electronic noise standard deviation) and high threshold.

the neighbour window is wider: 5 x 5. Here a strong zero suppression greater than 3.50electronic noise is applied but only in towers called suppressed. The average event size also fulfills the 100 Kbytes constraint: 95 Kbytes. Both schemes are not expected to introduce bias in shower energy reconstruction since regions where showers occurs remain untouched. 4. Conclusion

CMS electromagnetic calorimeter being a very fine granularity detector, its data volume at LHC high luminosity has to be reduced to an acceptable level. With the actual specification of ECAL raw data, only 3.5% of the calorimeter channels should be selected without loss of physics performances. The solution proposed in CMS is a combination of ”zero suppression” and regional selective readout based on trigger tower. The energy thresholds used in the selection have to be low enough so that no significant effect on physics performances is observed. References 1. ECAL TDR CERN/LHCC 97-33 (1997). 2. N. Almeida and J. Varela CMS-IN 2002/009 (2002). 3. P. Busson and P. Paganini CMS-IN 2002/006 (2002). 4. D. Stickland CMS-IN 1999/035 (1999). 5. S. Rutherford CMS-IN 2002/027 (2002).

PERFORMANCE OF CDF CALORIMETER SIMULATION FOR TEVATRON RUN I1

C . A. CURRAT Lawrence Berkeley National Laboratory, 1 Cyclotron road, Berkeley C A 94720, USA E-mail: CA CurratOlbl.gov (For the CDF collaboration)

The upgraded CDF I1 detector has collected first data during the initial operation of the Tevatron accelerator in Run 11. The simulation of the CDF electromagnetic and hadronic central and upgraded plug (forward) calorimeter is based on the Gflash calorimeter parameterization package used within the GEANT based detector simulation of the Run I1 CDF detector. We present the results of tuning the central and plug calorimeter response to testbeam data.

1. Introduction

The CDF I1 detector at the Tevatron is presently taking data after having undergone major upgrades to keep up with the new operating conditions at increased luminosity and center of mass energy1. A fast simulation of the calorimeters response is a valuable and implicitly recommended tool at a hadron collider, if one wants to gain control on the systematic errors on any high-pT related physics topic while accomodating the computing time limiting factor.

CDF calorimetry for run 11 The CDF I1 calorimeter^^^^^^ - covering the region 1171 < 3.6 - are of sampling 1.1.

type with separate electromagnetic and hadronic measurements as shown on Figure 1. In both sections the active elements are scintillator tiles read out by wavelength shifting fibers embedded in the scintillator. The calorimeters are segmented in q and 4 coordinates in order to have a projective tower geometry pointing back to the nominal interaction point. It has to be mentioned that the plug calorimeters have replaced the old gas calorimeters5 used during the first generation detector (referred to as CDF I).

345

346

Calorimeter type Central EM Central HAdronic Wall HAdronic Plug EM , Plug HAdronic

CHA (Fe) WHA (Fe) PEM (Pb)

Thickness 1x 19 4.5 A 4.5 x 21 1x 7x

xo,

xo,

Absorber/active 0.6 XO / 5 mm 1 in/ 10 mm 2 in/ 10 mm 0.8 X O / 4.5 mm 2 in/ 6 mm

Figure 1. Elevation view of one half of the C D F I1 detector with depth and sampling characteristics of the various calorimeter compartments.

1.2. The Gflash package Gflash is a fast calorimeter simulation using parameterized showers that is

interfaced with the standard GEANT-based' simulation of the detector. The program was developed by the H1 collaboration7 and is used as standard mass production Monte Carlo. Electromagnetic and hadronic showers are initiated when an incident particle undergoes an inelastic interaction inside the calorimeters. New GEANT pseudo-particle and track types are defined for the showers. Gflash uses a mixture-level GEANTgeometry for the calorimeters, i.e. the showers develop in a medium with effective density, 2, A, X O etc. It handles the energy loss of real particles inside the mixture geometry. For the electromagnetic longitudinal profiles, a standard gamma distribution is used:

where x = P z , z being the shower depth in units of X O . The average and width of CY and parameters have a typical logarithmic energy dependence. The correlation between a and P is properly taken into account when a shower

347

is generated:

with matrix C describing the correlation between (Y and B determined out of GEANTresults. The terms q1 and q2 are normal-distributed random numbers around the mean values pa and pp. The lateral profiles are described by the following expression:

where & is a free parameter that is a function of the shower energy and depth. The average profile and fluctuations of individual showers are described as (approximately log-normal) functions of the energy and shower depth. Distinction is made between the core and the increasingly slower growth of the radial extent of the shower with increasing energy. The parameterization of the hadronic longitudinal profiles is a superposition of up to 3 gamma functions H , F, L to correctly reproduce the 7ro energy dependence:a

Here the (Y and O, parameters for all the gamma distributions are allowed to fluctuate and their correlations are properly taken into account. Furthermore, to get the no fluctuations right, showers are divided into 3 classes, each with a given probability. The sampling fluctuations are reproduced by depositing a Poissondistributed number of spots per longitudinal integration interval according t o the radial probability function equation 3. Their energy and hence their number are given by the energy resolution that has to be reached. These energy spots are similar to GEANT“hits” or “visible” energy depositions. In a detailed GEANTcalorimeter geometry only the energy deposited in sensitive layers, as scintillators, are recorded as hits. For mixture-level GEANT calorimeter geometries, as is the case in Gflash, one needs to simulate sampling fluctuations and to explicitely convert the deposited energy Edep into “visible” energy Evisin the active medium

aNamely, H for the purely hadronic component of the shower, F for the no fraction originating from the first inelastic interaction, and L for the no fraction from later inelastic interactions occuring in the shower development.

348

where riz denotes the sampling fraction for minimum ionizing particles (MIP), and are the relative sampling fractions for electrons and hadrons, and respectively. The sum is over the electromagnetic (IC = e) and the purely hadronic ( k = had) components, with relative fraction Ck. The azimuthally symmetric spatial distribution function is given by fk(3= &fk(.z) . f k ( r ) . Further details on the implementation of Gflash can be found in the original paper7. In the following it is shown that by tuning the same shower parameterization in Gflash it is possible tQreproduce the physical response of the different sub-detectors forming the CDF I1 calorimetry.

1

1.3. Geneml method

~ O T tuning

Gflash

The tuning of Gflash is split according t o the two sets of parameters that control on the one hand the fraction of visible energy produced in the active medium and on the other hand the energy and depth dependent spatial distribution of the various components of the shower model. All along the tuning procedure, the consistency between data and simulation is assessed with a standard x2 estimator. The tuning of hadron showers proceeds as follows: (1) The first step consists in reproducing the position and width of the MIP peak. A few parameters at a time are involved (riz,ah for each specific volume) as the simulation is still handled by GEANTat this stage. The width of the peaks is controlled by adjusting the amount of intrinsic and sampling fluctuations in the active medium. (2) A sample of 7r+ with an incident energy of 57 GeV is used next to set the energy scale for the peak of full energy deposition (involving and ck).

(3) The energy dependence of the fk = fk(E)parameters defining the fractions of deposited energy is tuned to accomodate the data at all energies. This step is basically iterative once a partial deconvolution of Ansatz equation 5 set “pivots” typically being two points with a large energy gap. (4) The tuning of the lateral profile is performed almost independently of the tuning of the longitudinal energy dependence. 2. Tuning Gflash with testbeam data

The Gflash package is tuned using testbeam data8 including electrons and pions with energies in the range 8 < E < 230 GeV.

349

Focus will be made on the tuning of hadronic showers, as being the most involved part of the description. The tuning of the electromagnetic showers, although obviously valuable, presents less difficulty.

2.1. The MIP peak The MIP peak is measured in the electromagnetic compartment with pions. Besides of the position and width of the peak, the ratio of the peak content to the total number of measured pion showers provides an additional handle on the tuning, as it can be directly inferred from the geometry of the detector (see absorbtion/interaction lengths in Figure 1). Figure 2 shows the comparison between testbeam data and Gflash simulation for 57 GeV pions in the CEM calorimeter. The same level of agreement is reached in the plug calorimeter. 2.2. Setting the energy scale The hadronic energy scale in the tuning procedure is set by the response of 57 GeV pions both in the central and plug calorimeters. A set of reference plots shown on Figure 4 is used as a reference for adjusting the parameters steering the longitudinal development of the hadronic showers in the combined calorimeter compartments. Plots (a)-(d) respectively show the distribution of the measured total energy (electromagnetic+hadronic) for pions that are minimum ionizing in the PEM, the measured total energy for all the pions, the hadronic fraction of the deposited energy as measured in the PHA and the electromagnetic fraction of the deposited energy as measured in the PEM. The x2 estimator confirms that data and simulation are in fair agreement for each distribution. The same level of agreement is reached for hadronic showers in the central calorimeter. The purely electromagnetic showers are simultaneously well reproduced as illustrated on Figure 3. Plotted is the energy over momentum ratio for 11 GeV electrons. 2.3. Adjusting the energy dependence

Once the MIP distribution and the hadron energy scale are set, the other datasets can be reproduced by tuning the logarithmic dependence in energy of the shower components. Given the significant difference between the central and plug hadronic calorimeters construction, in particular when comparing their sampling structure (see Figure 1), two distinct parameterizations of the energy dependence for the fractions of deposited energy are kept ( c h , c f , ce in Equation 4).

350 300

57 GeV Pions

250 -

t 200 r

150L

100-

0.7

CEM Energy [GeV]

Figure 2. Peak for minimum ionizing par. ticles in the GEM for 57 GeV pions.

0.8

0.9

1.0

1.1

E(PW~

Figure 3. Ratio E / p =(calorimetric energy)/(beam momentum) for 11 GeV electrons in the PEM.

180 140 120

80 60

40

20 OO

10

20

30

40

w

w

70

80

Figure 4. Comparison of the energy profiles between testbeam data and simulation using Gflash showers parameterization for 57 GeV pions. (a) Total energy for pions that are minimum ionizing in the PEM (see text). (b) Measured total energy for all the pions. (c) Hadronic fraction Ehad of the deposited energy as measured in the PHA. (d) Electromagnetic fraction Eem of the deposited energy as measured in the PEM. Abcissas are in [GeV].

351

10

lo*

p [GeVk]

10

lo2

p [GeWc]

Figure 5 . Comparison of the linearity in the central hadronic calorimeter (CHA) between testbeam data and Gflash simulation. Top: Average ratio E / p =(calorimetric energy)/(beam momentum) as a function of momentum. Bottom: Width of the E / p distribution versus momentum.

On Figure 5 (top) the mean value ( E / p ) of the ratio of the measured total energy over the momentum of the incident particle in the central hadronic calorimeter is plotted versus momentum. Since the scintillator-based plug calorimeter is not compensating, the response is non-linear with the pion energy. The bottom plot shows the corresponding width a ( E / p ) / ( E / p )of the distributions. Both the peak position and the energy resolution are quite well reproduced by the current tuning of Gflash over the momentum range 8 < E < 230 GeV. The same level of agreement is reached for electromagnetic showers over the same range of momenta. Error bars are displayed on both plots although mostly covered by the marker size. The main contribution comes from the uncertainty on the determination of the beam line parameters.

352 2.4. Lateral profile

For the tuning of the lateral profile, the data consist of tracks obtained from minimum bias events and measured in the central part of the CDF I1 detector. The available energy range spans from 0.5 GeV to 2.5 GeV. At higher energies, tuning follow-ups are performed with the increased amount of data becoming available. The tuning of the lateral profile can mostly be done independently of the longitudinal one. The approach is typically iterative. Figure 6 shows the comparison between the minimum bias data and the Gflash simulation. The histogram of the simulation is normalized to the data to avoid a longitudinal dependency.

0.14

0.12 0.1 0.01)

0.06

0.M 0.02

0

Figure 6. The tuned plots of the lateral profile for CHA using tracks from minimum bias events. Th e lateral profile is expressed in units of pseudo-rapidity relative to the center of the tower the incident track is pointing to.

1 0

25

50

75

100

125

150

175 200

225

250

inc. particle momentum IGeV/cl

Figure 7. Comparison of cpu time consumption between GEANT and Gflash hadronic shower simulation.

3. Remarks and conclusions The typical gain in CPU time using the Gflash parameterization with respect to the standard GEANTsimulation to generate hadronic showers is up to a factor 100 as shown on Figure 7. This work allowed the integration of a fast shower parameterization based on the Gflash package in the CDF I1 simulation framework. The calorimeters response is reproduced in an unified description over the close to 47r coverage of the detector.

353

Acknowledgments The main author would like to thank the Swiss National Science Fundation (SNSF) for its support.

References 1. “The CDF I1 Detector Technical Design Report”, FERMILAB-Pub-96/390-E (1996). 2. L. Balka et al., “The CDF central electromagnetic Calorimeter”, Nucl. Instrum. and Methods A267,272 (1988). 3. S. Bertolucci et al., “The CDF central and endwall hadron calorimeter”, Nucl. Instrum. and Methods A267,301 (1988). 4. Ryutaro Oishi on behalf of the CDF Plug Upgrade Group, “New CDF end-plug calorimeter”, Nucl. Instrum. and Methods A453, 227 (2000). 5. S. Cihangir et al., “The CDF forward/backward hadron calorimeter”, Nucl. Instrum. and Methods A267,249 (1988). 6. R. Brun and F. Carminati, “GEANT Detector Description and Simulation Tool”, CERN Programming Library Long Writeup W5013 (1993). 7. G. Grindhammer, M. Rudowicz, and S. Peters, “The Fast Simulation of Electromagnetic and Hadronic Showers”, Nucl. Instrum. and Methods A290,469 (1990). 8. G. Apollinari et al., “Test beam results from the CDF plug upgrade calorimeter”, Nucl. Instrum. and Methods A409,547 (1998).

SIMULATION OF HADRONIC SHOWERS IN THE ATLAS LIQUID ARGON CALORIMETERS

A.E. KIRYUNIN', D. SALIHAGIC~, P. S T R ~ Z E N E C ~ Max-Planck-Institut fur Physik, Werner-Heisenberg-Institut, Fohringer Ring 6, 80805 Munchen, Germany

J. KISH Institute of Experimental Physics of the Slovak Academy of Sciences Watsonova 47, 04353 Kosice, Slovakia

P. LOCH University of Arizona, Tucson, Arizona 85721, USA

R. MAZINI University of Montreal, Montreal, Qc, H3C 3J7, Canada

Results of Geant4 based simulations of the response of the ATLAS hadronic endcap calorimeter t o charged pions are presented. T h e first results of hadronic simulations with Geant4 for the ATLAS forward calorimeter are shown as well. Predictions of Geant4 and Geant3 on energy response and resolution for charged pions are compared. Where it is possible, the comparison with experimental results of beam tests is done.

1. Introduction

The ATLAS Collaboration is building a multi-purpose detector for the future experiment at the Large Hadron Collider at CERN. The software is very important for the whole experiment's success. To assure its reliable maintenance over the long lifetime of about 20 years, the ATLAS Collaboration has decided to move from Fortran-based to object-oriented software. In this transformation it has started to use the Geant4 package' for the simulation of particle passage through matter. * On leave of absence from Institute for High Energy Physics, Protvino, Russia

t On leave of absence from University of Montenegro, Podgorica, Yugoslavia

* On leave of absence from Institute of Experimental Physics, Kosice, Slovakia

354

355

Geant4 is a freely distributed simulation toolkit, based on software engineering techniques and object-oriented technology. The package was developed in 1990’s by a world-wide collaboration, and since 1999 (after first production releases) it is maintained by the international Geant4 Collaboration. The implementation of the ATLAS detector in the Geant4 framework is going on. Part of this process is the validation of the physics models in Geant4, carried out for different ATLAS sub-detectors. This talk summarises the current status of the validation of Geant4 hadronic physics, done so far for the two liquid argon (LAr) calorimeters appropriate for the corresponding studies, namely the forward calorimeter (FCal) and the hadronic end-cap calorimeter (HEC). The validation is based on the comparison of Geant4 predictions with experimental results, obtained during beam tests of calorimeter modules. Comparisons to Geant3 simulations are also available. Geant32 is the simulation package, based on the “old” Fortran software. It was heavily used over last years for analysis and understanding of ATLAS test-beam data. 2. Forward Calorimeter

The forward calorimeter in ATLAS3 is a LAr sampling calorimeter consisting of three longitudinal modules. The first (electro-magnetic) one has copper as absorber, the two following hadronic modules have tungsten absorbers. The calorimeter covers the pseudorapidity region 3.1 < 1111 < 4.9. In 1998 quarter segments of the first two FCal modules have been tested in particle beams at CERN. Evaluation of the calorimeter performance requires the comparison of experimental data, obtained at beam tests, with detailed Monte Carlo simulations. Description of the FCal test-beam setup is done in Geant4 and Geant3 exactly in the same way. This guarantees that possible difference in predictions of both simulation packages is not due to the different geometries. See Reference4 for more details on the FCal simulation framework. First very preliminary results of Geant4-based simulations of the response of the two FCal calorimeter modules to charged pions are available. For these simulations version 4.0 of Geant4 was used. The range cut was set to 20 pm. In Figure 1 reconstructed energy distributions for 20 and 60 GeV pions are presented. They clearly indicate a larger signal in Geant4 than in Geant3 (about 19 % at 20 GeV and 14 % at 60 GeV). The relative signal fluctuations are smaller in Geant4 (-18 %) than in Geant3 (-26 %) for 20 GeV, a trend that is inverted at 60 GeV, with -15 % for Geant3 and -18 % for Geant4. This can be attributed to the larger tails in case of Geant4 at this energy. Computer time, necessary for simulations, is an important performance parameter of software products. Simulations of charged pions for FCal beam

356 n. 20 GSV 0.07

om

0.W

0.05

om

0.m

0.0)

0.03

0.03

0.02

0.02

0.01

0.01

0

1.3

3

10

12.3

15

17.3

20

22.3

0 ~

25

O

~

U

I

0.0s 0.01 0.07

L

)

--

M

~

&ant4

<E>

~~~~~

0.08

a

om 0.01

0.03 0.02 0.01 0 6

8

10

12

14

I6

10

?D

22

21

5. (GeV

0.M 0.03 0.02 0.01 0 20

m

UI

L)

M

49.04 GeV 8.7 GeV

70

M

L(W

Figure 1. Distributions on reconstructed (at electro-magnetic scale) energy for 20 GeV (left-hand plots) and 60 GeV (right-hand plots) charged pions, obtained with Geant3- and Geant4-based simulations.

tests with Geant4 are approximately 10 % faster than with Geant3 (for the used values of simulation cuts). 3. Hadronic End-Cap Calorimeter

The ATLAS hadronic end-cap calorimeter3 is a copper LAr detector with parallel plate geometry. It has four longitudinal layers and covers the pseudorapidity region 1.5 < 1771 < 3.2. In 1998-2001 extensive beam tests of pre-production and serial modules of the HEC were carried out at CERN. Various detailed analyses of large amounts of experimental data and comparisons with predictions from Geant3 version 3.21, allowed to evaluate the performance of the hadronic end-cap calorimeter. Main results of this work are summarised in Reference5. To validate physics of Geant4, simulations of the HEC test-beam setup were carried out. As for the forward calorimeter, geometry description of the setup was done as close as possible in Geant3 and Geant4. See Reference6 for more details on the HEC simulation framework. For charged pion simulations versions 3.2 and 4.0 (with patch 1) of Geant4 were used. The range cut was set to 700 pm. This corresponds to a 1 MeV energy threshold for particle production in copper, similar to the one applied in the Geant3 simulations. The default physics list, as provided by the Geant4 Collaboration for hadron simulations, was used. Within Geant3 GCALOR as a hadronic shower code was used. The response of the HEC modules to charged pions of different energies from 6 to 200 GeV was simulated. The

O

M

357

experimentally available energy range is 10-200 GeV. The Geant4 simulations are faster than Geant3, by a few per cents. One of the important characteristics of a sampling calorimeter is the amount of energy deposited in the active and passive parts. In Figure 2 the relative (with respect to the beam energy E B E A M response ) of LAr and the summed response (copper and LAr) are shown as functions of the beam energy. Certain problems are seen in the region of 20-50 GeV: instead of smooth behaviour, there are some kind of steps. This effect is present in both used versions of Geant4. 0 Geant4-4.0 A Geant4-3.2

2 ''9

Q

6

8

€?

6

P

1 ~ ~ ~ ~ 1 " " " " ~ " ~ " " 1 " " 1 " ~ ~ ' ~

80

a

im3

@

6

Qdea

7a

76

a

@

175

200

74

w \

A

W V

0

25

50

75

100

125

150 EBEAM

IGeVl

Figure 2. Relative response of LAr (top plot) and summed response of copper and LAr (bottom plot) as functions of the beam energy, obtained with different versions of Geant4.

The problem was recognised by Geant4 experts as a mix-and-match problem in the energy deposited by energy loss of charged particles in a pion shower. As a solution it was proposed to change the value of the maximal (minimal) energy for the low (high) energy pion model from 25 to 60 GeV (from 20 to 50 GeV). In Figure 3 the summed response is shown for the tuned version of Geant4. Evident improvement of the behaviour is observed. But there are still some small deviations around 50-60 GeV. In this Figure predictions of Geant3 are shown as well, which demonstrate good smoothness of the corresponding dependence. Another important calorimeter characteristic is the energy resolution. To

358

84 82 80 78 76

70 68 66

0 Geant4-4.0 0 Geant4-4.0 (tuned) 0 Geant3

-@

~ , , , , 1 , , , , , , , , , 1 , , , , 1 , , , , , , , , , 1 , , , , 1 , , , , 1

0

25

50

75

100

125

150 EBEAM

175

200

[GeV]

Figure 3. Relative summed response of copper and LAr as a function of the beam energy, obtained with different simulations.

reconstruct the energy, clusters of calorimeter cells are defined. The reconstruction procedure applied to simulated data is identical to the one used for experimental data. For direct comparison the electronic noise contributions were subtracted in quadrature from the experimental resolution. In Figure 4 energy dependences of the resolution, obtained experimentally and with Geant3and Geant4-based simulations, are presented. The energy dependence of the resolution can be parameterized by the following two term formula:

with a sampling term PI and a constant term Pz. Geant4 predicts a worse energy resolution than observed experimentally, while Geant3 gives better values than the experiment. At the same time deviations of Geant4 and Geant3 predictions from the experiment are at the same level. The HEC is a non-compensating calorimeter, which means that its response to electrons is different from its response to charged pions at the same energy. In Figure 5 the ratio of reconstructed energies of electrons and pions (e/7rratio) is shown as a function of the beam energy. It can be seen that Geant4 describes the experimental dependence significantly better than Geant3. The HEC has four longitudinal layers. Studying the energy depositions in those allows t o evaluate the longitudinal development of hadronic showers. Comparison of experimental results with predictions of simulation packages shows that both Geant3 and Geant4 are in reasonably good agreement with

359

30 0 Geant4-4.0(tuned)

,y2/ndf PI

36.59 / 10 75.20 f

1.262

150

200

25 20 15 10 5

0

L

0

25

50

75

100

125

175

Figure 4. Energy resolution of HEC modules as a function of the beam energy, obtained experimentally and with different simulations. T h e lines give fits with Equation (1)

\

Q)

1.5 1.45 1.4 1.35 1.3 1.25 1.2 1.15 1.1 1.05 1

0

20

40

60

80

100

120 %,AM

140

160

[GeV]

Figure 5. e/n-ratio as a function of the beam energy, obtained experimentally and with different simulations.

the experiment. In simulated data more detailed information about longitudinal energy depositions is available. Studies show that an average position of showers along the beam direction is closer to the calorimeter front face in Geant4 than in Geant3, at least for beam energies below 80 GeV. At higher energies average positions are very close.

360 4. Conclusions

First attempts to simulate the response of ATLAS hadronic LAr calorimeters to charged pions using the Geant4 package have been successful. After some tuning to match more accurately the low and high energy pion models, rather reasonable results were obtained. For the hadronic end-cap calorimeter, Geant4 results for the energy resolution, e/x-ratio, and the longitudinal shower profiles, are in agreement with experimental observations, at the level of a few percent. With this respect they are very comparable to Geant3 predictions. Further work on the validation of the Geant4 hadronic physics using simulations for ATLAS LAr calorimeters, is foreseen in the close collaboration with the Geant4 team. 5. Acknowledgments We would like to thank Hans-Peter Wellisch, a Geant4 expert, for the collaborative work in understanding and tuning of the hadronic code.

References 1. http://geant4.web.cern.ch/geant4/ RD44 Collaboration, “GEANT4: An Object-Oriented Toolkit for Simulation in HEP”, CERN/LHCC 98-44 (1998). 2. R. Brun et al., “GEANTS”, CERN DD/EE/84-1 (1986). 3. ATLAS Collaboration, “Liquid Argon Calorimeter Technical Design Report”, CERN/LHCC/96-41 (1996). 4. R. Mazini and P. Loch, in: Proc. IX Int. Conf. on Calorimetry in High Energy Physics (Annecy 2000), Frascati Physics Series, Vol. XXI, 485-491 (2000). 5. ATLAS Liquid Argon HEC Collaboration, Nucl. Instr. and Meth. A482,94-124 (2002). 6. A.E. Kiryunin, D. SalihagiC, P. Striienec, in: Proc. IX Int. Conf. on Calorimetry in High Energy Physics (Annecy 2000), Frascati Physics Series, Vol. XXI, 459-466 (2000).

MC SIMULATION OF THE ATLAS HADRONIC CALORIMETER PERFORMANCE

M. J. VARANDA FCUL and LIP, Av. Elias Garcia, 14, 1, 1000 Lisboa, Portugal E-mail: [email protected] [On behalf of the Tilecal/ATLAS collaboration)

Several MC Studies of the Tile Hadronic calorimeter (Tilecal) using GEANTJ and GEANT4 have been done after tunning the code with data from tests with high energy particle beams at CERN. The comparision between the two codes started with the study of the simulation of the electromagnetic interactions and results are presented. A preliminary study of the evaluation of the simulation of the hadronic interactions is also presented.

1. Introduction The ATLAS detector is composed of several subsystems. One of them is the Tile calorimeter (or Tilecal) which is the hadronic calorimeter in the central region of the detector (I 7 151.7). The Tilecal is a sampling calorimeter made of steel as absorber and scintillating tiles readout by wavelength shifting fibers as active medium. Basically, it is a periodic structure made of steel plates alternating with scintillating tiles along the beam axis (z) and in depth (R), figure 1. It is planned to detect and measure many different physics processes which require the ability t o measure a wide range of energies, from a few hundred MeV to the TeV scale. This requires a very careful construction and accurate knowledge of the calorimeter. The modules of the calorimeter are simulated in good detail1 in Geant3 (G3) and several simulations of their response to high energy particle beams have been done and compared with data from the tests of those modules. The simulation of the calorimeter in Geant42 (G4) started a few years ago and the response of the calorimeter to e.m. interactions have been compared with the predictions of Geant3. The study of the response of the calorimeter to hadrons started just during the 2001 summer.

361

362

Figure 1. Principle of the Tile Calorimeter. On the bottom it is the test beam setup in 2001: Two barrel modules and two extended barrel modules organized in three calorimeter layers.

2. Tilecal cell intercalibration The intercalibration of the Tilecal cells is done by running a 137Cs source inside a tube filled with water which is connected to a pump that keeps the liquid in motion. The tube passes through each tile and iron plate of the calorimeter. The mean free path of the 662 MeV photons emitted by the source is of the order of a period (18 mm), i.e, the distance between tiles (figure l), which allows to see the individual tiles and steel plates. Figure 2 shows the current produced in the PMT as a function of the source position when the source runs inside the tube with a constant speed. The peaks and valleys structure shown is due to the passage of the source through alternating layers of scintillating tiles and steel plates, respectively. The test beam data is compared with Geant3 simulation which agrees with the data with a precision of few a %. 3. Response to muons

The response of the Tilecal modules to muons was largely studied in many test beam periods. Muon beams are used in tests of Tilecal modules to intercalibrate the cells of the calorimeter and to study its uniformity.

363 C s D a t a a n d M C S i m u l a t i o n in C e l l A . 4 . T i i e 2

Data: module 24 (best RMS)

Tile No.

Data: module 7 (worst RMS) 0 I04

MC: individual tile coefficient applied 10s

ion

I10

I I2

I I#

Tile No.

Figure 2. Current induced in the PMT when the tiles of a cell of the calorimeter.

13’Cs

source passes though the stack of

Figure 3 shows the response of the calorimeter to 20 GeV muon beams entering the detector at q=-0.35 and 180 GeV muon beams entering at 90°, as predicted by G3 and G4. The observed differences between the two codes are of the order of 2-3% for 20 GeV muons and up to about 5% for 180 GeV muons. For 20 GeV muons, within the experimental errors, there is good agreement between Monte Carlo and data, not sensitive to the small differences between the two codes. 4. Response to electrons

The resolution for electrons is also shown in figure 4 as predicted by both codes and as measured in the test beam. In the left plot, the experimental effects include the electronic noise that was introduced with a gaussian smearing in the simulated data. The results of the non-smeared simulation are also shown. Neither optic non-uniformities nor other experimental effects are included in the simulation. The plot shows a good agreement between the predictions of G3 and G4 but both predict a stochastic term 4% better than the data. This is most probably due to the

364

Figure 3. Spectra produced by 20 GeV muons entering the calorimeter at q=0.35 (left plots) and 180 GeV muons entering at 90’ (right plots).

absence of many real fluctuations in the data that were not included in the simulation and maybe a better description of the electronic noise is also needed. In the right plot, the ratio between the energy deposited by electrons in the active material and the total energy lost in the calorimeter for beams of muons hitting the calorimeter at 90’ shows a predicted G3 signal about 1% higher than the predicted by G4. Figure 5 shows the foreseen e.m. showers in the Tilecal for 100 GeV electrons at 90°, that in G3 start later and are slightly wider than in G4. The differences cannot be checked with the Tilecal because its coarse segmentation does not allow to see which code describes better the e.m. interactions in the calorimeter. 5 . Response to hadrons

The response of the Tilecal to hadrons is shown in figure 6 and is compared with simulated data using several different hadronic packages. For the hadronic showers, the Fluka-standalone code was compared with Fluka running with G4. The e/h ratio predicted by Fluka in G4 is 1.26 while the predicted by Fluka-Standalone is 1.16 for energies between 5 and 100 GeV. The test beam data predicts e/h=1.36f0.013 for energies between 10 and 300 GeV. Gheisha re-writen for Geant4 was compared with Gheisha in Geant3. The resolution to pions predicted by Gheisha in G3 is about 3-4% higher than by

365

= Oeallt4 n Owin 3

I

.

I

. ....

,I

. .

I

I

. . I

Figure 4. Response to electrons at 90'. Left plot: Energy resolution with electronic noise included by a gaussian smearing in the simulated data. The results of the no-smeared simulation are also shown. Right plot: Ratio of energy deposited in scintillator to the total energy deposited in the active and absorber materials.

Figure 5. Geantd and Geant4 predictions for the e m shower development of 100 GeV electrons entering at 90'.

Gheisha rewriten for G4. The test beam data is in between the predictions of the two MC codes. 6 . Summary

In general there is a good agreement between the two MC codes and the data. Both codes describe well the interactions of high energy muon beams with the calorimeter. The agreement between them is better or equal to 5% depending on the beam energy. The codes predict slightly different characteristics for the e.m. showers produced by electrons entering the calorimeter at different angles. For electrons entering the calorimeter at 90' the e.m. shower predicted by G3 starts slightly

366

1.4

6

3

1.35 1.3 1.z

I.? 1.15 1.1

.m

I

I

Figure 6. Left plot: The e/h ratio, Fluka-Standalone (Fluka) is compared with Fluka running in Geant4 (Geant4). Right plot: Resolution for pions entering the calorimeter at 20°. Gheisha in G3 (Geant-Gheisha) is compared with Gheisha re-writen for G4 (Geant4).

later and is slightly wider than the predicted by G4. The segmentation of the calorimeter does not allow to check which code describes better the data. For the hadronic showers, the e/h ratio predicted by Fluka in Geant4 is 1.26 while the predicted by the Fluka-standalone code is 1.16 for energies between 5 and 100 GeV. The test beam data predicts e/h=1.36f0.01 for energies between 10 and 300 GeV. Gheisha rewriten for Geant4 was compared with Gheisha in Geant3. The stochastic term of the resolution to pions predicted by Gheisha in G3 is about 3-4% higher than the predicted by Gheisha rewriten for G4. The test beam data is in between the predictions of the two MC’s. Acknowledgments

The author would like to acknowledge the Tilecal/ATLAS collaboration and specialy to R. Leitner, A. Solodkov and A. Henriques for the useful discussions and comments t o this presentation. The participation in the conference was supported by ICCTI and FCT, Portugal. References 1. T . Davidek, New Tilecal Beam test simulation within Dice 95/Geant3 Framework,

ATL-TILECAL-2000-13. 2. A. Solodkov, Oral presentation in the Geant4 review, 2001 M. Bosman, ATLAS, Lund Physics Workshop, 2001 3. J. A. Budagov et al., The e/h ratio of the ATLAS Hadronic Tile Calorimeter, ATL-TILECAL-2001-001.

SIMULATION STUDIES OF THE JET AND MISSING TRANSVERSE ENERGY PERFORMANCE OF THE ATLAS CALORIMETERS

M. WIELERS TRIUMF, 4004 Wesbrook Mall, Vancouver, B.C., Canada, V6T2A3 E-mail: Monika. WielersOcem.ch (On behalf of the ATLAS Collaboration)

The measurement of jets and missing transverse energy reconstruction will play an important role for many physics channels at the Large Hadron Collider (LHC). The performance of the ATLAS detector for reconstructing jets and missing transverse energy has been evaluated using detailed simulations. In this paper results based on these simulations will be shown for the jet energy resolution, in addition t o some selected examples of the simulated jet and missing transverse energy physics performance. Special emphasis will be put on the experimental aspects like electronic and pile-up noise, non-compensation, and dead material, as well as their realisation in the simulation.

1. Introduction

Reconstructing jets and missing transverse energy (EFiss)will play an important role in LHC physics. Jets will be used in many different ways, for example the jet multiplicity and the ET spectrum will be measured in the context of QCD, SUSY and other models. They will be used for reconstructing resonances like W + j j , Z + bb, and t + bW decays, as well as in searches for possible SUSY or exotic particles for example in A / H -+ TT or W’ + j j decays. For some studies such as the heavy Higgs production via the Vector Boson Fusion process a central jet veto down to transverse momenta of p~ = 15 GeV and jet tagging at large rapidities 1 ~ > 1 2 is required. Large EFiss will be an important signature for new physics, for example H + Z Z + lluu. It will be as well used to reconstruct invariant masses in decays involving neutrinos such as A / H + TT or t + lub. More details on some selected physics channels will be discussed in section 5 and 6. A detailed overview can be found in1’.

367

368 2. The ATLAS Calorimeter System

The ATLAS calorimeter system is based on the liquid argon technology (LAr), with the exception of the hadronic calorimeter in the barrel. A schematic

ATLAS Calorimetry (Geant)

Figure 1. Overview of the ATLAS calorimeter.

overview is shown in figure 1. The electro-magnetic calorimeter is a Pb/LAr calorimeter with accordion geometry. It covers the rapidity region 1171 < 3.2. The total thickness is around 24 radiation lengths or 1.2 interaction lengths. To correct for the energy loss in the material in front of the calorimeter a preshower detector is preceding the electro-magnetic calorimeter. The barrel hadronic calorimeter (1.11 < 1.7) is a scintillating tile calorimeter which uses iron as passive material. The end-cap is covered by a hadronic Cu/LAr sandwich calorimeter (1.5 < JqJ< 3.2) followed by the forward calorimeter (3.1 < 1771 < 4.9), which is a dense LAr calorimeter with rod-shaped electrodes in a tungsten matrix. The total thickness of the hadronic calorimeter system is M 11 interaction lengths.

369

3. Simulation of Detector Response

For the performance studies presented here, events were generated using the Pythia event generator3. Subsequently, the ATLAS detector was fully simulated using GEANT34, which is to date still the “standard” simulation program in ATLAS. Studies are ongoing to fully validate GEANT4 and in time ATLAS will move to this simulation program. An important issue is to simulate correctly the effect of pile-up in the calorimeters taking into account the read-out time. At LHC on average 23

0

100

200

300

400

500

f 0

Time (or

Figure 2. Signal shape as produced in the detector (triangle), and after shaping (curve with dots). The dots represent the position of the successive bunch crossings at LHC.

minimum bias events will be present per bunch crossing at design luminosity ( L = 1034m-2s-1 ). In the calorimeters the read-out time is much longer than the bunch spacing time of 25 ns. In the LAr calorimeter system, the bipolar shaping functions have a response to a triangular signal that lasts for over 500 ns (see figure 2). They are designed so that the noise will peak at zero energy. The read-out in the tile calorimeter is much faster. The response is a unipolar shaping function which is approximated by a Gaussian distribution with a FWHM of 50 ns. Thus, to generate one signal event at design luminosity around 700 minimum bias events need to be added. Currently to generate these events in ATLAS two methods are used. In both methods the minimum bias events are read via a secondary stream. The first approach is to add pile-up in a fast way at reconstruction time to a given signal event via a direct access file which is kept in memory. This method only works for the calorimeter, and in case where correlations with other detectors need to be taken into account a much slower method is used. In this case pile-up is added in an additional simulation step for all detectors. Here the GEANT hits coming from the main bunch crossing are added for all ATLAS detectors. In addition, a third data

370

stream takes care of pile-up events from very early or late bunch crossings, which are not seen by all detectors. At present the pile-up simulation is done using FORTRAN, but work is ongoing to implement a similar mechanism in our offline framework which is written in C++. 4. Jet Reconstruction

The aim in jet reconstruction is, in most cases, to extract from the energy deposits in the various calorimeters the energy of the initial parton. To do so, many factors have to be taken into account which can be subdivided into two groups: experimental and physics related factors. Among the experimental factors are dead material in front of the detector and non-sensitive regions, non-compensation, longitudinal leakage, lateral shower size, read-out granularity, non-linearities, electronic and pile-up noise, and magnetic field effects. Physics related factors are initial and final state radiation, fragmentation, the underlying event, and the model to simulate minimum bias events. In addition, effects due to the jet reconstruction algorithm need to be taken into account. There are two basic groups of jet finding algorithms. In ”cone-like” algorithms a cone is drawn around a seed, which for example might be the centre of a calorimeter cell with a local energy maximum. The different methods vary for example in the strategy to calculate the jet directions or in the treatment of dealing with overlapping nearby jets. ” Cluster” algorithms are inspired by QCD and rely on pairing of ”particles” (approximated by calorimeter towers) starting from the closest particles. All these different algorithms suffer from a bias in the energy measurement as a function of ET due to the effects discussed earlier in this section. In summary, there is no unique strategy for jet reconstruction and the best approach will depend on the physics process t o be analysed, the jet reconstruction algorithm chosen, and the luminosity condition. For example, different jet algorithms will be used in ATLAS t o study the QCD jet multiplicity at low luminosity ( L = 1033c7r-2s-1) compared t o high p~ reconstruction of W -+ j j decays at design luminosity. Using a simple cone algorithm with AR = 0.7 the jet resolution based on the detector aspects is DE/E= 5 2 % / 0 @ 3 . 0 G e V / E @ 1 . 7 %at low luminosity. At design luminosity a cone size of AR = 0.7 cannot be used due to the effects of pile-up noise. The ET r.m.s. is 4.7 (14) GeV in a cone with AR = 0.4(0.7) at design luminosity. Using a cone of AR = 0.4 the resolution is D E / E = 81%/@ @ 3.9GeV/E @ 1.7%. These numbers will improve as we develop more sophisticated methods to correct for all the different detector effects. Figure 3 shows the resolution together with the contribution of the physics related factors on the resolution for a cone algorithm of size AR = 0.4. It can be seen that the intrinsic physics contribution is smaller but still of the same

371

0.15 0.2

-

0. I

0.05

0

"

'

"

~

0.1

"

'

~

'

0.2

"

'

1

0 Calorimeter 0

1

Total

l-

' 0

0 0

8 : "

0.3

1/JE (GeVln)

Figure 3. Left: jet energy resolution (calorimeter contribution) without pile-up (open squares) and with design luminosity pile-up events (stars) for a cone with A R = 0.4 obtained using fully simulated QCD di-jet events. Right: the intrinsic physics contribution t o the resolution (open circles) and the calorimeter related resolution (open crosses) found using Z o + j e t events at low luminosity. Combined in quadrature (black dots) they represent the resolution with which the parton energy is measured.

order as the contribution of the experimental factors. The algorithmic flow of the jet reconstruction package in the offline ATLAS C++ software which is currently being developed, can be divided into three distinct parts. The first one is the detector level. The calorimeter cell signals are reconstructed and calibrated to the electro-magnetic scale. Afterwards the information of the different calorimeters can be combined to form towers or clusters. The second part is the actual jet reconstruction, which consists of jet finding, reconstruction and calibration. The jet energy calibration will be a key issue due to the variety of experimental and physics related effects. Insitu physics processes like Zo+ jet and W + j j will be used, combined with information obtained in test beam runs. After all jets in an event are found, there is a jet classification step, in which they will be labelled as b-jets, Tjets, or light quark jets. One important requirement in the jet reconstruction software is flexibility and that for example any of the steps in the whole chain can be redone if desired. 5 . Physics Examples for Jet Reconstruction 5.1. Forward Jet Tagging and L o w - p ~Jet Veto

Forward jet tagging will be an important tool to select Vector Boson Fusion (VBF) processes. The signature of these processes is characterised by accom-

372

panying jets in the forward region and little hadronic activity in the central rapidity region. A tagging efficiency of x 90(80)% can be achieved at (771 < 4 with a fake rate of M 10% at low (design) luminosity (see figure 4). The effi0.9

.... .+.....

2 0.8

...*. , ....

j o Higgs with ATLFAST : + HiggswithDICE

.*...... . ...*...

40

I ..... .....

20

Pseudorapidity

,/'

.

0.4;

,.*'

,P

A A

0.2

f

ttbar with ATLFAST ttbar with DICE &..A

*.....&' *.---

-*"*--*

0.I A Full siniulmios wiIh pilc-up canslaal fake lag rate 10%

0.3

2

2.5

3

3.5

4

4.5

5

kudorapidity

10

15

20

25

30

35

40

Threshold (GeL

Figure 4. Left: average p~ of forward quarks and forward jet tagging efficiency as a function of rapidity. Th e efficiencies are given at low and design luminosity. Right: jet veto efficiencies in the central region ((171 < 2.5) for heavy Higgs production via VBF and for tf background. The efficiency is defined as the fraction of events with no additional jet with a p~ larger than the jet veto p~ threshold. Results are shown for low and design luminosity as found using either the fast simulation (ATLFAST) or the full simulation (DICE).

ciency decreases for 171 > 4 because the average p~ of the quark decreases. A jet veto in the central region helps t o reduce the background. This background mainly arises from t f events which typically have a strong hadronic activity in the central region of 171 < 2.5. The efficiency to veto jets in this region is shown in figure 4 for the heavy Higgs production process H -+ WW + Zvjj. For a 60% efficiency, the t f background can be reduced by approximately a factor of 10. 5.2. Reconstruction of Resonances

Jets will play an important role in reconstructing resonances decaying to jets. Two examples are shown in figure 5. The first example shows the jet-jet invariant mass peak from W decays at design luminosity. The typical ET of the jets, in that sample, is 120-150 GeV. The tail to lower masses is due to a bias in the jet direction when the jets overlap. The hadronic W decay will play an important role in many physics signals at the LHC, such as the search for SUSY particles or the heavy Higgs boson, the measurement of the

373

25

50

75

I00

125

mjj (GeV)

200

250

300

350 mbbbb

400

(GeV)

Figure 5 . Left: invariant mass distribution of the decay W --t j j at design luminosity. Right: invariant mass distribution of H -+ hh + b6b6 decays at low luminosity.

top quark mass, and QCD studies. A more complex example for a resonance reconstruction is the decay H -+ hh -+ bbbb, which is shown on the right hand side of figure 5. For this analysis it is assumed that the mass of the light Higgs is known. This mass is used as mass constraint to reduce the combinatorial background and t o recalibrate the b-jet energies. 6. E F s a Reconstruction

At LHC, the reconstruction of the missing transverse energy (EFi"")will be used for invariant mass reconstruction in decays like A / H -+ TI-. Some discovery channels for new physics, for example in SUSY models, involve a large EFiss. For these channels it is important to minimise the EFiss fake rate due to instrumental effects such as non-sensitive areas. EFissis reconstructed from cell energies over the full calorimeter coverage (1771 < 5). For a good EFiss measurement an accurate calibration is vital. ATLAS envisages to have different calibrations applied to cells within clusters and those outside. On the left side of figure 6 the resolution of the two components ( p ~ s s , p ~ i of s sthe ) EFiss vector is shown at low luminosity. The p:?" can be parametrised as p:?" = 0.46 . At design luminosity the pile-up noise contributes an additional 15 GeV to the resolution. On the right side of figure 6 the effect of badly reconstructed or undetected jets on the EFiss measurement is shown. Assuming the jet from Z+jet events is undetected large EFissmight be reconstructed. However, if these jets are reconstructed by the standard jet reconstruction program the EFiss is much

dm.

374 30

>

-0.46*J(LET)

h

.22

g

I

I

20

10

A minimum bias n

1

1

1

1

1

1

1

1

1

1

1

1

I

t

Figure 6. Left: Resolution of the two components of the EFissvector as a function of the total transverse energy in the calorimeters, for minimum bias events and A + TT events with mA = 150 GeV a t low luminosity. Right: EFiSSdistribution from Z+jet events with p$ > 200 GeV. The full line is represents the case for which the jets are assumed t o be undetected and the dashed line for fully simulated and reconstructed jets.

smaller. A rejection factor which is larger than 1000 is achieved for E F s s > 200 GeV.

7. Conclusions Jets and EFiss will play an important role in many physics analyses at LHC. Performance studies based on events with fully simulated detector response have shown that the calorimeters are well designed for jet and E F s s reconstruction even in the "high-noise" environment at design luminosity. Further improvements in many regions are expected over the coming years. References 1. The ATLAS Detector and Physics Performance Technical Design Report, CERN/LHCC/99-14( 1999). 2. The ATLAS Calorimeter Performance Technical Design Report, CERN/LHCC/96-40 (1996). 3. T. Sjostrand, Computer Physics Commun. 28, 227 (1983). 4. R. Brun et al., CERN/DD/EE/84-1 (1996).

JET ENERGY RECONSTRUCTION WITH THE CMS DETECTOR

SKUNOR1 University of Maryland, College Park M D 2074.2, USA E-mail: [email protected]

The CMS calorimeter is a non-compensating calorimeter and has non-linear response to jet energy. A 4 Tesla magnetic field in a tracking volume induces extra smearing of energy measurement by the calorimeter. Various algorithm to improve the measurement have been tested. A simple mapping of the calorimeter response to jets is implemented in the trigger and more sophisticated energy flow algorithm may be used at a later stage.

1. Introduction

The CMS calorimeter has been designed for precise energy measurements of photons and electrons in the psuedorapidity interval 1 ~ < 1 2.6, and moderate precision energy measurements of quarks and gluons in 1771 < 5.0 through the measurement of jets of p a r t i c l e s ' ~ ~ >The ~ . central calorimeter (-q1 < 3) consists of P b W 0 4 crystals for the electromagnetic calorimeter (ECAL) and scintillator-brass (70% Cu, 30% Zn) sampling calorimeter for the hadron calorimeter (HCAL). The forward calorimeter (3 < 171 < 5) is a quartz fiber calorimeter with iron absorber. The response of the calorimeter to photons is quite linear versus incident energy, while the response to hadrons depends strongly on incident energy4. This quite different response of the calorimeter t o photons and hadrons leads to a nonlinear response of the calorimeter to jet energy and a smearing effect in the energy resolution. Since the calorimeter response depends on the total energy of a jet (not the transverse energy), the transverse energy scale of jets showed a strong psuedorapidity dependence. In order to restore the energy scale of reconstructed jets and t o improve the resolution, we have studied various algorithms for correcting the jet energy. Simpler algorithms use only information from the calorimeter. More complicated algorithms require reconstructed charged tracks. In the following sections, we describe those algorithms and their performance.

375

376 2. Initial Reconstructed Jet Energy

We use the iterative cone algorithm with a cone size R=0.5 to reconstruct jets through this Monte Calro study. The energy of reconstructed jets is calculated as a simple sum of energy in ECAL and HCAL in the jet cone, i.e.

+

Ere,, = (EC H C ) (1) EC and HC are reconstructed energy in ECAL and HCAL, respectively. The energy scale of ECAL was calibrated with electrons. The energy scale of HCAL was calibrated with charged pions ( E ~ = 3 0 G e v of ) which shower started in HCAL. Therefore, E,,,, has a right energy scale for photons, but lower for most of pions in jets because of lower response of ECAL to hadrons. 3. Jet Energy Corrections

In order to correct the jet energy, we have tested several algorithms. All algorithms are intended to remove detector effects in jet energy reconstruction and not to make correction for parton energy. In order to translate energy of jets to energy of partons, we need additional correction for fragmentation process of partons and it is beyond the scope of this study. 3.1. Simple map of Calorimeter Response

The simplest correction algorithm5 uses a form of

EcoTT(&, 71) = ~ ( E T71), x Ere,,(&, 71) (2) A calorimeter response function, a(ET,q) =< Egene > / < Ere,, > was created using a Monte Carlo QCD jet sample. Since this approach corrected the energy scale by simply multiplying by a correction factor, it did not improve the energy resolution for individual jets, but did improve trigger turn on curves’ and reconstruction of dijet masses by equalizing the jet energy scale in the whole range of r] in the detector. For example, it provided 35% improvement in the resolution of di-jet mass for Higgs (115GeV) decaying to b-quark jets6. 3.2. Use of Longitudinal Segmentation

The CMS calorimeter has mainly two longitudinal segmentation- ECAL and HCAL. In order to improve the resolution of individual jets, our first attempt was to utilize the longitudinal segmentation’. We optimized two coefficients a(ET,q) and b(Et,71) in the following formula with QCD jets sample to obtain the best energy resolution.

EcoTT = ~ ( E T , vx) EC

+ b(ET,V) x HC

(3)

377

The improvement in relative resolutions before and after the correction was 4% on average in the wide range of ET and 77. For example, the resolutions before and after the correction were 14.3%and 13.6%,respectively at ET=80GeV and 77=0.2, i.e. 5% improvement in relative resolution. The improvement was not large. This lack of improvement may be attributed to the relatively thick ECAL compartment ( 1.1 interaction length) without any longitudinal segmentation. In general, a calorimeter can be divided into three longitudinal sections. The first section measures mainly showers due to photons in a jet, the second section measures a mixture of showers due to photons and hadrons, and the third section measures showers due to hadrons. By optimizing a weight for the second section one can somewhat compensate a different response of calorimeter to photons and hadrons, consequently improving jet energy resolution. The thick ECAL subsumes the second section completely. Therefore, there was not much room to improve the jet energy resolution by optimizing longitudinal calorimeter weights in the CMS calorimeter.

3.3. Use of Transverse Shower Size Contrary to the longitudinal segmentation, the CMS calorimeter has very fine transverse segmentation in ECAL (A7 x A@= .0174 x .0174). We tested to separate electromagnetic clusters (em) and hadronic clusters ( h a d ) in a jet cone using the transverse shower shape in ECAL and then apply correction coefficients to em and had, separately. A preliminary result shows that the separation of em and had clusters is not clean in terms of transverse shower size in ECAL. Therefore improvement of resolution is expected to be small8. 3.4. Correction f o r Out-of-Cone Tracks

Relatively soft tracks can be swept away from the jet cone by 4T magnetic field. For example a track with P ~ = 1 . 6GeV is deflected by the magnetic field by A@= 0.5 radian from the original direction when it reaches the calorimeter surface. Tracks with PT 5 0.8 GeV never reaches the barrel calorimeter. Fig. 1 shows a fraction of the energy of generated jets escaping from the jet cone in 4T field. One can see that on average about 15 % of the energy of the soft (- 30 GeV) jets escapes from the jet cone and there are big fluctuations of this energy. Adding the energy of the out-of-cone tracks to the calorimeter jet energy removes the fluctuations and therefore improve the jet energy resolution. Fig. 3 and Fig. 4 show the jet energy resolution and linearity as a function of the generated jet ET. The resolution is improved significantly with addition

378

of out-of-cone tacks(dots vs. opensircles). Also we have tested the correction using all Monte Carlo generated tracks and actually reconstructed tracks with our current track reconstruction program and found no significant difference between two cases. The efficiency of track reconstruction is 90% with a PT cut off at 0.9GeV. This efficiency and the cut off energy are sufficient to improve the jet energy resolution.

-

.- 0.25 0)

k.225

eA

rr!

lqsn-I

0.2

<0.5

0

a 0.175 S

c

"w 0.075

9

0 0.05

'0.025

1

'20 Figure 1 .

40 60 80 100 120 E, MC jet in cone 0.5, GeV

Fraction of energy of the generated jet escaping from the jet cone 0.5 in 4T field.

3.5. Energy Flow Algorithm A considerable improvement of the calorimeter jet energy resolution using reconstructed tracks (energy flow algorithm) has been already demonstrated in a number of HEP experiments such as LEPg, TevatronlO, ZEUS'l, Tesla12. Ultimate goal of such algorithm may be expressed in the formula, Ecom

= ECphtons

+ (EC + HC)neutral.h+ Trackscharged

(4)

The algorithm uses ECAL to measure the energy of photons, ECAL and HCAL to measure the energy of neutral hadrons, and reconstructed tracks t o measure energy of charged particles inside a jet cone.

379

A major issue of the algorithm is to remove energy due to charged particles and leave only energy due to photons and neutral hadrons in the calorimeter inside the jet cone. We performed the following procedure. First we reconstructed a jet with the standard iterative cone algorithm using calorimeter energy. Then we we reconstruct tracks inside a cone R=0.5 around the jet axis and propagated them to the calorimeter surface. For tracks going out of the jet cone, we simply added the energy to the jet. For tracks stayed inside the cone, we formed a energy cluster with 3x3 crystals ((0.017 x 3)2)and 3x3 HCAL towers ((0.087 x 3)2)around the impact point of the track. Then tested a match between the track and the cluster with the following matching condition. -0

< (Etracker - Ecalo) < 2 0

(5)

+

where (TIE= 100%/sqrt(E) 5%. If the matching condition is satisfied, we replaced the cluster energy with energy of the track. If not, subtracted an estimated calorimeter response Rave from the jet energy and added the energy of the track. We used GEANT3 to create a library of R a v e . The library has R a v e as a function of incident particle energy. Ravealso depends on starting point of y 30 0 g 27.5

8

25

0 0

2 22.5 h

8 20

v

.-$17.5 w

2 15

12.5 w' c '0 .-al

7.5

"0

20

40

60

80

100 120

E, MC jet in cone 0.5,GeV

Figure 2. The number of charged particles in the jet cone R=0.5 at the calorimeter as a function of the generated jet energy: dots for all charged particles, squares for tracks satisfied the matching condition Eq. 5 .

"0

20

40

60

80

100 120

E, MC jet in cone 0.5, GeV

Figure 3. The jet energy resolution as a function of the generated jet ET. dots for reconstruction with the calorimeter only and stars for with full correction. opencircles for correction with only out-of-cone tracks. Other two are the resolution by partial correction. (see text for more detail.)

380

hadronic shower. We crated it for showers started in ECAL and HCAL separately. In the correction process we identified whether a shower of unmatched track started in ECAL or HCAL by checking energy in ECAL, and applied R,,, accordingly. Fig. 2 shows the number of charged tracks in the jet cone of R=0.5 on the calorimeter surface and the number of tracks which satisfied the matching condition Eq. 5 as a function of the energy of generated jets. The matching efficiency is good at low energy, but getting worse as energy increases because of more tracks overlap with photons around the core of jets. Fig.3 and Fig.4 show the resolution and linearity of the jet energy as a function of the generated jet energy; dots for before and stars for after the full energy correction. Also shown are results with partial correction or with a simpler methods. The opencircles are after out-of-cone track correction. The squares simply replaced the energy in the calorimeter with R,,, without the track-cluster matching. The triangles used a simpler method to estimate Ravel3.Significant improvement is seen in both resolution and linearity with the correction.

1.2

.. .. .. .. .. 40 -ij............ ;... .....i......... i.......... i............. i............ ........... i............ 1 4 : < ( w c ~MET-MC MET),, GeV : . . I

v)

$ 1.1

.

20 -

C

0

0

.

-

1

* ~

6

. ... , ... .. .. .. . . .... . . ..... . ..... . . ... . . .... . . . . . . . :. : .. :. .:. . . . . : :; :; .: .. . i . .. - . ..

: *

~

-20 -

0.8

T

%c0.7

- 4 O - i . . 1 no coirectibns

W

+

.fin-

0.6

A type

. i ..

i

.: .

0

20

40

E,

80 100 120 MC jet in cone 0.5,GeV

60

Figure 4. The jet energy scale as a function of the generated jet E T . T h e definition of pointspoints are the same as Fig. 3

-'

type 2 corrections '

.. .. .

. ..

.............. j................ ; .. ; ... .

~

. ! ... . . .

i

.

....

.. ......... ....,..... ......... .. ; .. : ... ... . .... :... ~

: .. ..

:

.

.

.

..

i

~

:

..... . ~

'.. - . t .i ;. ;

1 corrections I

0.5

;*,

~

. . ... ............... ....... .)............ ?...................... ...~ . : - c . . . . .

0.9

0.4

. I .

&- 4 * -.

0.

Irt' . 0

i

<

-+4 ,

,

,

50 1W ' 150 '200'250'300 350 400'410 MC MET, GeV

Figure 5. Offset of the missing transverse energy as a function generated missing transverse energy. T h e type 1 correction apply correction to only inside jet cones and the type 2 correction applied correction to both inside and outside jet cones.

381

4. Correction for Missing Transverse Energy

The non-linear response of the calorimeter also distorts measurement of the missing transverse energy MET. We have extended the simple jet energy correction (Eq. 2) to the reconstruction of M E T . We took a sum of the corrected energy of reconstructed jets plus energy outside the jets cones. For the energy outside the cone, we used the correction coefficient for low energy jets (&=30GeV). Fig. 5 shows an offset of the missing transverse energy as a function of MET(gen) for a sample of SUSY events. Clearly the extension of the simple jet energy correction restored the energy scale of missing transverse energy above MET 100GeV. 5. Conclusions We tested various algorithms to improve the jet energy scale and resolution with the CMS detector. A simple mapping of the jet response in the q-ET space restores the jet energy scale. It has been implemented in the Level-1 jet trigger as well as the higher level trigger and offline analysis. Further improvement using only calorimeter information, namely more sophisticated use of longitudinal and transverse segmentation of the calorimeter, is found to be limited. Significant improvement is provided by combining calorimeter information and reconstructed charged tracks. The resolution of jet ET is improved from 24% to 14% at 20GeV and from 12% to 8% at 100GeV. The energy scale is accurate better than 2% in the ET range from 20GeV to 120GeV. A jet energy correction was extended to the reconstruction of the missing transverse energy. We observed significant improvement in both the energy scale end resolution with the extension of the simple jet energy correction. Those algorithms are still at primitive state and need to be tuned. We expect further improvement be achieved by the time we start taking data with the CMS detector.

References 1. The Compact Muon Solenoid (CMS) Technical Proposal, CERN/LHCC 94-38, December 1994. 2. CMS ECAL Technical Design Report. CERN/LHCC 97-33 CMS TDR4(1977), December 1997. 3. CMS HCAL Technical Design Report. CERN/LHCC 97-31. CMS TDR2, June 1997. 4. " S t u d y of the response of the prototype CMS hadron calorimeter, including magnetic field effects, t o pion, electon and m u o n beams." V.Abramov et al. CERN CMS NOTE 2000/03. (See figures 28 and 31 in the paper for example.)

382 5. ”Energy corrections for QCD jets.” S.Abdullin et al. CERN CMS IN 2001/001. 6. ”Jet Quality and H to bb Mass resolution”. V.Drollinger, Th.Muller. CMS NOTE 2000/073, Section 4.2, page8. 7. ”Jet energy correction using weights on calorimeter longitudinal readouts”. O.Kodolova, CMS NOTE 2002/023. 8. D. Green uEnergy Flow in CMS Calorimetry” Fermilab-FN-0709 9. see for example ALEPH, D. Buskulic et al. NIM A 360(1995) 481-506 10. S. Lami, A. Bocci, S. Kuhlmann and G. Latino, FERMILAB-Conf-00/342-E CDF January 2001 11. Matthew Wing, Precise Mesurement of the Jet Energies with the ZEUS Detector, In Proceedings of the IX Int. Conf. on Calorimetry in Part.Phys., Annecy, Oct.914, 2000. 12. P. Gay, Energy flow. In Proceedings of the Linear Collider Workshop 2000, Fermilab, Batavia, IL, USA, 2000. http://www-lc.fnal.gov/lcws2000 13. The simpler method assumed that a ratio of energy in ECAL and HCAL is 0.4/0.6 in case of shower started in ECAL. Then it used e/h=1.6 for ECAL and e/h=1.39 for HCAL to estimate the response of ECAL and HCAL for the energy deposit. These values were deduced from the CMS test beam data. See Ref.4.

Calibration & Monitoring Couener: M. Gataullin

M. Gataullin

Covener’s Report

C. Gatti

Calibration of the KLOE Electromagnetic Calorimeter

K. Miyabayashi

Monitoring and Calibration of the BELLE Electromagnetic Calorimeter

M. Barbi

Calibration and Monitoring of the ZEUS Uranium Scintillator Calorimeter at HERA

H. Torii

Calibration of the PHENIX Lead Scintillator Calorimeter

U. Bassler

D 0 Calorimeter Calibration

0. Lobban

Calibrating a Longitudinally Segmented Calorimeter

R. Nichol

The Calibration of the MINOS Detectors

P. Vahle

First Results from the MINOS Calibration Detector

A. Hamer

Calibrating the SNO Detector Response

K. Hanson

Time Calibration of AMANDA: Three Variations of a Theme of TO

L. Xia

Absolute Calibration of Electromagnetic Calorimeter at LHC with Physics Processes

L. Zhang

Monitoring Light Source for CMS Lead Tungstate Crystal Calorimeter at LHC

S. Danagoulian

LED based Light Monitoring System for the PRIMEX Experiment at Jefferson Lab

383

This page intentionally left blank

CALIBRATION AND MONITORING

MARAT GATAULLIN Physics Department, California Institute of Technology, Pasadena, CA 91125, USA E-mail: mamtOhep.caltech.edu (Convener’s Report)

Calibration and monitoring of an experimental apparatus is essential for the success of any physical experiment, and calorimeters are not exception to this rule. Nevertheless, it is not unusual that the importance of an adequate calibration is underestimated. Insufficiently precise calibration or systematic mis-calibration can potentially discredit the outcome of an experiment. In practice, it may lead to a significant deterioration of resolution, thus “wasting” the funds and effort spent to build a calorimeter. For example, a novel calibration technique’ based on the RFQ H- accelerator was first used several years after the beginning of the L3 experimental program. However, it was able to significantly increase the resolution of the L3 BGO calorimeter2 to a level previously seen only in beam tests prior to calorimeter installation in the underground cavern. There exists an extensive literature on how to design and build calorimeters, however, very few papers have been published on their calibration and monitoringa. Therefore a considerable effort was spent to attract papers related to this subject from different high-energy physics experiments. Overall, thirteen papers were presented in this session. Different aspects and challenges of the calibration and monitoring were covered both for existing collider experiments (KLOE, BELLE, ZEUS, PHENIX, DO CDF) and for neutrino physics experiments (MINOS, SNO, AMANDA). Three papers describing monitoring systems and calibration techniques to be used in future experiments (CMS and PRIMEX) were also presented. A short summary of each presented paper is given below. Of course, the reader is advised to read the respective papers.

aOne notable exception is a chapter on calibration in the monograph by Richard Wigmans3.

385

386

1. Collider Experiments

Claudio Gatti and Kenkichi Miyabayashi described calibration of the KLOE and BELLE electromagnetic calorimeters operating at e+e- colliders. Bhabha and yy events are used to derive the absolute energy and time (KLOE) calibration scales in both experiments, while the monitoring is performed using cosmic rays. Mauricio Barbi presented the procedure to monitor the ZEUS Uranium calorimeter. Natural uranium activity is used to calibrate and monitor the calorimeter response in situ, while the initial calibration constants were provided by the test beam. Different techniques of monitoring the readout system were described. Calibration of the PHENIX electromagnetic calorimeter was presented by H. Torii. Cosmic muon events were used to calibrate the calorimeter before its installation. Three different methods are used to check the calorimeter resolution an situ. The front-end electronics of the DO calorimeter have been replaced to accommodate the new requirements of the Tevatron Run 2. Ursula Bassler showed the first results on the on-line calibration and linearity determination of the DO calorimeter in the Run 2 environment. The absolute calibration scale was set using reconstructed Zo bosons, and the first measurements of E l p and jet energy scale were presented. Olga Lobban addressed in her paper a difficult task of calibrating longitudinally segmented calorimeter systems. Using test beam data from the CDF Plug Upgrade Calorimeter she compared three different methods of setting the hadronic energy scale, showing that one of the conventionally used methods introduced a number of undesirable effects. 2. Neutrino Physics Experiments Ryan Nichol discussed different concepts of MINOS detector calibration. The MINOS experiment consists of three detectors: a far and near detector to study neutrino oscillations; and a small calibration detector, which is being tested at CERN. The relative calibration between the detectors is done using cosmic muon events, while the absolute calibration is established with the calibration detector. Patricia Vahle discussed the goals and design of the calibration detector and presented the first results from its exposure to test beams at CERN. Andre Hamer gave a detailed and excellent description of the specialized devices used to calibrate the SNO heavy-water detector. The employed calibration techniques use sources of isotropic light, y-rays, neutrons, and P-particles.

387

The resulting systematic uncertainties in the neutrino flux measurements were also discussed. Kael Hanson describes the calibration procedures at the AMANDA-I1 neutrino telescope currently operating at the South Pole. It consists of an array of 677 optical modules deployed in the ice at depths between 1200 m and 2300 m. AMANDA timing laser and cosmic ray muons are used to calibrated the timing offsets. The calibration of the proposed IceCube detector will use digital trigger system. 3. Future Experiments

Lei Xia described a study of a calibration method which might be used to calibrate the CMS electromagnetic calorimeter at LHC. Several physics processes can be used to set an absolute energy scale. The estimate of data taking time necessary to achieve the required calibration precision is given. Liyuan Zhang gave a detailed description of the laser source system, which will be used to monitor the lead tungstate crystals of the CMS electromagnetic calorimeter. The performance of the custom manufactured tunable laser system and other issues related to the monitoring precision were discussed. Samuel Danagoulian presented the LED based multi-channel light monitoring system developed for the PRIMEX experiment at Jefferson Lab. After a six-month continuous run the prototype system demonstrated high reliability and good stability.

Acknowledgments I would like to thank all authors for their carefully prepared presentations and papers. This work was supported by DOE grant DE-FG03-92-ER40701. References 1. U.Chaturvedi et. al., IEEE Trans. Nucl. Sci. 47 (2000) 2101-2105. 2. L3 Collab., B. Adeva et al., NIM A289 35 1990. 3. R.Wigmans, “Calorimetry: Energy Measurement in particle physics.” Clarendon press, 2000.

CALIBRATION OF THE KLOE ELECTROMAGNETIC CALORIMETER

A. ALOISIO, F. AMBROSINO, F. CEVENINI, G. CHIEFARI, C. DI DONATO, A. DORIA, L. MEROLA, M. NAPOLITANO, G. SARACINO Dipartimento di Scienze Fisiche dell’Universitd “Federico II” e Sezione I N F N Napoli, Italy

A. ANTONELLI, M. ANTONELLI, G. BENCIVENN1,S. BERTOLUCCI, C. BLOISE, F. BOSSI, P. CAMPANA, G. CAPON,P. CIAMBRONE, P. DE SIMONE, S. DELL’AGNELL0,A. DENIG, M. DREUCCI, G. FELICI, M. L. FERRER, G. FINOCCHIAR0,C. FORTI, A. FRANCESCHI, S. GIOVANNELLA, S.W. HAN, L. INGROSS0,G. LANFRANCHI, F. LU, M. MATSYUK, W. MEI, S. MISCETT1,M. MOULSON, F. MURTAS, A. NEDOSEKIN, L. PASSALACQUA,P. SANTANGELO, B. SCIASCIA, I. SFILIGOI, T . SPADARO, P. VALENTE Laboratori Nazionali di Fmscati dell’INFN, Fmscati, Italy

J. LEE-FRANZINI Labomtori Nazionali di Fmscati dell’INFN, Fmscati, Italy, and Physics Department State Univerity of N e w York at Stony Brook, USA

V. PATERA, A. SCIUBBA Labomtori Nazionali di Frascati dell’INFN, Frascati, Italy e Dipartimento di Energetica dell ’Universita “La Sapienza”, R o m a , Italy

C. BACCI, P. BRANCHINI, F. CERADINI, A. FARILLA, A. FERRARI, E. GRAZIANI, F. NGUYEN, M. PALUTAN, A. PASSERI, E. SPIRIT1,L. TORTORA Dipartimento di Fisica dell’Universita “Roma f i e ” e Sezione I N F N , R o m a , Italy

C. BINI, V. BOCCI, R. CALOI, E. DE LUCIA, G. DE ZORZI, A. DI DOMENICO, P. FRANZINI, P. GAUZZI, F. LACAVA, D. LEONE, E. PASQUALUCCI,

388

389 E. PETROLO, L. PONTECORVO, E. VALENTE, S. VENEZIANO Dipartimento di Fisica dell ’Universitd “La Sapienza” e Sezione I N F N , R o m a , Italy

G. CARBONI, R. MESSI, L. PAOLUZI, E. SANTOVETTI Dipartamento di Fisica dell’llniuersitd “Tor Vergata” e Sezione I N F N , Roma, Italy

M. CASARSA, F. SCURI Dipartimento di Fisica dell ’Universitd e Sezione I N F N , m e s t e , Italy

V. CASAVOLA, G. CATALDI, E. GORINI, F. GRANCAGNOLO, M. PRIMAVERA, A. VENTURA Dipartimento di Fisica dell’Universitd e Sezione I N F N , Lecce, Italy

F. CERVELLI, S. DI FALCO, C. GATTI, M. INCAGLI, G. VENANZONI Dipartimento di Fisica dell’Uniuersitd e Sezione I N F N , Pisa, Italy

G. DE ROBERTIS, 0. ERRIQUEZ, F. RUGGIERI Dipartimento di Fisica dell’Universitd e Sezione I N F N , Bari, Italy

S. A. BULYCHJOV, V. KULIKOV Institute for Theoretical and Experimental Physics, Moscow, Russia

w. KLUGE, c . KUO, M. MARTEMIANOV, s. MULLER, B. VALERIANI Institut fur Experimentelk Kernphysik, Uniuersitat Karlsruhe, G e r m a n y

S. CONETTI Physics Department, University of Virginia, Charlottesuille, U S A

R. D. SCHAMBERGER Physics Department, State Uniuen’ty of N e w York a t S t o n y Brook, U S A

G. L. TONG, G. XU, G. W. YU Labomtori Nazionali di Frascati dell’INFN, Frascati, Italy, and Institute of High Energy Physics of Academica Sinica, Beijing, China

The KLOE detector was designed primarily for the study of C P violation in neutral kaon decays at DAaNE, the Rascati &factory. A lead scintillating-fiber sampling calorimeter has been built, providing good energy resolution and timing accuracy. The calorimeter is calibrated on-line using cosmic rays, Bhabha, and e+e- 4 yy events. q5 decays are then used to monitor the performance.

390

1. Introduction

The KLOE detector is designed for the study of C P violation in the decays of neutral kaons. The detector is composed of a cylindrical drift chamber1 surrounded by an electromagnetic calorimeter2 (EmC) both of which are embedded in a magnetic field of 0.52 T . The main task of the electromagnetic calorimeter is to reconstruct the CP-violating K L + 7r07ro + 4y decay and to efficiently reject the 200 times more abundant K L -+ 7ro~07ro+ 6y background. To accomplish this, the calorimeter must have a good detection efficiency for photons with energies ranging from 20 to 300 MeV, and must be able to measure the time and the position of the photons with high accuracy in order to reconstruct the decay point of kaons. The EmC is a fine-sampling lead scintillating fibers calorimeter with photomultiplier (PM) read-out. It is composed of a barrel and two end caps. The nearly cylindrical barrel consists of 24 modules 4.3 m long, 23 cm thick, and trapezoidal in cross section. Each end cap consists of 32 vertical “C”-shaped modules. All modules are stacks of grooved, 0.5 mm thick lead foils alternating with layers of cladded 1 mm diameter fibers. A module is composed of 200 such layers glued together. The ratio 1ead:fiber:glue is 42:48:10 by volume. The average density is 5 g/c7n3, the radiation length is 1.5 cm, and the overall thickness of the calorimeter is 15 radiation lengths. Granularity is defined by the (4.4 x 4.4) cm2 cross section of the light guides coupled to the PM’s. The signals coming from the PM’s are splitted and sent to the ADC’s for energy measurement, to the TDC’s for time measurement, and to the ADC’s for trigger.

-

-

2. Energy reconstruction and calibration For each cell, two amplitude signals S are recorded at the two ends, A and B, of the calorimeter modules. The energy signal, E , is obtained from S as:

where SO,^ is the zero-offset (ADC pedestal), S M , is~ the response for a minimum ionizing particle (mip), and kE is the energy scale in MeV. The cell energy is taken as the mean of the determination at both sides, after correcting for the attenuation along the fiber length (X,tt 4 m). Samples of cosmic rays crossing the detector are continuosly selected with and without circulating beam. In the latter case, the low occupancy of the calorimeter allows the determination of the zero-offsets as the mean of the distribution of ACD counts for each channel. In the former case, about 50-100 Hz in -24 hours of data taking allow the measurement of light attenuation lengths and the mip response with

-

391 1.1

- Q I 0s -

wn+n.no

o K -m+rino

- A et-y I -

-0.15 -Os20

50

100 150 200 250 300 350 400 450 500 J%,(MeV)

e* e-

Z0.3 0.2 0.1 OO

50

-+

095

-

09

-

085

-

-*-++

t

e' e' y

100 150 200 250 300 350 400 450 500

0.65

0

E(MeV)

Figure 1. Left: Differential linearity vs. E.,. (Upper plot); Energy resolution vs. E.,. The solid line shows the fit result (Lower plot). Right: Comparison of the photon detection efficiency obtained with three different samples of events.

an accuracy of 1-2% on each single cell. The channel equalization must be kept at the 5% level for trigger purposes. The HV's for each PM are then adjusted to maintain the required equalization level, since channel gains drift by a few percent per month. A refined calibration and energy scale determination are performed by selecting Bhabha and e+e- -+ yy events on-line. An integrated luminosity of 100 nb-' yields 150K Bhabha and 10K yy events. An iterative procedure uses Bhabha events to set the mean value of the energy measured in each column to the same level. The energy scale is set by imposing that the peak of the distribution of energy released by photons from yy events be at 510 MeV. A scale factor of 40 MeV/mip is obtained, corresponding to a light yield of 1 p.e. per MeV on each side. Energy resolution and linearity are checked using photons from radiative Bhabha events e+e- -+ e+e-y, by determining the photon energy and direction from track information (Fig. 1, left). The linearity in response is better than 1%for energies above 50 MeV, while a linearity variation of up to 5% is observed below this energy. The resolution obtained scales as c / E N 5 . 4 % / J E W . The photon detection efficiency is measured using samples of e+e- -+ e+e-y, q5 + n+n-n0, and K L -+ n+x-xo events (Fig. 1, right). An efficiency higher than 99% is reached for energies above 80 MeV.

-

-

N

3. Time reconstruction and calibration

As in case of the energy signal, two time signals T are recorded at both sides of a cell. After multiplying by the TDC calibration constants, the arrival time

392 $00

7

a

:

&ioo

1

P1 P2

r i; W

55.94 f

133.2

m4,

b 500 :

1.261 1.315

*

+ 7) y + 3 y 37

A@

no')'

oe+ e -

+ e+e - y

300 : 200 : 100 ;

E (MeV) Figure 2 .

Comparison of the time resolution obtained with three different samples of events.

t and the impact position s of the particle are determined from the sum and the difference of the two cell times:

"f ( t A - tB - t,A + t t ) s ( m )= -

(3)

2

where to is the time offset, L is the module length, and wf 17 cm/ns) is the velocity of propagation of light in the fibers. The difference between the time offsets at the two ends, Ato, and the velocity, wf, are determined fitting the raw t A - tB spectra for each cell. The sum of the two offsets is obtained by selecting straight cosmic ray tracks (p > 7 GeV/c) and imposing the relation AT = AR/c valid for particles moving at the speed of light, between the time and space distances of hit cells. With this procedure, the time offsets are determined with a precision of w 50 - 80 ps in half an hour of cosmic ray data taking. Fine corrections to the time offsets are obtained using yy events by minimizing the time residuals T - R/c between cluster times T and distances from the interaction point R. The accuracy reached is 20 ps for 100 nb-' of collected data. The time resolution is measured using the T - R/c distributions of prompt photons from radiative Bhabha and 4 decays (Fig. 2). The time resolution scales as ot N 56 p s / d m fB 133 ps. The constant term (W

-

393

-

E 2

4

$

8 d

5

0 Before calibration-, Slope -0.6~10'~

4

3

i..

0

After calibration+ Slope -0.5~10~'

2 -

;

1 :

U

I

20

3 10

4

040

45

XI

tc,-Rc,k (ns)

raAR ( a n )

Figure 3. Left: T - R / c distribution for e+e- + yy events. Peaks are separated by T R F . Right: Residuals between neutral and charged vertex positions in K L --f T + A - T O decays as a function of the K L path, before and after refined calibration.

is mostly due to the finite bunch length. Since the start to the calorimeter TDC's is provided by the level-one trigger, after synchronization with the radio frequency signal, all the measured times are known up to a multiple of the radio frequency period (TRF = 2.715 ns). The T - R / c distribution of yy events, reflecting the spread in trigger time formation, shows well separated peaks (Fig. 3, left). The distance between the peaks gives an estimate of TRF.The correction to the time calibration constants is obtained dividing the measured value of TRF by the expected value. Non linearities in the distribution of T - R/c as a function of the photon impact position, which lead to a systematic error in the time scale measurement of up t o 1%, have been carefully corrected. The time scale is then checked in K L -+ T+T-T' decays by comparing the position of K L decay vertex obtained using drift chamber tracks with the one obtained using the impact positions and arrival times of photons in the calorimeter. In Fig. 3 (right), the residuals between the charged and neutral vertex positions are shown ~ t as function of the K L flight path for two different samples, with and without correcting for systematics. The slope in the distribution disappears in the former case, showing the correctness of the time scale calibration. References 1. KLOE collaboration, M. Adinolfi, et al., submitted to Nucl. Znst. Meth. A. 2. KLOE collaboration, M. Adinolfi, et al., Nucl. Znst. Meth. A482,363 (2002).

MONITORING AND CALIBRATION OF THE BELLE ELECTROMAGNETIC CALORIMETER

KENKICHI MIYABAYASHI Department of physics, N a m Women's University Kita- Uoya-Nishi-machi, N a m 630-8506, Japan E-mail: miyabaya4hepl.phys.nam-wu.ac.jp (For the Belle Electromagnetic Calorimeter Group) We report monitoring and calibration issues to have stable operation of the Belle electromagnetic calorimeter. As a result, good mass resolutions for no and 7 are obtained to be 4.8 and 12.1 MeV/c2, respectively. The degradation of light output due to the radiation damage is small, about 3% for the radiation dose of 40rad.

1. Introduction The Belle detector' is working at the interaction point of the KEKB e+ecollider2 to detect B meson decay products to study C P violating phenomena. In order to cover various topics related to B decays, high resolution electromagnetic calorimetry is indispensable to reconstruct neutral particles such as y,T O , q and so on. Also the electromagnetic calorimeter plays an important role in electron identification3. In this report, we describe monitoring and calibration issues to achieve good performance to match physics requirements. 2. Calorimeter construction

2.1. CsI(Tl) crystal calorimeter The Belle electromagnetic calorimeter consists of 8736 CsI(TZ) counters. Out of these, the barrel calorimeter has 6624, while the forward and the backward endcap calorimeters have 1152 and 960 counters, respectively. The coverage in the polar angle(@)with respect to the electron beam axis is 12" < 8 < 155' in the laboratory frame. The barrel part has 1250mm inner radius. The mechanical support structure is comprised by aluminum inner wall and fins suspended from stainless steel made reinforcing bars and outer walls. In Ref.l, the details about the calorimeter construction are described.

394

395

Each CsI(T1) crystal is 30cm long and its cross section is 5.5cmx5.5cm at the front face where particles come in. The crystal is wrapped by 200pm white Gore-Tex film reflector, then is wrapped by a laminated sheet of 25pm thick aluminum and 25pm thick mylar. The aluminum layer works as an electric shield, while the mylar layer electrically separate the counter from the mechanical supporting structure. On the rear face(readout surface), two pieces of PIN photodiodes are glued via l m m thick lucite with an epoxy glue. Each photodiode has lcmx2cm active area. Just behind those, two preamplifiers are equipped inside a metal casing which is fixed on the crystal by screws. By a CsI(T1) counter, typically 1 MeV energy deposit yields about 5000 photoelectrons. In the operation of the first three years, there is no dead counter out of 8736 crystals. 2 . 2 . Readout electronics

The schematic diagram of the readout electronics is shown in Fig. 1. As described in the previous subsection, each photodiode signal is received by a charge sensitive preamplifier and integrated independently. These two outputs from each counter are summed at the first stage of a shaper-QT board located near the Belle detector. The summed signal is shaped with two different shaping times of 200ns and 1ps. The former goes to the calorimeter trigger system, and the latter is fed to a charge to time(QtoT) converter chip(MQT3OOA) for the energy measurement. The output of MQT3OOA is a ECL differential signal. The number of edges of pulse within the proper time window denotes the selected range and timing of the edge corresponds to the input pulse height. This is the mechanism of QtoT conversion with auto range selection. Finally, a signal is digitized by a FASTBUS TDC(LeCroy 18778). Using this scheme, a wide dynamic range corresponding t o 18bits is achieved, while incoherent and coherent noise contributions per crystal are 19OkeV and 17keV4i5,respectively. In order to reduce data size, a crystal hit which does not exceed 0.5MeV is suppressed on the TDC. In average, there are about 2000 hits in an event under this scheme. This treatment does not sacrifice energy resolution at all. 2.3. Monitoring tools for stable operation

Since a preamplifier consumes the electric power of 0.15W, in total the electromagnetic calorimeter generates a heat of a few kW inside the Belle detector structure. We have a cooling facility with water circulation and in total 312 thermistors are installed to monitor the temperature. Before installation, we tested a hygroscopicity of a CsI(T1) crystal and we found that low humidity less than 15% is necessary to avoid a damage on crystal surface. Based on

396 preamp. signal _ _ _ _ _ _ . _ _ _ _ _ _ _i _TWO __ ___ ___________.

i

Scintilation light

: are summed at 1st stage : of Shaper-QT board.

photodiode

I

Preamp.

.

;

'_-..Electronics Hut * ---_..I

Figure 1. The block diagram of the readout electronics.

this result, the dry air facility is installed and humidity monitoring is carried out by 104 humidity sensors(HUM1TTER). During a regular operation these environment parameters are kept to be quite stable; temperature change is less than 0.1"C and humidity is always less than 6%. Such stable conditions contribute a lot to ensure stable detector operation. Each CsI(T1) counter has a piece of LED inside the preamplifier casing. Since there are enough number of methods as described in the following section, it is used to check whether the counter is alive.

3. Calibration From the digitized signal of i-th counter(TDCi), the energy deposit(&) is obtained by the following formula;

where PEDi and ei are the pedestal and electronics gain constant of i-th counter which are obtained by the daily electronics calibration run, while Ci is determined by physics events. We obtained the initial Ci by cosmic rays before beam collision. It allowed the quick start of the calorimeter with good performance at the initial stage. After physics run started, we obtain Ci for e+e- collision data with Bhabha and e+e- -+ yy events by minimizing the x2 defined as follows.

x2=c

Esumj - Eexpj

3

*2

397

where j , Esumj, Eexpj and (T are the event suffix, the total energy of j th event, the expected total energy deposit of j-th event and the estimated energy resolution. This minimization results in the matrix inversion method. As shown in Fig. 2, the matrix inversion fairly converges. In actual calibration 1.2

,

-1.15

2

Crystal ID

0.85

0.8

2500

" " 1 " " 1 " " " " " " " ~ ' " " " " " " " " ' ' 0 1000 2000 3000 4000 5000

6000

7000

8000

9000 Cell ID

[ ir

2000

1 i L

1500 1000

1 L

500 0

0.8

0.85

0.9

0.95

I

1.05

1.1

1.15

1.2

Ci(after)/Ci(before) Figure 2. The ratio of the calibration constant Ci between before and after Bhabha calibration is plotted as a function of crystal identification number(upper) and projection onto vertical axis(1ower).

procedure, Bhabha calibration is carried out at first because of higher statistics. And then fine tuning is done by efe- + yy process since it has less systematics due to the material in front. Note that we use cosmic-rays calibration constant for inner-most crystals in the endcaps, because we have too small statistics in those crystals due to the degradation of shower reconstruction efficiency caused

398

by a quite thick material in front. In addition, some of them do not match back-to- back condition. Since each counter is calibrated by Bhabha and e+e- + yy processes, i.e. crystals are calibrated at the highest energy point, there can be a non-linearity of the reconstructed energy in low energy region(be1ow 1GeV). We carefully checked such effect using no signals and correct it by comparison between the data and MC no mass peak value. This correction amounts about 2% for a lOOMeV energy photon, as shown in Fig. 3.

Non-linearity correction (applied by div. for Exp. Data)

-0.939OE-& -0.1056E-l@

-

0.99

.................. .

..................-,

0.98

..................-.

.....................

0.97

.................*.,

.................__.

0.96

........

2.0

3.0 loglO(Ey in MeV)

Figure 3. Th e non-linearity correction is plotted as a function of incident photon energy. This correction is applied t o the showers found in e+e- collision events.

399

As the resultant performance, no and q mesons reconstructed by two photons are shown in Fig. 4, where each photon is required to exceed 50MeV. The obtained mass resolutions are 4.8 and 12.1 MeV/c2 for no and q, respectively.

Figure 4. The invariant mass distributions of two photons; #(left) and q(right) regions. Here each photon is required to exceed 50MeV. The obtained mass resolutions are 4.8 and 12.1 MeV/c2 for 7r0 and q, respectively.

4. Monitoring crystal's radiation dose and light output

Since the cosmic ray runs are carried out without beam, those datasets are suitable to monitor intrinsic light output of the crystal. The dark current of photodiodes is monitored and it gives us the radiation dose of the crystals. By these measurements, we can know the light output of crystals as a function of the radiation dose as shown in Fig. 5. During three years operation the crystals in barrel and endcaps have a radiation dose of -10rad and -40rad, respectively. The light output degradation is -2% in barrel and -3% in endcaps. AS a result, our CsI(T1) calorimeter has enough radiation hardness as tested before installation" 5. Conclusion The Belle CsI(T1) calorimeter has been working quite well for three years. The energy resolution for e+e- -+ yy is 1.7%. The no and q mesons are reconstructed by their decays into two photons with the mass resolutions of 4.8MeV/c2 and 12.1MeV/c2, respectively. All the 8736 CsI(T1) counters are

400

0

5

10

15

20

25

30

35

40

45

Dose (Radkrystal) Figure 5. The light output change plotted as a function of radiation dose.

alive. The light output degradation by radiation dose was found to small; about 3% even in the endcaps after three years running. We can conclude that we can enjoy much more B physics with neutral particle reconstruction. References Belle Collaboration, Nucl. Instrum. Meth. A479,117-232 (2002). KEK Report 95-7 (1995). K. Hanagaki, H. Kakuno, H. Ikeda, T. Iijima, T. Tsukamoto, hep-ex/0108044. H.Ikeda et al., “A detailed test of the CsI(T1) calorimeter for Belle with photon beams of energy between 20MeV and 5.4GeV”, Nucl. Instrum. Meth. A441,401426(2000). 5. H.Ikeda, “Development of the CsI(T1) Calorimeter for the Measurement of C P violation at KEK B-Factory” , Ph.D Thesis, Nara Women’s University, 1999. 6. K.Kazui et al., “Study of the radiation hardness of CsI(T1) crystals for the Belle detector”, Nucl. Instrum. Meth. A394 46-56(1997). 1. 2. 3. 4.

CALIBRATION AND MONITORING OF THE ZEUS URANIUM SCINTILLATOR CALORIMETER AT HERA

M. BARB1 Physics Department, McGill University, 3600 University Street, Montreal, Quebec, Canada, H 3 A ZT8 E-mail: barbiOmail.desy.de

(For the ZEUS Calorimeter Group) One of the main components of the ZEUS detector at the HERA storage ring is the Uranium Calorimeter (UCAL). It has been running successfully since ZEUS started data taking in 1992. The UCAL is a Uranium-Scintillator calorimeter with equal = 1.00 f 0.05 ), a linear energy response response for electrons and hadrons (

9 a

%

and a high energy resolution of = 18% for electrons and = for hadrons. It covers 99.7% of the solid angle and is able to handle bunch crossing rate of up to 10.4 MHz. This performance demands a very precise calibration and a constant monitoring of the detector. In this paper we present the procedure to achieve a calibration accuracy of better than 3% and to maintain it stable to better than 2% for more than10 years.

1. Introduction

The Uranium Sampling Scintillator Calorimeter1Ji3 (UCAL) is the main calorimeter of the ZEUS experiment3 at HERA, a electron-proton collider at DESY with a center of mass energy of 324 GeV2. This calorimeter type is unique in the world. The intrinsic uranium radioactivity and the high uniformity of the signal response over the entire calorimeter allow a simple, accurate and stable calibration, making this calorimeter a central component for all precision measurements of the HERA physics programme. The calorimeter provides precise energy and arrival time measurements which are used for different applications from trigger decisions to jet reconstruction. This contribution describes the methods used to calibrate and monitor the ZEUS Uranium Calorimeter.

2. ZEUS Uranium Calorimeter Architecture The main design criteria of the calorimeter were as follows:

40 1

402

(1) high energy resolution - essential for any high precise measurement; (2) high time resolution - to detect and suppress background like cosmic rays and beam gas events which are out of phase with the bunch crossing; (3) good spatial resolution - to be able to identify efficiently scattered electrons; (4) uniformity - equal signal response throughout calorimeter is essential for any precision measurement (5) stability (6) fast response - to be able to handle the 10.4 MHz of the HERA bunch crossing rate with a minimal readout deadtime.

These requirements are partially achieved by using the sampling principle which allows to tune the electromagnetic and hadronic parts of the hadronic shower to compensate. A small sample depth is used to guarantee a high resolution in electromagnetic shower measurements. Each layer is made up of 3.3 mm of stainless-steel-clad depleted-uranium and 2.6 mm of SCSN-38 scintillator, corresponds to 1.0 XO (0.04 X)I and is uniform throughout calorimeter. The thickness of the cladding of the DU is 0.2 mm. Figure 1 shows the ratio f of the electron to hadron signals as a function of the energy EK of the incident particle in the calorimeter. The UCAL has depleted-uranium (DU) as passive material. It has the advantage of producing slow neutrons by fission which helps in compensating the losses in the hadronic shower. The uranium acts also as an absorber of electromagnetic particles generated in the electromagnetic part of the hadronic shower, enhancing the compensation mechanism. As active material scintillator is used. It contains a large fraction of hydrogen atoms that produces the signal by interacting with the slow neutrons from the hadronic shower. The calorimeter is divided in three different regions, namely the forward (FCAL), the central (BCAL) and the rear (RCAL) sections, covering the polar angle ranges 2O - 40", 37O - 129' and 128' - 177". The FCAL and RCAL are subdivided in 23 modules each and the BCAL is subdivided in 32 modules. Each module consists of towers segmented longitudinally into two parts. The inner part constitutes the electromagnetic section (EMC) with a depth of 25 XO (1.1XI) throughout UCAL and the outer one constitutes the hadronic section (HAC) with a depth of 6 XI in FCAL, 3 XI in RCAL and 4 XI in BCAL. The EMCs are further segmented in 4 (2 in RCAL) sections and the HACs are subdivided in 2 (1 in RCAL) sections of 3.0 XI in FCAL (RCAL) and 2.0 XI in BCAL.

403 1.1

I

1.0

-

2 0.B

-

0.8

-

s

.”.-

0

QO

“ “ I

~ - - - p - - O - - l s eI

0-

%O o e/n

0

0 0.7 Oe6

-

0 0 -....... r--e/mip

3

+ I

0.5 0.1

0.5

....

I

1.0

I

5.0

. . . I

0

e/P

0

e/h

I

10.0

I

*... I

50.0 100.0

Figure 1. The ratio 2. The calorimeter is fully compensating for energies above 5 GeV and approaches the electron to mip ratio at low energies.

Due to the compensation and uniformity, the calorimeter has a high = 35% The electromagnetic energy is hadronic energy resolution of

9 a* measured with a resolution of 9= %. The UCAL can also measure the

time with a resolution better than 1 ns and the position with a precision better than 1.3 cm vertically and 0.8 cm horizontally. This allows a good identification of the scattered electron in DIS process and an efficient suppression of background events out of phase with the bunch crossing. 3. Calorimeter Readout

The calorimeter readout4 is designed such as to ensure a very stable system able to handle a high bunch crossing rate with very low deadtime. The light of each EMC and HAC cells is collected by 2 wave-length shifter guides (WLS) mounted at the two sides of a tower. The light is then sent to a photomultiplier (PMT) that is connected to a front-end card where the signal is splitted 5-fold: 0

0

0

0

to two shaping-sampling with different (high and low) gains; to a current sum node, serving the Calorimeter First Level Trigger (CFLT); to a current integrator averaging the input current over 20 ms and providing the uranium calibration signal measurement; to a termination resistor.

404

The signal is sampled and stored in a 58 cells deep pipeline chip continuously clocked at the 10.4 MHz bunch crossing rate until a first level trigger occurs. Then the readout stops for 10 ps and 6 samples for each gain are transferred to a 8 cells deep buffer chip clocked at 0.6 MHz to the digital cards (DC’s). This scheme allows 5 ps for a trigger decision and leads to a deadtime of 1% for a first-level trigger rate of 1 kHz. Each DC has 4 analog-to-digital converters (ADC’s), a digital signal processor (DSP) and a memory in which the calibration constants are stored. The latter are used subsequently in the DSP’s to correct the 12-bit digitization results from the ADC’s and to reconstruct the time and energy to be used in the second-level trigger and to be sent to an event builder. The 12-bit digitization, along with the 2 gain scales, allows an effective dynamic range of 17 (15) bits in FCAL and BCAL (RCAL). The dynamical range (high gain)(low gain) per Calorimeter cell is (0-24 GeV)(O-530 GeV) in FCAL, (0-18 GeV)(O-380 GeV) in BCAL and (0-18 GeV)(O-90 GeV) in RCAL, while the noise level is dominated by the uranium activity (UNO) of x15 (x25) MeV in the EMC (HAC) sectors. 4. Calibration Method and Monitoring

The calibration5 of the calorimeter is based on its uniformity and on the natural uranium activity (2-10 MHz per calorimeter cell) that provides the absolute energy calibration. A total of 16 modules were examined in test beams at CERN2, where the construction tolerance and its implication on the uranium signal calibration was extensively investigated. An averaged non-uniformity of less than 1%were measured, the main contribution occurring in regions across the modules and towers boundaries. Cell-to-cell calibration was measured to be better than 1.5% (2%) for EMC (HAC) sections. The nominal UNO currents are chosen such that the response of each cell to mip’s scales with the number of layers in the cell. The calibration constants were transported from test beam to ZEUS and then a set of corrections are applied in order to keep the UNO current at the nominal level. The corrections are obtained from a full electronic calibration, an adjustment of the high-voltage applied to the PMT’s and a measurement of offline energy normalization factors. In the full electronic calibration, the calibration constants are adjusted for the front-end cards and the digital cards. A digital-to-analog converter (DAC) on board of each front-end card allows to set a very precise reference voltage used to measure the pedestal and gain of the pipeline and buffer chips. In addition this voltage is injected into the buffer-multiplexer and used to measure the ADC-to-VOLT constants. A very stable and known amount of charge is

405

used to calibrate the gain of the shapers. Finally, the uranium noise offset is measured. All the constants are subsequently stored in the DSP on board of each digital card, as already mentioned in the previous section. The constants are found to be stable to much better than 1%over a week. Figure 2 shows a schematic of a front-end card.

I’M

41 J 4

CIN

CLK

COllT

IWLSE

444 CLK

fi

US7

CLK R S T H * M C

CONTHOL SIGNALS

Figure 2.

A calorimeter front-end card.

In addition to the full electronic calibration, the high voltages (HV) applied to the photomultipliers are adjusted such that the response of the detector to the UNO current is kept to the nominal one within less than 1%. The uranium current measured at DESY is used for such purpose. Figure 3 shows the relative deviation of a measured UNO current to the nominal one after a HV adjustment. Figure 4 shows the stability of the UNO current measured at different time in a period of 36 days. Ultimately, the UNO current is recorded daily for use as offline normalization correction factors. Beside the above corrections, a constant monitoring ensures a good quality of the signal. The calibration stability is checked daily and the bad channels are marked and removed from the readout. Dedicated runs are used for this purpose, namely:

406

z

400

E 350

% 300

2 250 U

g 200

E 150

2 100 f3

"

50 0 ~

~

25

"

50

'

~

75

~

100

~

125

'

~

~

150

175

Nominol UNO for o FCAL module

Entries $/ndf 21.58 Constant

FCAL module after an HV trimming

Si ma

z

~

"

~

200 225 PMT number

'

~

'

/

22 31.61

0.2754E-02

20 15

10

0-0.03

-0.02

-0.01

0

0.01

0.02

0.03

Relative Deviation from Nominal UNO

Figure 3. Nominal UNO current measured in test beam for a FCAL module and the relative deviation of a UNO current measured at DESY after a HV adjustment showing an agreement to much better than 1%.

(1) Pedestal (Ped) runs; (2) Charge injection (Qinj) runs; (3) Uranium noise (UNO) runs; (4) LED runs; (5) Laser runs.

The pedestal runs are used to check the whole readout chain from the photomultipliers to the digital cards. Any channel with bad pedestal is marked and removed from the readout. The UNO and Qinj runs are used for the same purpose of checking for bad channels. The first monitors the uranium activity tracing down malfunction of photomultipliers or HV systems and the latter checks the stability of the electronic readout. Figure 5 shows the results of the measurement from these three runs. We can clearly recognize single bad channels or groups of them. They are marked and fixed whenever possible. The LED and laser runs measure the light (LED) and laser pulse3 injected in the WLS guiders and can be used as a double check for the performance of the photomultipliers, the HV system and the electronic readout. They are also used to study the photomultipliers gain stability and linearity over a wide

~

407

RMS of 0.2 X Period of 36 days UNO token after calibration 0

80

-

60 -

40 -

20

-

0

~

Figure 4.

"

'

1

"

~

~

"

"

~

'

~

~

"

~

~

~

~

'

~

1

1

'

'

UNO current measured at different time showing high stability.

dynamic range. As a result of all corrections and monitoring, the calorimeter has typically 2% of the total readout channels declared as bad, the number of bad cells is usually less than 2 and the calibration is stable to 1 - 2%.

5 . Summary

The calorimeter uniformity and the compensation ensures a high hadronic energy resolution. The natural uranium activity provides the absolute energy scale and an in situ calibration is performed to correct for changes in the readout electronic. The nominal uranium noise current from the test beam measurement is re-established by adjusting the high voltage based on the measured uranium current at DESY. The latter is recorded on daily basis for use as offline normalization factor. A complete monitoring of the readout system is performed and bad channels are removed from the readout. The calibration is stable to 1 - 2%. However it is worth mention that the absolute calibration from the test beam to ZEUS is good only to 3 - 4% which is taken into account by using physics constraints6.

'

408

60 40

100

50

2000

150

20

0 Oinj x PMT number

100

50

150

Oinj RMS x PMT number

400

300 200 100

2

0

0

100

200

UNO x PMT number

100

200

UNO RMS x PMT number

2

0 50

-2 50

100

150

200

100

150

200

0

Pedestol x PMT number

Pedestol RMS x PMT number

Figure 5. Charge injection, U N O and pedestal runs used to monitor the readout channels. Those ones outside the calibration tolerance are marked as bad and removed from the readout.

Acknowledgments

I acknowledge the support of the ZEUS collaboration and, in particular, the ZEUS calorimeter group for the discussions, comments and materials used in this paper. I am thankful to Rik Yoshida for his excellent remarks and advises and to Wolfram Zeuner for his critical reading of the draft. References 1. M. Derrick et al, N I M A 3 0 9 , 77 (1991). 2. A. Andresen et al, N I M A 3 0 9 , 101 (1991).

3. The ZEUS Detector Status Report 1993, D E S Y , (1993) 4. A. Caldwell et al, ZEUS-Note, 92-022 (1992) 5. J. Crittenden, The Performance of the ZEUS Calorimeter, Proceedings of the I/

International Conference o n Calorimetry in HEP, Brookhaven National Laboratory, USA, 1994 6. M. Wing, Determination of Jet Energy Scale Using the ZEUS Detector, Submitted to the Proceedings of the X International Conference o n Calorimetry in H E P , Calthec, USA, 2002

CALIBRATION OF THE PHENIX LEAD SCINTILLATOR CALORIMETER

H. TORI1 Kyoto University, Oiwake-cho, Sakyo-ku, Kyoto, 606, Japan E-mail: htoriiObnl.gov (For the PHENIX Collaboration)

In the early summer of 2000, the PHENIX experiment for heavy-ion physics at RHIC began, and that for spin physics was started in 2001. In this experiment, the electromagnetic calorimeter (EMCal) plays an important role in detecting photons and electrons/positrons. In order to cover topics in both areas of physics, e.g., thermal photon measurements in heavy-ion physics, and prompt photon, ?yo and weak boson measurements in spin physics, the EMCal must cover a wide energy range from a few hundred MeV to 80 GeV. Spin physics also requires the energy measurement to be within 2% accuracy for measuring cross sections of prompt photons and ?yo production with 10% errors, because the cross sections have steep ) The PHENIX EMCal consists of a lead scintiltransverse momentum ( p ~ slopes. lator (PbSc) and lead glass (PbGl). In this paper, we will report the performance of the PbSc and the achievement of 2% accuracy.

1. Lead Scintillator CaIorimeter(PbSc)

The PbSc is a Shashlik-type sampling calorimeter made of alternating tiles of lead and scintillator. The basic block is a module consisting of four towers which are optically isolated and are read out individually'. The module has a 5.52 x 5.52 cm2 cross section and 37.5 cm length. The most important feature is its find segmentation, which is the 0.01 rapidity. The two photon from a r0 of p~ = 30 GeV/c will be separated. In this section, some basic performance known from beam test are will be shown.

1.1. Energy Linearity and Resolution From several beam tests at B N L 1 ~ 2 ~and 3 ~ 4 CERN5, the PbSc was found to have a nominal energy resolution of CTE/E = 2.1% @ 8 . 1 % / a from 0.5 to 80 GeV/c as shown in figures 1 and 2, where 69 represents the quadratic sum.

409

410

2

0

-2

-8 0

1

/

I

!

/

2

3

4

5

6 GeV

GeV

Figure 1. Energy linearity measured at BNL(1eft) and CERN(right) beam test.

s

-

14 -

-

A

.... '....i.I . *. ......,. 10 ..'.'..*k. ...... '.+.... . 0-...... 12 -

CERN 10-80GeVIc e

-

~. -..-.. 1.2%+6.2Y'

.i,

'.I...

it

--

%.

s::::::<,.:-. ............

4-

2-0

2.10/088.1Y'

.......

. . I .

6-

-

BNL 0.5-5.0GeVIc e

.

.....

'--'..I. .-sL... ......

'

1

'

1

1

1

'

I

....

1

I

Figure 2. Energy Resolution measured at BNL and CERN beam test.

1.2. Position Resolution The position resolution is very important for the high p~ 7ro measurement, because the distance of two photon are close. Figure 3 depicts the position resolution from BNL and CERN beam test. The both results are consistent within systematic error and are 1.4 mm + 5.9 m m / , / m .

2. Energy calibration in PHENIX Configuration In this section, we will show the strategy of energy calibration and how well the energy calibration has been achieved.

41 1

12

.,

E E

lo 8-

A

%..

CERN 1O%OGeVlc e.

**..,

..

0-

....... 1.4mm + S.Bmml\lELo.v)

-*..i...

+..4.a....

4-

'"+.-+-)...)..+_..__._ #

ao , , , , , . '

BNL 0.55.WeVk e .

~

,

, . . ' , ( '

s

'

l

.

s

'

l

0

'

Figure 3. Position Resolution measured at BNL and CERN beam test.

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

1

Figure 4. Comparison of 1.0 GeV/c K + energy spectrum (dotted line) at AGS beam test and 0.9-1.1 GeV/c .rr+ energy spectrum (line) at PHENIX Au+Au collision & = 130AGeV.

2.1. Stmtegy of Energy Calibmtion Before installing into PHENIX configuration, all towers are calibrated by using cosmic muon traversing laterally, which energy is 42 MeV. The calibration parameters are transported into PHENIX configuration by laser monitoring system. The UV laser is used for laser monitoring system6. The light is delivered through splitters into 3888 modules. The light intensity is monitored with PIN diodes at each intermediate splitter that are used for normalization. The additional misalignment is calibrated by charge pion from collision point traversing lengthwise. The nominal energy of ionization is 270 MeV. 2 . 2 . Absolute Energy Calibration

Figure 4 depicts the comparison of energy deposits by 1.0 GeVJc n+ at the = BNL beam test and 0.9-1.1 GeV/c n+ at PHENIX Au+Au collision 130AGeV with a 0.27 GeV minimum ionization peak. The n+ is identified by the momentum measured by the tracking system and the time of flight measured by the EMCal itself. The relative energy scale5) for all towers during Au+Au collision is already calibrated using the minimum ionization energy of a cosmic muon traversing laterally and that of a charged n traversing lengthwise.

412

The minimum ionization peak of 1.1 GeV/c(O.S GeV/c) x+ is predicted t o be higher(1ower) than that of 1.0 GeV/c x+ by 2%. The above comparisons are consistent within 2% accuracy. Figure 5 shows the E / p spectrum for the electron sample, which has a peak at 1.02, and the two-photon invariant-mass spectrum with a xo peak. The electron sample required a hit in Ring Image Cerenkov counter(R1CH). The background in the E / p spectrum is already subtracted. Because of the 2% error in the momentum measurement of the tracking system, the E / p spectrum indicates that the energy measurement has a 1.02f0.02 accuracy. The xo mass is evaluated after background subtraction by the event mixing method and is consistent with a nominal xo mass within a 2% statistical error.

0

0.2

0.4

0.0

0.1

I

1.2

1.4

w

0

0.1

0.2

0.3

0.4

0.5

MouCQW

Figure 5. Ratio of measured energy and momentum for electron sample (Left) and the two-photon invariant mass spectrum (Right) at PHENIX AuSAu Collision fi = 130AGeV.

3. Conclusion

From these three independent measurements, 1.0 GeV/c x+ minimum ionization peak, E/p spectrum, and xo mass, we can conclude that the absolute energy scale has 2% accuracy and our physics requirement is satisfied.

References 1. G. David. et. al.: IEEE Trans. Nucl. Sci. 42:306-310(1995) 2. G. David et. al.: IEEE Trans. Nucl. Sci. 43:1491-1495(1996) 3. G. David et. al.: IEEE Trans. Nucl. Sci. 45:692-697(1998) 4. E. Kistenev et al.: Proc. 5th Int. Conf. Calorimetry in High Energy Physics, World Scientific(l994) 211. 5. e-print nucl-ex/0202009 T.C.Awes, et. al. High Energy Beam Test of the PHENIX Lead-Scintillator EM Calorimeter 6. G. David et. al. : IEEE Trans.Nucl.Sci. 45:705-709,1998

DO CALORIMETER CALIBRATION

URSULA BASSLER LPNHE, 4 , place Jussieu, 75252 Paris Cedex 05, h n c e E-mail: basslerOinZp3.fr (On behalf of the

D 0 Collaboration)

The front-end electronics of the D8 Ur/liquid Argon Calorimeter has been replaced for the Run 2 of the Tevatron to operate with the new shorter time separation between particle bunches (396/132 ns compared to 3.5 ps for Run 1). The signal shaping was changed to accommodate this shorter time between bunches. That change led to the replacement of the existing electronics calibration system so that the injected calibration pulse shape would more closely match the actual detector pulse shape. First results using the online calibration system as well as the off-line calibration from reconstructed object are presented.

1. Introduction Since the start of the Tevatron Run 2 in March 2001, the upgraded DO detector is in operation. The DO Ur/liquid Ar-calorimeter, divided into a central barrel (CC) and two end-cap calorimeters (EC), with its excellent hermeticity, its fine granularity and quasi-compensation has been kept in place, although the reduced interaction time required an complete exchange of its electronics and its online calibration system (for more details see reference'). The original tracking system has been replaced by a Silicon Micro Vertex Detector, a Scintillating Fiber Tracker and the introduction of a superconducting solenoid delivering a 2 T magnetic field, allows now a precise measurement of charged particle momenta. First results on the linearity of the calorimeter electronics and systematics effects on the on-line calibration are detailed. An overview is given on off-line calibration studies using reconstructed objects to obtain an intercalibration of the calorimeter modules in the azimuthal angle cp, the absolute energy scale from the reconstruction of the 2 resonance and the comparison of E / p in W-events and the Jet Energy Scale.

413

414

2. On-line Calibration

The Calorimeter calibration system2 is divided in twelve identical units used for the liquid Argon calorimeter and one slightly modified unit for the Intercryostat Detector (ICD). Each unit is composed of one Pulser Board, its power supply and six active Fanout Boards and provides the calibration signal for one calorimeter quadranda. The Pulser Boards are controlled via a Serial bus to a VME I 0 register, in order to set the amplitude, the delay of the calibration signal and the channels to be pulsed. The Pulser Board delivers both a DC current corresponding to the chosen pulse height for each selected channel and a differential ECL command signal to the Fanout Boards located at 25 m in the preamplifier boxes on the top of the calorimeter cryostat. The pulse heights are set through a 18 bit DAC, with a 100 mA maximal current delivered, and the delays through 6 programmable 8 bit delay lines with a 2 ns step size. 16 switches located on each Fanout Boards generate a pulse on the reception of a command signal, converting the DC current to a calibration pulse, which is then distributed through the Preamp-Box backplane to 48 calorimeter channels. Both the Pulser Board and the Active Fanout have been tested and shown to provide a pulser signal with a linearity at the per mil level and all the currents delivered are uniform within 0.2% between all boards and 0.1% within one board (figure l a ) . After the final installation all the pulse shapes have been measured to estimate systematic effects on the signal amplitude, the timing and the charge injected (figure lb ).

-

?40

F

~

! I

j ~

-000+0 002

0

0 002

-LA c 004
Figure 1. Th e mean current/DAC delivered by the pulser boards is 825 mA/DAC with an uniformity of 0.2% for all 15 Pulser boards produced (a) and 0.1% within one board (b). (c) Pulse Shape measured a t Pre-Amp Input, Amplitude(mV) vs time(ns)

aOne calorimeter cryostat (CC or EC) is devided in 4 quadrads.

41 5

2.1. Linearity Determination The linearity of the calorimeter electronics and the pulser system is verified3 with a comparison of the electronics response (in ADC counts) vs. the pulse height (in DAC units) for each gain path ( x l and x8). The corresponding fitted slopes, al and as, are of the order of 0.25 ADC/DAC and the ratio g = a s / a l N 1, which leads to 1 DAC corresponding roughly t o 1 MeV. Figure 2 shows the residuals of a linear fit to the response after appropriate scaling by a factor 8 of one of the gain paths: similar structures in the residuals of the two gain paths are due to a non-linearity in the read-out electronics, where as differences are due to the pulser system. For all twelve pulser units we observe a similar structure: over a large part of the dynamic range, the electronics chain displays a very linear behaviour with deviation below f 1 0 ADC (< 0.3%),however for DAC settings below 1000 (8000) units for gain 8 (gain l),a non-linear behavior is observed. Within this region, differences between the two gain paths can be seen for pulser 1, 4, 5 and 8: more detailed studies revealed an offset at DAC = 0 for these pulsers, which has been corrected for in the following.

0

5

10

0

5

10

0

5

10

0

5

10

v/lOOO DAC units Figure 2. Residuals to a linear fit of the calorimeter response t o different DAC settings for gain x 8 (cross) and gain x l (circle) scaled by a factor 8 for the DAC setting and the response. Each figure corresponds to the mean response from channels pulsed from one pulser unit.

416

2.2. SCA Non-Linearity The observed non-linearity at low energy has been traced back to the Switched Capacitor Arrays, which are storing the analog calorimeter signal until the Level 1 and Level 2 trigger decisions have been taken. For the ADC t o energy conversion a universal correction function has been derived (figure 3a), using two free parameters that have been adjusted for all channels. After applying this correction function, the residuals are better than f 5 ADC over the whole range for both gain paths. 100

.

,

I

,

.

I

, . , .

I

-

50-

$ \

i

A-

O-<

M-

-lw

"

"

"

'

~

"

'

The effect of this correction on the energy reconstruction (figure 3b) results in a shift of 200 MeV for energies above 500 MeV, when read out with the gain 8 and a shift of 1 GeV when read out with gain 1. Reconstructing the mass peak for 2 + e+e- applying the non linearity correction results in a shift of about 10% with a peak position at 89.6 GeV before any further calibration factors have been applied. 2.3. Gain Determination

For all calorimeter channels, the gain calibration factors have been determined from the deviations of the slope ADC/DAC from its ideal value of 0.25. The factors obtained show a dependence with respect to the Pre-Amp type and the capacitance of a given cell. The dispersion of the coefficients for electromagnetic (hadronic) channels is about 5% (10%). Systematic shifts of the slope values can be observed for different Pre-Amp Types. Part of these differences are due to the injection of the calibration signal at the Pre-Amp input, outside of the cryostat. The signal is reflected towards the calorimeter cells and has therefore different shapes depending on the capacitance of the cell. The effect are biggest on hadronic cells with a high

417

f

Figure 4. Signal shape (ADC vs. delay unit/ns x 2) for an electromagnetic (a) and an hadronic (b) channel after read-out obtained by incrementing the delay of the pulser signal: signal start is at the right. (c) Response to reflection measurement for three different channels.

capacitance, where differences in the timing can produce large tails in the distribution of the calibration coefficients (figure 4 a,b). Corrections for these effects have been derived for the calibration coefficients and for an improved delay setting of the pulsers. Models of the electronics chain have been setup to evaluate the differences between the electronics response to a calibration signal and a detector signal4. To render these models realistic all stable parameters of the signal path from the detector to the Pre-Amp input have been determined from reflection measurements. The reflected response to a step function is shown in figure 4c for three different channel types. Quantitatively the values for the cable resistivity outside and inside the cryostat, the inductance of the feed-through and the signal-strips as well as the capacitance of each cell have been determined and used in a simulation model. Recently estimated corrections to the calibration coefficients evaluate these effects below the percent level for electromagnetic channels, when pulsed close to the signal maximum. 3. cp intercalibration

In order to take into account variations between the different calorimeter modules or inhomogeneous distributions of non-instrumented material in front of the calorimeter, the uniformity of the calorimeter response in cp is evaluated using the number of events and the energy deposited in each of the 64 - cp modules covering one ring of 17 = 0.1.

418

With N ( E 0 ) being the number of events for energies greater than EO and S(E0) being the summed energy above the same threshold, two calibration coefficients are determined ai = EiNi(E2) - S ( E i ) / E T e f N T e f ( E Te fS(E,.ef) ) and p = Ei - aiE,,j, with Ei being adjusted in each module such that Ni(Ei) = N,.,f(E,.,f). The validity of the 9 calibration coefficients is evaluated looking at the “flatness”, which is defined as the RMS of the energy distribution for all modules in one 9-ring. Figure 5 shows the improvement of this distribution for cells in Layer 1 for the CC after the calibration procedure.

Figure 5. Flatness distribution for 1111 < 1 before (thin line) and after (bold line) the pintercalibration in the first layer of the electromagnetic calorimeter. No calorimeter cells are located at the position 11 = 0.

4. Energy Scale from 2’-resonance

The absolute energy scale is obtained by reconstructing the mass of the 2’ resonance, and adjusting the value t o the LEP 2’-mass value. The measured Zo-mass value is obtained by fitting a Breit-Wigner distribution to the dielectron mass spectra, which is then compared to the value obtained from Monte-Carlo simulation. The calibration coefficients are obtained for different regions in 77 by maximizing the likelihood function of (M:e)tTue= M:e(l ~ k ) (+ l EL), where ~i is applied to a given 77 region of the calorimeter. The reconstruction of the Zo-peak after calibration is shown in figure 6 for all 2’ events and the events with both electrons in the CC. The indicated mean values are obtained from the fit to the peak region.

+

420

measurements, the response of the jet should be know at the particle level and corrected for all detector effect. As an example for the cone algorithm, the jet - E?ffset/(Rjc:~OR,c,O:e). Jet The offset energy, jet energy is defined as E:ttl = (Emeas E30eftfsetcorrects the measured energy for the contribution from the underlying event, pile-up and noise and is determined from minimum bias events. Rj",0reis derived from Monte Carlo simulations, and relates the energy measured inside the cone to the energy outside the cone but-belonging to the jet. To determine the calorimeter response R;::", y-jet events are selected and the transverse jet-energy compared to the calibrated photon energy. The response function obtained from a detailed GEANT simulation is shown in figure 7(b): different parameterization give a consistent results and are showing the response to jets above 50 GeV being above 90%. 7. Perspectives

A first on-line calibration of the Run 2 DO calorimeter electronics has been obtained. After correcting the response for non-linearity effects at low energy, only small scale corrections are needed. The dispersion of the electronics response is within 5% for em-channels and 10% for hadronic channels. Detailed studies of systematic effects on these coefficients aim to reach an on-line calibration with a precision of 1%. The procedures for calibration of the calorimeter with reconstructed events have been established, which allowed to verify the absolute energy scale of the calorimeter using the Zo resonance and the E / p reconstruction. Low mass resonances, J / P s i and T,are currently studied and will allow to verify the linearity of the calorimeter. Besides the use of y-jet events for the reconstruction of the Jet Energy Scale, the use of Zo+jet events is envisaged with high statistics and a dedicated calibration for bjets is in work. Acknowledgments Thanks to all the colleagues from the DO calorimeter group, the EM-id group and the Jet-Energy-Scale group who did the studies and produced the distributions shown in these proceedings.

References 1. N. Parua,these proceeding (2002). 2. P. Cornebise et al., DO-note 3731 (2000). 3. R. Zitoun, DO-note 3997 (2002). 4. R.Chiche et al., DO-note 3914 (2001).

C A L I B R A T I N G A L O N G I T U D I N A L L Y SEGMENTED CALORIMETER

0.LOBBAN Texas Tech University,Lubbock TX 79409, USA E-mail: [email protected]

Three different methods of setting the hadronic energy scale of a longitudinally segmented calorimeter system are compared with each other. The merits of these methods have been studied with testbeam data from the CDF Plug Upgrade Calorimeter. It turns out that one of the calibration methods introduces a number of undesirable side effects, such as an increased hadronic signal nonlinearity and a dependence of the reconstructed energy of hadrons on the starting point of their showers. These problems can be avoided when a different calibration method is used.

1. Introduction

Finding the correct energy scale of a calorimeter which is divided into an electromagnetic and a hadronic section is not trivial'. The energy scale is the constant which converts the signal from the calorimeter to units of energy. If a calorimeter is longitudinally divided into two sections with different material compositions, then an energy scale is needed for each section. The energy scale of a calorimeter is (at least initially) determined from a testbeam of particles with a well defined energy and of the type that the calorimeter will measure. In most longitudinally segmented calorimeter systems, since the electromagnetic section is deep enough t o fully contain the shower from an incoming electron, determining the energy scale of the electromagnetic section is straightforward. A testbeam of electrons is sent into the electromagnetic section and the ratio of the signal from the electromagnetic section (ADC counts) with the energy of the incoming electrons (GeV) is defined as the energy scale for that section. Determining the energy scale of the hadronic section, however, is tricky. In most calorimeter systems with two longitudinal segments, hadrons deposit their energy in both segments. We have studied three methods to determine the energy scale of the hadronic section.

42 1

422

2. CDF Plug Upgrade Calorimeter

The present study was carried out with testbeam data of the new endplup calorimeter of the Collider Detector at Fermilab (CDF). This calorimeter system was recently modified for Run 11 at the Tevatron. It covers the pseudorapidity range from 1.1 - 3.6 and consists of a 0.75Xi,t deep electromagnetic section and a 7.4Xint deep hadronic section (at 6 = 23’). The electromagnetic section has a lead/scintillator sandwich structure, with 4.5 mm thick lead plate: alternated by 4.0 mm thick plastic scintillator plates. In the hadronic section 6.0 mm thick scintillator plates are sandwiched between 50.8 mm thick iron absorber plates’. The electromagnetic and hadronic sections are both noncompensating devices with e / h values of 1.43 and 1.36, respectively. In this study we have used experimental data taken with a special module built for testbeam purposes. This module consists of four 15’ sections which are replicas of the actual Plug Upgrade calorimeter. One of these four wedges was built without an electromagnetic compartment.

3. The calibration methods

3.1. Electromagnetic section energy scale The energy scale of the electromagnetic section, A , was determined with beams of electrons (ranging from 8-180 GeV) sent into the electromagnetic section. The value of A is constant within experimental uncertainties over a wide range of electron energies. The weighted average was found to be A = 128.1 counts/GeV.

3.2. Hadronic section energy scale

-

Method I

In this method, a beam of pions with a well defined energy is sent into the electromagnetic section of the calorimeter (with, of course, the hadronic section directly behind it). Pions which pass through the electromagnetic section without undergoing a nuclear interaction (“Penetrating pions”) are selected. For only the penetrating pions, the ratio of the signal from the hadronic section (ADC counts) with the energy of the incoming pions (GeV) is defined as the energy scale of the hadronic section. Pions with energies ranging from 8-160 GeV were used to find the value of the hadronic section energy scale BI with this method. Figure 1 shows the value of BI as a function of the energy . deposited in the hadronic calorimeter section by the penetrating pions. Note the logarithmic scale of the horizontal axis.

423

120

lo2

10

Energy (GeV)

Figure 1 . Th e energy scale of the hadronic section, B I , found using Method I as a function of the energy deposited in the hadronic section of the calorimeter by penetrating pions.

3.3. Hadronic section energy scale

-

Method II

In this method, a beam of electrons is sent directly into the hadronic section of the calorimeter and the ratio of the signal in the hadronic section with electron energy is defined as the energy scale of the hadronic section. The underlying philosophy of this method is that the relationship between deposited energy and resulting calorimeter signal should be established in the same way for all segments of the calorimeter system. In practice this method is usually impossible to implement in an experiment, since the hadronic section is shielded from the particle source (2. e. the interaction region) by the electromagnetic section. However, in our testbeam setup we could study this method thanks t o the fact that part of the tested calorimeter was not equipped with an electromagnetic section. Electrons with energies ranging from 11-177 GeV were used t o find the value of the hadronic section energy scale B I I . It is constant, B I I = 173.5 counts/GeV, for a wide range of electron energies.

3.4. Hadronic section energy scale

-

Method 111

In this method, which to our knowledge has not yet been applied in any experiment, pions with a well-defined energy are sent into the electromagnetic section of the calorimeter, and the energy scale of the hadronic section is chosen such that the average energy reconstructed for penetrating pions is equal to that for nonpenetrating pions. The underlying philosophy of this method is to avoid any dependence of the reconstructed (hadronic) energy on the starting point of the showers. As

424

we will see later, such a dependence is an inevitable consequence of the application of Method I. Unlike Method 11, this method has the advantage that it can be implemented in practice and that it can be applied an situ using reconstructed tracks of isolated particles produced in the interactions studied by the experiment. In order to avoid the effects of shower leakage, we used several low energy (<20 GeV) pion runs to find the value of the hadronic section energy scale B I I Iusing this method. We found an average value B I I I=186.5 counts/GeV. 4. Experimental consequences of each method The reconstructed energy is defined as Ere,,,

=

EM signal (ADCcts) A

(ADCcts) + HAD signal B I ,11 , I I I

(1)

where A is the energy scale of the electromagnetic section and B I , I I , I Idenotes I the energy scale of the hadronic section according to Method I,II, or 111. If we chose Method I to set the energy scale of the HAD section, we would need to specify the energy at which the calibration constant BI was determined, since this value is energy dependent. The other two methods are based on energy-independent calibration constants. In practice, we used a value BI = 151.4 cts/GeV for our studies of the implications of Method I, i.e. the value obtained for the 56 GeV point. For each method, we (1) compared the reconstructed energy of penetrating pions (pions which penetrate the electromagnetic section without strongly interacting) with the reconstructed energy of nonpenetrating pions (pions which strongly interact in the electromagnetic section) and (2) plotted the ratio of the reconstructed energy with the beam energy as a function of the beam energy. 4.1. Dependence on the starting point of the showers Figure 2 shows the reconstructed energy distributions from an 8.6 GeV pion run for penetrating and nonpenetrating events. Method I was used to reconstruct the energy of the pions. The mean values of these two distributions differ by 15%. From these results, we conclude that using Method I to set the energy scale of the hadronic section of the calorimeter introduces a dependence of the reconstructed energy on the starting point of the pion shower. This is, in fact, a logical consequence of using this method with a noncompensating calorimeter. If penetrating pions are used to set the energy scale for the hadronic section, then it is only for this particular sample of pions that the energy will be reconstructed correctly. The value of Br found using Method I gives a larger

425

weight to the signal in the hadronic section. Pions which begin showering in the electromagnetic section do not fully benefit from this “boosting” of the hadronic section’s signal and will therefore have a reconstructed energy which is smaller than that of penetrating pions.

x’/ndf66.12

/

45

8.184 f 0.5566E-01

0

2.5

5

7.5

10 E,(CcV)

12.5

350

Entries

...

Mean

“E

I5

y’/ndf127.6

0

2.5

5

7.5

10

20

22.5

25

20

22.5

25

3904

RMS

xn

17.5

- yenelrating pions

12.5

I5

7.257 1.978 / 44

17.5

Emm(GcV) - nonyenetruling pions

Figure 2. Th e reconstructed energy of penetrating (top) and nonpenetrating (bottom) pions when Method I is used t o find the value of B.

Figure 3 shows the signal distributions for the same event samples, but this time Method I11 has been used to calculate the reconstructed energy = 186.5 counts/GeV). In this case, the mean values of the reconstructed energies of the penetrating and the nonpenetrating pions were found to be equal within the experimental uncertainties. When Method I1 was used, the mean values differed by 5%. Method I1 does not produce the same results as Method 111, because the electromagnetic and hadronic sections of the Plug Upgrade calorimeter have a different composition, which translates into different e / h and e / m i p ratios3.

-

4.2. Signal nonlinearity

For hadron showers, the average energy fraction carried by 7ros and other particles developing electromagnetic showers increases as a function of energy4.

426

t t

100

60

1607 6.761

I Mean

6.715f 1.663

*

Sigma

4oc

" 0

I Entries

n

n

I/

2

4

6

8

10

I

I

4

6 E,(CeV)

12

14

16

IS

.penelmlingpions

E,,(CeV)

2

3.1 14 0.4621E-01 0.3450E-01

8

Entries

10

I

3904

12

14

16

I8

- nonpenelrntlngpions

Figure 3. T h e reconstructed energy of penetrating (top) and nonpenetrating (bottom) pions when Method I11 is used t o find the value of B.

This causes an intrinsic signal nonlinearity for hadrons in all noncompensating calorimeters. This intrinsic nonlinearity appears when the pion energy is reconstructed using the calibration constant B I I I for the signals from the hadronic calorimeter section. If Method I is used to reconstruct the energy of pions, then one sees a larger nonlinearity than if Method I1 or I11 is used. This can be understood from the fact that as the energy of the pion shower increases, more and more energy is deposited in the hadronic section of the calorimeter. In Method I, the signal from the hadronic section of the calorimeter is "boosted" with respect to the signal from the electromagnetic section of the calorimeter. Figure 4 shows the ratio of the reconstructed energy and the deposited energy as a function of the latter for all three different calibration methods. Each data point represents the mean value of the distribution of E p y o n (Eq. l ) , where all pions, penetrating and nonpenetrating, have been taken into account. Each data set was fit with the function

The quantity k z l k l , which is a measure of the signal nonlinearity, is -40%

427

.=

1.1

1

1

W

\

A

E

1 -

P W" v

0.9

-

0.8

0 H

A

Method I Method II Method Ill

Figure 4. T h e ratio of the reconstructed energy and the deposited energy as a function of the energy deposited by pions showering in the CDF Plug Upgrade calorimeter. Results are given for the three calibration methods discussed in the text. Th e curves through the points are fits t o the fuction in Eq. 2.

larger when Method I is used as compared to Method 111. 5. Conclusions

We compared three different methods of setting the energy scale of the hadronic section of a longitudinally segmented calorimeter. The merits of these methods have been studied with testbeam data from the CDF Plug Upgrade calorimeter. It turns out that one of the methods (Method I) introduces a number of undesirable side effects, such as an increased hadronic signal nonlinearity and a dependence of the reconstructed energy on the starting point of the hadron showers. These problems are a direct consequence of the noncompensating nature of the calorimeters. They can be avoided when a different calibration method is used. References 1. 0. Ganel and R. Wigmans, Nucl. Instr. and Meth. A409,621 (1998). 2. CDF Collaboration, The CDF 11 Technical Design Report, Fermilab Internal Report 96/390-E (unpublished). 3. M. Albrow, et al., Nucl. Instr. and Meth., "Intercalibration of the longitudinal segments of a calorimeter system", accepted for publication (2001). 4. T.A. Gabriel, et al., Nucl. Instr. and Meth. A338, 336 (1994).

THE CALIBRATION OF THE MINOS DETECTORS

R. NICHOL~*,P. AD AM SON^, J. A L N E R ~ B. , ANDERSON~,D. AT TREE^, M. BARKER', A. BELIAS~,G. CRONE^, T. DURKIN~,N. FELT^, E. F A L K ~ , P. HARRIS^, L. J E N N E R ~ M. , KORDOSKY~,K. L A N G ~A. , LEBEDEV~, R. LEE^, N. L O N G L E Y ~M. , M A R S H A K ~P. , MIYAGAWA', D. MICHAELZ, R. MORSE~,J. MUSSER~,T. NICHOLLS~,J. OLI V E R~,G. PEARCE~, D. PETYT~,M. P R O G A ~B. , REBEL^, R. S A A K Y A N ~c, . SMITH^, P. SULLIVAN', J. THOMASa, P. VAHLEf A. WEBER', S. WOJCICKIk University College London, London W C l E 6 B T , UK 'E-mail: rjnOhep.ucl.ac.uk bRutherford Appleton Lab, Chilton, Didcot, Oxon, OX11 OQX, UK coxford University, Oxford OX1 3RH, UK Harvard University, Cambridge, MA 021 38, USA University of Sussex, Palmer, Brighton BN1 9QJ, UK f University of Texas at Austin, Austin T X 78712, USA g Macalester College, St. Paul MN 55105, USA University of Minnesota, Minneapolis MN 55455, USA California Institute of Technology, Pasadena, C A 91125, USA 3 Indiana University, Bloomington, IN 47405, USA Stanford University, Stanford, CA 94305, USA a

The MINOS experiment employs both a far and near detector t o study neutrino oscillations. Th e near detector is used t o characterize the beam before it has travelled far enough for oscillations to significantly occur. Oscillation measurements, including the energy spectrum of charged-current events at the near and far detectors, are made by comparing the events observed at the far detector t o those in the near detector. Th e relative calibration of the near detector and the far detector, and the different regions of the far detector itself, is paramount in the interpretation of the oscillation results. T h e absolute calibration is established using a small Calibration Detector while the relative calibration is established using cosmic-ray muons and a light injection system t o track short term variations in detector response. In this paper we present the overall strategy for calibration of the MINOS detectors and first results in Calibration using muons and light injection in the calibration detector.

1. Introduction

1.1. The MINOS experiment MINOS, Main Injector Neutrino Oscillation Search, is a long-baseline neutrino oscillation experiment. A beam of mainly muon neutrinos is produced at Fer-

428

429

milab, and directed at a 5.4kT detector, the Far Detector, located 735 km away in the Soudan mine, Northern Minnesota. A smaller 1kT detector, the Near Detector, will measure the beam on site at Fermilab before the neutrinos have travelled far enough for any significant oscillations to occur. Comparisons of these two measurements will enable the oscillation parameters to be accurately determined. The probability of a muon neutrino, of energy E , oscillating t o an electron neutrino over a distance L , in the simple two generation case, is given by:

1.2. Calibration Requirements

To accurately determine the oscillation parameters there are two requirements for calibration: (1) Relative calibration between the Near and Far detectors at the 2% level. (2) The absolute energy scale calibration at the 5%.

Differences in the response of the two detectors must be well understood. The second requirement is needed to accurately determine Am2. This is illustrated in Figure l, which shows the total energy in charged current beam neutrino events, for 3 different values of the oscillation parameters. The position of the dip gives Am2, depth of dip gives sin228.

1.3. The MINOS Detectors There are three MINOS detectors. The largest is the Far Detector, which is currently under construction in the Soudan mine in Minnesota. The Near Detector is being constructed on site at Fermilab. The smallest detector of the three is the Calibration Detector] CalDet, which was constructed at CERN in 2001, this is undergoing a series of test beam runs. All three detectors have been designed to be as similar as possible. They are tracking iron calorimeters, with plastic scintillator used as the active detector. Each plane of scintillator is divided into strips to give granularity. Wavelength shifting] WLS, fibre is embedded in the scintillator strips which captures the light. Optical fibres then transport the light to Hamamatsu multi-anode photomultiplier tubes, which are readout with custom electronics. The Near and the Far detector are magnetised which enables a better momentum measurement of muons.

430

CC mmsf. Evis (GcV)

200 150

100

50 0

CC mcmsf. Evis (GeV)

CC rawsf. Evis (GcV)

Figure 1. The total event energy in charged current beam related neutrinos, for different values of the mixing parameters.

2. Overview of Calibration Procedure The Near and Far Detectors are large underground detectors and as such an in-situ calibration is essential. Cosmic ray muons form the main tool for calibrating the detectors. Unfortunately the flux of cosmic muons in the Far Detector is very low, on the order of 500 muons per strip per month. Therefore calibration can only be performed on the timescale of a month or longer. A light injection system is used to monitor changes on a shorter timescale. The final part in the calibration is the Calibration Detector, which calibrates the hadron and electron energy response on the absolute scale. The most important use of muons is to provide the energy scale calibration across the detectors. Using the magnetic field or range, muons in the energy range 2-10 GeV can be identified by their curvature. The response of these

431

muons can be used to calibrate the energy scales across the Near and Far Detectors. The MINOS calibration system converts the response of the detector into Muon Energy Units, MEUs. An MEU is the energy deposited in lcm of scintillator by a minimum ionising particle at 90". The final component in the calibration is the CalDet. This is a small cut-out section of one of the larger detectors, and is running in a series of test-beams at CERN from 2001-2003. The main purpose of the CalDet is to determine the relationship between MEU and energy in the detectors, this is done through characterising the response of hadrons and electrons of known energies.

3. The Light Injection System 3.1. Overview

The light injection system was designed to account for short term gain drifts, and linearise the response of the electronics. Figure 2 is a schematic showing the components of the light injection system. Pih' relerencc

Ultra-hright bluc LED

' 7i i

dl&

1

I

Llgh! bJcCtlon Module

j i

t----i

Light distribution cont)cc!or

Figure 2.

Schematic of the light injection system

The principle behind the system is to inject light into the standard optical readout path, then calibrate the response using an independent light measure. Light from bright blue LEDs is distributed across 70 optical fibres in a pulser box. Each of these fibres transports the light to a different scintillator module. At the end of each scintillator module is the light injection manifold where the light from the fibre illuminates up to 10 of the WLS fibres within the module. From this point the injected light follows the same readout path as signals from the scintillator. Two of the fibres from the LED transport light to PIN

432

diodes. These diodes are used to independently monitor the light level, and there are 2 per LED to provide a degree of latency. At the Far Detector there are 32 pulser boxes each containing 20 LEDs. 3.2. Results

There are two main modes of operation of the light injection system. The first is used to account for PMT gain drifts which is achieved by flashing the LEDs at the same pulse height once an hour. Gain drift can be accounted for by comparing the response of the PMT and PIN diodes. At the CalDet the raw drift, before PIN correction, was measured to be only 2% over 4 days, as shown in Figure 3.

Time

Normaliesd A52

Figure 3. The raw PMT gain and LED drift at the Calibration Detector over a 4 day period.

The second use of the light injection system is to linearise the response of the PMTs and electronics. The LEDs are flashed over a range of different light levels, and the PIN diode is used to linearise the PMT response. The response of the PIN diode electronics is linearised using a charge injection system. Linearity curves taken at the CalDet have been used to test the performance of the light injection system. Over the course of 6 weeks, there were 4 linearity curves taken. For each channel successive curves were compared before and after calibration. The curves are calibrated by taking one point on the curve as a drift point and this point is used to remap a curve to take

433

Figure 4. Left: Procedure for testing the light injection system using linearity curves. Right: Light injection can correct for gain changes to the 1%level.

account of gain changes between the runs, as shown in the histogram on the left in Figure 4. The performance of the light injection system can be tested by selecting different points on the linearity curves and comparing the mean ADC for a given point with that predicted by the drift point calculation. The results of this procedure are shown in Figure 4. The upper right histogram shows the relative difference between measured, uncalibrated points and the lower right histogram the same for corrected points. The calibration corrects for gain differences to the 1%level, including raw differences of up to 20% which was due to an intentional high voltage change on the PMTs.

J RMS = 174.4 Chi2 I ndf = 86.43 I61 Landau Peak = 182.2 2.045 Landau Width = 41.64 1.972 Entries =7.576e+04 0 toRange= 594 0

0'

200 400 600 800 1000 1200 1400

ADCs Figure 5.

An example fit to the muon energy spectra.

434

4. Muon Calibration

Muons are used to calibrate the energy scale across the three MINOS detectors. They are also used to provide a strip to strip calibration within each detector. This strip to strip calibration accounts for the effect of: (1) Scintillator light yield in each strip. (2) Wavelength shifting fibre in the scintillator modules. (3) Optical readout fibres and connections. (4) PMTs and electronics. Calibration constants are determined from the muon energy spectra in each strip, either through fits to the peak of the spectra or by taking a truncated mean. An example fit is shown in Figure 5. The fitted function is an approximation to a convolution of a Landau and a Poisson. The muon calibration constants need to be stable over a time period of approximately a month. The stability of the muon constants has been tested with and without light injection corrections. In Figure 6 the upper histogram shows the fractional change in muon calibration constants for two runs taken two weeks apart. The lower histogram shows the difference after the data is

I

Fractional Change in MEU

I

50

= -0.0196

Mean

uy Fractional Difference

I Fractional Change in Corrected MEU I I

C';i

1 i h1 t cd

Mean

80 r

= -0.0 1245

60 r 4or 20 y y0.4

-0.3

-0.2

-0.1

-0

0.11-

'

012 0.2

'

'

013 0.3

'

'

20.4 4

Fractional Difference Figure 6.

Stability of muon calibration with and without drift point correction.

435

'0

30000

20000

"0

20

40

60

80

100

4ooaa

Raw ADCs

120

140

160 180 200

Raw MEUs

Figure 7. Electron event energy before and after muon calibration, the readout cable nonuniformity is corrected for. The lighter histograms are for WLS fibre readout cables, the darker for clear fibre cables.

corrected using the drift points. There is a 1%shift in the corrected differences due to a temperature difference between the two runs which can be corrected by a prescription: 0.3%/"C. Strip to strip calibration with muons has been shown to work at the CalDet using electron events, which are identified with a Cerenkov detector, event energy is plotted in Figure 7. Calibration accounts for a 20% difference in attenuation of the optical readout cables. 5. Conclusions The calibration of the MINOS detectors utilises cosmic ray muons and a light injection system to provide a relative calibration across the Near and Far Detectors. These methods have been shown to work at the CalDet with the light injection system correcting gain drift to the 1%level. The absolute calibration will come from the CalDet, after completing its test-beam running. Acknowledgments This work was supported in part by Particle Physics and Astronomy Research Council (U.K.), Department of Energy (U.S.) and National Science Foundation (U.S.) . References 1. P. Adamson et al, hep-ez/0204021 to be published in NIM.

FIRST RESULTS FROM THE MINOS CALIBRATION DETECTOR

P. VAHLE~*,P. AD AM SON^, J. ALNER', B. ANDERSON~,D. AT TREE^, M. BARKER^, A. BELIAS~,G. CRONE^, T. DURKIN~,N. FELT^, E. F A L K ~ , P. HARRISf, L. JENNERb, M. KORDOSKYa, K. LANGa, A. L E B E D E V , R. LEE^, N. L O N G L E Y ~ M. , M A R S H A K ~ P. , MIYAGAWA~,D. MICHAELZ, R. MORSEf, J. MUSSERj, R. NICHOLb, T. NICHOLLS', J. OLIVERe, G. PEARCE~,D. PETYT', M. P R O G A ~B. , REBEL^, R. SAAKYAN~, C. SMITHb, P. SULLIVANd, J. THOMASb, A. WEBERd, S. WOJCICKIk University of Texas at Austin, Austin T X 78712, USA 'E-mail: [email protected] University College London, London W C l E 6 B T , UK Rutherford Appleton Lab, Chilton, Didcot, Oxon, OX11 OQX, UK Oxford University, Oxford OX1 3RH, UK Harvard University, Cambridge, M A 02138, USA f University of Sussex, Falmer, Brighton B N l 9QJ, UK 9Macalester College, St. Paul MN 55105, USA University of Minnesota, Minneapolis MN 55455, USA ' California Institute of Technology, Pasadena, C A 91125, USA 3 Indiana University, Bloomington, IN 47405, USA Stanford University, Stanford, CA 94305, USA a

The MINOS Calibration Detector (CalDet) is a small version of the MINOS Near and Far neutrino detectors. A program of exposure t o beams of muons, electrons, pions and protons at the CERN PS will provide calibration of the calorimetric and topological response of the Near and Far detectors. In this talk, we briefly discuss the goals and design of the CalDet and present first results from the initial beam exposure

1. Introduction The measurement of neutrino oscillations in MINOS relies on the comparison of rates and spectra of neutrino interactions in two separate detectors. The MINOS Near detector is located at Fermilab, while the Far detector is located 735 km away in the Soudan iron mine in northern Minnesota. Since both detectors are large and assembled in-situ, neither can be exposed t o a calibration beam. Hence, a dedicated Calibration Detector (CalDet) has been constructed. The CalDet will be exposed to a beam of known particles at known energies so that

436

437

the response of the detector can be determined and characterized in terms of its response to muons. This characterization will allow the calibration to be carried over to the larger detectors, where a 5% absolute calibration is required to determine the neutrino oscillation parameters to the desired precision'. The CalDet data will also be used to provide more accurate physics input to the current MINOS Monte Carlo. Furthermore, since event shapes will be used to distinguish between neutral current and charged current interactions in the MINOS detectors, the CalDet will be used to study the topology of events with the goal of fine tuning pattern recognition algorithms. Finally, the measurements made at the CalDet allow the study of the hadronic response of sampling calorimeters at energies in the few GeV region. These interactions are poorly modeled in simulations, making such measurements interesting to the particle physics community a t large, but are particularly important for MINOS since the signature of oscillations will be most apparent at these lower energies. 2. Detector Specifications

The active medium of each of the detectors is 4.1 cm wide, 1 cm thick, coextruded plastic scintillator strips. In the CalDet 24 of these strips, backed by a 2.5 cm thick iron sheet, form a detector plane, 1 m x 1 m in area. The CalDet is composed of 60 of these planes, at a pitch of 5.9 cm. Consecutive planes of scintillator are rotated 90" with respect to the previous plane to allow for tracking in the horizontal and vertical directions. Each scintillator strip has a wavelength-shifting fiber, with a long attenuation length of approximately 650 cm, glued into a groove along the center of the strip. This fiber carries the scintillator light to a manifold at the edge of the detector. From there the light is transported via fiber ribbon cables to multi-anode photomultiplier tubes (PMTs). The effects due to the much larger size of the MINOS detectors are approximated in the CalDet by using 4 m long wavelength-shifting ribbon cables on one end of the detector and 6 m long clear ribbon cables on the other. All three detectors have identical segmentation and are constructed from identical materials. However, there are some differences in the read-out between the Far and Near detectors. The CalDet will run with each of the readout configurations to directly compare the two systems. When the CalDet is configured to mimic the Far detector, the light is measured at each end of the scintillator strip by Hamamatsu R5900-00-M16 photomultiplier tubes2, which are read out using the Viking VA chip, manufactured by IDE AS of Hovik, Norway. In future runs, the CalDet will be reconfigured to mimic the Near detector. The PMTs will be replaced by Hamamatsu R5900-00-M64 photomultiplier tubes2 read out by Fermilab QIE electronics3, and one end of the scintillator readout will be replaced by a reflector.

438

3. Beam Operation In 2001 the CalDet was run in the T11 beam line of the CERN PS complex. The beam consisted of charge selected pions, protons, electrons, and muons with a maximum momentum of 3.6 GeV/c and a momentum spread of 3%. The run program consisted of measurements at momenta ranging from 0.5-3.5 GeV/c, in roughly 0.5 GeV/c steps. The beam line was equipped with two means of particle identification, a Cerenkov detector and a time-of-flight (TOF) system. The threshold Cerenkov detector was used for electron identification. A sample spectrum from the Cerenkov detector can be seen in Figure 1. With a lever arm up to 13 meters, the TOF system was designed to distinguish between pions and protons. A sample spectrum from the TOF is shown in Figure 2.

Figure 1. Spectrum from the Cerenkov De- Figure 2. Spectrum from the time of flight tector taken from a 2 GeV/c run.

system taken from a 1.6 GeV/c run.

The MINOS Far detector electronics system was designed for the low interaction rates expected in the Soudan underground laboratory. Even in this relatively low intensity test beam, the detector performance suffered from intensity related electronics effects. The largest effect was caused by dead time induced by VA chip read-out. During read-out, a chip is dead for 4 . 5 ~ s More. over, since read-out is triggered by an above-threshold signal on the PMT dynode, sections of the detector were dead asynchronously. To overcome this problem, every hit was written out. A 4 out of 5 plane trigger was performed off-line, and chip histories were recorded in the data stream. In this way, events in which pieces of the detector were dead are removed from the sample. While the number of events that survive this cut is about 50%, the remaining sample has sufficient statistics for the calibration.

439

4. Results 4.1. Electrons

Figure 3 shows an example of the distribution of total energy deposited by 2 GeV/c electrons. The energy is expressed in terms of the muon energy unit (MEU), defined as the energy deposited by a normally incident, through-going beam muon read out from one strip end. On average, a 2 GeV/c electron deposits 91 MEU in the CalDet. The parameters from the Gaussian fit to the entire histogram range show that the energy resolution at 2 GeV/c is 16.5%. Further examination of events and studies of Monte Carlo electrons indicate that the events in the low energy tail are due to electrons that interacted upstream of the detector. htrmp Nont = 1233

Mean = 88.19 RMS = 21.45 ChP I ndf = Q7.78I68 Constant = M)Bh 2.338 Mean = 91.h0.4519

Figure 3. Energy distribution of 2 GeV/c positrons. Excess of events in low energy bins caused by interactions upstream of the detector.

Figure 4 shows the energy deposited in the detector versus the particle momentum. The response is linear, as expected, with a slope of 44 MEU per GeV/C. Figure 5 shows the square of the energy resolution versus the inverse plus a constant term of momentum. The fit gives the energy resolution as consistent with zero. This agrees well with the expected resolution.

%,

4.2. Hadrons and Muons

Figure 6 shows the number of hits in 1.6 GeV/c pion induced events, and Figure 7 shows the same quantity for proton events. While there is some muon contamination in both samples, these plots indicate that pions tend to induce more activity in the detector than protons. Figure 8 shows the energy distribution for pions at 1.6 GeV/c, while Figure 9 shows the energy

440 MEUs M. Momentum Ch12 I ndt = 5.154 I 3

Figure 4. Deposited energy of positron events in MEU vs. beam momentum. Errors along the x axis are 3% of the beam momentum, while errors along the y are the errors in the mean of the gaussian fit to the energy distribution. ResolutJonA2 M. 1lMornenIurn

= 4.000

* 0.026

0.w -

0.02 -

Figure 5. Energy resolution squared of positron events vs. inverse beam momentum.

distribution for protons. In general pions deposit more visible energy in the detector than protons, as is expected because of the difference in their kinetic energies. Further analysis of both the hadron and muon samples is currently under way. 5 . Conclusions

In summary, the first phase of the CalDet run has been completed and preliminary results have been obtained. Samples of electron and hadron events were collected at various energies. The electromagnetic shower energy resolution is

441 Hit Muitiplkity: n'. 1.6 GeV

100

Hit Multiplicity: p , 1.6 GeV

200

nil Multlplkily

Figure 6. Number of hits in pion events at Figure 7. Number of hits in proton events 1.6 GeVfc. at 1.6 GeV/c. Total MEU +1.6 GeV

Total MEU + l . B GeV

Figure 8. Deposited energy distribution of Figure 9. Deposited energy distribution of 1.6 GeVfc pions. 1.6 GeV/c protons

in good agreement with the expected resolution of 23%. A3

Acknowledgments This work supported in part by the UK Particle Physics and Astronomy Research Council, the US Department of Energy, and the National Science Foundation.

References 1. MINOS Collaboration, Fermilab Proposal P875, The MINOS Detectors, Technical

Design Report, Fermilab, NuMI-L-337, October, 1998. 2. K. Lang et al., Nucl. Instrum. Meth., A461,571-573 (2001). 3. R. J. Yarema et al., IEEE Trans. on Nuclear Science, 4, 750-752 (1993); R. J. Yarema et al., Presented at the 6th Pisa Meeting on Advanced Detectors La Biodola, Isola d'Elba, Italy, May 22-28, 1994.

CALIBRATING THE SNO DETECTOR RESPONSE

A. S. HAMER P-23, MS H803, Los Alamos National Laboratory, Los Alamos, NM, 87545, USA E-mail: [email protected]

(For the SNO Collaboration)

The Sudbury Neutrino Observatory (SNO) is a heavy water (DzO) Cerenkov detector designed t o detect 8B solar neutrinos. This is done via the charged-current, neutral-current, and neutrino-electron elastic scattering interactions with the target DzO. Success of the experiment necessitates an extensive calibration program for characterizing and testing the SNO detector’s response t o Cerenkov light producing particles as well as backgrounds. This has required the development of a variety of specialized calibration devices which includes sources of isotropic light, y-rays, neutrons, and b-particles. A detailed description of these devices along with the results of their application in the SNO experiment is given.

1. Introduction The Sudbury Neutrino Observatory (SNO) is a heavy water Cerenkov detector designed primarily for the detection of 8Bsolar neutrinos. SNO identifies neutrinos via the detection of Cerenkov light from energetic electrons produced by charged current (CC) and elastic scattering (ES) interactions, and neutron capture y-rays resulting from the neutral current (NC) interactions. It does this with an array of x 9500 photo-multiplier tubes surrounding the DzO target volume (1 kilo-tonne) which is contained within a 12 m diameter acrylic vessel. The entire assembly is immersed within a x 7 kilo-tonne water shield and located under x 2100 m of rock in INCO’s Creighton mine near Sudbury, Canada. A complete description of the SNO detector and its many subsystems is presented in Reference’. To extract event and signal information in the experiment a detailed understanding of the detector response is needed along with a comprehensive calibration program. The paper is therefore organized as follows: First a general overview of the SNO detector response is given. This is followed by a description of the calibration systems and sources. A discussion of the SNO calibration procedure and results is then given. Finally, the procedure for determining detector response related systematic uncertainties in SNO’s extracted physics result^^^^^^ is discussed.

442

443

2. The SNO detector response

The SNO experiment consists of three detector phases, each varying by the method at detecting neutrons for the NC interaction’. This paper focuses on the first phase of the experiment (the pure D2 0 phase) where the neutrons capture on deuterons resulting in the emission of a single 6.25 MeV y-ray. All the reaction channels are therefore detected via the Cerenkov light they produce and the primary event observables are energy, vertex, and direction. Also important for distinguishing against backgrounds is event isotropy or the directional distribution of photons in an event (neutrino events have Cerenkov ring characteristics). The above observables are extracted using the individual event’s PMT hit number, positions on the photo-array, and times. Reconstructing event observables therefore requires accurate calibrations of the P MT and electronic timing characteristics as well as a detailed understanding of the detector’s optical properties. Lastly, it requires an absolute scale calibration associating number of hits for a given event position and direction to an energy. Given the collection of events and their observables, and following data reduction, fiducial volume, and energy threshold cuts, SNO separates the remaining CC, ES, NC and background contributions statistically using a n extended maximum likelihood method. It employs characteristic probability density functions or distributions (PDFs) for each of the signals. This includes the energy distribution, the distribution of events in radius cubed (R3), and the distribution in the direction of events relative to the Sun-SNO vector (cosBsun). Example PDFs are shown in Figure 1 and illustrate the features of the individual signals that allow the separation. For CC and ES, the energy distribution is that of the CC and ES electron recoil energies for the neutrino spectrum, respectively. For NC, however, the energy distribution is well approximated by a Gaussian with peak value from the 6.25 MeV y-rays. The R3 distribution of events is nearly flat in the D20 volume for both CC and ES reactions, and extends further out into the H 2 0 volume for ES. The NC distribution, however, falls off away from the detector center due to the long diffusion length of neutrons in D2O and the competition of captures on deuterons versus captures on hydrogen as the neutrons approach the acrylic vessel and H2O volumes. The ES direction distribution is strongly forward peaked relative to the Sun, whereas the CC electrons have a (1-0.340~cosBSun) dependence before taking the detector response into account5. The NC distribution, however, is flat relative to the Sun’s direction. What is important to note here is that correct event reconstruction and construction of the above PDFs requires that the detector’s response to Cerenkov

444

3 11

8 = :. 2

f

.ulc..

d

2

i

5

PI

4

6

8

10 12 14 16 1 Energy (Me\

0

0.2 0.4 0.6 0.8 1 (WRav)'

-1

-0.5

0

0.5

1

cosmetasun

Figure 1. Probability density distributions for the C C (open circles), NC (closed circles) and ES (open triangles) reactions. Distributions were generated with Monte Carlo and have been scaled for improved illustration of the differences in the signals.

light and neutrons be well understood. 3. The calibration systems and sources

The calibration hardware consists of an internal electronic pulser system described in Reference' and sources of isotropic light, -prays, electrons, and neutrons. The individual sources are described here and include the laserball, 16N, 'Li, pT, and encapsulated sources. First, however, a discussion of the source deployment and manipulation system is given. Except for electronics calibrations, all calibrations require a method t o deploy sources into the SNO fiducial volume. This is achieved using a rope and pulley system (see Figure 2). Attached to the top of each source is a carriage with two side pulleys. Two Vectran ropes run down the vessel neck through the pulleys to fixed locations on the acrylic vessel permitting manipulation within one plane of the detector. A second set of side ropes can be attached allowing access t o another perpendicular plane. Also attached to the carriage is a third central rope for additional load bearing and for vertical only deployment. The system is computer controlled, can be remotely operated, and permits source positions in the two planes at M 65 % coverage. Position accuracy of the system in the lower hemisphere of the detector is better than 2 cm, whereas in the upper hemisphere it can be up to 5 cm due to a large d(position)/d(rope length) and an increased tension on the side ropes of the system. Sources can also be deployed vertically into the light water volume via six separate entry ports. Both D 2 0 and H2O entry ports are designed t o prevent light or radon

445

Figure 2.

The source deployment and manipulation system

leakage into the detector. Gate valves on each port can be closed when the sources are not in use. Many of the sources require umbilicals for various necessary connections to the deck6. Umbilicals have a coaxial design with an inner core consisting of a polyethylene tube within which can be placed fiber optic cables, gas transfer capillaries, or signal cables. Additional signal and low voltage power lines are wrapped in a helix around the central core tube. The helical geometry is necessary to prevent the wires from breaking when manipulated. This is placed within a clean, radioactively pure, flexible silicone sheath acting as a water seal and which is potted with silicone gel. The umbilicals are stored in a pulley array mechanism held under tension and fed out along with the central manipulator rope. The optical source must provide short pulse width light at varying wavelengths and intensities. This is achieved using a 337.1 nm nitrogen pulsed laser (model LN203C by Laser Photonics) with a 600 ps pulse width and a peak power of 150 kW. Pulse rates are tunable and can be set to as high as 45 '

446

H z ~ ?Longer ~. wavelengths (365 nm, 386 nm, 420 nm, 500 nm, and 620 nm) are achieved by pumping several dye lasers. Several orders of magnitude in intensity can be achieved using neutral density filters. The laser system is mounted on the deck above the SNO detector and the light transported through a fiber optic bundle through an umbilical to a quartz flask (the laserball) filled with silicone gel. The gel contains small hollow glass bubbles in suspension needed to diffuse the light. The result is a nearly isotropic source of light with a pulse width of about 2 ns (FWHM). As with the manipulator system, the entire system is computer controlled and can be operated remotely. SNO employs sources which use the short-lived radioisotopes 16N (71/2 = 7.13 s) and 'Li (71/2 = 0.838 s)8,9710. Both are produced using 14-MeV neutrons from a commercially available DT Generator (MF Physics model A320a) located in a lab utility room. The 16N and 'Li are produced via the 160(n,p) [Q = -9.64 MeV, u= 35 mb] and 'lB(n,a) [Q =-6.63 MeV, u = 30.5 mb] reactions, respectively". For 16N, the target consists of CO2 gas from compressed bottle flushed through a 7.9 cm long and 11.44 cm diameter annular target chamber surrounding the DT generator. The CO2 gas stream then transports the activity via thin Teflon capillary through an umbilical to a decay chamber in the SNO detector yielding several hundred events per second. For 8Li, the primary production target consists of 2 99 % enriched "B boron powder suspended in a polyethylene matrix. The target lines the inner wall of an annular target chamber that surrounds the DT generator and is similar in dimension to that used for 16N production. Because 'Li is very reactive, aerosol particles (NaC1) are needed to capture the 8Li recoils upon production. A helium gas stream, again supplied by compressed bottle, then transports the loaded aerosol to the detector providing x 0.5 'Li decay events per second. In the case of 16N, the design goal is to produce a high rate source of free gammas. 16N beta-decays, but often to excited states of l60which subsequently emit y-rays. The largest branch (66.2 %) produces a 6.13 MeV y-ray'. After production, the activity is send to a cylindrical decay chamber made of stainless steel (0.476 cm thick) and divided into two compartments. The lower compartment contains the decay volume which is surrounded by plastic scintillator. The upper compartment contains couplings to the umbilical and a photomultiplier tube facing the scintillator and decay volume. Tagging at greater than 90 % efficiency is achieved by detecting the decay beta with the plastic scintillator. The stainless steel walls effectively block the betas from exiting into the D2O volume providing a source of free gammas. For 'Li, the goal is to introduce a spectrum of beta particles into the SNO a M F Physics Corporation, Colorado Springs, Colorado, USA.

447

detector with as little interference as possible from any associated calibration hardware. 'Li beta-decays (central end-point value of 12.98 MeV) to the broad excited state of 'Be which in turn promptly decays to a-particles'. The activity is sent to a spherical decay volume with thin (0.06 cm) walls and is viewed by a photomultiplier tube placed above in a cylindrical end-cap which also contains umbilical couplings. For tagging, the 'Li source detects coincident alphas via the scintillation light produced in the Helium gas. This light is wavelength shifted in the monitoring PMT's visible range using tetraphenyl butadiene (TPB) lining the walls of the chamber as well as a small admixture ( M 0.1 %) of N2 gas introduced into the Helium stream. Tagging is greater than 90 % efficient. The p T source is a small ion source/accelerator that produces 19.8 MeV yrays via the 3H(p,y)4Hereaction12. The unit is designed to accelerate protons up to 30 keV onto a high-purity scandium tritide target. The accelerator components are compact, being only a total of 50 cm in length. The entire assembly is placed inside a stainless steel capsule for introduction into SNO. Normal operation is between 20 and 30 keV producing M 0.5 high energy photons per second. Various longer lived radio-isotopes are also used. This includes 252Cf,232U, and 226Rasources. These activities are encased in acrylic and mounted at the end of a polyethylene or acrylic rod that connects to the manipulator carriage. The sources are untagged but easily distinguished due to their strength and the ultra-low background environment of SNO. Attempts at developing tagged neutron and background sources are underway. 4. Calibration procedure and results The calibration procedure in SNO consists not only of establishing base constants (timing offsets, scales, efficiencies, etc.) necessary for analysis but also in testing the high level detector response. The base calibrations can be subdivided into five major categories which include electronics, PMT, optics, energy scale and neutron capture efficiency determination. The electronics calibrations consist of determining the transfer functions used to convert the stored digitized information, or ADC values, for each PMT channel into charges (picoCoulombs) and times (nanoseconds). These calibrations are achieved via pulsed charge injection methods. The PMT calibration deals with the individual PMT gains and time offsets, as well as the timing dependence on the PMT charge. These calibrations are done using the laserball source run at varying intensities. The optical calibration is also done with the laserball source run at various locations and at different wavelengths. From the resulting global data

448

analysis6, attenuation coefficients for the D20, acrylic, and H 2 0 are determined. The PMT (+ reflector assembly) angular response is also extracted. The last step in the calibration procedure for Cerenkov light detection is the determination of the absolute energy scale. The calibration consists of accumulhting data from the 16N gamma-ray source situated at the center of the SNO detector and comparing this to the result of Monte Carlo simulation. Agreement is then achieved by tuning the average Monte Carlo PMT collection efficiency. The determination of the neutron capture efficiency is done using a calibrated point source of neutrons placed at various locations in the SNO detector. The untagged 252Cfencapsulated source is used for this purpose. The deduced efficiency for neutron captures on deuterium is 29 f 1.1%for a uniform source of neutrons in the D20. However, given the SNO fiducial volume and energy threshold cut, the efficiency becomes 14.4%.

2

4

6 8 1 Energy (MeV)

0

-200 -100 0 100 200 Reconstructed Z Vertex (cm)

0

0.2 0.4 0.6 0.8 Direction Cosine

Figure 3. Comparisons between 16N d a t a (closed circles) and Monte Carlo (open circles) for energy, vertex reconstruction, and angular reconstruction distributions.

Tests of the high level response of the detector are done mainly with the Cerenkov sources. In most cases uncertainties are quantified by comparing distribution means of observables such as energy, vertex, and angular resolution for data versus Monte Carlo. Figure 3 illustrates such comparisons for a central position 16N run. To extract geometrical and temporal dependent uncertainties, the comparisons are done as a function of run position or time. This is illustrated in Figure 4 which shows the difference between Data and Monte Carlo energy distribution means as a function of detector radius for the 16N source. Similar plots can be made for energy resolution, vertex accuracy

449

and resolution, and angular resolution, and as a function of time as well. The overall energy scale uncertainty is determined by combining the temporal and position dependent uncertainties with 16N source related uncertainties and channel property uncertainties such as gain and threshold and is 1.21 %. To test the differential energy scale uncertainty (i.e. energy non-linearity) a microscopic study of detector effects such as electronic pick-up and the multiphoto-electron effect that could impact the higher energy response of the detector is done using 8Li and 16N source data. The results translate into a 0.23% uncertainty at p T energies. This uncertainty is tested with the p T source energy distribution compared to Monte Carlo prediction. The resolution uncertainties are determined by comparing the widths of the 16N and p T data energy (or Hit) distributions to that of Monte Carlo simulation. The uncertainty is 4.5% at 16N energies and increases up to 10% at p T energies.

3 0.05

k 0.04 V

3 e c

0.03

Source Radius (cm)

Figure 4. Fractional difference in 16N energy distribution means for Data and Monte Carlo as a function of source radial position.

The vertex reconstruction uncertainties includes uncertainties in both the accuracy and resolution. In both cases, the overall uncertainty is quantified as the difference between data and Monte Carlo for both 16N and 'Li source data. The uncertainty in the accuracy is 1 %. The vertex resolution is approximately Gaussian with an average width (in X, Y, and Z coordinates) of 18 cm but is energy dependent. The uncertainty is M 2.0 cm. Finally, the angular resolution of the detector is tested using the 16N source. This is done by comparing the reconstructed angle relative to the source-vertex vector for events occurring at least 1.5 meters away from the source.

450

5. Estimating systematic uncertainties in neutrino fluxes The most recent SNO flux results3 give CC, ES, and NC flux values for a kinetic energy threshold of T,ff2 5 MeV, a fiducial volume cut of 550 cm and are derived assuming the standard neutrino spectrum shape13. This is referred to as the results of the energy constrained signal extraction. Signal results in terms of electron and non-electron (mu and tau) neutrino fluxes are also provided as well as the flux results in which energy information is not used in the signal extraction (i.e. unconstrained results). Finally, day and night fluxes as well as an analysis assuming 2 x 2 MSW oscillations is presented in4. To illustrate the procedure for determining systematic uncertainties, however, the energy constrained results only are used below. As noted in Section 2, the CC, ES, and NC signals are statistically separated using the signal PDFs for energy, radius cubed, and direction. Also included are PDFs for the backgrounds. In the maximum likelihood fit, the signal amplitudes are allowed to float whereas those of the backgrounds are fixed. The signal PDFs are generated both using Monte Carlo simulation and analytic forms. The analytic forms are from integrating over a series of detector response functions in energy and, vertex and angular reconstruction, and are convolved with the analytic description of the neutrino interactions (energy spectra, interaction locations and neutron capture positions for NC, and direction). To establish systematic uncertainties, the detector responses are perturbed and the maximum likelihood signal extraction repeated. For the Monte Carlo method, observable values (energy, vertex, direction) in the Monte Carlo sample are shifted on an event by event basis in order to account for the possible ranges in detector responses as constrained by the calibration data. The difference in signal extraction results provide the flux uncertainties for the energy scale, resolution and non-linearity, the vertex accuracy and resolution, and the angular resolution. For the analytic method, parameters in the energy resolution, vertex reconstruction, and angular reconstruction response functions are varied. Lastly, the signal extraction is repeated varying the scale of the fixed background PDFs according to their uncertainties. The final result is a listing of systematic uncertainties as shown in Table 13.

6. Conclusions

SNO has developed a comprehensive calibration program. The various devices developed permit all the base calibrations to be performed and estimates of detector response related systematic uncertainties t o be made.

45 1 Table 1. Experimental uncertainties on fluxes for the SNO neutral-current paper. Here pd stands for photodisintegration and AV for acrylic vessel. Source Energy scale Energy resolution Energy non-linearity Vertex resolution Vertex accuracy Angular resolution Internal source pd background External source pd background D2O Cerenkov background H2O Cerenkov background AV Cerenkov background P M T Cerenkov background Neutron capture efficiency Data reduction cut acceptance Experimental uncertainty

CC Uncert. (percent) -4.2,+4.3 -0.9,+0.0 fO.1 fO.0 -2.8,+2.9 -0.2,+0.2 fO.0 fO.l

-0.1,+0.2 fO.0 fO.0 fO.1 fO.0 -0.2,+0.4 -5.2,+5.2

NC Uncert.

-0.0,+4.4 f0.4 fO.l f1.8 -0.3,+0.3 -1.5,+1.6 -1.O,+l.O -2.6,+1.2 -0.2,+0.4 -0.2,+0.2 -2.1,+1.6 -4.0,+3.6 0.2,+0.4 -8.5,+9.1

I aUTUncert. -0.0,+6.8 f0.6 f0.2 f1.4 -0.3,+0.3 -2.0,+2.2 f1.4 -3.7,+1.7 -0.2,+0.6 -0.3,+0.3 -3.0,+2.2 -5.8,+5.2 -0.2,+0.4 -13.2,+14.1

Acknowledgments This research was supported by: Canada: NSERC, Industry Canada, NRC, Northern Heritage f i n d Corporation, Inco, AECL, Ontario Power Generation; US: Dept. of Energy; UK: PPARC. We thank the SNO technical staff for their strong contributions. References The SNO Collaboration, Nucl. Inst. Meth. A449,172 (2000). The SNO Collaboration, Phys. Rev. Lett. 87,071301 (2001). The SNO Collaboration, Phys. Rev. Lett. 89,011301 (2002). The SNO Collaboration, Phys. Rev. Lett. 89,011302 (2002). J.F. Beacom and P. Vogel, Phys. Rev. Lett. 83,5222 (1999). B.A Moffat, The Optical Calibration of the Sudbury Neutrino Observatory, Ph.D Thesis, Queen’s University (2001). 7. R.T. Ford, Nitrogen/Dye Laser System for the Optical Calibration of SNO, M.Sc. Thesis, Queen’s University (1993). R.T. Ford, Calibration of SNO for Detection of * B Neutrinos, Ph.D. Thesis, Queen’s University (1998). 8. V.S. Shirley (Ed.), Table of Isotopes, 8th Edition, Wiley-Interscience, New York (1996). 9. M. Dragowsky et al, Nucl. Inst. Meth. A481, 284 (2002). 10. N. Tagg et al, arXiv:nucl-e~/0202024, (to be published). 11. V. McLane, C. L. Dunford, and P. F. Rose, Neutron Cross Sections, Vol. 2, Academic Press (1988). 12. A. W. P. Poon et al, Nucl. Inst. Meth. A452, 115-129 (2000). 13. C.E. Ortiz et al., Phys. Rev. Lett. 85,2909 (2000). 1. 2. 3. 4. 5. 6.

TIME CALIBRATION OF AMANDA: THREE VARIATIONS OF A THEME OF To

K. D. HANSON* Department of Physics €5 Astronomy, University of Pennsylvania 209 S. 33‘d Street, Philadelphia, PA 19130 E-mail: kaeld9amanda.physics.wisc.edu (For the AMANDA Collaboration)

T h e AMANDA-I1 neutrino telescope currently operating at the South Pole is an array of 677 optical modules (OMS) deployed in the ice at depths between 1200 m and 2300m beneath the surface. Calibration of the timing offsets of each OM is effected primarily by means of in-ice light pulses transmitted via optical fibers from a surface YAG laser. Discriminator walk, which is significant due t o the transmission of electrical signals over 2 km distances, is also calibrated using the YAG laser. Another way to calibrate the timing offsets is t o use downgoing cosmic ray muons. This method has the advantages of fuller coverage and year-round availability, i.e., it can be done anytime the detector is taking data. Finally, preliminary results of a technique used t o calibrate, with nanosecond precision, the local clocks in “digital optical modules” (DOMs), which digitize and timestamp PMT signals in situ, are presented using DOMs in operation in AMANDA-11. Th e DOM is part of the baseline design for the planned IceCube detector.

1. Introduction: Time Reconstruction in AMANDA

Phototube signals from the individual OMS in AMANDA propagate up electrical and optical paths and are digitized by TDC and ADC hardware on the surface. Because of the long signal paths, there is a considerable delayexceeding 10ns for the deepest OMS-between the arrival of Cherenkov light at the phototube and that signal’s digitization on the surface. Beyond these gross time offsets, much smaller delays, on the scale of ns, accrue along the signal path, such as PMT transit time and propagation delays in the electronics. The lump sum of these retardation effects is commonly termed the TOand must be subtracted away from the hit times recorded on the surface to obtain the true hit time at the phototube. * Now at University of Wisconsin

-

Madison, 1150 University Ave,Madison, WI 53706

452

453

There are also dynamic delays due to discriminator walk that must be compensated. The electrical cables in particular, but also to a lesser extent the optical fibers, are dispersive media which broaden the sharp phototube pulses which typically have risetimes in the range 5-101s (see Table 1). The finite Table 1. Typical risetimes for the three different signal cable technologies deployed in AMANDA. Cable Type

Coaxial Twisted Pair Optical Fiber

Risetime (ns) 100 50 10

risetime of the pulse causes pulses with smaller amplitude to cross the discriminator threshold at a later time than large-amplitude pulses, thereby adding an amplitude-dependent delay. The correlation with amplitude is extremely high (see later, Figure 2) and thus, even for the coaxial cables, most of the dispersive effects on timing precision can be eliminated as long as the pulse amplitude is recorded. The full time reconstruction equation is given in Eq. 1:

where p, the TDC ‘tick’ slope, is fixed by the TDC clock frequency of 960 MHz, TO,the total delay, is a parameter to be determined by calibration, and a , the timewalk coefficient, is also determined by calibration. The quantities T D C and ADC are the channel’s measured time and amplitude at the surface. Tests with montecarlo data sets imply that relative timing errors as large as 20ns do not significantly degrade the event reconstruction1. 2. Laser Calibration

The laser system employed in the AMANDAtiming calibration utilizes a single high-power Nd:YAG solid state laser (A = 532 nm) located at the surface in the electronics counting house. The laser light is delivered to the photocathode of each OM by means of optical fibers deployed in the cable bundles going down into the ice. At the OM, the light is diffused using an acrylic ball. The light intensity is controlled at the surface with filters and a mechanical attenuator to produce one to several photoelectrons at the PMT. Consult Figure 1 for a schematic representation of the calibration setup. The beam is split just after the laser aperture. One leg illuminates a photodiode which provides a reference trigger signal. The other is sent to the OM. By measuring the difference between the TDCs of the reference signal and the OM signal, the

454

w Figure 1.

Schematic of laser TOcalibration system.

total time delay between a photon falling on the photocathode and its amplified signal starting the TDC can be deduced. The total measured delay t o M - t R E F includes not only the signal time up into the TDC channel, but also the transit time of the laser light down the optical fiber; this latter contribution must be known a przorz and subtracted. It is measured using an OTDRa, a commercial off-the-shelf item used in the telecommunications industry. The variation of signal intensity allows the correlation of leading edge time to pulse amplitude-discriminator walk-to be mapped. Because of the functional form of the discriminator walk coefficient (see Equation l ) , the leading edge time is plotted in a scatterplot vs. 1 / d m and this correlation is fit to a simple linear model:

It now becomes apparent that the TO,which would be the leading edge time extrapolated to an infinitesimal risetime is just the y-intercept and the timewalk coefficient is the slope of the fit line. 3. Calibration with Cosmic Ray Muons

The AMANDA detector triggers on downward-going cosmic ray muons at approximately 100Hz. Annually, x 2 x lo9 muons events are written to tape. They are viewed by most to be an annoying background which must be picked over to obtain the signals of interest. However, the astute calibrator sees the opportunity to utilize these muons as a calibration source. In the laser TOcalibration the reference time of the originating pulse was provided by the photodiode at the laser. In calibrating with muons, the reconstructed muon track gives a time to which the OM hit time can be referred: aOptical Time Domain Reflectometer

455

Figure 2. Scatterplot of correlation of discriminator walk ( ’timewalk’) with pulse amplitude. The inset figure shows the population of timewalk-corrected TOS.The final calibration constant is found from fitting this curve with a gaussian.

the difference is called the time residual and is defined by ~RES

HIT

- t,

(3)

where t, is known from the kinematics of Cherenkov radiation. Scattering in the ice causes the distribution of timing residuals to assume a complicated, non-gaussian shape which is nevertheless easily measured from data: The fun-

Figure 3. (Filled curve): experimentally measured time residual of an AMANDA optical module. t R E s = 0 is the time when Cherenkov photons propagating without scattering (“direct” photons) reach the O M from the track. (Hollow curve): template function which is used in determining the effective offset of time residual distributions.

damental lemma of the muon TO calibration is that shifts in the TOof the OM appear as corresponding offsets in the distribution of the residual time.

456

These offsets are determined using the method of matched filter correlations: a template is generated from the average shape of all OMS’ time residual distributions. This template is (mathematically) slid over test distributions until the maximum point is found. The offset required to find this maximum is the offset of the test distribution. The algorithm proceeds as follows: (1) (2) (3) (4) (5)

Reconstruct muon tracks using current set of To values, Accumulate time residuals, Determine offsets, Apply offsets as correction to the To, Go to step 1,repeat as necessary.

The whole procedure must be iterated several times because changes in the To affect the track reconstruction. As the method finds a solution, the magnitude of the offsets should begin to die away. This convergence is monitored and a plot from a typical run is shown in Figure 4

Figure 4. Convergence plot of muon To calibration run, plotted as width of distribution of offsets to be applied t o TOversus iteration.

The algorithm was tested to see whether it actually converged to the correct solution. First, the TOvalues measured by the laser calibration were intentionally moved away from their true values by an amount indicated by the solid line of Figure 5. Then the muon calibration was applied and the To shifts that were found are plotted as dots in Figure 5 . 4. Calibration of Digital Optical Modules In the proposed IceCube detector, each digital OM (DOM) digitizes waveforms in situ. The start of a capture latches an asynchronous clock running at 40MHz. When the packet is sent to the surface, this local time must be

457

Figure 5. Test of convergence of muon TOagainst laser measured Tos in the AMANDA-I1 detector. The solid lines indicate the pattern of To shifts applied to the detector versus OM. The markers show the corrections found by the muon calibration.

-

converted to an absolute time scale a t the surface with an accuracy of ns. This conversion to such a high degree of accuracy requires implementation of a continuous calibration procedure to track oscillator drifts over time scales as short as 100ms.

Figure 6.

Schematic of R A P calibration for the digital AMANDA string.

The RAPCal method achieves this by periodically exchanging synchronization pulses along the communication lines. There are four stages t o the cycle: (1) At time

7'1,

the surface FPGA sends a RAP pulse down the cable,

458

simultaneously latching the surface clock. ( 2 ) At time T2, the DOM receives the pulse and latches the DOM clock. It then digitizes the RAP pulse waveform. (3) At time T3, some fixed delay after T2, the DOM FPGA send a RAP pulse up the cable, simultaneously latching the DOM clock. (4) At time T4, the surface receives the RAP pulse from the DOM and latches the surface clock. This waveform is also digitized. The ratio of the frequencies of the surface clock to the DOM clock is then given by

The RAPCal also can be used to obtain the cable delay, TO,of the digital signal: TO = 1 / 2 f S U R F ' (T4 - f S U R F / f D O M ( T 3 - T2) - T1)Tests with the String 18 prototype string of digital optical modules indicate that RAPCal cycles performed every few seconds are sufficient to establish timing accuracies below 5 ns, sufficient for event reconstruction with AMANDA or the IceCube detector.

5. Conclusions The laser To calibration persists as the mainstay for timing measurements in AMANDA as its analysis is very straightforward and the method itself is robust. Unfortunately, connections from the YAG to each of the approximately 750 optical fibers must be made manually. This is a tedious and timeconsuming process. The muon TOmethod shows promise t o be an automatic procedure which can be performed remotely and for any time period for which data exists. Currently it is not able to calibrate the discriminator walk but we are developing methods for including this as well. The data stream in IceCube will only contain time calibrated hits since the IceCube triggers mandate this requirement. The RAP calibration has been tested with existing digital hardware in operation now and provides this realtime calibration.

References 1. A. Biron, A M A N D A Internal Report, 20001101,(2000). 2. A more detailed report of the muon TOcalibration method may be found in D. Cowen and K. Hanson, Proceedings of the 27th ICRC, HE237, (2001). 3. All aspects of the RAPCal and DOM are covered to a much greater extent in the IceCube Preliminary Design Document, available on the web at http://icecube.wisc.edu/tech/PDD.pdf.

ABSOLUTE CALIBRATION OF ELECTROMAGNETIC CALORIMETER AT LHC WITH PHYSICS PROCESSES

TAO HU, LEI XIA, REN-YUAN ZHU California Institute of Technology, 1200 East California Blud. Pasadena, CA 91125, USA Physics processes Zo -+ e + e - , W + e * v e , J / $ -+ e+e- and T(1s) -+ e+e- were investigated for calibration of the electromagnetic calorimeter in site at LHC. The production cross sections of these processes, including high order corrections, were calculated by using PYTHIA 6.136 as well as CDF data. As an example, the Level 1 trigger efficiency was calculated by using events simulated with CMSIM 121 and ORCA-4-4-0-optimized. The result of this investigation indicates that although the process of J / $ t e+e- has a factor of 16,000 larger cross section as compared to that of Z o t e + e - , about 1 or 0.5 year data taking at relatively low luminosity of 1032cm-2s-' or 1 0 3 3 ~ m - 2 s - 'respectively is still required to reach a sub percent calibration precision by using electron pairs from J / $ . A combination of all physics processes thus may be needed a t the beginning of the data taking.

1. Introduction

A precise calibration is the key for maintaining the precision of an electromagnetic calorimeter (ECAL) in situ at LHC. A light monitoring system, for example, provides relative variations of the calibration in situ at LHC for the CMS lead tungstate (PbW04) crystal ECAL, while the absolute energy scale calibration is provided by physics processes, such as Zo -+ e+e- and W* -+ e*vel. The Zo + e+e- is a powerful calibration process with constraints of the Zo mass and width. An early study shows that a calibration precision of 0.4% (1%) may be achieved after iterations of fitting the Zo mass and line shape, if 100 (25) electrons per crystal is available2. Assuming a 10% divergence of the initial calibration constants, this algorithm fits the invariant mass distribution of electron pairs with a Breit-Wigner shape to determine the area (Aall),the mass ( M a l l ) and the width (Fall). With a fixed width rail, the algorithm further fits a similar invariant mass distribution for electron pairs, which have one electron hits crystal i, to obtain the area Ai and the mass Mi. The ( M a a / M i ) 2is the correction factor for the calibration of the crystal i after this iteration. The algorithm is converged after several iterations. The precision of

459

460

this algorithm is statistics limited. Figure 1 shows the gains and e+e- mass spectrum with a statistics of 100 electrons per crystal. It is clear that less than 0.4% calibration precision is achieved after 5 iterations. By using this algorithm, 25 electrons per crystal is needed t o achieve calibration precision of less than 1%.

100 electrons per cell 10

50

z 5 c

C

0 0.75

0

1

1.25

90

Cell gain

1

'

I',$/hdf

'

95

e-e mass '10.3\

I ' ' '

/ ' 174

1 ,$/ndf'

' 67111 /I

971

100

0

90

Cell gain

95

e-e mass 200

50

48.93 Sig a

0.3752E-02

Q,

.-

c

0 0.75

0

1

1.25

Cell gain Figure 1. Crystal calibration by using 2

90

95

e-e mass

+ e+e-'.

The Zo -+e+e- process, however, has limited cross section and thus may take too long to accumulate enough statistics. The W* + e*v, process has a relatively larger cross section and was studied e ~ t e n s i v e l y ' >The ~ . use of the

46 1

W* 4 e*v, process, however, relies on the performance of the tracker since the E/p matching is the only constraint for this process. We are thus interested in using additional physics process for the ECAL calibration, especially at the beginning of data taking, when LHC runs at cm-2 s-l. This report cmP2 s-' or relatively low luminosity, such as describes an investigation on absolute calibration using the J / $ -+ e+e- and "(1s) + e+e- processes. The advantage of these processes is the high cross section and that the physics constraints are entirely within reach of the ECAL. The disadvantage is the requirement of a very low p~ cut in Level 1 trigger. In this study, we try to calculate the cross sections of these processes, and simulate the trigger rate by taking CMS detector as an example. The result of this study indicates that about 1 or 0.5 year data taking at relatively low luminosity of 1032Cm-2s-1 or 1033Cm-2s-1 respectively is required to reach a sub percent calibration precision by using electron pairs from the J / $ production. A combination of all physics processes thus may be needed to achieve required precision quickly at the beginning of data taking. 2. Production Cross Sections

PYTHIA 6.136 is used to calculate the production cross sections of Z o ,W*, J/$J and Y(1s) at LHC. Since most processes in PYTHIA are calculated at the tree Level, CDF data were used to estimate the amplitude of corrections, or the Ic factor, which includes high order corrections. 2.1. Z o

+ e+e-

The CDF data for inclusive Zo production6, followed by muon pair decay, is ~ ( p -+p 2') . B r ( Z o -+ p'p-)

This cross section at o(@

= (233

* 18) pb.

fi = 1.8 TeV, calculated from PYTHIA, is + 2') . B r ( Z o -+ p'p-) = 172 pb.

(1)

(2)

Equations 1 and 2 indicate that the correction, or the k factor, to the Z o production cross section is 1.35, which is consistent with the estimation (1.3 1.4) by the QCD tools working group7. PYTHIA gives the cross section for the Zo + e+e- calibration process at LHC (6 = 14 TeV) as ~ ( p -+ p 2') . B r ( Z o + e+e-) = 1.85 nb.

(3)

Assuming a Ic factor of 1.35, we have our best estimate of .(pp

+ 2') . B r ( Z o + e f e - ) = 2.5 nb.

(4)

462

The energy (E), transverse momentum ( p ~ )pseudo , rapidity (77) and opening angle of two electrons from Z decay ( A R = distributions for electrons from Zo decays are shown in Figure 2. The peak of the p~ distribution is at 45 GeV and the opening angle of two electrons peaks at 180". A p~ cut as large as 10 or 15 GeV for the Level 1 trigger does not degrade the statistics.

d

AK

Figure 2. E, p ~17 and , A R distributions for electrons from Z o decays at LHC.

2.2. W

m

)

11

Figure 3. E, p~ and 7) distributions of electrons from Wf decays at LHC.

-+ ev

The CDF data for inclusive W* production*, followed by decays t o an electron and a neutrino, is ~ ( p -+ p W ). BT(W -+ ev) = (2.49 f 0.12) nb.

(5)

The corresponding total cross section calculated by using PYTHIA is

a ( p p -+ W ) . BT(W + e v ) = 1.88 nb.

(6)

Equations 5 and 6 indicate that the k factor for the W production cross section is 1.33, which is similar to the case of the 2' production. PYTHIA gives the cross section for the the calibration process of W* -+ e*v at LHC (fi = 14 TeV) as a ( p p -+ W ). BT(W -+ ev) = 20 nb.

(7)

463

Assuming a k factor of 1.33, we have our best estimation as a ( p p -+ W ) . Br(W -+ ev) = 26.6 nb,

(8)

which is a factor of 11 larger than that of 2 -+ ee. The energy (E), transverse momentum ( p ~and ) pseudo rapidity (77) distributions for electrons from W* decays are shown in Figure 3. These distributions are very similar to those of electrons from the 2' decays.

2.3. J / $

+ e+e-

The CDF data for inclusive J/$ productiong, followed by muon pair decay, is ~ ( p -+ p J/$,pT > 5 GeV, 1771 < O.~ ).BT(J/$-+ p+p-) = (17.4f0.1

ti::) nb.(9)

This includes near 20% from b hadron decays, about 10% from $(2S) decays and other 70% from the direct J/$ production and xc decays. In PYTHIA, the production of J/$ consists of 4 processes, 99 -+ J/$g,

99 -+ xcog, 99 -+ xclg, and 99 -+ xc2g,

(10)

where x C o , xcl and x c 2 can decay to JIG. Taking into account the branching ratio of J/$ -+ p+p-, we have a(pF -+ J/$,pT

> 5 GeV, lql < 0.6). Br(J/$

-+ p+p-) = 3.77 nb,

(11)

which is much lower than CDF data. By using Equations 9 and 11,the k factor for the J/$ + e+e- process is estimated to be 4.62, which is much larger than that for the Zo and W* productions. The PYTHIA cross section of J/$ production at LHC is a ( p p -+ J/$) . Br( J/$ -+ e+e-) = 8.53 pb.

(12)

By using the k factor of 4.62, we have a ( p p -+ J/$) . Br( J/$ + e+e-) = 39.4 pb,

(13)

which is a factor of 16,000 larger than that of 2 -+ ee. Figure 4 shows the energy (E), transverse momentum ( p ~ )pseudo , rapidity (17) and opening angle of two electrons ( A R ) distributions for electrons from J/$ -+ e+e- process at LHC. The peak of the p~ distribution is at 1.5 GeV, and the opening angle of two electrons from the J/$ decay has a broad peak between 120" and 180". To use most of the electrons from J/$ decays the p~ cut should be as low as 1 GeV.

464

AH

Figure 4. E, p ~7 and , AR distributions for electrons from J / $ decays at LHC.

2.4. Y ( l s )

Figure 5 . E, p ~q ,and A R distribution for electrons from T(1s) decays at LHC.

+ e+e-

The CDF data of inclusive "(1s) productionlo, followed by p pair decay, is da -(pp

dY

-i T(ls),y

=0 , p < ~ 16 GeV).Br(T(ls) -i p+p-) = (753f29f72) pb(

Since the Y(ls)'s y distribution is almost flat when 1y1 a ( p p -i T(ls),IyI

< 0.4, we have

< 0 . 4 , ~<~16 GeV) . Br(T(1s)+ p+p-)

= 602 pb. (15)

In PYTHIA, the productions of T(1S) also consists of 4 processes, -i T ( l s ) g ,

-i XbOg,

9.9 + Xblg, and

-i XbZg,

(16)

where XbO, Xbl and Xb2 can decay to "(1s). Although these 4 processes are not explicitly included in PYTHIA, we can use the calculation for J/$J production, by changing the code of state produced as well as the value of wave functions at the origin". Taking into account the branching ratio, we have

a ( p p -i T(ls), p~

< 16GeV, IyI < 0.4) .Br('Y(ls) -i p'p-)

= 250 pb, (17)

which is also much lower than CDF data. By using Equations 15 and 17, the k factor for the T(1s) -i e+e- process is estimated to be 2.41, which is larger than that of Zo -i e+e- and W* -+ e*v processes, but smaller than that of the J / $ e+ei process. The PYTHIA cross section of T(1s) production at LHC is

a ( p p -i T ( l s ) ) .Br(T(1s) + e'e-)

= 67.3 nb.

465

By using the k factor of 2.41, we have a ( p p + T ( l s ) . Br(Y(1s) + e+e-) = 162 nb,

(19)

which is a factor of 65 larger than that of Z -+ ee. Figure 5 shows the energy (E), transverse momentum ( p ~ )pseudo , rapidity (17) and opening angle of two electrons from the Y(1s) decay (AR) distributions for electrons come from "(1s) + e+e- process at LHC. The p~ distribution peaks at 4.5 GeV, so to use most of the electrons from Y(1s) decays the p~ cut should be around 3 GeV. In addition to the T(ls), the "(2s) production may be significant t o the Calibration too. 3. Event Rate and Time Needed for Precise Calibration

The key issue of using either J/$J or Y channels is the level 1 trigger efficiency. Since electrons from both processes are rather soft, a relatively low trigger threshold at Level 1 is necessary to get decent efficiency while keeping the total Level 1 QCD background trigger rate under control. In this section, we take CMS detector as an example to investigate possibilities of low Level 1 Trigger threshold and probe an option of requiring invariant mass of two electrons at Level 1. The CMS calorimeters cover a region of 171 < 5. The ECAL and HCAL cover 1171 < 3, and the forward HCAL completes the coverage by spanning the region of 3 < (71< 5. Each ECAL crystal covers 617 x 64 = 0.0175 x 0.0175. A quadratic cut view and orresponding trigger towers of the calorimeter is shown in Figure 6.

q=1.5660 q=1.6530

q=1.7400 q=1.8300 q=1.9300

q=2.0430 n=2.1720

*

Scale

0;5

'

(meters)

Figure 6. Calorimeter longitudinal section (one quadrant), showing the trigger towers in the r-z plane 12.

466

3.1. Level 1 Trigger Eficiency

The Level 1 trigger rate was calculated by using signal events generated by uisng CMSIM 121 for CMS detector response. ORCA-4-4-0-optimized was used to generate Level 1 ntuples for each physics process. The background QCD events have a pileup at luminosity of 1033~m-2s-1. This may be overestimated for the case of 1032~m-2s-1. The total level 1 trigger rate was calculated by weighting the event with the cross section multiplied by the luminosity, divided by the number of events in the p~ bin. We assume that the

Table 1. Event Rate for ECAL Absolutee Calibration by Using Four Physics Processes L

Level 1 Cuts ( G e V )

(1/cm2a)

E c u t l = 20 Ecut2 = 10 E c u t l = 10 Ecut2 = 6

1033

QCD Rate ( k H z ) Z

Level 1 Eff.(%)

1033

E c u t l = 20 Ecut2 = 10 E c u t l = 20 Ecut2 = 7 2.7 < Me, < 4.7 E c u t l = 10 Ecut2 = 6 E c u t l = 10 Ecut2 = 4 2.7 < Me, < 4.7

1033

7.4 f .8

69. f 1.

.47 f .02

.050 f .002

7.5 f .4

78. f .l.

.048 f .002

.0051 f .0002

1033

1032

-

--t

efv

7.4 f .8

48. f 1.

7.1 f .3

3.1 f .I

7.5 f .4

59. f 1.

.88 f .03

.39 f .01

7.4 f .8

,062 f .009

1.9 f .4

.21 f .05

7.8 -f .8

.052

* .008

2.0 dc 0.5

.22

* .05

7.5 f .4

.32 f .02

.87 f .09

.09 f .01

7.3

.29 f 0.02

1.2 f .1

.13 f .01

7.4 f .8

.04 f .01

.02 f .01

.003 f 0.001

*

.8

.12 f .02

.12 f .02

.013 f .003

7.5 f .4

.78 f .05

.034 f .004

.0036 f .0004

7.6 f .4

1.31 f .06

.ll f .01

.012 f .001

.4

\

E c u t l = 20 Ecut2 = 10 E c u t l = 20 Ecut2 = 5 8 < M e , < 11 E c u t l = 10 Ecut2 = 6 E c u t l = 10 Ecut2 = 4 8 < M e e < 11

Final Rate (H.)

+ e+e-

W*

E c u t l = 20 Ecut2 = 10 E c u t l = 10 Ecut2 = 6

Barrel Raw Rate (HZ)

7.6

I

467

maximum acceptable total trigger rate is about 7.5 kHz12. Only L L E l e ~ t r ~and n ’ 7“Dielectron” channels were studied. The “Electron” channel means at least one electron and photon candidate with ET greater than the cutoff “Ecutl”. The ‘LDielectron”channel means at least two electron and photon candidates with ET greater than the cutoff “Ecut2”. By lowering the Level 1 trigger threshold (Ecutl and Ecut2) the total Level 1 trigger rate is adjusted to about 7.5 kHz. The corresponding trigger efficiency for each physics process was calculated with numerical result shown in Table 1. Because of the softness of electrons in J / $ and T(1s) decays the trigger efficiency of both processes is low. We probed an option of requiring the invariant mass of electron pairs at Level 1 in a specific region, such as 2.7 to 4.7 GeV for J / $ and 8 t o 11 GeV for T.The result indicates that the trigger efficiency for J/$J doesn’t improve much, while that for T is increased by a factor of 3. The result of using this optional cut is also listed in Table 1. 3.2. Event Rate and Time to Reach Sub Percent Precision

The barrel raw event rate was obtained by multiplying corresponding Level 1 trigger efficiency by the luminosity, production cross section and the selection efficiency requiring all electrons in barrel, i.e. 1771 < 1.48. Table 1 lists barrel raw event rate from each physics process. For the final selection of physics events to be used for ECAL calibration, the following efficiencies were assumed. (1) The single electron identification efficiency, including track finding, isolation and a rough E / p matching, is assumed to be 44% in barrel1. (2) Including reconstructed invariant mass cut the efficiencies of electron pairs are 10.7% for both electrons in barrel’. We assume that the event selection efficiency of J / $ -+ ee and T + ee is identical to that for 2’ + ee used in ECAL TDR. A detailed study on electron selection and invariant mass reconstruction for these physics processes will be carried out in future work. Having the final rate, Table 2 shows the estimated time to accumulate 25 electrons per crystal from 2 , J / $ and Y decay, with 50% machine efficiency.

Table 2. Time Needed Topology ZO t

Days

J I Q -+ ee r-tee

cmP2 s-l 710

ee

I

Days Q1032 cmP2 s-l 6900

170

380

2700

2950

468

4. Summary

Four physics processes for ECAL calibration are investigated in this note: 2’ -+ e’e-, W* + e*v, J / $ + e+e- and “(1s) 4 e + e - . By using CDF data correction factors for each process were determined, and were used t o calculate production cross sections at LHC. Although the production rate of J / $ -+ e+e- and T + ee are about a factor 16,000 and 65 respectively larger than that of 2’ ,+e + e - , low Level 1 trigger efficiency is found for these processes because of the softness of the decay electrons. By uisng J / $ + ee channel about 1 or 0.5 year of data taking is needed at relatively low lumicm-2 s-l or cmb2 s-l respectively to obtain sub percent nosity of calibration precision. A combination of all physics processes thus is needed for CMS ECAL to achieve required precision at the beginning of data taking.

Acknowledgments Drs. T. Strojstrand and N. Glover helped for “(1s) + e+e- cross section calculation. Drs. P. Chumney, S. Dasu and W. Smith provided useful information on Level 1 calorimeter trigger and are kind to let us use their QCD Level 1 trigger ntuple files for total trigger rate calculation.

References 1. CMS collaboration, “The Electromagnetic Calorimeter Project Technical Design Report”, CERN/LHCC 97-33 (1997). 2. Ren-yuan Zhu, “On quality requirements to the barium fluoride crystals”, Nucl. Instr. and Meth. A340 442 (1994). 3. A. Nikitenko, “Study of E/p matching for E C A L calibration with W decays in CMS”, CMS NOTE 1998/021. 4. F. Martin and J-P. Pansart, “ECAL Barrel Inter calibration at Low Luminosity”, CMS NOTE 2000/033. 5. T. Strojstrand, “PYTHIA 5.7 and Jetset 7.4”, Comp. Phys. Comm. 82, 74 (1994). 6 . F. Abe et al., “Measurement of ZO and Drell-Yan Production Cross Sections Using Dimuons i n pj5 collisions at fi = 1.8 TeV”, Phys. Rev. D59, 052002 (1999). 7. C. Balazs et al., “Report of the QCD Tools Working Group”, hep-ph/0011122 (2000). 8. F. Abe et al., “Measurement of aB(W + e v ) and aB(2’ -) e’e-) in p p collisions at fi = 1.8 TeV”, Phys. Rev. Lett. 76 3070 (1996). 9. F. Abe et al., “J/$ and $(2S) Production in p p collisions at fi = 1.8 TeV”, Phys. Rev. Lett. 79 572 (1997). 10. F. Abe et al., ‘TProduction i n p p collisions at fi = 1.8 TeV”, Phys. Rev. Lett. 75 4358 (1995). 11. N. Glover, Private communication. 12. CMS Collaboration, “The TRIDAS Project Technical Design Report, Volume 1: The Rigger Systems”, CERN/LHCC 2000-38 (2000).

MONITORING LIGHT SOURCE FOR CMS LEAD TUNGSTATE CRYSTAL CALORIMETER AT LHC

L. ZHANG, K. ZHU, Q. WEI, R.-Y. ZHU Calafornia Institute of Technology, Pasadena, CA 91 125, USA E-mail: [email protected]

D. T. LIU Jet Propulsion Laboratory, Pasadena, C A 91109, USA

Light monitoring will track variations of the calibration of the CMS lead tungstate crystals in sztv a t LHC, which is crucial for maintaining crystal calorimeter's s u b percent constant term in the energy resolution. This paper presents the design of the CMS ECAL monitoring light source and high level distribution system. The correlations between variations of the light output and the transmittance for the CMS choice of yttrium doped PbW04 crystals were investigated, and were used t o study monitoring linearity and sensitivity as a function of wavelength. The monitoring wavelength was determined so that a good linearity as well as adequate sensitivity can be achieved. The performance of a custom manufactured tunable laser system is presented. Issues related to monitoring precision are discussed.

1. Introduction Because of its high density and fast decay time, lead tungstate (PbW04) crystal was chosen by the Compact Muon Solenoid (CMS) experiment t o construct a precision electromagnetic calorimeter (ECAL) at the Large Hadronic Collider (LHC). In the last six years an extensive R&D program has been carried out to develop radiation hard PbW04 crystals. As a result of this development program, a total of 11.2 m3 large size (25 X,) yttrium doped PbW04 crystals with fast light output will be produced. The crystals, however, still suffer from non-negligible radiation damage' 72,3,435,637.Our previous studies concluded that the scintillation mechanism in PbW04 is not affected by radiation, and the loss of the light output is due only to the absorption caused by radiation induced color centers?. The variations of transmittance can be used to deduce the variations of the light output. A light monitoring system, which measures variations of the transmittance, thus may track variations of the calibration in situ a t LHC, making a precision calorimeter possible even by using PbW04 crystals which suffer from radiation damage'.

469

470

In this paper, we report the design of the CMS monitoring light source and high level distribution system. A monitoring test bench was carried out to study correlations between variations of the light output and the transmittance for the CMS choice of yttrium doped PbW04 crystals. The monitoring linearity and sensitivity were investigated as a function of wavelength, and were used to determine the monitoring wavelength so that the best linearity and an adequate sensitivity are achieved. The performance of a custom manufactured tunable laser system in an 135 hours long term stability test is presented. Issues related to monitoring precision are discussed.

2. Choice of Monitoring Wavelength Figure 1 is a schematic showing a monitoring test bench used at Caltech to investigate monitoring sensitivity and linearity. Monitoring light from a xenon lamp went through a monochromator and injected to an integrating sphere.

Quartz Fiber Chopper

Intergrating Sphere

Step Motor Interface

I

i

MERLIN Radiometry System

/-

I $ I

I QVT Multictiannel L

Analyzer System -I

Figure 1. The monitoring test bench for the transmittance (position 1) and light output (position 2) measurement.

47 1

The light from one output of the sphere was coupled to the front end of a sample through a quartz fiber and an air gap. The light from the other output of the sphere was measured by a photo detector as a reference. The sample was wrapped with two layers of Tyvek paper and was optically coupled to a PMT. The output of the PMT is coupled to either a Merlin from ORIEL through a lock-in amplifier for the transmittance measurement (position 1),or a LeCroy QVT multi channel analyzer for the light output measurement (position 2). When the switch is on the position 1, the transmittance as a function of wavelength was measured by using the PMT output normalized to that of the reference detector. The reference detector was thermo-electrically cooled to reduce systematic uncertainties caused by fluctuations of the intensity of the light source. The PMT output was choppered and went through a lock-in amplifier to suppress the noise. Figure 2 shows a distribution of the PMT output normalized to the reference detector taken for a sample in 23 hours. A Gaussian fit shows that the stability of the transmittance measurement, defined as cr/p (width/average) of the fit to be better than 0.1%.

Q102

> 2

4

6

8

Time (hrs) 15

(b)

$2

10

5

Transmittance/Reference Figure 2. Distribution of the P M T output normalized to the reference detector on the sphere (position 1 in Fig. 5), taken in 23 hours.

L.Y. (p.e./MeV)

Tcorrected L.Y. (p.e./MeV)

Figure 3. (a) Light output measured by the P M T through a LeCroy QVT (position 2 in Fig. 5), taken in 9 hours, and its distributions (b) without and (c) with temperature corrections.

When the switch is on the position 2, the shutter at the input of the monochromator is closed so that there is no interference of the monitoring light source. The scintillation light output of the sample was measured by using a small 137Cssource and a LeCroy 3001 QVT in the Q mode. The Cs

472

spectrum was fit to a simple Gaussian to determine the peak position which then be converted to the light output in p.e./MeV by using a calibration of the single photoelectron peak. Figure 3(a) shows 20 measurements of the light output taken in 9 hours for a sample coupled to a PMT. Data were corrected by using the room temperature which has a variation of up to 0.5"C, despite central air condition of entire laboratory building and individual temperature adjustment and feedback in each room. Figs. 3(b) and (c) show raw and temperature corrected light outputs, respectively, and the corresponding Gaussian fit. A precision of 0.8 and 0.6%were achieved for the light output measurement without and with temperature corrections respectively. To simulate expected radiation environment in situ at LHC, samples were either irradiated by a 6oCo source under 15 to 1,000 rad/h or under recovery after the irradiation. Figure 4 shows correlations between the relative variations of transmittance (AT/T) and the relative variation of the light output (ALY/LY) for the monitoring light at four different wavelengths: (a) 410, (b) 440, (c) 490 and (d) 520 nm, for a yttrium doped sample SIC-S762. The correlation was fit to a linear function: ALY AT - = slope x T LY The result of the fit, including the x2/DoF and the slope, is also shown in the figure for each monitoring wavelength. The error on the slope is about 2 to Normalized Light Output (%) 4

3 2

.s 0

5

5

.G

-1

4

3 2 1

0

Wavelength (nm) Figure 4. Correlations between relative variations of transmittance and light output are shown at monitoring wavelength of (a) 410, (b) 440, (c) 490 and (d) 520 nm for a yttrium doped sample SIC-S762.

Figure 5. Monitoring sensitivity (solid dots), linearity (open dots) and emission spectrum (solid lines) are shown as functions of wavelength for samples (a) SIC-S347, (b) BTCP-2162, ( c ) SIC-S762 and (d) BTCP5658.

473

3%, dominated by the statistical error of the fit. While the slope represents the sensitivity for PbW04 monitoring, the x2/DoF represents the linearity of the fit. It is clear that the linearity is generally good when light output loss is less than 10%. Systematic deviations exist for monitoring light of 410 and 520 nm, as compared to 440 nm, since not all wavelengths are equivalent for the monitoring. AT ALY IT, Figure 5 shows the monitoring sensitivity, or the slope, defined as F and the linearity, defined as X2/DoF, as a function of the monitoring wavelength for four samples. Also shown in the figure is the PMT quantum efficiency weighted radio luminescence. All these samples have a better monitoring sensitivity at shorter wavelength and the best linearity around the peak of the PMT quantum efficiency weighted radio luminescence. The better monitoring sensitivity at shorter wavelength is understood because of the poorer initial transmittance as compared to that at the longer wavelength. The best linearity around the peak of radio luminescence is caused by two radiation induced color centers peaked at two sides of the radio luminescence with different damage and recovery speed, as discussed in details in our previous paper8. Table 1. Result of Monitoring Test Bench Sample Dopant ID SIC-S301 Y/Sb SIC23347 Y/Sb SIC-S392 Y SIC-S412 Y SIC-S762 Y Y BGRI-824 BGRI-826 Y BTCP-2133 Y/Nb BTCP-2162 Y/Nb BTCP-5615 Y/Nb BTCP-5618 Y/Nb BTCP-5658 Y/Nb

Sensitivity 440 nm 0.43 f 0.01 0.60 f 0.01 0.51 f 0.01 0.59 f 0.03 0.51 f 0.01 0.49 f0.01 0.42 f 0.01 0.34 f 0.01 0.28 f0.01 0.33 f 0.01 0.29 f 0.01 0.32 f 0.01

(q/w) 490 nrn 0.35 f 0.01 0.55 f 0.01 0.44 f0.01 0.49 f0.03 0.48 f0.01 0.49 f 0.01 0.42 f 0.01 0.31 f 0.01 0.25 f 0.01 0.30 f0.01 0.27 f 0.01 0.32 f 0.01

Linearity (x2/D0F) 440 nrn 490 nm 1.17 1.59 0.63 0.99 0.94 1.11 0.58 1.26 0.92 1.14 0.42 0.76 0.71 0.96 0.45 0.87 0.62 1.29 0.78 1.42 0.48 0.92 1.04 2.68

Table 1 lists result of our monitoring test bench for seven R&D PbW04 samples produced in China: Shanghai Institute of Ceramics (SIC) or Beijing Glass Research Institute (BGRI), and five preproduction samples produced in Russia: Bogoroditsk Techno-Chemical Plant (BTCP). Note, all are CMS full size samples of 25 radiation lengths, or 23 cm long, with a dimension of 2.3 x 2.3 cm2 tapered to 2.6 x 2.6 cm2. The sample ID, the dopant in the melt and the sensitivity and linearity at 440 and 490 nm are listed. All samples are mainly yttrium doped and have similar emission spectrum peaked at 420 nm.

474

We, therefore, choose 440 nm as the monitoring wavelength'. As discussed in our previous report', the 495 nm may be used as the monitoring wavelength for undoped PbW04 crystals. Since CMS uses only yttrium doped PbW04 crystal which has an emission peaked at 420 nm, this wavelength is now used as a cross-check wavelength.

3. Design of Monitoring Light Source and High Level Distribution System A precise calibration in situ is a key in maintaining the precision offered by a crystal calorimeter. For the CMS PbW04 calorimeter, the variations of calibration in situ is tracked by a light monitoring system, which injects light pulses into each individual crystal and measures variations of its optical transmission and uses that to predict variations of its light output. The monitoring light pulses produced by a laser system are distributed via an optical fiber system organized into three levelsg: an fiberoptic switch which sends laser pulses to one of 80 calorimeter elements (72 half super modules in the barrel and 8 groups of super crystals in two endcaps), and a two level distribution system mounted on each calorimeter element which delivers monitoring pulses t o each individual crystal. Combined with physics events, such as electron pairs from the Zo decays and single electrons from the W decays, the monitoring system is expected to provide calibrations with a precision of 0.4%. Figure 6 is a schematic showing the design of the monitoring light source and high level distribution (optical switch). The laser system is required t o be able to produce light pulses at two wavelengthsg. As discussed in our previous report, the choice of monitoring wavelength directly affects the monitoring sensitivity and linearity'. To track down variations of crystal light output caused by radiation damage and recovery in situ, the monitoring system must run 100% time during the data taking. To provide a continuous monitoring in s i k , a fraction of the 3.17 ps beam gap in every 88.924 ps LHC beam cyclelo will be used to inject monitoring light pulse into crystals. Our initial requirements to the laser light source are as follows''.

0

0

Two wavelengths: one close to the emission peak which provides the best monitoring linearity for the CMS choice of yttrium doped crystals, and the other provides a cross check. Spectral contamination: < Pulse width: full width at half maximum (FWHM) < 40 ns t o match the ECAL readout. Pulse jitters: < 3 ns for trigger synchronization to the LHC beam. Pulse rate: -80 Hz, which is the maximum rate allowed by the ECAL

475

(440or 495 urn)

Figure 6 . The conceptual design of the monitoring light source and high level distribution system for the CMS PbW04 calorimeter.

DAQ system. 0

Pulse energy: 1 mJ/pulse at monitoring wavelength which provides a dynamic range up to 1.3 TeV in single crystal. Pulse to pulse instability:
Our design of the light source consists of two sets of lasers with each consisting of an Nd:YLF pump laser and a tunable Ti:Sapphire laser. Two sets of lasers are necessary t o guarantee 100% availability of the monitoring system. The controls of the entire system, such as internal/external trigger mode, repetition rate and delay times, Nd:YLF laser’s shutter status, lamp status, pump current and cooler status, and Ti:Sapphire laser’s shutter status, wavelength choice, energy level and cooling temperature, are set by a central computer through GPIB and RS-232 interfaces. Each set of lasers has a main output and a diagnostic output. The diagnostic output is further split to two fibers by using a fiber splitter. One output goes to a monochromator for wavelength spectrum monitoring. The

476

other goes to a digital scope for pulse shape and timing monitoring. The pulse shape recorded by digital scope is analyzed by a computer. The histograms and history of laser pulse energy, FWHM, jitters and wavelength spectra obtained by the monochromator and the digital scope are stored in the computer. The selection of two main outputs is controlled by a 2 x 1fiberoptic switch. The output of this 2 x 1 switch is distributed via a 1 x 80 fiberoptic switch to 80 calorimeter elements through 150 m long quartz fibers. Before sending laser pulses to the level 2 splitters, a monitoring box is used to measure pulse energies in each channel. The history of pulse energy passing each switch channel is also stored in the computer. 4. Laser Performance

The first laser was manufactured at Quantronix Inc.13 and was delivered to Caltech at the end of 1999. After initial installation and tuning, an 135 hours long term stability test was carried out when the laser was run at 440 nm at close to the maximum power. Figure 7 shows distributions of (a) the pulse energy, (b) the FWHM, (c) the trigger time, which is defined as the half height of the front edge of the laser pulse, and (d) the pulse center, which is defined at the intensity weighted pulse timing. The instability of the pulse energy and FWHM are found to be 3.7 and 2.0% respectively. The jitters of both the trigger time and pulse center are found to be 3.7 ns. It is known that the long term stability is usually worse than the short

Energy (mJ)

FWHM (ns)

75w re o

20

40

111

m

rm

rw Time (hours)

Trigger Time (ns)

Pulse Center (ns)

Figure 7. Distributions of (a) pulse energy, (b) FWHM, (,-) Trigger tirne and (d) pulse center from monitoring Ti:Sapphire laser at 440 nm, obtained in a test run of 135 hours.

Figure 8. The history Of Ti:Sapphire laser pulse energy at 440 nm and (b) its dard deviation, obtained in a test run of 135 hours’

477

term performance. We also plot the histories of the pulse energy (Figure 8), the FWHM and the trigger time and find that the short term stability of the pulse energy and the FWHM are 2.0 and 1.6% respectively, and the jitters of the trigger time are 1.6 ns. We are pleased with the stability of the Quantronix laser system, which is much better than our 10% requirement. Figure 9 is a photo of the monitoring light source system at Caltech. The entire first laser system was installed and commissioned at H4 test beam site at CERN in August 2001, and has been used in beam test since then.

Figure 9. Photo of the 1st monitoring laser system at Caltech.

5 . Summary

We have designed a monitoring light source and high level distribution system to be used to provide crucial inter calibration in situ at LHC for CMS PbW04 crystal calorimeter. The system is capable to provide monitoring light pulses at monitoring wavelength of choice up to 100 Hz with dynamic range reach up to 1.3 TeV, and run continuously with 100% efficiency during data taking. A key issue of wavelength choice was resolved by a monitoring test bench. Seven yttrium doped R&D PbW04 samples grown by the Bridgman method in China and five yttrium/niobium double doped preproduction samples grown by the Czochralski method in Russia were measured. Consistent sensitivity and

478

linearity were found for crystals grown by the same technology. The better sensitivity at UV is iinderstood to be caused by lower transmittance and the best linearity around emission peak is understood to be caused by two color centers peaked at two sides of emission and with different damage and recovery speed. To achieve the best linearity and adequate sensitivity, 440 nm is chosen as the monitoring wavelength. An 135 hours long term stability test was carried out for the monitoring laser system. The result of the test shows that the laser system provides pulse energy of 1.0/0.6 mJ and FWHM of 25/35 ns at 440/500 nm. The long/short term stability for pulse energy and width are 3.7/2.0% and 2.0/1.6% respectively. The long and short term jitters are found to be 3.7 and 1.6 ns respectively. The system also provides clean laser spectral output with contamination of less than lop3.

Acknowledgments Drs. P. Lecomte and P. Lecoq provided samples from BTCP. Prof. Z.W. Yin provided samples from SIC. Useful discussions with Drs. J.-L. Faure, J.P. Pansart and J. b d e r are acknowledged. Drs. Q. Fu and S. Nikitin of Quantronix provided much help in the laser system, and always patient to our requests and questions.

References 1. H.F. Chen, K. Deiters, H. Hofer, P. Lecomte and F. Nessi-Tedaldi, Nucl. Instr. Meth. A414, 149 (1998). 2. P. Lecoq, I. Dafinei, E. Aufiay, M. Schneegans, M. Korzhik, V. Pavlenko et al., Nucl. Instr. Meth. A 3 6 5 , 291 (1995). 3. M.Kobayashi, M. Ishii and Y. Usuki, Nucl. Instr. Meth. A 4 0 6 , 442 (1998). 4. A. Annenkov, V. Ligun, E. Aufiay, P. Lecoq, A. Borisevich and M. Korzhik, Nucl. Instr. Meth. A 4 2 6 , 486 (1999). 5. M. Nikl, P. Bohacek, E. Mihokova, S. Baccaro, A. Vedda, M. Diemoz et al., Nucl. Phys. Proc. Suppl. 78, 471 (1999). 6. R.Y. Zhu, Q. Deng, H. Newman, C. Woody, J. Kierstead and S. Stoll, IEEE Dans. Nucl. Sci. 45, 686 (1998). 7. Ren-yuan Zhu, Nucl. Instr. Meth. A413, 297 (1998). 8. Xiangdong Qu, Liyuan Zhang and Ren-yuan Zhu, IEEE Dans. Nucl. Sci. 47, 1741 (2000). 9. CMS Collaboration, CERN/LHCC 97-33, 1997. 10. The LHC Study Group, CERN/AC/95-05 46, 1995. 11. Liyuan Zhang, Ren-yuan Zhu and Duncan Liu, CMS Internal Note 1999/014, 1999. 12. Ren-yuan Zhu, Nucl. Instr. Meth. A340, 442 (1994). 13. Quantronix, 41 Research Way, East Setauket, NY 11733,USA.

LED BASED LIGHT MONITORING SYSTEM FOR THE PRIMEX EXPERIMENT AT JEFFERSON LAB

SAMUEL DANAGOULIAN Department of Physics, North Carolina A bT State University 1601 E . Market St. Greensboro, NC 27455 E-mail: danagous4ncat.edu (For the PrimEx Collaboration)

A multichannel Light Monitoring System (LMS) based on super-bright blue LED, has been developed for the high resolution electromagnetic hybrid calorimeter (HYCAL) at Jefferson Lab. T h e central part of the calorimeter consists of 1200 P b W 0 4 crystal modules. Th e prototype LMS has been constructed and tested, and has demonstrated a capability of providing an adequate light pulse per each channel, equivalent t o the signal of 5 GeV electrons in the detector. Th e final Light Monitoring System with 1800 output channels and fiber optics light distribution system is presently under construction. T h e light source of the LMS represents an assembly of 31 super bright blue LEDs connected in series. T h e uniformity of the output light is achieved by mixing light from the LED assembly in an integrating sphere. Th e light intensity is being monitored using PIN photo diode and reference photomultipliers with Am241 radioactive source. In order t o achieve (0.1 - 0.3)% long term stability, the system including the light source, the photo diode with the preamplifier and the reference P M T , is thermally stabilized.

1. Introduction

The PrimEx (Primakoff Experiment)' at Jefferson Laboratory is an A- rated experiment which is designed to perform a high precision (one percent level) measurement of the partial decay width of the neutral pion in the reaction no + yy. This measurement will provide a stringent test of the predictions of quantum chromodynamics in the confinement scale regime. Photons from the Hall B photon tagging system will be used t o produce forward neutral pions in the Coulomb field of the target nucleus (Primakoff production). The two photons from the pion decay will be detected in the Hybrid Calorimeter2 (HYCAL). The Hybrid Calorimeter has been designed to provide high precision angle and energy information for the detected photons. Angular resolution is required to cleanly distinguish the Primakoff mechanism from no production from other processes, and the energy resolution will enable good invariant mass

479

480

reconstruction for rejection of background neutral particles. The performance of HYCAL is based on high energy resolution of the central core consisting of 1200 lead tungstate (PbW04) scintillating crystals. The schematic view of the detector is presented in the figure 1

-

Figure 1. Schematic view of the HYCAL electromagnetic calorimeter.

In order to maintain the high performance of the calorimeter, periodic energy calibration with the tagged photon beam will be necessary. This procedure is time consuming process requiring multiple measurements with the calorimeter in different positions, and therefore cannot be done often. Conse-

481

quently, an optical monitoring system will be used between successive beam calibrations as a relative calibration. During data taking, Compton scattering events, y "e" -+ y "e", where 1, . an electron from the atomic orbit, might be identified among the one e is photon events in coincidence with one electron in the telescope of the PrimEx pair spectrometer, and might be used for calibration.

+

+

77

2. PrimEx Light Monitoring System The main components of the Light Monitoring System (LMS) are: (1) a light source, (2) a mixing box, (3) a light distribution system, (4)reference detectors and (5) a dedicated computer to process the data for continuous stability control. The optical components and the reference detectors are mounted in a thermally insulated box whose temperature is controlled at a level of 0.I"C.

2.1. The Light Source The light source comprises an assembly of 31 NICHIA super bright blue LEDs (peak wavelength 470 nm) connected in series. The LED assembly is powered by 150 V, 8 nsec pulses generated in a circuit, based on transistor in avalanche mode. The light pulse width is on the order of 40 nsec on the output, which simulates the scintillation response of the lead tungstate crystal. The variation of the light pulse amplitude during a week period remains within 3.5% (see figure 2). Pulse variation over the period of one measurement (2 to 5 min) is in the order of 0.2% in average. Pulse to pulse variations are negligible. The prototype light source has been continuously powered during six consecutive months and no noticeable degradation of the performance has been observed. 2.2. Light Mixing and Distribution

Light mixing is a necessary attribute for a high precision (sub percent level) system, as it eliminates angular and spatial variations of the input light and provides a uniform light output for further distribution. For this purpose, a four inch diameter integrating sphere has been utilized. For the final design, a second, six inch sphere will be added in order to provide 2000 distribution channels. Light is distributed to the individual calorimeter modules via a bundle of plastic fibers, 3m long and 265 p diameter each fiber. In order to provide an adequate light pulse for each channel equivalent to the signal of 5 GeV electrons in the calorimeter, we had to attenuate the light after the mixing chamber by a factor of about ten. The radiation degradation of light transmission N

482

in the plastic fiber will be tested during the summer-2002 upcoming beam test of the HYCAL prototype. In general, the background level in Hall B is low and the radiation damage should not be an issue. Each fiber from the bundle is terminated on the detector end by a plastic ferrule of size 5 mm height and 4.6 mm diameter, with a central hole for the fiber. The fiber tip is permanently glued inside the hole. The ferrule is then attached to the front face of the detector module. For stability of the light transmission and simplicity of attachment, the ferrules will be glued to the surface using soft UV glue (bulk modulus 200). This provides a moisture free soft interface and enables the fiber to be easily disconnected and reattached without damaging the fragile surface of the PbW04 crystal. 2.3. Reference Detectors

For a stable reference detector, three types of photomultiplier tube (PMT) and a Hamamatsu PIN photo diode with low noise charge preamplifier are currently being tested. The reference PIN with the preamplifier is mounted directly on the light mixer. The light pulse is transmitted to the reference P MT via fiber cables from the bundle. Each reference PMT is provided with a radioactive a-source (Am241)on a scintillating YAP crystal. 2.4. Data Acquisition

A CAMAC based Data Acquisition System with LabView DAQ software (National Instruments) is being used for the readout and on-line analysis. The signals from the reference detectors are digitized in a LeCroy 2249W (11-bit) ADC and read-out through the GPIB-CAMAC interface. The events with injected light pulse ("LMS events") will be read-out in the same way as "real events", i. e. electromagnetic shower from particles in the calorimeter module. Each type of events will have appropriate ID tag in the data stream. The monitoring system will provide a reference point through an EPICS variable every 2-3 minutes which will be included in the data stream. The ratio of the response of each calorimeter channel to that of the LMS reference detector, for "LMS events", will be used to correct the "real data":

Light from LMS will be periodically injected into the detector modules between real events during data taking. The frequency of light injection is defined by the optimal performance of the light source and the stability requirement for the PbW04 crystals. On the basis of experimental tests and estimations, this

483

frequency is chosen between 10 and 20 Hz. A statistics of 1000-1500 "LMS events" is sufficient to determine the relative reference constant for each module. This will require 2 - 3 min data taking (in parallel with real experimental data) and will provide full relative calibration of the calorimeter.

...... 1-09 1. . .............

1.08

[

I

I

I

20

..............

'

40

I

I

.:...........

'

60

I

............................... .:................:... ............-;-

'

'

80

I

I

'

'

I

100

120

I

--

'

I

I

140

I

I'

160 Time (h)

1.18 1.17 1.16 1.15

cz2

1.14 1.13 1.12

-

1.11 -_ .................................................................................................................... 1.1

I

I

I

l

1

l

20

Figure 2.

40

1

I

1

60

l

l

I

80

l

l

I

100

l

I

l

l

120

Variation of light intensity of the LMS vs. time

l

-I

140

l

l

-

160 Time (h)

484 1.02 1.015 1.01 21.005

z

Ea

1 ........

p0.995

............................

0.99 0.985 0.98

20

40

60

80

100

120

140

160 Time (h)

600

I .......................................................................... I I I I I I I ~ I I I I ~ I I I I Il l

I I I I I I I I I l ........................................

l

500 400

300 200 100 0

0.98

0.985

0.99

0.995

1

1.005

1.01

1.015

1.02

Stability distribution Figure 3.

Stability of the light monitoring system over the period of 168 hours

2.5. Results of the Stability Test

A long term stability test of the prototype LMS has been performed and the results are presented in this section. For the test, the light intensity has been monitored with the PIN photodiode and three reference PMTs with Am241 radioactive source each. Signals from these detectors were digitized, read-out, analyzed, and variations in time were compared with each other. The following quantities per data point of 1500 "LMS events" are calculated for a stability N

485

factor:

s1 = g& and s 2 =

(3)

for i-th and j-th PMTs, where PIN and PMTi stand for the appropriate mean ADC value of the pulse height as measured with the PIN photo diode and with each P M T (the latter normalized to the average pulse height due to Am241source). In the figure 2, the variation of light intensity as measured with PMT2 and PMT3 versus time is presented. The exposition is over the time period of 168 hours. The distribution of the stability factor, PMTzIPMTs during the same period is shown in the figure 3. The standard deviation of the Gaussian fit to the stability distribution is about 0.1%. 2.6. Conclusion

A prototype Light Monitoring System for the HYCAL detector has been designed, constructed and successfully tested (in the laboratory) on the subject of stability. A stability level of -0.1% has been achieved during a week period test. A new type of light source, a group of blue LEDs connected in series and fired with a common high voltage pulse, has been developed. A long term, six months continuos run of the system, demonstrated very high reliability and good stability of the new light source. The prototype LMS provides up to 700 channels through the fiber optics distribution system. It is ready for beam tests. The final system with 2000 output channels is under construction and will be ready in the fall of 2002. N

3. Acknowledgments This material is based upon work supported by the National Science Foundation under Grant No. 0072466 and 0079840. References 1. D. Dale, S. Danagoulian, A. Gasparian (contact person), R. Miskimen et al.,

A Precision Measurement of the Neutral Pion Lifetime via the Primakoff Effect (PrimEx), Jefferson Lab Experiment E99-014 (1999) 2. Conceptual Design Report. A Precision Measurement of the Neutral Pion Lifetime via the Primakoff Effect. Jefferson Lab, March 3, 2000.

This page intentionally left blank

Cerenkov Calor irnetry Covener: S. White

S. White

Covener’s Report

Q. Deng

Influence of Phase Transition on the Optical Transparency of Lead Fluoride Crystals

0. Atramentov

Explicitly Radiation Hard Fast Gas Cerenkov Calorimeter

Y. One1

Present Status of CMS HF Quartz Fiber Calorimetry

S. Razzaque

Calorimetry of the RICE Detector

*D. Saltzherg

Radio Cherenkov Detection of High Energy Particles

$1. Dumanoglu

Radiation Hardness Studies of High OH- Quartz Fibres for a Hadronic Forward Calorimeter of the Compact Muon Solenoid Experiment at the Large Hadron Collider

*Written contribution not received $Not orally presented

487

This page intentionally left blank

CERENKOV CALORIMETRY

SEBASTIAN WHITE Brookhaven National Labomtory, Upton, N Y , USA E-mail: Sebastian WhiteOcern.ch (Conveners Report)

Cerenkov calorimetry started possibly with lead glass. High quality lead glass produced by Schott in Germany and Ohara in Japan was available in the early 70’s. It was used by Rubbia and co-workers in a C P violation experiment at CERN and soon after by Cool, Lederman and co-workers in the CERN-Columbia-Rockefeller experiment at the ISR. People were excited by the possibility of triggering on high transverse momentum particle production at newly available hadron machines FNAL and the CERN ISR. The lead glass module with its high transparency Cerenkov radiator was the detector of choice. The hunt was on for the W boson. But another discovery- high transverse momentum hadron production, evidence for the partonic structure of protons- lay in the path to this discovery. Soon after, the taste in high energy physics experiments moved to the 41r detectors such as UA1, UA2 and CDF and the style of a large coverage, projective detector made it attractive to use sampling calorimeters customized for the 41r geometry and to add a second, hadron calorimeter layer. In the 1990’s a second type of Cerenkov calorimeter appeared on the scene as a sampling calorimeter. The advantage of the Cerenkov sampling medium as opposed to scintillator was that it was faster, more radiation resistant and, in the case of Cerenkov fiber calorimeters, exploited the directional response to Cerenkov light in a fiber sampling layer. The first application of such a calorimeter was in the NA50 experiment which employed the technique in its “zero degree calorimeter”. An R&D programme in connection with the LHC experiment CMS was begun by Gorodetzky and co-workers at CERN. Under this project the technique called “Cerenkov fiber calorimetry” came into being and the project gave rise to a full description of the calorimeter response vs. incident angle, energy, etc. Several beam tests followed using, for reasons of cost efficiency, PMMA optical fibers but the radiation tolerance of large diameter (compared to the typical

489

490

62.5 pm diameter used in the communications industry) quartz core-quartz cladding and plastic cladding fibers was also studied. Today, Cerenkov fiber sampling calorimeters are an essential component of all RHIC detectors and are used in the “Zero Degree Calorimeters” as a triggering device and to study fragmentation in the very forward region. The Cerenkov sampling calorimeter has been used, as a case study, of the “ultimate non-compensating calorimeter” by Wigmans. The quartz fiber calorimeter has emerged as a key component of the CMS experiment at the CERN LHC. The current status of this project is described, in these proceedings, by the project leader of this effort, Yassar Onel. In another paper, Dumaglu presents the studies of radiation hard fibers evaluated for the CMS HF project at CERN’s LIL beam line. At this year’s Calor2002 conference a new direction in Cerenkov calorimetry was explored by David Saltzberg for study of Ultra-High Energy Cosmic Rays. Extending the detection of Cerenkov radiation to the Ghz radio-frequency region (S-band and C-band). Saltzberg and co-workers open up the possibility of studying the interaction of UHE cosmic rays with the Earth’s moon and in Antarctic ice “Rice” experiment (see presentation by Razzaque). A recent test at SLAC’s final focus test beam called the “kitty litter experiment” measured electromagnetic showers induced by photons with mean energy of -3 GeV. There appears to be very good agreement with the characteristics expected for Cerenkov radiation and the background from other sources, such as transition radiation at the interfaces, is well under control. At SLAC’s FFTB the radio signals from Cerenkov radiation are huge: -100 V/m, and this experiment used a fast digital oscilloscope and some attenuators as the data acquisition system. Perhaps this latest edition of the calorimeter conferences will lead to a new and un-expected session in future meetings on radio antenna design, or a session on the resolution of the mystery of the very highest energy cosmic rays.

Acknowledgments After the initial meeting at Fermilab this conference moved to a series of island venues and then an anti-island at Annecy. The meeting at Caltech was held at an island of science and technology in the week of the “Oscar” celebrations. The excellent organization of the meeting was very much appreciated by this convener.

INFLUENCE OF PHASE TRANSITION ON THE OPTICAL TRANSPARENCY OF LEAD FLUORIDE CRYSTALS

G.H. REN, Q. DENG, Z.K. LI, D.Z. SHEN Shanghai Institute of Ceramics, 1295 Dingxi Road, Shanghai 200050, P . R . China E-mail: rgh4mail.sic.ac.cn

Cubic lead fluoride (PbF2) crystals were grown by non-vacuum modified Bridgman method. Their optical transmission may deteriorate because of oxygen contamination during growth, annealing and machining, storing and carrying. It was found / + a phase in the that the role of oxygen is to promote the transformation of 3 initial cubic structured mono-crystals, which results in the formation of cryptocrystalline or microlithic texture composed of crystalline a-PbF2 grains. This texture is suggested to be responsible for the transmission loss of lead fluoride crystals.

1. Introduction

Dally and Hofstadter once predicted that the ideal material for Cerenkov detector would be a “transparent lead brick.” A near approximation of this brick is the lead fluoride crystal’. Its most outstanding properties are high density (7.77 g/cm3), short radiation length (0.93 cm), large average atomic number as well as a transmission extending to UV. Its Cerenkov light output is sufficient to allow an adequate electromagnetic energy measurements. Since Nineties, cubic lead fluoride (P-PbF2) crystal has been investigated as a Cerenkov radiator for electromagnetic calorimetry2i3. The parity violation experiment A4, which is currently being set up at Mainz microtron (MAMI) will comprise 1,022 pieces of lead fluoride crystals. Lead fluoride crystals are grown traditionally with Czochralski method4. At Shanghai Institute of Ceramics we grow lead fluoride crystals by using non vacuum modified Bridgman method5. No matter which method is employed, it is important to keep the growing system in an oxygen-free atmosphere. The optical transmittance of PbF2 crystal will degraded if a small amount of free oxygen, or oxygen-contaminated impurities, exists in the growth system. This phenomenon is called oxygen contamination for fluoride crystals. In addition, a white frosty layer would appear on the crystal surface if a grown clear lead fluoride crystal is exposed to open moisture atmosphere for a long time, leading to the deterioration of the optical transmission.

49 1

492

The number of Cerenkov photons (dN) at wavelength between dX is given by6 dN

0: dX/X2.

X and X

+

(1)

Good transmittance, especially in the short wavelength region, is important so that the Cerenkov photons yield. Transmission loss, particularly at short wavelengths, should be avoid. In this paper, the transmission loss in lead fluoride crystals as well as their origin are discussed. 2. Experimental

Lead fluoride (PbF2) crystals were grown from orthorhombic PbF2 powder by non-vacuum modified Bridgman method. In order to remove oxygen impurities, such as 02-and OH-, in the system completely, an scavenger, that helps to remove oxygen impurities but does not affect crystal properties and growth equipment, was doped into the starting materials. The raw materials and the scavenger were mixed thoroughly and then put into a platinum crucible. The temperature of furnace was controlled by a computer and kept above the melting point (822°C) of lead fluoride during whole growing process. T The optical transmission of a PbF2 crystal after each procedure was measured by using a Shimadzu UV-2501PC Spectrophotometer. The sample crystal structure was examined by using a RIGAKU D/MAX 2550 X-ray Diffractometer. The microstructure was observed with a CARL ZEISS JENA polarized optical microscopy. A VG Scientific X-ray photoelectron Spectrometer (XPS) ESCALAB MK-I1 was used to analyze the binding energy of different atoms in the crystals. Its excitation source is MgK-alpha (1253.6 eV). The test was carried out in vacuum of lowlopas and Cls energy of carbon atom (284.6eV) was used to calibrate the binding energy of other atoms. 3. Results and Discussion 3.1. Transmittance loss during crystal growth

With non-vacuum modified Bridgman method, it is necessary to keep the crucible completely closed during entire growing process. Usually, lead fluoride crystals grown by this method are transparent and colorless as shown in Fig. l a . XRD pattern shows that they are in cubic structure as shown in Fig. lb. However, with controlled modification of the the growth condition significant degradation of transmittance was observed. For example, the crucible was deliberately not sealed tightly so that oxygen in the atmosphere will enter into the growing system, or even the crucible is sealed tightly, but the amount

494

Figure 3. photograph of marble-like lead fluoride slice under cross polarized light

Five piece of wafer samples with size of 20 x 20 x 1 mm were cut from the same ingot. The numbers, 1 to 5, represent the position of the slices in from the semi-transparent seed (bottom) to the opaque tail (top). After being polished, they were measured with UV-2501PC spectrophotometer. Their transmission spectra are shown in Fig. 4, indicating that the transmission of these samples decreases from 1 to 5, i.e. from the seed to the tail of the ingot. Comparing with a clear sample (No. 0), the absorption edge shifts from 240 to 330 nm. 3.2. lhnsmission loss after annealing in the open atmosphere

A transparent and colorless lead fluoride crystal was cut into small slices with the size of 20 x 20 x 2 mm. They were put in a muff furnace for annealing in an open atmosphere at 2OO0C, 300"C, 350"C, 400°C and 500°C respectively for one hour. After annealing, the samples were cooled to room temperature and then polished. All of them were examined with spectrophotometer and their transmission spectra are shown in Fig. 5. It can be seen that samples annealed at temperatures bellow 300°C do not show any significant change in their transmission. However, if the annealing temperature goes up to 350"C, the transmission between 250 and 350 nm decreases dramatically. With the further increase of annealing temperature, the transmission continues to degrade and the absorption edge shifts toward longer wavelengths. The sample annealed at 500°C becomes milky and loses its transmittance completely. After being polished, the sample annealed at 350°C was evaluated with polarizing microscope. It was found that its transmittance was quite poor and great deal of cryptocrystalline or microlithic grains scattered near the

495

Figure 4. Transmission spectra of clear and milky lead fluoride slices cut from a ingot (The numbers from 1 to 5 stand for the position of slices in the ingot from semi-transparent bottom to the opaque top )

Figure 5 . Transmission spectra of crystal annealed at different temperature

cleavage plane of P-PbF2 crystal as shown in Fig. 6. These grains are different from the background of B-PbF2 phase, even their boundaries between different grains are obscure. XRD indicates that little amount of orthorhombic grains exist in the sample. Owing to the inherent optical anisotropy in orthorhombic PbF2, it is easy to distinguish them from the cubic phase. With the increase

496

of annealing time and temperature, more and more cryptocrystalline grains emerge and finally a microlithic texture is formed by precipitation of large amount of such grains. The boundaries between cryptocrystalline grains will scatter or reflect light and result in the transmission loss of entire lead fluoride crystals.

Figure 6. Optical microscopic photography of lead fluoride slice after annealing at 350 C (transmittance light)

4. Conclusion

We observed degradation of transmission in cubic lead fluoride crystal samples grown in oxygen-contaminated system or annealed in an open atmosphere at temperature higher than 300°C. The contamination of oxygen damages the crystal structure and promote the transition from cubic to orthorhombic phase. The condensation of large amount of small a-phase grains constitutes a cryptocrystalline or microlithic texture that is responsible for the transmission loss of lead fluoride crystals.

References 1. 2. 3. 4.

E.B.Dally and R. Hofstadter, The Review of Scientific Instruments, 39(5), (1968). D.F. Anderson et al., Nucl. Instr. and Meth. A 290(1990). R.D. Appuhn et al., Nucl. Instr. and Meth. A 350(1994). 1.1. Buchinskaya, et al., Proceedings of SPI, Solid State Crystals: Growth and Characterization, edited by Jozef Zmijia et al., Zakopane, Poland, 7-11 Oct. 1996. 5. Ren Guohao et al., Chinese Physics Letters, 2001, 18(7):976 6. Dyoshio Yoshimura and Akihiro Maki, Nucl. Instr. and Meth. 126, 1975: 541.

EXPLICITLY RADIATION HARD FAST GAS CERENKOV CALORIMETER

0. ATRAMENTOV Department of Physics and Astronomy, Iowa State University, Ames IA 50014 E-mail: [email protected] We describe the design and construction of a gas Cerenkov Calorimeter with three unique features: 1) it is constructed wholly of metal and gas and therefore indestructible by any dose of radiation; 2) the Cerenkov threshold is above 10 MeV so it is completely blind to any radioactivation as well as low energy e*y; 3)it has a time resolution of about 50 ps. The response is directionally sensitive and can be exploited, for example, in muon colliders - suppression off-axis muon and decay e halo backgrounds. The manufacturing is robust and inexpensive, making this calorimeter suitable as a luminosity monitor for existing hadron colliders and possibly as the only surviving technology for future colliders at high energies and luminosities above 1035cm-2s-1.

1. Introduction

/

Motivation

The next generation of colliders, with the luminosities and energies currently conceived, places great demands on fast and radiation hard detectors. Even existing colliders with their upgrades push detectors to their limits. Luminosities more than cm-2s-1, energies at the TeV scale, and bunch-to-bunch separations of order n s place rather severe constraint on applicable(suitab1e) technologies. In this paper we describe the conceptual design of a n electromagnetic Cerenkov calorimeter that is fast, radiation hard and has intrinsic electron energy cutoff of around 1OMeV which makes it insensitive to low energy e* ,ybackgrounds that will be rather large in future high energy colliders. Radiation hardness is difficult to achieve with atoms, especially those in solid form, since any damage remains fixed in place within the active medium. A liquid or a gaseous medium for the Cerenkov light can be recycled if damaged. Cerenkov light yield in the visible is given approximately

dN = 370 sin2 BC eV-lcm-l dx For isobutene, for example, it is dx

M

0.5 C(eV. cm)-l

497

498

which is by a factor of a thousand less than that in liquids. In spite of high light yield liquids are not suitable for our design for two reasons. Firstly - very low energy threshold makes them very sensitive to large background of low energy electrons and gammas. Secondly, large Cerenkov angle of around 40" implies a lot of mirrors or light channelling devices. In fact, the main problem with the light yield in quartz fibers is the large Cerenkov angle of about 47", which allows only electrons within a few degrees of this angle to generate photons that will be captured by the rather small numerical aperture of a fiber. In contrast, a gas has an index of refraction, n, which differs from one by a small amount, = 1+6,

6

a)

(3)

Cerenkov light is generated at a small angle(& M of few degrees, which for isobutane is 3.5". This particular property, small Cerenkov angle, is utilized in our design. Collection and transport of the Cerenkov light down highly reflective optical conduits can be accomplished in several ways, but the common feature is that gas gaps are aligned nearly parallel to the shower. One design is to use hexagonal metal rods, 1-cm in diameter and of a suitable length, polished and aluminized on six sides. These rods are arranged in the mosaic pattern such that the space between adjacent rods is about lmm, as shown in Fig. 1.

Figure 1. 3-d view of hex rods and metallic mirror. Cerenkov light generated in the gas between rods will typically make 10 bounces before emerging from the end o f 4 e rods at a depth of 30 cm.

The Cerenkov light is generated by shower of particles as they cross the

499

gas gaps between hex rods. Since the Cerenkov angle is small, a large fraction of the light is channelled down the optical gap conduits. The light makes typically 10-12 small angle reflections before exiting the calorimeter. It should be mentioned that the amount of light that reaches the end of the calorimeter is reduced roughly by a factor

where N is the average number of reflections and R refllectivity per bounce. This power law implies that in order to have 80% of the signal exit the calorimeter one would need to have reflectivity per bounce of around 98%. Such reflectivity is hard to achieve, but at greasing angles of incidence, for any sufficiently smooth surfaces, reflectivity approaches 100% . 2. Choice of Materials

In our design we intend to use only gas and metal. The reason is clear from the following arguments. Threshold energy for emission of Cerenkov radiation in gas is given by

for example, for isobutane at latm energy threshold

E f t M 9.8MeV

(6)

this automatically implies insensitivity to low energy background. Critical energy in metal can be very well approximated by EM

800MeV Z

(7)

which in tungsten is 10.8MeV (8) therefore most shower electrons will participate in the Cerenkov light generation. The main feature of electromagnetic showers (e.g. their longitudinal and lateral sizes) can be described in terms of one parameter, the radiation length X O , which is related to the characteristics of the material by the following relation 716 A M g cm-2 (9) Z ( Z 1)l n ( 2 8 7 / f l ) EW M

xo

+

where A is the atomic weight of the material. The longitudinal shower profile is to the first order independent of the type of material if expressed in terms

500

of X O . The shower maximum, i.e. the depth at which the largest number of secondary particles is produced, occurs at

where t,,, is measured in radiation length, E is the incident particle energy and t o = -.5(+.5) for electrons (photons). The detector thickness that is needed t o contain 95% of the shower energy is given by t95%

x

tmax+ .082 + 9.6

(11)

-

which indicates that calorimeter thicknesses of -25 XOare sufficient to contain TeV showers in the energy range up to a few hundred GeV. Therefore at energies calorimeters are very compact devises.

3. Geometry One of the designs is to use mosaic of polished hexagonal rods. In spite of nice looking geometry this approach might face certain engineering obstacles. Difficulty arises for full-depth hadronic calorimeter in which now l m long rods cannot be supported only at one end (Fig.2) but require some intermediate support that would block the signal. In addition, it is a challenge from engineering point of view to use tungsten as an absorber, since the price to manufacture tungsten rods of hexagonal cross-section can be rather high. Another solution that seems to be devoid of these disadvantages is "lasagna"-like geometry. As shown in the Fig.3 absorber consists of three elements: plate, parallel half-round rods evenly placed on one surface of the plate, and another array of rods on the other side of the plate but shifted with

m

Figure 2. Side view of hex rods and front support structure. This cantilevered support is adequate for 30-cm rods, but not 1.2m rods.

501

Figure 3. shims.

Separate elements of one layer: plate, 2 sets of half round rods, and 2 reflective

respect t o the first by half of a period. This assembly is covered with polished stainless steel shim as shown. The detector itself consists of several copies of such assembly evenly spaced as show on the Fig.4. Lasagna design is more preferable from the following points of view: 1)it allows use of any absorber material (including tungsten); 2) support problem is easily solved; 3) easy and inexpensive to build - it does not require complicated machining, can be assembled out of on-shelf materials (plates, rods, etc.).

Figure 4. Several copies of assembled layers stacked to form calorimeter mass. Space in between is filled with gas.

502

4. Readout

In order to minimize noise in PMTs it is preferable to place them off shower axis. As shown on Fig.1 and Fig.2 spherical mirror can be used to forward Cerenkov light off the shower axis. This detector permits bunch-by-bunch detection of particles, for example, at LC with 1.4ns bunch spacing, at the Super LHC with 12.5ns bunch spacing. Therefore requires fast photodetectors, and digitizers at the GHz frequencies. Hamamatsu provides very fast PMT (R3809U-57), that has rise time of 200ps, and goes as low as 200nm in wavelength. Commercially available digitizers exist, Maxim maxl04 2.2 GHz ADC, for example is fast enough, but we also need to transfer digitized signal to fast memory in order to use bunch-by-bunch information for the whole train. We suspect that a XyLinx with fast memory and integrated logic capabilities will be able to handle this traffic. 5. Expected performance We used Geant3 to simulate 100 GeV electrons in an Fe mass with exact optics in reflective ss tubes. The tubes are 2-mm inner diameter, centered every 5mm. The e- beam is incident at angles from 0.1 to 0.01. Figure 5a shows the photoelectron (pe) pulse height distribution of 100 GeV e- in 2 atmosphere of isobutane with a reflectivity of R = 1; the high side tail is from showers near B x 0.01, as shown on the Fig.5b. These can be ameliorated by increasing the angle of incidence with respect to the optical conduit. The lower two figures, F i g . 5 ~and 5d, show the same distributions but with air as the gaseous medium and a reflectivity per bounce of R = 0.90, a rather poor calorimeter. There is expected decrease in pe yield due to both a smaller 6 and a smaller R N , and a clear degradation in resolution. As we can see on Fig.5a resolution of such detector is around 12%, which is not much but for some applications like luminosity monitors this might be ok. It should be mentioned that resolution improves with energy as 1 I a .

6. Summary/Status We presented generic design of electromagnetic calorimeter that is suitable for future high energy colliders. Geometry that we described is very flexible and such detector can be placed very close to beam. In fact such calorimeter is proposed as luminosity monitor for Next Linear Collider. Its performance can be optimized by proper choice of materials.

503

2oo 160

20

0

70

-.

ID tmr*.

".On

M

1 Isf 4.m 1.611

0.02

::1

0.04

0.06

0.08

0.1

E(GeV) vs. Theta

10

0

0

0.02

0.04

0.06

0.08

0.1

N(pe)/lO vs. Theta

Figure 5 . (a) The pulse height distribution in approximate GeV units (1 GeV % lope) for 100 GeV e- in 2 atm. isobutane ( 6 M 3.8. with a reflectivity R = 1; (b) the pe response as a function of the e- angle of incidence. The high-side tail is from showers near 0 M 0.01. (c) The pe response for 100 GeV e- in a calorimeter filled with air at N T P and with reflectivity per bounce of R = 0.90. (d) The pe response of (c) vs. angle.

Acknowledgments This work was done in collaboration with John Hauptman, Mark Kane, Nural Akchurin, Vladimir Atramentov and supported by ISU and DOE High Energy Physics.

PRESENT STATUS OF CMS HF QUARTZ FIBER CALORIMETRY

Y . ONEL Department of Physics and Astronomy, University of Iowa, Iowa City, IA 52242 E-mail: Yasar-Onel%uiowa.edu (On Behalj of the CMS Collaboration)

The experiments at the Large Hadron Collider will have to deal with unprecedented radiation levels. The design of the CMS forward calorimetry detector (HF) is now finalized. The present design of CMS calls for the HF calorimeter to be based on quartz fiber technology. It consists of two modules, located symmetrically at about 11 meters from either side of interaction point. They cover the pseudorapidity range 3-5. The length along the beam is 1.65m or 10 nuclear interaction lenghts. Each calorimeter consists of a large steel block that serves as the absorber. Embedded quartz fibers in the steel absorber run parallel to the beam and constitute the active component of the detector. In order to optimize energy resolution for E and ET flows and forward jets, the calorimeter is effectively segmented longitudinally by using two different fiber lengths. The present status will be discussed.

1. Introduction

The Large Hadron Collider (LHC) together with its major detectors, ATLAS and CMS, will provide the premiere instrument to search for new high energy physics phenomena in the early years of the 21st century. In particular the CMS detector is designed to exploit the full range of physics at the LHC up to the highest luminosities 1034cm-2s-1 for p+p. With this detector it will be possible to push the standard model to its limits and look for new physics. Of all the CMS physics goals the ones we find most intriguing are the heavy Higgs and SUSY searches. The Forward Calorimeter (HF) (Figure l ) , which covers the pseudorapidity range 3 < 7 < 5, relates directly to these searches. It improves the detection of missing transverse energy by an order of magnitude in the range below 100 GeV and enables very forward jets t o be identified. These jets are instrumental for standard model heavy Higgs as well as SUSY Higgs searches.

-

504

505

2. Physics Goals of the Very Forward Calorimeter

The HF improves the detection and measurement of missing transverse energy ( E Z i s s ) below 100 GeV and enables very forward jets to be identified. These jets are a distinguishing characteristic of several important physics processes. The measurement of (EZiss) is important for the study of top quark production and for Standard Model (SM) Higgs searches for m x 500 GeV in the H -+ ZZ -+ llvv and H -+ WW -+ lvjj channels. It is also important in SUSY Higgs searches (MSSM) for A -+ tt -+ ep EZiss and A -+ tt -+ l*hk EZiss, allowing one to reconstruct the mass of A. It is clear that if the coverage in q of the detector falls below 5.0 the missing ET measurement is seriously compromised below 100 GeV. Studies of WWg and W W Z gauge boson couplings and searches for supersymmetric particles are also improved by the presence of the HF. Many of these processes are accessible at low luminosities ( L 1032m-2s-1). The search for a heavy Higgs boson ( m H lTeV), with the decay H -+ WW -+ lvjj,H -+ ZZ -+ l l j j , requires high luminosity, ( L 1034cm-2s-1). The production of the Higgs in this mass range, through the WW or ZZ fusion mechanism, is characterized by the appearance of two forward tagging jets. The jets are energetic ( p ~ 1TeV) and have a transverse momentum of order m w . They are emitted in the pseudorapidity range 2.0 < 1771 < 5.0. The detection of the tagging jet greatly enhances the signal and, for a given missing ET cut, the QCD background p p -+ 2 + j e t s is strongly suppressed by extending the 77 coverage to 5.0. It is mandatory in order to suppress the large QCD W,2 + j e t s background. Detecting forward tagging jets is also useful for the 80 to 140 GeV mass range of the Higgs. The study of longitudinal intermediate boson (WL) scattering (whether resonant or not) will also require the detection of forward tagging jets. This study will be important if WL scattering becomes strong, particularly if the SM Higgs does not exist or if it is extremely massive. It is clear from the above discussion that the extended 17 coverage provided by HF is crucial to the above physics. In addition the HF is required to have moderate energy resolution, with a constant term 10% for single hadrons, and sufficiently fine granularity to tag forward jets. The granularity is needed to reject fake jets formed by particles coming from superimposed minimum bias events and for adequate tagging of jet angular resolution in the transverse plane. The HF is required to have a very fast response time and to be radiation hard.

+

+

-

-

-

-

506

3. The CMS Forward Calorimeter Detector (HF)

The present design of CMS calls for the HF calorimeter to be based on quartz fiber technology. It is to consist of two modules, each with an active radius of 1.4 m, located symmetrically at 11.1m from either side of the interaction point. They cover the pseudorapidity range 3-5. (see Fig 1) Lo

*

g

q51.1

Figure 1.

. n-1

Y 3

P

a

3" %lea5

G

H

3

3

The q coverage of the CMS detector.

The length along the beam is 1.65 m or about 10 nuclear interaction lengths. Each calorimeter consists of a large steel block that serves as the absorber. Embedded quartz fibers in the steel absorber run parallel to the beam and constitute the active component of the detector. HF is constructed such that there are two different lengths of fibers inserted into the absorber from the back end. The fibers that run the entire length of the absorber (165 cm) are called Long fibers. The Medium length fibers are shorter by 22 cm (15 Xo)from the front of HF. Consequently, Long fibers constitute the electromagnetic (EM) section, Medium fibers, the hadronic (HAD) section. The two section are readout by different photomultipliers. Quartz-fiber calorimeter technique is based on the detection of Cherenkov radiation that is emitted by a charged particle traversing a fiber with a velocity greater than speed of light in quartz. The opening angle of the Cherenkov radiation is a function of the velocity of the particle, ,b :

507

where n is the refractive index of quartz (1.45). Particles entering the absorber produce showers of particles. Charged particles with a p greater than the threshold value emit Cherenkov radiation. Showers produced this way are narrow. For showers produced by electrons, for example, the transverse dimensions of the shower is smaller than the corresponding Moliere radius which is defined as RM = XoEB/Ecwhere X O is radiation length, E, is 21 MeV and E, is the critical energy. E, is defined as the energy at which an electron’s energy loss by bremsstrahlung and by ionization are equal and characterized approximately as E, = (800) MeV/(Z 1.2). Producing the light by the Cherenkov effect has a number of advantages. First, HF is only sensitive to relativistic charged particles. Consequently, only electrons and positrons produced in e.m. showers give rise to an appreciable shower signal. Hadron showers predominantly register through their e.m. shower core, which has a twofold benefit. First, the shower profile is considerably narrower than in conventional calorimeters. Measurements we have performed in 1995-96 with prototypes confirm this feature. The width needed to contain 90% of the Cherenkov signal was measured to be more than a factor of three smaller than the width of the dE/dx distribution. Secondly, the instrumental depth is considerably smaller than that required for full containment of the hadron shower (8X vs 12X). Another advantage of this technology is that HF is insensitive to low-energy (MeV) neutrons which will traverse it in large numbers. It is also largely insensitive to the effects of induced radioactivity. Quartz fiber also has the advantage of being radiation hard. Finally, the detector is extremely fast. The production of Cherenkov light is an instantaneous process and timing tails resulting from shower thermalization are absent. This feature was also confirmed by our CERN measurements which showed that all the light produced in hadron showers was collected in 12 ns.

+

4. Prototypes

We have built several HF prototypes. A hadronic prototype (HAD95) designed to test the quartz fiber technology was tested in the H4 beam at CERN during 1995-98. The primary purpose was to ascertain the response to pions and electrons among other characteristics special to quartz fiber calorimeters. A wide range of particle energies was studied; for electrons, 10 5 E, 5 200 GeV; and for pions, 35 5 E, 5 350 GeV. An electromagnetic calorimeter was built in 1996 (EM96), and the same year a combined test (EM96+HAD95) was carried out. In 1999, we have designed a Pre-Production-Prototype and tested its performance at the CERN H4 test beam. The results are presented in the following section.

508 5 . Pre-Production-Prototype I

The PPP-I is an iron matrix, serving as an absorber, embedded by quartz fibers of 300pm core diameter which serve as active material. The iron matrix is composed of 2.5mm thick iron layers which have grooves at every 2.5mm in both transverse directions. The length of the absorber is 165cm with a cross sectional area of 18cm x 18cm. In each groove, a single quartz fiber is inserted. The 6000 fibers are grouped into 27 bundles. There are three different lengths of fibers to achieve a longitudinal segmentation. They are called electromagnetic (EM) , hadronic (HAD), and tail catcher (TC). The physical spacing between fibers is the groove 2.5mm spacing. The pattern used for the fiber insertion is shown in Fig. 2. There are two EM fibers for every HAD and T C fiber. PPP-I is divided into nine physical regions (towers) each with a 6cm by 6cm cross sectional area (see Fig. 2). Each fiber bundle (EM, HAD and TC) from a tower is coupled to a separate photomultiplier tube (PMT) via a light-guide and read out as a separate channel. The performance of the PPP-I was tested during a beam test in September 1999. The detector was placed on a platform which could move in three dimensions with respect to the H4 beam line of the Super Proton Synchrotron at CERN. The results are summarized in the following sections. 6. Measurements 6.1. Spatial Uniformity of P P P - I

In order to study the spatial uniformity of the PPP-I, a 120GeV electron beam was moved with 1 centimeter steps across the face of the detector, and the signal of three adjacent towers (towers 4, 5 and 6. See Fig. 2) were plotted as a function of the beam position. The beam spot size was 2cm2 and, with the help of drift chambers located upstream, the impact position of every single particle of the beam on the detector face was known with a precision of 200pm. In Fig. 3, the response of the detector is shown for three adjacent towers. As the beam moves from one tower to another, the measured signal amplitudes for adjacent towers change. A sharp tower to tower transition is seen due to the narrow lateral profile of Cherenkov-light generating particles. The sum of the signals of three towers is also shown with connected data points in the same figure. A similar beam scan was also carried out with a 120 GeV T - beam that was moved vertically across the face of the PPP-I. The signal amplitudes recorded

510

Figure 4. Vertical scan with 120GeV pions.

Figure 3. Horizantal scan with 120GeV electrons.

6 . 2 . PPP-I Energy Resolution

The energy resolution of a calorimeter can be parametrized as

(

a)2= ( E

a

a

-)2

+ b2

The first term is the sampling term and characterizes the statistical fluctuations in the signal generating processes. The second term is a constant term that is related to the imperfections of the calorimetry, signal generation and collection non-uniformity, calibration errors, and fluctuations in the energy leakage from the calorimeter. The energy resolution of PPP-I has been studied as a response to both electron and pion beams at different energies. The response of the PPP-I has been recorded as a function of beam energy. The response of the calorimeter to electrons is seen to be Gaussian. However a deviation from Gaussian behavior is seen in the response to n- beam. This is a result of the different natures of electromagnetic and hadronic showers. A hadronic shower contains an electromagnetic core. The Cherenkov signal generating mechanism of the hadronic shower is, in fact, mostly this electromagnetic core since most of the charged pions are too slow to generate Cherenkov light. The electromagnetic core is carried by no’s and the number of nos fluctuates from shower to shower. At low energies, the number of nos in the shower is characterized by a Poisson distribution which becomes more and more Gaussian at high energies.

51 1

Figure 5. EM fiber locations revealed with 120GeV electrons. The source tube location is also visible.

Figure 6. PPP-I electromagnetic energy resolution as a function of I/@.

Fig. 6 shows energy resolution for electrons as a function of energy. The resolution ( a / E ) is plotted against 1f fi and fitted to equation 2. The fit yields

Hadronic energy resolution of PPP-I as a function of energy is shown in Fig. 7. At 1 TeV the energy resolution reaches 20%. 6.3. PPP-I Energy Linearity Energy linearity is another important characteristic of a calorimeter. The energy response of PPP-I was normalized to the beam energy and this normalized response is plotted against the beam energy. (Fig. 8). We have finalized our detector design. The results of the test beam with the hadronic and electromagnetic prototypes specify the following main features and form the basis of the H F design; a) The electromagnetic energy resolution is 200%E where E is the particle energy in GeV for a packing fraction of 0.85% quartz. b) The light yield is 0.25 photoelectrons/GeV for electromagnetic showers for the same packing fraction. For hadronic showers the light yield increases with the energy. c) The hadronic energy resolution contains an intrinsic component due to the fact that the Cherenkov mechanism responsible for the signal

512

1.025

1

0.8

0.6

0.975

0.1

0.2

0.925 00.9 0

25

50 .

75

1W

125 g I50

175

2W

5

Enerw ( C d t

Figure 7. PPP-I hadronic energy resolution as a function of s- beam energy.

Figure 8. Normalized response of PPP-I t o e - s as a function of beam energy.

generation essentially selects only the 7ro component of the developing showers. This irreducible component amounts t o 25% at 100 GeV extrapolated from the test beam data t o 1 TeV, this component is 10%. d) The calorimeter response was found t o be dependent on the impact position of the incident particles. e) The energy resolution of the quartz calorimeter contains contributions from the following components; Photoelectron statistics, sampling fluctuations and a constant term.

7. Radiation D a m a g e Studies After the analyses of the data from the prototypes, a major concern was the effect of radiation damage to the quartz fibers. In 1998 we attacked the problem of radiation damage on two fronts: The forward calorimeter in CMS will experience unprecedented particle fluxes. On average, 760 GeV for each p+p collision is incident on the two forward calorimeters, compared to only lOOGeV for the rest of the CMS. Moreover, this energy is not uniformly distributed, but has a pronounced maximum at the highest rapidities. At 7 = 5 and integrated luminosity of 5 x lo5 pb-' ( 10 years of LHC operation), the HF will experience 1 Grad. The neutron rates (En x 14MeV) at the front of the detector will be of order of lo8 Hz/cm2. The charged hadron rates will also be extremely high, especially at the shower maximum of the HF, the rate will reach 1013 to

513

10l6 Hz/cm2. This hostile environment presents a unique challenge in particle detection techniques. (1) By studying the effect produced on the optical properties of single quartz fibers: We investigated darkening of nine high OH- fiber types irradiated with 500 MeV electrons from the LIL at CERN. The transmission of Xe light was measured in situ in the 350-800 nm range. The induced attenuation at 450nm is typically 1.52 f 0.15 dB/m for 100 MRad absrobed dose. (2) By building a new module to study the degradation of its performance in a test beam after subjecting it to intense radiation: A new module, nick-named Raddam 98, was designed and built at Iowa in 1998 specifically to study the degradation of performance in a beam test after it was subjected to intense irradiation at CERN’s LIL facility. It was subjected to a total dose of 700 Mrad, which corresponds to approximately 7 years of operation in the LHC environment at q = 5, the worst case for HF. When tested before irradiation with electrons at 80 GeV, the resolution was 9%. After irradiation, the resolution worsened to 15%.

Figure 9.

H F wedge structure

514

8. Mechanical Overview and Final Parameters

The HF is presently under construction. HF will sit on a riser to align HF with the beam height. The HF calorimeter is built up of several componets as shown in Fig. 9. The absorber is made up of 18 wedges along with their base plates, PMT boxes and, shielding. Each wedge encompasses 20 degrees. The fibers are connected into bundles that go into light guides. A PMT sits directly on a light guide. The HF fiber matrix is optimized for the required performance and ease of optical assembly. The fiber spacing is on a 5mm x 5mm grid, and the quartz fiber packing fraction is 0.85%. The fiber outer diameter is 800pm with a 600pm core. The 11 x q5 segmentation is 0.175 x 0.175 and the length of the EM and HAD fibres are 165cm and 143cm respectively.

Acknowledgments This project was supported by the US Department of Energy (DE-FG02-91ER 40664) and NSF (NSF-INT-98-20258), the Hungarian National Fund (OTKA T026184), the Scientific and Technical Research Council of Turkey, TUBITAK (TBAG-1590), the International Science Foundation (grants M82000 and M82300) the State Committee of the Russian Federation for Science and Technologies, and the Russian Research Foundation (grant 95-02-04815).

References 1. Test Beam Results of CMS Quartz Fiber Calorimeter Prototype and Simulation of Response t o High Energy Hadron Jets, N. Akchurin et al., Nucl. Instr. and Meth. A409 (1998) 593-597. 2. On the Difference Between High-energy Proton and Pion Showers and Their Signals in a Non-compensating Calorimeter, N. Akchurin et al., Nucl. Instr. and Meth.A408 (1998) 380-396. 3. Test Beam of a Quartz-fiber Calorimeter Prototype with a Passive Front Section, N. Akchurin et al., Nucl. Instr. and Meth. A400 (1997) 267-226. 4. Test Beam Results from a Fine-sampling Quartz Fiber Calorimeter for Electron, Photon and Hadron Detection, N. Akchurin et al., Nucl. Instr. and Meth. A399 (1997) 202-226. 5. Quartz Fiber Calorimeter, N. Akchurin et al., Nucl. Instr. and Meth. A379 (1996) 526-527. 6. Results From the Beam Test of the CMS Forward Quartz Fiber Calorimeter PreProduction-Prototype (PPP-I), A. S. Ayan et al., CMS N O T E 2002/021.

CALORIMETRY OF THE RICE DETECTOR

SOEBUR RAZZAQUE Department of Physics and Astronomy, University of Kansas, Lawrence, KS 66044 E-mail: soebOkuhep5.phsx.ukans.edu

Radio Ice Cherenkov Experiment (RICE) is an ultrahigh-energy (UHE) cosmic ray neutrino detector for neutrino energies greater than a PeV. This pilot project explores the radio detection technique for UHE particles. Deployed at the Antarctic polar ice cap, RICE antennas have been operational since 1996. Basic calibrations of the antenna array have been done using data taken mostly in situ. The calibration results and an upper limit on electron neutrino flux based on one month of analyzed data are reported here.

1. Introduction

UHE neutrinos (v) can be detected from their charge current neutrino-nucleon (vN)interaction with matter. Up to 80%of the energy of an electron neutrino (ve)is transferred to a secondary electron in such an interaction. The electron then produces an electromagnetic cascade shower inside the material. Highenergy charged particles in the shower radiate a Cherenkov signal, as their speeds are greater than the speed of light in the medium. Due to processes like Compton, Moller and Bhabha scatterings, and positron annihilation, a net negative charge is created in the shower which propagates and emits coherent Cherenkov signal at long wavelengths' UHE v detection requires large target volume because of small vN crosssection and small v flux. Detection of coherent radio signal requires the target medium to be radio-transparent. Antarctic ice fulfills these condition^^*^^^ The attenuation length for radio signals is 1 km at 1 GHz and 10 km at 100 MHz. Detailed analyses for an experiment to detect UHE v's using coherent radio signals was encouraging5 and prompted design and implementation of the RICE prototype detector6 The concept of RICE is illustrated in Fig. 1. It is a multipurpose experiment; the goals include UHE v detection, testing of the standard model UHE vN cross-section and pilot application at UHE of the coherent radio Cherenkov signal emission, which has been tested at GeV energies in the laboratory7 The effective detection volume ( V e ~for ) a radio array is much bigger for the same instrumented volume than an optical array

515

516 RICE Concept

UHE V e Radio ! transparent ice

300 rn

v N CC-interaction -80% of Wenergy to e

Radio Antenna

Cherenkov pulse

Figure 1. A cartoon of UHE v, detection by RICE using radio antennas buried in ice.

such as Antarctic Muon and Neutrino Detector Array (AMANDA) above 1PeV shower energy8 This makes RICE an attractive complimentary experiment to AMANDA at UHE. 2. Detector Array and Data Acquisition

RICE antennas are deployed in a 200 m cubic array which is 100-300 m beneath the ice surface, approximately 1 km away from the geographic South poleg Currently there are 20 dipole radio receivers (Rx), 5 under-ice radio transmitters (Tx) and 3 surface horn antennas (TEM) in the array. 14 Rx’s are co-deployed with photomultiplier tubes in AMANDA-B holes and 6 Rx’s are deployed in the holes drilled for RICE. Each Rx is a 30 cm dipole with good sensitivity in the frequency range 0.21 GHz. The signal from each Rx is immediately amplified by a 36 dB in-ice amplifier (see Fig. 2), then carried by 300 m coaxial cable to the surface electronics board. The signal there is first filtered by a 200 MHz high-pass filter to reduce noise from AMANDA photo-tubes and 149 MHz continuous wave (CW) background from the South pole station and then re-amplified either by a 52 dB or by a 60 dB amplifier. The re-amplified signal is then split into two copies; one copy is fed into a CAMAC crate (which houses the LeCroy 3412 discriminator), and the other copy is fed into a digital oscilloscope channel. The waveform from the oscilloscope is recorded in the hard-drive of a P C when a signal from the discriminator via trigger logic is passed on to the N

517 Scope (HP54542)

200 MHz high-pass filter

7 -300 m cable

I

t

52/60 dB amp. 36 dB in-ice

amp’

I

30 cm dipole antenna

DAQ: Antenna to Trigger

Figure 2. Schematic diagram of the RICE signal path from an under-ice antenna to the data acquisition electronics.

xcilloscope. 2.1. Event Trigger and Vetoes

The RICE event trigger is based on a preset 1.2 ps time window ( A t ) (time required for electromagnetic signal to pass through the array)and an adjustable discriminator threshold voltage above the noise level which determines the rms voltage seen on the oscilloscope channels. A general RICE event trigger is defined when, in a time window of At: (i) st least 4 Rx’s record voltages above 5a x V,,,, or (ii) one Rx records voltages sbove 50 x V ,, and 30-fold AMANADA-B phototubes trigger, or (iii) one Rx records voltages above 50 x V ,, and a high multiplicity SPASE trigger. The event is vetoed if (i) one or more of the surface horn antennas are also hit within At, indicating a noise generated on ice-surface, or (ii) the pattern 3f Rx-hits matches with known noise patterns (checked in software). If an event passes these criteria, waveform data (8912 ns traces) from the xxilloscopes is written on a disc (- 10 s/event). Besides the general triggers, 3, waveform measurement is also recorded every 600 s (unbiased event) and the V, is updated. The raw event trigger rate before veto is about 30/s and the Lypical livetime of the experiment is about 80%. 3. Calibrations

The source vertex of an event is determined by solving an equation: I&, tS O U r C e l = ct/n for each of the 4 hit Rx’s, where n is the refractive index for ice. To calibrate antennas for better timing and vertex resolution, short duration d s e s were broadcasted from an in-ice T x to the Rx-array. Measured Rx-to-Rx

518

(ij)timing delays btij from hit Rx’s and expected timing delays (after vertex reconstruction) were used to perform a minimum x2 and hence calibrate Rx’s afterward. Typical timing resolution is 1.5 - 2.0 ns for good channels and typical vertex resolution is 5 m after calibration. The gains of the Rx’s were calculated using measurements taken in sitv with full circuit (amplifiers, cable, splitters etc.) configuration. To calibrate the Rx gains, a 1 mW CW signal was broadcasted from a Tx in 0-1 GHz frequency range with 1000 bins. The response of each of the Rx’s was recorded and was compared to the expected gain from the University of Kansas antenna testing range (KUATR) data. At the KUATR, the gain of an Rx (similar to the ones deployed at the pole) was measured from the broadcast signal from a ) ) the Rx, (ii) angular Tx taking into account: (i) the effective height ( h e ~ ( w of Tx/Rx efficiency, (iii) cable loss, and (iv) thermal noise. The uncertainty in full circuit gain is f6 dB.

-

-

E(v) rmlutan: NhiblO (.) I N h k = l O (. -)

-1.0

-0.5

0.0

0.5

1.o

bplo(E(v.remnalruaed~E(~,~nerated))

-400

-200 0 2w A(z(remnaIrucId) .~(Ime)).meters

4m

Figure 3. Left pannel: Histogram of number of events vs. logarithm (base 10) of Reconstructed/true neutrino energy, as obtained from simulated events. The sample is divided into high hit multiplicities (>lo) and low hit multiplicities (510) with solid and dashed lines. The energy resolution is obviously better for events with higher hit multiplicities. Right panel: True interaction depth vs. deviation between reconstructed and true interaction depth (Az) of the simulated events. The reconstructed vertices here are obtained using the analytic 4-hit vertexing algorithm. The size of the squares corresponds to the number of reconstructed events in the Monte Carlo simulation.

4. Monte Carlo Event Simulation

Monte Carlo Event Simulation: A Monte Carlo which takes into account the gross features of the calibration results has been developed to simulate

519 neutrino events. Fake u, events are simulated in a rectangular box shaped region 2 km on a side around the RICE array and 1 km in depth. After specifying a vertex location and the u energy, a radio signal from the shower created is simulated using GEANTlO Thermal noise is also simulated at each Ftx location. The event is reconstructed using a 4-hit vertexing algorithm after taking care of the amplifier gain and the cable loss, and smearing the signal by uncertainties in the gain ( f 6 dB) and the Ftx time ( f 2 ns). The source direction is determined by fitting a Cherenkov cone of 56" half width to the hit Ftx's. The results from 10,000 simulated Y, events with energy 10 PeV each are plotted in Figs. 3 & 4.

0

50

100

150

d0 (true-reconstructed direction)

Figure 4. Monte Carlo prediction for RICE array angular resolution in units of degrees. More than half of the simulated 10,000 events are reconstructed within 10 degrees 3f the actual event locations.

Figure 5. RICE 95% CL upper limits on the u, flux (shown with heavy lines). Also shown are different flux models used to calculate the limits. The flux models are, at l o 7 GeV from top to bottom, correspond to (i) Stecker & Salamon, (ii) Protheroe, (iii) Mannheim-A, (iv) Protheroe & Stanev, and (v) Engel, Seckel & Stanev.

5. Results

The RICE array was operating in a stable configuration during August 2000 and data from that period have been analyzedll The total experimental live time for August 2000 was 333.3 hrs, during which time the experiment detected no Y event. Based on non-observation of a Y event, 95% C.L. limits have been put on the Y, flux corresponding to various flux models13 (see Fig. 5).

520

6. Summary and Outlook RICE is signal, not background limited. It attains the vertex, angular and energy information necessary to identify a Y event. RICE demonstrates the radio technique to detect UHE v’s works. With one month of fully analyzed data, RICE has achieved the upper limits on v, flux competative with other experiments in the 100 PeV - 10 EeV energy range12 Future improvements under consideration for RICE include (i) deploying more antennas in future AMANDA/ICECUBE holes to increase the exposure, (ii) adding a hadronic shower component in the simulation, (iii) transmitting signal via optical fiber and/or in-ice waveform processing, (iv) rejecting surface background in hardware and (v) improving signal t o noise ratio using crosspolarized antennas.

Acknowledgements We thank AMANDA for logistic support. RICE was also Supported by grants from the National Science Foundation Office of Polar Programs, the University of Kansas General Research Fund, the NSF EPSCoR Program, the University of Canterbury Marsden Foundation, and the Cottrell Research Corporation.

References 1. 2. 3. 4.

5. 6. 7. 8. 9. 10. 11. 12. 13.

G. A. Askar’yan, Sov. Phys. JETP14, 441 (1962). M. A. Markov and I. M. Zheleznykh, Nucl. Inst. Meth. A248, 242 (1986). A. L. Provorov and I. M. Zheleznykh, Astropart. Phys. 4, 55 (1995). J. P. Ralston and D. W. McKay, in Arkansas Gamma Ray and Neutrino Workshop 1989, edited by G. B. Yodh, D. C. Wold, and W. R. Kropp [Nuc. Phys. B (Proc. Suppl.) 14A, 356 (1990)], reprinted in Proceedings of the Bartol Workshop on Cosmic Rays and Astrophysics at the South Pole, AIP, New York (1990). G. Richter, J. P. Ralston, and D. W. McKay, Phys. Rev. D53, 1684 (1996). D. Besson in Radio Detection of High Energy Particles 2000, AIP Conf. Proc. No. 579, AIP, Melville, NY (2001). D. Saltzberg et al., Phys. Rev. Lett. 86, 2802 (2001). P. B. Price, Astropart. Phys. 5, 43 (1996). I. Kravchenko et al., astro-ph/Oll2372, (2002). S. Razzaque et al., Phys. Rev. D65, 103002 (2002); in Radio Detection of High Energy Particles 2000, AIP Conf. Proc. No. 579, AIP, Melville, NY (2001). I. Kravchenko et al., result paper, (2002). J. Learned, in Neutrino 2002, Munich, Germany (2002). F. Stecker and M. Salamon, Space Sci. Rev. 75, 341 (1996); R. Protheroe, ASP Conf. Ser. V121, 585 (1997); K. Mannheim, Astropart. Phys. 3, 295 (1995); R. Protheroe and T. Stanev, Phys. Rev. Lett. 77,3708 (1996); R. Engel, D. Seckel and T. Stanev, Phys. Rev. D64, 093010 (2001).

RADIATION HARDNESS STUDIES OF HIGH OH- QUARTZ FIBRES FOR A HADRONIC FORWARD CALORIMETER OF THE COMPACT MUON SOLENOID EXPERIMENT AT THE LARGE HADRON COLLIDER

I. DUMANOGLU, E. ESKUT, A. KAYIS-TOPAKSU, N. KOCA, A. POLATOZ, G. ONENGUT Cukurova University, Adana, Turkey

J.P. MERLO, N. AKCHURIN~,u. AKGUN, s. AYAN, P. BRUECKEN, I. SCHMIDT, Y. ONEL University of Iowa, Iowa City, U.S.A

A. FENYVESI, K. MAKONYI, D. NOVAK Atomki, Debrecen, Hungary

M. SERIN, M. ZEYREK Middle East Techn. University, Ankara, Turkey (This paper was accepted, but not presented in the Calor.200.2)

Darkening of various types of high OH- fibres were studied by irradiating them with 500 MeV electrons. The transmission of Xe light was measured in situ in the 300-700 nm range. The induced attenuation at 450 nm is typically (1.52 f 0.15) dB/m for 100 Mrad absorbed dose. Two-parameter fits for darkening were presented. After irradiation the tensile strength remains essentially unchanged. For Polymicro quartz core fibres the tensile strength is typically (4.6 f 0.4) GPa.

1. Introduction The Large Hadron Collider(LHC) will collide protons at 14 TeV centre of mass energies. The Compact Muon Solenoid(CMS) experiment' is one of the two general purpose detectors, and is optimised for searching new phyiscs at the LHC. The Hadronic Forward(HF)2 calorimeter is a sub-detector of CMS and will cover the pseudorapidity of 3 < q < 5. The HF consists of quartz fibres Now at Texas Tech University, Lubbock, USA

521

522

embedded in an iron absorber. Particles hitting the absorber generate showers, and subsequently some of these secondary particles produce Cherenkov light in the fibres. At LHC during the high luminosity running, the HF will receive doses up to 100Mrad/year at q = 5. The radiation hardness of the fibres to various type of particles is crucial for satisfactory operation of this detector. Darkening, recovery and mechanical strength of the various types of fibres were studied by irradiating them using 500 MeV electrons which were delivered by the Linear Injector(L1L) for Large Electron Positron (LEP) collider at CERN. 2. Experimental Setup and Data Taking

LIL delivers 500 MeV electrons to LEP in supercycles of 19.2 s which consist of 16 cycles each with 1.2 s duration. The fibres were irradiated in the LIL experimental area (LEA) in parallel with LEP operation. To get an aproximately constant dose rate, only 4 cycles of supercycles were used. Each cycle delivers 120 bursts of 10 ns duration(fwhm). At 6 x lo9 e-lbwst, the mean rate is 1.5 x 10l1 e-/s and the instantaneous rate is 4.8 x 1017 e-/s. The mean dose over the fibre length for an exposure to 10l6 e- is about 40 Mrad; the corresponding dose rate is 600 rad/s. The fibres were embedded in an iron bar (40 mm high, 12 mm thick and 1 m long) which was placed on a remotely controllable table. Two fibres were placed at 4.5 mm depth from the front surface of the iron and a set of radiophotoluminescence (RPL)3 dosimeters were inserted behind them for measurement of dose during irradiation. The iron absorber was installed at a slope of 8% relative to the beam direction. Electrons will see an effective thickness of 5.5 cm of iron in front of the fibres. This configuration allows the maximum of 500 MeV electron shower to be in the iron where fibers were installed and to transfer the maximum possible dose to the fibres. Sweeping the beam horizontally within 8 cm allows us to irradiate 100 cm of the sample fibres. The optical setup uses components from Ocean Optics, Inc. A two-channel spectrometer (SD2000) triggers a Xe lamp and reads the transmitted light intensities at the different wavelengths. The Xe lamp generates a spectrum in the range of 160-1000 nm. Spectrometer is sensitive to light between 300nm to 800nm. All optical components and fibres were coupled with SMA connectors. All measurements were done in situ. The Xe light pulses were sent through a 35 m long fibre, to a Y fibre which splits the incoming light into two irradiated fibres. The transmitted light coming from two sample fibres is transmitted along two other long fibres and terminates at the slave and master channels of the spectrometer. This

523

setup allows direct comparison of two sample fibres as a function of time. The light source and spectrometer were kept at constant temparature (10f 1)OC. The experimental area where the fibres were irradiated, was at room temperature. The integration time was set to 100 ms. The beam was turned off every 2 minutes to prevent any light generation due to beam crossing. The beam was turned on as soon as the sample spectra were recorded. The fibres were exposed to a dose of about a 100 Mrad. The number of electrons, time t, and table position were recorded and monitored until the end of the irradiation.

3. Analysis and Results Table 1 shows the charecteristics and origin of the the tested fibres. Fibres were supplied mainly from two different companies. All fibres have a quartz core (9) but quartz(q) or plastic(p) cladding. able 1. Characteristics and origin of the tested fibres. Fibres were manufactured in Polymicro echnologies Inc., US (PT), and in Hesfibel, Turkey (H). SSU type preform were produced either Russia (R) or by Heraeus in Germany. F-110 is only produced by Heraeus. Diameter (pm) c/cl/b

Quartz/Origin

Core( c)/clad( cl)/buffer(b)

SSU (G)/Heraeus

Silica/F-silica/Kapton

300/330/370

SSU (R)/Russia

Silica/F-silica/Kapton

300/315/345

F-110 /Heraeus

Silica/Polymer/ Acrylate

300/330/500

F-110 /Heraeus

Silica/Polymer/ Acrylate

300/330/350

SSU (G)/Heraeus

Silica/F-silica/ Acrylate

300/315/345

SSU (G)/Heraeus

Silica/F-silica/Kapton

400/440/480

SSU (R)/Russia

Silica/F-silica/Kapton

400/420/460

FllO /Heraeus

Silica/Polymer/Acrylate

400/430/730

FllO /Heraeus

Silica/Polymer/ Acrylate

400/430/450

3.1. Direct Comparison of Two Types of Fibres The ratio of the two transmissions p y x , t ) = RQ, t ) / R j ( X ,t )

gives a direct comparison of the two fibres tested, independently of dose rate, Xe lamp fluctuation or other systematic effects. Ratios p33(A,t ) and p77(A, t ) were measured and were close to 1 as expected. Using these ratios, we estimate the systematic error of the ratio R to be f3% at 100 Mrad.

524

Figure 1 shows comparison of fibres 2 and 4 (see table 1 for fibre specifications). As can be seen from the figure, while the difference is small at 450 nm it becomes larger a t 610 nm. Different quartz preforms (Heraeus for qp and Russian for qq) were used in the core of the fibres. Both fibres were manufactured by PT. After the beam stop, we continued to take data for recovery measurement. During the recovery phase, the ratio stays almost constant following some recovery adjustments.

“‘

“Q 1.3 1.2

1

Darkening

1.1

Figure 1. Ratio of normalised transmission of qq (no : 2) and qp (no : 4) fibres (300prn) drawn by Polymicro with G(qp) and R-quartz(qq). Upper curve at 450 nm, bottom curve at 610 nm. One hour of darkening correspond to 2.16 Mrad.

3.2. Darkening of Fibres At our level of precision the relative attenuation of light is negligible for the wavelengths longer than 720 nm. To remove the effects of changes in the fibre geometry due to table motion or the few percent fluctuations in lamp, the spectra were “renormalised” to the part of the spectra in the range 720-780nm. Induced attenuation of the fibre as a function of dose and wavelength is expressed as: A(X1 D)= -(10/L)los[I(X, D ) / I ( X , 0)l

(2)

525 where A(A, D ) is attenuation in dB/m at dose D and wavelength A, L is the length of fibre in meters and I ( X , D ) is transmitted light intensity at dose D and wavelength A. A(X, D ) is usually represented with the following two-parameter function:

A(A, D) = a(A)Da(’) We fit the ratio of spectra to extract the parameters a ( A ) and

(3)

p(X) :

I(A,D ) / I ( X ,0) = e ~ ~ [ - 4 . 3 4 3 L ( ~ ( X ) ( D / D , ) ~ ( ~ ) ]

(4)

For a dose unit D , = 100 Mrad, a ( A ) is attenuation in dB/m at 100 Mrad. Figure 2 shows relative transmitted spectra versus dose for PT qp fibre (no:4). The fit parameters a ( A ) and p(X) are presented in Figures 3 and 4, respectively. The systematic relative error on the ratio I ( X , D ) / I ( X ,0), estimated as f 3 % (see section 3.1), results in an absolute systematic error on the attenuation (and then on the parameter .(A)) of AA = 0.13 at 100 Mrad. Comparing the fits for different periods of irradiation, we have estimated the relative systematic error Ap/p to be 0.03 at 450 nm and 0.08 at 610nm. As observed with the ratios of normalized spectra, the two types of SSU quartz gives different attenuations. The noticeable difference at 450 nm becomes large at the absorbtion bump of 610 nm. Below are the means over the two sets of quartz fibres “G” (1,3-6,8,9) and “R” (2 and 7) “G” < ( ~ ( 4 5 0>= ) 1.52 f 0.02 dB/m < ( ~ ( 6 1 0>= ) 6.08 f 0.04 dB/m “R” < ( ~ ( 4 5 0>= ) 2.11 f 0.04 dB/m < ( ~ ( 6 1 0>= ) 11.89 f 0.04 dB/m 3.3. Mechanical Strength of Fibres The mechanical strength of three types of irradiated fibres was measured. Twenty masurements were carried out for each type of fibre. The results are given in Table 2. There is no significant radiation effect on the tensile modulus. The PT fibres present the highest tensile strength with minumum RMS. 4. Discussion and Conclusions

We performed a set of tests for the mechanical and optical properties of nine types of fused-silica core fibres. We measured in situ the darkening, while irradiated, and the recovery of two fibres after the irradiation. This arrangement allows a direct comparison of the optical properties of two different fibres that is independent of dose, injected light or electron beam fluctuations. There is no significant difference in the optical properties of the polymerclad (qp) fibre drawn by Polymicro Technology and the fluorine doped silica

526

0

I

0.2

"

'

~

Number of electrons (X10'') 0.6 0.8

0.4

'

"

~

"

'

"

"

~

"

'

~

1.2

1

"

'

~

l

Figure 2. Relative transmitted spectra versus dose for fibre (no 4). (a-) at 450 nm and (b) a t 610 nm bandwidth. The curves through the points correspond to the fit with the function 4.

Table2. Fibre tensilestrengthvalues aregiven inGpa. In parentheSis are the minimum and maximum measured values over 20 measurements per sample. Fibre type

No irradiation

30 Mrad

lOOMrad

300 Mrad

2 qq (PT)

4.5f0.6

4.2f 0.6

4.3f 0.6

4.7k 0.4

(2.5,5.1)

(3.2,5.1)

(2.6,5.0)

(3.9,5.2)

4.9 AZ 0.5

5.1 f 0.6

(3.6,5.2)

(3.5,5.5)

3 qq (PT) 5 qq (PT)

1.6 k 0.8

1.7f 0.9

2.2 f 0.8

1.9 f 0.7

(0.5,3.0)

(0.5,2.9)

(0.4,2.9)

(0.4,2.9)

clad fibres drawn by Polymicro or Hesfibel from Heraeus preforms under irradiation up to 100 Mrad. The quality of the core material strongly affects the fibre performance. We measured very different attenuation in fibres drawn, by a single company

527 20

I

I T 1

0

2

0

3

A 4 0

5

L

0

Figure 3. The results of fits with function 4 are shown for the a(X) parameter for the five 300prn core diameter fibres. The a(X) values correspond to the relative attenuation at 100 Mrad.

(Polymicro), where the preforms came from two different suppliers (Heraeus and Russia). The attenuation and recovery as a function of time are well represented by two-parameters fits given by Griscom4. Tensile strength measurements show that Polymicro fibres exhibit the highest tensile strength with minimum RMS. There is no significant radiation effect on the tensile modules, which remains at 4.6 f 0.4. Three bands of luminescence are observed in fused silica at 280nm, 470 nm and 650 nm with lifetimes around 4 ns, 2-10 ms, and 2 0 p , re~pectively~3~. The characteristics of the luminescence, mainly the 470 nm band, in the fibres have to be further measured in order to evaluate fully the effects of the luminescence on the performance of the HF calorimeter.

Acknowledgments We thank Louis Rinolfi, Simon Baird and the LIL operators for the delivery of a high-quality electron beam; Bernard Amacker for the connector and fibre

528

L 0

Figure 4. The same fitting procedure as in Figure 3 results in determination of the parameter for same 300prn core diameter fibres.

p(X)

installation; Andre Braem for providing optical components; Minna Santaoja, Hugo Munoz, and Marc Tavlet for the dosimeter measurements; and Florence Pirotte, Andre Muller, and Guy Roubaud for safety controls. We also thank Thomas Ruf (LEP) and Emmanuel Tsesmelis (CMS test beam) co-ordinators. This work was supported by the US Department of Energy (DE-FG0291ER 40664) and NSF (NSF-INT-98-20258), the Hungarian National Fund (OTKA T026184), and the Scientific Research Council of Turkey, TUBITAK). References 1. THE COMPACT MUON SOLENOID Technical Proposal, 1994, CERN/LHCC 94-38. 2. N. Achurin et al., Nucl. Instrum. Meth. A399, 202 (1997). 3. K. Becker, Solid State Dosimetry C R C Press., 334 (1973). 4. D.L. Griscom et al., Phys. Rev. Lett. 71,1019 (1993). 5. L. Skuja, J. Non-Crystalline Solids. 239, 16 (1998). 6. A.L. Tomashuk et al., IEEE Frans. Nucl. Sci. 47, 693 (2000).

Scintillation Calorimetry Covener: M. Cavalli-Sforxa

M. Cavalli-Sforza

Covener’s Report

A. Henriques

Status of the ATLAS Tile Hadron Calorimeter Production

S. Nemecck

Studies of the ATLAS Tile Hadron Calorimeter Performance

S. Katta

An Overview of CMS Central Hadron Calorimeter

A. Benen

Performance and Calibration of the Forward Plug Calorimeter at ZEUS

A. Attal

Plug Shower Maximum Detector for CDF Run I1

S. Dell’Agnello

CDF I1 Integrated Calorimetry Environment

L. Miramonti

Borexino: A Real Time Liquid Scintillator Detector for Low Energy Solar Neutrino Study

L. Mualem

The MINOS Far Detector Construction and Quality Assurance Testing

S. White

A New Hermetic Electromagnetic Calorimeter Design for Future Collider Experiments

V. Korbel

The Tile HCAL Calorimeter for the TESLA Detector, a Status Report

529

This page intentionally left blank

SCINTILLATION CALORIMETRY

MATTE0 CAVALLI-SFORZA IFAE, Universitat Autonoma de Barcelona, E-08193 Bellaterm (Barcelona), Spain E-mail: caval1iOifae.es (Convener’s Report)

Scintillation Calorimetry remains to-date a popular and cost-effective technique to create large calorimeters a t colliders, stationary-target facilities, and solar neutrino observatories. While sampling calorimeters - usually with WLS fiber readout - represent the greatest majority of the entries, fully active calorimeters for the very low-energy solar neutrinos are also present. The ten contributions that appear in this session indeed represent a broad variety of detectors, spanning the range from the upgrades of the more mature facilities to major calorimeters currently under construction, but also including conceptual studies and prototypes for future machines. To the first category belong the Zeus Forward Plug Calorimeter, presented by Arno Benen, the Plug Shower Maximum Detector for Run I1 of CDF, described by Alon Attal, and the comprehensive report by Simone dell’Agnello on the CDF2 Integrated Calorimetry Environment. The Liquid Scintillator Detector of the Borexino experiment - an unsegmented, fully active detector, was described by Lino Miramonti. The work in progress on an other calorimetric neutrino detector, for MINOS, was reported on by Leon Mualem. The sampling techique was chosen by both general-purpose LHC experiments, ATLAS and CMS, for their massive hadronic calorimeters in the lowto-moderate 77 regions. Ana Henriques and Stanislav Nemecek (standing in for Ilya Korolkov) showed the construction and testing activities of the ATLAS Tile Calorimeter, while Sudhakar Katta covered these aspects for the CMS Central Hadron Calorimeter. Finally, addressing the requirements of hermeticity, high segmentation and energy-flow performance of the next generation of calorimeters for future colliders, Sebastian White showed the design of a hermetic electromagnetic calorimeter, while Volker Korbel reported on the R&D work in progress and test program on a hadronic calorimeter prototype.

531

STATUS OF THE ATLAS TILE HADRONIC CALORIMETER PRODUCTION

A. HENRIQUES CERN CH 1211 Geneva, Switzerland E-mail: ana.henriguesOcern.ch (For the ATLAS Tilecal collaboration)

The status of the construction of the ATLAS TILECAL hadron calorimeter is reported. The various aspects of the construction started at the end of 1998: mechanics, optics, instrumentation, certification and final integration will be presented. At present 80% of the 3 cylinders: 1 barrel and 2 extended barrels is fully instrumented and stored at CERN. Various quality control steps are done during the components production and during the modules instrumentation. An evaluation of the modules uniformity extracted during the final certification using a radioactive 137Cssource is shown. The status of the electronics production and the modules performance extracted during the calibration with particle beams are described in other talks of this conference presented by M.Varanda, F.Martin and S. Nemecek.

1. Calorimeter design

The ATLAS Barrel calorimeter will include a Pb-Liquid Argon (LAr) electromagnetic calorimeter with accordion-shaped electrodes, and a large scintillating Tile hadronic calorimeter, with iron as absorber material and scintillating plates read out by wavelength shifting fibres. The iron to scintillator ratio is 4.67 to 1 in volume. The main function of the Tile Calorimeter is to contribute to the energy reconstruction of the jets produced in the pp interactions with an energy resolution of 5 0 % / 0 3% and, with the addition of the end-cap and forward calorimeters, to provide a good p ~ measurement. ~ i ~ ~ The Tile Calorimeter consists of a cylindrical structure with an inner radius of 2280 mm and an outer radius of 4230 mm. It is subdivided into a 5640 mm long central barrel and two 2910 mm extended barrels as shown in Figure 1 a. The barrel covers the region -1.0< 1771 <1.0, and the extended barrels cover the region 0.8 < 17) < 1.7. An innovative feature of this design is the orientation of the tiles which

+

532

533

Figure 1. a) View of the Tile calorimeter barrel and extended barrel in ATLAS. b) Principle of the Tile hadron calorimeter.

are aligned parallel to the incident particles at 17 = 0 plane and staggered in depth, see Figure 1 b. Fibres running radially collect light from the tiles at both of their open edges and are thus accessible from the side faces of each module. Readout cells are then defined by grouping together a set of fibres into a photomultiplier (PMT). Thus each calorimeter cell is read out by 2 PMTs. The calorimeter is longitudinally segmented into three layers, approximately 1.4, 3.9 and 1.8 interaction lengths thickness at = 0; the transversal segmentation A17 x A 4 = 0.1 x 0.1 in the two first longitudinal samplings and 0.2 x 0.1 in the last radial layer, tail catcher. This calorimeter layout has been successfully proven since 1993 in beam tests of large-size prototypes as part of the RD34 project1. This was followed by the construction and test of full size prototypes (module O’s), between 1996-1998’. The production of the final calorimeter components started in 1998. The modules mechanics assembly and optics instrumentation started in 1998-1999 and should be finished at the end of 2002. The calibration of all the modules with a cesium source will start in summer 2002 following the readout electronics construction. Most of the components are constructed and assembled in the participating Institutes. The overall schedule is driven by the requirement that the tile calorimeter will be ready and calibrated to be lowered in the ATLAS pit in spring 2004 to accomodate for the most recent LHC schedule to start the beam in 2006. 1.1. The mechanics

Each detector cylinder is built of 64 independent modules along the azimuthal direction. The azimuthal design gap of 1.5 mm between modules corresponds to

534

a loss of fiducial coverage of less than 0.5%, substantially smaller than for other scintillator-based hadron calorimeters (e.g. UA1, UA2, CDF and ZEUS). The mechanical structure consists of a stack of a large number of trapezoidal steel plates of 4-5 mm thickness of only a few shapes, which periodically repeat along the z direction. On a first step sub-modules of 30 cm width are assembled. There are at present 93 % of all the sub-modules produced. On a second step the modules are assembled out of sub-modules mounted on a strong-back support girder which will be used to support the read-out fibres and provide the space for read-out electronics (19 sub-modules for a barrel module and 9 sub-modules for an extended barrel module). There are at present 85% of the modules mechanically assembled. An important advantage in this technique is that we may unfold completely the mechanical work from the subsequent optics and electronics instrumentation.

1.2. The optics system

80 tons of injection molded scintillator tiles of 3mm thick (460000 tiles), have been produced in 1998-2001 by an injection molding procedure; Eleven sets of tiles of radial sizes from 97 mm to 187 mm form the longitudinal segmentation of the detector. The injection molding procedure is both efficient and economical: in the production only 1 minute is required to produce a finished tile without additional machining or polishing steps required. The basic plastic component is polystyrene, to which primary and secondary wave-length shifting dopants (PTP 1.5% and POPOP 0.03%) are added. The scintillator tiles are wrapped with Tyvek paper sleeves with black strips printed in the paper to improve the light yield and light uniformity. The peak of ligth emission is at 420 nm and have an attenuation length of 40 cm. The tiles light yield and light transmission were measured during the production for one scintillating tile out of 20, the rms spread was typically 8-10%. As this value was high tiles have been sorted in various groups to keep the rms of the tiles inside the same module to 2-4%. This sorting was determinant to keep the module cells fluctuations (including all the optics components: fibres, tiles, fibre-tile coupling fluctuations) around 5-6%. The measured number of photoelectrons per GeV per cell (2 PMTs) is about 64, which corresponds to an increase of a factor of 2.5 with respect to the very first prototypes tested in the beam in 1994. This is the result of the improvement of the tiles-fibres quality and increase of the PMT quantum efficiency. 1100 km of Y11 (200)s multicladding fibers (650000 fibres) have been pro-

535

duced in 1999-2001 by Kurary. They are used to read out the light produced in the scintillating tiles. These fibers absorb the blue light from the tiles and re-emit in the green (peaking at about 495 nm). The fibers emission has a decay time of about 10 ns; 2.5% of the produced fibres have been measured; The light yield and attenuation length fluctuations are below 3%, well below the specified fluctuations limit of 5%. After the fibre acceptance the fibres have been polished and aluminized by sputtering at the extremity far from the PMT obtaining a reflectivity of about 85%. The light attenuation length measured after aluminisation is 2.9 m. After aluminisation the fibres are inserted in special plastic profiles in an automatized machine that allows a fast insertion of 3 or 4 fibres in the same profile, providing the cell longitudinal segmentation. This is done at a speed of 1 profile/2 minutes. 90% of the profiles are finished, distributed to the instrumentation plants and this activity started in 1999 should finish in summer 2002.

Figure 2. a) One barrel and b) one Extended barrel module instrumented with all the optics components. A mock-up with a fibre routing model is used in front of the barrel module for guidance.

2. The modules instrumentation

After the steel mechanics assembly scintillating tiles and plastic profiles are inserted into the modules. The fibers extremity corresponding to each readout cell are bundled, and routed to each individual hole situated in the girder. The fibre bundles are then glued and polished. An instrumented barrel and extended barrel module can be seen in Figure 2. The whole operation takes on average 2 weeks per module. This activity started in 1999 and is planned

536 65.00 60.00 55.00 50.00 45.00

40.00 35.00 30.00 25.00

20.00

15.00 10.00

5.00 0.00

AWQ

Mar-00

Oct-OO

AprM

NOMl

Mav-crr

DeC-oe

Date Figure 3.

The production curve of the Tile barrel and extended barrel modules.

to finish at the end of 2002 a s seen in Figure 3. 3. Modules certification

In order to verify that after the instrumentation of each module there are no broken fibres, bad fibre-tile couplings or fibres being routed to the wrong cell, the modules are certified with a moveable 5 mCi 137Cs source. This is the same source that will be used later to do the cell calibration of all the modules in the surface and in the ATLAS pit. Since the production of the final readout electronics drawers will only start in summer 2002 a test electronics readout system constructed for the module 0 tests is used for the certification. Each fibre bundle is air coupled to a PMMA light mixer, which in turn is air coupled to a PMT tube. The air gaps are 1 mm wide; The source is a 5 mm long source embedded into a thin stainless steel tube (outer diameter of 0.7 mm) which can be inserted into the holes that exist in the scintillating tiles and steel along the t direction (see Figure 4 a). Figure 4 b shows a typical current measurement along the z dimension of one cell with an evident multiple peak/valley structure; the local peaks correspond to the passage of the source across each individual scintillator tile. The suppressed peak observed is due to a bad fiber tile coupling or damaged

537

Figure 4. a) Concept of the 137Cssource system used during the modules certification and calibration. b) Current measured in a P M T as a function of the source position along the Z axis.

fibre. During the certification of each module typically 1% of the fibres need to be replaced as they have a light reduction bigger than 25% with respect to the cell signal average. After bad fibres replacement, the modules cell RMS is typically 5-6% well below the 10%upper limit set by the ATLAS requirements. There are at present 80% of the modules fully instrumented and stored at CERN with the last arrival expected at the end of 2002. In summer 2002 the modules will start to be calibrated with the cesium source system following the electronics mass production. 12% of the modules are calibrated in the SPS beam, an activity that should finish in 2004. Another ambitious step to start in summer 2002 will be the preassembly in the surface of the three barrel and Extended barrel cylinders before starting to transport the individual modules in the ATLAS pit in spring 2004. References 1. F. Ariztizabal et al. (RD34 collaboration), Nucl. Instr. and Meth. A349, 384

(1994)

and

other

references in the Web DETECTORS/TILE/tileref/tileref.html. 2. ATLAS Tile Calorimeter Technical Design Report, CERN/LHCC/96-41. http : //atlasinfo.cern.ch/Atlas/SUB-

site:

STUDIES OF THE ATLAS TILE HADRON CALORIMETER PERFORMANCE

ILYA KOROLKOV IFAE, Universitat Autonoma de Barcelona, E-08193 Bellaterra (Barcelona), Spain E-mail: [email protected] presented by

STANISLAV NEMECEK Institute of Physics of ASCR, Praha, Czech Republic

Extensive studies of the ATLAS Tile hadron calorimeter, TileCal, performance were conducted between 1996 and 1998 on full size prototypes. The construction of the calorimeter started in late 1998. Research activities since then focussed on controlling the energy resolution and sensitivity to muons. Most of the activities take place at the SPS H8 beam line at CERN. The status of these studies is presented, with comparisons to results on prototypes. An update on prospects of using the calorimeter signal in muon triggers in ATLAS is reported. The calibration program of TileCal modules with test beams started in 2001. A brief overview of the calibration and monitoring program of the calorimeter is given.

1. Detector

ATLAS (together with CMS) is one of two general purpose detectors utilizing the LHC. There is an abundance of exciting physics involving jets and missing energy at the LHC1. The main tasks of the hadronic calorimeter are jet and missing transverse energy reconstruction, making sure that the jet energy is contained, and protecting the muon system from excessive backgrounds. Some sensitivity to muons is used in the first level trigger. ATLAS has adopted the sampling calorimetry concept in different configurations. The TileCal’, the barrel part of the hadron calorimeter in ATLAS, is a sampling device made of steel and scintillating tiles. It is designed to optimize jet response. The detector is composed of one barrel and two extended barrel cylinders. Each cylinder is built of 64 independent modules aligned along the azimuthal direction. Between the barrel and the extended barrels there is a gap of about 600 mm, which is needed for the services. The barrel covers the region lql < 1.0, and the extended barrels cover the region 0.8 < 1171 < 1.7. Part of the gap is

538

539

covered by a calorimeter plug in the region of 0.8 < (171 < 1.0. Each A 4 = 0.1 azimuthal slice has 73 cells arranged into projective towers with A17 of 0.1. The TileCal segmentation allows an efficient hadron leakage cut, needed for electron and photon identification. To avoid low energy tails in the reconstructed jets energy, the depth of the active calorimetry is at least nine interaction lengths. Each of the modules has three depth segments called samples A, B, and D, which allows, with appropriate weighting technique, to restore the linearity of the energy response to hadron showers at the level of (l-2)%'. The discovery potential of LHC is directly connected to the center of mass collision energy (14 TeV) it will be able to reach4. To explore this potential the calorimeter response must be understood to the highest jet energies, which may reach 3-4 TeV. With the measured photo-tube yield of the TileCal, 63 to 70 phe/GeV5, and the nominal photo-tube gains a response of 1.2 pC/GeV of EM energy at 90" was measured in 2001. This corresponds to a dynamic range of 1.5 TeV/cell, with a non-linearity of PMTs at the high end of about 2%6. Online laser calibrations are foreseen to monitor non-linearities of photo-tubes. The dynamic range may be extended above 1.5 TeV/cell using non-saturated energy samples7. The technique is under study. With the designed LHC luminosity of cm-2sec-1, events are to be resolved over a background of 23 minimum bias (MB) events per bunch crossing every 25 nsec. An optimal filtering algorithm (OFA) is to be applied on-line to reduce the effect of the MB events on the data. Applying OFA to the 1998 TB data, a 30% reduction in noise was obtained8. Data with 25 nsec beam structure were taken during 2001 TB. The observed dispersion in timing everywhere in the module was found to be better than 1 nsec which includes the beam spread. For ATLAS the timing is expected to be about 0.5 nsec, compatible with the bunch size. To use the MB events as a performance monitor each readout channel in the TileCal has a dedicated chain able to integrate MB signals with the integration time of 0.5-10 msec. Tiles and fibers are expected to lose less then 5% of the light yield after being exposed to 400 Gy (10 years of ATLAS)g. Natural ageing of optics is estimated to be at the same level. This loss will be monitored with dedicated Cs scans during ATLAS shut-downs and with MB currents during the data taking.

2. Studies of the TileCal Performance

Starting from 1993 several generations of TileCal prototypes went through a series of beam tests". In 1994 the LAR and the TileCal calorimeter prototypes were first tested in a combined mode. An azimuthal sector of the ATLAS barrel

540

calorimeter was reproduced by placing the EM prototype inside a cryostat with the hadronic TileCal prototypes located downstream. The energy resolution of pions in the energy range from 20 to 300 GeV at an incident angle 0 of about 11” is well-described by the expression

a/E = [(46.5 f 6 . 0 ) % / a + (1.2 f 0.3)%]@ (3.2 f 0.4) GeV/E,

(1)

where E is in GeV and the symbol @ indicates a sum in quadrature. The above result was obtained using a weighting technique. Shower profiles, shower leakage, and the angular resolution of hadronic showers were also studiedll . A second combined test was done in 1996. The results obtained in 1994 were confirmed and improved12. An analysis of the 1996 pion data taken in the ‘benchmark’ framework demonstrated that the fractional energy resolution of the combined calorimeters is

a/E = [(60 f 3 ) % / a + (1.8 f 0.2)%]@ (2.0 f 0.1) GeV/E.

(2)

In the ‘benchmark’ algorithm, a two step procedure is adopted to reconstruct the nominal beam energy: first, the energy of the particle is obtained as the sum of several terms, using only a few parameters which are optimized by minimizing the fractional energy resolution a/Eo. This first-pass energy EOis re-scaled to the nominal beam energy in a second step. A more satisfactory approach is to use a weighting technique, based on a concept developed by the H1 callaboration, which relies on the fact that EM showers are much denser than hadronic ones. Hence, high energy density cells are assumed to contain mainly EM signals and are given small corrections, while low energy density cells are assumed to have hadronic signals and are given correspondingly larger corrections. The fractional energy resolution obtained using this weighting technique is

a/E = [(42 f 2 ) % / a + (1.8 f 0.1)%]@ (1.8 f 0.1) GeV/E.

(3)

It can be concluded that the cell weighting technique improves the energy resolution obtained with a simpler approach. The linearity of the calorimeter response to pions was studied for both the 1994 and 1996 data sets using the weighting technique. The response to pions is well within f 2 % over the full energy range of the data. An analysis of the 1996 combined T B data performed in the framework of a nonparametric method has shown that the reconstructed mean values of the hadron energies are within 1% of the true values and the fractional energy resolution is a/E = [(56 f 3)%/&

+ (2.7 f 0.3)%]@ (1.8 f 0.2) GeV/E.

(4)

The nonparametric method utilizes only the known e / h ratios and the electron calibration constants and does not require the determination of any parameters

541

by a minimization technique. Thus, this method lends itself to be used in a first level trigger. A stand-alone beam test of the full size TileCal prototype module, denoted as MO, was carried out, and the ratio ( e / h ) T i l , = 1.30 f 0.03 was measured13. The stand-alone tests of the MO prototype continued through the years 19971999. The first stand-alone test of the production TileCal module was performed in 2000 and the calibration program started in 2001. The purpose of the stand-alone tests was to verify the effectiveness of various techniques meant to improve different aspects of the calorimeter. Close attention was paid to the level of the electronics noise, the uniformity of the production modules instrumentation, and the quality of the calibration scans. The inter-calibration and gain equalization for the TileCal cells is done by a dedicated system. It consists of a capsule with a radioactive 137Cs source movable longitudinally through the iron and scintillating tiles, orthogonal to the tile planes. The induced PMT current, referred to as the Cs response, is measured by an integrating ADC. For reliable inter-calibration of the TileCal modules it is crucial to insure a tight correlation between the Cs response and the calibration constants measured with T B particles. The Cs system allows to measure the non-uniformity of module response, which was 8% for MO prototype and was reduced to 5% for the production modules. Based on the analysis of the T B data collected in 2000-2001, an RMS of (2-3)% for the Cs and muon signals ratio is obtained14. In 1996 the same ratio was 5%. A non-direct confirmation that the TileCal hadronic energy resolution is improving over the years comes from the electron data analyses, in which both electron energy resolution and the uniformity of the signal across the module are improving15. The TileCal detector contributes to the first level trigger with the fast analog signal coming from the trigger summing boards. The TileCal trigger towers are built by adding linearly several signals from the three sampling layers, usually, two from the sample A, two from the sample B and one from the last layer, D. The adders provide two types of analog signal to the first level trigger system. The first one is a trigger tower sum with the granularity Aq x Aq5 = 0.1 x 0.1. Together with the tower sums from the electromagnetic calorimeter it allows t o perform the online event selection based on the transverse energy threshold for single or multiple jets and/or missing transverse energy16. The signal from the third TileCal depth sample, D,are used by the muon trigger, to reduce the rate of fake triggers17 due to low energy particles background. The analog signal from the adder muon output was extensively studied during the 2001 TB. With the best settings, an 86% efficiency of the muon signal can be achieved with only 10% fraction of the noise even at low pseudo-rapidities (17 = 0.2, where the muon signal in the D sample is smallest).

542

Further improvements are foreseen in 2002 by additional noise reduction in the readout chain and an increase of the photo-tube gains by 20% in the sample D of the barrel". 3. Calibration and Monitoring in the TileCal

A very precise calibration of the ATLAS calorimeters is essential for minimizing the energy resolution and setting the right energy scale. The calorimeters in ATLAS are calibrated at the so-called electromagnetic (EM) scale, i.e. with muons or electrons, which in the absence of further corrections would lead to systematic underestimates of the jet energies and deterioration of jet energy resolution. The first correction step implies inter-calibration of various cells and parts of the calorimeter followed by an appropriate weighting technique. The studies showed that weighting algorithms may both correct the mean response and improve the resolution. Final in-situ jet calibrations are necessary to correct for remaining jet non-linearities due to longitudinal leakage, energy lost in the dead material, noise from MB events and electronics, magnetic field effects, and finite granularity of the detector'. The above methods will use limited statistics of some channels and many parameters which represent the whole ATLAS calorimeter. The intention is to achieve 1%accuracy in the measurement of the absolute jet energy and to recover linearity to l%level. The TB calibration should provide accurate measurements of the input parameters to the fit. A preliminary step in the TileCal calibration program is to measure the response of 12% out of all 192 modules that constitute the TileCal. This part of the program will take place over the next three years. Most of the actual measurements take place at the SPS H8 beam line at CERN where the TileCal modules are tested in the beam of the high energy electrons, pions, and muons. The responses of the module to the particles are referred to as calibration constants. To propagate the calibration constants to the remaining 82% of the modules measurements done by the Cs system are used. All the TileCal modules will undergo Cs scans after been assembled in ATLAS. The RMS of (2-3)% for the Cs and muon signals ratio demonstrates the accuracy of the method when applied to muonsI4. From the installation of TileCal detector in the experimental pit to physics start-up the Cs source will be used to tune and verify the TileCal cell intercalibration. Similar scans to monitor (and recalibrate if needed) optics inside of the cells, taking into account extremely detailed effects from individual tiles, fibers and couplings, are foreseen during ATLAS shut-downs. MB currents will be used to monitor cell optics during data taking. MC studies have shown that depending on the cell position 5 to 500 measurements are enough to reach 1% level of accuracy in measuring the MB currentslg. The laser system

543

will monitor the time behavior of the PMT block optics and PMT gains. The charge injection system will monitor the behavior of each read-out channel over its full dynamic range utilizing the empty bunch crossings. To monitor aging effects one production barrel, two extended modules, and module0 prototype will be kept at the TB area after TileCal commissioning. 4. Summary

The detector geometry and principles of TileCal were optimized for the best hadronic performance in combined operation with other ATLAS calorimeters. An extensive check of the stand-alone and combined performance was carried out on a set of prototypes from 1993 to 1999 mostly at the H8 test beam at CERN. The latest full size prototypes, called MOs, complied with all the requirements set for TileCal. The research efforts since then concentrated on calibration of the production modules which started in 2001. From all the data collected during TBs 2000-2001 we conclude that the performance of the production modules is the same or better than the performance of MO full size prototypes. Special efforts are applied toward future integration of the TileCal in ATLAS such as transportation of the calibration constants to the ATLAS environment and developing several monitoring systems. References 1. M. Bosman, ATLAS Collab., CMS Collab., In: 9th International Conference on

Calorimetry in High Energy Physics, Annecy, France, 533-543 (2000). 2. ATLAS Collab.: ATLAS TDR 3, CERN, (1996). 3. M. Cobal et al., ATL-TILECAL-98-168, (1998). 4. G. Azuelos et al., SN-ATLAS-2002-018, CERN , (2001). 5. S. Nemecek, TileCal internal presentation, (2002). 6. R. Lefevre et al., ATL-TILECAL-2000-018, (2000). 7. Ch. Roda, TileCal internal presentation, (2002). 8. E. Fullana, TileCal internal presentation, (2002). 9. A. Karyukhin et al., ATL-TILECAL-94-025, (ATL-L-PN-25), (1994). 10. F. Ariztizabal et al., RD34 Collab., Nucl. Instr. and Meth. A349 384 (1994) 11. Z. Ajaltouni et al., ATLAS Collab., Nuc. Instr. and Meth. A387,333 (1997). 12. S. Akhmadaliev et al., ATLAS Collab., Nuc. Instr. and Meth. A449,461 (2000). 13. J. Budagov et al., ATL-TILECAL-96-72, (1996) 14. 0. Norniella, TileCal internal presentation, (2002). 15. Y . Kulchitsky, TileCal internal presentations, (1996-2002). 16. ATLAS DAQ Collab.: Level-1 Trigger TDR, CERN, (1998). 17. A. Nissati, ATL-DAQ-98-123, CERN, (1998). 18. A. Cerqueira et al., ATL-TILECAL-2002-002, (2002). 19. X. Portell, TileCal internal presentation, (2002).

AN OVERVIEW OF CMS CENTRAL HADRON CALORIMETER

SUDHAKAR KATTA Tata Institute of Fundamental Research, Mumbai-400005, India (On behalf of CMS

- HCAL

Collaboration)

The central Hadron Calorimeter for CMS Detector is a sampling calorimeter with active medium as scintillator plates interleaved with brass absorber plates. It covers the central pseudorapidity region (1171 < 3.0). The design and construction aspects are reported. The status of construction and assembly of various sub-detectors of HCAL are presented.

1. Introduction The CMS (Compact Muon Solenoid) detector' is one of the two general purpose detectors being built at Large Hadron Collider (LHC) at CERN. The hadron calorimeter for the CMS detector will be used for measuring the energies and directions of particle jets and the missing transverse energy flow. The determination of missing energy flow is crucial in searches for new particles and phenomena. Adequate granularity, resolution and shower containment are essential in attaining these goals. The hadron calorimeter (HCAL) consists of the central barrel (HB), the two endcaps (HE), the outer barrel (HO), and the forward calorimeters (HF). Since HF is covered by another talk at this conference, we restrict here only to the central hadron calorimeter. 2. Central CMS Hadron Calorimeter

The central Hadron c a l ~ r i m e t e r consists ~ ~ ~ , of barrel calorimeter (HB) covering pseudorapidity region 1171 < 1.3, the endcaps (HE) covering the region 1.3 < 171 < 3.0 and an outer barrel (HO) covering the region 171 < 1.2. All these subdetectors have similar principle of having scintillation tiles with fiber readout as active medium and brass as absorber and form a sampling calorimeter. The 7 - 4 segmentation of HB, HE, HO is 0.087 x 0.087, except near 1171 = 3.0 where the size of segmentation is doubled. HB and HE are inside the 4-Tesla solenoid coil. Since HB is only 6.5 interaction lengths thick, the Outer Calorimeter (HO)

544

545

is placed outside the solenoid coil just below the muon barrel system and forms an additional sampling layer. HO utilizes the solenoid magnet coil and muon absorber plate as additional absorber. HO improves the shower containment as HB is not thick enough to contain the showers particularly those which developed deep inside HB. Monte Carlo studies have shown that HO improves energy resolution and useful in studies using missing ET information. Figure 1 shows the longitudinal view of CMS detector and location of various hadron calorimeter sub-detectors.

Figure 1. Longitudinal view of CMS detector.

The longitudinal segmentation for HB and HO is one unit while for HE it varies from one to three. To compensate the radiation damage at 1771 > 2.0, HE has extra longitudinal segments to allow correction for signal loss. Various sections of the calorimeter have been tested using pion, proton, electron and muon beams. In all these sub-detectors, scintillators are in tile form and readout using Y11 double clad wave length shifting (WLS) fibers from Kuraray company, embedded in the key hole grooves in the tile which are of c shape. In case of HB and HE, each tile has one (T groove but in case of HO there are 4 c grooves per tile. To decrease the attenuation loss of light in transmission, the WLS fibers are spliced to clear fibers outside the groove and brought to the connector placed at the edge of the megatile (tray). Besides the active and passive elements, the central hadron calorimeter has optical readout system using 19 pixel Hybrid Photodiodes (HPD), 12 KV high voltage system,

546

low voltage system, specially designed ADC (QIE), calibration systems using moving radioactive wire source, laser and LED and special purpose electronics. Details can be found in previous conference proceedings4y5. 3. Design aspects

HB consists of two half barrels and each barrel is divided into 18 wedges of 20" each in 4. Each wedge has 4 megatiles each covering 5" in 4. Length of a megatile varies from 3.7 m to 4.3 m. There are 17 layers of megatiles in a wedge. Each megatile has many tiles each of which covers 0.087 x 0.087 in r] - 4. All the tiles with same r] - 4 from the 17 layers are summed up for the signal and sent to HPD for readout. The brass absorber wedges are fabricated by the Felguera company in Spain. The megatiles are made at Fermilab, USA and brought to CERN. The megatiles are inserted at CERN into these wedges. HE consists of two endcaps each having 18 sectors. Each sector covers 20" in 4. The brass absorber is fabricated at the factory MZOR, Minsk in Belarus. The megatiles are made at Protvino in Russia and sent to CERN. The megatiles are inserted at CERN into the absorber wedges. There are 19 sampling gaps filled with megatiles. Each sector has two megatiles and they are logically divided into four 5" regions in 4. HO is geometrically located just below the muon rings. A muon ring has 12 sectors each covering 30" in 4. In each sector there are 6 trays with each tray covering 5" in 4. There are 5 muon rings- -2, - 1 , O , +1,+2. Ring 0 has an additional layer of HO trays below the tail catcher iron. The scintillator used is 10 mm thick Bicron (BC408) and the fibers used are Y11 double clad Kuraray fibers. To increase the light output as the readout is only from one or two tiles, there are 4 (T grooves per tile. The WLS fibers are spliced to clear fibers and are brought to the two connectors situated at the edge of the tray. In case of Ring 0 trays, the signals from the two layers are summed up. The trays for HO are fabricated and assembled in India. They will be shipped to CERN and be kept ready for installation. 4. Status of various sub-detectors:

All the wedges are completed at Felguera, Spain. All the megatiles are produced and brought to CERN. The wedges for HB- are stuffed with megatiles and one half barrel is assembled and installed at SX5 hall in Sept 2001. Figure 2 shows the picture of already assembled half barrel HB- at SX5. The wedges for the other half barrel HB+ are being stuffed with megatiles and will be installed at SX5 in Sept 2002. The readout boxes are in production at Mississippi and should be complete by April 2002. The ODUs (Optical Decoder

547

Unit) for the readout boxes are completed at Notre Dame.

Figure 2. Assembled half barrel HB- at SX5.

The absorber wedge HE- is already at CERN and the wedge for HE+ will be delivered in Oct 2002. The megatiles production is completed already. The assembly and installation of H E will be done by end of 2002. The readout box is in design and should be complete by April 2002. Production of readout

548

boxes will start in late Spring 2002. ODUs will be built in the summer of 2002. The HO trays are in production and will be shipped in batches to CERN. They will be installed at SX5 in Oct-Dec 2003. Design of Readout box will be finalized soon and production to start in summer 2002. ODUs will be made in Fall of 2002. 5. Conclusions

CMS hadron calorimeter is progressing well. One half barrel HB- is already assembled at SX5 in Sept 2001 and the other half barrel HB+ will be installed in Sept 2002. HE is also progressing well and HE- will be installed by end of 2002. HO has gone into production stage. Front-end electronics and higher level electronics are also progressing well and are going to be tested in the summer of 2002 in the testbeam run at CERN. References 1. The Compact Muon Solenoid(CMS) Technical Proposal, CERN/LHCC 94-38, December 15, 1994. 2. The CMS Hadron Calorimeter Technical Design Report, CERN/LHCC 97-31, June 20, 1997. 3. Outer Hadron Calorimeter Engineering Design Review Note, TIFR/CMS-9901, June 8, 1999. 4. V. Hagopian in Calorimetry in High Energy Physics, ed. G.Barriera, B.Tome, A.Gomez, A.Maio, M.J.Varanda (World Scientific, Singapore, 2000). 5. V. Hagopian, Conf report CMS CR 2002/001.

PERFORMANCE AND CALIBRATION OF THE FORWARD PLUG CALORIMETER AT ZEUS

A. BENEN Universitat h i b u r g , Fakultat f i r Physik Hermann-Herder-Str.3, 79104 h i b u r g , Germany E-mail: [email protected] The ZEUS Forward Plug Calorimeter performed with an energy resolution of U E / E= (0.41f0.02)/[email protected] for electrons and U E / E= ( 0 . 6 5 ~ 0 . 0 2 ) / ~ @ 0 . 0 6 for pions. An independent calibration using beam halo muons confirms the results obtained from test beam measurement and “Co-scans within an uncertainty of 8.2%.

1. Introduction The Forward Plug Calorimeter (FPC) was installed in the ZEUS detector in order to extend the calorimetric coverage in pseudorapidity from 7 5 4.0 to q _< 5.0. This increases the physical potential in deep inelastic electron-protonscattering (DIS). For diffractive events the mass range of the dissociated system is enlarged by a factor of about two. For forward-jets measurements the FPC allows a region to be probed which may distinguish between different parton evolution models, e.g. DGLAP or BFKL. 2. Structure and Properties The FPC consists of a lead-scintillator sandwich calorimeter with wave length shifter (WLS) fibers and photomultiplier (PMT) read out. It was installed in the 20 x 20 cm2 beam hole of the forward uranium-scintillator calorimeter (FCAL) of the ZEUS detector at HERA’. The beampipe with its 63 mm diameter is completely surrounded by the FPC which has a depth of 108 cm. In the FPC, lead plates of 15 mm thickness alternate with scintillator layers of 2.6 mm. The WLS fibers have 1.2 mm diameter and pass through 1.4 mm diameter holes in the lead and scintillator layers. The effective lead to plastic ratio by volume is 5.2:1, taking into account the WLS fibers and the fiber holes in the lead plates. The FPC provides an equal response to electrons and hadrons (compensating calorimeter, e / h = 1) which is based on the results from a lead-

549

550

2.4

cm

co * . ... . .. . .... . . . . .. . .. ,

,

*

. . .. . . . , . . . . . . . . . . . .

19.2 cm

Figure 1. Structure of the FPC (looking in proton direction)

scintillatorgv calorimeter of similar composition2. It has approximately the same radiation length Xo and nuclear absorption length X as the FCAL: Xo(FPC) = 0.68 cm and Xo(FCAL) = 0.74 cm, X(FPC) = 20 cm and X(FCAL) = 21.0 cm. The FPC is subdivided longitudinally into an electromagnetic (EMC) and a hadronic (HAC) section which are read out separately. The electromagnetic section consists of 10 layers of lead and scintillator corresponding to 26.5 XO and 0.9 A. The hadronic section of the FPC consists of 50 layers and represents 4.5 X leading to a total for the FPC of 5.4 A. The scintillator layers consist of tiles and form cells which are read out individually. The cell cross sections are 24 x 24 mm2 in the EMC, commensurate with the Moliere radius, and 48 x 48 mm2 in the HAC section. The polystyrene based scintillator SCSN81T2 from Kuraray shows good light yield and radiation stability3. The PMTs, Hamamatsu R5600U, are relatively insensitive to magnetic fields. All 4 (16) fibers corresponding to an EMC (HAC) cell are connected to the same PMT through a light-mixer bar. The total number of readout channels is (EMC HAC): 60 + 16 = 76.

+

551

Figure 2.

Sideview of the FPC.

3. Performance and calibration The FPC was tested in the X5 beam of the CERN SPS with 10-100 GeV electrons and pions as well as 100 GeV muons. It was installed between modules of the FCAL prototype, providing similar conditions to the ZEUS-detector set-up.

3.1. Performance

For the study of the energy resolution, electrons incident uniformly on the area of a cell (24 x 24 mm2) have been selected where the total signal has been obtained by summing the signals from the cluster of 3 x 3 EMC cells centered on the cell containing the point of incidence. In Figure 3 the ratio RMS to mean of the pulse height distribution is plotted versus the beam energy. Restricting to the 8 x 8 mm2 region on the center of the central cell, a substantial improvement is obtained in the resolution which contains the contribution of non-uniformities. The pions provided by the test beam showed a muon contamination which has been substantially rejected from the data sample using a muon-veto counter situated behind the FPC. In Figure 4 the reconstructed beam energy dependance and the energy resolution are shown where each point is the result of gaussian fit to the corresponding pion with different energies. The resolution is affected by transverse leakage. When installed in ZEUS the FPC was completely surrounded by FCAL modules. As a result there is no transverse leakage except into the beam hole and, according to MC, the energy resolution in ZEUS improved by 20%.

552 IC- beam

+.e/

9

*m rn

e- beam

.

w 0.3 c?

0

0.25

02

+

24 x 24 mm2 (RMS/MEAN nE / E 0.34 / dE % 0.07

i

0 8 x 8 mm2 (Gauss flt) uE I E = 0.32 / dE @ 0.03

!

8 FPC prototype

(DESY test beam)

0.15

FPC E , = 0.75 A FPC+FCAL rolotype E,/ ,E I c\ FPC + FCAL (MC) Em, / Ebem = 1.01

$60

0.8

al 40 ri 220

0

2 0 4 0 6 0 8 0 1 0 0

beam energy [GeV]

. w

0.4

'

0.3

H FPC oE/ E = 0.88 / JE @ 0.10 A FPCrFCAL oE/ E = 0.65

w

/%'%&

A

0

0.1 0.2

0.05

0

.

0.1

20

40

60 80 100 beam energy [GeV]

Figure 3. Electron energy resolution obtained from a gaussian fit to the signals measured with a 3 x 3 EMC cell cluster and from the RMS of the signal distributions. The beam electrons are distributed either over a 24 x 24 mm2 square (cell size) or over a 8 x 8 mm2 square. The curves show fits to the CERN data. For comparison also data from the prototype measured at DESY are also shown.

FPC + FCAL (MC) nEI E I 0 . ~ /9:'E w 0.06

0

.

2 0 4 0 6 0 8 0 1 0 0

beam energy [GeV] Figure 4. The energy measured for T - in the HAC section of FPC, in the FPC and FCAL prototype modules. in the FPC The lower plot shows the energy resolution (uE/E) for T - as measured by FPC alone FCAL prototype as a function and FPC of the T - beam energy.

+

+

3.2. Calibration For the analysis of the CERN-test data the FPC has been simulated by Monte Carlo (MC) using the GEANT 3.21 package4. It has been implemented as a sampling calorimeter consisting of scintillator and absorber layers where light attenuation in the WLS fibers, photostatistics, noise and cross talk were taken into account. The accuracy of the simulated response to electrons and muons has been estimated to be about 3%. For pions the accuracy of the simulated energy signals, which have been determined with the hadronic package GHEISHA' , is about *lo%. The calibration of the FPC was done differently for the EMC and HAC section as follows.

553

Test beam electrons of 60 GeV distributed over the complete FPC surface were used for the calibration of the EMC cells. The calibration constants have been adjusted such that the mean of the summed signal of all EMC cells is independent of the point of incidence. The absolute calibration were adjusted to the beam energy. Muons have been used for the calibration of the HAC cells since the energy of pions,.is not fully contained in the FPC. A Landau function convoluted with a gaussian function, to take into account photostatistics and electronic noise, has been fitted to the muon signals. The calibration constants have been computed in order to adjust the peak value obtained from the fit, to the value predicted by MC. Since the distribution of the muon signals is not a gaussian shape, it is crucial that it is well described by MC. A good overall agreement is observed for the HAC section after calibrating with muons, Fig. 5 . A monitor system using a 6oCo source allows the detection of changes in the performance of the scintillator tiles and the WLS fibers as well as drifts in the gain of the PMTs. By measuring the ratio of response to 6oCo and beam particles the absolute and cell-to-cell calibration constants can be transported from the test beam to ZEUS and the stability of the calibration can be monitored. In order to measure hadronic energies, the following two corrections have been applied. The ratio between the visible and the total energy deposited in the FPC (sampling fraction) is different for an incident pion and an incident muon. Assuming compensation the sampling fraction obtained with electrons was used for pions. The effect of the light attenuation in the WLS fibers is different for muons and pions. Therefore, the energy scale of the HAC section has been raised by a factor of 1.33. The accuracy of the absolute calibration thus obtained is estimated to be 5%. The error is dominated by the uncertainty of the MC prediction. 4. Beam halo muons

Due to inelastic interactions of the beam protons with beam gas or halo with components of HERA collider, charged pions are produced which mainly decay into muons. These beam halo muons have trajectories mostly parallel to the beam axis and traverse the detector parallel to the direction of the proton beam. Using these events, an independent cross-check on the calibration of the FPC HAC cells is possible. In order to reduce electron related background, special runs with only protons in the HERA collider were performed. The trigger condition was a coin-

554

gG O W0055 P

9 0.M5 0 01

100 GeV muons

5 300

FPC-HAC

0 075 0 0.5

Om 0 0.I

5 250

0.075

data

00.5 OM5

0 01

0 075 0 05 OD2

0 0

2

4 0

2

4 0

2

4 0

2

4

energy [GeV]

Figure 5. The reconstructed energy measured in the HAC sections of the FPC for 100 GeV muons is shown for data and MC.

Figure 6. Energy distribution wihtin the FPC HAC of beam halo muons after applying the isolation cut explained in the text.

cidence of a hit in the FPC-presampler tiles, a FPC-EMC cell cluster and a signal in the proton remanant tagger. In the Monte Carlo the full ZEUS detector set-up was used. The muon momentum spectrum was taken from previously measured beam halo muons in the uranium calorimeter6. For the final event selection a muon isolation cut was introduced. Only events with single HAC cell hits, whose adjacent cells measure less than three times of their particular noise threshold and whose coincide with a muon type signal in their corresponding EMC cells, were taken. This cleaning cut allowed the beam related hadronic background to be significantly reduced. The mean values were determined using a gaussian fit in the peak region of the distribution, Fig. 6. Assuming higher hadronic background activity in the cells closer to the beampipe, the mean peak values in the outer cells for data and MC deviate by only 5%, Tab. 1. Nevertheless, a slight asymmetry in the outer cells, likely due to small mis-alignment of the FPC, can be observed. Taking into account that the muons are not distributed uniformly in the

555 Table 1. Energy mean peak values in GeV corresponding to FPC HAC cells

data outer ring

1.779

0.068

data all MCIdata

1.722 1.082

0.124 0.073

detector and varying the muon momentum spectrum taken for the MC an overall systematic uncertainty of 8% was estimated. In the data, the mean peak values still show a reasonable agreement within a combined uncertainty of 8.2%although the simulation tendd to overestimate the measurements, Fig. 7.

2.5

n

>

8

2.25

h

2

3

1.5

Y

!8? 1.75 i%

1.25 1

0.75

0.5 0.25 0 1

2

3

4

5

6

7

8

910111213141516

No. of HAC cell Figure 7.

Mean peak energies for outer and inner cells, the band gives the total uncertainty.

556

5. Conclusions The FPC shows an energy resolution of U E / E= (0.41 f 0.02)/@ 69 0.062 f 0.002 for electrons and U E / E= (0.65 f O.OZ)/@ @ 0.06 f 0.01% for pions. The calibration of the FPC HAC cells with 100 GeV muons and its monitoring with 6oCoscans can be confirmed within an uncertainty of 8.2% using beam halo muons.

Acknowledgments I would like to thank all members of the ZEUS-FPC group for the hard work related to the Forward Plug Calorimeter. It is also a pleasure to thank the organizers of CALOR2002 for providing a very interesting and enjoyable conference in Pasadena. References 1. The ZEUS detector, DESY (1993). 2. ZEUS. Collaboration, E. Bernardi et al., Nucl. Instr. and Meth. A262 (1987) 229. 3. T. Hasegawa et al., Nucl. Instr. and Meth. A311 (1992) 498. 4. CERN Application Software Group, G E A N T 3.21 Detector Description and Simulation Tool, CERN Program Library Long Writeup W5013 (1993) 5. H.C. Fesefeldt, Simulation of Hadronic Showers, PITHA 85-02, RWTH Aachen (1985) 6. A. Freidhof, A. Fiirtjes, Muons from the Proton Halo: A Precisison test to monitor the F/RCAL Calibration, ZEUS-Note 93-076 (1993)

PLUG SHOWER MAXIMUM DETECTOR FOR CDF RUN I1

A. ATTAL University of California, Los Angeles, Los Angeles C A 90024, USA E-mail: [email protected]

A new forward/backward endplug calorimeter, covering the eta range between 1.1 and 3.5, has been installed for CDF Run I1 at the Fermilab Tevatron. One of the key components is a scintillating strip/wavelength shifting fiber detector located at the electromagnetic shower maximum. Its purpose is to measure the position of electromagnetic particles and to help discern early showering hadronic particles from electromagnetic decays. I will discuss the design, functionality and calibration of the Shower Maximum Detector and present performance results from the beginning of Run 11.

1. Introduction

The Collider Detector at Fermilab (CDF) has gone through extensive upgrades to deal with higher bunch crossing rates and luminosity. Among the changes is an entirely new endplug calorimeter made up of an electromagnetic (EM) and a hadronic section. The EM component is a non-compensating, sampling calorimeter composed of plastic scintllator tiles, sandwiched between layers of lead and read out by wavlength shifting (WLS) fibers. One of these layers is located a t shower maximum and is made up of scintillating strips. This shower maximum detector (SMD) is read out independendtly form the EM calorimter. The SMD enables us to collect much more information about the shower shape and is used for accurate 2D position measurement and improving hadron rejection. 2. Design

We considered several factors during the design phase of the SMD, the most important beign hermiticity, granualarity, signal amplification and cost. To take advantadge of the azimuthal symmetry, the detector is broken up into 8 identical wedges. Strips are oriented along the sides of each wedge so that the full area is covered. There are 2 layers on a wedge labeled u and v with the u layer being closest t o the interaction point (IP). Strips from different layers cross at a 45' angle. With this construction, strip lengths vary considerably

557

558

and have dimensions 5mm x 6mm x (2.5 - 109.3 cm). Layers are divided into 2 regions defined by their distance from the IP. In all there are 200 strips per layer. Figure 1 shows the wedge layout.

12.6-27.9 Longest Strip (cm)

109.3

Figure 1. Wedge layout. 1 megastrip = 10 strips.

Wedges are covered on each side by lexan panels. 4 source tubes are present

559

on one of the panels for in situ calibration of the SMD. A wire driven Co60 source can be inserted into these tubes thereby irradiating any strip within the detector. Since the source tubes are present only in one of the panels, the strip response of the u layer is less than the v layer. Sourcing the detector over periodic intervals is a good way of detecting signal degredation due to scintillator aging and radiation damage.

3. Component Selection Monte carlo was used to study position resolution and hadron rejection, the two main goals of the detector. They indicated a minimum light yield of 1 photoelectron (pe) per minimum ionizing particle (MIP) is necessary to meet these goals. After testing several products, the materials chosen were Bicron BC408 scintillator and Kuraray Y11-350ppm WLS fibers. Each component had to pass quality tests, otherwise it was sent back to the manufacturer. To read out a large number of fibers, Hamamatsu R5900-M16 multi-anode phototubes (MAPMT) are used. This is a 3rd generation tube, which is much more suited for our purposes than any other one available at the time. The key MAPMT features are it's compact size and it's low cross-talk levels. In all 416 are used. The MAPMTs in the low eta region operate at a nominal gain of 5E5 while the high eta MAPMTs operate at a gain 5 times smaller. A light yield of 3 pe/MIP is reached.

4. Calibration 4.1. Test Stand We initially calibrated all MAPMTs at UCLA. The MAPMT high voltages (HV) were determined for both gains settings along with channel variation, dark current, cross talk and non-linearity. 4.2. Test Beam

The radioactive source calibration system was first used during the test beam, where 5-220 GeV positrons were directed at one of the 45" wedges. A comparison of the source response to the antiproton response for every strip agrees very well. Only one source tube was used at the time.

4.3. Source Calibration After the insertion of the detector in it's final destination at CDF, source calibrations was performed in 3 iterations. For the first 2, HVs were adjusted

560

which decreased the strip gain variation (RMS) by 35% in the low eta region and 68% in the high eta region. The last run was used to define individual strip corrections that goes into the offline code and reduced strip variation by another 24%. An energy scale was also determined by comparing the average strip response to a smilar measurement taken during the test beam.

4.4. Other Methods

The radioactive source method is the first option because it can theoretically illuminate every strip with a constant signal. In reality not all strips are irradiated evenly, while some aren’t at all. Also, this method can only be used during CDF shutdown times. To calibrate these remaining strips, minbias data is used. For every strip, there are 15 others which are identical. By comparing the average energy, the offline corrections can be calculated. Enough events are used, so that statistical fluctuations are minimized.

5 . Performance

5.1. Strip Response We are interested in understanding the performance of the SMD on several levels. The first step is to study individual strip response. Occupancy and energy distributions are plotted to look for strips or groups of strips that stick out. By doing these types of comparisons we see that the overall response agrees with our specifications. Currently the number of dead or low channels is approximately 1.5% of the total. And by looking at the response as a function of time we can measure the fluctuations and drift if any. A comparison was done of two min bias data sets taken l-month apart and found the overall difference to be 2%.

5.2. Shower Profile When an energetic particle tranverses the detector, a cluster will be formed if the highest energy (seed) strip passes a threshold cut. The cluster is currently 9 strips wide centered about the seed. To study the clusters, we accumulate them from electrons coming from W and Z candidates. The resulting shower profile, like that shown in Figure 2, is then compared to the ones from the test beam. Overall the shape is typical for energetic EM objects. The only concern is that cross talk between neighboring channels is higher than expected. This is currently under investigation.

561

Figure 2.

Average shower profile from 3900 W

-> e v

events

5.3. Energy Matching

A nice performance check for the SMD is matching cluster energies to the EM calorimeter for the same particle. For EM clusters with more than one SMD cluster, the closest one is chosen. Figure 3 demonstrates a good correlation between the two detectors. The SMD energy is on average 30% higher, but is not a concern since determining the energy scale is more difficult than in the EM calorimeter. At about 90 GeV the energy flattens out, which is due to MAPMT non-linearity and the fact that it takes more material for more energetic particles to reach shower maximum. 5.4. Position Resolution

With each strip 5mm wide, the SMD should have excellent position measurement. The best way to measure the resolution is to measure the distance between the SMD clusters and extrapolated tracks. CDF has a Silicon Vertex Detector (SVX) which can measure tracks to 17 = 3. Unfortunately, the SVX hasn't been fully integrated into CDF. In the mean time, we substitute the EM calorimeter for the SVX. This comparison yields a position matching of about 1.5 cm which is dominated by the size of the EM calorimeter. 6. Conclustions

A new Shower Maximum Detector using new technology is now being used at CDF which extends the 17 coverage out to 3.5. The detector has been cali-

562

Figure 3.

Energy matching between EM calorimeter and SMD.

brated using a Co60 source and over 98% of the channels are working properly. Progress is being made to improve the uniformity of the detector by replacing damaged components and more carefully calibrating channels were necessary. SMD tests, including analyzing shower profiles, energy matching and position resolution studies show that the detector is working well. And work is continuing so that we can better understand shower shapes and make the proper corrections. References 1. The CDF Detector for Tevatron Run 11, Technical Design Report, The CDF I1 Collaboration, FERMILAB-Pub-96/390-E (October 1996). 2. G. Apollinari, et al, Shower maximum detector for the CDF plug upgrade calorimeter, Nucl. Instr. and Methods in Phys. Res., Sect. A 412 (1998) 515526.

CDF I1 INTEGRATED CALORIMETRY ENVIRONMENT

S. DELL’AGNELLO Labomtori Nazionali d i Frascati dell ’INFN via E. Fermi n. 40, Frascati (Rome), I-00444, ITALY

e-mail: [email protected]

(For the CDF II Collaboration) The Run IIa integrated calorimetry environment of CDF I1 comprises: sampling scintillator calorimeters, e.m. pre-shower and shower-maximum detectors (gas based in the central, scintillator based in the forward), crack-filler detectors, radioactive source calibration systems, light pulse calibration systems, dedicated triggers for critical low energy calibrations, new FEE, a Windows N T slow control system, the master online-offline Oracle Database and an online data validation framework, distilled over the previous years of CDF data taking (now based on the CERN Root). The pre-existing central calorimeters are the bridge between the energy and time measurement of the past and current physics runs. New forward calorimeters (the “plugs”), built with the modern scintillating tile-fiber technique, replaced the old gas calorimeters. Together they form an integrated, generalpurpose calorimetry system which has been successfully commissioned with the 2000-2001 collider data. The time measurement, before present only in the central hadron, has now been extended the plug hadron calorimeter. To complete the integration in Run IIb (>2004), the central gas pre-shower will be replaced with a tile-fiber detector and the time information will be added also to the e.m. calorimeters. Select jet energy topics are described which show how the CDF I1 calorimetry has the capability to improve the measurement of the W and the topquark masses, and to enhance the search for dijet mass peaks in conjunction with btagging.

1. Evolution of the Calorimeters and Run I1 Commissioning

This paper describes the evolution of the CDF I1 calorimeters into an more integrated environment for the physics Run I1 at the upgraded Fermilab Tevatron. The Tevatron energy was raised from 1.8 to 2.0 TeV (which increases the top-quark production cross section by 40%) and the bunch crossing time was shortened from 3.5 psec to 132 nsec (which forced a complete replacement of the FEE’). Peak (integrated) luminosities are expected to reach 5 ~ 1 0 ~ cmW2sec-’ or more after the year 2004 (10-20 fb-’ before the turn on of the

CERN LHC). A review of the central calorimeters can be found in2. The upgrade of the

563

564

forward calorimeters (the “plugs”) is described in the CDF I1 TDR3. Since the new plugs are scintillator-based an integrated design for the new FEE could be chosen. The central e m . shower detectors are the only ones using a customized version of the new ADCs. The plugs are now instrumented to measure the time associated to the energy release in the hadron compartment (like the central was since Run I), and they have a better shielding from beaminduced background, which caused large noise in Run I. The e.m. shower measurement has now been extended to the plugs4. The commissioning of the calorimeters was performed in 2000 and 2001. The absolute energy scale was calibrated using test beam data and is mantained in situ with radioactive source systems (137Csin the central, 6oCoin the plug) and special control samples from collision data. Relative variations of the PM gains are tracked using combinations of light pulsers, like Lasers, LEDs, PIN diodes and Xe flashers. The whole central source system was heavily refurbished. In particular, the mechanical source drives of the hadron calorimeter towers at 0.66< 1771 <1.32 (the “endwall”, see reference2) were all replaced. Since the 1983-85 pion test beam, these particular towers have had a 20% decrease and a 20% broadening of their light yield. In 2001, many source calibrations were taken to equalize the PM gains (by changing their high voltage) in order to set the response of these towers to be in the middle range on the input dinamic range of the ADCs. Similarly, the central e.m. looses l%/yr, both with and without beams, mostly due t o the shortening of the attenuation lenght of the scintillator. For the remaining hadron towers this decrease since 1985 is less that 10%. None of these effects are a concern for the stochastic resolution term. One of the main surprises of the energy calibration in Run I1 was due to the reduction of the ADC integration gate from 700 nsec of Run I to the current 132 nsec: the central hadron calorimeters loose about 6.5% of the energy outside the 132 nsec, while all the others loose on average only about 2%. This was measured using control samples of low energy jets and muon MIPS, in which data from 4 consecutive 132 nsec gates were acquired. This effect is believed to be due to a longer 2nd time component of the signal of the central hadron scintillator, whose material is different from all the others. The value of the MIP from J/$ muons in Fig. 1 shows how the central hadron energy scale was restored to the Run I value at the % level at the end of the commissioning period in November 2001. The top plot of Fig. 2 shows that also the energy scale of the central e.m. calorimeter has been restored at the % level. This was achieved with an initial source calibrations to a few % ’ accuracy, then refined with high statistics E / P calibrations from the inclusive electron sample. The bottom plot of Fig. 2

-

-

-

565 M,, of Z + e

"

0.035

d

8'

can idates, central+central

O.s.mb.rwlJ.n"rym

15

0.03 0.025 40

*RB.U.

60

80 M..

lGeVlc 1% 1

120

140

0.02 0.015

M,. of 2 + e' e. candidates,centrakplug

40 R

0.01

-

35 30

0 25 t

0.005

20

.5

15

0

10

I

0 CENTRIL

3

2

t!ADRa( CNCR J/F'SI Kw

MIP

Figure 1. The Run I and the Run I1 muon MIP at the end of commisioning agree at the % level. The Run I1 distribution before 15%. commissioning is lower by

5 40

Figure 2. bution.

W

80

19

Me,(GeVlc )

120

140

The CDF I1 2 t ee mass distri-

N

shows that the new plug provides a large increase of the number of Z'S, which will allow a cross calibration of the central and plug e.m. calorimeters. Both central and plug electrons are contained in the W sample of Fig. 3. Note that the Run I top-quark discovery and W mass measurement used only central electrons and that for M(W + ev), the e.m. energy scale was ultimately set using the 2 + ee sample. The plug hadron energy scale is currently under study using the traditional technique of dijet balancing in the transverse place. By selecting one jet in the central and one in the plugs one can transfer the known energy calibration of the central to the new plugs. Other useful control samples include photon-jet balancing and MIPS from the increased CDF I1 muon coverage. The central pre-shower and shower-maximum2 gas detectors are important for e/r/.lr separation and for optimized jet energy corrections. The Run I performance and functionality has been restored. For example, the W electon acceptance measured from 2001 data is 99% for the wires and 100% for the strips. Hardware problems are down to < 1%single channel problems. Lab test confirm that for the shower-maximum no ageing is expected up to the highest Run I1 luminosities. The occupancy growth from 1%to few % will be acceptable. On the contrary, because of high occupancy problems, the preshower will be replaced in 2004 (Run IIb) with a tile-fiber scintillator similar

566

to the plug. This will integrate into the design also the +crack detector, which uses tungsten bars as crack fillers.

M, of W

120

+ e v candidates

ecernber 20D1-January 2002

20 Figure 3.

40

60

80 100 M, (GeV/c2)

120

140

Th e CDF I1 W transverse mass distributions, with central and plug electrons.

2. Select Jet Energy Measurements over 20 Years The standard CDF jet energy corrections use mainly calorimeter information. The good performance of these corrections is shown by the wealth of published papers involving jets. In the 80’s (4.4 pb-l) and 90’s (120 pb-l, “Run I”) new algorithms have been developed to improve the jet energy resolution, c ( E j e t ) , by exploiting additional information. Because the central hadron calorimeter is non-compensated, its response to n* of E< 10 GeV is non-linear (lower, up to 40% at 0.4 GeV). This has been measured in situ with minimum bias data, because the lowest energy of the 1983-85 test beam was 10 GeV. In Run I1 to measure more accurately this non-linearity an isolated track trigger is used. Correcting jets for this effect track-by-track, using their measured momentum provides a 15%improvement

567

of g ( E j , t ) compared t o the standard energy corrections5. CDF also searched for the mass peak of W / Z + j e t - j e t , to test the effectiveness of any jet energy correction. U A 2 observed this signal6, but the mass resolution was not enough t o separate the W from the Z, thus preventing direct jet energy scale calibration. Unfortunately, at the Tevatron the S/N is worse and the CDF trigger thresholds were too high. A further step was achieved in the 90's. Before forming jets, tower energies are corrected using tracking, shower-maximum and calorimeter information, with an approach based on the energy flow concept7. This algorithm included the correction for non-linarity of the 80's. The improvement over the standard jet corrections shown in Fig. 4 is 25%. N

Photon

3

+ Jet P, Balancing in CDF Data

E!

v

Typiul CDF Id Rcrolurion urim

-2.5

ClrlDriserryonly

A New CDF Id A l g o t i i h Udna Tmckhg, CdotinulryadShourrM u Ddcdorr

1

ct z

1

8 1.5

0.5

20

25

30

35

40

45

50

55

60

SO

'0

Photon P, (GeV)

Figure 4. Improvement of a ( E j e t ) with optimized corrections shown in photon+jet data.

la, 150 Ua 250 Duet Invariant Mass (GeVlcz)

:

Figure 5. The reconstructed W mass in top events using the standard CDF jet corrections.

Using Run I data, the 2 + bb peak was observed in the inclusive muon trigger sample. A btagged jet in the event was required and special &jet corrections were developed to account for the presence of a b semileptonic decay. Note that a 30% improvement in r(Ej,t) gives 25% increase in the M ( H + bb) peak significance with 10 ft-'. CDF I1 can increase the sample of 2 + bb by directly triggering on secondary decay vertices using the Silicon Vertex Tracker3, which is one of the major novelties of Run 11. In Run I an important check of the standard jet corrections came from the reconstruction of the W mass in the tt + e / p 4 j e t s channel, using the two non-b jets. The signal is shown in Fig. 5. The reconstructed mass is M ( W ) = 77.2 f 3.5(stat) f 2.9(syst) GeV. With 2 fb-', CDF 11 will observe 500 W + qq' events.

-

N

+

-

568

0.5

0

0.5

I

1.5

meKG-Y

Figure 6. Check of the energy scale of standard CDF jets using the 2 1 jet sample.

+

2

2.5

3

3.5

&

4.5

Iml

Figure 7. The Run I1 time resolution vs. energy has been restored to the Run I value, and extended from the central to the whole hadron compartment.

Another equally important check is given by the Pt balancing of 2’s (and 7’s) decaying to dilectrons recoiling against 1 “standard” jet. The mean of the fractional unbalance (see fig. 6 ) indicates that the the jet energy scale is correct at the 2% level, which is also comfirmed by the CDF simulation. Despite this, the systematic error on the jet energy scale in the measurement of the topquark mass has been conservatively taken to be 10%. With 2 fb-’ CDF I1 will 27,000 Z(-+ ee) 1 j e t events. The standard collect a control sample of jet corrections are being used to cross calibrate the old central and the new plug calorimeter and to produce the first physics results, because they provide a reliable energy scale and are well tested. On a longer time scale, however, the corrections which improve a(Ej,t) will have to be adopted to exploit as much as possible the detector and physics capabilities of CDF 11.

-

+

3. Time measurement

In Run I the central hadron timing has been used to remove beam-gas, Main Ring splashes and cosmic ray background. Rare SUSY searches involving the detection of electrons and photons have also shown that the rejection of cosmic rays is mandatory. Until now this is achieved by measuring the time of the e.m. shower leakage into the hadron compartment. This provides good rejection, but the efficiency is so low that is makes many interesting searches hopeless.

569

In order to improve this efficiency for Run IIb, the timing infomation will be added both to the central and to the plug e.m. calorimeters. In Run IIa all hadron calorimeters measure energy and time (with a 0.4 GeV threshold in the central and 1.0 GeV in the plug). The associated FEE has been the last one to be successfully commissioned. The endwalls showed very high RF pick-up noise, which was eliminated by making new custom discriminators with filtering transformers at the inputs. Hardware problems are down to few % single channel problems. Fig. 7 shows that the time resolution vs. energy of Run I has been restored over the full solid angle calorimeter.

Acknowledgments The speaker wants to thank warmly the INFN national scientific committee for high energy physics (“Gruppo I”), and especially M. Calvetti and U. Dosselli for having supported (and made possible) the work of INFN researchers on the upgrade and commissioning of the central hadron calorimeters.

References 1. C. Nelson, these Proceedings. 2. S. Bertolucci et all NIM, A267, 301 (1988); L. Balka et al, NIM, A267, 272 (1988). 3. The CDF I1 Collaboration, FERMILAB-Pub-96/390-E. 4. A. Attal, these Proceedings. 5. The CDF Collaboration, PRL 65, 968 (1990). To date this is the only CDF paper which does not employ the standard jet corrections based on calorimetry. See also B. Hubbard, Ph.D. thesis, University of California at Berkeley, CA, USA and S. Dell’Agnello, Laurea thesis, University of Pisa, Italy. 6. C. Rubbia, Proceedings of the 7th Topical Workshop on Proton-Antiproton Collider Physics, p. 822, Fermilab (1988), ed. by R. Raja, A. Tollestrup, J, Yoh, World Scientific. 7. S. Lami, A. Bocci, S. Kuhlman, G. Latino, FERMILAB-CONF-00/342-E. Published Proceedings 9th Conference on Calorimetry in High Energy Physics (CALOR 2000), Annecy, France, October 9-14, 2000.

BOREXINO: A REAL TIME LIQUID SCINTILLATOR DETECTOR FOR LOW ENERGY SOLAR NEUTRINO STUDY

LINO MIRAMONTI Physics Department of Milan University and INFN Via Celon'a 16, 20133 Milano E-mail: Lino.Mimmonti4mi.infn.it

Borexino is a large unsegmented calorimeter featuring 300 tons of liquid scintillator, contained in a 8.5 meter nylon vessel, viewed by 2200 PMTs. The main goal of Borexino is the study, in real time, of low energy solar neutrinos, and in particular, the monoenergetic neutrinos coming from ' B e , which is one of the missing links on the solar neutrino problem. The achievement of high radiopurity level, in the order of 10-16g/g of U/Th equivalent, necessary to the detection of the low energy component of the solar neutrino flux, was proved in the Borexino prototype: the Counting Test Facility. The detector is located underground in the Laboratori Nazionali del Gran Sass0 in the center of Italy at 3500 meter water equivalent depth. In this paper the science and technology of Borexino are reviewed and its main capabilities are presented.

1. Introduction

Our Sun is essentially a self-confining nuclear fusion reactor whose production rate is regulated by the weak nuclear interaction. All the reactions can be summarized by a single one in which four protons combine into a 4 H e nucleus: 4p -+a + 2e- 2u,. In this process two neutrinos are emitted and energy is released (E 26.6 MeV). About 6.10'' neutrinos of solar origin hit a centimeter square of the Earth per second. The discrepancy between the u, expectations of the solar neutrino flux, as calculated by the Solar Standard Model (SSM)l, and the experimental results, is known as the Solar Neutrino Problem (SNP). The recent results from helioseismology confirms the SSM to an accuracy within 1%. A possible explanation of the SNP is that u, emitted from the fusion reactions oscillate while traveling from the Sun to the Earth and transmute in non electronic flavour neutrinos, with cross sections null or lower than those of electron neutrino. This oscillation is possible if neutrino has mass and mass eigenstate do not coincide with flavour eigenstate. The probability for an electron neutrino of energy E to conserve its flavor

+

570

571

traveling a distance L is: P = 1- sin228 . sin (LAm2/(4E)),where 8 is the mixing angle and Am2 the difference between the squared masses. This scenario (Vacuum Oscillation scenario or VO), can explain the experimental data if the mixing angle is such that sin228 N 1 and the Am2 N 10-11eV2. In case of resonant oscillation in the core of the Sun (Mikeyev-Smirnov-Wolfenstein effect or MSW effect), two different scenario are obtained; Large Mixing Angle solution (LMA) for sin228 N 1 and Am2 N 10-6eV2, or Small Mixing Angle solution (SMA) for sin228 N and Am2 N 10-6eV2. 2. The Borexino detector

2.1. The Borexino design The construction philosophy adopted by our collaboration is that of a graded shield of progressively lower radioactivity material approaching the detectors core, ending in the definition of a fiducial volume. The apparatus is located deep underground at about 1780 m of overburden rock (about 3500 meter water equivalent) in the Apennines in the Laboratori Nazionali del Gran Sass0 of the National Institute of Nuclear Physics (INFN) in the center of Italy. The muons crossing the detector is about 1.1per m2 h-l corresponding to a reduction of about six order of magnitude. The core of the detector consists in 300 tons of liquid scintillator contained in a 125 pm transparent nylon vessel (Inner Vessel) of 8.5 m diameter, viewed by 2200 photomultiplier tubes supported on a stainless steel sphere (SSS) of 13.7 m diameter. This sphere is than enclosed in a cylindrical tank 17 m height with a diameter of 18 m. The sketch of the apparatus is presented in Figure 1. The zone between the Inner Vessel and the photomultiplier tubes is filled with pure pseudocumene (with a 5 g/l shift wavelenghter) in order to compensate the buoyancy force and to ensure good light coupling; furthermore in pseudocumene the radiopurity achievable is greater than in water. The region between the stainless steel sphere and the external tank is filled with ultrapure water in order to reduce neutrons and gammas coming from the surrounding rock. Furthermore the ultrapure water contained in the external tank serves also for muon veto. Between the stainless steel sphere and the Inner Vessel a second nylon vessel (Outer Vessel) is planned in order to stop the radon emanating from the outer part of the detector and in particular from the photomultiplier tubes. There are three different classes of photomultiplier tubes; the first one (1800 inner PMTs) is composed by tubes equipped with a light concentrator designed in such a way to observe only photons coming from the Inner Vessel. A second class is composed by 400 PMTs without light concentrator, that in addition to the light from the Inner Vessel are able to detect the

572

\.

L HoldingStrings Stainless Sleet Water Tank

tam0

Figure 1.

\

Steel Shielding Piales Bmx 8mx locm and 4m x 4m x 4cm

Schematic view of Borexino.

light produced by events in the pseudocumene buffer. The task of these inner PMTs is that of survey for the light produced by muons that could be confused with neutrino signal. The last class of photomultiplier tubes is composed by 200 outer PMTs placed on the outer surface of the stainless steel sphere. The goal of these devices is to identify muons both traversing and not traversing the pseudocumene buffer. In the first case, they allow a double tag to identify such events, and for the second case they give an alert to search afterward for muon induced events which could reach the scintillator.

2.2. The photomultiplier tubes The photomultiplier tubes employed in Borexino are 8 inchs 9351 made by ETL (former EMI). The main characteristics are summarized in Table 1. Among its performances we can notice a limited transit time spread (jitter) that is less than 1 ns, a good peak to valley ratio (2.5) a very low dark noise rate (1 kHz), a low after-pulsing probability ( 5 4%). hrthermore, the strong requirements about radiopurity led us to manufacture these PMTs with special low radioactivity glass and internal parts2. The 1800 light concentrators are realized with anodized aluminium in order to have a good reflectivity in the 400 nm region and a good compatibility with pseudocumene.

573

Table 1. Characteristics of Borexino 8 inch photomultiplier tbes. Characteristics Quantum efficiency

26% at 420 nm 1 ns

Transient time spread Peak/valley

2.5

Dark noise

1 kHz

lo7

Gain

Low radioactivity glass

schott 8246

2.3. The liquid scintillator

The liquid scintillator is composed by Pseudocumene as solvent and PPO at the concentration of 1.5 g/l as solute. The pseudocumene (PC) is 1,2,4-trimethylbenzene CsH3(CH3)3, and the PPO is 2,5-diphenyloxazole C I ~ H ~ ~ NThe O .refractive index of the scintillator is equal to 1.505. The excitations, produced by electrons (b)/gamma rays (7) and alpha particles ( a ) ,have different properties (see Table 2). During the research and development period a great numbers of tests were performed in order to study the intrinsic optical properties of the ~cintillator~. At the concentration of 1.5 g/l, the light yield results to be about 11000 photons per MeV, with a fast decay time of about 3.6 ns (the intrinsic decay time of the PPO is 1.6 ns, and those of pseudocumene without solute is about 28 ns). The attenuation length is about 30 m and the scattering length is about 7 m, both for a wavelength of 420 nm. This mixture gives also a good alpha/beta discrimination (see Figure 2).

Table 2. The decay times and the relative quantities of the decay components of the liquid scintillator PC+PPO at 1.5 g/l. c1

Second

17.61 ns

Third

59.49 ns

Fourth

330.4 ns

574

:

FC * PPO ( I . 5 g i i. I

, .*

i

c

?OO

i

406

6co

BoEi

1000

1200

1400

I600

T i m (11s)

Figure 2. Experimental light profiles measured for the scintillator (PC with PPO at 1.5 g/l for alpha and beta particles.

2.4. The Borexino detector performance

The PMT's total coverage is about 34%, hence, taking into account the photocathode efficiency, the photoelectron yield is about 450 pe/MeV (at 1 MeV), giving an energy resolution of 8% (at 1 MeV) and a spatial resolution of 12 cm (at 1 MeV). These good performances, high energy and spatial resolution are very important for the experiment; the energy resolution is essential to disentangle the recoil spectrum of the scattered electrons from the incoming neutrinos via the sharp edge at about 660 keV. The high spatial resolution permits a good spatial reconstruction of the events and hence to reject the background coming from outside. 2.5. Calibrations and monitoring of the detector

In order to assure a careful understanding of the detector operational conditions during the data taking, different calibrations and monitoring systems are envisaged5. For a precise event reconstruction, it is needed a precise time measurement of the PMTs signals. We plan to perform this measurement by illuminating each photomultiplier tube with an optical fiber coupled to a laser generating a fast pulse of the order of 50 ps. It will be also monitor the gain stability of the PMTs since these last are illuminated at the level of single photoelectron. In order to monitor in time the stability of the scintillator, we plan

575

to locate an external 228Thsource just outside the stainless steel sphere. The scintillator will be monitored also via radon sources inserted inside the inner vessel through organ pipes. In addition, a continuous check of the stability of the optical properties of the buffer is performed by means of laser light that induce photoexcitation process. 3. The physics of Borexino

The main goal of Borexino is the detection of solar neutrinos coming from the electron capture of 7Be that gives a monochromatic line at 862 keV through the u e- + u e- electroweak scattering reaction. The high radiopurity level and the high yield of light of the scintillator make it possible to reach a detection threshold as low as 250 keVa. The scattered electron produces a continuum recoil spectrum, with a maximum energy of about 660 keV. The expected event rate, according to the Solar Standard Model in the case of non-oscillation, is about 55 events per day in 100 tons of fiducial volume. The ultimate background component is represented by the scintillator itself; the U and Th content must be kept at the level of 10-l6g/g, while for the natural K at the level of 10-14g/g. This level of impurities will give, in 100 tons fiducial volume, about 10 events per day. Figure 3 shows the estimate background spectrum and the Solar Standard Model neutrino signal. In Table 3 the solar neutrino counting rates expected per day in the Borexino neutrino window (i.e. 250 keV - 800 keV) for 4 scenario are reported: The Solar Standard Model (SSM), the Large Mixing Angle solution (LMA), the Small Mixing Angle solution (SMA) and the low mass solution (LOW).

+

+

4. The Counting Test Facility

In order to prove the capability to reach a so high radiopurity level, a prototype of the detector was constructed: the Counting Test Facility (CTF). The main goal of CTF was to develop solutions directly applicable to operational issues for Borexino(for further details see reference4). In CTF 4 tons of liquid scintillator are contained in a 2 m diameter transparent nylon vessel mounted at the center of an open structure that supports 100 PMTs. These last are coupled to optical concentrators viewing the nylon vessel with 20% optical coverage. The whole system is placed within a cylindrical tank (10 m of height and 11 m of diameter) that contains about 1000 aThe 250 keV energy threshold is dictated by the The 14C/12C must be in the range of 10-l8.

14C content

in the organic scintillator.

576

Event Energy

Figure 3.

(MeV)

The estimate background spectrum and the Solar Standard Model neutrino signal.

Table 3. Solar neutrino counting rate per day in Borexino fiducial volume in the neutrino energy window (250 keV - 800 keV) for 4 scenarios: The Solar Standard Model (SSM), the Large Mixing Angle solution (LMA), for the Small Mixing Angle solution (SMA) and for the low mass solution (LOW). These rates are calculated for a mean Sun-Earth distance and 113 m3 fiducial volume. The Am2 and sin228 used parameters are: Am2 2 1 . 8 . 10-5eV2 and sin228 = 0.76 for LMA, Am2 N 5.4 . 10W6eV2 and sin229 N 5.5 . for SMA and Am2 2 7 . 9 . 10-seV2 and sin228 = 0.96 for LOW. 15 0 17F 8B total -- -

c

I

SSM

0.22

LMA

0.15

0.95

SMA

0.08

0.39

LOW

0.13

1.03

24.4

5.5

0.07

0.08

2.27

2.86

0.03

0.03

0.87

1.12

0.01

0.04

2.13

2.86

0.03

0.04

tons of ultra-pure water in order to shield against gamma rays coming from the PMTs and other detector materials and neutron from the rock. With 300 pe/MeV the CTF reachs an energy resolution of 9% at 825 keV (214P0 line) and a spatial resolution of about 12 cm at the same energy. Concerning the a l p discrimination capabilities, the (Y identification efficiency is 95% and the

577

inefficiency is of few percents. The radiopurity levels evaluated in CTF are:

238U5 3.5 f 1.3. 10-l6g/g 2 3 2 ~ h= 4.4+1.5. 10-16 9/9 -1.2 14C/12C = 1.94 f0.09. 5. Conclusions

The demands on the radiopurity of the detector materials, especially for the scintillator itself, are very challenging. The CTF has demonstrated the capability to achieve unprecedented levels of radiopurity in the liquid scintillator, opening in this way the construction of Borexino. Today, Borexino is in its construction final phase, and probably it will be able to take physical data from the spring of the next year playing in this way a crucial role in the understanding of the solar neutrino puzzle.

Acknowledgments The author acknowledges the support of the Italian National Institute of Nuclear Physics.

Collaboration list Belgium: I.R.M.M. European Joint Research Center - Geel. Canada: Queen’s University - Kingston. France: Coll6ge de France. Germany: Max-PlanckInstitut fuer Kernphysik Heidelberg, Technische Universitaet Muenchen. Hungary: KFKI-RMKI Research Institute for Particle and Nuclear Physics Budapest. Italy: Dipartimento di Fisica Universita and 1.N.F.N di Genova, L.N.G.S., Dipartimento di Fisica Universita and I.N.F.N. di Milano, Dipartimento di Fisica Universita and I.N.F.N. di Pavia, Dipartimento di Chimica dell’universita and 1.N.F.N di Perugia. Poland: Institute of Physics, Jagellonian University Cracow. Russia: J.I.N.R. DUBNA, Kurchatov Institute Moscow. United States: Bell Laboratories Lucent Technologies, Massachusetts Institute of Technology, Princeton University, Virginia Polytechnic Institute.

References 1. 2. 3. 4. 5.

J.N. Bahcall et al., Astrophys J.555:990-1012, (2001) and references therein. G . Ranucci et al., Nucl. Inst. and Meth. A 333 (1993) 553-559. F. Elisei et al., Nucl. Inst. and Meth. A 400 (1997) 53-68. G . Alimonti et al., Nucl. Inst. and Meth. A 406 (1998) 411-426. L. Miramonti, Progress in Part. and Nucl. Phys. 48 (2002) 27-28.

THE MINOS FAR DETECTOR CONSTRUCTION AND QUALITY ASSURANCE TESTING

L. MUALEM Tote Lab of Physics, 116 Church St SE, Minneapolis, MN, USA E-mail: mualem4hep.umn.edu (For the MINOS collaboration)

The MINOS experiment will study neutrino oscillations using the Fermilab Main Injector neutrino beam and both near and far detectors. The detectors are finegrained sampling calorimeters with 1 inch thick steel absorbers with scintillator as the sampling elements. The scintillator planes are segmented in 4 cm wide strips for tracking and event topology measurements. The very large size of the far detector (26,000m2 of scintillator) requires that the cost per unit of detector be kept low. A combination of extruded solid scintillator, wavelength-shifting fibers, multi-pixel PMTs and low-cost electronics meets this challenge. In this talk some important details of the system design will be discussed along with test results from fabrication and initial performance of the first 100 installed far detector planes.

1. The MINOS Detector

The MINOS far detector is a part of a long baseline (735km) neutrino oscillation experiment between Fermilab and Soudan, Minnesota. The detector is made of 8 m octagonal planes of plastic scintillator strips mounted on steel planes andis being built as two magnetized supermodules. At this time the first supermodule is 2 / 3 complete. The first magnet coil will be installed during the summer of 2002, with the completion of the rest of the detector being completed in the spring of 2003. The total mass will be 5.4kT. The detector has to serve as a target for neutrinos, but also has to be affordable. It has to function as a calorimeter, because the energy of the events is critical to determining the oscillation parameters. The specification on the energy determination is that it has to be correct to 2% between the near and far detector and 5% on an absolute scale. Muon neutrino charged current events will be the most common beam events. These will have the energy measured with range and curvature, since the detector is magnetized. Other events, like the v, charged current and neutral current events will cause electromagnetic and hadronic showers in the detector. Calibration of these

578

579

Figure 1.

Completed MINOS detector plane being lifted for installation.

events will come from test beam data taken at CERN.

2. Detector Composition

The scintillator planes that make up the active part of the detector are built of 8 aluminum skinned scintillator modules shown in figure 1. These are mounted on the 2.54cm thick steel which makes up most of the mass of the detector. Inside each of the modules are strips of extruded scintillator, which is key to making it affordable. The strips are 4.lcm wide and lcm thick, with the longest covering the full 8 m. In order to aid in light collection, the strips have a co-extruded Ti02 coating which provides a diffuse reflector. The light is actually collected by a WLS fiber that is glued into a groove, which is also part of the extrusion profile. Figure 2 shows a section of the extrusion with the fiber glued in.

580

Figure 2. End on picture of a section of a 4.lcm x l.Ocm scintillator strip with wavelength shifting fiber glued in groove. White outer coating is co-extruded Ti02 in polystyrene.

3. Module Construction

These strips are packaged in an assembly line type of operation at the module factories located at the University of Minnesota, Caltech, and at Argonne National lab. The first production step is to trim the extrusions to their final length. All the extrusions for a given module are cut at one time with a standard circular saw. This is the only operation performed on the extrusions, which minimizes the handling and consequently, the cost of the modules. The strips are then placed in a thin bed of epoxy insie a pre-formed aluminum light case. The whole assembly tray is then covered with plastic and evacuated to a pressure difference of 6” of water to laminate them to the aluminum light case. After the glue has cured for a day, the WLS fibers are glued into the strips. This is done by a machine, designed and built at Argonne National Lab, that follows along the groove of each strip dispensing glue, fiber from a spool of about lkm of fiber, and covering them with aluminized tape. The motion along the strip is motorized and automated, with the operator threading fiber through the manifolds at each end. Once all 28 strips are done, the module is place in a rack to cure for a day. At least four modules can be produced each day at each factory. The next day an aluminum top is placed on the module and vacuum laminated again, to complete the light case. Once the top is cured, the light case is crimped. This is done with a machine designed and built at the University of Minnesota which folds over flanges on the top and bottom of the light case to make a light tight labyrinth seal. The optical connector is then flycut with a diamond tipped flycutter bit to finish the connector and fiber ends. 4. Module Testing Once assembled, the module is tested for its light output to see if it meets the light output standards. This is done on a source scanning table which moves a 5mCi radioactive source (137Cs) continuously across the module, at longitudinal steps of 8cm. The light from each fiber is determined by measuring

581

the current from a pixelated Hamamatsu M-16 PMT with one fiber on each pixel. The collimation of the 13'Cs source gives a triangle profile in the current where the peak determines the light output at that position, and the intercepts of the triangle locate the strip edges.

-10

-

f

1

%

-8

-

1

-8 C

2

-6

-

c

a

-4

0.02 -

--

-

-2

-

Figure 3. Light output as a function of distance for a representative strip from a module read out from both ends.

Figure 3 shows the light output as measured from both ends for a particular strip in a module. Most of the anomalies are due to slight variations of the depth of the fiber in the groove. A summary of the module is also produced which allows one to identify any fibers with anomalously low output which might preclude its installation in the detector. All of this information goes into the experiment database for use in analysis of data. Figure 4 shows one way to look at the end result performance of the modules. It shows the variation in light output for a slice at 2m for all of the modules that have been built. With over 60% of the module production complete a Gaussian fit shows an 11%sigma. The fraction of strips with light output less than half of the mean, which is how we define a damaged fiber, is at 0.2%, while the design goal was less than 1%. 5. Plane Construction and Installation

After passing quality checks at the factory the modules are loaded into shipping crates, each of which holds 16 modules. They are then shipped on trucks, as many as 5 crates high, to Soudan. On arrival they are unloaded, lifted

582 Relative Light Distribuition

I

Figure 4. Relative light output measured by a slice through all constructed modules. Th e Gaussian fit has a sigma of 11%.

on-end into the hoist, and taken N 1/2 mile down to the mine through a l m x 2m shaft. Once underground, they are pulled out the back end of the

1-

s

0.1

C

e

5

c 80.00

0

c

a 0.06

0.04

0.02

'0

100

200

300

400

500

600

700 000 Distance (cm)

Figure 5. Light output as a function of distance for the strip shown in figure 3 as circles, with the second underground measurements shown as points with error bars.

583

hoist for transport into the lab. After all this the modules are again tested. To accomplish this, there is an identical module mapper underground which has been used to check the first 10% of the detector that was delivered. It showed that they arrive essentially unchanged. In figure 5 the black circles are the ,original factrory measurement and the points with error bars are the underground measurements. They agree very well. The scintillator modules are then assembled into planes, being placed on the 2.54cm steel. The entire plane and support system is then lifted and carried down the MINOS hall for installation on the detector. This happens about 7 times a week currently. Shortly after the plane is in place, it is connected up to the readout system, and it is immediately live, collecting what are mostly cosmic ray muons. These muons will provide the basis for the long time scale calibration of the detector, cross calibration with the near detector, and alignment of the detector. There will also be a low rate of neutrino induced events that can be used to investigate properties of atmospheric neutrinos. 6 . Construction and Installation Status

Table 1 shows some of the highlights of the production as compared to the design goals from our 1998 technical design report. Installation of both supermodules will be complete in June 2003. Table 1. Comparison of Design goals and schedule for MINOS far detector construction Design goals ( TD R 1998) Light Output (> 2.5 required from physics)

> 5.0 pe/mip

summed

Light Uniformity

Current 10 pe/mip 11%

Module Production Rate

lB/week

20-25/week

Module Production Time

22 hrs/module

18 hrs/module

Total Production Time

27 months

26 months

Production complete

Jan 2003

Feb 2003

Installation Rate

5.5/week

6.5/week

Installation Time

24 hrs/plane

18.75 hrs/plane

Total Installation Time

23 months

22 months

Acknowledgments This work was funded in part by the United States Department of Energy.

A NEW HERMETIC ELECTROMAGNETIC CALORIMETER DESIGN FOR FUTURE COLLIDER EXPERIMENTS

E. KISTENEV, P. NEVSKI, C. WOODY Brookhaven National Laboratory, Upton, N Y 11973, USA E-mail: [email protected]

V. KOCHETKOV, I. KOROLKO, E. MELNIKOV Institute for Theoretical and Experimental Physics, Moscow, Russian Federation

I. MAJATSKY TECHNOPLAST, Vladimir, Russian Federation

We present in this paper the initial status of a research and development project that will result in the construction and testing of a new prototype electromagnetic calorimeter consisting of a composition of corrugated (accordion shaped) Lead plates and scintillator sheets with embedded fibers for light collection. This novel design is simultaneously addressing a few of the major concerns of collider experiments - projection geometry and hermeticity while keeping mechanical structure relatively simple.

1. Introduction It is a well established fact that discoveries made by most of the completed and currently running collider experiments of the last 30 years relied heavily on the excellent calorimetry employed by those experiments. While actual techniques were quite varied, one may easily compile almost identical lists of issues which preoccupied calorimeter designers: 0 0 0

0 0

0

near 4n hermeticity; particle identification (implies granularity and good timing resolution); energy and angular resolution appropriate for electron/jet measurements in the kinematic range of the experiment (typically up to few hundreds GeV); fast response and short memory time; in-situ energy calibration; triggering.

584

585

Electromagnetic and hadronic calorimeters employing scintillators as an active media have always been the preferred solution which provides an almost perfect match t o this list, except if performance requirements are at an extreme. The first large-scale detectors were used in cosmic-ray studies'. The technology has made amazing progress in recent years, and probably the best collection of review material on the subject can be found in the recently published book of R.Wigmans2. Scintillator based calorimeters are the most abundant, since they are reasonably inexpensive, could be industrialized in their production and are robust in their operation. Energy and position resolution of a calorimeter are tunable parameters (by varying the sampling rate and frequency towards the limits set by homogeneous calorimeters). The actual performance limits are often set by the financial constraints. The break-through in technology of hermetic, large area, low cost electromagnetic calorimeters happened when plastic fibers doped with wave-length shifting fluors were first used to collect and transport the light from scintillating tiles/layers to photon detectors (photomultipliers, avalanche photodiodes e t ~ ) ~ . By decoupling two aspects of the calorimeter design: production of the measurable signal (scintillation light) and transporting it to the detector (photomultiplier), the WLS technique solved the major obstacle to implementation of the new approaches to structuring the scintillator based calorimeters. Two well known examples are the PHENIX lead scintillator calorimeter with penetrating fiber read-out ( ~ h a s h l i k )and ~ , the STAR parallel plate lead-scintillator calorimeter5, where the readout is provided by fibers embedded into groves in individual tiles. The characteristic energy resolution for this kind of detector is of the order of 8% at 1 GeV, with the position resolution being defined by the granularity of the individual detector. However the technique is still in need of further improvements. Existing attempts to accommodate the readout to the specifics of the particular detector design (pointing geometry, directionality, hermeticity) either failed to reach design goals or resulted in cost increases, complexity in production and assembly (in huge numbers of individual components, problems with automatization of the assembly process, resulting nonuniformities in the response of the assembled detector, etc.). The calorimeter design presented in this paper is based upon the prehistory of similar developments in ionization (with liquefied crypton and/or a r g ~ n and scintillation* electromagnetic calorimetry, where a new absorber-active media configuration known as the accordion geometry was proposed. This configuration avoids the problems with fine segmentation and pointing inherent to the classic planar geometries (even with penetrating fiber readout) and

586

allows flexibility in the formation of towers and absolute hermeticity, while keeping the number of mechanical components at a minimum. The interest in this geometry is further advanced by the progress in scintillator production: large inexpensive extruded sheets of high quality polystyrene based scintillator are easily available today from commercial sources. High precision computer controlled grove cutting machines solved the problem of a fiber embedding. Thermal shape modification of the scintillators and fibers has long been successfully used in particle physics experiments. Progress in technology and availability of low cost base materials allow one to consider the accordion geometry as a prime candidate for future experiments planning for a large area calorimetry coverage. The purpose of this work is to show that such a calorimeter can be built in industry, and to build a proof of principle prototype. Results of the prototype beam-testing will be published elsewhere. A complimentary goal has been to develop the Monte-Carlo simulation of the Pb-Sc accordion elecrtromagnetic calorimeter to the level available for ionization accordion calorimeters and thus to gain confidence in the design optimization procedures with respect to energy and position measurements, and to nonuniformities.

2. Pb-Sc accordion electromagnetic calorimeter design Given the proof-of-principle nature of this work, we have chosen to build a simple rectangular calorimeter module with dimensions 32 x 30 x 45 cm3 using 3.2 mm thick 32 cm wide extruded scintillator sheets available at TECHNOPLAST. The extrusion was made from melted granulated polystyrene similar to one used in the injection molding technique". The accordion calorimeter structure is shown schematically in Figure 1. Based upon extensive simulation and test results available from the ATLAS experiment" we have chosen the solution characterized by the ratio of the amplitude of accordion wave to the cell thickness (lead and scintillator combined) close to 4. Most of the rays orthogonal to the front of the calorimeter will pass through 4 cells per accordion period. To avoid light loss from the fibers, the folds have a lcm radius of curvature which introduces a modulation in thickness. In the future, the amplitude of the folds and radius of curvature at the folds will be adjusted to minimize response nonuniformities. One particular aspect of this work which is of specific interest is the solution we used to contain the light within scintillator sheets. The scintillator sheets were chemically etched as an alternative to classical techniques of light containment by wrapping, painting or coextrusion. At the time of the prototype construction little was known about the aging behavior of the refelectivity of the etched polystyrene surfaces. Later tests in the beam at CERN have shown N

N

587

Figure 1. Schematic representation of the accordion electromagnetic calorimeter structure.

Figure 2. A single shaped scintillator sheet with embedded fibes.

Figure 3. A prototype accordion Lead Scintillator calorimeter.

no obvious degradation in the light yield. Scintillator sheets were extruded with preformed grooves (fiber-to-fiber spacing of 8.9 mm). After etching, the groves were mechanically cleaned to insure the quality of light collection. A typical accordion shaped scintillator sheet with fibers ( using BCF-99A fibers, similar to those used in the PHENIX shashlick calorimeters4) embedded into groves is shown in Figure 2. Similar shaped lead plates of 3.2 mm thickness were stacked together with

5888

corrugated scintillator sheets. TYVEK reflecting paper was used t o improve light containment around the fibers and protect the etched scintillator against direct contact with the rather rough untreated lead surface. The assembled prototype calorimeter is shown in Figure 3. The fibers were arranged in 6x5 (fibers f rmz one cell x number of cells) groups and were bundled together to form a single readout tower. We intentionally avoided the temptation of creating light barriers between the towers in the direction orthogonal to fibers. By design, the accordion calorimeter has a built-in shower crosstalk between towers due to cell-to-cell coupling. Allowing for light crosstalk along the cells will serve to improve position resolution in both directions. The prototype was finally equipped with PMT housings identical to those used in the PHENIX electromagnetic calorimeter and moved to X7 beam line at CERN for testing.

3. Simulation The accordion shaped calorimeter was relatively easy to build but rather complicated to describe using standard geometry description tools available in GEANT package. In our attempt to create an initial simulation framework for the PbSc accordion, we largely benefited from the availability of the ATLAS simulation framework ATLSIMg. With minimal modifications, the ATLSIM framework was used to study

I

Figure 4. A Geant3 implementation of the accordion electromagnetic calorimeter. Superimposed are two electromagnetic showers of 10 GeV each.

Figure 5. Simulated response of the accordion electromagnetic calorimeter to 10 GeV/c electrons.

589

10

5 W 8

6 1

2

3

4

5

6

7

8

9

1

0

Cell nunber Figure 6. Variations in the response of the accordion electromagnetic calorimeter to 10 GeV/c electrons (simulation).

the energy resolution and uniformity of the response with a grossly simplified accordion calorimeter module (no light collection effects, mechanical tolerances are all set to zero, and no shape optimization). Figure 4 illustrates the representation of the accordion geometry in GEANTS. Also shown are superimposed electron tracks from two electromagnetic showers each of 10 GeV energy. The distribution of the shower energies sampled by the calorimeter (corrected for the calorimeter sampling fraction) is shown in Figure 5 . It has a width corresponding to 17%/@ energy resolution, and it is perfectly symmetrical with no tails. No attempt was made either in the prototype design or in the simulation to reduce the variations in the sampling fraction of the calorimeter for particles crossing it at a different impact points with respect to the cell boundaries. One way of accomplishing this would be to add a second harmonic in the modulation of the waves sufficient to avoid the wave crests lying on a common radii. The response variations predicted by simulation are of the order of 8% (maximum to minimum) with a period equal to the projection of the cell size on the calorimeter front surface (Figure 6 ) . We are certain that optimizing the cell geometry will greatly reduce this response nonuniformity.

-

4. Conclusions and further prospects

We have presented the current status of the development of a Pb-Sc electromagnetic calorimeter with an accordion geometry. A prototype calorimeter

590

with this new geometry was successfully built by TECHNOPLAST in Vladimir (Russia), and is now installed in the X7 test beam area at CERN. Results of the test beam measurements will be available soon. A very limited number of identical components used in the prototype construction made the detector easy to construct and fit well into industrial techniques of mass scale production. We developed a GEANTS based framework for simulation of the electromagnetic showers in the Pb-Sc accordion calorimeter. The energy resolution and nonuniformities predicted from the first simulation runs are consistent with expectations for this kind of structure. The parameters of the simulation will be fine tuned using data from the first test-beam exposure of the prototype calorimeter. We are planning to rely heavily on the simulation and further test-beam studies for the future optimization of calorimeter design. Acknowledgments

We thank the staff at ITEP engineering department and at TECHNOPLAST for their vital contribution. We thank A.Soldatov and 1.Shein (both from IHEP, Protvino, Russia) for the help with ideas and prototype design. We thank S.White (BNL) for presenting this talk to this Conference. We acknowledge support from the Department of Energy (U.S.A.) and RMS (Russia). References 1. Murzin V.S. 1967. Progress in Elementary Particle and Cosmic-Ray Physics, Vol

IX,p.247. Amsterdam: North-Holland. 2. R.Wigmans, Calorimetry Energy Measurements in Particle Physics, 0-19-8502966 (2000). 3. M.G.Albrow et al., Nucl. Instr. Meth. A256 (1987) 23. 4. E.Kistenev et al., "PHENIX PbSc Electromagnetic Calorimeter: Results of Test Beam Studies", Proceedings of the Fifth International Conference on Calorimetry in High Energy Physics, World Scientific (1994) 211-223. 5. STAR Collaboration, The STAR EMC Technical Design Report, May 10, 1998 6. B.Aubert et al., Nucl. Instr. And Meth. A309 (1991) 438. 7. B.Aubert et al., Nucl. Instr. And Meth. A 321 (1992) 467. 8. A.V.Inyakin et al., Beam tests of the bayan electromagnetic calorimeter, Preprint IHEP 93-157 (1993). 9. P.Nevski, ATLSIM, generic PAW-based framework to run ATLAS detector simulations using GEANT3, ATLAS computing webpage, 1998 10. VSemenov, Proc. Of the IX Conference on Scintillators, p.86, Kharkov 1986. 11. CERN/LHCC/96-41, ATLAS TDR 2, 15 December 1996.

THE TILE HCAL CALORIMETER FOR THE TESLA DETECTOR, A STATUS REPORT

VOLKER KORBEL Deutsches Elektronensynchrotron DESY, Notkestr. 85, 0-22603 Hamburg, Germany E-mail: korbel%mail.desy.de The TESLA linear collider detector is described in a short overview and his unique feature of optimised particle flow measurement is underlined. A review of the technical concept, design and readout of the hadronic tile calorimeter HCAL is followed by a presentation of some actual R&D results for the optimisation of the scintillator tile wavelength shifter system. Concluded is with a discussion on further efforts and plans for a lm3 HCAL prototype for beam tests and calibration studies.

1. Particle Flow

At TESLA the collisions of point-like leptons produce unique final states with minimal underlying background. The strong and homogeneous 4 T magnetic field covering the detector volume allows excellent momentum resolution for charged tracks and bends them, in the dense, pencil-like jets, away from the neutrals, improving such their separation and individual cluster reconstruction inside the calorimeter volume. Momenta and impact angles of the charged particles are measured in the large TPC. In the calorimeters (ECAL and HCAL) the shower pattern of such identified charged particles as confined in a cone along the particle path is reconstructed and subtracted from the global pattern observed. The remaining clusters measured with the calorimeter resolution only are assigned to photons (in the ECAL) or neutral hadrons (mainly in the HCAL). Such a particle flow reconstruction allows energy and angle measurements superior to pure calorimetric measurements and results in an excellent mass determination of the produced new heavy particles. More detailed descriptions of the intended procedures are described in reference1, 2 . 2. The TESLA detector The inner detector, optimised for vertex- and 4-momentum measurement of the charged particles, contains a Si-CCD vertex detector, many layers of intermediate Si-pad trackers and a large volume TPC, Figure 1.

591

592

J

Figure 1. The layout of the complete TESLA detector

The tracking system is surrounded by the calorimeter assembly with the combined ECAL and HCAL segments split longitudinally in 2 barrel half cylinders. The ECAL with 20x0 depth (- 1.1X) is a Si-W sandwich sampling calorimeter with very high granularity, see H. Videau’s talk3. For the HCAL two different options are proposed and studies are underway to get the optimal particle flow reconstruction. The first option, a stainless steel/scintillator-tile sandwich with 38 active layers and 1/4 million of active calorimeter cells represents a further development from existing modern tile calorimeters as in construction for the CMS and ATLAS collaborations at the LHC, but with significant smaller tile sizes and calorimeter cell volumes. Such an approach favors optimal cluster separation and the possibility of improved e/n compensation by software weighting. The second HCAL option aims at a digital measurement of the shower clusters within 38 barrel sampling layers also but with pad sizes as small as lcm2. To compete with the energy resolution of the Tile-HCAL about 30 million pads have to be read out. Presently RPCs with single pad sensors are favored. The calorimeter barrel is closed hermetically by 2 end-cap disks with similar ECAL and HCAL structure. The calorimeters are enclosed by a 1.8X thick magnet coil. Shower frac-

-

-

-

-

593

-

-

-

tions leaking through the 5.5X inner calorimeter and the coil will be identified and measured - with a moderate energy resolution of l O O % / a - in the 6X deep tail catcher and muon tracking system assembled from up to 8 layers of 10 cm Fe interspersed with scintillator and/or RPC detector layers.

3. The Tile-HCAL

3.1. Structure and mechanical design The status of design and R&D from summer 2001 can be found in reference4. The basic structure implies a 4.4 X deep calorimeter stack assembled from 38 stainless steel plates, 20 mm thick each, but the ability of brass, W and Pb plates will also be investigated in further simulation studies. As detector layers serve 5 mm thick plastic scintillator tile plates in between. At present a solution with tile sizes of 5 x 5 m 2 enlarging with calorimeter depth to 16 x 1 6 m 2 is studied. The calorimeter cell depth increases from 3 summed tile layers (in the first 3 segments) to 8 layers (in the last of 9 HCAL segments). The longitudinal arrangement of cells is roughly vertex pointing, but more effort is put on choosing an acceptable fine 3-dimensional granularity. The barrel wheels are assembled from 16 single wedge shaped HCAL modules Figure 2, optimised in construction to have a minimum of leaking holes or additional dead material and spacers for the sensor layers. Two assembled HCAL wedge modules carry a single ECAL module in front. The HCAL end-cap section has more sandwich layers (45-53) to improve the containment (from 4.4X to 5.7X) for the higher jet energies expected in that forward range. The following inner flux return yoke with 4X depth is instrumented with 4 additional tile layers also. More details can be found in the TDR5 and the special LC-notes'.

-

3.2. Signal read out and data acquisition

The scintillation light produced in the tiles enters in the attached wavelength shifter (WLS) fibres (actually a diameter of 1 mm is favored) where it is absorbed and re-emitted as green light with a maximum around 500 nm wavelength. About 5% percent of the emitted green WLS light is - by total reflection - accepted in the double clad WLS fibre. Half of this light is emitted in backward direction and has to be recovered by reflection on an appropriate reflector to be deposed on the downstream WLS fibre end. The collected light is guided to sensitive photo-detectors outside the calorimeter volume at a distance between 2 to 5 m. To minimise the light losses the WLS fibres with attenuation length Xatt 2.4m (Kuraray Yll(200)) are connected to double clad clear

-

594

Figure 2. Cut through the barrel calorimeter with charged particle trajectories, the module arrangement and sandwich structure.

optical fibres with Xatt > 8m. In front of the photo-detectors the light of all tiles of a cell is summed in light mixers. Different types of photo-detectors are studied for optimal photocathode efficiency, gain, minimal noise and large signal to noise distance. The cost/channel, pixel sizes, power consumption and integration possibility for large quantities have to be taken into account. Possible choices could be: rn

rn

0

Avalanche Photo Diode Arrays or Silicon Photomultiplier Arrays (Si-PM or Metal-Resistor Photodiodes) which can both operate in large magnetic fields and Multi-anode photomultiplier arrays or CCD single photon counters which operate outside the field.

-

Modern avalanche photodiodes have a photo-conversion efficiency of 80% and a gain of 200-500, whereas Si- and Multi-anode PMs offer, with a reduced conversion efficiency of 12 - IS%, gains of up to lo5 to lo6. Efforts are underway to improve the Si-PM conversion efficiency towards 30%. The photodetectors should be linear in the signal range from MIPS (- 30 MeV), which are used for online calibration, up to 15 GeV shower energy as deposited

-

-

595

-

maximally in a single cell. This requires a dynamic range of 500. For larger signals amplitude measurement with reduced precision should be possible. The photodetector signals are amplified in charge-sensitive preamplifiers with signal shapers t o about 50 mV/GeV. Shaping times 5 150 ns are required to avoid pile up from the next bunch crossing. For the signal digitalisation ADCs or Flash-ADCs can be used. For ADCs a dynamical range of 12 bits is acceptable if different signals with gain difference of 4 are produced by the preamplifiers. Signal Sample&Hold followed by multiplexed readout will allow a further reduction of the required number of ADCs. With Flash-ADCs of 60-80 MHz sampling, enough samples for each signal are available to measure the amplitude, to subtract the actual pedestal from the previous bunch and to reconstruct the event time with a precision better than 5 ns. This allows the rejection of cosmic background by a factor of 50 beyond the rejection achieved with vertex reconstruction and helps significantly in the search for rare SUSY events. For each cell a single Flash-ADC with 12 bit range is required, making such a solution probably more expensive. The digitised signals will be fed to pipelines from where they are taken after the bunch train has passed and when a gap of 200ms is available for software trigger application, re-calibration, data compression and read-out. N

-

3.3. Calibration and monitoring

The precision of the online calibration of all individual calorimeter cells dominates the constant term contribution in the energy resolution. All 250 000 calorimeter cells can be calibrated with cosmic MIPS (barrel part) or beam halo muons (end caps) to a precision of 5 2%. The final energy resolution will than be limited by the sandwich sampling fluctuations. To profit from such a precision of calibration, the lateral non-uniformity across the tiles has to be smaller than 5% and the relative light yield of the individual tiles assembled to cells has to be equalized within similar limits. The gain and dynamic range of all channels will be continuously monitored by injection of LED light pulses in the front window of the photo-detectors. These measurements can be carried out in the large time gap between the beam bunch trains. Sufficient statistical calibration precision can be achieved in about 24 hours. N

-

4. Actual R&D studies

In autumn 2001, after the presentation of the TDR5 a detailed R&D program has been initiated by the CALICE collaboration in order to optimise the components of the Tile HCAL detector and to study the alternative digital HCAL version and has been presented to the DESY Physics Research Committee7.

596

The status of these studies together with a comparison of the physical performance of both HCAL options will be reviewed in autumn 2002. A joint effort of Czech, German and Russian Research Institutes* has been started and concentrates first on: -Study of appropriate absorber material for the calorimeter structure, -Search for cheep plastic scintillators with sufficient light yield, finally 3000m2 of such material will be needed, -Selection of appropriate WLS-fibres, -The best geometrical configuration of the tile-WLS-fibre coupling, -Optimal reflectors for the tiles and the open WLS-fibre end, -Ageing stability of the selected tile-fibre system, -The choice of photo-detectors capable to operate in the strong magnetic field, -The study and development of preamplifiers and signal shapers, -The selection of ADCs or Flash-ADCs for digital signal conversion and -The specification of the buffering, multiplexing and DAQ system. In the following some new results from the tile-fibre system studies are discussed. A part of the fibre R&D could profit from previous CMS studiesg. The light yield of scintillator tiles of 5 x 5 m 2 size was measured with WLS-fibres attached along a tile side. The tiles and fibres were wrapped with TYVEK paper. The best yield, namely 7.5 photoelectrons from the photocathode of a photomultiplier with 12% efficiency were found with a Bicron BC-408 scintillator viewed by a 1 mm diameter Yll(200) double clad WLS-fibre from Kuraray as shown in Table 1. N

-

Table 1. Light yield studies for various tile material and WLS-fibres. WLS-fibreltile material: B C F - 92M, 0.8mm

Vladimir

Y 1 l(200)

2.47

2.06

BC-416

1

BC-408

Protvino 2.98

B C F - 91A,0.8mm

3.21

3.53

Y11(200), 0.6mm

2.13

2.44

2.50

Y11(200), 0.8mm

3.33

3.65

3.63

Y11(200), 1.Omm

3.75

4.08

4.00

4.58 2.81 6.57

4.63 5.19

-

A cheaper solution would be the use of Russian scintillators as e.g. polysty70% of BC-408. role based scintillator from Protvino with a light yield of Light yield improvement studies for such and other plastic scintillator materials are continuing. Optimal geometrical tile-fibre coupling was studied with WLS-fibres arranged in various geometrical shapes, directly attached to the tile surface or

inserted in grooves, Figure 3. Some results for light yield with air coupled WLS-fibres are shown in Table 2.

Figure 3.

Various geometrical arrangements for study of optimal tile fibre arrangements.

Table 2. Light yield WLS-fibre configuration lightyield

meas.error

1

2

3

4

70.1 1.3

5

6

72.2

67.1

1.6

1.7

7

8

81.0

83.3

96.3

100

2.0

1.6

2.2

2.6

9

1

0

38.6

72.6

82.4

0.9

1.5

1.8

The best results from 10 different configurations were found with diagonal grooves, straight or bent to circle shape with radius of the tile size, both nearly equal within the measurement errors. A straight WLS-fibre in a deep groove in the center of the tile still gives a yield of 82%, straight fibre read out without a groove along the small tile edge yields 70% of the optimum. With better, 20% more light optimised reflector (TYVEK, "Synthetical special paper") was caught. A new super-reflector foil (3M, "Radiant Mirror Film") allowed a further increase of 30%. Simulation results for the tile-fibre system point to a further increase in light yield when the air coupling of the WLS-fibre to the tile is replaced by gluing. Such studies are going on. At present from a cell with 3 tiles (5 x 5 x 0 . 5 ~ 250 photons are seen at the photodetector window, sufficient to allow a good MIP/noise separation. Further factor 2 improvement in light collection seems possible. The spread in lateral uniformity observed for the various configurations is between f4to7%, best uniformity was found for configuration 1. Also WLS-fibre loops with radius as small as 2.4 cm can be used when embedded in machined tile grooves. A maximal light yield is observed for 2 WLS-loops, but

-

N

598

-

the yield as measured in 2 institutes differs by a factor of 1.5. More detailed studies are needed to explain the different results. Possibly, the fibres with lower response were from different quality production badges or were damaged during bending. Ageing studies for such loops are prepared.

5. The future R&D program 5.1. The ”Minical” test array

The actual R&D program has to be continued to study the light yield, stability of response, cross talk to neighbored tiles and ageing of a larger sample of similar tile-fibre arrangements. Thus a small calorimeter test stack has been built with 27 sandwich layers where a larger number of tiles of various size can be inserted and assembled and continuously tested with LED light injection, cosmic muons and electromagnetic showers from a test beam at DESY. Such a structure, called “Minical” test array, is shown in Figure 4.

1

I

Figure 4. T h e layer and cell structure of the “Minical” test array.

The larger tiles in the upper and lower part of the “Minical” are used as triggers for muons. In between an appropriate arrangement of individual tiles or machined multi-tile plates can be inserted resulting in a total of up to 165

599

individual tile-fibre elements or 55 cells with 3 tiles each. Also a central larger scintillator-plate with 2-dimensional strip read out can be inserted and will allow localization of the particle trajectory within a grid with lcm2 size for uniformity studies. The WLS-fibres ( w 50cm long) will catch the light and guide it to the photo-detectors installed nearby. At the beginning 4 HAMAMATSU multi-anode PMs, with 4x4 anodes each, will be used to read the signals. Later they will be replaced by 2 HAMAMATSU 4 x 8 Si-APD matrix arrays and/or new Si-PMs from the development carried out in 2 contributing Russian Institutes. The low rate of useful cosmic muons, with 0.1 Hz for a 5 x 5cm2 tile, requires continuous data-taking with the “Minical” for some weeks. The shower measurements in the electron test beam probe the linearity along the required dynamic range, the cross talk between tiles and cells and last not least the energy resolution acheavable. Later measurements are scheduled to study the change in response of the scintillatorWLS system in strong magnetic fields (up to 4 Tesla) and the widening of the shower shapes (up to 1.2 Tesla).

-

5.2. The prototype HCAL

The actual expected HCAL performance in identification and separation of clusters has to be proven by prototype studies in a test beam for a couple of hadron energies and angles. To contain the high energy showers such a prototype needs an active sandwich volume of lx1x1.1m3 as shown in Figure 5a.

-

k

30 cm

Figure 5. A 3-D view and possible tile arrangement for a full size realistic prototype of the HCAL for hadronic beam tests.

Showers leaking through the 1.8 X deep coil and cryostat system have to

600

be recovered by an appropriate tail catcher prototype. First simulation and reconstruction studies were made to find, along the HCAL depth, the optimal geometric arrangement of a minimal number of cells with appropriate tile sizes. and suggest a minimum of 1200 cells with tiles distributed as indicated in Figure 5b. After some HCAL stand-alone measurements a CALICE Si-W ECAL prototype has to be placed in front. The final design, construction and assembly of such a prototype HCAL for the insertion of analogue and/or digital detector layers, the set up of the. appropriate readout and the installation and operation of a realistic leakage detector system is a large effort. Also more studies are needed for optimisation of size, number and arrangement of the calorimeter cells to improve the particle flow measurement as well as the calorimetric energy resolution by e / r software compensation (“energy weighting”, Hl). In addition the required beam particles, beam energies and angle settings have to be specified. Given this enormous effort and the also limited access to appropriate high energy particle test beams some agreement was reached that such an activity is unique for all actual LC calorimeter proposals and it was proposed that prototyping studies, operation in beam tests, data acquisition and analysis should be envisaged as a common worldwide project.

-

6. Concluding remarks

The concept of the TESLA Tile-HCAL is based on the realisation of a dense arrangement of fine granular cells with WLS and clear fibre read out to sensitive photo-detectors outside the calorimeter volume which are insensitive to the strong magnetic field. Present F&D results prove that the number of photons read out from the Tile-HCAL cells will be > 250, thus avalanche photodiodes with intrinsic gain of 200 - 500 could be used and online calibration of all cells with muons from cosmic rays and halo is possible. The properties of new competing Si-PM detectors will also be investigated. A small “Minical” prototype with up to 64 channels is prepared for system studies of the performance of tile-fibre readout and photo-detectors and has started its operation at DESY. A larger HCAL prototype with 1200 cells containing hadronic showers in the final jet energy range is proposed to proof the particle flow reconstruction concept. The large effort needed for hardware and software preparation, simulation and reconstruction studies and the limited future access to high energy test beams require a common, worldwide effort.

-

-

601

References 1. Ray Frey, invited talks at this conference 2. Vasily Morgunov, invited talks at this conference 3. Henry Videau, presentation at this conference 4. V. Morgunov and V. Korbel, Snowmass 2001 Conference, http://snowmass2001.org, conference proceedings to be published, see: http://www.slac.stanford.edu/econf/C010630/forweb/E3l0~morgunovl.pdf and: http://www.slac.stanford.edu/econf/C010630/forweb/E315~morgunov.pdf. 5. TESLA, Technical Design Report, Part IV, A Detector for TESLA, DESY 2001-011 and ECFA-2001-209, 2001 6. P. Hassler, V. Korbel et al, “The Hadronic Tile calorimeter for the TESLA Detector, Design, construction and Installation”, pages 2108-2130, V. Korbel et al, “Upcoming R&D, Design and Construction Studies for the HCAL Tile Calorimeter HCAL” , pages 2193-2202, V. Korbel and V. Morgunov, “Calibration and Monitoring”, pages 2203-2214 and V. Korbel, “The Optical Read Out of the HCAL Tile calorimeter”, pages 22152221, all in Physics and Experimentation at a Linear Electron-Positron Collader’ DESY 01123F, vol. 3, (2001). 7. “A calorimeter for the e+ e- TESLA detector”, Proposal for RBD, PRC R&D-01/02, and “Memorandum to the PRC”, DESY, Hamburg, 2001 8. contributing institutes: Czech Institutes: -Charles University, Prague --Institute of Physics, Academy of Science of the Czech Republic, Prague German Institute: --Deutsches Elektronen Synchrotron, DESY, Hamburg Russian Institutes: -Institute of Theoretical and Experimental Physics, ITEP, Moscow -Institute of Nuclear Power Engineering, INPE, Moscow, -Lebedev Physics Institute, LPI, Moscow --Moscow Engineering and Physics Institute, MEPHI, Moscow 9. H. S. Butt et al, “Fiber R and d for the CMS HCAL”, SCIFI Conference , Notre Dame, Indiana, 1997

This page intentionally left blank

Electronics Covener: J . Elias

J. Elias

Covener’s Report

F. Martin

The ATLAS Tile Calorimeter Front End Electronics

E. Ferrer Ribas

Overview of Liquid Argon Front End Electronics

*s.Los

First Results with the QIE8 ASIC

A. Go

Front-End Electronics for CMS Preshower Detector

D. Breton

The Front-end Electronics for the LHCb Calorimeters

C. Nelson

Front-end Electronics Upgrade for the CDF Calorimeters

V. Boudry

The Electronics of the New H1 Luminosity System

I.G. Eschrich

The BaBar Electromagnetic Calorimeter in its Third Year of Operation

D. Hoffmann

H1 Calorimeter DAQ Upgrade for HERA-I1

*Written contribution not received

603

This page intentionally left blank

ELECTRONICS

J. ELIAS Fermilab, Batavia, I L 60510, USA E-mail: [email protected] (Convener’s Report)

Even though front-end electronics is what typically comes to mind, calorimeter electronics actually comprises a chain of signal processing, transmission and digital data storage systems. Each piece of the chain has its own unique challenges and unique requirements. In this 10th conference on Calorimetry in high energy physics, there were six contributions on front-end electronics, one on data acquisition, and two on design and performance of the entire electronics chain. Ivo Eschrich’s report on the first three years of operation of the BaBar electromagnetic calorimeter at PEP-11, we clearly saw the effort needed to make the transition from a few hundred channels working in a test beam to the full system working at high luminosity. This contribution should be required reading for those developers reflecting hubristically on their test beam triumphs. In a different vein, Vincent Boudray ’s discussion of the electronics chain for the luminosity calorimeter in the H1 experiment at HERA brought out clearly the complexity of present day systems. Calorimeter electronic systems have a natural useful lifetime of a decade or so. Upgrades and even wholesale replacements are necessary either because technological obsolescence makes them difficult , if not impossible, to maintain or accelerator performance improvements vitiates their utility. Both factors were at work in the CDF experiment at Fermilab as described in Charles Nelson’s report on the design and development of a multi-range pipelined frontend electronics system to replace the original sample-and-hold electronics. Responding to a similar situation, Dirk Hoffmann presented the recently upgraded data acquisition system for the H1 calorimeters at HERA. Another trend was evident here; the new system makes extensive use of commercial data processing equipment as opposed to the custom VME or FASTBUS equipment commonly deployed 10 to 15 years ago in high energy physics experiments. Calorimeter front-end electronics must meet some very stringent perfor-

605

606

mance requirements, more severe than for say wire chamber or tracker electronics. The combination of wide dynamic range, high speed, and extremely low noise needed presents a formidable challenge to the designer. In addition, there are the issues of precision calibration, radiation hardness, and uniformity of response over the whole device, that impact the design. In the presentations on some of the front-end electronics systems for the LHC detectors, 0 0 0

0

ATLAS hadronic calorimeter by fianck Martin, ATLAS electromagnetic calorimeter by Esther Ribas, CMS hadronic calorimeter by Sergey Los, and CMS preshower detector by Apollo Go

it is evident that custom application specific integrated circuits form the core of todays solutions. Two very different approaches are apparent. In one, the signals are amplified, shaped and stored in a switched capacitor array pipeline at the beam crossing frequency and only digitized on receipt of a first level trigger. In the other, the signals are amplified, shaped and digitized with a multi-range technique at the beam crossing frequency and stored in a digital pipeline. In all cases reported, the performance specifications (as dictated by the physics objectives) were met by the design teams. In future editions of this conference we shall see how these successes translate into systems with tens of thousands of channels.

THE ATLAS TILE CALORIMETER FRONT END ELECTRONICS

F. MARTIN LPC Clermont-Ferrand, IN2P3-CNRS, FRANCE E-mail: martinOclermont.in2p3.fr

After a short description of the ATLAS^ tile calorimeter front end electronics, the quality control procedure is presented. It is required both to ensure that the electronics match the ATLASrequirements and to face the complexity of any maintenance access in ATLAS.The test benches dedicated to tests of more than 10000 photomultipliers and all 256 entire electronics modules are described, and some results about the radiation hardness are given.

1. Introduction

Tile calorimeter modules2 are made of a steel structure (passive absorber) in which 3mm thick scintillating tiles pointing radially to the beam line (active medium) are inserted. Emitted light is collected on tile’s edges using wavelenght shifting fibres. Set of fibres are grouped in bundles to define cells, and connected onto a photomultiplier (PMT) through a light mixer to insure a very good light uniformity on the photocathode. The front end electronics is supported by a structure called ’super-drawer’ (SD), inserted on the external edge of each module. There are 2x64 extended barrel (EB) modules (0.8< <1.7) and 128 barrel modules (17 <1) in the whole calorimeter. 32 PMTs are needed for an EB, and 45 for a barrel. About 460000 tiles, 64000 fibres and 10000 channels of front end readout electronics are then needed for this detector. This contribution focus on the SD and PMT tests. 2. The front end electronics: requirements and description The design and performances of the front end electronics readout are presented in elsewhere4. At LHC,the proton bunches collide at a 40 MHz frequency. The front end electronics must provide pulses of 50ns FWHM, with a rise time of 20ns or less. Sampled data must be stored in pipeline memories for 2 . 5 ~ until 3 the LVLl decision. If the event is triggered, data are formatted and send to ROB through an optical link.

607

608

-

-

Dynamics of 16 bits are needed, because energy deposition varies from 30 MeV to 2 TeV per cell. As more than 50 pe/GeV are collected, the maximal signal is of the order of 800 pC. This leads to the choice of two gains, with an amplification ratio of 64. A precision of 10 bits is sufficient. The dynamics for trigger cards is also 10 bits. The non-linearity of the full front end "electronics chain must be less than 2%, to avoid simulating some fake compositness effect for example3. The rise time of the light emission in the tile is about 5.511s. The chosen PMT is the R5900 from Hamamatsu. It fits all the ATLASTilecal conditions: small size (26x26 mm2, 20mm height), low dark current (100pA at 680V), quick answer, small sensitivity to magnetic field (less than 1%gain variation for a 500 Gauss magnetic fiel) and temperature variation (0.25%/"C) and light sensitivity in the [300,65O]nm window, with the best response at X=420nm (X=480nm in ATLAS). The bialkali thickness on the dynodes has been optimized to minimize the PMT drift, as well as the number of dynodes (8) and the high voltage distribution to maximize the P MT linearity'. Each of the 10000 PMT is submitted to 3 steps of tests6: the PMT alone, the PMT with its divider, the PMT with the divider and its 3inl board7. This board is directly plugged onto the divider, to reduce cabling and noise. PMT, divider and 3inl board are protected by a shielding tube. Shaping and amplification, charge injection (precision better than 0.4%) and integration (encoded in a 12 bits ADClO,precision -0.4% for C S ' source ~ ~ calibration) are done by these cards. The power dissipation of all the 3inl in a SD is 70W. Low and/or high gains are digitized (encoded with 2 x 10 bits ADC per PMT) and stored in pipeline memories onto the digitizer* until the LVLl decision. Maximum signal for the high gain is 800/64pC, which gives 12.2fC per ADC counts. Low gain covers the entire range. The noise is at the level of 1.3 ADC counts and 0.6 ADC counts for high and low gain respectively. Linearity is well below the goal of 1%. The analog low gain signal of up to 5 cells are linearly summed by the adder cardg, and this information is used in the LVLl trigger. For muon, only the D-cell (the more external, see figure 3) are used. If the LVLl decision is positive, data are sent to the interface board (IB), formatted and transmitted to ROB through optical fibres. The IB also receives the T T C signal (timing and control), and transmits it to the mother board. Orders can be tranmitted through TTC, or through Canbus.

-

-

3. High voltage regulation, tests of the PMTs.

The design and performances of the high voltage (HV) regulation system are described in reference14. A bus board for HV distribution runs along the drawer. Orders and control commands are treated by the microcontroller (one

HV-MICRO/SD). Two HV-OPT0 boards/SD supports the 48 HV regulation loops. The power dissipation in one SD is -35W. A regulation at the level of 0.4V is required, to stabilize the PMT gain at 0.5%. All the boards are individually tested (functionality tests, burn-in, new functionality tests, HVO P T 0 calibration). Concerning PMTs, a full review can be found in reference6. PMT must reach the collection efficiency plateau before lOOV, and 90% of it before 50V. The dark current (DC) must be less than 2nA at 800V, and less than 8nA at 9OOV (more precisely, 5 4nA or 5 10DC800). Gain must be lo5, with a nominal high voltage in the [600,8OO]V window. Drift must be less than 2% for an anode current of 3pA (figure 1). The quantum efficiency (QE) must be greater than 15%. For an anode current of 50mA, the non-linearity must be less than 2% (figure 2). If this criteria is not fullfilled, the associated divider is changed. About 98% of the PMTs are accepted, the main cause of rejection being the DC and the drift. There are also some criteria on the mean value for PMT batches: QE? IS%, DC<250pA at 800V, drift and non linearity less than 1%.

4. Readout electronics: radiation tests; adder boards.

All the cards are individually tested, submitted to burn-in (one week at 65"), and tested again. Radiation tests are also performed. Typical simulated levels of radiationll for 10 years in ATLAS^^ are 8.5838 to 5.69310 p.cmV2 (Single Event Effect, SEE), 2E10 to 2.3Ell lMeV n cm-2 (Non Ionizing Energy Loss, NIEL, with a neutron spectrum from 100 KeV to -2 MeV), 2.5E-1 to 2.6 Gy (Total Ionizing Dose, TID) at the level of the IB and of the most exposed 3inl board respectively. Safety factors are applied, to take into account simulation uncertainty (5 for SEE, NIEL and 3.5 for TID), the number of components/cards tested (4) and that the actual exposure is at a much lower dose rate than is practical for testing (5 for bipolar components/TID). Most of the tests have been carried out. Board affected by radiation are the 3inl card, the digitizer and the IB. Only the 3inl calibration function is affected by SEE, and a power cycling is required to cure the faults. Digitizer and IB are affected by SEE but at an acceptable rate, and errors are cured by power cycling. On IB, NIEL produce errors, but again at an acceptable rate. All the boards support the TID. Response of the adder cards has been checked with a muon beam at ~ = 1 . 0 5 (EB part, fig. 3). Good signal efficiency is observed in the EB, but extrapolation to the whole calorimeter13 leads to a small efficiency (N SO%, fig. 4) in the barrel. Consequently, adder design has been slightly modified before final

-

610

0

20

60

40

80

0

100

20

80

60

40

Photocathode Voltage (V) (a)

100

Photocathode Voltage (V) (b) 3

v

8

2.5

o.l 0.08

-

............... ,...............................

t-.

t:

~ . . , ~. ~

~.

~~

~

~

O.

~

~

~ 2,

~

~

................

. ..

I

, I

:

.

:~ , : , : A , . . . .................. ,......,........,.. ~

~

,

0.06

I .5

0.04

I

~

,

: ,

.............

~

,;* . ,

I I

0.02

0.5

I I

0

500

I 1 1 1 1 1 1 1 1 1 / I I I I / I I I

600

700

800

,

:, ... ~~.~~~~~

~~

.......

:..

..

I_:............... :,.....<

! ......; ............... ;..

5

15

--*-

0

900

HV (V) (C)

-

E

1

c

0.5

0

-0.5

-1

0

5

10

15

20

0

10

Time (h) (e)

20

Time (h)

(0

Figure 1. PMT test bench, stepl. Top plots: collection efficiency as a function of the photocathode voltage, a good (left) and a bad (right, no plateau reached) PMT are shown. Middle plots: dark current (left) and ratio of G / G , , , (right) as a function of the HV. Bottom plots: PMT on/off drift as a function of the time (3 periods of 3 hours lighting, separated by 2 periods of 3h without light. Drift is calculated wrt the first point). A good (left) and a bad (right) PMT are shown.

production to increase the ratio S/N for the p output, and the high voltage of the D-cell PMT is let to its nominal value, instead of being ajusted during Cs

61 1

0

500

1500

loo0

0

ZOtUI

20

40

60

80

PMT anode current (mA) (b)

PMT anode current (ADC counts) (a)

Figure 2. PMT test bench, step2: the PMT is now associated to its divider. Left plot shows linearity and right plot deviation from linearity, as a function of the anode current.

TILECAL CELLS

$ 1 2

6

3.0

2.5 2.0 I

I

I

I

I

I

I

I

I

I

I

I

I

0.

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

5.5

6.0

Figure 3.

Figure 4. Signal efficiency for the muon, as a function of 7. Cell structure in the ATLAStile calorimeter. cellstructure is clearly seen,

calibration (where HV is set to match a response of 1.2pC/GeV, which leads to a mean AVz-30V wrt the nominal HV). 5. Tests of the Super-Drawers (SD).

The SD testing is divided in four steps. First, SD configuration is checked, as six different configurations are possible in ATLAS,depending of the SD position in the detector. Bar code readers allow to know which PMTs are mounted on a SD, and a cross-check is done with the PMT database information^'^. Next, a short test of the electronics is done: all the connection and currents are checked and card functionnalities are controlled. Pulsed and continuous light or CIS are used to control all the chain from the light emission to the data. In the third step, stability of the electronics on a longer time scale (- 12h) is checked.

612

The cooling system (depressurised water at 18'C) is also controlled. End, the temperature probes are calibrated. A quality sheet associated to each certified SD allows to trace back the history of the SD in case of failure in ATLAS.

6. Conclusion The ATLAStile calorimeter enters now in the construction phase. About 75%of the modules (steel and scintillating tiles) are in~trumented'~. The production of the SD just starts. Dividers, 3inl boards, bus board, HV-MICRO and ADC integrator production are done. 80% of the PMTs have been provided by Hamamatsu, and 50% have been tested. Half of the HV-OPT0 cards are ready. Digitizer, IB and adder final production starts. Almost all the cards have been submitted to TID, SEE and NIEL radiation tests, and have been prooved to support the radiation levels in ATLAStilecal environment. First modules have been calibrated16, and the full calorimeter must be assembled ends 2003.

References 1. ATLASdetector and Physics performance TDR, CERN/LHCC/99-14, Vol I1

(1999). http://atlasinfo.cern.ch:8O/Atlas/Welcome.html 2. ATLAStile calorimeter TDR, CERN/LHCC/96-42 (1996) 3. P. Brette, Ph.D thesis, Clermont-Ferrand, 1996. A non linearity of 2% fakes compositness at a 20 TeV scale, observed at a 4 TeV PT. 4. K.Anderson e t al.) Rome LEB paper, http://ifae-cl.ifae.es/Tilecal-Electronics/ 5. C. Hebrard, Ph.D thesis, Clermont-Ferrand, 1999. 6. R.Lefevre, Ph. D thesis, Clermont-Ferrand, 2001. 7. http://hep.uchicago.edu/atlas/electr/electronics.html 8. S.Berglund et al., 'The ATLAS tile calorimeter digitizer', CERN-OPEN-2000056, CERN LHCC 99-033. http://www.physto.se/-ker/designreview/dr.html 9. A. Santiago Cerqueira, 'Analog Trigger Towers for the Tilecal: Radiation Tolerance and Production Tests', talk at CALOR2000, Annecy, France. http://wwwlapp.in2p3.fr/Calor2OOO/http://montreal.lps.ufrj.br/Ncern/ 10. http://ifae-cl.ifae.es/Tilecal-Electronics/ 11. http://atlasinfo.cern.ch/Atlas/SUB_DETECTORS/TILE/production/

electronics/radiation/rhawgnew.htmi 12. M.Shupe, http://atlas.web.cern.ch/Atlas/GROUPS/FRONTEND/radhard.htm

13. G.Usai, 'Muon analog signal in the Tile calorimeter, analysis of 2001 Test Beam data', ATL-COM-TILECAL-2001-011. 14. R.Chadelas e t ai., 'high voltage distributor system for the tile hadron calorimeter of the ATLAS detector', ATL-TILECAL-2000-003. 15. http://tilecal.in2p3.fr/pmt/accueil/accueil.php 16. S. Nemecek for I. Korolkov, these proceedings. 17. A. Henriques, these proceedings.

OVERVIEW OF LIQUID ARGON FRONT END ELECTRONICS

E. FERRER RIBAS DSM/Dapnia/SPP, CEA/Saclay, 91191 Gif-Sur- Yvette, FRANCE E-mail: esther.ferrer.ribasOcern.ch (for the ATLAS-LARG collaboration)

The ATLAS Liquid Argon Front End rad-tolerant electronics chain is described, and the main requirements that have led to today's architecture are discussed. Performance characteristics of the rad-soft prototype system, based on results obtained from about 6000 read-out channels installed in the test beam, are presented. During the past year, significant progress has been made in the transition to radtolerant electronics, which is based on a number of ASICs both in DMILL and DSM technology. Initial measurements on the prototype of the final version of the front-end board are presented.

1. Introduction The LHC environment sets strong constraints on the ATLAS calorimetry and on its electronics. The very high energy and luminosity of LHC require a large number of channels with a large dynamic range and operating at high frequency. The Liquid Argon technique is used for all electromagnetic calorimeters (electromagnetic barrel and end-cap calorimeters, presampler and forward calorimeter) and the hadronic end-cap calorimeters adding up to a total of -180000 electronic channels to be read'. Common electronics is used everywhere except the very front-end where the HEC uses cold preamplifiers. This standardisation minimizes design effort and will facilitate maintenance. Most of the elements of the system are now in the production phase and boards are ready to undergo the final tests. In this note, an overview as well as the status of the system are given. 2. Requirements The signal delivered on the detector electrodes is a triangular shaped current with a fast rise time (- Ins), decreasing linearly to zero at the end of the drift time of ionization electrons in liquid argon (- 450 ns in a 2 mm gap at 10 kV/cm electric field). A typical value of the current is 2.8 pA/GeV for the

613

614

EM calorimeters. This signal is delivered on the detector impedance which can be modeled, to a very good approximation, as a pure capacitance. The main requirements for the read-out electronics can be summarized as: 0

The dynamic range to be covered is as least of 16 bits as the energy to be measured can be as large as 3 TeV and the mean energy deposited in a single cell coming from pile-up interactions is of 50 MeV. The physics goals to be reached need a good energy resolution (for the electromagnetic calorimeters) e 0 . 7 %. In order to keep the constant term below 0.7 %, the contribution of the calibration to this term should be less than 0.25 % over the whole energy range. Since the energy of a cluster is the sum of many cells, the coherent noise over many cells should be kept small. It is required that the coherent noise per cell should be less than 5% of the level of the incoherent noise. Cells should be summed in trigger towers of granularity Arj x A+ = 0.1 x 0.1 and the result sent to the Level-1 processor. The read-out system should sample at 40 MHz; to this end a pipeline with a depth of at least 2.5 ps should be provided; in addition a large derandomizing buffer and a fast read-out should allow for a maximum Level-1 trigger rate of up to 75 kHz. Since the system will be located in a region of limited access, it needs to be highly reliable. In addition, although radiation levels at that location are not very large (1012n/cm2/yr; lo1' ionizing particles/cm2/yr) the electronics needs to be radiation tolerant. A number of ASICS (Application Specific Integrated Circuits) have been designed in DMILL and DSM rad-tolerant technologies. N

0

0

0

3. System Design

In Figure 1a schematic view of the front-end and read-out electronics is shown. This figure shows the logical flows of information and the different boards with the important components. The sensitive analog electronics is on detector, in a front-end crate (FEC) attached to the cold to warm feedthrough. This crate is on top of the warm feedthroughs, in the crack between the barrel and the end-cap to provide an extension of the cryostat Faraday cage. This should protect the readout electronics against external electromagnetic radiation and minimize pick-up noise. In addition, this location keeps the warm part of the signal and calibration cables to a minimum length, minimizing the associated attenuation and noise. The front-end crate houses the following type of boards: 0

The Front-end board amplifies and shapes the analog signals, sums

615

0 0

cells within layers to prepare the input for the tower builder board, stores the signals in analog memories waiting for the Level-1 trigger decision, digitizes the selected pulses and transmits the digital results on optical fibers. The Calibration board generates pulses with a precision of 0.2%; The Trigger Builder board does the final analog summation of the different layers in depth to form trigger towers and to transmit the analog signals to the Level-1 cavern for digitisation and processing. The Control board receives and distributes the 40 MHz clock, the Level-1 accept signal and other signals to configure and monitor the boards in the crate.

The limited space in the FEC constrains the access and power dissipation and cooling. As already said, radiation levels at that location are not very large however the electronics needs to be radiation tolerant. ATLAS has defined standard Radiation Tolerance Criteria (RTC) that all components need to pass in order to qualify. The criteria are defined taking into account the expected dose after 10 years of LHC running increased by a large safety factor. These doses in the FEC are the following: 0

0

0

NIEL (non ionizing energy loss) test with neutrons at the level of 1-3.6 1013n/cm2; TID (total ionising dose) test with X-rays with a dose of 50 to 330 krad, depending on the component to be tested; SEE (single event effects) at the level of 7.7 10"h/cm2/yr;

4. Fkont End System 4.1. Fmnt End Board

The Front End Board deals with 128 channels in 16 groups of 8 channels. Each group contains 2 four-channel preamplifiers, 2 four-channel shapers, 2 twelvechannel analog memories (SCA, Switched Capacitors Arrays) and one 12-bit ADC. The preamplifiers amplify signals above noise level. They need to accept the whole signal dynamic range (216 bits) and deal with high speed inputs requiring them to have a low input impedance. There are two types of preamplifiers. The warm bipolar hybrids preamplifiers (electromagnetic calorimeter) are more than 50% produced and the overall noise performance is en = 0.4 nV/&; in = 6 PA/&. For the cold Ga As monolithic HEC preamplifiers the production and testing is now completed giving a yield of 84%. The circuits have

616

! ExternalTrimen

E

a

rm98 -

CTP

I

LEVEL 1 PROCESSOF

E

TTC A

b

LEVEL 1 INTERFACE 4

32 Bits 40 MHz

A

- Calorimeter Monitoriq

( I

I

.

i

I I

I

Preamplifiers

I

I

I

I

I

Y

In

7

Cryostat

I

Figure 1. Block diagram of the front-end electronics. The drawing shows warm preamplifiers in the front-end board which are present for the electromagnetic and forward calorimeters. For the hadronic end-cap calorimeters, the preamplifiers are located in the liquid argon.

617

been tested up to 3 kGy and 1014 N. The AMS bipolar shapers (rad tolerant by design) limit the system bandwidth to match the 40 MHz sampling frequency. A bipolar filter CR(RC)2 is adopted to remove the signal tail and limit the bandwidth. In order to optimize the electronic noise and the pile-up noise at high luminosity of the LHC, the time constant ( r = 13 ns) has been optimized to minimize the total noise. In order to achieve a 16 bit dynamic range in a linear scale, 3 ranges are used to split the dynamic range. The noise is of 390 pV for the medium gain (850 and 250 pV for the high and low gain). The integral non-linearity has been measured to be less than 0.2%. The full production is now completed. Signals from the shapers are stored in a circuit using Switched Capacitors Arrays (SCA). The SCA chip contains 16 analog channels. Among them, 12 are used to store the signal coming out from the shaper; the 4 others store a reference level. During the read-out operation, an off-chip amplifier subtracts the closest reference channel from each signal channel. This offers a noise rejection ratio improvement larger than a factor 4 during the simultaneous read and write operations of the chip. Two prototypes of the circuit were carried out: a CMOS version intended for the test beam electronics and the final version in DMILL technology. The performances of both prototypes have been measured on test bench and are summarized in Table 1. The DMILL chip was irradiated with 3 kGy and 2 1013N/cm2. No measurable change was observed after both irradiations. The analog pipelines are followed by a 12bit 5 MHz ADC (AD9220)a which digitizes the output of two SCA after a level 1 trigger. The gain of the signal is either chosen automatically by the hardware (free gain) or the digitization of one, two or three gains can be programmed. Table 1.

Summary of the measured SCA performances.

fixed sequence noise dynamic Range cell to cell variation drift

< 0.2 pv 13.3 bits < 0.02% < 3mV/ms

aAnalog Devices, 1 Technology Way, Norwood, MA 02062, USA

619

4.4. Controller Board

The Controller Board loads, updates and checks the different registers and parameters for all the different front-end boards. A special 10 Mb/s bus has been designed, the SPAC bus, with read and write lines doubled for safety. The final boards will be ready in October 2002. 4.5. Optical link

The optical link will transfer the Front End Board data to the Read Out system. The chosen link is the HP Glink, a radiation resistant Gigabit ethernet speed digital Optical link. A lot of effort has been put in developing a system with a low error rate2. Extensive radiation studies were performed with irradiations of 3 kGy and 51013 N. The single event effects were of 0.47 bit flip per link and per hour which corresponds to 8 hours of dead time in 10 years of LHC running. 5. Test beam results

Final electronics have been tested on test beam during the past two years. For the electromagnetic calorimeter one pre-series module and 4 series modules have been tested with final electronics using 6000 channels, 50 front-end boards, 12 calibration boards and 1 Tower Builder Board. The boards have all functionalities but are not completely rad-hard. The noise per standard electron cluster has been measured to be 140 MeV (high gain) and 240 (medium 5 - -7% and the power dissipation 0.7 W gain). The coherent noise is per channel. The automatic gain selection has been tested (the lower threshold only since the electron energy is limited to 245 GeV) and is behaving as expected. Full test beam results can be found elsewhere in these proceedings3. For the Hadronic End Cap 12 pre-series modules and 24 series have been tested on test beam using 400 cold channel^^>^.

-

6. Conclusion and perspectives

The successful use of more than 6000 front-end channels in test beam with full functionalities show the validity of the technical choices and have proven that the prototypes fulfill the required specifications (except for radiation criteria). A test of the full system crate with final radiation hard boards is scheduled for the fall 2002. This test will provide further measurements (coherent noise, system stability, timing jitter ...) qualifying the system for the production phase.

620

The next major steps are expected for April 2003 with the production of the 1650 FEBs, in October 2004 with the complete assembly and finally, installation is scheduled in February 2005. Acknowledgments

I would like to thank the organisers and staff of CALOR02 for the nice atmosphere of the Conference. Thanks to P. Borgeaud and B. Mansoulib for their help preparing this talk and for the careful reading of this note. References 1. ATLAS Collaboration, Liquid Argon Calorimeter Technical Design Report, CERN/LHCC 96-41. 2. M-L. Andrieux et al, Nucl. Instrum. Methods Phys. Res., A 456 (2001). 3. D. Zerwas, these proceedings. 4. M. Fincke-Keeler, these proceedings. 5. A. Kiryunin, these proceedings.

FRONT-END ELECTRONICS FOR THE CMS PRESHOWER DETECTOR

A. GO National Central University, Chung-Li, Taiwan E-mail: apollo.goOcern. ch

P. ASPELL, D. BARNEY, P. BLOCH, A. PEISERT, B. LOFSTEDT, S. REYNAUD CERN, Geneva, Switzerland

S. BORKAR, S. LALWANI BARC, Mumbai, India

The Front-End readout system PACE2 for the CMS Preshower detector consists of two chips: Delta is a 32 channel pre-amplifier and shaper that provides low noise, charge to voltage readout for large capacitive silicon sensors over a large dynamic range (up to 400 MIPS); PACE-AM contains a 32-channel wide, 160cell deep, analog memory with a 32 to 1 multiplexer for serial readout. These chips are designed in .8 pm BiCMOS DMILL radiation tolerant technology. The performance in terms of dynamic range, linearity, noise, peaking time and memory uniformity are presented.

1. CMS Preshower Detector

1.1. Physics Motivation

One of the most important physics goals of the CMS (Compact Muon Solenoid) experiment is the search for the Higgs boson, the missing particle in the Standard Model. For a Higgs mass below 130 GeV/c2, the most promising detection channel is via its decay to two photons ( H o + yy). An important background comes from ?yo + yy where the two photons are close to each each other and fake a single photon. In the barrel region the granularity of the Electromagnetic Calorimeter (ECAL) is sufficient to resolve the two photons with adequate effiency, but in the forward region, where the photon separation is smaller (a few mm) neutral pion background rejection decreases strongly. Therefore, a Preshower detector is included in the endcaps of CMS, in front of the ECAL'>2.

621

622 1.2. Micromodule

The CMS Preshower detector consists of two layers of lead converter and silicon strip sensors. The strips of the two planes of silicon sensors are orthogonal to each other. The sensor size is 6.3 x 6.3cm2 with 32 strips at 1.9 mm pitch and 320pm thick3. The basic detection unit, a micromodule, consists of a silicon sensor and the front-end readout electronics, both mounted on a ceramic substrate and an aluminium holder for mechanical support (Figure 1)2.

\bias

i

voltaqe

i

Figure 1.

The micromodule and its components.

2. Front-end electronics: PACE2 The front end electronics4 includes two chips: the preamplifier and shaper chip, Delta, and the analog memory and multiplexer chip, PACE-AM (Figure 2). Having two chips helps reduce any coupling between the digital part of the PACE-AM chip from the sensitive analog Delta chip. These chips are made in 0.8 pm BiCMOS DMILL technology with requirements for radiation hardness of up to lOMrad of ionizing radiation and up to 2 x 1014n/cm2 of neutron fluence.

623

I

I

PACE 2

I

Figure 2.

Block diagram of the PACE2 assembly.

Delta chip has 32 channels of preamplifier with leakage current compensation. Due to the harsh radiation environment, the silicon sensors will have increasing leakage current after few years of running. The preamplifier is DC coupled to the silicon sensor and can compensate for leakage current up t o 150pA. The CR - RC2 shaper is designed to have a peaking time of 25ns and can be switched between two gains. The low gain (LG), with a dynamic range up to 400 MIPs (minimun ionizing particles) is used for physics data taking while high gain (HG), with a dynamic range up t o 50 MIPs is used for absolute calibration of the MIP using particles and pulse injection. There are also 9 programmable DACs for internal biasing and current settings and calibration. Since the biasing condition changes with radiation dose, the biases and currents can be tuned by changing the appropriate DAC value, thus maintaining optimal performance. A calibration DAC is used to inject charges across the whole dynamic range and to cross-calibrate the two gains. PACE-AM contains a 32 channel wide, 160 cell deep analog memory (switched capacitor matrix). Signals from the shaper are sampled at 40MHz into cell memory specified by a write pointer. A read pointer, separated from the write pointer in time by the trigger latency, is used t o block 3 consecutive time samples when the level 1 trigger arrives. These blocked cells are read by an amplifier, multiplexed out at 20 MHz and then digitized externally by an ADC. The output signal is divided by 2 in order t o fit the 1V dynamic range of the ADC. The PACE2 chips were submitted at the end of 2000 and received in Spring 2001. A special motherboard was built containing the ADC (Analog Device 9042) t o digitize the serial data, an FPGA (Altera FLEX 10k) t o generate

624

all necessary fast signals to the PACE2 chip and a microprocessor (Mitsubishi M16C) t o control PACE2 and to interface with a P C via RS232. This set-up is self-contained and simple to use since it does not require complicated and expensive VME and NIM crates.

2.1. Initial Perfonnace Initial measurements were done on the chips without bonding to a silicon sensor. Digital functionality was checked. Some minor bugs were found but these did not affect the functionality. After fine tuning the biases and currents of the chips, the analog pipeline memory uniformity was measured. Figure 3 a shows the pedestal values for a few channels vs. memory cell number. The pedestal variation has an RMS spread of around 10 ADC counts (2.5mV). They have a similar shape, differing mainly in the overall DC shift. This can be easily corrected with look-up tables (Figure 3 b) .

Figure 3. a) Pedestal vs. memory cell number for a few input channels. b) Solid line: pedestal variation vs. channel before any correction. Dashed line: pedestal variation after applying cell-to-cell corrections.

The gain and signal-to-noise ratio (S/N) can be measured by injecting calibration pulses into the input of the preamplifier. For a fixed memory cell, we obtain 15.8mV/MIP for HG and 2.3mV/MIP for LG with a noise of l.OmV for HG and 0.5mV for LG. This results in S/N ratios of 15.8 (HG) and 4.6 (LG) respectively. By varying the delay of the trigger, we were able to map the pulse shape. The rise-time (10%-90%) is 15.5ns for LG and 18ns for HG. The shaping time is roughly 25ns, satisfying the design.

625

2 . 2 . Performance with Silicon Sensor and Radioactive Source

We have bonded a Preshower silicon sensor to the PACE2. In order to study the real response to particles, we placed a lo6Ru source in front of the sensor and two small plastic scintillators behind to act as a trigger. The single MIP signal was measured in HG (Figure 4 a ). The noise is larger than without a sensor (7 ADC counts=1.75mV instead of lmV), due to the large capacitance loading (52pF) and noise pick-up. A S/N ratio of 5.6 was obtained. This can be improved by better grounding and by using the 3 time samples. Knowing the absolute calibration from the radioactive source, we can inject charges to simulate large numbers of MIPs. The gain is measured (the slope of the straight line fit) to be 9.4mV/MIP (HG) and 1.4mV/MIP (LG). The gain remains constant within the dynamic range of 400 MIPs for LG and 50 MIPs for HG (Figure 5).

+

0.0813E-01 6.W f 233.4 f

0.5710E4.7 0.6959603 0.46Y9E-01

40

60

Figure 4. a) Pedestal and MIP signal from radioactive source fitted with two gaussians. b) Pulse shape for the MIP signal. Both are in HG.

We again mapped-out the pulse shape (Figure 4b). The rise time is around 20ns, slightly slower than the case without silicon but still satisfactory. 3. Conclusion and Outlook

The PACE2 was designed to be used to read out the silicon sensors of the CMS Preshower detector. Analog performance has been measured. One iteration is still needed to correct minor bugs and to extend the length of the analog memory from 160 to 192 in order to give sufficient safety margin for the level 1 trigger latency. Because of the uncertainty in the yield of the DMILL technology, a parallel development in .25 micron technology has started.

626

,

4w Injeclim [MIPs]

0

20

40

I

I

,

60 bljrcrmn [MIPS]

Figure 5. Output pulse height vs. injection signal for a) low gain and b) high gain. The flatten response above 400 MIPS in a) is due to ADC saturation from the high value of the pedestal which can be reduced easily.

References 1. D. Barney et. al. An artificial neural net approach to TO discrimination using the CMS Endcap Preshower, CMS Note 1998/088, (1998). 2. P. Wertelaers et.al. CMS Preshower EDR, CMS ECAL EDR-4 vol. 2, (2000). 3. A. Peisert, N. Zamiatin, Silicon sensors for the CMS Preshower, NIM A479,265 (2002). 4. P. Aspell, Conception et mise au point d e l'e'lectronique frontale du de'tecteur de pied de gerbe (Preshower) de l'expe'rience CMS, thesis, Universite Claude Bernard - Lyon 1, (2001).

THE FRONT-END ELECTRONICS FOR LHCB CALORIMETERS

DOMINIQUE BRETON Loboratoire de I’Acce‘lCrateur LinCaire - Centre Scientifique d’Orsay - B.P.34 - 91898 ORSAY CEDEX, FRANCE

For the readout of the calorimeters of the LHCb experiment at CERN, specific front-end electronics have been designed. In particular, three different front-end analog chips were studied respectively for the ECAL/HCAL, Preshower and Scintillator Pad Detector. We will present the three front-end electronic chains, point out their specific requirements together with their common purpose, and describe the corresponding ASICs.

1. INTRODUCTION The LHCb calorimetry is based on an electromagnetic and an hadronic calorimeter (ECAL/HCAL), a Preshower (PS) and a Scintillator Pad Detector (SPD). This set of four detectors takes place between M1 and M2 muon chambers. It provides high transverse energy hadron, electron and photon candidates for the first level trigger which makes a decision 4 us after the interaction. Its other essential function is the detection of photons to enable the off-line reconstruction of B-decays. These physics goals define the general structure of the calorimeter system and its associated electronics in terms of resolution, shower separation, selectivity and fast response. The ECAL and HCAL are lead-scintillator and iron-scintillator sandwiches read by light shifting fibers. The output of the plastic fibers is equipped with phototubes. The readout system will have about 6000 channels for the ECAL and 1500 for the HCAL. For economic reasons the ECAL and HCAL calorimeters will be equipped with the same electronics including fibers and PMs. The PS and the SPD are pad detectors read by the two ends of a simple fiber and multi-anode PMs. Both have 6000 channels. The crates and the backplanes will be the same for the four detectors, whereas the front-end electronics will be specific for a part and common for the other part.

627

628

2. REQUIREMENTS

The main requirement for LHCb electronics is the pile-up rejection. To ensure a satisfactory independence of successive sampling, it implies fast fibers, fast PMs and fast shaping. In ECAL/HCAL, the shaped signal has to be sampled at 40MHz over 12 bits to cover the resolution over the full dynamic range of the two calorimeters (100 Mev to 10 Gev of transverse energy). Data is then transcoded into energy over 8 bits for trigger data and 12 bits for readout data. The latter has to be buffered during the LO latency of 4psec, derandomized and then rebuffered for the level 1 latency of 256 psec'. The corresponding maximum rates are lMHz for the level 0 and 40kHz for the level 1. After the level 1 trigger, an extended zero suppression has to be performed before sending the formatted event to the DAQ. The aim of PS is twofold: for the level0 trigger, it has to perform the identification of electrons and photons while rejecting the charged pions. In parallel, it has to measure the part of energy lost in itself to correct for the global calorimeter information. The SPD has to perform the separation between photons and the charged particles. There are consequently trigger elements sitting in the front-end crate. The first stages concern the search for local energy maxima inside groups of 512 channels, with a validation by the PS and SPD data. The main difference in the shape of the signals outing the different subdetectors is actually due to the mean number of photoelectrons at their source. The SPD and PS that have to cover the 1 to 100-MIPS range receive only 20-30 photoelectrons per MIP. Their signal is consequently very unpredictable, especially at low levels. Conversely, in the HCAL, we get 50 photoelectrons per GEV, whereas it even goes up to 500-1000 in the ECAL. So they offer a much smoother and reproducible shape. The overall consequence is that the analog front-end electronics will greatly differ among the different sub-detectors. Its description is the aim of the following chapters.

3. FRONT-END OVERVIEW The front-end electronics will be situated on the top of the calorimeters as shown on figure 1. The total radiation dose expected there is about 1-2 krads over 10 years thus allowing the use of commercial components, provided that the most critical of them have been tested in a beam. 14 ECAL and 4 HCAL crates receive respectively 6000 and 1500 channels. 8 P S crates receive a total of 6000 channels both from the PSs and the SPDs very front-end elements. Figure 2 describes the principle of the front-end setup and presents a standard calorimeter front-end crate and its main interconnections. The ECAL/HCAL PM signals are connected to the front-end boards

629

Figure 1. overview of the LHCb detector and location of the calorimeter electronics)

Figure 2. overview of the front-end elements and of the front-end crate

through 10 meters of coaxial cables, whereas differential pairs are used for PS/SPD. There are 16 FEBs in the crates, each receiving 32 signals for ECAL/HCAL crates and 64 from each detector for PS/SPD crates. The output of these boards are connected to the standardized custom backplane, sending signals using LVDS levels to the Calorimeter Readout Controller (CROC) and the Validation boards. The CROC performs the event formatting after the first trigger level. Data is then sent to the DAQ through optical links. The CROC also receives the ECS2 and the TTC signals and therefrom distributes clock, global commands and the serial link that is used for loading the hardware over the whole crate. From its 32 signals, and using also neighboring cells, each ECAL/HCAL FEB computes, in pipeline mode, the maximum of the 32 sums over every 2x2 cell area3. This maximum is sent to the Trigger Validation Board assuming 8

630

FEBs for one Validation. The PS and the SPD validate the ECAL candidates. The energy of the latter is then added to the corresponding HCAL candidates. The output is sent to selection crates via optic links to get the highest (and second highest) of the candidates, for each type of particle. 4. THE FONT-END ELEMENTS OF THE ECAL/HCAL As explained here above, the purpose of those elements is to shape the PM pulses in less than 25ns to avoid electronics pile-up. The corresponding requirements are the following4: 0

0 0

0 0

At the PM output, the maximum current is 20mA over 25 ohms. At the ADC input, the dynamic range has to be 1V under 250 ohms. The residue after 25ns should be smaller than 1%. The sampling area should cover f2nsec with a 1%precision. The RMS noise should be < 1 ADC count (250uV)

To fulfill the requirements listed here above, two problems had to be solved. The first one concerns the PM signal. If one looks at figure 3, which shows a PM signal, the PM output current has a fast rise time but a slow decay that goes over at least the two consecutive samples at 40 MHz. It thus needs to be pulled to zero after 1Ons to ensure the zero pile-up requirement. The remaining area is on the order of 60% of the original one. To realize this cut on the signal, the clipping circuit of figure 4 will be used. I

-

g

3

I

I

1 Qa

1 a&

a4

... .

...

. ; /;\ j

j

I

'

!.............. I

,

.

I ...................... ,

I

,

,

"!.".-...........

.

).

.

I

(

: A :

;

...I.._.......:..-.-..........I..

02 .

.

..........

i

-

....

............

i.--.

j

......................

i

_.

..i

Figure 3. P M output pulse with and without the clip, and integrator output.

It consists in a short 5ns cable located at the output of the PM base. The latter sends part of the signal towards a variable network, which will send back an inverted part of the signal. As both the source and reflected signals are, on average, negative exponential, their superposition gives an almost zero signal, as shown in blue line on figure 3. Now that the input signal has been shaped, we have to measure the energy deposited in the calorimeter. The corresponding

631

Figure 4.

ECAL/HCAL front-end electronics.

information is the area of the PM signal. The best way to measure it without deteriorating too much the statistic fluctuation is to integrate this signal in a capacitor (Cf) as shown in figure 5. The difficulty then becomes to empty this capacitor. Two ways are possible: Use a switch mounted in parallel but this system induces a dead time when the capacitor is being emptied, which implies the use of two integrators in parallel and of a multiplexor. But due to the inevitable injection of charge from the switches, pedestals are generated which can be the sources of drifts at the 0.1% level. Subtract in a linear way the signal to itself thanks to a specific configuration. The latter was the chosen solution. The corresponding configuration appears in the middle of figure 4. The input signal of the analog chip is diverted, delayed by 25ns, then subtracted to itself thanks to the differential buffer. The latter also has in charge the division of the input current to adapt it to the small value of Cf. This solution is the one formerly proposed in the technical proposal2. Between the buffer and the integrator, an external AC coupling allows us to separate the DC levels. The integrator has a fast rise time and offers a satisfactory plateau at the top of the signal. Physically, as shown on figure 5, the main data path inside the board starts with the four 8-channel coaxial input connectors. The signal goes into the cable compensation, a pole zero network compensating for a 10% -20nsec signal tail, before entering the analog chip. At its output, after the 12bit ADC conversion, data undergoes a subtraction of the smallest of the two previous samples. This subtraction is intended to reduce the high bandwidth noise of the integrator that is never reset, and the two samples are used to decrease the probability

632

-

_I

-One channel

-. . --

_ I

L

-

I

Figure 5. ECAL/HCAL front-end physical implementation

.

that a signal is present in the sample that is subtracted. This subtraction is performed at 40MHz. Figure 6 shows the schematics of the chip. It consists of a differential input buffer that supplies the two opposite charges to the integrator. The buffer is linked to the integrator by an AC link, which is at the same occasion a filtering network. The integration capacitance value of 4pF was chosen for noise reasons. The feedback resistor is a dynamic circuit that provides a high resistance (ICON). The buffer gain was chosen in order to get 1.4V of dynamic range at the output. Its collector 2k resistor, which is used for replacing a 1.5mA current source, was chosen in order to reduce noise and simulate a current output as well as possible. The integrator is based upon a very low noise cascode amplifier with an open loop gain of 65dB. A first output emitter follower is integrated in the chip whereas a second one is mounted on the board with discrete components. This aims at avoiding any feedback from the ADC input to the amplifier, in particular to get rid of any potential saturation at high signal levels. Table 1 shows the performances of the chip. These results show a good adequation between simulation and real circuit, except for the gm of the input transistor. This is due to simulation models and also explains the mismatches in the rise time and the input impedance of the integrator. Figure 7 offers both a view of the 4-channel chip layout and a picture of the board. On the left side of the latter, one can see the 2 x 8 channel input connectors, followed by the PM and cable compensation networks and the delay lines. The four 4-channel analog chips are followed by half the ADCs and LUTs. All the big FPGAs (10K50 and 6K16) which perform the pedestal subtraction and the readout buffering are actually located on the other side of the board, together with the other half of the ADCs.

633 INTEGRATOR Vdd

Vdd

External AC coupling I

I

BUFFER VhS

Figure 6. ECAL/HCAL chip schematics Table 1. Chip performances

Integrator open Loop

65dB

6538

gain Crosstalk

<0.4%

~0.6%

Power Consumption

60mW /channel

57mW /channel

5. THE FRONT-END ELEMENTS OF THE PRESHOWER

As explained in chapter 2 and shown in figure 8a, the signal outing the PS and SPD photomultipliers is rather unpredictable. It is therefore impossible to perform a nice and smooth shaping thereon. Moreover, like for ECAL/HCAL, the signal spreads over more than one clock period of 2511s. And similarly to ECAL/HCAL, the only way to get the useful information is to make an integration. The smaller SNR requirement allowed us to use here a configuration based on two interleaved integrators per channel, one being read then reset

634

Figure 7.

4-channel ECAL/HCAL chip layout and top side of the prototype FEB.

while the other one is performing the integration (see figure 9)

Figure 8.

a) Typical 1-MIP signals.

b) percentage of charge in a 5-MIP signal.

As shown on figure 8b, the mean charge deposited within two consecutive periods for a given event is of the order of 85 and 15% for a 5-MIP signal that corresponds to the local trigger threshold. Moreover, the sigma on the 85% is rather small (4%), which will allow us to consider the 85% as a fixed value. Thus to prevent pile-up, 15% of the previous sample will systematically be subtracted to the present one. Concerning the energy measurement, the range to cover spreads between 1 and 100 MIPS. The precision aimed at for the MIP being of the order of lo%, a 10-bit 40MHz ADC will cover properly the overall necessary dynamic

635

range. But to ensure properly the SNR, the chip dynamic range will have to be as high as $-2V, which wont be easy seen the maximum power ratings of the technology (5.5V). As the double integration solution chosen wasnt requiring any big external component, and as the 64 channels outing a PM are physically grouped, it was decided to put the analog front-end chip5 as close as possible to the detector to gain on the noise level, more precisely on the board housing the 64-channel PM bases. The integrated and multiplexed signal can then be sent to the front-end board that houses the ADCs over a 15 to 20-meter long analog differential pair. This setup is shown on figure 9. The 5-MIP trigger threshold is consequently applied digitally beyond the ADC.

)A

One amona 8 channel of the PS ASIC.

-=

Figure 9.

block diagram of the PS front-end setup.

The ASIC has to perform a clean integration every 25ns. But conversely to the ECAL/HCAL one, there are digital signals therein. Moreover, the 40MHz clock is divided by two to perform the multiplexing between the two halfchannels, which is a source of bothering pick-up. Thus to avoid as much as possible the potential crosstalks between digital and analog parts, the latter are fully differential. In order to reduce the offsets, bipolar transistors will be used in the input stages, whereas CMOS gates will be used in the digital part to reduce the power consumption. The reset also has to be of high quality to avoid to spoil the trigger with the leftover from the previous sample. The chip technology is actually the AMS 0.8um BiCMOS. As shown on figure 10, the first stage of the chip is a common to differential amplifier, followed by the differential integrator. The clever AOP design allows us to fix the common mode voltage of the output at zero thanks to an extra feedback loop which is located in the dashed area of the drawing. The extra switches located at the inputs of the AOP are intended to pull their voltages to zero during the reset phase. Special care has been taken everywhere in the chip to reduce the loss of dynamic range due to temperature compensation

636

Figure 10. PS chip input stage and integrator.

and to improve the linearity. For instance, a parallel scheme has been realized in the multiplexor as shown on figure 11. Another advantage of this scheme is that it can be designed to over-compensate the concerned stage to also take the following one into account. OutDut buffer :

Q

J1

Figure 11. PS chip multiplexor stage and output buffer.

The design of the output buffer also was a challenge. It has indeed to be able to drive up to 15 meters of cable with a f l V dynamic range while offering the smallest possible power consumption. The thing is there are 64 channels on the small Very Front-End board, and the differential output requires two buffers per channel. Moreover, no good PNP transistor is available in the technology, only N P N ones. Thus the PNP was replaced by the network shown on figure 11, which actually behaves the same way. This scheme has the other advantage not to be very sensible to the sizing of the transistors, and thus to

637

always ensure a good stability. The performance obtained is rather amazing: with only 16mA of total quiescent current, the buffer is able to drive a 15-meter 100-ohm cable with a rise time of 3.5ns and less than 0.5% of non linearity. The overall performances of the ASIC appear on figure 12a and 12b. For instance, figure 12a shows the treatment of a nice test bench pulse. The spreading of the signal over a few clock periods at the input can be easily correlated with the corresponding integrated charge for those periods. Figure 12b shows the measurement of an actual 1-MIP signal after 15 meters of cable during a test beam at CERN. The plots are almost identical concerning the shape of the signal, and the amplitude of the 1-MIP signal is about 10 ADC counts, which is in perfect accordance with the chip requirements.

(a) test bench measurement

(b) an actual 1-MP signal Figure 12.

Figure 13a shows the plot of the pedestals at the output of the chip. One can easily distinguish the two discrete peaks corresponding to the two half channels. The noise distribution there around is at the level of 0.74 ADC count (equivalent to 750pV, which is well within the requirements (one ADC count). These two offset values will be subtracted further in the chain. Fig.13b shows the chip linearity, which actually remains within a 1%window over the whole dynamic range. Figure 14 shows both the layout of the 8 channel chip and the top side of the Very Front-End prototype board. The 800mW chip power consumption was a tricky problem to solve, especially when concentrated on such a small surface. A special package was chosen therefore.

638 8 channels prototype linearity e

1% window

90

IOO

110

L30

130

trfO

ISO

160

170

LgO

ADO count

0

V)O

(a) pedestal distribution

200

300

<00

BOO

COO

TOD

(b) chip lineqrity Figure 13.

•~~~~~~

{iggjgWggg

——-

Figure 14. 8-channel PS chip layout and top side of the prototype VFE board

6. THE FRONT-END ELEMENTS OF THE SCINTILLATOR PAD DETECTOR As explained in chapter 2 and chapter 5, the shape of the signal outing the SPD like the PS is rather unpredictable. The signals shown on figure 8a are indeed valid for the two detectors. However, the SPD only aims at giving to the trigger system a particle identification information. Its one bit per channel outputs thus report if the crossing particle was a charged one or a photon.

639

Figure 15a shows the corresponding energy deposition. Applying a threshold at 1.5MeV ( 0.5MIP) is a good compromise to reject the photons with not too bad an efficiency for the trigger. There is anyway a bothering problem correlated with this very low threshold as shown of figure 15b: as the signal spreads over more than one clock period, the potential tail of an high amplitude event peaking in the previous clock period could cross the threshold and fire a fake trigger. To avoid this kind of situation, the signal integrated during the previous clock period will be systematically subtracted to the current one. Conversely to the PS situation where this can be done beyond the ADC, and as there is no reason here to send the analog data towards the front-end crates, it will be done locally in the front-end chip6, the latter being located on the PM-base board like for the PS. This moreover allows us to reduce the amount of interconnections between the very front-end boards and the crates through the use of fast LVDS serializers. SPD Threshold (1.5MeV

0.5MW)

SPD dep. e n e q (MeY)

(a) SPD energy distribution

(b) integrated signal Figure 15.

The corresponding physical implementation can be seen on figure 16. The block diagram of the SPD chip looks like that of the PS on the input side. As in the PS chip and for the same reason, the input is unipolar and the first operation consists in passing to differential mode. Then the signal is integrated and stored in a sample and hold. The next stages are different. As explained here above and before being compared to the 1.5MeV threshold ( 200mV), the integrated signal undergoes a subtraction where 17% of the previous sample is removed in order to reject the potential pile-up. The low threshold being applied locally, the internal noise and the digital to analog crosstalk were things to be carefully taken into account. The chip main blocks schematics are shown in figures 17 and 18. The input stage is a simple differential amplifier feeding current in the differential

640

Figure 16. one of the eight channels of the SPD chip

integrator whose capacitance value is about 1pF. The scheme of the latters amplifier takes advantage of the BiCMOS process to obtain low noise and high gain. The track and hold design is optimized concerning the pedestal, drop rate and input to output feedthrough. Moreover, only one clock phase is necessary to make it work, which reduces the digital to analog crosstalk.

Figure 17.

SPD chip schematics

The differential subtractor is shown on figure 17. Its both outputs are connected to the latched comparator to which the threshold is also sent in a differential way. The latter has to perform a clean decision without injecting any charge towards the analog part. Therefore, its output is differential and this digital signal will remain differential up to the physical output of the circuit. The comparator receives the threshold from a 7-bit DAC (1 bit for sign and 6 bits for module) also located within the chip. Due to the individual offsets of the half-channels, there are actually two DACs per channel. They will be written and read back thanks to an 12C interface internal to the chip. An external adjustment is also available to finely tune the DAC range.

641

Latched comparator Figure 18. SPD chip schematics

The circuit has been thoroughly simulated using Spectre and the first prototypes have been tested both on test bench and in test beam. As shown on figure 19, the overall performances are well adequate to the requirements. The non-linearity of the analog part is at the level of 0.5% over the whole dynamic range that is of the order of +-1V. The noise level is about 1mV. The high gain of the comparator allows the latter to present a very good precision. Threshold crossing

.,

Figure 19.

Threahdd preclslon

.. . .~

SPD chip performances

Figure 20 shows two simulations of a MIP, one with a very low and the other with a high threshold. The uppermost plot shows the 20MHz clock and the 1-MIP input signal. Just below, one can see the output of the two half-channel integrators, then the same signals stored in the track-and-hold (TH1 And TH2). On the following plots, the lowest signal that corresponds to the integrated MIP crosses the threshold and the comparator switches. The last plot shows the output of the multiplexor that validated one of the two comparators, so that a 25ns digital pulse appears at the output. Its interesting to notice here that the output signal is the same whereas the chosen thresholds were extreme. The standard value should be of the order of 0.5MIP. Figure 21 shows the layout of the four-channel prototype and the corresponding test board. New prototypes are currently under design. The goal is to better adapt the gain of the chip to the characteristics of the PM and

642

Figure 20. 4-channel SPD chip simulation plots

.

Figure 21. 4-channel SPD chip layout and test board

to reduce the power supply voltage which is currently of 5V down to 3.3V for power dissipation purposes. The latter could then indeed go from 1.2W down to 0.6W per chip, which would be fruitful for the circuits will be located in a place where their cooling will be difficult. The final 8-channel prototype is targeted for end of 2002.

7. CONCLUSION As it was explained here above, the specifities of the different sub-detectors constituting the calorimeter of LHCb led us to design three different front-end electronics. For this purpose, three different ASICs were developed in the AMS

643

0.8um BiCMOS technology. The LAL in Orsay was in charge of designing the chip for ECAL/HCAL, the LPC in Clermont-Ferrand of designing the Preshowers one and the group of Barcelona the SPDs one. The two first chips have actually passed the PRR and their production (2500 pieces for each) has been launched. The SPD chip is still at the level of the last design review. Despite these differences, we tried to standardize the readout elements as early as we could in the chain, and most of the electronics located beyond the ADCs will be common, especially the crates and the backplanes that are densely interconnected because of the trigger. There are indeed about 1000 cables carrying a total of 1200Gbit/s between the crates, each backplane having itself to transport 200Gbit/s. Another tricky problem to solve was due to the radiations. Indeed, even if the total dose remains rather small (1-2 krads over 10 years), it will cause Single Event Upsets, Latch-up and Transients that might be a source of errors or damage in the electronics. The latter is thus being designed in order to remain as immune as possible to those annoying effects. For instance, all the currently used Altera FPGAs will be replaced by ACTEL ones based on the antifuse technology. The use of RAMS was also strictly limited, and there will always be redundancy encoding in the ones remaining. All the sensitive elements will also have their schematics hardened thanks to triple voting schemes. Moreover, the ECAL/HCAL and PS chips had to undergo radiation tests and proved they could handle up to 50kRads of dose without bothering degradation of their performances. With the PRRs for all the boards located in the front-end crates about to take place in mid-2004, there are two years remaining now to finish the final design. References 1. J.Christiansen, “Requirements to the LO front-end electronics”, LHCb 99-029 2. D.Breton et al., “SPECS: Serial Protocol for the Experiment Control system of LHCb ’ I , LHCB note in progress 3. C.Beigbeder et al., “A Joint Proposal for the Level 0 Calorimetric Triggers”, LHCb 99-017 4. C.Beigbeder et a1,”The Front-End Electronics for LHCb calorimeters” LHCb 99-053 5. J.Lecoq et al., “The mixed analog/digital shaper of the LHCb Preshower ‘ I , LHC Electronics Board 2001, Stockholm, G. Bohner, ”A mixed analog-digital shaper for the LHCb preshower”, LHC Electronics Board 2000, Snowmass 6. A.Dieguez et al., “A BiCMOS discriminator interface for the SPD “, LHC Electronics Board 2000, Stockholm

FRONT-END ELECTRONICS UPGRADE FOR THE CDF CALORIMETERS

C. A. NELSON, T. M. SHAW Fennilab, P.O.Box 500, Batauia, IL 60510, USA Upgrades made to the Fermilab Collider to increase luminosity have made it necessary to replace the original CDF sample-and-hold calorimeter electronics with a synchronous system based on the Fermilab QIE series of ASICs. The QIE is an auto-ranging, gated integrator with eight binary-weighted ranges. We have designed and produced a front-end readout system, which digitizes the CDF calorimeter signals, produces trigger information and operates within a two-level trigger system. The architecture of this system and some results from the commissioning period are reported.

1. Introduction Changes made to the Fermilab collider to increase luminosity have necessitated upgrades to several of the CDF detector components and to all the readout electronics. This paper describes the new readout electronics for all the CDF calorimeters. CDF also uses this system to read out the luminosity monitor, the time-of-flight detector and the mini-plug detector. The new readout system digitizes the charge from every calorimeter channel, at a 7.6 MHz rate (132 nanoseconds). It also provides prompt trigger outputs that are sums of up to four channels, weighted by sin(0). It operates within a two-level trigger system in which prompt signals from certain detector components (including the calorimeters) are processed to form a provisional, Level 1 trigger decision, and then a Level 2 decision, involving much more information, is made on events accepted by Level 1. Events accepted by the Level 2 trigger are read out of the front-end crates. The readout has a dynamic range of approximately 18 bits. An LSB at the low end of the response is 5.8 fC, and the full-scale charge is about 1300 pC. Noise is around 10 fC, and the system permits charge from the detectors to be measured to an accuracy of within 0.5%, or better, depending on how the measure of accuracy is defined. The system consists of two components: a small (4.25”x0.9”) card in the 72-pin SIMM format, and a 9Ux400mm VMEbus VIPA’ format board. The

644

‘

645

SIMM card, called the C A F E Card (CALorimeter Front End), contains the electronics for digitizing a single calorimeter channel. The VMEbus board, called the A D M E M ( A to D MEMory), contains sockets for 20 CAFE Cards, plus electronics that implement the pipeline registers, the Level 2 buffers, the trigger signal formation, and the VMEbus readout and control interfaces. The ADMEMs are housed in VMEbus VIPA crates located near the detector. These crates also contain cards servicing other components of the detector, and also commercial VME computers that perform control and readout operations. The system supports an event readout rate of up to 1000 events per second. Operations performed by the VME computers are asynchronous with the 7.6 MHz beam-crossing clock driving the analog-to-digital processing. A major challenge in the development of this system was minimization of noise from this asynchronous digital traffic.

+

2. The C A F E Card

2.1. Block Diagram

Figure 1 shows a block diagram of the CAFE card. A 50 ohm input resistor terminates the input cable, and the Current Buffer section passes the input current to a charge-integrating ASIC, called the QIE. The QIE analog output is digitized with a ten-bit ADC, and digital data from the QIE and the ADC are presented to the address inputs of a FlashRAM lookup table which translates the piecewise-linear raw data format into a linearized representation of the input charge. A dc current source in the Calibrator section can be switched in to measure the QIE response. The Source Monitor section allows the experiment to measure small dc currents produced by the detectors in response to radioactive sources that are used periodically to monitor overall detector response. 2.2. The QIE

The CAFE cards are based on the QIE series of charge-integrating ASICs designed at Fermilab2. The CDF version divides the full input scale into eight binary-weighted ranges. The QIE puts out an analog voltage and five bits of digital data. Figure 2 shows an idealized version of the output voltage versus input charge. Only the lowest four ranges are shown. The lowest, RANGE 0, is shown spanning an input charge Qo. Each range spans twice the charge of the previous range, so the QIE full scale is 255 x Qo.

646

Figure 1. CAFE Card Block Diagram

Figure 2.

Idealized QIE Response.

Within each range the output voltage is linear with charge, rising from zero to V,,, . This output is amplified, level-shifted, and presented to an ADC. The

647

CAFE card ADC has ten bits, so the full scale charge is 1024 x 255 times the LSB of RANGE 0. Thus the dynamic range is essentially 18 bits. The QIE creates the binary-weighted ranges by splitting the input current f , and so on, down to There into eight paths; the current splits are are two paths with & of the input current. The eight currents are integrated onto capacitors at the splitter outputs. All the capacitors are the same size, 2.5 pF, except for the second path, which has a 5 pF capacitor. Thus, in terms of its voltage output, this last path functions as though it has & of the input current. After an integration cycle, switches disconnect the eight capacitors from the splitter stage and connect them to a set of comparators. The outputs of the comparators determine which voltage will be the analog output. Three “RANGE” bits indicate which of the eight capacitors has been selected. The output cycle is next. The capacitors are disconnected from the comparators and connected to a multiplexer which drives out the selected voltage. The RANGE bits are also driven out. After the output cycle, the capacitors are disconnected from the multiplexer and reset with switches. Reset lasts for an entire clock cycle. After the reset cycle, the capacitors are again connected to the current divider stage, and another integration cycle begins. A single set of eight capacitors will integrate for only one clock period in four, so the QIE must contain four sets of capacitors. At any given time, one set is connected to the current divider stage, one to the comparators, one to the output buffers, and one is in reset. In addition to the analog output and the RANGE bits, the QIE puts out two “CAPID” bits to indicate which set of capacitors is at the output stage.

5, 5,

A.

2.3. The Current Bufler, Calibrator and Source Monitor

Due to limitations in the QIE fabrication process, the device’s input impedance is not well controlled and cannot adequately terminate a 50 ohm cable from the detectors. It was necessary to place a circuit, called the Current Buffer, in front of the QIE to provide proper cable termination and to deliver the input charge to the QIE with minimum noise and distortion. For best charge reconstruction, it is necessary to calibrate the QIE response for each RANGE and CAPID combination. The Calibrator section contains a precision voltage-controlled current source and switches to permit the source to be connected to the Current Buffer input. A programmable calibration voltage is supplied by the ADMEM motherboard. The CDF calorimeters contain radioactive sources that can to moved so as to illuminate all detector sections. The detector response can be regarded

648

as a dc current of order 100 nanoamperes. This integrates to a charge that is too small to be measured well by the QIE. The Source Monitor section of the CAFE card contains a xl000 current amplifier and switches that allow the amplifier to be switched into the signal path. The resulting amplified current integrates to a charge large enough to be measured accurately. 2.4. The FlashRAM Lookup Table

The CAFE card can transform the raw, piecewise-linear QIE/ADC data into a calibrated, linearized representation of the input charge. This is necessary for trigger sum formation by the ADMEM; the ADMEM logic that sums data from four channels to form one trigger tower output cannot work with uncalibrated, raw data. The transformation is done on the CAFE CARD with a lookup table stored in a FlashRAM memory device. Using the CAFE card’s onboard calibrator, all RANGE and CAPID combinations of the QIE can be calibrated. The FlashRAM can then be downloaded to produce a calibrated, linearized output. Since the dynamic range of the QIE/ADC combination on the CAFE card is 18 bits, and the FlashRAM output has only 16 bits, CDF has chosen a representation with a 15-bit value and a one-bit exponent. If the exponent is set to one, the value is to be multiplied by eight. The FlashRAM has 16 address inputs, but the QIE/ADC code contains only 15 bits, so the lookup table occupies only the bottom half of memory. The top half is used as a pass-through table; each location contains its own address. The high order address bit is programmable, so the calibration procedure can read raw QIE/ADC data by selecting the top half. The FlashRAM can be downloaded and read back under computer control. The highest location is reserved for an ID number, which is unique to each CAFE card. The top half of the memory is protected. Only the lookup table section can be erased and rewritten.

3. The ADMEM 3.1. Opemtion The ADMEM is the motherboard for 20 CAFE cards, and it contains the logic for forming the trigger sums, and for implementing the Level 1pipeline storage, Level 2 buffers, and VMEbus interface. Figure 3 shows the data flow. Data from four CAFE cards are presented to a trigger-tower summing section and to Level 1 pipeline registers. In the summing section, the 16-bit word from each channel is changed into a straight 18-bit number by multiplying the low-order 15 bits by eight

649

Figure 3.

Calorimeter Data Flow

if the high order bit is set. A programmable, eight-bit pedestal is subtracted from each channel. The four 18-bit words are then summed, and the result is divided by eight, truncated to 15 bits and set to full scale it is if too large to truncate. This scaled sum is then presented to a FlashRAM lookup table which effectively multiplies it by sin(8) to produce a value proportional to the transverse energy E t . Ten bits are used to represent this weighted trigger output. Trigger words are transmitted to the Level 1 trigger system by LVDS drivers, through connectors on the ADMEM front panel. Data entering the pipeline registers emerge a fixed number of cycles later, at which time the Level 1 trigger decision has been received. An accepted event is stored in one of four buffers, pending a Level 2 decision. The accept command from Level 1 specifies which buffer will receive the data. On receipt of a Level 2 accept command, the VME computer reads the event stored in the buffer indicated by the command. The computer sparsifies the data and writes it to another VMEbus module, which serializes the data stream and sends it to a higher level part of the CDF readout system. The trigger summing sections, Level 1pipeline registers and Level 2 buffers are implemented in FPGAs. Configuration data for the FPGAs is held in

650

FlashRAM memories. These can be erased and rewritten via the VME interface, and the FPGA configuration cycle can also be initiated, so the entire structure of the data flow can be altered remotely. Similarly, the contents of the FlashRAMs which apply the sin(8) weighting to the trigger sums can be changed remotely. The VME interface is also implemented in an FPGA, but this is not remotely re-configurable. The ADMEM conforms to the A32D32 VMEbus VIPA standard. The P1 and P2 connectors have standard VMEbus assignments, and user-definable pins on P2 are used for clock and trigger signals. The VIPA PO connector carries analog power supplies and returns. The P3 connector brings the detector signals onto the ADMEM. Signal traces are routed on an internal layer and sandwiched between two analog ground planes. Also, signal traces are separated by analog ground traces. CAFE cards are aligned vertically and arrayed across the bottom of the ADMEM.

3.2. Control of Digital Noise Much effort went into minimizing noise from asynchronous digital activity occurring during VMEbus readout and control operations. After much testing and a few design iterations, a successful configuration was found. It had been decided early in the project to have separate power and ground distribution for analog and digital sections. Analog ground and digital ground were to be connected in one place only - behind the VME crate, at a local ground reference point for the (floating) power supplies. This was found to be important. When accidental connections were made at other places, pedestal distributions developed long tails that were associated with computer activity. It was also found to be important to avoid any overlap between analog and digital power and ground planes, and especially between analog planes and traces carrying asynchronous digital traffic. This applied to both the CAFE and the ADMEM. A particular problem on the ADMEM was that analog power and ground must pass by the P2 connector on their way from the PO connector to the CAFE cards at the bottom of the board. Overlap here between analog planes and digital signal traces is avoided by the use of laminated power bus bars that permit analog power and ground to run above the surface of the ADMEM board. At the PO connector, analog power enters through EM1 filters, which are essentially common mode chokes with integrated, low-inductance bypass capacitors. These provide a measure of rejection against high-frequency noise picked up by the analog power planes in the backplane of the VME crate.

651

4. Operational Experience

The ADMEM calorimeter readout is now in use by CDF in Collider Run 11. Reliability has been high. The CAFE failure rate of 0.7% per year has been due mainly t o blown fuses resulting from certain improper operations involving the PM T high voltage distribution system. The ADMEM failure rate has been about 3% per year, due mainly to some failing tantalum capacitors. QIE calibrations are performed about once every two days. Pedestals are found to be stable to within f l LSB per month, gains to within f0.15% per month. The FlashRAM lookup tables in the CAFE cards are updated about once every two weeks. The entire process takes 15 minutes. There is as yet no precise information from Run I1 on the accuracy with which charge from real PMT pulses is reconstructed using QIEs calibrated with dc current, but studies were carried out before Run I1 in which light pulses from a laser were attenuated by a variable density filter wheel and viewed by a PMT, read out by an ADMEM, and also viewed by a pin diode, read out with a slow, precise ADC. Comparison of the ADMEM readout with the pin diode indicated that charge reconstruction for pulses is accurate to within 0.5%, or better. Initial noise levels were 2.5 LSBs (about 15 fC). Noise has now been reduced to 1.5 LSBs (about 9 fC), by shielding the signal cables between the detector and the ADMEMs. The ADMEM is working well for CDF and meeting all specifications.

Acknowledgments The success of this project owes much to the participation of physicists in testing the early ADMEM and CAFE prototypes and in providing feedback which helped guide the design work. Also important was the very careful work done by those who installed the system. We thank Mike Lindgren, Steve Hahn, Hao Wei, Howard Budd, Willis Sakumoto, Rick Tesarek, and many others for their contributions to this project. This work was supported by the U.S. Department of Energy under contract NO. DE-AC02-76CH03000.

References 1. “VME Extensions for Physics / VME64xP,” International VMEbus Trade Association publication ANSI/VITA 23-1998. 2. “A Pipelined Multiranging Integrator and Encoder ASIC for Fast Digitization of Photomultiplier Tube Signals,” R. Yarema, et al., Fermilab-Conf-92/148. “A Second Generation Charge Integrator and Encoder ASIC,” T. Zimmerman and M. Sarraj, IEEE NS 43-3,1683 (1996).

THE ELECTRONICS OF THE NEW H1 LUMINOSITY SYSTEM

v. BOUDRY, F. MOREAU, A. E. SPECKA, CH. RENARD+ L. L.R., Ecole Polytechnique, Palaiseau, France E-mail: Vincent.BoudryOin2p3.fr

E. BARRELET, PH. BAILLY, H. LEBBOLO, A. VALLEREAU L . P .N.H.E., Universite' Paris VI- VII, Paris, France

R. CHICHE, CH. DE LA TAILLE L.A.L., Universite' Paris-Sud, Orsay, France In the scope of the upgrade of the HERA collider, the H1 luminosity system was rebuilt anew. The analog electronics has been designed to transmit photo-multiplier pulses with a repetition rate of 10.4 MHz through 125m cables, to compensate for the cable skewing and to do a fast shaping avoiding pile-up. The corresponding acquisition is based on custom digitizing cards using the 41 MHz, 12 bits AD9042 ADC chip, a custom 48Mb/s readout bus, and a commercial computing board (MFCC from C.E.S.) performing fast histogrammation of data. Preliminary results show the ability to readout data at a rate of 0.6MHz and process them at 0.4 MHz.

1. Introduction

At HERA, the luminosity is determined by measuring the flux of bremsstrahlung photons emitted at zero degree off electrons in the field of protons. For H1, the detection is realized by a new 2 x 12-strip quartz fiber calorimeter, called Photon Detector (PD), situated 104 m downstream the interaction region' and detailed at the previous session of this conference cycle2. A small 12-channel calorimeter, the Electron Tagger (ET), at 6 m completes the system. The HERA upgrade realized in 20013 increases the luminosity by a factor 5 and implements the longitudinal polarization for the experiments. The measurement of the luminosity then requires4 to record the complete deposited energy spectrum, rather than mere rates, on a bunch by bunch basis and once per minute to match the 20minute build-up rate of polarization. t Now at SubaTech, Nantes, France.

652

653 In that scope the new quartz fiber PD presents some very appealing features such as fast signal, insensitivity to synchrotron radiation photons and very high radiation hardness. It nevertheless challenges the front-end electronics because of its low light yield (about 130 photo-electrons per GeV). Due to the high bremsstrahlung rates - a significant signal is expected every 96ns, the HERA bunch crossing (BC) period - the photo-multipliers (PM) have to be run at low gain (a few lo5). Moreover, transmission of the pulses through 125 m cable implies attenuation and distortion which have to be compensated. Front-end cards have been designed to scope with these problems and are first described. The associated acquisition hardware and performances are addressed in section 3 and 4 respectively. 2. Analog Electronics

The main features required from the analog electronics are to provide a substantial gain to the signal and to avoid the pile-up of signal from successive BC while keeping a 12-bit dynamical range. Its two parts, a pre-amplifier card at the back of the detectors and a shaper card in the experimental hall have been realized by the electronics group of the LAL, Orsay. 2.1. Pre-Amplification

The output of the PM is fed by a 20cm-long 50Q LEMOcable to the preamplifier card, with an impedance match at both ends. A gain of 10 on a 5 0 0 load is provided and allows to run the P M with a reduced gain. A card processes 14 channels (two being spare ones), provides their analog sum and a reference channel (for subtraction) in order to reduce the picked-up noise. Three such cards serve the X and Y projection of the PD and the ET. The mechanical design has been optimized to facilitate the maintenance and to minimize the time spend in the tunnel in exchanging the card: its small dimensions (10 x 15 cm2) and the placing of the 1/0 on opposite faces allows it to be fixed on the door of the casing containing the PM tubes and to achieve an efficient shielding. 2 . 2 . Correction

tf.4

Shaping

The signals are transmitted to the shaper cards through 125m of RG58 (low loss CERN type) coaxial cable. The 10ns-width signal emitted by the preamplifier card (fig. l a ) is skewed by the transmission with the higher frequencies damped relative to lower ones, resulting in an amplitude loss of lOdB and a prolongated tail (fig. lb ), largely exceeding the BC period. The 16-channel card design includes i) two Pole Zero Compensation (PZC)

654

5

0

20

40

60

-.a0

E

80

100

time Ins

a0.6 0.4

0.2

0'

20

40

60

80

100

time Ins

Figure 1. Evolution of an optical calibration signal, very similar to a physical one, a) at the output of the pre-amplifier card, b) after 125m of coaxial cable, c) at the output of a shaper card.

tuned on the measured skewness of the cable, restoring fully the initial preamplified signal, i i ) a unipolar shaping with a typical rising time of -8ns and a return below 1% ' within 72ns (fig lc.), iii) the subtraction of the reference channel. By these means, a reduction of the low frequency coherent noise, dominant at 150Hz, from -80mV to -8mV is achieved, the residue originating from the local noise pick-up. The white intrinsic noise at the shaper output is of (T 0.55 mV on single channels and of 0.96mV on the sums, pointing t o a dominant component coming from the shaper card. The linearity of both cards were measured on test bench to be better than 1%.

-

3. Acquisition Due to the new requirements recalled in the introduction a completely new acquisition, parallel to the H1 central readout (running at -50 Hz), has been designed and realized, featuring: i) a fast readout and histogramming (at typically 1MHz) of the analog sums, aiming the percent statistical precision on individual BC per minute, ii) a complete readout (at typically 100 kHz) of all the channels for detailed off-line analysis, iii) a slow control permanent monitoring of the systematics in order to keep the energy calibration at the

655

1% level. In addition, stability of the sampling position with respect to the shaped pulse (fig lc) requires precise timing monitoring. The acquisition system is physically based on custom designed VME-sized cards and an extended VME crate; it includes 8 ADC and 1 TDC cards, a fast readout bus, named lumi bus, and a commercial mezzanine computer board, the MFCC. A Service Module provides all the external signals (clocks, triggers, ...). The full system is globally cadenced at 10.4MHz, the frequency of the HERA clock (HCk). Each component is described in detail in the following sections. 3.1. A D C €3 TDC Cards The ADC cards are based on a chip AD9042 from Analog Device able to digitize a differential signal in the range 0-2 V in 12 bits at frequencies as high as 41 MHz. One card treats 8 channels in parallel, with 4 samplings per BC. The data produced are stored independently in two pipelines of 512 pairs of sample (256 BC), each driven by its own Pipeline Enable (PEn) line; one pipeline is readout through the VME and used by the central H1 readout. The other one is readout via the dedicated lumi bus for the local acquisition. A digital comparison on the 4 samples of two channels on the card provides the signals for triggering the local acquisition. The TDC card is based on a chip developed for the BaBar collaboration by the team of the LPNHE, Paris, who designed the ADC and TDC cards. Each chip records as many as 16 inputs with a resolution of 0.5ns inside a programmable time window anterior to the trigger signal. Up to 8 formated events per channel are kept at a time in an internal buffer. One TDC card processes 8 digital and 8 internally discriminated analog inputs, digitized by 2 chips in order avoid buffer overflowing. On an external trigger, the internal buffer of the TDC chips are transfered to an external FIFO for a delayed readout, using either of the buses.

3.2. The Lurni Bus The lumi bus support is a backplane board extending on 14 slots of the VME crate and connecting the 52 “A” and “C” rows. The lines have a 5 0 0 impedance and critical control lines have matched terminations. The bus, fully controlled by the MFCC, contains 24 data and 16 address and sub-address synchronous lines, cadenced at 20.8MHz (2 x HCk), plus several asynchronous control lines (Reset, Synchronization, Pipeline Enables, TDC trigger, ...). The addresses and data bits are offset by one clock cycle in order to improve the performances.

656

3.3. The MFCC Acquisition Board In form of a mezzanine board, the MFCC 8441 (Multi-Function Computing Core), a commercial product developped by C.E.S.5 for communication, is a fully developed computation unit based on a 300MHz PowerPC, 32MB of RAM, a 66MHz bus and two fast FPGA (Altera 10K50). One of them ensures the communication with the carrying VME station (via a PCI bus). The other one is dedicated to lumi bus control and data I/O. In this “user” FPGA, the following tasks are realized: 0

0

0

0

Trigger Logic: 16 maskable bits originating from the ADC cards can launch a readout sequence (standard “triggered” mode). The start can also be given by a specified BC number and allows to record a number of consecutive BC in a row (“train” mode). Synchronization: internal counters for the BC number and the ADC pipeline index (commanded by the “lumi” PEn) are kept up to date. Readout Sequencing: on a readout command, the addresses and subaddresses are send on the bus. They are formed from the required mode (“triggered” or “train”) sequence table, modifiable by the PowerPC on the fly, and from the pipeline index. Data Formatting & Output: the corresponding data fetched from the ADC cards are completed by the local information (BC number, trigger bits, ...), formatted and put into the FIFO using the internal dual port memory of the FPGA. The latter provides the buffer needed between the 20.8 MHz input and the 66 MHz output.

The “train” mode, limited to 55 BC by the FIFO size, allows to accumulates rapidly a high statistic on the analogically summed energies and avoid completely any trigger screening effect, an otherwise overwhelming problem at high luminosity. A full HERA cycle (220 BC) is completed with 4 such trains. 4. Preliminary Performances The data available in the FIFO of the user FPGA are processed by a standalone program (no 0s) on the MFCC PowerPC and histogrammed. The present acquisition time in “train” mode is of 2.5 ps/BC (readout only: 1.6ps/BC). Limited by the CPU (192ns expected from bus, 460ns from sequencing chaining), it will be improved by software optimization. A typical distribution obtained is shown on figure 2. In the “triggered” mode the total energy is calculated from individual channels at a rate of 27 kHz. Once per second the histograms are sent to the carrying board for analysis, extraction of the luminosity numbers and checking of the calibration.

657

10’

Total Energy Spectrum accumulated in 60s

6

10

I 0’ 10‘ 1o= 10‘

10

1 0

500

1000

1500 2000 Energy I ADC count

Figure 2. Typical bremsstrahlung spectrum accumulated on the Y projection of the PD in one minute by 60 interleaved “train” (clear) and “triggered” (dark) acquisitions (see text), reading respectively the analog sums and all the channels. Pure processing rates were of 400 kHz and 27 kHz.

Conclusion & Perspectives

A completely new electronics ranging from pre-amplification to acquisition has been realized and is fully operationnal for the new H1 luminosity system. A 12bit dynamical range for analog pulses is currently provided at the end of 125 m cables. The acquisition developped allows currently to process the 42 MHz digitized data at a rate of 2.5 ps per bunch crossing on a train of 55 successive bunches and to read full events for the control of the systematics. Since the conference the system has been commissioned and is now (June 2002) in operation while HERA is tuning the beams for luminosity production. The very same electronics and acquisition is used for the transverse and future longitudinal polarimeter of HERA. References 1. I. Abt et al. [Hl Collaboration], “The H1 Detector At Hera,” Nucl. Instrum. Meth. A 386 (1997) 310. 2. A. E. Specka et al., “Test beam results of the tungsten/quartz-fibre calorimeter for the luminosity measurement in Hl,” . Published in *Annecy 2000, Calorimetry in high energy physics* 121-130 3. M. Seidel, “HERA luminosity upgrade,” DESY-M-99-02ZB Prepared for IEEE Particle Accelerator Conference (PAC 99), New York, N Y , 29 Mar - 2 A p r 1999. 4. H1 Collaboration, DESY PRC 98/05. 5. CES - Creative Electronics System S.A. http://www.ces.ch

THE aABAR ELECTROMAGNETIC CALORIMETER IN ITS THIRD YEAR OF OPERATION

IVO GOUGH ESCHRICH Imperial College of Science, Technology and Medicine Blackett Laboratow, Prince Consort Road, London S W 7 2B W, England E-mail: [email protected]

(For the BABARCollabomtion)

The BABAR experiment at the SLAC B-Factory has recorded more than 80 fb-' of integrated luminosity since 1999. Its electromagnetic calorimeter which consists of 6580 CsI(T1) crystals has to detect both photons below 20 MeV as well as electrons in the 0.5-9 GeV range with a few percent resolution. Status and performance of the readout electronics including reliability issues and operational experience after the first three years of operation are presented.

1. Introduction

The BABAR detector1 at the SLAC B-Factory has been taking data since May 1999, recording more than 80 fb-l of integrated luminosity by May 2002. It is interesting at this point to review the performance of its electromagnetic calorimeter (EMC) as well as the steps that were necessary to obtain this performance. At this conference we have reported on calibration2 of the EMC, radiation damage to its 6580 CsI(T1) crystals3, and extrapolations towards significantly higher luminosity4. This contribution will focus exclusively on the front end electronics. The BABAR electromagnetic calorimeter maintains an energy resolution of a few percent over the range from 0.5 MeV to 9 GeV. At the same time it copes with the high rates of photon background at an instantaneous luminosity exceeding 4 x 1033~m-2s-1.This is accomplished by employing an analog shaping time constant of 800 ns to lessen pileup effects, and a 3.71 MHz digital sampling rate using 10-bit ADCs. A custom range encoding curcuitry5 preserves 18 bits of dynamic range by using four different analog amplification stages. A latched comparator curcuit is used to pick the best range, two rangebits are set accordingly, and the analog signal routed t o a 10-bit ADC. These 12-bit samples of 24 channels are multiplexed into one 1.5 Gbit/s fiber

659

optic link to off-detector readout modules. Here they are converted back to the full energy range, stored in a pipeline, and copied to the first level trigger. If the latter accepts the event, the corresponding samples are read from the pipeline. Peak energy and time of the original pulse are determined (“feature extraction”) and passed on t o the next trigger level.

2. Feature Extraction and Digital Filter

A parabolic fit to the waveform determines the energy. The hit time is calculated from the ratio of the first to zeroth time-weighted moment of the raw waveform sum. A matched digital filter is applied to samples of less than 30 MeV in order to reduce background and electronics noise. It achieves up to 50% reduction of total hit multiplicity under normal beam conditions. The digital filter enhances the signal-to-noise ratio for known signal shape and background characteristics. It is implemented as a linear transformation of the raw waveform given by a set of samples x i into filtered samples yi = C ja j x i + j . The optimal weights aj are determined for each crystal using clean pulse shape measurements and auto-correlation coefficients of the noise. For data taking the noise consists of constant electronics noise and beam-related background which may change its characteristics with luminosity and other beam parameters. The weights are calculated based on non-triggered waveforms and can be easily adjusted whenever necessary. For calibrating the EMC with a radioactive source2 the weights are optimized to eliminate electronics noise.

3. Diagnostics and Monitoring

Live monitoring of currents, voltages, temperatures, etc. is provided using EPICS‘ software tools. Characteristic parameters like crystal occupancy, hit timing and multiplicity are extracted from the data as it is recorded. The shift crew is automatically alerted when a parameter drifts out of tolerance. With a few hours latency, the now fully reconstructed events provide higher-level benchmark parameters like mass and width, energy-momentum ratio, and track-cluster matching efficiency. Daily calibrations using a lightpulser system check the complete readout path. A xenon flashlamp combined with spectral filter, attenuator, and light mixer delivers a spectrum equivalent to all but the lowest energies via optical fibers to every crystal. A reference crystal calibrated with radioactive sources provides normalization. The lightpulser system is also used to test the linearity of the electronic calibration by scanning the spectrum.

660

4. Readout Electronics

4.1. Calibration

Calibration of the readout electronics is performed by injecting a known charge into the preamplifier. Two capacitors - one small, one large - are used to cover the entire energy range. Each of the CARE chip’s four amplification ranges is calibrated individually first, then a simultaneous fit is performed over all ranges which produces the lookup table for recovering the full 18-bit dynamic range out of 12-bit (10 mantissa, 2 range bits) words. Electronic calibration data has proven to be useful to diagnose even subtle hardware defects, often down to a particular pin or solder pad. 4.2. Electronics Noise

3000 ~

j

2500

Mean RMS

4 971 0 4055

2000 1500 1000 500 0 0

5

10

15

20

25

x256 Ped RMS low-gain 1800 1600 1400 1200 1000 800 600 400 200

j

30

35

40

(ADC counts)

Mean

7.518 1.016

! RMS

0

0

5

10

15

20

25

x256 Ped RMS high-gain

30

35

40

(ADC counts)

Figure 1. Distribution of raw pedestal RMS widths across the calorimeter after turn-on in May 1999 (a), after low-voltage power supplies were equipped with filters in October 1999 (b), and after correcting an unstable curcuit on the 1 / 0 Boards and raising the bias voltage to 50V in December 1999 (c).

661 a, U .-0 U

a, U .-0

U

t Q

,

.

.

.

t Q

.

P

P

Endcap a, .-0 U

..

.. .. .. .. .. .. .. .. .. ..

..

, ,

. ,

..

..

. ,

..

..

Barrel Bwd

-0

..

t Q

..

..

..

..

0 ..

..

..

..

..

N 0

N 0

Barrel Fwd Figure 2. Average leakage current per readout diode over the period from June 1999 to March 2002.

The average electronics noise level today is equivalent to 320 keV, effectively reduced by digital filtering from 450 keV. Initially the noise level was much higher due t o switching noise from the low voltage power supplies (Fig. 1). This was overcome by implementing line filters. Correcting an unstable curcuit on the 1/0 Board and increasing the diodes’ reverse-bias voltage from 30 V t o 50 V yielded another 10% improvement. There has been no notable change in pedestal width after three years. On the other hand, the leakage current of the reverse-biased readout diodes have increased at a fast pace (Fig. 2). This effect, seen in the Belle calorimeter as well7, is currently under investigation.

662

10'

L.. .

.... ............ j ...

;

r........................ ..........................................

j

............

;

L....................

..........

;,.i .............................

.....

...........ill............. ^.

................. ..;.............. ..................... ............ ....................

i

.i ................

.............

................... .........................

....... .....

^

; ............ .:. ........... -..

3..

..........

:.

.............. ................. :....I.. ............ .-...:.........................

.i.............

.

...i.

l , , , , l . , , , l

-2

-1.5

-1

6.6

0

0.6

loglqE~Dlgl(0e

-2

-1.6

-1

6.6

I

.

I

. !

0

I

,

,

I

:'c

. I I

loglo(&,

>_&

0.");

Figure 3. Energy spectrum of all channels, corrected and uncorrected, 2000 (left) and 2002 (right) data.

4.3. Linearity

The electronic calibration corrects nonlinearities of the preamplifier response. However, at first the EMC data still showed nonlinear behavior in certain energy windows (left of Fig. 3) which had to be corrected offline. This effect resulted from the overlay of three independent problems. A curcuit on the ADC board could be driven into an oscillating state when the board was running under high load, for example during an electronic cdibration when all channels fire simultaneously. This was the strongest contribution to the observed nonlinearity and required energy corrections of up to 10%. In the 2000/2001 winter shutdown all ADC boards were modified to remove the oscillation. The second largest contribution ( 5 4% correction) turned out to be crosstalk between neighboring channels, propagating through the preamplifier cable shields. It is now corrected for as part of the electronic calibration procedure. The crosstalk pattern was surveyed for every channel inside an ADC board and its neighboring boards during the 2000/2001 shutdown. As part of the calibration, the magnitude of the crosstalk is determined by pulsing only every other channel while reading out the crosstalk signal in the unpulsed channels. The correction is then applied to each channel's calibration according to the pattern expected for this channel, scaled with the magnitude measured as part of the same calibration. After removing these two contributions, the EMC data was left with nonlinearities requiring corrections of up to 2%, however only in very narrow energy windows (right of Fig. 3). They appeared precisely where the range encoding curcuitry switches from one energy range to the next one. The underlying effect is still under investigation, however a correction procedure has been deployed recently which removes the nonlinearity.

663 4.4. Reliability

ADBs replaced

106s replaced

n

TRBs replaced Figure 4. Number of frontend electronics board failures per quarter since October 1999, for ADC Boards (top, serves 12 channels, 560 in system), 1/0 Boards (center, 72 channels, 100 in system), and Transition Boards (bottom, 720 channels, 10 in system).

Only a small fraction of the electronics components - readout diodes and preamplifiers - are not easily accessible. They are implemented in twofold redundancy to allow for at least 10 years of continuous operation without any major intervention. To date, there is only one out of 6580 channels which has been unusable since installation. One of two diodes/preamplifiers has been disabled for 11 other channels. The readily replacable components include ADC boards, 1/0 boards, and transition boards. While the failure rate was substantial immediately follow-

664

ing installation or reinstallation during a shutdown, it has subsided rapidly (Fig. 4). In the first months of operation the fiber optics transmitters were dying at an alarming rate. It was found the vendor had supplied CD lasers, unsuited for continuous operation under BABAR conditions. All transmitters were replaced with vertical cavity surface emitting lasers from a different vendor. After an initial infant mortality of 1% in the first two weeks there has been no failure since. The most common failure modes observed on front end electronics boards so far were, in descending order of occurrence: (1) Multipin connectors where one or many pins were disconnected from their solder pads. This was mostly due to the combination of weak solder joints and mechanical stresses from (un)plugging the connector, or bad alignment of the connector and its counterpart. Reinforcing the solder joints and fixing the alignment remedied this problem completely. (2) Unstable electrical contacts due to faulty solder joints. As was the case above, cold solder joints were a frequent flaw of the manufacturing process, and were easily fixed by resoldering the contact. (3) Failures of individual electronic components, apart from the fiber optics transmitters described earlier, have been limited to a few instances with no particular pattern as to the type of component involved. 5. Conclusion In its first three years of operation, the BABAR electromagnetic calorimeter has overcome a number of initial surprises and teething pains. It is now running in a most stable fashion, with little need for maintenance.

References 1. B. Aubert et al. [BABARCollaboration], The B B A R detector, Nucl. Instr. Meth. A479,1 (2002), SLAC-PUB-8569, hep-ex/0105044. 2. M. Kocian, these proceedings. 3. T. Hryn’ova, these proceedings. 4. W. Wisniewski, these proceedings. 5. D. R. Freytag and G. Haller, Analog Floating-point BiCMOS Sampling Chip and Architecture of the Babar CsI Calorimeter Front-End Electronics System at the SLAC B-Factory, BABAR-Note-285; IEEE Trans.Nucl.Sci.43:1610-1614,1996. 6. http://nun.aps.anl.gov/epics/ 7. B. Shwartz, these proceedings.

H1 CALORIMETER DAQ UPGRADE FOR HERA-I1

DIRK HOFFMANN, PIERRE-YVES DUVAL, CLAUDE VALLEE CNRS

-

Centre de Physique des Particules IN2P3 - Univ. Me'diterrane'e, Marseille, France E-mail: Hoflmann OCPPM.In2p3.fi

The H1 collaboration has performed an upgrade of its data acquisition system for the calorimeters in view of the HERA-I1 programme. A heterogeneous system based on 29K/VRTX, 68klOS9 and Vax/VMS was replaced by an integrated Unix cluster composed of two PPC/LynxOS VME boards and Sparc/SunOS stations, using TCP/IP protocols for inter process communication (IPC) and POSIX standards in general. Software transcription consisted of porting three essential functions: hardware setup, calibration datataking with a high serial data throughput and online datataking which emphasizes low frontend deadtime through a three level buffering by means of POSIX threads and messages. Low performance control tasks were programmed in Perl, the user interface has been written in Java. Although the very frontend electronics remain unchanged, a factor two increase in performance was obtained together with a manifestly improved environment for monitoring and diagnostics.

1. System Context and Environment

1.1. H l Calorimetry The H1 experiment' is a multi-purpose 47r detector, which has measured electron-proton and positron-proton collisions on the HERA-I accelerator since 1992. Since 2000, the accelerator underwent a major upgrade2 with the aim to increase instantaneous luminosity by a factor five and to deliver longitudinal lepton polarization. The detector was modified according to the new conditions around the interaction region, and the collaboration also used the time for further enhancements of the experiment. After the upgrade the Calorimeter Subdetector group henceforth comprises 0

0

0

a Liquid Argon (LAr) Calorimeter, which contains 44,352 cells and covers the major part of the polar angle, a Spaghetti Calorimeter (SpaCal) with 1,300 photomultiplier channels in the backward direction for scattered electron detection a small PLUG Calorimeter in the very forward direction for particle measurements at very low angle, which contains eight tiles for a rough

665

666

0

calorimetric measurement, and the analog readout part of the instrumented iron, called Tail Catcher for the measurement of hadronic shower tails leaking out of the LAr calorimeter, with a total of 3,888 channels.

They are readout by a common data acquisition (DAQ) system, the Calorimeter DAQ3 (CaloDAQ), which delivers in turn the data to the Central H1 DAQ.

1.2. Energy and T i m e Measurement The analog data of in total 49,548 channels are shaped in different electronic equipment, according to their origin and time behavioura. The final digitization for the energy measurement is performed by identical 12-bit ADCs, which are multiplexed, sequenced and read out by a total of 70 custom made DSP boards4 equipped with CPUs of the 56k generation5. Three additional DSPs of the same type treat special detector signals from photodiodes, trigger electronics and TDCs in the SpaCal partition. Certain channels with large signal dynamics are fed to the ADC battery through two amplifiers with different gains, thus allowing for an effective 14-bit precision. A third order polynomial calibration correction and zero-suppression are performed by the DSPs on the fly, without adding overhead to the generic sequencing time of the ADCs. In parallel to the precise and stable energy measurement, the LAr cells are used in the LAr Trigger' branch to deliver fast signals for time measurement and trigger purposes. To this end, all channels are fed into a total of 512 big tower sums according to their geometrical location and separately for the cells of the electromagnetic and hadronic calorimeter parts. Their signals are sampled after fast shaping by 8-bit Flash-ADCs at the HERA bunch crossing rate of 10.4 MHz. They fill a digital pipeline delivering a trigger decision after 2.3 ps, but internally keep a history of 256 bunch crossings. The readout is performed by eleven 56k DSPs, mounted on a different type of custom made VME board7. These are in charge of the calorimeter data transmission to the higher trigger levels and perform the readout of the big tower signal history around the nominal interaction time and the calculation of time weighted energy sums for signal pileup monitoring. The trigger raw data amount to 150 kB, which is comparable to the data volume in the calorimetry ADC branch.

*The pulse length in the SpaCal is of the order of 1 ns, whereas the drift time in the LAr Calorimeter nears the 0.5 ps.

667

1.3. Inter Crate Communication and VME Tree The CaloDAQ readout electronics consist of several hundred boards in VME standard, accommodated in 47 VME crates, which communicate through

17 DSP CALORIMETRY fonvardbmel IR

CALIBRATION

Figure 1. VME tree of the readout electronics in the two CaloDAQ branches: Crates on the left side contain the DSPs of the ADC/TDC branch. DSP crates for the trigger readout are placed on the right. Two big arrows indicate the main directions of the readout data flow. The Event Builder PPCs on top of these arrows feed their data into the Central DAQ Taxi boards, which are linked by an optical fiber ring to the other H I DAQ branches. All VME crates of the frontend electronics are connected by VICbus branches, starting from either of the Event Builder crates. The PPCs are connected to each other and a local Unix cluster fileserver through Fast Ethernet.

668

VICbusg connections on two and three vertical busses respectively in two separate branches. Fig. l illustrates this segmentation into the ADC/TDC branch and the Trigger branch, each equipped with a Taxi' connection to the Central DAQ. The readout hardware properly speaking is completed by crates for setup, calibration, interface to the Central H1 Trigger and monitoring. The address space in this VMEtree is managed by a custom-made VMEtree compiler, which allocates VME pages dynamically and translates the tree into the setup of the master VIC boards. Extensions of the Calorimeters like the Jet Trigger" are supposed to use the faster, smaller and more reliable PVICll connections for the same price. The PVIC setup will be integrated in the VICVMEtree compiler as well, thus providing the option to replace single VICbus branches transparently in future. The Event Builder PPCs are connected to each other and to a local Unix cluster by 100baseT connections (Fast Ethernet) on two Gigabit IP switches, which communicate through a glass fiber backbone. This is the gateway to all other Internet hosts and the switches can isolate the CaloDAQ hosts completely or limit the bandwidth for outsiders, if needed. 1.4. Hardware Choices f o r the Upgmde

The newly installed upgrade parts are hatched in fig. 1;foreseen extensions are shaded in light gray. Two PPCs on a R10212 VME board replace the former3 AMD 29k processors; and a Unix cluster replaces the anterior OS/S-Vax system in the central control. Control and management tasks (cf. sec. 2) have been partly shifted into the frontend, due to significantly higher computing power of contemporary CPUs. All PPCs run the real-time operating system LynxOS13. The extension by the Jet Trigger" project will demand additional capacities in the readout of the Trigger branch, which can be provided through a satellite P P C station on a PVIC'l bus. The Unix Cluster was extended from an existing Sparc station running SolarisOS. It is used as a fileserver for all LynxOS systems and included in the central DESY backup for maximum data integrity. Tests with a Pentium 111 under a recent version of Linux showed that there are no drawbacks when using this lower-priced alternative, except for the different byte order in the CPUs, which might imply some careful rewriting of offline analysis programs. 2. Software Design 2.1. Servers, Clients and Protocols

The Event Builder CPUs can be accessed through Internet protocol (IP). Firewalls and server internal host selection algorithms protect against unwanted

669

accesses. Both Event Builders have an identical client/server program structure (see fig. 2), which enables them to call each other in case of requests which need synchronization. The “entrance” for all DAQ requests is a Unix socket,

Figure 2. Software architecture of the new CaloDAQ for one Event Builder: On request from a DAQ client, the DAQ server daqd can launch one or several service processes, which access the VME bus. Service tasks which send back data, can write into a file, connect in turn to a data receiving server, or dump data into the Multi Event Buffers of the Central DAQ Taxi. The post office process daqpo is explained in the text.

which is bound by the daemonizingb server daqd. Depending on the kind of request, daqd selects VME service programs, which do the actual hardware access. In this way the exclusive access to VME can be granted if necessary. The daqd request protocol is not time critical and therefore plain text, which has proven to provide efficient ways for debugging or test of clients and servers independently from the current status of other parts of the project. Service programs include request for setup (rqs) of the VME boards, request for readout in standalone mode ( r q r ) , expert setup of the LAr Trigger ( t r i g l o a d , ssm) and high performance event building for real time data taking (eb, sec. 3). They are implemented as standalone programs, taking their configuration data from the standard input channel, which allows once more uncoupled (object oriented) testing of single blocks. In case they produce any data for the requester, these are formatted in H1 standard BOSi5 format and sent in binary form through an extra IP channel. As the response of daqd is strictly sequential and prioritized, the task of asynchronous status and error reports between processes has been delegated to the DAQ post office (daqpo) daemon. It allows message exchange in the DAQ through an unlimited number of sockets from various clients with low priority compared to other tasks on the CPU. It broadcasts incoming messages to all connected clients. In addition, daqpo translates Central DAQ run requests ~~

bdetached from its calling (parent) process and running in background, typically started at system boot time

670

coming through the traditional VMEtaxi channel, as long as the IP protocol is not yet implemented there. 2.2. Choice of Pmgmmming Languages

The different elements of the DAQ software have been encapsulated to the most possible extent during design and are only seen by others through their interface definition. This enabled us to choose different technologies according to the aim and the requirements of each of the programs. daqd: modest requirements on reactivity and data throughput on one hand and high flexibility for the implementation of new requests on the other led to the decision for a compiled scripting language, Perl 5 , which is optimal for the recognition of text patterns and reliable for long uptimes. VME service programs: The high demand for performance and the usage of system level C routines made it necessary to use compiled programs for these tasks. The chosen standard for message queues and threads was POSIX'8. daqpo: was designed in the style of all other VME service programs. DAQ Clients: Up to now, and not including the trivial method by a telnet command, three different technologies have been used to design clients for the new DAQ. The main user interface (cf. sec. 4) has been written in Java; a very complex trigger analysis and calibration package had been inherited in FORTRAN and was adapted to the TCP/IP protocol by means of a C wrapper; and some debug clients and receiver servers have been written in Perl. The combination of all programs for the standard electronic offline calibration resulted in a gain in performance of an approximate factor two. This will allow to control and perform the calorimeter calibration more often and more reliably, as the regular idle time of the accelerator without beam is usually not more than 30 minutes.

3. Real Time Behaviour The most challenging project in DAQ systems is the readout part, where optimal performance is needed in interplay with many other systems of the detector. The event building programs of the two CaloDAQ branches are split into three main threads, which are shown in fig. 3 for the ADC/TDC branch'. CThe Trigger branch is designed identically with the exception that DSP and IT routine have three front end buffers.

671

Figure 3. Program threads (in ovals) of the real time event building program, from left to right with decreasing priority: The IT provides the buffering for unbuffered front end registers. The bulk of data is stored in one of two DSP buffers. The BM gathers the primary buffer blocks and writes them to a circular buffer, from where the EB dispatches them among possible consumers after formatting.

As not all frontend hardware contains internal buffers like the DSPs described above, an interrupt routine (IT) is used to readout these registers and hands them over t o the block mover (BM) in two buffers. The BM task acts, when the first order dead time has finished, because at least one front end buffer for the next event is available in the DSPs. It stores the data from the front end buffers into a ring buffer, which can hold approximately twenty events, depending on their size. A third task is in charge of the actual event building (EB), which means data formatting and delivery to the Taxi board of the Central DAQ or optionally a monitoring file or IP client. We recorded fig. 4 as a benchmark for the new CaloDAQ. The slope of the

0.35 7

jo ,

j

,

A

0.3 :

0.25 7

<

<

,

I

6 ,

:

:

Aj

I

$

d

<

A

A

-:

5

L4

,

Output Rote / Hz

Figure 4. Deadtime behaviour of a single CaloDAQ partition and of the complete CaloDAQ system in comparison to the behaviour of the DAQ for the whole experiment: The first order or synchronous deadtime leads to a linear rise for moderate rates, while events are sufficiently distant in time to allow for the asynchronous tasks to be executed in between. The hypothetic lines for 1.0 and 2.0 ms are drawn for comparison. The available time for the secondary tasks becomes sparser with increasing datataking rate; the deadtime “clambers the wall of saturation”. The maximum rate is determined by the average overall time for treating a single event. The theoretical limits for some values are drawn vertically as guidelines.

672

first order deadtime for moderate rates is in agreement with the independently measured deadtime of less than 1.2 ps. Inclusion of all partitions yields a performance which is better than the overall experiment limitations. Saturation presently occurs at more than 400 Hz for the ADC/TDC branch and 210 Hz for the Trigger branch. 4. User Interface

A graphical user interface for a comprehensive series of standalone tests and hardware check programs has been implemented in Java. The physical and statistical evaluation of the results is done with routines of the Java Analysis Studio14. It performs also the regular calibration level ramps including an elaborate algorithm for sequential pulsing of the generators, in order to avoid crosstalk effects. Sums over 100 events at the same level are performed in the DSPs, which minimizes data transmission over the VICbus. The present limitation is however exactly the speed of this transmission.

5. Conclusion The speed improvement by a factor two is a convincing and desired achievement. The performance is now limited by the unchanged ADC/DSP architecture and VME interconnection on the front end. The manifest benefit of the upgrade is the innovative and nevertheless 100% backward compatible redesign of the core readout and control software of the CaloDAQ in an integrated LynxOS-Solaris Unix cluster, using exclusively Unix standards and conforming wherever possible to POSIX conventions. It puts a comfortable environment for the development of future survey and analysis programs at the disposal of the expert physicists.

Acknowledgements We thank all our colleagues between the Mediterranean and the Baltic Sea, who participated with smaller and bigger contributions in the preparation and completion of our data acquisition system. We would like to mention in particular Jean-Baptiste Chabane for his work on the Java user interface, Daniele Laugier for the fundamental studies on the Unix data transfer, Sepp Huber, Tibor KurEa and Burkhard Reisert for the implementation of the trigger loading software in the new framework, FrantiSek Krivan for the continuous support for the frontend hardware and the Sun cluster service, and last but not least Etienne Barrelet for the original plan of “our new DAQ”, his unequaled way of thinking and his continuous supply with new ideas.

673 References 1. H1 Collaboration, The H1 Detector, Preprint DESY-H1-96-01 and Nucl. Inst. and Meth. A 386,310-347 & 348-396 (1997)

2. L. Suszycki et al., HERA Luminosity Upgrade, DESY-Report (July 1997); updated version available at D http://vuv.desy.de/”parker/IR/Report970702/Report970702.html The HERA Luminosity Upgrade, ed. U. Schneekloth, DESY HERA 98-05 (1998) 3. F. Descamps, C. Vallke, Data Acquisition for the H1 Calorimeters, Internal note H1-10/92-256 and Proc. of the 111. International Conference on Calorimetry in HEP, Corpus Christi (1992) 4. Digital Signal Processor Board DSP 8510, User’s Manual, Revision 0.2, Creative Electronic Systems, Geneva (1990) 5. Motorola DSP56000 Digital Signal Processor, User’s Manual, Motorola Inc. (1986) 6. T. Carli et al., Performance of the H1 LAr Trigger in 1994, Internal note H107/95-445 and Contr. to the Int. Europhys. Conf. on HEP, EPS0644, Brussels (1995) 7. F. Blouzon et al., Manuel utilisateur du DSPT et de la carte t o , LPNHE Paris, unpublished (1992) 8. W. J. Haynes, Experiences at HERA with the H1 Data Acquisition System, Proc. Int. Conf. Computing in HEP, Annecy, France, 151-161 (1992) 9. VIC 8250 VMV to VME One Slot Interface, User’s Manual ver. 5.0, DOC 8250/UM, Creative Electronic Systems, Geneva VICbus, VME Inter-Crate Bus, ISO/IEC JTC 1/SC26, ISO/IEC 11458 10. H1 Coll., Proposal to Upgrade the LAr Calorimeter Trigger: The Jet Trigger, DESY-PRC-99-02 (1999) 11. PVIC Systems, PCI to PCI connections, User’s Manual, v. 4.0, DOC 3925A/UM, Creative Electronic Systems, Geneva (2000) 12. RI02, PowerPC based RISC 1 / 0 Board, User’s Manual, v. 7.0, DOC 8062/UM, Creative Electronic Systems, Geneva (2000) 13. LynxOS releases 3.0.1 and 3.1.0 by Lynux WorksTM, presented on D

http://vuv.lynxos.com

14. Java Analysis Studio, Java library for physics analyses; available from D

http://vvv-sldnt.slac.stanford.edu/jas/

15. V. Blobel, The BOS System. Dynamic memory management (1977) 16. G. Cozzika, The H1 Detector, Proc. of the 111. International Conference on Calorimetry in HEP, Corpus Christi (1992) 17. M. Fleischer, The Performance of the H1 Liquid Argon Calorimeter, Proc. of the VII. International Conference on Calorimetry in HEP, Tucson (1997); Talk and Proc. available from D

http://vvv-hl.desy.de/calo/conf

18. Institute of Electrical and Electronics Engineers (IEEE), Portable Operating System Interface (POSIX) - Part 1: System Application Program Interface (API), IEEE 1-55937-061-0; I S 0 9945-1 (1990) “POSIX 1” IEEE, POSIX - Part 2: Shell and Utilities, IEEE 1-55937-255-9 “POSIX 2” IEEE, POSIX - Part 1: API - Amendment: Realtime Extensions, IEEE 155937-375-x “POSIX 4”

This page intentionally left blank

Ionization Calorimetry Covener:

P. Schacht

P. Schacht

Covener’s Report

A. Besson

Argon Purity Measurement of the DO Calorimeter

N. Parua

The Run I1 D 0 Calorimeter : Electronics Upgrade and its Performance

S. Rodier

ATLAS LAr EM Calorimeter: Construction and Uniformity of Response

D. Zerwas

Performance of ATLAS EM Modules in Test Beam

M. Fincke-Keeler

The ATLAS Hadronic Endcap Calorimeter

A. Kiryunin

Performance of the ATLAS Hadronic End-Cap Calorimeter in Beam Tests

*KK. Joo

ATLAS Forward Calorimeter (FCAL)

G. M. Jones

A High Resolution Luminosity Monitor for SLAC Experiment E 158

*Written contribution not received

675

This page intentionally left blank

IONIZATION CALORIMETRY

P. SCHACHT Max-Planck Institut fur Physik, (Werner Heisenberg Institut) Fohringer Ring 6, 80805 Munich, G e m a n y E-mail: p ys Omppmu.mpg. de (Convener’s Report)

Eight talks have been presented in this session, mostly devoted to the Liquid Argon calorimetry. The DO upgrade to cope with the high luminosity in Run I1 of the Tevatron at FNAL and ATLAS, entering the final phase of construction and beam calibration, are pronounced examples where ionization calorimetry has employed successfully novel technologies to reach new frontiers in high energy physics detectors. Argon purity has a strong impact on the ionization signal in liquid argon calorimeters. Auguste Besson presented the argon purity measurement of the DO calorimeter. The goal is to achieve a purity of better than 0.5 p p m oxygen equivalent and to monitor the purity with high precision. To achieve this goal a and B sources are used. With well defined pollution samples a calibration can be carried out yielding finally errors at the level of 0.15 p p m . Nirmalya Parua presented for the DO calorimeter the electronics upgrade and its performance. The reduction of the minimum bunch spacing from 3.5 ps to 396 n s and finally to 132 ns in Run I1 B required major modifications of the readout. Preamplifiers, shapers and SCA’s as well as the calibration and timing control system have been replaced. The rapid progress in understanding the new system has been demonstrated from the analysis of timing, noise and calibration studies as well as from recent studies of physics benchmark processes. Stephane Rodier described in his talk in detail the ATLAS Liquid Argon Electromagnetic Calorimeter. To guarantee the required uniformity and linearity of the response, low noise and energy resolution - and here in particular the constant term contribution - the quality control throughout the whole production and assembly phase is crucial. The related measurements prove that the high quality in production can be maintained through the ongoing construction phase.

-

N

677

678

The test of the ATLAS Liquid Argon Electromagnetic Calorimeter modules in beam is essential to prove that the ambitious performance goals can be achieved. This is demonstrated in the talk of Dirk Zerwas. The precise reconstruction of the signal shape is one of the crucial elements in this chain. The signal to noise ratio for muons of 7.11 f 0.07 demonstrates the low noise level achieved. For electrons an energy resolution with a sampling term below 10% and a constant term below 0.3% has been obtained. These results are in good agreement with MC simulation. The Hadronic Endcap Calorimeter of ATLAS was the subject of Margret Fincke-Keeler’s talk. The production is well advanced and nearing completion. Again, quality control during the production phase plays an important role. This holds for the cold electronics, preamplifier and summing amplifier system - operated for the first time in liquid argon - as well as the for the mechanics. Regular beam tests with electrons, pions and muons prove the required uniformity of the response. The precision of the signal reconstruction is at the level of 0.5% - 1%. The temperature and high voltage dependence of the electron response has been studied in detail and allows for a precise extrapolation to the ATLAS operation regime. Andrej Kiryunin discussed the results of the beam calibration of the ATLAS Hadronic Endcap Calorimeter in detail and compared the results with MC predictions. For electrons and muons the data agree fully with the expectations from simulation. For pions there is substantial progress in the simulation, but the data for the energy resolution or electron to hadron ratio show still some deviation from the prediction. The status of the ATLAS Forward Liquid Argon Calorimeter has been presented by Kyuang Kwang Joo. Radiation hardness in this forward rapidity region (3.2 < q < 4.9)and compactness are the prime requirements. The small argon gap size - varying from 250 p up to 500 p - keeps the ion build up at an acceptable level. Again, quality control during the production is one of the key issues. In the last talk Gary Mark Jones described the High Luminosity Monitor for the E-158 experiment at SLAC. The goal is a precision measurement of the weak mixing angle in Maller scattering. To cope with the extremely high electron intensities radiation hardness and robustness are key issues in the design. Aluminum chambers with parallel plate ionization collection using nitrogen as buffer gas have proven to match the requirements. The present level of precision is 150 ppm in the intensity asymmetry resolution and a nonlinearity well below the 2% level.

ARGON PURITY MEASUREMENT OF THE DO CALORIMETER

A. BESSON, G. SAJOT Institut des Sciences Nucle'aire 53 avenue des Martyrs, 38000 GRENOBLE, FRANCE E-mail: abessonOin2p3.fr, [email protected]

The DO Argon purity Test Cell (ATC) measures the 02-equivalent pollution of a liquid argon sample extracted from the calorimeters. The cell consists in two radioactive sources, a (241Am) and p (lo6Ru), immersed in liquid argon which produce ionizations. Then, the created charges are drifted by an adjustable electric field. Due to absorption of e- by electronegative impurities, the total collected charge depends on the 0 2 pollution. The device is an upgraded version of Run I DO ATC. Its present sensitivity is estimated to be better than f 0.15 ppm. A pulser with adjustable frequency is used to calibrate the electronics. The electric field is ramped step by step between 5 kV cm-' and 15 kV cm-'. Furthermore, for calibration purpose, the ATC is equipped with a system to pollute a pure Argon sample, with a given amount of 0 2 . The setup and method of analysis of a and B measurements will be described. Measurements of the purity of the gas Argon from the three calorimeters (Central, North and South End Cap) will be presented.

1. Introduction The signal of drifting electrons in liquid argon is highly affected by recombination with 0 2 (or other electronegative molecules)l, so that the Argon purity needs to be precisely monitored before filling the DO calorimeters. Pollution above 1 ppm would deteriorate the signal significantly2. The principle of the measurement is the following : two radioactive sources ( a 241 Am and lo6 Ru) produce ionizations in the liquid Argon between two electrodes where an electric field can be adjusted (fig. 1). The collected charge depends both on the electric field and on the 0 2 pollution. 1.1. The ATC and the modifications for Run 11

The ATC developed, built and operated for Run has been upgraded5i6 and tested at Grenoble from January to June 2000. Main modifications concern : 1314

(1) the electronics which is based on NIM modules (developed for the AT-

679

680

LAS experiment7). A pulser with adjustable frequency and with the same shape as the signal (width 1.5 ps, amplitude 50 mV) is used for calibration purposes, (2) the possibility to pollute at a given level a pure Argon sample, to calibrate the system, (3) the data acquisition interface card (National Instrument AT M I 0 16E10 ) and the acquisition program written in LabWindows/CVI 0.

-

The cell ( fig. 1) contains two sources : an a source (241Am, 5.4 MeV, activity 2 kBq which has already been used for Run I) and a /3 source (losRu, 3.5 MeV, with an activity of 40 kBq). These sources have been manufactured by Isotope Products in Burbank (USA) by electro-deposition on a 2.5 cm diameter stainless steel electrode. The gap between cathode and anode is 2.15 mm. In addition for the /3 source, a second gap of the same width is used as a trigger. The HV is varied between -5 and 15 kV/cm for the a and between -1 and 15 kV/cm for the ,B. The field in the trigger gap is set at 10kV/cm.

To Pre-Amplifier Figure 1.

Alpha and Beta sources layout

1.2. Cooling procedure Argon is filled in gaseous phase by “cryogenic pumping” It is cooled down by liquid Nitrogen. By adjusting the Nitrogen pressure, one can get a stable pressure of Argon at 5 f 1 PSI. For cross-checking purposes, three temperature probes (Pt100) are placed close to the cell and show that temperature (T=89.0 f 0.5 K ) is stable with variations below 1 K. This is consistent with the Pressure-Temperature phase diagram2. The 10 liters of liquid Argon, which are necessary for the measurements, are usually obtained in 3 hours.

-

681

1.3. Electronics and Data acquisition

The signal goes through the following electronic chain: (1) a preamplifier (designed at ISN), which is placed outside of the cryostat as close as possible to the cell, in order to minimize noise, (2) a shaper, a constant fraction discriminator (CFD), a Charge Voltage Convertor, (3) a data acquisition card (National Instrument) to allow data analysis on a PC, using LabWindows/CVI.@. In addition to the source signal, the calibration pulser signal and the pedestal are also measured on line (with typically 40000 events in total for each high voltage value). For a given LAr sample and for twenty values of the electric field, the acquisition time is around 1 hour. The pedestal, the pulser and the a signals are fitted with a Gaussian distribution while the B signal is fitted by a Moyal fits. The mean values are used to compute the pollution. 2. Measurements with a source 2.1. Absorption factor a particles deposit almost all their energy over 10 - 20pm. This induces a constant current over a period of time equal to the drift time t d between the electrodesg. For a sample with pollution p and for each value of the electric field E , a normalized ratio X ( E , p ) is defined as:

X ( E 1 p )=

< signal(E,p) > - < pedestal > < calibration > - < pedestal >

(1)

Where < signal(E,p) > is the mean value of the signal produced by the source, < Calibration > and < pedestal > are the mean values of the calibration signal and of the pedestal. In this ratio possible fluctuations of the gain of the electronic chain cancel. Collected charge Q ( E , p )and absorption A b s ( E , p ) are related bylo:

+

where QO is the total charge produced, ln(1 5) describes initial recombination of electron-pairs, and A b s ( E , p ) is the absorption factor. Since no theoretical model provides a satisfactory description of this recombination, a parameterized formula obtained by the ISN-ATLAS group7 is used :

<E = a ( 1 - c x e-bE)

(3)

682

with a = 474 f 1.4 kV/cm, b = 0.143 f 0.006 cm/kV and c = 0.403 f O.OIO1l Furthermore, the absorption factor is:

with:

X(E,P)= a x E/P

(5)

where X(E,p) is the absorption length, d is the gap between the electrodes, and a = 14.2 f 1.4 mm2 ppm/kV(12). Eventually, assuming that X ( E , p ) is proportional to & ( E , p ) ,one gets:

C,,,., is a normalization term that, in particular, takes into account charge to voltage conversion factor. The equation 6 is fitted as a function of E , where p and CnOrmare the parameters. The pollution p affects mostly the curvature of Abs(E,p)vs E . 2.2. Calibration and e m r s

2.2.1. Main sources of uncertainties on p measurement The main sources of errors on the 02-equivalent pollution are due to the :

fit error and statistical error on the peak determination for signal, pulser and pedestal : f 0 . 0 7 ppm. error on the gap measurement : 2.5 % precision of the high voltage settings : 2 % error on parameters a, b, c and on trapping constant a. systematic errors (electronic fluctuations, non linear response of the preamplifiers ...) are more difficult to evaluate and can be estimated with a calibration, as explained below. 2.2.2. Calibration Equation 6 shows that p is an absolute measurement. However, to crosscheck the validity of the fit, the possibility of doing calibrated pollutions has been added. First, the cryostat is filled with high purity Argon (for example certified with less than 0.1 ppm 02). Then, a given volume of pure 0 2 is added, followed by an Argon gas flow (coming from the cylinder) during 30 minutes, so that 100% of the 0 2 is transfered into the cryostat. Fkom the quantity of

683 0 2 and LAr in the cryostat, a nominal (i.e. expected) value for the pollution is derived. Experimentally, one has found that measurements get stable and are consistent with the nominal pollution after one hour. The relative uncertainty of the nominal pollution is 10 %. Figure 2(a) shows different calibration measurements. The standard deviation of measurements varies from 0.05 to 0.1 ppm depending on the pollution range. We choose to set it at 0.1 ppm for all measured values to be conservative. To cross check the consistency of the measurements, one can compare nominal and measured pollutions. A linear fit is then realized (pnominal = a X pmeasured b ) , giving the following results:

-

+

( 1 ) a = 0.95 f 0.05 (2) b = 0.03 f 0.04 Where a and b are the fit parameters, pnominal is the nominal pollution and pmeasured is the measured pollution. Nominal and measured pollutions are in very good agreement within a large range (from 0.1 to 5 ppm). So the calibration validates the procedure and provides an estimation of the systematic errors by adding quadratically the error on a and on pmeasured. Figure 3 shows the total error of the measurements: for low pollution measurements ( p < 1 ppm), uncertainties are lower than 0.15 PPm. 3. Measurements with a 3.1. Properties of the

p source

p source

/3 are minimum ionizing particles so that most of them cross the gap depositing energy along their path producing a linear decreasing currentg. The signal of the second gap is used as a trigger to provide a good rejection of the electronic noise. Because of this minimum ionizing behavior, concentration of ion-electron pairs along the track is small, so that recombination effects are much lower than for a particles. 3.2. Calibration

As for a measurement, a linear fit is performed and the estimation of the systematic errors is obtained with the same method (fig. 3). The large error at high pollution is due to a lack of calibration data taken in this domain. For each calibration sample an empirical fit is done using the following expression:

< signaZ(E,p)> - < pedestal > < pulser > - < pedestal >

=a

+ ( d + c x E + g x E 2 ) x e(-b*E) (7)

684

(a) Q: Absorption meas-enb ofssmplea with known pollution va electric field. (notation measured/namhal is used for the results on the 6gure. (b) B: central Calorimeter measurnmmt (Dec.2001). F i g u r e 2.

Where the same conventions as for equation 1 are usad. a, b, e, d and g are the parameters of the fit. More precisely, the fitted quantity is the expression on the left hand side of the previous equation divided by its aymptotic value reached for high electric fields (E > 12 kV/cm). The calibration has shown that d is a linear function of the pollution p . Figure 2(b) s h m a measurement of the Central Calorimeter compared with two calibration samples.

3.3. Results

Figure 3 shows that for low pollution, @ source measurementsare slightly better than a source measurements. During the years 2000-2001,several campaigns of measurements have been performed. Results are summarized in table 1. It clearly appears that the purity of all the three calorimeters are far better than 0.5 ppm. This is very satisfactory after 5 year$ of storage in the Dewar, proving that the Argon cryogenic system has no leak. So we conclude that the Ar purity of the d e m has not been degraded and can be wed for Run I1 wihout any correction. The analysis of the first data of the DO edorimeter are in good agreement with expectation and confirmed a p s t e h r 9 this r e ~ d t ~ ~ .

685

-k

Aloha ond Beta. Error vs Measured Dollution

Q

v

+-

C

0.2

P ln 0 e,

E I

0 L L

W

0.15

0.1

0.05

0

0

I

I

I

I

I

0.2

0.4

0.6

0.8

1

I

I

1.2 1.4 Measured pollution (pprn)

Figure 3. Errors on a and p measurement as a function of the pollution.

Acknowledgments We are particularly grateful to Yves Carcagno and Gabriel Mondin for their great work on the ATC, and to Philippe Martin (ISN-ATLAS group) for his help during the tests of the ATC in Grenoble. Many thanks to John Kotcher for his support of the project, to Gerald Blazey, Christian Zeitnitz and Bob Hiroski for their advices, and to Dan Markley for his help during installation and data taking, and for the numerous discussions we had with him during measurements. Germain Bosson and Robert Foglio have realized the preamplifiers and electronic modules. Solveig Albrand provided the initial version of

686 Table 1. Summary of all the measurement performed with Sample

Extracted Phase

ALPHA

gas

liquid

0.34 f 0.12 0.33 f 0.12 0.25 f 0.12 0.37 f 0.12

0.18 f 0.10 0.19 f 0.10

gas gas

0.49 f 0.12 0.16 f 0.12

0.38 0.21

gas

0.07 f 0.12 0.11 0.12 0.17 0.12

dewar dewar dewar dewar C.C. N.E.C. C.C. N.E.C. S.E.C.

liquid gas

I

-gas gas

I

*

(Y

BETA

I

I

f 0.11 f 0.10 0.10 f 0.10 0.09 f 0.10 0.14 f 0.10

and

0 sources. date

I

July 2000 July 2000 October 2000 October 2000 December 2000 December 2000 December 2001 December 2001 December 2001

the Acquisition soft. References 1. C. Brassard, Liquid ionization detectors, NIM 162, 29 (1979). 2. ATLAS Technical Design Report, p.76, (December 15th 1996) Gas encyclopedia. Elsevier Scientific Publishing Company (1976). 3. G. Blazey for the DO collaboration, Monitoring liquid argon purity at DZERO, Proceedings of the first Conference on Calorimetry in high energy physics, World Scientific October 2gth - November lst 1990. 4. G. Blazey, DO Liquid Argon Monitoring Hardware, DO Note 000940, 3/1/91. 5. Hardware details and pictures of the ATC can be found at: http://isnwww.in2p3. fr/dO/cryostat.html 6. A. Besson and G. Sajot, dO internal notes 003799 and 003827, (December 2000). 7. M.L. Andrieux et al, Response of an (Y source mounted in a liquid argon ionization cell and read out in full charge collection mode, NIM A 427, 568 (1999). 8. J.E. Moyal, Theory of Ionization fluctuations, Phil. Mag 46, 263 (1955) 9. W.J. Willis and V. Radeka Liquid-Argon ionization chambers as total absorption detectors NIM 120,221 (1974) 10. J. Thomas and D.A. Imel, Phys. Rev. A 36, 614 (1987). 11. Parameters fitted by P.Martin ATLAS group ISN Grenoble (private communication) 12. W.Hofmann et al., Production and transport of conduction electrons in a liquid argon ionization chamber, NIM 135, 151 (1976). 13. A. Juste, DO Status and first results from Run 11, Moriond, (March 2002).

THE RUN I1 DO CALORIMETER : ELECTRONICS UPGRADE AND ITS PERFORMANCE

NIRMALYA PARUA State University of New York, Stony Brook, USA E-mail: [email protected] (For the 00 Collaboration)

The front-end electronics of the DO calorimeter has undergone a complete redesign to operate successfully with the shorter bunch spacing time at the RunII of the Fermilab Tevatron. After the successful installation of the new electronics, first results of the commissioning, the calibration and the performance of the calorimeter will be presented.

1. Introduction

The DO detector performed extremely well during Run I (1992-1996) at the Fermilab Tevatron. The calorimeter is a key component of the DO detector and has provided excellent accuracy and resolution in measuring the position and energy of electrons, photons, jets and inferring the presence of neutrinos in events. The discovery of top quark and subsequent measurement of its mass and production cross section relied heavily on the calorimeter. Also calorimeter has been instrumental in the precision measurement of the mass of the W boson; the search for new particles and anomalous couplings; verification of quantum chromodynamics; and important measurements of electroweak physics. During RunI Tevatron delivered to both CDF and DO, about 125 pb-l of integrated luminosity with peak instantaneous luminosity of 1.6 x 1031 cm s-I. The accelerator has undergone major upgrade, in order to deliver the larger luminosities needed to study low cross section and high p~ processes such as top, Higgs, W/Z and physics beyond the standard model. The goal of this upgrade is to deliver at the first phase of RunII, 20 times more luminosity compared to RunI. Two main components of the accelerator upgrade are the addition of the Main Injector and the anti-proton recycler. The center of mass energy of collision has also been raised from 1.8 TeV to 1.96 TeV. The Tevatron has been operational since March 2001. The higher instantaneous luminosity

-'

687

: H>m
Figure 1. (Left) Side view of the upgraded D0 detector. (Right) side view of one half of the tracking system.

requires a reduction in the bunch spacing time from the RunI value of 3.5 ^s to 396 ns for the first phase and then to 132 ns for the second phase of Runll. The upgrade of the D0 detector is designed to enhance its capabilities from RunI and to accommodate the new beam structure in its readout. In the following we will briefly overview major upgrades of the D0 detector and then discuss the upgrades for calorimeter in more details. 2. Overview of the D0 Detector Upgrade Figure 1 depicts the overview of the upgraded D0 detector. The tracking system is completely replaced with a Silicon Micro-strip Tracker (SMT) and a Central Fiber Tracker (CFT) with scintillating fiber readout. The entire tracking system is contained inside a 2.8m superconducting solenoid that produces a 2 T magnetic field. To compensate for the energy loss in the solenoid, central and forward preshower detectors are introduced. Both these detectors consist of scintillator strips fitted with wavelength shifting fiber readout. Although the uranium liquid-Argon calorimeter remained untouched in Runll, most of the readout system was replaced to handle the smaller bunch spacing time. The Photo multiplier tubes used for the scintillator based Inter Cryostat Detector (ICD) were relocated to a region of lower magnetic field. The Runll muon system includes scintillator based detectors used to provide a fast trigger. Also, in the forward muon system the Proportional Drift tubes were replaced with planes of plastic Mini Drift Tubes (MDTs). Massive shielding is used to reduce backgrounds originating from scattered protons and antiprotons.

689

To deal with the high event rate, the trigger system needed a full replacement with a 3-tier system and fast pipelined front-ends. The newly designed first stage (L1 trigger) incorporates the signals from the scintillator detectors, tracking and the calorimeter detectors. L1 trigger reduces the input rate of 1.7 MHz to 10 kHz. L1 trigger decision time is 4.2 ps and it is essentially deadtimeless. The L2 trigger consist of dedicated processors that must make the decisions in 100 ps and further reduce the rate to 1 kHz. The third stage trigger (L3) consist of a farm of commercial processors that utilize fully digitized information to reduce the L3 accept rate to 50Hz. 3. Inter Cryostat Detector

In the intermediate region of the calorimeter (1.1. 5 1 ~ 51 1.4) energy measurements are degraded due to un-instrumented regions and the rapidly changing material profile. The Inter Cryostat detector (ICD) restores the energy resolution of jets in this region and the missing transverse energy ($T) of an event by providing additional sampling. The ICD detector consist of 384 scintillator tiles with wavelength shifting fibers (WLS) read out using photomultiplier tubes. The tile signals are transported to a region of low magnetic field by 5-6 m long clear fiber waveguides. The 384 tiles each cover a region of 0.1 x 0.1 in - 4 space and are grouped into two sets of 16 supertiles per end cap (EC). 4. Liquid Argon Calorimeter

A detailed description of the DO calorimeter can be found elsewhere'. In this section we briefly summarize important features of the calorimeter. The DO calorimeter uses liquid argon as the active medium and with depleted uranium plates as the primary absorber material except in the the outer layers where stainless steel and copper plates are used. The schematic design of the calorimeter is shown in Fig. 2. The central calorimeter (CC) covers roughly a region in 1171 5 1 and the two end calorimeters (EC) extend the coverage up-to 1171 M 4. The Inter-cryostat Detector (ICD) covers the overlap region between CC and EC. The calorimeter is modular and is finely segmented in the transverse and longitudinal directions of the shower. Three distinct types of modules are used in the calorimeter: a finely segmented electromagnetic section with 2mm uranium absorber plates; a fine hadronic section (FH) with thicker uranium absorber plates than EM section; and a coarse hadronic section (CH) with thick copper and stainless steel plates. Each module consists of a stack of interleaved absorber plates and signal boards as shown in the Fig. 2. The absorber plates are separated from signal boards by a liquid argon gap of 2.3 mm. Signal

690 I

C# UQUID ARGON CALORIMETER

Figure 2.

The DO Calorimeter.

boards are constructed by laminating a copper pad with two 0.5 mm thick pieces of G10. The outer surface of these boards are coated with resistive epoxy. An electric field is applied by grounding the metal absorber plate and by biasing the resistive coat to 2kV. The electron drift time across the gap is M 430ns. This sets the time scale for signal charge collection. Signals from several pads at approximately the same 77 and 4 are ganged together in depth to form a readout cell. The readout layers are arranged into pseudo-projective towers. 4.1. Upgraded Calorimeter Electronics

The goal in replacing the calorimeter electronics was to maintain the same overall noise performance achieved in RunI. The noise is the sum of electronic, uranium and pile-up contributions. To achieve this goal it was necessary to optimize the relative noise contributions and to add analog storage device on which to store the calorimeter signals until a L1 trigger decision could be made. This required a new strategy for baseline subtraction. A schematic diagram of the Run11 calorimeter readout is shown in the Fig 3. The signal from each of 55,000 cells is brought to the feed-through port (a sealed interface between the inside and outside of the cryostat) on a 30R co-axial cable. The original 115 52 twist and fiat cables from the feed-through port to the preamplifiers (,- 5-7 m) were replaced with 30 R coaxial cables to provide: impedance matching to the preamplifiers; faster charge collection; individual channel timing equalization. About 55000 individual hybrid charge sensitive preamplifiers were built for the calorimeter signals. The new design has lower noise and improved drive capability. Two additional driver stages were added to drive the single ended 115 R cable to the shaping circuitry, the base line subtracter (BLS), and to allow for the option of differential outputs. The preamplifier uses two jFET’s (2sk369) in parallel to reduce the electronic

691

Figure 3.

Readout chain of the calorimeter in Run11

noise (at the cost of additional power consumption). This reduces the noise level by a factor of 1/& and partly compensates for the increase in noise level arising from the reduction in shaping time (which scales as 1/J shaping time). Fourteen different species of preamplifiers are used in Run I1 to compensate for the effect of varying detector capacitances. Forty-eight preamplifier hybrids are socket mounted on a single preamplifier mother board that connects to the backplane of the preamplifier box. Each box houses 96 preamplifier motherboards and are powered by low noise 2kW commercial switching power supplies. Calibration pulses are injected through 0.1 % precision resistors. The preamplifier output is brought to the BLS unit through 115R twist and flat cables of length 25 m. BLS unit is located in the platform below the detector. The output pulse of the preamplifier has rise time of 430 ns and decay time of 15 ps. In order to reduce pileup effects the preamplifier output is differentiated with a time constant of 250 ns. The shaped output is sampled at its peak at about 320 ns, and the voltage signal is stored in an analog pipeline called the switched capacitor array (SCA). The SCAs are 48 channels deep, and can thus hold the signals until a L1 trigger decision can be made 4.2 ps (32 samples taken every 132 ns) later. The SCAs are not designed for simultaneous read and write operations. Two SCAs are used for the same channel so that one SCA can be readout while the other is being written. At a L1 trigger rate of lOKHz this provides for a deadtimeless operation. The SCA has a 12 bit precision.

692

In order to achieve the required 15 bit dynamic range, shaped signals are passed through two different gain paths ( x 1and ~ 8 ) Both . are simultaneously sampled and stored into two parallel SCAs. So a total of 4 L1 analog buffers are used for every channel. In order to remove slowly varying offsets, and pileup of events from neighboring bunch crossings, samples taken 396 ns earlier are subtracted from the sample taken at trigger time and the subtracted signal is stored in a fifth SCA of same depth as the L1 pipeline, where it awaits a L2 trigger decision. After the L2 trigger accept is received, the signal is readout from the appropriate depth in the L2 SCA for digitization by the ADC. The ADC is a successive approximation digitizer and has been retained from RunI. The ADC has 12 bit of dynamic range but a low and high gain path ( x l and x8) maintains the 15 bit dynamic range. The calorimeter L1 trigger uses a fast trigger pickoff signal obtained by hard differentiation of the preamplifier output by a shaper hybrid located on the BLS unit. Trigger towers are formed by summing over all channels with AQ x A$ of 0.2 x 0.2. The summation is done in two stages on the BLS unit, the second stage is also used to drive the signal to the L1 trigger system. Coordination between the trigger framework and the calorimeter readout is done by the timing and the control system. It receives trigger, accelerator and clock information through the serial command link (SCL) and oversees that the appropriate depth in the pipeline is readout at every trigger level. 4.2. Commissioning of the Calorimeter Electronics

The installation and commissioning of new electronics was nearly complete at the time of the conference only awaiting the full trigger coverage. At the time of the conference the L1 calorimeter trigger coverage was complete up to 1 ~ 51 0.8. At present it is extended up to 1771 5 2.4. The rest of the calorimeter trigger is expected to be finished shortly. We have measured the liquid argon impurity to be less than 0.15 ppm2 which is less than the 0.5 ppm specification. At the beginning of the commissioning we checked continuity of the cables inside the cryostat and from the feed-through port to the preamplifier by injecting a step pulse and measuring the reflected pulse from the calorimeter cells. The reflected pulse shape can be modeled by using appropriate detector capacitances and cable impedances. Using this method we have found less than 0.05 % of the cables have problems that can not be repaired. These bad cables are randomly distributed, and thus have minimal impact on the calorimeter data. We have also debugged the entire readout chain by reading out the means of pedestals and the noise for every channel. The response of every channel to the injected calibration pulses has been checked. After removal of bad channels we are left with 5 0.05% bad channels.

693

These bad channels are also randomly distributed over the entire detector and are being suppressed during data taking. We have measured the distribution of correlation coefficients for pairs of channels. From the deviation of this distribution from the expected Gaussian distribution we have estimated that the correlated noise is 5 0.05 ADC counts. A correlated noise study is underway to measure the effect on the calorimeter readout due to other detector systems (SMT and CFT). As mentioned earlier, the calorimeter signal is sampled in every 132 ns and stored in analog pipelines (SCAs). It is very important to study the variation of this stored signal at all depths in the pipeline. We have done detailed study to measure the variation. About 1 9% of all SCA chips showed sizable variation and those have been replaced with good chips. We have also confirmed that we are sampling the shaped calorimeter signal at its peak. This was done by taking 3 samples in an event (averaged over many events), with the nominal peak sample taken at the point as defined by the trigger, a second sample is taken at 132 ns earlier and a third one 132 ns later. The nominal sampling position with respect to the peak of the shaped signal can be accurately determined from the ratio of these samples when compared to a spice3 model fit to the signal. Using this method we have determined that the channel-to-channel variation of the nominal sampling time is less than 10 ns. This implies less than 0.1% variation in the peak sampled value due to timing differences. A preliminary calibration study shows excellent linearity for calorimeter energy deposits greater than 1 GeV (the non-linearity is found to be < 0 . 3 V ~ ) For ~ . smaller energies the nonlinearity is non-negligible, but is corrected by using a simple parametrization. 5. Physics Studies

We have started several physics analysis using calorimeter data. Preliminary results on W boson and Z boson production and their subsequent decay to electronic channels, W decay width etc. will be presented in the ICHEP conference. In Fig 4 preliminary plots of inclusive jet p~ spectrum and dijet mass spectrum is shown. 6. Outlook

We have recently finished commissioning most of the new calorimeter electronics. The remaining part of the trigger readout is expected to be done soon. The new calorimeter electronics is performing well. Benchmark physics processes using the calorimeter are being studied. Preliminary results are encouraging. In the near future we will be studying the calibration and jet energy scale in more detail.

694

i0

Figure 4.

Inclusive jet p~ spectrum (left) and Dijet mass spectrum (right)

Acknowledgments Thanks to my DO colleagues for doing the hard work that brought the detector to its present state. Also, their help in preparing the manuscript is highly appreciated. Thanks to the organizers for arranging a very enjoyable conference.

References 1. R. D. Schamberger, The DO Calorimeter Performance and Calibration, in: Proc. fifth Calorimetry in High Energy Physics (ed. H. Gordon and D. Rueger, Upton, NY, September 1994), 39-50, (World Scientific, 1995). DO Collaboration, S. Abachi et al., Nucl. Instrum. Methods Phys. Res. A 338,185 (1994). 2. A. Besson, Proceedings in this conference. 3. www.psspice.com 4. U. Bassler, Proceedings in this conference.

ATLAS LAR EM CALORIMETER: CONSTRUCTION AND UNIFORMITY OF RESPONSE

S . RODIER Universidad Autdnoma de Madrid, Cantoblanco, 28049 Madrid, Spain E-mail: [email protected] (For the ATLAS LARG EM calorimeter group) The status of the construction of the ATLAS electromagnetic calorimeter is presented. The quality controls of the module production, their results and consequences are rewiewed. Electron test beam data are used to further assess the uniformity of the modules.

1. Introduction The ATLAS LAr EM calorimeter' is first presented and its main parts are introduced. Then the production, the stacking and the diverse tests which allow to control the quality during the production are rewiewed and eventually the uniformity is treated of. 2. The calorimeter: general geometry

The electromagnetic calorimeter is based on the accordion geometry (See Figure 1). It consists of accordion shaped lead absorbers interleaved with electrodes. The honeycomb spacers separate the absorbers and the electrodes. At the end, the gap is filled with liquid argon. The calorimeter is split up into three parts, a cylinder (the barrel) and two wheels (the end-caps) and covers the pseudorapidity region up to q inferior to 3.2.

2.1. The barrel The barrel2 is subdivided into two half barrels centered over the z-axis, in their turn subdivided into 16 modules. Each half barrel covers a zone in pseudorapidity lower than 1.475 (the boundary corresponds to q = 0). Each modules is composed of 64 absorbers and 128 electrodes (subdivided into two types of electrodes: one for q < 0.8 and the other one for q > 0.8). The granularity depends on the pseudorapidity q, the azimuthal angle 4 and the

695

696

Figure 1.

Detail of the LAr EM calorimeter structure.

depth. The electrodes (see figure 2 on the left) define the granularity in depth (three compartments named front, middle and back) and in pseudorapidity 77. The granularity is thinner in the front than in the two others compartments. The line at 77 = 0.8 is the boundary between the two types of electrodes. Table 1 summarizes the barrel granularity. Table 1.

Granularity in the barrel

r ComDartment I

An

I

Ad

I

2~1256

697

Figure 2.

Left: view of a barrel electrode. Right: Sketch of an end-cap module.

2.2. The end-caps

The end-cap3 is composed of two wheels with a projective crack a t q = 2.5. It covers a pseudorapidity from 77 = 1.375 to 3.2 and it is subdivided into 8 modules (See figure 2 on the right). In one module, 96 absorbers and electrodes are stacked in the outer wheel and 32 in the inner wheel. The gap increases with the radius and therefore decreases with 77. Table 2 summarizes the granularity in the end-cap. Table 2. Wheel Outer

Inner

0 range [1.375,1.425] [1.425,1.5] [1.5,1.8] [1.8,2.0] [2.0,2.4] [2.4,2.5] [2.5,3.2]

Granularity in the end-cap Front

Middle

0.05 x 0.1 0.025 x 0.1 0.03 X 0.1 0.04 X 0.1 0.06 X 0.1 0.025 x 0.1

0.05 x 0.025 0.025 x 0.025 0.025 X 0.025 0.025 X 0.025 0.025 X 0.025 0.025 x 0.025 0.1 x 0.1

N

N

N

Back

0.050 X 0.050 X 0.050 X 0.050 x 0.1 x

0.025 0.025 0.025 0.025

0.1

The outer wheel is devoted to precision physics so that it is segmented into three longitudinal sections. The inner wheel is segmented only into two sections. The main difference between barrel and end-cap comes from the high voltage set up. It is constant everywhere in the barrel. The detector signal is proportionnal to the drift velocity and inversely proportionnal to the liquid argon gap thickness so that an 77-independent detector response could be achieved with a continuously varying HV with the pseudorapidity. Figure 3 on the left exhibits this variation of the high voltage with respect to 77 (open circle).For technical reasons, a high voltage varying by steps (full triangle) was chosen.

698

Inner Wheel 0

? l 132 ” F - T -

A * A

~

0 0

130

A - A A 0 0

126 124 00

nponse

122

--0

1.4

1.6

1.8

2

2.2

2.4

2.6

2.8

3

3.2

lzo10

15

20

Figure 3. Left: High voltage variation along residual of step effect: energy versus 11.

I

4

II 7)

for a constant calorimeter response. Right:

Figure 3 on the right shows the raw energy versus the pseudorapidity for the 2001 test beam data. The residual of step effect in each HV zones is not corrected yet. 3. The calorimeter: mechanical components 3.1. The absorbers Table 3.

Components specification for an absorber Barrel TI

< 0.8 I n > 0.8 I

End-cap Outer W. I Inner W.

0.2 mm 0.15 mm adhesive

I absorber The absorbers are made of lead plates jacketed in two layers of stainless stell to ensure the absorber rigidity. These layers are glued using a glass-fibre adhesive. Table below summarizes the thickness of each component. The lead plates thickness changes at 77 = 0.8 in the barrel. To minimize the contribution of the passive material to the constant term in the energy resolution, stringent tolerances on the lead plate thickness must be imposed so that all the plates were thickness mapped using a technique based on ultrasound. Deduced from

699

the r.m.s of the sliding mean thickness, a contribution to the constant term of 0.12% and 0.18% is expected in the barrel. In the end-cap, a raw simulation indicates that a thickness distribution with a r.m.s inferior to 17pm in the outer wheel and 22pm in the inner wheel is required to achieve a contribution of the order of 0.2% to the local constant term. For all measurements, a r.m.s of 9pm was obtained, well below tolerances. In the barrel, for each absorber a 3D measurement is carried. In the end-cap, absorbers are controlled by optical inspection, thickness and width measurements at prefixed positions. For 10% of its, a 3D measurement is carried. Requirements on the liquid argon gap uniformity is at the 2% r.m.s level. Absorbers production goes on normaly.

3.2. The electrodes An electrode consists of three layers of copper, insulated by two kapton polyimide sheets, as shown in figure 4. The outer layers provide the HV while the

i(t)

Figure 4.

Schematic view and photo of an electrode

700

inner one allows the signal collection by capacitive coupling. The granularity in Q and in depht is obtained by etched patterns on the different layers. All resistors and capacitances are measured and a high voltage test is performed. Since module0, the series production have been strongly improved. Now less than 10% of the electrodes are rejected during the fabrication cycle. At the moment more than 80% of the electrodes are delivered. 4. Stacking and Tests

In a cleaned room successively absorber, spacers, electrode, spacers and finally another absorber are stacked (this entire group is named gap). Of the order of 4 gaps a day are stacked. After stacking one gap, the thickness of the stack is measured and a capagap test is carried out. Regularly two electrical tests are ~ e r f o r m e d ~the > ~ low : frecuency test and the high voltage test. Now the nominal HV is applied during the night to bake out. 4.1. Overthickness measuremnents

To control the overthickness of every absorber, the relative heigh of each wave is measured. 4.2. Low frequency test

It allows to check the continuity of the electrical circuit and the electrode connections including the high voltage distribution. A low frequency sinusoidal signals is injected on the HV lines. A current is induced on the signal layer. The decoupling capacitance is calculated.

4.3. High voltage test With this test, any high voltage problem is detected by measuring the leakage current of each electrode side and each high voltage sectors so that the absorbers, electrodes and honeycomb spacers cleanliness are checked as well as any surface default. A tension is applied to the electrodes and the leakage currents are recorded. At the beginning, the charge currents are dominated by the decoupling capacitance charge so that they are roughly constant (they are proportionnal to the sector surface. See equation 1).

Above a given threshold, the voltage is stabilized until a significant decrease of the whole currents.

701 4.4. Gap measurements

This test is carried out to measure the distance between the electrode and the absorber. In the barrel, the gap have to be constant while in the end-cap, they have to be all the sa,me for a given 77. A sinusoidal input signal is injected on one cell. With the measured impedance, the gap capacitance value is deduced. With the production modules, the contribution to the constant term is of the order of 0.2%.

4.5. Cabling

After asembling, the module is fully cabled with summing boards, mother boards and high voltage boards. Summing boards: They group the signal in q5 to the desired granularity.

[

&ell

I

Barrel 16 electrodes

I I

End-cap 12 electrodes

I I

( End-cap OW) 3 electrodes

1 1

+

(Barrel End-cap IW) 4 electrodes

Mother boards: They are connected to the summing boards and they route the output signals to the readout cables and they include precision injection resistors for the calibration system. HV boards: They supply the high voltage to electrodes. Eventually, at room temperature, the following electrical test are performed: a low frequency test and a high voltage test. Then modules are tested at cold (at CERN for the end-cap, at Saclay, Annecy and CERN for the barrel). At the moment 11 modules are already produced in the barrel (first full-equiped half barrel before the end of this year) and 3 in the end-cap (first full end-cap at the end of this year). 5. Uniformity

In figure 5 the energy versus the pseudorapidity in the barrel as well as the energy distribution are represented. The RMS over the mean energy is of the order of 0.9%. The lead plates transition at 77 = 0.8 is not corrected yet and there is X-talk in the testbeam feedthrough. Further detailed studies of elaborating calibration are in progress in order to improve the RMS over the mean energy.

702

245

MODULE M10

240

s

235

W

Q 230

b 225 220 215

10 Figure 5.

20

30

40

0

50

q (middle)

Mean energy versus pseudorapidity.

6. Conclusion

Four series modules have been beam tested so far an three more are expected to be beam tested in 2002. At the moment eleven modules have been produced in the barrel and three in the end-cap. Up to now the production goes on normally. Improvements of analysis are in progress. References 1. ATLAS Collaboration, A T L A S Calorimeter Performance Technical Design Report CERN/LHCC/96-40 2. ATLAS Collaboration, Performance of the Barrel Module 0 of the A T L A S Electromagnetic CalorimeterNucl. Instrum. Meth. A to be published. 3. ATLAS Collaboration, Performance of the Endcap Module 0 of the A T L A S Electromagnetic Calorimeter, Nucl. Instrum. Meth. A to be published. 4. N. Massol et All Test bench of the barrel calorimeter modules ATL-LARG-2001-007. 5. P.Baxrillon et Al, Test bench and R L C measurements of the electromagnetic endcap calorimeter modules to be published.

PERFORMANCE OF ATLAS EM MODULES IN TEST BEAM

D. ZERWAS Labomtoire de I'Acce'le'mteur Line'aire, B.P. 34, 91898 Orsay Cedex, France E-mail: zerwasOlal.in2p3.f r (On behalf of the ATLAS EM Group)

Barrel and endcap modules for the electromagnetic calorimeter of ATLAS have been built and exposed to electron beams of up to 245 GeV. Their performance in terms of noise, energy and position resolution is discussed and compared with Monte Carlo simulations.

1. Introduction The pre-series (module 0) and series modules (barrel: M10, M13; endcap ECCO, ECC1) of the future electromagnetic barrel and endcap calorimeters' of ATLAS have been tested in electron beams at CERN's H6 (H8) beam line with energies from 10 GeV up to 180 GeV (245 GeV). The construction of these modules is described elsewhere2. The goal of the module 0 tests, performed from 1998 to 2000, was the validation of changes since the previous prototypes built and tested in 1997314, i.e., the use of large electrodes, use of lead sheets with a constant thickness in the endcap, cold electronics (summing and mother boards). In the course of these tests, the design of the summing and mother boards was improved and validated. In addition, the test beam also served to validate the ATLASlike electronics (Front End Board, Calibration Board) which are described elsewhere5. The beam tests of the series modules are part of the quality and acceptance control for ATLAS modules. In the following the module 0 results are final, whereas the series module results are preliminary. The paper is structured as follows: first the test beam setup will be described, followed by a discussion of the signal reconstruction. The modules' response to muons and electrons will be presented and compared to Monte Carlo simulations. The results are summarized in the last section.

703

704

2. The Test Beam Setup Each beam line is equipped with four beam chambers. The coincidence of three scintillation counters (in front of the cryostat) is used as trigger. The cryostat can be moved around a virtual ATLAS interaction point to scan the uniformity of the whole calorimeter module. Behind the cryostat muon and pion counters are installed behind about 5 A0 of iron and lead. The dead material in front of the calorimeter is 1.1 (1.4) XOfor the endcap (barrel). In 2001 in the barrel beam line additional material, present since the begin of the tests, due to a incorrectly installed beam tool, amounting to 0.3 Xo was discovered. Data are read through feedthroughs, of which only one has gold-plated pin carriers to operate the transition cold to vacuum and vacuum to warm. The signal was sampled with ATLAS-like electronics installed in the warm on the feedthrough at 40 MHz, i.e., every 25 ns. The signal maximum is reached typically after -40 ns. Seven samples were recorded. As the trigger was asynchronous, the phase of the 40 MHz clock with respect to the trigger was recorded. The variation of the energy response as function of the temperature was measured to be -2%/K. The typical variation of the temperature (89 K) of the liquid over a beam period of two weeks was (rms) 7 mK. 3. Signal Reconstruction

The physics signal is generated in the detector gap, whereas the calibration signal is injected in the mother boards. While most of the signal path is common to the calibration and physics signals, any non-uniform inductance seen in series by the physics signal and in parallel by the calibration signal can affect the uniformity of the module. There are several potential sources of inductance in the calorimeter which were studied in detail: the summing board, which was modified to have equalized inductances in 4, and the electrode. 3.1. Electrical Detector Modeling

The capacity and inductance of the electrode were calculated and compared to measurements with an RLC-bridge. While the measured capacitances agreed quite well in form as function of 17, a disagreement was observed for the inductance of the middle compartment, where the signal is brought to the connector on a thin line past the back section: every 8th cell the measured inductance was greater than the calculated one. The jump was traced to insufficient grounding of the electrode connector: every other connector has the structure (G=Ground return, M=Middle, B=Back) GMBMMBMG, while the other one

705

1.05

1.045

1 . .

-

1.04 -

.

1.ms -

1.03 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .I 24 26 28 30 32 34 50 38 40 Lrnlddb [""

Figure 1. Left: signal shape for calibration- (dashed curve) and physics-like (full curve) injection, Right: ratio of physics to calibration amplitude as function of the inductance.

is GMBMMBM. The missing ground return was added subsequently on all electrodes. The effect of the inductance on the calibration and physics signal was studied in a mockup, where holed polypropylene was used instead of Argon. The mockup permitted to inject physics-like (on the electrode) and calibration-like (at the summing board) of input signals of the same amplitude. The inductance, as shown in Figure 1, modifies the signal shape on its rising edge and reduces the amplitude of the calibration signal. Quantitatively, the inductance induces a bias of 0.2%/nH. 3.2. Optimal Filtering

The signal amplitude (Ama,)was reconstructed as a linear combination (digital aiSi). The coefficients ai were filtering6) of five samples in time (Ama, = Efz1 determined from the noise autocorrelation function, the pulse shape and its derivative. The main difficulty was the prediction of the physics pulse shape. As physics and calibration signals share most of the signal path, after a Fourier transformation the common signal path cancels in the ratio of calibration to physics signal. The physics shape then follows from calibration shape by correcting for the different input currents, triangle current for physics and exponential decay for the calibration signal, and the inductance Lo. Only three parameters were needed in this approach: the starting times of the physics and calibration signal and W D = 1 / d m . With this method the shape was fitted with residuals of k2% in the middle

706

compartment, even on the rising edge of the signal, where the effect of the inductance is strongest. The measured ratios between physics and calibration amplitudes, typically 4%, showed the characteristic jump every eighth cell (correction 6%) as expected from the insufficient grounding. For the endcap data, an approach based on sigmoid functions and neural networks was used, leading to residuals of about 3~2%. 3.3. Noise performance and Cross talk studies

The noise behavior, after application of the optimal filtering coefficients, was flat as function of 71, except for the back due to the variation of the detector capacitances. The coherent noise per FEB (128 channels) was measured to be less than 5% in high gain. For an electron cluster, e.g., at 77 = 0.26, the noise was determined to be 143 MeV (261 MeV) in high (medium) gain of which 95% (80%) was incoherent noise. Typically the application of OF coefficients reduced the noise by a factor 1.4 to 1.8, depending on the detector capacitance, with respect to a single sample measurement. The crosstalk between the different sections was studied extensively. In the barrel in particular, the crosstalk between front and middle sections (via the High-Voltage-resistors) was stable over the beam periods (cool up and down) and well correlated with the electrode resistances. The crosstalk middle to back (inductive crosstalk) was reduced by the addition of the missing ground return (before: 1%every 8th cell, after: homogeneous at 0.5%). In the endcap the front-front crosstalk was large as expected due to the fine segmentation (capacity). The middle-middle (inductive) crosstalk was measured to be 1-2% in the module 0. However for the series module, the summing and mother boards were redesigned and this type of crosstalk was reduced to typically less than 0.5% 4. Response to Muons

To reconstruct the energy deposited by muons in the calorimeter a cluster of size A77 x A# = 1 x 2 in the middle compartment was used. With the beam chambers the impact point in the calorimeter was predicted. The signal to noise ratio was measured to be 7.11 f 0.07. In Figure 2 (Left) the variation of response over one physical gap is shown. The peak to peak variation is about 6%, which is due to the variation of the electrical field in the gap (folds). Using the OF-coefficients for muon signal reconstruction (Figure 2 (Right)) improved the uniformity in 77 with respect to a single sample reconstruction. Thus the muon analysis confirmed that the determination and correction for the inductance is well under control.

707

= 1.1

:

5.075

-s! L

+pl5

a 1.025

1 0.975

0.95 0.925

0.9

Figure 2. Left: muon signal variation over a physical gap in simulation (circles) and data (triangles). Right: Uniformity of muon response with a single sample (full circles) reconstruction and OF (open squares).

5. Response to Electrons The electron energy was reconstructed in a cluster of size Aq x A 4 = 0.075 x 0.075 in the barrel and Aq x A$ = 0.125 x 0.125 in the endcap (due to the smaller geometrical cell size). The cluster was centered on the most energetic cell in the middle compartment. The weights applied to the presampler and back compartments’ response to compensate for energy losses in front of the calorimeter and leakage were determined by minimizing (TIE. The weights vary as function of q between 6 and 8 for the presampler and for the back from 1.5 to 2.5 for a barrel module. In the endcap, the setting of the high voltage by finite sectors induces a slope of the energy response in a sector. The slope was corrected for with a linear correction as function of q. The correction coefficient was obtained by the minimization of a / E in a HV-sector. The energy was then corrected for lateral leakage in q and 4. The 4 accordion shape induces a modulation of the response. Due to the particular geometry of the endcap, this modulation changes as function of q as shown in Figure 3. The effect is well reproduced in simulation. A residual dependence of the response as function of the phase with respect to the 40 MHz clock was also corrected for.

5.1. Energy Resolution The energy resolution as function of the energy is shown in Figure 4. The noise was subtracted at each energy point. The back section was not used for beam

708

0 Simulation

0.98'

"

4A

"

"

0.2

0 Simulation

"

'

I

'

'

I

0.2

"

" '

0.4

I

4.4

4.2

0

0.2

@

0.4

@

Figure 3. Modulation of the energy response within a detector cell as function of q5 for (Left) 9 = 1.6 and (Right) 9 = 2.9. Full circles: data, open circles: simulation, full curve: fit result.

energies less than 40 GeV. In the barrel the sampling term was less than 10% in three points ( q < 0.8) and the constant term was less than 0.3% in two of the three points. The energy leakage at q = 0.0875 degraded the constant term to 0.5%. Similar results were obtained with the series modules. For the endcap data, (Figure 4 (bottom)), a sampling term of 10.4% in data, compared with 10.1% in simulation was obtained at q = 1.9. The measured local constant term of 0.3% was in good agreement with the 0.4% expected from simulation. In all other positions, where energy scans were recorded, the local constant term was less than 0.6% and the sampling term better than 13%. The prediction of the absolute energy scale from simulation was correct at the level of 5%. The linearity is better than hl%,for the barrel partially affected by the additional material upstream of the calorimeter. 5.2. Position and Angular Resolution of the Endcap module

To reconstruct the position resolution in the endcap module 0 the energy weighted barycenter was calculated. In the front section in q only three cells were used. The S-shape was corrected for and the beam chamber resolution (about 0.35 mm) was subtracted. The position resolution of the front section was stable, essentially independent of q at 2.7 m m / a as the granularity varies. The position resolution of

709

n

iE

4.5 L 4 '

q=0.3625 0

"0

-cr

noise contribution subtracted resolution

50

100

150

o(E)/E = a/dE 0 b a = 9.24 omlo% b = 0.23 k 0.04%

200

250

300

3.5

odE=aJdE 0 b,

oEIE=a&E @ b,

aE=10.35f 0.05%

a,=10.14+ 0.03% GeVIR

$ 3

2.5

b,=

027f 0.02%

b,= 0 . 4 2 0.02%

2

1.5

1

0.5

0

D

Figure 4. Top: energy resolution at 7 = 0.3625 barrel data. Bottom Left: Energy resolution endcap data at 17 = 1.9. Bottom Right: Energy resolution endcap 7 = 1.9 simulation.

the middle section improved as function of q (Figure 5) from 8.3 mm/& to 3.3 m m / a due to the decrease of the physical cell size. The two measurements of q were combined with the longitudinal shower barycenters to determine the resolution of the polar angle. The beam dispersion of 0.1 mrad was negligible. At q =1.9, 50 m r a d / f i were expected from simulation and achieved. In R4 direction, only the middle was used, as the granularity of front section is too coarse. The position resolution showed the same qualitative behavior as the q resolution. At q = 1.6, 6.5 mm/& were measured and at q = 2.4, 3.2 m m l a were obtained.

710

%= 5.m020 mn esv'" b,,= 0.1w 0.W mn

c"= 9.47* 1.47 mn esv 0.5

5

Figure 5. Left: 7 = 1.9 position resolution middle section. Right: position resolution middle section as function of 7.

5.3. Uniformity

The uniformity of the energy response of the barrel series module M10 is shown in Figure 6 . The response is flat as function of q even in the transition region (7 = 0.8) of the two lead thicknesses. Deduced from one measurement per middle section cell, the global constant term was determined to be 0.93%, in agreement with 0.94% deduced from the fit of the energy distribution of the whole module. For ATLAS a global constant term of less than 0.7% is required. This is to be achieved by obtaining regional constant terms (AT x Ad = 0.2 x 0.4) of less than 0.5% and inter-calibrating these regions with Z boson decays. This has been achieved in module M10 for several regions of Aq x Ad = 0.2 x 0.15 in the gold plated feedthrough. The module 0 barrel uniformity in Aq x A$ = 1.2 x 0.075 was measured to be 0.8%. For the endcap module 0, the uniformity of the inner wheel was determined to be 0.6%.

Enlnes 7ieor-233 10 GeV

Energy (Gel

Figure 6.

Left: Energy response as function of 7. Right: Energy spectrum module M10.

71 1

6. Conclusions and Perspectives

Since 1998 extensive beam tests of module 0s and series modules have been performed. The hardware (summing boards, mother boards, ground return) has been improved and the ATLAS-like electronics have been validated. In the barrel modules the signal to noise ratio for muons was measured to be 7.11f0.07. A sampling term better than 10% was obtained. The local constant term is less than 0.3%. Preliminary results on the uniformity of the series module are encouraging. The sampling term for the endcap module 0 is less than 13%. The angular resolution is in agreement with ATLAS specifications. A global constant term of 0.6% was obtained in the inner wheel. In 2002 two modules of second half-barrel and one module of the endcap will be tested in beam tests. A combined run of an endcap module and a hadronic endcap module will take place at the end of the year. Combined runs of endcap and forward calorimeter modules and of a barrel module with the hadronic barrel calorimeter (tiles) will be done.

Acknowledgments The author would like to thank the organizers for the well organized conference. It is a pleasure to thank my ATLAS-Larg-EM colleagues for providing me with figures and results. I am indebted to Laurent Serin for discussions in the preparation of the talk and the manuscript.

References 1. ATLAS Collaboration, Liquid Argon Calorimeter Technical Design Report, CERN LHCC/96-41. 2. S. Rodier, these proceedings. 3. RD3 Collaboration, Nucl. Instr. and Methods A411 313 (1998). 4. RD3 Collaboration, Nucl. Instr. and Methods A389 398 (1997). 5. E. Ferrer-Ribas, these proceedings. 6. W.E. Cleland, E.G. Stern, Nucl. Instr. and Methods A338,467 (1994).

THE ATLAS HADRONIC ENDCAP CALORIMETER

M. FINCKE-KEELER University of Victoria, Dept. of Physics and Astronomy Vzctoria B . C., V8 W-3P6, Canada E-mail: mgfOuvic.ca

(For the ATLAS HEC Collaboration)

The construction of the ATLAS Hadronic Endcap Calorimeter is nearing completion. Beam tests of the series modules have been completed. The performance of the modules yields a resolution for electrons of o ( E ) / E = (21.4 f 0.2)%(GeV)1/2/v% @ (0.3 f 0.2)% and for pions of a ( E ) / E = (70.6 f l.5)%(GeV)1/2/v%@( 5 . 8 f 0 . 2 ) % ,where @ denotes a quadratic sum. The uniformity and linearity have been verified. The details of the electronic readout chain are well understood and allow precision predictions for the performance of the calorimeter under the final ATLAS operating conditions.

1. Design of the hadronic Endcap Calorimeter

The ATLAS experiment for the LHC has chosen liquid argon (LAr) as the active material for its endcap calorimetry. The Hadronic Endcap Calorimeter (HEC) will be sharing the endcap cryostat with the Electromagnetic Endcap Calorimeter (EMEC) and the Forward Calorimeter (FCAL). The HEC is a Cu/LAr sampling calorimeter, constructed of 2 wheels at each end with 32 wedge-shaped modules in each wheel. It will cover a range in pseudorapidity of approximately 1.5 < 1771 < 3.2, and both wheels together are designed to provide a nuclear absorption length of 1OX in depth. The modules for the front (rear) wheel consist of layers of 25 mm (50 mm) thick copper separated by liquid argon gaps of 8.5 mm thickness. The gap size is maintained by stainless steel spacers. Each gap is instrumented and further divided with 3 polyimide boards, resulting in 4 subgaps that form an EST structure' (see Fig. 1). The size of the subgaps is maintained by honeycomb mats of a nominal thickness of 1.85 mm. All polyimide boards have an outer carbon-loaded layer that hold the operating HV of 1800 V for the 4 subgaps. The central board is a sandwich that contains a copper-pad readout structure in between the outer polyimide layers. This readout structure defines the granularity of the calorimeter to be Aq x A$ = 0.1 x 0.1 for a pseudorapidity less than 2.5 and

712

713

operating H V : 1800V

cold prsnrnplifierr

Figure 1. Artist’s view of a HEC module and a close-up of the substructure of a liquid argon gap.

Aq x Ad = 0.2 x 0.2 for 1771 more than 2.5. The pads are arranged to provide a semi-pointing geometry with respect to the ATLAS interaction point. Further details of the design of the HEC have been published elsewhere2. 2. Status of Module Production and Assembly Tooling The construction of the Hadronic Endcap Calorimeter is well under way. At the time of the conference (by end of March 2002) 77% of all modules had been built world-wide and 63% of the final number of modules had been cold tested. Beam tests have been completed, with a total of 24 production modules having been exposed to beams of electrons, pions and muons. At the same time the wheel assembly table and rotator have been completed and put into operation. Assembly of the first HEC wheel will commence during the summer of 2002.

3. Beam Tests 3.1. Setup

The beam test cryostat provides enough space to hold a set of 3 front modules and 3 rear modules. Production modules were tested during 6 beam periods of 2 to 3 weeks each. For the tests, the modules are arranged such that they

714

are “standing” on the outer circumference of the wheel (see Fig. 2). Beam tests have be performed with electrons in the energy range of 6 GeV to 193 GeV, pions in the range of 10 GeV to 200 GeV and muons of 120 GeV, 150 GeV and 180 GeV. The horizontal movemeilt of the cryostat and vertical steering of the beam allow the testing of a number of different readout channels and detailed position scans of the modules. Fifteen well defined “impact points” (see Fig. 2) are used to compare the response of different modules and different beam periods.

Figure 2. Three modules in a typical beam test setup. The diagram illustrates the accessible impact points for the beam (A to 0 ) .

Longitudinally, the LAr gaps are grouped into 4 readout-channel segments. The first longitudinal segment (LSEG) comprises 8 LAr gaps of the front wheel, while the second LSEG comprises 16 gaps. The rear wheel is divided into two longitudinal segments of 8 LAr gaps each.

3.2. Module Performance Signals are recorded typically in 16 or 32 time samples, separated by 25 ns each. The amplitude is determined with optimal filtering3 over the 5 time samples in the region of the signal peak. Pedestals are obtained from the 5 signal free time samples preceeding the signal. After subtraction of the electronic noise in quadrature, the energy resolution can be parametrized as

where @ denotes a quadratic sum.

715

e0 :, Least square fits t o the data recorded with electron beams and averaged over several impact points yield the following results:

0.05

-

0.04

-

0.03

-

JuneOO

-

Au@

:! ,

* July01

1

AugOI

a = (21.4 f0.2)%-

b = (0.3 f 0.2)% E (GeV) Figure 3. Resolution for electrons as a function of energy for impact point G for four different beam periods. Each beam period is represented by a different symbol.

% 0.25 W

!

50.225

1

3.2

-

0 a(%)

70.7f1.5

b(%)

5.9+0.1

a(%)

70.5+1.5

a(%)

71.2f1.4

0.2 ; 0.175

Least square fits to data recorded with pion beams and averaged over several impact points yield

:

0.15 : 0.125

:

A

0.1 : 0.075 :

u = (70.6 f 1.5)%b = (5.8 f 0.2)%

0.025 0.05

t 0

25 50 75 100 125 150 175 203

Energy (GeV)

Figure 4. Pion resolution for 3 different impact points. Shown are the measurement points as well as the fits to equation 1 for each of the 3 impact points.

The uniformity and linearity of the calorimeter can be demonstrated by the response to electrons measured at various impact points and the average

716

t

; ; . f -~ l ~~ ; . + l l

a'

1

-03 E .02

'.04 1.02

Q)

....................

~

~

~...................... ? .

$ 0.99

0.98

0.98 0.96

0.971, G

H

I

J

K

L

M

N

0.96

O

Impact point

Figure 5. Ratio of the electromagnetic scale factor a,, to the average electromagnetic scale factor (a,,) for several impact points.

I

50

,

,

,

I

100 150

Energy (GeV) Figure 6. A typical distribution of (aem)/aem as a function of energy for one impact point.

response to electrons as a function of energy (see Figures 5 and 6). Muon 150 GeV X scan Y=.O August MOO

............

i......................

IM

[,,,, i + ..................

i

;........... .j............

............

.f

..............

i

t.1

............. ..................

W

0 X ( 4

Figure 7. Response to muons as a function of horizontal position in the four longitudinal segments (see text).

717

Figure 7 shows the amplitude of the signal in the 4 longitudinal segments as a function of horizontal position for muons of 150 GeV energy. The second longitudinal segment has twice the number of LAr gaps compared to the other 3 segments and therefore contributes a signal that is twice as large as the signal of the other segments. The boundaries between the different modules at x M f 1 7 cm can be detected with the narrow shower of the muon beam.

3.3. Signal Readout

A detailed study of the signal response in the electronic readout chain has been performed for individual ADC channels. The behaviour of distinct parts of the readout chain (preamplifiers, shaper, cables) has been determined separately and was then combined to form an analytical model describing the entire chain5. A comparison between the measured signal shape and the analytical model is shown in Fig. 8. The residual between the two distributions is typically within f l % .

Figure 8. A typical shaped pulse from one readout channel and the residual between the pulse and the analytical model prediction.

718

Based on this detailed understanding of the signal pulse, a refined analysis can be performed to determine the dependence of the drift time and the initial ionization current as a function of the electric field applied in the gap. 45

.._ 0

0 1

2

4

3

5

7

8

8

9

2

4

6

8

1

0

electric field, kV/cm

10

Electric field, kVlcrn

Figure 10. Initial ionization curFigure 9. Drift velocity versus electric field. The rent as a function of electric field. continuous line indicates a model prediction by W. The continuous line represents a fit Walkowiak 2 . to the Thomas-Imel Model.

Figure 9 shows dependence of the drift velocity on the electric field in the LAr gap. Figure 10 shows the initial ionization current as a function of the electric field. Both show good agreement with theoretical expectations.

m

87

88

80

8(1

81

sz tamperalure, K

93

m

BI

88

8s

w

ni

92

m

temperature, K

Figure 11. Ionization current and drift time as a function of LAr temperature. The dots represent the measurement, while T B indicates the nominal operating point of the beam test cryostat .

719

The LAr operating temperature of the ATLAS endcap cryostat (87K to 89K) cannot be reached in the beam test cryostat. However, a limited temperature interval can be mapped out, that allows an extrapolation from the testbeam conditions to the operating conditions of ATLAS. Figure 11 shows the dependence of the initial ionization current and the drift time as a function of liquid argon temperature. A linear extrapolation toward ATLAS temperatures indicates that in ATLAS the initial ionization current will be slightly larger while the drift time will be slightly shorter compared to the beam test measurements. 4. Conclusion

The production of modules for the ATLAS Hadronic Endcap Calorimeter is nearly complete. Assembly tooling for the calorimeter wheels is in place and construction 0s scheduled to start in the summer of 2002. Beam tests confirmed that the calorimeter performs as expected. A detailed model of the electronic properties reproduces the data and has proved very useful for the understanding of the detector. References 1. 2. 3. 4. 5.

J. Colas, M.Pripstein, W.A. Wenzel, NIM A294,583 (1990). The ATLAS HEC collaboration, NIM A482,94 (2002). W.E. Cleland and E.G. Stern, NIM A338,467 (1994). W. Walkowiak, NIM A449,288 (2000). Leonid Kurchaninov, private communication.

PERFORMANCE OF THE ATLAS HADRONIC END-CAP CALORIMETER IN BEAM TESTS

A.E. KIRYUNIN* Max-Planck-Institut fur Physik, Werner-Heisenberg-Institut Fohringer Ring 6, 80805 Munchen, Germany (On behalf of the ATLAS Liquid Argon HEC Collaboration)

Serial modules of the ATLAS hadronic end-cap calorimeter have been successfully tested in particle beams at CERN in 2000 and 2001. Main performance parameters of calorimeter modules, obtained by analyses of electron, muon and charged pion data, are presented. Detailed comparisons of experimental results with predictions, based on Geant3 simulations, are done.

1. Introduction The hadronic end-cap (HEC), a copper liquid argon (LAr) detector with parallel plate geometry, is being built now for the future experiment ATLAS at the Large Hadron Collider at CERN. It has four longitudinal layers and covers the pseudorapidity region 1.5 < 1771 < 3.2. In 1998-2001 extensive beam tests of pre-production and serial modules of the HEC were carried out at CERN. Various analyses of large amounts of experimental data and comparisons with predictions, based on Monte Carlo (MC) simulations, allowed to evaluate the performance of the ATLAS hadronic end-cap calorimeter3. This talk summarises the main performance parameters of HEC serial calorimeter modules, obtained in beam tests in 2000 and 2001. A detailed description of the test-beam setup and discussion of the physical program of module tests are presented in Reference2. The talk is organised as follows. In Section 2 the description of the simulation package and simulation procedures, used for data analyses, are presented. In Sections 3 and 4 the main results of electron and muon studies are given. Section 5 is devoted to analyses of pion data. *On leave of absence from Institute for High Energy Physics, Protvino, Russia

720

721

2. Monte Carlo Simulations for HEC B e a m Tests

The evaluation of the HEC performance requires the comparison of experimental data, obtained at beam tests, with detailed Monte Carlo simulations. To fulfill this task a special software package has been prepared4. It is based on Geant35 (version 3.21), a toolkit for the simulation of particle passage through matter. The package allows to simulate the response of the HEC modules to various particle beams of different energies. The geometry of the test-beam setup is described there in full detail. For example, the HEC module description includes not only copper plates and gaps of LAr, but also polyimide electrodes, copper pads and tie-rods. All elements of the beam-line, such as the cryostat, multi-wire proportional chambers and scintillating counters, extended over a distance of 30 meters, are included as well. The GCALOR code for the hadronic shower development is used for main part of simulations. To estimate the influence of the energy leakage (due to the limited acceptance of HEC modules) on the performance of the calorimeter, “virtual” leakage detectors are implemented in the simulation framework. In this leakage detectors the kinetic energy is summed up for all particles leaving the HEC modules through lateral and back sides. 3. Analysis of Electron Data

Energy scans in the range 6 GeV < EBEAM< 193.1 GeV have been performed for electrons at 15 different impact points covering the area accessible through the beam window of the cryostat. To reconstruct the energy, clusters of cells have been defined. The cluster size has been kept fixed for each impact point so that the noise contribution can be better studied and controlled. Read-out channels near each impact point were included if they had an average signal of more than 200 nA at 175 GeV. This selection yields a typical cluster size of 6 read-out channels. Simulation studies give an estimated energy leakage of only 0.5 % outside selected clusters. To estimate the electronic noise contribution, the noise from the cluster channels has been summed. Thus any correlation between different channels or any coherent part in the noise is automatically taken into account. The equivalent noise amounts to 0.6 GeV for electron clusters. N

-

3.1. Energy Response and Resolution

For the energy reconstruction a single parameter C Y E M ,connecting the measured current and the nominal beam energy, has been used. In Figure l the relative response is shown as a function of the beam energy for four different

722

sp"

8 ---. 0.rh..

21

0.99

$ 1s m IS 17

0

50

100

150

[ ~ : c ~

E m

g I6

0.96 0

50

100

150

8

15

; ; ; ;I

Exp.

C

D

E

F

G

H

J

K

imDact ooint

0.99 0.98 0.96

0

50

100

150

0.99 0.98 0.97 0.96

Point K 0

50

100

0 4.2

150

44

C

D

E

F

G

H

J

K

Impact point

Figure 1. Linearity of the signal response of electrons as a function of the beam energy at four different impact points.

Figure 2. The sampling term (top plot) and the constant term (bottom plot) of the energy resolution for electrons at different impact points (results of the September 2001 run period).

impact points. The data demonstrate good linearity (mostly 310.5 %) and are in agreement with Monte Carlo predictions. Variation of the scale factor Q E M from point t o point is also within 1 % (which shows the good homogeneity of HEC modules). Another important calorimeter characteristic is the energy resolution. For the direct comparison of the resolution at different impact points and with Geant3 predictions, the electronic noise contributions were subtracted in quadrature from the experimental resolution. The energy dependence of the resolution can be parameterized by the following two term formula:

with a sampling term A and a constant term B . In Figure 2 experimental and predicted values of the sampling and constant terms at different impact points are presented. It can be seen, that corresponding values are very close. In average, the energy resolution for electrons is characterized by the sampling term A = 21.5 f 0.2 %GeV1/2 and the constant term B = 0.1 f 0.2 %.

723

3.2. Ionization Current and Visible Energy

One of the important parameters, obtained during analyses of test-beam data, is the conversion factor between the visible energy deposited in LAr and the measured ionization current. This factor can be obtained by the direct comparison of the reconstructed signal in one channel and the amount of visible energy in the same channel, as predicted by simulations. Electron energy scans at eight different impact points were used for this study. Signals in the most loaded channels were compared. In total 149 experimental runs, taken in 2000 and 2001, were used for this analysis. In Figure 3 the distribution of the ratio between the measured current in one channel and the visible energy in this channel, obtained from simulations, is shown. The mean value of the distribution is 7.188 nA/MeV, its r.m.s. is 0.074 nA/MeV. -1%

7.135

+1%

v)

C

3 25

L rc

0

20

ij

0

E

15

3

z

10 5

0 7

7.1

7.2

7.3

7.4

7.5

C [nNMeV] Figure 3. Ratio between the measured current and the expected visible energy in one channel. Lines mark the theoretical value with uncertainties.

The theoretical value of such a conversion factor can be calculated according to the following formula, based on the Thomas-Imel model6:

where q is the electron charge, ~,=23.2 eV/pair the ionization energy per electron-ion pair, ~d,.=450ns the drift time, E0=0.84 kV/cm a model param-

724

eter, E the electric field in LAr and FH = 0.973 the correction for the volume of honeycomb spacing mats (which were not included in simulations). The theoretical value has some uncertainties (the drift time is not exactly known for each impact point, for example). Decreasing cuts in Geant3 can lead to a certain increase of the simulated visible energy. The overall estimate of uncertainties due to known factors is at the level of f l %. As it can be seen from Figure 3 the values of conversion factor, obtained by analysis of HEC test-beam data and by theoretical calculations, are in good agreement. 4. Analysis of Muon Data The muon data are very useful because of the following reasons. At first, they allow to study peculiarities of the reconstruction of low energy signals in HEC modules. At second, horizontal and vertical scans test the homogeneity of the calorimeter at full depth. Muon beams of 120, 150 and 180 GeV have been used in HEC beam tests. The main results of muon analyses are the following: - Signal-to-Noise ratio (obtained for a four channel cluster) is 5.4 (5.9) for 120 GeV (180 GeV) muons. So, muon signals are clearly seen in the HEC. - Electron-to-Muon ratio (when most probable values of muon signals are used) equals to 0.93f0.02 (0.92f0.02) for muons at 150 GeV (180 GeV). - Electron-to-MIP (minimum ionizing particle) ratio equals to 0.99f0.02. Errors presented here are pure statistical. The systematic error has been estimated using the maximum and minimum values obtained using different calibration runs or data at different impact points and varying the analysis cuts used. From these studies an estimate of the systematic error of the Electronto-Muon and Electron-to-MIP ratios is f0.03. The agreement between the data and the simulation is within errors. 5. Analysis of Charged Pion Data

Energy scans in the range 10 GeV < EBEAM< 200 GeV have been performed for charged pions at the standard impact points. Energy reconstruction was done within clusters of fixed size at each point. Channels with an average signal of more than 15 nA at 180 GeV beam have been used. The typical cluster size for the pion reconstruction is 50-60 read-out channels. The electronic noise contribution in such clusters is at the level of -6 GeV. Simulation studies give an estimated energy leakage of 4.5-5.5 %. This value consists of 1-2 % leakage outside selected clusters within calorimeter modules and 3.5 % leakage through their lateral and back sides (as “measured” in the leakage detectors).

725

5.1. Energy Resolution In Figure 4 the sampling and constant terms, obtained by fits with Equation 1 at different impact points, are presented. Results are in agreement. The average values are A = 70.6f 1.5 %GeV1/' and B = 5.83~0.2%. It can be seen that Geant3 gives a somewhat too optimistic energy resolution in comparison to the experiment. The obtained values of the resolution are affected by the energy leakage. Special studies7, based on the analysis of simulated samples, allowed to derive the HEC intrinsic energy resolution. For the sampling term A = 62.2 f 1.8 %GeV1/2 and for the constant term B = 5.2 & 0.2 % have been found typically.

5 . 2 . Ratio e / h An important intrinsic characteristic of a calorimeter is the ratio of the electron to hadron response elh. The deviation of this ratio from 1, typical for a noncompensating calorimeter, leads to a worsening of the energy resolution and

........... ...........

.......... ........................... ...........

1

C

E

F

G

H

J

.. .. .. .. ... ... .... .... ............................................... ............. .............i.......................... i .. . . . . . . ..

0.9

K

25

50

75 100 125 150 175 200

Energy (GeV)

Impact point ~

8

l

.

~

~

'

5 6.5 =

g $

6 5.5

5 4.5 4

C

E

F

G

H

J

K

25

50

75 100 125 150 175 200

Impact point

Energy (GeV)

Figure 4. T h e sampling term (top plot) and the constant term (bottom plot) of the energy resolution for pions at different impact points.

Figure 5. T h e energy dependence of the e/n-ratio for uncorrected data (top plot) and after correcting for energy leakage (bottom plot). T h e lines give fits with Equations 2.

gives rise to a constant term of the energy resolution. The elh-ratio and the

726

measured e/n-ratio are related by the following equations8:

here fir. is the mean fraction of the initial pion energy deposited via electromagnetic cascades. For pions typical values are E(, M 1 GeV and m M 0.85. The measured eln-ratio for one of the impact points is shown in Figure 5 (top plot) as well as the result of the fit using Equations 2. These results are biased due to the energy leakage outside the chosen cluster and the given calorimeter acceptance. Correction factors were extracted from simulated samples and applied to experimental data. In the bottom plot of Figure 5 the corrected dependence of the elr-ratio on the beam energy is presented. The applied correction considerably improves the quality of the fit. Estimates of the systematic errors were done taking into account the precision of the leakage correction and the variation of the e/h-ratio from point to point. The final result for the HEC is e l h = 1.4950.01f O . l O , where systematic errors dominate. The prediction of the Geant3 package (with GCALOR) is e l h = 1.32

5.3. Longitudinal and Transversal Shower Profiles

The HEC has four longitudinal layers. Studying the energy depositions in those allows to evaluate the longitudinal development of hadronic showers. Figure 6 shows the energy dependence of the mean energy fraction (normalized to the total energy in a cluster) deposited in the individual longitudinal layers of the HEC module. The energy fraction in the last part is increasing noticeably with the beam energy. Comparison of experimental results with predictions of the Geant3 package shows reasonably good agreement. The HEC granularity itself is not fine enough to see details of the transversal shower shape. But they can be studied somehow using the horizontal scan data obtained with pions at 200 GeV. The signals in the vertical (77) channels related to a given horizontal (4) read-out channel have been summed to towers. In Figure 7 the signal in one of the &towers as a function of the x-coordinate of the impact point is shown for the four longitudinal layers. The increase of the hadronic shower size width with the longitudinal propagation of the hadronic shower is clearly visible. The Geant3-based simulation (with GCALOR as a hadronic shower code) gives a good description of the data, except for the extreme tails where Geant3 predicts a slightly larger shower size than observed in the data.

727

0.6

10

exp. 0.55 0.525 0.5 0.475

10 1

C

0.45

10

Layer 2 0.1

0.25 0.225 0.2 0.175 0.15 0.125 0.1 0.075 0.05 0.025 0

0

50 100 150 200

0.4

0

50 100 150 200

lo

-20

0

20

10

1

2

-20

0

20

0.07 0.06 0.05 0.04 0.03

0.02

Layer 3 0

50 100 150 200

0.01

0

Layer 4 0

50 100 150 200

E,,

lGeVl

Figure 6. The mean energy fraction deposited in the four longitudinal layers as a function of the beam energy.

X b m

(cm)

X h m

(cm)

Figure 7. Transversal shower shape: the mean energy fraction in a 4 tower (ztower in scan, centre at -25.5 cm) in dependence on the 2-coordinate of the beam impact position in the the four longitudinal layers.

6. Acknowledgements

I would like to thank all my colleagues from the ATLAS Liquid Argon HEC collaboration for their work on building and testing HEC calorimeter modules and analysing test-beam data. References 1. ATLAS Collaboration, “Liquid Argon Calorimeter Technical Design Report”, CERN/LHCC/96-41 (1996). 2. Talk by M. Fincke-Keeler at the “Ionization calorimetry” session of this conference. 3. ATLAS Liquid Argon HEC Collaboration, Nucl. Instr. and Meth. A482, 94-124 (2002). 4. A.E. Kiryunin and D. SalihagiC, “Monte Car10 for the HEC Prototype: Software and Examples of Analysis”, Internal ATLAS HEC Note-063 (1998). 5. R. Brun et al., “GEANTS”, CERN DD/EE/84-1 (1986). 6. J. Thomas and D.A. Imel, Phys. Rev. A36 N.2, 614 (1987). 7. A.A. Minaenko, “Analysis of Test-Beam Data, Obtained with Module Zero of Hadron End-cap Calorimeter”, Internal ATLAS Note LARG-99-011 (1999). 8. R. Wigmans, Nucl. Instr. and Meth. A265, 273 (1988).

A HIGH RESOLUTION LUMINOSITY MONITOR FOR SLAC EXPERIMENT El58

G . M. JONES Stanford Linear Accelerator Center, 2575 Sand Hill Road, Menlo Park C A 94025 E-mail: markjOslac.stanford.edu (For the E l 5 8 Collaboration)

A calorimetric detector based on ionization has been employed as a low-angle luminosity monitor for the parity violation experiment E15S1 at SLAC. The experiment utilizes a 50 GeV polarized electron beam on a liquid hydrogen target. The detector looks at high energy Mott and Moller scattered electrons, with a per pulse flux of 4 x los particles. This large signal allows the device to serve the dual role of monitoring target density fluctuations, as well as detecting false asymmetries. In the first physics run of the experiment, the detector has achieved a per-pulse intensity asymmetry resolution of 170 parts per million. The linearity of the device also has been verified to 51%.

1. El58 Overview 1. l . Physics Goals The electroweak theory has been probed to an astounding level at the Zo resonance through the joint contributions of LEP and SLC. The purpose of El58 is to complement these results by measuring the weak mixing angle at a much lower Q 2 . This will allow new insight into the running of sin20w and form a more complete test of the Standard Model. Figure 1is an adaptation of a plot by Czarnecki and Marciano2 displaying the Standard Model prediction for the running of sin20w. Here, uN refers to the NuTeV experiment at Fermilab3, and APV denotes the result of the atomic parity violation experiment on cesium conducted at the University of Colorado at Boulder4. The NuTeV result is actually 3a above the Standard Model. Since the publication of the Boulder experiment, the interpretation of their result has evolved as new theoretical treatments have been applied576t7i8. In the plot, the APV point is now roughly 2a below the theoretical valueg.

728

729 eff sin2ew

I

0.01

0.1

1 10 Q IGeV/c]

100

1000

Figure 1. Standard Model running of sin'tlw

1.2. E l 5 8 Layout The experiment utilizes a longitudinally polarized electron beam provided by the 50 GeV linac at SLAC on an unpolarized target of liquid hydrogen (LHZ) to measure the parity-violating asymmetry in Moller scattering. Following the target, a dipole chicane and a quadrupole package are used to focus the Moller electron signal onto the main detector. This setup is depicted in Figure 2. From Standard Model calculations', the expected asymmetry for El58 is -150 parts per billion (ppb). Due to the small size of this number, it is extremely important that all sources of false asymmetries be understood and quantified.

+ Beam 48 GeV Lmac

LUMINOSrrY MOmOR

LH2

Quadtupoles

Target

Dipole Chcane

J ,

Moller

Detector

Figure 2.

1.3. Purpose of Luminosity Monitor

The luminosity monitor (lumi) is located 70 meters downstream of the hydrogen target at an angle of approximately 1 milliradian. From GEANT simulations of a nominal beam rate of 5 x 10l1 e- per pulse, the lumi is expected to see a flux of 4 x lo8 e-. Based purely on counting statistics, this extremely high counting rate sets the scale for the maximum resolution of the detector at the 50 ppm level. Simulations done at SLAC" show that the signal of the lumi is comprised of 70% Mott scattered electrons, with the remaining 30% made up of high

730

energy Moller electrons. The average energy of these electrons is -40 GeV. In this kinematics region, the theoretical physics asymmetry is only 5 ppb. This is the defining feature of the lumi: it should measure no asymmetry within the proposed statistics of the experiment. This would be a powerful test against the presence of false asymmetries. 2. Detector Design

2.1. Concerns and Constmints

The ultimate design of the lumi was dictated by the strict requirements of E158. The main concern was in keeping the intensity asymmetry resolution at the scale of 100 ppm. For E158, custom 16-bit adcs were constructed on site with noise properties sufficient for the experiment. Minimizing systematics due to non-linearities was also a concern. Over the full course of the experiment, the charge asymmetry is expected t o be at the level of 200 ppb or below. Because of the small size of the Moller asymmetry, it is desirable to keep any individual systematic effect at the 1 ppb level. This translates to a demand on linearity at the 0.5% level. An interesting requirement on lumi design occurs because of asymmetries found in synchrotron radiation, due to a possible presence of transverse polarization of the electron beam. From simulations, it is known that about 17W of synchrotron radiation hits the lumi, originating mainly from the final dipole magnet of the spectrometer. This is a non-negligible portion of the 150W total signal. Inserting El58 spectrometer parameters into the theoretical formulas", and assuming a transverse beam polarization of 1%nets the conclusion that the contribution to the total detector signal must be less that 1%in order to keep the observed asymmetry on the few ppb level. The final requirement on the detector design was that it had to be robust against radiation damage. Over the course of the experiment, the detector will get a dose on the order of a gigarad. 2.2. Physical Construction The lumi is comprised of 16 individual chambers, arranged into two separate, complementary detectors. These are simply called the "front" and "back" lumi. The front lumi serves as the primary device, while the back lumi is used mainly as a cross-check. Upstream of the front lumi is 7 radiation lengths of aluminum, which performs the dual role of showering the signal and attenuating the synchrotron radiation background. Between the front and back lumi detectors is an additional 4 radiation lengths of aluminum. Figure 3 depicts the full detector layout.

731 Side View Chamber5

Aluninurn

5

Figure 3.

Inside of each chamber, there are 11 parallel plates, positioned transverse to the beamline. These are alternately kept at lOOV and ground, so that when a charged particle traverses the chambers, the electrons it liberates through ionization are collected. A signal is actually read out from each plate in the device, since they are all capacitively coupled. By reading in the signals differentially, noise pickup from the cables is reduced. The separation between the plates is approximately one millimeter. This small distance was chosen to ameliorate possible non-linearities induced by unwanted recombination of liberated electrons and ions in transit to the plates. The buffer gas was chosen to be nitrogen, since it provided adequate signal sizes and did not require the,special attention required by a flammable gas. Simulations in EGS show that the front degrader produces a x40 amplification of the incident signal. The lcm of nitrogen gas in each chamber also produces a similar amplification factor. The signal sizes seen from a single front chamber for typical running conditions are between 4 to 8 volts. This allows the detector to be run entirely without amplifiers. The ratio of the observed signals in the front to the back lumi is about 3:1, which agrees with the simulations. Since the back chamber has a much reduced signal, it can serve as a linearity check on the front ring of chambers. The fact that the front and back rings of the detector measure the same signal flux allows the back chambers to be a crosscheck for the front chambers for the measured asymmetries as well. 3. Detector Performance 3.1. Resolution

The ultimate per-pulse intensity asymmetry resolution achieved during the first run of El58 was at the level of 170 ppm for the front ring of the detector, and 250 ppm for the back ring. To achieve this level of resolution, it was necessary to remove beam related

732

effects from the detector signal. This is accomplished by finding the slope of the correlation between asymmetries in the detector and a beam position monitor (bpm) and then subtracting out this relation. After implementing these corrections, the detector becomes less sensitive to beam parameters and approaches its maximum possible resolution. For E158, seven individual bpms are used to remove the dependence on x and y position and angle on target, as well as energy related effects. Before this regression procedure is done, it is common for the raw resolution of the front detector to be at the 250 ppm level, while the back was often near 425 ppm. By noting the sensitivity of the detector to fluctuations in these beam parameters, and the corresponding resolution on that parameter provided by the bpms, the predicted final regressed resolution of the detector can be calculated. In this way, all sources of noise for the detector can be itemized. It should be noted that the bpms for El58 attained an excellent position resolution, at the level of 2 microns. Taking into account the noise contributions from bpm and toroid resolution, as well as including counting statistics and pedestal fluctuations, it is found that the resolution of the detector should be 110 ppm. This is significantly below the observed 170 ppm, pointing to an unknown noise source around the level of 100 ppm. The source of this extra noise contribution is still unidentified, though a likely candidate is electronics crosstalk. It has been observed that the resolution of the detector is greatly affected by adjusting timing parameters on the adcs. The 170 ppm resolution is only achieved for a very specific timing setting. For most other settings, the resolution worsens to 280 ppm. Research into the implications of this is still ongoing.

3.2. Linearity Conceptually, the linearity of a detector is a simple parameter. In practice, however, determining the linearity of the luminosity monitor to the level of 0.5% has proven to be quite a challenge. Though several methods have been employed, the most promising technique involves making cuts on beam parameters below bpm resolution and then plotting lumi asymmetry (not normalized by toroid) vs. toroid asymmetry. By making tight cuts, and working with asymmetries, the large effects coming from beam motion are mitigated. About 0.1% of the data survives this cut. The deviation of the observed slope from unity is then a direct measure of the linearity of the lumi-toroid system. Figure 4a shows the slopes obtained from this method for each chamber in the front ring of the lumi. Clearly, there is still a systematic effect related

733 u a

," 1 . m .

u

1

KO u o

<

I . 1

. .

.

.

,

.

2 3 4 5 6 7 F r o n t Chamber Number (a)

.I 8

2 0.97 F r o n t Sum

B a c k Sum

(b)

Figure 4. A slope of 1 indicates 100% linearity. Notice the azimuthal dependence to the observed slopes in 4a.

to position, since the azimuthal dependence is quite pronounced. One way to decrease this effect is to sum up an entire ring of the detector, and perform the same analysis. Since the gain of each chamber is essentially identical, by summing the full azimuth of a ring, position effects are greatly reduced. The suppression in sensitivity is seen to be at least factor of 10. This is quantified by comparing the response of a single chamber to that of an entire ring for a given deviation in a beam parameter. Figure 4b shows the results obtained by employing the same method as before on an entire ring of chambers. The error bar in the plot comes purely from the fit of the slope between lumi asymmetry and toroid asymmetry. As an estimate of the remaining systematic effect due t o beam position fluctuations, we note that the amplitude of the azimuthal oscillation in the Figure 4a was about 8%. With the order of magnitude in beam sensitivity suppression afforded by the sum, this brings the systematic error down to about 1%. The front and back rings are both consistent with being linear, then, with the slope of the front ring found to be 0.987f0.004(stat)f0.OlO(syst).

3.3. Synchrotron Radiation

To set an upper limit on the fraction of the detector signal coming from synchrotron radiation, the average signal size is compared with the target in and out of the beam. Figure 5 shows the ratio of these two configurations, on a per chamber basis. All chambers show that the maximum background is below the 1%level required by the experiment. Moreover, since chambers 3 and 7 are in the horizontal plane (the bend plane of the last spectrometer dipole magnet), the amount of synchrotron radiation they receive is orders of magnitude above the other chambers. It is precisely these chambers that show a pronounced spike in the above plot. Therefore the actual synchrotron background is more appropriately quantified as the height of these spikes above the level of the other chambers. The synchrotron background is then demonstrated to be well below the 1%requirement.

734

c

0.5

J-,

5

A v 4 1 -0.3 0 .

;0 . 2

1 2 3 4 5 6 7 8

1 2 3 4 5 6 7 8

Chamber

Figure 5. The ratio of average signal with target out to target in, versus channel number

3.4. Target Density Fluctuations Pulse-to-pulse target density fluctuations would present themselves as a correlation between the Moller and Lumi detectors. This is due to the fact that each detector has a sufficiently unique signal flux that the only real connection between them is the target itself. Figure 6 displays the Moller detector asymmetry vs. the lumi asymmetry. No significant amount of target density Holler Detector vs. Lumi

-1000-500

3

Asymmetry

500 1000

(wm)

Figure 6. There is no correlation between the two detectors.

fluctuations are observed. 3.5. Lumi as a BPM

Because the individual chambers of the lumi are exceptionally well gain matched, the detector was able to be used like a bpm for alignment purposes. The beam position is figured from a "center of mass" type calculation, using the signal size in each chamber as the mass, and the chamber position in space as the weighting factor. For deviations of less than a few centimeters from the center of the detector, this was found to be accurate to better than lmm. Figure 7 shows the calculated beam position in the front lumi ring, vs. different angle setpoints of the beam on target. They indicate that the beam would be centered in the front ring of the detector for the X-angle set to 6pR and the

735

Y-Angle set to OpR. This extra feature of lumi, derived from the simple nature

-2

-4 -20

-10

0

10

Y-Angle (pR)

;

Figure 7.

of the detector, assisted in the initial positioning of the El58 beam. 4. Conclusion

The luminosity monitor for El58 has proven to be a powerful, yet simple detector. The device has shown an exceptional per-pulse intensity asymmetry resolution of 170 ppm. The linearity of the device had been demonstrated at the 1%level for the first run of the experiment. The background contamination of synchrotron radiation has been shown to be at an acceptable level. The detector has been critical in looking for target density fluctuations. As more data from the first run of El58 is processed, the lumi will become an important monitor for false asymmetries. For the final run of E158, the goals will be to better understand the relation of the adc electronics to resolution and to demonstrate linearity to 0.5%.

References 1. K. S. Kumar et al., [Online] Available: http://www.slac.stanford.edu/exp/el58 /documents/proposal.ps.gz, (1997). 2. A. Czarnecki and W. Marciano, Int. J. Mod. Phys. A15, 2365 (2000). 3. G.P.Zeller et al., Phys. Rev. Lett. 88,091802 (2002). 4. C.S. Wood et al., Science 275,1759 (1997). 5. C.S. Wood et al., Can. J . Phys. 77,7 (1999). 6. S. C. Bennett and C. E. Wieman, Phys. Rev. Lett. 82,2484 (1999). 7. A. Derevianko, Phys. Rev. Lett. 85,1618 (2000). 8. V. A. Dzuba et al., [Online] Available: http://xxx.lanl.gov/abs/hep-ph/OlllOl9, (2001). 9. V. A. Dzuba et al., [Online] Available: http://xxx.lanl.gov/abs/hep-ph/O204134, (2002). 10. L. Keller, [Online] Available: http://www.slac.stanford.edu/exp/e158/ documents/technotes/LUMIAcceptance.ps.gz,(2001). 11. A. Bondar and E. Saldin, Nucl. Instr. Meth. 195,577 (1982).

This page intentionally left blank

Jet Measurement Cowener: J. Terron

*V. Zutshi

Simulations and Prototyping Studies for a Digital Hadron Calorimeter

M.N. Minard

Jet Measurements with the Aleph Detector at LEP2

H. Videau

Calorimetry Optimised for Jets

C. Schwanenberger

The Jet Calibration in the H1 Liquid Argon Calorimeter

M. Wing

Setting the jet energy scale for the ZEUS calorimeter

M. Tonnesmann

Jet Reconstruction at CDF

A. Savine

Suppression of Pile-up Noise in a Jet Cone

J. Krane

D8’s Recent Results and Experiences with the kT and Cone Jet Algorithms

I. Nakamura

Jet Measurement in OPAL

A. Kiiskinen

Developments on Jet Reconstruction by DELPHI

S. Magill

E-Flow Optimization of the Hadron Calorimeter for Future Detectors

R. Wigmans

On the Energy Measurement of Hadron Jets

*Written contribution not received $Not orally presented

737

This page intentionally left blank

JET ENERGY MEASUREMENT WITH THE ALEPH DETECTOR AT LEP2

M.N MINARD LAPP/IN2P3, Annecy le Vieux BP11074941, h n c e E-mail: [email protected]

The impact of the jet energy and direction measurement performances is studied through the 4 jets ( W pair production) at LEPZ. Emphasis is put on the sensitivity of the algorithms to the calibration of the jet component and to the the jet fragmentation modelling. The implication upon the systematics for the main physics channels are derived.

1. Introduction One of the main evolution between LEPl and LEP2 concerns the way the jet are used in the analysis. At LEP1, topological studies on jets were used in QCD related analysis as well as the analysis of their content, jets were also used for flavor tagging either by identifying their content or from their lifetime; at LEP2 in addition to the LEPl studies, jets are heavily used to identify and measure 2 or 4 jet final state. Among the subject of interest at LEP2, one of the most challenging one is the precise W mass measurement, reconstructed from jets, for which the expected statistical error from the 4 LEP experiment, using the whole LEP2 statistic, will be about 22 MeV. Therefore particular care is given to the evaluation of the systematic errors. Among systematic error sources those issued from the uncertainties on the jet reconstruction covers several aspects:

(1) (2) (3) (4)

the absolute calorimeter calibration. the jet direction and mass determination the jet hadronization scheme the quality of the detector simulation ( 5 ) the choice of the the jet algorithm

To achieved the best measurement from the 4 LEP experiments, the sensitivity of jet algorithm upon the final state modelling, which concerns effects correlated between experiment, has to be evaluated and minimized.

739

740

2. Hadronic j e t s

As mentioned above, at LEP2, physical processes involve jets in their final state as W+W- pair production inducing 4 jets when both W decay in quarks or 2 jet when one of the W decays semi-leptonically, or Higgs production in association with a Z boson. A crucial point in those analysis is to determine the di-jet invariant mass. The determination of both energy and direction with the best accuracy relies upon a precise measurements of all its components; using the Aleph energy flow algorithm technique, 60% of the jet energy is carried out by charged tracks, the remaining 40% coming from neutral particles, photons or neutral hadrons, can only be measured by the calorimeters3.

2.1. Calorimeters calibration The Aleph electromagnetic calorimeter is made of 36 modules (12 covering the central region, and 12 in each of the 2 endcaps), of 45 layers of lead and proportional tube, finely segmented, each element covering a solid angle of about 1 degree in polar and azimuthal dimension and achieving a resolution of o ( E ) / E= 0 . 1 8 / a 0.009. The gain of the gas in the proportional tubes is monitored by an Fe55 source, which allow to correct for time variation due to pressure or temperature fluctuations. The intercalibration of the 36 modules is done with electron in the range from 2 to 15 GeV issued from yy processes taken during the Aleph running. This procedure allows to reach a 0.3% precision upon the module intercalibration and take in account shower leakage and saturation effects measured in test beam data. The absolute calibration is performed with events recorded when the LEP operate at the Z, and controlled with bhabhas from high energy running; the quoted overall uncertainty upon the electromagnetic calorimeter is of 0.7%. The hadron calorimeter has 36 modules where iron layers are interleaved with streamer tubes, the whole arranged in projective towers with a granularity of 3.7x3.7. Analog readout on towers and digital one on cathode strips along the streamers tubes give a two dimensional information used for muon tracking. The energy resolution for charged pions measured in test beam was measured to be ( r ( E ) / E= 0.85/fi. The gas monitoring system' allow to correct for time dependence due to temperature or pressure fluctuations, the level of the correction can reach 10%. The overall calibration comes from the energy deposited by a muon crossing the calorimeter at normal incidence normalized to test beam energy response. The residual time dependence is monitored from yy events during the high energy run and found to be kept within 1.5%. To derived systematics due to calibration uncertainties, calorimeter response for both electromagnetic and hadronic calorimeter are compared to Monte-Carlo;

+

74 1

the Monte Carlo is re-scaled to the data taking in account the polar dependence and the response of each calorimeter element and is shaken according to the resolution. 2.2. Energy flow algorithm

Tuned at LEP1, the Aleph energy flow algorithm2 has been designed to optimize the total energy measurement and intensively used at LEPl for the Higgs search; it uses the good performance upon the track momentum measurement and takes advantage of the finely segmented calorimeters to disentangle the different contributions: (1) charged tracks and identified leptons contributions are taken from their tracking measurement (2) y and 7ro from the electromagnetic calorimetry (3) neutral hadron from both calorimeter measurement (4) the last component being the residual from charged hadrons or y which should be kept at the lowest level

The performance of this algorithm has been tested at LEPl using Z decays into 2 acoplanar jets accompanied by an high energy photon. The invariant di-jet mass reconstructed from the energy flow objects can be compared to the recoil mass against the high energy photon. This method allows to measure the jet resolution as a function of the jet energy to be: a(E)/E = (0.59 -0.03)/* (0.6 -0.03) GeV The dependence upon the jet energy is shown on Figure 1 and the expected resolution at high energy is derived from this measurement. For Z events among 60% of reconstructed energy comes from tracking measurement, 32% from the electromagnetic calorimetry; from LEP2 study the sharing of the energy remains identical.

+

+

+

2.3. Jet reconstruction performance

To reconstruct jet the energy flow objects are clustered using Durham algorithm where objects for which the distance yij = 2min(E:, E?)(l-cosBij)/E:m is smaller than a certain threshold are assigned to the same jet. According to the topic studied either the maximal distance is fixed or adjusted in order to force the event into a certain number of jets ( for exemple 4 in the case of W pair production decaying into 4 quarks). The jet performance resolution in term of energy and direction has been studied with Z data. The jet energy reswhich correspond to 10% at 45 olution behave as a(Ejet)/Ejet= 0.67/*jet, GeV where a 6-7% resolution is expected for a perfect detector measured from

742

t

ALEPH

uE= 0.59&+0.6

GeV

2 = 6.6 for 6 d.0.f

\.

'. \. \.

/

\

on MWs '. '. Resolution on MREC

- Resolution

I

0

\.

s

I

I

20

I

40

,

I

I

I

t

'.

'.

'. \.

,

I

I

,

k ,

80 100 Invariant mass (GeV/c*)

60

Figure 1. Measured resolution on qq-y from Z decays

Monte Carlo parton shower. The angular resolution is studied from Z data where the 2 jets are expected to be back to back, an angular resolution of 0.9 is observed for energy flow jets in good agreement with Monte-Carlo expectations which represents a large improvement compare to the 1.6 and 1.4degree obtained respectively for jets build from charged tracks and calorimeter objects. 3 . W mass reconstruction

W mass measurement is based on a precise reconstruction of the jet-jet mass. The whole method uses large simulated events inputs and requires a fine and detailed matching between all the components; therefore the understanding of detector effect as well as the physics modelling of parton shower will allow to improve the measurement precision. 3.1. W Reconstruction method

In the case of W pair production where both pair decays into quarks, the event is forced into 4 jets and the procedure used takes advantage of the kinematical

743

constraints from the known total energy available in the center of mass, rescaling the jets energy and momentum as CBiEi = Ecm,with the momentum balance constraint Czl'iipipi = 0, leaving the jet direction unchanged. In a more elaborate recursive fitting procedure the jets direction and momentum are allowed to vary within their expected errors while a 4 momentum conservation constraint is applied (4C-fit), jet masses can be forced to be equal in a 5C-fit procedure. By using kinematic fitting procedure the resolution improves as shown on Figure 2 The fitting procedure implies a large use of Monte-Carlo

Wmass Figure 2. Reconstructed W m a s from raw jets (dashed), energy re-scaled jet (dot), and 4c fit (full ).

samples, the di-jet mass is fitted by changing the underlying W mass parameter and applying weights calculated from the matrix element ratio when changing the W mass parameters. The accuracy of the method is based on the control of the agreement between simulated and real events. 3.2. Jet mass dependence

For massless partons the invariant mass of 2 of them is defined as M i f = 2EiEj(1- coseij, therefore the energy of each of the jets and the angle determination axe important feature. The uncertainty linked to a jet energy mea-

744

.-dz 1.04 c.l

L

L

0

u

1.03 I.02 L

II

t

I

1-01

0.99 0.98 0.97 0.96

0.95

Figure 3. The correction factor to the jet energy determined from jet energy comparison at the 2 is plotted as a function of the polar angle

surement difference between data and Monte-Carlo is estimated by re-scaling the jet energy taking in account the angular dependence shown on Figure 3 As mentioned above the jet angular resolution is in good agreement between data and Monte-Carlo4. But the hypothesis of massless parton is too trivial, and reconstructed jets have a mass which is sensitive to several effects: (1) jet multiplicity due to fragmentation scheme or induced by detector effects (2) particles not assigned to the right jet (3) final state modelling and among them color reconnection effect One should notice that systematic differences in jet mass between real and simulated events will induce systematic upon W mass determination : ~ ( M w= ) mi/lMwG(mpta- myC)Oij,typically a 100 MeV shift in the jet mass determination will induce a 40 MeV systematic effect on the W mass. To estimate the corresponding systematics detailed Monte-Carlo/Data comparison are conducted in order to only use for the mass determination the component for which a good agreement between simulated and real data is

745

achieved. Detailed comparison of simulated and real jets had been done, on the energy spectrum and multiplicity of their different energy flow components, it as shown that at low energy the multiplicity of energy flow objects coming from hadron and photon residuals in Monte-Carlo poorly match the data ( a factor 1.3 for element below 1 GeV). Removing from the jet definition such energy range for this type object does not affect the expected mass resolution as shown in Figure 4 and still improve the determination compare to a massless hypothesis for which the expected error will 122MeV.

189 GeV - 4qChannel

8 0.13

T

50.12 2 0.11

Q)

Figure 4. W mass resolution evolution as a function of the an energy cut upon the energy flow object. In addition the expected uncertainty for a massless hypothesis is quoted

The systematics uncertainties derived from the above studies are summarized in Table 1. The uncertainty quoted for the calorimeter simulation is expected to decrease with the improvement in the understanding of the jet energy flow component. The last line of this table concerns the uncertainties arising from hadronic scheme, it concerns: the fragmentation scheme for which the associated uncer-

746

Table 1. Systematics uncertainties for W mass measurement for fully hadronic W pair and for events where one of the W decay semileptonically. 4q

Calorimeter calibration Jet calibration

4 MeV 7MeV

h B 5 MeV

lOMeV

Jet angle

5 MeV

4 MeV

Calorimeter simulation

10 MeV

15 MeV

FSI(CR,BE,Fkagm)

48 MeV

20 MeV

tainty is derived using several scheme ( Ariadne-Herwig-Jetset). (1) the fragmentation scheme for which the associated uncertainty is derived using several scheme ( Ariadne-Herwig-Jetset). (2) the jet particle association. (3) the bose-einstein effect (4) color reconnection level for W pair going to 4 quarks These effects lead to a low weight in the W mass determination of the 4quarks channel (27%). This uncertainty should be reduce using alternative jet algorithm designed to be less sensitive to the final state interaction. 4. Conclusions

At LEP2 the use of energy flow technique for the jet measurement allows physics measurements involving two and four hadronic jets final state. On top of it the use of kinematic fitting procedure requires a good matching between real and simulated data, leading to a more demanding understanding of the detector feature. Recent progress in this field will allow improved and more robust uncertainties determination for the W mass measurement.

Acknowledgments

I would like to thank my ALEPH colleagues, in particular Andrea Venturi, Eric Lancon and Stephane Jezequel for their help and disponibility. References 1. 2. 3. 4.

A.Messimeo et a1,Nucl. Znstr. and Methods A320,177 (1992) ALEPH collaboration,Nucl. Znstr. and Methods A360,481(1995) D.Decamp et al, Phys Lett. B 2 4 6 , 306 (1990). ALEPH collaboration, Eur-Phys J. C17, 241 (2000).

CALORIMETRY OPTIMISED FOR JETS

HENRI VIDEAU, JEAN-CLAUDE BRIENT Laboratoire Leprince-Ringuet Ecole polytechnique - F-91128 Palaiseau, France E-mail: Henri. Videau@ini!pZpJ.fr The physics programme for a coming electron linear collider is dominated by events with final states containing many jets, dijets from H, W, 2 . We contend that, in the energy range under consideration, the best approach is to optimise the independent measurement of the tracks in the tracker, the photons in the electromagnetic calorimeter and the neutral hadrons in the calorimetry, together with a good lepton identification. This can be achieved with a good tracker and a high granularity calorimetry providing particle separation, through an efficient energy flow algorithm. But we do not contend that this is a universal panacea. Following that programme from the calorimetric side on hardware and software issues is the goal of the CALICE collaboration.

In this paper we will discuss four topics, one deals directly with reconstructing jet energies in the energy range of the linear collider, the second presents a possible implementation of a calorimeter for that purpose (namely one TESLA TDR implementation), the third presents current results on reconstruction and specifically on photon reconstruction, the last gives some status of the developments going on in the CALICE collaboration for a digital hadron calorimeter hardware. 1. Measuring the energy of jets, the requirements

1.1. The physics

The coming linear collider will cover an energy range from the Z mass to the TeV. Among the subjects of interest, aside the top production, the Higgs production in e+e- + Z H , the Higgs self-coupling e+e- + Z H H , the production of bosons e+e- -+ W + W - , e+e- + WWvO or e+e- + ZZvO, not to speak about supersymmetric channels. To extract at best all the physics we will have to manage the multijet (dijets) events and to identify and measure well the leptons, electrons and muons but also the taus which are slightly more challenging. It may be of interest to have a glance at the object of our study. How do these jets look like. This can be seen in figure 1.

747

748

Figure 1. Having a look at jets, on the left a radiative qtj at 500 GeV, on the right a W from e+e- + WW at 800 GeV.

On the left image you see a qq at 500 GeV, in a frequent configuration with a large radiative photon leaving to the right. The two jets with a Z mass are strongly boosted toward the end cap. On the right a W from e+e- -+ WW at 800 GeV is also boosted toward the end cap due to the t-channel exchange process. The direction of the view optimises the separation of the dijet particles. The approach we consider to reconstruct the jets is called the “analytical energy flow method” (AEF). It has been used extensively at LEP in particular by ALEPH. An illustration of the ALEPH detector behaviour is shown on figure 2. This is a peculiar tau event where the tau decays through T -+ vT7rK,Kl. The Kl is easily seen in the HCAL, the pattern is obtained by the streamer tubes parallel to the beam and read digitally, the energy can be measured using analog signals from cathode towers or counting the tubes. The LEP example is relevant because, the multiplicity increasing, the energy spectrum of the different particles is quite similar at LEP and at the higher energies of the linear collider. But other lessons can be learned from the LEP detectors: if you want to properly measure the neutral hadron energy it is particularly nasty to insert a thick coil in front of the HCAL, empty projective cracks should be avoided in ECAL and HCAL, the longitudinal segmentation has to be good enough, and, if the ALEPH tau shows the interest of the digital pattern, to make use of the pattern in more complex events a 3d read-out is mandatory.

749

Figure 2. A remarquable tau decay seen in ALEPH: T t V , T K , K ~ The . Ki is seen as a pattern in the HCAL and an imbalance in the energies.

1.2. Impact on the physics programme of the jet resolution The first argument in favour of the effort to ameliorate the jet resolution is in the impact a given resolution has on specific studies. Only few channels have been up to now studied and this is a place where more work is absolutely needed. To quantify, the resolution has been parametrised in a very classical way as A E = a / a . The absence of a constant term may seem inadequate, but the value we use in the most favourable case ( a = 0.3) has been obtained with that formula in a simulation study of jets at the Z. For e+e- -+ Z H H it has been shown that, for a luminosity of 1ab-l, going from a = 0.6 t o 0.3 improves the significance of the observation of the signal from 3 to 6 u , the difference between a hint and a discovery. For e+e- + Z H where Z + qQ and H + W W * going from 0.3 t o 0.6 is equivalent t o loosing 45 % of the luminosity. A visual evidence is shown on figure 3. The reactions e+e- + WWuii and efe- + ZZuii have been simulated in the TESLA detector through a fast simulation taking properly into account the imperfections of the detector, we can remark that the typical energy of the jets making the W’s is not dramatically different from the energy of a jet

750

at the Z peak and then not introducing a constant term has little effect. The separation between WW and ZZ is clear on these scatter plots for 0.3 and purely statistical for 0.6. This can be quantified as an equivalent loss of luminosity of 40%. Thus the price of a more elaborate calorimeter is to be compared with the corresponding additional running cost.

Figure 3. 0.3/8?.

Separation between e + e - t WWuP and e+e-

+ ZZuP

for a jet resolution of

1.3. Reminder of the basics of the analytical energy flow

method The first point is to remember that we want both to measure well the jets and properly identify the leptons even inside jets. The fact that muon energy can not be measured by calorimetry as well as the distorsion introduced in a global calorimetric energy flow by the action of the strong field on charged particles are, aside the resolution obtained, the arguments for an analytic energy flow rather than a purely calorimetric one. The presence of a lepton in a jet is the signature of a neutrino, even signing t m s in jets may be worthwhile. The energy flow of a jet is written as the sum of its components: pjet = C Pchargedparticles + C P y + C Pneutralhadrons, The basic argument then goes as follows: the charged particles make about 60% of the energy and, being of rather low energy, are much better measured by the tracker, the goal is just to isolate the 10% neutral hadron energy. Such a method relies much more on the separation of particles than on the calorimeter intrinsic energy resolution.

751

The basic requirements for a calorimeter design are easy to derive. The calorimeter has to be far enough from the interaction point, entirely inside a coil providing a strong field to widen the jets, it has to be compact with a small radiation length, a small interaction length but large compared to the radiation length in the first part dedicated to the measurement of photons and electrons. Then the read out has to have a granularity matched to the radiation length plus the capability to follow minimum ionising particles as a tracker. We need a good tracker, not so much on momentum resolution but with good track efficiency, small rate of fake energetic tracks, good Vo identification and small rate of reinteraction. We need a good electromagnetic calorimeter, not so much on resolution but with good photon efficiency, even close to hadrons, a small rate of fakes from hadronic debris, a good electron identification (prompt electrons, this is a requirement for the tracker). We need a good hadron calorimeter to identify muons in particular at energies where they do not reach the muon system, to disentangle neutral hadronic showers from charged ones in order to measure their energy.

2. Elements for a calorimetric design (from the TESLA TDR) Few words to recall the calorimeter design as seen in the TESLA TDR. The electromagnetic part of the calorimeter is made of a sandwich of tungsten radiator and silicon diodes detector. A total depth of 24 Xo in 40 layers provides a resolution around O . l l / f i for a total thickness of 20cm. The way to dispose the modules in an eightfold way (figure 4) eliminates the cracks between modules. The cell size matched to the mean Molihre radius is about lcm'. The efficiency to minimum ionising particles is obtained by a mips to noise ratio of about 10. The hadronic part has a radiator adapted to the optimisation between intrinsic resolution and separation. This is illustrated in figure 5 where a jet is shown developping in an iron structure and in a mixture incorporating tungsten. In the second version the electromagnetic subshowers are kept much smaller inducing a loss in resolution but better separation. The cells are kept very small lcm2 but this granularity is payed by a yes/no (digital) read out. In fact we will see in the following section that we do not loose anything. In the CALICE collaboration another HCAL solution using scintillator tiles with an analog read out is being studied as well. Note that the cracks between modules are kept projective, but they are very small compared to a shower and for muons which would escape detection, it is evident that a particle going through iron without any interaction is a muon.

752

Figure 4. A view of the barrel calorimeter showing the eightfold structure.

Figure 5. Where it appears that, even for a same mean interaction length, stainless steel and wolfram do not produce similar showers.

3. Some results about reconstruction

The calorimeter design requirements have to be validated by a full simulation. In the following studies the simulation used is MOKKA, an application

754

Visible Energy (MC truth - recons.) GeV Figure 7. Resolution as the difference between the Monte Carlo truth and the reconstructed energy for jets at the 2 mass.

models for example. A proper reconstruction is the cornerstone of jet calorimetry. In the following figures some results on photon reconstruction at the highest energies are shown, specifically from efe- + WW at 800 GeV. They have been obtained with REPLIC a code you can find on the CALICE web site. First we discuss the information on an event by event basis, then on a photon by photon basis. The figure 9 shows the comparison between reconstructed and generated photon multiplicity in an event. On the left a scatter plot shows that the number of fake photons equals quite well the number of lost photons. Their energy distribution is shown on the right. A similar information regarding the energy is shown on figure 10. It is clear that the total amount of photon energy in an event is properly obtained. On a photon by photon basis, the figure 11 shows the energy spectra for generated and reconstructed photons. The agreement is good in general but an excess of reconstructed photons appears below 500 MeV. These are fakes from hadronic debris. The figure 12 shows a similar distribution but the “true” photons only have been kept. It measures the efficiency which drops strongly below 250 MeV. The way to assess the fact that a reconstructed photon is a true photon is by requiring that more than 75 % of its energy comes from

755

Figure 8. Looking at the impact of a W dijet from e+e- + W W on the first 4 Xo of the electromagnetic calorimeter. This is a 0 - C#J view, the size of the image is 100 mrad. The crosses are the generated photon impacts (8), the pluses the imapct of the charged tracks (4), the stars those of the neutral hadrons (l), the circles are the reconstructed photons.

i

I03

d

120

Figure 9.

Photon multiplicity per event reconstructed versus generated.

a generated photon. The efficiency can then be derived and is presented for the low energy part in figure 13. Above 2 GeV it is essentially 1. Finally

756

..-

Figure 10. Photon energy per event reconstructed versus generated

I

Figure 11. Photon energy distribution, generated (histogram) and reconstructed (points), the left figure is in linear scale, the right one in log scale.

the energy resolution for the photon contribution in e+e- + WW at 800 GeV is presented in figure 14. It shows the difference between generated and reconstructed photon energy in the jet. The distribution is well symmetric, it has been fitted with two gaussians, the narrow one has a mean of 0.23 GeV and a width of 7.01 GeV, the wide one, which corresponds t o a third of the narrow, has a mean of -.02 GeV and a width of 18.5 GeV. Even if the photon reconstruction appears in a rather good shape, a more complete reconstruction of the jets at high energy has to be done and is under way.

757

'0

0

5

10

15

20

25

0.5

30

35

I

1.5

2

40

45

50

Figure 12. Energy distribution for generated photons and true photons reconstructed.

-YO0 -80 -60 -40 -20 0

Figure 13. Photon efficiency at low energy.

20 40 60 80 100

Figure 14. Energy resolution for the photonic contribution to jets.

4. Few more informations about the digital HCAL solution The results presented above rely on the simulation of a very granular HCAL. Is there a technology for such a detector? We discuss here quickly a solution

758

Hv layer

glass electrodes

/ G10

/pads

glass electrod?

/ ground layer

/mylar

my17

spacer

HV l a y 7

spacer,

A=-

/ bakelite Figure 15.

//pads

/varnish

/ground layer

Scheme of a RPC with pads outside or inside the chamber.

under development. The requirements for a sensitive detector are to be compact, efficient to mip, with a high output signal and cheap. A straight idea is to consider a gas detector, either a streamer or Geiger wire detector or RPC’s. Currently the last solution is actively pursued in a subcollaboration of CALICE involving IHEP, Interphysica, LLR, MEPhI and Seoul university. The figure 15 shows two designs for a glass RPC. In the first one, on top, the read out pads are outside the glass panels. To get a larger signal and to reduce cross-talk between adjacent pads a solution with pads inside is also considered, bottom. The gap is 1.2 mm, the glass plates are lmm thick and the pad size is 1x1 cm2. With a mixture of tetrafluoroethan, nitrogen and isobutane in the proportion 80/10/10 a signal of 3V is obtained on 500 and

759

Figure 16. Scheme for implanting the read out directly on the RPC.

the efficiency to mip is better than 98%. The figure 16 shows the structure of the HCAL gap. Between the two radiator layers made of stainless steel ot heavier material, a classical glass RPC is installed. The pads are put on a multilayer G10 board containing the connections of 64 pads to a chip seated on the board, as well as the power, command and read out lines. Figure 17 shows then a possible read out scheme for a 64 channel chip followed by the drawing of a token ring reading the chips serially. The current idea is to have the 64 pad lines arriving in parallel to the chip. If one line at least is up, the 64 bits are recorded together with an identification of the bunch crossing time. Between trains, the chips memories are read one after another, the chip address being added. Remember that, if the number of channels is as large as 64 million channels, the occupancy of the cells is very low, typically a 20 GeV pion shower generates between one and two hundred hits. 5. Perspectives and conclusions

To extract the physics produced in an electron linear collider below 1 TeV, a measurement of the jet energies with a stochastic term at a level of 0.3 or below seems mandatory. Such a precision does not seem out of reach with an adequate calorimetric hardware and a proper software.

760

Clear

n ,

I

.

Figure 17. Scheme for a RPC digital front end chip (above) and its read out (bottom).

We have a roadmap with hardware developments and prototypes ( in beam in 2004) and with a lot of software imagination. If you like to play this game, join the worldwide effort of the CALICE collaboration. http://polywww.in2p3.fr/tesla/calice.html

THE JET CALIBRATION IN THE H1 LIQUID ARGON CALORIMETER

C. SCHWANENBERGER Deutsches Elektronen-Synchrotron DESY, Notkestrajle 85, 0-22607 Hamburg, Germany E-mail: [email protected] (On behalf of the H 1 Collaboration)

The jet calibration of the Liquid-Argon-Calorimeter of the H I Detector a t HERA is described. In the measurement of high jet transverse energies systematic uncertainties as low as 2% can be reached in deep inelastic scattering with a high photon virtuality (0')and in photoproduction. Furthermore, the concept of a new energy weighting scheme of HI is presented. First applications with a high &' neutral current deep inelastic scattering sample show that the resolution of the balance in transverse momentum between the hadronic system and the electron is improved.

1. Introduction

In the H1 Experiment' at HERA a Liquid Argon (LAr) Calorimeter is used to measure the energies of particles emerging from the interaction of 27.5 GeV positron (or electron) and 920 GeV (or 820 GeV) proton beams over an angular range of 4 O 5 0 5 154O." The H1 Liquid Argon Calorimeter' is a non-compensating sampling calorimeter which is divided into 8 self supporting wheels each built out of 8 octants in the barrel part or two half shells in the forward part. It consists of electromagnetic and outer hadronic sections. In the electromagnetic modules lead is used as absorber material which adds up to 20 - 30 radiation lengths ( X O ) .The hadronic part is built out of stainless steel absorber plates which corresponds to a total of 4.5 8 interaction lengths (A) including the electromagnetic section. ~

~~

aH1 uses a right-handed coordinate system, where the direction of the proton beam defines the positive z-direction. The polar angle 0 is measured with respect t o this direction.

761

762

2. Calorimeter Calibration using H1 Physics Data

Knowledge of the LAr calorimeter energy scale can be improved using neutral current (NC) deep inelastic scattering (DIS) physics data in off-line analysis. With the increased amount of data collected in the past years the precision in this calibration method has reached the design level. 2.1. Energy calibration with DIS data The over-constrained kinematics of NC DIS events at HERA allow the prediction of the energy E, of the scattered electronb from the electron beam energy Ee-beam,the scattering angle of the electron 8, and the effective angle of the final state 8 with the double angle method (DA)3:

The remaining final state can be a hadronic shower (DIS-DA) or a photon (QED-Compton-DA). Using the predicted energies E, , position dependent calibration factors for the electromagnetic scale are derived. The hadronic energy scale can be adjusted using the known electron energy. The scale correction factors for the electromagnetic and hadronic sections are obtained wheelwise from the ratio of transverse momenta of the calibrated electron and the hadronic final state.

&3.9251,, 0.90

, I , , ,

20

,

40

I

,

,

,

60

,

, ,

I , ,

80

,

I , ,

100

,

I , ,

120

,

,

1

140

8/" Figure 1. The ratio of data and Monte Carlo prediction for the pt-balance ratio as a function of the mean hadronic scattering angle 8 , for N C DIS with high momentum transfer Q 2 and for hadronic transverse momenta between 12 and 25 GeV.

The uncertainty of the hadronic energy scale is defined by the difference between the correction factors in data and Monte Carlo simulation. As can be seen in Fig. 1, the uncertainties are below 2% in a wide angular range of bIf not particularly emphasized, electron can mean either an electron or a positron.

763

calorimeter acceptance. Larger deviations can be found in the backward part where mostly low energy hadrons are accepted and where energy leakage occurs due to the missing hadronic section in the most backward wheel. 2.2. Dijet Data in Photoproduction The calibration procedure obtained from DIS data is applied to the measurement of the dijet cross section in photoprod~ction~. The comparison of the ratio of the transverse energy of the highest energy jet Et,max and the transverse energy of the rest' Et,rest for data and Monte Carlo simulations indicates the quality of the hadronic calibration. This ratio is shown in Fig. 2 (a) as a function of the highest energy jet. The ratio of E t , m a x / E t , r e s t in data and simulations, which is shown in Fig. 2 (b), is consistent with uncertainties of the hadronic energy scale below 2%.d Detailed studies5 demonstrate that at large transverse momentum these scale uncertainties are independent of the angular distribution and the mass of the jets as well as of different data selections such as direct, resolved or diffractive processes. Et,,,,

- - HERWIG

4-

1.2

=8=+ 1

Jet-Rest balance 0.8

40

60

Et,max

80

(GeV)

0.9

40

80

60

Et,max

(GeV)

Figure 2. (a) The ratio of the transverse energy of the highest energy jet and the transverse energy of the rest for data and two different Monte Carlo simulations as a function of the transverse energy of the highest energy jet. (b) The ratio of data and Monte Carlo prediction.

In order to illustrate the impact of these uncertainties on a cross section measurement, the relative difference between the measured and the theoretical cross sections6 is given in Fig. 3, as a function of x7, the longitudinal momentum fraction of the photon taken by the interacting parton. The correlated CThe rest is given by the energy of the second jet plus remaining energies in the event. dThis ratio is larger 1 due to the bias from the selection of the jet of highest Et and losses for Et,rest in the beam pipe.

764

errors due to the uncertainty in the calorimeter energy scales are shown as a shaded band. Fig. 3 shows that the assumed scale uncertainties in the nextto-leading order QCD calculation are the dominant source of uncertainties in the comparison of data and theory. 25<%3m,x<35GeV

0.4 0.2

35
GeV

M data: correlated uncert.

0.2 0 -0.2 0.4 0.2 0

0 -0.2

0.4 0.2 0 -0.2

4.2 0.2

0.4

0.6

0.8

1

0.2

0.4

0.6

x;.

0.8

1

xr

Figure 3. The z-,dependence of the relative difference of the measured dijet cross sections (Q2 < 1GeV2) from the NLO prediction, with hadronization corrections applied (here u . ~ h ~The ~ ~ symbol ~ ) . u stands for du/dxy. Figures a) and b) show the relative difference for a lower Et,max and a higher Et,maxregion respectively. The inner error bars denote the statistical error, the outer error bars denote all statistical and uncorrelated systematic errors of the data added in quadrature. The correlated systematic errors are shown in the middle plots as a shaded band. The bands in the lower plots show the renormalization and factorization scale uncertainties of this NLO prediction.

3. Towards a New Energy Weighting Scheme Because the H 1 LAr calorimeter is non-compensating, a software weighting method7 is applied for the energy reconstruction. In order to overcome some deficits in the low energy regime of the current energy weighting scheme a new weighting procedure was studied using CERN test beam data*. The reconstructed energy E,ieCin a calorimeter cell i is derived from the measured cell energy Ei by

EieC= w ( E t / VOP,Egroup ) . E; .

(2)

The weighting factor w depends on the deposited energy density E i / Volz (VoEi being the cell volume) which is different for an electromagnetic and a hadronic deposition, and on the total energy t o be reconstructed (energy of the group of selected clusters Egroup).The latter accounts for the fact that

765

both the relative difference between the energy response to electrons and to pionse and the fluctuation of the electromagnetic component in the hadronic showerf depend on the energy. The weighting factors w are tabulated and derived wheelwise, separately for the electromagnetic and the hadronic parts of the whole calorimeter, using a more detailed simulation of single pions as standard for H1 physics analysesg. Furthermore, noise corrections are applied. The method is valid from the highest energies down to the noise level. For the final calibration real DIS data are used: the pt-balance is adjusted wheelwise to ppd/pE -+ 1. Fig. 4 depicts the change in the pt-balance distribution if either the current (a) or the new (b) energy weighting scheme is applied. In the current energy weighting scheme there are too many entries for large p:ad/pp values which originate from neutral pions. In the new energy weighting scheme there are fewer entries in the tails of this distribution. Furthermore, the shape is more Gaussian-like and the resolution is somewhat improved. The jet energy dependence of the pt-balance is also given in Fig. 4. In the new energy weighting scheme an improvement of the hadronic energy response at small jet energies can be observed (c) and the agreement between data and Monte Carlo simulation is very good (d). 4. Conclusions

It has been discussed how physics data from electron proton scattering can be used for the jet calibration of the LAr Calorimeter. In analyses of photoproduction and NC DIS with high momentum transfer Q2,the calibration uncertainties are of the order 2%. A new software weighting scheme was discussed. First applications in DIS with high Q 2 show an improvement in the hadronic energy response at low jet energies. Using this energy weighting scheme, the distribution of the ptbalance between the hadronic system and the electron gets more Gaussian-like compared t o the current weighting.

Acknowledgments I would like to thank the H1 Calorimeter and the Energy Scale Working Groups for many discussions and the opportunity to present these results at this very interesting conference. My thanks to the organizers. eFor instance, a t an energy of 10 GeV, the response to electrons is a factor 1.35 larger than the response to pions. ‘This is due to neutral pion production in nuclear reactions in the hadronic component of the shower.

766

. 5

0.12

P

L

0.4

aw 0.w 001

am I

Figure 4. Distribution of pt-balance for the current (a) and a new energy weighting (b). The data are compared to the Monte Carlo simulation prediction (histogram) and to a Gaussian fit (curve). (c) Mean pt-balance ratio as a function of the jet energy. Data (dark points) and Monte Carlo simulation prediction (dark squares) in the new energy weighting scheme are compared to data (light points) and Monte Carlo simulation prediction (light squares) in the current energy weighting scheme. (d) The ratio of data and Monte Carlo prediction in the new (dark points) and the current (light points) energy weighting scheme as a function of the jet energy.

References 1. I. Abt et al. [Hl Collaboration], Nucl. Instrum. Meth. A 386, 310; ibid. A 386, 348 (1997). 2. B. Andrieu et al. [Hl Calorimeter Group Collaboration], Nucl. Instrum. Meth. A 336, 460 (1993). 3. S. Bentvelsen et al., “Proceedings of the Workshop Physics at HERA, Vol. l”, eds. W. Buchmuller, G. Ingelman, DESY (1992); ibid. K. C. Hoger, p. 43. 4. C. Adloff et al. [Hl Collaboration], “Measurement of dijet cross sections in photoproduction at HERA,” arXiw: hep-ex/O201006. 5. S. Caron, Ph.D. Thesis, “Jets in photoproduction at HERA”, RWTH Aachen, to be published. 6. S. Frixione and G. Ridolfi, Nucl. Phys. B 507, 315 (1997); S. Frixione, Nucl. Phys. B 507, 295 (1997). 7. I. Abt et al. [Hl Collaboration], Nucl. Instrum. Meth. A 386, 348 (1997). 8. C. Issever, Diploma Thesis, “Die Entwicklung eines alternativen Gewichtungsverfahrens fur das H1-Kalorimeter”, Interner Bericht, D E S Y FH1-96-06, August 1996. 9. J. Marks, Internal Report, DESY, to be published.

SETTING THE JET ENERGY SCALE FOR THE ZEUS CALORIMETER

M. WING Bristol University, DESY F1, Notkestmsse 85, 22607 Hamburg, Germany E-mail: [email protected] ( O n behalf of the ZEUS collaboration)

A much improved determination of the transverse energy of jets has been carried out in ZEUS, using a correction procedure based on two independent methods. The first is based on a combination of tracking and calorimeter information which optimises the resolution of reconstructed kinematic variables. The conservation of energy and momentum in neutral current deep inelastic e+p scattering events is exploited to determine the energy corrections by balancing the kinematic quantities of the scattered positron with those of the hadronic final state. The method has been independently applied to data and simulated events. The second method uses calorimeter cells as inputs to the jet algorithm. Simulated events are then used to provide a correction for the energy loss due to inactive material in front of the calorimeter. A detailed comparison of the jet transverse energy and the transverse energy of tracks in a cone around the jet provides the final correction. This procedure relies on an accurate simulation of charged tracks and so is less reliant on simulating the energy loss of neutral particles in inactive material. Final comparisons of the data and simulated events for both methods allow an uncertainty fl% to be assigned to the jet energy scale.

1. Introduction

The energy scale uncertainty of the calorimeter (CAL) coupled with differences between data and Monte Carlo (MC) simulations has traditionally been the dominant systematic uncertainty in jet measurements from the ZEUS collaboration. Energy scale uncertainties of f ( 3 - 5)%, lead to uncertainties of &( 10 - 20)% in the cross-section measurements'. The HERA accelerator collides positrons of 27.5 GeV with protons of 820 (or 920) GeV leading to heavily boosted final states. Neutral current deep inelastic scattering events with high momentum transfer, Q2,provide both interesting physics and the opportunity to study and calibrate the CAL energy scale. Quark-parton model type events, in which the positron scatters off a quark in the proton producing a final state jet back-to-back with the scattered

-

767

768

positron, have been selected. After setting the electromagnetic energy scale, as discussed in section 2, a comparison of the scattered positron energy with that of the hadronic jet allows the determination of the uncertainty on the hadronic energy scale as discussed in the rest of this paper. 2. Electromagnetic energy scale uncertainty

The electromagnetic energy scale has been studied in detail2 by taking the ratio of the energy of the scattered positron measured in the calorimeter, Eel with the track momentum or the electron energy reconstructed via the double angle (DA) method, EDA.The DA method predicts the electron energy from the angular information of the scattered positron and the hadrons and is, therefore, t o first order, independent of the absolute energy scale of the CAL3. The difference between the energy ratio in data and MC is shown in figure 1. The agreement is within kl%.

-

BCAL + corr +smear + ZrSOcm BCAL + corr + %,<1.5 GeV +b44GeV

P

h

OQ

UiFJ

..... ............................................................. ~

v

7

8 910

20

30

40

50 60 70 8090100

rGeV1

Figure 1. Difference between data and MC of the energy scale for scattered positrons as a function of the energy, E D A .

3. Jet energy scale uncertainty

Due to inactive material between the interaction point and the CAL, jets need to be corrected for the energy loss. Typically, 20% of the jet’s transverse energy is lost and is the major factor t o be accounted for in order to produce an accurate determination of the uncertainty in the jet energy scale. TWO different methods for correction have been developed.

769

3.1. Method 1

The first method uses energy flow objects (EFOs) to reconstruct the final state in which a combination of tracking and calorimeter information is used4. Clusters of cells are formed and combined with tracks originating from the primary vertex and a decision made on whether to use the cluster or track. In the case of isolated clusters or tracks, the decision is trivial. For a matched cluster-track system, the resolutions of each object and ratio of energy to momentum are considered. Using this procedure, a list of track-EFOs and CAL-EFOs was obtained, where the track-EFOs are assumed to be an accurate measure of the particle energy and the CAL-EFOs are subject to energy loss in the inactive material and must, therefore, be corrected. The conservation of energy and momentum in NC events was exploited to determine the CAL-EFO energy-correction functions by balancing the momentum of the scattered positron with that of the hadronic final state5i6y7. Two samples of events were used, both with Q2 > 100 GeV2; one sample had high positron p~ and the other sample had high y. The variable y is the fraction of the lepton energy transferred to the proton in its rest frame and is a measure of the effective longitudinal momentum. Using the two samples, full angular coverage of the detector was achieved. The kinematic variables of the positron were reconstructed using the DA method. The hadronic final state four-vector was calculated from the EFOs reconstructed as above and its momentum components balanced with that of the scattered positron. The CAL-EFOs were corrected for energy loss as a function of the cluster energy in several angular regions (reflecting the detector geometry). The difference between p~ and y for the hadronic system and scattered positron was minimised and correction factors obtained separately for data and ARIADNE' and HERWIG' MC simulations as shown in figure 2. It can be seen that the data and MC show similar

-

E

2.2

2

1.8 1.6 1.4 1.2 1

A

E

2.2 2 1.6 1.6 1.4

1.2 f

Figure 2. Energy corrections as a function of cluster energy in bins of 0. The corrections are shown separately for data (solid line) and HERWIGMC (dashed line) and ARIADNE MC simulations (dotted line).

770

trends but differ in detail, justifying the need to perform the fits and apply the corrections separately for data and MC. To test the validity of the procedure, the correction functions were applied to an independent photoproduction MC sample, where the scattered positron is not detected in the CAL. Jet quantities were reconstructed using both EFOs with and without correction and the transverse energy, EFt, compared to the hadron-level, EFADas shown in figure 3. Using calorimeter cells as shown in

-0.4

' ' ' -1

-0.5

'

L---LL..-L

0

1

0.5

I , .

U A 1

1

I5

2

9

Figure 3. Fractional difference between hadron-level jet ET and that reconstructed with (a) corrected EFOs, (b) uncorrected EFOs and (c) calorimeter cells as a function of the transverse energy. T h e shaded band shows the width of the distribution.

figure 3c, the deviation from the true value is 20% which is reduced to 10 - 15% when using EFOs due to the use of tracking information as shown in figure 3b. After correction, as in fig. 3a, the transverse energies are significantly closer to the true values, demonstrating that the energy correction helps to reproduce the true quantities when applied to an independent MC sample. To determine the jet energy scale uncertainty, the difference between data and MC after the application of these corrections was considered; this is discussed in section 3.3.

3.2. Method 2 In the second method jets are reconstructed using calorimeter cells and a correction for energy loss is derived from MC simulation". The reconstructed jet energies are corrected on average to the value of the jets from hadrons as a function of transverse energy and in regions of pseudorapidity. The correction factors are applied to both data and MC events. After this procedure

771 ZEUS Data 7p 42.7 pb-’

~z v

~

O

3000 O

o

~

~

3000

2000 2000 1000 0

0.5

0

L

V

v

1.5

1

0

2

0.5

1

1.5

r

4

-0.1 -F -I

2

, “ { , , u ,

-0.5

0

, , I

0.5

!

* ‘ i

1

1.5

2

2.5

qW

Figure 4. Comparison of (a) TTRACKS and (b) TDIJET for data (points) and MC simulation (histogram) and (c) the difference between data and MC as a function of pseudorapidity.

the calorimetric jets in the data and MC simulation are compared by utilising tracking information in a cone around the jets. The ratio TTRACKS of the jet transverse energy, E F , and transverse energy of tracks in a cone around the jet axis is shown for data and MC in figure 4a. This quantity can only be calculated within a certain angular region corresponding to good acceptance for the central tracking chamber. For a jet outside this region, the ratio, TDIJET, of its transverse energy to that of a central jet was calculated; this is shown in figure 4b. The mean value in data and MC for these ratios was found in different regions of pseudorapidity of the jet, $et; the difference between data and MC is shown in figure 4c. The MC agrees with the data to within f2%; this deviation is then used as a further correction. This procedure relies on the accurate simulation of charged tracks and so is less reliant on simulating the energy loss of neutral particles in inactive material. This jet-correction procedure was applied to an independent sample of neutral current DIS events. As for method 1, the ratio of the transverse momentum of the positron and hadronic jet was calculated and the difference between data and MC determined.

772

3.3. Jet energy scale uncertainty The jet energy scale uncertainty is shown in figure 5 as a function of pseudorapidity (for method 1) and transverse energy (for method 2). The difference between data and MC is within *l%.

I

-0.02

-0.03 -1 -0.5 0

-0.1

0.5

1

1.5

2

2.5

L

A

20

.

y

30

d

40

.

d ~ , , , L ~ 50 60 70 80 I

~ 90 i

<* (GeV)

Tid

Figure 5 . Jet energy scale uncertainty as a function of (a) qJet and (b) E F

4. Conclusions

Two independent methods have been developed for correcting jet energies for energy loss in inactive material in the detector. Both methods give an improved reconstruction of the hadronic final state and understanding of the jet energy scale. The uncertainty of the jet energy scale for EFt > 10 GeV is &l%. This leads to uncertainties in measured cross sections of f 5 % , significantly smaller than current theoretical uncertainties. N

References 1. ZEUS Collab., J. Breiwteg et al., Eur. Phys. J. C11,35 (1999); ZEUS Collab., J. Breiwteg et al., Eur. Phys. J . C18,625 (2001).

2. A. Lopez-Duran-Viani, Ph.D. thesis, Humboldt-Universitat, Berlin, DESYTHESIS-2001-056 3. S. Bentvelsen et al., Proc. Workshop on physics at HERA vol. 1 (1992) 23. 4. G. M. Briskin, Ph.D. thesis, Tel Aviv University, DESY-THESIS-1998-036. 5. J. H. Vossebeld, Ph.D. Thesis, University of Amsterdam, (1999) (unpublished). 6 . A. H. Ochs, Ph.D. Thesis, McGill University, Montreal, (2001) (in preparation). 7. ZEUS Collab., S. Chekanov et al., Euro. Phys. J. C23,4, 615 (2002). 8. G. Marchesini et al., Comp. Phys. Commun. 67,465 (1992). 9. L. Lonnblad, Comp. Phys. Commun. 71,15 (1992). 10. ZEUS Collab., S. Chekanov et al., Phys. Lett. B531,9 (2002).

J E T RECONSTRUCTION AT CDF

M. TONNESMANN Max-Planck-Institut fur Physik, Fohringer Ring 6, 80805 Miinchen, Germany E-mail: matthiasOmppmu.mpg.de

Jet clustering algorithms which have been developed for the analysis of Run I1 data are presented and compared. A shortcoming of the new cone algorithm is presented and discussed in the framework of an analytic model.

1. Introduction

Run I1 of the Tevatron collider started in 2001. The present goal is to collect data corresponding to an integrated luminosity of 2 f t - l till 2004 at an = 1.96TeV. This will allow to carry out increased center-of-mass energy a rich program of physics analyses, which includes precision measurements of the fundamental input ingredients of QCD, namely the parton distribution functions and the value of the strong coupling constant as over a wide range of energies.

2. Old and new jet algorithms

An important facet of preparations for Run I1 at the Tevatron has been the study of ways in which to improve jet algorithms'. Historically cone jet finders have been the jet algorithm of choice for hadron collider experiments. The algorithms employed in the analyses of Run I data were based on the Snowmass accord2. However, the incomplete specification of the Snowmass algorithm and the necessity to save computing time led to different jet clustering algorithms used in the experiments CDF and DO. These algorithms exhibit several shortcomings: They had to be modeled in perturbative calculations, and, even worse, as higher order calculations become available, they develop a marked sensitivity to soft radiation (infrared and collinear safety). I t has been proposed' to use improved cone algorithms and k~ clustering algorithms in Run 11. This report contains an overview of the jet algorithms that will mainly be used in analyses of Run I1 data taken with the CDF experiment.

773

774

2.1. JetClu

The JetClu algorithm is the cone jet finder which has been used by CDF for the analysis of Run I data. It consists of three steps. In the first step preclusters are built from adjacent “seed towers” (calorimeter towers with ET > 1GeV). The size of these preclusters is limited to 2Rcone x 2R,,,, in the r] - 4 plane, where r] and 4 denote the pseudorapidity and the azimuthal angle and R,,,, is the parameter of the jet algorithm which controls the size of the jets. After that for each precluster a cone is defined by all seed towers inside the precluster and all towers with AR = ~ ( A V (A4 ) ~ )2 < R,,,, with respect to the highest ET tower. The centroids of the cones are calculated. The identification of the members of the cones and the calculation of their centroids is repeated until the old centroids (the cone axes) agree with the new ones. Note that every seed tower of a precluster is kept in its cone even if the angular distance to the cone center exceeds Rconedue to the migration of the centroid during the iteration. This (unusual) prescription is called “ratcheting” and is unique to the JetClu algorithm. In the last step overlapping stable cones have to be treated because each calorimeter tower may only belong to one jet. In the JetClu algorithm a pair of overlapping cones is merged if more than 75% of the transverse energy of one of the cones is shared by the other one. Otherwise they are separated using an iterative algorithm. The towers are redistributed to the cone whose centroid is closer, and the centroids are recalculated until a stable configuration is reached. The JetClu algorithm is neither infrared safe nor collinear safe. Moreover the cone iteration process uses ratcheting, which is difficult to simulate in perturbative calculations.

+

2.2. Midpoint

The Midpoint algorithm has a substantially reduced infrared and collinear sensitivity compared to cone algorithms used in Run I. This was achieved by adding seed locations for trial cones between pairs of stable seed-based cones. The algorithm works as follows: 0

a

Iterate cones starting at each seed tower (no preclustering). No ratcheting is done in the iteration process. For each pair of stable cones whose centers pi and p j lie within an angular distance A R < 2 . R,,,,, iterate a cone starting at the midpoint pi i-p j .

775 0

Split or merge overlapping cones. Cones are merged if they share at least 50 % of the transverse energy of the less energetic cone.

2.3. KtClus Both CDF and DO will employ ICT clustering algorithms3 in analyses of Run I1 data. Their advantage is that by design they are infrared and collinear safe and can be applied in the same manner to calorimeter towers, hadrons and in perturbative calculations.

3. Comparison of cone algorithms

In this section the dependence of physical observables on the choice of the jet algorithm is discussed to get an understanding of the differences between the jet algorithms presented in Section 2. To this end representative jet algorithms are applied to a data set that was generated with the HERWIG 6.1 Monte Carlo program and then run through the CDF detector simulation. It consists of 100,000 QCD 2 + 2 parton reactions inside pp collisions at a center-of-mass energy f i = 2 TeV. The two primary outgoing partons are required to have a minimum p~ of p ~ , =~100 i GeV. ~ The jet algorithms used are the JetClu and the Midpoint algorithm with a cone size R,,,, = 0.7.

0

50

100

150

200

250

300

350 400 E, [GeVl

Figure 1. The difference in ET between matched pairs of JetClu and Midpoint jets. The left hand plot shows the AET distribution for leading jets, the right hand plot shows the mean values ( A E T ) as a function of E p C 1 u .The distributions plotted in grey were calculated using JetClu in its original version, while the black histograms were calculated using JetClu with the ratcheting switched off.

In the shaded histogram of the left hand plot in Figure 1 the difference

776

of the transverse energy ET of the leading JetClu jet and the corresponding Midpoint jet is plotted. The matching of the jet pair requires their angular separation t o meet A R < 0.1. The distribution exhibits a peak at AET = 0 for which the leading JetClu and Midpoint jets are identical. However, a bias towards positive AE, values is also clearly visible. On average, leading JetClu jets have about 1GeV higher ET values than leading Midpoint jets. A similar conclusion can be drawn from the distribution plotted in grey in the right hand histogram in Figure 1. Here the mean difference ( E p c * "EFidPoint) for matched jet pairs is plotted as a function of EPC1". The observed differences are due t o the ratcheting which is used in the iteration process of the JetClu algorithm only. Since calorimeter towers belonging t o the preclusters never leave the cones during the iteration process, the use 0 of ratcheting can lead to stable cones, and hence to jets, which exceed the cone size R,,,,. Switching off the ratcheting in JetClu leads to a much better agreement between the transverse energies not only of matched leading JetClu and Midpoint jets, Figure 2. Event display pictures of a simulated QCD event but also over the whole in the C D F calorimeter. The upper picture shows the jet ET range covered by the configuration as reconstructed with the JetClu algorithm, the lower picture shows the one obtained with the Midpoint data set (black distribu- algorithm (R,,,, = 0.7). T h e black (seed) towers in the tions in Figure 1). lower picture are not included in any jet. Since ratcheting is difficult to simulate in perturbation theory, it has not been implemented in the Midpoint algorithm. However, it ensures that all seed towers in an event are included in stable cones and hence in jets. Figure 2 shows an event in which the lack of ratcheting in the Midpoint algorithm leads to a substantial

777

part of the transverse energy not included in any jet. Cones which are initiated by the seed towers at (77 NN -0.2, $ NN 100') invariably migrate towards the nearby more energetic towers, which in the end form the leading Midpoint jet. The black towers remain unclustered. In the JetClu case these seed towers are not released from the cone and in the end are part of the leading jet. 4. Improving the Midpoint algorithm

To provide insight into the issues raised by Figure 2 a simple, but informative analytic picture will be discussed4. It describes the dynamics of cones under the influence of partons in the iteration process and will serve to illustrate the impact of showering and hadronization on the operation of cone jet algorithms. Consider a distribution of partons with (transverse) energy

(For simplicity, the 4 coordinate is suppressed in this 1-dimensional model.) The total energy contained in a cone C whose center is placed at 77 is given by: i

i€C

Since in the iteration process a cone moves under the influence of forces which are created by the partons and which are linear functions of the parton energies and their distances to the cone axis, a potential can be assigned to a cone with its center at 77: 1 ET,i . ((77 - yi)2 - R 2 ) const. Vc(7)= - . (3)

+

iEC

Stable cones can be found in the minima of Vc(q) and can be calculated by requiring the force FC (77) to vanish:

For a scenario involving two partons of (scaled) energies E T , ~= 1 and E T , ~= 0.7 at 711 = 0 and 772 = 1.0 the functions defined in Eqs. 1-4 are plotted in Figure 3 (bold solid lines). The potential V C ( ~ )exhibits three minima corresponding to the expected positions of the stable cones found by the Snowmass and the Midpoint cone algorithms (R,,,, = 0.7). Since both partons are entirely within the center cone, the overlap fractions are unity, and the usual splitting/merging routine will lead to a single jet containing all of the initial energy. In fact, an additional parameter Rsep is included in the NLO calculations5 such that stable cones containing two partons are not allowed for

778

angular separations AR > Gep.Rcone.Its value was chosen to be &ep = 1.3 to yield reasonable agreement with the Tevatron Run I data. The specific parton configuration in Figure 3 will thus yield two jets in the theoretical calculation, while the Midpoint algorithm will produce only one jet.

1 Energy dlstribution I P

3.5

a = 0.25

0 -1.5 -1 -0.5

I

0

1

1.5

2

2.5 tl

I

Potential

-1.5 -1 -0.5

0.5

0

-7.5 I 4.5

1

1.5

2

2.5 tl

1

1.5

2

2.5

11

I

Force

0.5

0.5

0

-1.5 -1 -0.5

0

0.5

1

1.5

2

2.5

11

Figure 3. A study of the “cone dynamics” in the Midpoint algorithm (R,,,, = 0.7). The four plots show the quantities defined in Eqs. 1-4. The different line styles correspond to different values of the smearing parameter Q. In addition, the dotted lines denote the case u = 0.25 and R,,,, M 0.495. The Vc(9)curves are shifted by (unphysical) constants for better readability.

A major difference between the perturbative and the experimental level is the smearing that results from perturbative showering, non-perturbative hadronization and detector effects. These effects are simulated in the model using gaussian smearing. After replacing the delta functions in Eq. 1 with gaussians of width (T = 0.10, the middle minimum of Vc(v) and thus the

779

middle stable cone has been washed out (thin solid lines in Figure 3) resulting in a two jet configuration. However, no energy is missed by the Midpoint algorithm in the case (T = 0.10. Only with increased smearing, e. g. by setting (T = 0.25, a value which is supported by jet shape measurements, also the second stable cone, corresponding to the second parton, has been washed out (dashed lines in Figure 3). This explains why the Midpoint algorithm fails to include the energetic towers shaded in black in the lower picture of Figure 2 into the neighboring or separate jets. The iteration of any cone containing these towers invariably migrates to the nearby higher ET towers. This behavior can be avoided by diminishing the influence of the energy at the edge of the cones. The simplest fix is to use two values for the cone radius R,,,,, one during the search for the stable cones and the second during the calculation of the jet properties. As an example, a reduced search cone radius of R,,,,/t/Z x 0.495 is used in the dotted lines in Figure 3. The Vc(q)curve indicates that both outer stable cones at the positions of the smeared partons are found leading to a two jet configuration, in agreement with the result of the Snowmass algorithm with Rsep= 1.3. To understand more deeply the behavior of the JetClu algorithm, a similar analysis can be carried out after modifying Eqs. 2-4 to take the ratcheting into account. This analysis suggests that two different stable cones are found, independent of the amount of smearing. The first stable cone is at the position of the more energetic parton, while the second one can be found at the less energetic parton in case of little smearing and in the middle position in case of (T = 0.25. In the latter case the two stable cones will be merged into one jet, as observed in the upper picture of Figure 2. The result of this analysis is that the Midpoint algorithm (and any other cone algorithm based on the Snowmass accord without ratcheting) , applied to experimental data, can fail to reconstruct entire partons if there is a more energetic parton within a distance R,,,, < d < 2.R,,,,. This deficiency can be eliminated by modifying the prescription of the Midpoint algorithm presented in Section 2.2 as follows: 0

0

In the cone iteration process reduce the cone radius, e. g. by a factor of 1/&. After reaching a set of stable cones, recalculate their contents using the nominal cone size R,,,,. Then go to the splitting/merging process.

Figure 4 indicates that the suggested fix indeed improves the behavior of the Midpoint algorithm in the case of Monte Carlo events which were run through

780

the CDF detector simulation. It contains the distributions of the difference of the total transverse energy contained in JetClu jets and in jets reconstructed with the Midpoint (shaded histogram) and the fixed Midpoint algorithm (solid line) respectively. The JetClu algorithm was chosen as a reference because by design it includes all seed towers (ET > 1GeV) in jets. Entries in the tail of the shaded histogram are caused by events in which ; f

the Midpoint algorithm fails to include a large fraction of the transverse energy in the event in jets. The suggested fix is able to reduce the tail considerably by recovering the stable cones missed in the original Midpoint algorithm due to the influence of nearby more energetic clusters of calorimeter towers. Moreover it slightly shifts the whole distribution to the left because of an increased efficiency in finding low ET jets.

Midpoint with fix

YO'

lo' 10'

10

1 -20 -10

o

10

;;EF~?u-,;?,,,d~$,.,

60 70 ) [GeVl

Figure 4. The difference of the total transverse energy contained in JetClu and in Midpoint jets (Rcone = 0.7). The shaded histogram shows the distribution for the original Midpoint algorithm, the solid line histogram shows the one for the fixed Midpoint algorithm with R,,,, % 0.495 in the iteration process. Only jets with ET > GeV were 'Onsidered for the calculation of the total E T . Y

5. Summary and Outlook

As suggested in the R~~ 11 workshopl CDF and D0 are working On the development of common tools for the reconstruction of jets, which will be employed in the analyses of data that both detectors are actively taking. References

1. G. C. Blazey et al., Run 11J e t Physics: Proceedings of the Run 11QCD a n d Weak Boson Physics Workshop, FERMILAB-Conf-00/092-E (May 2000). 2. J. E. Huth et al., in: Research Directions for the Decade: Snowmass 1990, July 1990, edited by E. L. Berger (World Scientific, Singapore, 1992), p. 134. 3. S. D. Ellis and D. E. Soper, Phys. Rev. D48, 3160 (1993); S. Catani, Yu. L. Dokshitzer and B. R. Webber, Phys. Lett. B285, 291 (1992); S. Catani, Yu. L. Dokshitzer, M. H. Seymour and B. R. Webber, Nucl. Phys. B406, 187 (1993). 4. S. D. Ellis, J. Huston and M. Tonnesmann, On Building Better Cone J e t Algorithms, hep-ph/0111434. 5. S. D. Ellis, Z. Kunszt and D. Soper, Phys. Rev. Lett. 69, 3615 (1992); S. D. Ellis, in: Proceedings of the 28th Rencontres de Moriond: QCD and High Energy Hadronic Interactions, March 1993, p. 235; B. Abbott, et al., FERMILAB-Pub97/242-E (September 1997).

SUPPRESSION OF PILE-UP NOISE IN A J E T CONE

ALEXANDRE SAVINE University of Arizona E-mail: savin~physics.arizona.edu Multiple low-pT (min-bias) interactions within a beam crossing at a high luminosity hadronic collider contribute to pile-up noise in the calorimetric measurements of jets. I show how to minimize this noise by taking advantage of correlations in these background events. Substantial reductions are possible

A cross-section for inelastic p p interactions (including diffractive) at LHC energies is estimated as 80mb. During high luminosity operation ( L = 1034cm-2s-1)there will be 20 low - p in ~ every beam crossing. Those interactions create a substantial fluctuating background in the detector subsystems, including calorimeters. Since the beam crossing are separated by only 25ns, min-bias signal from previous crossings also contributes to this background, depending on the signal shaping and data processing algorithms. As any noise does, the pile-up background affects the energy resolution and forces to increase p~ thresholds in the calorimeter. Present research was aimed on reduction of this negative impact. There are few sources of the pile-up fluctuations: (1) Variation of the beam crossing conditions like amount of protons in each bunch, beams relative alignment and focusing variations. (2) Poisson fluctuations of interactions number (crossing to crossing). (3) Physics of each individual proton-proton interaction. (4) Shower development:

(a) Interactions in the beam pipe elements, tracker, calorimeter support/cryostat (b) Shower development in the calorimetric system itself (c) Albedo of the calorimeters

First two factors on this list determine number of interactions and are same for every calorimeter cell. Fluctuations introduced by variation of interactions number behave exactly like a coherent noise, while fluctuations coming from individual interactions are responsible for a stochastic noise. Indeed, low - p~

781

782 N,=

Figure 1.

15

Nn.221

N,=27

Sample min-bias events. E g S in Aq x A# = 0.2 x 0.2 towers

events are independent, but are randomized by same QCD rules. Figure 1 shows three events with different number of interactions. There are numerous peaks from individual showers, but it is also clear that they rest on top of a well-defined 'table', proportional to the number of interactions.

Figure 2. q = -0.9

Correlations between the visible energies in the jet cone at q = 0.9 and q-slice at

For large groups of cells (like jet cone) coherent component of the pile-up increases in proportion with number of cell, while the 'stochastic' component is proportional to square root of this number. Figure 2 demonstrates such a combination of coherent and uncorrelated fluctuations. It also provides a clue for the pile-up reduction algorithm: the calorimeter shall be used an instant

783

'luminosity monitor'. Of course, the regions of interest with their large signal shall be excluded from this the analysis. N=32

*

N=32-N.

3

7

I

0

I

7

-3

Figure 3. Scheme of the Dynamic Pileup Suppression: C, and CfoT' are raw and corrected cone sums, Aj is an average 0.2 x 0.2 tower response at q j . w i j is a set of coefficients minimizing the pile-up RMS in the cone while keeping average at zero.

Figure 3 shows how the optimal weights were determined for pile-up noise reduction. lOOk Lou, - - p ~interactions were generated by PYTHIA 6.136. Simplified detector (beam pipe and bulk lead calorimeter) was used to run GEANT 3.2111. Segmentation - Aq x Ad, = 0.1 x 0.1. Interactions were saved one-byone. To simulate min-bias events, random sets of event were chosen. Number of interactions in each event was determined by Poisson random generator. In each generated event several jet cones were taken at different (random) q and d,. Pile-up from previous beam crossings was not included. Used procedure is linear and as a result - luminosity-independent. Despite (some) amplitudes from previous crossings come with negative amplitudes, they do not break the linearity of described procedure. Compareson of pile-up spectrum before and after suppression is presented at Figure 4. As was expected, pile-up suppression algorithm reduces the Pois> 7GeV (and son assimetry of the spectrum. Rate for events with E$" EFrS < -7GeV) is reduced by 213, rate above 20GeV - by 314. This region is marked at the semi-log plot as 'True Pileup'. At the same time, even the min-bias events may deliver some well-collimated jets. They contribute to the high energy tail: there is no improvement at E$IS > 50GeV. Indeed, this is

784

region where no global coherence exists.

60000 40000

20000

-25

0

25

SO

7.5

0

-25

0

y105

25

SO

7.5

Raw Suppressed j RMS 7.48 5.01 HWHP 7.88 4.65 Peak 1.0 1.61

8 i;10' 10

102

10 1

-25

0

25

.TO

-25

75

0

25

50 Raw

6W00

7s Suppressed

6.51 6.07

4.38 4.04

50

77

40000 20000

-2.5

0

2.5

so

75 E, visible, GcV

O - 3

0

25

E, visible, CeV

Figure 4. EFrs pileup spectrum before (shaded) and after the suppression. Cone A R = 0.5 with added hadronic shower margin was used. Raw spectrum is shifted so the average is zero. Average number of the interactions per event corresponds to L = 1034~m-2s-'.

In addition to the RMS and peak width reduction, Dynamic Pile-up Supression improves the shape of the pile-up noise. Though still not a perfect Gaussian, corrected spectrum is more symmetric. Choice of the Jet Cone size for measurements at the high luminosity is also affected by a pile-up noise. Though it is commonly recognized that A R = 0.7 is much better from the acceptance point of view, much smaller A R = 0.5 is considered maximum acceptable when accelerator is running at L = 1 0 3 4 ~ - 2s- 1 What can the Dynamic Pile-up Supression do here? Since this method is aimed on the coherent component of the pile-up, it works better for larger cones. Indeed, it turns out that (Figure 5) the Dynamic Supression reduces the pile-up in the AR = 0.7 Cone down to the A R = 0.5 scale.

785

4

yi 4-

f 104 w

m

30000

10.3

20000

10

lDO#

1

0.97

1.0

eok

102

n

Y

-2.5

0

25

50 75 E, visible, GeV

-25

0

2.5

50

0

25

50

7.5 visible, GeV

-25

0

25

50

75 E, visible, GeV

yi

wB 1 0 4 I0 I0 10

I -25

7.5

ET visible. GeV Row 0.5 Suppressed 0.7

4B 104

6.09 5.38 1.19

RMS 6.51 HWHP 6.87 Peak 1.0

10 3

10 10

I

0

-25

25

50 75 FT visible, GeV

-25

0

25

50

75 E, visible, GeV

Figure 5. E$IS pileup spectrum for Jet Cone A R = 0.5 without Dynamic Pileup Supression (shaded) is compared with Jet Cone O R = 0.7 with Dynamic Pileup Supression applied. Raw spectrum is shifted so the average is zero. Average number of the interactions per event corresponds to L = 1 0 3 4 ~ m - 2 s - 1 .

Conclusions (1) Pile-Up noise may be reduced on the event-by-event basis using calorimeter as an instant luminosity monitor. (2) Suppression allows to:

(a) Increase the Jet Cone for better energy measurements (b) Lower the ET threshold for higher efficiensy (Taggin jet in the FCal) (3) Dynamic Supression requires very small volume of data:

(a) 50 digitized sums (Aq = 0.2 rings) (b) 50 integer tower counts (to exclude region of interest)

(4) Described procedure is independent of any other noise reduction technics

D0’S RECENT RESULTS AND EXPERIENCES WITH THE KT AND CONE JET ALGORITHMS

J. KRANE Iowa State University, Fennilab MS 357, P.O. Box 500, Batavia, IL 60510 E-mail: [email protected]

(For the D 0 Collaboration)

This paper presents recent results, current problems, and possible solutions for analyses that use both the kT and cone jet algorithms. Hadronization of final-state partons can improve the level of agreement between NLO QCD predictions and the inclusive jet cross section observed using k~ jets. The dijet transverse thrust , a jet measurement in two kinematic regions analysis, which also uses k ~ provides where NLO has little predictive power. The results suggest both resummation and higher-order predictions can improve the theory in their respective regions. Finally, the cone jet algorithm, including the recent “midpoint” improvement, contains an inherent weakness, as identified by CDF (see Matthais Toennesmann’s presentation in this session). The D 0 Collaboration is exploring the suggested modification to this algorithm, in the hope that both experiments will use a common algorithm in Run I1 of the Tevatron.

The phenomenology of jets in QCD matured alongside experimental results obtained with the cone jet algorithm. The UA2 experiment used fixed cones of dimensionless size R = 1.3 (in azimuth and pseudorapidity), to compare to leading-order QCD predictions. The large cone size tended to collapse all events into a dijet topology. Years later, the CDF and DO experiments used a version of the cone algorithm (with R = 0.7 and a merge/split function) to compare jet data to NLO predictions. This second generation of jet results enjoyed a factor of two or more improvement in precision. Toward the end of the Tevatron’s Run I, DO implemented a recombinant jet algorithm, the Ellis and Soper’ k~ algorithm, and revisited the inclusive jet cross section measurements previously done with cone. The jet data have been incorporated into the standard parton distribution functions; this article attempts to summarize the open issues facing jet physics today, and how they impact D8’s plans in Run 11. The DO calorimeter2 provides the majority of the jet information for the analyses. The calorimeter is hermetic, consists of more than seven nuclear

786

787

interaction lengths, has fine segmentation (0.1 x 0.1 in q - (6) and covers the pseudorapidity region from -4.2 to 4.2. The active medium is liquid argon, separated by absorbers of spent uranium in most places and isolated regions of copper or steel. DO defines jets with two well-known algorithms. The cone algorithm employs a fixed-radius q - C$contour to select calorimeter cells for energy summation. The energy-weighted center provides a new jet centroid. The jet is redrawn, the contained energy is recomputed, and a new centroid is found, etc. until the jet has stable energy. After this iteration, jets that are very close to one another are either merged or split depending on their fraction of shared energy. The k~ algorithm1 compares the pT-weighted distance of a l o w - p ~ energy cluster to the next largest cluster with a resolution parameter D.If the comparison of clusters i and j ,

is less than p g i / D 2 of the smaller cluster, then the two clusters are combined into a single object with that pT-weighted location. Otherwise, the small cluster is saved to a list of stable jets. The procedure continues until all clusters have been combined or saved. This algorithm exhibits infrared and collinear safety. A sophisticated jet energy scale algorithm3 corrects jet energies for calorimeter response, noise, underlying event, and other effects. By design, the energy scale corrects jet energies back to the “particle level”, such that all detector effects are removed but natural hadronization effects are preserved. D 0 conducted three analyses with the k~ algorithm. First, the multiplicity of subclusters within jets4 revealed a difference between jets of like energy but produced in collisions with different center-of-mass energy. This result implies a difference in the jets from quarks as compared to gluons. The other two analyses are described in the following sections. The last section of this article describes recent work on the cone algorithm. 1. Inclusive Jet Cross Section with ICT

The information from this analysis requires accurate and effective statistical comparisons. The DO collaboration takes great care to build a full covariance matrix that appropriately manages the correlations in energy of each individual uncertainty. The availability of the covariance matrix allows DO data to determine the parton distribution functions5@(PDFs) at high-x. Whereas the earlier cone results display clear agreement with available NLO predictions (even before integration into the PDFs), the k~ result (Figure 1) differs from

788

100

0 Figure 1.

200 300 400 500 Jet Transverse Momentum (GeV)

D0's inclusive jet cross section with the kT algorithm and 171 < 0.5.

c 1

-

0.8

I1 0.6

n

0.4

0.2 0

-0.2 -0.4

-0.6 -0.8

-1

1

Cone (R-0.7) PRL

~ " ~ " " " ' ~ " ~ " " ~ " " ' ~ " '1 ' ~' J 50

100

150

200

250

300

350

L 400

L u d 450 500

Jet Transverse Energy (GeV) Figure 2. The fractional difference between DO data and NLO QCD predictions. The cone result is compared to the prediction for cone and the kT result is compared to the prediction for kT.

789

the prediction by roughly two sigma. Figure 2 compares cone and kT cross section results to their respective predictions. Because the uncertainties, which largely result from the energy scale correction, are highly-correlated in ET,the normalization differences are not remarkable. Notice instead the shape difference at low ET in the kT cross section; this departure in shape is not consistent with the uncertainties in the data. If the first four data points are ignored, the k~ predictions are consistent with the observed cross section at the 77% probability level (CTEQ4HJ). We conclude that the marginal agreement of the full 24 point comparison results entirely from the effect of the first four points. The actual x2 values (24 degrees of freedom) resulting from a comparison of D0’s inclusive jet cross section with the k~ algorithm and 1771 < 0.5 with NLO QCD predictions from JETRAD appear in Table 1. Table 1. Example of the x2 values. Last row is JETRAD plus a hadronization correction derived with HERWIG.

I x2 I

I

CTEQ4M CTEQ4HJ CTEQ4HJ+Had.

I

31.2 27.2 24.3

I

1 I

Probability (%) 15 29 44

I

A little investigation reveals that the cross section for k~ differs from the cone result because the individual kT jets in each event possess more energy than the matching cone jet. The jet-by-jet energy difference (Figure 3) accounts for the entire difference between the cross sections, as demonstrated by artificially removing the extra energy jet-by-jet. Particle level Monte Carlo simulations7 indicate that the hadronization process scatters some of the original parton energy outside a 0.7 radius cone. The opposite occurs with the kT algorithm, where the found jet contains more energy than the original parton. The shape and general direction of the simulated effect are suggestive, but the magnitude ( k energy ~ - cone energy) is only half the size of that observed in data. Because the jet phenomenology of QCD was largely developed with cone algorithms, it seems possible that unexpected effects unique to the cone aigorithm have been inadvertently incorporated in the PDFs. In that sense, agreement between theory and cone results is not surprising, nor is the marginal agreement between the predictions and the relatively new k~ algorithm’s results. The DO results indicate that a previously ignored missing component in the prediction has now become important. Hadronization effects appear to represent at least some of this missing piece, but in their current form they are not enough to remove the observed discrepancy.

790

0

& -l 50 -2

100

150

200

250

300

PT(kT)GeV Figure 3. Difference in transverse momentum for a jet found with the ICT algorithm and the same jet for cone. Jets were “matched” in calorimeter position for this comparison.

2. D i e t Transverse Thrust

This analysis employs a novel event shape variable to probe kinematic regions that NLO predictions fail to describe. DO defines the dijet transverse thrust as

where the summation of transverse momentum, projected onto vector ii, occurs over i jets, with ii selected such that TT takes its maximum value. To avoid distortions caused by calorimeter noise, the summation involves only the two leading jets in the event. In this definition, an event with the leading jets back-to-back in azimuth results in TT = 1.0, and a three-jet final state can take any value in the range 11. With more final states, the minimum angle of the leading two jets can be less than 120 degrees, and even lower values of thrust are possible. Although the observed thrust distribution is well-described by NLO QCD predictions in the range [0.01,0.1], values above and below this range result in large departures of theory from the observed distributions. Figures 4 and 5 display the thrust distributions of two jet samples with different total energya. The right-hand side of the distributions represents the kinematic limits of NLO

[Z,

aThis quantity, HT3, characterizes the hardness of the event with the energy sum of the leading three jets. Again, the quantity considers only the leading few jets to avoid distortions caused by noise.

791

calculations; the predictions fail as they approach this limit. The left-hand side of the distributions shows a region where resurnmation might have a large effect in the predictions. Both of these effects are larger than the initial estimate of the uncertainties. 430 < H13 <700

160 C HT3 <260

-

10

15

-+

Jetrod --t

--ct

-t--

430

I80 < HT3 <2M1

< HT3 <7W

D h

7 5 v c

10-

Figure 4. D0’s dijet transverse thrust, as computed with the kT algorithm for events with low total jet energy (in GeV). Solid line indicates NLO QCD predictions.

Figure 5. D0’s dijet transverse thrust, as computed with the kT algorithm for events with high total jet energy (in GeV). Solid line indicates NLO QCD predictions, which exhibit somewhat smaller deviations as compared to the low total energy bin (Figure 4).

3. Modification to the Cone Algorithm

Whereas the AT algorithm is recombinant, the cone algorithm is iterative, and therein lies its weakness. In some events, a small energy cluster can “hide” near a large cluster of energy, with the cone algorithm consistently overlooking the small cluster in favor of the large one. (Cf. Matthais Toennesmann’s talk in this same session.) The DO data do not exhibit a strong tendency toward this effect, perhaps because the calorimeter noise contributes a sufficiently varied baseline or because the calorimeter is finely segmented. Searches for clear examples of unfound jets in DO data continue. CDF has proposed a solution in the form of smaller search cones during iteration, with expansion to full size only after a stable centroid is found. This simple scheme has the advantage of seamless incorporation to the existing midpoint algorithm, but has the disadvantage of adding another parameter to the algorithm. DO is currently testing the performance of the proposed

792

algorithm. If DO does not discover some unforseen disadvantage, they will most likely adopt the change whether or not significant numbers of lost jets manifest in the search. 4. Conclusions D0’s results from Run I hold a worthy place in the history of jet physics. Cone algorithm results have unprecedented ability to determine PDFs at high-s, largely due to the successful covariance matrix technique to quantify uncertainties. The k~ algorithm is less well-understood in terms of the phenomenology of data and theory agreement, and thus we should expect to learn more basic phenomenology before we can extract the same quanitites as with the cone algorithm. In fact, the k~ inclusive jet cross section indicates that a missing quanitity, perhaps hadronization effects, must bridge a gap between data and the NLO QCD predictions by providing extra energy to each jet. One possible implication is that the excellent agreement between QCD and the cone cross section result is partially circular and hides a similar effect. The dijet thrust analysis probes kinematics outside the region of validity of NLO predictions. At low values of thrust, higher perturbative orders in QCD predictions should improve their accuracy. At high values of thrust, other techniques will be required, such as resummation. Finally, the cone jet algorithm occasionally fails to cluster significant energy deposits into jets. This effect can occur during the iteration of cone position to maximize jet energy. DO is now testing a modification to the jet algorithm proposed by CDF that purports to alleviate this problem.

References 1. 2. 3. 4. 5. 6. 7.

S.D. Ellis and D.E. Soper, Phys. Rev. D 48 3160 (1993). B. Abbott et al., Nucl. Instrum. Methods Phys. Res. A338, 185 (1994). B. Abbott et al., Nucl. Instrum. Methods Phys. Res. A424, 352 (1999). B. Abbott et al., Phys. Rev. D 65, 052008 (2002). H.L. Lai et al., Phys. Rev. D 55, 1280 (1997). A.D. Martin et al., Eur. Phys. J. C 4, 463 (1998). G . Marchesini et al., Computer Phys. Commun. 67, 465 (1992).

JET MEASUREMENT IN OPAL

NAKAMURA ISAMU University of Pennsylvania, Philadelphia, 19025, Pennsylvania, USA E-mail: [email protected]

The Jet Measurement in OPAL is summerised. In OPAL, jet is reconstructed from the energy flow “object” which are the output of the energy flow algorithm. The algorithm tries to subtract track momentum from calorimeter energy and tries to do software compensation. For better measurement of particle mass such as Zo W or Ho, jet re-association technique is used. The detail of energy flow algorithm and jet re-association is described.

1. The OPAL Detector

The OPAL detector is a multipurpose apparatus having nearly complete solid angle coverage. The detailed description described of the OPAL detector can be found in elsewhere’. The central detector consists of a silicon strip detector and tracking chambers, providing charged particle tracking for over 96% of the full solid angle, inside a uniform solenoidal magnetic field of 0.435T. The energies of particles are measured by a lead-glass electromagnetic calorimeter located outside the magnet coil and the iron-streamer tube hadron calorimeter outside the the lead-glass calorimeter. The characteristics of the OPAL detector which are related to jet reconstructions are listed in Table 1. Table 1. The characteristics of the OPAL detector.

material resolution

tracking Silicon + j e t chamber 1.25 x lOP3pt

-

electromagnetic lead-glass 10% 1%

N

/a+

hadronic iron/streamer tube N

120%/a

An important feature for the jet energy reconstruction is that the lead-glass calorimeter is not compensated at all, hence has a large e/n ratio. So one need to apply software compensation in the offline analyses, which is described in the following section.

793

794 2. Energy Flow Algorithm

The idea of the algorithm is t o remove double counting of energy measurement in calorimetry and momentum measurement in tracking device by subtracting track momentum from calorimeter cluster. To get best energy flow measurement, not only energy but also direction measurement, the algorithm was designed to maximise the usage of tracking information. Because the tracking system has better energy/direction resolution calorimetry. The problem in the OPAL calorimeter is that the energy responses t o electron and hadron are very different. To solve the situation, the energy is re-calculated by the following function

E = Pem ( E e m )

’

Eem

-k Phad ( E h a d ) ‘ Ehad

where P is the compensation factor as a function of measured raw energy, E . The factor is tuned with the single particle (e, T) Monte Carlo simulation. The functions are shown in Figure 1. h 0

s 2 ccz 1.75 0 3 1.5 g1.25

0.8

a 0.75

0.4 -.........................................

1

. I

c u w

0.5 0.25 0

0.6

0.2 10 20 30 ECAL cluster raw energy [GeV]

0

-

-. ..........................................................

I , , ,

4

I , ,

0

,

-

Figure 1. Compensation functions for EM(Left) and Hadron(Right) calorimeters.

Track is associated to cluster if the track is extrapolated inside the cluster or the distance between track and cluster boundary is smaller than specific threshold. The thresholds are determined to have the same fraction(f) of tracks are associated to the clusters. The track is associated by several different classes of “goodness” of matching as functions of the fraction, f . Since the expected energy deposition of electron or muon is well known, the expected energy is subtracted from cluster for all identified electrons or muons in early stage of algorithm. The isolated track and cluster pair which has E/p compatible t o electron

795

are considered t o be either electron, conversion photon, or hadrons which deposit most of the energy in electromagnetic calorimeter. Track momentum is subtracted from cluster. After this step calorimeter energies are re-calculated by compensation function described above. The track momenta are subtracted one-by-one from hadron calorimeter to electromagnetic, from lowest momentum track to highest, and from highest matching class to lowest. Since lower momentum track has smaller matching region, this ordering gives best track-cluster association. After the subtraction, remaining hadron and electromagnetic clusters are treated as neutral hadron and photon, respectively. Energies are recalculated accordingly. Figure 2 is the total visible energy distribution for Zo event taken at fi = mp. The energy resolution is 9.0% and 8.6% for all detector region and central([cos&hrustl <0.7), respectively. This results are about 10% better than that obtained by older algorithm without energy compensation.

r.m.s. 12.8GeV 0

lo00

-

9.OGeV

r.m.s. 12.8GeV IS

85GeV

-

500-

E,,(GeV) Figure 2. Total energy for 2' event at (Monte Carlo).

fi = mZo.

Points(Histogram) are OPAL data

Figure 3 is distribution of differences of angles for two back-to-back jets in Zo event. Angular resolution in both polar and azimuthal angle are estimated to be about 30 mrad.

796

r.m. 0.llOrad a O.0333rad

-I

6.75

-05

-0.25

r.ms. 0.0693rad

a

0

0.23

0.5

0.75

I

-1

0.034Snd

6.8 4.6

4.4

43

0

0.2

Od

A$ @ad)

Figure 3.

0.6

0.8

I

A8 (rad)

Compensation functions for EM and Hadron calorimeters.

3. Jet Reconstruction

In OPAL variety of jet reconstruction algorithms are used in various analyses for their purpose. Among them the Durham algorithm2 is the most widely used especially in particle searches. In new particle search, such as the SM Higgs boson search, mass resolution plays crucial role. The particle mass is directly reconstructed from four momenta of two measured jets. For the better mass resolution, it is important to reduce mis-association of tracks/clusters between different particles. The mis-association also has effect to b-tagging performance, because sometimes significant track is lost due to the mis-association. However Durham algorithm tries to combine particles from smaller momentum, the final jet depend on the situation of the small energy track or clusters. To reduce mis-association, particle-jet re-association is performed. Particles are released after the first path of jet reconstruction. Then they are re-associated to the jet with smallest parameter defined as

where Ejet,E are the energy of jet and particle and 0 is the angle between jet and particle. Four momentum of jet is then recalculated. The correct association rate is illustrated in Figure 4. Figure 4 shows the improvement of W mass resolution and Higgs search sensitivity. The improvement of Higgs search sensitivity come both from improved mass resolution and b-tagging performance.

797

moa O

o

0.1

0.2

03

0.4

05

0.6

0.7

oa

0.9

I

Number of particle fraction

Energy fmction

Figure 4. (Top) Fraction of number of correctly associated particle. (Bottom) Fraction of correctly associated energy.

Figure 5. (Left) mass resolution for reconstructed W. (Right) Higgs search sensitivity. Number of SM background events per 100 pb-' as a function of selection efficiency.

4. Summary

The energy flow algorithm and jet-particle re-association scheme are summerised. The energy flow algorithm achieved the total visible energy resolution for Zo event of 8.6 GeV. The jet-association scheme has visible impact on jet-mass resolution. References 1. The OPAL Collaboration, K. Ahmet el al., Nucl. Instr. Meth. A305 (1991) 275; S. Anderson el al., Nucl. Instr. Meth. A403 (1998) 326; B.E. Anderson el al., IEEE Trans. Nucl. Sci. 41 (1994) 845. 2. S. Catani el al., Phys. Lett. B269 (1991) 432.

DEVELOPMENTS ON JET RECONSTRUCTION BY DELPHI

A. KIISKINEN Helsinki Institute of Physics, Helsinki, Finland E-mail:ari.kiiskinenOcern.ch The most relevant techniques used by DELPHI to identify jets in multihadronic final states are reviewed. The performance of jet reconstruction algorithms is analysed together with the additional use of energy and momentum conservation in order to allow for a precise reconstruction of the event kinematics. Also jet flavour tagging methods are summarised. Applications in some analyses like searches for new particles such as Higgs bosons, W mass physics and QCD studies are presented.

1. Introduction 1.1. LEP physics with jets

In most of the physics processes studied at LEP, the final states consist of hadronic jets. These include Zo measurements, QCD studies, b physics, WW and ZZ physics and various searches for new particles such as the Higgs boson(s) or Supersymmetric particles. This presentation summarises some of the basic methods of jet reconstruction and discusses new developments in jet measurement tools used in the physics analyses in the DELPHI collaboration. In addition, a simulation study of heavy charged Higgs boson mass reconstruction at a high energy linear collider is shown in order to demonstrate how the tools developed and used at LEP can be applied for higher energies at the future colliders. 2. Jet reconstruction

When studying the properties of heavy particles such as Zo, W or Higgs bosons in their hadronic decays, jets are used as an approximate representation of the high energy partons, which are the initial decay products of the primary heavy bosons. In order to obtain maximal information about the initial partons, the jets must be reconstructed with high efficiency and their properties should coincide with the parton properties as precisely as possible. This is done using the following procedure (using the following tools): 0

Measurement of the energy and momentum of detectable particles

798

799

(tracking & calorimetry) Combination of tracking and calorimetry information into energy flow of the event, i.e. sharing calorimetric energy between charged and neutral particles Determination of the energies and momenta of the initial partons (jet clustering, kinematic fits) Particle identification (tracking, calorimetry, muon identification, dE/dx, Cherenkov detectors) Identification of the flavour of the initial parton (jet quark flavour tagging, gluon identification) Various kinds of jet clustering algorithms are available and in use in LEP analyses. Some of them differ from each other fundamentally in the basic concept and some are just different versions of the same basic method. The list of algorithms commonly used in DELPHI includes e.g.: Luclus, Jade, Durham, Cambridge and Diclus. They all have their strengths and weaknesses and are therefore optimal for different kinds of analysesa and in many analyses the results obtained using one algorithm are cross-checked using other algorithms in order t o estimate the systematic effects related to the clustering algorithm. One example of LEP analyses with jets, is the extraction of the energy defrom the four-jet rate of hadronic pendence of the strong coupling constant, as, events' as shown in Figure 1. After the jets have been reconstructed, additional information, such as jet flavour tags, is often assigned to the jets and the precision of the estimates of the jet energies and momenta can be improved using energy rescaling methods or kinematic fits based on energy-momentum conservation. A good reference process showing the precision of DELPHI jet measurements is the decay of the Zo boson into a qQ pair at LEP12: 0

Measured di-jet collinearity 1.3" (jet direction resolution about 0.9.) Zo total energy measurement precision about f8 GeV in barrel and f10.5 GeV in forward regions (order of 10% energy resolution)

2.1. Jet flavour tagging Jet flavour tagging is mainly based on two methods, secondary vertex reconstruction and particle identification. The former is based on the fact that hadrons consisting of heavy quarks fly measurable distances before they decay. This method relies on the high impact parameter measurement precision of &Agood review of these jet algorithms can be found for example in the following publication: S. Moretti, L. Lonnblad and T.Sjostrand, JHEP 9808,001 (1998)

800

-

200 GeV

$

DELPHI prehinary

0.13

I /

0.12

10

0.11

I0

0.I

10. 104

10.'

JADE JADE

Yt,

.

10-1

O(*

0.09

(StaiSliCal

xp: sxp. opt.

crmrs)

y,=o.wo

0.08

,',

100

120

140

160

Durham 0.13

loJ

DURHAM

,-'

10

1

, 180

,

t,

,

, ,

m EpFeV

DELPHI preliminary

0.12

Y, 0.11

0.I

0.09

0.08

Figure 1. Plots on the left: QCD jet rates (R) at 200 GeV as a function of ycut compared to the prediction of JETSET 7.4 PS. Plots on the right: Energy dependence of a, as obtained from jet rates. The errors shown are statistical only. The band shows the QCD expectation, when extrapolating the world average to other energies.

the silicon vertex detectors. The latter approach compares the rates of identified particles inside jets to the known expectations of the particle contents in hadronisation processes of different initial quark flavours. Particle identification uses combined information from all detector types: trackers, calorimeters, muon taggers, Cherenkov detectors etc. Secondary vertex reconstruction methods enable efficient identification of b jets with high purity, which is essential for b physics studies and for Higgs searches. Using a combination of all flavour tagging methods, c , s and gluon jets can also be identified with efficienciesthat are useful for some analyses. 2 . 2 . Inter-jet phenomena One type of analyses using flavour tagged jets are the QCD studies of the difference between gluon and quark jets. A non-biased sample of gluon and quark jets can be found using three-jet bbg events, where the quark jets have been identified using b tagging. Events with identified quark and gluon jets have been used for studying QCD string effects. In the inter-jet region the 'jet=parton' approximation is

801

not valid as the particles do not originate from independent partons but from the colour fields between parton pairs. Therefore a conventional jet clustering approach is not sufficient for these analyses but one needs to study also the distributions of individual particles.

'y

Charged particle flow

Figure 2.

Particle multiplicities as a function of a normalised inter-jet angle.

Figure 2 shows the particle multiplicity distributions as a function of the angle between the jets3. Most of the particles are in the hard cores of the jets but many particles also lie in the inter-jet regions. The multiplicities of the inter-jet regions of quark-quark pairs have been found to be lower than the multiplicities in the quark-gluon pairs as predicted by QCD. The inter-jet multiplicities in quark-quark pairs qqg events have been also compared to the multiplicities in quark-quark pairs in qq-y events with similar topology3. This ratio has been measured to be: R, = Nqqg/Nqq,= 0.56 f O.OG(stat)f 0.02(acc. p r . ) f O.Ol(jetaZg.) which is in good agreement with the asymptotic perturbative QCD expectation: 0.65n2- 1 R,(QCD) = Nqqg/Nqg, N n z ~ l N 0.60

+

802

3. W and Higgs boson physics 3.1. Heavy boson mass reconstruction

LEP2 has provided a large sample of W bosons for measurement of its properties and the energy of the collider has given a chance to look for the Standard Model Higgs boson in the mass region predicted by the electro-weak fits. The masses of the W and possible Higgs bosons are reconstructed by measuring the mass of the di-jet pair, which originates from the decay of the initial boson. In these analyses kinematic fits are used to improve the mass resolution. They compensate detector inefficiencies and indetectable particles, such as neutrinos. Following constraints are most commonly used: full energy momentum conservation with precisely known centre-of-mass energy, equal masses of bosons (e.g. WW, HA with some model parameters and H+H-) and known masses of particles, e.g. Zo mass in the Standard Model Higgs search. 3.2. Flavour tagging in Higgs searches and W measurements

Figure 3 shows the power of b tagging in separation of b jets and light quark jets in the DELPHI Standard Model Higgs boson search. No signal has been found and a 95% CL lower mass limit €or the Standard Model Higgs boson is set at 114.3 GeV/c2 (with expected mass limit of 113.5 G ~ V / C .~ ) ~

'-4

-2

0

2

4 6 B-tagging variable

Figure 3. Separation power of b tagging for light quarks (dark histogram) and b quarks (open histogram).

Identification efficiencies for c and s quarks are much lower than for b quarks but c and s tagging is used successfully for WW background rejection in DELPHI e+e- + H+H- + cSEs analysis5 and a 95% CL lower mass limit for H* is has been set at 73.8 GeV/c2. c and s tagging also enables a direct V,, measurement in flavour tagged W decays. The DELPHI measurement is: lVcsl = 0 . 9 1 + ~ : ~ ~ ( sf t a0t .)0 5 ( ~ y s t .) ~

803 3.3. Mixed Lorena boosted #"s

i n W mass measurement

A new technique, mixed Lorenz boosted Zo (MLBZ) method, has been developed for estimation the systematic effects of hadronisation and particle detection and jet reconstruction in the W mass measurement'. Conventional methods compare real Zo decay events, collected during detector calibration runs at the Zo resonance energy, to simulated Zo events. Uncertainties in the simulation of jet measurements are estimated in these comparisons and are then propagated through W analysis chain. The effects of these different uncertainties for the W mass measurement are estimated separately and finally combined to obtain the total uncertainty of the mass measurements. MLBZ method uses a different approach. Artificial 'pseudo WW events' are first created by overlapping two boosted ZO's. Then a W mass measurement is performed on these 'pseudo WW events'. The same procedure is repeated both for real and simulated Zo's and the difference between the mass estimators can be used as an uncertainty estimate. The strength of this method is that it takes automatically into account correlations between different effects. However, some interference effects between the two W's, such as the Bose-Einstein correlation and colour reconnection, can not be evaluated using this method. MLBZ results show that the measured W mass values in data and simulations agree within a few MeV/c2, which means that conventional error esti1998 ZO run final processing

c

6-100

+ J n

" " ' " " " ' ' ~ ' " ' " * ' ~ ' ~ ' " ' ~ ' " ' ~ '

0

5

10

15

20

25

30

35

HOUR IN SECOND ZoRUN Figure 4. Stability of MLBZ mass.

804

mates have been conservative. The stability of the DELPHI mass reconstruction is demonstrated in Figure 4, which shows the reconstructed MLBZ mass in separate data sets taking during LEP2 calibration runs. 4. Extensions to linear collider physics

At the high centre-of-mass energies of planned future linear e+e- colliders, many processes will have higher jet multiplicities than the processes at LEP energies. For example, the following Standard Model and non-Standard Model cascade decays lead into final states with six to eight jets (with additional gluon jets):

0

e+ee+e-

0

e+e-

0

+ t f - i 6 jets + tfHo -+ tfbb + 8 jets -+ H+H- -+ tbEb + 8 jets

Efficient b-tagging is needed for rejection of multi-jet background processes. Effective b-tagging can also help significantly in finding correct jet pairings in mass reconstruction, which is otherwise very complicate because of the huge number of possible jet pairing hypotheses in e.g. eight-jet events. DELPHI multi-jet analysis tools have been used in feasibility studies for linear collider physics processes. An example is the search for heavy charged Higgs bosons at an 800 GeV linear colliderg. Full reconstruction of 8-jet events H+H-300 GeV/cz (tbtb)

H+H-300 GeV/c2(tbtb)

150

175

200 Us 250 275 Joo 325 350 375 400 Raw @lb mgll CCeV/c?

?SO

I75 200 225 250 275 300 325 350 375 400 Rued t4lb mpp8 (CcV/C5

Figure 5. Mass distribution of the charged Higgs bosons in e+e- + H+H- + &b -+ 8 j e t s process at an 800 GeV linear collider (with TESLA parameters) before and after the kinematic fits. A 1.5% mass resolution can be obtained.

805

has been shown to be possible and a 9-constraint kinematic fit, which uses algorithm developed for DELPHI analyses improves the mass resolution significantly as can be seen in Figure 5.

5. Conclusions There has been a long chain of developments from the first jet measurements, performed on events with two back-to-back jets, to sophisticated LEPP analyses. Large variety of different clustering algorithms have been used and new algorithms are still being developed. Sophisticated jet flavour tagging methods are a key element in many jet measurements and physics analyses. Studies of inter-jet phenomena require complementary approaches in which jets can only be used as tools for approximate parton direction and energy measurement. In these studies particles have to be considered as objects that do not originate from independent partons but from the colour fields between partons. Jet physics will become even more and more demanding at future experiments where the jet multiplicities will be higher. Experience and tools of LEP experiments can and will be used and developed further in these experiments.

Acknowledgements The author would like to thank all the DELPHI collaboration members who have helped in preparation of his talk.

References 1. DELPHI Collab., DELPHI 2001-059 CONF 487, contributed paper for EPS HEP

2. 3. 4. 5. 6. 7. 8. 9.

2001 (Budapest) and LPOl (Rome). DELPHI Collab., P. Abreu et al., Nucl. Instr. and Meth. A378,57 (1996). DELPHI Collab., P. Abreu et al., Zeit. Phys. C70, 179 (1996). DELPHI Collab., P. Abreu et al., Phys. Lett. B499, 23 (2001). DELPHI Collab., DELPHI 2001-071 CONF 499, contributed paper for EPS HEP 2001 (Budapest) and LPOl (Rome). DELPHI Collab., P. Abreu et al.,’Phys. Lett. B439, 209 (1998). DELPHI Collab., DELPHI 2000-51 CONF 366, contributed paper for “Rencontres de Moriond”, Les Arcs, France DELPHI Collab., DELPHI 2001-103 CONF 531, contributed paper for EPS HEP 2001 (Budapest) and LPOl (Rome). A. Kiiskinen, M. Battaglia and P. Poyhonen, Study ofe+e- + H + H - at a 800 Gel/ Linear Collider, Physics and experiments with future linear e+e- colliders, Editors Adam Para, H. Eugene Fisk, AIP conference proceedings (2001), Vol 578, p. 237.

E-FLOW OPTIMIZATION OF THE HADRON CALORIMETER FOR FUTURE DETECTORS

S. MAGILL, S. CHEKANOV, G. DRAKE, S. KUHLMANN, B. MUSGRAVE, J. PROUDFOOT, J. REPOND, R. STANEK, R. YOSHIDA Argonne National Laboratory, 9700 S. Cass Ave., Argonne, IL 60439, USA E-mail: [email protected]

A. BAMBERGER h i b u r g University, Hauptstrasse, Freiburg, Germany

The status of R&D at Argonne National Laboratory on the design of a future Hadron Calorimeter (HCAL) is presented here. This includes work done on optimization of the HCAL for best energy resolution for jets reconstructed using E-flow techniques and initial tests of the use of Resistive Plate Chambers (RPCs) as a calorimeter readout device.

1. Motivation At a future e+e- linear collider, it is essential to separately identify the vector bosons, W and Z, as reconstructed in their dijet decays. The ability to use the dijet decay mode significantly enhances the understanding of these processes as backgrounds to new physics as well as known processes in which they appear in the final state. To be able to distinguish the W and Z in jet mode, the dijet mass must be measured to within 3 GeV. This corresponds to a jet energy resolution of 3 0 % / a - better than any present calorimeter result. In the recent past, it was thought that the best performance of a calorimeter could be obtained if hardware compensation was achieved. However, very good results have also been obtained with non-compensating calorimeters as long as sufficient longitudinal segmentation is retained, so that showers can be identified by their starting points in the calorimeter. So far, neither hardware compensation nor longitudinal segmentation has been able to achieve jet energy resolutions approaching the above requirement. Knowing that hadronic jets are primarily made up of charged hadrons (- SO%), if the momentum of these particles can be determined by the tracker, a substantial improvement in energy resolution of the jet should be obtained. Of course, to achieve this, substitution of the particle momentum for the mea-

-

-

806

807

sured calorimeter energy must be done. This requires 3-D reconstruction of the calorimeter showers for both neutral and charged hadrons separately and an algorithm which can uniquely match tracks with calorimeter clusters or cell hits. In hardware, this translates into fine-granularity in the transverse dimensions as well as adequate longitudinal segmentation to define shower starting 25% of a hadronic jet. By themselves, these points. Photons make up particles can be easily measured in an electromagnetic calorimeter with energy resolutions approaching lO%/D.However, since hadronic showers sometimes start in the electromagnetic section, adequate shower reconstruction must be achieved to separate the charged and neutral hadronic showers from the photons. Again, transverse cell sizes should not be larger than R M ,the Moliere radius of the calorimeter absorber material. Finally, if the charged hadrons are all reconstructed by the tracker and the photons by the electromagnetic calorimeter, the remaining neutral hadrons (- 15%) are all that a hadronic calorimeter has to measure. This measurement technique is usually called Energy Flow, or E-flow, which represents the collection of hardware characteristics and reconstruction algorithms (software) necessary to measure jets in high energy physics processes. It is expected that optimization of the calorimeter design and construction suitable for use in an E-flow jet reconstruction program will result in superior jet energy resolution.

-

2. Single Particle Studies in Simulation

The first step in the optimization procedure is to understand the present calorimeter designs. For this purpose, the SD detector of the NLC study group is used. This detector is designed to be used as an E-flow calorimeter - it has very fine-grained transverse and longitudinal segmentation. The ECAL is constructed as a sandwich with the active material consisting of 0.04 cm thick silicon pads of size 5 mmx 5 mm. The absorber is 0.25 cm thick tungsten. There are 30 layers of this construction making up 20 Xo(- 0.8 Xi). The HCAL is constructed of 34 layers of 1.0 c m thick scintillator pads of size 1 cmx 1 cm with 2.0 c m of stainless steel as absorber. The total number of interaction lengths in the HCAL is 4Xi. Both calorimeters are located inside the solenoidal coil which produces a B-field of 4 Tesla. It has been shown in simulations that digital readout of the (fine-grained) HCAL, when combined with track substitution and photon measurement in the ECAL may be able to achieve the required energy resolution for jets in W and Z decay1. The optimization procedure will, therefore, include comparisons of both analog and digital readout for the HCAL. For this paper, tests of “perfect” analog and digital readout of both of

-

-

-

-

-

-

808

these calorimeters were done to understand how to proceed with the HCAL optimization and to establish the goal for the E-flow algorithm. Calorimeter clusters in all cases here consist of all hits associated with the particle initiating the shower, so these measurements correspond to an ideal case for both the analog and digital methods. Reported here are the results of comparing digital and analog readout for photons in the ECAL and digital and analog readout for neutral K; mesons in both the ECAL and HCAL.

2.1. Photons in the ECAL As a test of the utility of digital calorimetry, a look at both digital and analog readout options for photons is done. It was expected that the analog readout would perform best (best energy resolution) since these electromagnetic showers are very dense and probably can not be adequately reconstructed hitby-hit. Figure 1 shows the response of the ECAL to photons from Z decays to jets comparing the analog measurement to the true photon energy.

1,6001.500-1.400.1,3W-1.200-1.100-1.000--

sow-

f!!

800-700--

600 - -

5w

+

100 0 0.0

- 8

0.2

0.4

0.8

0.6 G-

1.0

1.2

cMcr/prtiEk

1.4

*,-. 1.6

- _, 1.8

EM^

Figure 1. Analog energy resolution for photons in the ECAL.

The resolution (alrnean) for these photons in the SD ECAL is seen to be 15.5%. The slight shift of the mean to a value less than one is due to some leakage of photon energy out of the back of the ECAL. This can be compared to a digital readout for the photons. Figure 2 shows the correlation between the number of calorimeter hits in the ECAL cluster to photon energy. Note that the relationship is not linear - the number of hits shows saturation effects at high energies. This is what was expected due to the denseness of the photon showers. Nevertheless, Figure 3 shows the average

-

809

N&dlWs 240

. .

..

.. .. .

.

..

.. . .

... . :. : . . .. .. .

~. . . . . .. .. . . .. . .. ... .. ... ...

180

.

.

i . . . : . : . : ... ...- . ..-. ’.... ..! . . .. ... .. .

.....

04

0

Figure 2.

2

4

6 G-a

8 Em.

10

12

14

(GeVJ

Correlation of number of hits to photon energy in the ECAL.

ratio of number of hits to photon energy in order to extract the digital readout resolution. 1,000900-800.700--

600-500--

Figure 3.

I/

Digital energy resolution for photons in the ECAL.

The energy resolution (cT/rnean) in this case is the analog case, again as expected.

N

25%, much worse than in

2.2. KZ Mesons in E C A L and H C A L

According to the e’e- -+ Z Z simulations used in this analysis, the dominant component of the neutral hadron part of the jets is due to K i mesons. A comparison of analog and digital readout for K i was done, using the “perfect”

810

Kg calorimeter cluster as defined only in the HCAL. By requiring that the Kg energy deposition both starts and ends inside the HCAL, a comparison of analog and digital measurements can be made. In addition, the "perfect" cluster for the Kg, as defined in the ECAL, can also be measured digitally - thus providing a comparison of a finer-grained, higher sampling fraction calorimeter with denser absorber than the default SD HCAL. Figure 4 shows the analog readout of Kg mesons for contained HCAL energy deposits. Lonp A N o p Rcsolulion EV

1.4308

It

Figure 4. Analog energy resolution for K! mesons in the HCAL.

-

The resolution (almean) obtained is 27%. Figure 5 shows the corresponding digital measurement of these Kg mesons in the SD HCAL, comparing the number of hits to the Kg energy. Note that the dependence is linear with no evidence of saturation of hits. A calibration procedure using muons from Z decay to set the energy level of a mip deposition in an HCAL cell yields essentially the same number as obtained from the inverse slope of the Kg hits versus energy, indicating that the granularity (1 x 1 cm2) is sufficient to measure (on average) mips in the reconstructed calorimeter hits. Figure 6 shows the ratio of number of hits to the Kg energy. The resolution (a/mean) for the Kg mesons, if measured digitally, is 29%, this time only slightly worse than the analog resolution. The calorimeter hit correlation with energy is also seen for the fraction of Kg energy deposited

-

811

Nurber:a!Hits *OT 750-700-650--

...

COO--

550500--

.. .

450--

400.-

. . . . .. .. . .. . . . .-...

350--

-

300-250--

ZOO--

. .. . . . . ~ - .- . .-=. . .. .. . ::: . : ..m.

150--

0

loo--

e.OD0..

Figure 5. Correlation of number of hits to Kg energy in the HCAL. Nukr 1 3 0 ~

1ZW-

110--

loo--

mean :

2453Om.256

30--

no-70-60.50.-

o

5

lo

15

20

25

30

35

10

45

50

55

60

mis nHasE-a

Figure 6. Digital energy resolution for KE mesons in the HCAL.

in the ECAL. Figure 7 shows that the digital measurement for K: in the ECAL also shows a strong linear behavior. 3. RPC R&D

Work has also started on possible HCAL readout options. In particular, one option could be the use of Resistive Plate Chambers (RPCs) - particularly as a digital readout, but maybe as an analog readout if operated in streamer

812

0

!

wm

Figure 7.

I

2

4

6

10

8

12

14

16

18

I

20

Urrp Enam KirVJ

Correlation of number of hits to Kg energy in the ECAL.

mode. RPCs are presently operating successfully in the B-factories - Belle and BaBar2. Some attractive features for a future HCAL RPC readout are :

0

Cheap construction (main component is ordinary window glass) Small readout cell size with pads Thin cross-section suitable for sandwich calorimetry

Some drawbacks are :

Gas system manifolds, etc. causing dead areas 0

Very high HV requirements (- 10 kV)

At ANL, several window glass RPC prototypes were obtained from FNAL3 and tested with a cosmic ray trigger. High voltage is applied through a resistive electrode coating applied on the outside of the glass plates. The plates are separated by a gap of 2 mm, which makes up the gas volume. The chambers were operated in both avalanche and streamer mode. Figure 8a shows CR muon signals from avalanche mode operation for H V < 8.5 IcV and Figure 8b shows the observation of streamers at higher voltages. Note that in streamer mode, the charge in the signal is larger by a factor of 200 compared to that in avalanche mode operation. An efficiency of > 95% has been achieved. Different pad size configurations, where the pads are located behind the resistive electrode, were readout. F’rom the pad size tests, it was determined that the effective charge radius spread over 6 cm. It is expected that this can be N

-

813

' 5

. 8o

1

J50

9.OkV. slreorner mode

8.2 kV. avalanche mode

Figure 8. a) CR muon signal charge for RPC operated in avalanche mode, and b) CR muon signal charge in streamer mode.

-

reduced by increasing the resistance of the conductive coating on the outside of the glass. Crosstalk in a pad located 1 effective charge radius away from the signal is small. Work continues to demonstrate the capabilities of RPCs for HCAL readout by minimizing the thickness of the chamber, using small readout pads with small effective charge areas, and investigating the proportionality of streamer signals to particle energy. 4. Summary

A first look at a comparison'of analog and digital readouts of a future HCAL at an e+e- linear collider has been reported here. Analysis techniques and shower reconstruction algorithms for E-flow are being developed in simulation to be used in the optimization of the HCAL design. A particular readout scheme using RPCs for digital and possibly analog readout of the HCAL is being investigated. References 1. TESLA Technical Design Report, Part IV: A Detector for TESLA, Editors: T . Behnke, S. Bertolucci, R.-D. Heuer, R. Settles, pgs. 85-86, (March 2001). 2. A. Abashian, et al., Belle Preprint 99-04, (8 October 1999). The BABAR Collaboration, The BABAR Detector, N I M A479,1 (2002). 3. P. Mazur, Third International Workshop on Resistive Plate Chambers and Related Detectors, Scientifica Acta Vol 11, 1 (1996).

ON THE ENERGY MEASUREMENT OF HADRON JETS

0. LOBBAN, A. SRIHARAN, R. WIGMANS Department of Physics, Texas Tech University, Lubbock, TX 79409-1051,USA E-mail wagmans4ttu. edu

The elementary constituents of hadronic matter (quarks, anti-quarks, gluons) manifest themselves experimentally in the form of jets of particles. We investigate the precision with which the energy of these fragmenting objects can be measured. The relative importance of the instrumental measurement precision and of the jet algorithm is assessed. We also evaluate the “energy flow” method, in which the information from a charged-particle tracker is combined with that from a calorimeter in order to improve the jet energy resolution.

1. Introduction Matter as we know it consists of leptons and quarks. Whereas the properties of leptons such as electrons or muons can usually be measured with a very high degree of precision, the same is not true for quarks. Quarks are “locked up” inside mesons or (anti-)baryons and any attempt to isolate them creates more such particles. In high-energy scattering experiments aimed at studying their properties] quarks, diquarks or anti-quarks fragment into j e t s of hadrons. The precision with which the properties of the fragmenting object can be measured depends on two factors: The jet-defining algorithm and the detector quality. Usually, a jet is defined as the collection of particles that fall within a cone with opening angle R emerging from the interaction vertex. Typical values of R, when expressed in terms of an interval in r ] , q5 space ( R = J-), range from 0.3-0.7. If the chosen R value is large, the cone may be contaminated with particles that have nothing to do with the fragmenting object, if R is small, some jet fragments may be located outside the cone. Fluctuations in the jet energy contained within the jet-defining cone form an irreducible component of the jet energy resolution. At energies below 100 GeV, the contributions of this irreducible component are substantial and in practical experiments they are the main factor limiting the jet energy resolution. However, at higher energies, jets become more and more collimated and the effects of the jet algorithm on the energy resolution diminish correspondingly. In Section 2 of this paper, we investigate the energy

814

815

dependence of these effects. One of the problems in designing calorimeter systems for modern experiments is the fact that the requirements for excellent energy resolution for single hadrons and jets are orthogonal to those for high-resolution electromagnetic (em) calorimetry'. High-resolution hadronic shower measurements require compensating calorimeters. And compensation (2. e. equal calorimeter response to the em and non-em components of hadron showers, e / h = 1.0) is only achieved in sampling calorimeters with a very small sampling fraction, e.g., 2.3% in lead/plastic-scintillator structures. On the other hand, highresolution em shower detection requires an instrument with a large sampling fraction, e.g., 100% in crystals or > 40% in detectors such as the NA48 LKr calorimeter2 In order to solve this dilemma, it has been proposed that one could significantly improve the performance of a poor-resolution hadronic calorimeter system by combining its information with that of an upstream tracker system. In this approach, sometimes referred to as the Energy Flow Method, the momenta of the charged jet fragments measured with high precision by the tracker serve as a first-order estimate of the jet energy. Second-order corrections, intended to account for the neutral jet component, are derived from the calorimeter signals. Of course, the contributions of showering charged particles to the calorimeter signals have to be discounted properly for this method t o work. Methods of this type have been successfully used to improve the resolution of jets from 2-decay a t LEP3. In Section 3 of this paper, we investigate the prospects of such methods at higher energies. Concluding remarks are given in Section 4.

2. Effects of the Jet Algorithm We have studied the effect of a jet-defining algorithm on the energy resolution for fragmenting quarks with a Monte Carlo program that we developed for this purpose. This Monte Carlo program is based on a highly simplified representation of the physics processes taking place in practice. However, it does contain the essential elements necessary to evaluate the energy dependence of the contributions of the jet algorithm t o the resolution. In our program, the fragmentation process is governed by a fragmentation function

D ( z ) = ( a + 1)

(1 - z ) Q 2

in which D ( z ) denotes the probability that a jet fragment carries a fraction z of the energy of the fragmenting object. The parameter a can be chosen as desired. It has been demonstrated that a function of this type gives a

816

reasonable description of the fragmentation processes measured at LEP and at the Tevatron, for parameter values Q = 3 and 6, respectively4. In our Monte Carlo program, jet fragments are generated with energies zEjet, with the values of z chosen from a distribution representing Equation 1. Each fragment is assigned a mass m, a charge and a transverse momentum p l . Ten percent of the particles are assumed to be kaons and ninety percent pions. One third of the particles are electrically neutral, the rest are charged. The transverse momentum is chosen from an exponentially falling distribution with a mean value of 0.3 GeV/c. If the chosen parameters yield an unphysical result, e.g., if the chosen mass is larger than the fragment's energy zEjet, or if the transverse momentum is larger than the total momentum J(ZEjet)2 - m2,the fragment is discarded and a new one is selected. The selection of jet fragments is continued until the jet energy is exceeded. In that case, the energy of the last fragment is reduced so that the total energy of all fragments combined equals the jet energy.

1

.

D

. 5

0

20

40

60

80

I00

Energy (GeV)

Figure 1. Energy distribution of particles generated in the fragmentation of a 100 GeV jet according to Equation 1, with cr = 3 and 6, respectively.

0 0

10

20

30

40

o

%

50

60

70

Multiplicity Figure 2. Multiplicity distribution of fragments from 100 GeV jets, fragmenting according to Equation 1, with a = 3.

We used this program to generate jets with fixed energies, ranging from 10 GeV to 1000 GeV. For each energy, 10000 jets were generated for two different values of the fragmentation function parameter: Q = 3 and Q = 6. First, we show some general results that give an impression of the characteristics of the generated jets. Figure 1 shows the energy distribution for the particles

817

that constitute a 100 GeV jet, fragmenting according to a = 3 or a = 6 . In Figure 2 , the distribution of the jet fragment multiplicity is given for 100 GeV jets and the average multiplicity is shown as a function of the jet energy in Figure 3. In spite of the large numbers of particles constituting the jets, only relatively few particles contribute substantially to the total energy. This is also shown in Figure 3. For example, in a = 3 jets the 10 most energetic particles carry 90% of the total jet energy. For a = 6 jets, that takes 15 particles, on average. This is true at all energies, which is of course a direct consequence of the very concept of a fragmentation function that depends only on z. We defined the cone parameter R that formed the basis of the applied jet algorithm as

+

R = J(Aq5)2 ( A Q ) ~ (2) where Aq5 and AQ denote the spread around the nominal direction of the fragmenting object in the azimuthal and polar angles, respectively. The fate of a jet fragment was decided on the basis of the ratio of its transverse and longitudinal momenta, p l / p ~ If~ . arctan(Pl/PII) > R / 2 , then the fragment fell outside the cone, otherwise it was considered to contribute to the measured jet characteristics (energy, momentum, composition). We thus implicitly ignored the effects of an eventual magnetic field, which has the tendency to sweep soft charged particles out of the cone. Therefore, the results to be presented are somewhat too optimistic, especially for low jet energies. At the high energies which are the focus of our study, jet fragments that are susceptible to the sweeping effects of a magnetic field represent only a small fraction of the total jet energy. The effects of a magnetic field are discussed in more detail in the context of the Energy Flow Method, in Section 3.3. Figure 4 shows the average fraction of the jet energy that was found to be contained in a jet-defining cone, as a function of the jet energy. We used two different cone sizes, R = 0.3 and 0.5, respectively. We also used two fragmentation functions which differed in the value for the parameter a, as before: a = 3 and a = 6. The error bars on the data points in Figure 4 indicate the spread in the jet containment resulting from this degree of freedom. These data show that for jets of about 20 GeV, on average some 30% of the energy was carried by fragments that travelled outside the cone. However, as the jet energy increases, the containment rapidly improves. For energies above 100 GeV, typically less than 10% of the energy is unaccounted for when the chosen jet algorithms are applied.

818

__ a = 6 totll multipliciry a = 6 multipliciiy for 0.9 E,,, a = 3 multiplicity for 0.9 E,.,

2 40-.

L t l :

" . . - . .. .

or...,,

10

'

. . . ..

* '

"""'

'

100

'

"""'

loo0

.

1

Jet energy (GeV) Figure 3. The average fragment multiplicity and the average number of fragments that is minimally needed to account for 90% of the jet's energy as a function of the jet energy, for two different values of the fragmentation function parameter a.

I ....

I

10

. . . . ....

I

.

100

. . .

. . . I

loo0

Energy (GeV) Figure 4. The average fraction of the jet energy contained in the jet-defining cone as a function of the jet energy. Results are given for cones with R = 0.3 and R = 0.5. The error bars indicate the spread in the results caused by the choice of the value of the fragmentation function parameter a.

The energy resolution caused by fluctuations in the energy carried by particles travelling outside the cone is shown in Figure 5. For jet energies of 45 GeV, as found in the decay of Zo bosons produced at the e+e- collider LEP, the contribution to the energy resolution from jet algorithms such as those discussed here amounts to N 10%. Therefore, there was no compelling reason to install detectors measuring hadrons with a precision better than that in the LEP experiments. However, as the energy increases, the situation changes. The jets become more and more collimated and, as a result, fluctuations in the energy contained inside the jet-defining cone are reduced. For jets of 500 GeV and higher, the contribution of the jet algorithm to the jet energy resolution is of the order of 1%,smaller than the instrumental energy resolution achieved with any hadron calorimeter that has ever been tested. Therefore, a high-resolution hadron calorimeter would in practice make a crucial difference for the precision with which high-energy jets can be measured. This is illustrated in Figure 6, which shows the contribution from the "irreducible" fluctuations for a cone with R = 0.3 as a function of the jet energy, which is plotted here on a scale linear in E-1/2 (the solid curve). For comparison, the measured hadronic energy resolutions are given for two experi-

819 Energy (GeV)

I

0.25

Energy (GeV)

Figure 5. The contribution of fluctuations in the total energy carried by particles escap ing the jet-defining cone to the jet energy resolution, as a function of the jet energy. Results are given for cones with R = 0.3 and R = 0.5. The error bars indicate the spread in the results caused by the choice of the value of the fragmentation function parameter a.

0.20

0.15

I

I

I

I

-

0.10

,

0.05

0

1 m

Figure 6. The hadronic energy resolution of three calorimeter systems and the contribution of a jet-defining cone with R = 0.3 to the jet energy resolution, as a function of energy.

ments at the future LHC at CERN (ATLAS5 and CMS') and for the SPACAL calorimeter7, which currently holds the world record in terms of hadronic energy resolution. The latter calorimeter would represent a significant advantage (compared to the LHC ones) for the detection of jets with energies above 100 GeV, but at lower energies the quality of the measurements is dominated by the jet algorithm. It should be emphasized that in this analysis we have deliberately ignored the effects of particles that do not belong to the fragmenting object but migrate into the jet-defining cone. Of course, these effects tend to extend the importance of the contributions of the jet algorithm to the jet energy resolution. They are extremely dependent on the type of experiment (e+e- or p p colliders, fixed target experiments) and also on factors such as the luminosity and the kinematic region under study. These underlying event effects are largest in the high-q regions of high-luminosity p p collider experiments, such as those planned for the future LHC at CERN, where every interesting event is accompanied by 25 other events taking place in the same bunch crossing.

-

820

On the other hand, in high-energy e+e- colliders disturbances of this type are not very important. The conclusions derived from Figure 6 are thus primarily valid for the latter type of experiments. 3. The Energy Flow Method 3.1. The basic idea

In the previous section, we have shown that as the high-energy frontier of particle physics is pushed to higher and higher values, high-resolution measurements of fragmenting hadronic constituents such as quarks, anti-quarks, diquarks and gluons becomes increasingly possible, since the limitations imposed by jet-defining algorithms become less of an issue. Therefore, the intrinsic qualities of the instruments with which the jet energies are being measured become the determining factor in this respect. The techniques that have been used until now in calorimetry make highresolution em and hadron shower detection mutually exclusive propositions'. High-resolution hadronic shower measurements require compensating calorimeters. And compensation ( i e . equal calorimeter response to the em and nonem components of hadron showers, e / h = 1.0) is only achieved in sampling calorimeters with a very small sampling fraction, e.g., 2.3% in lead/plasticscintillator structures. On the other hand, high-resolution em shower detection requires an instrument with a very large sampling fraction. The ZEUS Collaboration, which currently operates the highest-resolution hadron calorimeter in the world8, pays a price for that in the form of a rather mediocre performance for em shower detection: a / E = 18%/@. Calorimeters such as the ones that will be used in the LHC experiments emphasize excellent electromagnetic resolution, at the expense of hadronic resolution, as is illustrated in Figure 6 . In future experiments, e.g., at a proposed linear e+e- collider in the 0.5 - 1 TeV range, which was recently defined as the most desirable future machine for particle physics researchg, one will want to be able to measure all constituents of matter with resolutions at the 1%level. The question is how that can be achieved. One of the solutions pursued in this context involves the so-called E n e r g y Flow Method (EFM), in which the information from the calorimeter system is combined with that from an upstream tracker system. The momenta of the charged jet fragments, measured with high precision by the magnetic tracking system, serve as a first-order estimate of the jet energy. The calorimeter signals are used to obtain second-order corrections to that energy, caused by the neutral jet component (ys, Kos and neutrons). With methods of this type, several LEP experiments improved the resolution of jets from 2-decay from 12% N

821

-

to 9%. We have studied the merits of such methods, and in particular the energy dependence of these merits, with the same Monte Carlo program that was used to investigate the contributions of jet algorithms to the jet energy resolution (Section 2).

3.2. N o calorimeter The Energy Flow Method exploits the fact that the charged fragments of jets can be measured much more precisely with a tracker than with a calorimeter. However, the calorimeter information is still needed to account for the contributions of neutral particles, mainly ys from xo decay, but also Kos and neutrons. In the absence of calorimeter information, based on tracker information alone, the jet resolution would be determined by the fluctuations in the fraction of the total jet energy that is carried by the charged fragments. Table 1 lists these fluctuations for jet energies ranging from 10 GeV to 1 TeV, for simulated jets with a = 3 and a = 6 , respectively. Table 1. The total energy carried by the charged fragments of jets and the fluctuations in this energy (urms/Echarged) are listed for jet energies ranging from 10 - 1000 GeV. Results are given for two different values of the fragmentation function parameter a. Jet energy (GeV) 10 20 30 40 50 100 200 300 400 500 1000

a Charged fragments 6.83 f 2.06 13.2 f 4.13 19.8 f 6.13 26.5 f 8.10 33.4 f 10.0 66.6 f 19.9 133 f 40.1 200 f 59.8 266 f 80.4 332 f 99.9 663 f 201

= 3 Fluctuations 30.1% 33.6% 32.6% 30.6% 30.0% 30.4% 30.2% 29.9% 30.3% 30.1% 30.3%

a Charged fragments 6.88 f 1.68 13.6 f 3.27 20.2 f 4.89 26.9 f 6.46 33.5 f 8.10 66.6 f 16.3 133 f 32.0 200 f 48.4 266 f 64.2 332 f 80.5 665 f 160

= 6 Fluctuations 24.3% 24.2% 24.2% 24.0% 24.2% 24.4% 24.1% 24.2% 24.2% 24.2% 24.1%

The average energy carried by the charged jet fragments is 213 of the jet energy. However, the event-to-event fluctuations are large, the arms amounts to 30% of the average value for a = 3 jets and 24% for a = 6 , independent of the j e t energy. One may wonder why these fluctuations do not become smaller at higher energies, given the fact that the number of jet fragments increases. The reason for this is that the observed increase in multiplicity is uniquely caused by the addition of more soft particles. The bulk of the jet energy is invariably carried by a small number of the most energetic particles (cf. Figure 3). This means that the fraction of the jet energy carried by charged particles is strongly

822

dependent on the extent to which these particles participate in the "leading" component of the jet. Therefore, the event-to-event fluctuations in this fraction are large and do not significantly improve with energy. As an aside, we mention that the same thus necessarily applies for the event-to-event fluctuations in the fraction of electromagnetically interacting particles (mainly 7ros). These fluctuations are responsible for the poor jet energy resolution of non-compensating calorimeters, especially at high energy', since the response of such calorimeters is usually considerably larger for em showers than for non-em ones. In the absence of a calorimeter, one should therefore not expect to be able to measure jet energy resolutions better than 25-30% on the basis of tracker information alone, at any energy. And since the contributions of showering charged particles to the calorimeter signals have to be discounted properly for the EFM to work, the quality of the calorimeter information is in practice important. 3.3. Magnetic field effects The proponents of this method claim that the key to its success in a LinearCollider experiment is determined by the granularity of the detector. A high granularity would make it possible to recognize and eliminate all contributions of the charged particles to the overall calorimeter signal. The remaining calorimeter signal could then be attributed to the neutral jet componentslO. However, the question arises whether the jet fragments, by the time they reach the front face of the calorimeter, are sufficiently separated from each other in order to individually recognize their showers. The lateral develop) the fragments ment of the showers is governed by the Molihre radius ( p ~ for that develop em showers and by the nuclear interaction length (&) for the fragments that develop hadronic showers. If there is significant overlap between the showers initiated by the various jet fragments, then even the finest detector granularity would not make it possible to disentangle the different shower profiles. The problem one faces may be illustrated with Figure 7, which shows an event display of the SPACAL calorimeter''. A beam of pions was sent onto a thin target placed 1.5 m upstream of the calorimeter. Interacting pions were selected and the reaction products were recorded in the calorimeter. SPACAL was a fine-grained calorimeter, the effective radius of each readout cell was 0.19 Xint (2.0 p ~ ) Readout . cells of hadron calorimeters used in most experiments are typically 20 - 30 times larger. The figure shows several individual reaction products that can be clearly distinguished. However, there is also considerable overlap between the showers initiated by these particles. Making the granular-

823

0

20

40

60

80

100

Figure 7. A SPACAL event display of the reaction products from a pion interaction in an upstream target. The numbers denote the energy (in GeV) deposited in the individual calorimeter cells''.

ity smaller would not help to resolve the energy deposit pattern, since the cell size is such that even electromagnetic showers usually ended up sharing their energy among several cells in this detector. The question whether or not the jet fragments are sufficiently separated is mainly determined by two factors:

(1) The distance between the calorimeter's front face and the beam line, where the fragmentation takes place, and (2) The strength of the magnetic field that is used to separate the charged jet fragments from each other.

In addition, the jet energy and the type of jet (fragmenting quark, diquark or gluon) may play a role. We investigated this issue for the proposed TESLA experiment12, which is equipped with a 4 T magnetic field, while the calorimeter front face is located at a distance of 1.68 m from the beam line. According to Figure 3, a particle produced in the fragmentation of a 100 GeV quark carries typically 9 GeV (90% of the energy is carried by the 10 most energetic particles). If this particle is electrically charged, it will deviate from a straight path by 19 cm before reaching the TESLA calorimeter, as a result of this magnetic field. That is 3 times as much as the average deviation

824

from the jet axis resulting from the intrinsic transverse momentum of the jet fragment. For the softer jet fragments, the effect of the magnetic field, compared to that of the intrinsic p l , increases further. For example, a 3 GeV/c 7r+ travelling in a plane perpendicular to the magnetic field will deviate by 65 cm from its original direction upon arrival at the calorimeter's front face, 4 times as much as the effect of the intrinsic p l . And pions with a momentum < 2 GeV/c will not reach the calorimeter at all.

8.6GeVx-10

-100

-50

0

50

-201"

'

'

100 -20

'

'

-10

'

'

'

'

'

0

' '

'

'

'

10

'

'

'

'

I

20

$I direction (cm) Figure 8. Fragmentation of a 100 GeV quark jet with a leading T O in the calorimeter of the TESLA experiment. The circles indicate the characteristic lateral dimensions of the showers developed by the fragments, and the numbers represent the energies of the fragments, in GeV. The point (0,O) corresponds to the direction of the fragmenting quark. See text for more details.

An example of an event in which the EFM might work as anticipated is shown in Figure 8. This is an event with a leading 7ro meson. The charged fragments are bent away by the magnetic field, to such an extent that their showers do not interfere with the (most energetic) photon ones. The circles indicate the characteristic size of the showers initiated by the jet fragments, i.e. p~ for em showers, Xint for hadronic ones. We have assumed a calorimeter with the highest possible density, ie. with the smallest possible values of pn.i and Xint, 1 cm and 10 cm, respectively. The showers from hadronic fragments are indicated by open circles, the electromagnetic ones are represented by the shaded circles. The energies of the fragments (in GeV) are indicated in the figure. Only particles carrying more than 1 GeV are shown in this display, = 0 (perpendicular to the beam line). Quarks which concerns a jet at

825

travelling in other directions will develop jets that are somewhat broader in the r] direction. Figure 8b shows a close-up of the central 40x40 cm2 area surrounding the (0,O) point, which represents the jet axis. However, the problems with this method arise for jets with energetic charged jet fragments. Compared to soft charged fragments, the effect of the magnetic field on these particles is small and, therefore, they enter the calorimeter in the same region where also the (7s from) 7ros deposit most of their energy. The most energetic particle in our 100 GeV quark jet carries, on average, 29 GeV. If it is charged, it reaches the calorimeter's front face at a distance of 6 cm from its straight-line-extrapolated original momentum vector. The distance between the impact point and the jet axis is, on average, 7 cm. Figure 9 shows 4 examples of events which pose serious problems for the EFM. In all cases, one or several energetic em showers fall within the region covered by showers generated by electrically charged jet fragments. One might argue that the longitudinal shower development of these two types of jet fragments is very different and that one could use this information to disentangle the energy deposit profiles. However, it is important to keep in mind that hadron showers typically deposit one third to one half of their energy in the first nuclear interaction length, which constitutes usually the electromagnetic calorimeter section of the detector. It is, therefore, very likely that it is impossible to disentangle the detector hit patterns for events of this type into contributions from the charged jet fragments and from the other components. It should be emphasized that the depicted events were not specially selected to illustrate this point, 70% of all high-energy jet events resemble those shown in Figure 9, which were taken from the sample of the first 10 events generated in our Monte Carlo simulations. 3.4. Importance of the calorimeter quality

For reasons described in the previous subsection, the calorimeter system needs other qualities besides a high granularity. In particular, it needs a good hadron energy resolution in order to measure jet energies with good precision. This resolution will determine how well one can determine the contribution of the precisely measured charged jet fragments to the total calorimeter signal and, therefore, the precision of the neutral energy obtained after subtracting this contribution. 3.4.1. Monte Carlo simulations

In order to quantify the above statements, we have studied the merits of the EFM for a calorimeter system with a hadronic energy resolution n / E =

826

L

-20

-10

0

10

20

$ direction (cm) Figure 9. Four examples of 100 GeV quark jets with energetic charged fragments, detected in the calorimeter of the TESLA experiment. The circles indicate the characteristic lateral dimensions of the showers developed by the fragments, and the numbers represent the energies of the fragments, in GeV. The point (0,O)corresponds to the direction of the fragmenting quark. Only showers initiated by particles carrying at least 1 GeV and developing in the region of 40x40 cm2 surrounding the jet axis are depicted. See text for more details.

+

7 0 % / 0 5% and an e / h value of 1.5. These parameters are typical for the calorimeters that were used in the LEP experiments. The jet resolution that could be expected on the basis of the Energy Flow Method applied to such a calorimeter system was evaluated in the following way. We distinguish three types of jet fragments: a) Particles that develop em showers in the calorimeter (mainly ys from ro decay)

827

b) Soft hadrons that do not interfere with the calorimetric jet measurements as a result of the magnetic field. This field either prevents them from reaching the calorimeter at all or bends them to such an extent that they end up outside the jet-defining cone. c) Hadrons that do contribute to the calorimetric jet signals. If the jet consisted only of particles of types a and b, then the EFM would be a perfect tool t o determine its energy. It would be no problem at all to get sub-1% energy resolutions for jets with energies in excess of 100 GeV. The contributions of these particles to a realistic jet resolution may, for all practical purposes, be considered negligible. This resolution will be completely determined by the particles of type c and, more in particular, by the fluctuations in the signals they generate in the calorimeter. Therefore, in our Monte Car10 simulations, we determined the “EFM” signal for a given jet by smearing the energies from the fragments of type c with the hadronic calorimeter resolution and adding to these the exact energies of the particles of type a and b. Jet fragments were selected as before according to a fragmentation function of type (1). The distinction between hadrons of types b and c was made on the basis of the momentum of the particles. Usually, we considered hadrons with momenta < 1GeV/c particles of type b, and the rest particles of type c, but we also varied this threshold to study its effect. For each hadron of type c, we drew a random entry (E,!)from a Gaussian distribution with a central value given by the fragment’s energy (E,) and a width given by the resolution function, e.g., 0 = 2.7 GeV for a 10 GeV hadronic fragment. Then, the contribution of this fragment to the calorimeter signal was determined taking into account the effect of the e / h value. The em shower fraction was taken to be fern = l-E;o.ls and the signal was calculated as S, = E,,[fe, (1 - fem)h/e]l. The total “EFM” signal was found by summing over all fragments, as follows:

+

n

i= 1

Examples of the resulting signal distributions are shown in Figure 10. For Q = 6 jets, the resolution was found to be 9.3% at 100 GeV and 6.6% at 500 GeV. Compared to the resolution that may be expected from the calorimeter system alone, this represents a relative improvement of 23% and 19%, respectively. We have studied the effects of the EFM over a wide range of energies. We also varied the parameter that defines the jet ty p e - ( a ) , as well as the momentum threshold for the distinction between hadrons of types b and c (pb,). The results are summarized in Figures 11 and 12. Figure 11 shows the calorimetric jet resolution of the generic LEP detector

828

0

200

400

600

Jet signal (a.u.) Figure 10. Signal distributions for 100 GeV ( a ) and 500 GeV ( b ) jets in a LEP detector, after application of the Energy Flow Method. Results of simulations described in detail in the text.

(the dashed line), as well as the jet resolution one might expect when applying the EFM to jets measured with this detector, assuming the momenta of the charged jet fragment are precisely known. The results are shown as a function of the jet energy. At low energies, the EFM is seen to improve the jet resolution by 35%. As the energy increases, the relative improvement slowly decreases, to 18% at 1000 GeV. The main reason for this is the fact that the jets become increasingly collimated at higher energies, the slow hadronic fragments that are swept away by the magnetic field represent a decreasing fraction of the jet energy. This may also be illustrated by the fact that the relative resolution improvement achieved with the EFM increased with the momentum threshold p b c , and with the value of a (a= 6 jets contain more soft particles than a = 3 ones). The error bars in Figure 11 indicate the effect of the choice of the a parameter on the results, and pbc was 1 GeV/c in these simulations. Figure 12 shows how the improvement of the energy resolution that can be achieved with a tracker that measures the momenta of the charged jet fragments depends on the various parameters used in the simulations. This improvement clearly benefits from a stronger magnetic field, which is equivalent to a larger value of pbe (Figure 12a). The benefits of the EFM also have a tendency to decrease when a better calorimeter system is used. We concluded

--

-

Energy (GeV) 20

I

\ -

\

I

I

I

1

\

n

0

Y

2 ;h 9

w2

1

I

I 1

I

- - - OE= 70%/@

\

\

+ 5%

\ \

\

p = 3

3

.rl

a

'

\

@ 15W G 0

1

\ \

3 I-4

829

\

\ \

\

+ +

- 01=6 10-

\ \

+

\

-

\

\ \

4

\ \

* + \ \ '* \

......__ .._....

......_... .+.......-... -..._

5-

..__ ..... +....... -.. .-..+.. *......

.

\

\

\

-

-

\

-..__. ..._

.-I._.

...

0

"

"

~

'

"

"

"

'

"

'

'

Figure 11. The jet energy resolution as a function of energy, obtained after applying the Energy Flow Method (the black dots), using simulated data from a calorimeter with a jet resolution given by the dashed curve. For comparison, the jet resolution of a compensating calorimeter is given (SPACAL', the dotted curve).

this from simulations, in which we replaced the calorimeter by one with a hadronic resolution of 4 3 % / a 3% and e / h = 1.3. On the other hand, they increase when a more inferior calorimeter system is used. We checked that by simulating a calorimeter with a hadronic resolution of loo%/@ 8% and e / h = 2.0. These tendencies can be understood by considering the extreme cases: For a perfect calorimeter, there is nothing left for a tracker to improve upon, while for no calorimeter at all, the tracker still gives a 30% resolution for the jets (Section 3.2). However, as can be seen from Figure 12b, for the three systems we simulated the differences are relatively small.

+

+

830

Energy (GeV) 60r

30

50 7

I

.

100

loo0

00

30

50

I00

I

I

a = 6, pbc = 2 GeVIc o

a = 3, pbc= 1 GeVIc a = 6, pbc = 2 GeVIc

0

. . . - 0 .

0

O

000

0 ~ " ' " " " " " " ' ' " 0.20 0.15 0.10 0.05

0

-

0 0.20

0.15

0.10

0.05

0

IIE

Figure 12. Relative improvement of the jet resolution by using the Energy Flow Method, as a function of the jet energy. Results of Monte Carlo simulations with different parameter choices. The black diamonds were obtained using testbeam data from the CDF Plug Upgrade calorimeter. See text for details.

3.4.2. Experimental data In order not to rely exclusively on Monte Carlo simulations, we also analyzed some testbeam data taken with the CDF Plug Upgrade calorimeter13. This calorimeter consists of two sections, which we will label EM and HAD. We used these data to build "libraries" of jet signal distributions for jets of a variety of energies, ranging from 30 - 1000 GeV. This was done as follows14. Jet fragments were selected as before according to a fragmentation function of type (1) with a = 6 , which is favored by the CDF data. For each jet fragment, we used the measured signal distribution for testbeam pions or electrons of the nearest energy in order to determine what the calorimeter signal would have been had that fragment actually deposited its energy in the calorimeter. For each jet fragment i with energy Ei,we randomly pulled an EM signal, sem,and a HAD signal, shad,from the corresponding signal distributions for a testbeam run of electrons (if the fragment was neutral) or pions (if the fragment was charged) whose energy was closest to the energy carried by the jet fragment. For instance, for a 10 GeV charged jet fragment, we used the experi-

831

mental signal distributions for an 8.6 GeV pion testbeam run for that purpose. This jet fragment was then attributed an EM signal SFm= (10/8.6)sem and a HAD signal S:ad = (10/8.6)shad, respectively. For a 10 GeV neutral fragment, the same procedure would be followed, but the signals would be taken from an electron run rather than a pion run. Charged hadrons with energies below 2 GeV were assumed not to reach the calorimeter and, therefore, they did not contribute t o the calorimeter signals. The described process was repeated for each of the n fragments that made up the jet of energy Ejet, The energy the calorimeter would have reconstructed for this particular jet then is simply

where A and B denote the calibration constants used to convert the signals from the EM and HAD calorimeter sections into energies. Such Sj,t signal libraries were generated for each of the chosen jet energies. For a given jet of a certain fixed energy, e.g., 100 GeV, the experimental pion signal distributions were used to determine the calorimeter signals S;" and (in the EM and HAD sections) for individual charged jet fragments j , following the same procedure. These signals were also converted into energy units using the calibration constants A and B . The total calorimeter signal from the m charged components of a given 100 GeV jet was thus found as

and the calorimeter signal (or rather its energy equivalent) representing the neutral jet components of our 100 GeV jet (EneUt) was found by subtracting Echarged from the average value of the Sjet distribution for 100 GeV jets. Finally, the energy found with the EFM for this particular jet was calculated as EEFM

= 2 E i

+

Eneut

(6)

i=l

where Ei (i = 1,2, ...rn) represent the exact energies of the chosen charged jet fragments, including the soft ones swept away by the magnetic field. The relative effect of the EFM on the jet energy resolution was determined by comparing the fractional widths of the Sjetand EEFM distributions. The improvement of the jet energy resolution found this way is included in Figure 12b. The data agree well with the results of our simulations for a calorimeter system with the properties of the CDF one.

a32

Both our simulations and the experimental data show that the EFM does offer a beneficial effect. However, this effect should not be exaggerated. The improvement in the energy resolution is typically 30%. Poor calorimeter systems benefit more than good calorimeter systems, and a strong magnetic field also helps. It is important to note that the EFM does not work as well at high energies as at low energies. Therefore, the 30% improvement in the mass resolution obtained for hadronically decaying Zos at LEP is probably an upper limit for what may be expected from this technique at a high-energy LinearCollider experiment. At high energies, the hadronic calorimeter resolution is dominated by fluctuations that result from the different calorimeter response to em and non-em energy deposit. These fluctuations are not addressed, nor cured by the EFM. For comparison, we show in Figure 11 the jet resolution measured with the SPACAL calorimeter7. Thanks to the compensating character of this device, the jet resolution scales very well with E - 1 / 2 . This feature, combined with the diminishing resolution improvement achieved with the EFM at high energies, is responsible for the much better performance that may be expected in the high-energy region. 4. Conclusions

In this paper, we have studied some of the factors that limit the precision with which the energy of fragmenting quarks may be measured. We found that at energies below 100 GeV, a dominating role is played by the algorithm that is used to identify the components of the fragmenting quark (the jet algorithm). However, at higher energies, the jets are increasingly collimated and the precision with which the quark's properties may be measured is increasingly dominated by the quality of the experimental equipment. We have shown that the so-called Energy Flow Method, in which the momenta of the charged fragments form the basis of the jet energy measurement, provides a modest improvement of the resolution that can be obtained with stand-alone calorimeter systems. The relative improvement is about 30% for jets from decaying W,2 bosons and decreases at higher energies. Claims that much better results may be achieved for highly granular calorimeter systems, in which the showers generated by the individual jet fragments may be recognized and separated from each other are unsubstantiated. We have shown that for most of the showers in practical detectors, the overlap between the shower profiles rather than the detector granularity is the factor that limits the benefits of this method. A better solution for high-precision measurements of fragmenting quarks is a high-resolution calorimeter system. Traditional compensating calorime-

-

833

ters, which allow for high-resolution jet measurements, are limited in their electromagnetic resolution. An optimal solution might be a dual-readout calorimeter15,in which the em shower fraction is measured event-by-event and which does not require the small sampling fraction needed for compensating devices.

Acknowledgments This study was carried out with financial support of the United States Department of Energy, under contract DEFG03-95ER40938, and of Texas Tech University's Clark Scholars program.

References

- Energy Measurement in Particle Physics, International Series of Monographs on Physics, vol. 107, Oxford University Press (2000). G.D. Barr et al. , Nucl. Instr. and Meth. A370 (1996) 413. D. Buskolic et al. , Nucl. Instr. and Meth. A360 (1995) 481. D. Green, Dijet Spectroscopy at High Luminosity, Fermilab Report FermilabConf-90/151 (1990). ATLAS Collaboration (1996). The ATLAS Liquid Argon Calorimeter Technical Design Report, report CERN/LHCC/96-41 (1996). P. de Barbaro et al. , Nucl. Instr. and Meth. A457 (2001) 75. D. Acosta et al. , Nucl. Instr. and Meth. A308 (1991) 481. G. Drews et al. , Nucl. Instr. and Meth. A290 (1990) 335. Press release after the 2001 APS Workshop on the Future of Particle Physics, Snowmass (Colorado), July 21, 2001. V.L. Morgunov, Calorimetry Design with Energy-Flow Concept, talk presented at the 10th Int. Conf. on Calorimetry in High Energy Physics, Pasadena, March 25-29, 2002. Acosta, D. et al. (1991), Nucl. Instr. and Meth. A305,55. TESLA Technical Design Report, report DESY 2001-011, DESY, Hamburg, Germany (2001). M. Albrow et al. , Nucl. Instr. and Meth. A480 (2002) 524. M. Albrow et al. , Nucl. Instr. and Meth. A487 (2002) 381. R. Wigmans, Quartz Fibers and the Prospects for Hadron Calorimetry at the 1%Resolution Level, Proc. of the 7th Int. Conf. on Calorimetry in High Energy Physics, Tucson, 1997. World Scientific, Singapore, 1998, p. 182-193.

1. R. Wigmans, Calorimetry

2. 3. 4. 5. 6. 7.

8. 9. 10.

11. 12.

13. 14. 15.

This page intentionally left blank

Perspective Chairperson: R. Wagrnans

D. Green

The Future of Calorimetry in High Energy Physics

J. Krizmanic

Future Experiments in Astrophysics

t W. W. Moses

Synergies Between Electromagnetic Calorimetry and P E T

-

~~

~

t Presented in the medical application session

835

This page intentionally left blank

THE FUTURE OF CALORIMETRY IN HIGH ENERGY PHYSICS

DAN GREEN MS205, Fermilab, Batauia I L 60565, USA E-mail: [email protected]

The paper begins by defining the role of calorimetry in high energy physics experiments. Then the status of the present state of the art is briefly examined. Recent improvements in calorimetry, e.g. “energy flow” are examined, particularly in the light of fundamental limitations in the calorimetric technique. Directions of possible future developments in calorimetry as new facilities are planned and designed are examined.

1. Introduction

The author denies all responsibility for the grandiosity of the title of this invited talk. In addition, as a practicing CMS physicist1t2,the focus of my talk will, of necessity, be on LHC issues. Finally, I will concentrate on those topics where I have firsthand knowledge. Therefore, the scope of this paper cannot be all inclusive, for which you have my apologies. First let us consider the guidance of the past which our science offers. For the last 35 years, the energylmass scale at which we work with has increased exponentially with time as seen in Fig. 1. That is why we call our particular science “high energy physics” As the energy frontier advances in the future, we can predict that, with the energy increase, calorimetry will become increasingly important since the fractional energy resolution dE/E is, at worst, constant with energy, while tracking systems have a momentum resolution which scales as, dP/P P2 In addition, in whatever form it takes, we know that “New Physics” will appear at high mass scales and decay ultimately into the Standard Model (SM) particles shown in Fig. 2. Therefore, we need to look at the role of calorimetry in the measurement of the properties of SM particles at high energies. Calorimetry is used to trigger the detectors, to perform particle identification (id) and to measure the energy and position of the SM particles produced in high energy collisions4. Calorimetry will be used for jets of quarks and gluons as well as showers of electrons and photons as shown in Fig. 3. Neu-

-

’.

837

838

LO

Starting Year Figure 1. Plot of the energy available in the C.M. frame for particle production as a function of the year of start of operations at an accelerator.

trinos (or other new and exotic particle states, e.g. SUSY) will be inferred from missing transverse energy (ET) measured in calorimeters. Muon particle id will rely on the minimum ionization energy deposit and use calorimetry for isolated muons (but the critical energy is, Ecrit 300 GeV). Tau lepton id will come from “narrow” jets arising from the decays, T -+ vW,W + T , p, A. An example of the identification capabilities of calorimetry is shown in Fig. 4. The test beam data arise from a beam of mixed e, p and T , incident on a calorimeter with two longitudinal segments, electromagnetic and hadronic, ECAL HCAL. Particle id uses the differences in interaction length, X , and X(e and T ) , which is a factor 30 in lead (Pb), and in the radiation cross section of leptons (e, p ). To lowest order pions deposit energy only in HCAL, electrons/photons only in ECAL, and muons deposit minimum ionization in both segments. N

+

839

Figure 2.

A tabular list of the fundamental particles of the Standard Model (SM).

ECAL

HCAL

T e

Figure 3. A tabular representation of the role of calorimetry in triggering, particle id, and measurement of photons, electrons, muons, jets (quarks and gluons), and missing energy.

At some point with increasing energy scale, when it is above the critical energy, the muon becomes much harder to identify as an object that “only ionizes”. In general, we have not yet reached that situation, as we see in

840

Figure 5 . Muon energy loss in iron by ionization and radiation as a function of the muon energy5.

Figure 4. Energy deposited in the ECAL and HCAL segments of a test calorimeter for a beam containing electrons, muons and pions.

Fig. 5, although operation at small angles at the LHC now approaches this difficult regime. For quark and gluon jets, what is actually measured is an ensemble of photons and hadrons. If we assume a measurement error for single particles which is characterized only by a "stochastic coefficient" ,a, and ignore a "constant term" in the fractional error dE/E, then the jet energy error is single particle error. Similarly, the dijet mass error is determined by the jet energy error. Finally, the missing energy error for dijets is also determined by single particle error. The derivation of these statements is indicated in Eq. 1. Of course, we assume here that we can correctly identify the ensemble, with no spurious additions or deletions. This assumption is discussed later in this paper. N

M 2 = 2E1E2(1 - ~ 0 ~ 8 1 M2 ) 4ElE2

dMIM El

$T d$T

N

E2

a / a

-

N

N

-

4E2

a/m

MI2

(1)

El - El N

a m

- ad=

2. Status

After that briefest of introductions it is time to take stock of where we are in calorimetry. At present several new collider experiments are in data taking mode. There are two main types of collider experiments. The electron-positron

841

colliders Belle/BaBar are running very successfully and are beginning to explore CP violation in the B sector. However, major luminosity increases will not be possible with the existing calorimetry. In the area of hadron colliders CDF and DO are data taking and aim for high luminosity to search for the Higgs boson and other new physics such as supersymmetry (SUSY). Upgrades to CDF and DO do not envision major changes to the calorimetry. Other detectors are completing the R&D phase and are going into production. The CERN Large Hadron Collider (LHC) detectors are being built, ATLAS/CMS, and have some “headroom”. However, an increase of lox in the LHC luminosity, which has been discussed by the accelerator teams , cannot be simply accommodated because radiation damage becomes insupportable. In the realm of R&D, the Linear Collider (LC) has been judged to be the next new facility that should be built. Therefore, R&D has begun and LC detectors are being designed for high precision measurements. Let us look at the fruits of the R&D programs which were mounted for the design of the new detectors at the LHC. For electromagnetic calorimetry, sampling calorimetry has made great strides. With fine sampling, i.e. a sampling fraction (ratio of active to total sampling), W 0(1),and small source capacity (accordion, ATLAS R&D) a high precision , high speed, EM calorimetry is possible as indicated in Fig. 6 . N

t

W Figure 6. Stochastic coefficient as a function of sampling fraction for several electromagnetic calorimeters3.

10

100

low

E[GeV]

Figure 7. Energy resolution in CMS electromagnetic calorimetry6 with the various contributions due to noise, photostatistics, and intrinsic factors labeled.

An approximate expression for the stochastic coefficient as a function of W is shown in Eq. 2. Note that a+O as W + l , which ignores errors outside of

842

sampling.

W = 6E/(6E + A E ) Fully active devices have no sampling fluctuations. However, there exist also noise, photon statistics, and light collection non-uniformity . For example in CMS6 the P b W 0 4 crystals (Fig. 8) have an energy resolution of, d E / E ~ 0 . 7 %at 100 GeV even though stochastic coefficient is only -2.3% as seen in Fig. 7. At these levels of precision, control of the contributions to the constant term is crucial. At higher energies the constant term dominates the error in measurement of the energy. Note that the transducer used, the APD, is also the result of SDC and CMS R&D. The ATLAS accordion electrodes are the result of LHC R&D as are the CMS crystals and transducers. Clearly, a vigorous R&D program was the vehicle that allowed high precision EM calorimetry to be designed for the LHC.

Figure 8. CMS crystal, transducer (APD = Avalanche PhotoDiode), and test beam results for electron energy resolution6.

For the hadron calorimetry, R&D has also contributed to the understanding of hadron showers. Shown in Fig 8. are the longitudinal profiles from test beam data for six pions in a finely longitudinally segmented and deep lead (Pb) sampling calorimeter.

843

Layer Number Figure 9. Longitudinal profile for pions incident on a P b calorimeter with scintillator sampling. Note the electromagnetic clusters localized to a few samples and the fluctuations in the shower development (SDC data, unpublished).

Figure 10. Dijet mass resolution for the L3 experiment in WW events a t the W mass, and for CD F with "b tagged" jets a t the Z mass.

The present state of the art for dijet mass resolution is shown in Fig.10. In the LEP case a beam constraint7 can be used and the result is that dM/M -3.6% at the W mass. In the case of hadron colliders the fundamental quark initial state is not constrained. The CDF data' with no cuts to reduce final state radiation (FSR) but also with no pileup, is that dM/M -14%. These are big differences which need critical examination, which will be attempted later. There are clearly implications for Linear Collider calorimeter R&D.

844

3. Improvements in Detectors

We now look briefly at improvements in the analysis of detector data that have arisen in considering the projected performance of the new LHC detectors. This is not an exhaustive list but should be indicative of the continuing effort to improve calorimeter performance. 3.1. Non-compensation %onstant term”

One issue for calorimetry is that of compensation. It is well known4 that precision electromagnetic (EM) calorimetry is incompatible with compensation. Thus, if we require excellent EM energy resolution, we must explore the impact of non-compensation. For a device with hadronic response h and electromagnetic response e, the response to a hadron shower with neutral fraction “fo” includes an error term which is present when e is not equal to h.

E

N

+ h(1

[e”f{

-I’

~:)]EIN

Since the fluctuations in the neutral fraction, “dfo”,are not zero, there is an energy resolution factor which is energy dependent. In fact, non-compensation leads to dE/E which decreases as ln(E), following the dependence of the particle

8

.-0

80

70

5 50

3

40

k?

30 20

-

1

l

l

l

l

I

I

I

1

I

1

’

resolution (rms) photodectron statics subtracted

0

-

n

-

0

-<

-

0

-

10-

0

1

-

~ t -

0

-

60 -

I

I

0

a

[I:

I

‘i I

I

I

I

l l l l l

> I

I

I

I

I

I

I

845

multiplicity

Data taken at the CERN H2 test beam as part of CMS R&D1 is shown in Fig. 11. Note the component of the energy resolution at high energy which falls as expected if the main contribution is e not equal to h. Because this device has quartz fiber sampling (sensitive only to electrons and thus the EM part of the shower), we expect to be dominated by errors of the type indicated in Eq. 4. 3.2. Mixed Media

In many cases, the electromagnetic and hadronic compartments of a calorimeter system may be of different construction. In this situation we have “mixed media”, devices with different (e,h) in different compartments. Pions will deposit energy in both compartments, see Fig. 4, and have poor energy resolution unless the mixed effect is properly corrected. Basically we need to determine e/h for both the ECAL and HCAL compartments of the full system.

R = response = eE,

+ hEh

combined setup

ECALIHCAL calib to electrons - eE, eH (e/h) = (e/.)(1 - Fo)/[l- Fo(e/.)l For HCAL, pions which do not interact in the EM compartment can be selected. The hadronic compartment can also be exposed to electrons. In this way, e/h for the HCAL can be determined. For the ECAL, electrons can be used for calibration. Then, by using pions, and by using the known energy of the test beam, the e/h for the ECAL can also be found. In that way, as shown in Eq. 5, the effect of “mixed media” can be removedg. Clearly, these corrections improve the overall resolution, although the effects of non-compensation remain. The pion non-linearity effect of e/h not equal to one is also removed. In order for this method to be valid, particle id is needed so that corrections appropriate to pions are not applied to photons. Particle id in CMS uses

846

Figure 12. Test beam datag on the energy deposited in ECAL and HCAL for 300 GeV pions without e/h corrections and after e/h corrections for both the HCAL and ECAL compartments. These corrections must be made on an event by event basis because EE/EH varies.

e -

l Figure 13. Transverse r.m.s. size for test beam data on pion and electron showers in the CMS crystals used in ECAL. Units are crystal size (CMS data, unpublished).

the transverse crystal size in ECAL (- X o ) and the ECAL/HCAL energy partition as a discriminant. It is expected since the crystals have transverse dimension XO,that photons will be contained within 1 crystal, whereas pions interacting in ECAL will be more widely spread. Test beam data in ECAL on e and pion transverse r.m.s. size, R, is shown in Fig. 13. Indeed, there are marked differences between the pions and the electrons. However, there will be basic limitations to particle id at the LHC due to pileup which then causes id errors due to the intrinsic size of hadron showers which accidentally overlap the photon shower. Perhaps increased longitudinal segmentation would reduce the overlap effect, as has been concluded by those doing R&D on the LC calorimetry.

-

-

847

3.3. Energy Flow

In the past, the calorimeter was treated as a “stand alone” device. This is not an optimal use of the totality of a collider detector. For example, taking typical tracking errors from CMS, an ECAL with a 5% stochastic coefficient, and a 1%constant term, and an HCAL with a 50% stochastic coefficient and a 3% constant term leads to the plot shown in Fig. 14.

:m - Tracking

,,-’

/‘

.

10.’ :

,)

.%

I

-- -*- - \

-..

-

//

-.

--

Y %

z

-.--

/ ---7

/-fflr

--------..-

:

/,,

A:.

i

----..--___ __

--- - -

,-= ./-

r”

,’

,/”/’rx

z-

10-3

/’

f‘.

,

,

, , , , ,,

,

I

2

3

Figure 14. Energy measurement error as a function of energy for a "generic" tracking system, and a "generic" ECAL and HCAL. The crossover is at `(100/400) Gev for ECAL/HCAL.

-

Note that a jet is an ensemble of particles where the most energetic has a momentum fraction, on average, of (zmaz) 0.22. For charged particles < 100 GeV (jets < 0.5 TeV) the momentum resolution of the tracks is better than the calorimetric energy measurement E. Therefore, for present energy scales at the LHC and LC the best measurements are made by the tracker and they should be used if possible. At a VLHC this will not help, as we see from Fig. 14, but that is only a future worry.

848

Guided by these considerations, an examination of “energy flow” was begun in CMS. That means, fundamentally, the replacement of the charged particle energy deposit in the calorimeter by the track momentum. The basic insight is that, even in a jet, the towers are only sparsely occupied. An example comes from a study of 120 GeV Z’ bosons decaying into dijets done for CMS. A typical event is shown in Fig. 15

Figure 15. Energy deposits in a jet cone from Z’(l20) decayslO. ECAL and HCAL deposits, with HCAL segmentation, are shown in a) and b). T h e particle id for photons, pions interacting in ECAL and pions interacting in HCAL are shown in c), d ) and e).

A jet cone of radius 0.9 has -400 towers of HCAL and -10,000 towers of ECAL with CMS segmentation. Note the tower energies in ECAL and HCAL shown in Fig. 15. The energy deposits are clustered transversely and then are sorted into three categories, photons, pions interacting in ECAL and pions interacting in HCAL. Clustering is performed in 3 x 3 towers to avoid edge effects. Of the -400 towers in the cone, only -24 clusters are occupied - i.e. the towers are sparsely populated. The distribution of number of clusters in the cone is shown in Fig. 16. The cone size was chosen t o minimize the mass resolution, as shown in Fig. 17. At low luminosity where this data was simulated (no pileup) a shallow minimum exists at R 1.0. For a Monte Carlo sample of 120 GeV Z’ the attempt was made after clustering and particle sorting id t o match tracks in both 7 and C#J to “hadronic” clusters within the jet - cone size -0.9. The units used in Fig. 18 are HCAL tower sizes. The matches are shown as a function of the transverse energy of

-

849

Figure 16. Distribution of the number of clusters found in a jet from Z'(120) decaylo.

Figure 17. 2jet mass resolution (dM/M) and the ratio of reconstructed mass versus generated m a s (MJJ/Mo) are shown as a function of jet cone radius for Z'(l20) decaydo.

Figure 18. Matching of tracks to calorimeter "hadronic" clusters in both function of the track momentum".

and

as a

the tracklo. There is little ambiguity in matching clusters to tracks for transverse energy > 10 GeV, but many accidental matches at low transverse energy. After replacing the energy and angles of the matched tracks, the mass is recomputed. There is a -22% improvement in the dijet mass resolution. This is welcome but the fact that it is a small improvement clearly implies that calorimeter resolution is not the whole story. At CMS there must be other major contributions t o the dijet mass error beyond those contributed by the calorimeter resolution. Note that final state radiation (FSR) is active in this Monte Carlo sample. CDF have studied energy flow in photon + J events using the EM "shower

850

dlWM = 21%

dMfM= 17%

Figure 19. CMS study of the dijet mass, M J J , distribution for Z’(120) events using purely calorimetric energy measurements or using “energy flow” t o replace matched clusters with the momentum of the associated trackslo.

max” detector for aid in particle id along with the tracking information. A similar ~ 2 4 % improvement was seen. The results” are shown in Fig. 20. There is an improvement in measurement accuracy, but again, not a dramatic one. Note that pileup is small in the CDF data.

+ Jet PrBalancing in CDF Data

Photon

COF Preliminary I

20

. . . ,

25

I

.... ....

30

, , , ,

I

I

35

40

I . . . ,

45

I

. . . . , , . . ,

50

55

E

Photon Pr (GeV)

+

Figure 20. Jet energy resolution for photon jet events in CDF. Th e two curves shown as a function of photon transverse energy are for pure calorimetry and calorimetry augmented by track momentum in the case of “energy flow” matches”.

851

3.4. Missing ET Given that energy flow improves the mass resolution of high mass dijets, it is appropriate to see if a “global” variable, computed using all the towers of the calorimetry, such as missing transverse energy is also improved. A study has been done for CMSl2, first using only inclusive inelastic, or minimum bias events. Three major sources of missing ET were identified - incomplete angular coverage, B field “sweeping” of charged tracks to small angles and calorimetric energy resolution. A plot of the missing transverse energy as a function of the maximum pseudorapidity coverage is shown in Fig. 21. Note that “complete coverage” only occurs a t 9 units of 1171 in the model of minimum bias events used in this study.

E w n t Mllllng & - 6 . 7

Figure 21. Mean missing transverse energy in minimum bias (i.e. inclusive inelastic) events at CMS as a function of the maximum pseudorapidity coverage12.

GOVTObl

I

Figure 22. Fractional contributions to the missing transverse energy in minimum bias (inclusive) events at the LHC. The three contributions are to be added in quadrature to obtain the mean missing transverse energy”.

The contributions to the total missing transverse energy in a “minbias” event are shown in Fig. 22. The sense of the “pie” chart is that the three contributions should be added in quadrature. Note that in this model the incomplete angular coverage of CMS, 1171 < 5, is the major contribution. However, since the other contributions are not small, CMS have chosen not to extend the coverage at small angles due to the limited payoff and greatly increased radiation dose with which smaller angle detectors would have t o contend. The “energy flow” methods can also be used a t the LHC for minimum bias events. In the case of pileup, one could use the tracker t o remove calorimetric deposits due to charged hadrons in pileup events, sorting on the vertices. Within the desired event, one could use the tracker in “energy flow” mode to

852

Figure 23. Missing transverse energy significance for badly measured dijet QCD events for purely calorimetric measurements versus “energy flow” measurements. The improvement of the latter is clear, but limited significance (CMS study, unpublished).

reduce the %weeping” effects of the B field and the effects of poor calorimeter resolution at low energies. The plan would be to reduce the effect of pileup at high luminosity (not with neutrals though) and then to reduce the effects of B field sweeping and calorimetric energy error just as in the dijet case (energy flow), but now for all energy deposits over the full tracking coverage of CMS, ( ~ <1 2.3. If these effects can be reduced, then the angular coverage could more profitably be extended down to smaller angles. At future higher energy colliders, with a longer rapidity “plateau”, smaller angle coverage will be needed in any case. This study has just started. There is a 16% improvement in the missing ET significance, S = $ ~ / d musing , energy flow (no pileup) applied to a data set of badly mismeasured high ET dijets. The results of this first study of missing transverse energy are shown in Fig. 23. The improvement is somewhat encouraging. However, the matching of tracks to clusters is not particularly good at low transverse energy at present, see Fig. 18, which limits the matching reliability t o those deposits with high energies. Work will continue to attempt to improve on the matching algorithms.

853

4. Intrinsic Limitations Despite the improvements that have recently been made, there remain intrinsic limitations in the calorimetric technique. These follow directly from the physical processes used in detection or from the objects to be measured themselves, e.g. jets. Transverse position is measured with an error set by the transverse shower extent, either X o or A. In turn, this sets a limit to the useful smallest tower size. Longitudinal depth is set by the requirement that shower leakage not degrade the energy resolution of the device. The required depth is 20 X O for electromagnetic calorimetry and 1OX for hadronic calorimetry. Greater depth is not called for based on intrinsic losses set by jet leakage. Speed of signal formation is presently limited by the 25 nsec bunch crossing at the LHC. No reduction in pileup is achieved if the signals are faster. Jet energy resolution is limited by several factors. Final state radiation (FSR) at the LHC dominates over calorimeter energy resolution. Therefore, a “better” HCAL may not lead to better Physics at the LHC. Let us look a bit more deeply at the above list of limitations.

-

N

4.1. Transverse Position

Very fine grained EM calorimetry with multiple samples in depth can be used to make angular measurements of electron or photon showers using the center

‘t Figure 24. Angular resolution expected in the Tesla EM calorimeter as a function of photon energy13.

0

10

ZU 30 Radius [cm]

40

50

Figure 25. Distribution of transverse energy in hadronic showers integrated over all shower depths. There are two components of the shower evident, and EM “core” and a wider hadronic part.

854

of gravity of the EM shower at different depths. The accuracy of the EM shower angle is ultimately limited by stochastic shower fluctuations. The resolution expected in the proposed Tesla EM calorimeter13 is shown in Fig. 24. For hadronic calorimetry the jet position is also measurable. The hadronic shower size of the hadrons in a jet limits the number of resolvable “particles” in a jet, especially in the dense “core” of a jet near the cone (-parton) axis. In turn, the hadronic shower size then puts limits on the “energy flow” algorithms. A plot of the transverse shape of a hadronic shower, integrated over all depths and energy weighted is shown in Fig. 25. Clearly, there are two components, the EM core of the shower and the hadronic remainder. As we might expect from Fig. 9, there are very large fluctuations in the shape event by event, which limit the transverse position measurement resolution3. 4.2. Leakage and Depth

Good hadron energy measurements will require a depth > 1OX due to late developing shower leakage and the fluctuations in the energy of those showers5. In Fig. 26 is shown the average depth in iron needed for 95% and 99% hadron shower containment as a function of hadron energy.

a-F

A

#

4

8

8 0

Figure 26. Depth in iron required for hadron containment of 95% and 99% of the energy on average as a function of the hadron energy5.

855

The results of Fig. 26 are true on average. However, we know there are tremendous fluctuations in hadronic showers. There is, therefore, a finite probability for a single hadron to “leak” a large fraction of its’ energy due to late developing showers which have small neutral fractions, fo. Test beam data’ from CMS on 300 GeV pions incident on a calorimeter of depth 7X followed by a “tail catcher of depth 4X is shown in Fig. 27. Since there are 5000 pions in the plot, there is a finite probability, 0.5%, for there to be more than 200 GeV in the tail catcher. Because CMS will study rare events, such as SUSY with missing energy, a depth of 11X was chosen in CMS to reduce leakage to acceptable levels.

Figure 27. Test beam data for 300 GeV pions incident of a calorimeter of depth 7X followed by a compartment of depth 4X’.

There is another effect, illustrated in Fig. 283. The probability for there to be a hit deep in a calorimeter as a function of the pion energy is shown as a function of depth at different energies. Clearly there is a component that dies off exponentially with a length scale A. However, at the -1% level there is another component that dies off only slowly with depth. It is due to pions in the hadron shower decaying via, IT -+ pv, before they are absorbed. Thus, hadrons ”leak“ no matter what the depth used. Clearly, it is best to make the

-

856

P

Figure 28. Probability to observe a particle as a function of depth for pions incident with depth independent component which appears at the -1% different energies. There is a level, for 300 GeV pions, deep in the calorimeter3. N

calorimeter to be as compact and dense as possible. Finally, there is another effect; jets “leak” themselves. Consider the measurement of a gluon jet. Jet “splitting”, g+QQ, with subsequent decay, Q+qlv and the neutrino carrying off energy, puts an intrinsic limit on the required depth. A simple Monte Carlo model14 was made with results shown in Fig. 29. From that figure we can conclude that -0.1% of all gluon jets will lose > 1/2 of their energy due to splitting. Clearly, this means that there is a limit t o the depth of calorimetry which one need not exceed.

857

#of jets

10

MlsSlng Energf (GeV)

Figure 29. Number of jets with energy greater than the missing energy for a 500 GeV gluon jet for 10,000 generated gluon jets14.

4.3. Signal Speed

The LHC operates with bunch crossings every 25 nsec. At design luminosity there are 20 minimum bias events on the average within each bunch crossing. Therefore, the pulse formation time of LHC calorimetry needs to be < 25 nsec, but no faster. The results of a vigorous R&D program are that LHC calorimetry, both liquid argon and scintillator based, is fast enough to minimize pileup effects at the LHC since all calorimeter signals at the LHC are contained in 1-2 bunching crossings. N

N

Figure 30.

Pulse formation in a liquid argon calorimeter with preamplifier-shaper.

858

10

m

40

41

83

m

80

92

100

ti-4

Figure 31. Observed pulse shape in the CMS hybrid photodiode used in the hadron calorimetry (HCAL). At bias voltages yielding fields well over depletion, the pulse is contained in a single 25 nsec bunch crossing16.

The pulse shaping for a liquid argon calorimeter is shown in Fig. 30. If the source capacity can be controlled (ATLAS accordi~n’~), then a pulse shape with width less than a single bunch crossing can be obtained. The impulse response16 of a hybrid photodiode (HPD) is shown in Fig. 31. The HPD used in CMS for the hadron calorimeter is sufficiently fast to contain the pulse within 25 nsec. Note that both the accordion electrode structure and the HPD are the fruits of the LHC program of R&D. 4.4. Energg/Mass Error

In searches for new physics, jet measurement and jet-jet invariant mass are crucial elements. For example, the Higgs decay into b quark pairs is the preferred search mode at CDF and DO in Run 11. In order to understand the optimal search strategy at CMS a series of Monte Carlo studies17 were done in order to identify the elements contributing to the mass error. For example, at low 13% , was found in events PT, Z -+ JJ, a fractional mass error of, dM/M without final state radiation (FSR). The contributions to the mass error are shown in Fig. 32. The total mass error is obtained by adding the errors shown in the “pie” chart in quadrature.

-

859

No high luminosity pileup contributes t o the errors indicated in Fig. 32.

Figure 33. Fractional mass error as a function of jet cone radius for crossings with and without the full LHC pileup of minimum bias events for decays of 2 with high PT into dijets.

Figure 32. Contributions to the fractional mass error for 2 decays into dijets (light quarks only). The contributions should be added in quadrature in order to find the total error.

The FSR is clearly the biggest effect. Ideally a cone could be chosen t o be large enough to contain the jet fragments. However, often the parton itself radiates wide angle and hard gluons prior to hadronization. This process is intrinsic t o the physics and can only be constrained by using very large cone radii. Increasing the cone size would reduce the effects of fragmentation/radiation at wide angles with respect to the jet axis. However, the cone size cannot increase too much or the fluctuations in the underlying event energy found within the cone become large. The plot of Fig. 32 refers to an optimized cone size, one which minimizes the overall mass resolution. At that cone radius the underlying event is the second largest error ( R 0.7) while the competing fragmentation effect is the third largest. Calorimeter resolution is a minor effect. At high luminosity (LHC) there is a minimum dM/M at a reduced cone radius, R 0.618because the pileup due to minimum bias events adds t o the underlying event energy. The optimization balancing jet fragments falling out of cone with inclusion of underlying event and pileup events energy and the accompanying fluctuations shifts to a smaller cone size. Pileup is small for “boosted” Z -+ JJ if a cone radius R 0.6 is used, as seen in Fig. 33. Note that, in this study there was no FSR, so that the mass resolution was only , dM/M 9% for “boosted” Z. Pileup is not the dominant effect. The dominant effect of FSR is illustrated in Fig. 34. Radiation is, indeed, N

N

N

N

860

Figure 34. Distribution of the ratio of the reconstructed dijet mass to the generated mass. a) no FSR, b) FSR turned

soft and collinear on average. However, there is multiple gluon emission and there is always a finite probability to emit a hard gluon at a large angle. The ratio of reconstructed to generated mass, MJJ/Mo, for dijets in CMS (Z decays) with and without FSR are shown. The dominant effect of FSR is clear. The d(M/Mo)/(M/Mo) r.m.s. in the fractional mass error rises from 11%to 19%, and the distribution shifts to smaller MJJ/MO due to radiation outside the jet cone. A radiative low mass tail also becomes evident. It is difficult in CMS to reduce the effects due to FSR by increasing the cone to include the radiated energy since that would include too much energy from the underlying event and pileup events.

-

5. Future Developments and New Physics

Clearly, future progress in calorimetry for high energy physics will be the result of a comprehensive program of R&D. Higher mass (see Fig. 1)will mean the use of higher luminosity which, in turn, means there is a need to address increased radiation damage and occupation (pileup). The fundamental 2 body production rate goes as square of mass as does the needed luminosity, L , in the best case which occurs when x is << 1, [ xf(x) const. 1. The basic Drell-Yan production of a state of mass M is described in Eq. 6. Clearly, since the typical x value of the distribution functions is, (x) M / & , the mass dependence of the needed luminosity can be much stronger, for example for gluon production L M2/[1 - M / a 1 2 .

-

N

-

861

2 -+ 1, "D - Y" 6

:

< 4nX2(2J + l), partial wave unitarity

J(6)dd = 7r2(2J + 1)(rI2/M),integrate mer narrow width

+

~2(d+y)~= =~c[.f1(.)zf2(.)1,=~[n2rl2(2~ 1 ) / ~ 1

-

I'lz/M

(6)

LL~12"

M2(dD/dY)y=0= q.11

(.).ti (.)1z=fi[~2w2(2J + 111

A plot of the luminosity required to produce 1000 events in one year in gluon - gluon fusion ( i.e. f(x) = g(x)) production of a state of mass, M = 1, 5 and 10 TeV, with electroweak coupling is shown in Fig. 35. Clearly, if there is new physics to be explored at high masses, luminosities in excess of that designed at the LHC will be required or the C.M. energy will need to be greatly increased. For masses C.M. energy, the required L rises rapidly and thus the C.M. energy increase is the most important. For masses << C.M. energy, L goes as the square of the mass. Therefore, even if the C.M. energy is increased without limit, the required luminosity at future accelerators must increase in order to address new physics issues at higher masses. Higher mass states or higher L in hadron colliders will require calorimetry which can withstand > 10 Mrad (ECAL) and > 2 Mrad (HCAL) for lql <3. N

~m'nosityNNeededfcri0M Eventsin i y r t b r g + p N (EW) PYPmductim \,''', \, ' . ' " " '

10

I

10"

\

-

M = 1 TeV M = 5 TeV M=lOTeV

$

I'

L 1

I

1 .

,6t. 10"

1o3

!,

'

*

, , ,~

1o4 C.M.Energy(GeV)

,

-..,;

10"

Figure 35. Plot of the required luminosity to produce masses of 1,5 and 10 TeV in gluongluon fusion production with electroweak coupling as a function of the C.M. energy of a p-p collider.

862 S t a n d a r d SCSNBl/BCFBiA(or ?7)

Figure 36. The light yield of plastic scintillator as a function of the radiation dose.

Figure 37. Cross section for 2 photon production of a state of mass M at the LHC. A formation width proportional to the fine structure constant is assumed.

These present limits do not change appreciably with C.M. energy as they are on the minimum bias “plateau”. However, hermiticity will require coverage to smaller angles as the C.M. energy increases and the “plateau” extends to smaller angles. Already the angular truncation is important at LHC as was shown above. Forward calorimetry will need to endure > 1 Grad. Note that the dose at small angles goes as 1/03, so that extension to smaller angles will lead to a major increase in radiation dose. Clearly R&D on new technologies is needed if new p-p colliders at higher C.M. energies are contemplated. What about scintillator technology? This technology will not survive in the endcap HCAL of CMS if the LHC L increases. A plot of the light losslg, with data taken as part of the SSC R&D program, as a function of dose is shown in Fig. 36. Clearly, the exponential loss of light due to color center formation means that a change of technology would be required if large increases in luminosity are planned. Just as an amusement, it is noted that at the LHC the beam synchrotron radiation plays a significant role in the heat load that must be taken out of the magnets. In a related observation, the proton is ultra-relativistic and can readily radiate photons. With two very small angle recoil tags, two photon physics can be studied in a p-p machine at LHC energies or higher. A cross section estimate is shown in Fig. 37.

863

'I

- *

a

1

Figure 38. Loss of gain with time and increase in the occupancy in the EM calorimetry of BaBar with increasing luminosityz0.

What about increases in luminosity for electron - positron colliders. In this case too, an increase in luminosity will also stress the calorimetry. In the case of ECAL CsI crystals there is light loss due t o radiation damage and increased occupation due to pileup. A 100 fold increase in L could not be tolerated by the present detector choices according to a BaBar study2'. The effects of light loss and pileup are shown in Fig. 38. The effects are such as to compromise the physics reach, thus requiring a new detector design if a luminosity of 1036/cm2sec is to be well utilized. Instead of increased luminosity in electron-positron colliders, consider increased C.M. energy. In this case, there have been many studies associated with the physics of the linear collider (LC). At the LC, cross sections with respect to LEP are down by large factors. Therefore, high L is going to be needed. For example, double Higgs (HH) production is important to study to explore HHH coupling. Because the cross section is small, that measurement requires a very large integrated luminosity. A set of interesting LC cross sections as a function of C.M. energy is shown in Fig. 39. The shaded portion of the plot corresponds to C.M. energies already covered by the LEP program. Studies of dijet mass resolution at the LC have indicated very good resolutions, better even than the LEP results shown in Fig. 10. As seen in Fig. 40, the American LC groups find that boosted Z have a fractional mass resolution 3% while jets have an energy error , dE/E 1 8 % / a . Energy of dM/M flow calorimetry achieves -3% mass resolution using very fine grained towers,

-

-

864

Q

Figure 39.

Cross section of representative processes at the LC as a function of C.M. energy.

-

-

of transverse size Xo, A, in the EM and hadronic compartments and many depth segments resulting in 3-d shower development information21. This result is to be compared to a stochastic coefficient for jets of -60% at the LHC, or a dijet mass resolution of -13%. The difference presumably resides in the ability at the LC to use very large cone sizes, unencumbered by the underlying event or the pileup events, In that case FSR and jet fragmentation effects can be largely reduced without paying any penalty. In this situation, the effects of calorimeter resolution and magnetic field sweeping become important. Since both these errors can be reduced by using tracking information, a big improvement in jet resolution is possible. Another example worked out by the Tesla groups is shown in Fig. 41. Note that much rides on the quality of the tracking. Spurious tracks can-

865

!ay-. .

, ... , .

.

, . ~. , -

* ?-

Figure 40.

.

-

I-

F B

.

\:

f

-

Study of Z mass resolution at the American Linear Collider21

Figure 41. Study of W and Z mass resolution by the Tesla groups using the energy flow te~hnique’~.

not be tolerated and inefficient track finding is also detrimental. It is also the case that the tracker of choice is that ideal object, the massless detector. 0.1 nuclear interaction length In comparison, the CMS detector contains of material, which leads to interactions of the hadrons in the tracker itself. Such interactions are rather detrimental to the “energy flow” algorithm, and will need to be tightly controlled. They clearly place limits on “energy flow” improvements in CMS.

-

6. Summary 0

0

Calorimetry will be increasingly important to HEP in the future as the energy frontier moves to higher masses. Detectors now being built or designed have made and will make im-

866

0

0

provements to the state of the art of calorimetry. Vigorous R&D is needed if progress is to continue. Intrinsic limits due to fluctuations in transverse position, longitudinal position, energy deposits, signal formation time and jet leakage will remain. Studies of higher mass states will require operation at yet higher luminosity which will put in premium on radiation resistance and high detector segmentation. A serious R&D program will be needed for the next generation of detectors. Calorimetry sufficient to do the physics should be the goal, not the “best” possible calorimetry.

References 1. CMS Collaboration, The Hadron Calorimeter Project, Technical Design Report, CERN/LHCC, 97-31. 2. V.V. Abramov et al., Nucl. Inst. €4 Meth. I n Phys. Res.A457, 175 (2001). 3. D. Green, The Physics of Particle Detectors, Cambridge University Press, 2000. 4. R. Wigmans, Annu. Rev. Nucl;. Part. Sci. 41, 133 (1991). 5. D. Groom et al., The European Physical Journal C15, 1 (2000). 6. CMS Collaboration, The Electromagnetic Calorimeter Project, Technical Design Report, CERN/LHCC, 97-33. 7. D. Glenzinski, U. Heintz, hep-e~/0007033July 18, 2000. 8. M. Karena et al., Report of the Tevatron Higgs Working Group, FermilabCONF-00/279-T. 9. D. Green, Calibration of the CMS Calorimeters, Fermilab FN-704, July 2001. 10. D. Green, E N E R G Y FLOW IN CMS CALORIMETRY, FERMILAB-FN0709, Dec 2001 11. U. Baur et al., QCD and Weak Boson Physics in Run 11, Fermilab-Pub001297, Dec. 2000. 12. D. Green, MINIMUM B I A S E V E N T S A N D MISSING E N E R G Y AT T H E LHC, FERMILAB-FN-707, Jul 2001. 13. Tesla Technical Design Report, http://TESLA.DESY.DE/ 14. D. Green, J E T S A N D ASSOCIATED MISSING E N E R G Y , FERMILAB-FN682, Sep 1999. 15. D. Fournier, Status of Calorimetry at LHC, this conference. 16. D. Green, Pulse formation in a Hybrid Photodetector, Fermilab - FN, Jan 1998. 17. A. Beretvas, D. Green, J. Marraffino, W. Wu., Dijet mass resolution at high luminosity in the CMS calorimeter, Fermilab - FN626, Oct 1994. 18. D. Green et al., CMS Note 1998/026, Feb 1998. 19. S.W. Han et al., Nucl.Instrum.Meth. A365 337 (1995). 20. W. Wisniewski, Considerations for Calorimetry at a Super B Factory, this conference. 21. Resource Book for Snowmass 2001, Fermilab - Pub’-01/058-E.

FUTURE EXPERIMENTS IN ASTROPHYSICS

JOHN F. KRIZMANIC Laboratory for High Energy Astrophysics USRA/NASA GSFC, Code 661, Greenbelt, Maryland 20771 E-mail: jfkOwsmicm.gsfc.nasa.gov The measurement methodologies of astrophysics experiments reflect the enormous variation of the astrophysical radiation itself. The diverse nature of the astrophysical radiation, e.g. cosmic rays, electromagnetic radiation, and neutrinos, is further complicated by the enormous span in energy, from the 1.95K relic neutrino background to cosmic rays with energy > 1020 eV. The measurement of gravity waves and search for dark matter constituents are also of astrophysical interest. Thus, the experimental techniques employed to determine the energy of the incident particles are strongly dependent upon the specific particles and energy range to be measured. This paper summarizes some of the calorimetric methodologies and measurements planned by future astrophysics experiments. A focus will be placed on the measurement of higher energy astrophysical radiation. Specifically, future cosmic ray, gamma ray, and neutrino experiments will be discussed.

1. Introduction

The measurement of the properties of the astrophysical radiation is one of the key components in understanding the underlying astrophysical phenomena. Unlike experiments at terrestrial accelerators, the specific properties of the astrophysical particle beam are usually not known. This translates into experimentally determining the energy, arrival direction, particle type, and arrival time of the incident radiation and possibly correlating these with measurements in other wavelength bands, e.g. optical or infrared. The exact experimental arrangement needed to perform these measurements is dependent upon nature of the radiation under study and on the energy scale of experimental interest. The situation can be further complicated by sources of background that can overwhelm the signal of interest or by attenuation effects of the atmosphere. For example, gamma ray measurements need to be performed above the atmosphere in orbiting experiments, which forces strict mass and power requirements on the detector systems. This paper discusses the future experiments that plan on performing cosmic ray, gamma ray, and neutrino measurements with a focus on higher energy. Nearly all of the these experiments employ calorimetric principles to determine

867

868

the energy of the incident particles. At high energies, calorimetry is the only technique available to measure the energy of the astrophysical radiation. As the general tendency of the flux of cosmic radiation is to decrease with increasing energy, there is a certain energy scale where it becomes prohibitively expensive in mass, volume, etc. to directly measure the radiation. Thus, astrophysical experimental configurations fall into two classes, those that directly measure the particles, usually from sub-orbital (balloon) or orbital (satellite) platforms, and those that use the atmosphere, Earth, or moon as the detection media. Each has its own unique set of technical challenges to overcome. In particular, the calibration of large-scale detectors at extreme energies requires input from Monte Carlo simulations as the energy is beyond any terrestrial particle accelerator test beam. Furthermore, the properties of the detection medium, e.g. the atmosphere, must be monitored and well-understood to accurately understand the energy resolution. Systematic effects on detector energy resolution can have devastating effects when measuring a steeply falling spectrum as is the case for many astrophysical observations. 2. Cosmic Ray Experiments

The measurement of the cosmic ray spectrum highlights the nuances of measuring the cosmic radiation. The differential flux of cosmic rays is shown in Figure 1 and displays a power law behavior over more than 11 orders of magnitude, from GeV to more than 10l1 GeV. Several features due to changes in the spectral index are evident in the figure. The first, at 1015 eV (knee), is identified by a steepening of the spectrum. The second, 10l8 eV (ankle), shows a recovery of the spectrum. The question of whether or not there is a feature at lo2' eV (big toe), is currently an issue and several future cosmic ray experiments are planned to make decisive measurements at this extreme energy. Figure 1 also illustrates the variation in cosmic ray integral flux at a particular energy: from 1 particle per m2-s at 100 GeV to 1 particle per km2century at 10'' GeV. It turns out that the energy below which direct cosmic ray measurements are feasible with current technology is approximately 1015 eV, i.e. the location of the knee. Above this energy, indirect measurements using the Earth's atmosphere as the detector medium are required in order to obtain an appreciable event rate. At the extreme energy of lo2' eV, even terrestrial experiments that employ the atmosphere as a vast target become rate limited. The possibility of using space-based experiments to monitor even larger atmospheric volumes offer the opportunity to increase the experimental sensitivity to the highest energy particles known in the universe. Astro-particle physics experiments that measure the lower end of the cos-

-

--

-

-

-

869

s

-

r\

,02L

-

h

k x

G

-1-

10

fluxes of Cosmic Roys

%

.a, -A

7

0,

<

(1 particle per m2-second)

'72

Figure 1. The differential flux of cosmic rays as a function of incident energy (Source: Simon Swordy).

mic ray spectrum require only a fraction of a m2-sr geometry factor to achieve a substantial event sample in several hours operation. This experimental size fits well into a balloon-borne payload. As an example, the CAPRICE instrument1 has had two successful flights performing measurements on the proton and helium spectra (0.4 - 200 GV), antiproton measurements (0.62 - 49 GV), elec-

870

trons and positrons (0.85 - 14 GeV), and atmospheric muons. The experimental configuration employs a detector suite using a superconducting magnetic spectrometer, RICH detector, time-of-flight (TOF) system, and a segmented calorimeter. The 48 x 48 cm2 imaging calorimeter is 7 X O and is constructed of interlaced planes of tungsten absorbers and silicon strip detectors. This configuration achieves an energy resolution of approximately 15%/&. The CAPRICE instrument is to have a dedicated flight to perform atmospheric muon measurements, as an aid in determined the atmospheric neutrino flux, in the 2003. Balloon-borne payloads offer the opportunity to develop technologies for incorporation in satellite-based experiments and the W-Si calorimeter used by CAPRICE is a prime example. This calorimeter has been further developed and is an integral part of the PAMELA experiment213that is scheduled to be launched in 2003. The experiment's scientific goals are the determination of the proton spectrum to 700 GeV, the electron spectrum to 400 GeV, positron spectrum to 270 GeV, the antiproton spectrum to 190 GeV, and perform searches for anti-nuclei. The experiment will be attached to a Russian Resurs-DK1 satellite and will be launched in an elliptical, quasi-polar orbit. The PAMELA detector suite includes a magnetic spectrometer with silicon strip detector tracker, a transition radiation detector, TOF detector, anti-coincidence detector, and a W-Si imaging calorimeter. The 24 x 24 cm2 calorimeter, which provides electron-hadron separation, is 16.3 X Oand 0.6 XI^^ with an energy resolution of 5 6% for electrons ( E , 2 25 GeV). Several other cosmic ray experiments using large area calorimeters are planned to be flown (and re-flown) in the future include the AMS-02 experiment4 which will be placed on the International Space Station and experiments that will be flown on long and ultra-long duration balloons, ATIC5 and CREAM'. The AMS-02 experiment is to be launched in 2005 and will include a superconducting magnetic spectrometer, transition radiation detector, RICH detector, time-of-flight system, and a 16 X O lead-scintillating fiber calorimeter in its detector suite. The AMS-02 experiment will perform charged-particle spectroscopy and anti-matter searches, extending these measurements to higher energies and sensitivities. The ATIC experiment, which uses a 18 X O segmented BGO calorimeter, a silicon charge detector, and scintillator detectors, has had a successful long-duration balloon flight (15 day) and plans another this year. The CREAM experiment plans an ultralong duration (loo+ day) balloon flight in 2003 and employs a 20 X O sampling tungsten-scintillating fiber calorimeter along with a timing-based scintillation charge detector and transition radiation detector. The science goals of ATIC and CREAM are to perform composition measurements of the cosmic ray spectrum to greater than 1014 eV. N

N

87 1

There has been recent development on realizing the goal of placing an instrument with a large calorimeter in orbit with the science goal of measuring the cosmic ray composition at the knee, 1015eV. Two versions of an instrument known as ACCESS7 have been proposed and each includes a deep, segmented, large-area calorimeter in their design. Under the ACCESS development program, prototype TRDs, charge detectors, and calorimeters have been tested in a variety of CERN test beam experiments along with detector modules from the PAMELA, ATIC, and CREAM experiments. Above the energy of approximately 1015 eV, it becomes extremely difficult to construct a direct measurement experiment that can have an appreciable cosmic ray event rate. However, the atmosphere can be used as a target and detection medium for cosmic ray induced airshowers. Two different methodologies are employed8. The first uses a sparse array of ground detectors t o measure the charged particle content of the traversing airshower. The energy of the airshower can be extracted from the measured signal at some distance from the shower core by comparing the results t o airshower simulations. The incident direction of the airshower can be obtained via detector timing. The second method, pioneered by the Flys Eye experimentg, uses the atmosphere as a calorimeter by measuring the near-UV nitrogen fluorescence excited by the traversing charged particles in the propagating airshower. A detector with sufficient spatial, angular, and temporal segmentation can image a large portion of the profile of the airshower, which facilitates energy, angular, and incident particle identification measurements. Air fluorescence detectors can only operate on dark, moonless nights and thus have 10% duty cycle as compared to the 100% of ground arrays. Furthermore, the transmission properties of the atmosphere must be well understood in order to accurately reconstruct the airshower energy. The ground array method has been employed by the KASCADE experimentlo (and others) to measure the cosmic ray spectrum from around 1015 eV. In addition, the ground array and air fluorescence methods allow the construction of detector systems with detection apertures large enough t o measure the cosmic ray spectrum to energies > 1020 eV. The results of the HiRes air fluorescence experiment" and the AGASA ground array experiment12 have indicated cosmic rays in excess of 1020 eV albeit without full agreement on the absolute flux. These events have fueled the mystery of their source as their energy places them above the Greisen-Zatsepin-Kuzmin (GZK) cutoff, caused by the interaction of ultra-high energy protons and the cosmic microwave background, and they do not point back t o any obvious local (< 100 Mpc) sources. In order to resolve the mystery of Ultra High Energy Cosmic Rays

-

872

(UHECR), several experiments are under construction or being developed. The first is the Pierre Auger Observatory13, currently under construction in Argentina, will employ a large ground array of detectors along with air fluorescence detectors. This hybrid measurement technique allows for the crosscalibration of a sample of airshower events using the two measurement techniques. The Auger ground array will contain 1600 water Cherenkov tanks spaced 1.5 km apart along with 4 air fluorescence telescopes. This arrangements leads to full efficiency at 1019 eV with an aperture of 7000 km2-sr, which is nearly an order of magnitude larger than currently running experiments. The aperture of the Auger array which uses the hybrid technique is 700 km2-sr. Simulation studies indicate that the angular resolution should be 1 deg with an energy resolution (at 1020 eV) of 25% for the ground array and 10% for the hybrid mode. The array is expected to be complete in 2005, and an engineering array, comprised of 40 water tanks and an air fluorescence detector, is currently in operation. Plans also exist for constructing an Auger site in the Northern Hemisphere that will have a similar aperture as the Southern site. There is also another ground-based experiment, the Telescope Array14, that is developing an array of air fluorescence detectors to yield a large aperture UHECR instrument. Ground-based experiments such as Auger push the limits on the size of a UHECR experiment. However, space-based experiments using the air fluorescence technique can substantially increase the detection aperture of UHECR15. The Extreme Universe Space Observatory (EUSO) 16, currently under European Space Agency (ESA) Phase A study, entails placing a large air fluorescence detector on the International Space Station. The experiment is comprised of a telescope with wide-angle optics that will collect and focus the near-UV air fluorescence signal on a segmented focal plane array. The baseline instrument design in an orbit of 400 km with a 10% duty cycle leads to an effective detection aperture of 50,000 km2-sr and an energy threshold of slightly more lo1' eV. Simulation studies indicate an energy resolution of better than 20% at lo2' eV and good angular resolution. EUSO is planned to be deployed 2009. The Orbiting Wide-angle Light-collectors (OWL) experiment17, currently under study by NASA, will employ two air fluorescence telescopes in quasiequatorial, 1000 km orbits. Each telescope will use large, wide-angle optics to image the air fluorescence signal onto a segmented focal plane array. The two telescopes will orbit with a nominal 500 km separation leading to stereo viewing of UHECR induced airshower events. Stereo reconstruction, as proved by the Fly's Eye and HiRes experiments, offers a powerful methodology in both reconstructing the events and understanding the properties of the atmosphere.

-

--

-

-

-

873

-

-

Simulation studies indicate an energy resolution 15% at lo2' eV and an angular resolution 1 deg. The event detection aperture, assuming a 10% duty cycle, is more than 200,000 km2-sr with an energy threshold slightly less than 1020 eV. Thus, the OWL mission offers the opportunity to obtain a substantial sample of trans-GZK events and unravel the mystery of the source(s) of the highest energy particles known in the Universe.

3. Gamma Ray Experiments The flux of astrophysical gamma rays exhibits an effect similar t o that observed in the cosmic ray spectrum: the flux decreases with increased energy. Aside from effects caused by the astrophysical generation mechanisms, the gamma ray flux is further depleted at higher energies, with a strong dependence on the redshift of the source, due to the interaction with the bath of I R photons in the Universe18. Thus, there is an energy where it becomes difficult to obtain an appreciable gamma ray event rate with direct detection. However, since high energy photons induce airshowers, the atmosphere can be used as a target and detection medium. For example, this phenomena is used t o measure high energy gamma rays by imaging the Cherenkov radiation induced by electromagnetic cascades in the atmosphere. The Swift missionlg, scheduled for launch in 2004, is designed t o make sensitive measurements of the Gamma Ray Burst (GRB) phenomena with arcsecond precision. The instrument includes a UV and optical telescope, an X-ray telescope (0.2 5 E7 5 10 keV), and a Burst Alert Telescope (BAT, 15 5 E7 5 150 keV). The BAT is comprised of 256 modules each with 128, 4 x 4 x 2 mm3 cadmium-zinc-telluride (CZT) detectors. The > 0.5 m2 CZT detector array images gamma rays that pass through a coded aperture mask with a 2 sr field of view. Laboratory measurements have demonstrated an energy resolution 5% at 60 keV. At higher energies, two missions are planned to improve upon the pioneering measurements provided by the EGRET instrument2' that was part of the Compton Gamma Ray Observatory. In particular, these missions will extend the energy range of gamma ray measurements. The first is the AGILE experiment21 which is t o be launched in 2003. This instrument is comprised of a pair-conversion telescope employing silicon strip detectors and tungsten absorbers followed by a shallow, 1.5 X O CsI calorimeter formed in two, segmented layers. The AGILE instrument will have a large, 3 sr field-of-view imaging gamma rays in energy range 30 MeV - 50 GeV. The GLAST mission22, to be launched in 2006, is to perform gamma ray astronomy with high sensitivity in the energy range 5 MeV - 300 GeV.

-

-

N

-

-

874

The large field-of-view (- 75% sky coverage every orbit) instrument consists of silicon microstrip detectors with lead absorbers to form a pair-conversion telescope with an on-axis area of over 2 m2, a gamma ray burst monitor, a segmented anti-coincidence system, and a CsI segmented calorimeter. The 9.5 X o ~ a l o r i m e t e will r ~ ~ be constructed in a modular arrangement mimicking that used for the silicon, pair-conversion telescope with each of the 16 modules containing 8 layers of CsI logs. Test beam and simulation results indicate energy resolutions of < 20% (5 - 100 MeV), < 10% (100 MeV -10 GeV), and < 6% (10- 300 GeV). A novel approach to perform gamma ray astronomy in the 100 keV 1 MeV range is offered by the Fresnel Lens Gamma Ray T e l e s c o ~ e ~This ~. visionary mission is based upon the principle that a several meter diameter phased-F'resnel lens constructed of aluminum can be employed to concentrate gamma rays on a segmented detector array leading t o orders of magnitude improvement in gamma ray sensitivity. Furthermore, angular resolutions of a micro-arcsecond can be achieved enabling the imaging of gamma rays emitted from the horizons of extragalactic black holes. This imaging capability in gamma rays will complement that in X-rays that will be provided by the MAXIM mission25. As the focusing of the lens is weak, the focal length of the telescope is of the order of lo6 km and requires the formation flying of two spacecraft, one with the lens and another with an array of gamma ray detectors. The detectors must have very good energy resolution, 1%,in order to exploit the full angular resolution potential of the lens which has energy dependent focusing. Above the energy of a few hundred GeV, it is difficult t o construct an instrument t o directly measure gamma rays with an appreciable event rate. However, on dark, quasi-moonless nights, the atmosphere can be used as a target and calorimeter by imaging the Cherenkov radiation generated by the induced electromagnetic cascade. The Cherenkov light fills in a cone with a characteristic angle of 1 deg and illuminates an area 50,000 m2. At ground level, the Cherenkov signal arrives in several nanoseconds with a strength of approximately 100 photons/m2 at 1 TeV. Pioneered by the Whipple experiment26, Imaging Airshower Cherenkov Telescopes (IACTs) use large mirrors t o collect this Cherenkov light and focus the photons onto a segmented detector. The incident shower direction is obtained from the orientation of the image in the detector while the shape of the image allows for rejection of hadronic showers. The intensity of the signal in the detector yields a measurement of the energy of the primary photon. One of the goals of the next generation IACTs is t o lower the gamma ray detection threshold to the level of 10's of GeV and thus provide an overlap with

-

N

-

-

875

the GLAST measurements. The Magic telescope27 will have a collecting area of N 240 m2 and thus will able to lower the energy threshold to < 15 GeV. The energy resolution is expected to be 10% at 1 TeV, and the angular resolution is anticipated to be 0.2 - 0.5 deg. Other experiments plan on constructing arrays of IACTs and employ stereoscopic imaging to improve on the angular resolution to < 0.1 deg. Included in this group are HESS28, C a n g a r ~ o - I I I ~ ~ , and Veritas30. The energy threshold for these arrays is given as N 50 GeV with an expected energy resolution of 10% at 1 TeV. Another methodology to measure high energy photons is to employ Solar collector farms (during dark nights) to harvest the Cherenkov light onto a segmented array of detectors. As the individual heliostats have collecting areas of 10’s of m2, the incorporation of a large number of the available heliostats in the measurement leads to a large effective collecting area and thus a lower energy threshold. Experiments currently using this technique include STACEE31, CELESTE32, GRAAL33, and the Solar Two O b ~ e r v a t o r y ~The ~. performance goals of these instruments are an energy threshold in the 10’s of GeV, angular resolution in the range 0.1 - 0.25 deg, and energy resolution of 25% at an energy 50 GeV. Gamma ray astronomy can also be performed by measuring the charged particles in the photon induced electromagnetic cascades. The benefit of this type of measurement is that a large area can be instrumented with 100% duty cycle detectors that can achieve large solid angle acceptance. Experiments such as M i l a g r ~and ~ ~ARGO-YBJ36 are located at high altitude where the charged particle density in the airshower is relatively large. Milagro is a large pond of water instrumented with two layers of photomultiplier tubes that detect the Cherenkov radiation from the charged particles in the shower. The area coverage is enhanced to 40,000 m2 with the incorporation of 170 “outrigger” Cherenkov water tanks. An energy threshold of N 200 GeV is eventually anticipated with an energy resolution of 30 - 50% for E7 > 1 TeV and an angular resolution < 1 deg. The ARGO-YBJ detector, with an energy threshold 100 GeV, will employ a single layer of RPC’s covering an area of 6500 m2.

-

-

-

-

-

4. Neutrino Experiments

The ability to measure the flux of extraterrestrial neutrinos offers a unique window to the astrophysical processes of the universe. Neutrinos are unaffected by magnetic fields and propagate virtually unattenuated. Thus, astronomy using neutrinos promises to expand the study of astrophysical phenomena, particularly in energy ranges where photon astronomy is limited due to absorptive effects, e.g. at energies 2 1 TeV. Neutrinos are difficult to measure because

876

of their small interaction cross sections and the existence of large backgrounds due to sources such as the much more numerous products of cosmic ray interactions. These lead to the need for large detector volumes to obtain an appreciable event rate and requirements to reduce the backgrounds such as locating the neutrino detector deep underground. eV (1.95K relic The energies of astrophysical neutrinos range from neutrinos) to > 1020 eV for neutrinos from the interaction of UHECR with the cosmic microwave background. Between these energy extremes are neutrinos from the Sun ( E , 5 15 MeV), neutrinos from supernovae (< E, >% 10’s of MeV), atmospheric neutrinos which follow the spectrum of the inducing cosmic rays, and high energy neutrinos with ( E , 2 100 GeV). In the arena of Solar neutrino measurements, the Super-Kamiokande experiment37 (water Cherenkov detector) is to be rebuilt, the Sudbury Neutrino O b ~ e r v a t o r y(heavy ~~ water Cherenkov detector) will continue to take data, and the ICARUS experiment39 (liquid argon TPC) is to be constructed. The next generation of solar neutrino experiments have a focus of reducing the neutrino energy threshold to well below 1 MeV in order to measure the monoenergetic neutrinos from the Be solar process and obtain better determination of the neutrino oscillation parameters40. The reduced energy threshold also allows for measurement of neutrinos from the p p process which is more insensitive to solar modeling. The BOREXINO experiment41 will use 300 tons of liquid scintillator as the neutrino target and detector to achieve an energy threshold of 250 keV and an energy resolution of 5% at 1 MeV. KAMLAND42 employs 1000 tons of liquid scintillator and plans to extend its primary science of a reactor neutrino oscillation search to solar neutrinos. Other experiments include the LENS experiment43 which will use Yb loaded liquid scintillator, HERON44using superfluid helium, and HELLAZ45which will employ a helium gas TPC. Galactic supernovae occur with a frequency of one every 10-30 years. Thus, experiments sensitive to the observation of neutrinos from supernovae need to be designed for long, many-year operation in order to guarantee an observation. Currently operating experiments such as Super-Kamiokande and SNO are sensitive to neutrinos from supernovae. In addition to the future Solar neutrino experiments that will have supernovae-neutrino sensitivity, multi-ton experiments have been proposed to use high-Z neutrino targets and detect the products, particularly neutrons, from the neutrino-nucleus interactions. The OMNIS experiment46 will use iron and lead as the target materials. The different energy thresholds for the various interaction channels leads to a methodology to distinguish between v, charged-current and neutral current interactions as well as ux neutral current interaction^^^. The event rate dependence of

-

-

-

877

the different interaction channels leads to a sensitivity to neutrino oscillations. OMNIS and another experiment of this type known as LAND could be located at a dedicated, underground neutrino laboratory. At energies 2 1 TeV, astrophysical objects such as Active Galactic Nuclei (AGNs)~*and Gamma Ray Bursts4’ could be a source of neutrinos via the decay of mesons produced from an accelerated hadronic component. However, detectors with 1 km3 of water-equivalent volume are required in order t o obtain an appreciable event rate50. At these energies, the neutrinos can be detected by measuring the Cherenkov radiation of the interaction products using the technique pioneered by the DUMAND51 and BAIKAL52 experiments. The ICECUBE experiment53 will expand upon the successful AMANDA array54 by deploying 80 strings totalling 4800 PMTs to instrument 1 km3 of Antarctic ice. The energy resolution from measurements of the muon from vcl interactions is expected t o be O.SIog(E,) with an angular resolution of 1 deg. For shower-type events, e.g. v, interactions, the energy resolution is expected to improve while the angular resolution will be degraded. Several experiments are planned to construct large, underwater neutrino telescopes with an eventual volume 1 km3. These include ANTARES55, NESTER56,and NEM057. These experiments have a nominal energy threshold of 1 TeV and a planned upgrade to the BAIKAL experiment would provide neutrino measurements at lower energies. As the Cherenkov light-scattering length is much longer in water than ice, the underwater experiments expect an angular resolution 0.5 deg for muons from vcl interactions. These experiments will also have a sensitivity to bursts of lower energy neutrinos such as ve from supernovae. At energies 2 1 PeV, the neutrino and antineutrino cross sections become virtually equivalent5*. Furthermore, the average energy of the lepton in a neutrino interaction is more than 70% of the initial neutrino energy, increasing to 80% at E, = lo2’ eV with the remaining energy given to a hadronic shower at the neutrino interaction point. These kinematics lead to an unique, “double-bang” signature for high energy, charged-current v, interactions where the produced r-lepton has a sufficient Lorentz boost to lead t o a separated shower induced by the r decay5’. Thus experiments, such as ICECUBE, have a methodology to detect v, events that could arise from the oscillations of astrophysical v p neutrinos. The importance of v, sensitivity is enhanced when considering the fact that the Earth attenuates neutrinos with energies 2 40 TeV, but v, can effectively propagate through the Earth, albeit with degraded energy, due to regeneration60. Neutrinos a t ultra-high energies, 10” eV, are expected from the decay of pions produced from the interaction of UHECR with the microwave background61, i.e the GZK effect. Furthermore, the speculative, “top-down”

-

-

-

-

-

N

N

-

-

878

processes, such as topological defects62 or Z - b u r ~ t sthat ~ ~ , are proposed t o be the source of the trans-GZK cosmic rays also predict hard neutrino spectra at ultra-high energies. The next generation UHECR experiments, Auger, Telescope Array, EUSO, and OWL, monitor a large atmospheric volume and thus have a sensitivity to ultra-high energy, neutrino induced airshowers. These events can be separated from the more numerous UHECR events by using the fact that neutrino-event interactions can occur much deeper in the atmosphere. Thus deep, horizontal airshowers offer a signature for ultra-high energy neutrino interactions. An ingenious method for detecting high energy u, interactions in the Earth (or moon) uses the radio Cherenkov signal generated by the subsequent electromagnetic shower64. Proposed by Askaryan, the intense coherent, Cherenkov radio pulse is caused by a 20% charge asymmetry in developing electromagnetic showers and the coherence leads to an energy dependent power enhancement. The Askaryan effect has been verified in a SLAC test beam65 and has been employed by the RICE array66 in Antarctica to search for high energy u, interactions in the ice and the GLUE experiment? which uses radio telescopes to search for ultra-high energy neutrino interactions in the moon. The strength of the radio Cherenkov technique is that for materials with good radio transmission properties, extremely large neutrino detectors can be instrumented. The use of 25 km3 salt domes is being investigated as well as flying a radio Cherenkov experiment on a long duration balloon over Antarctica. The latter experiment, ANITA64, would have lo6 km3 of ice as the neutrino fiducial volume and be sensitive to neutrinos above 1017 eV. The experiment is expected t o have an energy resolution of 1 at lo1* eV and an angular resolution of 10 deg. The potential performance combination of the relatively low energy threshold and large neutrino-detection aperture of these experiments surpasses the neutrino detection capabilities of other techniques for ultra-high energy neutrinos.

-

-

-

-

- -

5 . Summary

Future experiments in the area of cosmic rays will extend charged-particle spectroscopy measurements and provide a sensitive search for antimatter in the cosmic radiation (PAMELA, AMS) , provide elemental composition measurements t o 1015 eV (ATIC, CREAM, and possibly ACCESS), and try to unravel the mystery of the source of ultra-high cosmic rays (AUGER, Telescope Array, EUSO, OWL). Future gamma ray missions promise to study the phenomena of gamma ray bursts with superb angular resolution and sensitivity (SWIFT) and close the energy window that exists between 20 GeV and 1

-

879

TeV (GLAST, Air Cherenkov Telescopes, MILAGRO) while making dramatic improvements in angular resolution and sensitivity. The potential to perform high energy neutrino astronomy will be realized with the construction of km3 neutrino detectors (ANTARES, NESTER, ICECUBE). These combined with the neutrino measurement potentials of the UHECR experiments and experiments that employ the radio Cherenkov technique (ANITA) promise to open a new window to the astrophysical processes of the universe. References 1. http://www.roma2.infn.it/research/comm2/caprice/ 2. http: / /WiZard.roma2.infn.it/pamela/ 3. V. Bonvicini, these proceedings 4. R. Kossakowski and P. Maestro, these proceedings 5. J. Isbert and T.L. Wilson, these proceedings 6. 0. Ganel, these proceedings 7. http://lheawww.gsfc.nasa.gov/ACCESS/ 8. for an overview see http://hires.physics.utah.edu/background.html 9. http://hires.physics.utah.edu/flyseye.html 10. http://iklaul.fzk.de/KASCADEhome.html 11. http://hires.physics.utah.edu/ 12. http://www-aken0.icrr.u-tokyo.ac.jp/AGASA/ 13. A.K. Tripathi, these proceedings 14. http://www-ta.icrr.u-tokyo.ac.jp/ 15. J. Linsley, Proc. 19th ICRC (La Jolla), 3, 438 1985) 16. K. Arisaka, these proceedings 17. http://owl.gsfc.nasa.gov/ 18. F.W. Stecker, astro-ph/0010015 19. http://swift.gsfc.nasa.gov/ 20. http://lheawww.gsfc.nasa.gov/docs/gamcosray/EGRET/egret.html 21. http://agile.mi.iasf.cnr.it/Homepage/ 22. http://www-glast.stanford.edu/ 23. A. Chekhtman and R. Terrier, these proceedings 24. G.K. Skinner, Astronomy and Astrophysics, 375, 691 (2001) 25. http://maxim.gsfc.nasa.gov/ 26. http://egret.sao.arizona.edu/index.html 27. http://hegral.mppmu.mpg.de/MAGICWeb/ 28. http://www.mpi-hd.mpg.de/hfm/HESS/HESS.html 29. http://icrhpS.krr.u-tokyo.ac.jp/ 30. F. Krennrich, these proceedings 31. http://hep.uchicago.edu/-staceel 32. http://wwwcenbg.in2p3.fr/extra/Astroparticule/celeste/e-index.html 33. http: / /hegral .mppmu.mpg.de/GRAAL/ 34. http://solartwo.ucr.edu/solar2.html 35. http://www.lanl.gov/milagro/ 36. http://wwwl.na.infn.it/wsubnucl/cosm/argo/argo.html

880

37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68.

http://www.phys.washington.edu/-superk/ http://www.sno.phy.queensu.ca/ http://www.aquila.infn.it/icarus/ http://www.sns.ias.edu/Nj.b/ http://almime.mi.infn.it/ http://www.awa.tohoku.ac.jp/html/KamLAND/index.html http :/ flens.in2p3.fr/ http://www.physics.brown.edu/research/cme/heron/index.html http://sgl.hep.fsu.edu/hellaz/ http://www.physics.ohio-state.edu/OMNIS/ J.J. Zach et al., NIM A484, 194 (2002) F. Stecker and M. Salamon, Space Sci. Rev. 75, 341 (1996) E. Waxman and J. Bahcall, Phys.Rev.Let. 78, 2292 (1997) F. Halzen, astro-ph/9605014 A. Roberts, Rev.Mod.Phys. 64, 259 (1992) http://www.ifh.de/baikal/baikalhome.html J. Lamoureux, these proceedings http://amanda.berkeley.edu/amanda/amanda.html http://antares.in2p3.fr/ http://www.uoa.gr/Nnestor/ http://nemoweb.lns.infn.it/ R. Gandhi et al., Phys.Rev. D58, 093009 (1998) J.G. Learned and S. Pakvasa, hep-phf9408296 F. Halzen and D. Saltzberg, Phys.Rev.Lett. 81, 4305 (1998) R. Engel et al., Phys.Rev. D64 093010 (2001) G. Sigl et al., Phys.Rev. D59, 043504 (1999) T.J. Weiler, Astropart.Phys. 11, 303 (1999) D. Saltzberg, these proceedings D. Saltzberg et al., Phys.Rev.Lett. 86, 2802 (2001) S. Razzaque, these proceedings http://www.physics.ucla.edu/Nmoonemp/public/index.html D. Saltzberg, these proceedings

Conference Pictures

This page intentionally left blank

AUTHOR INDEX

Abbiendi, G., 287 Adams, J. H., 89 Adamson, P., 428, 436 Ahn, H. S., 89, 133 Akchurin, N., 521 Akgun, U., 521 Alford, R., 133 Alner, J., 428, 436 Aloisio, A., 388 Ambrosino, F., 388 Anderson, B., 428, 436 Andersen, V., 95 Antonelli, A., 388 Antonelli, M., 388 Aspell, P., 621 Atramentov, O., 497 Attal, A., 557 Attree, D., 428, 436 Auffray, E., 240 Ayan, S., 521

Bencivenni, G., 388 Benen, A., 549 Bertolucci, S., 388 Besson, A., 679 Bini, C., 388 Bloch, P., 621 Bloise, C., 388 Bocci, V., 388 Boezio, M., 101 Bonvicini, V., 101 Borkar, S., 621 Bossi, F., 388 Boudry, V., 652 Branchini, P., 388 Breidenbach, M., 304 Breton, D., 627 Brient, J.-C., 309, 747 Bruecken, P., 521 Bulychjov, S. A,, 388 Cadoux, F., 108 Caloi, R., 388 Campana, P., 388 Capon, G., 388 Carboni, G., 388 Carminati, F., 95 Casarsa, M., 388 Casavola, V., 388 Case, G., 89 Cataldi, G., 388 Cavallari, F., 223 Cavalli-Sforza, M., 531 Ceradini, F., 388 Cervelli, F., 108, 114, 388 Cevenini, F., 388

Biischer, M., 215 Bacci, C., 388 Bacelar, J., 215 Bailly, Ph., 652 Bamberger, A., 806 Barbi, M., 401 Barker, M., 428, 436 Barney, D., 621 Barrelet, E., 652 Bashindzhagyan, G., 89 Bassler, U., 413 Beatty, J. J., 133 Belias, A., 428, 436 Benchekroun, D., 331

889

890

Chambert-Hermel, V., 108 Chang, J., 89 Chekanov, S., 806 Chekhtman, A., 121, 127 Chen, G., 108, 114 Chen, H., 108, 114 Chiche, R., 652 Chiefari, G., 388 Choi, M. J., 133 Christl, M., 89 Ciambrone, P., 388 Coignet, G., 108, 114 Collot, J., 274 Conetti, S., 388 Coutu, S., 133 Crone, G., 428, 436 Currrat, C. A., 345 Danagoulian, S., 479 De La Taille, Ch., 652 De Lucia, E., 388 De Robertis, G., 388 De Simone, P., 388 De Zorzi, G., 388 Deiters, K., 231 Dell’Agnello, S., 388, 563 Deng, Q., 190,491 Denig, A., 388 Di Domenico, A., 388 Di Donato, C., 388 Di Falco, S., 108, 114, 388 Djannati-Atai, A., 127 Doria, A., 388 Doring, W., 215 Drake, G., 806 Dreucci, M., 388 Dubois, J. M., 108 Dumanoglu, I., 521 Durkin, T., 428, 436 Duval, P.-Y., 665 Duvernois, M. A., 133

Elias, J., 605 Empl, A., 95 Erriquez, O., 388 Eschrich, I. G., 658 Eskut, E., 521 Falchini, E., 108, 114 Falk, E., 428, 436 Farilla, A., 388 Fasso, A., 95 Fazely, A. R., 89 Felici, G., 388 Felt, N., 428, 436 Fenyvesi, A., 521 Ferrari, A., 95, 388 Ferrer Ribas, E., 613 Ferrer, M. L., 388 Fincke-Keeler, M., 712 Finocchiaro, G., 388 Forti, C., 388 Fougeron, D., 108 Fouque, N., 108 Fournier, D., 17 Franceschi, A., 388 Franzini, P., 388 Frey, R. E., 54, 304 Freytag, D., 304 Futo, E., 95 Gottlicher, P., 296 Gallin-Martel, M. L., 274 Ganel, O., 89, 133 Gasparian, A., 208 Gataullin, M., 385 Gatti, C., 388 Gauzzi, P., 388 Giovannella, S., 388 Girard, L., 108, 114 Go, A., 621 Gorini, E., 388 Goy, C., 108, 114

891

Grahl, J., 231 Grancagnolo, F., 388 Granger, D., 89 Graziani, E., 388 Green, D., 837 Grove, J. E., 127 Gunasingha, R., 89 Guzik, T. G., 89 Haller, G., 304 Hamer, A., 442 Han, S. W., 388 Han, Y. J., 89 Hanson, K. D., 452 Harris, P., 428, 436 Hejny, V., 215 Henriques, A., 532 Hermel, R., 108 Hoek, M., 215 HofEmann, D., 665 Hryn'ova, T., 175 Hu, T., 459 Huang, H.-C., 161 Huffer, M., 304 Incagli, M., 388 Ingram, Q., 231 Ingrosso, L., 388 Isbert, J., 89, 95 Jan, S., 274 Jenner, L., 428, 436 John, M., 127 Johnson, W. N., 127 Jones, G . M., 728 Karpetian, G., 331 Katta, S., 544 Kayis-Topaksu, A., 521 Kellogg, R. G., 287 Kiiskinen, A., 798 Kim, H. J., 89, 133

Kim, K. C., 89, 133 Kim, S. K., 89, 133 Kiryunin, A. E., 331, 354, 720 Kish, J., 331, 354 Kistenev, E., 584 Kluge, W., 388 Koca, N., 521 Koch, H. R., 215 Kochetkov, V., 584 Kocian, M., 167 Korbel, V., 591 Kordas, K., 331 Kordosky, M., 428, 436 Korolko, I., 584 Korolkov, I., 538 Kossakowski, R., 108, 114 Kouznetsov, E., 89 Krane, J., 786 Krennrich, F., 139 Krizmanic, J . F., 867 Kronqvist, I., 231 Kuhlmann, S., 806 Kulikov, V., 388 Kunori, S., 375 Kuo, C., 388 Kuznetsov, A., 231 Lohner, H., 215 Lacava, F., 388 Lalwani, S., 621 Lanfranchi, G., 388 Lang, K., 428,436 Lebbolo, H., 652 Lebedev, A., 428, 436 Lecoq, P., 262 Lee, K., 95 Lee, M. H., 133 Lee, R., 428, 436 Lee-F'ranzini, J., 388 Leltchouk, M., 331 Leone, D., 388

892

Li, Z. K., 491 Liao, J. Y., 190 Lieunard, B., 108 Litvin, V., 325 Liu, D. T., 469 Liu, L., 133 Liu, Z., 108, 114 Lobban, O., 421, 814 Loch, P., 331, 354 Lofstedt, B., 621 Lomtadze, T., 108, 114 Longley, N., 428, 436 Lu, F., 388 Lu, Y., 108, 114 Lutz, L., 133 Miiller, S., 388 Machner, H., 215 Maestro, P., 108, 114 Magill, S., 806 Majatsky, I., 584 Makonyi, K., 521 Malinine, A., 133 Mao, R. H., 190 Marrocchesi, P. S., 108, 114 Marshak, M., 428, 436 Martemianov, M., 388 Martin, F., 607 Martin, P., 274 Matsyuk, M., 388 Mazini, R., 331, 354 Mei, W., 388 Melnikov, E., 584 Merlo, J. P., 521 Merola, L., 388 Messi, R., 388 Michael, D., 428, 436 Minard, M. N., 739 Minnick, S. A., 133 Miramonti, L., 570 Miscetti, S., 388

Miyabayashi, K., 394 Miyagawa, P., 428, 436 Mocchiutti, E., 101 Moreau, F., 652 Morgunov, V. L., 70 Morse, R., 428, 436 Moses, W. W., 251 Moulson, M., 388 Mualem, L., 578 Murtas, F., 388 Musgrave, B., 806 Musienko, Y., 231 Musser, J., 428, 436 NgmeEek, S., 538 Nakamura, I., 793 Napolitano, M., 388 Naqvi, S. A., 89 Nedosekin, A., 388 Negroni, S., 331 Nelson, C. A., 644 Nevski, P., 584 Newman, H., 325 Nguyen, F., 388 Nichol, R., 428, 436 Nicholls, T., 428, 436 Novotny, R., 215 Noyak, D., 521 Nutter, S., 133 Oliver, J., 428, 436 Onel, Y., 504, 521 Onengut, G., 521 Paganini, P., 339 Palutan, M., 388 Panasyuk, M., 89 Panov, A., 89 Paoletti, R., 108, 114 Paoluzi, L., 388 Park, I. H., 133 Parnell, T., 87

893

Parrour, G., 331 Parua, N., 687 Pasqualucci, E., 388 Passalacqua, L., 388 Passeri, A., 388 Patera, V., 388 Pearce, G., 428, 436 Peisert, A., 621 Petrolo, E., 388 Petyt, D., 428, 436 Pilo, F., 108, 114 Pinsky, L., 95 Polatoz, A., 521 Pontecorvo, L., 388 Pretzl, K., 3 Price, B., 89 Primavera, M., 388 Proga, M., 428, 436 Proudfoot, J., 806

Qu, X. D., 190 Ranft, J., 95 Razzaque, S., 515 Rebel, B., 428, 436 Ren, G. H., 491 Renard, Ch., 652 Renker, D., 231 Repond, J., 806 Reucroft, S., 231 Reynaud, S., 621 Rodier, S., 695 Rosier-Lees, S., 108, 114 Ruggieri, F., 388 Rusack, R., 231 Russell, J. J., 304 Saakyan, R., 428, 436 Sajot, G., 679 Sakhelashvili, T., 231 Sala, P., 95

Salihagic, D., 331, 354 Samsonov, G., 89 Santangelo, P., 388 Santovetti, E., 388 Saracino, G., 388 Savine, A., 781 Schacht, P., 677 Schamberger, R. D., 388 Schiavon, P., 101 Schindhelm, E., 133 Schmidt, I., 521 Schmidt, W. K. H., 89 Schwanenberger, C., 761 Scian, G., 101 Sciascia, B., 388 Sciubba, A,, 388 Scuri, F., 388 Seez, C., 323 Seligman, W., 331 Seo, E. S., 89, 133 Serin, M., 521 Seyfarth, H., 215 Sfiligoi, I., 388 Shaw, T. M., 644 Shen, D. Z., 190, 491 Shevchenko, S., 325 Shwartz, B. A,, 182 Sina, R., 89 Singovski, A., 231 Smith, C., 428, 436 Sokolskaya, N., 89 Soukharev, A., 331 Spadaro, T., 388 Specka, A. E., 652 Spiriti, E., 388 Sriharan, A., 814 Stanek, R.,806 Stewart, M., 89 Stroher, H., 215 Striienec, P., 331, 354

894

Strom, D., 285, 287 Sullivan, P., 428, 436 Swain, J., 231 Swordy, S. P., 43, 133 Tonnesmann, M., 773 Terrier, R., 127 Thomas, J., 428, 436 Tong, G. L., 388 Torii, H., 409 Tortora, L., 388 Tournefier, E., 274 Tripathi, A. K., 151 Turini, N., 108, 114 Vacchi, A., 101 Vahle, P., 428, 436 Valente, E., 388 Valente, P., 388 Valeriani, B., 388 Valle, G., 108, 114 Vallee, C . , 665 Vallereau, A., 652 Varanda, M. J., 361 Venanzoni, G., 108, 114, 388 Veneziano, S., 388 Ventura, A., 388 Vialle, J. P., 108, 114 Videau, H., 309, 747 Voronin, A., 89 Wang, J. Z., 89, 133 Weber, A., 428, 436

Wefel, J. P., 89, 95 Wei, Q., 469 White, S., 489 Wielers, M., 367 Wigmans, R., 814 Wilson, T., 95 Wing, M., 767 Wisniewski, N., 325 Wojcicki, S., 428, 436 Woody, C . , 249, 584 Wronska, A., 215 Wu, J., 89, 133 Xia, L., 459 Xu, G., 388 Yan, D. S., 190 Yin, Z. W., 190 Yoshida, R., 806 Yu, G. W., 388 Yu, Z., 108, 114 Zampa, G., 101 Zampa, N., 101 Zatsepin, V., 89 Zerwas, D., 703 Zeyrek, M., 521 Zhang, L., 190, 469 Zhu, K., 469 Zhu, R. Y., 190, 459, 469 Zhuang, H., 108, 114

LIST OF PARTICIPANTS ABBIENDI

Giovanni

INFN

Bologna

Italy

ARISAKA

Katsushi

UCLA

USA

ATRAMENTOV

Oleksiy

Iowa State Univ.

USA

ATTAL

Alon

UCLA

USA

AUFFRAY

Etiennette

CERN

Switzerland

BAMBERGER

Andreas

ANL/Freiburg Univ.

USA

BARB1

Mauricio

McGill Univ.

Germany

BASHMAKOV

Yuriy

P.N. Lebedev Physical Inst.

Russia

BASSLER

Ursula

LPNHE

France

BENEN

Arno

Univ. of Freiburg

Germany

BESSON

Auguste

I.S.N.

France

BLANCHARD

Rick

Hamamatsu

USA

BOEZIO

Mirko

INFN - Trieste section

Italy

BONVICINI

Valter

INFN

Italy

-

-

Sezione di Trieste

BOUDRY

Vincent

CNRS/INSP3

France

BRETON

Dominique

LAL Orsay

France

CAVALLARI

Francesca

INFN

Switzerland

C AVALLI-SFORZA

Matteo

IFAE - Barcelona

Spain

CHEKHTMAN

Alexandre

NRL/GMU

USA

CHOUDHARY

Brajesh

Caltech

USA

COLAS

Jacques

LAPP

France

CURRAT

Charles

LBNL

USA

DANAGOULIAN

Samuel

NC A&T State Univ.

USA

DELL’AGNELLO

Simone

INFN

Italy

DENG

Shanghai Inst. of Ceramics

P.R. China

DI CREDICO

Qun Alessandra

LNGS/INFN

Italy

DYCHKANT

Alexandree

Northern Illinois Univ.

USA

EIGEN

Gerald

U. Bergen

Norway

ELIAS

John

Fermilab

USA

ELLIOT

Lipeles

Caltech

USA

EREDITATO

Antonio

INFN Napoli

Italy

FABBRI

Franco

Frascati - INFN

Italy

895

896 F E R R E R RIBAS

Esther

SPP-CEA-Saclay

France

FINCKE-KEELER

Margret

Univ. of Victoria

Canada

FLYCKT

Esso

Photonis

France France

FOURNIER

Daniel

LAL-Orsay

FREY

Raymond

Univ of Oregon

USA

GANEL

Opher

Univ. of Maryland

USA

G ASPARIAN

Ashot

NC A&T State Univ.

USA

GATAULLIN

Marat

Caltech

USA

GATTI

Claudio

Universita’ Pisa and INFN

Italy

GO

Apollo

CERN

Switzerland

GOETTLICHER

Peter

DESY

Germany

GOUGH ESCHRICH

Ivo

Imperial College

USA

GREEN

Daniel

Fermilab

USA

HAMER

Andre

Los Alamos National Lab.

USA

HANSON

Kael

Univ. of Pennsylvania

USA

HENRIQUES

Ana

CERN

Switzerland

HITLIN

David

Caltech

USA

HOFFMANN

Dirk

DESY

Germany

HRYN’OVA

TETIANA

SLAC

USA

HUANG

Hsuan-Cheng

National Taiwan Univ.

Rep. of China Italy

INTROZZI

Gianluca

Univ. of Pavia - Italy

ISBERT

Joachim

Louisiana State Univ.

USA

JOHANN

Collot

ISN

France

JONES

Gary

Caltech

USA

JOO

Kyung Kwang

U of Toronto

Canada

KATSAVOUNIDIS

Erik

MIT

USA

KATTA

SUDHAKAR

TATA INSTITUTE

India

KIISKINEN

Ari

Helsinki Inst. of Physics

Switzerland

KIRY UNIN

Andrei

Max-Planck-Inst. for Physics

Germany

KOCIAN

Martin

SLAC

USA

KORBEL

Volker

DESY

Germany

KOSSAKOWSKI

Roman

L A P P - IN2P3

France

KRANE

John

Iowa State Univ.

USA

KRENNRICH

Frank

Iowa State Univ.

USA

KRIZMANIC

John

USRA/NASA/GSFC

USA

a97 KUNORI

Shuichi

U. of Maryland

USA

LAMOUREUX

Jodi

LBNL

USA

LECOMTE

Pierre

ETHZ

Switzerland

LECOQ

Paul

CERN

Switzerland

LIAO

Jingying

Shanghai Inst. of Ceramics

P.R.China

LOBBAN

Olga

Texas Tech Univ.

USA

LOCH

Peter

Univ. of Arizona

USA

LOKAJICEK

Milos

Acad. of Sci. Prague

Czech Republic

LOS

Sergey

Fermilab

USA

LUDWIG

Jens

Univ. Freiburg

Germany

MAGILL

Stephen

Argonne National Lab.

USA

MA10

Amelia

FCUL and LIP/Lisbon

Portugal

MANSOULIE

Bruno

DAPNIA-SACLAY

France

MA0

Rihua

Caltech

USA

MARTIN

Franck

L P C Clermont Ferrand

France

MELCHER

Charles

CTI, Inc.

USA

MICHAEL

Douglas

C alt ech

USA

MINARD

Marie-Noelle

LAPP

France

MIRAMONTI

Lino

Milano Univ.

Italy

MIYABAYASHI

Kenkichi

Nara Women’s Univ.

Japan

M 0RG U N OV

Vasily

DESY-ITEP

Germany

MOSES

William

LBL

USA

MUALEM

Leon

Univ. of Minnesota

USA

NAKAMURA

Isamu

Univ. of Pennsylvania

USA

NELSON

Charles

Fermilab

USA

NEMECEK

Stanislav

Acad. Sci. Prague

Czech Republic

NICHOL

Ryan

Univ. College London

UK

NOVOTNY

Rainer

Univ. Giessen

Germany

OBERLACK

Horst

MPI fuer Physik

Germany

ONEL

Yasar

Univ. of Iowa

USA

PAGANINI

Pascal

LLR Ecole Polytechnique

France

PAOLO

Maestro

INFN-Pisa & Univ. of Siena

Italy

PAPPAS

Stephen

C a1t ech

USA

PARNELL

Thomas

Univ. of AL in Huntsville

USA

PARUA

Nirmalya

SUNYQ Stony Brook

USA

898 PORTER

Frank

Caltech

USA Switzerland

PRETZL

Klaus

Univ. Bern

RAVEL

Akhmetshin

Novosibirsk INP

Russia

RAZZAQUE

Soebur

Univ. of Kansas

USA

RODIER

Stephane

UAM

Spain

RUSACK

Roger

The Univ. of Minnesota

USA

SALTZBERG

David

UCLA

USA

SAVINE

Alexandre

Univ. of Arizona

USA

SCHACHT

Peter

Max-Planck-Inst. fur Physik

Germany

SCHWANENBERGER

Christian

DESY

Germany

SEEZ

Chris

Imperial, London

UK

SHEN

Dingzhong

Shanghai Inst. of Ceramic

P.R. China

SHEVCHENKO

Sergey

Caltech

USA

SHWARTZ

Boris

Budker INP

Russia

SIMON

Swordy

Univ. of Chicago

USA USA

SKUJA

Andris

Univ. of Maryland

STROM

David

Univ. of Oregon

USA

TERRIER

Regis

P C C Collbge de France

France

TERRON

Juan

Universidad Autonoma Madrid

Spain

TOENNESMANN

Matthias

MPI Munich

Germany

TORI1

Hisayuki

Kyoto Univ./RIKEN

USA

TRIPATHI

Arun

UCLA

USA

TU

Qinwei

Shanghai Inst. of Ceramics

P.R. China

VAHLE

Patricia

Univ. of Texas a t Austin

USA

VARANDA

Maria

LIP

Portugal

VIDEAU

Henri

LLR Ecole polytechnique/IN2PB

France

WE1

Qing Sebastian

Caltech

USA

WHITE

BNL

USA

WIELERS

Monika

TRIUMF

Canada

WIGMANS

Richard

Texas Tech Univ.

USA

WILKINSON

Richard

Caltech

USA

WILSON

Thomas

NASA, Johnson Space Center

USA

WING

Matthew

Bristol Univ., UK

Germany

WISNIEWSKI

William

SLAC

USA

WOODY

Craig

BNL

USA

899

wu

Weimin

XIA

Lei

Caltech

USA

YOSHIDA

Rik

Argonne National Lab.

USA

YOSHIMURA

Sam

Kuraray America, Inc.

USA

ZERWAS

Dirk

LAL-Orsay

fiance

ZHANG

Liyuan

Caltech

USA

ZHOU

Li

IHEP Beijing

P.R. China

ZHU

Ren-yuan

Caltech

USA

ZUTSHI

Vishnu

Northern Illinois Univ.

USA

Fermilab

USA

Frontiers of particle physics. Proceedings of the tenth Lomonosov conference on elementary particle physics

3rd International Symposium on Quality Electronic Design: Proceedings 2002 March 18-21, 2002, San Jose, California

Calorimetry in Particle Physics: Proceedings of the Tenth International Conference, California, USA 25-29 March 2002

Frontiers of particle physics. Proceedings of the tenth Lomonosov conference on elementary particle physics

3rd International Symposium on Quality Electronic Design: Proceedings 2002 March 18-21, 2002, San Jose, California

Mathematical Foundations of Programming Semantics: 7th International Conference, Pittsburgh, PA, USA, March 25-28, 1991. Proceedings

Evolutionary Programming VII: 7th International Conference, EP'98, San Diego, California, USA, March 25-27, 1998: Proceedings

Particle Physics on the Eve of LHC: Proceedings of the Thirteenth Lomonosov Conference on Elementary Particle Physics

Particle Physics at the Year of Astronomy: Proceedings of the Fourteenth Lomonosov Conference on Elementary Particle Physics

Particle Physics in Laboratory, Space And Universe: Proceedings of the Eleventh Lomonosov Conference on Elementary Particle Physics

Particle Physics And Cosmology. TASI 2002

Recent Advances in Visual Information Systems: 5th International Conference, VISUAL 2002 Hsin Chu, Taiwan, March 11-13, 2002. Proceedings

Astroparticle, Particle And Space Physics, Detectors And Medical Physics Applications (Proceedings of the 9th Italian Conference)

Hybrid Systems: Computation and Control: 5th International Workshop, HSCC 2002, Stanford, CA, USA, March 25-27, 2002, Proceedings

Spin in Particle Physics

International Mathematical Conference 1982: Proceedings

Concepts of particle physics

Mathematical Results In Quantum Physics: Proceedings of the QMath11 Conference

Mathematical Results In Quantum Physics: Proceedings of the QMath11 Conference

Diagrammatic Representation and Inference: Second International Conference, Diagrams 2002 Callaway Gardens, GA, USA, April 18-20, 2002 Proceedings

Particle Physics

Particle Physics

Dynamics in logistics first international conference; proceedings

Advances in Information Systems: Second International Conference, ADVIS 2002, Izmir, Turkey, October 23-25, 2002. Proceedings

Handbook of particle physics

Handbook of Particle Physics

Concepts of particle physics

Particle Physics

Physics at the Japan Hadron Facility: Proceedings of the Workshop Adelaide, Australia, 14-21 March 2002

Financial Cryptography: 6th International Conference, FC 2002, Southampton, Bermuda, March 11-14, 2002, Revised Papers

Handbook of Particle Physics

COTS-Based Software Systems: First International Conference, ICCBSS 2002, Orlando, FL, USA, February 4-6, 2002, Proceedings

Particle Physics

Calorimetry in Particle Physics: Proceedings of the Tenth International Conference, California, USA 25-29 March 2002

Frontiers of particle physics. Proceedings of the tenth Lomonosov conference on elementary particle physics

3rd International Symposium on Quality Electronic Design: Proceedings 2002 March 18-21, 2002, San Jose, California

Mathematical Foundations of Programming Semantics: 7th International Conference, Pittsburgh, PA, USA, March 25-28, 1991. Proceedings

Evolutionary Programming VII: 7th International Conference, EP'98, San Diego, California, USA, March 25-27, 1998: Proceedings

Particle Physics on the Eve of LHC: Proceedings of the Thirteenth Lomonosov Conference on Elementary Particle Physics

Particle Physics at the Year of Astronomy: Proceedings of the Fourteenth Lomonosov Conference on Elementary Particle Physics

Particle Physics in Laboratory, Space And Universe: Proceedings of the Eleventh Lomonosov Conference on Elementary Particle Physics

Particle Physics And Cosmology. TASI 2002

Recent Advances in Visual Information Systems: 5th International Conference, VISUAL 2002 Hsin Chu, Taiwan, March 11-13, 2002. Proceedings

Astroparticle, Particle And Space Physics, Detectors And Medical Physics Applications (Proceedings of the 9th Italian Conference)

Hybrid Systems: Computation and Control: 5th International Workshop, HSCC 2002, Stanford, CA, USA, March 25-27, 2002, Proceedings

Spin in Particle Physics

International Mathematical Conference 1982: Proceedings

Concepts of particle physics

Mathematical Results In Quantum Physics: Proceedings of the QMath11 Conference

Mathematical Results In Quantum Physics: Proceedings of the QMath11 Conference

Diagrammatic Representation and Inference: Second International Conference, Diagrams 2002 Callaway Gardens, GA, USA, April 18-20, 2002 Proceedings

Particle Physics

Particle Physics

Dynamics in logistics first international conference; proceedings

Advances in Information Systems: Second International Conference, ADVIS 2002, Izmir, Turkey, October 23-25, 2002. Proceedings

Handbook of particle physics

Handbook of Particle Physics

Concepts of particle physics

Particle Physics

Physics at the Japan Hadron Facility: Proceedings of the Workshop Adelaide, Australia, 14-21 March 2002

Financial Cryptography: 6th International Conference, FC 2002, Southampton, Bermuda, March 11-14, 2002, Revised Papers

Handbook of Particle Physics

COTS-Based Software Systems: First International Conference, ICCBSS 2002, Orlando, FL, USA, February 4-6, 2002, Proceedings

Particle Physics

Recommend Documents