1:/j

I,LBL-15973

UC-37CONF-830224

PROCEEDI NGS

OF THE

1983 DPF WORKSHOP

ON

Collider Detectors:Present Capabilities

AndFuture Possibilities

February 28 - March 4, 1983

Lawrence Berkeley LaboratoryUniversity of California

Berkeley, California 94720

Prepared for the U.S. Department of Energy under Contract DE-AC03-76SF00098 and for theHigh Energy Physics Section of the National Science Foundation.

LEGAL NOTICE

This book was prepared as an ac/::ount of worksponsored by an agency of the United StatesGovernment. Neither the United States Govern­ment nor any agency thereof, nor any of theiremployees, makes any warranty, express or im­plied, or assumes any legal liability or responsibilityfor the accuracy, completeness, or usefulness ofany information, apparatus, product, or processdisclosed, or represents that its use would notinfringe privately owned rights. Reference hereinto any specific commercial product, process, orservice by trade name, trademark, manufacturer,or otherwise, does not necessarily constitute orimply its endorsement, recommendation, or favor­ing by the United States Government or any agencythereof. The views lUld opinion~ of authors px­pressed herein do not necessarily state or reflectthose of the United States Government or anyagency thereof.

Printed in the United States of AmericaAvailable from

National Technical Information ServiceU.S. Department of Commerce5285 Port Royal RoadSpringfield, VA 22161Price Code: AID

Lawrence Berkeley Laboratory is an equal opportunity employer.

)

•

.. ~

•

o

o

o

o

o

()

()

LBL-15973April 1983

PROCEEDI NGSOF THE

1983 DPF WORKSHOPON

Collider Detectors:Present Capabilities

AndFuture Possibilities

February 28 - March 4, 1983

Lawrence Berkeley LaboratoryUniversity of California

Berkeley, California 94720

Edited ByStewart C. Loken and Peter Nemethy

Prepared for the U.S. Department of Energy under Contract DE-AC03-76SF00098 and for theHigh Energy Physics Section of the National Science Foundation.

/)

')/

)

)

)

\/

o

o

o

o

o

FOREWORD

The "Workshop on Collider Detectors: Present Capabilities and FuturePossibilities" was sponsored by the Division of Particles and Fields ofthe APS and hosted by Lawrence Berkeley Laboratory. It was held at LBL fromFebruary 28th to March 4th, 1983.

The organizing committee consisted of A.K. Mann (Chairman), C. Baltay,R. Diebold, H. Gordon, D. Hartill, P. Nemethy, D. Ritson and R. Schwitters.The local organizing committee was R. Cahn, S. Loken and P. Nemethy.

The workshop focused on the problems posed by high luminosities athadron colliders, considering luminosities on a continuous range from 1029

to 1034 cm-2 sec- l , picking two specific center-of-mass energies, 1 TeVand 20 TeV. The participants divided into the five working groups tabulatedbelow.

These proceedings contain three sections. Section I consists of inputto the workshop, the introductory comments of the organizing committee chair­man (A.K. Mann); two out of our three invited talks (W.J. Willis, M. Banner,C. Rubbia) on collider experience; finally two documents, which were invaluablein getting the workshop started, theoretical estimates of relevant cross sections(R. Cahn) and of high P~ jet behavior (F. Paige).

Section II contains the working group summary reports from the five work­ing groups; this is the meat of the workshop. Section III is a rich mix ofcontributed papers relevant to the workshop.

I want to thank Jeanne Miller, our workshop secretary, and Peggy Little, theLBL Conference Coordinator, for all their help; Donna Vercelli, Judy Davenportand Loretta Lizama for their work on the proceedings. I also thank our workinggroup leaders and scientific secretaries for their dedication and all the parti­cipants for a lively and spirited workshop. Support for this workshop was pro­vided by the Department of Energy and the National Science Foundation.

Peter NemethyWorkshop Organizer

()WORKING GROUP GROUP LEADER SCIENTIFIC SECRETARY

Tracking Detectors Doh Hartill David °Herrup

Calorimetry Bernie Pope Melissa Franklin

() Triggers Mel Shochet Mike Ronan

Particle Identification Dave Nygren Rem Van Tyen

Detector Systems Barry Barish Mark Nelson

\ iii~_J

)

.)

)

)

)

)

o

o

o

o

o

o

o

u

u

~)

TABLE OF CONTENTS

Foreword. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. iii

List of Participants.................................................................... ix

SECTION I. WORKSHOP INPUT

Introductory Comments................................................................... 3A.K. Mann

Experience of the Axial Field Spectrometer at the CERN ISR.............................. 5

W.J. Wi 11 is

Search for Hadron Jets and Large Transverse Momentum Electrons at the SPS pp Collider... 21

M. Banner

Cross Section Estimates for Multi-TeV Physics....................... 28R.N. Cahn

Monte Carlo Simulation of High ~ Jets.................................................. 30

F.E. Pai"ge, S.D. Protopopescu, and D.P. Weygand

SECTION II. WORKING GROUP SUMMARY REPORTS

Tracki ng Detectors...................................................................... 43D. Harti 11

Calorimetry............................................................................. 49B. Pope

Tri ggers. . . . . . . . . . . . . . . . . • . . . . • . . . .. . . . . . . . . • . . . . . . • . . . .. . . . . . . . .. . . . . . . . . . •. . • • . . . . . . . . 58

M. Shochet

Particle Identification................................................................. 62

D.R. Nygren

Detector Systems........................................................................ 64

B.C. Barish

v

SECTION III. CONTRIBUTED PAPERS

Track Chambers for High Luminosity...................................................... 75T. Ludlam and E.D. Platner

The Performance of A 16000 Wire Mini-Drift MWPC System.................................. 81P. Chauvant, R. Cousins, K. Hayes, A.M. Smith, R. Bonino, J.G. Zweizig,J. Alitti, B. Bloch-Devaux, A. Borg, J.B. Cheze, A. Diamant-Berger,I. Giomataris, B. Pichard, F. Rollinger, M. Tararine, J. Zsembery

Breakdown Processes in Wi re Chambers, Prevention and Rate Capabi 1ity .. .. .. . .. .. .. •.. . . .. 88M. Atac

The Concept of a Sol id State Drift Chamber.............................................. 97E. Gatti and P. Rehak

Large Area Non-Crysta11 i ne Semiconductor Detectors...................................... 99V. Perez-Mendez, T. Mulera, S.N. Kaplan, P. Wiedenbeck

The Design and Expected Performance of a Fast Scintillator Hadron Calorimeter •.•••..•... 101R.B. Palmer and A.K. Ghosh

Scintillation Chamber Operation of Calorimeters for Colliding Beam Detectors ..........•• 106L.W. Jones

High Energy Electromagnetic Shower Position Measurement by a Fine GrainedScintillation Hodoscope .........•.•.•.•.•..•....•...........................•......... 108

B. Cox, G. Hale, P.O. Mazur, R.L. Wagner, D.E. Wagoner, H. Areti,S. Conetti, P. Lebrun, T. Ryan, R.A. Gearhart

Scintillators and Waveshifters for Fast Scintillator-WSB Calorimetry••.•••.••..•••..•.•. 111G.E. Theodosiou

Scintillator Calorimetry with Vacuum Photodiode Readout. 113

W. Selove and G.E. Theodosiou

A Measurement of the Response of an SCG1-C Scintillation Glass Shower Detectorto 2-17.5 GeV Positrons 115

B. Cox, G. Hale, P.O. Mazur, R.L. Wagner, D.E. Wagoner, H. Areti,S. Conetti, P. Lebrun, T. Ryan, J.E. Brau, R.A. Gearhart

Calorimetry at L = 1033•••.••..••..•...•••.••..•.•••••.•......••...•...••..••.••••..••.. 119W. Selove and G.E. Theodosiou

Triggering at High Luminosity: Fake Triggers from Pile-Up.............................. 122R. Johnson

vi

C)

C)

o

)

)

)

)

)

o

o

o

o

o

o

o

u

u

. . 33 -2 -1Fast Asynchronous Levell Pre-Trlgger for Electrons at L = 10 cm sec 125M.J. Tannenbaum

Prospects for Cerenkov Ring Imaging at Hadron Colliders................................. 127M. Goldberg and D. Leith

Use of Synchrotron Radiation for Electron Identification at High Luminosity 132S. Aronson

TOF for Heavy Stable Particle Identification ...........•................•.......•....... 134C.Y. Chang

Operation of a High-PT Spectrometer Arm at 1033 cm-2 sec- l with ParticleIdentification........................................................................ 137

S. Aronson, M. Goldberg, M. Holder, E. Loh

t4uons and Electrons in General.......................................................... 140L. Nodulman and J. Bensinger

Large Solid Angles Solenoid Spectrometer with Particle Identification by dE/dx.......... 142D.R. Nygren

Detecting Heavy Quark Jets.............................................................. 143G. Benenson, L.L. Chau, T. Ludlam, F.E. Paige, E.D. Platner,S.D. Protopopescu, P. Rehak

Angul ar Coverage of Detectors........................................................... 157R. Diebold

A Detector with a 3 T Solenoid for a 10 TeV Collider.................................... 159R. Kephart, A. Tollestrup, M. Mestayer

Sta te of the Art........................................................................ 161L.M. Lederman

Experience with High Luminosity Running at the CERN ISR................................. 163J. T. Li nnemann

Operation of the AFS at L = 1.4 x 1032 cm-2 sec- l : A First Look at Data atHigh Luminosity from the CERN ISR..................................................... 169

The Axial Field Spectrometer Collaboration

Rates and Time Structure Expected from is = 40 TeV Colliders............................ 172R. Diebold

vii

Advantages of Spatial and Temporal Segmentation for Detectors at High LuminosityCW Colliders.......................................................................... 174

M.J. Tannenbaum

Monte Carlo Simulation of the Effect of Pileup.......................................... 175H.A. Gordon, R.A. Johnson, S.A. Kahn, M.J. Murtagh, D.P. Weygand

Calorimeteric Physics in the Presence of Multiple Events................................ 179J. Yoh

Virtus: AMulti-Processor Event Selector Using Fastbus................................. 185J. Ellett

viii

)

)

)

)

'0

o

o

o

()

()

o

u

Jeffrey Appe1Fermilab

Samuel AronsonBrookhaven National Laboratory

Muzaffer AtacFermilab

Charles BaltayColumbia University

Marcel BannerCEN Saclay

Barry Bari shCalifornia Institute of Technology

Wulfrin BartelDESY

Eugene BeierUniversity of Pennsylvania

Jim BensingerBrandeis University

Harry BinghamUniversity of California Berkeley

James BransonMIT

Martin BreidenbachSLAC

Alan BrossLawrence Berkeley Laboratory

Robert CahnLawrence Berkeley Laboratory

Leslie CamilleriCERN

William CarithersLawrence Berkeley Laboratory

Terry CarrollSLAC

Chung Y. ChangUniversity of Maryland

William Chinowsky .University of California Berkeley

James ChristensonNew York University

Bradley CoxFermilab

Robert DieboldArgonne National Laboratory

Mark EatonHarvard University

Guido FinocchiaroState University of New York

LIST OF PARTICIPANTS

Melissa FranklinLawrence Berkeley Laboratory

Emilio GattiIstituto di Fisica Politecnico di Milano

M. GilchrieseCornell University

Marvin GoldbergSyracuse University

Howard GordonBrookhaven National Laboratory

Douglas GreinerLawrence Berkeley Laboratory

Richard GustafsonUniversity of Michigan

Donald HartillCornell University

David HerrupLawrence Berkeley Laboratory

Werner HofmannLawrence Berkeley Laboratory

Martin HolderUniversity of Siegen

Walter HooglandNIKHEF

Frederick Russell HusonFermilab

Randy JohnsonBrookhaven National Laboratory

Lawrence JonesUniversity of Michigan

John KadykLawrence Berkeley Laboratory

Richard KassUniversity of Rochester

Robert KephartFermi 1ab

Richard KoflerUniversity of Massachusetts

Leon LedermanFermilab

David LeithSLAC

David LevinthalFermilab

James LinnemannRockefeller University

Eugene LohUniversity of Utah

ix

Stewart LokenLawrence Berkeley Laboratory

Thomas LudlamBrookhaven National Laboratory

Richard MajkaYale University

Al MannUniversity of Pennsylvania

Jay MarxLBL/DOE

Mac MestayerVanderbilt University

Peter NemethyLawrence Berkeley Laboratory

Mark NelsonLawrence Berkeley Laboratory

Laurence NodulmanArgonne National Laboratory

Davi d NygrenLawrence Berkeley Laboratory

Renee OngSLAC

Frank E. Paige, Jr.Brookhaven National Laboratory

Robert PalmerBrookhaven National Laboratory

Sherwood ParkerUniversity of Hawaii

Edward PlatnerBrookhaven National Laboratory

Bernard PopeMichigan State University

Veljko RadekaBrookhaven National Laboratory

Pavel RehakBrookhaven National Laboratory

David RitsonSLAC

Mi chae1 RonanLawrence Berkeley Laboratory

Jerome RosenNorthwestern University

Carlo RubbiaCERN

Bernard SadouletCERN

Fabio SauliCERN

Rafe SchindlerCalifornia Institute of Technology

x

Peter SchleinUCLA

Roy SchwittersHarvard University

Walter SeloveUniversity of Pennsylvania

Kirk ShinskyLawrence Berkeley Laboratory

Mel ShochetUniversity of Chicago

Eric SiskindNYCB Real-Time Computing, Inc.

A.J. Stewart SmithPrinceton University

Stephen R. SmithFermilab

Lee J. SpencerBrandeis University

Herbert StelzerGSI

Harry Sti ckerRockefeller University

Michael TannenbaumBrookhaven National Laboratory

George TheodosiouUniversity of Pennsylvania

Dennis TheriotFermilab

Alvin TollestrupFermilab

Rem Van TyenLawrence Berkeley Laboratory

Hugh Will iamsUniversity of Pennsylvania

Bill WillisCERN

Gunter WolfDESY

John YohFermilab

John ZweizigUCLA

:)

:)

:)

)

)

)

o

o

o

o

o

()

o

I ,\..~)

,)

SECTION I: WORKSHOP INPUT

)

)

,)

j

o

INTRODUCTORY CO~fMENTS

THE DPF WORKSHOP ON COLLIDER DETECTORS:

PRESENT CAPABILITIES AND FUTURE POSSIBILITIES

A. K. MannDepartment of Physics

University of PennsylvaniaPhiladelphia, PA 19104

our work here toworkshop in the

of elementary

o

o

o

o

()

()

It is useful before beginningrestate briefly the purpose of thislight of the present circumstancesparticle physics in the U.S.

The goal of our field is easily stated in ageneral way: it is to reach higher center of massenergies and higher luminosities while employing moresensitive and more versatile event detectors, all inorder to probe more deeply into the physics of ele­mentary particles. The obstacles to achieving thisgoal are equally apparent. Escalating costs of con-

'struction and operation of our facilities limitalternatives and force us to make hard choices amongthose alternatives. The necessity to be highlyselective in the choice of facilities, in conjunctionwith the need for increased manpower concentrationsto build accelerators and mount experiments, leads tocomplex social problems within the science. As thefrontier is removed ever further, serious technicaldifficulties and limitations arise. Finally, compe­tition, much of which is usually healthy, now mani­fests itself with greater intensity on a regionalbasis within our country and also on an internationalscale.

In the far (£ 20 yr) future, collaboration onphysics facilities by two or more of the major eco­nomic entities of the world will possibly be forth­coming. In the near future, we are left to bypass orovercome these obstacles on a regional scale as bestwe can.

The choices we face are in part indicated in thelist of planned and contemplated accelerators shownin Table I. The facilities indicated with an asteriskpose immediate questions: (1) Do we need them all andwhat should be their precise properties? (2) How arethe ones we choose to be realized? (3) What is thenature of the detectors to exploit those facilities?(4) How do we respond to the challenge of higherluminosity as well as higher energy in those col­liders?

The decision-making process in this country andelsewhere depends on the answers to these technicalquestions. Those relating to the accelerators havebeen and continue to be addressed in many workshopsand studies. For example, a workshop organized by~f. Tigner will begin to study the means of achievinga very high energy (10 TeV x 10 TeV) hadron collider;this is scheduled at the end of March at Cornell Uni-

3

versity. If it seems desirable, continuity in theform of subsequent workshops on technical questionsrelating to accelerator facilities might be providedby the Division of Particles and Fields (DPF) of theAmerican Physical Society, as it is doing here forcollider detectors.

The workshop we are about to begin is intendedto address questions (3) and (4) above. It is anattempt to look at those questions from a broad pointof view by assembling a wide spectrum of experts fromuniversities and national and international labora­tories. It is planned to make the proceedings of theworkshop available to the 1983 Woods Hole Sub-Panelof HEPAP which is charged with the responsibility forrecommendations concerning the choices that face theU.S. program. This is the main reason that we aremeeting at the present time.

In this connection, it is worth emphasizing thatthe DPF is an organization of roughly 3000 physicistsfrom universities and national laboratories. It isan independent organization not affiliated with anylaboratory or government agency. Most important, itis not a decision-making body or a lobbying group.Its aim is to provide neutral arenas for scholarlydiscussion of the salient issues of our area ofscience, e.g., the DPF Summer Study on ParticlePhysics and Facilities in Snowmass, Colorado, in thesummer of 1982. For this reason it concentrates ontechnical questions such as those of this Workshop.

Lastly, let me comment on the structure of theWorkshop.

First, note that according to the registrationrolls there are 89 American physicists here, dividedabout equally between universities and national labo­ratories, and 15 physicists from Europe, a sign ofof the beneficial international cooperation that ispresent in our field. In addition, there are tworepresentatives from industry.

All of us are indebted to the three (and onlythree) invited speakers, W. Willis, C. Rubbia andM. Banner, who have brought us the results of theirextensive experience in the subject before us. Weowe thanks also to D. Shirley and D. Jackson for thehospitality of Lawrence Berkeley Laboratory; and tothe local organizing committee for the Workshop:P. Nemethy, Chairman, S. Loken and R. Cahn, who havelabored mightily in our behalf.

Table I . . Planned and Contemplated New Facilities for High Energy Physics.

West Germany DESY

Japan KEK

U.S.A. SLAC

FNAL

BNL

?

U. S. S.R. Serpukhov

Switzerland

Laboratory

CERN

!.acility Properties

SPS Collider1 270 GeV x 270 GeV; 1029 -2 -1pp: cm sec

LEP + - 50 x 50; 1031e e :

* pp or pp: > 5 TeV x > 5 TeV; ?

HERA e p: 30 x 800; 0.5 x 1032

Tristan + 30 x 30; 1031e e:

SLC + - 50 x 50 6 x 1030e e

TeV I pp: TeV x TeV 1030

TeV II fixed tgt p: TeV; > 2 x 1013/60 sec

* CBA pp: 400 x 400 1033

* ? pp or pp: > 5 TeV x > 5 TeV; ?

fixed tgt p: 3 TeV;

, )

===============================================================================================

1 In operation and yielding data

4

)

)

'.)

)

EXPERIENCE OF THE AXIAL FIELDSPECTROMETER AT THE CERN ISR

William J. Willis

o

()

o

o

()

o

1. Introduction

The Axial Field Spectrometer is a facilitydesigned with an emphasis on flexibility and measure­ments of energy by calorimeters of very good energyand angle resolution for both electromagnetic andhadronic particles. It was proposed by BNl-CERN­Copenhagen-lund-Rutherford collaboration and startedoperating in 1979. It was designed for moderateluminosities, 10 31 - 32 cm-2 s-1 or interaction ratesfrom 0.4 - 4MHz, and charged particle multiplicitiesup to 50. It would certainly not operate satisfac­torily at a luminosity of 1033 , but the rates arehigh enough to bring out the important differencesfrom the operation of low rate experiments, such asthe e+e- or pp colliders. In fact, in the surveypresented to this Workshop by lederman, the highestinstantaneous rates in a full solid angle, "open"spectrometer leading to a published paper are from theISR experiment RllO, with a conventional magnet low-8 insertion, and a luminosity of ~ 6.5 x 1031 cm- 2 S-1.A luminosity of 1.4 x 1032 an-2S-1 was obtained lastDecember in a short test of the superconducting low -8insertion installed for the Axial Field Spectrometer.Some first results on an energy flow high PT jetanalysis will be described in these proceedings byHoward Gordon. (On the last day of this Workshop, westarted a successful 60 hour run under these condi­tions which will yield publishable results.) Thehistory of the record luminosity attained in the ISRover its history is shown in Figure 1. The arrowshows the design luminosity. I hope the newercolliders will have the same fortunate history. Someimportant consequences of the relatively high ratesare:

i) The mutability of the experiment.ii) The dominance of trigger considerations.iii) A related technical matter: the overlapping of

the data acquisition with trigger processing,in time.

2. Permutations,Combinations

The high rate allows several different experi­ments to share the solid angle and the core detectors,with parallel triggers which are sufficiently selec­ti ve to keep th eli vetime hi gh. These experimentscan be completed in a reasonable time, to be succeed­ed by others, a steady change of confi gurati on whi chI have called the "mutability" of the set-up. Therate of these individual experiments would of coursebe higher if they covered the full solid angle, butin a world of finite resources it is often verydifficult to provide for that, unless it can be builtinto a universal detector of sufficiently highperformance or into a device we still have not seen,even on paper.

Below we describe eight mutations of the AxialField Spectrometer, with brief descriptions of thephysics aims of each, and the list of collaboratinginstitutions involved. Note that the originalcollaboration did not undertake all these efforts, norwould that have been possible. Some institutionsjoined the core collaboration after undertaking aspecialized experiment, others participated only in aspecific topic.

5

A. AFS Central Tracking Detector and UraniumCalorimeter - BNl-CERN-Copenhagen-lund-Rutherford­Pennsylvania-Tel Avis

The apparatus is shown in Figure 2, with somelater additions. The physics aims were the study ofhigh ET events, high PT jet production, high PTelectrons and pairs. There was a great emphasls ongood measurement of energy flow, with 800 independenttowers over two units of rapidity, with electromagneticand hadronic readout, and position interpolation inthe towers. The single hadron energy resolution in0.37E~ (GeV). Figure 3 gives an idea of theconstruction, and shows two of the four walls coveringthe whole ezimuth. Tracking is provided by a JADE­type drift chamber (Figure 4) with 42 layers whichidentifies electrons and slow hadrons by ionization(0 = 11%), as shown in Figures 5 and 6.

Typical physics results from this apparatus arethe single inclusive jet cross sections shown inFigure 7, or the comparison of high -PT single particleproductions by pp and pp collisions in Figure 8.

B. Identified High PI Charged Particles - Copenhagen­Lund-Rutherford

This experiment was described in the originalproposal, with a spectrometer including three Cerenkovcounters and PWC's provided by the above institutions.The set up is shown in Figure 9. The physics aim isjet fragmentation and particularly quantum numberdependences, and preliminary results were presentedat the Paris Conference.

C. Direct Single Photon Production

This process (which had not been forseen) turnedup in the data of ISR experiment R806 during theconstruction of the AFS, and the liquid argon photondetectors were installed to run in conjunction withthe AFS from the beginning of its operation (Figure l~.This proved extremely interesting since it allowed thestatistics on the photon production to be greatlyincreased as well as providing a complete measurementof the associated event. The photon cross section asa function of PT is shown in Figure 11, with a slopecharacteristic of an elementary QCD process, comparedwith the pi-zero cross section measured from two photonevents in the same apparatus. The ability to measurethe whole event was essential to demonstrate that thephotons come from the elementary QCD "Compton" process,rather than something like quark bremstrahlung.Evidence for this comes from several measurements, butmost clearly from the distribution of charged particlesassociated with single photons and with pi-zeros(Figure 12). The pi-zero is accompanied by otherparticles resulting from the fragmentation of a highPT parton, but there is absolutely no sign of otherparticles associated with the photon, showing it is thephoton in this process. Other measurements agree withthis picture, so that this process can be used as adirect measurement of the gluon distribution function.Further information comes from comparison of pp and ppinteractions.

D. Forward Direct Single Photon Production - BNL­Pisa-Pittsburgh

Measurements near 90 0 do not provide enoughinformation to determine distribution functionsaccurately. A small group of physicists built aphoton detector of very fine granularity to measuredirect photons at angles of 11° to 17°. The detectorconsists of 10mm wide scintillation and 1.7mm thickuranium plates, and provides a spatial resolution of2mm, separating single photons from pi-zeros forenergies up to ~ 25 GeV, at only 1.8m from the source,Figure 13. The inclusive photon spectrum is shown inFigure 14, with two theoretical predictions and theanalysis of the associated charged particles inunderway.

E. Forward Hadron Calorimeters - Johns Hopkins­Pennsylvania

'These were installed to carry out an initialdouble~Pomeron trigger, and added to for studies witha-a and a-p interactions, as shown in Figure 15. Theyallowed many measurements on the correlations ofproton and a fragmentation with central rapidityregion multiplicity, etc. Typical distributions areshown in Figure 16. Note that these calorimeters arevery close to the outgoing beams and will be mentionedwhen the question of radiation damage arises later inthis talk.

F. Double Diffractive Proton Trigger - Cambridge­Queen Mary College-Rutherford

This experiment is a search for gluonium,presumably enhanced in two gluon exchange, and thusin the double diffractive region. The protons aretagged by a system of small drift chambers just out­side the beam pipes 5-6m from the interaction point,Figure 17. If elastic scatterings are eliminated, andevents with forward particles vetoed, the coincidencerate is small enough to allow this to function as aparallel trigger.

G. Single and Double Direct Photon Detector with NaIand Vacuum Photodiodes - Athens-Moscow EngineeringUniversity-Lebedev Institute, Novosibirsk-Pisa­Pittsburg

This experiment is an improvement over the R806photon detector in the following respects:

i) better resolution, especially at low photonenergies;

ii) tower readout instead of strips;iii) larger solid angle;iv) (more important!) compatible with the

full-coverage

It uses two arrays of (35 mm)2 NaI crystals inprojective geometry, 5 radiation lengths thick andreadout in a magnetic field with a vacuum photodiode,giving a noise level of about 1 MeV, as shown inFigure 18 and 19.

H. High PI Muon Detector

The PWC's from the Cerenkov arm have been re-usedto create a muon trigger covering about a tenth of thesolid angle, using the calorimeter and 80cm of iron asan absorber, a total of more than eight absorbtionlengths. The energy threshold and rejection areenough to allow the trigger to run in parallel with aminimum of other conditions.

6

This shared mode of operation ~equires.

i) a sUfficiently high interaction rate,

ii) a powerful trigger system to pick theparticular processes of interest in a selective,efficient manner so that a number of parallel triggerscan be used with a high live-time fraction,

iii) enough live-time per year to make acceptablethe overheads due to more complex situation,

iv) a certain amount of good will among theexperimental teams, each of whose projects will rm abit slower than if they had sole use of the facility.

I believe that the degree of parallelism describedabove is not usually or ever achieved in fixed targetspectrometers, due (I suppose) to a combination o'r. allthe above factors. For example, on point i), weshould recall that the continuous duty cycle of ISRgives a factor of four or so more livetime. Thecharacter of storage rings is such that if any runs,everyone runs. This has tended to provide more hoursper year than fixed target programs. The fact thatthe arrangements were made with a minimum of inter­vention of a program committee has probably avoidedexcerbating the problem iv). Of course the questiondoes not come up in e+e- machines where the idealtrigger selects all real events. All these pointsconsidered, the provision of a suitable trigger systemis still the most important.

3. Triggering

The trigger system involves a number of levels,schematically indicated in Table I. A much simplifiedschematic of the first two levels is shown in Figure20, showing only the part having to do with theuranium calorimeter. The system is one dimensional,in that sums are performed for fixed azimuthal angle ¢,with weights as a function of polar angle and ¢ (totake out the center of mass motion due to finitecrossing angle at the ISR) so that the sums representPT (¢). These 48 electromagnetic and 48 hadronic sumsare transmitted on fast cables to the counting roomtogether with a number of other fast signals used inthe trigger, while the data aquisition signals travelon longer slow cable, giving a delay of 250ns whichallows the first level of trigger processing to benearly deadtimeless. The first level consists of anumber of digitally controlled thresholds on globalsums and other simple functions.

The second level is also carried out by analogcomputation, and can consequently be quite fast. Infact, a moderate increase in the data acquisitiondelay would have allowed it to be overlapped with thesignal transmission, effectively eliminating itsdeadtime, but this did not seem to be necessary at ourluminosities. There are about 700 computer controlleddiscriminators, whose outputs are ~ncoded onto about50 parallel lines, on a fast ring bus, which drives upto 48 logic modules which use those lines to formdecisions, consulting fast memories loaded by computer,and "prescalers" if desired. Up to this level, thesystem is completely parallel, rather than organizedserially, so that this addition of additional triggershas no effect on timing, or any other function ofanother trigger. A complete monitoring and checkingsystem is provided.

)

)

)

\/

o

u

()

\ )

The timing of the next level was partly dictatedby electronic processing or in the PWC single particletrigger, and partly by intrinsic detector delays, asin the case of the drift chamber or NaI detector.

The next level uses the ESOP microprogrammedcomputer developed at CERN and used in a number ofexperiments. A typical use is a sagitta calculationon a track in the drift chamber, given a pointer froma shower detector at level 2, which takes 80-200 ~s,

depending on the number of tracks in the road, thefirst number being mainly the data transfer time.

This system is used with 10-20 parallel triggers,with about 6Hz of events written to tape (8000 tapesper year for 107s !). Each trigger then must average0.3 Hz, or about ~10-7 of the total interaction rate.High PT thresholds can of course give as small a rateas one likes; the trick is to select the desiredreactions without very high thresholds - consequentlythe need for sensitive, multi-level triggers.

4. Processing

Going to higher luminosities while maintainingthe same sensitivity to rare, but moderatelY low PTprocesses, seems to require the maintenance orextension of the concepts of parallelism, analogprocessing and overlapping seen in the first twolevels of our trigger. It is difficult to envision atime when summing and weighting can be done as fastwith a digital system as with an analog one. True, itdoesn"t matter if the trigger processing takes a bitlonger as long as it is overlapped with the dataacquisition through some sort of delay. This delaybecomes very formidable though, as the number ofchannels in the detector increases; the dynamic rangerequired increases (for example because the relativeprecision of our calorimetric detector is increasingas E-~) and the time resolution of the detector isbeing improved to cope with the higher luminosity.

The equivalent of the first two or three levelsof our system need to become more powerful to provideincreased selectivity, unless one is content toaccept much higher rates (of much larger events!). Itwill be important to keep the processing in a fullyparallel mode, while dealing with more detectorelements. This will require wider buses - perhapsseveral hundred lines for example, preferably operat­ing faster. New ideas on connectors and mechanicaldesign are needed. (Our ring bus represents anuneasy compromise between compactness and mechanicalsolidity.)

I stress this subject because it is extraordinar­ily elusive in the context of a workshop. It clearlydoesn't correspond to a definite physical limit assome detector issues do. It is hard to identify theproblem unless the detector and its use are specifiedin some detail. It sounds like it is just a matter ofexpense, but one should keep an eye on it.

5. Detector Issues Associated with High Rates

First, I should note that the detector we haveused at the ISR have all been optimized for resolution,in energy or angle, with attention to high ratecapability as a secondary consideration. Indeed, theexperience in our field has shown that resolution isso important that I would not give up as much as afactor of two in resolution to get only a factor oftwo in response time, for example. Of course priceand delivery time have often played a role. Forexample, the optical components in our scintillator

7

calorimeter could now be replaced with faster oneswith no sacrifice in performance. Figure 21 shows asmall increase in hadron resolution in our calorimeterwhen the gate is as short as 50ns, with the electronresolution . unaffected. This is presumably due to acombination of time of flight of slow hadrons,particularly neutrons, which will be dependent onabsorber material, and of slow components in thescintillator induced by heavily, ionizing particleswhich will depend on scintillator type. In optimizingthe system, one will effect the linearity of thehadron calorimeter, and its resolution, particularlyfor jets. This is shown, in the form of the ratio ofresponse of e's and TI'S of equal available energy,in Figure 22.

6. Pileup

Working with detectors with non-negligibleresponse times for the rates we deal with, we have hadto devote a lot of attention to detection ofaccidentals, and pile-up. Our initiation to thisproblem was with our LA calorimeters, which had abi-polar response 1200ns in overall length. We startedusing this system with interaction rates of about50kHz, and expected to have to fight a serious pile-upproblem as the ISR luminosity rose by an order ofmagnitude. Accordingly, we equipped ourselves withsophisticated pile-up detectors built by V. Radeka.In fact, we never encountered any problem even at thehighest rates, because all the physics programs withthat detector turned out to involve at least onerelatively high PT particle which could be used in thetrigger through the analysis might involve CMS energiesas 50 MeV. The energy of the high PT particle in thetrigger was required to be localized in a solid angleof about 0.0075 steradians, which leads to a lowsensitivity to accidentals. The solid angle quoted waswith a detector with strip type readout. One withtower type readout, as in the case of our present NaIdetector, has a cell solid angle an order of magnitudesmaller, and our double direct photon experiment wouldfunction at very high luminosities, even with the slowNaI.

The solid angle occupied by a jet is larger, andwe have prepared three systems for use in our plannedexperiment to measure jet production near the kinematiclimit with L ~ 2 X 1032 cm- 2 S-1.

i) Time measurement on a set of scintillatorsforming a central barrel, and read out at each end.The error is a few ns.

ii) A set of simple fast scintillator counterstowers filling 30° cones forward and backward. Theresolution is 0.7ns.

iii) The accidentals can be adequately rejectedat high luminosity with the previous two systems, butwe want to accept most events with "harmless"accidentals and make a small pileup corrections. Forthis purpose, we have installed a set of 50 fast flashADC's looking at the p ($) and other sums, with 10nsbins. l~ith these we can usually establish whether anevent has been appreciably affected by accidentals.

In order to apply this technique to ET triggers atour luminosity, we might have to digitize the shapeof the signal from our 800 towers, which would not beunreasonable, but since the ISR has only one more year,we have no plans to do that.

7. Radiation Damage and Lifetime

This can be a problem at ISR. The forwardcalorimeters near the beam, which were shown inFigure 14 received significant radiation damage aftersix months, at which point they were removed. Theyused acrylic scintillator, while current practicewould use polystyrene based scintillator which has alonger lifetime by a factor of about five. Note thatthe dose delivered to a calorimeter goes up with theenergy of the beam as well as the luminosity. Forexample, 10MHz of a 6 TeV forward jet is 10 watts, anintense radiation field, since most of it is depositednear the beam pipe in a limited angular region. Sincethere are many constraints which limit the length ofthe drift space, this will be a limited area as well.It maybe tempting to dismiss the forward energy flowas irrelevant to high PT phenomena, but this wouldeliminate a very important tool for many experimentsone will want to do. Even if you did not wish to knowPlongitudinal of the beam jets there is a substantialtransverse energy flow in this area, compared to thepotential errors on this quantity from other sources,and this consideration alone should require thismeasurement. The measurement of Plongitudinal andthus the total energy, is a powerful handle on thepile-up problem, among other things, and will beextremely useful if it can be measured with extremelygood time resolution. I believe that it can be, athigh energy, because the very size of the energydeposit broadens the range of techniques available tomeasure it. The use of plastic scintillator seemsruled out in the high radiation zone. Liquidscintillator might be used if a means of reading itout can be devised.

We have been faced with exactly this problem indesigning a fixed target experiment recently. Wedecided.to look again at liquid· ion chambers, butwith a fast response, unlike the usual liquid argoncalorimeter. The increased noise level can betolerated because the amount of energy is large, 200­400 GeV in our case. We have designed a calorimeterwith readout in (15mm)2 towers, with a bipolarresponse full length of lOOns, using LA-CO or LA-CH 4mixtures. It should be proof against radiationdamage. Further reductions in response time should bepossible, if neutron lifetime effects permit,particularly at higher energies. At 20 TeV, 100MHzrates might be possible -- this will required forcedcooling of the forward detector!

8

The lifetime of drift chambers at high rates isnot very well understood. Sometime ago, gas mixtureswere found which seem to eliminate the problem ofcathode deposits. Most present chambers fail due todeposits of silicon on the anodes. The source of thesilicon is clearly in the environment, and since ittends to be ubiquitous in the forms of pump oil, dustand fiberglass, most practical chambers have not beenwell enough controlled to analyze quantitatively. Wecan say that the hottest wires in the Axial FieldSpectrometer chambers have accumulated almost 104

vamp-hours/meter of wire without signs of trouble.D. Cockerill, E. Rorro and H. Hilke made carefulstudies of small chambers which seem to show that thisproblem does not occur in sufficiently clean chambers,which seems logical. The CHARM collaboration uses acontrolled addition of H20 vapor to suppress thephenomenon. My conclusion is that drift chambersrunni ng at low gas gains (<,104 ) wi 11 be 1imited byaccidentals rather than destruction, if suitablyconstructure and operated. --

8. Outlook

As luminosity is increased above 1-2 x 1032 cm- 2S-l, we might forsee the following trends:

i) calorimeters can be speeded up - beyond acertain point some price will be paid in resolution asthe neutron response is clipped, but this may beacceptable in the forward direction where fast responseis really needed.

ii) drift chambers with small drift distances canbe pushed to higher rates. They will be mechanicallymore complex as support for the wires is probablyrequired, and a "clean" construction to avoid lifetimeproblems.

iii) the very fast response of low pressureparallel plate gas chambers would be attractive ifthere were a way to get signals. Examples are theBaF2 counters of Anderson and Charpak.

iv) don't forget, lots of R&D on systemproblems!

v) none of this is free.

)

)

)

)

Uranium CalorimeterNA I nHEX CAL

1 30 ns 'Minimum Bias' CountersSPT Matrix (Counters + PWC)Muon'Double Pomeron Trigger'

Above Combined With: !'l<P

CAL. ET M1 M2 M3 i M4 360°

1 Hadron N1 N2 N3 1 N4 15°

2 200 ns 1 Em. II II II

iII 7.5°

Wall II II II,

II 90°I ii ,

Jet III II I II,

II 45°I I !I

NA I LocalizationPWC Matrix Look - up

3 800 ns DC Sector· SPT Pointer'e' DC Sector • EM Pointer'e' DC Sector· NA I Pointer

'ESOP' (Fast Digital Proc.)Routines: - sagitta1,2 k

4 100 ]JS - u-CAL Localization

5 168 E Filter

o

o

u

()

)

LEVELDEADTIME

Table I. TRIGGER COMPONENTS BY LEVEL

COMPONENT

9

()

)838281807977 78

Year

7675

1.6

1.4

1.2

1.0

NC')0po 0.8X...I

0.6

0.4

0.2

072 73

XB L B34·15OO

Fig. 1. History of the ISR record luminosity. )

)

)

CERENKOV A.P.CHARIOT

UR. CAL.CHARIOT

IURANIUM CAL.

+ DRIFT CHAMBER. CHARIOT

un. CAL.CHARIOT,

;ltlOI1IMOER'~ 0;=1---1I

i

(.i ",.~---j

I'HotON DEl.L,!!"

Fig. 2. AFS Central Tracking Detector and UraniumCalorimeter

10

o

o

()

LIGHTGUIDES'

PHOTOMULTIPLIERWITH SHIELDING

LIGHT TIGHT BOX

SCINTILLATORLIGHTGUIDE

/fiLSIBACKSECiJONJ

FRONT-·PLATE - --Y

'AVQH Ii it

~~Z ".'''''ml ~

""1H1,'f! /

FONG -A·

---.-..

-" --- _~..-\r...,. i-·.. ·~;··:;·~:~:~···~···l- - --~ .

---- - . ··,t·;;·...,,··»·;:····, - j-••••••• .. ••••••··•··· .. t-_~

' -\] _ 1' 1'.1< 1'."•• , .

.-...-..~ , ,t~·····~·······~···~!~11.............. •.. ~· .. b..·1'.. " II It n 2t If I' ''I U"1'Cl:..::C:l.._~ ~.~ -f. , [ ·..·T·..·,..t·'l'..·".· •~. ·~:'I::l:~l: .. P."~".\\.,, Ii " II " n n n n n~·.........:··.-....·.....~ll.........~.i '_ : •••••••••• • •• 'I·l··~·········~~·!·t"

. , ~.~ ·~'t"T.,... ~.':.t-~.•'t.... 'i

Fig. 3. Calorimeter Construction

u

Fig. 4. Tracking Chambers

/j

JUrI IGQQI,. MH1' 14JG'. '1lIRJI Mrs..10,.- ---,

)

/.eo

..

..fIUlIC:.QI1I.C'r.NT ~&8. IHCO[ft HITS...

"

'"./

[ '"-",

71

"t ",.c.r,.V(, Y .....

oJ

\.50\\

;",.i

227

i '. /~O 4

..."

"- ".\,. " ".

• \,.. ....... .,,

",.

..~L-_-J..__....--'-__'--_--'-_-J..'--_~_--''-_...l

• (e»)

... L-_--L._-L_-l._-.JL._.L-_..L_--L_-J~ -40 ...:10 020

X (Of)

Fiq. 5. Typical events reconstructed by computer

,)• iOi

135 ..126 :)

\K \

i05~ TRITONS0~ ~096

90 \ 20~8..-.:c ,'"' e , 102~

)...UJ 15 9 ,:c 512

UJ,

<fI \ 256oJ I);:, ,CL 60 , 128,

6~

, l:-4S , 32f) __

16 )"

~O

15

0LC'Q 1.20 I.~O 1.60 1.80 2.00 2.20 1 2.~0 2.60 ! 2.80 3.00 3.20 3. ~O 3.60 3.80

LO(;IP) 1~0 +0.2 .5 1:5 GeV/c

Fig. 6. Pulse height vs momentum )

12

o

o

+ PP -JET·X

+ pp-7TO.X

10- 31 I- n

I

----ISAJET (BOler et all

--- ISAJET (O,,~ns .Reyo)

o.c.o~'!--c-_....

_--- (1M

1I30Q..o.F

(0)fs=31 (jeV

------10

-

-

-

"'. ",'.. ', '-

"\,-

I-

1034- ..

~)~

N

>4>(;)

1:u

'--'

I,J 0",.,

bl'"'" a."U "U

W

10-36I- -

U +1

1()37I .

6 7 8 9- 10 II 12 13 14

P,. (GeV)

,)

0·5

2 3

t

4

P, (GeVlel

Fig. 7. Single Inclusive Jet Cross Sections Fig. 8. Comparison of high PT_single.p~rticleproduction by pp and pp colllslons

- ) 13

()

)

)

,.'. .. -/

////

..'

COP CAL.CHARIOT

fERENKOV A.P.CU"RIOT

I

\

' URANIUM CAl.UR. CAL. • DRIFT CHAMBERCHARiOT CHARlOT,'------..,...---J·'------~vr_--------'

OUTSIDE INSIDE

fJ- ryi).-_J1\ I-IiI--..,,...-"">'-(·-!'-'T_-··,-l:f-i~+<'J_O •.L~~_.J:-::::=::---'----I

\ "'~".J •.•r·~",#,···" Y"~~::~'~.RRIt.tETERS I. CENtRAL ~.ASCE~EL\ / ..I:[} -- ~... , "~

;~ "

Fig. 9. Detector for High PT Charged ParticleExperiment

lOT

L.A. - E.M.CAlORIMETER -14-_'1:1

)

Fig. 10. Direct Single Photon Production Experiment

14

()

()

n

o

C)

o

u

MU

NI

>OJ~

N

E<:.J1

b I' 0-M M"0"0

W

to-37

§

o

o

o

o

\IS =63 GeV

o n;o• y

¢

o

6 8 10P

T(GeV/c)

12

0.25

• !to

oY

0.20

~>Qj-U 0.15:;:..(,:)

0.. ..;

"....s-UI 0.10'"uel-

•0.05 •

Y +? ~<»•, ,

0 24 48

1I

t

12 96 120114l (oJ

. \"'-./

.,)

\, J

Fig. 11. Photon cross section as a function of PT

15

Fig. 12. Distribution of charged particles associatedwith single photons and with pi-zero's

Y-f;LO-lif;.IllON

PHOTOMULT1PLIER

BBO BARS FORFRONT-STRIPS -

SUPPORT ROD

Fig. 13. Detector for forward direct single photonproduction

GUIDE

)

)

:)

0.30

0.20

0.10

o

{ s=63GeV

+This experiment2.00<y<2.75

¢ Ref [111 y-O

;(b)

""././

././

./;'

;'/

//

II

II

(a)

)

)

PT IGeVIe)

Fig. 14. Inclusive photon spectrum

16

.)

o

o

()

o

o

a -SIDECALOR IMETER

BEAM HOl:lOSCOPE _(4 SCINTIl-,l.,ATORS)

Fig. 15. Forward Hadron Calorimeters

0.12 r----..,.----...,.---...:----r-----,-------.

•o

-15 GeV p - 15 GeV p

• 15 GeV/Na -t5 GeV/N a

o 15 GeV/Na-32GeV p

a SIDE

iIi i

. liil 5l ~_li-

600 800 • 1000

•o

,400

•o 0

o •o

'0.0

>-1- 0.1 0-' 00

ill-~ 0.08o0::(L'006t!.' .LiJ:> 0' Od1-' ...<C ';..J "W ' '-.)0:: .:-:,J.~~

()

!)

u

PULSE HEIGHT

Fig. 16 Pulse height distributions

17

Fig. 17. Double Diffractive Photon Experiment)

)

)

)

)

II I U-WALL

:'1I ,tI I

~,

I \I , I

_\ ,, I II I

; I II I, t I I -HI I I

" I I I I I I I I, I I I I I I I II I I I I

I I"

I II I ,

1200I

II

, II

I! I

I I I I I II I I; t I I

I I I I I II I I I I

I I I I I

..!..I I I I I I

I I I I I I

1-1·>---880 ---I-I

-r2400

_1

NT SECTION

CALORIMETER

RIFT CHAMBER

I{

1m

SR

I 0II

VPD11 \\ l\\\\~ No I

~ FRO

/ u-

- --. f

Fig. 18. Sodium Iodide Detector Array Fig. 19. Detail of Sodium Iodide Array )

18

o

o

()

o

oSingle

hadron Jet sint/> cost/> ~

C

CalorimetersSinE>·I

Calibrate

~ ETOT Pre trig

T ~ ETOTDC

o

Flattened

Otherinputs

Delay

Strobe

XBL 834-1497

Fig. 20. Schematic of first and second level triggerfor Uranium Calorimeter

19

()

1.1

1.0

0.9 ()15c 0.8C)

CiS0.7

0.17

e 0.16

e

.::-----._---.'Tr

)

0.15

20 40 60 80 100

T (nsec)

120 140 160

)XBL 834-1496

Fig. 21. Calorimeter response and resolution as afunction of gate width

)

1.6

1.5

1.4

1.3

GIll< 1.2

1.1

1.0

0.9

0.8

00.7

0.5 2 5 10 20

)

)

EavaR (GeV) )XBL 834-1508

Fig. 22. Ratio of response of e's and IT'S of equalavailable energy

20

()

o

SEARCH FOR HADRON JETS AND LARGE TRANSVERSE MOMENTUMELECTRONS AT THE SPS pp COLLIDER

Marcel Banner

()

o

()

()

o

u

1. Introduction

The search of high transverse momentum electronsneeds the use of all the different elements of the UA2detector; therefore the description of this search isa good way to understand the features of the apparatus,and its performance in a collider environment.

We present a preliminary analysis of the UA2data collected during the last Collider run (20 nb- lintegrated luminosity) with particular emphasis onlarge transverse momentum hadron jets and on electronshaving the configuration expected from the decay ofelectroweak bosons.

2. Detector and Data Taking: General Description

2.1 - Detector

The UA2 detector 1,2, shown in Figure 1, was de­signed mainly with the aims to observe electroweakboson decays and to study final states containing largetransverse momentum hadron jets: such phenomena areexpected to result from the collisions at very shortdistance accessible to the SPS pp Collider, the onlyexisting facility providing a sufficient centre ofmass energy, Ys =540 GeV. The expectation that thecollision products relevant to the study of such pro­cesses populate mostly the central rapidity region ledus to restrict the detector coverage to -3.5 rapidityunits, within which, for example, -2/3 of ZO decays areexpected to occur. The cones corresponding to scatter­ing angles e < 20°, e > 160° are not covered by UA2but are left open to house the detectors of experimentUA4 designed to measure the elastic and total pp crosssections. 3

Despite their small branching fractions the lep­tonic decay modes of the electroweak bosons

W± ->- 2\) (BR:::=: 8% per 1epton type)

and ZO ->- 2+2- (BR:::=:3% per lepton type)

are supposed to be the least difficult to detect be­cause of the very small expected background contamina­tion. For this reason UA2 was designed to detect andidentify electrons over its whole acceptance; elec­trons were preferred over muons because excellentenergy resolutions can be achieved in compact calori­meters.

Hadron detection is also implemented over thewhole UA2 acceptance but different approaches havebeen adopted in the central (40° < e < 140°) and for­ward (20° < e < 40°, 140° < e < 160°) regions. Thecentral region is instrumented with a highly segmentedhadron calorimeter having a cell configuration wellsuited to the observation of hadron jets independentlyof their mode of fragmentation. The forward regions,where important W->- ev charge asymmetries are expectedto occur, are equipped with two magnetic spectrometers,each consisting of a toroidal field magnet followed bynine drift chamber planes.

21

A set of cylindrical wire chambers densely packedaround the beam in the collision region provides mea­surements of the position of the event vertex and ofthe directions of the charged particles produced inthe collision. It is made of four proportional cham­bers with helical cathode strips and of two drift cham­bers of 24 azimuthal cells each, with six drift wiresper cell.

and of two drift chambers of 24 azimuthal cells each,with six drift wires per cell.

The measurement of a track pointing to an energydeposition localised in one of the electron calori­meter cells provides a powerful means of selectingelectron candidates but it leaves a significant con­tamination of narrow TIo- charged hadron pairs (over­lap background). This is strongly reduced by usingpreshower counters in front of the electron calori­meters to provide improved space resolution. Thecentral preshower counter is a cylindrical proportionillchamber with helicpreshower counters are proportional tube planes. Eachis preceded by a -1.5 radiation length thick converter.This chamber is called C5 in the text.

2.2 - Data Taking

The data discussed in this report were recordedusing triggers sensitive to events with large trans­verse energy in the central and forward calorimeters.They were of three types:

-- The LET trigger required a total transverse energy(LET) measured ln the central calorimeter (electronand hadron cells linearly added) in excess of-35 GeV,-- The Wtrigger required the presence of at least onequartet (2 x 2) of electron calorimeter cells (centralor forward) in which the measured transverse energyexceeded 8 GeV,~- The ZO trigger required the presence of two suchquartets, each having a transverse energy in excessof 3.5 GeV, and aximuthally separated by ~ ¢ > 60°.

In addition the LET and Wtriggers (but not theZO trigger) required a coincidence with two signalsobtained from scintillator arrays coverning an angularrange 0.47 < e < 2.84° on both sides of the collisionregion. This additional condition is satisfied bynearly all non-diffractive collisions Early signalsmeasured in these scintillator arrays were used to tagbackground events induced by beam halo particles in­teracting in the detector.

3. Large Transverse Energy Hadrons

3.1 - The Central Calorimeter

In the present Section we restrict the analysisto events which satisfy the LET trigger. The centralcalorimeter is the part of the UA2 detector which ismost relevant to the present study. It is segmentedinto 200 cells, each covering 15° in ¢ and 10° in ¢,and built in a tower structure pointing towards thecentre of the interaction region. The cells (Figure 2)

are segmented longitudinally into a 17 radiationlength thick electromagnetic compartment (lead-scintil­la~or) followed by two .hadronic compartments (iron­sClntlllator) of two absorption lengths each. Thelight from each compartment is collected by twoBBQ-doped light guide plates on opposite sides of thecell.

The energy resolution for electrons is measuredto be 0t/E = 0.14 I E (E in GeV). In the case ofhadrons, 0E/E varies from 32% at 1 GeV to 11% at 70

GeV, approximately like E-1/4 . The resolution formulti-hadron systems of more than 20 GeV is similar tothat of single hadrons. Calibration of the energyresponse was done with PS and SPS beams, then trackedcar~full~. The systematic,uncertainty in the energycallbratlon for the data dlscussed here is less than2% for the electromagnetic calorimeter and less than3% for the hadronic one.

Simplifications in the very preliminary dataanalysis add further errors of ±10%.

3.2 - Two-Jet Dominance

Figure 3 shows the distribution of the observedevents as a function of their total transverse energy,EET > 100 GeV. The increased statistical accuracy isnow sufficient to give evidence of a departure fromexponential when EET exceeds 60 GeV. This departurecorresponds to the emergence of two-jet dominance atlarge values of EET' Immediate evidence for two-jetdominance is obtained from a simple inspection ofenergy distribution in the e-~ plane: examples areshown in Figures 4a to d.

Figures 5,6,7 show details of the jet distribu­tions.

4. Search for W+ e v Decays

,T~e decay into e±v of W± produced at rest in ppcolllslon would generate a monochromatic electron withenergy E(e) = ~ Mw' Mw being the rest mass of the Wb?$On: In practice,W's,are expected to be producedwlth lmportant longltudlnal momenta. This does notaffect the transverse momentum distribution of the de­cay electron which peaks near its end point at ET (e)= 1/2 Mw (Jacobian peak).

W's are also expected to be produced with impor­tant transverse momenta, similar to that of a two-jetsystem of a same mass (see Figure 7; however part ofthe measured transverse momentum is of instrumentalorigi~). This results mainly in a smearing of theJacoblan peak but does not induce significant correla­tions between the transverse momentum of the decayelectron and that of the W(or of its associated re­coil particles). We shall therefore search for largetransverse momentum electrons which are not accompaniedby other particles at small angles to the electron mo­mentum. This simplified approach will strongly reducea possible two-jet background and should not affectseriously the W+ e v signal. However it precludesany search for large transverse momentum electronsamong jet fragments.

We shall deal separately with the central and for­ward regions in which different experimental methodsare used. In both cases the initial event sample isthat of W-triggers.

4.1 - Search for W+ e v in the Central Region

The full data sample recorded using the Wtrigger

22

corresponds to a total integrated luminosity of~9.0 nb- 1 • A first event selection is made by search­lng for clusters of energy with the configuration ex­pected from isolated electrons. For each cluster sat­isfying some initial cuts a transverse energy E' iscalculated using the position of the cluster ce~troidand assuming that the event vertex was in the centerof the apparatus. The events with E' > 15 GeV (363events) are fully reconstructed and the exact locationof the event vertex (as measured in the vertex detec­tor) is used to obtain the correct value of the trans­verse energy ET. The ET distribution for these eventsis shown in Figure 8a. The falloff for ET < 17 GeVresults from the selection requirement E+ > 15 GeVapplied without knowledge of the exact event vertex.For E~ > 17 GeV the distribution of Figure 8a has nothresnold bias. There are 7 events with ET > 30 GeVthe highest ET value being 40.3 GeV. '

This sample is further reduced by requiring that?ne, and only one charged particle track reconstructedln the vertex,detector points to the energy cluster.Such a cut, wlth the ET distribution shown in Figure8b.

, We then require that the track produce a showerln the t~ngsten converter, with an associated chargecluster ln chamber Cs , as expected in the case ofelectrons. This condition applied to our sample re­du~es'~he number of events from 96 to 35. The ET-dis­trlbutlon for these events is shown in Figure 8c.

In order to further ensure that the electroncandidate is isolated, we require that no other chargeclusters are present in Cs in a cone of 10° half-aper­ture around the track. Such clusters could resultfrom the conversion of high-energy photons accompanyingthe charged particle in the tungsten. We estimate thatthe loss of events due to radiative corrections is lessthan 6%. .

, :en events satisfy this requirement. The ET dis­trlbutlons for these events are shown in Figure 8d.

As a final selection criterion we use the highsegmentation of the central calorimeter to check ifthe shape of the energy cluster is consistent withthat expected from an isolated electron impinging alongt~e direction of the observed track. The E distribu­tlon for these events, which represent our final sampleof electron candidates, is shown in Figure 8e.

As a check of this analysis all of the initials~mple o~ 363 e~ents was carefully scanned by physi­C1StS uSlng a hlgh resolution graphics terminal. Thesame three events were found.

There are three main sources of background con­tamination:

a) Single, isolated high-PT nO-or n-mesons un­dergoing Dalitz decay, or high-PT photons (both singleand from nO(n) + yy decay) converting in the vacuumchamber wall. The expected number of background eventsfrom this source is < 0.04 for ET > 25 GeV.

b) Single high-PT charged hadrons interacting inthe tu~gsten converter and depositing a large fractionof thelr energy (> 90%) in the electromagnetic calori­meter. We estimate that the expected number of back­ground events from this source is < 0.04 events forET > 25 GeV.

c) "Overlap" events, consisting of either atl~ast one high-PT pho~on accompanied by a charged par­tlcle or of several hlgh-PT photons of which one con­verts. We estimate this background to be approximately

/) ,

)

)

)

)

o

o

o

C)

o

()

()

u

,)

0.1 event for ET > 25 GeV.

In conclusion, the total background contributionto the three electron candidates amounts to less than0.2 events for ET > 25 GeV. Furthermore, backgroundsources would have an ET-distribution similar to thatof Figure 8a, whereas the distribution of the threeelectron candidates is inconsistent with it. This isan independent indication that the background contri­bution is indeed small.

4.2 Search for W+ e v in the Forward Regions

The search for W+ e v decays in the forwardregions follows the same general guidelines as in thecentral region.

Each of the two forward detectors covers a rangeof scattering angles 20° < e < 37.5° and ;s instru­mented in 12 azimuthal sectors with magnetic spectro­meters (Figure 9). The magnetic fields are generatedfrom two "lamp-shade" magnets of 12 coils each. Thefield integral is 0.38 T m on the average and the lossof azimuthal coverage caused by the magnet coils is

18%.Depending on the sign of their charge, particles

are bent towards or away from the beam and are mea­sured in three triplets of drift chambers, havings~nse ~ires at _7°, 0° and +7° to the magnetic fielddlrectlon. In the present preliminary analysis the

. resolution achieved in momentum measurement is~(i) ~ 2% GeV-l. Each magnetic spectrometer sector is

follo~ed by a preshower counter and an electromagneticcalorlmeter. The preshower counter consists of a 1.4radiati?n length thick lead converter preceding twoproportlonal tube chambers. Each electromagnetic cal­orimeter sector is subdivided in 10 cells, each cover­i~g.a rapidity x azimuth domain ~y x M~0.17 x 15°slmllar to that covered by the central calorimetercells .. E~ch cell is s~gmented in two compartments, one24 radlatlon length thlck in which most of electromag­netic showers are contained, the other 6 radiationlength thick serving as a veto against hadronic showers.The response and performance of the preshower countersand calorimeters have been extensively studied in a10 GeV electron beam. A comparison of the two photo­tube signals receiving light from the wave-shiftinglight guides on each side of a cell allows the locali­zation of the impact point to within ±2 cm.

The following are the main selection criteriaapplied to the sample of Wtriggers to select possibleelectron candidates:

- a forward track with more than 10 GeV associ­ated transverse energy.

- leakage in the calorimeter veto compartmentmust be less than 2%.

- the quality of matching between track, pre­shower counters and energy in the calorimetermus t be good.

- the momentum p measured in the magnetic spectro­meter must be consistent with the energy Emeasured in the calorimeter.

- the electron candidate must be isolated: theenergy measured in adjacent calorimeter cells(including the contribution of track momenta)must not exceed 3 GeV, no other track mustpoint to the same cell as the electron candi­date and no other signal facing this cell mustbe measured in the preshower counter.

23

Ten events are found to survive these straight­forward selection criteria for an integrated lumino­sity of 16 nb- 1 • The corresponding ET distributionis shown in Figure 10.

Backgrounds in this sample are expected to origi­nate from the same sources as described in Section 2.Above ET = 15 GeV, 3 events survive (see Figure 10)and we estimate that less than 0.2 events could be abackground event.

4.3 Missing Transverse Momentum, for ET > 15 GeV

We now verify another feature of reaction W+ e vnamely, the presence of an undetected neutrino withtransverse momentum similar in magnitude to that ofthe electron but opposite in azimuth.

In order to estimate the missing transverse mo­mentum carried away by the neutrino we reconstructthe total momentum vector from the available calorime­tric and spectrometric measurements. To each calori­meter cell, including those in the lead glass wall,we assign a vector p' with magnitude equal to theenergy deposited in the cell and direction along theline joining the event+vertex to the cell centre. Thetotal momentum vector p is then projected onto thedirection of ·the electron transverse momentum vectorPeT to define the missing transverse momentum PTm1ss

The ratio p~iSS IE (where ET = It TI is shownin Figure lla for the eTectron candidat~s found bothin the central calorimeter and in the forward detec-tors~ The events with p~iSS lET 1 are consistentwith reaction W+ e v.

The ET distributi.on of the four events with p~iSS

lET> 0.8 is shown in Figure llb. These electron can­didates are those having the highest ET values.

In order to verify that the large missing trans­verse momentum of the four electron candidates is notdominated by the limited coverage of the detectors, wehave studied a sample of background events obtainedwith the original Wtrigger. This control sample isdominated by events containing a high-E jet,whichare not expected to show a large missin~ transversemomentum. We find that the fraction of such eventswhich satisfies the condition p~iSS lET> 0.8 is only

20% for both the central calorimeter and the forwarddetectors.

Figure llc shows the cell energy distribution ine and ~ for the highest ET event. The only signifi-cant energy flow within the detector acceptance iscarried by the electron candidate. The other threeevents have the same spectacular configuration.

5. Conclusions

The application of the various selection criteriaused in the analysis to identify single, isolatedelectrons has reduced the original event sample to fourevents with ET > 15 GeV and large missing transversemomentum. Both the configuration of the events andtheir number are consistent with W+ e v. Detailedanalysis of all the information obtained from a care­fully designed detector allows the observation of avery small signal in a very large sample of interac­tions.

References

1. M. Banner et al., Phys. Lett. 118B (1982) 203.

2. M. Banner et al., Phys. Lett. 115B (1982) 59.M. Banner et al., Phys. Lett. 121B (1983) 187.M. Banner et al., Inclusive charged particle pro­duction at the CERN pp Collider, to be publishedin Phys. Lett.

3. R. Battiston et al., Phys. Lett. 117B (1982) 126.

TUNGSTEN CONVERTER --=::,~~IIlIl!IIII11I1!!IIIII::;~~~,AND CHAMBER_C=--5 _ * . _

FORWARD CALORIMETER

')

)

.. )

)

Fig. 1. UA2 Detector

)

)

200ISDIDD

tEl IlitVl

\,0-

),~

>~~ ,01':::;1.:;

10'

,0*

Fig. 2. Central calorimeter cells

Fig. 3. Observed (uncorrected) EEr distribution inthe central calorimeter.

24 )

TRANSVERSE ENERGY DEPOSITlor~

o

o

, l.U... ILl'\ "'1lH[ReT A80Yt ~.D CE'I- ----.e m::m

UA'l

. 'IN[RcY AliM M CEV...... ­--1J ':: tUft

~~ I10 GEV(RIIlUCED)

UA2

o

o

. a) b)

()

TP.ANSVERSE

[NER::Y ISCI.:.!

OGEV

ENERGY DEPOSIT~9~~~. , ~;,. .... :.~. «.1"

ENER::Y AliOVi: ~.D CE\I............'DC'\AI"" OC'~

'~ 1ft lUll

UA2

TRANSVERSE ENERGY DEPOSITION!'!'Dr , 1:-'0)-", ..... rll

o

c) d)

u

Fig. 4. Typical 8-¢ distributions of the transverseenergy in large EET events.

100

~

!so

'06C> IdeOrtu)

)

Fi g. 5. Azimuthal separation ~¢ between two jets,each having ET > 15 GeV, in events withE ET > 30 GeV.

25

)

Fig. 8. Transverse energy distribution of the eventsamples in the central calorimeter:a) After requirements on energy cluster;b) After association with a track;c) After association of the track with a

shower in the tungsten converter;d) After requiring only one such shower;e) After further cuts on the quality of

the track-energy cluster matching.This is the final electron sample.

TOO 150

"H IG.V/~1

SO

100 nSO

.100

~10 al

60 1363 eventsl..5" 4

j ')20

10

64

)40 60 ao lOO

E, (G.VI

20 bl

Observed (uncorrected) transverse energy> 196 events)

Fig. 6..,

l:> 10distribution of jets having ET > 40 GeV. N.... 6OIl

C4 U.,

>....2

IIi-) • ao G.V/~1086 )

42

,0242

':!:TO 20

=t PI Ij-jl4

e) )'" 13 eventsl~ 10' 2

i 10 20 50 60~...

Fi g. 7. Two-jet invariant mass (M .. ) distributionJ-J

(uncorrected). The observed (uncorrected)transverse momentum distribution of two-jetsystems having M.. > 80 GeV/c 2 is shown inthe insert. J-J

\. /

26

!O

o

o

o

fltld "'" pilch5 mm

r.lI "'dth! 5 , ..

H.!nelk 1I.ldfDd!, 0.38 Tm""---.

Prf(onvtrttr1.4 •.1. Pb/fo

m.1I • Ilmm S,. Imm PbJ

CALORIMETER

o Fig. 9. Forward detector sector.

o

()

.20

[T IGtVI

Fig. 10. Transverse energy distribution of the 10electron candidates in the forward detec­tors.

4

2

o

6

4

2

al

bl

·0 Central. calorimeter

~ Forward detectors

o Central calorimeter

~ Forward detectors

()

Fig. 11. a) p~iSS distribution for electron candi­

dates.b) ET distribution for events with

p~i ss lET > 8.c) Theta-Phi distribution for highest ETevent.

27

o

c)

ENERGY ISCALE

10 GEV

CROSS SECTION ESTIMATES FOR MULTI-TEV PHYSICS

This note provides rough estimates for cross

sections of interest in the multi-TeV regime.

Summary

Introduction

Robert N. CahnLawrence Berkeley Laboratory

University of California, Berkeley, CA 94720

of about 65 mb.2 Still, it is possible to produce a fit

with all the required analyticity which while giving

a In2 s behavior at present energies, eventually be­

comes constant.3 These two kinds of fits give very

different predictions at -./8 = 20- 40TeV. Calling the

pure In28 fit # 1 and the one which becomes constant

asymptotically #2, we have

Lepton Pair Production

It is worth noting that even for the total cross sec­

tion where we have excellent low energy data and

the theoretical constraints of analyticity, when we ex­

trapolate to the 40 TeV region, there are uncertainties

of 100%. Similar extrapolations 4 for the slope para­

meter of elastic scattering, B, suggest that its value

which is about 13 GeV-2 at the ISR will be about

twice that at -./8 = 20 TeV .

Lepton pair production has been a tremen­

dously important process because it has led to the dis­

covery of the JN and the 'I' and because the con­

tinuum production is an especially important process

theoretically. Using a crude model for the quark dis­

tributions, it is possible to obtain a simple estimate!

for the cross section to produce a dilepton pair with

an invariant mass mp/i = -./rs :

In designing detectors for the multi-TeV re­

gion, it is worthwhile to consider what cross sections

might be anticipated on the basis of present experience

and theoretical prejudice. Such projections must be

treated with a good deal of skepticism since they re­

quire very large extrapolations from the ISR-FNAL

energy range or slightly lesser extrapolations from the

fragmentary SPS Collider data. Even worse, some

predictions are founded nearly entirely on theoreti­

cal pictures which have not been extensively tested.

Nevertheless, this is the best we can do. Depending

on the process, the uncertainties may be as little as a

factor of two, or as much as an order of magnitude or

two.

It should be emphasized that the projections

given here are not only very approximate, but further­

more do not represent what is most likely to be inter­

esting at these very high energies. After all, we want

to look for new phenomena, not just the standard

catalogue of jets, dileptons, and the like. Nevertheless,

these mundane items do furnish us with benchmarks,

standards for the cross sections which are potentially

interesting and perhaps examples of signatures which

could help identify truly new physics.

These estimates are based on my SLAC Sum­

mer School Lectures of 1982 to which the reader is

directed for more complete discussion and references.!

J80.54 TeV

2 TeV

20 TeV

40 TeV

Fit #1

71 mb

100 mb

173 mb

200 mb

Fit #2

66 mb

82 mb

107 mb

114 mb

/)

: )

)

)

)

Total Cross Section

Ever since the rise the in pp cross section

was first discovered at the ISR, it has been popular to

interpret the data as saturating the Froissart boundform, In2 s. Such a fit is indeed possible and is not

inconsistent with the measurement at the SPS Collider

28

( _ (540GeV)2a mp/i > -'/TQ8) f'::j 10 37cm2 _ 2-20fiQ •mpp

We see that the last factor is nearly unity at

20 TeV for reasonable cross sections, so the behavior

is nearly as m;i. The cross section to produce a pairwith mass greater than 180 GeV is about 1O-36cm2 .

)

o

Wand Z Production

High Transverse Momentum Jets

If past history is a guide, Wand Z productionwill be an important background for something inter­esting. Their masses are small enough compared tothe available energy at multi-TeV machines that thereis little energy dependence in their production crosssections. Using quark distributions appropriate to theSPS Collider without modifications for possible non­scaling effects, we find that the cross sections shouldbe roughly 5· 1O-33cm2 for each species in either ppor pp machines.

The early data from the SPS Collider in­dicate that high transverse momentum jets do be­come prominent, well-defined features at such energiesin contradistinction to the situation at ISR energieswhere jets are often obscured by fiuctuations in thenon-jet events. QCD predicts cross sections for theseprocesses but with great uncertainties. The uncer­tainties arise from two sources: the quark distributionsand radiative QCD corrections to the basic partoniccross sections. To get a crude idea of the cross sec­tions, we finesse these problems by ignoring them: we

use scaling distributions for the quarks and lowest or­der calculations for the cross sections. 1. R. N. Cahn, Expectations for Old and New Physics

at High Energy Colliders, LBL-15469 and forthcom­ing Proceedings of the 1982 SLAC Summer Institute.

2. M. Haguenauer, UA-4 Collaboration, Journal dePhysique, 43, 21st International Conference onHigh Energy Physics, Paris, 1983, p.579

3. M. M. Block and R. N. Cahn, Phys. Letters 120B,

224(1983).

4. M. M. Block ·and R. N. Cahn, Phys. Letters 120B,

229(1983).

5. B. L. Combridge, Nucl. Phys. B151, 429(1979).

. Heavy flavor production has been considered

by Combridge in a QCD model.5 If we restrict ou~selves to the two mechanisms gluon+gluon -+ Q Qand qq -. QQ, it is clear on dimensional grounds thatthe production cross section must go as Ci~ / M2 whereM is the heavy quark mass. There may in addition bea factor representing the suppression for making thequarks at finite 8. At the multi-Tev machines, this

suppression is not important for M < /8/100, sowe expect a cross section nearly equal to the naive es­

timate above. At more detailed analysis suggests thatthe numerical factor which multiplies the dimensionalquantity is about 5. For M = 40 GeV , this gives across section of about 5· 1O-32cm2 •

Heavy Flavor Production

References

Typical results are shown in the Figure. Thecurves show the cross section to produce a jet with

transverse momentum greater than a given amount,considering only the subprocess gluon + gluon -. gluon

+ gluon. The solid curve is for /8 = 20TeV and thedashed curve is for /8 = 40 TeV.

" " " " " " " " "" " "-

"""-\

\

" " " "

" "- 110

- 210

- 310

- 410

- 510

~ - 6'-' 10tl

- 710

- B10

- 910

-1010

10

o

o

o

o

o

o

,)

jet transverse momentum (GeV)

29

MONTE CARLO SIMULATION OF HIGH PT JETS

Frank E. Paige, Serban D. Protopopescu, and Dennis P. WeygandBrookhaven National Laboratory, Upton, New York 11973

We discuss the properties of high PT jets

expected at high energy hadron colliders.

Perturbative QCD predicts that high PT hadronic

reactions should contain jets of hadrons from the

fragmentation of quarks and gluons. Evidence for jets

with about the predicted cross section as been found

at the ISR1, although the jets are not clearly separ­

ated from the background of low-PT particles. At

the SPS Collider, the jets stand out dramatically2,

and some clear multi-jet events are seen. Since jets

are expected both from QCD hard scattering and from

the decay of heavy particles, we will surely want to

study them at any future hadron collider.

It is easy to guess the qualitative properties of

jets at very high energies. The mean kT of parti­

cles with respect to the jet axis must be determined

by the jet PT' the only scale in the problem:

approximation keeps only the parts of the matrix ele­

ments which dominate in the collinear regions but uses

exact, non-collinear kinematics. Since the parts re­

tained involve no interference terms (in the right

gauge), they can be recast as classical probabilities

for a quark or gluon to split into two. Thus this ap­

proximation is very suitable for a Monte Carlo ap­

proach.

Retaining the dominant terms in the collinear re­

gion gives the correct leading-log scaling violations

in the jet fragmentation and qualitatively the right

cross section for multiple jets at large angles. It

also leads to a very rapid increase in multiplicity6:

')

)

dN dNz dz = dy - 10

implying a significantly higher density of particles

along a jet than in a low PT beam jet. While the

To avoid generating too many particles, the 11°' sand

n's have not been decayed at the highest energy.

Fig. 1 shows the z distribution for all parti':"

cles, charged and neutral, in a jet. At small z,

)

o(cm2)

5.8 x 10-34

3.0 x 10- 34

1.1 x 10-35

Is (GeV) PT (GeV)

800 100- 110

2000 200- 220

10000 1000-1100

A more correct calculation7 summing all powers of

as(PT2)~n2 PT2 gives a similar form but with the

coefficient of [~n PT211/ 2 decreased by a factor

of 12. The physical reasonS is that the leading-log

approximation treats all gluon emissions as indepen­

dent, while a gluon of momentum q can only be radiated

when its parent parton is separated from the others by

a distance of order l/q. Hopefully, a more correct

treatment of soft partons will be included in a future

version of ISAJET.

We have generated three samples of high PT jet

events at energies for which the cross sections are

large:

Hence the mean opening angle is

This opening angle is of course a consequence of the

radiation of extra gluons and quark-antiquark pairs.

As PT increases, more and more of these extra par­

tons will have energies large compared to 1 GeV and so

will be observable above the background of soft parti­

cles. Therefore the number of visible jets will in­

crease while the opening angles between jets decrease

very slowly.

To make these remarks more quantitative, we have

produced samples of high PT jet events using ISA­

JET3 , a Monte Carlo event generator for hadronic reac­

tions. The program uses an approximation to perturba­

tive QCD introduced by Fox and Wolfram4 • In lowest

order, QCD predicts just two jets. In higher orders,

extra gluons and quark-antiquark pairs can be pro­

duced, giving multiple jets. The multi-jet cross sec­

tions are suppressed by extra powers of as (PT2)

except when the partons are collinear. The collinear

regions give contributions of order as(PT2)~npT2

producing the leading-log scaling violations of the

jet fragmentation functions5 • The Fox-Wolfram

30

o

3. F.E. Paige and S.D. Protopopescu, BNL 31987, to

be published in the Proceedings of the DPF Work-

shop on High Energy Physics and Future Facilities

(Snowmass, 1982) •

4. G.C. Fox and S. Wolfram, Nucl. Phys. B~, 285

Fig. 1:

~

Na

t:-o.~

+>uco '0

Lt

1.0

z(dN/dz) VB z. (a) PT = 100 GeV.

(b) PT = 200 GeV. (c) PT = 1000 GeV.

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

Z

(a)

c

c.....•••

II!!!

tt ft1t t

11 Iljl II

'0

'0

'0

References

(1980).

J •• Owens, Phys. Letters 76B, 85 (1978);

T.A. DeGrand, Nucl. Phys. Bli!, 485 (1979).

D. Arnati, A. Bassetto, M. Ciafaloni,

G. Marchesini, and G. Veneziano, Nucl. Phys.

B.!Zl, 253 (1979).

A.H. Mueller, Phys. Letters 104B, 161 (1981).

Yu. L. Dokshitser, V.S. Fadin and V.A. Khoze,

Leningrad-82-789 (1982).

Axial Field Spectrometer Collaboration, Phys.

Letters 118B, 185, 193 (1983).

UAI Collaboration, CERN-EP/82-134 (1982);

UAZ Collaboration, Physics Letters 1I5B, 59 (1982).

7.

8.

5.

6.

1.

2.

Acknowledgment

program is tuned to give the right multiplicity at

PETRA, the cascade of soft par tons will produce an

overestimate of the density of low-z particles at

higher energies.

Fig. 2 shows the energy-weighted opening angle

distribution with respect to the direction of the

original parton. Fig. 3 shows the multiplicity as a

function of this angle. These plots are based on the

jets as defined by ISAJET, not on jets reconstructed

by some algorithm.

Fig. 4 shows the charged particle multiplicity

distribution in a jet. The multiplicity obviously

grows very rapidly with PT' While this growth is

overestimated, nevertheless high multiplicities are

expected.

Figs. 5 and 6 show the distributions in opening

angles between pairs of particles in a jet and between

pairs each having PT > 3 GeV. Figs. 7 and 8 show

similar distsributions on a finer scale for the charg­

ed particles. These plots indicate the segmentation

needed to track individual particles in a jet. While

such tracking is not very important for the gross fea­

tures, it is essential if one wants, for example, to

look for prompt leptons indicating the presence of

heavy quarks.

At very high energies, jet events contain many

sub-jets, so qua)1tit;l.es like the mean opening angle

are not very meaningful. In Fig. 9 we show the first

ten events generated at ,Is = 10 TeV with PT = 1000­

1100 GeV and Iyl < 1. The particles from these events

have been put into a simulation program for a calori­

meter segmented into towers with 6y = .1 and 6~ = 5°.

The energy in each tower is indicated by a solid line,

and the energy of each parton is shown by a dotted

line. In these ten events there is one with seven

distinct jetsl For a typical event the program gener­

ates about 50 partons, so the visible jets are not

coming from partons close to the infrared cutoff. Of

course when extrapolating to such high energies we can

not expect to predict correctly such details as the

rate for seven jet events, even if we had a good algo­

rithm for defining jets. Nevertheless, it seems like­

ly that events in this PT range will show a rich and

complicated structure.

The submitted manuscript has been authoredunder contract DE-AC02-76CHOOOI6 with the U.S. Depart­ment of Energy.

o

u

()

o

o

o

o

o

31

'0,,..- ----,

Fig. 2: z(dN/d6) VB 6.

7~ SN

~,<:0........,U

'" ':'s... SlJ..

~

N"t:i,Z"t:l~ ':'N S

'0

(b)

0

0

,.......~H

tTtt t

jljl!11Ii

0.0 0.1 0.2 0.3 O.i 0.5 0.6 0,7 0.8 0.9 1.0

Z

~

No,<:o........,u

'"J: ~

+~ (a)

~ ~++H

t+tHt+

H+ H++tT

t Tt t tttTtttT T

t +tT tT Ht tt tt

r)

r)

,)

~

N_~'o,­<:o........,U

'"s...lJ..

'0

(c)

,

,

••~

t+ttt t t

! II jIll

'0-f--r---r-..,.--r---r-..,.--,...----,--...,.----l0.0 0.1 0.2 0.3 O.i 0.5 0.6 0.7 0.8 0.9 1.0

e(rad)

,)

0.0 0.1 0.2 0.3 O.i 0.5 0.6 0.7 0.8 0.9' 1.0

Z

32

o

o

o

o

o

o

o

u

u

,)

(c)

~ 7N S0

........<=

++tt+0.~.....u Hn:>I-

H+hHttt+tl.L.

~

<D 0

tt ttttttttt"0

tt........

ttz"t:J

t t ttt~

N

tt ttt tttt

~0

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

a(rad)

Fig. 3: dN/dO VB O.

00-r- --.

~

No........U1QJ

·O+--r_---.,-.....,.-_,_-...,_--r--r_---.,-.....,.---l0.0 0.1 0.2 0.3 0.4 0.5 0.6 0,7 0.8 0.9 1.0

a(rad)

33

-0.,.... --.

:(b)

N0

........U1QJ..... °0U.~.....l-n:>c.

'0-t--~-r__.....,.-_,_-...,_-_r_-r___,-.....,.--l

-0-r- -,...----.(c)

N

c:........U1QJ.....

°0U.~.....l-n:>c.

'O+--~___,r__.....,.-_,_-_r_-_r_-r___,-.....,.---i0.0 0.1 0.2 0.3 0.4 0.5 0.6 0,7 0.8 0.9 1.0

a(rad)

Fig. 4: Charged multiplicity.

~

'0-l--..,.....-r----r---r---r-L...J.-r--,.----,r---.,.--f0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0

nch

~

°(b)

it jttttt j jII I II HI j

tt t tc 'l'

I jI I0

°........UtU~

l.L

0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0

nch

34

·0.,.... --,-..,........(c)

':'O+--.,.....---,r----..--r-...,....L..-..,...--.,.J.;----,--,J_-l0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0

nch

Fig. 5: Pairs vs 912'

-0..-- ...,(a)

mmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmm.. '.

....o.......Vl~ 0.... °~

'0-1--..,.....-,......--.---.--...,..--.--..,.....-,......--.--10.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

6 12

)

:)

)

)

)

)

o

0, -;--:-_,(b)

Fig. 6: Pairs each with Pr > 3 GeV vs 6.

o

o

o

o

~

M0

..... -0Vls.-It1

0..

~

N.....CD"0.....2:"0

°O+__,--_--r__..,.....;..-,-,_....,.__,.._--r__..,...__,--_-l0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

S12(rad)

0, ------ ...,(c)

-0....- --,

(a)

M0

.....Vls.- ".It1

..0.. ..

•~ °0 ••N.....

•••CD"0..... ••z: •"0 ••• •H

+++++!++!!!!+!+!+++++!++++

7

° 0.00 0.05 0.10 0.15 0.20 0.25 .0.30 0.35 0.40 0.45 0.50

S12(rad)

o Mo.....Vls.-It1

0..

mmmlllmlllmmmmlllmillmmmlllmlllm

. mmmmmillmmmmmm lllmmlllmmlllm

mllIllI III lIl lD lD ell.-0.,... ...,

(b)

.CII al

illm...

~

N.....CD"0.....Z"0

N.....CD"0.....z:"0

-O+--,-----r--..,...--,--_..,..__..,..._-,,......-....,.--,..--I0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

S12(rad)

~

M

~.....Vls.-

°0It1

0..

....".....

lb 4l ¢l III

•

35

'0-t--,..---,---,,--r--""T"'"--,----,--;---,---t0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

S12(rad)

0-,- ----,

(C)°0:-,- .....;.. -.

(b)

~

C'l....<0"C-.....2:"C~

~

~-..... -0III~

rtlc.,

...... .. ......... ..lIllDlIlm

IlIlDlDmllllJlllllDlIIllI

III 11I 11I lIllDlIl lDlII

III III III III III III III III

~

oo-.....III~

rtl0.. '0

/)

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

612(rad)

~+-_r_-_r_-,.._-,_---,r____r-__,.-_._-_._-__10.00m.005 0.010 0.015 0.020 0.025 0.030 0.035 0.040 0.045 0.050

612(rad)

:)

Fig. 7: Charged pairs VB 812.

-0:-.-------------------:--:--,(c)

)

)

0.00m.005 0.010 0.015 0.020 0.025 0.030 0.035 0.040 0.045 0.050

612(rad)

'0 -l-_...,-_ _r_-,.--,_---,r____r---r---,---r---l

~

C'l ­..... '0

<0"C-.....Z'0~

°0

(a)

~

0

~-..... ~

+ ++++ + ++++++++++++++III 0~

tH+++++++rtl+H + +

0-

~ tttttC'l

ttt tt....<0"U-.....

jll I"'""C ~0

'0+--_r_-,.--,_---,r____r---r---,---r--,---!0.00m.005 0.010 0.015 0.020 0.025 0.030 0.035 0.040 0.045 0.050

612(rad)

36 )

o

Fig. 8: Charged pairs with Pr > 3 GeV vs 812·-0:;- -......,;----- ---.

(c)

o

'0-l--r---r---y---r--r----r---y--.,.--,..---f

0......0

~......VIS-

({j ':'ll. S~

'".-<<D"'0......z:"t:l

~......oo......VIS-

({jll.

11111111111111111111111111111111, I IIII '111 1 " 11 I

T

!

0.00lD.005 0.010 0.015 0.020 0.025 0.030 0.035 0.040 0.045 0.050

812(rad)

0.00lD.005 0.010 0.015 0.020 0.025 0.030 0.035 0.040 0.045 0.050

812(rad)

'O-+--~--r---Y---r--r---r---Y---r--r--lo

o

Fig. 9: Calorimeter plots of 10 events with

Pr = 1000 GeV.

QO...- ---:":-:---,

(b)

0

......00

...... HVI ~

S- ° HtHf t({jll. tt t()~

'".-<tt

<D"0......z:"0

jI,)

"00.00lD.005 0.010 0.015 0.020 0.025 0.030 0.035 0.040 0.045 0.050

812(rad)

1000.0

800,0

~

>. 600,0~t.J

100,0

'0°,0

830302 - 180652 - 1

.J

37

')1000.0 830302 - 180652 - 2

\000.0830302 - 180652 - 58()0.0

1100.0;;:

:)~

600.0

;;:600.0'" ~100.0

'" 100. 0100.0

loa'O

~)

0.0

0.0

()

)

1000.0 830302'- 180652 - 31000.0 830302 - 180652 - 6

800.0

800.0)

~

~" 600. 0

" 600. 0~

~'"'"

100.0

100.0100.0

100.0

)0.0

0'0

)

1000•0 830302 - 1806F2 - 4 1000.0 830302 - 180652 - 7

)800.0800.0

~

;;:600. 0

" 600.0

~~

'" '" 100.0

100. 0

100.0100.0

0.00'0

38

)

0

1000.0 830302 - 180652 - 8

800.0

0 ;:800. 0

~t..J

100.0

'00.0

0 0.0

o

1000.0 830302 - 180652 - 9

0800.0

,'" 800.0~t..J

100.0

0 '00.0

o

U 1000. 0 830302-180652-10

800.0

,'" 800.0~

~)t..J

100.0

'00.0

39

:)

: )

)

)

)

o

o

o

o

o

o

, \,-y

'-)

SECTION II: WORKING GROUP SUMMARY REPORTS

)

)

)

o

o

TRACKING DETECTORS

Dona1d Hart i 11

Working Group Summary Report

D. Harti11, M. Atac, W. Bartel, M. Breidenbach, A. Bross, W. Carithers,E. Gatti, D. Greiner, D. Herrup, B. Kephart, D. Levintha1, T. Ludlam,

M. Mestayer, R. Ong, S. Parker, E. Platner, V. Radeka, P. Rehak, F. Sau1i,P. Schlein, K. Shinsky, D. Theriot, and A. To11estrup

o

o

o

u

I. Introduction

The problem of tracking charged particles in thecores of high energy jets in a high rate environmentis a difficult one and will require a great deal ofRand D to solve in an affordable way. During theworkshop the tracking group concentrated on threeproblems to provide a focus for our efforts. Thesewere 1) the general vertex detector, 2) the centraland forward regions for 1 TeV pp collider with aluminosity of 1033 cm- 2 sec-I, and 3) possible track~

ing devices for a 20 TeV hadron hadron co11ider. Thelargest effort went into the second problem in orderto help assess the desirability of building such aco11ider. After one day of preliminary work the 20TeV subgroup joined with the 20 TeV Systems subgroupto form a larger group that could make some progresson a first look at the whole detector question at thisenergy. Most of their effort was devoted to modellingwhat jets at 20 TeV might look like in order to deter­mine detector design parameters. Their efforts aresummarized in the report from the Systems Group led byBarry Barish. The final section of this reportconcerns the research and development needs to provideadequate tracking for the next generation co11iderdetectors. -

II. 1 TeV with L = 1033 cm- 2 sec- I

The interaction diamond for a pp co11ider with1 TeV total c.m. energy and a luminosity Of)1033 isapproximately 40 cm long by a few mm wide. 1 Sincethe primary physics interest is in high transversemomentum phenomena a magnetic solenoid for the centralregion has several advantages. Because of the smalltransverse size of the interaction diamond, it ~an beused as a constraint (point1ike) for a fast hardwaretrackfinder for trigger purposes and facilitates usingfast software trackfinders to do rapid event filtering Ito find the interesting high PT events. The otherpossibility is to use a dipole field transverse to thebeam direction for the central region. This config­uration aids the detection of particles in the forwarddirection by sweeping them away from the beam line sothey can be detected. The disadvantage is that thevertex constraint from the beam in the bend plane isnow 40 cm in length and hence not very useful as aconstraint in fast trackfinding. What seems moresensible is to construct a hybrid system using asolenoid in the central region with open frame dipolescovering the forward and backward directions. Thegranularity requirements imposed by rate considera­tions and particle densities in jet cores will be thesame with either a central solenoid or a centraldipole implying no clear preference. Because of thefast triggering and trackfinding advantages we willchoose the solenoid with forward-backward dipoles asthe basis for determining the tracking requirements.

43

A. Central Detector

The minimal requirements for the central trackingdetector are:

i. Rapidity coverage y = ± 1.5 (26° ~ e ~ 154°)

ii. ~p/p = 10% at p = 100 GeV/c, sufficient tomeasure the charge

iii. Operation at L = 1033 cm- 2 sec- I

iv. Efficient detection of individual particles inthe core of a 100 GeV/c jet.

The luminosity requirement implies an interaction rateof 50 MHz with an average of 3.6 charged particles perunit of rapidity in the central region. This gives a500 MHz charged particle rate into the detector.Present drift chamber technology limits particlefluxes to 1 and 2 MHz per sense wire in chambers with::; 3 mm drift distances. 2J To satisfy this rate limiteach layer of a cylindrical chamber must have at least250 sense wires. The fourth requirement also impliesfine granularity as is illustrated in Figure 1 whichsummarizes the fraction of tracks missed due to over­lap in a cell for PT =40 GeV/c jet generated by theISAJET Monte Carlo assuming a total center-of-mass of0.8 TeV. As can be seen from the graph 300 - 500 cellsgive a 70 to 80% efficiency for seeing a track in aPT = 40 GeV/c jet. Figure 2 shows how this trackingefficiency scales with PT. The other way of deter­mining the granularity is to limit the number of eventsoverlapping during the maximum drift time in thechamber. For a drift distance of 2 mm this drift timeis 40 nsec giving an average of two overlapping eventsfor the 50 MHz primary interaction rate.

Figure 3 shows a drift chamber configuration thatwill satisfy our four design criteria. Assuming aspatial resolution of 200~ in the drift direction, 10%momentum resolution at 100 GeV/c can be achieved bymaking 45 measurements over a radial path length of130 cm in a 1.5 Tes1a field. The basic design issimilar to the AFS drift chamber3) in use at the ISR atCERN. Figure 4 shows the basic cell configuration inmore detail. The sense wires are all parallel to thebeam axis and the z-coordinate is obtained by chargedivision to 1% on every sense wire to aid in patternrecognition. The layers are grouped in four rings withthe inner one containing fifteen - _ sense wire layerswith the inner most layer at a radius of 20 cm from thebeam line. The second, third, and fourth rings contain15, 10, and 5 layers respectively. To minimize thetotal number of sense wires, the drift distances alsoare increased to 3 mm, 4 mm, and 6 mm respectively.The 1% charge division requirement imposes a minimumgas gain of greater than 105 giving an avalanche sizeof "" 3 x 106 electrons. For the inner layers theparticle rate is 1.7 MHz per wire which leads to acurrent of "" 1 ~A per sense wi re and "". 25 ~A persense wire in the outer layers.

1,o,.-------,,.--------r-------....---------,

)

)

;;)

Track resolution in JETSPT (JET) = 40 GeV/c

0,8

0.2

1'sen 0.6

~....'0§ 0.4

'fif

LL

0 ......----:-:10~0~------:7::-::-----~:-=-:::-:=------:1:':'00::-'..000

Fig. 1. Track Resolving Inefficiency vs. Cell Density

,)

O,8r---.--.----,-r-r-T'TTTTT---r-I'""'"T~T"T'"1,.,.,

'g:g 0.6

's~(J

f.... 0,4'0co'fif 0.2

LL

)

)

. )

Fig. 2. Track Inefficiency for Different PT

To improve the z resolution from the ~l cm valuein the first ring and ~4 cm value in the fourth ring,one layer in each ring is equipped with a cathodestrip system. Figure 5 schematically illustrates thissystem. The strips are a few mm wide and 16 mm to48 mm long (covering four sense wires) and connectedtogether by resistors to form a pad. If each end ofthe resistor chain is connected to a charge sensitiveamplifier and the pulse height is measured theavalanche position along the sense wire can bedetermined to ~ 1 mm. Three thousand pads per speciallayer on the inner two rings and six thousand pads perspecial layer on the outer two rings would provideuseful granularity for high PT jets.

Table 1 summarizes the number of wires andgeneral properties of each ring. For the sense wireseach end must be instrumented with an electronicchannel which must measure a time to better than2 nsec accuracy and a pulse height to a part in 1024(10 bit resolution) to provide the 1% z coordinatemeasurement by charge division. The 10 bit precisionis necessary to accommodate the large dynamic range ofthe pulses from the sense wires due to Landaufluctuations in dE/dx and the large angular acceptanceof the central detector. Although by existinggeneral-purpose-detector standards, this device isnearly an order of magnitude increase in the number ofelectronic channels and mechanical complexity, itbarely meets our design goals stated in the beginning

" )

44 )

()

T:::1:::::; R;,,4. a=6mm

-- Cathode pads: I1z E! 1 mm

I I

.~£ Drift wires, charge division: u(> = 200 I'm

Uz = 1%

()

()

o

"./ -'" :Y

....--/ -....-- -

~,-

25· --f I I

1 2 3z (meters)

Ring3a=4mm

~Ring2~a=3mm

B. Forward Tracking System

The forward direction covers the rapidity intervalfrom 1.5 to 7.0 (the kinematic limit) corresponding topolar angles from e = 260 to e = 0° and typically hasmore than twice the number of charged particles perevent as compared to the central region considered inthe previous section. In addition, to measure a givenPT = p sin e, larger momenta must be accuratelymeasured. As in the case of the central detector, weassume an upper limit of 2 MHz per sense wire, amaximum drift distance of 2 mm, 3.6 charged particlesper unit of rapidity and a 50 MHz interaction rate.Transforming from rapidity to distance from the beamline we find that the nearest sense wire.in a planenormal to the beam line can never be closer than 20 emto the beam line. To get to a larger rapidity(smaller e) you have to go farther from the interactionpoint. As mentioned previously, a dipole fieldtransverse to the beam line will aid in sweeping theparticles from the beam direction in addition tomeasuring the particle momentum.

Figure 6 illustrates a forward detector whichcovers the rapidity interval from 1.5 to 5.5(neglecting the bend of the dipoles) using two windowframe dipoles to be used with the central solenoid.If a central dipole is used the smaller first dipolecan be omitted since the central dipole provides thesweeping function of this magnet. Figure 7 shows theorientation of the sense wires in the plane perpendic­ular to the beam line. The design is a mini-drifttype similar to the MPS chambers at Brookhaven with a2 mm drift distance instead of the 3 mm drift used in

Partical End View of Drift Chamber withBlow-up of Sector

o

o

o

I)

Fig. 3. Quarter Section of Cylindrical Drift Chamber

of this section. The UAI central detector operatingat the CERN pp collider with a total cm energy of 0.54TeV has an effective granularity of 1.5 cm in thedrift direction. They plan to decrease that to 5 mmby installing shaping networks on their preamp inputsand feel that this will still be too coarse agranularity to carry out their next round of physicsanalysis in a reasonable way. Their experienceconfirms our feeling that the proposed centraldetector is a minimal detector to do the job.Certainly ease of trackfinding and better particleisolation in jet cores in the off-line analysis wouldpush for even finer granularity and more channels.When planning a detector, the software costs areoften overlooked and it is very likely that morecomplexity in the hardware would payoff in decreasedsoftware costs and provide a cheaper overall detectorsystem.

Table 1. Central Tracking Chamber

Ring 1 r 1 = 20cm a = 2 mm314 sense wires/layer 4700 sense wires15 layers 3000 pads

Ring 2 r = 50 cm a = 3 mm .523 sense wires/layer 7850 sense wires15 layers 3000 pads

Ring 3 rl = 90 cm a = 4 mm722 sense wires/layer 7220 sense wires10 layers 6000 pads

Ring 4 rl = 140 em a = 6 mm1460 sense wires/layer 7330 sense wires5 layers 6000 pads

Total sense wires 27,000Total pads 18,000Total electronic channels 2 x 45,000 = 90,000

One concern with a chamber system with thisnumber wires is the total thickness in radiationlengths and interaction lengths of the chamber. Ate = 90° the total thickness is .02 radiation lengthsand .003 interaction lengths assuming that the fieldwires are AI, the sense wires are stainless steel andthe gas is 50% A - 50% Ethane. The pulse heightmeasurement will aid in detecting converted photonssince they will appear as 2x minimum ionizing particlesuntil the magnetic field separates the tracks.

Fig. 4.

a Ring 1~_J a=2mm

a\ \ .• •.- ............ ..

a .......... • •••• • •••••• • •••••• • • •••••• • • •• • • •• • • •• • • •• • • •• • • •• • •• ••

,) 45

Fig. 5. Schematic Layout of Cathode Strip System

)

, )

, )

Table 2

Module Position Area No. of Sense Wi res

1 4 m 1.6 x 1.6 m2 32002 8m 3.2 x 3.2 m2 64003 12 m 4.8 x 4.8 m2 96004 16 m 4 x 2 m2 36005 20 m 5.6 x 2.8 m2 63006 25 m 7.2 x 3.6 m2 7700

Total 36.8K

the MPS design. Each module would consist of 2X, 2U,2Y, 2V planes or a total of eight planes. The trackingsystem would consist of six modules located as shownin Figure 6. Table 2 summarizes the sizes of thechambers and their sense wire counts for the system.Since charge division is not used, the electronicchannels only need to measure the time, so a simpleelectronic system similar to the one used on the MPSchambers would probably be sufficient.

An identical forward detector system would berequired for the other beam direction giving a totalof 73.6K sense wires for the forward tracking system.The granularity is probably sufficient for detectingindividual particles in high PT jets since they haveroughly constant size in rapidity and llcj>. Moretroublesome is whether the system measures enoughspace points to carry out the pattern recognitionefficiently in complex events. Detailed Monte Ca~lo

calculations similar to the ones carried out for thecentral solenoid will have to be performed to showthat this system is adequate. It certainly is aminimal system and even so it has ""75K electronicchannels.

Anode

Evaporated ordiscreet resistors

Anode wires-600

}Sector-100

h = 2.4cm

rCa!hOdestriPS

I

Preamp connections(60/sector)

_---~ =150cm---

T__ 1 cm

L.---I -I_--l_---Iol

Avalanche - 4 X 105 electronic\.!

1~·5m~1 tT2 cm Cherge1 ampl

1--12cm--l

)

5m 10m 15m 20m 25m

Fig. 6. Layout of Forward Detector Window FrameMagnet is 4 x 2 x 2m3, Second Magnet (DashedLine) is (2m)3.

46

()

o

o

o

o

o

o

18

F' 7 Sense Wire Orientation:2x, 2u, 2y, 2v planes19. .

III. Vertex Detector

Constructing a vertex detector with 'V 20 II spatialresolution and an inner radius of a few cm which willhandle the particle rates for L = 1033 cm-2 sec- 1 at1 TeV is impossible with present day techniques. Theonly realistic possibility is to use a vertex detectorwith the collider optics adjusted to give a 2 cm longinteraction region with a maximum luminosity of 1032

cm- 2 sec-I. If we take the point of view that themain trackfinding will be done by the centraldrift chamber and that the vertex detector will onlybe used as a vernier to identify isolated vertices, areliable vertex detector can be constructed.

Silicon strip detectors can provide spatialresolution of 5 II or better. The principal drawbacksto using them are their cost of $300 per cm 2 for thestrips and their readout electronics, their thicknessin terms of radiation lengths, and their radiationdamage sensitivity. Presently available devices cantolerate only 1013 minimum ionizing particles per cm 2

before their response is substantially reduced. At1032 cm- 2 sec- 1 this corresponds to about six monthsto one year of operation before the detectors wouldhave to be replaced. Clearly substantial Rand Deffort should be devoted to trying to produce radia­tion hardened devices. The Reticon readout elec­tronics fails at about the same radiation exposure andshould also be replaced with a radiation hardenedsystem. Figure 8 shows a possible layout along witha more detailed sketch of one of the strips. About700 strips would be required for the detector in thefigure. The total radiation length, thickness of thestrips, plus their holders is .03 - .05 at 8 = 90°

A lower resolution option for a vertex detectoris a high pressure drift chamber either of t~e. .microjet type developed by Va'vra4 ) or the m1n1dr1fttype under development at CERN. Figure 9 shows thepossible wire configuration for both the microjetchamber and the minidrift type along with how theymight be assembled to form a chamber. Either typewill probably give 20 - 30 II resolution at four tofive atmospheres using seven micron diameter sense

47

wires. The pressure vessel will be the major contri­bution to the radiation length thickness and canprobably be limited to a fewer percent of radiationlength. Substantial Rand Dare still required tomake either type into a reliable vertex detector with20 - 30 II spatial resolution.

Another long range possibility for a vertexdetector is a silicon drift chamber. It is a relativ­ely new idea and initial tests will be carried outduring the spring of 1983 by Rehak at BNL. Itpromises 5 to lOll resolution over mm sizes with simpleelectronics. At an electric field of 1KV/cm theelectron drift velocity in silicon is i3ll/nsec whichshould give good resolution with fairly simpleelectronics. The resolution will be dominated byelectronic noise and will require fully depleted 300llthick junctions to provide large enough signals togive lOll resolution.

VI. Future Research and Development Needs

From the general discussion of the tracking groupand from our attempt to outline a realistic trackingsystem for a 1 TeV L = 1033 cm- 2 sec- 1 hadron colliderseveral areas which would benefit substantially fromincreased Rand D emerged. These areas are:

i. Longevity of Drift Chambers in a High RateEnvironment

The TASSO Drift Chamber operating at PETRA seemsto survive with current draws of up to 0.5llA per sensewire which is about a factor of 5 above what otherlarge chambers will tolerate. In the designsconsidered above an additional factor of at least twois needed. A substantial effort must be devoted tounderstanding the failure mechanisms and what choiceof gas, construction techniques, gas purificationsystems, etc. produces the most reliable chamber.

ii. Front End Electronics

Because of the sheer numbers of channels, usingpresent day electronics, the electronic costs couldeasily represent 30 - 50% of the cost of a generalpurpose detector. In addition, because of the sensewire density the electronics must be integrated andmounted directly on the chamber since there is simplyno room to bring out the signals on cables to theoutside world. The present MPS integrated electronicsis a step in the right direction. However, it onlymeasures the drift time and not both the drift timeand the pulse amplitude. In addition, it dissipatesmore than 400 mW/channel which implies more than 40 kWdissipation for the central tracking system leading tovery substantial cooling and longevity problems. Theother problem is radiation damage of the circuits. Atpresent, conventional bipolar transistors are the mosttolerant to radiation; however, they also consume themost power. Clearly, a coherent and national approachto the front end electronics should be the highestpriority item on our Rand D list.

iii. The Whole Vertex Detector Question

As mentioned in the section dealing with vertexdetectors, there is a strong need to develop radiationhardened silicon strip detectors and their associatedreadout electronics. The high pressure gas driftchamber options still require considerable Rand Dbefore they are reliable detectors. For the future,the silicon drift chamber certainly deservesconsiderable effort.

4.67 3

6.0 4

)

Interaetiondiamond

~2cm

""" 3.33 2 /7"""'R = 2.0 cm 17"

Fig. 8. Silicon Strip Vertex Detector

1 1.... 50mm .1 ·2X 100channel reticon

5

Tmm I ~~ ==~i ====~ == ==~ == == ==~ == = ~I!J readout

Strip detail

Electrode pitch 50 mm (top)Back side: (stereo 0< 6°)

Fig. 9(a). Microjet Vertex Detector

iv. New Ideas

There should be substantial support made availableto develop new ideas in detectors. This support shouldnot be entirely attached to particular experiments ordetector facilities. However, if there is not someattachment to experiment, one is likely to end up withsolutions in search of a problem to solve.

Our future depends on our present Rand D effortsand if we don't invest that effort now, we willrapidly lose what remaining initiative we have inparticle physics experiments.

)

)

References

1. Hadron Hadron Collider Group Report, Proceedingsof the 1982 DPF Summer Study on ElementaryParticle Physics and Future Facilities.

2. E.D. Platner, The MPS II - A Tracking DetectorSystem for Large High Rate Experiments, Proceed­ings of the 1982 CPF Summer Study on ElementaryParticle Physics and Future Facilities.

P. Schlein, et. al., The Performance of a 16000Wine Mini-Drift MWPC System CERN-EP/82-198 Reprint.

3. W. Willis, The AFS Detector System at the ISR,this workshop.

4. J. Va'vra, Measurement and Simulation of the DriftPulses and Resolution in the Micro-Jet Chamber,SLAC-PUB-3045, Jan. 1983.

x =Sense wiree =Field wire

Fig. 9(b). Minidrift Vertex Detector

48 )

o

CALORIMETRY

Bernard Pope

Working Group Summary Report

C. Baltay, L. Camilleri, B. Cox, M. Eaton, G. Finocchiaro,M. Franklin, M. Gilchriese, H. Gordon, R. Gustafson, R. Huson,

J. Kadyk, R. Palmer, B. Pope, D. Ritson, R. SchindlerW. Selove, L. Spencer, G. Theodosiou, G. Wolf

o

o

o

o

)

,)

I. Introduction

We judged that a limited goal that could beachieved in such a short workshop was an overallreview of properties of various types of calorimeterswith special attention being paid to the challenges ofhigh luminosity and high energy. One difficulty thatarose immediately was the perception that the subjectcould be approached from a number of directions; theperformance properties of calorimeters, the physicsrequirements of calorimeters, various types ofcalorimeters. It was felt that the group should bedivided up into subgroups corresponding to one of theaxes of the (at-least-three-dimensional) matrixcorresponding to these variables. Each subgroup wouldthen have the task of filling in the two-dimensionalmatrix corresponding to the subtopic. It was almostas hard to do this as it is to describe!

There was some discussion on the most appropriatedirection on which to subdivide. There are goodarguments for "starting with the physics" or forcomparing a single performance aspect (e.g. energyresolution) for a number of detectors but eventuallyit was decided that the most logical grouping was toform a subgroup for each of four types of uetectors.These subgroups (corresponding to 'continuous'calorimeters, scintillation sampling, gas sampling andliquid sampling· calorimeters) would have the task ofreviewing the performance properties and, if possible,the physics requirements of each type of calorimeter.Special attention would be paid to new developments ineach field with the hope being that areas requiringfurther study would be singled out for Rand 0 effort.

Section II of this summary contains the reportsof the four subcommittees, with particular attentionbeing paid to new developments. Section III exploressome limiting factors involved in experiments at highluminosities (or at high energies) and offers someexamples of detectors capable of operating at thehighest luminosities (energies). Section IV itemizesspecific areas where substantial research anddevelopment work is needed. The conclusions of thegroup are stated in Section V.

II. Reports of Subgroups

A. 'Continuous' Calorimeters

A summary table comparing properties of contin­uous sampling calorimeters is given in Table 1. It isstressed that a direct comparison of all propertiesof these calorimeters is difficult to make and some ofthe entries in Table 1 are no more than educatedguesses.

In addition to Table 1, this subgroup expressedthe following opinions which were believed to hold forall calorimeters in this category. A spatialresolution of CI "'5.6 mmrV"E has been measured for1 cm segmentation placed at depth of 3.5 radiationlengths. 6 A reasonable approach would be to segmentwith pads of approximately 2 cm x 2 cm at this depth.

49

The resolving power between two showers would then beapproximately one pad width. Tower (and projective)geometry appears to be desireable in order to cope

with high luminosities and the large multiplicitiesexpected at high energy. Transverse segmentation(tower size) should be approximately 2 r.l. x 2 r.l.Longitudinal segmentation involving at least threesamples is thought to be necessary. These continuoussampling calorimeters are most suitable for electro­magnetic calorimeters. Using the segmentationsuggested above, it is anticipated that n/e rejectionsat -10- 3 could be obtained with any of these devices.

B. Scintillation Sampling Calorimetery

Table 2 summarizes the work of the subgroupstudying scintillation calorimetry. Some of theentries in this table require some elaboration.Several groups are presently engaged in the developmentof fast scintillators and waveguides for calorimetery.Bob Palmer described a study in collaboration with theR-807 group of Bill Willis in which he enumerated thevarious factors contributing to the FWHM of the outputpulse of the R-807 uranium-scintillator calorimeter.He measured the following contributions: scintillator(acrylic napthalene) 15 nsec, waveshifter (BBQ)12 nsec, phototube 5 nsec, effects of shower depthand dispersion in the waveshifter 5 nsec. The overallFWHM thus predicted (by adding in quadrature) is about20 nsec, in reasonable agreement with a measured valueof 23 nsec. Palmer further studies possible improve­ments in these factors and measured the following:NEllI, pilot Uor polystyrene scintillator 1.5 nsec,BBOT waveshifter 5 nsec, RCA C31024 phototube 1 nsec.Combined with the shower fluctuations these improve­ments would suggest that overall FWHM's of 7 nsec arepossible. W. Selove and G. Theodosiou also reportedencouraging measurements of the light yields andattenuation lengths of fast scintillators andwaveshifters. 7 This group used polystyrene doped with1% B-PBD as a scintillator and experimented with BBOTand POPOP of various concentrations in Plexiglass as awaveshifter. Light yields for these materials wereseen to be 20% higher than from NEII0. Attenuationlengths were in the range of 80-140 cm and typicaltimes (convol~tion of scintillator and waveshifter)were 5-6 nsec.

Another development, of particular interest forlarge area calorimeters were good time resolutionproperties are required, is a recent series of testsof vacuum photodiodes by the University ofPennsylvania group.8 These devices, used withscintillators or scintillating glass, would providegood se~mentation capability (especially longitu~dinally), have excellent gain stability, and could runwith gates of 30 nsec or less. Such an array could betimed to better than a nanosecond by making specialtiming runs which would result in a map of timings foreach tower. Coupled with a system of multiple gates(or perhaps flash ADC's) energy depositions from eventsoverlapping within 10 nsec could be resolved.

'J

lead Glass1 Scint i llat ion2 4 lead Glass/5(SF5) Gl ass ISCGI-C) BG03 !!ll

BaF2 Scinto Composite

4.5% .!.,.!! 1% 1% l:Z! 9.1%Ener9Y Resolution « (Ejli ,ff ,.ff(Statistical Tenn Only) ,fE ,fE'""

Typical Integration Time 40ns lOOns 300ns 250ns 40ns )Possible Integration TIme <IOns <5ns

Rad. length 2.5cm 4.35cm 1.12cm 2.5cm 2.1+cm 3.2cm

Abs. length 42cm 45cm 23cm 41cm 43cm

light Output5.1 103 104 700 9.5Relative to SFS

"Most )Radiation Oamage (10% Ion) 2500rads 8.5xl04rads 104-105rads' 105rads Resistant

Scinto" 2500rads

Cost per m2$97K $170K $2.2M $1. OM $400K $70Kper20r.1.

Photodiode? X I I I NfA I

Number of ce11 s/m22000 400 570 250 ~)Assuming 2 rol.x2 r.l. 400 130

Problems. Comments Radiation New Technique Cost, Temp. Hygroscopi c New Techni que. More like a samplinsDamage long rado Dependent Compl icated device

length Readout

TABLE 1

Perfonnance Properties of Continuous Sampl ing Calorimeters

)

C. Gas Sampling Calorimetry

The report of this subgroup is summarized inTable 3. Again, comparisons are difficult and someassumptions have to be made for specific geometries.The essential conclusions are that MWPC systems havethe advantage that their timing can be improved by theuse of faster gases while non-linearities can bereduced by a reduction in gain. .

D. Liquid Ionization Calorimeters

These calorimeters consist of alternating metalplates and liquid gaps. Alternate plates are held atdifferent voltages. Ionization produces electrons inthese gaps inducing a current whose time distributionis shown schematically in Figure la. Clipping of sucha signal, for instance to 1/4 of its length, losesonly 50% of the charge (Figure lb). The driftvelocity depends on field gradient as shown in Figure2 for argon, methane and a mixture. High purity isrequired for the use of methane or its mixtures. Whatcomplications are involved are not yet known.

The induced charge in argon is about 4000electrons per mmof gap per minimum ionizing particle.Amplifier noise varies as the square root of the gapcapacitance and bandwidth~ Signql-to-noise thusvaries as gap-3/2 area- l/G timel / 2. There is thus atrade-off between low noise and short pulse time. Forexample, the calorimeter used by the TASSO'collabora­tionll has low noise but a 1.4~s pulse whereas thesmall gap fast SPS calorimeter has 25-50 ns pulsesbut much higher noise. 12

Table 4 summarizes the properties of fourarbitrarily selected examples; the first two of whichhave operated and the others are only proposed at this

50

time. In summary the advantages of these calorimetersarea) very good radiation resistance allowing its usein the forward direction, b) ease and stability ofcalibration allowing realization of high resolution athigh energies, c) cracks are not needed for lightshifters etc., d) they will work in any magnetic field.Disadvantages include: a) relatively noisy especiallyif fast pulses are required, b) cryogenic temperaturesrequire thermal insulation which is space-consumingand technically complicated,c) care is needed toavoid shorts.

III. Limiting Factors at HighLuminosity and High Energy

Figures 3a and 3b show the results of testsperformed by the E-705 collaboration at Fermilab inwhich samples of SF5 lead glass and SCGl scintillatingglass were exposed to ionizing radiation. It has beendetermined that a dose of 8.5 x 104 rads will cause a10% loss in transmission for scintillation glass incomparison with 2.6 x 103 rads which.will cause asimilar loss of transmission in lead glass. Clearlyoperation of these detectors at high luminositymachines might pose the problem of radiation darkeningsimply from the luminosity of the machine i.e.excluding backgrounds caused by beam losses. In orderto examine these effects we have developed a simplemodel for the production of particles as a functionof angle, energy and of course luminosity. The modelis derived from experimental observations by the UA2group at the SPS pp collider with assumed scaling ofthe cross section and rise of the rapidity plateau.The results are shown in Figure 4 as a function ofenergy and angle for an instantaneous luminosity of1033 cm- 2 sec-I. Also shown on the curves areestimates of the amount of radiation that would berequired to produce a 10% drop in transmission forvarious sensitive detectors. It can be seen from

.)

)

)

o

TABLE 2

SCINTILLATOR SAMPLING CALORIMETRY

o 1. Time Resolution

a) Existing Systemsb) Fast Scint/Waveshifterc) Diode Readout

Width(FWItl)

25ns8ns

~8ns

Resolution'(Jitter)

(for hadrons)

±2ns±2ns

~±lns

GateWidth

70·100ns20ns20ns

()2. Space Resolution-(Electromagnetic)

10 GeV(Pb or U)/Waveshifter 1cm

100 GeV0.5 em (Diode readout

is 50% worse)

()

1 cm scinto strips orwire plane sampl ing at4 r.l.

3. Space Resolution-(Hadronic)

FeU

0.2cm

10 GeV3em2cm

0.2em

100 GeV2em2cm

()

o

o

.)

~)

4. Energy Resolution (To at Least 100 GeV)EM(Ja r.1. sampling) 14%/[f Hadronic: Iron (1 ") 70%/Jf Uranium(3mm) 35%//f

5. Cell Size

Coupled to shower sizeTowers E-M > 2x2cm2

Hadrons > 5x5cm2

6. Cal ibration and Stabil ity

Present systems 1%-2%Diode readout: Needs R&D. but could be <1%

7. Radiation Damage or Other Deterioration

Acrylic is bad (Kienzle; R807; NIM ill. 517 '79). UA2: 4% Oown in 1 year

Polystyrene at 10cm and 2xl05rad have 10% light reduction

8. Magnetic Field

PMT's shielded to ·1kgaussDiodes less sensitive (more R & 0 needed)

9. Dynamic Range and Linearityno problem to 100 GeV hadrons

40 GeV electrons(No Measurements Yet Beyond)

10. Cost

a) $lDOK to $200K/m2(R807) (E609)

b) Segmentation Costs1) considerable for mechanically separate towers2) minimal for diodes

51

TABLE 3

PROPERTIES OF GAS SAMPLING CALORIMETERS

MWPC and Saturated Streamer HPC (High9

/)Proportional Tubes Avalanche Mode Tubes Density

Projection ChamberMinimum Thickness/Layer 0.7 cm 0.7 cm 0.5 cm 0.9 cm

Pul se Width 200 ns 200-400 ns 300 ns 3-5 u·s(with faster gasperhaps 50 ns l

Time Jitter of 20 ns (perhaps 10 ns) 20 ns 20 ns no fast trigRise Time signal

')Maximum rate for gain 105 => 1

~~ 1 M~ 1MH.e -30 KHz1 TeV shower hitting gain 104 => 1 10 Msame spot J

Lif7time (assume10-100 years 10-100 years 10-100 years >10 years10 e-/cm)

Non-1 i neari ty 10% at 100 GeV 10% at 100 GeV 10% at 20 GeV no saturatreduce ga in =>

10% at 1 TeV C)Energy Res ~ for 13%. 16%. 22% la, 13%. 19% 20% for 7mm 11 %, 13~;. 192mm/3mm/6mm of lead R'W If lFlE'"lF ~ lead .(r'{E';:;E

Rad, Length 2.7. 2.0. 1.3 2.7.2.0,1.3 2.0, 1.6. 1.1 3.3, 2.3. 1.

Stabi1 ity 2% a 1% 1%

Effect of B perp. =>no Bffect RequireMagnetic Field B paralle1·=>r X 1.4 Same No Effect B~.L <20 )Gauss;

Cost per m2$500 $500 $350per layer

Cost of e1 ectronic 10

channel $25-50 $25-50 $20-50

()

TABLE 4

PROPERTIES OF LIQUID CALORIMETERS)

TASS011 R-B06 SPS 1_14412 1-144 + Methane

Material A/Pb A/Pb A/U A + CH4/U

Gap 511'111 211'111 111'111 111'111

Plate Thickness 211'111 1.511'111 1.5 mm 1.511'111 .)Rad. Length 2em 2em 0.6 em 0.6 cm

Abs. Length 60 cm 60 em 20 cm 20 cm

Volts/1I'I11 0.4 KV 1 KV -2 KV -2 KV

Collection Time 1400 ns 600 ns 50 ns 25 ns

Position)Resolution

EM/Hadroni c 311'111 311'111 1 11'111/511'111 1 11'111/5 mm

Energy g% 12% ~2B% *2B%ResolutionEM/Hadronic .If -If {iif -lr-ff

Noi se per 50 cm2

EM/Hadronic 5 MeV 10 MeV 100 MeV/BOO MeV 30 MeV/300 MeV

cost/m2 $75K13 $75K13

52 \/

0

2

(a)

:JL0 i~.. ..

ClippedE

Collection~ 2.6 X 1020 methane/cm3

signal .~time

( 1/4 length but )8i

approx 1/2 charge~

0 ..Q

Pure argon

Fi g. 1. Charge Collection in Liqui d Argon ~0 6 8 10

0 KV/cm

Fig. 2. Drift Velocity vs. Field Gradient

Fig. 3. Radiation Damage (E-705 Collaboration)

0=900

0=300

.JSGeV

Dose received in one year from theluminosity (~c 1033 ) as a function of angle

107 ...------.,..-----.,---------,

}.. 436 nm

2X 104 3 x 104 4 x 104 5X 104

Reds

1 2 3105

Energy deposit 110'OGeV!cm2 )

-II>(b) 'C

~510

Radiation damage SCGl glessQ 104

1 2

Energy deposit 11012 GeV/cm2 )

1.00.90.80.6

~0.7

0.5l-e 0.4

:§E 0.32S

0.2

Ia:

0.1

0

1.00.9

f=0.80.7

i=: 0.6e.2 0.5./aa 0.4

S 0.3..~ 0.2..a:

0.10

o

u

()

Fig. 4. Dose Limits in Detector Materials

) 53

Figure 4 that even for perfectly clean machines thereis a limit on the use of these detectors as a steepfunction of angle and less- steep function of energy.This conclusion is quantified in Table 5. Inaddition it is cautioned that the estimates of"luminosity-'induced" doses for the CERN ISR were lowby about a factor of 6 from those that were actuallyobserved.

In a similar way we have attempted to use theintegration times for detectors given in Tables 1-4 inorder to place further limits on detectors being usedat high luminosity machines. These additionalconstraints are also itemized in Table 5. There is aconsiderable degree· of sUbjectivity in the conclusionsof Table 5.

While Table 5 indicates the angular regions wherecertain detectors would be able to operate at highluminosities, two of the subgroups chose to 'design'a detector capable of operating at luminosities of1033 cm- 2 sec- l and energies at 20 TeV. These studiescould be taken as an "existence proof" or certainclasses of detector and, while they are certainly notdefinitive or exhaustive proposals they indicate someof the choices that will have to be made. The work issummarized briefly in the next two paragraphs.

The subgroup on gas sampling calorimeteryproposed a detector using proportional tubes for thecentral region i.e. from e = 30° to 150°. Using 6 mmof lead sampling it is estimated that an electro-

LIE = 22%magnetic energy resolution of - ;~ + 2% could beE l' E

obtained. In order to· keep saturation below 10% at4 TeV (maximum energy of a jet) a gain of 103 would beused. Using a pulse width of 50 nsec, pile-up would berecognized by recording the time of the leading edge of

pulses. This would give a resolving time of about 10nsec. The center of gravity of the shower could belocalized to better than 5 mm. This would be done atthree longitudinal depths (each 10 r.l.) givinga measurement of the direction of a photon with anaccuracy of about 15 mrad. The hadron calorimeterwould also be composed of proportional tubes with 5 cmiron sampling. This would give a hadronic energy

. t f LIE 100% .measuremen 0 T =ff + 2%. Tower SlZes would beabout 300 cm2 with a total thickness of 12 interactionlengths subdivided longitudinally 3 times. It is esti­mated that the center of gravity could be located toabout 1 cm and the direction of a jet could be measuredto 5 mrad. Assuming a bunched beam structure (3 nseclong bunches separated by 10 nsec), 100 mb crosssection and a luminosity of 1033 cm- 2 sec· l implies 5events in a gate time of 50 nsec. The probability ofhaving 2 events overlap would be considerably reducedby imposing a transverse energy requirement of ET xsin e > 2.5 GeV. Total cost13 of the detector isestimated to be $52K per m2 •

A detector proposed by some members of thescintillation sampling calorimeter subgroup14 is shownin Figure 5. The basic characteristics are as follows.The electromagnetic calorimeter would consist of BariumFluoride (see later) or perhaps scintillating glass.Three modules in depth would give a total of 30radiation lengths. The hadron calorimeter would bemade of fast scintillator/iron sampling. Polystyrenewould be used with vacuum photodiode readout. Threemodules in depth would correspond to 10-12 absorptionlengths. The essential feature of the detector wouldbe to do very fast gating (10 nsec) accepting that thismay degrade the energy resolution because slow neutronswould not be collected. With careful relative timingand mapping within a sector (group of towers) one can

:)TABLE 5

SUITABLE CALORIMETERS FOR HIGH LUMINOSITY(At YS. 1 TeV)

Effect of Effect ofQllillir. Rad. Damage Integ. Time

Lead Glass X N/A

Scinto Glass 8>10· 8>30·

BGO 8>30· X

NaI 8>20· X

BaF2 8>10·1 8>10·

Acryl ic Scinto 8>40· 8>40·

Polystyrene Scinto 8>10· 8>10·

Polystyrene + P.O. 8>5·1 8>5·

Prop. Tubes (Fast Gas) 1 8>30·

Sat. Avalanche 1 X

Streamer Tubes 1 X

Hi gh Dens. Proj. Chamber 1 X

Liquid 1 8>10·

liquid Argon + CH4 1 11

Conclusions Region Candidate

8<5· LA + CH48>5··10· Polystyrene. BAF21. LA

Central Scinto Glass. Prop. Tubes

54

)

)

)

oI

90° R =4m

o

20°

D Magnet r::J 4° l+ T,---~- -- 1.5m

-r---f10m~ 1__,~_5m 1°1~ f~ ~_ Calorimeter _~ ~

oFig. 5. Sampling Calorimeter Geometry

Fig. 6. Barium Fluoride Chamber

IV. Suggestions for Research and Development

In the course of the workshop several areas werementioned where either Rand Dwork was activelyunder progress in a field or it was urgently neededto determine the performance properties of a detector.We itemize some of these promising projects in thissection. A brief description of the area is given insome of the cases.

Barium Fluoride

An exciting development for electromagneticcalorimeters is being pursued by D.F. Anderson et al. 4at CERN coupling BaF2 with a low pressure wirechamber. 16 This device is shown schematically inFigure 6. There is a fast scintillation component inthe UV region (2000-2500 A) which is quite fast(100-500 psec). The wire chamber consists of ametall~c cathode surface coated with a liquid layer ofTMAE. This layer is deposited by evacuating thechamber and then allowing the vapor of liquid TMAE toenter. The chamber is then filled with ~3 Torr ofpure isobutane. The low pressure counter has a gainof ~106 with little time jitter and a time responseof 540 ps. Such chambers have been operated in largemagnetic fields. Although the quantum efficiency ofthe TMAE photocathode is only l%,the number ofphotoelectrons/GeV is ~3000 which should lead to anexcellent energy resolution of better than 2%/~.There has only been a test of a single 26 mm thickcrystal but within this year a full electromagneticcalorimeter will be built and tested at CERN. Thecost of the BaF 2 is about 3 cents/cc for the rawmaterials and will probably cost $l/cc for crystals.

~~\TMAE

photocathodeIsobutanechamber(3 Torr)

The CERN-Oxford-Rockefeller Collaboration (R-110)has used timing information15 to deal with multipleevent pileup at luminosities up to 6 X 1031 cm- 2 sec-I.Thirty-two lead-scintillator shower counters and abarrel hodoscope covered ±1.1 units of rapidity andgave times from the average of both ends. These timeswere found to cluster within a 12 nsec wide window.Multiple interactions could be distinguished on anevent-by-event basis by searching for secondaryclusters of 2 or more counters falling outside themost populated window. The cross section for a totalenergy trigger was found as a function of luminosityfor luminosities in the range 1.5-6.0x1031 cm-2 sec- 1 •The raw data were seen to show strong luminositydependence with the "cross section" at the highestluminosity being 4.5 times the value extrapolated toL=O. By using the timing information to rejectmultiple events it was seen that 3/4 of the events atthe largest luminosity had an interaction within thetrigger resolving time. Obviously for this totalenergy trigger, one could not successfully insist onsin~le interactions for luminosities beyond -103 cm- 2 sec- 1 without better time resolution. Thetechnique could however be extended to individualcalorimeter cells. By contrast, a trigger looking fortwo cl usters of M =12° by lin = ±1.1 uni ts had amaximum 10% contribution due to multiple events atthe highest luminosity.

separate individual events in time using transversemomentum information of individual tower segments. Thenumber of towers would be ~10,OOO (each would be10 em x 10 em). A magnet is assumed to cover the cen­tral region and the total cost is estimated as $100 M.

Two experiments which actually constitute --"existence proofs" for calorimetric experimentsrunning at the highest existing luminosities atcolliding beams are ISR experiments R-110 and R-807.R-807 (the Axial Field Spectrometer) has comparedtriggering rates and cross sections for varioustriggers at two very different luminosities; a veryhigh luminosity (1.4 x 1032 cm-2 sec-I) and a relativelylow luminosity (3 x 1030 cm-2 sec-I). A jet trigger(ET in a limited region (cluster) being greater than9 GeV) was seen to trigger at a rate 50% higher withthe high luminosity run. A total energy trigger(ET > 28 GeV) had a trigger rate for the highluminosity run equal to 5.5 times the rate for the lowluminosity run. An analysis of the timing of countersnear the beam pipe corroborates the pileup at highluminosities. The group strove to make a directcomparison of cross sections after minimal cuts. Forthis comparison they chose the jet trigger since adefinite signature could be expected. After makingcuts on the time of calorimeter hits (20 nsec) thegroup is confident that they can extract an unbiasedjet cross section.

o

()

o

55

During discussions at this workshop two newideas emerged. First it was suggested that a wave­length shifter could be used (either in the crystal orin a bar adjacent to the crystal) to shift the fastUV component to a longer wavelength so that conven­tional photodetectors could be used. Second a UVsensitive secondary emission layer of CaF or CeF couldbe deposited on the crystal. This could be thecathode of a photodiode.

Thallium Formate

Some research is underway on heavy liquids whichwould replace lead glass as a continuous samplingcalorimeter. Thallium Formate (Te(HC02 )) is one suchliquid. Its properties are as follows: density =3.27 gm/cm 3 , refractive index = 1.59 and radiationlength = 2.57 cm. It has about the same transparencyas lead glass but with a 30 nm lower edge. It appearsto suffer very little radiation damage, 3 x 106 radsbeing reported to have had no effect. Its cost iscomparable with and probably lower than lead glass butsuffers from the disadvantage that it is a toxicliquid.

Xenon

Xenon has been suggested as a liquid scintillatorfor electromagnetic calorimetery at 180c K. Its densityis 3.1 gm/cm 3 and radiation length is 2.8 cm. Itsemission spectrum is in the range 150-185 nm with 3 timecomponents (3 ns, 22 ns and 700 ns). Its present costis about $4/cm3 and appears to give approximately thesame output as Sodium Iodide.

Use of Methane in Liquid Argon12

Faster Scintillator/Waveshifter7

Shorter Radiation Length Scintillating Glass

One disadv~ntage of the new scintillating glassdescribed above is its significantly longer radiationlength. Studies are at present underway to reduce thisvalue. 18

Radiation Resistant Lead Glass

Rand D is presently underway by a'variety oftechniques to improve the resistance of lead glass toradiation. 19 One such technique is the addition ofsmall amounts of Cerium to the melt.

Photodiodes8 and Phototriodes 20

Silicon Detectors

V. Radeka has suggested that silicon detectorsmay be an appropriate device for small areas. Thesignal is about 13 times higher than that from liquidargon and the signal/noise is 15 times higher. Thecharge collection time is about 20 ns (for a typicalthickness of 300 ~). Ninety-five percent of thischarge would be collected with 10 ns. The presentcost is estimated as $20/gm for the material with anadditional $20-40/gm for processing. This would givea total estimated cost of $100K/m2 per layer.

Bismuth Silicate21

56

V. Conclusions

I hesitate to use the word 'conclusions' forsuch a short workshop. Clearly the review ofcalorimetry undertaken in this report illustrates arapidly changing field and a strong research effortwhich has also been reported in many previousreviews. 22 ,The particular emphasis of this workshopwas to explore the challenges posed by high luminos­ities and high energies.

'The difficulties introduced by pileup problemsand by high radiation doses were certainly broughtout clearly by this workshop. Considerable care willhave to be taken with spatial segmentation and withtime resolution in order to reduce the problems causedby high rates and high multiplicities. Some energyresolution may have to be sacrificied.

However the consensus of this working group wasthat it appeared that a calorimeter capable ofhandling luminosities of 1033 cm- 2 sec- 1 with (almost)411 coverage could be built. Such a detector will havemany more channels (an order of magnitude?) and willcost considerably more (factor of 2-5) than presentday detectors.

Calorimeters will become increasingly importantat higher energies. As the statistical term drops(proportional to l/E), systematic effects willdominate. Sampling calorimeters will suffice andsampling can become more coarse. Paradoxically it maybe that calorimeters will become cheaper!

One final recommendation: the short duration ofthis workshop prevented an in-depth'study of the fieldof calorimetry. The end effects inherent in definingthe problems and laying out the guidelines reduced thetime even more. Perhaps it would be appropriate tocreate a "commission" with the charge of centralizinginformation available on calorimeters, recommending Rand D for promising projects, "designing" a detectorcapable of performing at high luminosity and/or highenergy, arranging topical workshops or conferences,and periodically reporting to the high energycommunity.

)

)

)

)

)

)

o

()

o

o

References

1. J.S. Beale et al .• Nucl. Instrum. Methods 1lI. 501(1974) .

2. B. Cox et al •• Fermilab preprint FNAL Conf. 82/75­Exp.

3. M. Cavalli-Sforza. SLAC Conference on Instrumenta­tion for Colliding Beam Physics (1982). p. 216.

4. D.F. Anderson et al •• (Charpak group) submitted tothe Vienna Wire Chamber Conference. February 1983.

5. C. Cozza et al •• University of Padova preprint.submitted to Nuclear Instruments and Methods.

6. B. Cox et al •• Fermilab preprint FNAL Conf.82/76-Exp.

7. W. Kononenko et al •• Nuclear Instrum. Methods 206.91 (1983).

12. W. Willis talk at this workshop. Letter of Intentto the SPSC. C.W. Fabjan et al .• (CERN/SPSC83-8 I 144 (1983)).

13. Europe~n generated numbers: multiply by 2?

14. W. Selove and G. Theodosiou. private communication.

15. J. Linnemann. R-II0 Memo 131 (1983).

16. D.F. anderson. Phys. Lett. 188B. 230 (1982).

17. D.F. Anderson IEEE transactions on Nuclear ScienceNS-28. 842 (1981).

18. Ohara Optical Glass. Inc •• B. Cox. privatecommunication.

19. M. Marx. private communication.

20. M.D. Rousseau. Opal Collaboration at LEP.8. W. Selove. G. Theodosiou. Nucl. Instrum. Methods

186. 585 (1981). See also W. Kononenko et al.. 21. M. Kobayashi. Nucl. Instrum. Methods 205. 133University of Pennsylvania Report UPR 0098E (1982). (1983).

o

o

o

u

9. O. Ullaland. SLAC Conference on Instrumentationfor Colliding Beam Physics (1982) p. 212.

10. $10 seems possible. see Ball et al .• Nucl. Instrum.Methods .!2Z.. 371 (1982).

11. G. Wolf talk at this workshop. See also A. Ladage.SLAC Conference on Colliding Beam Physics (1982)p. 180.

57

22. See for example: SLAC Conference on Instrumenta­tion for Colliding Beam Physics (1982).Proceedings of the SLC Workshop on ExperimentalUse of the SLAC Linear Collider (1982). Proceed­ings of the 1982 DPF Summer Study on ElementaryParticle Physics and Future Facilities. See alsothe excellent review "Calorimetry in High EnergyPhysics" by C. Fabjan and T. Ludlam. Annual Reviewof Nuclear Science (1982).

TRIGGERS

Mel ShochetWorking Group Summary Report

T. Carroll, W. Hoogland, R. Johnson, R. Kass, J. Linneman, S. Loken,R. Majka, M. Ronan, M. Shochet, M. Tannenbaum, B. Willis,. J. Zweizig

The major task of the Trigger Group was toexamine the possibility of trigger systems which couldoperate at a luminosity of 1033 cm- 2 - sec- 1 • Weconcentrated on two luminosity related problems: thefast trigger decisions which must be made because ofthe high interaction rate, and the effect of multipleinteractions within the integration time of thedetector. Little time was spent on the higher triggerlevels where the final selection of events to bewritten on tape is made, because the difficulty ofthe problem is very dependent on the specific signa­ture of the physical process under study.· If one islooking for events with multiple large PI leptons andjets and significant i1 imbalance, then the eventselection will not be very difficult. On the otherhand, if the events of interest are chiracterized bylow PI leptons and hadrons and little PI imbalance,then the signal to noise will be poor and the eventselection problem will be an extremely hard one.

We concluded that a trigger system for highluminosity operation can be constructed. The systempresented here as an existence proof, while notnecessarily optimal, will, we believe, work. It is amulti-level trigger system, where the number of eventswhich can be retained at any level is only limited bythe requirement that the deadtime at the next levelnot be excessive. Before describing the overallsystem, we first discuss multiple event pile-up,tracking, electron identification, and transversemomentum imbalance.

Multiple Event Pile-Up

At a very high luminosity co11ider, a substantialfraction of events will contain more than one inter­action within the integration time of the detector.For example, in a DC machine with a luminosity of1033 cm- 2 - sec- 1 , and with an inelastic cross-sectionof 50 mbarn, interactions occur with an averageseparation of 20 nsec. For ten percent of the inter­actions there will be another· interaction within 2nsec of the first.

A crucial question is how the presence ofmultiple interactions affects trigger decisions. Aconvolution equation can be written which gives thedistribution in transverse energy (ET) for multipleevents in terms of the ET distribution for a singleeven~. If PN (ETmin) is the probability that N over­1applng events glVe a transverse energy ET> ET . ,then mln

j ETminPN (ETmin) = P1 (ETmin) + dE P~(E) PN-1 (ETmin-E)

oThe probability of obtaining at least ET' from Nevents is equal to probability that one mln eventprovide the necessary ET plus the product of theprobability that N-1 events give ET . -E and thedifferential probability that the mln Nth event giveE. Using this equation and P1 (ET)' probabilitydistributions for any N can be obtained.

58

Figure 1 shows a probability distribution,P1(ET)' which is exponential at low ET. The effect ofmu1tlp1e interactions within the integration time isseen to be extremely large and to grow with increasingET. The difference between low luminosity and anaverage 5 events per integration time is almost sixorders of magnitude at ET=20 GeV. Fortunately,however, the cross-section at large ET has a power lawbehavior. Hence the effect of multiple interactionsdoes not grow with increasing ET, but rather the effectis a uniform shift in the ET scale. This shift isequal to the mean number of interactions multiplied byan average ET per interaction.

Such a shift can be minimized if the incrementalET from extra interactions is low. This can beaccomplished if a cluster algorithm is used in thetrigger. Figure 2 shows an ISAJET simulation of hadronjet production. The jet cross-section is increasedgreatly from its low luminosity level if the full solidangle ET produced my multiple interactions is included.If ET is added only from the solid angle near the jet,the effect is .great1yreduced. If, in addition, onlythose calorimeter modules containing ET > 1 GeV/c areincluded, the increase in cross-section becomes verysmall. .

This method for limiting the effect of multipleinteractions is used in our trigger system.

Central Tracking

Large transverse momentum tracks can be foundquickly and easily because of the structure chosen forthe central drift chamber (see Tracking Group report).The narrow drift distance not only gives rapid chargecollection «100 nsec), but allows us to use driftcells as simple hodoscope elements. Since the triggersystem does not use drift time, the hardware consistsof simple digital coincidence logic.

The use of 4 to 6 planes in each of three radialregions provides a local minimum PT of 4 GeV/c whichis determined by the cell size. The minimum PT for atrack can be varied from 4 GeV/c to 20 GeV/c when thelocal track segments are placed in coincidence. Thisinformation is available in a few hundred nanosecondsand thus can be used in the earliest trigger levels.If a few layers of drift wires are equipped for chargedivision, then the track can be associated with asingle calorimeter tower of the first trigger level.

Muon Tracking

We assume that the detector is 8mx8mx10m longand that we have muon detection on the four side walls.If these are four planes of 2.5m long wires with 1cmdrift space, then the muon system contains approx­imately 25,000 wires. In a half cell staggered wirearrangement, the first signal appears within 100 nsecof the interaction. A muon can be correlated with alarge PT inner track and with a single calorimetermodule if two muon planes are equipped with chargedivision read-out.

/)

)

)

)

)

)

o

70

ISAJETJs =1 TeV

<n>=O

Isajet Simulation of Jet ET ProbabilityDistri buti on

Fig. 2.

10.0

10-1 1.0

10-2 0.1

10-3 0.01

10-4 '"":'EE 10-3

W

-. Q:"

•• 10-5E 10-4w·-a..

10-6 10-5

10-7 10-6

10-8 10-70

10-9

10-100 10 20 30 40 50

o

o

o

()

Fig. 1.

ET (GeV)

Probability of exceeding ET. for differentevent overlaps mln

deposited in the first segment. There would then beapproximately 2000 fake electrons per second, which islow enough for the first two trigger levels. If theelectromagnetic calorimetry were to have no depthsegmentation, then the electron PT threshold wouldhave to be increased by a few GeV/c.

,)

.)

Electron Identification

We have considered two methods for electronidentification: traditional calorimetric separationand active electron tagging using transition radiationdetectors.

(l) Calorimetric Separation of Electrons

In the early levels of the trigger, the back­ground comes from the uniformly distributed low PTparticle production as well as from particles asso­ciated with jets of transverse momentum ~10 GeV/c.Events containing jets of much larger PT will beretained in the first few trigger levels whether ornot there is an electron present. The low PT parti­cles are not a serious contaminant since theytypically deposit less than 1 GeV/c transverse energyin a single calorimeter module. The low to moderatePT jets are so wide compared to a single calorimetermodule, that we can use the single particle inclusivecross-section to estimate the hadronic background.

For charged pions with PT > 5 GeV/c, the integralcross-section is 4xlO- 29 cm2 as measured at the SPScollider. At a luminosity of 1033 cm- 2 - sec-I, therewould be 40,000 such particles produced per second.If the electromagnetic calorimetry had two segments indepth, we could get at least a factor of 20 pionrejection by requiring most of the energy to be

The integrated neutral pion cross-section forPT > 5 GeV/c is 2xlO-29 cm2 , giving 20,000 particlesper second at 10 33 luminosity. We estimate thatrequiring a stiff track (PT > 4 GeV/c) pointing tothe single calorimeter module would give at least afactor of ten pion rejection. Thus, again we wouldbe down to the acceptable level of a few thousand fakeelectrons per second.

It this appears that for the first few triggerlevels, an electron threshold as low as 5 GeV/c mightbe achieved.

(2) Transition Radiation Detector

If identification of electrons within jets isdesired, active electron tagging is required. Atransition radiation system can provide both thenecessary segmentation and rapid decision time so thatthe information can be used in the trigger. Theassumption here is that the electromagnetic calorim­eter cells would be the size of an electromagneticshower. Several LEP proposals have such segmentation(105 elements). To reject IT'S and K's, the transitionradiation detector should have similar segmentation.One of six sections in depth is shown in Figure 3.The electron requirement would be satisfied if atleast 4 wires out of the 12 in a line record energydeposition greater than 4keV (five times minimumionizing).

) 59

A rough estimate of the electonics cost for sucha system was made under the assumption that analoginformation from the wires would not be read out.Rather there would be a single bit for each segmentindicating the presence of an electron. A circuitboard would contain the amplifiers and comparators for48 channels as well as PROM circuitry to perform themajority logic. The estimated cost is $15/channel or$1.5xl06 for a 10 5 channel system.

Missing PI Trigger

There might be a class of events for which themost prominent characteristic is a large transversemomentum imbalance. If events are to be retained atthe early trigger levels solely on the basis of PTimbalance, what thresholds would be required? Toanswer this question we used UAl distribution formissing Px or Py, which is well described by a gaussianof -0.4 ET' where ET is total transverse energy in theevent. We also used the UAl ET distribution forminimum bias events and then calculated the missing PTspectrum when the average number of overlappinginteractions is between 2.5 and 5. A threshold of15-20 GeV/c would suffice for the first trigger levelwhile a 20-25 GeV/c threshold would be required at thesecond level. Of course these thresholds could bereduced if there were additional requirements ofleptons or jets in the event.

Trigger System

The trigger system is an extension of the onedesigned for the CDF detector. The major difference isthat at the Fermilab Collider, beam crossings occureach 3.5 ~sec. Thus the first level trigger decisioncan take a few microseconds without incurring anydeadtime. At a CW collider with 1033 luminosity,interactions occur on average every 20 nsec. Thisrequires a very fast and deadtimeless first leveltrigger.

A sketch of the calorimeter level 1 trigger logicis shown in Figure 4. The calorimeter signal firstpasses through an integrating amplifier. The amplifieroutput enters a 500 nsec lumped delay line in order toallow time for the level 1 trigger decision to be made.Since the latter stages of the trigger take manymicroseconds, the calorimeter signals must be held.This is accomplished by the BEFORE-AFTER circuitrywhich is fed by the lumped delay line. Both switchesare normally closed and thus both capacitors follow

Fig. 3. Transition Radiation Detector

-t50mm

-1-10 mm

--L••••

••••• •• •• •• •

• • •• • •• • •• • •

=r.-F=3mm

CH2 wool

60

the output of the integrating amplifier. If the level1 trigger is satisfied, the BEFORE switch is openedjust before the interaction associated signal leavesthe delay line. The AFTER switch is opened approx­imately 100 nsec later. The output of the differenceamplifier is proportional to the charge collected bythe detector element during that 100 nsec period.

The level 1 trigger logic begins with a differenceamplifier connected ,to the integrating.amplifier andto a 50 nsec tap in the lumped delay llne. The output,which is the charge collected during the previous50 nsec, goes to a number of devices. It is connectedto two current summing operational amplifiers. Thefirst (second) amplifier has input resistors propor­tional to the sine (cosine) of the detector element'sazimuthal angle. These amplifiers thus give the totalPx and Py respectively for the calorimeter cellsconnected to that circuit board. By adding the signalsfrom the various circuit boards, we get the total Pxand Py for the event. If the magnitude of either isabove some threshold value, then the PT imbalance level1 trigger is satisfied.

The output of the level 1 trigger differenceamplifier is also used in a conditional sum to lookfor large ET deposition in the detector. If thesignal is above a PT threshold, typically 1 to 1.5GeV/c, then that calorimeter cell is considered tohave large ET. In that case, a linear switch isclosed and the signal enters a current summingoperational amplifier with identical input resistor~.

When these amplifier outputs are added over the entlredetector, we get the total ET for all cells above PTthreshold. If this is above a predetermined value,the large ET level 1 trigger is satisfied.

The output of the PT threshold comparator isused for one other purpose. A five bit ring counterhas a single one-bit which is shifted each 20 nsecsynchronously throughout the detector. When theoutput of the comparator goes true, the value in thecounter is shifted into a storage register. (If thistiming is not precise enough, the comparator outputcan be used to gate the phototube leading edge wh~ch

would determine the timing.) When the level 1 trlggeris satisfied, these timing counters can be used toreduce contamination from multiple interactions. Thetime from the calorimeter cell containing the largestPT is defined as the event time. This five bit codeis sent to all circuit boards. Only those channelswhose time equals the event time are used in latertrigger decisions.

The level 1 trigger, which occurs in 300 nsec,must reduce the event rate to 50khz. We have usedSPS collider data as well as ISAJET and CDF montecarlo results to estimate the required level 1thresholds. These studies indicate that if anyoneof the following is satisfied, the event can beretained and the BEFORE-AFTER switches opened:

1. Greater than 8 GeV transverse energy'isdeposited in those towers with PT threshold of1-1.5 GeV/c.

2. An electron candidate is present with ~T > 5GeV/c. The electron requirement is a large PT 1n asingle logical calorimeter cell either ~ith ~ transi­tion radiation count, if such a system lS bUllt, orwith a high PT track pointing to the cell and ~heproper energy ratio in the two depth segmentatlons ofthe calorimeter.

3. There is a PT imbalance greater than 15-20GeV/c.

,)

)

o

o

o

o

Before

500 nsec

30 ­50

nsec

opthresh1

() Fig. 4. Schematic of Level 1 Trigger for aCalorimeter

C)

()

,)

)

4. Another possible signature is the presence ofa high PT muon candidate in the muon chambers.Details of the steel shield are needed to determinewhether this requirement would be restrictive enough.

Level 2 can take approximately 1 ~sec and mustreduce the rate by an order of magnitude, to 5khz.During this period, cluster finding hardware willlocate the largest ET electromagnetic cluster andhadronic cluster in the detector. The event can beretained if either:

1. There is a jet with PT > 20 GeV/c;

2. There is an electon (using improved trackinginformation) with PT > 5 GeV/c.

3. There is a PT imbalance of 20-25 GeV/c;,

4. There is a muon, as defined by the muonchambers, central tracker and hadron calorimeter, withPT above a level which depends on the thickness ofthe iron shield, or;

5. If various combinations of requirements aresatisfied such as PT imbalance 15 GeV/c and a jet withPT ""'10 GeV/c.

Level 3 requires an average of 10 ~sec and mustreduce the event rate by at least another order ofmagnitude to a few hundred per second so that the fulldetector readout which takes 2 msec can be initiated.In this level the overall topology of the event isdetermined. All the moderate and large PT energy

61

clusters are found, with hadronic and electromagneticET' mean rapidity and azimuth, and the widths inrapidity and azimuth calculated. Clusters arecorrelated with high PT tracks to produce an ETordered list of energy clusters with tags for e, ,and jet. In addition there is a list of muons in theevent.

With the topology thus determined, fast bit-sliceprocessors are used to decide whether it is a topologyof interest. Events containing a large PT electronand missing PT 1800 away in azimuth will clearly beaccepted. An event with a multi-jet system of largeinvarient mass on one side of the detector and alepton on the other side of the detector is a likelyheavy quark decay event and thus also will be accepted.Events with two jets (PT > 20 GeV/c) approximately1800 apart in azimuth are likely QCD scattering eventsand because of their abundance will be accepted at aprescaled rate. These are obvious examples ofinteresting topologies; signatures for other physicsprocesses would also be included.

During the 2 msec required to read out the event,these processors will continue to analyze the informa­tion to help determine whether this should be one ofless than 10 events per second to be written onto tape.

It is quite possible that additional preprocess­ing will have to be done to reduce the data to amanageable level. A group of people from the TriggerGroup and Systems Group considered this problem. Theresults are included in the report of the SystemsGroup.

PARTICLE IDENTIFICATION

Davi d R. Nygren

Working Group Summary Report

S. Aronson, J. Bensinger, C.Y. Chang, M. Goldberg,M. Holder, R. Kofler, D. Leith, E. Loh

P. Nemethy, D. Nygren, S.J. Smith, R. Van Tyen

I. Introducti on

This group addressed a wide variety of subjectswithin the general context of searching for limita­tions in capability due to high average rates. Theluminosity taken for reference was 10 33 • The groupmade an effort to identify areas of technique whereinspired Rand Deffort is especially needed.

Because of the short time scale of the workshop,considerations were limited mainly to the 1 TeVregime. The group felt that the physics of the 20 TeVregion is not sufficiently obvious at present to de­fine the relationship between rate questions andparticle identification.

Within these constraints, a general conclusionwas that no intrinsic "brick walls" for luminositiesor the order of 10 33 were discovered, other than theusual compromises arising from financial limitations.In fact, rates were in some instances rather modest,arising from the high degree of segmentation needed apriori to swallow the high multiplicity jets expected,and evident now at the SPS collider.

Topics receiving attention included the followingarray of possibilities: Cerenkov ring imaging, tran­sition radiation, synchrotron radiation, time-of-flight(yes!), Hi gh P...L spectrometer, heavy quark taggingwith leptons, general purpose ~ and e detector, anddE/dx. The individual reports on these subjects re­present the main work of this group. Some commentsregarding these contributions are included here toprovide a perspective.

II. Cerenkov Counters

Cerenkov ring imaging remains an attractivetechnique, realized currently in beam line geometriesbut not yet in collider configurations. Previousstudies (1,2) have identified the problems which stillremain to be solved before the technique can be accept­ed as a practical large scale tool. An intrinsiccharacteristic of Cerenkov effect is the rapid varia­tion of 0c just beyond y threshold, rising to 97% of0i max at only about 3 y threshold. Therefore, toobtain the widest range of sensitivity to y, greatprecision in optical imaging and photon detection isessential. The relatively small number of Cerenkovphotons in a frequency interval small enough to avoiddispersion places a severe requirement on the effi­ciency of photon detection devices. The discovery ofsuperior molecular photoelectron emitters with highquantum efficiency in a wave length regime suitablefor quartz windows would advance the practicality ofthis technique considerably. It is easy to nurture

technological fantasies marrying advanced photocat­hodes with micro-channel plate electron multipliersand CCD imaging arrays in a single device, yet suchdevices are not likely to be developed with the usualfinancial resources devoted to instrumentation forhigh energy physics. Nevertheless, because the yrange can be adjusted through choice of radiator tospan a large dynamic range, by cascaded detectors,this technique deserves a larger investment of effortthan it has received.

For details, see M. Goldberg and D.W.G.S. Leith,contributed paper in these proceedings.

III. Transition Radiation

An unconventional utilization of transition radia­tion (TR) was examined, as an alternative to the con­ventional method of muon momentum measurement by mag­netized toroids. The analysis showed that the conven­tional technique gives better momentum resolution thanthat obtained with TR, however, the TR method mightbecome interesting if magnetized iron were precludedby other considerations.

IV. Synchrotron Radiation

The use of synchrotron radiation to tag electronsin the multi-GeV regime was evaluated with the conclu­sion that straightforward techniques such as xenon­filled MWPC's could reliably convert and measure theoutboard halo of x-rays emitted by electrons. Themagnetic field strength (and hence length) can betuned for a particular energy range of interest.

See S. Aronson, contributed paper in theseProceedings.

V. Time-of-Flight

Time-of-flight techniques assuming contemporaryperformance characteristics were studied in the con­text of a search for new massive, lon~-lived objects.It was found that a luminosity of 10 3 could be uti­lized if some redundancy in measurement is provided,

Details are given in C.Y. Chang, contributedpaper in these proceedings.

62

: )

)

)

()

o

o

()

()

()

VI. Jet Spectrometer

Starting with a design developed at an ISABELLEWorkshop, the problem of designing a high PJ. spectro­meter to measure jets and identify jet components wasreexamined. New insight into the phYsics of jets atthese energies was provided at the workshop by somegraphic examples of events produced by ISAJET, a MonteCarlo incorporating the best current theoretical pre­judices. The use of a relatively modest magneticbend region deflects the majority of the particles,i.e. Low P~, away from the sensitive elements of thespectrometer. The corresponding rates are on theof a kilohertz/wire, but the bite in rapidity andazimuth must be much larger than was previously thoughtnecessary due to the structure of the jets.

Details are found in S. Aronson, M. Goldberg,M. Holder, and E. Loh, contributed paper in theseproceedings.

VII. Quark Tagging by Leptons

The problem of heavy-quark tagging through theidentification of the decay leptons was studied. Thissignature would be observable in relatively low multi­plicity jets as an outboard lepton, i.e. a leptonisolated from the core of the jet but close enough tobe identified as a jet component. Values of PJ-ex­pected, relative to the jet axis would be of the orderof 10 GeV/c for top in the expected range. Back~grounds from pion and kaon decay limit the usefulnessto higher energy jets. Simultaneous muon and electronidentification capability enhance the efficiency ofthis signature considerably since decay chains couldbe observed.

VIII. Muon/Electron Detection

A general purpose specific muon and electro~

detector was defined and evaluated. Based on thlSstudy, the conclusion was made that systems of thetype used for the Fermilab CDF are adequate forL=1033, but that single ~ and e signatures are suffi­ciently frequent (a few KHz) that these have to beused in conjunction with other information or at ahigher trigger level. Backgrounds below 10 GeV PJ.are large.

63

See J. Bensinger and L, Nodulman, contributedpaper in these proceedeings.

IX. Energy Loss Sampling

Finally, a whimsical study of a large solid anglespectrometer employing ~ultiple dE/d~ sampling was.made to explore the limlts of what mlght be done wlthbrute force. This spectrometer, using the conven­tional axial-wire soleroid field confirguration re­quires the stringing of -lxl06 wires and instrumenta­tion of 25xlOs cells. The strength of the dE/dx tech­niques lies partly in the fact that the informationis contained in the track itself, whereas the weaknesscomes from the slowly varying logarithmic sensitivityto y. Furthermore, discrimination is essentially lostabove a y of 200-300 as hadrons and muons merge withelectrons. The maximum luminosity supportable by theconfiguration examined in probably is excess of 10 32if machine backgrounds are low. Despite the formid­able number of channels, a true optimization mightresult in a manageable project. An amusing option isthe possibility of balancing the several hundred t?nsforce from the wire tension with an overpressure wlth­in the chamber volume. The overpressure, on the orderof 1/2 atmosphere would improve the dE/dx resolution.

See D.R. Nygren, contributed paper in these pro­ceedings.

X. Conclusion

The time available allowed sufficient progress toestablish a belief that particle identification willprobably not represent a primary obstacle to utiliza~

tion of 10 33 luminosities.

References

1. Proceedings of the Workshop on Possibilities andLimitations of Accelerators and Detectors, FERMILAB,April 1979. ICFA.

2. Proceedings of the Second Workshop on Possibili­ties and Limitations of Accelerators and Detectors,Les Diablerets, Switzerland, October, 1979. ICFA.

DETECTOR SYSTEMS

Barry C. Barish

Working Group Summary Report

J. Appel, B. Barish, J. Branson, R. Cahn, T. Carroll, J. Christenson,B. Diebold, H. Gordon, W. Hoogland, R. Johnson, L. Jones,

L. Lederman, J. Linnemann, M. Nelson, F. Paige, J. Rosen,·B. Sadoulet,R. Schwitters, E. Siskind, H. Sticker, A. Tollestrup, J. Yoh

,)

1. Introduction

This report represents a brief review of the workof the "Sys terns" group for the DPF Workshop onCollider Detector. There are many systems consider­ations involved in large particle detectors. In thisshort workshop, only a few select topics could beconsidered. I have divided this report into twogeneral parts: (1) Detectors at 20 TeV and (2)Detector Considerations for a High Luminosity Hadron­Hadron Machine.

2. Detectors at 20 TeV

2.1 The Character of Events at 20 TeV (F.E. Paige)

In order to get some feeling what events mightlook like at very high energies, plots of ISAJETevents at -vs = 10 TeV were made for high PT jetevents.

Two examples are shown in Figures lA and lB.Primary jets have

1000 GeV < PT < 1100 GeV-1 < n < 1

The vertical scale is in GeV. Solid lines are energydeposited in calorimeter cells (tin = 0.1, M = 5°).Dotted lines are parton energies after QCD jetevolution.

The events contain very sharp jet clusters, butoften there are several clusters or side clusters.The central structure is very sharp.

2.2 20 TeV Detectors (B. Sadoulet, F. Paige,J. Bronson, R. Cahn)

Several important physics considerations werenoted for doing physics at 20 TeV:

(1) We add high PI W± and ZO to our standardset of indicators of interesting events. For example,four jet events could be used to search for a heavyHiggs.

HO + W+ W- (Reconstruct W-mass)+ 4 Jets

Note the jets from Whave few low momentum particlesand are characteristically narrower than QCD jets.

(2) The highest P1 leptons which we might expecthave P1 "-' 3 TeV. This is from considering ahypothetical Z' + Jl+Jl- with a mass of 5 TeV and thesame dimensionless coupling on the Zoo

(3) QCD subjects ~ be narrow. We wish tomeasure jet widths to distinguish W's and T'S.

6.4

Detector A. A sample detector was designed bythis group and is shown in Figure 2. The mainfeatures are the following:

Detect: high PI jetsT, eW±, ZO jetsmissing PI

Give up: Jl - momentum above "-'300 GeVe - charge above"" 50 GeV

It uses calorimetric methods. Note that sincee.m. shower and hadronic decay depths do not changerapidly with energy, this kind of detector is verysimil ar to a 1 TeV detector.

The 4TI hadronic calorimetry is done with small.towers: tly "-' 0.1, M"-'40 (10cm x 10cm + 2.5 K cells).The electromagnetic shower counters have finersegmentation (2.5cm x 2.5cm + 45 Kcells).

The central tracking is used to identify stiffparticles (for electron identification), find thevertex, measure multiplicity, identify Wand T jets.The iron toroids are used to measure muon momentum.

Detector B. (B. Sadoulet) A detector wasproposed, somewhat conventional in nature, for a 20TeV collider.

Physics Goals - (Assume L ~ 10 31 , 107 sec ~ fL1038 cm 2 )

(1) Heavy new objects in TeV range.

e.g. (i) heavy Wor ZO at 4 TeV a ,~ 10- 36

->- e+e- al3"-' 10-38

(ii) Higgs ->- W+W-If m (H) = 300 GeV a "-' 310- 36

with W->- 2 jets

(iii) new families: jet + e, jet + Jl

(2) Extremely violent collisions

e.g. very high PI jet

Fermilab collider will study jets forPI "-' 400 GeVA 20 TeV collider at L = 1031 will studyPI ,,-, 2.5 TeV

( )

()

( )

o

o

1000.0 830302'- 180652 - 3

o

o

()

o

600.0

()1000.0 830302 - 180652 - 8

600.0

~

". 600.0~

U l..J100.0

'00.0

Fig. 1 (a) and (b). Isajet Simulation ofVS = 10 TeVevents

") 65

20 TeV detector = (1.5 - 2) *CDF

Muon chambers

Em shower

1)=3

Beam pipe

:)

()

I I I I I I I I I I18 16 14 12 10 8 6 4 2 0

Meters)

Fig. 2. Calorimetric 20 TeV Detector

Detector Considerations -

(1) The detection of leptons. v. and single y's areimportant for tagging. The electron is easy. single ymaybe hard. ~ is extremely difficult above 50-100 GeV,v is identified by missing transverse energy. To

measure ~E~~~~ing to 30 GeV will require coverage down

to ""1°mrad.

(2) The presence of the Wmay be used as tag. Forexample, W+ 2 jets boosted gives narrow jets and canreconstruct W-mass.

(3) Jet energy will fragment into "lumps" as we seefrom the Paige events. The jet "width is """'0.5 rad(~ angle), but many lumps have typical width ~~ ......,0.1rad, ~y """0.1. These lumps represent the trulyelementary objects.

(4) The approximate angular range for a jet of PI ispolar angle,

e > 3 x kinematic limit = 3 PI€i2

So for 400 GeV, e ~ 120 mrad· implies y.> 2.6.

66

Desired Detector Characteristics

(1) Match calorimetry to jets. The typical hadronicshower diameter ~ = 0.2m should be matched to ~d = 0.1and ~y = 0.1.

(2) Making the downstream detector a "reasonable"size yields Iyl < 2.

(3) Missing energy considerations imply detector must

go down to ......,10 mrad for E~r;~ing """'30 GeV/c.

(4) Inside one needs tracking to separate singleparticles and jets (best two particle resolutionavailable). Magnetic fiera-can be moderate (......,IT),and is needed to identify low energy particles andmeasure ~PI """'0.3%.

An illustration of such a detector is shown in Figure3. The main feature are summarized below:

Magnet:coil length 12m~ = 2m externalField 0.7 - IT conventional

)

)

o

/.I chamber

o

o

o Fig. 3. Solenoidal 20 TeV Detector

Tracking:JADE type or UA1 ~ 10K wi res(volume divided by electronics in ~106)

Muons:350m2 x 2 x,y

2 layers

2500 towers

10K towers x 2samples in depth

Consider this method at 900 • Then the impactparameter b =~ (. = angle of bend) and at 1 TeV,b = •15~p2 = l;m. Note that the impact parameter

b = 4 x Sagitta, so this scheme in principle is simpleand more accurate. This, however, depends on limita­tions from scattering in the coil, and precision ofknowledge of the vertex position.

First consider the scattering in the coil =J.i~ =ffllr at 1 TeV. This gives 6b = 3011 implying~Ssmearing due to scattering of 6p/p = 3%. This is,of course, at 90 0 and becomes worse for more forwardangles.

A schematic view of the end view of such adetector is shown in Figure 4. Consider a coil 2Xothi ck, p = 1. 5m, and B = 3T. Then the transversemomentum kick of the magnet is PT = 0.35 GeV/c andparticles of .675 GeV/c will curl up in the solenoidalfield.

12505000 x 2 samples

in depth

Ca 1orimetry-End Caps:size 2 x 25m2

hadron 6X = 6Y = 0.2me.m.

Calorimetry-Barrel:hadron 6. = 0.1

6y = 0.1e.m. 6. = .D5

6y = .05

o

o

o

u

u

u

Conclusions:

A 20 TeV detector is not necessarily a monsterand might look a lot like a 1 TeV detector, especiallyif the role of the magnetic field is restricted torecognizing hard tracks and measuring ~ < 50-1000 GeVparticles.

Several areas requiring R&D can be noted.There are a large number of cells required forcalorimeters (>30K minimum). We need a good idea,similar to time projection for tracking, in order to'subdivide the volume electronically. Tracking in avery long solenoid may also present problems.

Other Considerations for 20 TeV Detector A pos­sible alternative to measuring the sagitta of a veryhigh energy track is to use the field free regionoutside the solenoid coil. The technique then is tomeasure ~he straight trajectory in that region,extrapolate to the vertex, and measure the impactparameter.

Field free )region

Fig. 4. Momentum Measurement byImpact Parameter

Cal

67

The second limitation comes from the knowledge ofthe vertex position. The emittance of the beam isinvariant £0 = 20n mm mr. For Smachine = 150m,a = 0.2mm and Smax = 3000m, a = lmm, the size of thebeam pipe is --20a = 2cm. Consider low S quads at15m, then Smin = 15m, giving a = 70~ at the interactionpoint (correspondingly, Smin = 3m gives a = 30~ at theinteraction point).

Overall, this technique has the advantage oftracking in a field free region, where we have straightline pattern recognition. By this scheme we estimatethat all charged particles can be measured to op/p5-10% to 1 TeV. This technique will only work in thecentral region where the effective thickness of thecoil can be kept small.

3. Aspects of High Luminosity

3.1 State of the Art (L. Lederman)

A relevant question to ask in considering highrate detector systems is what is the "state of theart?". What limits the present generation of experi­ments and at what effective luminosity? One way toget a handle on this question is to look at a widevariety of high rate experiments at FNAL, ISR, AGS,etc. and try to average over the excuses (i.e. toomuch halo, trigger death, ...• ). This has been donein Table I and an attempt has been made to translatethese state-of-the-art experiments to a 1 TeV colliderin order to find what such a collider detector, builtin 1978-82, could have taken in luminosity.

/)

:)

TABLE I

STATE OF THE ART

~y?

Interaction Target/Experiment Sec Beam

FNAL E-

ABS

Rapidity Effectiven

cmLuminosity**

400

537

326*

615

673

609

623629515

288*

105

105

106

4 x 105

5 x 1011*

Si/400neutronsW/125 GeV

n,pVar. 225GeVn" 250 n-

Be 200 n

H2 400 P

200n,400pc/200 p, nBe/200lTP

eu/400 pPol. ~ 2 GeV/c

o

60"eu

10'Fe

8' Be +16' co

o

ooo

22'Be

1,6 n

2 IT

3 n

2 IT

2 IT

2.9 n

1, 8n2lT @ 900

2.5 n

•007 (~lT)

at 90

01' 1030

< 1030

01'

4 x 1030

1030

5 x 1030

2 x H30

+ 10 302 x 10

312 x 10311 x 10

2 x 1031

trigger deadtimeBeam tagging

DriftChamberstrigger deadtime + PWC'spb glasspileup PWC,mg.Halo muonsotherwise x5SC & D. C. jam.Liq A pileupLiq A pileupC ConfusionScinto counter+PWC rates

, )

)

1SR1-1(CCOR)

1-6(AFS)

30 x 30

30 x 30

o

o

2 IT

oI'lT

> 5 x 1031 No problemsexcept in E

t

2 x 1031

6 x 2030 ~P83-01

1.4 x 103 EP82-139 "10%pileup"

AGS726

732

13 GeV n o

o

large

1/5 of 4n 1031

datahandlingtriggeringbeam taggingbeamtagging

)

33 34* These experiments correspond to 4 x 10 and 1.2 x 10 resp. in a closed ge~~etry

with ~~ measurements before the absorber. At 1 TeV these numbers become 01' 1 x 10 and3 x 10 respectively.

** Not yet corrected for multiplicity.

68

o

This represents the added energy (due to multipleinteraction) superimposed on top of a given jetcluster. A second interesting question is how well adetector will respond to hard scattering. For example,consider 50 GeV x 50 GeV PT and use the simplealgorithm of adding all clusters with ET ~ 4 GeV. Forsingle events this works well as is shown in Figure 6.However, the same algorithm gives a very large shiftfor multiple events. It is evident that the challengein the presence of multiple interactions is to developsmarter criteria that can determine what energy tokeep and what energy to throwaway.

For the purpose of this study, it is assumed thatthe calorimetry covers ""4n and is divided into smalltowers and that tracking is divided into small cellsin the B-fie1d. So far this study has included MonteCarlo studies of the effects of mUltiple interactionson the calorimeter trigger and measurements of~~ for leading jets. Work is in progress on theef~ects on mu1tijets ~.g. W+ TB, QCD Studies) and theeffect on tracking for the high PT trigger. Otherinteresting questions that could be studied includethe effect on measurements of missing ET' isolatedshowers, pattern recognition and resolution, process­ing, mu1tievent tagging, etc.

Calorimetry Studies Consider the question ofp~bS vs. p~ea1 and how this is effected by thepresence of multiple interactions. The algorithm isto choose the largest PT inside a window. This hasbeen done for two windows:

~n = 1, ~q, = n/2 (Gordon)~n =< .6, M = n/6 (Yoh)

Due to QCO effects, a single window algorithm onlypicks up ""80% of PT' even for the narrower of the two"original jets" (see Figure 5).

The effect of multiple events is to give someadditional PT to the observed p~bS in a jet. Sincethe algorithm is to maximize EPT inside a window, thecenter of the window shifts and increases this effect.

o

o

o

o

o

o

There are four issues that need to be consideredin making this decision:

(1) Solid angle or acceptance must be translatedto a "standard 4 " collider detector.

(2) Environments may be quite different.

(3) Multiplicity varies with~

(4) Open versus Closed geometry.

Consider the first issue of translating the solidangle. For a physics thrust of the UAl or UA2 type,one needs to see correlated jets, leptons, missing PT,etc. So one must reduce the L by a fraction of solidangle which the fixed target detector subtended.(e.g. Accidenta1s A = R2 T n(n-l) for n separatepieces of a 4~ detector. Thus, we must use R/n tokeep A tolerable (P1 cuts have been neglected).

The third issue is the different multiplicities.(AGS -5 charged particles, FNAL -9, ISR -13, SPS -25,1 TeV -30). Since this burdens the detector theremust be corrections. This is partially adjusted forin the solid angle correction.

The fourth issue is how to account for closedvs. open geometry experiments. The closed geometrycan be translated to open geometry by correcting forthe hadrons absorbed. Absorbers typically give afactor of -2 10 3 to 104 and each experiment needs anadjustment in the luminosity according to the density.

Overall, the results tabulated in Table I,indicate that the present state of the art seem to bethat experiments have been rate limited at effectiveluminosities of L ~ 5 10 30 - 10 31 •

3.2 Effects of Multiple Interactions (Diebold, Gordon,Johnson, Linnemann, Paige and Yoh)

A study has been made of the physics consequencesof multiple events. The attempt has been to comparesingle interactions vs. <n> = 10. The characteristicsof physics at VS = 1 TeV are dn/dn ""6 particles(including neutrals), <~PT> "".4 GeV yielding~~ET z 2.5 GeV. The total transverse energyEET 'Ot 25 GeV, so that if <n> = 10, <EET> "" 250 GeV ~

50 r-----,---,---.-----,----r-----,(al

GordonYoh

Additional PT Due to <n> = 10n/4n EPT/4n Acutal Shift1/40 6 GeV ""8 GeV1/200 1.2 GeV ",,2 GeV

Ratio

""1. 3

""1. 7

pOBS vs pREALT T

<n>«1

Response to 50 + 50 GeV PT hard scattering

co

:110~

10-8 L __..L----=-_--L----l_-L__----.L__--l__--::-::

20 40 80 120 160 200 300

(~ PTI all clusters (YOH alogorithml

Fig. 6. Overlap effect on Apparent EPT

10-5 .----,.---..,.----,----,----,------,

60504020 30P real __..~

T

10

Fig. 5.

40

/\3(1

110l-e..

V 20

10

00

u

u

69

By tightening requirements it should be possible to geta fast trigger that has a rate ~l kHz (see theTrigger Group summary report). For the purpose ofstudying the processing problem,we assume that thisfactor can be obtained by asking for simple signaturesin the trigger.

20.000

16.000

12.000

en....CCII> 8000w

4000

00

pp -+ w± + XI.ev

<n> %below cut

-0 95%7.5 80%

II • Trigger

~~ Pr

I2 3 4 5

Look For (on-line)W± evW± ]JV

W+ y

gluinostt

new physics

Signature

e: PI > PI min & missing PI]J

e + y

missing PIe + jet + missing PI

?

,)

L PT in 2 rings around trigger cell

Fig. 7. Overlap Effect on W± events

Another physics problem to consider in evaluatingthe effects of multiple interactions is the reaction

pp.... W± + X

Lev

Figure 7 shows the added PT that affects the cutsfor isolated electrons. The plot shows that the effi­ciency of identifying isolated electrons is reducedfrom 95% to 80% in the example. given.

Conclusion for Multiple Events In conclusion, nofundamental difficulty from multiple interactions athigh luminosity has been uncovered. Although effectsexist, the trigger for hard scattering seems alright;detection and measurement of leading partons isalright, however the detailed picture including gluonshas problems for ptlUOns < 5 GeV. In the case oftracking, determining the jet core should not be muchworse, but the overall tracking will be more difficult.

Now, let's consider the problem of handling thistrigger rate.

{

1 event = 105 byteschannels = 2 10 5

Assumption Channels/ADC = 100

Channels are 10% occupied

The compaction and digitizing time (ADC) is then100 x 1 ]JS (below threshold) plus 10 x 5 ]JS (abovethreshold) giving a total time of 150 ps/event.Therefore the readout rate for full events ~ 5 KHz.This is alright for the assumed trigger rate, but thisreadout level could be a bottleneck on livetime.

Next, let's consider the data storage limits,

(1) Conventional Technology:1 25 IPS 6250 Tape Drive

= 5 x 10 5 bytes/sec= 5 events/sec

(2) Videotape SystemsMay go up to 108 bytes/sec

= 10 3 events/sec

If videotape systems are used, the problem isshifted to off-line processing. For the purpose ofthis study we assume a storage rate of ~50 Hz.

The added on-line problem is to do processing toreduce the rate written on the data storage systemfrom 1 KHz .... 50 Hz. Events must be thrown away inatime of 1 ms.

)

)

UAl Experience: 7 ms = high PI track1 ms =all ]J tracks

~50 ms = general algorithm

So we assume the average processing time to handle andfilter these events = 50 msec.

3.3 Processing Requirements at L = 1033 cm- 2 sec- I(T. Carroll, W. Hoagland, E. Siskind, H. Sticker)

The experience we start from is from UA1. Thefast trigger for UAl ~l/sec, which scales to ~201 KHzat 10 33 cm- 2 em-I. This trigger includes:

a) electron P > 10 GeVb) jet ET > 15 GeVc) tota1 ET > 40 GeV, InI < 1.4d) ]J

70

Define: Processor Unit = IBM 1683081 E

=3-4 VAX 11/780

)

)

()

o

,,)

)

Therefore, we estimate that to handle a 1 KHzrate requires:

= 50 PO= 50 3081 E= 200 VAX 11/780

At first sight, this is an awesome amount of on-linecomputing. However, this appears technicallyfeasible and even the cost is not outrageous. Onecould conceive of Fastbus module-processor (3081E) +1 Mbyte memory (cost ~$10-15K). So the cost of50 PU = $lM. The array of processors need to bearranged in an array that can pass this rate.

Data Rate = 103/sec x 10 5 bytes= 10 8 bytes/sec

Need ~3 Fastbus segments~3 Crates of microprocessors

This whole area of microprocessors on a Fastbus card,fast switch, and segment interconnects is an areawhere R&D could be extremely useful.

The off-line requirements are also very large,UAl 0.1 Hz on CDC 7600

= 0.03 Hz/l PU

Therefore the analyzed data rate/acquired datarate =10. This means to process the 50 Hz off-linewill require ~160 PU or roughly ~50 CDC 7600.

In conclusion, it appears technically feasibleto handle the rates from L = 1033 in terms of pre­processing and processing. In order to have flexibletriggers and power in data selection and analysis ina general purpose detector without strongly biasingthe physics at the trigger level various R&D effortsshould be strongly pursued.

a) arrays of microprocessor systemsb) data storage technologyc) software for preprocessingd) fast ADC's

3.4 Magnet Option (J. Appel, J. Christenson,L. Jones, J. Rosen)

The requirement of using a magnet in a colliderdetector was examined. The ideal was to determinewhat a magnet "buys" in terms of physics and what it"costs".

Abilities and advantages with a magnet include:

(1) Momentum Determination(2) Sign of charge determination (including

correlations)(3) Sweeping of low Pt tracks(4) Aid in particle identification

71

Abilities and advantages without a magnet include:

(1) Simplicity of construction(2) No field effect on photomultipliers(3) No coil (radiation lengths) before e.m.

calorimeter(4) Reduced need for track length and number

of layers in tracking(5) Opportunity to move calorimeters closer,

make smaller(6) Reduced cost

. v(7) Reduced smearing of radial llnes (C

signals, towers)(8) Reduced decay length (muon backgrounds,

and missing energy calorimetry)(9) Simplified track reconstruction (trigger

and off-l i ne) •

An example detector was considered thatidentified jets; measured Ejet' Etjet' M, Z; measuredmultiple jets (e.g. W+ t6, t + 3 jets); foundelectrons and isolated y's for .special jets andreconstructed "M" + Wy, yy, WW.

The conclusions of this study were that nooverwhelming benefit was found for considering ageneral detector with no magnet. The cost wasestimated to be .5-.7 of a similar capability magneticdetector. No unique physics capability was identifiedfor the nonmagnetic detector. No inherent extra ratecapability was found, except for analysis where thenumber of channels is reduced. Lastly, thereprobably is an advantage in flexibility or mutabilityfor a non-magnetic detector.

3.5 Cost Considerations for a High LuminosityDetector (B. Barish and R. Schwitters)

Introduction: During this workshop, technicalsolutions for the various detector components (track­ing, calorimetry, triggering, preprocessing, etc.) havebeen presented. A complete detector utilizing varioussolutions described by the working groups remains tobe done. At that point, the detailed work of measuringthe detector against possible physics goals needs to bestudied, in order to evaluate the actual ability to dophysics. That is probably the most importantunfinished job after the completion of this workshop.

In the absence of such a specific detector andphysics goals, we have ignored the detailed design ofa detector, and instead tried to assess the difficultyof incorporating the technical solutions posed in thisworkshop into a modern general purpose detectorcapable of handling the high luminosities. As a guidewe have used the Fermilab CDF detector and scaled fromthe known costs of that facility. We are aware that anew detector would be done differently and do thisexercise to identify the scope of such a facility(total cost) and the most costly areas where R&Dmight be most helpful.

In general most solutions posed at this workshopinvolve more segmentation, speed, preprocessing power,etc., rather than totally new directions for futurehigh luminosity detectors. The method we have usedto cost is to scale all costs relative to CDF. Newdetector components, like precision vertex detectors,or particle ID methods, have been ignored.

Cost Comparison We have compared the cost of CDFand a high luminosity detector in Table II. The ratioof HLD/CDF cost considerations were the following:

Magnet: Identical (includes coil and cryogenics)Factor x 1.

Tracking-Mechanical: We assume from the cell size5 times as many cells. Cince the cost in labordominated we assume 3 times the CDF cost.

(For reference, the CDF central detector has 12K cells,the vertex and intermediate 8K and the forward 8K.)

Shower Calorimeter - Mechanical: The ratio of fastliquid Argon gas we estimate 5: 1 (e.g. Tasso$75K/m2 x 2 (labor) x 1/2 (mechanical fraction) vs.Gas (CDG $500/m2 x 30 layers). For the scintilla­tion, we guess 1.3:1 (extra cost for photomulti­pliers, scintillator, etc.).The mix of gas:scintillator we assume is 1:1 likeCDF. Factor x 3.25.

Hadron Calorimeter - Mechanical: We assume slowgas, fast scintillator and use the CDF mix 1:1.Factor x 1.15.

Muons - Mechanical: We assume 2 x - more cells andcomplexity. Factor x 2.

Front-End Electronics: For CDF the mix is 1/2tracking and 1/2 other. For tracking in HLD wehave 5 x cells and 3 x complexity (multihit, chargedivision, etc.). For the rest of the front endelectronics, due to short gates, flash ADC's, morechannels, we assume an additional x 1.2 for addedsystem complexity for the necessary bandwidth.Factor x 10.

Data Acquisition: Scaling from CDF. Factor x 5.

Levell, 2 Trigger: Same as level (1,2) plus afactor x 1.2 for increase in system complexity as infront end electronics. Factor x 2.4.

Installation, Administration, etc.: Take overallHLD/CDF ratio. Factor x 3.7.

72

Conclusions: The overall cost of a CDF typedetector built for high luminosity appears to be ~3.7

x CDF in cost (complexity). This gives a total costof ....... $1 46M. A large general purpose detector facilitylike this could be considered for a high luminositymachine like (CBA). The detector facility wouldpresumably service ....... 250 physicists, making its cost/physicist not so different from LEP detectors or CDF.Technological advances and clever design could reducecertain costs.

Two items are of particular interest in terms ofpointing out areas most in need of R&D. The largenumber of channels and complexity have made the front­end electronics inordinately expensive ($69M). Acoherent plan for developing cheap front endelectronics is urged. Secondly, a total of $18M isestimated for Level Triggers and Data Acquisition.In addition, a large amount of off-line analysispower is needed. R&D on preprocessor systems, datastorage, and analysis systems is crucial for meetingour future needs, both in terms of sophistication andcost.

TABLE II

SUMMARY - HLD COSTS ($M)

System HLD/CDG CDF Cost* HLD Cost

Magnet 1 4.1 4.1Tracking 3 2.6 7.8Shower Cal. 3.25 5 16Hadron Cal. 1. 15 9 10~1uons 2 1.4 2.8Front-end Electronics 10 6.9 69Data Acquisition 5 1.4 7.0

ElectronicsLevell, 2 Trigger 2 3.0 6.0Level 3 Trigger 0 5.0On-line Computers 2.4 1.1 7.6Cable Plant 2.4 1.0 2.4Misc. l:L l:L 13.7

TOTALS <3.7> $39 M $146 M

*Includes Japanese and Italian Contributions

C)

)

o

()

o

G

o

()

,)

SECTION III: CONTRIBUTED PAPERS

,)

)

r )

TRACK CHAMBERS FOR HIGH LUMINOSITY*T. Ludlam and E.D. Platner

Brookhaven National Laboratory, Upton, New York 11973

o

o

o

C)

o

:J

I. Introduction

We examine the requirements for wire chambers aslarge-solid-angle tracking devices in a hadron col­lider at L = 1033 cm- 2 sec-l. This luminosity corre­sponds to an event rate of 50 MHz, and we assume anaverage charged particle density dn/dy = 3.6. Wetake as a starting point the known performance of theAxial Field Spectrometer drift chamber l and the MPSII drift chamber system. 2 The criteria which we haveapplied for a "safe" extrapolation of these systemsare:

i. The rate per wire should not exceed 2 MHz.ii. The sensitive time of the chambers (i.e.

maximum drift time) should be short enough to limitthe average number of pile-up events to no more thana few.

These requirements impose a high degree of seg­mentation: The cgntral and forward tracking systemseach require ~ 10 channels. This fine granularityof detection elements, imposed for rate capability,is also necessary if the tracking system is to havegood efficiency for resolving individual tracks injets.

Since the number of readout channels in such asystem is an order of magnitude greater than forexisting detectors, the development of new elec­tronics approaches is an R&D item of high priority.The probable cost of a readout system based on customdesigned integrated circuits should be much less thanestimates obtained by simply scaling the costs ofexisting detectors according to the number of chan­nels.

II. Central Tracking Chamber

We take a solenoidal geometry, and adopt thebasic philosophy of the Axial Field Spectrometerdrift chamber (Fig. 1). This chamber, which operatesat the CERN ISR with a typical luminosity of ~ 3 x10 31 cm-2 sec- l , uses "bicycle wheel" geometry with42 cylindrical layers of sense wires and 4° azimuthalsegmentation in each layer. The active volumeextends from an inner radius of 20 em to an outerradius of 80 em. With a position accuracy a~ =200 ~m in the drift plane, and a magnetic field of 5kilogauss, the momentum resolution is aPT/PT =.02 PT' The sense wires are staggered to resolveleft-right ambiguity, and the z-coordinate isdetermined by charge division. A principal benefitof this design is that the radial geometry andunambiguous space-point measurement makes the patternrecognition extremely robust: The chamber has ademonstrated ability to handle event multiplicities~ 50 tracks, and has operated at a luminosity of ~ 5times the design value with no loss of efficiency forthe track-finding software. 3

For a central detector at CBA we require a widerrapidity coverage than is afforded by the AFSdetector, and better momentum resoltuion for hightransverse momentum tracks. In table I we give theparameters for a conceptual design for a chambercovering ± 1.5 units of rapidity (25° ~ a ~ 150°).The chamber layout is illustrated in Fig. 2. Aspace-point accuracy of 200 pm is assumed, in amagnetic field of 15 kilogauss. Drift chamber wires,in the bicycle wheel configuration, are arranged in

75

radial groups, "rings", with 15 layers of sense wiresper ring (10 layers in ring 4). Each wire has chargedivision readout.

If only the first three rings are used, so thatthe chamber extends to a radius of 1 meter, the

momentum re:~~Ujtion :s.0027 PT'

PT R=lmThe addition 0 the fourth ring, extending the radiusto 1.5 meters, provides a 10% momentum measurement atP = 100 GeV/c. We imagine, however, that some com­promise might be necessary in the rapidity coverageof the outermost ring, which would otherwise call forwire lengths of 6 meters.

The cell size, a, in Ring 1 is 2 mm, corres­ponding to a maximum drift time of 40 ns and anazimuthal segmentation of 314 sense wires per layer.In the outer ring the maximum drift time is 120 ns,with 750 anodes/layer. At a luminosity of 10 33 cm-2

sec-1 , and dn/dy = 3.6, the rate per wire is 1.6 MHzin Ring 1 and 0.7 MHz in Ring 4. For 1% charge diviGsion accuracy we require an avalanche size of 5 x 10electrons (gas gain ~ 1.0 x 105), so that the signalcurrent will be 1 pamp per wire at the innermostlayers.

The charge division measurement will be rathercoarse, with an uncertainty of several centimeters inthe outer layers, but it is also highly redundant andshould be adequate for pattern recognition and ini­tial track reconstruction. For precise measurementof the Z coordinate, each ring has one special layerin which the position along the wire direction isdetermined from the induced charge distribution oncathode pads. 4 These "pad layers" are subdividedinto many sections, as given in Table I. Each sec­tion consists of a row of pads which are resistivelycoupled so that the position along the pad row isfound by charge division (Fig. 3). The accuracy ofthe measurement is ~ 2% of the length of the pad row,provided the section is not occupied by multipletracks. Thus the number of pad sections is largeboth because of the desire for millimeter accuracyover large surface areas and the necessity to resolveindividual tracks.

Such a chamber is an order of magnitude morecomplex than those of present generation detectors,approaching 105 readout channels each of whichinvolves pulse height measurement of ~ 8-bit accu­racy. Mechanical techniques for assembling alargedevice with such a high density of wires are beingstudied. Feasible solutions, based on methodsalready used in large chamber construction, have beenidentified and will be studied further with prototypedevices. The readout for such a· chamber calls for amajor R&D effort. Systems based on miniaturizedfront-end amplifier circuits and high speed flashencoders for time and pulse height measurement can bedesigned using existing technology, but a much higherlevel of circuit integration seems possible, andshould be developed. A further discussion of thispoint is given in Section V.

III. Tracking in Jets - A Study of Segmentation

In order to withstand high rates we are led to ahigh degree of segmentation, taking the form of a

Second coordinate measurement:Charge division: a .01 L (L wire length)

Pad layers: a 5~ 1 romAvalanche gain: 1.0 x 10Magnetic Field: B = 15 KgaussMomentum Accuracy:

R = 1.0 m (rings 1-3): 6PT/PT = .003 PTR = 1.5 m (rings 1-4): 6PT/PT = .001 PT

TABLE IParameters of Central Track Chamber

Position accuracy in drift coordinate: a~

200 11m

i. Squares: The cylindrical surface is sub­divided into square detection elements (68 = ~~).

ii. Pads: Each detection element subtends 20 mrin~. This would correspond to wires with 4 romanode-to-anode spacing at a radius of 20 em. Furthersubdivision is achieved by segmenting the readoutalong each wire, e.g. with cathode pads.

iii. Wires: No segmentation in 8. Each detec­tion elem~orresponds to a wire running the fulllength of the cylinder.

Our measure of efficiency for resolving tracks is thefraction of tracks missed due to an overlap of two ormore tracks' in the same detection element ("cell").

Thickness of Material at 8 = 90°(wires, gas, pad layers): .02 rad. len.

.003 coIL len.

Starting Cell Size No. Anodes No. TotalRing Radius (cm) (a ,rom) per layer Layers Anodes---

I 20 2 314 15 4,7102 50 3 523 15 7,8453 90 4 720 15 10,8004 130 6 750 10 7,500

30,855

Charge Division Pad Sections(each wire) (l layer each ring)

Ring L(cm) aZ(cm) Section Size No. Sections aZ

1 120 1.2 2 x 2.5 em2 4,000 .5 rom2 230 2.3 4 x 3 cm2 6,000 .6 rom3 400 4.0 5 x 5 cm2 8,000 1 rom4 600 6.0 5 x 10 cm2 10,000 2 rom

28,000

very high density of wires. The desire to resolveindividual tracks within jets also calls for a fine­grain detector, and for this purpose it is evidentthat some kinds of segmentation are more effectivethan others. A jet consists of a high-densitycluster of tracks which occupies a small fraction ofthe detector area. Thus, for a given number ofdetection elements, it is best to have them as fine­grained as possible in~ coordinate, acceptingcoarse granularity in the other coordinate. This isso because the distance between jets is large com­pared to the distances within jets:

The results from 40 GeV jets are shown in Fig.4. Tracking in jets is best achieved with fine wirespacing, and the wire densities arrived at in Table I(> 300 cells/layer) correspond to efficiencies of, 80% for resolving tracks in jets. Figure 5 showshow the results, for segmentation in wires, varywith jet momentum. All results are for a 15 kilo­gauss solenoidal field.

IV. Forward Tracking Drift Chambers

There is a 25° half angle cone around the beamsthat is not covered by the central tracking cham­bers. This cone is most simply filled with high-ratechambers such as those in MPS II which use simpleplanar frame construction. Here again we impose theconstraint that no anode shall have more than a 2 MHzevent rate. A sim~le approximation to the event rateper cm2 gives R/cm '" 3 x 107/r2 at~= 10 33 where ris the distance from the beam pipe, Fig. 1. Theintegrated event rate per wire is:

2SR/d tan-1 LIdwhere S is wire spacing, 2[ is anode length, and d isdistance of closest approach. For i » d and s = 4mm, this limits d to 19 cm. If chamber modules areemployed as in Fig. 2 using 4 rom anode spacing out to50 cm, 6 rom spacing between 50 and 100 cm and 10 romspacing outside of r = 1 m the 2 MHz rate limitationper wire will not be exceeded. Applying MPS IIexperience we would construct modules having pairs ofX, Y, U and V chambers as shown in Fig. 3 where eachpair is staggered by a drift space. Since thesemodules are in a field free region, 4 sets of suchmodules will cleanly generate point slopes and elimi­nate left-right ambiguities. Table 2 is a listing ofthe anode wire count.

)

, )

vv

Xx;< )If\'

X70 XxX

TABLE 2

Module Anode Length Anodes/Plane

D1 3 1.0 KD2 4 1.2 KD3 6 1.6KD4 7 1.8 KD5 8 2.0 KD6 9 2.2 KD7 10 2.4 KD8 11 2.6 K

As a quantitative study we have used ISAJET togenerate jet events into a cylindrical detector andexamined the two-track resolution for three types ofsegmentation:

76

This table adds up to 14.8 K wires per plane forthe 8 modules. However there are 8 planes per moduleor 120 K anodes for the spectrometer. In order tocover the full 4n, including the central detector,one needs 2 spectrometers or a grand total of 240 Kchannels. This electronics requires only the TOCfunction so inexpensive MPS II-type electronics isapplicable.

References

The cost estimate of the tracking detectoramplifiers and discriminators is a hard number takenfrom MPS II experience. The ADC cost is an estimatebased on an analysis of development costs for otherstate-of-the-art IC production history. It should benoted that at the lOOK part level we are dealing withindustrial levels of production which bring veryattractive pricing.

2.7 M

11 M2 M

$ 3.3 M

$19.0 M

* Research supported by the Department of Energyunder Contract # DE-AC02-76CH00016.

1. O. Botner et al., Nucl. Instrum. Meth. ~, 315(1982). --

2. E.D. Platner, Proc. Int. Conf. on Instrumentationfor Colliding Beam Physics, stanford, California,February 17-23, 1982, p. 24.AFS Collaboration, "Operation of the AFS at 1.4 x1032 cm-2 sec- 1: A First Look at High LuminosityData From the CERN ISR," these proceedings.V. Radeka, "Centroid Finding of Induced Charge inGas Proportional Chambers," these proceedings.

4.

270K

TABLE 4

3.

ADC's

TABLE 3

Energy HadronGeV {Energy Resolution II of Bits

400 20 2.5% 9100 10 5 7 ADC1

50 7 7 610 3 17 4 94 2 25 81 1 50 6 ADC2

.5 .7 70 5

.1 2

Drift Chamber Amplifiersand Discriminators

330KTDC's

11OKIC Development

V. Electronics Development

Time and amplitude electronics should be dead­timeless and pipeline delayed by at least 1 ~s inorder to facilitate trigger decisions. Because ofthe desire to reconstruct jets as well as providehigh rate capability, with good efficiency, it isnecessary to have fine granularity for both trackingchambers and calorimetry. The number of channelsinvolved is ~ 105 for both the TDC and ADC functions.

Systems of this size affect the design philo­sophy in that one can consider substantial investmentin custom IC design that can radically reduce theactive component count, reduce crate count and im­prove reliability. As an example, the MPS II driftchamber electronics required the development of 3custom IC's, the development phase costing about 1/2the total electronics expeditures. However, due tothe design freedom inherent in fully custom IC tech­nology the design goals of a fully pipelined time­to-digital conversion in which there are no adjust­able and thus potentially unstable parameters thedetectors described in this paper the only further ICdevelopment we would recommend would be the engi­neering and production of a very low noise multi­channel amplifier IC. The other active componentsrequired have been in use at MPS II for a year.Remaining development would include hybrid packagingof these components since the active circuit powerlevel is low. This should allow a spatial compres­sion of up to a factor of 4 over that of MPS II.

The detector system we describe here will alsorequire a pipelined ADC system probably with samplingrates of 200 MHz. The charge division ADC shouldhave 8 bit resolution with 9 or 10 bit dynamicrange. The current state-of-the-art is availabilityfrom 3 or 4 sources of 6 bit 100 MHz flash ADC'scosting $50-100. There is also an 8 bit 30 MHz ADCwith 9 bit dynamic range currently availale for ~

$8. These are examples of commercial products. Itis our belief that with a reasonable engineeringbudget, i.e. $1-2M, flash ADC's satisfying ourrequirements can be produced in a time frame of 3 to4 years. Once such devices are developed, productionquantities should be less expensive than the ADC'snow available for HEP because of the economics ofdedicated IC systems.

()

o

o

()

( )

The ADC requirements imposed by high speedcalorimeters are similar to that of the trackingcharge division channels except that the dynamicrange is much greater. Here one requires a dynamicrange of .1 GeV to 400 GeV with 2 bit resolution at.1 GeV improving to 8 bit accuracy and resolution at400 GeV. Table 3 shows how 2-9 bit ADC's can be usedto cover this dynamic range while contributing negli­gible error to the measurements in a hadron calori­meter.

)The large general purpose detector described

above will require 30K TDC channels for the AY =± 1.5 detector and 240K TDC channels for the < 25°forward detectors. It will also require 90K channelsof 9 bit ADC for the charge division and pad readoutin the AY = ± 1.5 detector. In addition, the Calori­meter with 5K channels each for the electromagneticand hadronic calorimeters require 2-9 bit ADC'stotaling 20K channels. Table 4 is a compilation ofthis total electronics requirement.

.) 77

r."..l·.~· ...r'_xZ (Nonlonol it

dltf(.hon )

:)

·:::::::::::::::::::::::..:\. \ .••.1 : :......................................; .; ...

" I i

I

\ t ·•·..···-·· ··~: .

\L:ffi:~~..·.~~~~~·.~·.·.h-~~-.1 n 3S 'J1 )9 " :: '.. . .; :-.:.-I" / .,

z:·-v-·~·~~~Y

Figure 1The Axial Field Spectrometer Drift Chamber (Ref. 1).

ELECTRODE STRUCTURE' x=ANODE WIRES

~j31!::::::::a=CELL SIZE

1.6DRIFT WIRE;

RING 4

1.4

1.2 !PAD LAYER

~ 1.0 RING 3~ 0.8Q)

oS 0.6 RING 2a::0.4

0.2 RING I

0 2 3Z (meters)

Figure 2Schematic Layout of the Central Tracking Detector.

78

)

< )

()

o

o

o

CATHODE STRIP

SECT\

} SECTOR

ANODE

/WIRES

o

\'. j

--------------~

Figure 3Induced Charge Measurement for Precise Z Coordinate on Pad Layers (Ref. 4).

Figure 4Track resolution in jets, for various kinds of segmentation in the central rapidityregion (y ~ 1.5). The points are ISAJET calculations for PT(jet) ~ 40 GeV.

79

)1.0

lawVl 0.8~::;: Pr (JET): J.Vl 200

1J~

~ 0.6100 ()0::

l-LL.

~ 0.4QI-0

~ 0.2LL.

,-) ,

100 1000NO. WIRES/LAYER

Figure 5Track resolution in jets, for segmentation bycylindrical layers of wires. Figure 6

Geometry for flux calculation in forward spectrometers.)

SPECTROMETER

bLE

01 02 03 04 05 06 0708

(

CALOR IMETER

. CORRECTIONOIj:)OLE

Figure 7Layout of forward spectrometer.

I)' I( I

)

2I I4 6 8 meters

")

Figure 8Orientation of wire planes in one module of theforward spectrometer.

80

()

o

o

THE PERFORMANCE OF A 16000 WIRE MINI-DRIFT MWPC SYSTEM

P. Chauvat, R. Cousins 1 , K. Hayes, A.M. Smith

CERN, 1211 Geneva 23, Switzerland.

R. Bonino, J. Ellett, M. Medinnis, A. Russ, P.E. Schlein 2 ,

P. Sherwood, J.G. ZweizigUniversity of California*), los Angeles, California 90024, USA

J. Alitti, B. Bloch-Devaux, A. Borg, J. B. Cheze, A. Diamant-Berger,I. Giomataris, B. Pichard, F. Rollinger, M. Tararine, J. Zsembery.Centre d'Etudes Nucleaires-Saclay, 91190 Gif-sur-Yvette, France~

o

u

(}

C)

The Mini-Drift chamber system and associated electron­ics employed in experiment R608 at the CERN-ISR aredescribed. The performance of the system which uti­lizes 3mm spacing MWPCs produces a spatial resolutionof a = 180 1.1 under experimental conditions.

INTRODUCTION

While it has been believed since the early daysof Multi-Wire- Proportional-Chamber usage that spatialinformation Was available in the arrival time of chamberpulses [G. Charpak et a!. -NIM 62, 235(68)], no full­scale experiment has, to our knowledge, made use ofthis information to improve resolution. This was duepartly to the vast amount of electronics necessary totime-digitize signals from a typical experiment withtens of thousands of channels, and partly to the ad­vent of larger cell drift chambers.

ty logic which has a data flow rate of several hundredMegabytes per second. In particular, we have con­structed and gained experience with a large systemconsisting of 514 32-channel amplifier-cable-TDC units.

Besides the electronics demands made byMini-Drift, the inherent problem of left-right ambiguityresolution is exaggerated because the tracks are alwaysclose to the nearest sense wire ( 1.5 mm in our case).Redundant measurements are necessary to insure ahigh probability of correct solutions.

The charged-particle spectrometer of experi­ment R608 at the CERN ISR has been in operationsince February 1981. The system performs well evenat the highest ISR luminosities, with good multi-trackefficiency and momentum resolution. The global spatialresolution, measured to be a = 180 1.1, results in mo­mentum resolution of ap/p = 5.8 % AT 26.7 Gev/c.

In our application at the CERN Intersecting­Storage-Rings, we were faced with a relatively highrate (,,1Q6/sec) and particle density in the forward di­rection of j5p and pp interactions. Moreover, the avai­lable magnetic field integral was low ("0.3 Tm), so thata = 200 1.1 was needed for good momentum resol ution.Thus, the Mini-Drift solution was chosen.

Nonetheless,high rates or highwould like to retainwhile achieving thedrift chambers.

in experimental situations withspatial density of particles, onethe narrow cells of MWPC systems

a = 200 1.1 resolution typical of

SPECTROMETER AND CHAMBERCONSTRUCTION

The layout of the double septum magneticspectrometer is shown in figure 1. The wire chambersof the spectrometer, constructed at Saclay, consist offive pairs of MWPC packs placed above and below the

ISR beam pipe with 3 pairs upstream and 2 pairsdownstream of the magnet. The chambers above thepipe and those below the pipe can be thought of asconstituting two independent spectrometers, which werefer to as the UP and DOWN spectrometers, respec­tively. There are 16448 wires in the system.

,)

The NEVIS group of Sippach et al. pioneeredin the development of a compact 16 channel 6-bit TDCusing ECl electronics. In the system described here,we have carried on in this direction in developing a32-channel 4-bit ECl TDC module. We have also de­veloped a fast readout system and associated multiplici-

Each chamber pack has four planes of wires,X, X', Y and U, the latter ,being inclined at approxi­mately 37 degrees to the vertical. The X' and X wiresare vertical and are displaced horizontally with respectto one another by 1/4, 1/2 or 3/4 of the 3 mm wirespacing.

81

The chambers are constructed with 7 mm bet­ween a sense wire plane and each of its neighboringcathode planes. The cathode planes are constructed ofmylar with graphite coating on both sides. The cham­bers are operated with the cathode planes at -3.3 kVgiving the potential distribution shown in Figure 2.This distribution is typical of MWPCs, and is differentfrom the standard drift chamber field configuration. Inorder for the pulse arrival time to measure the dis­tance in the sense wire plane between a sense wire andthe traversal point of a track in that plane, the ioniza­tion electron nearest the sense wire plane must not betoo far from that plane. With 40 or more ionizationcenters per track traversal and with electronics sensi­tive to the first electrons, this condition appears to besatisfied. Clearly the velocity is not near saturationin the well-known low-field region midway between thewires but this appears to pose no problem as discussedin section 1.4 below.

The chambers are filled with 60% argon, 39%isobutane and 1% of a mixture of argon containing 1%freon. The argon is bubbled through isopropyl alcoholat 4 degrees celsius. While we have not made exten­sive tests of the optimal gas mixtures for this applica­tion, it is noteworthy that the desired results havebeen obtained with this rather typical unexotic gasmixture.

CHAMBER ELECTRON ICS

The amplifier cards consist of four amplifica­tion stages each utilizing a Motorola MC10115 Emitter­Coupled-Logic (ECL) integrated circuit, followed by aMC10116 line-driver stage. The inputs of the 10116'sare biassed to 200 mv so that an amplified chambersignal of -300 mv produces a full differential logicswing at the output. With four cascaded stages of am­plification, a. small change from the nominal VEE of-5.2 V allows the overall gain to be adjusted from amaximum of 1500 down to 800 or less. In ou r system,the amplifier cards are operated at -4.8 V, resulting ina gain of 1100 and a lower operating temperature forthe cards. At this setting, there is a 0.3 mV thres­hold which produces differential ECL logic pulses of1.2 V at the end of the 70 meter cable. The cablesare Ansley 101-wire flat polyethylene cables with everythird wire grounded.

The TDC system, shown schematically in Fig­ure 1, was designed and manufactured at UCLA forthis experiment. It digitizes the drift times and en­codes the addresses of those wires in the system whosesignals arrive during an 80 nsec Levell trigger gate.

It provides a high-speed path for transfering the wire

adresses and drift times to central Randem Access Me-

82

mory (RAM) on receipt of a Level 2 trigger signal fromthe experiment. Finally, it makes its own Level 3trigger decision based on multiplicity logic on thechamber hit data. The Level 3 decision is the compu­ter interrupt.

The TDC cards also have 32 channels so that·there is a one-to-one correspondance between TDC andamplifier cards. Up to sixteen TDC cards are mountedin a 19" crate along with their read out module, alt­hough a slave crate with up to eight additional modulescan also be read out serially by the READOUT modulefor the larger chamber planes.

The time digitization is accomplished by latch­ing the state of a 4-bit Gray-code counter at the arri­val time of a chamber pulse. The Gray-code signals ina given crate are generated by a central clock startedby a Level 1 signal from the scintillation hodoscope up­stream of the spectrometer. The time bins are 5 nsecwide. Only one "hit" per wire can be registered foreach event.

There are 40 READOUT modules, one for eachof the wire planes in the system. After an event hasbeen gated into the TDC's, the Level 2 trigger logiceither causes the event to be strobed into the MEMORYmodules or the entire system is reset. On receiving agood Level 2 ("start readout") signal, the addresses ofthe wires hit and the digitized times are strobed intothe MEMORY modules at the rate of 75 ns/hit with all40 planes read out in parallel. The total time after agood Level 2 signal to read out all TDC modules to the

MEMORY modules is typically 375 nsec for a 5 ·track

event.

There are 20 MEMORY modules, each of whichreceives data from a pair of wire planes in the upperand lower spectrometers. Each MEMORY module countsthe number of wires hit in each plane of the pair sepa­rately as the data is received, and compares the hitcounts with upper and lower limits which have beenpreviously down-loaded to the modules from the PDP-lldata aquisition computer. Each module also comparesthe sum of hits from both pairs to a set of upper andlower limits (each plane in the system has its own setof limits).

At the end of the readout process, the results ofthese comparisons for all MEMORY modules are trans­ferred to a hard-wired processor module called VI RTUS(Very Intelligent Real Time Uvent Selector) which car­ries out the Level 3 majority logic.

!j

", /

)

, )

)

o

C)

SYSTEM PERFORMANCE

Because of the large size of the system(77,366 IC's), there was initial concern that normal fai­lure rates would make smooth, reliable data acquisitiondifficult. Experience has shown that this is not thecase. The hardware in the system works well. Typi­cally, there are 20 channels dead at the amplifier card

level.

Perhaps once or twice a month, a TDC board (ormore likely one of the 41 switching supplies that powerthem) develops a problem. They are easily replacedby spare modules.

The typical drift velocity of 50 lI/nS impliesthat the time spread of hits in the chambers should bearound 30 nsec. A histogram of raw hits recorded ona typical wire for 1000 events (figure 4) shows thatmost hits occur within 5 or 6 time bins, as expected.Variations from the nominal 5-nsec time bins have beencalibrated with a computer-controlled pulser in order tocorrect for the effects of imperfections in the Gray­code transmission lines.

The small tail at large times in Figure 4 is duemainly to late arrivals from tracks which pass midwaybetween two neighboring sense wires. The drifttime-distance dependence has been studied by usingtracks which have zero drift time in an X-plane wireover a large range of incident angles. The calculatedtraversal point in its neighboring X'-plane is comparedwith the drift time information from the struck wire inthat plane. The results are shown in Figure 5, whichis seen to be linear up to about 1.25 mm from thewire. The last .25 mm or so, where the curve levelsoff, corresponds to the very low field region seen inFigure 2. The observed non-linearity presents no ana­lysis problems - large times correspond to large dis­tances - but it is necessary that the Level-l gate belong enough to have full efficiency. Our measuredchamber efficiency is typically 99% with the 80 nsecgate.

About 5% of the time, a particle will cause twoadjacent wires to fire (i .e. a cluster). Figure 6 is ascatter plot of the times (in units of tdc bins) for the

two adjacent wires in two-wire clusters. It can beseen that both hits have times which lie in the tail offigure 4. Most clusters then result when a trackpasses through the low field region (see fig. 2) near acell boundary. (Other mechanisms such as <S-ray pro-'duction or the presence of a nearby track would usual­ly result in a short time for at least one of the wires.This fact is useful since it allows immediate resolutionof the left-right ambiguity (discussed below) of any2-wire cluster. Closer examination of events in Figure

6 in which both recorded times were less than 30 nsecshows that these are mainly due to the presence of asecond track.

OFFLINE ANALYSIS

The offline pattern recognition program findstracks using only the wire hit information. The mo­mentum is then determined by fitting to the coordinatesdetermined from the drift-time information. The spatialresolution can be illustrated in the following way.Consider the hits in the X and X' planes for one trackcrossing a particular chamber. The time informationcan be used to calculate the distances, d 1 and d 2 re­spectively, of the track from the hit wires in theseplanes (see Figu re 7). If we assume that the trackpasses on particular sides of its two wires, d 2 can bepredicted using the known slope and the measured dis­tance d 1 • Figure 7 also shows a histogram of the dif­ference between the predicted and measured d 2 for allhits in one chamber for the case where the track hasbeen assumed to pass on the right side of both wires.The peak centered at zero comes from those instances

where this assumption was correct, while the shoulderat larger values contains all the cases where the as­sumption was wrong. The width of the peak is \1'2times the actual resolution which is typically cr = 180 II·The relative areas of peak and background in Figure 7depends on the relative offsets of the X and X' planesand on the angular distribution of the tracks used. Adirect measurement of the momentum resolution isshown in Figure 8, which is a histogram of the recon­structed antiproton momentum in elastic scattering at26.7 Gev. This distribution has crp/p = 5.8%.

Finally we comment on the question of left­right ambiguity resolution with our system. The solu­tion of the left-right ambiguity in any given plane re­quires that we can predict, using all the availableinformation for a track, where it passed through thatplane. The ambiguity will be solved if one and onlyone of the two possible solLitions is consistent with the

data from all the other planes.

We solve the ambiguity problem in two steps.First we calculate the track trajectory using only thecoordinates of the hit wires. This gives us, to goodapproximation, the local direction cosines of the trackat each chamber pack. The ambiguities are then simul­taneously solved for all four planes in a chamber byexploiting the two constraints which relate the fourmeasurements of the track coordinates in the chamberpack. This is done by considering each of the 16 pos­sible solutions. If only one combination satisfies thetwo constraints, all four ambiguities are solved. If

83

two or more solutions satisfy the constraints, then theambiguities for one or more planes remain unsolved,depending on the specific combinations.

Using this technique, typically 84% of thecases have at least one X or X' ambiguity solved andabout 44% have both solved. Hits for which theambiguity is not solved are usually those where thedrift distance is small ( < 4001-1 ), but these cases areunimportant as we already know that the track passedclose to the wire. Typically 44% of y-view ambiguities

and 56% of u-view ambiguities are solved.

We have attempted to resolve the remainingunsolved ambiguities with a global track fit to the 20hits in the 5 chamber, using the drift-distance cor­rected coordinates for those hits whose ambiguitieswere solved using the method described above. We ob­served only a small improvement in chamber. 2 and noimprovement elsewhere. This is because at the cham­ber level, our track measurements are not highlyover-constrained; a different choice of solution results

in slightly different track parameters, and not neces­sarily a different i for the fit. This illustrates theneed for redundant coordinate measurements in achamber pack if a high efficiency of ambiguity resolu­

tion is to be obtained.

The time-distance relationship and the resolu­

tion are systematically controlled in the experiment us­ing straight tracks recorded with the ISR operatingbut with the spectrometer magnet turned off.

ACKNOWLEDGEMENTS

We very much appreciate the excellent work atS.T.I.P.E (Saclay) in the design and construction ofthe chambers. We are grateful to the U.S. NationalScience Foundation and to the UCLA Physics Depart­ment for the support which allowed the development ofthe chamber electronics. We thank CERN and the ISRExperimental Support Group for their important help insetting up the entire system. P.E.S. thanks BruceKnapp and Bill Sippach for many invaluable discussionson their experiences with Mini-Drift.

'}

)

VETO COUNTERS

Figure 1

MAGNET

M.W.P.C.

SHOWER COUNTER

HODOSCOPE

)

)

)

Layout of the R608 spectrometer.

84

800

600

2000

2400

3200

I~ 1,5mm.r-CATHODE PLANE

2

1

c:= .§Q:::::>C =400 =:>800

133

1

o

1600

2000

C) 2400

2800

3200

FIELD DISTRIBUTION3mm WIRING SPACING

o

C)

2. The electric field distribution around one an­ode wire.

COMPlJTlRU.lTE RUPl

3. Block diagram of the chamber TDe and read­out system.

85

)

)

50 (nsec)

DRIFT TIME403020

TIME BIN NQ105

0,5

50

100

')

DRIFT(mm) DISTANCE

COUNTS PER TIME BIN

-)

250 1,5

"CELl:'SIZE

200

}1,0

150

4. A histogram of the times recorded in the

TDG's of one chamber.

5. The time/distance relationship obtained from

fitted tracks.

)

2- WIRE " "CLUSTERS TIME CORRELATIONS

14

12

10ru

.!:: 8~

*-.c. GO't.;:c 4oruE 2i=

30nsczc

I~ I ~ ~ 5 ~ ~ i .1 ~

1 4 I r ~ 3 ; ~ ~ ~ ~2 ~ I ~ ~ ~ ~ 1 1 ~

1 2 ~ 3 I ~ ; 1 ; ~ ~ Z ~_1_3_2_3_1_3+2 3_1_4_4_30nse'"

1 3 1 1 2 1 2 1 1 1 ~

2131 ~I 1

~ 2 1 ~ ~ I

~ 1 I I --+--Ij-:I-+I-.:..-t-I--+1---+1-""'11---1-1--+-1--+1---+1-2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Time on left wire

)

)

6. A scatter plot of the times recorded when twoneighbouring wires both register hits (for hitsused on found tracks).

86

()

NUMBEROF EVENTSPER 200pbin

03000-

2000-

1000-

0

:)

o 1 2 3 4 5 6 mm( d2 calculated - d2 measured)

-

-

~FWHM

'3 Ge Y

:)X

X'

0

... 1loIIII... j.. . •

_.• I

C)

.J

)

. )

7. An illustration of the method employed to mea­sure the resolution.

87

8. Measured momentum distribution of elastic an­tiprotons at 26 Gev.

BREAKDOWN PROCESSES IN WIRE CHAMBERS,PREVENTION AND RATE CAPABILITY

M. Atac

Fermi National Accelerator Laboratory.

)

)

Summary

Breakdowns were optically and electronicallyobserved in drift tubes and drift chambers. Theyoccur at a critical gain for given intensity in a gasmixture when ultraviolet photons are not completelyquenched. It was observed that the breakdownsdepended critically on average current for a givengas mixture independent of the size of the drifttubes used. Using 4.6% ethyl alcohol vapor mixedinto 50/50 argon ethane gas, breakdowns areeliminated up to 7 ~A average current drawn by pulseson a 1 em section of an anode wire under an intensesource. Pulses with an avalanche size of 106

electron rates above 106 pUlses per centimeter perwire may be obtained with the elimination ofbreakdowns.

Introduction

Operation of drift chambers and proportionalchambers has been a painstaking struggle for many.groups until the end of their experiments. For everycase it took a learning period to accomplishsatisfactory operation. Some people trained wires byallowing a small current for burning offcontamination on cathode wires, some found certaingas mixtures improved the situation, some were ableto operate at low gain by reducing pickup noise, thusincreasing amplifier discriminator sensitivety, etc.In most cases, especially where beam rates were high,hot wires showed up in a random way. These hot wireshad to be replaced or cleaned or disabled.Unfortunately no one has written a few paragraphs toexplain what helped them for satisfactory operation.For this reason no references can be given for theabove cases. All information was obtained byfriendly private communications.

Most of the problems may have been due toarbitrary chamber designs and casual selection ofgases. One took someone elses design, because thatworked. It may have worked with certain gain and forcertain rates, but it may have failed at higher beamintensities. Detailed studies of breakdown processesin gases were done 1 earlier using mainly parallelplate geometry. There has been no methodic study ofbreakdowns with very commonly used gases in multiwiredrift chambers or drift tubes.

This paper is the result of some systematicstudies of breakdown processes and prevention ofchamber breakdowns related to beam rate and gain.

a. Breakdown in Drift Tubes

The following tests were carried out to helpunderstand intensity and gain related breakdownprocesses in wire chambers. For simplicity twoaluminum tubes of different sizes, a conductiveplastic tube 2 and a small drift chamber with a fewwires were used for the tests. All the tubes and thechamber had transparent windows for optical

• Operated by Universities Research AssociationInc. under contract with the United States Departmentof Energy

observation. These window~ were made of In-Sn Oxidetransparent film on mylar. The film is sufficientlyconductive to serve as a cathode material. Thealuminum tubes were 9.5 x 9.5 mm 2 and 12 x 12 mm 2

cross-section, and the conductive plastic tube was 7mm x 10 mm. Configuration of the drift chamber isshown in Fig. 1. The anode wires in the tubes and ofthe chamber were 50 ~ thick gold plated tungsten.The gas mixture for these tests was 50/50 percentpremixed (by volume) argon-ethane flowing throughethyl alcohol (CH

3CH20H) at OoC at a rate of 200

cc/min.

A narrowly collimated (large fraction of theintensity within 5 mm diameter circle) Sr 90 source of1 mCi strength was held at a short distance from thetubes. Intensity of the source was regulated byaluminum shim absorbers.

As shown in Figs. 2-a,b and c, independent ofthe source position along the wires in the tubes, aspontaneous breakdown occurred at a critical gain anda critical intensity. It occurred randomly within aminute (from a few seconds to a minute). Thus thepoints were taken at three minute intervals. Themanifestation of the breakdown is a sudden increasein the high voltage power supply c~rrent that slowlyreaches to a value around 10 ~ which is limited bythe gas impedance (there was no current limitingresistor in the circuit). This current is sustaineduntil the high voltage is lowered considerably.

Average de current drawn through the powersupply was monitored. An interesting fact is thatthe current was approaching 0.4 to 0.5 ~A when thebreakdown occurred independent of the tube size andintensity. The conclusions from the results are thatthe breakdown is critically dependent on totalaverage current (i.e., gain x intensity) for thiscollimation of the source illuminating mainly 1 em ofthe anode wire, and the value of the current is 0.5~ for this gas mixture. It is an important matterto note here that after so many breakdowns the wiresheld the voltage values at the intensities given.The wires and the cathode surfaces were examinedunder a microscope showing no observable damage. Thegain characteristics of the 9.5 mm x 9.5 mm tube isgiven in Fig. 3 using a Fe 55 source to get an ideaabout the gain values for Fig. 2a. Primaryionization characteristics of the Sr 90 source israther different; Sr 90 results in a wide range ofprimary specific ionization.

As an exercise let us take Fig. 2a. Breakdownvoltage is around 2.4 kV for the source intensity of1.25 x 10 5 sec- 1

• The gain value obtained from Fig. 3for 2.4 kV is 10 5

• 0.5 ~A average de currentcorresponds to 5 x 10- 7 Coulomb of charge. This is ~3.1 X 10 12 electrons. An average avalanche size(number of total electrons after multiplication) pertrack is

~ 2.4 x 107 electrons •

Then the average primary ionization per track is

88

)

)

)

)

o

o

o

c)

J

)

,)

2.4 x 107= 240 electron-ion pairs.

105 -

This is a reasonable number for the source.

The same tests were attempted with the smalldrift chamber using the same gas mixture and the samesource strength. No breakdown up to 7 ~A de at 3.1kV applied voltage was found. Self quenchingstreamers showed up at a voltage around 2.6 kV and at3.1 kV mainly streamers were observable.

b. Optical Observations and Rate Capability

Using an image intensifier video cameradescribed in an earlier paper,~'5 photons from theactive area were observed while the average decurrent was measured as a function of applied highvoltage. Pulses from the anode wires were alsoobserved to detect streamer transitions when theapplied voltage was sufficiently high. High voltageversus measured current characteristics is shown inFig. 4. Space charge saturation begins at 2.3 kV(avalanche saturation). The streamer threshold isaround 2.75 kV.

Once the breakdown is eliminated the driftchamber wire can take very high rates. Figure 4 alsoshows that up to 0.4 ~A current gas gain isexponential. This current corresponds to over 2 x10 6 pUlses per second per one centimeter length ofthe anode wire for a uniform flux of a-rays havinggas gain above 10 5:

0.4 ~A = 0.4 x 10-6 coulomb/sec, corresponds to

4 x 10-7 = 2.5 x 10- 12 electrons

1.6 x 10-19

Avalanche size of a typical drift chamber pulseis about 10 6 electrons, thus

2.5 x 1012 6----- - 2.5 x 10 pulses/sec. 1 em wire106-- -

Photon activity around the anode wires of thedrift chamber makes the wires visible as shown inFigs. 5a,b and c. The source rate was .1'10 5 countsfor each wire per second. The higher the gain themore photons are detectable. Some short streamersare visible at 2.8 kV (Fig. 5b). Many more can beseen at 3.1 kV (Fig. 5c). These and further picturesare the negatives of the polarized photographs forbetter visibility in copying processes. Thephotographs are 1/20th second time exposed picturestaken from a video monitor. The author believes thatthese photons come from some recombination processesnot from normal avalanche processes. A hint for thisis at low rates (less than 10~/sec) only streamerscan be seen when the voltage exceeds 2.7 kV; nophotons could be detected below the streamertransition. A large number of photons are emittedfrom the space surrounded by cathode wires when thesource intensity is high and the gas gain issufficiently high, and the space is filled by an ioncloud. Even then no breakdown, (i.e., no sustainedactivity) was observable up to 3.1 kV with the gasmixture when the source was pUlled away.

89

c. Breakdown in the Drift Chamber

This time the 1.4% ethyl alcohol component ofthe gas mixture was removed; only 50% A - 50% C2H6mixture was used. The S-source was kept at the sam~

place from the chamber, thus providing about 10counts per second per wire over mainly 1 cm oflength. In this case, wires were visible at 1.8 kV.A boiling hot circle appeared (fig. 6) at 1.9 kV.The circle became very visible at 2.1 kV and itstayed active when the source was pulled away. Thehigh voltage had to be lowered below 1.8 kV for theactive circle to disappear. This and other phenomenawill be explained further in the paper. The highvoltage was turned off and on each time the hot spotappeared (at the same place). Each subsequentdischarge appeared at successively lower voltages.Thus the first breakdown left a scar on the wires andit was getting worse at each repeated breakdown.Figure 7 shows the relation between the average decurrent and the applied voltage. The breakdowncurrent of 0.31 ~A is interestingly close to what wasfound with the tubes previously.

The above results indicate that CH3

CH20H is animportant element in the gas mixture in preventingthe breakdown and probable surface damages related tointensity and gain.

d. Preventing Breakdowns in the Conductive PlasticTube

More tests were carried out for understandingthe effectiveness of the ethyl alcohol component inthe A - C H gas mixture in preventing breakdownsmentioned2 ~n the previous sections. The conductiveplastic tube was chosen for the following experimentsbecause of its small size (7 mm gap).

As explained in the previous sections,breakdowns did occur at critical gains and criticalintensities (Fig. 2c) with the conductive plastictube in the gas mixture of 49.3% A - 49.3% C2H6 ­1.4% CH~CH20H, but there was no breakdown in thedrift Ohamber until the ethyl alcohol vapor wasremoved. To help understand these phenomena, theethyl alcohol vapor concentration was varied from 0to 4.6% and breakdown current was meas~red with asource intensity exceeding 1.5 x 10 pulses persecond. The results are shown in Fig. 8. Breakdowncurrent is very low (around 0.1 ~A) without CH~CH20H

and the breakdown is practically eliminated witfi 4.6%CH

3CH20H in the gas mixture. There was no breakdown

up to 7 ~A. This is a very respectable current(i.e. gain x rate per second), therefore no attemptwas made beyond this intensity and the gain.

It was clear from the results above that ethylalcohol did enable the tube run at much higher gainwithout breakdown. Then the next logical experimentwas to study other alcohol vapors, namely, methanol(CH~OH) and isopropanol «CH~)2 CHOH). Figure 9sho~s partial vapor pressUres of the alcoholsmentioned as a function of temperature. The numbersare taken from "Handbook of Chemistry and Physics".From these curves we find that we need to keep CH~OH

at 3°C and (CH3) CHOH at 35° C to have 4.6% of e~chin the A - C H6 mixture. Clearly, the latter wasfound not €o be practical. Tests also showed thatbubbling A - C2H6 through (CH

3)2 CHOH at room

temperature was not sufficient for preventing thebreakdown. CH OH had to be kept around 10°C to beeffective. TRis may be due to the reason that CH~OHis a less complex molecule than CH

3CH20H, and tfius

would have less rotational and vibrational absorptionlevels for quenching ultraviolet photons which couldreach the cathode and knock out electrons. CH~CH20H

should also be preferred because it is not an actrvevapor for plastics, epoxy, etc.

Optical studies also confirmed the resultsabove. Some details are given in the next section.

e. Optical Observation of Geiger Mode

The conductive plastic tube was studied throughthe image intensifier camera to gather moreinformation about the breakdown processes. Asdiscussed in the previous sections, breakdowns couldoccur in the tube when the ethyl alcoholconcentration in the argon-ethane gas mixture wasbelow 4.6%. Such a breakdown was observed on thevideo display. Bright spots around the anode wireshowed up. After a few seconds more bright spotsappeared and then the brightness spread quickly allalong the wire. At this time about 10 ~A current wasdrawn through the high voltage power supply eventhough the source was pulled away. It continueduntil the voltage was lowered below 1.5 kV. Thissequence is shown in Fig. 10a,b and c.Interestingly, the holding voltage of the wire stayedthe same after many breakdown tests even afterallowing 10 ~A of breakdown current to continue formore than 10 minutes with the existence of ethylalcohol vapor in the gas mixture. After the tests,surfaces of the anode and cathode were examined undera microscope and no visible damage was found.

Is this a true Geiger mode? This question willbe discussed further in the paper. There was noobservable breakdown on the screen when 47.7% A ­47.7% C2H6 - 4.6% CH~CH OH gas mixture was used.Figure 11a and b show thS pfioton activity around thewire at currents of 0.7 and 3~, respectively.Figure 11b shows projections of the self-quenchingstreamers around the anode wire.

The Geiger mode-like spread along the anode wireof the conductive tube could not be seen in thealuminum tubes. The breakdowns were around localspots. A reason for this may be due to the aluminumtubes being considerably larger in size. All theevidence found with this and earlier works indicatethat breakdown spread along the wire is mediated byelectrons knocked out of the cathode walls byultraviolet photons, unlike the well known Giegeroperation. 6

f. Epoxy Droplets on Cathode Wires Affect Operationof the Drift Chamber

Some cathode wires of the beam drift chamber(mentioned in Sections a and c) were smeared withepoxy which formed droplets around the wires as shownin Fig. 12. Optical observation on the video displayshowed that at 2.2 kV (0.16 ~) a hot spot on thewire appeared with 1.6% ethyl alcohol in the gasmixture and it got brighter at 2.7 kV (2 ~) seen inFig. 13. The hot spot disappeared when the sourcewas pulled away. The position of the spot was in theregion where the epoxy droplets were. This could notbe observed under the same conditions given abovewhen the same anode wire plane and a clean cathodewire plane were used. A probable explanation forthis phenomenon is that the epoxy charges up with thepositive ions, polarizes and becomes an electronejector (Malter Effect). If this is so it couldexplain why the spot disappeared when the high flux

of positive" ions stopped (removing the source).

g. More Boiling Hot Circles

It was discussed in Section c that circlesshowing boiling-like activity in the drift chamberwere observed with the image intensifier camera using50% A 50% C2H6 gas mixture and this breakdowncondition was completely prevented by the addition ofethyl alcohol vapor. The tests were done with anewly wired chamber frame. This time identical wireplanes of a damaged beam drift chamber were used forthe tests. The chamber had some breakdowns while.operating with the A - C2H6 gas only. There wereclearly seen scars on the anode wires.

Hot spots were observed in this chamber up to anethyl alcohol concentration of 3.3% in the gasmixture. Figure 14a shows the hot spots around theanode wire with the B-source illuminating the chamber(breakdown current of 3 ~). Figure 14b shows thatthe hot spots remained after removing the source. Aboiling circle showed up near the cathode that feedsone of the hot spots on the anode when the camera'ssensitivity was increased (Fig. 14c). Hot spotscompletely disappeared when 4.6% alcohol vapor wasused, as seen in Fig. 14d. The conclusion is thateven a damaged chamber can be successfully used whencomplete UV quenching is achieved.

h. Some Hints from the Experimental Results

The following conclusions can be summarized onbreakdowns and their prevention from the experimentalresults given in this paper:

1. The breakdown is strongly dependent upon theaverage current for a given gas mixture(i.e., higher gain for smaller beam rate).

2. The current =gain x primary ionization xrate can be increased with additionalquenching vapor.

3. For all types of the chambers examined hereethyl alcohol, CH CH OH is an excellentquencher in reduCiag 6reakdown probability.4.6% ethyl alcohol vapor in 47.7% A - 47.7%C2H~ mixture appeared to be sufficientlygood quenching gas for currents up to 7 ~

without any breakdown, even with damagedwires.

4. Epoxy droplets may produce hot spots onanode wires.

5. In conductive plastic tubes of 7 mm size, aGeiger-like mode was observed.

6. There was no detectable damage on the anodeor cathode surfaces after successivebreakdowns using the Ar - C2H6 - CH

3CH20H

gas mixture.

7. A lifetime test was made with the driftchamber using 49.2 A - 49.2% C2H6 - 1.6%CH CH20H and a high flux Sr 90 B-source.ThJre was no detectable damage on the wiresfor integrated pulses of 10 11 with a gain of10 5 • This amounts to a total of 2.4 x 10 18

electrons/em on the wires (see Section a).

8. Electron drift velocity is affected by theaddition of ethyl alcohol vapor. Thisremains to be measured in a systematic way.

90

)

()

9. Last, but not least, there is a great dealmore to be learned on the subject.

1. Breakdown Processes

Some explanations can be given on how and whybreakdowns occur in wire chambers or drift tubesusing the results given in this and the earlierpaper. s These explanations may be supplementary tothe work edited by H. Raether 1 for the parallel plateconfiguration.

1. In the avalanche process, a critical chargedensity is reached before the streameraction so that radiative recombination canoccur. s

scars may be left on anode and cathode surfaces.Some tests were done using 70% argon 30% isobutane(instead of ethane) mixture bubbling through ethylalcohol at temperatures up to 17°C, adding about 4.6%alcohol to the mixture, using the conductive plastictube. Every breakdown left damage on the anode andthe cathode surfaces, thus high voltage had to belowered. Why couldn't the alcohol prevent breakdownwith isobutane mixture and damages? This should beinvestigated.

Acknowledgement

The author would like to express hisappreciation to Dr. M. Mishina for many discussionson the sUbject.

2. What makes the streamer grow is mostprobably a reproduction of electrons nearthe streamer head by the recombinationphotons.

References

1. H. Raether, Electron avalanches and breakdown ingases, Washington Butterworths (1964).

Fig. 6 - Hot breakdown circle in the drift chamber

5. M. Atac, A. Tollestrup and D. Potter,Nucl. Inst. and Meth. 200 (1982) 345-354.

6. S. A. Korf, Electron and Nuclear Counters (VanNostrand 1955).

2. M. Atac et al., IEEE Trans. on Nucl. Sci.,Vol. NS-29, No.1 (1982) 368.

3. M. Atac, NucL Inst. and Meth. 176 (1980) 1-8.

Fermilab Report,

- Photon activity around the anode wires ofthe drift chamber at 2.5, 2.8 and 3.1 kV.The pictures are negatives of thephotographs taken from the video displayof the image intensifier camera.

- Gain characteristics of the 9.5 x 9.5 mm2

aluminum drift tube in the saturatedavalanche and streamer regions.

- Gain characteristics of the drift chamberusing the high intensity Sr 90 a-source oflOs pulses/sec for 1 em wire. The gasmixture is 49.3% A - 49.3% C2H6 - 1.4%CH~CH20H. Drift chamber wire can take veryhigh rates once the breakdown iseliminated, as indicated in the figure.

Fig. 5

Fig. 4

Fig. 3

Figure Captions

Fig. - Cross section view of the drift chamber.

Fig. 2a - Breakdown curve for 9.5 x 9.5 mm 2 sizealuminum drift tube. Independent ofsource position along the wire there was abreakdown at a critical high voltage(i.e., gain) at the given intensity whenaverage current reached around 0.5 ~. Thegas mixture was 49.3% A - 49.3% C2H61.4% CH

3CH20H.

Fig. 2b - Same as 2a for 12 x 12 mm 2 size aluminumtube.

Fig. 2c - Same as 2a for 7 x 10 mm2 size conductiveplastic tube.

4. M. Atac and A. Tollestrup,FN-337 (1981).

Conclusions

3. Another probable way of streamer growth mayalso occur when a highly intense beam orsource is used. Coincidentally, someelectrons happen to be present from anothertrack near the tip of an avalanche conealthough the critical charge density may nothave been reached for regeneration ofelectrons by the radiative recombinationphotons. These accidental electrons couldin succession enable the avalanche to growinto a full streamer. The streamer may growclose to the cathode. When this happenselectron regeneration from the cathode couldfollow by the photons knocking outelectrons. This condition could end up witha continuous feedback from the cathode asfound in the breakdown conditions when thequenching was insufficient. This feedbackmechanism continues even after removing thesource (Sections c and g). Figure 15 showssome of these cathode electrons fed backbreakdown circles dramatically (boilingcircles). Interestingly, the circles areidentical in size, about 3.6 mm in diameter(real size). Top views of some of theindividual self quenching streamer pictures svery much resemble the circles (Fig. 16).

One should not allow breakdown to occur withouta protective vapor like ethyl alcohol, otherwise

If this is all correct then the anode cathodeseparation and cathode configuration play animportant role in electron regeneration by photonsfrom cathode surfaces. Wire cathode configurationreduces the probability of such photoelectronproduction, thus it should be preferred relative tocontinuous cathode coverage (simple solid angleconsideration). Work function of the cathodematerial may also play an important role in thismatter.

A suggestion resulting from this work is thatone should do a careful study of gain characteristicsof a chamber designed for a specific purpose under anintense source using a gas mixture to match withelectronics (mainly amplifier- discriminatorcircuits). It should not be assumed that even forrelatively low intensities one can increase gainarbitrarily.

o

,)

o

91

when the ethyl alcohol component of thegas mixture was removed. Diameter of thecircle is about 3.6 Mm. It remained thereeven after removing the source.

Fig. 7 - Average current as a function of theapplied voltage with the existence of thebreakdown circle shown in Fig. 6. Therewas no ethyl alcohol vapor in the driftchamber.

Fig. 8 - Breakdown current as a function of ethylalcohol concentration in A - C2H6(50/50)for the conductive plastic tube. It showsthat no breakdown occurred up to 7 ~A whenthe ethyl alcohol concentration was 4.6%.

Fig. 9 - Vapor pressure curves of ethyl, methyl andisopropyl alcohol as a function oftemperature.

Fig. 10 - (a) Hot spots on the anode wire of theconductive plastic tube when the gain isabove a critical value for the hot source(see Fig. 2c). 1.5% ethyl alcohol in thegas. (b) More hot spots appear after amoment. (c) Geiger-like spread along thewire.

Fig. 11 - Breakdown was eliminated in the conductiveplastic tube by 4.6% ethyl alcohol in theA C2H6 (50/50) gas mixture up tostreamer operation with the hot source.(a) Photon activity at 0.7 ~, (b)streamer photons at 3 ~.

Fig. 12 - One of the cathode wire planes of thedrift chamber with epoxy droplets whichwere deliberately smeared on the wires forthe study.

Fig. 13 - A hot spot on the anode wire of the driftchamber in the vicinity of the epoxydroplets. Gas mixture contains 1.6% ethylalcohol. HV = 2.7 kV. Average currentwas 2 ~.

Fig. 14 - An identical drift chamber with damagedwires. (a) hot spots on the anode wireeven with 3.3% ethyl alcohol vapor in thegas; (b) hot spots remain after removingthe source; (c) showing one of the hotspots that was fed by a boiling circlenear the cathode wires; (d) hot spotscould not be observed with 4.6% ethylalcohol in the gas.

Fig. 15 - Breakdown circles and hot spots on theanode wire of the drift chamber moredramatically seen when the ethyl alcoholwas removed from the gas mixture. (a) Thesource is on, (b) source is off. At arather low HV = 2 kV and 2.1 kV, averagecurrent is around 0.12 ~.

Fig. 16 - (a) A pair of branched streamers grownlike a cone with the cathode fed electronsknocked out by ultraviolet photons. (b)An imaginary cone around (a). Top view ofthe cone is about the same size as the hotcircles.

92

)

: )

)

)

)

o

1.5

7 l( IOrnm2

conduClive ploSllC lube

1.0Source Intensity(counts/sec)

.5

2.56

2.48

2.40

2.32~<II

2.24'"0~.c 2.16.S!':r

2.08

2.000

Transparent windowIn-Sn Oxide filmon mylar

·················E•.•••.•. • 12mm

...... OOjl;;;j

100fLm Ag-plated-Cu-SeField shaping wires

o

()

o Fig. 1Pig. 2c

o3.5 9.5><9.5mm2 aluminum rube

o

~ 3.0

<II

'"J?~.c£' 2.5

5040

30

20r- Streamer Thresholdl for 5.9 keV

10

o2.0L---..l.5----1J.,'O-------:I~.5----:!2x-l:::0·5---'

Source Intensity(counts/sec)

;:;Q,

3

Fig. 2il

0.5 L-.---,r---r--,---,---,-----.---2.72.6

Fig. 3

2.3 2.4 2.5High Vollage (kv)

2.2

2

12)( 12 mm2 aluminum lube

.'\". )

3.5

~ 3.0<II

'") J?,..~.c

'"£2.5

) 2.0 L-----!.5L----,l1.o,,--------,�;,.5,---~2':'x"10"'5:-----J

Source Intensity(counts/sec)

Fi<J. 21.1

93

,) !

!)

. )

U0

«::l..

c 10'1

t:::>

)U>-0.a.:::>

(j)...Q);:

C£

)

162

1.8 2.0 2.2 2.4High Voltage (kV)

Fig. 7 .)

Fig. 6

._ ..-.--"-_.

(a)

------ ----'-1'12.5mm

... . ... .J .

2.5kV

2.8kV~__..~.",ifI"'-----

10",A

/-'"'"/ 0

/ g/

/ IDa./ lJ)

//\streamer

c:eu

/ threshold Q)

0;

1 "Q,/ '0

1 Q)

'I .!::llJ)

U Q).c0 0c:

<! 0

:i. ~0

c ...st E:::> 0U10"1 U

Q)lJ)

......Q)

0a:::C:::>0U"U

~Q)a.)(

W

102

2.0 2.2 2:4 2.6 2.8 3.0 3.2

High Voltage (kV)

Fig. 4

3.1kV

(b)

(el

.. 5:J:

N0

.t 4

.:

is 3N

:J:0..,:J: 2u

~

No breakdown up to 7Jl A---------------,

..."//'

I"1

1I

11

/I

)

2Breakdown Current (Il A) DC

3 4

Fig. 5Fig. 8

94

()(b)

Fig. 11

o

o (a)

Fig. 9

.,-.. .., _.-) Fig. 12

(hi

)

• •(e)

Fig. 10

95

.', ., ,.·.·f .11I,

;.>r.~~~':', '~.: /::, Hot Spot." ~. '·0 ,;... , '.

,." "II" ..;" ,,'

Fig. 13

I,

(a)

I~__•__·_·_I(b)

(c)

I" ~:i'~'::' .·1

~ .';,.'•·..'..:,;,j:illlii~··.··IO!tlii.i~.::".~.~:.".".,' ,

(d)

Fig, 14

rig. 16

(bl

')

') I

)

. )

)

)

Anode ::.::' : '. " •Wires ; "

\'.t": ,:,,,

.• ' :"! •..••. ". :-~ .. " .'.I··l,t'·..·........... ~

. . . .

fa'

-HOI SpO)

HOI Circles

Fig. is

.~.

(bl

96

)

)

oTHE CONCEPT OF A SOLID STATE DRIFT CHAMBER*

E. Gattit, and P. Rehak

Brookhaven National LaboratoryUpton, New York 11973

I. General Description

The Solid State Drift Chamber (SSDCH) is a thinwafer of a high purity n-type silicon (few cm2 x a fewhundreds ~m thick) with a single small-area, sma11­capacitance anode readout (Fig. 1). The drift voltageis supplied to an array of "drift" electrodes on bothsides of the wafer to produce a uniform drift fieldparallel to the surface of the wafer and to ensure thecomplete depletion of the wafer. 1

FRONT SURFACE

loo~·c--l~~:Iso.;-;oo;-~~h.o ~.o 34.0 4Jt.O 4~.MICRO"

·,1~ ~

~ !SURF,A<: .

"100.0 -so,o 0,0~ICRONS

TOWARD COLLECTING ANODE

DRIFT FIELD ELECTRODES

II"I

DRIFT FIELD ELECTRODES

/1"-// I

I

DRIFT DISTANCf

ANODE

o

o

()

I:10 mm

Fig. 2. Profile of the central "gutter" in a SSDCH.The parabolic slope of the channel is due to the re­maining fixed charges in the n-type silicon.

() the gas drift chamber resolution (o-rays and fluctua­tions in the ionization density) are largely reduced.

()

Fig. 1. Solid State Drift Chamber. Silicon wafer isabout .3 mm thick and about cm2 in the front area. Themajority of the surface is covered by drift field elec­trodes (only electrodes at the extremes of the waferare shown). The ionization produced by a high energyparticle drifts toward the anode which is the onlyreadout channel on the wafer.

The wafer is in fact completely depleted throughthe anode which sits on the highest potential of allelectrodes (minimum electron potential energy). Byvirtue of remaining fixed volume charges and the ap­plied potential to the drift electrons, the potentialenergy of the electrons inside the wafer has the formof a parabolic "gutter" inclined towards the anode(Fig. 2).

The following position resolution limiting factorswere considered:

i) amplifier noiseii) electron diffusion

iii) silicon leakage current

The theoretical resolution is expected to be in the re­gion of (2 - 5) ~m, i.e., in the region previously ob­tainable only by emulsion detectors. The test of thedevices presently under the construction at BNL will becarried out in the spring of 1983.

II. The Solid State Drift Chamber in High Rate,High Multiplicity Environment.

We would like to evaluate the utility of a SolidState Drift Chamber as a vertex detector for thestorage ring experiments.

, )

, )

The electrons produced by a fast charge particleinside the wafer moves to the center of the "gutter"and drifts towards the anode. The time difference be­tween the particle passage (scintillation signal) andthe arrival of the electrons to the anode is the mea­sure of the drift distance in a similar way as in agas drift detector.

Due to the much higher density of the silicon com­pared to the gas, the two fundamental limitations of

* This research was supported by the U. S. Departmentof Energy: Contract No. DE-AC02-76CH00016.t Present address: Instituto di Fisica, Po1itecnicodi Milano, Piazza Leonardo da Vinci 32, Milano, Italy.

The ability to measure the position of particleswith high precision is not the only requirement of avertex detector. The multi-particle resolution and theability to handle the high rate are the additional re­quirements for a vertex detector.

These additional requirements can be satisfied ina SSDCH by paying a certain price in the degradation ofthe position resolution. (The drift velocity has to beincreased to reduce the memory time of the chamber andalso to reduce the diffusion of electrons to facilitatethe double track resolution.)

Let us state the merits and limits of a vertex de­tector for a planned hadron colliding beam accelerator.

) 97

A special intersection region with. the length ofa few cms only is required to match the dimensions ofthe vertex detector. The same beam intensities whichproduce the luminosities of the order of 1033/cm2s willproduce the luminosity of a few times 1031/cm2s ; stillhigh enough to carryon many interesting experiments.The mean time between two interactions is about .5 - 1.liS.

A drift field of 103V/cm in silicon corresponds tothe drift velocity of 13 IIm/ns. With the maximal driftdistance of 1 cm the chamber memory time is about 0.7liS.

The calculated double track resolution is about100 11m. For' tracks closer than 100 11m it is still pos­sible to measure their total number (the total chargemeasurement) and the center of gravity.

The cal~ulated precision of the position measure­ment is down to the (5 - 10) 11m region limited by theamplifier noise.

The surface covered by a . SSDCH is of the order ofa few cm2 • The full vertex detector thus would consistof a few thousands SSDCH with the same number of read­outs. More resolutions along the second coordinate canbe achieved by an anode segmentation at the expense ofa larger number of readout channels.

References

1. E. Gatti and P. Rehak, to be published in NIM.

~y acceptance of this article, the publisher and/orrecipient acknowledges the u.s. Government's right toretain a nonexclusive, royalty-free license in and toany copyright covering this paper.

98

'J

)

)

)

()

()

o

o

u

~)

.J

Large Area Non-Crystalline Semiconductor Detectors

V. Perez-Mendez, T. Mulera, S.N.Kaplan, P. WiedenbeckLawrence Berkeley Laboratory, University ofCalifornia, Berkeley, CA 94720

Summary

The properties of various non-cryst~l~ine semi~~n­

ductors are considered for use as pOSitIOn sensitivedetectors. Amorphous silicon and conducting plasticcan be doped to form NP depletion regions similar tothose in single crystal silicon, but without the limita­tion of single crystal size. Chalcogenide glassymaterials such as Te-Se-Ge compounds as well as somemetallic oxides such as the Vanadium oxides haveswitching and memory properties. They could serve asx,y location identifying devices when trig&ered byamplified pulses from a parallel plate or multIst~p gasfilled detector stage in order to resolve the multItrackambiguity for x,y readout schemes.

I. Introduction

Single crystal semiconductor detectors have beenwidely used in Nuclear Physics resea~ch as energymeasuring devices. For Yspectrometry In the few KeVto MeV energy range, back biased NP junc:ionGermanium single crystals of ultra-pure germaniumhave given the highest en7rgy resolution of any device.Position sensitive detectIon of charged partIcles hasbeen done with back biased NP junctions in siliconcrystals with a plated,metallic g~id to give th~ positioninformation. In high energy phYSICS these deVIces havecome into use as. vertex detectors in colliding beamaccelerators, with spatial resolutions in the few tens of·microns range.

All of these applications of 'silicon and germaniumsemiconductor detectors are limited to small areausage by the fact that they are slabs of single crystalmaterial. Larger area devices could be made by use ofa mosaic of crystals at a high cost and with a great dealof electronic complexity. Furthermore, if the eventswhich traverse the crystal have a high multiplicity, theusual multi-track ambiguity from crossed grid detectorsis still present.

We propose to develop the properties of threeclasses of non crystalline semiconductors for uses asradiation detectors, which would not be subject to theabove mentioned limitations. These are (a) Amorphoussilicon; (b) semiconducting plastics and (c) chalcogenide- Metallic oxide compounds.

Both amorphous silicon and semiconductingplastics - primarily polyacetylene - can be doped bysuitable materials to form N or P type layers and thuscreate a depletion region by back biasing in the usualmanner. Since both of these materials are non-crystal­line the limitations as to size do not apply; however,the' difficulty with the multi-track ambiguity stillremains.

Amorphous silicon can be deposited in lar?~ areasby various proccesses such as the decomposItIon ofSiiane-SiH41. Doping with trivalent and .p~ntavalentimpurities can be done by vacuum depOSItion, or bysputtering techniques. The mobility of the elecro~s andholes in this material is appreciably slower than In the

99

single crystal silicon,2 and thus there may be ~eriouslimitations to the resolving time of such layers If theyare more than a few microns thick. Multiple layers canpresumably be made in order to. h~v7 a sU.ffi~i7nt

stopping power for detection ~ft minImUm IO~IZIn.g

particles. Amorphous. si.licon, being non::rystalhne. ISless susceptible to radIation damage than ItS crystalhnecounterpart.

The development of conducting plastics hasoccurred primarily during the last few years. Amongthe various types3 the most promising at this point isPolyacetylene which can be made by vacu~m

techniques4 in layers up to a few m.m. thick. Dopingwith trivalent arid pentavalent impurities for PNjunctions formation can be done by electrolyticmethods5 or by deposition from the gaseous phase.6One present difficulty with the conductingPolyacetylene is that it is unstable in the ~resence ofoxygen; this difficulty can, however, be readIly resolvedby encapsulation with thin mylar sheets.

A third approach to large area solid statedetectors is to use the switching properties of somechalcogenide (Se-Te-Ge-Sj) and metallic oxidecompounds. These materials switch from a high resis­tance state to a low resistance state; some under theaction of an electric7 field and some at a criticaltemperature8• The Vanadium Oxides V203 and V02 forexample switch at temperatures of 14?OK and 6~oC

respectively. Some of the Se-Te-Ge-SI chalcogenideshave been used as electronic latches in prototypecomputer memory arrays9. These. ~evices can ~e madein large areas by vacuum depOSItIon, sputterI~g ~nd

similar techniques. The temperature SWItchingcomponents are slower than the electronic switchingcompounds 00-100 Jlsec compared to 1-20 nsec).

The switching compounds have the followinguseful features.

(a) They are readily made and there is somecommercial background on their use

(b) They are more radiation resistant than thesemiconducting plastics

(c) The switching properties and their memorycan be used to resolve the multi trackambiguity when x,y crossed grid arrays areused.

Figure 1 shows the voltage current relation fortypical electronically switched chalcogenides. Sincetraversal of a semiconducting layer by a chargedparticle will not produce a net potential differenceacross the layer, these materials could be used inconjunction with a gas filled chamber and serve as amemory at each node of an orthogonal grid of wires orplated strips similar to a glow chamber memorydevice 10.

Figure 2 shows the configuration of an argon orneon filled chamber (with a suitable quenching gas),either working as a parallel plate chamber or in amultistep chamber configuration. The pulse ofelectrons (107 to 108 electrons) can be drifted onto aplated layer of chalcogenide material, causingswitching at locations where charged particlestraversed the chamber. A readout scheme in which then2 locations are sampled is shown in Fig. 3. Each xline is pulsed sequentially to switch off any conductingelements on that line. The corresponding y locationsare read out in synchronism. The speed of such areadout can be as short as 10-20 n/sec per x wire,which is appreciably faster than the corresponding glowmemory device.

References

1.2.3.

4.

5.

6.

7.

8.

9.10.

W.E. Spear, Adv. Physics 26 (1977) 811.R.S. Grandall, J. App. Physics 52 (1981) 1387.A.G. MacDiaimid and A. J. Heeger, SyntheticMetals 1 (1979/80) 101.C.K. Chiang et al., App. Phys. Letters 33 (1978)18.Y.W. Park et al., SoHd State Commmunications 19(1979) 747. -­J.F. Kwak et al., Synthetic Metals 1 (1979/80)213. -T. Tani et al., Solid State Communications 33(1980) 499. ­N.F. Mott and E.A. Davis, Electronic Processes inNon-crystalline Materials, Clarendon Press,Oxford (1979).J.K. Higgins et al., J. Non-Crys. Solids 18 (1975)77. --S.R. Ovshinsky, Phys. Rev. Lett. 20 (1968) 1450.T. Mulera, M. Elola, V. Perez-Mendez, and P.Wiedenbeck, IEEE Trans. Nuc.Sci. NS-24 (1982)425.

Fi g. 2.

J T_ _________ GAS. AMPLIFICATION

lSECTION

~~~ . MEMORY LAYER

-+VDC

Schematic of proposed detector. The gasamplification scheme shown here is a multi­step avalanche chamber.

)

The correlated x-y readout scheme. Thecircles represent memory elements which havebeen switched on by the passage of a chargedparticle at that location.

)

)

)

c=:JR

received pulse..n..X-Stripes

rr-"-- VI

'"WcJl.rl'

0

VIUZ

'" Ww:= w

Stripes w ~

rio' u~w

'" c'"w c

VI 0-' u

r'"::> Ia. >-

tted L.-e-H-

I~ +YDC uY II

IX-COORDINATE ENCODERS I lIXIlCOMPUTE3 PULSE TRANS:mTERS I

Y-

transmipuls.n.

Fig. 3.

,,,,,On ..,,,,,

Off'

'\,\

\ ,\ ,,,,

This work was supported by the Director, Office ofEnergy Research, Office of High Energy and NuclearPhysics, Divison of High Energy Physics of the U.S.Department of Energy under Contract No. DE-AC03­76SF00098.

Fig. 1. Current voltage characteristic of achalcogenide glass switch.

100

o THE DESIGN AND EXPECTED PERFORMANCE OF A FAST SCINTILLATOR HADRON CALORIMETER*

R.B. Palmer and A.K. GhoshBrookhaven National Laboratory, Upton, New York 11973

o

o

o

o

()

()

I. Introduction

We consider a calorimeter of the type develoredby W. Willis, et al., for the ISR 807 experiment.Figure 1 shows the basic concept. The calorimeterconsists of a stack of alternating metal (copper oruranium) and scintillator plates 20 cm wide, 120 cmlong and 1.5 mm thick. Light generated in the scinti­llator passes out through the edges and into wave­length shifter bars that lie between the stacks. Thelight from the scintillator is absorbed in the shift­ers and excites lower frequency emission that thenpasses down the shifter to phototubes in the rear.The 807 geometry has in fact separate shifters for thefront electromagnetic and rear hadronicparts of thecalorimeter but we will not be considering this com­plication here. We will also restrict the discussionto copper plates.

A typical pulse from the 807 calorimeter is shownin Figure 2. This was generated by 4 GeV electronsbut the pulses from hadrons and at different energiesare not significantly different. The width and shapeof this pulse comes from the convolution of a numberof sources:

a. The time spread of energy deposition by ashower including time of flight of slow pro­tons and neutrons,

b. scintillator phosphor rise and decay times,c. shifter rise and decay times,d. phototube response,e. time delays in the light collection from

different parts of the calorimeter and timedispersion in transmission.

The objective of the first phase of this studywas to isolate these separate contributions, estimatehow they could be speeded up and find what costs areinvolved. For this phase we have run Monte Carlo for(a), made measurements using single photon countingfor (b), and (c). Observe (d) with a fast scope,observe (e) and (f). From these observations we esti­mated that a fast calorimeter could be made withpulses whose full width half maximum would be of theorder of 7 nsec.

In the second phase we constructed an extremelycrude calorimeter whose pulses should have the samecharacteristic as in a real device. With this we haveobserved signals whose mean width was 7 nsec and whosewidth at 10% of maximum height was 15 nsec. Clippingcould reduce these widths to 6 and 12 nsec respective­ly. We conclude that gate times of less than 20 nsecwould be appropriate for such a calorimeter.

A third phase is unqerway to build a good testcalorimeter to confirm these results.

II. Discussion

a. Energy Deposition

A Monte Carlo program used and modified by AlanStevens has calculated the time distribution of energydeposition for both a copper and a uranium calori-

* Work performed under the auspices of the U.S.Department of Energy.

101

meter. The results in 5 nsec bins are shown below.For copper they are plotted in Fig. 3.

Time Interval Copper Uranium

0- 5 nsec 97.2 915-10 2.2 6.0

10-15 0.5 2.115-20 0.1 0.720-25 0.225-30 0.1

These results show a fast initial spike 2 nsec widefollowed by an exponentially falling slow componentwhose time constant is 4 nsec and whose total fractionis about 10% for copper and 20% for uranium.

b. Scintillator

ISR 807 uses an acrylic scintillator that con­tains among other ingredients Napthalene whose pub­lished3 decay time is 96 nsec. We have not measuredthe response of this scintillator without shifter butthe observed long tail on the shifted signal (Fig. 2)is presumably from this decay.

More conventional scintillators contain somecombination of fast UV scintillating chemicals such asPBD (emission max ~ 370"nm) and wavelength shifterssuch as POPOP that lower the emission frequency to amore convenient visible range ~ 430 mm. These scin­tillators are faster than Napthalene scintillators butslower than scintillators without the shifter. Suchscintillators (eg NEllI and Pilot U) emit in the UV(370 and 390 nm respectively) and have shorter attenu­ation lengths than conventional scintillator but theseare not a disadvantage when the light path to a sepa­rate shifter is short.

We have looked at the pulse distributions fromtwo fast UV emitting scintillators: NEllI and BC420(similar to Pilot U). Both give pulses less than 2nsec FWHM and less than 5 nsec at 10%. Publisheddata2 gives pulse distribution for NEllI as shown inFig. 4. We expect that polystyrene scintillatorsusing on the chemical PBD without shifters have simi­lar response at lower cost. The cost of scintillatorfor a hadron calorimeter using BC420 or Pilot U isabout 140 K$ per m2 in a 0.4 m2 lot. The cost forpolystyrene should be about half this.

c. Wavelength Shifter

Conventional calorimeters (such as that in theISR 807 experiment) use the shifter BBQ whose absorp­tion maximum at 430 mm matches many conventional scin­tillators. We measured (see Sec. 3) the response ofBBQ illuminated by NEllI and observed a F~1 of 12nsec and width at 10% of 30 nsec. We also measuredthe response of POPOP and BBOT each illuminated withNEllI and observed FWHM's of 6.4 and 4.9 nsec respec­tively and width at 10% of 12.5 and 17.2 nsec respec­tively (see Fig. 5). Clearly both BBOT and POPOP aremuch faster than BBQ, with BBOT preferred over POPOP.The absorption maxima for the POPOP and BBOT are at360 and 398 nm suggesting that good combinations wouldbe NEllI (or Polystyrene UV) and POPOP or Pilot U (orPolystyrene UV-390) and BBOT. The latter combinationis preferred because of the faster wavelength shifter

and also because the further UV emitting scintillatorsNEllI and Polystyrene UV have shorter attenuationlengths (8 cm and 16 cm respectively) compared withPilot U and Poly UV-390 (100 cm and 20 cm).

It hasnsec.and isis now

a FWHM of only about 1 nsec and at 10% about 3The tube has a rather standard 2" photocathodevery well suited to calorimeter use. Its costjust over 1000$ but would be less in quantity.

e. Light Collection Time Delays

The Table below gives the estimated or measuredcontributions (FWHM) to the expected output pulses ofthe ISR 807 calorimeter and a possible faster ver­sion.

The measured mean velocity of light in the wave­length shifter bars is 0.5 times the velocity oflight. The velocity of the shower itself is equal tothat velocity. As a result light from the back of theshower arrives at the phototube earlier than that fromthe front. If the calorimeter is 1 m deep this causesa time spread of 3 nsec.

Time dispersion arises because of the finiteaperture of the shifter bars. Light travelingstraight down the bar takes a time t = ~N/C (where Nis the refractive index), yet the measured mean timewas 2 ~/c: slower because of the finite angles ofrays in the shifters. For a shifter and light guideabout 4 ft long the resultant FWHM time spread isabout 3 nsec.

:)

Summary of Time Constantsf.

Published Measured (w/NE111)A

( fallabsorption emission time FWHM

A. nm A. nm nsec nsecShifter

BBQ 430 505 12POPOP 360 415 1.1 6.4BBOT 398 432 1.6 4.9

A "Conventional" ScintillatorA.

( rise fall '\emission A. time time FWHM Attenuation

nm nsec nsec nsec A. cmScintillator

NEllI 370 1.7 1.55 8Pilot U 391 0.5 1.36 1.97 100NE110 434 1 3.3 3.3 400

Table I

Table I summarizes the response time data2 - 3 forboth .scintillators and shifters discussed plus NE110for comparison.

Acrylic is the conventional base for the shift­ers. It is easy to handle but less radiation resis­tant than PVT. PVT is harder to handle, being inclin­ed to craze. Polystyrene could possibly be employedalthough most polystyrene has an unsatisfactorilyshort attenuation length.

807 Fast Cal.

a. E. deposition 2 2 , )b. Scintillator Acrylic ~ 15 Pilot U 1.5c. Shifter BBQ 12 BBOT 5d. Phototube standard 5 C31024 1e. Light Collection 5 5

Overall 20 7Observed 23 7

! )thethe0.3

Table II shows attenuation lengths measured byPenn and ISR groups. For the Penn measurementsbar was 1 cm x 2.5 cm. For the ISR it was 20 cm xcm. In both cases the shifters are in acrylic.

Table II

Numbers withothers measured.fast calorimeterbelow.

approximate sign were estimated, theThe observed overall value for the

was obtained from the test reported

d. Phototube

The concentration of shifter should in general beas high as possible consistent with a sufficientattenuation length. These measurements are criticallydependent on the surface condition of the bars andthus show some inconsistencies. The values shouldthus be taken as lower limits on the bulk attenuation.

The set-up used is shown in Fig. 7. The hole infront of the phototube #2 was adjusted to keep thecounting rate from tube #2 to less than 1% of tube #1.The signals from both tubes were taken to constantfraction discriminators. The tube #1 high voltage wasset so that the counting efficiency for a minimumionizing particle passing through the trigger scintil­lator was about 70%. The tube #2 high voltage was setto obtain a similar efficiency for single photons.The time difference distribution between tubes 1 and 2were obseved on a "QVT". The distributions obtainedare shown in Fig. 5 and reflect effects from the scin­tillator, shifter and dispersion down the shifter bar,but not from the phototubes except for the jitter inthe system (about 1 nsec).

Shifter Concentration

BBQ 80 mg/litre

POPOP 300 mg/litre100 mg/litre

30-40 mg/litre

BBOT 100 mg/litre

AttenuationPenn

97131

70

Length cmISR

152

114190157

a.

III. Measurements

Single Photon Counting Measurements )

Typical 12 stage phototubes have responses simi­lar to that measured6 for RCA 8850 and shown in Fig.6a. FWHM is about 5 nsec and width at 10% about 12nsec. The width comes primarily from variation intransit time in the dynodes. The RCA C31024 is a tubeemploying only 5 dynode stages. Gain of over 106 isachieved by using Gallium phosphide on all stages.The output from the tube is taken from a special co­axial pin further improving the high frequency re­sponse. A pulse from this tube is shown in Fig. 6b.

Since the scintillator and this jitter are smallcompared to the observed width we can interpret thiswidth as that of the shifter without correction.

b. Calorimeter Experiment

In order to check the estimates given in SectionII(f) prior to the construction of a full scale testwe set up a crude experiment. Sixteen lead bricks(each 3-1/2" x 7-1/2" x 1-1/2") were placed in a row

102

o

o

o

with 15 4" x 4" x 1/8" pieces of NEIll scintillatorbetween the bricks (see Fig. 8). Two shifter barswere tried: one 30" x 4" x 1/8" of POPOP the other24" x 8" x 1/8" of BBOT. Both were blackened at thefront end. A single RCA C31024 phototube was buttedup to the back end of the shifter without a lightguide, glue or grease. This device was placed in asomewhat broad (~ 6" diameter) 22 GeV 11- beam. Atrigger was formed from the coincidence of one small1/2" x 1/8" scintillator in front of the calorimeterand a second phototube looking at one of the 4" x 4" x1/8", 1/3 of the way down the row. The second triggerphototube level was set to fire on at least one mini­mum ionizing particle at that point in the shower andthus vetoed triggers from a background of low energytracks not in the beam direction.

1.

2.

3.

4.

References

O. Botner, et al., Nucl. Inst. & Methods 179, 45(1981); also Ibid 196, 303 (1982), and privatecommunication. ---

J. Pronko, et al., Nucl. Inst. & Methods, 163,227 (1979); F. Galligaris, et al., Ibid 171, 617(1980).

Catalogs: National Diagnostics, Somerville, NJ,USA; Nuclear Enterprises, Edinburge, Scotland.

W. Kononenko, W. Selove, and G.E. Theodosion,UPR96E, U. of Pennsylvania, internal report; RA­Ong, DD23/0G563, Brookhaven Lab internal report.

o

o

o

It must be noted that this "calorimeter" had across section that contains only a small fraction ofthe approximately 8" diameter of a hadron shower. Italso has limited sampling with a very high Z material.It is a very poor calorimeter. It should howeverprovide a reasonable estimate of the pulse length seenby one 4" x 4" tower in a real calorimeter of the samedepth.

The observed pulse height distribution is shownin Fig. 9 (for POPOP shifter). The cr for this distri­bution is 38% corresponding to 180% / IE. Under thecircumstances this is a surprisingly good energy reso­lution.

A typical pulse is shown in Fig. 9 (for BBOT).Considerable pulse to pulse variations were observed,as expected in a calorimeter that does not contain theshower. The Table below gives the widths for succes­sive pulses using both BBOT and POPOP.

Widths at half height nsec

5.

6.

Private communication, W. Willis and W. Wenzel,CERN, Geneva, Switzerland.

H.R. Kroll and D.E. Persyk, IEEE Trans. Nuc. Sci.NS19, 45 (1972); B. Leskovar and C.C. Lo, LBL6456 Lawrence Berkeley Lab internal report.

SHIFTER

Figure lOa shows the mean shapes that hopefullyapproximate the real shape that a larger calorimeterwould give. Figure lOb shows the mean shape from BBOTafter shaping by a 4 nsec 23% clip. This pulse hasonly 6 nsec full width at half maximum and only 12nsec at 10%.

The question was raised as to whether the signalfrom neutrons at larger distances from the 4" x 4"shower case might be much slower. A further experi­ment was performed in which the beam was directed at asecond stack of uninstrumented lead bricks set upalongside the instrumented stack. Observation of thesignals in this instrumented stack showed no indica­tion of signals arriving significantly (i.e., morethan 5 nsec) later than had been observed before,i.e., we saw no evidence for a significant slow signaloutside the core.

Pulse observed in ISR 807 calorimeter from4 GeV electron.

-

60 nsec

100 nsec806040

GATES USED 120 nsec

20

Schematic representation of calorimeterbeing considered.

o

PARTICLES

Figure 1.

Figure 2.

5,12,8,14,10,16,11,11; mean 11;8,9,6,5,7,5,7,10; mean 7.

POPOP:BBOT:

()

()

IV. Conclusion

We conclude that a scintillator with suitablychosen available components could give pulses withwidths comfortably less than 10 nsec at half heightand less than 20 nsec at 10% of height. Gates of 20nsec seem not unreasonable. Work is needed to confirmthis result with a full scale calorimeter and on elec-

,) tronics to digitize the signals. This is underway.

103

100

Figure 3.

C) I

I@sec

~4nsec

(a)

8 10 12 14 16 18

TIME nsec

Pulses obtained using (a) POPOP and (b)BBOT shifters together with NEll12scintillator •

Figure 5.

5~

o 2 4 6

>­t:~ 10%wI­Z

l­I<.')

::Jw>i=<!-lWn:

6 8 10 12 14 16 18 20 22 24TIME (n sec)

Monto Carlo time distribution of energydeposition from a 5 GeV negative pion inthe test calorimeter.

~__4:...:n.:....s~ec AT 10 %

2 nsec AT 50 %1-----

uwen 100::W0-

ZQI­eno0­Wo>­<90::WZW

W>~ 0.1--lw0::

..J<lU

to­0...>­01-l.&JU'>>z-l.&J1-0<l..Jl.&JD::

o

NEill

NO FILTER

1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0

TIME (ns)

Figure 4. Light pulse from NEllI.

-l«z<.')

iii-l«t)

a:l­t)W-lW

W>i=«-lwIX:

(a)

(b)

~sec

)

)

o 5

TIME (nsec)

10

Figure 6. Output pulses from (a) RCA C31024 and (b)RCA 8850.

104

7 n sec FWHM

()

o

o

()

()

SMALL HOLE

PHOTOTUBE#2

PHOTOTUBE#1

Figure 7. Arrangement for single photon measurementof scintillator and shift pulse length.

16 LEAD BRICKS TRIGGER p.T.

~SCIN\LATORS 0#2 [fiR PoT

D~D~D~---~----~D 0 T°JD DD D BEAM

~~RT. --------RCA C31024

16~EAD BRICKS (DUMMY CAL.)SHIFTER

(a)

(b)

o

o

10

10

20

20

TIME30 n sec

15 n sec AT 10 %

TIME30 nsec

o

u

,)

Figure 8. Calorimeter experiment to observe pulselength with fast scintillator shifter andphototube.

Figure 9. Typical signal from calorimeter experimentusing BBOT shifter.

105

Figure 10. Average pulse shapes from the calorimeterexperiment using BBOT (a) without and (b)with a 23% 4 nsec chip.

SCINTILLATION CHAMBER OPERATION OF CALORIMETERS FOR COLLIDING BEAM DETECTORS

Lawrence W. JonesDepartment of Physics

The University of MichiganAnn Arbor, Michigan 48109

(II) ,

optical path length from each scintillator edge to theobjective lens equal. The set of strip mirrors alsoserves to collapse the image to eliminate the spaceoccupied by the iron on the image plane, utilizingmore completely the photo cathode area of the imagedevice. As an example, the system used in Huggett'scosmic ray experiment of Ref. 1 is reproduced inFi g. 1.

)

)

')

100 em L.............. ...~_~

la)

C:::J GLASS[=_-_~ SCINfiLLATOR

~"'~ :,'.ECROM'''R .: .. .•. .

PM~

'vII,lAGE •

, I'NT(I;-)Inr~

[ffq~l~~:O: S:~b:~ . M'RROH SySTEM

OBJECTIVE. ro, ' 1183 cr» ~;I"I; I.- IMAGE ItHE"NS1FIER

100 <m .-~.~ D RECORO"I" CAM'"

FIG. 1. Apparatus for photographing stereoscopicallyhigh energy cascades in a calorimeter. '(a) side view,(b) plan view. Taken from Ref. 1. .

A scintillator of reasonable qual ity with 2 MeVper centimeter of energy loss by a minimum ionizingparticle produces about 2x 104 blue photons withone- si xth of these bei ng transmitted to each of thesix surfaces of the scintillator. Consider an imageintensifier tube with a cathode such that the imagefrom the calorimeter must be demagnified by a factorof 20. Let us also assume an objective lens with anf-stop of 1.5. The fraction of the light this lenswill subtend from the scintillator is approximately1/900 of the light coming out of the scintillator fromthat surface. The photocathode efficiency of theimage intensifier tube may be about 20 percent. Forthis numerical example, approximately two-thirds of aphotoelectron would be recorded per centimeter ofscintillator per minimum ionizing particle in thescintillator, and this independent of the depth of thetrack in the scintillator. This is probably anoptimistic figure as there would be at least a factorof two lost in transmission through the optics and inreflection losses. Thus, let us take one photo­electron per three centimeters (or 6 MeV of energyloss) in scintillator. If the calorimeter is madewith four centimeters of iron for each centimeter ofscintillator, however it is divided, then there willbe about 50 MeV of energy loss in iron per centimeter

A Prototype System

Concept

Using photoelectric image intensification, lightfrom a scintillator can be recorded preserving spatialinformation of the source of light in thescintillator. In the late 50's the scintillationchamber or luminescent chamber was developed whereintracks and details of nuclear interactions werephotographically recorded from sodium iodide and fromplastic scintillators, using photoelectric imagei ntensifi ers and cameras. Si nce those years the imageintensifier has evolved a great deal and now findsapplication routinely in astronomical research,although the scintillation chamber was totallysupplanted by spark chambers (and later proportionalchambers and other digital readout devices) in highenergy and elementary particle physics. Oneapplication of the scintillation chamber which isparticularly relevant to the present discussion wasmade by R. W. Huggett and colleagues in a c~smic rayexperiment carried out at Climax, Colorado. In thisapplication they had a calorimeter measuring 5m x 1.8mx .9m containing 20 scintillator sheets and thiscalorimeter was photographically recorded using animage intensifier and photographic film. The concepthere is that a sheet of scintillator will pipe thelight to its four narrrow edges by total internalreflection. If the edges are cut straight then lightwhich emerges from those edges maintains the spatialinformation of the coordinate of the source of lightin the two dimensions of the scintillator sheet area.A sampling calorimeter containing alternating absorberand scintillator can be viewed from one or two of itsside faces using a system of mirrors and a lens toimage the light onto an image intensifier and theimage will then preserve the profile of the cascade ofa hadronic or electromagnetic shower within thecalorimeter.

Summary. It is suggested that the scintillationchamber, a technique first discussed almost thirtyyears ago, might find application in colliding beamdetector systems, in particular as a means ofefficiently extracting detailed spatial and energyinformation from a sampling calorimeter.

Consider a downstream or endplug calorimeter withtransverse dimensions of about one or two meters andextending along the beam pipe one or two meters,consisting of iron plates of the order of one inchthick and plastic scintillation sheets of the order ofone quarter of one inch thick. It is desired to viewthe plastic scintillator in order to resolve separatehadronic cascades and to determine the separatecascade energies with best possible spatialresolution. If photomultipliers and wave bars andtowers are used inevitable compromises must be made.However, the image intensifier system might achievethe best possible separation and resolution of such asystem. Strip mirrors could be mounted beside eachscintillator and a second set of strip mirrors used tocollect the light onto one or more large area mirrorsto direct in turn through the objective lens onto thecathode of an image intensifing photoelectric tube. Atransfer lens or fiber optics might then bring thelight from the image intensifier onto a CGD or similardigital readout device of lesser senitivity. The twosets of strip mirrors might be necessary to make the

106

o

o

o

o

o

o

o

u

,)

,J

. )

of scintillator and consequently one photoelectronwould be recorded per 150 MeV of energy loss in thecalorimeter. The resolution of such a calorimeterbased only on photoelectrons statistics would betherefore ahout 40/1E percent, with E measured in GeV.The resolution would, of course, be improved with afaster lens or less demagnification and/or more imageintensifier systems. At the very least, one would usetwo image intensifier systems perhaps looking at thesame system of mirrors but from different corners, inorder to provide a stereo image. Image tubes could beplaced on more than one side of a rectangularcalorimeter, but this may not be necessary.

Comments

It is amusing to remark that a scintillationchamber detector was discussed in the context of onethe early paperszon the experimental utilization ofcolliding beams. The concept implemented by theLouisiana state group, however, is almost identicallythat which would be most appropriate for the collidingbeam detectors under current discussion. The timeresolution of an image intensifier-scintillationchamber system in the applications of the late 50'swas determined by the storage time of the phosphor ofthe first stage of the image intensifier. One can useany number of phosphors with storage times of from amillisecond to much less than a microsecond, howeverthe convenient application of the image intensifierscintillation chamber was made possible by gating asecond or later stage of an image intensifier systemusing the first stage phosphor as a buffer storage, soto speak. A continuously-on image intensifier coulduse the properties of the CCD or other imaging systemin the electronic readout to serve as a buffer store,retaining only information from an event of particularinterest. In any case, auxillary triggering, forexample from photomultipliers viewing the samescintillators, would probably be necessary, and thetime incumbent in this auxillary triggering would bethe determining factor in the ultimate time resolutionof the system. Greater time resolution might bepossible by using image intensifier systems in pairs,one recording a static image as described here, thesecond utilizing an image intensifier with sweepelectrodes incorporated such that the images would beswept across the recording field at a convenient rate.T~is would produce a disp,lacement of one image in thestatic system relative to the other in the sweptsystem related to the relative time delay and thesweep rate. In this way, a time resolution might beachieved comparable to the intrinsic time constant ofthe scintillation process in the plastic scintillator.It is clear that the topology of the scintillatoroptics image-tube systems could become quite tediousfor a central detector, but perhaps not insoluble.Except for such details it seems surprising and indeedamusing that this detector teChnique, dormant for overtwenty years, might find application in the newest andmost ambitious detector systems of the next decade.

This work was supported by the U.S. NationalScience Foundation.

References

lR. W. Huggett, T. Sander, and J. E. Webb "ImageIntensifier Photography of Ultrahigh Energy Cascadesin an Ionization Spectrometer. 1. Apparatus," Rev. Sci.Jnstr. 41, 1007 (1970).

L. W. 'JOnes "The Luminescent Chamber and Other NewDetectors," Proceedings of the Intern. Conf. OnHigh-Energy Accelerator and Instrumentation, 561 CERN(1959); L. W. Jones "Recent U.S. Work on CollidingBeams," Intern. Conf. on Hi gh Energy Accel era tors ,Dubna, USSR (1963) •

107

HIGH ENERY ELECTROMAGNETIC SHOWERPOSITION MEASUREMENT BY A FINEGRAINED SCINTILLATION HODOSCOPE

B. Cox, G. Hale, P.O. Mazur, R.L. Wagner,D.E. WagonerFermi National Accelerator Laboratory, Batavia, Illinois 60510

H. Areti, S. Conetti, P. Lebrun, T. RyanMcGill University, Montreal, PQ, H3A 2T8, Canada

R.A. GearhartStanford Linear Accelerator Center, Stanford, California 94305

Summary

')

We have measured the centroids of high energyelectromagnetic showers initiated by positrons in theenergy range 2 to 17.5 GeV with a fine grainedscintillation hodoscope composed of seven 1 cm wideelements placed behind a 3.6 radiation length (15 cm)converter composed of SCG1-C scintillation glass. Asimple first moment calculation using the ionizationobserved in each element of this hodoscope yields ashower position resolution as a function of energy of:a(mm) ~ 0.7 ± 5.6/IE(GeV). We present results on theenergy dependence of the shower profiles and theionization measured by this hodoscope.

Introduction

The determination of the position of a photon bymeasurement of the profile of, its shower in anelectromagnetic shower detector is a criticalrequirement of any experiment which attempts toreconstruct photon energies and directions. Inpreparation for Tevatron Experiment 705 1 at Fermilabwe have tested a fine grained (1 cm wide elements)scintillation hodoscope positioned 3.6 radiationlength deep in electromagnetic showers initiated by 2to 17.5 GeV positrons. We have determined theposition resolution attainable by this device as afunction of energy. In addition, we have measured theshower shapes and ionization seen by the hodoscope asa function of energy. Comparisons of the measuredshower shapes with the EGS electromagnetic showerMonte Carlo2 have been performed.

Test Beam----

FINGER HODOSCOPE

TEST APPARATUS

ACTIVE' CONVERTER(SCGI-C)

~ SF5

e+_~ SF5BEAM ~

SF5

PLAN VIEW

fIGURE I

RCA 0055

')

.)

)

This test was performed using the positron testbeam 19 in the B end station at Stanford LinearAccelerator Center. The small phase space of the beam(90% of the positrons included within a a ~ 1 mm spot)allowed us to make a relatively precise determinationof the position resolution of the scintillationhodoscope. The momentum of the beam was tunable from2 to 17.5 GeV and the contamination of the beam byhadrons and muons was less than 10 -S. The beam wastypically operated at 10 pulses per second with lessthan 0.3 positrons per 1.6 ~sec pulse. Beam pulseswith two or more positrons were tagged and laterrejected in the off-line analysis.

Experimental Apparatus

The arrangement of the scintillation hodoscope,active converter and SCG1-C scintillation glass/leadglass array3 used in this test is shown in Fig. 1.The incident positron beam showered in a 15 cm thick(3.6 radiation lengths) SCG1-C scintillation glassactive converter. 3 A finger hodoscope composed ofseven 1 x 1 x 15 cm , NEl14 scintillators coupled to12 stage, 2 inch diameter EMI 9807B phototubes wasplaced 3 cm downstream of the active converter. Pulseheights from these counters were digitized using aLeCroy 2249W ADO. The main shower detector array,

which was composed of a central 15 x 15 x 80 cm3

SCG1-C scintillation gla~s shower counter surroundedby eight 15 x 15 x 45 cm SF5 lead glass counters, waspositioned approximately 4 cm downstream of the fingerhodoscope. The entire apparatus rested on a tablewhich could be positioned vertically or horizontallyto ± 1 mm.

Test Results

With the active converter removed, the positronbeam was scanned horizontally through each of theelements of the hodoscope to locate the hodoscopeedges and to determine the pulse height of minimumionizing particles in each of the finger hodoscopeelements. With the minimum ionizing pulse heightsdetermined, all results could be expressed in units ofthe average pulse height for a minimum ionizingparticle. After normalization of the gain of eachcounter, a 17.5 GeV positron beam was centered in thecentral finger hodoscope element (F4). Effects ofbackscatter from the main array were then studied.Less than 5% of the showers had greater than one unitof minimum ionization pulse height in either of theadjacent counters (F3 and F5). This did notsignificantly change when 2.5 cm of polyethylene

108

r ,

F7F6

o 2.0 GeV

G 17.5 GcV

6 10.0 ~eV

+ 6.0 GeV

EGS PREDICTION17.5 GcV

F3 F4 F5

FINGER COUNTER

FIGURE 4

F2

o

+

6(>

IONIZATION DISTRIBUTIONFINGER COUNTERS

17.5 GeV IONIZATIONDISTRIBUTION

0 t<>I- 0oft::>6 0+,

g+ • fI!' ~i\0

2 3 4 5 6 7FINGER COUNTER FIGURE 3

.1

.1

.6

.6

zo~!::! .5zQ0:w~.4:cVI'

.J

~~ .3Ii.oz~.2~0:U.

We have performed a first moment calculation ofthe position of the positron for each shower using theformula

the 17.5 GeV shower i~nization distribution togetherwith an EGS Monte Carlo prediction of the ionizationthat the finger hodoscope should detect. As shown thedata is consistent with the EGS prediction. The barson the 17.5 GeV/c data once again are the width of theionization distributions seen in the various fingercounters.

~IGURE 2b

2 4 G 8 10 .12 14 16 18 20E (GeV)

c:;- 55zN 50ZQ 4·z:E 40u.0(I) 35t:z 303z0

~NZQ.Jgt-:-

I,!)zNZQz~

,~

(I).1-Z3zoEit::!zQ<tIi.

absorber was inserted between the main array and thefinger hodoscope.

FIGURE 20

The energy of the beam was then varied from 2 to17.5 GeV. The total ionization observed in thehodoscope is shown in Fig. 2a as a function of energy.The bars on the data points in this figure are noterrors in the determination of the centroid of theobserved ionization distributions but rather the fullwidth at half maximum of these distributions. Figure2b shows the ionization in the central counter, F4, asa function of energy.

18 20

Figure 3 shows the shower profile as seen by thishodoscope for 2.0, 6.0, 10.0 and 17.5 GeV showers. Asshown, the shower profiles at a depth of 3.6 radiationlengths are narrow with full widths of less than onecentimeter. The shape of the shower was slowlychanging with energy. This can be seen in Fig. 3 inthe slight tendency of the 2 GeV showers to be lesscollimated than the 17.5 GeV showers. Figure 4 shows

The active converter was then inserted in frontof the finger hodoscope. In this configuration wesearched for possible effects of a wide angle, lowenergy electron component of the 17.5 GeV showers byexamining the average pulse heights seen in F1 and F7(the finger counters most distant from the showercenter) as a function of polyethylene thicknessinserted between the active converter and thehodoscope. The average pulse height for a 17.5 GeVshower in either extreme counter was approximately 1.5times minimum ionizing. In comparison, pulse heightsaveraging approximately 25 times minimum ionizing wereobserved in the central counter, F4. These resultswere essentially independent of polyethylene absorberthickness.

o

o

o

o

o

o

()

()

,J

109

where P is the pulse height in the i th finger counterand Xi is the position of the center of the counter.The aistribution of <X> for a set of 2 GeV and 17.5GeV showers is "shown in Fig. 5a and 5b•

)

)

CY(mm)= .7+Ji"

FINGER HODOSCOPEPOSITION DETERMINATION

(leM HODOSCOPE-lsT MOMENT)

References

M. Binkley, et al., Fermilab Proposal E-705, 1981.R.L. Ford and W.R. Nelson, SLAC-210 (1978).B. Cox, et al., A Measurement of the Response ofan SCG1-C Scintillation Glass Shower Detector to2-17.5 GeV Positrons, Paper 2H7, 1982 NuclearScience Symposium, Washington, D.C.

1.5

> ~ > ~QJ QJC> C> C> C>10' a q qr::: Q lD C\l

~ t ~ ~.5

.1 .2 .3 .4 .5 .6 .7 .8I (GeV'!)

JEBEAM FIGURE 6

4.0

3.0

3.5

4.5

1 •2.3.

We wish to thank Fermilab Physics Department andto acknowledge the considerable efforts of the SLACoperating staff. We would also like to thank theU.S. Department of Energy, the Natural Sciences andEngineering Research Council of Canada and the QuebecDepartment of Education for their support.

Acknowledgments

material even if the material has a relatively longradiation length (rt = 4.35 cm in the case of theSCG1-C scintillation glass). The position resolutionwhich contains the finite size (a '" 1 mm) of thepositron beam position is observed to be a (mm) = 0.7+ 5.6/1E(GeV). The shower shape is less than 1 em fullwidth at this depth and is slowly varying with energy.The ionization observed in this hodoscope varieslinearly with energy with half the ionizationappearing in the central element. At 17.5 GeV a totalpulse height of approximately 50 times minimumionizing is observed.

E 2.5E~

b2.0

105·10

._._.__ .._---250 FIGURE 50

17.5 GeVe+SHOWERS

a'=2.12mm200 FIRST MOMENT

DISTRIBUTION

(/)....15zw

>w~II: 100l~

co:z:>z

50

100

5002.0 GtV e+SHOWERS

40

·5 0X lmml

The a of the distributions of first moment isapproximately linear with 1/IE(GeV) as shown inFig. 6. These data are approximately described by a(mm) = 0.7 + 5.6/IE: This a contains the a '" 1 mmphase space of the incident positron beam whichcontributes to the energy independent term. Theerrors assigned to these data points are not thestatistical error in the a of the distributions offirst moments but rather the spread in the a's ofthese distributions when repeated measurements of thepositron showers are done under different experimentalconditions.

Conclusions

A fine grained (1 cm) scintillation hodoscope ofthe type described can be used to determine withprecision the centroid of electromagnetic showers whenplaced behind 3 to 4 radiation lengths of converting

110

SCINTILLATORS AND WAVESHIFTERS FORFAST SCINTILLATOR-WSB CALORIMETRY

George E. rheodos iOLi

Department of PhysicsUniversity of Pennsylvania

Philadelphia, PA 19104

o

o

AbstractWe dis cuss some properities of fast sci nti llators

and waveshifters, and their use and limitations incalorimetry using the wavelength shifter bar(WSB) read­out technique.

Introducti on

The BBQ WSB readout technique in scintillatorcalorimetry, developed recentlyl,2, has been up to nowthe dominant calorimetry method for experiments at highenergies. The high event rates in future highluminosity Colliders necessitate the development ofcalorimetry with scintillators and waveshifters fasterthan BBQ. We have carried out, more than two years ago,extensive studies 3 on the light output and attenuationlength of several inexpensive acrylic and polystyrene(fast) scintillators 4 coupled with fast waveshifters(POPOP, BBOT) with very encouraging results. We brieflypresent some of these results and discuss theirimplications on fast calorimetry.

Measurements

Figure 1 shows a test set up with a RU 106 source,the scintillator and waveshifter bar test samples andan 8850 Quantacon photomuliplier.

Npe

1.0

X sc i=7.5cm

SCINTI LLATOR =POLY UViSl (POLYSTYRENE)

{

' 100 mg/~ POPOPWSB = ° 100 mg/.D. BBOT

+ 30 mg/.D. POPOP

.0.10 20 40 60 80 100 120 140 160 180

YWSS(cm)

Fi~ure 2

Figure 3 shows the same thing, but for an acrylicscintillator sample, ACR3, which is considerablyslower than polystyrene4 .

oEND

BLACK

~ XSC1 ---'! Ru ,06

r~ JI SCINTILLATOR ~AR

¥\ ENDTRIGGER COUNTER BLACK

CDVl

.;!

J - OPAQUE MASK

~o

1.0

~t~~'''-.~0"'.,~""-

"'b'"-.,

X."75~(

'2% NOPh:h) ,SCINTILLATOR =ACR3 .:% B-pao

{

' 100 mg/.D. POPOPWSB = ° IOOmg/.D. BBOT

. + 30 mg/.D. POPOP

,,-)

Figure 1

Figure 2 shows the number of photoelectrons plotted(Npe) as a function of light travel distance in the WSB(Ywsb) for a fixed position of the source relative tothe scintiHator(Xsci)for a 6.3 mm thick polystyrenescintillator sample, Sl, and several kinds of 10mmthick WSB's (1,2,3): UVT acrylic doped with differentkinds of wavelength shifters of varying concentrations.

111

0.1 L---'---'_-'---'-_'----'---'_-'-~o 20 40 60 80 100 120 140 160 180

YWSS (cm)

Fi gure 3

A much more detailed and complete account for thetests and the results is given elsewhere 3

•

Discussion of Results

The table below gives Npeo, the number of photoelectronsextrapolated to Ywsb=O and the attenuation length(Awsb)for each scintillator + wsb combination as well as thecomposition of each of the materials.

Scintillator Waveshifter Npeo Awsb(cm)(Xscic7 . 5cm)

Poly UV WSB1 5.7 131WSB2 5.5 101WSB3 5.0 71

ACR3 WSB1 4.1 131WSB2 4.3 93WSB3 3.5 68

Poly UV = 99% Polystyrene + 1% B-PBDACR3 = 87% PMMA + 12% Naphthalene + 1% B-PBDWSB1 = P~4A + .003% POPOPWSB2 = PMMA + .01% POPOPWSB3 = PMMA + .01% BBOT

We note that the attenuation length increases as thewaveshifter concentration decreases so thatwith POPOPconcentrations of .001% - .002% one might be able toobtain attenuation lengths of 1.5m or more, Thepolystyrene scintillator is best in the number ofphotoelectrons (pe) produced. With a 1 cm thickscintillator sample and at a distance of Xsci = 5cm,from our data we can calculate that one should expectabout 10 pe(Npeo)instead of 5.7.

Materials and Costs

Cast acrylic and extruded polystyrene scintillatorshave been routinely produced in large quantities inrecent years for several experiments at FNAL and CERNand they are relatively cheap. Their cost is severaltimes lower than that of PVT aromatic scintillators.

Calorimetry Considerations

The large signal levels and attenuation lengthsmeasured by us with the polystyrene and POPOP combina­tion make its use for fast calorimetry quite feasible.A 100 cm2 area, 10 gap sampling calorimeter module·with the proper polystyrene + POPOP combination, readout two meters away could produce as many as 40-60p.e. per minimum ionizing particle.

The feasibility of fast calorimetry has also beeninvestigated recently in tests at BNL by R.B. Palmer,where calorimeter pulses shorter than 10 nsec havebeen obtained with a small prototype sampling calor­imeter module 5 •

The WSB readout technique has however some limit­ations pertinent to experiments with high luminosities:The difficulty to obtain very fine granularity, thehighlight attenuation with distance, of fast wave­shifters, and its sensitivity to radiation damage. 4 6However, all three limitations can be greatly reducedif this technique is used for the outer parts of acalorimeter detector surrounding an interaction regionand vacuum photodiodes (VPD) are used for read out 7for the inner part of the calorimeter, located closerto the beams and the interaction point.

This combination is quite attractive at present,since VPD readout is conslderablymore expensive thanWSB readout and the volume of a spherical-like detectorgrows as R3 (R = distance from the interaction point).

Acknowledgements

The tests were planned and performed withW. Kononenko and W. Selove. We are indepted toB. Musgrave and L. Nodulman for providing to us theKSH polystyrene samples, to V. Fischer and U. Widhamof Roehm GmbH Chemische Fabrik for the WSB samplesand to DOE for support. We would like to thank theWorkshop Organizers for providing the appropriateclimate for exploration of such ideas. The exchangeof important information and ideas with R.B. Palmer,C. Baltay and R. Gustafson is gratefully acknowledged.

References

1. G. Keil, N.I.M. 87(1970)111 and references therein.

2. W. Selove et.al, NIM 161(1979)233; B. Barrish et.al,Preprint, CALT68-623.

3. "Scintillator-Waveshifter Light Measurements".W. Kononenko, W. Selove and G.E. Theodosiou,University of Pennsylvania Report UPR-0096E,March 9, 1982. To be published in NIM.

4. C. Aurouet et.al., NIM 169(1980)57; J.C. Theveninet.al, NIM 169(1980)53.

5. R. B. Palmer, Proceedings of this Workshop.

6. UA1Collaboration, private communication.

7. "Scintillator Calorimetry with vacuum photodiodereadout" by W. Selove and G. Theodosiou, Proceedingsof this Workshop.

:)

( )

112

()

SCINTILLATOR CALORIMETRY WITH VACUUM PHOTODIODE READOUT

o

W. Selove and G.E. TheodosiouDepartment of Physics

University of PennsylvaniaPhiladelphia, Pennsylvania

AbstractPROTOTYPE SPED e-m CALORIMETER MODULE

MUON PULSE HEIGHT DISTRIBUTION

19 LEAD PLATES OF 6 mm '"~---19SCINTILLATORS OF 10mm--------.

114 PHOTODIODES (6)( 19)66 em23 RADIATION LENGTHS

Fig. 2

Introduction

Measurements

Vacuum Photodiodes (VPD) are discussed as a newreadout technique for fast scintillator samplingcalorimeters.

The feasibility of this technique has beendemonstrated with light yield mfasurements usingpolystyrene scintillators and diodes and measurementswith a 2 prototype scintillator sampling e-mcalorimeter. Some of its properties and results arepresented here. The amplifier noise levels and thepossibility of the use of fast signals are brieflydiscussed as well as the sensitivity of the techniqueto radiation damage effects. Finally, the numerousvirtues of the technique are reviewed pointing to theurgency of rapid manufacturing development of low costVPD's.

o

o

o

SPb PED

80

80

I' SIGNALI•

1,,,.""..""""~u"II"XIlItIC

lOll.UU....uuuu

lunxxx..lUlU1lIt1OlAli.tIl .....cuux

).ItI.UUII'tJlhllUItll

'1 I"

PEDESTALI

Fig. 3

Fig. 4

20 GeV ELECTRONS

20 40 60CHANNEL NUMBER

20 40 60

CHANNEL NUMBER

-l

~VIWoWGo

Z 1000i=::>

80mii: Ql

1;;"8 60

~~:X:c 40e>-w&:x: 20wen-l::> 0a. 0

100

200

o

COUNTSKA

,- ~2cm Typical

~

ONE CELL

Fig. 1

PARTICLEDIRECTION

Figure 1 illustrates the basic principle of thetechnique, called SPED, where thin, large area vacuumphotodiodes(PED) are used to collect light produced inthe scintillator(S) in each cell of a samplingcalorimeter. The diode signals of several layers areadded together and fed into an amplifier/shaper forreadout.

Figure 2 shows an e-m calorimeter prototype,which was used to measure muons and 20-38 GeVelectrons at FERMILAB. The apparatus description,tests 2and measurements are given elsewhere, indetail.

Figure 3 shows, for this prototype, muon andpedestal pulse-height distributions with a ratio ofSignal/Noise=2.4 and Fig. 4 a 20 GeV electronpulse-height distribution, with a pulse rise time oft r , e = 150 nsec and a 1 ~sec gate, consistent with anen~~gy resolution of alE = 0.22/1El (1 r. 1. Pbplates). 113

o

o

,)

,)

The muon signal measured in this prototype wasabout 20000 photoelectrons (pe) and corresponds to a0.2 GeV signal equivalent energy loss or lOOk pe/Gev.Then the rms noise is 85 MeV or 30 MeV per 10 diodesper amplifier. From these measurements we estimatethat, with properly designed VPD's (1 cm anode-cathodeseparation) and amplifier electronics, a 25 cmsquared, 10 gap, calorimeter module, read out by oneamplifier, will contribute a noise of 15 MeV permodule, or 30-40 MeV per tower consisting of 4 suchmodules. Therefore the noise contribution to e-m orhadronic showers of a few GeV energy is negligible.In order to be able to handle the high ev~~t r~~es Ofa high luminosity Collider, i.e., L = 10 cm sec­(DC mode), it would be necessary to use pulse risetimes shorter than the 150 nsec used here. The VPDreadout technique seems quite promising in thisrespect, if very fast pulse shaping is used and thesignals are large. If faster si~nals are used, thenoise increases as l/lt. • With optimization ofparameters it appears feasrB~~ to go to rise timesmuch shorter than 150 nsec and still have reasonablysmall noise levels for high energy hadrons or jets,for example 50 GeV and higher.

Radiation Damage Effects

Measurements made with polystyrene scintillatorsby J.C. Thevenin et al.,4 show that the amount oflight loss5due to radiation damage is less than 2% fora 2 x 10 rad dose for light travel distances of theorder of 1 cm, but increases rapidly as the lighttravel distance increases (20% for 20cmtravel). Thisindicates that the effect of radiation damage on thesignal can be expected to be considerably smaller withthe use of the VPD readout technique than with WSBreadout.

Merits

The VPD readout technique, applied to fastscintillator sampling calorimetry as well as to highresolution "continuous media" e-m calorimetry, is ofhigh potential at high luminosity machines, mainlybecause:

We also note a very important development needfor thin, large area vacuum photo-"triodes" with again of 10 or so. With such devices, virtually allamplifier noise problems would become negligible.

Acknowledgments

·This work has been carried out all along withW. Kononenko. We would like to thank B. Robinson forexchange of important ideas, R. Van Berg for work onthe prototype design and the electronics and R. Rulandfor help with the diode tests•. We are indebted toJ. McCormick and T. Hayashi of the Hamamatsu Co. forthe VPD development, to R. Boie and V. Radeka foressential discussions and for providing the amplifiersand to DOE for vital VPD development support.

References

1) W. Kononenko et al." NIM 186(1981) 585-591.2) IEEE Trans. on Nucl. Sci., Vol. NS-30, No.1,

Feb. 1, 1983; Univ. of Pennsylvania ReportUPR-l01E, to be published in NIM.

3) V. Radeka, private communication.4) J.C. Thevenin et al., NIM 169 (1980) 53-56.

)

)

)

! )

Calorimeter pulses can be read out very fast.Pulse time resolution can be very good, of theorder of 100-200 psec (rms).Radiation damage effect on light yield isexpected to be minimal.High segmentation capability.High gain stability.

)

Other important properties, not directly related tohigh luminosities, include

6) Readout compactness.7) Low sensitivity to magnetic fields.8) Good energy resolution.9) No high voltages.

10) Simplified calibration/monitoring.11) Relatively simple construction and operation.

Cost and Development

At the present time, VPD's with the desiredproperties can be available at $600 per sq.ft. forquantities of the order of 10k sq.ft. This cost mustgo down to $150-200 in order for the overallcalorimeter cost to be very competitive with that ofgas calorimeters.

114

\.j

o

A MEASUREMENT OF THE .RESPONSEOF .AN SCG1-C SCINTILLATION GLASS

SHOWER DETECTOR TO 2-17.5 GeV POSITRONS

B. Cox, G. Hale, P.O. Mazur, R.L. Wagner, D.E. WagonerFermi National Accelerator Laboratory, Batavia, Illinois 60510

H. Areti, S. Conetti, P. Lebrun, T. RyanMcGill University, Montreal, PQ, H3A 2T8, Canada

J.E. Brau,t R.A. GearhartStanford Linear Accelerator Center, Stanford, California 94305

o

o

I,, J

o

()

Summary

We have measured the response of anelectromagnetic shower counter constructed from thenew scintillation glass (SCG1-C, Ohara Optical Glass,Inc.) to positrons in the energy range 2 to 17.5 GeV.We have measured the energy resolution of this 18.4radiation length detector plus its attendant SF5 leadglass shower counter array to be alE = (1.64 ± 0.14)%+ (1.13 ± 0.33)%/1E with the constant term dominatedby variations in the conversion point of the positronand shower leakage. We found this counter to belinear over the energy range examined. We have alsomeasured the light output of the SCG1-C counterrelative to light output of the SF5 lead glass guardblocks using 17.5 GeV positrons. We find that theSCG1-C counter produces 5.10 ± 0.30 more light at thephoto tube than the SF5 lead glass counters.

Introduction

A new type of high density glass (SCG1-C fromOhara 0ft~cal Glass, Inc.) has recently beendeveloped' which can be used in electromagneticshower counters for high energy physics. This glassdiffers in two major ways from the various types oflead glasses which have been used for over a decade inhigh energy physics in two major ways. This glass hasbarium oxide rather than lead oxide as the high Zmaterial and contains Ce20~ which acts both as ascintillator and a wavel~ngth shifter for Cerenkovradiation. The blue and near ultraviolet Cerenkovlight is absorbed and reemitted at longer wavelengthswhich survive absorption by the glass. The observednet gain, relative to that possible with lead glass,in the number of photons which reach thephotomultiplier results in an improvement in the partof the energy resolution due to photon statistics.

This scintillation g3~gs, along with earlierversions, has been tested with low energy photonsand electrons « 300 MeV). We have constructed anelectromagnetic shower counter which is suitable forhigh energy shower measurements from this new glassand have tested it using high energy (2 to 17.5 GeV)positrons at the Stanford Linear Accelerator Center(SLAC).

Beam

This test was performed using the positron testbeam 19 in the B end station at SLAC. This positronbeam had a momentum bite, Ap/p (FWHM) = 0.25%, and asmall phase space with 90% of the beam particlescontained within a beam spot of radius 1 mm. Thecontaminatio~ of this beam by pions and muons wasless than 10 over the entire momentum range. Themomentum of this beam was tuneable from 2 to 17.5GeV/c. We typically operated the beam at 10 pulses

per second with an average of less than 0.3 positronsper 1.6 ~s pulse. Beam pulses with two or morepositrons were tagged and later rejected in theoff-line analysis.

Experimental Apparatus

The arrangement of shower counters used for thistest is shown in Fig. 1.

SCG1-C SCINTILLATION GLASSTEST ARRAY

.RCA B055 PHOTOTUBES7

FIGURE

The array consisted of a central counter composed ofthe scintillation glass, SCG1-C, surrounded by eightSF5 lead glass guard counters to capture thetransverse leakage of the high energy showers. TheSCG1-C counter was 80 cm in length (18.4 radiationlengths) and 15 cm x 15 em in cross section and hadthe composition shown in Table I.

Table I

Composition of SCG1-C Scintillator Glass2

(by weight)

43.4%42~5%4.0%3.3%3.3%2.0%1.5%

The heavy compound which is the major contribution tothe 4.35 em radiation length is BaO and thescintillating, wavelength shifting component is Ce20~.

This particular counter, constructed from two 40 ~m

t Present address: Department of Astronomy and Physics, Univ. of Tennessee, Knoxville, Tennessee 37916.

115

Experimental Results

The pulse shape obtained from the SCG1-C counterresulting from a 17.5 GeV positron,shower is shown inFig. 2 along with the counter's response to a 40 nspulse from a green LED (Monsanto MV5252).

t. (n sec). 50 100 150 200250 300 350 )

()

theto

2.0 GrN e+

j~ CljkPik)

jri

minimization leads

N

rk=1

The resolution and linearity of the SCG1-Ccounter and its accompanying guard counter array weredetermined by positioning the positron beam at thecenter of the SCG1-C counter (to ± 1 mm) and recordinga few thousand showers at energy settings of 2,4,10and 17.5 GeV. Pulse height spectra for 2 and 17.5 GeVenergies obtained by summing the pulse heights abovepedestals in the array are shown in Fig. 3a and 3b.

(/)

!z 150

~lL.o0: 100'w.~

~

200

250"""''--'''''-''--'''---'---'':'---''-''''''''

50

Using this formula and a data sampletRf N showers witha 4.0 GeV beam centered in the i counter, the Cicalibration constant was calculated. This C. thenreplaced tfihe initial (or the current iteration} valuefor the i counter for the next iteration through thearray. Using this technique, the calibrationconstants for the array stabilized after threeiterations.

. wi th respect to Ci ~ Theformula for Ci :

Nr (EPik ­k=1

GREEN LED

SCGI-C SCINTILLATION GLASSPULSE SHAPES

(RCA 8055 PHOTOTUBE)

5

20

4

30

90

100

!C. 6ClW:cw~ 80::30.

10

pieces, was viewed toward the incoming positron beamby an RCA 8055 photomultiplier. Epotek 305 epoxy wasused for the glass-glass joint and for the tube-glassjoint in this counter. The eight SF5 guard blockswere 45 cm in length (18.0 radiation lengths) and 15cm x 15 cm in cross section. These counters were alsoviewed end on by RCA 8055 tubes. The shorter guardblocks were positioned so that their upstream endswere 25 cm from the upstream end of the SCG1-Ccounter. This positioning was chosen for the maximumcontainment of the transverse shower leakage from theSCG1-C counter. Each of the nine counters of thearray had a red LED (Monsanto MV10B) mounted on theend of the counter opposite to the phototube forpurposes of gain monitoring.

FIGIJRE 2

1.64 1.88 1.92 I.S6 2.00 lW4 2.08 2.12EIIoEASUIlED (GW) .FIGURE 30

. 250r--,,---.--r-..,...-..,....-,....~,....-,

The time response of the RCA 8055 tube is a majorcontributor to the 300 ns length of these pulseshapes. We have estimated the exponential fluorescentdecay time of the glass from the difference of the twopulse shapes to be approximately 70 ns. Both 400 nsand 1 ~s gates were used with a LeCroy 2249W ADC withno observed difference in the energy resolution. Themeasurements reported in this paper were performedwith the 400 ns gate.

The calibration constants, C., for each elementof the array were defined by E'k~= C.P. where E' k isenergYtgeposited and P'k th~hob~erved~pa!seheigh~ inthe i counter for ~ne k shower. These constantswere determined from a set of +~O GeV runs using aniterative, bootstrap technique. Initially, a startingestimate was made for the calibrat~on constant of eachof the nine counters. For the i individual counterthe calibration constant C. was then determined byminimizing the chi-square ~

200

(/)

~ 150

~lL.o0: 100wm:E::I'Z

50

16.7 16.9 17.1 17.3 17.5 17.7 17.9 10.1 18.3EMEASURED (GcV) FIGUnE 3b

)

E beam energy The majority of the energy of the showers was

116

The energy resolution of the test array (which isdominated by the SCG1-C counter response) is shown inFig. 5 plotted against 111m.

Estimated Contributions to theEnergy Independent Term of the SCG1-C Resolution

Conversion Point Fluctuations 1.1%Undetected Energy Leakage Fluctuations 0.8%Resolution of Guard Blocks for Detected

Transverse Leakage 0.3%Momentum Spread of Test Beam 0.1 %Fluctuation of Shower Across Glue Joint 0.1%

Each of these sources of the energy independent termare to the first order independent and therefore canbe added in quadrature to get the 1.4% estimate of theconstant term. The largest contribution comes fromthe different optical attenuation of the light frompositrons converting at different depths in the glass.This contribution may be minimized by an appropriatefilter at the phototube to cut out the component ofthe light at short wavelengths which fluctuatedbecause of differential absorption. In addition,corrections for the position of the conversion pointmay be applied on a shower by shower basis if thelongitudinal development of the shower is sampled byuse of an active converter. The second largestcontribution to the constant term, the undetectedenergy leakage fluctuations, can be reduced byincreasing the length of the SCG1-C counter and byproper matching of the lengths of the guard countersto the central counter in order to minimize undetectedtransverse shower leakage.

Table II

Finally the relative light yield of the SCG1-Ccounter and one of the SF5 guard counters wasmeasured. The relative gain of the photo tube of theSCG1-C counter and the guard counter was determined bycomparing the pulse heights from a red and green LEDwhich were viewed by both these counters. Therelative gain determined in this manner has beencorrected for the different absorption of the LEDlight by the SF5 and SCG1-C glass. The two counterswere then exposed to a 4 GeV positron beam and theresulting pulse heights adjusted for the relativegains of the tubes. The ratio of the light outputs ofthe SCG1-C counter to the SF5 counter determined inthis way was found to be 5.10 ± 0.30. Most of theerror in this ratio originates in the mechanicalreproducibility of the gain measurement which requiredmoving the LED's from one counter to the other.

These data fit the form alE = a + bilE with a = 1.64%± 0.14% and b = 1.13% ± 0.33%. This resolution is tobe compare3 with the result a of' 1.65%11m obtained byBartalucci usin§ 60 MeV photons and a of' 2.5%IIEobtained by Chiba using 20 to 120 MeV electrons.These experiments were in the energy regime in whichsensitivity to a constant term was smalS. We haveused theEGS shower Monte Carlo program to estimatevarious sources of contributions to the constant term.Of the 1.6% which we observe, 1.4% can be expectedfrom the sources listed in Table II.

2X

20

.

III

10E SEAM (GeV)

FIGURE 4

I

SCGI- C SCINTILLATION GLASSENERGY RESOLUTION

I

5

3%1-

15

SCGI- C SCINTILLATION GLASSLINEARITY

contained in the SCG1-C counter with onlyapproximately 3.6% of the visible shower leaking intothe guard blocks. (This fraction was nearlyindependent of energy.) The linearity of the arraywith the beam centered on the SCG1-C counter is shownin Fig. 4.

20,..-,-,--,-..,.......,.-r-'"!:-r-,-..,.......-r-,r--r-r--,--r-.......-r-I

We find that these data fit a straight line with aof 3.4 for 2 degrees of freedom.

o

o

u

()

Conclusions, i ~

> > ~0>',._)

(!) ClI(!) (!) (!)It) 0 0 0~ 2 .f N

1% f- ~ l J ~I I I I I I

.1 .2 .3 .4 .5 .6 .7 .6 .9I/.IE (GeV-1/2)

.FIGURE 5

A counter constructed from Ohara SCG1-Cscintillation glass has been tested in a high energypositron beam and has been found to have resolution(alE = 1.64% + 1.13%/1E) and good linearity. Forcomparison the re~o9ution for lead glass measured byother experiments' is of order alE of' 1% + 4.511E. Ina complete scintillation glass detector the energyindependent term can be reduced by reducing leakageand by compensation for conversion point variations.The intrinsic decay time of glass is consistent withan exp(-t/70 ns) behavior. The light from a 4 GeV

.) 117

shower arr~v~ng at the photomultiplier of this oounteris larger by a faotor of 5.10 ± 0.30 than that for aoomparable SF5 oounter. We were enoouraged by theseresults and are designing a larger soinrbllation glassdeteotor for Fermilab Experiment E-705.

Aoknowledgments

We wish to thank Fermilab Physios Department andto aoknowledge the oonsiderable efforts of the SLACoperating staff. We would also like to thank theU.S. Department of Energy, the Natural Soienoes andEngineering Researoh Counoil of Canada, and the QuebeoDepartment of Eduoation for their support.

Referenoes

1. Y. Yoshimura, et al., Nuol. Inst. and Methods,137 (1976).

2. Ohara Optioal Glass, Ino., privateoommunioations.

3. S. Bartaluooi, et al., Laboratori Nazionali diFrasoati preprint, LNF-80/10(P), (1980).

4. M. Kobayashi, et al., KEK Preprint 81-8 (1981).5. M. Kobayashi, et al., Nuol. Inst. and Methods,

184 (1981).6. M. Chiba, et al., Nuol. Inst. and Methods, 190

(1981).7. J.E. Brau, et al., Nucl. Inst. and Methods, 196

(1982).8. R.L. Ford and W.R. Nelson, SLAC-210 (1978).9. J.A. Appel, et al., Nuol. Inst. and Methods, 127

(1975).10. M. Binkley, et al., Fermilab Proposal E-705

(1981).

118

n

)

)

)

/ )

)

)

n

CALORIMETRY AT L = 1033

W. Selove and G. Theodosiou

o Department of PhysicsUniversity of Pennsylvania

Philadelphia, PA 19104

i J

Abstract

We show that existing scintillation calorimetrytechniques make operation at collision rates of

108/sec feasible for most rare events. Thus for pp33colliders at L = 10 ,with DC operation, most such

events will be analyzable. The question of possiblemisleading effects due to pile-up is discussed.

Contents. 1. Introduction and summary2. Handling events of small time separation3. Time resolution of an individual event4. Calorimeter segmenting considerations5. Timing considerations for triggering6. Some final comments

The fundamental point is this: In any givensector of the calorimeter--say e.g. some 0.5 srsector, with several hundred towers--the timingsignals from these particles will form a tightlybunched group on a time plot--a group of FWHM of­order 0.3 to 0.5 nsec. (The a in this group will bemuch smaller than the pulse duration from eachcalorimeter module--see below.) If two events occur,within say a 10 nsec AUC gate, they will in generalform two groups on a PT-vs.-time plot; and for most

triggers these two groups will be (a) quite distinct,(b) very different in integrated PT.

We illustrate this point in Figure 1, whichrepresents a scatter plot of PT(tower) vs. time, for

1. INTRODUCTION AND SUMMARY

8 To work efficiently at collision rates of10 /sec, calorimeter signal duration, including anyjitter, must be short enough to fit into an AUC gateof order 10 ns or less. Moreover, one must be ableto cope with the fact that more than one interactionwill normally contribute to the set of AUC signalsresulting from a single trigger.

PT(GeVlcl

30

x20

rEVENT A

321T(ns)

o

10

Fig. 1. Expected scatter plot fortower PT vs. time, for an example

with 2 events depositing particles ina 0.5 sr group of towers. Event A isa jet with 100 GeV/c total, carryingmany high-PT particles; event B has

only 3 particles in this sector,carrying a total of 1.5 GeV/c. (Onlytower signals of 0.2 GeV/c or more areshown. )

Finally, for very rare events, which might be ofunusual physics interest, we discuss briefly pile-upeffects. Pile-up effects may be extremely small ifthe event signature is sufficiently unique (as forexample the signature for W+ ev); if they are notthat small, then pile-up contamination can of coursebe readily recognized by taking runs at differentluminosity.

2. HANDLING EVENTS WHICH OCCURWITH SMALL TIME SEPARATION

Both of these requirements can be met. First,very fast scintillation calorimeter signals are nowavailable. Secondly, we describe a technique wherebyone can expect to recognize very easily whether asecond interaction (or more) occurs within the gatetime, at least in 80% of the cases and probably in99% or more of the cases. This same technique canbe expected to allow one, in most cases, to use thetriggering event with little or negligibleinterference from the second event.

Our conclusion is that as far as calorimetry is33 -2 -1concerned, luminosities up to 10 em sec are

likely to be highly usable to search for rareprocesses. One might find that rare processes foundin this way are in fact simply due to pile-upeffects--but it is also possible that pile-up effectswill play no significant role.

o

\....)

Consider an event occurring in the interactiondiamond at some coordinate z along the length of thediamond. (Define z = 0 at the center of the diamond.)This event sends out particles in various directions.The particles reach different angular regions of thecalorimeter at different times (unless z = 0).

two events both delivering particles into one sector.This Figure shows what will typically be seen if thetwo events have a) 1.5 nsec time difference in theaverage arrival time of the particles, b) totalintegrated PT values of 100 GeV/c for event A (a

119

typical jet) and 1.5 GeV/c for event B (a typicalnon-jet set of particles). Clearly if the high-PT

event was the event of interest, and caused thesystem to trigger, than from this plot one would knowthat there was a second event present in the gate-­and that it was separable, if not in fact ignorable.

In the following sections of this note we treatfurther details of the time resolution of anindividual event. Here we make only a few furthercomments. (1) The two events ~hich would typicallycorrespond to Figure 1 will in general occur atdifferent zls. Thus recognition of 2-eventoccurrence, and separation of the events, is possiblenot only by PT-vs.-time analysis, but additionally by

tracking, given a tracking system of comparable timeresolution. Time analysis alone would allow' one tointerpret some 80% of high-PT triggering events as

probably clean, for the timing resolution indicatedabove; time and tracking analysis together wouldprobably allow over 99% of high-PT triggering events

to be given an initial interpretation as probablyclean. (2) The time separation between two events,shown as 1.5 nsec for the example in Figure I, willbe different in different parts (sectors) of a 4ncalorimeter array, since the relevant collisionsoccur typically at different zls. This variation isin general helpful in showing whether two (or more)events did occur within the gate time.

The point to be emphasized in connection withFigure 1 is this:

Most collisions result in relatively smalltotal E.r into the calorimeter array--but

more particularly they produce mainly low-PT

particles. Hence if the events of interesthave a major part of the information carriedby high-PT particles, then a second event

occurring within the ADC gate will generallygive virtually no distortion of an event of'interest.

A plot like Figure 1 merely facilitates recognizingthat the 's~cond' event, in this example, is ofalmost negligible PT' at least in this region of thecalorimeter.

3. TIME RESOLUTION OF AN INDIVIDUAL EVENT

We remark here on two separate points: a) Theduration of sampling calorimeter signals, and b) thepulse timing accuracy.

a) Two techniques have been reported at this Workshopwhich can give signals of 3 to 6 nsec FWHM. (Thereis a longer duration pulse "tail" effect, which wediscuss below.) One is the convetional SC-WSB-­scintillator-waveshifter bar--technique, used withmaterials having very short decay times and with PMs

which have fast rise time. 1 The second is the VPD--2vacuum photodiode--technique. Other potentially very

fast techniques have also been reported, but are notdiscussed here.

The fast SC-WSB technique is discussed inreference 1. Here we make a few further comments onthe VPD technique. The VPD readout system usesinexpensive fast scintillator and relativelyinexpensive thin large-area diodes. The major partof the signal can be obtained as a pulse of a fewnanoseconds duration. The low noise amplifiers usedwith such a device have normally been used with

shaping circuits which give rise times of order 100ns or more, which would require ADC gate lengths of200 ns or more. However, very much faster shapingcircuits can be used, and gates as short as 10 ns orless, at the cost of higher noise in proportion to

T-1/ 2 , where T is the gate length. The noise level,at 10 ns, would still be reasonably small in a hadroncalorimeter working with high PT events, involving

3say jets of 50 GeV/c or more.

There is a longer time "tail" component, in anycalorimeter, due to long-lasting physical processesin the development of the cascade energy. Restrictingthe signal collection time by use of a short gatewill thus generally worsen the energy resolution.This effect will vary substantially according to thedetailed calorimeter design and operating conditions.For non-uranium calorimeters, at high PT' the

resolution worsening can be expected to be relativelysmall; for uranium calorimeters, particularly atmodest p , the effect can be a factor of perhaps 1.5

4 Tto 2.

b) Pulse timing accuracy. We remark principally onthe VPD technique. Signals of say several GeV in atower correspond to 200,000 p.e. or more. Thus usingzero-crossing discriminators the time jitter for aTDC comes entirely from amplifier noise and not fromphotostatistics. For 0.1 GeV noise per module, andsay 4 nsec rise time, the time jitter would be lessthan 0.1 nsec for 5 GeV energy deposition, and would

vary like E-1•

For the SC-WSB Technique one expects about 100to 200 p.e. per GeV; in this case the TDC jitter comesfrom photostatistics rather than noise. For 4 nsecrise time the time jitter for this technique wouldalso be about 0.1 nsec, and in this case would vary

like E-1/ 2•

Other sources of time jitter include internaltime jitter in a module, and, for a calorimetersector (say 30· in theta and 60· in phi), variationin the flight time for particles from events with thevertex z different from zero. With suitable designall of these effects, combined, can be kept down to0.2 to 0.3 nsec, rms, thus providing the tight timegrouping, for events of say 20 GeV or more, indicatedin Figure 1.

4. CALORIMETER SEGMENTING CONSIDERATIONS

Individual tower size should be minimized,consistent with shower sizes, so as to reduce theprobability that the two events indicated in Figure 1will both deliver particles into the same tower.Precise granularity design will depend on the exactmachine energy, physics objectives, and overalldetector design. For a design in which at 90· theshower maximum occurs at about 150 cm from the beaminteraction point, and in which steel plates of 3 to5 cm thickness form the main part of the calorimeter,we estimate that 200 to 500 towers per steradian at90·, and about 2000 towers per unit rapidity elsewhere,would probably be appropriate. This would give about30,000 towers for a 10 TeV + 10 TeV collider.

To carry out the event separation indicated inFigure 1, the analysis of each event, whether on-lineor off-line, must be performed using "sectors"corresponding to a limited theta-band, so that thevarying flight distances for different particles fromthe interaction vertex will not unduly broaden the

120

()

)

, )

,)

n

o

()

()

peak in Figure 1. The tolerable size of each theta­band depends on the ratio of interaction diamondlength to radius of the calorimeter, as well as onwhether the calorimeter array is spherical orcylindrical. Theta-band widths of order 30 degreesappear satisfactory for the case of diamond lengthsof +/- 25 cm and calorimeter radius (spherical case)of 150 cm to the cascade shower peak.

5. TIMING CONSIDERATIONS FOR TRIGGERING

We have been discussing pulse durations of order5 nsec FWHM. For an interaction vertex which may haveany z value in the diamond, there can be substantialvariation in the signal time for "forward" particlesrelative to "backward" particles, up to +/- 2 nsec.If triggering is to use a combination of forward andbackward segments, as is the case e.g. for a missing­PT trigger, and if the trigger is to operate on

the combined amplitude of the calorimeter signalrather than on the combined area, then the signalsproviding the trigger may have to be re-shaped to havea flat top of length roughly 3 nsec. The signalswhich go to the ADC need not have this shaping--theADC gate length simply has to be wide enough to allowroughly +/- 1 nsec variation in the signal locationrelative to the gate, from event to event.

6. SOME FINAL COMMENTS

The approach we have described requires pushingpresent techniques to about the limit, to permitworking with 10 nsec gate lengths. One might fail toreach this limit, by a factor of 2 or perhaps even3. On the other hand, if the technology improves onecould reach this mode of operation without seriousproblems. A particularly important development inthis connection would be the development of thinvacuum photo-"triodes," with a gain of 10 or so.

We have discussed the fact that PT-vs.-time

analysis, combined with tracking information, wouldprobably allow some 99% of all high-PT events to be

analyzed even though additional interactions, but oflow PT' overlap in the ADC gate. The remaining 1%

121

involve events in which there are two interactions soclose together in both z and time that they cannot beseparated by the techniques described here. For thoseevents one can indeed mis-interpret the character ofthe event--the multiplicity of jets, the effectivemasses of combinations of jets, etc. Ultimately, anysuch mis-interpretation can be identified only by thetechnique described in section (1): take runs atdifferent luminosities. This limitation, on doingclean physics analysis at the highest luminosity,will always exist at some level, at any luminosity.As one pushes toward higher energies, and searchesfor ever higher masses and/or processes with weakercoupling constants, one can only try to developtechniques to improve the prospects for clean analysisat higher luminosity. It is in that spirit that wehave written this note.

7. ACKNOWLEDGMENTS

We wish to thank the organizers of the BerkeleyWorkshop--A. K. Mann et al. of the DPF, and PeterNemethy, Melissa Franklin, and the other members ofLBL, for the stimulating atmosphere at this Workshop.We also thank B. Pope, R. B. Palmer, and W. Willis forseveral interesting discussions.

REFERENCES

1. R. B. Palmer, this Workshop.

2. W. Kononenko et al.: NIM 186, 585 (1981); IEEETrans. on Nucl. Science, Vol. NS-30, No.1, p. 125(1983); U. of Pa. Report UPR 101E, Dec. 1982, tobe published in NIM. See also Proceedings of thisWorkshop.

3. W. Selove and G. Theodosiou, Proceedings of thisWorkshop, report about 5000 electron equivalent

noise charge for 20 cells of 20 cm2 , in a testsetup, at 150 nsec rise time, with a signal of20,000 p.e. from a muon. With optimization ofparameters, one could expect that with 5 nsec risetime amplifier noise equivalents of 0.1 to 0.2 GeVcould be obtained in hadron modules of this size.

4. W. J. Willis, this Workshop.

TRIGGERING AT HIGH LUMINOSITY: FAKE TRIGGERS FROM PILE-UP

Randy JohnsonBrookhaven National Laboratory, Upton, NY 11973

)

)

, )

:)

5040

\\

\\

\\

\\

\ ,

\ ,

\

20 30

Et GeV

\

\

10

1O-l,.::j0'---...,----,.---l-,..--l....;-----I

o

Figure 1. Trigger probability ~er interaction for thecross section P'O(E) = .84 e-· 8 E. Solid line isthe integral production cross section (n = 0). Dashedline is for n = 1 and dashed-dotted line for n = 5.

10° =r.""'~---------'"

Before moving on to some examples, let me make adisclaimer. I am not trying to predict rates for anyspecific experiment in a high rate erivironment wi.ththese examples, although they are taken from ISRexperiments and the ISAJET Monte Carlo ca1cu1ations. l

Instead, I will show what generic types of crosssections can and cannot be measured, and what can bedone .at the trigger level to reduce the pile-uptrigger rate.

The first example is a pure exponential:Pl(Et ) = .84 e-· 84Et (shown in Figure 1). Ascan be seen in Figure 1, any amount of pile-up duringthe measurment is disastrous. Increasing the triggerthreshold reduces the fraction of interesting triggers(triggers from single events with Et > Emin)'

r do/dE dEEmtn

probability of an individual event.= trigger

1Po =

°TOT

and

Then

Changing the order of integration gives

Triggers based on a cut in transverse momentum(Pt) have proved to be useful in high energy physicsboth because they indicate that a hard constituentscattering has occurred and because they can be madequickly enough to gate electronics. These triggerswill continue to be useful at high luminosities ifoverlapping events do not cause an excessive number offake triggers. In this paper, I will determine ifthis is indeed a problem at high. luminosity machines.

The trigger probability for high Pt processeswith n additional overlapping events can be calculatedas follows. Let Pn(Emin) be the probability thatn overlapping events give an energy E ~ Emin withPn(O) = 1. (E can be total energy, transversemomentum along an axis in a hemisphere, or any othercontinuous parameter of an event which is positive andadditive among overlapping events.) P'n(E) =dPn/dE = the differential probability for getting noverlapping events between E and E + dE.

1P' 0 = 0TOT = do/dE

The probability that n interactions give a trigger isthe probability one interaction triggered the systemplus the convolution integral for the probability thatone interaction combined with the group of (n - 1)interactions to give a trigger. With this formula,the probability for n interactions can be calculatedby successive convolutions of the single interactionprobability distribution.

Convolutions of exponential distributions getequal contributions from all portions of theintegration interval; convolutions of distributionswhich fall less steeply will tend to get the mostweight from the limits of the integration. Therefore,it is not surprising that the most probable high Pttrigger from n overlapping events will come from onemoderately high Pt event plus (n - 1) low Ptevents.

For triggering purposes, the mean number ofinteractions during the integration time of the analogsum is the appropriate measure of the luminosity.High luminosity at the ISR has meant .2 to .4interactions per ~rig~er integration time. At aluminosity of 103 /cm /sec the mean number ofinteractions might be between 1 and 5. The .triggerprobability for an average multiplicity of nandtrigger threshold of Emin is the sum of Pn(Emin)!eighted by the Poisson statistic for n with a mean ofn.

122

()

A parton picture of the nucleons implies that thehigh Pt scattering cross section eventually deviatesfrom a pure exponential. As the second example, letthe cross section follow the exponential to Et = 10GeV/c and then fall like Et - B beyond that. Thetriggering probability distributions for this crosssection are shown in Figure 2. The low Et portionis still dominated by pile-up from multiple eventsfrom the exponential region, but, as Et goes up, thetrigger rate eventually deviates from this exponentialpile-up and stays at approximately a constant factorabove the true cross section. In the high Etregion, the pile-up just lowers the effectivethreshold from Et to Et - n Etaverage.

These first two examples are consistent with theEt ~ross section measured by R807 at the ISR. 2 Theyobserved an increase in the trigger rate of about 5when going from a mean multiplicity of 0 to .2 with anEt threshold of 10 GeV/c. The exponentialdistribution of example i gives a factor of3.3 increase for these 'parameters.

At higher energy storage rings, the flattening ofthe Et cross section is expected to begin sooner.The Et cross section predicted by ISAJET for Ecm =800 GeV is shown in Figure 3 along with the triggerrates for n = 1 and n = 5. Because the cross sectionis flatter, the deviation from exponential pile-upoccurs at a lower Et • The Etaverage per eventhas not changed, however, and the shift in theeffective trigger level is about the same as in Figure2.

50

\

\

40

\

\

\

\

,

\\

20 30

Et GeV

\

\

\

\

10

,

\·············.···.....................................

,,,

,

,

10-6+----r-----r--,----r----lo

10° --,...-----------,

504020 30

Et GeV10

10-1'.:::.J0~_---'---'-----"--r--"""'r---!o

o

()

Figure 2. Trigger probability eer interaction for thecross section P'O(E) = .84 e-· B E for E ~ 10 GeVand P'O(E) = 1.8 x 10Q·E-B'for E > 10 GeV. Solid lineis the integral production cross section, dashed linefor n = 1 and dashed-dotted line for n = 5. Thedotted lines are the curves from Figure 1 for the pureexponential distribution.

Figure 3. Trigger probabilities per interaction forthe ISAJET production cross sections at Ecm = 800GeV/c. Solid curve is for the integral productioncross section, dashed curve for n = 1 anddashed-dotted curve for n = 5.

123

)

)

)

)

'j

References

3. H.A. Gordon, M.J. Murtagh, R.A. Johnson and D.P.Weygand, Monte Carlo Simulation of the Effect ofPile-up, these proceedings •

1. F.E. Paige and S.D. Protopopescu, ISAJET: A MonteCarlo Event Generator for pp and pp Interactions,Version 3, Proceedings of DPF Summer Workshop onFuture Facilities, 1982, Snowmass, Colorado.

technique, the accidental to actual rate is 1:1 downto 30 GeV/c for 5 overlapping events. Placing athreshold requirement on the subelements of thedetector reduces the average Et even further.Although this technique cannot be calculatedanalytically as ,the previous examples have, HowardGordon and Dennis Weygand have written a Monte Carloprogram using ISAJET to simulate this process. 3 Theyfind that with a minimum cut of 1 GeV/ c the highluminosity trigger rate (n = 10) is only 30% more thanthe low luminosity rate down to Et = 10 GeV.

What do these examples teach us? First,theamount of pile-up is dependent on the cross section tobe measured. Exponential cross sections are verydifficult to measure at anything but extremely lowluminosities; cross sections which flatten out at highPt are easier. For those flatter cross sections,the ambient background from multiple events tends toshift the energy scale of the trigger. A trigger setfor 50 GeV/c will trigger, on events produced at Pt= 50 - fi ptaverage. n can be reduced by reducingthe resolving time of the apparatus. ptaveragecan be reduced by reducing the area covered by thetrigger, or by summing only the showers generated byparticles coming from a given vertex region. Ifthese techniques are used, triggering on high Ptevents can be quite efficient even at highluminosities.

2. The AFS Collaboration, Opertion of the AFS at 1.4x 1032/cm2/sec: A First Look at Data at HighLuminosity from the CERN ISR, these proceedings.

\\

·······················\..... \

100~-------------..,

... \

\ .....\\<\

\<\"'<.\

"'<.~.

The way to reduce the number of pile-up triggersis to reduce either n or Etaverage. Obviously,better time resolution and smaller integration timeswill reduce n. Etaverage can be reduced in a ,number of ways. Rather than summing the Et over theentire solid angle, just the Et in the small areaaround the jet need be summed. Then the Et from thejet would be unaffected, but the ambient Et would bereduced (Figure 4). The probability distribution forone jet is the same as in the previous example, butthe probability distributions for all subsequentevents are reduced. By reducing the integration areafrom ay= 4, a$ = n (as in Figure 3) to ay = I, a$ =n/2 the average Et for an accidental coincidence isreduced from 3 GeV/c to .7 GeV/c. With thistechnique, the accidental to actual rate is 1:1 down

10-6:+---,----;----::.""""""T--r----Io 10 20 30 40 50

Et GeV

Figure 4. Trigger probabilities per interaction forthe reduced solid angle trigger described in the text.The solid line is the integral production crosssection used for PO(E); the dotted line is the randomjet probability distribution used for the convolutionintegrals. The dashed curve is for n a 1 and thedashed-dotted curve for n = 5.

124

FAST ASYNCHRONOUS LEVEL 1 PRE-TRIGGERFOR ELECTRONS AT L = 10 33 cm-2 sec- l

Michael J. TannenbaumBrookhaven National Laboratory, Upton, NY 11973

The pre-trigger is made as follows: A singleparticle solid angle cell is defined by linearlyfanning-in (100) independent groups of 4 x 4 towersin the 40 x 40 array. The output of each fan-in goesto a discriminator with several thresholds and alsoto the next level linear fan-in (3 x 3 cells of the 4x 4 'towers, for a total coverage 8~ = 0.611, 8y = 1.2which is suitable for a jet trigger, etc.). Each ofthe cells of 4 x 4 towers is fully efficient forsingle particles in the central 2 x 2 towers, andless efficient in the outer 12 towers. Thus for fullefficiency over the whole array, four sets of cellsof 4 x 4 towers can be formed, each centered on theintersections of the edges of alternate rows andcolumns.

The trigger will have a time slewing of 20 nsecdue to the dynamic range so that it will have to beretimed with a zero-crossing or constant fractiondiscriminator to a precision of ~ 2 nsec. At thispoint other fast trigger information like TransitionRadiator Electron identification1 or a nearby trackl

with PT > 2 GeV/c can be added. Leading edgetiming from the calorimeter l can be used to determinewhich towers were struck in-time (5 to 20 nsec) toavoid pile-up during the longer ADC, gate.

The triggering cell size is sufficiently smallthat triggering rates can be estimated byintegrating the measured6 single 11° cross section atIs = 540 GeV.

o

o

,)

Asynchronous Triggers have long been used atfixed target accelerators and the CW High LuminosityCERN ISR. Bunched beam colliders have tended to usetriggers which were synchronized with the time of thebeam crossing. The CDF trigger schemel has 3.5 ~secbetween such crossings to decide whether to furtherprocess any events which occurred during thecrossing. The level 1 trigger can accept a rate of50 KHz without appreciable dead time. l The level 2trigger uses fast bit-slice processors to selectevent topologies in 10 ~sec and thus can accept atrigger rate of 5 KHz l • Readout of the system whichtakes 1/2 msec is required for level 3, thus limitingthe triggering rate at level 3 to ~ 100 Hz. l Thepurpose of this note is to show that this sametri~ger scheme would work at a CW luminosity of 1033

cm- sec- 1 by the addition of a conventionalhard-wired asynchronous pre-trigger.

A pre-trigger for relatively low PT electronsis discussed, since electrons have long been used asa signature for interesting hadron interactions. LowPT is emphasized, since such interesting particleslike J/W and T tend to produce electrons with PT~ 1 GeV/c and ~ 4 GeV/c respectively. If theseelectrons are lost at the trigger level there is nohope of getting them back afterwards.

One imagines2 a typical central region detectoras covering the full azimuth, a rapidity range - 2 <y < + 2 and being segmented into 40 x 40 = 1600 ­calorimeter towers each covering 8~ = .16, 8y = 0.10.The calorimeter will be divided into electromagneticand hadronic compartments. For the purpose of thelow PT electrons discussed here, a 17 radiationlength counter is superior in charged pion rejectionto the 26 radiation lengths used in UA1,3 so it isassumed that the electron compartment will besubdivided into at least 2 compartments, with thefirst ~ 17 radiation lengths being used in thistrigger.

It is important to distinguish the timeresolution required to obtain analog information withprecision suitable for triggering from the timerequired to integrate the signal for the ultimateenergy resolution. For the R-807 calorimeter~, thepulses could have been clipped to ~ 20 nsec fortriggering, although a 125 nsec integration time wasrequired at the ADC. A high impedance fanout nearthe photomultiplier would tap-off, clip anddistribute the fast signals for the hard-wiredtrigger while the main signal would proceed to theprecision ADC through a delay of ~ 500 nsec. Duringthis time the main signal would have its rise timedegraded by the long delay but this predominantlydispersive process would have little effect on thepulse area.

The tower size of the calorimeter discussedabove2 is well matched to single particle detectionfor PT < 15 GeV/c. CCOR5 used a tower size of 8~ =0.10, 8y = .10 for single particle studies at theISR. Furthermore, the jets at these low PT valuesare much wider than those at higher PT' It isexpected that the background conditions for singleparticle triggers with PT < 15 GeV/c will be verysimilar to conditions at the ISR where extensive dataexist on single particles and jets with PT < 15GeV/c.

E4.9 x 10-25 cm2

(PT + 1)8.31

,)

1. M. Shochet, Trigger Summary, These Proceedings.2. H. Gordon, et a1., "Reasons Experiments can bePerformed at L = 1033 ," DPF Snowmass Proceedings.

125

Figure 1. Trigger Rates for Low PT electrontriggers.

For a detector covering b~ = 2n, by = 4 and forL = 1033 cm-2 sec-l, the nO production rate is shownas the solid curve in Figure 1. If there were notracking or particle I.D. information, this would bethe trigger rate. The threshold is determined byeither the 20 KHz maximum trigger rate at level 1 orthe 5 KHz of level 2. With tracking, or particleI.D., the threshold can be reduced.

The principal background is due to chargedhadrons, n+, which deposit most of their energy inthe electromagnetic calorimeter. For a fixed energydeposition,this background is down by a factor of 20from nO (dashed line). This would be 'the limit ofthe trigger improvement with fast tracking. l A simi­lar but smaller background is a nO which deposits itsenergy accompanied by a charged track with PT > 2GeV/c coming from the same jet and landing on thesame 4 x 4 cell. This is at the levelS of ~ 1.5% andis less of a problem (long-short dashed curve). Thedotted curve shows the real electron background7 dueto Dalitz and external conversions (1.5% XO)' which isdown a factor of 500 from the nO rate. If theelectron energy is not measured independently of theparent nO, then the relevant rejection factor is only5% which is the probability that the energy of a nOincludes any conversion electrons. With a magneticfield to sweep aw~y soft electrons, a SegmentedTransition Radiator l could provide a trigger levelrejection of 1% (long-short dashed curve). Thiswould allow a threshold of8 4 GeV/c for singleelectrons at 5KHz triggering rate. Level 2 can thenbe used to find interesting event topologies like 2or more electrons, electron + jet, electron + missingET etc.

)

)

)

3.4.

5.

6.

7.

8.

C. Rubbia, These Proceedings.B. Pope, Calorimeter Summary, These Proceedings,quoting R.B. Palmer. ,A.L.S. Angelis, et al., Physics Letters 79B, 505(1978); Physica Scripta 19, 116 (1979). ----UA2, J,P. Repellin, PariS-Conference 1982, asquoted by R. Cahn, These Proceedings.G. Bunce, et al., W, ZO Production, DPF Snowmass1982N.B. the threshold of 3 GeV/c from Figure 1 wasmultiplied by a factor of 1.3 to allow forsystematics. This factor will be refined withfurther investigation.

126

PROSPECTS FOR CERENKOV RING IMAGING AT HADRON COLLIDERS

Marvin Goldberg

oSyracuse University. Syracuse. New York 13210

David W.G.S. Leith

Stanford Linear Accelerator Center. Stanford. California 94305

1. Introduction

... l. I I··I·~~i1

:Fig. 3. Schematic of 16m Helium radiator of E605 atFermilab. Photon detectors are shaded.

PI)

The fixed target E605 arrangement at Fermilab isshown in Fig. 3. Here the radiator (length 16m) islonger than the focal length of the mirror segments.which focus the U.V. photons on the outboarddetectors. which never see the beam particles. Heliumis the radiator of choice in this experiment.

Fig. 2. Detector of SLD proposal at SLC. The DELPHIdetector at LEP will have individual photon detectorsassociated with each mirror. A liquid Freon radiatoris shown in addition to gaseous isobutane.

. .........t··········

The imaging properties of the mirror can easily be~nderstood by realizing that the two sets of parallelCerenkov rays in the plane of the page look like asource at infinity. and thus form two points on a ringof radius R=f tan8 where f=RM/2 is the mirrorfocal length. A photon detector sits in the focalplane of the mirror.

The above arrangement can be inconvenient whenother apparatus is needed. This leads to a"retracted" geometry where segments of the mirror aremoved away from the interaction region. More mirrorsurface area is required at the larger distance. butthe total area of the photon detectors can remainesentially the same due to the focusing of rings fromnon normally incident particles. This arrangement isused in the SLC and LEP proposals (see Fig. 2) whereisobutane is the gas radiator.

particle 2Cherenkovrodioting medium /

Mirror radius /.

RM /

/Oelectorradius RD

Summary

The developing techniques of Cerenkov ringimaging detectors are reviewed. with emphasis on theirperformance in high luminosity environments.

II. Geometry

Fig. 1 illustrates the "classic" geometry wherethe interaction region is surrounded by a mirror ofradius RM• Charged partiples of velocity S orLorentz factor Y radiate Cerenkov photons at an angleo which is focused on a sphere of radius RM/2.

Cerenkov ring imaging is a rapidly developingtechnique which may be used to identify chargedparticles over a wide momentum range. Such detec10rshave recently been utilized in E605 at Fermi~ab(1 •and(h)ave been proposed for detectors at LEP(2 and theSLC 3. The requirements of high quantum efficiencyhave led to the use of gaseous detectors containingphotoionizing vapors.

Since the ring imaging qetector can measure theangle of radiation of the Cerenkov photons. thedirection of motion of the radiating particle and thenumber of radiated photons. it Acan extract far moreinformation than conventional Cerenkovs. and needs onlytwo-dimensional readouts as opposed to dE/dxtechniques which require far larger numbers of wires.These features lead to relatively compact detectors oflarge dynamic range.

Fig. 1. Schematic view of the ring images producedby two different particles emerging from a target oran interaction region.

)

C)

127

One can also imagine a "hidden" geometry. wherethe photons are reflected to a deteotor shielded fromparticles emerging from the interaotion region. Thistechnique could be use(ul at small angles inconjunction with high Cerenkov thresholds (Lorentzfactor YT) in cases of high luminosity. so that onlyhigh momentum particles are seen by the detector (seeFig. 4).

V. Precision of Position Measurement

Fig. 6 shows the contrib~~tons to the uncertainty ,/:,8 due to various sources for an Argon radiator.

For regions of interest here, chromatic dispersion isa significant limitation. We note that an uncertaintyof /:,8 = 0.3mrad will provide the upper limit to thedynamic range. Since 8", =26mrad, we conclude that a1% angular measurement is possible. Since gases withless inherent dispersion (smaller /:,8 I 8) generally

IR•

Fig. 4. Schematic of possible "hidden" geometry.where photon detector'is hidden by shielding frominteraction region.

III. Useful Equations and Approximations(4)

With f3 the partiole velocity and n the radiatorindex of refraotion. then

(1) oos 8 = 11 f3n and small 8 then(2) sin 8'::, tan 8 :; 8 For relativistic

particles(3) 1- f3 =1/2y 2•

Then for small n-1(4) 8", = J2(n-1l ~ 1/ Y T where YT is

threshold Y • 8", is 8 as Y -> '" •The number of detected photons per cm of radiator isthen

Fig. 5. In thisAplot the left curve shows howthe normalized Cerenkov angle 8 n= 8/8", grows fromzero at the normalized velocity y n=1 up towards themaximum value 1 at large velocities. Taking theordinate to be a relative energy scale normalized tothe threshold energy for pions, the three curvescorrespond to 11, K, and p as indicated. The ordinateon the ri$ht-hand side shows the normalized number ofemitted Cerenkov photons.

have lower indioes of refraction and are thereforelonger, the required position resolution does not varysubstantially with the choice of gas radiator in theclassic or retracted geometry.

,)

)

)

(5)where

(6)

Fig 6. Aberrations of an argon gas radiator with thelisted geometry and detector condition versus theLorentz Yfactor. The threshold of such a counter isYT=40; s is the wire spacing or equivalent.

VI. Readout

A collider detector suited to study narrow jetsshould maximize the number of radiated photons so thatoverlapping rings are clearly resolved. While twodimensional readout could accomplish this task, thenumber of elements and complexity of readout requireddoes not presently allow this technique. Electronpositron collider detectors have chosen to driftelectrons perpendicular to the direction of incidentphotons, determining the associated co-ordinate from

, )

! •

110100

E,=6.25 ev

Ez=7.25 ev

S=lrrm

90

ARGON GAS (I bor, 20·CI

geometric,

R.. =200cm

Rz =100.5

Ro =100

multiple SCOl1ering (r I

! optiC !

60 70 80y

, chromaticH

~<D<l 0.4

The ring radius is (8) R = f8 where (9)f = L = RM/2 is the focal length for a radiator oflength L in the classical configuration with mirrorradius RM• For two particles of the same momentumwhere Yma~efers to the heavier particle

(10) fiR/Roo ~ fi8/8", :; Y¥/2y2max

IV. Dynamic Range

The curves of Fig. 5 illustrate(5), in a radiatorindependent manner, the angle of emission of Cerenkovradiation as a functionAof y. The detected ring sizeis used to measure the Cerenkov angle, and so a kaon,for example, oan be identified from just above itsthreshold YT up to ym ' the energy where 1- 8 K istoo small to measure wr~h precision. The range forklon ideRtif1cation. YT to Ymax' or correspondinglyE T to E max' yields a range of Ifcomplete" 11 - K-pseparation. Equation (10) indicates that a 1%measurement in radius yields a range in energy from ETto El x = 7 E* for such separation.

rdditional partiole identification is possibleover other energy intervals. Pion-electron separationis available as long as the pion ring isdistingUishable from that of the electron. The energyinterval range is then from E~ to 7 ET.

N", =NoL 8J = NoLI Y~ = 2NoL! n-1 )

N =370eV-1cm-1 £ fiE where fiE is~he energy interval for detected photons andE an overall efficiency.

For slower particles (7) N=N", (1 - Y,fl y2).

128

o

o

()

the drift time and anode wire positio~ (See Fig. 7)A third oo-ordinate oould be determined by oathodereadout. This teohnique is not suited to highluminosity oolliders beoause of the large timeaperture assooiated with the long drift distanoe(>100m). Constraining ourselves to reasonable numbersof MWPC wires leads to a system of eleotron deteotionbelow the oonversion region of Fig. 7, instead of onthe side. Fig. 8 illustrates this type as used in thedeteotor of Fermilab E605. The preamp plus PWCoombination allows suffioient gains while minimizingproblems due to avalanohe produoed photons. The tworequired oo-ordinates are measured by a oombination ofanode and stereo oathode wires. Fig. 9 shows anexample of this type of readout. Even photons distantfrom the ring of interest oan oause oonfusion, and thenumber of photons per ring must be limited to -6.Improvements in this readout may be aohieved by:oharge division for the distant photon problem; pulseheight oorrelations for ring overlap problems; timeslioing to improve the resolution on non-normallydrifting eleotrons.

Fig. 7. Sohematio view of the single-photon driftdeteotor, Delphi proposal.

Fig. 8. Photon deteotor struoture.

VII. Choioe of Radiator-Deteotor System

A luminosity of 10330m-2seo-1 at 1TeV willproduoe -3.5 partioles/interaotion in a unit rapidityinterval (or 21Tster) for polar angles 30·< e <150· atan interaotion rate of 50 MHz. If we oonsider adeteotor of solid angle -1str, a flux of -3x107Hz mustbe handled.

With ourrent materials, we have a ohoioe of twophotoionizing vapors, TMAE and TEA. The time apertureof a TMAE system is 0.5-1 usee due to the 0.5-10m meanfree path of photons in the gas, while a TEA systemoan provide a mean free path of 0.5-1mm and aoorresponding time aperture of 50-100nseo.

The above oonsiderations have led to a ohoioe ofa TEA deteotor for E605, and TMAE proposals foreleotron positron oolliders. As Fig. 10 indioates,TMAE has the advantage of quartz windows, variety inliquid and gaseous radiators and higher quantumeffioienoy. If heated to inorease its vapor pressure,the time aperture may be reduoed somewhat. A magnetmay be Plaoed

6in front of the deteotor, reduoing the

flux to -5x10 Hz, and if a oell of photon dete~tor

sees 10mstr, the flux through it oould be 5x10 Hz.Very few of these partioles will produoe rings in thegas deteotor due to the radiation threshold.Providing that other deteotor elements label traoksfrom previous events, a TMAE system like that of theDelphi proposal with the addition of E 605 typereadout with time slioing, may work. The limits for1~-K-P separation are indioated in Fig. 11 for a s3stemof liquid freon and isobutane (n-1 = .277, 1.7x10­respeotively) •

A TEA deteotor oould have a time resolution anorder of magnitude better than that for TMAE: oaloiumfluoride windows with speoial mountings (oost $200for 100m x 100m window) are then neoessary, and noble

Fig. 9. An example of photon reoonstruotion for anevent with four photons and a beamAtraok. The beamtraok is not at the oenter of the Cererikov ring whiohis determined by the partiole direotion. Two ghostpoints appear, but they are easily eliminated beoausethree ooordinates are required for other, real points.

from wireaddress

C photon-4 mm

CoF

6

CPAIIII I

:~:,,....- I

I II II I

: :I I

4

Incident pholonsyy

=0

18

Drift

x from drift lime

IIIII .(

: ;l'II,,I

3.2 3.2

? ? ?y u x

PWC

Quartzwindow

)

)

()

()

129

Fig. 10. Quantum efficiency of various photoionizingvapors and window cutoff frequencies.

Table I would suggest Argon as the preferred radiator.A 10 mstr cell at -3 meters from the intersectionwould enclose most rings.

The above detection can help identify electronsup to E~ax=37Gev. or distinguish pions from ~s andprotons rrom 6 GeV. and provide complete 1f - K-Pseparation from E¥=20GeV to the highest momentum ofinterest.

(n-1)x10-5 YT Length for 8detected photonsassuming No = 80cm-1

cm

. )

3

8

4

12

ClassicGeometry

cm

19

56

15

12 42

16

5.3

4.191

1340

690

147

30

116

83

3834

55

3.7

Ne 7.2

Ar

Kr

He

Table I

Noble Gas Characteristics(7)

9 g tVI I

~o

I

I1001-

Q. E.

gases must serve as radiators.

Realizing that the energy range for ~k-p

separation is -E~ to 7E~. and that the upper limit ofinterest at proposed hadron colliders is <100 GeV.

)

)

)

X. Conclusions

Although there are difficulties. one can conceiveof a detector with freon. isobutane and argonradiators. covering a momentum range from less than 1to over 100 Gev with 1f-K-P separation abilities. Wewish to thank F. Sauli for many valuableconversations.

VIII. Cell Content on Triggered Eyents

IX. R&D Actiyities

The following items would clearly be of interestA. TEA matched radiators with indices of

refraction larger than isobutane butwith no cryogenic or pressurerequirements.

B. Thin photocathodes of high quantumefficiencies (CsI and liquid TMAE areunder study).

C. Two dimensional9 fast and efficientreadout schemes (needles. pads. CCD's).·

D. Other TEA matched windows.Additional studies should be done to assess thefeasibility of long liquid helium ring imagingdetectors near the beam lines.

If the density of rings per event per cell ismore than 1. reconstruction will be difficult with thetype of readout proposed. In the gaseous radiationdetectors. preliminary ISAJET studies indicate that ajet trigger will provide a mean number of -1 ring plusionization of -1 extra particles if used incombination with a sweeping magnet in front of thedetector. This type of ring density will also beencountered in LEP and SLC detectors.

Extending the momentum convergence downward withTEA presents unpleasant choices: r:ressurized Argon(where YT varies as (pressure)-1 2. or liquid Helium(n-1 =0.02). The cryogenic problem may be easier tolive with than high pressure on CaF2 windows. ThesimPlest(gylution seems to be a series of thresholdcounters •

K

• LIQUID'. RICH

\\

"---\------------- -----

"'"

\\\

K

0) 71"-K

(/) 10".,.,.-"'---,-rr'T'"'"'T'"'".,-.,..-,-rr-----.-,.-,.-,.......,......,Zo~:> B~

~ 6~Z«lii 4

~ 3 ---------

8 2

~a:~ Oc.......L..L.'-'--_ __J.---l_'_.L...J~~.n__ __J._ _'_.L...J~J..."-LJ

w 10,,,...,...,...,.----,--,-.,.....,.,..,.,.-.,,,.--,-,--,-.--,-.-.-.,..-,-T"IW •~ b)K-p \~ B \

~ \~ 6 - \~ LIQUID \u RICH \

~ 4 ! \z 3 -~ ---------------\OJ'ti 2 i~ i!JJ 0, ';-'-.........."'7-'----:!:--'---'--:--'-..J....l.-"':---'::'=--'---'-:~..J....l..LJ

.5 5 10 20 50 100P (GeVlcl

Fig. 11. Pion-kaon and kaon-proton separationsgiven by the forward Delphi counters. For both pairsof particles the two counters match. giving many morethan the conventional three standard deviationseparation in wide momentum ranges: 0.3-35 and 0.7-60GeV/c respectively. The thresholds (i.e. the verticallines) correspond to 95~ detection efficiency for thelighter particles. The arrows indicate the thresholdfor the heavier particles.

130

o

o

o

)

)

References

(1) Paper submitted to IEEE Transactions on NuclearScience, E605, Fermilab. Also see R. Boucher etal CERN EP 82-83. A. Breskin et al IEEETransactions on Nuclear Science NS-28, 1981, 429.G. Coutrakon et al IEEE Transactions on NuclearScience NS-29, 1982, 329

(2) DELPHI 82/1 and COLLEPS 82/23, Proposals to CERNfor the LEP detector.

(3) The SLD proposal of Cal. Tech, J~hns Hopkins,MIT, SLAC, U. of ILL., U. of Wash.

(4) Proc of 1981 Summer Workshop, Isabelle, BNL 50885

(5) T. Ekelof et al Phys Scripta 23, 718, 1981

(6) T. Ypsilantis, Phys Scripta~, 371, 1981

(7) T. Ekelof, ICFA Workshop, 1979, Ed by U. Amaldi(CERN, 1980) P 396

(8) Proc of 1978 Summer Workshop, Isabelle BNL 51443

(9) B.Robinson, Phys Scripta 2.3,., 716, 1981

131

USE OF SYNCHROTRON RADIATION FOR ELECTRON IDENTIFICATIONAT HIGH LUMINOSITY

S. AronsonBrookhaven National Laboratory, Upton, NY 11973

Fig. 2. The number of X-ray photons produced versusphoton energy, Ex. The curves are labelled accord­ing to electron energy in GeV. The shaded verticallines delimit the region of good X-ray detection effi­ciency in Xenon.

Introduction

Synchrotron radiation has been used successfullyto identify electrons of

l10 to 30 GeV traversing a

field length of 30 kG-m. Since comparable fieldlengths are a feature of many proposed collider detec­tors, and since this is an electron energy range ofinterest at Is ~ 1 TeV, we consider ~~ether such adevice could be useful in the L = 10 environment. 2

Figure 1 shows the arrangement. The electrontrack and its "tail" of X-rays are detected in a

DETECTOR

2.5

iii'<I>!0

Q...,'"~

z

0.1

2 5 10

E. (keV)

20

B-30kG

1- 1m

50 100

)

2

3010

B-30kG1-lm(IOem Xe)

B

c)

lIJ...0 6lIJ...lIJC

~

z4

20

E.IGeV)

Fig. 3. The number of photons detected in 10 cm of Xeas a function of electron energy.

with 2.4m long wires parallel to the magnet~c field.The charged single particle rate at L = 10 3 (takinginto account the PT cutoff of the field) is about0.25 MHz/wire (or 4 ~sec mean separation). Since oneshould be able to operate such a chamber with a gate <100 nsec, rate does not appear to be a problem in thisconfiguration.

For example d is 15mm for a 30 GeV electron with B =30 kG and ~ = 1m. The width of the X-ray tail isdetermined by the spatial resolution of the detectorin the direction perpendicular to the bend plane.

Xenon-filled MWPC. Figure 2 shows the spectrum ofX-rays produced in a 1 meter distance at 30 kG forseveral electron energies. The total energy emittedin a distance ~ and field B is given by

6E ~ 0.013 E2B2~,x e

The electron identification depends on thespatial correlation between the electron and X-rayhits; the X-rays lie in the bend plane of theelectron, spread over a distance d(mm), where

d = 15 B~2/Ee.

Fig. 1. Schematic depiction of electron identifica­tion by synchrotron radiation. The length of theX-ray "tail" is denoted by d. The detector is amultilayer wire chamber, the large dots representhits.

where 6Ex is in keY, Ee is in GeV, ~ is in metersand B is in kG. Despite the fact that the totalenergy is proportional to Ee

2 , the number of detec­ted photons peaks at Ee = 5 GeV (see Fig. 3). Thisis due to the detection efficiency of Xenon and to theshift of the spectrum with Ee depicted in Fig. 2.

Rates and Particle Densities

We assume a detector which covers about 2 unitsof rapidity (centered at y = 0) and is composed of 4mmmini-drift cells at 1m from the interaction diamond,

A more serious question (albeit independent ofluminosity) is that of fake electron triggers ortrigger losses due to the spatial overlap of particleswithin the same event. To investigate this questionwe use predictions of particle densities within 100

132

o

C)

()

o

o

u

GeV jets from ISAJET. 3 To estimate the resolutionperpendicular to the bend plane, we assume this isdone with cathode strips which yield 5mm localizationof the hits. Taking any spatial coincidence of twocharged particles within an area = 5mm x d to be an"electron," we find about 0.1 fake electrons above 10GeV per 100 GeV jet. This can be suppressedsignificantly by seeing if the tracking system finds 1or more than 1 track pointing to the "electron."Further suppression could be based on the fact that acharged track and a collection of X-rays would producevery different energy depositions on the drift wires.With dedicated processors these reduction factorscould probably be obtained in the few ~sec ava\lable.A crude estimate of these factors yields < 10- fakeelectron triggers above 10 GeV per event.

Using the same input from ISAJET one candetermine the loss of real electrons due to overlapsof other charged tracks. In 100 GeV jets this loss is1 to 2% for 10 to 30 GeV electrons. (A brief look at1000 GeV jets reveals that for comparable values ofEe!E(jet) the losses are at most a factor of 2worse. )

Compatibility with Other Detector Components

Note that AEx « B2~ and d «B~2. Thus onebenefits (at fixed fBd~) from larger B and smaller ~.

This yields more X-rays in a smaller impact area.Tracking systems, on the other hand, prefer large ~

and smaller B. This gives larger sagittas (also «B~2) and fewer trapped soft tracks. Thus compromisesneed be made in a system that includes both. Itshould be kept in mind that if the tracking chamber iscarefully designed to keep the mass low, then it willbe transparent to those X-rays which can be detectedin the Xenon chamber. Thus tracking and electronidentification can coexist in the same field volume.

Research and Development

At Ee = 10 GeV the number of detected X-rays isalready falling (albeit slowly) because the X-rayspectrum peak has shifted out of the Xenon detectionwindow. The energy range of identifiable electronscould be expanded (or the B,~ requirements lowered) ifa detector with .good efficiency for Ex > 100 keVcould be used. Thus development of a practical liquidor solid detector might be very useful.

This research supported by the U.S. Dept. ofEnergy under Contract No. DE-AC02-76CH00016.

References

)

1.

2.

3.

J. Dworkin, Univ. of Michigan Ph.D thesis(unpublished). In this case electrons from A-adecay in a neutral hyperon spectrometer wereidentified with a Xe-filled proportional chamberfollowing the analysis magnet.J. Kirz and H. Walenta, Proceedings of the 1981ISABELLE Summer Workshop, BNL 51443, p. 1406.F.E. Paige and S.D. Protopopescu, "ISAJ]!;T: AMonte Carlo Event Generator for pp and ppInteractions, Version 3," BNL 31987 (Sept. 1982).

133

TOF FOR HEAVY STABLE PARTICLE IDENTIFICATIONC. Y. Chang

University of MarylandCollege Park, Maryland 20742

Searching for heavy stable particle productionin a new energy region of hadron-hadron collisions isof fundamental theoretical interest. Observation ofsuch particles produced in high energy collisionswould indicate the existence of stable heavy leptonsor any massive hadronic system carrying new quantumnumbers. I Experimentally, evidence of tts productionhas not been found for PP collisions either at FNAL2,3or at the CERN ·ISR4 for IS = 23 and 62 GeV respec,"·tively. However, many theories beyond the standardmodel do predict its existence on5a mass scale rang­ing from 50 to a few hundred GeV. If so, it wouldmake a high luminosity TeV collider an extremelyideal hunting ground for searching the productionof such a speculated object.

To measure the mass of a heavy stable chargedparticle, one usually uses its time of flight (TOF)and/or dE/dX information. For heavy neutral particle,one hopes it may decay at some later time after itsproduction. Hence a pair of jets or a jet associatedwith a high Pt muon originated from some places other.than the interacting point (rp) of the colliding beamsmay be a good signal. In this note, we examine thefeasibility of TOF measurement on a hea~ stableparticle produced in PP collisions at Is = 1 TeV anda luminosity of 1033 cm-2 sec- l with a single armspectrometer pointing to the Ip.

To measure the tillle of flight (TOrI for thestable heavy particles produced at large angles, weshall use the S = 1 muons as a reference. Figure I-ashows the TOF for a 100 GeV/c2 mass particle with50 GeV/c momentum in comparison with that of S = 1ref~rence particles. At 1 TeV C.M. energy and 1033cm- sec- l luminosity, the PP system collides at anaverage rate of -80 MHZ. It is not infrequent tofind a second event occurring within 10 to 20 nsecafter the current event took place. To distinguishthe TOF of heavy particle from that of the S = 1 back­ground arising from different collisions, we shallsample a set of TOF measurements for the particles atdifferent stages of the spectrometer. Figure I-bshows the time separations among heavy particles andthe S = 1 background with respect to the referenceS = 1 muons at various distances. It is clear thatthe TOF of the heavy particle is highly constrained.While at small angles, the acceptance of the spectro­meter will be severely reduced, because every par­ticle, light or heavy, moves with a high speed, onehas to depend on a nicely bunched beam and a verylong flight path for measuring the TOF of the heavyparticles. Figure 2-a shows the angular distributionof a 100 GeV/c2 heavy lepton produced by pp collisionsat 1 TeV. This result is obtained for PP + L+L- +anything via ordinary Drell-Yan process. When a TOFwindow cut (e.g. 3 < 8 (TOF) < 20 nsec at 6 m~ters

from IP) and a mass weighting factor [(Q2- 4ML )/Q2]1/2,are applied to the produced events, one noticed thatin Figure 2-b, the small angle enhancement of eventsattributed by the kinematic collimation of L± in theforward region is lost due to poor time separationbetween the signals and the S = 1 reference muons.In the 90° region, presumably, a few meters of pathlength would be sufficient to identify the heavyparticles. However, there are too many slow junksflooded in this region, one may not be able to dealwith it at such a high rate. An optimum angle forthe spectrometer axis to be oriented with respect tothe beam axis of the collider is probably 40°.6

We would like to thank Dr. R. G. Glasser forprOViding us many handy routines to generate variousspectra of Drell-Yan process at Maryland.

134

Footnotes and References

l. New and Unthinkable Ideas, SUllunarized by L. M.Lederman, Proceedings of the 1975 ISABELLEStudy, July 14-25, 1975 (BNL 20550) page 84.See also page 356 for detail calculations byChang and Samios.

2. D. J. Bechis, et al., Phys. Rev. Lett. 40,602 (1978). -D. Antreasyan, et al., Phys. Rev. Lett. 39,513 (J.977). -

3. J. A. Appel, et aI" Phys. Rev. Lett. ~' 428(974) .D. Cutts l et al.I Phys. Rev. Lett. Q, 363 (1978).

4. M. G. Alqrow, et al., Nucl. Phys. B97, 190 (1975).5. Beyond the Standard Model, G. L. Kane and M. L.

Perl (See it for references); Proceedings of theDPF Workshop on Future High Energy Physics andFaciIities, Snowmass, Co., July (1982).

6. In reference 2 (D. J. Bechis, et al.), we havedumped 109 protons per beam spill at 400 GeV ontoa 5 meter long stainless steel beam dump pluggedinside the sweeping magnet of the FNAL neutralhyperon beam. We found a 72% dead time for thedown stream p.w.c. read out system because ofthe existence of the high energy muons. At 45°c.m. of angle, we believe that we have enoughS = 1 muons for reference.

Figure Captions

Figure I-a. TOF for a 100 GeV/c2 massive stableparticle with momenta 10 and 50 GeV/cin comparison with that of a S=l bunchof particles.

Figure I-b. Time separations observed among heavyparticle and the S = 1 backgroundarising from different event. One cansample the 8(TOF) measurements atvarious locations of the spectrometer.

Figure 2. Angular distribution of stable heavyleptons produced via Drell-Yan mechanism.We assume IS' = 800 GeV, mL = 100 GeV/ c2,distance from IP is 6 meters.a) All L± 2 2 2 1/2b) Weighted by [(Q -4ML )/Q ] and in

the TOF window (3 < 8t < 20 nsec.)

)

)

0

7(Q..)

6

0 -...~5Q)

E-l=i4E

0 elL.3Q)()c:0-.!!! 20

0

o Pigure ha

o

A\ Ih

.1~ ~ j

~ All ~

.l~ ill l

.l!¥td 1 I~

1M J Wi

~lIJ I I I JJi I I I

6

7

-l'!~5Q)

E-f=i4EoIt 3Q)()c:

'*2C

o 5 10 15 20 25 30Separation of TOF (nsec)

35

, )

135

9Et

(

(

q-Z,,~naM

(;?;l.\~C<l)e

e-zo~naM

(~"vgC)Cf.)e

Ob09O~o06°9Of;oII~€

ISS9sa1:016119S1SSIOll£al~Ol:

II:~

aSI:SSI:I:£l:661:90SSI:SO~S

£set£E16Sao~

SI:~

I:~~

6St9£t€6tOIS£~.:I

~tS

19.:1a£.:IS6.:11:196~9

9t9€99089£69tilleta~£

.:19£I:a£66£91aseaosa

<l .....

xxxxxxxXxxxxxx~xxxxxxxxxxxxx*xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxXxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxXXXXXXXXXX)(XXXXXXXXXXXXiCXXXXXxxxxxxxxxxxxxxx+xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx++xxx+xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

SX+Ixxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx£xsxxxxxXxxxxxxxxxxxxxxxxxxxxxxxxxxxxX

xexxIx~xxxxxxXxxxxxxxxxxxxxxxxxxxxx+X+Xxxxxxxxxxxxxxxxxxxxxxxxxxxxx~6XXXXXXXXXXXXXXXXXXXXXXXXXXXX

£9XXUxxxxxxxxxxxxxxxxxxxxxxXXx+xxxxxxxxxxxxxxxxxxxxxX+Xxxxxxxxxxxxxxxxxxxxxx~Xxxxxxxxxxxxxxxxxxxxxx

+xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxXX++9xxxxxxxxxxxxxxxxxx

+xxxxxxxxxxxxxxxx+XXXXXXXXXXXXXXXX£~xxxxxxxxxxxxxx

XXXXXXXXXXXXXXxxxxxxxxxxxxxxxaxxxxxxxxxxxxIxxxxxxxxxxxx

XXXXXXXXXXXX+X9XXXXXXXXXXxxxxxxxxx£xxxxxxxxx

xxxxxxxxx~xxxxxxx+

xxxxxxxxxxxxxx~xxxxxx

xsxxxxxxxxxI:XXXX

+xxxXX6ax

S

d

y

e96~I

SI~91I~

~~

£~

'oe l€e

9t'6€~~

Is~a~

IS~S

£S'09'e9

I:: I~£

1

Sta£la~a

£a!06e696'66I~ol

I:; 1111~Il

LIIO~I

e~1

19~1~I

el'Iselael'I~I

~~I'Il~1!;t·

;~=~;;~;;;;;;I;;;:;;:;:;;:;;~~~;~~I::;;:;XxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxXxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxIXXxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx~

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx~xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxXXXXXIXXxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxIXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxXxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx~xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxIxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx~

I~xxXXXXXXXXXXXXXXXXXXXXXXXXXXXIXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXIXXXXXXXXXXXXXXXXXXXXXXXX~

XIXXXXXXXXXXXXXXXXXXXXXXX~

XXXXXX~XXX~XXXXXXXXXXXXI~XXXXX~XXXXXXXXXXXXX

XIXIXxxXXXXXXXXXXXXxxXXXXX~I~

IXXxxXXXXXIxxXXXX

xxXXX~X~~IXX

XxxXX

X~

u

c

c

(

(~

()

o

o

1033 _2 _1

OPERATION OF A HIGH-PT SPECTROMETER ARM AT cm secWITH PARTICLE IDENTIFICATION

S. AronsonBrookhaven National Laboratory, Upton, NY 11973

M. GoldbergSyracuse University, Syracuse, NY 13210

M. HolderUniversity of Siegen, D5900 Siegen 21, W. Germany

E. LohUniversity of Utah, Salt Lake City, UT 84112

1. What rate ~~abi~itie~ are required to copewith L = 10 cm- sec- ?

2. What segmentation is needed to deal with the

0particle densities expected in high PTjets?

3. Is the resulting device within the reach ofpresent technology?

o

o

I. Introduction

In the Proceedings of the 1978 Isabelle SummerStudy1 a high-PT spectrometer pair is presented as aprototype high-luminosity experiment. We considerhere an updated version of this apparatus with thefollowing questions in mind:

Section II presents the current version of the deviceand the expected rates. Section III discusses therates and segmentation of the components. Section IVcontains the results of calculations related to eventpile-up and triggering. The main conclusion, thatparticle identification appears to be quite feasibleat these rates, is discussed in more detail in SectionV.

Since 1/sin8 varies by only 10% over the aperture, wecan treat this flux as uniform over the solid angle.

For present purposes, it is assumed that thespectrometer performs tracking, momentum analysis,calorimetry and particle identificat~on of all parti­cles in the aperture. We use ISAJET predictions ofparticle multiplicity and density within PT = 100GeV/c jets produced in Is = 800 GeV pp collisions.The cross-section for production of 1007110 GeV/c jetsinto the spectrom~ter aperture is =10- mb (i.e.O.l/sec at L = 10 3cm-2sec-1). The spectrometer ele­ments are as follows (see also Table I):

Tracking

Interspersed in the detector are MWPC or driftplanes as shown in Figure 1. These are relativelysparse, as only straight-line tracking is involved.Sufficient information to resolve left-right and x-yambiguities is available at each wire chamber loca­tion. Momentum analysis is derived from the apparentmiss distance in the bend plane of the track and theinteraction point. 1

II. Spectrometer Components and Total Rates Calorimetry

The total rgte is estimated with the assumptionsLainel = 50 x 10 /sec, and dn/dy(ch) = 4.1 on theplateau at Is = 800 GeV. For the aperture given, wefind the rates

R = 30 x lOG/sec =Ry'ch

With a 30 kg-m field integral all charged particleswith PT < 0.45 GeV£c are swept out of the aperture.Assuming dn/dy dPT «exp(-6PT), actual ratesreaching the spectrometer are

We consider a single arm of the apparatus.Physics considerations will dictate the most efficientuse of the rest of the solid angle. Some choices arean identical arm in the opposite direction as in Ref.I, a larger acceptance arm with more modest particle10 and tracking, or an "energy-flow" device (Le.,mainly segmented calorimetry) covering as much of theremaining solid angle as possible.

Figure 1 shows the spectrometer arm. Theacceptance, as defined by the magnet aperture, is asfollows:

()

o

,)

AQ =0.9 steradiansAy = ± 0.5M/2Tl = 0.15

(1)

(2)

Electromagnetic (> 20 r.l.) and hadronic (> 6int.l.) is provided over the whole aperture, plus wideangle coverage for charged particles with PT down to2 GeV/c which are bent away from the spectrometeraxis.

Particle ID

1. Electrons and photons: The electromagneticcalorimetry is assumed to be finely enough segmentedin depth to reject hadrons relative to electrons atthe 10-3 level. Independent electron identification

3is provided by either transition radiation or synch-rotron radiation. 4 For this purpose the first track­ing chamber following the magnet is assumed to befilled with Xe for good X-ray detection efficiency.In the transition radiation case this would be a stackof several fiber or foil layers interleaved withXe-filled MWPC's.

2. Muons: The hadronic calorimeter is followedby 6 more interaction lengths of steel withinterleaved proportional drift tube arrays for muonidentification.

3. ~drons: A system of 3 ring-imaging Cherenkovcounters, provide Tl/K/p separation from about 1 GeV/cto over 50 GeV/c.

137

III. Component Segmentation

The charged particle flux given in Eq.(2) is notlarge by accepted standards for fixed-target spec­trometers of comparable area. For example, the first[Xe-filled] MWPC, with 2 mm wire spacing to keep gatewidths to ~ 50 ns, would be comprised of ~ 1000wires/plane. The charged particle rate is then ~

5kHZ/wire. The photon flux is about an order of mag­nitude higher so some care must be taken to minimizethe mass of detectors. If the Xe-filled chamber ispart of a transition radiation detector (~ 5% of aradiation length) then the flux per wire from conver­sions is ~ 3kHZ, still giving a very manageable totalrate. In this front-most chamber it is seen that seg­mentation is driven by the desire for speed, not bythe singles rates on individual wires. This is, ifanything, more true of the tracking chambers.

From the stand-point of calorimetry also, rate isnot the driving force behind fine segmentation. The

~~:~:~~~~ei~i~~eO~a~:~~:~~e~~:::a; ~a~u~a~m~r~: ~~:-electromagnetic and 10 x 10 cm or larger in thehadronic parts, respectively. The density of parti­cles in-high PT jets likewise gives a naturalangular segmentation for calorimetry on individualparticles. Using the angular correlations between jetparticles from Ref. 2, we find

A6em ~ .01 (all particles)A6had ~ .025 (charged particles)

In other words, for calorimetry on individualparticles in jets of PT ~ 100 GeV/c, 4 em/.Ol = 10cm/.025 = 4 meters is a natural distance betweensource and calorimeters. For the aperture of thisspectrometer we then get the following rates inindividual towers:

Electromagnetic:~104 towers; rate ~ 3kHZ/tower(all particles)

Hadronic~

~10 towers; rate 5kHZ/tower(charged particles)

For the ring-imaging counters there is also a

~:~u~:lr~~:m:~~e~!zeinb~~~: ~~s:h~Od::~~t~e~ist(~~~~about 100 per counter in this aperture) are appro­priate. This gives a small probability for 2 or morerings per cell and an average of 1 additional (soft)charged particle per cell within the jet. The totalcharged particle rate (assuming the photoionizationdetectors cover the aperture) is about 50 kHZ/cell andnot all of these produce Cherenkov light. This rateis further divided among the ~100) sense wires perdetector.

In all of the detector components associated withparticle identification there are 1-2 x 10 4 elementsand rates 1-50 kHZ/element. These numbers are notdaunting; in fact the detector is not different inscope from many existing fixed target experiments.

IV. Event Pile-Up

Table II gives the contribution to ET ofminimum bias event pile-up as a function of gatewidth. Due to the field, a small fraction of chargedparticles from pile-up events contribute to ET,while photons (with half the <PT> of chargedparticles) contribute in full force. Per pile-upevent about 0.6 photons and 0.15 charged particlesenter the aperture, contributing on average ~ 0.24 GeVto ET per pile-up event. From the table one can see

138

that even for a 100 ns gate pile-up at the ~ 1% level(i.e. 18 events in the gate 1% of the time) produceonly about a 4 GeV addition to ET' Since one mayexpect ET of 50 GeV or more into the aperture for a100 GeV jet down the axis of the spectrometer, theeffect on the trigger is manageable. One will be ableto do even better off line. Event pile-up in theCherenk6v counters is potentially the most seriousproblem for the particle ID components, because of therelatively coarse segmentation (100 cells) and thememory time (~1 \lsec) for the TMAE photoionizationdetector. At 1 \lsec the mean pile-up is 50 events;these contribute < 0.1 charged particle per Cherenkovcell and so constitute less of a problem than that ofparticle density within jets mentioned above.

V. Summary

The technologies employed in the particle identi­fication components of this detector are either avail­able or very close at hand. In particular, rate andparticle density are not issues of fundamental impor­tance in a limited solid angle apparatus such asthis. The price paid for this rather intensive parti­cle identification and analysis over ~ 1 steradian isin numbers of detector elements (i.e. complexity).However detector systems with thousands of calorimetercells and tens of thousands of wires are already inroutine use. Extending a system such as this to covera larger $ range and a somewhat larger y range(centered around y = 0) is only a matter of cost andcomplexity.

This research supported by the U.S. Dept. ofEnergy under Contract No. DE-AC02-76CH00016.

References

1. S. Aronson, et al. Proceedings of the 1978 SummerWorkshop, BNL 50885, p 278.2. F.E. Paige and S.D. Protopopescu,_"ISAJET: A Monte

Carlo Event Generator for pp and pp Interactions,Version 3," BNL 31987 (Sept. 1982).

3. M. Holder and E. Loh, these Proceedings.Transition radiation is presented there in adifferent context; see references containedtherein.

4. S. Aronson, these Proceedings and referencescontained therein.

5. M. Goldberg and D. Leith, these Proceedings andreferences contained therein.

,)

)

o

o

o

o

Component

A. Tracking Chambers

B. Calorimeters (e-m)

(hadron)

C. Particle ID (e)

(IT/K/p)

Table I. Spectrometer Components

II of Elements

5 x 101t wires

1.5 x 101t towers

2.5 x 103 towers

4 x 103 wires

1.5 x 103 wires

300 cells

Comments

MWPC's or minidrifts(2-4 mm drift)

Liquid argon with pad readout,lead glass bars with interleavedpatch chambers, etc.

Iron-scintillator with waveshifter bar readout.

Transition radiation modules(carbon fibers or plastic wool),or synchrotron radiationdetector.

1 m iron with interleavedproportional drift tube planes.

3 ring-imaging Cherenkovcounters: Liquid Freon (2 em),Isobutane (50 em), Argon (150em). TMAE photoionizing gasdetectors.

0Table II. Event Pile-up Contributions to ET

At Gate .!! n(CL".005) lIET

20 ns 1 4 0.6 GeV() 40 2 6 1.2

100 5 11 2.4200 10 18 4.0 I

P2 JL

T2~ .----PI~[b?l em

1j C2 C3CI .-=

(J

BEAMPIP.~ .~.-Ht-.Ilt1--ltH---

u

Fig. 1. Schematic representation of the high-PTspectrometer. The components discussed in the textare labelled as follows: E = electron identifier;Ti - tracking chambers; Ci = ring imagingCherenkov counters; em - electromagnetic calorimeters;h - hadronic calorimeters; Pi =proportional drifttube arrays; p = iron muon filter.

139

MUONS AND ELECTRONS IN GENERAL

L. NodulmanArgonne National Laboratory, Argonne IL 60439

J. BensingerBrandeis University, Waltham MA 02254

We address the problem of simultaneouslyidentifying electrons and muon~3in_~ ge~rral purposedetector at a luminosity of 10 cm sec • Thosediscoveries and important measurements which arereasonably well predicted now are under. way and likelyto be fairly complete before turn on. The signaturesof new effects to be found at high rates are not wellpredicted, and for flexibility it may be necessary tolook simultaneously for some combination of jets,missing E electrons, and muons. This leadsimmediately to an open geometry with magnetic trackingand calorimetry. At high luminosity, getting outtrigger information quickly is a prime concern. Notethat if a given signature requires isolatingindividual events then even for an optimistici~tegratiou time of 20 ns, the optimal luminosity isabout 2xlOj2cm-Zsec-l • We have not had theopportunity to be very specific in design or toconsider the extended momentum range implied by 10-20TeV collisions.

In order to be specific, we have assumed tdetector which is a variant of the CDF design.Detailed calorimetry design, and solving the problemsof magnetic tracking at high rates are left toothers. We assume a solenoidal detector with an openflight path of 1.5 m at 900 and calorimetry of 100radiation lengths corresponding to 1.5 m of iron2 m thick. The calorimeter is separated with theelectromagnetic portion preceding the hadronic.The electromagnetic calorimetry is assumed to be ina tower geometry with separate front and back segmentswith x-y strip chambers near shower maximum.

The mu~n system we consider is a variant of theCDF system. We add scintillators with similarsegmentation to the calorimetry in order to providefast trigger information and in order to eliminatecosmics immediatly in a situation where, in effect,there is always an event. The system involvesaccurate tracking in azimuth to enable a fastprojection to the beamline to give an implied PTcut. This reduces decay background and cuts outshower splatter. The increase from 1 (CDF) to 1.5 mof iron equivalent reduces both splash andpunchthrough, with no claim to being optimal.

The dominant source of background for muons is+ +decay in flight of ~- and K-. Extrapolating from a

CDF study by G. Ascoli,3 the decay probability at 5Gev/c PT is 1.4% varying as l/PT' An effectivethreshold of 10 GeV/c PT should give a trigger rateof about 1000/sec with overall misiden~ification

probability improving from about 7xlO- at 10 GeV/c.Lesser backgrounds are from punchthrough, about2xlO-4 and splash. The splash probability incr~asesfrom 2~10-2 at 10 GeV/c to 1.3xlO-l at 30 GeV/c.Splash is readily removed by tracking, and bothpunchthrough and splash make little contribution tothe rate seen by the muon system. The 1.8 GeV/c PTcut given by the 1.5 m of iron will keep theinteraction related muon system counting rate down toabout 10 /sec, roughly comperable to cosmic rays.Radiation damage is no problem, but the muon triggerruns much too fast to be used alone.

140

A fast coincidence, strobed by beam/beam countersor whatever, is available for muons for an initialpipelined deadtimeless trigger level. This wouldrequire an outside scintillator' corresponding tomimimal calorimetry signals. Note that this tends toantiselect muons in a jet. This fast informationshould be sufficient to allow reasonably long driftsin muon tracking. Staggered pairs of wires in drifttubes, aligned to the beamline,can be put into timecoincidence corresponding to an azimuthal impactparameter. The drift time and coincidence time andpossibly charge division ratio can be made availablefor sample and hold even if no coincidence triggerpulse is produced. Perhaps central tracking canproduce a stiff track signal in rough alignment, alsoon the time scale of a few microseconds. At a stillhigher level, a detailed track match may be required.

The details of the system are dominated bymultiple scattering considerations. The rms projectedimpact parameter to the beamline at 10 GeV/c PT is5.6 cm. The softness of the'resulting PT thresholdcan be aided by the stiffness required of the track.The rms misalignment in azimuth of the inside trackprojected to the back of the calorimeter with theoutside track is 1.9 em at 10 GeV/c. This gives anul timate limit to pic~ing out a muon in a jet,although the few xlO- rejection may not be generallysufficient.

The electrons are identified by a combination ofmomentum measurements of incoming track in centraltracking chambers, proper electron signature in theelectromagnetic calorimeter, and small energydeposition in the hadron calorimeter. In a CDP test asetupS similar to the one described here, and assuminga central tracking momentum measurement ofop/p=.002 x p (GeV/c) achieved a pion rejection facto~

of 7xlO-4 for 30 GeV/c electrons and pions, with a 70%effieiency for electrons. This level is sufficientsince there is no point pushing this beyond the levelset'by Dalitz decays and photon conversion.

Overlap between a charged pion and a photon can beeliminated if position resolution of both the chargedparticle track and the shower are better than 1 cm ata radius of 1.5 meters. In that case even in a jet ifthe particle density is 100 times background (assumedto be 5 charged tracks and 2.5 neutral particles perunit of ~seudorapidity) the probability of an overlapis ~ 10- •

Triggering involves several levels ofinformation. At the first level pulse height fromeach calorimeter can be compared with a predeterminedPT' A second level can identify clusters and compareinformation between electromagnetic and hadroniccalorimeters. A third level could match stiff tracksfrom central tracking with calorimeter towers.

Time limitations prevented more than a super­ficial look at forward muons, but between say 35 0 andhowever small an angle may be tolerated~ a toroidsystem such as has been proposed for CDF can givesimilar rejection and fixed op/p of less than 20%.

)

)

)

()

o

o

o

()

o

u

Although we did not give adequate considerationto 10 TeV collisions. the problems are apparent.How to measure even the sign of the high end of theavailable lepton momentum spectrum is a problem.Also. more dense jet cores will make picking outleptons even less likely.

In conclusion. a muon and electron system similarto that Of CDF can ~rovi~e siygle lepton tag rates oforder 103 /sec at 10 3 cm- sec-. The muonmisidentification probability is a few x10-3 above 10GeV/c. dominated by decay. This would be improved byhigher quality tracking allowing shorter flight paths.probably to the detriment of electron triggering.

References

1. CDF-Design Report. CDF-111 (unpublished).

2. Ibid. section 6.

3. G. Ascoli. CDF-57 (unpublished).

4. Gabrial_ and Bishop. NIM 155. 81 (78). quotedin CDF-15!.

5. L. Nodulman et. 81 •• NIM 204. 251 (83) andR. Diebold et. al •• CDF-106 (unpublished).

141

LARGE SOLID ANGLE SOLENOID SPECTROMETERWITH PARTICLE IDENTIFICATION BY DE/DX

David R. Nygren

Lawrence Berkeley Laboratory(J

PARTlCLB IDBNTlFICATION

20 Bem atm/ . 4.0 ,), ,,"

"10,

E/0.5cm atm

"~

5 5%4

£:3 ,)'"~ ,

e 2.;;:QI

0

""1 ,/ // X

" ISIS 1 ,

~", , 12%" "0.5 " 15%"

30 50 100 200 300 500 1000Number of samples

Summary

In this note we consider a brute force approachto achieve simultaneously useful particle identifica­tion for hadrons and electrons, good momentum resolu­tion and 100 nsec sensitive time apertures. The con­figuration envisaged is a cylindrical gas volume (withsolenoid magnetic field) sampled in entirety by axial­wi re drift ce 11 s instrumented for pul se hei ght measure­ments as well as timing. The 100 nsec aperture impliesa drift length of 1 cm or less, depending on gaschoices and on the cell geometry.

The conventional lorel about dE/dx implies atrade-off between number of samples and sample thick­ness for a given resolution (Fig. 1). To achieve asatisfactory level of particle identification it isnecessary to obtain at least 6% FWHM in the usual trun­cated mean estimator. For this performance, a choicecan be made between the extremes of 500 samples of atotal of 4 meter-atmospheres and 100 samples of 8meter-atmospheres. Minimizing the physical size of thedetector pushes the choice toward the large number ofsamples. With the further choice of inner radius of0.5 meters and outer radius of 4.5 meters, the 500sample configuration implies nearly 5 x 10 5 individualcell s.

This number is perhaps a little uncomfortable, soanother tactic can be considered: reduce the numberof cells and wires by-2, and pressurize to regain therequisite product of atmosphere-meters. An amusingalthough potentially very practical by product ofpressurization is the possibility of balancing therather formidable force on the endcaps due to the wiretension: -200 tons. This possibility implies thatwire stringing would be done with very low tension;pressurization would then provide most of the wiretension. Thus the cylindrical structures would needto be equipped with a bellows to permit the endcaps tomove about-5 cm or 1/2-1% of the total wire length.

The press~rization needed is about 1/3 of anatmosphere to· provide the balancing force. The combi­nation of 250 samples and 4. x 1.3 = 5.2 meter-atmo­spheres falls handily on the 6% resolution contour. Itis worth pointing out that R&D effort to ensure theabsence of snapping wires on first pressurization is avery good investment! Pattern recognition should bea tractable problem because of the very high degree ofsegmentation.

References

Figure Caption

Figure 1. The ionization resolution (% FWHM) of amultisampling detector filled with pure argon, calcu­lated with the photo absorption-ionization model des­cribed in reference 1. The dashed lines are Loci ofconstant sample thickness.

)

1. W.W.M. Allison and J.H. Cobb, "Relativistic Charg­ed Particle Identification by Energy Loss", AnnualReview of Nuclear and Particle Science, Vol. 30 p. 253(1980) .

142)

()DETECTING HEAVY QUARK JETS

II. A Heavy Quark Detector

It is natural to imagine repeating the detectorarm shown in Fig. 1 four times to achieve fullazimuthal coverage, particularly to be able to detect

The inner vertex detector is shown (actualsize) in Fig. 2. It consists of four planes of highresolution position sensing elements. We give adetailed discussion of this device in Sec. IV. To beeffective, this detector must be extremely close tothe beams. We assume that the first plane is 1 cmfrom the beam axis. We have discussed this withISABELLE accelerator physicists and it does not seemto be a fundamental problem provided the chamber isplaced either above or below the beams (i.e., not inthe horizontal plane). The chamber would be in arough vacuum separated from the beam by a thin skin« 250 lim) of titanium. It would have to be retractedduring stacking and acceleration of the beams.

CBA will also produce enormous numbers of T 's(>109) compared to 105 at LEP. For most+decay~ theywill be impossible to distinguish from IT or F"" •Nonetheless thjre are s~e decay modes worth lookingfor, such as ll- e+e- or ll-<I>. These decays areforbidden in the standard model but could be as largeas 10-4 with horizontal gauge symmetries, andefficiencies as low as 10-4 will be sufficient toreach that level. Finding a signal into such a modewould have momentous impact.

In this exercise we examine the performance ofa detector specifically configured to tag heavy quark(HQ) jets through direct observations of D-mesondecays with a high resolution vertex detector. To op­timize the performance of such a detector, we assumethe small diamond beam crossing configuration asdescribed in the 1978 ISABELLE proposal,(2) giving aluminosity of 1032 cm-2 sec-I. Because of the verylarge backgrounds from light quark (LQ) jets, mosttriggering schemes at this luminosity require high PIleptons and inevitably give missing neutrinos. If al­ternative triggering schemes could be found, then onecan hope to find and calculate the mass of objectsdecaying to heavy quarks. A scheme using the highresolution detector will also be discussed in detail.The study was carried out with events generated bythe ISAJET Monte Carlo(3) and a computer simulationof the described detector system. Many of the re­sults that follow were presented at the DPF SummerStudy at Aspen.(4)

In Fig. 1 we show what a "modest" size heavyquark detector (HQD) may look like. It needs basi­cally five sections: an inner vertex detector asclose to the beam as feasible, a charged particle de­tector consisting probably of drift chamber planesinterleaved with transition radiation detectors tohelp identify electrons, an electromagneticcalorimeter, a hadronic calorimeter and a II detector.Because the expected angular spread of jets is large(> ± 5°) we requ ire ca lorime ters covering <I> = ± 45° andin rapidity y =±1.

G. Benenson, L.L. Chau, T. Ludlam, F.E. Paige,E.D. Platner, S.D. Protopopescu, P. Rehak

Brookhaven National LaboratoryUpton, New York 11973

level) with 10-7 efficiency to observe m1x1ng. Thesimplest way to accomplish this is to detect equalsign leptons that come from semi-leptonic B decays inback-to-back jets.4

I. Introduction

The ability to identify c, band t jets and toseparate them from the much more abundant u, d, sandgluon jets could lead to a wealth of new physics. Avery promising technique is the possibility ofmeasuring vertices near the interaction region witha resolution of the order of 10 lim, making one sensi­tive to decay lifetimes of a few x 10-13 sec and sotagging charmed particles and possibly T 's.

The immediate result, of course, would be themeasurement of d2 a /dPtdy for such jets. This couldbe compared with QCD predictions, and it would alsoshine some light on the question of the amount ofintrinsic ce, bb and tt in the qq sea of the pro­ton. Detailed study of these jets may allow us toseparate t, band c jets. At sufficiently high Pt(high compared to the t mass) the relative cross sec­tions are expected to be 1/1/1; significant devia­tions from the expected ratios could lead to theuncovering of new phenomena.

Another interesting possibility is to attemptfull reconstruction of B mesons. Given the largebackground and multiplicities this may seem utopic;however, at L'" 1032 cm-2 sec-I, close to 1010 Bmesons are produced in a year (10 7 sec). This is tobe compared with <105 expected at LEP. So any schemethat selects b jets with >10-4 efficiency produces>106 events for further study. f good example of anInteresting decay mode is It- + It" e+e-, which isexpected from the standard model to have a branchingratio "'10-6 • While such a small branching ratio maype impossible to observe, horizontal gauge symmetrymodels(l) can raise the branching ratio to "'10-4,which may be possible at CBA but out of reach at LEP.A branching ratio substantially higher than 10-6 can­not be explained in the standard model and would bea very important discovery. Since in this case thedecay does not involve charm, one could look for itin the jet accompanying the tagged one; having anidentified heavy quark jet will give a tremendous re­duction in background. The study of two or more jetswill require the capability of doing electromagneticand hadronic calorimetry over a large solid angle,but the charmed particle tag may be more restrictivein solid angle without a very detrimental effect oneffic iency.

Another observation of interest is that ofBOB" m1xmg. An initial state of BO (B") can havea finite time-integrated probability of becoming aii" (Bo). The mixing is maximal when there is equalprobability for the final state to be BO or B"irrespective of the initial state (as is the case forIf K" ). There are reasons to be lieve that, althoughthe mixing for DOff and t'T' is small, the mixingfor BOB" could be maximal. (2 ,3) In this case onefinds:

the factor is 1/10 if only Bdo or Bso have maximalmixing and 1/5 if both do. Thus, one can expect thatbetwee~_20% and 40% of_back-to-back b jets end up asbb or bb rather than bb. It is sufficient to sepa­rate bb from bb or bb jets clearly (at the 10%

o

o

()

o

()

,)

u

)143

more than one jet. However, for the reason givenabove, it does not seem feasible to achieve fullazimuthal coverage with the vertex detector.

part, 500 MeV for the hadronic part) the rate due topile-up of "minimum bias" events is suppressed wellbelow the jet trigger rate for ET ~ 5 GeV.

Muon Trigger

For calorimeter-triggered events, the mean num­ber of charged particles entering the calorimeter is~9 (see Fig. 3), of which 18% have P> 2 GeV/c for acalorimeter threshold of 30 GeV.

Given these numbers we can construct the follow­ing table of minimum background rates for a muon trig­ger in coincidence with the calorimeter trigger (at,luminosity = 1032 cm-2 sec-I):

In Fig. 4 we show the momentum spectra ofcharged tracks in the detector for the 15 GeV thresh­old setting. Note that the leptons shown in 4b and4c come from both B meson and D meson decay. (Fig.4b also includes Dalitz electrons.)

)

)

)

)

9

.5

62 sec-1

'IT ,K Decay Muons

.05

1.5

Punch-Through

10 sec-1

15

30

The probability for a pion to traverse thecalorimeter without interacting (punch-through) andthus fake a muon is .0025. The probability for apion (kaon) to produce a decay muon before absorptionin the calorimeter is .029/Prr (.22/PK). We take aK/IT ratio .2, and assume that only particles with P> 2 GeV/c will produce a fake muon trigger bypunch-through or decay.

Thus the trigger rate can, in principle, bereduced by a factor of 30-40 from that given by thecalorimeter alone (Table I). To achieve this, how­ever, we must deal with additional severe backgrounddue to leakage of shower particles out the back ofthe calorimeter. This can be reduced to levels compa­rable to those in the table above by the followingtwo means, both of which require a fast processor ifthey are to be implemented at the trigger level:

The calorimeter trigger rates for our 3 chosenET thresholds are given in the first line of Table I.The corresponding rates for b-quark jets among thesetriggers is given on the second line. It will beseen that the ratios of triggers/b-quark jets are~2000, 800, 600 for ET thresholds of 8 GeV, 15 GeV,30 GeV, respectively. For the 15 GeV threshold wehave ~l b-jet/sec among the triggers, and we need tobring the trigger rate down by another factor of10-20 to reach a reasonable rate for data recording.

For the detector configuration shown in Fig. 1,with a 6 absorption length calorimeter, the energyloss suffered by a muon is 1.5 GeV. Thus a triggeron muons traversing the calorimeter is limited to P~

~ 2 GeV/c, or about 70% of the muons from Band Dmeson decay (see Fig. 4).

Fig. 3 shows the multiplicity of charged tracksinto the detector for minimum bias events and for atrigger threshold of 15 GeV. These include onlytracks which traverse all four planes of the vertexdetector. Because of the small diamond size the frac­tion of tracks which do otherwise is small: For ET> 15 GeV the mean number of hits in the first planeof the vertex detector is 9.8.

!T Threshold

8 GeV

Calorimeter Trigger

The philosophy here is to implement a totalenergy (ET) trigger with a calorimeter, and examinethe response of the vertex detector for events thustriggered -- i.e., the effectiveness for tagging Dmeson decays. As we shall see, this effectiveness in­creases with increasing momentum of the trigger jet.In order to maximize the yield of B mesons, however,we wish to keep the ET threshold as low as possible.

The detector sketched in Fig. 1 will have excel­lent tracking capability --the vertex detector alonemeasures track angles to an accuracy $1 millirad.It is designed to handle high densities of low momen­tum tracks. Thus a very modest magnetic field willsuffice for adequate momentum measurements. It maybe argued that no magnetic field is necessary,supplanting momentum measurement with calorimetricenergy measurement. However, some momentum informa­tion will be useful for evaluating multiple scatter­ing errors in the precise vertex measurements; itwill be helpful to have measured muon momenta; and in­formation on the signs of tagged leptons may be cru­cial in some studies. For the present study we havenot included the effects of a magnetic field in ourcalculations.

III. Trigger

With the calorimetric trigger alone the majorbackground is due to light quark jets. Therefore weconsider a "low" ET trigger (15 GeV in thecalorimeter) with lepton triggers in coincidence.Since the leptons of interest are relatively soft,triggers on leptons are not straightforward, and willrequire some real-time processing. The choice of 15GeV ET threshold is guided by our estimates of leptontrigyer capability, for the luminosity of 1032 cm-2sec-. We also consider a "high" ET threshold of 30GeV, and a "low-low" threshold of 8 GeV.

The hadron calorimeter is 1.5 meters from theinteraction diamond and is 3x3 m2 in area. Fine seg­mentation is important. It is subdivided into 20x20cm2 towers (~250 cells). The electromagnetic showerdetector (approximately the first 10 radiationlengths of the calorimeter) is subdivided into 1000cells. The full calorimeter is 6 absorption lengthsdeep.

Since we are triggering at relatively low ET'the energy resolution of the calorimeter is a criti­cal determinant of trigger rates. A calorimeter withhadronic energy resolution GE = .alE (typical ofiron/scintillator devices) would give a trigger rateseveral times the rate at ET = 15 GeV. ~sing the"best" hadron calorimeter, with GE '" .YE (e.g.uranium/scintillator), the trigger rate will be 30­50% higher than the true rate. For the resultspresented here we do not include the effect ofcalorimeter resolution on the rates.

At L = 1032 cm-2 sec-1 we have 6xl06interactions Isec, most of which send something intothe calorimeter. Hence the mean time between eventsis ~200 nsec for a calorimeter trigger. For reason­able calorimeter gate widths (~100 nsec) this resultsin a substantial contribution to the trigger rate dueto pile-up. If, however, we require that each cellof the calorimeter have a minimum energy beforeadding it to the trigger sum (250 MeV for the EM

144

o

o

o

()

i) Require a mLnLmum ionLzLng signal in thecalorimeter segments traversed by the''muon.''

ii) Require that the position and angle of thetrack exiting the calorimeter match, withinmultiple scattering limits, the trajectoryof a charged particle incident on thecal.or-imeter.

For lack of a detailed study (but guided by a similarstudy done by the R807 Group) we take the muon trig­ger rate to be roughly twice the "minimum" rate dueto punch through and decay. Th.is crude estimate ofrates is entered in Table I for calorimeter plus muontrigger. Note that the addition of a muon trigger isuseless for ET> 30 GeV (for this detectorconfiguration) and in fact the signal-to-backgroundratio is not greatly improved by the muon trigger forany of the three ET threshold settings.

Electron Trigger

For the e-trigger we choose transition radiationdetectors (TRD) because of the high degree of segmen­tation required and the desire to separate electronsfrom hadrons over a wide range of energies with a com­pact device.

thickness, with charge collection time $ 50 nsec fora 300 ~m thick detector. This, coupled with theintrinsically high degree of segmentation, leads toexcellent rate capability. The lifetime of such de­tectors has been measured 'to exceed 1Q14/cm2 forfluxes of relativistic' charged particles. For the de­tector geometry discussed here, the innermost planeof the vertex detector would see an integrated fluxof .r2xl012 charged particles/cm2 in a year of runningat a luminosity of 1032 cm-2 sec-I. The effects ofbackground radiation need more study: slow neutronsand heavily ionizing particles are much more damagingthan minimum ionizing particles. Tests carried outwith silicon surface-barrier detectors placed nearthe beam crossing at the ISR indicate that these back­grounds will not be a problem, however.

The most serious technical difficulty forimplementing these detectors in colliding beams isthe extremely high density of output connections ina situation where we need to cover (relatively) largearea. An output on each strip implies thousands ofconnections per centimeter along the detector edge.One way out, which is being studied at BNL, is to useresistive charge division to interpolate the positionamong groups of strips. An analysis by V. Radeka(8)gives the following formula for the optimal number ofoutputs as a function of detector characteristics:

where

170 + 317 + 462 + 640", 1600

For the detector illustrated in Fig. 2 this gives,for the four planes,

IV.2

IV.l

_ AN - 47 w2/ 3

strip length

1.4xl0-2 cm1/ 3 for silicon detectors

number of signal outputs.

detector area

resolution

detector thickness

w

t

A

N

r,2/3a

For Ox 10 ~m and t = 300 ~m (a standard waferthickness for semiconductor devices) one obtains

We estimate that .rl/20 of the calorimetertriggered events will have a conversion electron withp > 1 GeV/c. This background can be virtuallyeliminated by requiring that the electron track ap­pear in all four planes of the vertex detector, butit is unlikely that this can be accomplished at thetrigger level. Some improvement can certainly be hadwith an on-line processor, e.g., by requiring theelectron track to appear in the first of the two TRDmodules. Guided by these considerations, the entriesin Table I are based on the assumption that the elec­tron trigger reduces the calorimeter trigger rate bya factor of 30, and is 90% efficient for electrons.

We assume a total length of .r80 cm of TRD,subdivided into two separate modules as shown in Fig.2. Each module is made up of.r5 planes of radiator(Li foils or Carbon Fibers), each with MWPC readout.Such a device should be capable of reducing the n/erate by a factor of.r 103 , with good efficiency forelectrons of momenta ~ 1 GeV /c. (We are specifi­cally guided by the configuration tested in Ref. 5).Such a device, coupled with the EM calorimeter willgive a very powerful electron tag. At the triggerlevel, however, the counting rate will be dominantedby conversion electrons, as the TRD thickness will be.r5% of a radiation length.

u

IV. Vertex Detector total readout channels.

,)

.J

The use of semiconductor detectors as very highresolution tracking devices in high energy physics ex­periments has been a subject of intense developmentover the past few years.(6) A number of such detec­tors are being constructed (one is operational)(7)which give .rIO ~m space-point resolution in the hightrack density environment of heavy quark searches atfixed target ma'chines. These "micros trip" detectorsconsist of silicon wafers whose surface area isfinely subdivided into strips; each strip may be readout as a separate detection element. The strip-to­strip spacing is typically 20-50 ~m. This gives thecharacteristic position resolution, which is achievedalong one coordinate of the detector.

The signal charge for a mLnLmum ionizing trackis .r8xl04 electron-hole pairs per mm of detector

Note that our detector measures one coordinateonly. We take this to be the azimuthal coordinate.To obtain space points in 2 dimensions (using stripdetectors) would require at least 3 times as many de­tector planes. As we shall see, the cost in multiplescattering errors would outweigh the usefulness ofsuch a scheme. Thus the proposed detector really pro­vides a tag for charmed particles and not a fullyreconstructed vertex.

It should be pointed out that other schemes forrealizing semiconductor devices as track detectors ofthis type are being considered and developed by vari­ous groups. For instance, CCD devices could, in prin­ciple, solve both the "connection" and the 2­dimensonal readout problems; however, these (as trackdetectors) are in avery early stage of development

) 145

and are fundamenta l.1y unsui ted for the ratesencountered in CBA experiments. A device with stripelectrode geometry but serial readout (hence fewconnections) is being developed at the University ofPittsburgh.(9) This is based on a very interestingtechnique in which signal charge is stored in shallowimpurity traps in the i-region of a PIN diode detec­tor at cryogenic temperatures. These and other devel­opments may point the way to better devices than thatproposed here. For the moment, we confine our anal­ysis to the "known" technology of micros tripdetectors.

For a 4-plane detector, the error (a v ) in theextrapolation of a track to the vertex position isgiven by (see Fig. 5):

c/ '" ai +(a x Ll)2 +(8.7 x 10-4 )2v 4 73 L2 B P (GeV) Ll

The first term is the position resolution, the secondis the effect of the angular resolution and the lastis the multiple scattering contribution. For ax = 10~m, Ll = 1 cm, L2 = 2 cm:

Multiple scattering dominates for p$ 2 GeV/c eventhough we have been at some pains to place the detec­tor as close to the beams as feasible (Ll = 1 cm).Nonetheless, for the average momenta of tracksthrough the detector in ET-triggered events (Fig. 4)we obtain av '" 10 ~m.

The average multiplicity of charged tracks intoour detector, for calorimeter triggered events, is ~9

(Fig. 3), and the distribution ranges up to about 20.From these tracks we must reconstruct the primary ver­tex and determine which, if any, tracks originatefrom a secondary decay vertex. For our Monte Carlosimulations we conservatively included only nt decaysand used a lifetime for them of 8xlO-13 sec. The dis­tribution of projected miss distance, ~, is shown inFig. 6. The criteria for resolved decays were as fol­low: (Note that a v is momentum-dependent.)

i) If only a single decay track is visible, itmust have

~ > max 000 ~m, 4 a v )

ii) If two decay tracks are visible,

~ > max (50 ~m, 4 a v)

iii) If more than two decay tracks are visible,

~ > max (50 ~m, 3 a v )

iv) If a decay lepton is tagged,

a > max (50 ~m, 3 a v)

for all visible track multiplicities.

The results are shown in Table II for variousdecay topologies and ET thresholds. Roughly 40% ofthe charged D mesons entering the vertex detector areresolved. This means that ~20% of the triggered HQjets have a visible decay. The rates foraccumulating events with resolved decays are shown inTable I.

Cut (i) is the least restrictive; a 4 a cutkeeps only ~1/16000 tracks. However, since the aver­age number of tracks per event is 10, the probabilityof a fake decay is ~1 per 1600 events. Since thecuts reduce the signal by a factor of 6 (a factor of2 by requiring nt and another factor of 3 from life­time cut) the overall improvement in signal-to­background ratio (S/B) is ~200. From Table I we cansee that once a lepton trigger is used the vertex de­tector cuts increase SIB to 1/2. Off-line the leptonidentification can be substantially improved, particu­larly'for electrons by correlating TRD and showercounter information. Therefore, it is possible to re­duce the LQ jet background to the 10%-20% level. Theonly significant background to b jets at this pointis from c jets.

V. Physics with Two Detectors

Two detectors, like the one shown in Fig. 1,placed opposite each other, offer very interestingphysics possibilities. In Table III we give therates for various triggers using both detectors. Thetriggers require a minimum total energy (ET) deposi­tion in the calorimeter and in triggers, 3, 4, 5there is also an electron with Pe > 1.0 GeV/c (two intrigger 5).

Consider, for example, trifger 4, which has acomfortable trigger rate (8 sec-). After a vertexcut on the jets in the detector with an electron trig­ger the signal-to-background ratio SIB is 1/1. Thiscan be reduced by at least another factor of 5 to 10by additional off-line requirements on the electron,such as matching momentum measured in the drift cham­bers with energy deposition in the electromagneticcalorimeter. This gives an unbiased sample of b jetsin the opposite detector with a background of c jetsonly (roughly 2 to 1). In 107 sec we have then 105heavy quark jets of which 3xl04 are b jets. This isto be compared with ~105 b jets that are produced atLEP in a similar period of time before ~ selectionsare made.

As will be discussed in the next section, it ispossible to reduce the data rate by a factor of ~10by using information on the multiplicity and patternof space points in the vertex detector at the triggerlevel. If so, then triggers 2 and 3 can ultimatelyprovide as many as 106 tagged b jets.

Trigger 5, which requires 2 electrons in one de­tector, is not very efficient for b jets, but it isof great interest for searching for rare decays, suchas B± -+- K±e+e+ or T± -+- ~±e+e-.lf all gt decayed toR±e+e- we would have 2.2 events/sec (trigger 1).Requiring Pe > 1.0 GeV/c and PK > 1.5 GeV /c reducesthe rate to 0.20 sec-1 so after a vertex cut on thedetector opposite to the one triggering on 2 elec­trons we would collect ~3xl05 Kee decays in 107 sec.The only significant background at this level are thefew percent of b jets which produce 2 electrons.After requiring PK > 1.5 GeV/c (no K identification),Pe > 1.0 GeV /c and mCKee) = m(B) ± 50 MeV the b jetbackground is down to$ 1/40,000. Thus a branchingratio for gt -+- K+e+e- > 10-5 could be observed. Aslightly higher branching ratio could be seen for T

-+- ~ee (using T'S from Band F decays).

In the search for BB mixing we can use trigger3 with the additional requirements of observing alepton in both arms. Because of the chain decaysb+c+-J!, and b+ccs with c+-J!, the leptons observed couldbe of equal sign even in the absence of mixing. How­ever, the leptons that come from b+J!, have substan­tially higher momentum than the secondary leptons, as

)

)

)

)

)

146

o

b) Algorithm

c) Backgrounds

For the trigger consider three planes withhits Xli, x2k , X31 • If those hits lie on a straightline they must satisfy the requirement

There are two effects that can increase thebackground from LQ jets: Multiple scattering tail(hard scattering) and multiple events in the time win-

Vr.lx k = a x i + a3x31

2 1 1

The second part of the algorithm makes useof the fact that the beam is all contained within±200 J.l. The intersection xnk of the selected linesthrough 5 or any small number of planes in the inter­action diamond can be quickly computed using equationVI.l w~th the appropriate constants al n , a3n • The av­er!ge xn = ~l xnk/N and a pseudo chisquare X2E(xn - xnk)!-can then be calculated for each plane.

The third part of the algorithm is tochoose the plane n for w~ich X~ is smallest and keepthe event if 50 J.l < maxi xn - xn I < 1000 J.l.

The algorithm was checked with Monte Carloevents (generated with ISAJET) and seems quite effec­tive in rejecting LQ jets by a factor of 50-100 whileaccepting ~20% of the HQ jets. To eliminate confu­sion one must require that two readouts with signalsbe followed by at least one with none; this impliesthat the hits must be separated by at least 300 J.lbefore they can be used in the trigger. Table Igives the rates for LQ and HQ with one and twodetectors. In Appendix A we give an example of howthe previously described algorithm can be implementedso as to keep the time required for calculations toless than 2 msec.

where al and a3 are fixed constants. F9r all pairs,i,1 we can calculate a corresponding X2 1,1. The dis­tribution of IX21,1 - X2k! is sgown in Fig. 9. It isobvious that imposing IX2 - X2 1,11 < 70 J.l will re­move all the wrong combinations. Note that inside amagnetic field of 5 Kgauss the sagitta for tracksthrough 3 detectors (separated by 2 cm) is <40 J.l forp > 200 MeV/c; so only soft particles will fail thecut. Since they also have large errors from multiplescattering and for a linear extrapolation to a ver­tex, this is not unwelcome. If redundancy is felt tobe necessary the chosen pairs can be checked againsta 4th plane.

that each Si plane have at least 4 hits, then a 500sec-l rate is obtained with ET > 8 GeV in each armand now adding the vertex trigger to either arm givesa rate of <10 sec-l while the HQ rate is 0.25 sec-I,i.e. HQ/LQ ~ 1/40 at the trigyer level. This setupgives for fLdt = 1039 cm2 sec- a sample of 106 unbi­

-ased b quark jets in the arm that did not use the-ver­~trigger. A great advantage in keeping ET low isthat the track multiplicity in the jet is also low,consequently the combinatorial background whenreconstructing either D or B mesons is less severe.This is to be contrasted with roughly 105 b jets atLEP (in the same amount of running time) with ET = 45GeV. In Fig. 8 we show the type of event that trig­gers the vertex detector: a) shows all the chargedtracks from the event, b) an expanded view of onlythose tracks which are reconstructed in the vertex de­tector. Figure 8 illustrates the point that althoughthe track multiplicity in the event is high, only asmall number of tracks go through each vertex detec­tor and are used for triggering.

Because of the large backgrounds from lightquarks and gluon jets (LQ jets) the limiting factorin studying heavy quark jets (HQ jets) is the abilityto trigger selectively. The cross section for HQjets increases rapidly with decreasing ET (as long asET > mass of the heavy quark); therefore, to getlarge numbers, it is important to trigger with thelowest feasible ET in the calorimeter. As one cansee from Table I to keep the trigger rate at a manage­able level with an electron trigger the lowest ET inthe HQD is ~ 15 GeV. The electron trigger require­ment improves the ratio HQ/LQ from 1/800 to 1/150while the rate goes from 600 sec-l to 20 sec-I. Thisis a substantial improvement but still leaves anenormous number of background events that need to berejected before one can analyze HQ jets. A very sig­nificant reduction in the background level isobtained using the high resolution Si detectors nearthe interaction diamond to tag charm particle decaysas discussed in Sec. IV. Obviously, there is much tobe gained if one could use these detectors at thetrigger level.

a) Motivation

With th!! above cuts one expects .1'20,000 pairsfrom direct BB leptonic decay and less than 2,000where one lepton comes from the decay chain ~c+1 (in107 sec at L=1032 cm:2 sec-I). Unequal sign leptonscan also come from cc pair; however, the simultane­ous requirements that a D be tagged and that thelepton be within 3~ of the primary vertex reduce theexpected number to .1'3,000. This number can bereduced further by studying the PT distribution ofthe leptons with respect to the jet axis, so they arenot a significant background. A possible source ofbackground of equal sign leptons is due to cc or Cc jets. For Iyl<l these are exp!!cted to have a muchsmaller cross section than the bbjets and thus negli­gible.

VI. Vertex Trigger

shown in Fig. 7. We make the requirements that bothleptons have Pe>4.0 GeV/c, that a D be tagged in eacharm, and finally that the leptons be within 3~ fromthe primary vertex (to reduce the number of leptonsfrom D decay). Then the equal sign pair are reducedto less than 10% of the direct pairs.

Given the large number of lepton pairs from di­rect B decays (.1'20,000) satisfying the selectioncriteria described above and the relatively low back­ground, detailed studies of the characteristics ofthese events are possible and should enable one toshow convincingly the existence or non-existence ofBB oscillations.

In the next section we will describe analgorithm and a scheme for implementing the algorithmthat selects HQ jets with high efficiency in lessthan 2 msec. Using the vertex trigger and requiringET > 15 GeV in the HQD gives a rate of <10 sec-l androughly 0.2 b jets/sec. Further offline cuts lowerthis to ~ 106 b jets with HQ/LQ ~ 1/2 - 1/3 afterusing only the vertex detector information. This isto be compared to 2 x 105 jets with an electron trig­ger -- however, with better HQ/LQ « 1). Althoughthe background level is still high this system givesa large number of events which could be studied inconjunction with other detectors. The limitation onET comes mainly from the trigger rate before usingthe vertex trigger, i.e. 500 sec-l which forces ET >15 GeV (Table I). If additional requirements areimposed, ET can be lowered. Let us, for example, usetwo detectors (as discussed in Sec. V) and demand

o

o

o

o

J

,)

( \\,._J

)

) 147

3. Total dead-time is not to exceed 2 ms.

Approximate system requirements are as follows:

The proposed algorithm separates neatly into 2parts:

1. Input data originates from 3 detectorplanes, and is linearly encoded with 10 ~ resolutionover a maximum range of 5 em.

)

)

)

10/event; maxi-2. Average track multiplicitymum = 30.

Each microcomputer would have 2 ms to performapproximately 10 N arithmetric operations per event;or about 300 maximum. Assuming 1.5 ms for this task,a high-performance device, such as Motorola 68000,would be required. The cost would be roughly $5K permicro, or about $25K for 5 projection planes in thedecay region.

Hardware processing techniques are not as appro­priate to the second part of the algorithm. For onething, the speed requirement simply is not there: Theprocessing time scales with N, rather than N2. Fur­thermore, there are multiplication and divisionoperations, which would be expensive in hardware, butnot well justified. Finally, a hardware vertexfinder would incur high design costs for an algorithmwhich is relatively untested. For these reasons, thesecond part of the algorithm is better implementedusing an array of microcomputers, each executing thesame algorithm on one of the projections (see Fig.11). The output of each microcomputer will consistof two pieces of information: chi-square for the pro­jection; and maximum deviation (or, more compactly,the result of the limit test: yes or no). Themicrocomputer outputs would be made available via abus structure to a hardware trigger logic box and/ora host minicomputer.

2. Analyze the sets of decay-band interceptsfor the signature of a secondary vertex. An examplealgorithm would be as follows: For each projectedplane, calculate the chi-square of the intercepts.For the plane with lowest chi-square, apply a windowtest to the maximum deviation from the mean; if themaximum deviation for that plane is within limits,produce a trigger; otherwise, reset the detector forthe next event.

The problem of high-speed, high-multiplicitytrack-finding in a segmented multi-plane detector sys­tem has been studied extensively,(12) and a system iscurrently under construction for experiment 766 atthe AGS. The first part of the algorithm could beimplemented using essentially the same techniques, al­though perhaps at lower speed and cost. Figure 10shows a possible configuration. The cost of such aprocessor is estimated to be $20-$30K. Processingtime is dominated by production and testing of allpairwise combinations of hits from planes 1 and 3.Assuming the maximum multiplicity N=30, there are N2~ 1K combinations, so that even at TTL cycle times,only about 200 ~s are required for track finding.

1. Find all tracks which pass through all 3 de­tector planes, and project back the intersection re­gion. Calculate the intercepts of each track with aseries of planes (~5) parallel to the detectorplanes, but within a narrow decay band of the inter­section region.

VII. Conclusion

dow needed to read the information from the Sidetectors. The procedure for selecting straighttracks tends to eliminate from consideration trackswith noticeable scattering in any plane except thefirst one. Limiting the range for maximum deviationbetween 50 ~ and 1000 ~ also substantially reducesthe probability of selecting an event in which one ofthe tracks underwent a hard collision. It, inciden­tally, also reduces the probability of triggering onK decays, less than 4% of the K~ + n+rr- decays willsatisfy the trigger. Multiple events can be a moreserious problem. The readout time for the Si detec­tors is unlikely to be less than 50 nsec. At L =1032 cm-2 sec-l the average time between events is200 nsec, so 25% of the time there will be anotherevent in the 50 nsec window. One could veto thoseevents using a device with faster response such as ascintillator hodoscope. Another approach is to ori­ent the readout along the beam direction as shown inFig 8. Since the interaction region is 2 em long andtracks are selected with deviations less than 1 mmonly ~ 1/20 of the double events will satisfy thetrigger requirement, i.e. 1/80 of the LQ jets willtrigger. Better timing off line will still berequired to reduce the background further. The multi­ple events problem then can be eliminated by suffer­ing a 25% loss. The disadvantages of orienting thereadout along the beam direction (Z) are: 1) thevertex detector can no longer be used as part of a mo­mentum measurement in a solenoidal field; 2) it is im­possible to join tracks in the vertex detector to therest of the system if that system does not have goodresolution along Z. The natural field with this ar­rangement is a dipole with field lines parallel tothe ground.

We have shown that it is possible in principleto design a detector of modest dimensions capable ofdetecting heavy quark jets with good rejection oflight quark (and gluon) jets. In order to achievethis, good lepton identification and the ability totag a charmed particle decay with-a-high resolutionvertex detector are indispensable. To keep the ver­tex detector within realistic dimensions we make useof the possibility at CBA of a small interaction dia­mond (± 2 em full width) with high luminosity, L=1032cm-2 sec-I.

Appendix A

With two detectors placed opposite to eachother it is possible to collect at least 105 unbiasedheavy quark jets (in the detector opposite the taggedjet) in a year of running with practically no back­ground from other processes. More sophisticatedon-line data processing can raise this number to 106The ratio of c to b jets in this sample is 2 to 1.With such large numbers of B mesons it is conceivablethat some decay modes may be fully reco~structed. Ofparticular interest is the decay rf + K""e+e- forwhich an upper limit of 10-5 is achievable, easily anorder of magnitude better than what is possible atLEP. The standard model predicts 10-6 for thisbranching ratio; however, horizontal gauge symmetriescan increase it to 10-4 which would be observablewith the detectors envisaged here.

This research was supported by the U.S. Depart­ment of Energy under contract DE-AC02-76CH00016.

The trigger would be implemented as a front-endprocessor of data from a silicon strip detectorarray. The proposed architecture makes use ofdata-flow hardware processor techniques.(lO,ll)

148)

oTable I

Rates with One Detector at L

Table III

Rates with Two Detectors at L-1 -2

sec cm

oTrigger Configuration

Detector 1 Detector 2Trigger Rate

sec-l b jet Rate*

o

()

Calorimeter Only:

Trigger rateB Jets +

B -+- Resolved D-

Calorimeter + MuonTrigger:

Trigger rateB Jets . +

B -+- Resolved IJ

Calorimeter + eTrigger:

ET > 8 ET> 8 7500 1.1

7000. 560. 12. 2 ET> 15 ET> 15 130 0.123.0 .7 .020.5 .24 .004 3 e+ET > 8 ET> 8

120 0.20" ET> 8 e+ET > 8

4 e+ET> 15 ET > 15350. 40. 2. 8 .02

.7 .16 .004 ET> 15 e+ET> 15

.1 .03 .00125 2e+ET > 8 ET> 8

4 .0025ET > 8 2e+ET > 8

Table II

±Resolved D ET Threshold+ vs.

Total D- in Trigger Jet

ET > 8 ET> 15 ET> 30

1 Visible Track 20% 24% 25%

2 Visible Tracks 11 9 8

>2 Visible Tracks 7 8 14Total 38% 41% 47%

Visible Lepton 12% 12% 13%

o

()

o

()

)

)

Trigger rateB Jets +

B -+- Resolved IJ

Calorimeter + VertexTrigger:

Trigger rate

250.0.5

.08

120.1.5

.8

20.0.12

.02

10.•3.15

.4•0035.0007

0.2.01.007

*The b jet rate is for b jets in the detector thatrequires no electron in the trigger •

References

1. D.R.T. Jones, G.L. Kane, J.P. Leveille, Nucl.Phys. B198, 45 (1982).

2. J. Sanford et ~, BNL 50718 (January 1978) •

3. Frank E. Paige and S.D. Protopopescu, BNL-29777. This study was done using the latest ver­sion, ISAJET 3.24.

4. L.L. Chau et al., DPF Workshop on High EnergyPhysics an~Future Facilities (Snowmass Co.,1982).

5. C. Fabjan et al., Nucl. Instr. Meth. 185, 119(1981) •

6. For the proceedings of a recent workshop on thesubject, see "Silicon Detectors for High EnergyPhysics," T. Ferbel, Ed. (Fermilab, 1982).

7. B. Hyams et al., CERN Exp. NA-14 (see Ref. 2, p.195). --

8. V. Radeka and R. Boie, IEEE Trans. Nucl. Sci.,NS (1979); see also Ref. 2, p. 21.

9. P. Shephard ~~, Ref. 2, p. 73.

10. "A Hardware Architecture for Processing DetectorData in Real-Time," G. Benenson, et a1., inProceedings of the 1978 Summer WorkS~ BNL50885.

11. "Real-Time Processing of Detector Data," W.Sippach, et al., 1979 Nuclear Science Symposium,IEEE TranS: on-Nucl. Sci. NS-27:1, Feb. 1980.

12. Appendix, FERMILAB proposal 627, B. Knapp, ~a1.

149

MUON DETECTOR

/)

RIFT~ / CHAMBER~ TRD 2

"=;~~~~~~~~.~DRIFT CHAMBER

""'":'Z'FT CHAMBER~VERTEX DETECTOR

HADRON CAL.

I I I I I I IEM CAL. I I I I I I I_0

)

I METER

Figure 1. Single arm heavy quark detector at e =90°, covering ±1 units of rapidity and45° in azimuth.

,)

~@",,,--

@ ®" DIAMOND INTENSITYCD--==----==-- PROFILE

BEAM I -- I I , I I --,BEAM6420246

Z (em)

)

)

yPLANE XX z OUTPUTS1.., /

/

@' ~ / CD lem 2x6em2 170@ ,

® 1.7 3.4x8 317®CD , 45· @ 2.3 4.6x10 462/

--- -- --xBEAM @ 3 6xl2 640

1589

)

Figure 2. Four plane vertex detector, shown fullsize.

150 )

()

0.3

MINIMUM BIAS<n> =2.00.2

()

()

0.1

>-t-..J-0 CD 0.0<t

0 5 10 15 20 25CD00::Q.

Er> 15 GeV() 0.10 <n> =9.0

0.08

100.06

0.04

0.02

0.00~) 0 5 10 15 20 25n

,)

Figure 3. Multiplicity of charged tracks throughthe detector.

151

40

Ic)MUONS<p> =4.0 GeV Ic

)

2 4 6P IGeV/c)

12)

Figure 4. Momenta of charged tracks in the detectorfor trigger threshold ET > 15 GeV.

(a) All charged tracks

(b) Electrons (including Dalitz decay of1TO,n)

(c) Muons

152

o

o

o

o

FITTEDTRAJECTORY

__VERTEXPOINT

.I

1

}

DETECTOR(4 PLANES)

o

Figure 5. Resolution at the vertex for a singletrack (not to scale).

o

(}

( \

60

EJ- 40(\J

.....a::LLJm:::i: 20::JZ

200 400 600

8 (fl-m)

800 1000

,, /

Figure 6. Projected miss distance for resolveddecay tracks of n± mesons, using thecriteria given in the text.

153

(0 )')

,}

100 200IX~al_X~eas I(JLm)

10.08.04.0 6.0Pe (GeV/c)

.....IIIIII

.-=!::c::::l

I

Z

/

::>.ci...>-C

II:«II:!::CDII:«

IJ.

Figure 7. (a) Momentum distribution of leptonsfrom semi-1eptonic B decays.

(b) Momentum distribution of leptonsfrom the chain decay b + c + t.

Figure 9. Difference of calculated and measuredhits on plane 2 using all combinations ofhits from planes 1 and 3 in the vertexdetector.

154

o

~,------------------r (a)

o

q(')

I

o.,.;I

7.05.03.01.0

(em)-1.0

z Axis-3.0-5.0

o'1'+--,---,---...,..--,---,---..-------1

-7.0

o§,-- .., (b)

13000110007000 9000

z Axis (fL)5000

...., .

......

/.......... '...........

; \.•f \"

:::::::'.!

)

\...... ~" ..../""_...J\\./ /

o

~I +_--,.-_-.,-_--,._-,._--,__,.-_,..-_..--_,---l3000

isgI

ois(')

u

Figure 8. A cc event selected by the vertex trig­ger.

(a) All tracks from the event and thevertex detectors.

(b) An expanded view of the vertex withonly the tracks detected in the ver­tex detector.

155

I)

)<.,I,I

~~r~

\:..::d)"I-c.

~:t"~r ...

.I:!!llte=~

Go-:=~AAIiTC't.. ~ ?2 cof)vl:F- ..... u ... ,"', Zv.JlS'1S

c:.e.,.-t~I""A."o....J$ Ot=- ~"'\)I('~::i

(r rF:» r:;;"S"=l} eN 'I!.""'-N t:r Ce:tv,,"''- c.~ S'r-.3'?..,'t"""

~~l.~ l?IT:~,:'#' Jc.v_P.o..:):c.._-j--------w------

~~.:"jTo~ ::r-...~ (-t";/~o) "'<;

'\ /'-. 'FtJ~o"" OF ""....vr:= (')l.,) i "Z'FA~LI4ji -7 f'-i. C'owT'C.ve,,",S LD~""O"..,J'S IN VtCINIT""'/

COUNTf."I..: 'STOl.'; r ....')~, 7.o',It<; ,.-.... :r"~ l- ../ oc:: "{t~?,t'.,.~'9 V"'-U~ .:f, 1'0 0'="--

S~~v~•..n\""''''''''1 A~i)!J~ D~":'j\' .,....~lt.""1t....a.~ ~"IS~ ....Crt' Ct=- '5T"ft. ........ f'... rr-TlIc;:'''1 (F'Ift..l)T" .;:-..... "'i':;) OI<J S''=1.~~''''';) .__~._~..!..t"_'!.-:.'!~~ _'1'0':5 I'" S-.o·,,'1 01= 1< (Po'''''-) 01=""10.,,1 ......" ..-.. ,,It(t.,,,,= ,,-r"' ';'t.F.... " -e.",H::·~~: t, - 1]e'ftf~ ~,F'"O "l3vF~ftZ'it.

'f7"2T", F.".C'T. ",'I")', ,',I I ('",,, 'tteCSlu'~~1) .,.0 w;;.'-'~'o.,J"T?:' Q'E"""f;)-

~1~1it. C)rCl'--'C:O-V"It: (~ltit .....~·.C"")C:. I Pj ,""c.c.S ~~Dv""'J -n.1~ L-e:toP' CON "7C:;Ai\(''"''''TO~II.w r,.,Tp:, '··Alr5, I SIS"orIN"'.... 0= ~C.)G,'-,N,+/1H.

".~"..l~,"\.A"T\'T) E',..U... Frt.. O.......-v""..- ~

j\,.l"" \&C' 'P"In..c; C...,rrSS-o:..); SIC.'iJI'-LL k-:-,'t"AI7$ (N':t'.·~,,":,

)

Figure 10. Hardware processor configuration fortrack finding.

)

)

l-i'O'to <:o'1P~~: 7';11<"0;1.'1 5T1\.'T1STIC....L.

A"'" AL-"1S. S O~ "Pn.D:S~<"f'f;,?.,-,e.-, e~ T,.Jrr;::(.­

CIF?T~ 0.... ~ "....,......,.~ I~ -n.\1'f' IJr:rc"",",'t ~H:t,'Sf1E~ 1-f;o...N "UF,/Cr)

l,(v~..IH""

it.I .... & '!!.\Jt="4="e:-il..S: STOIl.'! 1l:1f'\..u V';,,"-A ....,,("''"1JT-..JCo:'';:;-·''l.CS'i -P,..SS A"'l- "P',..,.....A ~ "1"'1'l""(U••

rl""'~/tIr·ir-. O!Jr?",~ """"'-1\. r--rl.r.1 ..... ACCE""?'T~~

~\I~,.J,'t --n::> C"Joo1'?u~~jL. "e...JS \J7t>~ ('DI--tHI\""":'J

!="1t<>""1 C"''''"T "A.... rIZoCo.=SS""'''.

'I'''l (}r-"R.)I

1'. x..,

Cf?~

":'''...01<.-1 :

C::~-.\rl'U~

7rzc,c~£t..:

MI\-lC."'::S "ntl~.&f:.;;t..

"Or=-CI'SIOr-J ~"'Sl:;"7) ohJ

:Tt-J,,-aiLI-1AT)el' .... 1="tlQ""" t-1 \("'flcS

v)

Figure 11. Array of microprocessors for vertex find­ing.

156

oANGULAR COVERAGE OF DETECTORS

R. DieboldArgonne National Laboratory

n

, ),/

Introduction

The total rapidity range at high energy pp and ppcolliders is very large. For-yrotons, Ymax =~n(1S /Mp) = 7.0 (10.0) for Is = 1 (20) TeV. If onewere to cover this entire range using pseudorapidity n= - ~ tan e/2 as a guide, the small angle limit wouldbe 2 (0.1) mrad (corresponding to a PT cut off of 0.94GeV/c for elastic scattering). Heavy particles tend togive decay products at much larger angles, however,and for a given amount of money one can ask how tooptimize the general purpose "411" detector and wherefine segmentation is most likely to be of use ingetting out important new physics.

Parton-Parton Interactions

Here we will assume that a primary purpose ofhigh energy colliders is to explore large subenergiesin the parton-parton rest frame:

where xl and x2 are the fractions of the beam energycarried by the two interacting partons and s is theoverall center-of-mass energy squared. The largesubenergy can then be used to produce massiveparticles, cause high PT jets, etc. For simplicity,we will think of this as the production of a heavysystem with M = IS. Since the structure functionsfall rapidly with x, the cross section for a givenmass tends to be largest when xl = x2' and fallsrapidly in the asymmetric case, i.e., for large Iyl.

To take a specific example,l we look at gluon­gluon interactions with G(x) «(l-x)S/x. For a simplematrix element, we then have

~; «G(x1) G(x2) « (1 - x1)5(1 - x2)S/x1x2

and at fixed IT = M/ iii:

do « (1 _ IT eY)S (1 - IT e-y)Sdy

This function is traced out in Fig. 1 for severalvalues of IT. For IT = 0.1 (100 GeV at IS = 1 TeV,2 TeV at IS = 20 TeV), the distribution falls to halfheight at y = 1.0. For a y = 1 heavy particledecaying isotropically into relativistic particles,the median angle of the decay products will be 400

(corresponding to a pseudorapidity of 1.0) and 76% ofthe decay products will be within %1 units ofrapidity.

The half-height width of the distribution inFig._1, Y1/2' is shown in Fig. 2 as a functionof IT. Even for a relatively light object withM/ IS = 0.01 (10 GeV at IS = 1 TeV, 200 GeV at IS= 20 TeV), Y1/2 is only 2.7 •. Allowing an extra unitof rapidity for the decay gives the result that mostof the decay products from such an object will emergeat e > 30

• Heavier objects produced by parton-partoninteractions will give decay products primarily atlarger angles.

dissociation. The rapidity of the heavy object isthen simply given by

y = ~n (Is /M)

For e~ample, a Centauro of M = 200 GeV diffractivelyproduced at the Fermilab Collider would have y = 2.3(median angle of 120 for relativistic decay products).

Missing PT

Depending on the physics being investigated, animportant consideration in determining the small anglecut-off of the detector may be the apparent PTimbalance resulting from particles escaping at smallerangles. For a heavy system produced with a rapiditycorresponding to the cut-off angle and decaying intotwo relativistic particles, one of these particleswill always be observed and the other not, leading toa PT imbalance of as much as M/2. This is a worstcase since decays into more particles will tend toreduce this imbalance. Particles produced 0.5 unitsof rapidity away from the cut-off would have a ~ SO%chance of balancing momentum in the case of two-bodydecays.

As an example, we estimate from Fig. 1 that aboutS% of the particles with M/ Is = 0.02 would beproduced with y = 3 % O.S, and that roughly half ofthese would give a significant imbalance of PT' up to0.01 IS. At this level, cracks and inefficiencies inthe detector are likely to give effects of a similarmagnitude. This would suggest that a cut-off angle ofabout SO should be adequate to hold the apparentimbalance to PT ~ 0.01 IS ; this calculation is, ofcourse, just a rough estimate and must be checked witha Monte Carlo program for the specific detectorgeometry and physics under study.

Inclusive Distributions

Even for minimum bias inclusive events at the SPScollider, the half-height of the pseudorapiditydistribution is not moving out as fast as allowed byphas~ space. A comparison of UAS streamer chamberdata taken at the ISR and SPS collider is shown inFig. 3; the half-height point has moved out about 1.2units of rapidity, only half the change in Ymax'

Summary

Decay products from large PT scatters and heavyparticles will show up at "large" angles (more than afew degrees) and emphasis should be placed on thisregion when designing a general purpose "411"detector. In addition to optimizing the system for afixed cost, this means that less space is requiredalong the beam than might be first estimated bylooking at the nominal rapidity range. Smaller anglescan be covered with a crude detector to tag any eventswith significant PT at angles less than that coveredby the high-quality detector.

References

Diffraction Dissociation

In principle, rather heavy systems can beproduced with nearly the full beam momentum (smallfour-momentum transfer) by single diffraction

157

1.

2.

R. Diebold, "Report from the CDF Workshop Groupon New Particles," CDF-88 (1981).

K. Alpgard et a1., Phys. Lett. 107B, 310 (1981).

5d<r!(da:\dy dy)y=o

5

4

2

30·

60·

O~OL,O-'-I--'---'---'c...L..L.uO""'.O-I~---'---'-'-..L.W.Ou.:-1----'----'--'-'-~I.090·MIlS

Fig. 1 Normalized rapidity distributions for heavyobjects produced from gluon-gluoninteractions with a simple matrix element.

Fig. 2 Half-height widths of the distributions shownin Fig. 1; 6 is the corresponding angle givenby the usual formula for pseudorapidity,n = - ~n tan 6/2.

31 dO"adl\

2

1

, ," ",

\\

\\,

\\\\\

)

54321o+---..---,---...,----.---,....-l

o11\ I

Fig. 3 Pseudorapidity distributions for inclusive particlesfrom minimum bias events (UA5, Ref. 2). The open pointsare from the ISR at IS = 53 GeV while the closed arefrom the SPS at IS = 270 GeV. The broken curves are pre­dictions for these, two energies from a limited-PT phasespace model with < PT > = 350 MeV/c; the solid curve isthe same model at 540 GeV, but for < PT > = 500 MeV/c.

158

()A DETECIDR WIlli A 3 T SOLENOID FOR A 10 TeV COLLIDER

R. Kephart and A. Tollestrup

Fenni National Accelerator LaboratoryBatavia, IL 60510

M. Mestayer

Cornell universityIthaca, NY 14853

()

It has been frequently suggested that a long highfield solenoid could be the basis for a high perfor­mance detector at a high luminosity machine. we ex­plore here the properties of such a detector for a 10TeV collider using a 3 neter dianeter superconductingsolenoid with a 3 T field. Such a magnet will curlup particles with P.L less than 675 nev/c so that theseparticles will not penetrate the coil. Originally, weexplored a very long solenoid (20 neters) with theidea that end effects could be min:illli.zed. HCMever, itbecane apparent that the coil %ckness (about 2 Xcand 0.4 absorption lengths at 90 ) would be intoler­able at angles fo:rward of 300

• Since it would be veryhard to calorineterize the superconducting coil, weabandoned this effort and retreated to a nore "mXIest"roil of 'V 11 m in length whose end is crnq;>lete1y open(Le., no end plug) yet is covered by fo:rward calori­netry. This detector is shCMll in Fig. 1.

3. The IlIU1tip1e roulOlTb scattering in a coil of2 XC is small conpared to the bend angles ofparticles with nmenta of interest (> few GeV) •For instance for a 1 Tev/c track:

t, = 1 nm = 1000 ~ (proportional lip)

M = 21 ~ due to multiple coularb scattering(proportional lip)

M = 28 ~ beam size (independent of p)

If this track is neasured at R = 1. 7 m and R = 2.2 mwith 'V 20 ~ accuracy (e.g., 10 layers of 65 ~ highpressure drift chanbers), then the expected error inthe iIrpact paraneter t, is:

ot, = 44 ~.

Measurenent with "standard" atrrosphere pressure cham­bers giving 200 ~ resolution would yield:

Such a detector has several interesting properties.Since the coil cuts off all IlOIllEmtum be1CM 675 neV/c,we estimate that this reduces the number of chargedparticle tracks outside the solenoid to 29 percent ofthe field off value. This feature could make atracking system outside the solenoid particularlyattractive at high luminosity. On the other hand, thecurling tracks presumably make tracking inside the roilnore difficult and will also cause difficulties at theend regions. we will also shCM be1CM that all chargedtracks in the central region can have their IlOIllEmtumneasured to quite high precision by such an "external"tracking schene.

'!hus, the overall IlOIllEmtum resolution is:

-J (21 ~)2 + (28 ~)2 + (~;-Jl)2 '"1000 ~

5.6 percent

The arrangenent of the tracking chanbers proposedis shCMll in Fig. 1. As indicated above, tracking forthe central region, InI < 1. 5 is located outside thecoil but inside the E.M. calorinetry. This has threeadvantages:

1. The single rates on chanber wires should belCMer.

2. Pattern reoognition should be easier since itis done in a field free region in which thetracks will be straight lines.

3. The nonentum of all charged particles can beneasured with high accuracy.

To aeeatplish the latter, we observe the follCMing:

1. For an invariant normalized emittance of 24 7f

mn/mr (95 percent) and a S* of 2 m the trans­verse beam size, cr, is 28 ~.

2. The outside tracking system will neasure theiIrpact paraneter which is four tines as largeas the nore camon1y neasured sagitta.

The IlOIllEmtum resolution at 1CMer nnrentum is betterthan this and approaches the 2.1 percent limit set bythe multiple cou1anb scattering.

In sUlllllarY, we believe that very long high fieldsolenoid is not a good candidate for a high energy,high luminosity detector, but that a sorrewhat shorterhigh field magnet with tracking chambers located out­side the coil may give both good nnrentum resolutionand easier pattern reoognition than standard solenoidswith tracking inside.

,) 159

11.5m.

L~~.........~

10 TEV DETECTOR

I~~~~~~~=~~==~=~~~~::~~~V~E~R~YFORWARDTRACKING

.,.,__-=.:-C::Al:::ORIMETER

.)

J'o

I . II.ll 3.0

METERS

)

Fig. 1. Detector for a 10 TeV co11ider.

160

)

o

STATE OF THE ART

Leon M. Lederman

Fermi National Accelerator LaboratoryBatavia, Illinois

()

-)

C)

,)

There is a large body of experience in "highluminosity" data taking in fixed target research. Wetry to consider a wide variety of high rateexperiments which were limited by the detector (not byavailable beam) to a preset number of collisions persecond. We then attempt to translate thesestate-of-the-art experiments to effective colliderexperiments. To this end, we extend the chosendetector to a comparison 4n collider detectoroperating near 1 TeV.

There are several issues: (1) effective solidangle must be translated to ~ 4n, (2) environments maybe quite different, e.g., beam dump near fixed targetor beam halo muons vs. collider backgrounds, (3) themultiplicity varies over the experiments selected and(4) we have to treat open vs closed geometries. Thelarge variety of experiments selected is designed toaverage over the causes for detector limitation.

Finally we chose detectors which have producedphysics in order to gauge the "state of the art."

(1) Effective Solid Angle

If we use 4n in order to see "the whole event"then we must correct by the solid angle. One way tosee this is via accidentals. Suppose there are npiece~ of a detector, each subtending 1/n th of 4n. Ifthe maximum sustainable rate in a piece is Rets/sec,then the accidental rate is:

(Dead times, pile up etc. are all forma ofaccidentals). If the R rate itself is arrived at bytaking the maximum allowable accidental rate, then inthe collider mode, each piece must be reduced by n tohave acceptable accidentals. This means we mustreduce the luminosity by the factor of n, Le., thefraction of the solid angle in the fixed target run.Some fixed target experiments install a momentum or p~

cut. If this is modest it probably conforms tothe p~> 1-1.5 GeV/c requirement imposed in SPScollider calorimetry. However, in colliders, theseevents usually burden the tracking even if theirinfluence on calorimetry is relatively benign.

(2) Environment

Fixed target environment can have beam halo, fluxfrom nearby beam dump or poor duty cycle during theslow spill. Collider environments are alsocomplicated by beam gas, beam wall, side splashes andthe debris from forward jets striking flanges etc. C.Rubbia estimated that UA1 had 40% additional tracksper event which is event originated but not from thevertex. We'll call this a wash. The time structureof the comparison collider (and ISR) have a 100% dutycycle-- most fixed target experiments average > 50%duty factor due to spikes or poor duty factor. -

(3) MUltiplicity

The FNAL experiments have mean charged particlemultiplicities of 9, ISR is 13 and pp at SPS of 25.The AGS mean multiplicity is about 5. These are justthe properties that burden detectors (one has to becareful to correct for thick target effects) and sincewe are aiming at a 1 TeV collider detector, we shouldmake a correction to the luminosity.

(4) Closed Geometry

These are a valuable variety of experimentsdesigned for high rates. we can treat them in twoways - as a limit of the state of the art in closedgeometry experiments and, by adding back the absorbedhadrons, reduce the luminosity accordinglY. A perfectabsorber decreases the rates by > 10 i.e. directmuons to hadron ratio near y=O. More often there iseither finite density ~bsorber an~ some decay space.We use factors of 2 x 10 to 5 x 10 accordingly.

Conclusions

OUr results are. presented in Table I. One couldcorrect all the achieved luminosities by dividing by ~

30/multiplicity and by looking at the correspondingrapidity interval rather than solid angle. However,the lesson seems clear enough - the state of the artof published physics before 1983 seems to hitconsiderable resistance above 1031 cm-2 sec-1• Thesole exception is the I-1 experiment at the ISR.Closed geometry experiments also suffer limitationssomewhere between 10 33 and 1034 • This wide variety ofexperiments, designed by competent physicists towithstand high rates, indicates that d~tectors

designed for much higher luminosity than 103 , mustbreak new grounds in invention, complexity and/orcost. Finally, there is the concern which, to myknowledge, is not addressed by this workshop: how canwe be sure that the new, non-state-of-the-art detectorcan extract physics?

161

STATE OF THE ART

Why?Interaction

Experiment Sec

FNAL E-

Target/Beam ABS

RapidityQcm

EffectiveLuminosity**

400

537

326*

615

673

609

623629515

288*

ISRI-1(CCOR)

105

105

106

4 x 105

5 x 10 11 *

Si/400neutronsW/125 GeV

IT,PVar. 225GeV IT" 250 IT-

Be 200 IT

H2 400 P

200lT,400pc/200 p, ITBe/200lTP

cu/400 Pp.... "J. 2 GeV/c

30 x 30

o

60"eu

10'Fe

8' Be +16' co

o

ooo

22'Be

o

1.6 IT

2 IT

3 IT

2 IT

2 IT

2.9 IT

.007(~lT)

at 90

2 IT

01' 1030

< 1030

01'

4 x 10 30

1030

5 x 1030

2 x lq30... 10 302 x 10312 x 10311 x 10

2 x 1031

> 5 x 1031

trigger deadtimeBeam tagging

DriftChaml:>erstrigger deadtime + PWC'spb glasspileup PWC,mg.Halo muonsotherwise x5SC & D.C. jam.Liq A pileupLiq A pileupC ConfusionScinto counter+PWC rates

No problemsexcept in E

t

,)

I-6(AFS)

30 x 30 ooI'lT

2 x 1031

6 x 20303~P83-011.4 x 10 EP82-139 "10%

pileup"

AGS726

732

13 GeV IT o

o

large

1/5 of 4lT 10 31

datahandlingtriggeringbeam taggingbeamtagging

)

* These experiments correspond to 4 x 10 33 and 1.2 x 1034 resp. in a closed ge~~etrywith n~ measurements before the absorber. At 1 TeV these numl:>ers become 01' 1 x 10 and3 x 103 respectively.

** Not yet corrected for multiplicity.

162

)

)

()

EXPERIENCE WITH HIGH LUMINOSITY RUNNING AT THE CERN ISR

James T. Linnemann

The Rockefeller University

()

o

I will discuss experience of the CCOR and COR colla­borations at the ISR at the steel low ~ intersectionregion, with luminosities up to 6x10J1 cm- 2 s- 1 • In

general, this luminosity level has caused only minorinconvenience for a detector covering polar angles 45to 135 degrees in the center of mass, except for aspecial unrestrictive trigger on total transverse ener­gy.

The ISR

Lead Glass

Two arrays of lead glass are located outside the mag­net coil, 1.2 m from the intersection region. Theyare retracted to 4 m during filling of the rings tominimize radiation damage (lead glass sli'vers are usedas long-term dosage monitors at the ISR). This radi­ation damage causes a calibration drift of ~1% permonth, which is monitored by the 450 MeV single par­ticle line in the blocks. The back arrays are each168 blocks of 15 x 15 cm 2 x 17 Xo , and the front lay­ers are 34 blocks of roughly 10 x 50 cm 2 x 4 Xo •

Apparatus

Figure 1: The apparatus viewed along the beam axis.

The major elements of our detector are shown in Fig­ure 1. I will summarize their characteristics, withemphasis on timing behavior. A 1.4 T magnetic fieldis provided by a superconducting solenoid[1].

The front glass blocks use a flasher system (EGG

Krytron KH22) which, in response to a trigger pulse,sends light through optical fibers to all blocks and toa well-shielded reference counter which also views aNal crystal illuminated by a 137Cs source. A 700 ns

gate is required for the Krytron signal. During thisgate time, 1.3 interactions are expected, but few de­posit energy in any individual counter, and one couldeasily take flasher calibration data with the beams on.

All blocks were calibrated with a CERN PS elec­tron beam. The back blocks maintain their calibra­tion [2] with Na I(TI) crystals doped with the ex emitter241 Am, which gives a light pulse of about 1 GeV

equivalent energy. The phototubes are 5" RCA 8055,inexpensive, but with a 30 ns signal halfwidth. Theevent gate is thus 200 ns, and the calibration gatefor the slow Nal signals is 400 ns. Since this cali­bration system is not triggered, there is some back­ground problem in taking calibration data with thebeams circulating. This can be circumvented by run­ning the low-threshold calibration trigger with the Bcounters in veto (see Figure 1). The long gate timesled us to use Lecroy 2249W ADC's, AC coupled withT = 30 IlS, to achieve good linearity. The 9 mm 2 Nalsources are hit by charged particles at a rate givingan apparent cross section comparable to the lTo cross

section above PT = 8 GeV. We remove these spu riousevents by making use of the slower Nal signal shape.Analogue sums of the two arrays are read in a pair ofADC' s with 130 ns and 500 ns gates. Spu riousevents have much larger pulse heights in the longgate.

Strip chamber50cm

Shower counter

The ISR is a separated ring machine, with the beamsintersecting at 14.77°. It is normally operated as anunbunched pp collider, the ribbon-like beams inter­secting in a diamond-shaped region, at our intersec­tion roughly 10 x 50 x 0.4 cm J • The steel low ~ sec­tion gives our intersection a luminosity 2.2 times thatof the standard intersection region.

,./

u

C)

.J 163

Shower Counters Triggering

Drift Chambers

Eight to ten layers of cylindrical drift chambers[3]filled the remaining magnetic volume. The coordinatealong the wire was derived (0" = 1 cm) from comparingtimes at the ends of a delay line (velocity 0.44 cm/ns) glued near each sense wire. The TDC systemdesigned by W. Sippach at Nevis Labs can record upto 14 hits per channel with 1.5 ns resolution. Threechannels (sense and 2 delay line ends) are used foreach of the 580 sense wires. The delay lines andmultihit capability allow several tracks in a singledrift cell, and give 3-space points as input to thepattern recognition programs, reducing the combina­toric problems. The drift wires are in paired gaps,staggered by a half cell as shown in Figure 2. Thisstaggering assures that out-of-time tracks will have abad X2 in first order. Our typical resolution of 350 I.land the drift velocity of 50 I.l/ns give 30" deviationsfor events 20 ns out of time.

Inside the magnet were 32 lead-scintillator sandwichshower counters, each 1.5 m long (along the beam di­rection) and covering 60 in azimuth. They were di­vided longitudinally into 3.9 Xo and 10.5 Xo compart­ments. Each end of each compartment had aphototube, 128 tubes in all. These phototubes weremuch faster (3-6 ns halfwidth) than the lead glasstubes, allowing a 130 ns event gate. An even short­er gate could have been used had not the analoguesection of the ADC's saturated at 2 V (we sent thesignals to the ADC's through dispersive cables toslow them down and decrease the peak). Each tubehad a TDC, Lecroy 2228A, with 0.02 ns bins and arange from -20 to +80 ns relative to the trigger time.The calibration system was a Krytron system like thatused for the front lead glass. The long term drift ofthe calibration was <1% per month, monitored by the175 MeV deposited by a non-interacting particle.This monitor was necessary because of the long timespan between electron beam calibration of the coun­ters. This energy implied a dynamic range of 100 : 1in the ADC's.

!I ••

= in time= out of time

•

• = sense wire

The analogue trigger electronics must slow showercounter signals in order to match the glass signals,and to provide even response along the full length ofthe counters (10 ns transit time). The circuitry has

a fall time T = 50 ns and a 20 ns flat-top.

Two triggers commonly used in the experiment of­fer a useful contrast. The ETOT trigger asks for asimple energy deposition greater than 20 GeV, sum­med over all shower counters and glass. It uses theentire solid angle and its purpose was to search forjet structure. It proved quite sensitive to doubleevents. The PAl RS trigger demands a 2-cluster massabove 7.5 GeV. The calorimeter cluster was definedas a wedge in azimuth of fl<j> = 0.22 rad x fly = ±0.6(glass) or ±1. 1 (shower counter). The PAl RS trigger

proved to be little affected by double events.

Delay Curves

To demonstrate concretely the response of the appara­tus to out of time events, delay curves were mea­

sured. Drift chamber data were altered in the soft­ware by adjusting the raw time data (thisdiscriminates slightly against late events). The re­sults are shown in Figure 3 for ETOT and PAIRSdata. The space poi nts a re rejected in softwa re forfalling outside the drift cells at a rate consistent withthe drift velocity (maximum drift time for sense hitswas 300-400 ns, and up to 800 ns for delay times).Track fitting, even with very loose cuts, rejects mosttracks 60 ns out of time with the trigger . Noticethat the ETOT data is more asymmetric in time thanthe PAl RS data, offering evidence that it is more con­taminated by early events than PAIRS data.

Delay cu rves for the calorimeter data were ob­tained by comparing signals summed from groups ofcounters sent to a monitor ADC, with the softwaresum of the individual counter ADC's, as the timingfor the individual ADC gates was moved. Events withonly a single interaction in the sensitive time were se­lected on the basis of timing information. The selec­tion procedure is described in more detail below.The results are shown in Figure 4. The apparentenergy of signals out of time by a full gate length is

less than 10-20% of the true energy .

,)

)

)

)

)

)

)

Fig u re 2: The effect of staggered gaps for out of

time events.

164

Figure 3: Delay curves for drift chamber data.3a) <Space points / event>

3b) <tracks / event>

~(b)

a518 •> •LU • • ETOT" •~10 •• • • • •u • •« • •a:t:;. • • PAIRS2 • • •

-90 -60 -30 0 30 60 90At(ns)

(a)

(a)

• • • • •• • ETOT• • • • • •• • • • •• • • • • PAIRS• • •

range min. #

±200 ns 2(-20, +80) ns 2

width12 ns12 ns

Time Cluster Cuts

tL-t R10 ns12 ns

Table 1:

Counters

Ashower

Detection of Multiple Events

t L-t R is the cut on end to end time differencewidth is the time window defining a clusterrange is the range in which hits are consideredmin. # is the number of counters to define a hit

Multiple interactions were searched for by examin­ing the time data for multiple clusters. The overallscheme of the time clustering is as follows. A goodcounter was defined as one having a coincidence bet­ween left and right end time consistent with the coun­ter length. For each counter system (A and showercounters, separately), the average of the left and

right end times of good counters was histogrammedfor each individual event. A cluster was defined as agroup of 2 or more times (a precaution against noiseor induced radioactivity) which fell within a 12 ns in­terval (see Figure 5 below). This time window wasslid along the histogram to find the most populatedcluster. If this contained the required 2 or morecounters, the counters in the cluster were removedfrom the histogram and the process repeated. A"double" event was defined as one with 2 or moresuch clusters in either the A counters or in theshower counters. Table 1 below summarizes the cutsused for the time clustering.

The timing information used to detect multiple interac­tions was that from the shower counters, and from abarrel scintillation hodoscope (labelled "A" in Figure1). The A counters covered the entire solid angleseen by the drift chambers and electromagnetic show­er detectors. There were 32 A counters, with a pho­totube at each end. Radiation damage to the lightguides incurred in the 7-year stay near the ISR beampipe has been observed. Each of the 64 A photo­tubes is read by a channel in the Nevis TDC system,allowing multiple hits (minimum dead time about 40 ns)and a range of ±300 ns about the trigger time. Theshower counter TDC's fired with a deposit of 150(500) MeV in the front (back) longitudinal compart­ments. The shower counters covered 60% of 2rr in <p.

60 90Late events

200events

(b)

Pb- Scinto130 ns gate

•gate length

• t

gate length

t

•

•

•

80 160

• Pb Glass• 210 ns gate

•

o 30At(ns)

•• •••

• •

•

•

••

•••-160 -80 0

Delay M (ns)

-200 -120 -40 0 40 . 120Early events Delay M(ns) Late

-90 -60 -30Early events

o

1.0

0.8

~ 0.6 •":;::<l •w 0.4 •

•0.2

1.0

0.8

0 0.6W~.....<l 0.4w

0.2

,.J 0

()

.-.I-~300>LU

"') ~200z(5

Q.100LUU«Q.Vl 0.......

')

o

Figure 4: Delay curves fora) Lead Glass and b) Shower counters.

.)165

Figure 5 shows the distribution of good A counter

times about the mean time of the cluster, for eventswith only one time cluster. For this plot only, the

cluster width definition was widened to 15 ns. Thefinal 12 ns width is shown on the plot. These cutsshould be able to recognize multiple interactions downto a time separation even somewhat less than 12 ns.

lal

400

1200 Ie)

300 (bl 600 ldl

Figure 6: The number of good counters/cluster:

6a) single cluster events, A counters6b) single cluster events, shower counters

6c) extra clusters, A counters6d) extra clusters, shower counters

)

)

5 10 15 20

200

400

5 10 15 20 25 30 35 40

100

200

600

800

OL1--L-ClLC:C..-...1---L-L....l..-~==r::::::r:::::c=~

-20 -15 -10 -5 0 5 10 15tcounter - (t) (ns)

1000

200

400

Figure 5: Deviations of A counter times from the clus­

ter mean for events with a single time cluster.

Tests of the Method

Additional characteristics of the clusters found inETOT data are shown in Figure 6. Secondary clus­ters are much less populated than the primary clus­ters. The multiplicity distribution of primary clusterstogether with Figure 5 allows an estimate of the prob­ability of falsely getting two clusters from the tails ofhigh-multiplicity events. For example, an event with25 A counters set (rare) should give 2 apparent clus­ters less than 10% of the time. One might hope todistinguish double events separated by less than thewindow width on the basis of the r.m.s. inside thewindow, but the distribution of this quantity in larg­est clusters is very similar in events with or withoutsecondary clusters. In any case, one must know the

time constants of the counters to about 1 ns/end to

achieve stable results.

The validity of the method for separating single from

multiple interactions was tested in several ways. Theaverage times of cl usters inA cou nters was fou nd tocorrelate well with the times for shower clusters whenmultiple clusters were found in both systems. ETOTdata were examined for "close doubles", double eventswith two time clusters in ±30 ns of the trigger time(in order to have good trackfinding efficiency accord­ing to the delay curve in Figure 1). The "close dou­ble" events were found to have 35% more space pointsand tracks than the remaining events, the "no closedouble" events. The average spatial resolution infer­red for these close double events was 500 I!, while"no close double" events had 350 I!, the same as thatobserved for all PAIRS events. Of the "close double"events, 52% had multiple vertices, compared with 17%in the "no close double" class. The distribution of

the number of tracks in the extra vertices is shownin Figure 7, demonstrating that the extra vertices in"close double" events have significantly more tracksthan the extra vertices from the "no close double"class. One may remark that minimum bias events with2 A counters firing have a vertex in our detector 60%of the time, and that "no close double" events maysti II have an extra event just outside the 30 ns cuts.Taken together, these data support the idea that thetiming information is able to recognize multiple inter­

actions.

)

)

166

6.x1031

4.53.0.Y'

1.5

SHOWERALONE

_____0-

LINEAR -------:..0-----:;0-~o=---o t40ns

:~o A ALONE~~;; ====~:===== 0------"--t200ns

20

ETOT= 22 GeV

16

8

o

Figure 9: As for Figure 8, with other event selectionsadded. For explanation of cu rves, see text.

24RAW 14

ETOT =22 GeV ETOT =23 GeV20

16 10

l-I-

a ~ 8I- 12 UJUJ "0"0

~ 6"-t> "0"0

8

SINGLE EVENT 4

4

I

1.5 3.0 4.5 6.0 1. 3.0 4.5 6.0.fe' xl031 .!£ x1031

10 10

ETOT =25 GeV8

I- 6 I-a al- I-UJ UJ"0 "0"- 4 "- 4t> t>"0 "0

2

1.5 3.0 4.5 0 1.5 3.0 4.5 6.0if xl031 if xl031

Figure 8: The cross section deduced for severalETOT bins as a function of the luminosity at whichdata were taken. The upper points are raw data.The lower points are events selected as single inter-actions (see text).

I­oI­W-0

';:; 12-0

140 140

(a) (b)

100 100"No close "Close double"

double"

60 60

20 20

0 2 0 2

The effects of the single event selection cuts werestudied by varying the luminosity at our intersectionwith a relative vertical displacement of the beams. Ateach of 4 luminosities data were taken with the ETOTtrigger, and the apparent cross section dN/dETOT +

(fSfdt) was measured for all events and for eventswith the single event selection criteria (no extra timecluster in ±200 ns). The result is shown in Figure 8for several bins of ETOT. The data without cutsshow a strong luminosity dependence, while the datawith the single interaction requirement are nearly in­dependent of luminosity. The small downward slopecorresponds to the effect of the random overlap rateexpected from the 22 mb cross section to which the 2A cluster definition is sensitive, perhaps diluted byan -5% probability for double events to be misidenti­fied as single events. The roughly linear extrapola­tion to zero of the single event curves agrees reason­ably with the extrapolation of the unselected events,which exhibits a Sf dependence between linear andquadratic. At the highest luminosity, up to 80% ofthe triggers are due to multiple interactions.

Effects of Timing Cuts with Varying Luminosity

Figure 9 gives further details of the luminositydependence. Various versions of the cuts lead to ex­trapolations to the same cross section. Most (,,75%) ofthe extra apparent cross section is due to the extraevents within ±40 ns of the nominal trigger time.The small time differences max imize contribution ofthe extra event to both the analogue trigger and inthe ADC gate. The shower counters alone can detectmuch of the spu rious cross section due to reasonablesegmentation, but they are insensitive to times out-

side the -20 to +80 ns range. It is also seen that theA counters alone carry most of the burden.

Figure 7: Tracks per event in extra vertices.7a) events with "no close double" (defined in text)7b) events with a "close double" (defined in text)

o

\,j

')

<)

.) 167

In contrast with the ETOT data, the fraction ofPAl RS events removed by a single event requirementis in close agreement with the predicted rate from ful­ly random overlaps (contributing to neither triggernor offIine fi Iter, but perhaps interferi ng with moredetailed topological investigations). At most, 10% ofthe PAIRS triggers are due to multiple interactions atthe highest luminosity.

Summary

Some general comments have been made on operatingexperience at ~ = 6xl031 cm- 2 s- 1 at a \Is = 63 GeVpp collider. The problems presented by the luminosi­ty have not been great, and pileup contributions tomost triggers used have been small. An exceptionhas been a sea rch for jet structu re with a 211 electro­magnetic calorimeter[4], where the geometry readilyaccepts random contributions; these can be removedby use of timing information. The random contribu­tions could also have been minimized at the triggerlevel by faster signal shaping, use of tower struc­ture, or vetos on non-triggering interactions. Sincethis trigger was usually taken in parallel with otherdata, the main reduction was done offline with littlepenalty. Multihit readout and good segmentation wereimportant for the success of this effort, and are ex­pected to be more important still at higher luminosityand higher multiplicity.

References

[1) M. Morpurgo, Cryogenics 17 (1977) 89.

[2] J. S. Beale et. al. Nucl. Instrum. Methods 117(1974) 501.

[3] L. Camilleri et. aI., Nucl. Instrum. Methods 156(1978) 275.

[4] A.L.S. Angelis et. aI., submitted to Phys. Lett.

168

')

)

)

)

)

OPERATION OF THE AFS AT L = 1.4 x 10 32 cm- 2 sec- i :A FIRST LOOK AT DATA AT HIGH LUMINOSITY FROM THE CERN ISR

The Axial Field Spectrometer Collaboration

(BNL - Cambridge - CERN - Copenhagen - LUND ­Pennsylvania - Pittsburgh - Queen Mary College, London

- Rutherford - Tel Aviv)

We have compared several trigger rates at the twoluminosities to see if the observed rates scale withthe luminosity. This is given below for three differ­ent triggers as the ratio of the trigger rates dividedby the luminosity for the two runs.

The high relative rate for the ETOT trigger showsthat this type of trigger is particularly sensitive totriggering on pileup events. The more selectivelocalized energy triggers are much less sensitive topileup and scale more closely with the luminosity.

In December 1982 a run was made at the CERN ISRwhich utilized the superconducting low beta quadru­poles in intersection f~ at ~he ISR and achieved aluminosity of 1.4 x 10 cm- sec- i for 26 x 26 GeV ppcollisions. At this luminosity the mean time betweeninelastic collisions is about 200 ns. A comparisonrun was alsg made at thy same energy with a luminosityof 3.0 x 10 0 cm-2 sec-. The luminosity under normalrunning conditions is typically 1.5 x 10 31 cm-2sec_i. Data were collected with the Axial Field Spec­trometer with a variety of calorimeter triggers. The

~~~~r;:~:~:n~~haS~~~~~~m~:~~:~~11:~~r c~:~~:i~~et~~~arangle range 50° < a < 130°. The shaping amplifiersused in the trigger have an integration time of 60 nsand the ADC gate for the photomultiplier signals has alength of 120 ns. The triggers ranged from a non­selective total transverse energy trigger (ETOT) tomore selective jet and single particle triggers. Thejet trigger summed the transverse energy in an azi­muthal range A~ ~ 45°.

Jet(ET in a limited region> 9 GeV)

Hi L ETOT 66%

Lo L ETOT 7%

Hi L Jet 28%

10L Jet 4%

Minimum Bias,

Normal L 4%

As an initial step we have analyzed the data fromthe two luminosities with a minimal number of cuts inorder to see if it is possible to extract the samephysics from the high luminosity data as from the lowluminosity. For this we have chosen the jet triggerdata where we expect a clean signature for theevents. We attempted an analysis with the centraldrift chamber, but unfortunately the data sufferedfrom a hardware readout problem unrelated to the high­er luminosity. Therefore, after a short review of thechamber performance, we will concentrate on the analy­sis of the data from the 2n calorimeter.

or multiple events for the ETOT and jet trigger arelisted below, along with the fraction found for aninelastic minimum bias trigger at normal luminosity.

These numbers suggest a lower pileup rate thanwould be calculated from the previously mentioned num­bers since 1) not all inelastic events have hits'inthe counters which were used and 2) only those timeswithin ± 17 ns of the event time were included in theanalysis. Since the window for double events is ap­proximately one half of the shaping amplifier integra­tion time, we can estimate the expected ratio of thetrigger rates by doubling the observed rate of pileupevents. By this method, we obtain the values of 4.5and 1.7 for the ETOT and jet triggers, respective-ly. This procedure could be used to remove most ofthe pileup events from the high luminosity data. Thetiming information from elements of the barrel hodo­scope and calorimeter could also be used in this man­ner to signify out of time events. In the case of jetand single particle triggers we expect the rate ofclean events to scale roughly with the luminosity,while for the ETGT trigger the fraction of eventswhich must be rejected due to pileup may outweigh thegain due to the increased luminosity.

1.0

1.5

Trigger Rate at ~igh/Lhigh

Trigger Rate at Llow/Llow

4

Trigger

Two Electromagnetic SingleParticle Clusters(Espc > 4 GeV)

Total Energy(ET > 28 GeV)

)

()

)

\- ,

A large fraction of the pileup events were ident­ified with the use of the ti~ing information from twoarrays of 15 1/2 x 15 1/2 cm scintillation counterssurrounding each outgoing beam pipe. There are 39separate counters in each array. After time slewingcorrections the RMS time resolution for these countersis 0.7 ns. An algorithm was developed to recognizedouble events by first pairing counters whose time waswithin 3 ns for pairs within one array, or within 4 nsfor a pair with one hit in one array and one in theother. Then the RMS for the mean time for the pairswas calculated from the times of all counters whichwere included in such pairs. A large value of thisRMS indicates the presence of a double event. Thefraction of triggers which were identified as double

Drift Chamber

The drift chamber was operated at reduced gainfor the high luminosity run in order to maintain thechamber current at its nominal value. The wire layersat small radius, which received the highest flux, wereoperated at 18% of their normal gain, while the outerwire layers were operated at 80 - 100% of their fullgain. We found the drift velocity remained unchangedat its normal value of 52 mm/ns. We were able to useour standard calibration programs with only a slightmodification and calculate a position resolution fortracks found in the chamber. This increased to 320 ~

169

per point from its normal value of 220~. The averagenumber of found tracks per event was 15.7 for the highluminosity data, compared with 14.4 for the low lumin­osity. The average track length was 30.1 cm for thehigh luminosity compared with 37.6 em for the low lum­inosity. The most difficult problem with the driftchamber data was a malfunction in one of the discrim­inator crates which rendered one quarter of the cham­ber inoperative for both luminosity runs. Mainlybecause of this we decided not to follow through witha physics analysis of the drift chamber data at thistime.

of the jet. Fig. 2 shows the PT distribution ofjets defined in this way for the two luminosities.Again the data have been normalized using the inte­grated luminosities and no cuts have been made toeliminate multiple events. One can see that above thetrigger threshold (PT ~ 10 Gev/c), the two sets ofdata agree reasonably well. Below the trigger thresh­old there is an excess of events in the high luminos­ity data as one would expect from the overlap of a lowPT event with a high PT event slightly below thethreshold causing the trigger. Again, at the highestPT there is a contribution from cosmic rays.

Calorimeter

10' -:r---r--...,...---r--;---r--r---r---:l

Pl Or .leI,We compared the PT distribution of energy clus­ters found in the uranium calorimeter for the two lum­imosities for the jet triggered data. A descriptionof the cluster finding algorithm can be found in Ref.2. In order to set an upper limit on the effect ofpileup events, we made no explicit cuts to eliminatemultiple events. Fig. 1 shows the PT distributionof the clusters appropriately normalized using theintegrated luminosities of the two runs. At low PT,there is an excess of clusters found in the high lumi­nosity data as one would expect from the overlap of anextra inelastic event with a true jet event. However,at higher PT (PT ~ 5 Gev/c) there is reasonableagreement between the two spectra. Above PT ~ 15GeV, there is a contribution from cosmic rays, partic­ularly in the low luminosity data, as we have seenbefore in other low luminosity runs. 3 Rejection ofcosmic ray background requires the data from the driftchamber. When the requirement that high PT chargedtracks from an event vertex point to the calorimeterjet is imposed, the cosmic ray background will be sub­stantially reduced.

10"

~'2

102cZf

c

~

10'

,-' .'1 = 3.0 x 10"

•• 0 •0

0

•0 e

•?*

,?t t t

y

)

)

Fig. 1. PT distribution of clusters in the AFScalorimeter at high and low luminosity.

)

)

)

18 201810 12 11

PI GeV/c

This research supported in part by the U.S. Dept.of Energy under Contract DE-AC02-76CH00016.

Fig. 2. PT distribution for jets in the AFScalorimeter. See text for details.

In conclusion, we found that selective triggerssuch as jet or single E~rticle triggers work well at aluminosity of 1.4 x 10 and do not suffer severelyfrom triggering on pileup events. In fact, the twosingle particle triggers will probably not be suscept­ible to pileup at even higher luminosity, allowing thesearch for high mass e+e- states, for example. Lessselective triggers such as the total transverse energytrigger are sensitive to pileup and contain a largefraction of triggers due to the overlap of two lowerET events. We found that our drift chamber was ableto operate in a high luminosity environment and itsperformance was not seriously degraded. This isnotable since the chamber was not designed for suchhigh luminosity. The maximum drift time at the outerradius is 560 ns. With more closely spaced wires, theperformance at high luminosity would be improved. Wewere able to compare the PT spectrum of clusters andjets found in the calorimeter and found that whilepileup events affected the number of events at lowPT due to our trigger definition, the spectra agreedat high PT'

:I. - 14 Yo IO~'

(' :I '" 30)( IO~o

10 12 I~ 18 18 ~ ~

1\ (;(,V/c

J(J~ ~

101,.

~1f1',

!,Jf"·d.>

If( ,t"

101

10"_n

We have also defined jets using the clustersaccording to a simple algorithm in order to comparethe jet PT distribution. A thrust direction was

A ifound which maximized the quantity fT'PT/ETOT' whereA +iT is the thrust direction, PT is the transversemomentum of each cluster within 6y = 1/2 and 6$ = 45°of Tand ETOT is th~ total energy. Once this direction was found the PTAof the clusters within 6y =1/2 and 6$ = 45° of T were summed to define the PT

170

o

()

o

1.

2.3.

References

H. Gordon et al., Nucl. Instr. & Meth. ~, 303(1982) •T. Akesson et al., Phys. Lett. 118B, 185 (1982).T. Akesson et al., Phys. Lett. 121B, 439 (1983).

171

RATES AND TIME STRUCTUREEXPECTED FROM IS = 40 TeV COLLIDERS

R. DieboldArgonne National Laboratory

Summary magnets), the total number of antiprotons would be

Luminosities and time structure for pp and ppcolliders were examined at the DPF Snowmass meeting,and the results are summarized here as input for thedesign of detector components and systems. Forbunched beams, the need to separate the beam orbitsbefore the next bunch meeting gives a spacing betweenbunch collisions of about 800 nsec for pp and 100 nsecfor pp. For the beam emittance assumed, as many as 25interactions per bunch collision sh3~d be possibl~~

giv~ng luminosities of up to 3 x 10 and 2.5 x 10cm- sec-l for pp and pp, respectively. Unbunched ppbeams give a better duty cycle, but require greaterbeam intensities for a given luminosity. Synchrotronradiation will damp the beams at these high energiesand may allow one to maintain constant luminosityduring a run.

Bunched Beama

For bunched beams, it is convenient to considerthe luminosity per bunch collision,

L = < n >/Otot '

where < n > is the average number of interactions perbunch colli~~gn; ~or convenience we will take 0tot =100 mb = 10 cm. If 8t is the time intervalbetween bunch collisions, the luminosity is given by

!l?= < n >/Otot 8t •

In the pp case, the beams are easily split bybending magnets at each end of the interactionregion. Allowing ±10 meters clear space for thedetector, and taking into account the relative motionof the two beams, a bunch spacing of 30 m = 100 nsecallows adequate separation at the next bunch meetingpoint. This gives

!l?(pp bunched) = < n > x 1032 cm-2 sec-l •

The optimal value for < n > depends on the detectorcapability, as well as the particular physics understudy a~d its sensitivity to backgrounds from multipleevents.

The separation of pp beams is harder and requireselectrostatic separator plates. For optimalefficiency, the separators need to be placed where the~ function is large and can thus not be located tooclose to the interaction point. Space must also beallowed for the beam orbits to drift apart (betatronphase advance), so that the minimum distance betweenbunches is roughly 240 m = 800 nsec, and

!i?(pp bunched) = 1.25 < n > x 1031 cm-2 sec-l •

At Snowmass the "conventional" collider grouplassumed a normalized emittance of En = 10~ x 10-6meters; for 20-TeV beams, this gives

N = 1.4 x 1010 ~/bunch

as the number of antiprotons (or protons) required perbunch. For a 60-km circumference ring (8 to 10 Tesla

172

3.5 .f<U> x 1012

With an extension of the technolo,y beingdeveloped for TeV I, it was estimated that it shouldbe possible to collect 1012 p/hour. It is traditionalin the design of hadron colliders to assume a limitfor the so-called beam-beam tune shift of 0.005.Using this criteria, values of < n > up to 25 arepossible with the emittance assumed; this wouldrequire 18 hours of p collection.

Even higher values of < n > are possible if theemittance is suitably increased to keep the tune shiftunder control, but at the expense of a linear scalingof the number of pIS with < n >. A low-field ring of2.5-Tesla magnets would have three or four times thecircumference and thus require as many times thenumber of p's. In C&nsiderations of such a machine,the "low-cost" group at Snowmass assumed that a beamemittance a factor of 10 smaller could be obtained(N ~ 1/ IE for fixed luminosity). At this smalleremittance, however, the beam-beam limit already comesinto play at < n > = 2.5.

Unbunched Beams

For unbunched beams, the duty cycle for thedetectors can approach 100%, but with the need formore beam particles effectively excluding the ppoption. While the luminosity is inverselyproportional to a, the crossing angle, this angle mustbe ~ 60 ~rad if the beams are to separate cleanlyS(each beam is diverging from its tightly focussedwaist at the center of the interaction region). Thisleads to a "diamond" with rms length OJ/, = 0.28 m and

l4~N (unbunched) = 4 x 10 {~/lO~~/beamtot

for En = 10~ x 10-6 m and a 10-Tesla ring. Includingthe long-range contributionsS to the beam-be~m tu~eshift gives a nominal limit of !l? ~ 1.3 x 103 cm­sec- , but this would require 1.2 A in each ring, or1.4 x 1015 protons/ring (5000 megajoules) for a 60 kmcircumference.

A summary of luminosity and the number of beamparticles required is given in Table I for thedifferent modes of operation. If the total number ofbeam particles turns out to be the limiting factor(for example, if the abort system can only handle somany gigajoules), the bunched beam would give about 20times the luminosity of the unbunched beam.

Synchrotron RadiationAt the SPS collider there is a contribution to the

duty factor from the luminosity lifetime of about 16hours; the luminosity falls from its peak value at thebeginning of the run to give a lower average value.At 20-TeV gynchrotron radiation damping becomesimportant, and if no other factors were at work, theluminosity would increase by a factor of e = 2.72every six hours for a 10-Tesla machine (four days for2.5 Tesla). With a high field machine, one may beable to set the luminosity to the desired value andthen keep it constant for the length of the run.

)

)

)

o

()

References

1. R. Diebold et al., '" Conventional' 20-Tev, 10-Teslap:1'p Colliders", Snowmass Proceedings, p. 307.

2. R. Diebold, "Detectors and Luminosity for HadronCo11iders," Proc. Maryland DPF Meeting, Oct. 1982.

3. G. R. Lambertson and Ch. W. Leemann, "AntiprotonProduction and Accumulation for a 20-TeV ppCo11ider", Snowmass Proceedings, p. 338.

4. R. Huson et al., "20-TeV Colliding BeamFacilities: New, Low-Cost Approaches," SnowmassProceedings, p. 315.

5. L. Smith, "Luminosity of Continuous Beams withCrossing Angle," Snowmass Proceedings, p. 351.

6. L. W. Jones, "Synchrotron Radiation in Multi-TeVProton Synchrotrons," Snowmass Proceedings,p. 345.

Table I. Typical time structure, luminosity, and number of particles per beam forthree modes of operation. For bunched beams, At is the time between bunchcollisions; in the unbunched case, At is the memory time of the detector,taken to be 30 nsec as an example for this table. A circumference of 60 kmwas assumed to calculate N.

: )

()

\, /

)

< n >= 1 < n >= 25At Sf N Sf N

(nsec)( 10

32)

(1014/beam)( 10

32)

(1014/beam)

cm-2 sec-1 cm-2 sec-1

Bunched pp 800 0.12 0.04 3 0.2

Bunched pp 100 1.0 0.3 25 1.4

Unbunched pp (30) 3.3 2 80 11

173

ADVANTAGES OF SPATIAL AND TEMPORAL SEGMENTATION FOR DETECTORSAT HIGH LUMINOSITY CW COLLIDERS

Michael J. Tannenbaum

Brookhaven National Laboratory

The highest rate hard process at high energycolliders &/s > 540 GeV) is high PT jet production.At 10 33cm- sec- 1 luminosity, the production rate forjets with PT > 100 Gev/c is greater than 1 persecond. 1 The-typical 100 Gev/c jet has a meanmultiplicitr of 22 charged particles in a cone of N15°half angle, which corresponds to a mean chargedparticle density in the jet of 80 particles per8y 8~. This is over 100 times the mean mimimum-biascharged particle density of 0.6 particles per 8y 8~.

A major detector problem at high energy collidersindependent of the luminosity is the ability toresolve particles inside jets. Only in very specialcases, e.g., UA1 W± production 2 with PT N Mw/2will enough of the signal be clear of jets, so thatevents with tracks in jets can be summarily rejected.Eventually, detectors will have to cope with tracks injets. The high track density requires good track-pairresolution and efficiency, which implies wire chamberswith drift distances of 2 mm or less, independent ofthe luminosity.3 The small drift distances requiredby the jet physics are very helpful at highluminosities4 since the maximum drift time of 40 nseclimits the pile-u~ to an average of 2 events at aluminosity of 103 cm-2sec- 1•

Pile-up reduction in ET triggers by using fasttiming information was presented at this meeting bythe COR R-110 experiment at the CERN ISR. 11 Beforetiming cuts, the ET spectrum had a huge luminositydependence, increasing linearly with luminosit~ br afactor of four over the range 1 to 6 x 10 31cm- s- •The use of precision timing (N 10 nsec resolution) toreject events which had additional, out-of-time,events within the 200 nsec ADC gate eliminated thepile~up. The spectrum became independent ofluminosity at the expense of 35% dead time at6 x 10 31cm-2s- 1• It should be noted that the majorpart of this detector had no rapidity segmentation[8~ = 0.11, 8y = 2.2] and furthermore that timinginformation was not available on all segments. Ifthere had been finer segmentation (in rapidity) andprecision timing on all the segments in the R-110detector, then the energy from individual out-of-timesegments could have been eliminated rather thanrejecting entire events. This would have eliminatedboth the pile-up and the dead time.

This research was supported by the U.S.Department of Energy under contractDE-AC-02-76CH00016.

)

ReferencesThe segmentation of "calorimeters is also set by

jet physics and the desire to resolve and meas~re

collimated jets. UA2 uses a calorimeter tower with8~ = 0.25, 8y = 0.18; CDF 6 has 8~ = 0.25, 8y = 0.11;and Sadoulet has discussed 8~ = 0.10, 8y = 0.10 for a20 TeV detector. 7 Segmentation of this order3~ssufficient to eliminate pile-up effects at 10 • Adetector considered at Snowmass4 covered the fullazimuth and the rapidity interval -2 < y < + 2 with a40 x 40 array of towers each covering-8~ ~ 0.168y =0.10. Pile-up from an average of 10 overlappingevents in a 200 nsec ADC gate could be eliminated forjets by requiring individual towers to have greaterthan 1 Gev/c in PT' This simple algorithm worksbecause jets are collimated clusters of high PTparticles while the pile-up comes from large numbersof particles with <PT> NO.4 Gev/c which areuniformly distributed over the detector.

An additional feature of these highly spatiallysegmented detectors which can be exploited in CWcolliders is to combine the spatial segmentation withtime segmentation. The fast timing from the leadingedge of the calorimeter signals8 can be used tolocalize the time of the energy deposition in thecalorimeter to N5 nsec. Calorimeter towers with noin-time hits to a precision of 5 nsec are simply

9ignored, with no loss in physics. Pile-up during theN 200 nsec ADC gate is only a problem if one of theN 30 out of 1600 towers4 which has an in-time hit isalso struck out-of-time. This reduces the pile-upeffect considerably, even for ET triggers. Thepile-up can be further reduced by using "flash" ADC'sto analyze the wave-form2 and extract the in-timeenergy or by speeding up the calorimeter integrationtimes. 10

174

1.2.

3.4.

5.6.7.8.9.

10.11.

F.E. Paige, "High PT Jets," These Proceedings.C. Rubbia, These Proceedings; see alsoCERN-EP/83-13 to be published in Physics Letters,and CERN-EP/82-170 IEEE Nuclear Science Symposium1982.D. Hartill, Tracking Summary, These Proceedings.H.A. Gordon, et al.~ "Reasons Experiments can beperformed at L = 10 3cm-2sec-1," DPF, Snowmass1982.M. Banner, These Proceedings.CDF Design Report, August 1981, FNAL.B. Barish, Systems Summary, These Proceedings.M. Shochet, Trigger Summary, These Proceedings.See however: C.Y. Chang, Search for Heavy StableParticles, These Proceedings.R.B. Palmer, et al., These Proceedings.J. T. Linnemann, These Proceedings; see alsoA.L.S. Angelis et al., Observations of JetStructure in High ET Events, submitted toPhysics Letters.

)

)

oMONTE CARLO SIMULATION OF THE EFFECT OF PILEUP

H.A. Gordon, R.A. Johnson, S.A. Kahn, M.J. Murtagh and D.P. Weygand

()Brookhaven National Laboratory, Upton, NY 11973

()

()

The work reported at Snowmass l showed that thepileup of even an average of 10 minimum bias events(Poisson distributed) would not affect the trigger ofeither jet events or localized electromagnetictriggers ( ~o or e±). We have continued this typeof study with a modification to ISAJET2 which allowssoft jets to be generated above PT = 1 GeV such thatthe integral of the two jet cross section equals thetotal inelastic cross section of '" 60 mb at Is = 800GeV. Each event with two jets also has a so-calledminimum bias event superimposed which essentiallyarises from the beams fragmenting. With thismodification it was easy to assess the effect of thepileup of events on a jet trigger.

We considered a calorimeter covering Iyl < 2subdivided into 40 cells in y (Ay = 0.1) and into 40cells in $(6$ = 9°). The trigger used the searchlightmethod, looking for the section of the calorimeterwhich had the largest transverse momentum. Thissection was defined as 6y = 1 and 6$ = 90°. The sizeof this secion was always large enough to contain thecore of the jet. The curve in Figure 1 shows thecross section for triggering vs. PT' Also shown isthe cross section for an average number of 10 pileupevents. The major contribution to pile-up triggers inISAJET type events at large transverse momentum comesfrom one jet near the trigger threshold in conjunctionwith several low PT jets, as opposed to two mediumPT jets. 3

For the cross section shown, the first jet was chosenfrom the full jet spectrum and then n jets with PT <10 GeV/c were added to the event (n was chosen from aPoisson distribution with mean n =10). This back­ground can be reduced by demanding that each of thetowers contributing to the trigger have PT > 1 GeV.The cross section obtained after imposing thisrequirement is given by the lower set of points inFigure 1.

Of course the PT plotted in Figure 1 is onlythat observed in the selected section of thecalorimeter. To obtain the jet cross section is amuch more complicated job. 4 One must make correctionsfor the geometric boundary, the energy leakage, thenon uniformity of the electromagnetic versus hadronicresponse, etc. These effects are at least asimportant as the effect of pileup. Figure 2a showsthe average value of the parton PT obtained with agiven PT trigger threshold.

50,0 ....-------------------....

(a)

1,11(J , f---,---,- --,----,---,.---1

Btl 1110 1-,:11 I-Ill tliO IHIi :'IJO

1\ C;"\,l

4010 20 30

P t (trigger) GeV IeO.O+----....,~---..,..----T"""---_l

o

20.0

30.0

10.0

40.0

\"\

,.

e\

•10""

lor'°'1 ".)

>vCJ0

"n:

"(Jt)''0

U

Figure 2b shows the efficiency of measuring thetrue parton PT with this trigger.

: )

Figure 1 Cross section for triggering on jetsversus PT' Curve is for the alogorithmdescribed in the text. Solid circles arefor events with <n> = 10 events in addi­tion to the high PT jet. Open circlesare for <n> = 10 events in addition tothe high PT jet with PT > 1 GeV/crequired for each cell.

Figure 2a) Average PT parton vs. PT of triggerusing the algorithm described in text.

,) 175

100.0_-----------------~::II

)

)

MIN. BIAS OVERLAP <~>=7.5

NO MIN. BIAS OVERLAP

3 4 5 6[P, IN 2 RINGS AROUND TRIGGER CELL

eliminate electrons from jets would be to requirelittle energy in the two rings of cells adjacent towhere the electron hit. The distribution of thetransverse energy found in such events where only oneevent is analyzed at a time is shown in Fig. 3a. Morethan 95% of the events have EPT < 1 GeV in thesurrounding cells. The effect of pileup is shown inFig. 3b where the distribution of associated energy isbroader. However only ~ 20% of the events would belost with a EPT < 1 GeV cut.

·'0

i'!z\3! 20. -l.U

w-e17.5

ciZ

IS.

12.5

10.

7.5

5.

2.5

'-------lO.

.10

i'!Z 12\3!l.U

w-eci

10

Z

605030 40

P l (true) GeV Ie20

Efficiency as a function of PTof the primary parton for a triggerthreshold of 25 GeV/c. Both a) and b)have <n> = 10 events in addition to thehigh PT jet events.

O.O~-~-,....:....-...,.---'"I'""---,__--~10

20.0

A. The Effect of Pileup for w± Physics

40.0

80.0

(b)

60.0

Figure 2b)

Two other physics topics were studied at BNL toassess the effect of pileup:

Now that it is apparent that the W+ eV has beendiscovered at CERN one may ask whether high luminosityis required for further study of the Wand if so whatpossible effects of high rate might hamper such a highstatistics study. There are at least three topicswhich require a large number of W' s : 1) a precisiondetermination of the ZO - Wmass difference is ofcritical importance to determine the ~ parameter inthe standard model; 2) the study of W-y productionwould measure the magnetic moment of the Was well astesting the non-Abelian character of the gauge theoryand 3) the production of w+w- pairs would also testthe non-Abelian gauge theory as well as open a newdecay channel for heavy particles, for example, aheavy standard Higgs HO + w+w-. In the latter twocases the cross sections are rather small « 10- 36cm2)therefore high luminosity is required.

J 4 5 6[P, IN 2 RINGS AROUND TRIGGER CELL

Figure 4 shows the effect of pileup on themeasurement of the PT of the electron from thecalorimeter. The PT spectrum is broadened slightly,however an excellent determination of the Jacobianpeak is still possible.

Kahn et al. 5 have studied the effect of extraminimum bias events on the identification of W's. Ina slightly different calorimeter with cell size of AS= 5° and A~ = 9° covering the central 4 units ofrapidity, the trigger would be electromagnetictransverse energy deposit in one cell greater thansome threshold, e.g., 20 GeV. Then a track withmomentum close to that energy would be required topoint to the correct cell and a particle identifiersuch as a transition radiation detector would signifythe presence of an electron. The next selection to

Figure 3 (a) Transverse energy in the vicinity ofelectrons from W± + e±v.(b) Same as a) but with <n> = 7.5 eventsin addition to the W.

176 )

403530

:t=2.xlO"

BACKGROUND

X.< .5

15 20 25

Pout GeV/c

M~ = 100 GeVg

\\\

10

Gluino signal compared with backgroundswith and without pileup.

510-1

,,\2--.,....-.....,..--~-....,...._....l...r--- .......----..---1o

Figure 5

This research was supported by the U.S.Department of Energy under contractDE-AC-02-76CH00016.

We conclude that the effect of extra events for amissing PT trigger worsens the signal to noise--butthe signal for gluinos is still clear. There has beensome objection to this analysis in that there may beother more perverse backgrounds. However it appearsthat the effect of pileup would not be very §evere.Those unknown backgrounds would presumably obscure thesignal at any luminosity.

background. This can be understood since it isvirtually impossible for a normal two jet event tohave, for example, one jet at PT ~ 100 GeV and theother at PT = 40 with a relative 20 GeV oftransverse momentum with respect to the first. Theeffect of pileup due to overlapping minimum biasevents is included in the background calculation. Aluminosity of L = 2 x 1032cm-2sec-1 which correpondsto an average of 2 added events (Poisson distributed)is assumed. As the luminosity is increased thebackground increases significantl~, but as shown inFigure 5, even at L = 1 x 10 33cm- sec-I, beyond Pout=25 GeV/c there are less than 500 background events ascompared to 9000 signal, for Mg = 100 GeV/c 2•

••

50 60P,(e') GeV/c

•o

o 8•

•o~

.03020'0

o

PT of the electron from W± + etvwith and without <n> = 7.5 additionalevents piled up.

••• •0 00•_.0 0

·0 •••••00 0

0 0 °0 •

•• • ~. 0• 0 0 0 •00· . 0

•• 0.0.0. -.0 0

•• eoo• 00

00

.~

o

@ W+"'e+1I with min. bios overlap <n>=7.5

o

oo

-I'0

10-2 f-

-.10

Figure 4

XE

the component out of the plane is called Pout.Events are selected with XE < 0.5, and the Poutdistribution is plotted for both the gluino signal andthe background in Fig. 5. Above Pout = 25 GeV; thegluino signal is orders of magnitude above the

)

) B. The Effect of Pileup onSupersymmetric Particle Detection

Aronson et al. 6 have shown that a simpleexperimental signature exists for finding gluinoswhich are pair produced in hadron interactions.Basically each gluino decays into a photino, which isnot seen in the detector, plus high PT jets. Thisresults in a large amount of missing PT' Thebackground comes from normal QCD jet events where onejet is lost geometrically or an energetic neutrinocarries a large amount of PT' To select gluinoevents, a sphericity analysis is used to find twohadron jets with momenta PI and P2' A plane isdefined by the beam and the larger momentum jet Pl'The PT balance in the plane is measured by

, )

o

o

o

o

()

. ) 177

References

1. H.A. Gordon, T. Ludlam, E. Platner andM.J. Tannenbaum, Reasons Experiment~ can beperformed at a pp machine at L = 10 3cm-2sec-l,Proceedings of DPF Summer Workshop on FutureFacilities, 1982, Snowmass, Colorado.

2. F.E. Paige and S.D. Protopopescu, ISAJET: AMonte Carlo Event Generator for pp and ppInteractions, Version 3, Proceedings of DPFSummer Workshop on Future Facilities, 1982,Snowmass, Colorado.

3. R. A. Johnson, Triggering at High Luminosity:Fake Triggers from Pileup, these Proceedings.

4. See, for example, T. Akesson et al., Phys.Lett. 118B, 185, 1982.

5. P. Grannis, T. Killian, S. Kahn and M.J. Murtagh,BNL Report (in preparation).

6. S. Aronson et al. DPF Snowmass Proceedings... 1982. p. 505

7. L. Littenberg and D. Weygand, The Effect ofPileup on the Gluino Signal, BNL Internal Report,1983.

178

)

.)

)

o

()

o

()

: )

CALORIMETRIC PHYSICS IN THEPRESENCE OF MULTIPLE EVENTS

John YohFermilab

Hadron co11iders of the present andfuture generations will have to face theproblem of having multiple events withinthe resolution time of the detector. Acritical issue is the effect of multipleevents on physics capabilities. This noteaddresses this issue. Most of theconclusions are based on Monte Carlostudies by H. Gordon, R. Johnson and J.Yoh, as well as on other considerations.H. Gordon has separately submitted hisresults and conc1usions7 the conclusionsexpressed in this paper are my own.

Several of the areas addressed by thisnote are listed below and are covered inthe following pages7 we give here briefconclusions of each of these areas:

1. ASSUMPTIONS AND PROCEDURE --Most of theMonte Carlo studies involveElectromagnetic (EM) and Hadroncalorimetry towers covering ~ andfinely segmented into bins in Phi andEta (the pseudo-rapidity) 7 theconclusions are based on variouscluster criteria applied to eventsgenerated by various monte carlo modelssuch as ISAJET, FOX (full QeD) and MB1(a minimum bias Monte Carlo fitted toISR and SPS co11ider data) at a centerof mass energy of 1 TeV.

2. SINGLE JET PHYSICS --One can triggerand measure the highest Pt narrow jetin an event using a simple windowalgorithm to define a c1uster7clusters with Pt above 10-15 GeV arenot faked by multiple minimum biasevents and thus constitute a clearsignature for hard scattering orproduction of massive partic1es7 theeffect of multiple events is to simplyadd some energy (typically 1.5 timesthe ambient background of 2.5 GeV Ptper unit rapidity per extraneous event)to the Pt of the cluster, which canreadily be corrected for by MonteCarlo.

3. MULTIPLE JET PHYSICS --However, QeDinvariably leads to gluonbremsstrahlung, many of which will besoft (e.g., below 7 GeV in Pt) 7 theselow energy clusters can frequently befaked by mu~tiple ev7nts. Typically,use of a s1ngle wlndow in eachhemisphere only picks up 80-90% of thehard scattering Pt7 this leads to agradu~l threshold in triggering and acomp11cated model dependent unfoldingprocedure to obtain the real hardscattering cross section if one onlyme~sur7s the two leading clusters.T~1S 1S even more serious for theslgnal of heavy mass decaying into many

179

jets (e.g., Higgs, W or Z into heavyquarks) • Preliminary Monte Carlostudies of the process W decay into T +Bbar shows that while th W mass can bereconstructed to an accuracy of betterthan 14% in the absence of multipleevents, adding 2 Minimum Bias eventsinto the W event seriously smears theresolution as well as contributesserious tails which could obscure thesignature. 10 extra events totallyobscure the signal.

4-. MISSING ET PHYSICS --The Pt ofneutrinos could be determined from themissing Et of the event and wouldconstitute an interesting signature for1eptonic decays of W, Z and gluinos.In a well designed detector, themeasurement accuracy of missing Etshould be dominated by energyresolution, in which case multipleevents would contribute more Et andthus larger error in Et7 e.g., anevent with Z into 2 neutrinos couldhave Et totalling, say 50 GeV (withperhaps a 30 GeV gluon), while 10 extraminimum bias events would giveadditional total Et of 250 GeV! On theother.hand, there are many other sourceof mlsmeasurement of missing Et, suchas geometry, leakage, missing neutrinosfrom conventional sources, etc., andthese could well dominate7 in thatcase, the effect of mUltiple events maybe less crucial.

5. ISOLATED LEPTON AND PHOTON PHYSICS--Isolated leptons from W or Z decaysand isolated photons from W-~production are some of the mostinteresting physics. Multiple eventswould increase the probability that thesignal tower or adjacent ones wouldhave particles hitting them, thusvetoing the signal. As a particularexample, demanding no other particleswithin a 5 degree cone of the signalwould throw out 2% of the signal perextraneous event7 one would thus haveto investigate the signature which mayallow low Pt particles nearby.

6. MISCELLANEOUS COMMENTS We brieflycomment on four other topics,advantages of continuous vs. bunchedbeams, effect of mUltiple events ontracking and on data reduction, and onrecognition of multiple events.

In conclusion, while multiple eventswould not obscure the fact that hardscattering or heavy mass production hasoccurred, the measurement precision, andthus the signal to background will beseriously compromised. Most newcalorimetric physics in hadron co11idersare crucially dependent on signal tobackground7 thus the effect of multipleevents could be fata17 at a particularmass, the ratio of parton-parton luminosityto minimum bias events can be increased bygoing to higher hadronic energy, thusminimizing this problem.

(Ib)

100 200

p.,. (D'S)

)

)

:)

to+10

_+0-·-+10

(IQ.)

10 20 30I'r(DIJ$.) ~

.37/0

Figure 1 - Distribution of observed Ptof the highest Pt cluster observed in aevent. (la) - Minimum bias and softjet events from MBI using Yoh's narrowwindow algorithm. (lb) -ISAJET 2 Jetevents using Gordon's wider windowalgorithm.

10-4

(1) ASSUMPTIONS AND PROCEDURE

H. Gordon and J. Yoh independentlystudied the effect of multiple events oncalorimetry physics7 the size of windowused and the event generation Monte Carloare different, but the conclusion on singlejet performance is similar.

Events are generated by various Montecarlo programs. Minimum bias events aregenerated by ISAJET and MB17 ISAJETassumes exponential-like Pt dependence

while MBl fits the Pt dependence to ISR andUAl observed distributions and thusincludes, on the average, some contributionfrom soft jets. Minimum bias events fromISAJET program does not include enoughmedium Pt (e.g., 1-5 Gev) particles asindicated by UAl results and is thusinadequate7 one would have to include somefraction of events with soft jets in orderto realistically simulate the expected realminimum bias events. These minimum biasevents are mixed into signal events to seethe effect of multiple events.

The studies involve finding clustersdefined by a certain size window in a towergeometry calorimeter. H. Gordon uses awide window of 90 degrees in azimuth and 1unit of rapidity, while J. Yoh uses anarrow window with 20 degrees in Phi and .4in Eta.

The effect of multiple events arestudied by comparing signal distribution inPt, mass or Sum Pt with the signal eventsonly ( the +0 case), with an average of 2extraneous minimum bias events mixed inwith the signal event (the +2 case) andwith 10 extraneous events (+10). Thenumber of extraneous events are determinedby poisson distribution. The case of +2would correspond to a resolution time of 40nsec and an instantaneous luminosity of l~~

Signal events investigated includes2-JET events from ISAJET and FOX montecarlos, and W into T + Bbar from ISAJET.The FOX monte carlo, developed by Field,Fox, Wolfram and others, includes effectsof both initial state and final state gluonbremsstrahlung 7 although ISAJET onlyinclude final state bremsstrahlung, theeffect in the central region is probablynot too significant.

JlL SINGLE JET PHYSICS

The simplest trigger and analysisa~gorithm one can imagine is to set a fixedSlze wlndow in Phi and Eta and search thecalorimeter for the largest Pt deposited bya event within that window at all possiblelocations of the window (limited by thetower boundaries). Figure la and lb showsthe observed Pt spectrum of

cluster. The effect of 10 extraneousminimum bias events is-merely to shift theenergy scale of the spectrum by an amountroughly 1.5 times the expected ambientbackground from the average energy flow of2.5 Gev transverse energy per unit rapidityper event. This corresponds to about 1 Gevin Yoh's algorithm and 8 Gev in Gordon'salgorithm7 the slope of the spectrumremains unchanged. While this represents alarge factor in rate at fixed Et thresholddue to the steep slope, a change in thethreshold by the given amount wouldmaintain the trigger rate.

180

:0

1.0

.8

.,0

t\U

While this technique of finding thewindow with the maximum transverse energyin the event is excellent in recognition ofhard scattering or heavy mass production(and thus quite adequate for triggeringpurposes), the observed cluster energy onlyrepresent a fraction of the real hardscattering Pt, as is shown in figure ~.

Typically, 20% of the transverse energy 1Soutside the window due primarily to gluonbremsstrahlung and parton hadronization.Thus the threshold, in terms of real Pt ofthe hard scattering, is rather gradua11one need to go to at least 1.5 the nominalthreshold before efficiency becomes greaterthan 95%.

o

o

o

20

Figure 2cluster PtFOXjets vs.scattering.

Efficiency atthreshold of

the Pt of

a nominal50 GeV fdr

the hard

changes significantly, even for the +2case. Figure 4a and 4b shows thepreliminary attempt to reconstruct the Wmass by forming the mass combining allclusters above a certain threshold, for thethree intensity cases. Figure 4a uses acriteria requiring that the clusters haveminimum of 2 Gev Et and 1.5 Gev in a singletower. Figure 4b raises the cluster Etthreshold to 4 GeV. The percentage error(determined by FWHM divided by 2.34) is 13%for the first case and 22% for the secondcase. For the first case, any extraneousevents apparantly will wash out the signa11the second case appears immune to the 2extra event case, but is washed out in thecase of 10 extra events. Thus, theconclusion is clear at minimum,extraneous events will force one to adopt aalgorithm with a higher threshold so thatmass resolution will be deteriorated.

We have also studied an algorithm tomeaS1ure the hard scattering Pt of FOXjetevents. Figure 5 shows the observed Sum ofPt of all clusters found for the threeluminosity cases. Again, a deteriorationis observed with multiple events. Thissignal is less sensitive to mUltiple eventsthan for the mass signal since low Ptclusters at small polar angle contributesslightly to Sum Pt but could contributeenormously to Mass due to the cosinefactor.

While these algorithms are quitesimplistic, one can envisage moresophisticated algorithms, perhaps withvariable size windows, which can perhapspick up energy near the clusters. Suchalgorithms could result in better

The most interesting jet physics arethe understanding of the QeD nature of hardscattering and the search for heavyparticles in their decays to jets (andleptons) • This physics requires theprecision measurement, not only of .theleading partons, but also of the other Jetsin the events. Here, extraneous events,being able to fake clusters with transverseenergy up to approximately 7 GeV, leads toa serious deterioration in physicscapabilities.

We have studied (i) W decays into a 35Gev T and a 5 Gev B generated by ISAJET and(ii) 2 JET production at Pt of 50 + 50 Gevgenerated by FOX'S full QeD Monte car~o.These two types of signals are stud1esunder three luminosity cases -- +0, +2, and+10 extraneous events.

Figure 3 shows the number of clustersfound as a function of cluster Et for the Wevents for three cases -- +0, +2, +10. The

1 t Per event near 5 Gevn~mber of c.us ers

Figure 3 - Number of clusters found perW into T Bbar event vs. the observedcluster Pt. The three luminosity casesare +0, +2 and +10.

- +0-- +2-_.- +\0

10 20fr~

.,

fO.D

physics isextraneous

Hence, the leading partonnot seriously affected byevents.

,)

,)

:)

) 181

)

)

)

l!L MISSING ET PHYSICS

Another serious consequence ofmultiple events is that it would make thetask of lepton identification moredifficult. Many proponents of missing Ptphysics stress the absolute necessity ofbeing able to identify events with chargedleptons (at least above a certain Pt) andthus being able to discard these events;these charged leptons often imply heavyquark production and thus imply that theevent will also likely have neutrinos.Multiple events would enable electrons,through overlap with hadrons, to simulatehadrons; in addition, photon conversionwould create large numbers of electronpositron pairs. This problem is likely tobe critical for non-magnetic detectors.

Resoution for missing Et can besmeared by many sources besides the energyresolution of the calorimeter. Othersources of smearing are leakage down thebeam hole, punch through leakage, missingneutrinos from conventional sources (whichmay be vetoed by throwing out all eventswith electrons or muons, a formidabletask), dead or inefficient areas, andnon-uniformity due to attenuation and dueto difference in EM and hadronic showerresponse.

Assuming that the dominant source· ofsmearing is in fact the energy resolution,then the error in missing Et scales assquare root of Sum Et. Typical minimumbias events have Sum Et of 25 Gev, whilethe two signals mentioned above, assuminganother 20 Gev Et ambient background in theevents, would be 50 Gev and 170 Gevrespectively. Thus, the increase in errorin missing Pt for +2 (+10) for the twocases would corresponds to factors of 1.4(2.5) and 1.14 (1.6). This deterioration

could be crucial.

One of the most exciting physicspossibilities in hadronic colliders is thesearch for new heavy particles which decayinto jets and invisible particles.Examples are Z into 2 neutrinos producedalong with an energetic gluon, productionof gluinos, squarks, leptoquarks, etc.These events are charaterized by one ormore jets with large missing Et. Twoexamples are a single 30 Gev gluon with 30Gev missing Et, and two jets 100 Gev and 50Gev in Pt and non co1inear in the azimuthalview; these are signatures of Z into 2neutrinos and gluino pair production.

- ..... _--. ...-..--.-.

(4-b) -+()-- +2-.- +10

t~ \

\\" '".~.--.-

4b goM-+-

Figure 5 - Reconstructed Summed Pt in50 + 50 Gev FOXjet events for threeluminosity conditions; all observedclusters with Pt above 2 Gev is summed.

-+0--+2-.- +10

- +0_.. +~_.. ""0

Figure 4 Reconstructed W massdistribution for the three luminositycases. (4a) -With cluster threshold at2 Gev Pt. {4b)-With cluster thresholdat 4 Gev Pt.

resolution, perhaps 10% for the W mass.However, it is likely that these algorithmswill be even more sensitive to multipleevents; we have tried a simple algorithmwhich attempt to sum all neighboors as longas they exceed a.4 Gev threshold; thisalgorithm fails abysmally in the case of+10, by having the cluster grow to a sizecomparable to the entire array.

~ ISOLATED LEPTON AND PHOTON PHYSICS

Another important physics signature isthat of an isolated energetic lepton orphoton. Background rejection is usuallymuch better for leptons which are isolated,although one could in some instancestolerate low energy particles nearby.

)

182

o

()

o

o

()

()

The effect of multiple events caneasily be determined by the fact that themultiplicity is about 5 charged particlesand 3 neutral particles per unit rapidity.In the case of a tower calorimeter withtower size 10 degrees and .1 in rapidity,the number of towers per unit rapidity is360, leading to an occupancy rate of 1 in45 towers. Thus, demanding that the signaltower have no other particle would cause a2% loss per extraneous event, if we wantto demand absence of particle in the 8neighboring towers as well, theinefficiency increases to about 20%.Conversely, if one could tolerate particleswith pt below 1 Gev, then the situation isroughly 4 times better. The problems ofhadronic showers are somewhat worse sincemost showers would illuminate severaltowers.

Of greater concern is the situationwith the tracking requirement for thelepton. Backgrounds to electrons such ascharged-neutral overlap may be moreserious. It may also be more difficult toremove photon conversion background bysearching for the other electron.

Jil MISCELLANEOUS COMMENTS

We briefly comment on several otherissues of relevance to mUltiple events.These includes thoughts on (i) Advantagesof CW beams vs. bunched beam, (ii) Effectsof multiple events on tracking, (iii)Effects of multiple events on datareduction and processing, and (iv)Recognition of multiple events.

(i) Advantages of continuous beam vs.bunched beams

algorithms, whose time requirements goes asthe square of the number of hits, must thusbe discarded.

Another effect is the inefficiencyinduced by the extra tracks, this seems tobe not a very important effect, in ourstudies of an isolated high Pt track in aminimum bias event, the efficiency for thehigh Pt track at large anq1es exceed 99%,thus extra events are un11ke1y to causesignificant loss. For the case of the jetcore, the density there is almost entirelydue to the jet itself, extraneous eventsare unlikely to be too important, althoughit may be the straw that breaks the camel'sback. A more serious problem is that extratracks always have some probability ofrobbing a good hit from an important track,this can be ameliorated by increasing thenumber of measurements.

A third issue is the effect on a highPt trigger, we have investigated a high Pttrigger algorithm involving using chambercells as hodoscopes, and demanding astraight track inside a solenoid. Aparticular implementation would give a10%-90% threshold of 4-7 Gev in Pt, using11 planes, less than 1 in 500 minimum biasevents would pass this criteria. However,typical occupancy in the intermediateplanes within the search road is about 8%,thus, adding 10 extra events is bound toswamp this particular algorithm. Ofcourse, a more sophisticated algorithm(presumably much more expensive) couldprobably provide a good rejection.

No dedicated studies on theseimportant issues have been made. It wouldbe important to determine how much moresophisticated the tracking system have tobe, and how much more computer time onewould need to reconstruct the events.

(iv) Recognition of multiple events

In a scenario in which the averagenumber of events per resolution time 1Sless than 1, one could try to bypass all of

Having multiple events is definitelygoing to increase the load on dataacquisition and reduction. Not only is theinformation per event significantly larger,but also the tendency to triggerpreferentially on multiple events and thenecessity of having larger numbers ofchannels will likely cause an order ofmagnitude increase in data rate comparedwith current large detectors. Thus, it iscrucial to be able to process the data at alower level.

Data processing effortto be an order of magnitudediscussed earlier, tracktime will greatly increaseof extraneous hits.

andreduction

is also likelyhigher. As was

reconstructionin the presence

data(iii) Effects onprocessingOne could reduce the resolution time

down to 20-50 nsec, this would reduce theaverage number of events within theresolution time, in the case of continuousbeam (CW), one could further use flash ADCto resolve contributions to the events frommore than +-5 to 10 nsec away if one has100-200 Mhz flash ADC's. While this isexpensive, it is a serious advantage.Nevertheless, shortening the gate time andpile up is bound to degrade the energyresolution.

(ii) Effects of multiple events on tracking

The extraneous hits from multipleevents will clearly have some effect onreconstruction, inefficiency, triggering,etc. The amount of reconstruction timewould be increased significantly in thepresence of excess hits, this is ourexperience when we add noise hits to ourmonte carlo data sample. Certain

On the other hand, one would have toprovide a t-zero for drift chambers in thecontinuous beam case, leading to largeamounts of fast scinti1lators.

o

,.)

u

183

the above problems by trying to recognizemultiple events and then throwing them out.Many ideas exists; one such idea is tohave a tracking system near the beam pipewhose purpose is to measure the multiplevertices; one could then hope to be ableto recognize multiple interactions on lineand discard them.

CONCLUSIONS

The effect of multiple events withinthe resolution time of a detector must beinvestigated much more thoroughly. Ourstudies show that capabilities for physicsmeasurements are seriously compromised,especially for mUltiple jets.Unfortunately, due to QCD and the existenceof heavy flavors, almost all interestinghadronic final state physics involvedmultiple jets. Whether this deteriorationwill wash out a particular signal willdepend on signal to background for thatprocess. In hadronic colliders, where jetphysics signals are not particularlystriking, the effect of multiple eventscould well be fatal for many physicssignals. Signals involving leptons only,however, will likely not be washed outsince signal to background is usuallybetter.

184

)

)

oVIRTUS: A MULTI-PROCESSOR EVENT SELECTOR USING FAST BUS

J. Ellett

o University of CaliforniaLos Angeles, California 90024

o

o

o

This report describes a large, parallel array ofM68000 microprocessors, designed with a throughputcapacity of 10,000 events per second. FASTBUS seg­ments interconnect the microprocessors, the experimentdata readout system, and the PDP-ll data acquisitionsystem. A single M68000 receives the complete set ofdata for one event and processes the event indepen­dently of other events and other processors.FASTBUS provides the medium and the protocol for theefficient transfer of data-blocks to and from processorsthat become available in random order. Without modifi­cation, VIRTUS is also capable of being used off-line

for full event analysis.

crate can be constructed to test the concept and thehardware. The basic system can then be expanded tothe desired size. A system of about 250 processors ispresently envisaged, although much larger numbers arepossible.

The general architectural features of the proposedsystem are the following;

1. Programmable

2. Simple in concept

3. Applicable either on-line or off-line

4. Processor independent

INTRODUCTION5. Host CPU independent

5. Efficient use of FASTBUS

3. Small prototype system to test concept

1. Standardized FASTBUS system

2. Large number of powerful, low-cost micropro­cessors

featu res of the proposed

6. Expandable.

Some of the keyimplementation are:

4. Required modules can be designed for generalFASTBUS use

This report describes a programmable multi-proces­sor system that will make the final decision in theevent selection process of an experiment (VI RTUS =Very-Intelligent-Real-Time-Event-Selector). This sys­tem, designed with a throughput capability of 10,000events per second, consists of a large, parallel arrayof M68000 microprocessors. FASTBUS segments inter­connect the microprocessors, the experiment data read­out system, and the PDP-ll data acquisition system.Events selected by the trigger logic of the experimentare distributed via FASTBUS to an available M68000.A single M68000 receives the complete set of data forone event and processes the event independently ofother events and other processors. If an event is re-

jected, the processor becomes available for a new one.Accepted events are transferred via FASTBUS to aPDP-ll. At any point in time a large number of inde­pendent events will reside in the memories of theM68000s and wi II be processed concu rrently.

o

,)

A fundamental design problem with this type ofsystem is having efficient transfer of data-blocks toand from processors that become available in randomorder. FASTBUS provides the medium and the proto­col for solving this problem.

The proposed processor system is designed so thata small, 16-processor version in a single FASTBUS

6. Fully buffered data transfers

7. Efficient processor allocation scheme.

The initial test application of the 16-processor pro­totype system will be in experiment R608, which isrunning at the CERN-ISR. The spectrometer data areread out with a combination of standard CAMAC hard-

) 185

ware and a 16,448 channel TDC system designed andconstructed at UCLA for R608. The processor systemwill make possible the efficient real-time selection ofrare event topologies in cases where there is no othermeans of triggering on these events.

TABLE 1

COMPONENT LIST OF BASIC SYSTEM

SYSTEM DESCRIPTION

QTY DESCRIPTION

This section describes the overall system operation insome detail. The next section describes the individualmodules of the system. A block diagram of the pro­posed multi-processor system is shown in Figure 1;each block represents a hardware module. The initialprototype version of the system will consist of theFASTBUS crate segment, shown on the left side of thefigure, along with the modules connected to it. Thecrate will include five modules that control the transferof event data from the experiment to a designated pro­cessor module, 16-processor modules, and a host-com­puter interface module. The system may be expandedby using Simple Segment Interconnect modules to con­nect the original FAST BUS segment to additionalFASTBUS crates filled with processor modules. Table1 lists the modules and quantities necessary for the16-processor basic system in the prototype applicationto experiment R608.

The basic concept of the proposed system is quitesimple. It is composed of three primary parts, theevent readout and buffering system, a large number ofprocessors connected in a parallel array, and a connec­tion to a host-processor. Events are received, stored,and distributed to the processors by the readout sys­tem. Each processor receives the entire data-set of anevent, and processes that event independently and as­ynchronously from all other processors and events.Analyzed events are either discarded, if of no inter­est, or sent to the host-computer for recordi ng ontape.

Control of the distribution of incoming events tothe processors and outgoing events to the host is man­aged with the use of two queues, one in the EventReadout Control Module and the other in the Host Pro­cessor I nterface Module. These queues a re written toby the individual processors when they have availablememory space for a new event or when a processedevent is ready for transfer to the host. The proces­sor writes a single word containing the processor ad­dress and block-mode register to use for the transfer.

FASTBUS crate with ancillary logicand power supply.

1 TDC Buffer Readout Control Module(TBRC)4 CAMAC Crate Readout Control Module(CCRC)5 Event Data FIFO module(EDF) I1 Event Readout Control Module(ERC) I

16 M68000 Processor Element Module(M68KPE) I1 Host Processor Interface Module(HPI) I1 PDP-11 Unibus Adapter Module(PDPAM) I1 FASTBUS Monitor Module(FBM) I

II

The block-mode register will be preset by the

processor to the internal memory address. In the caseof incoming events, the ERC will become bus masterand execute a primary address cycle using the topelement in its queue to make the connection to the pro­cessor. Then it will use a non-handshaked, block­mode write to send the event data at maximum speed tothe processor.

In the case of outgoing events to the host, thehost will become bus master, make the connection tothe processor using the top element in its queue, andthen execute a non-handshaked, block-mode read toreceive the processed event data. In both cases theprocessor acts as a slave during the transfer. An in­ternal processor interrupt will be generated at thecompletion of the block-transfer.

The advantages of this method of distributingevents are the following:

1. It efficiently handles the routing of data toand from processors which become available inrandom order.

2. Only a single FASTBUS write cycle is requiredto specify the transfer of each data block,thus minimizing bus traffic.

3. The method is independent of the number ofprocessors in the system and allows processorsto be added (or removed) with little, or no,software changes.

C)

)

)

)

)

)

186

o

The FI FOs in the EDF modules are large enough tobuffer several events, thereby allowing event data in­put to overlap in time with event data output, viaFASTBUS, to a processor. The data will be formattedinto 32-bit words for FASTBUS transfers. Assuming anon-handshaked, block-mode transfer rate of 200 nano­seconds per 32-bit word, the event transfer ~ime fromthe EDFs to a processor is 60 Ilsec (0.20*600/2). Ta­ble 2 gives a summary of the different types of trans­fers that occur on the FASTBUS, the maximum antici­pated rate, the length of time for each transfer, andthe total bus loading in msec/sec.

In experiment R608, the initial application of the

processor system, each event generates about 20016-bit words of wire chamber data and 400 16-bitwords of CAMAC data. With a minimum-bias triggerfollowed by the existing multiplicity logic, the maximumevent rate available to the processor system is 10 KHz.

In order to minimize the readout time, event data willbe transferred to the processor system in 5 parallel

channels (one channel for the wire chamber data wordsand one for each of four CAMAC crates). The eventreadout time is the time required to read the slowest of

the five channels. Assuming a 0.5 Ilsec/word rate forthe CAMAC channels and a 0.25 Ilsec/word rate to readfrom the TDC buffer memory, the time required toread the event into the Event Data FI FO modules is 50Ilsec (200*0.25 or 100*0.50). If 50 Ilsec digitizationtime is allowed for the CAMAC system, the total eventtime is about 100 Ilsec. Thus, entire events can betransferred into the processor system at about a 10KHz rate, matching the fastest rate in the experiment.

TABLE 2

FASTBUS TRANSACTION SUMMARY

This section gives a general description of eachmodule. These descriptions are intended to give a

general understanding of the function and ope~ation ofthe modules and are not a final specification. Thenine modules are listed below. Of these nine, eightare general purpose modules that could be used in oth­er FASTBUS applications. The only module that is notgeneral purpose is the UCLA TDC Buffer Readout Con­trol Module. The Simple Segment Interconnect moduleis not necessary for the initial 16-processor system butis necessary when the system is expanded to a largernumber of processors. The following system modules

are described in this section:

3. Event Data FIFO Module (EDF)

The memory size of the processor modules, 512KByte, should be large enough for relatively sophisti­cated compiled FORTRAN programs. Floating pointhardware is optional for the processor modules, andexpansion of the memory space is possible through a

connector to an external memory module.

MODULE DESCRIPTIONS

1. TDC Buffer Readout Control Module (TBRC)

2. CAMAC Crate Readout Control Module (CCRC)

Although the primary use for the proposed proces­sor system is as an on-line event selector, it is clearthat the system can also be used off-line to processevents that have already been recorded on tape. Theonly functional difference would be that the host be­comes the event source as well as the destination forprocessed events. The distribution of events to andfrom the processors would be handled in exactly thesame manner as before, except that both queues wouldbe maintained in the Host Processor Interface module.

As seen in Table 2, the projected maximum numberof bus transactions is on the order of 20,000 per sec­ond, and the maximum expected bus loading is 62%.

IIIIII

Load I(msec/sec)

Length(Ilsec)

Rate(Hz)

Type of Transfer

o

()

u

o

o

o

o

I,) EDF to M68KPE 10,000 60 600

Host from M68KPE 100 100 10 IM68KPE to ERC queue 10,000 1 10 IM68KPE to Host queue 100 1 0.1

ITotals 20,200 620.1 I

,) I

4. Event Readout Control Module (ERC)

5. M68000 Processor Element Module (M68KPE)

6. Host Processor Interface Module (HPJ)

7. PDP-ll Unibus Adapter Module (PDPAM)

8. FASTBUS Monitor Module (FBM)

9. Simple Segment Interconnect (SSI)

187

TDC BUFFER READOUT CONTROL MODULE

This module is installed in the Memory Crate of theexisting UCLA wire chamber TDC readout system. Itcontains the logic required to transmit the TDC data inthe Memory Crate to the Event Data FI FO module uponreceipt of the proper signal from the EDF. The datais transmitted in the proper format for the EDF mo­dule. When not transmitting data, this module pro­vides access by FAST BUS to internal registers in theMemory Crate via the EDF module.

CAMAC CRATE READOUT CONTROL MODULE

This module performs the equivalent function as theprevious module but for a CAMAC crate. It physicallyreplaces the crate controller in a CAMAC crate. Itshould have a fast read mode that only uses the S1CAMAC pulse. When not transmitting to the EDF, this

module will allow FASTBUS to execute individualCAMAC cycles via the EDF module.

EVENT DATA FIFO MODULE

This is a FASTBUS module that is installed in acrate segment. It receives data over an external busthrough a front panel connector and is controlled bythe Event Readout Control Module via the auxiliaryconnector. This module performs two functions, whichmay be active simultaneously - writing a data-block tothe internal FIFO (fi rst-in, fi rst-out) memory from anexternal device, and transmitting a data-block from theFIFO memory to a destination on the FASTBUS innon-handshake, block-mode. Both functions are cont­rolled by signals from the ERC module; the signalspass through the auxiliary connector. The FIFO me­mory is 32-bits wide and is large enough to hold sev­eral events (1 K to 2K 32-bit words). When idle, thismodule will allow FASTBUS access to an internal regis­ter' whose contents can be used to access and controlthe external device over the external bus.

. EVENT READOUT CONTROL

This is a FASTBUS module that controls the read­out of event data from the external devices into theEDF modules and the transfer of data from the EDFmodules to a destination on the FASTBUS. Both typesof data transfer can occur at the same time. The ERCuses a simple Auxiliary Bus connected to the EDFs viathe auxiliary connectors to control the EDFs. Fromthe trigger logic, the ERC module receives signals that

indicate when an event is available for transfer to theEDFs. When an event is available, a signal is sent tothe EDFs, via the auxiliary connectors, to initiate the

transfer. The completion of the transfer is indicated

by the return of signals from the EDFs. As long asspace remains in each EDF FIFO, a new event will beaccepted by the ERC.

The ERC maintains a count of the number of com­plete events present in the EDFs. If this number isone or more, then the ERC will initiate a block-modetransfer from the EDFs to a destination processor.The selection of the destination processor is made us­ing the contents of a FIFO memory on the ERC board.Each processor in the system that has available memoryspace for an event, will write its address to this FIFO.The ERC then uses the top value in the FIFO as thedestination address. To execute the transfer of theevent data, the ERC arbitrates for, and gains controlof, the FASTBUS. It then executes a primary addresscycle to set up the transfer. Via the Auxiliary Bus,the ERC then enables each of the EDF modules in turnto transmit its data in non-handshaked, block-mode.When the final EDF has completed its transmission, theERC interrupts the destination processor and dropscontrol of the bus. This process is repeated as longas at least one event is present in the EDF modules.

M68000 PROCESSOR ELEMENT MODULE

This module contains the M68000 microprocessoralong with its memory, peripherals, and an interface tothe FASTBUS. The M68000 bus is used internally as aprivate bus to interconnect all the devices on the pro­cessor module, thereby allowing the M68000 to executeinstructions independently of any transactions on theFASTBUS. This module can be either a slave or mas­ter FASTBUS device. In the slave mode, theFASTBUS interface becomes master of the internalM68000 bus, thereby allowing FASTBUS access to ad­dresses on that bus. Two sets of incrementing ad­dress registers are available for use in block-modetransfers.

In the FASTBUS master mode, two methods of ac­cess are provided. In the first method, M68000 ad­dresses within a certain range will be mapped intoFAST BUS address space and will automatically causeFASTBUS cycles. The second method gives the M68000direct control over FASTBUS cycles through a set ofspecial purpose registers. These registers can beused to generate single cycles or block-mode transfersfrom M68000 address space to FASTBUS address space.

The M68KPE will contain two types of RAM memory.

188

.)

)

)

)

)

)

o

o

()

o

()

Space for up to 512 KByte of program memory will be

available, as well as a smaller amount of faster RAM,to facilitate high-speed block transfers of data. Theinternal bus can be extended through a front panelconnector to allow additional memory or peripheral dev­ices to be added.

HOST PROCESSOR INTERFACE MODULE

This module provides the connection between thehost-processor and the FASTBUS system. It is in­tended to be a general purpose interface, which canconnect to any minicomputer through special purpose

adapter modules. It is essentially the same module asthe M68KPE without the M68000 and has the sameFASTBUS master and slave capabilities. The internalbus is connected to the host-processor through thespecial purpose adapter module for that host, therebyallowing the host to, in effect, take the place of the

M68000.

PDP-11 UNIBUS ADAPTER MODULE

This module, installed on the PDP-11 UNIBUS, al­lows the PDP-11 to access the internal bus of the HostProcessor Interface Module~ Various modes of accessto the internal bus are provided to the PDP-11. Theseinclude registers to gain master control of the internalbus and address mapping logic, which allows direct ac­cess to internal bus addresses from Unibus addresses.Interrupts generated by devices on the internal buswill cause PDP-11 interrupts.

FASTBUS MONITOR MODULE

This is a diagnostic module that monitors signals ona FAST BUS segment, thereby facilitating tests ofhardware and software. The module contains a largememory that continuously stores the states of theFASTBUS signal lines. The sampling time can be pro­vided externally or derived internally from cycles onthe FASTBUS.

SIMPLE SEGMENT INTERCONNECT MODULE

This is a FASTBUS module that implements a subsetof the standard Segment Interconnect Module. It isdesigned to connect two, and only two, FASTBUS cratesegments together using a simple cable, which is not aFASTBUS segment. In addition, it only provides theminimum logic necessary to effect segment to segmentFASTBUS cycles.

SOFTWARE

The software components required for the processorsystem fall into the catagories listed below:

1. Cross software,

2. M68000 resident operating system,

3. Processor interface driver,

4. Host utility programs,

5. On-line application programs,

6. Off-line application programs,

7. System diagnostics.

CROSS SOFTWARE

Several software packages for the host computerare required to allow it to be used for development ofthe programs that will be run by the M68000 proces­sors. These packages will allow assembly language andFORTRAN programs to be compiled and linked togetherinto programs that are executable by the M68000. Thefollowing software will be required:

1. M68000 assembly language cross assembler,

2. M68000 FORTRAN cross compiler,

3. Qbject module linker,

)

Writing to the memory can be stopped N sample 4. Downline loader.periods after a certain pattern of signals is detected,where both N and the pattern can be set. The con-tents of the memory can be read via the FASTBUS or

through an auxiliary connector. M68000 RESIDENT OPERATING SYSTEM

An operating system is required that will perma­nently reside in each M68000 module and will contain

189

those routines that are necessary for operation of theprocessor system at the most fundamental level. Thisoperating system will consist of a ROM portion and adown-line loadable RAM portion. The ROM residentpart will be a program which contains the initializationcode that is entered whenever a power-up occurs or areset command is received. This code will includediagnostic routines that test for proper operation ofthe M68KPE module and routines that allow down-lineloading of other programs from the host.

The RAM resident portion will provide routines thatcan be called from application programs to execute sys­tem level functions, such as peripheral data I/O, dataI/O to the host or another processor in the system,memory allocation, floating point functions, etc.

The operating system software will also contain theroutines which receive commands, from the host-pro­cessor, to control the loading and execution of theM68000 programs.

PROCESSOR INTERFACE DRIVER

The processor interface driver is an assembly lan­guage program that runs (on the host machine) as adriver under the host's operating system and providesthe routines that allow control of, and communicationswith, the processor system via the Host Processor In­terface module. All programs running on the host willcall this driver to do data I/O to the processor sys­tem.

HOST UTILITY PROGRAMS

The utility programs are complete programs thatrun on the host and allow the operator to perform spe­cific functions involving the processor system. Onesuch program might allow an operator to down-loadspecified diagnostics to one or more M68000 processorsand receive and display diagnostic messages retu rnedfrom these processors.

ON-LINE APPLICATION PROGRAMS

lection criteria. This program will run under theM68000 resident operating system. It will probably be

written in M68000 assembly language to optimize

execution speed.

Associated with the on-line application program arethe changes that will be required in the existing ex­

periment data acquisition software, which runs on thehost during on-line data taking. These changes,which incorporate calls to the processor interface dri­ver, allow the processor system to be used as the

event sou rce.

OFF-LINE APPLICATION PROGRAM

The off-line application program is a program thatruns on the M68000 processor and does full event mo­mentum analysis. It receives unprocessed events fromthe host computer via the Host Processor Interface andreturns analyzed events to the host computer. Thisprogram will probably be written in FORTRAN andcompiled on the host computer using the FORTRANcross compiler. It will run under the M68000 residentoperating system just as the on-line application pro­

gram does.

SYSTEM DIAGNOSTICS

To verify proper system operation and to aid in lo­cating faulty system components, a complete set of sys­tem diagnostic programs will be required. The follow­ing programs must be provided as a minimum set of

diagnostics.

1. UCLA Readout System Diagnostic - Modifica­tions to allow the existing diagnostics for theUCLA TDC Readout System to run properly

with the processor system.

2. Processor System Exerciser - An overall sys­tem exerciser, which is comprehensive enoughto exercise each system component. Success­ful completion of this program should indicatea properly functioning system. Error messag­es generated should indicate the general loca­

tion of the faulty component.

)

)

)

)

)

)

The on-line application program is the program thatruns on the M68000 and implements the event selectionalgorithm. This program will input an event data-setfrom the readout system, analyze the data using theappropriate algorithm, and transfer the event to theHost Processor Interface, if the event satisfies the se-

190

3. Module Diagnostics - Individual diagnostic pro­grams for each system component. These pro­grams would exhaustively exercise specificparts of the system.

r6

f8

f6

TOt EveNT~1 M"VlC. ",..aK

~ BUFfeR , 0",,.,,, I PE PE0 • •Rio .. FIFO ICn.:.. ..oOL I

1 1 IICMAt E"UlT I-i !;L- CIl.fl,n: ~ p",,.1\RIo FIFOCoNTAoL

JI EJ-I ....1• · I• •

~ I •• • I •"'t\AC E\itHT ~ !

~ C!tATE ~ OATAf(.jD . I'" .1 r1f'l> I

•

I •I

[\leNT I ·, RIC> H I

C "' ....ON"1lOL I

1- ___________ -_ ---I H&.81<. Ml.Vl/.PE . • · P£

POP-II HO"'T I If- UNIBuS I" Pltocess.I\OI\PT£1 I/F I

1$1 B-IMlo'llK I

Pt"J II ,.I EllPAW~IO'"

• I 5£'"t~1$• I

I

•MIoBKPE~/"

'-

UCI.:A Toet\trtOIl.~

U,..6U5

o

o

CAttAC C"",,.E

o

Ex Pf.~ttf!tJTo TAI'UR Lolil

o

u

,)

191

....-----------------------------------------

This report was done with support from theDepartment of Energy. Any conclusions or opinionsexpressed in this report represent solely those of theauthor(s) and not necessarily those of The Regents ofthe University of California, the Lawrence BerkeleyLaboratory or the Department of Energy.

Reference to a company or product name doesnot imply approval or recommendation of theproduct by the University of California or the U.S.Department of Energy to the exclusion of others thatmay be suitable.

)

TECHNICAL INF'ORMATION DEPARTMENT

LAWRENCE BERKELEY LABORATORY

UNIVERSITY OF CALIFORNIA

BERKELEY, CALIFORNIA 94720

- - •./

~ .. • •