TheMACROdetectoratGranSassokats/publications/pub32_nucl_instrum... · modules(overalldimensions6 6 2m3),made of10cmthickpolyethylenefoamradiatorsand proportionalcounters.EachTRDcountermea-sured6

Nuclear Instruments and Methods in Physics Research A 486 (2002) 663–707

The MACRO detector at Gran Sasso

M. Ambrosioa, R. Antolinib, R. Assiroc, G. Auriemmad,1, D. Bakarie,f,A. Baldinig, G.C. Barbarinoa, E. Barbaritoh, B.C. Barishi, G. Battistonij,2,Y. Becherinie, R. Bellottih, C. Bemporadg, P. Bernardinic,*, H. Bilokonj,

V. Bisik, C. Bloisej, E. Bottazzie, C. Bowerl, M. Brigidah, S. Bussinom, F. Cafagnah,M. Calicchioh, D. Campanaa, A. Candelab, M. Carbonij, S. Cecchinie,3,

F. Ceig, A. Ceresh, V. Chiarellaj, B.C. Choudharyi, S. Coutui,4, M. Cozzie, P. Cretic,G. De Cataldoh, L. Degli Espostie, H. Dekhissie,f, C. De Marzoh, I. De Mitric,

J. Derkaouie,f, M. De Vincenzim, A. Di Credicob, D. Di Ferdinandoe,R. Diotallevid, O. Erriquezh, C. Favuzzih, C. Fortij, P. Fuscoh, M. Gebhardl,G. Giacomellie, R. Giacomellie, G. Gianninig,5, N. Gigliettoh, M. Giorginie,R. Giulianig,i, M. Gorettii, M. Grassig, H. Graui, L. Grayb, A. Grillob,F. Guarinoa, C. Gustavinob, A. Habigl,6, J. Hansoni, K. Hansonp,7,

A. Hawthornel, R. Heinzl, J.T. Hongi, E. Iaroccij,8, E. Katsavounidisi,9,I. Katsavounidisi, E. Kearnsn, H. Kimi, S. Kyriazopouloui, E. Lamannad,10,

C. Laneo, A. Leonec, D.S. Levinp, P. Liparid, G. Liui, R. Liui, N.P. Longleyi,11,M.J. Longop, F. Loparcoh, F. Maaroufie,f, G. Mancarellac, G. Mandriolie,S. Manzoore,12, V. Marrellii, A. Margiottae, A. Marinij, D. Martelloc,A. Marzari-Chiesak, M.N. Mazziottah, D.G. Michaeli, S. Mikheyevb,i,13,

L. Millerl,14,{, P. Monacelliq, M. Mongellih, T. Montarulih,

*Corresponding authors. Tel.: +39-0832-320443; fax: +39-0832-325128.

E-mail addresses: [email protected] (P. Bernardini), [email protected] (C.W. Peck).1Also at Universit"a della Basilicata, 85100 Potenza, Italy.2Also at INFN Milano, 20133 Milano, Italy.3Also at Istituto TESRE/CNR, 40129 Bologna, Italy.4Currently at Department of Physics, Pennsylvania State University, University Park, PA 16802, USA.5Also at Universit"a di Trieste and INFN, 34100 Trieste, Italy.6Also at Dipartimento di Scienze e Tecnologie Avanzate, Universit"a del Piemonte Orientale, Alessandria, Italy.7Currently at Bartol Research Institute, University of Delaware, Newark, DE 19716, USA.8Also at Dipartimento di Energetica, Universit"a di Roma, 00185 Roma, Italy.9Also at Dept. of Physics, MIT, Cambridge, MA 02139, USA.10Also at Dipartimento di Fisica dell’Universit"a della Calabria, Rende (Cosenza), Italy.11Currently at Macalester College, Department of Physics and Astr., St. Paul, MN 55105, USA.12Also at RPD, PINSTECH, P.O. Nilore, Islamabad, Pakistan.13Currently at Institute for Nuclear Research, Russian Academy of Science, 117312 Moscow, Russia.14Also at Department of Physics, James Madison University, Harrisonburg, VA 22807, USA.{Deceased.

0168-9002/02/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved.

PII: S 0 1 6 8 - 9 0 0 2 ( 0 1 ) 0 2 1 6 9 - 6

M. Montenok, L. Mossbargeri, S. Mufsonl, J. Musserl, D. Nicol "og,R. Noltyi, C. Okadan, M. Orsinib,i, C. Orthn, G. Osteriaa, M. Ouchrife,f,O. Palamarab, S. Parlatii, V. Pateraj, L. Patriziie, R. Pazzig, C.W. Pecki,*,G. Pellizzonim, M. Perchiazzih, L. Perronec, J. Petrakisl, S. Petreraq,

N. Pignatanoi, C. Pintoc, P. Pistillim, V. Popae,15, A. Rain "oh, J. Reynoldsonb,F. Rongaj, A. Rrhiouae,f, A. Sacchettih, P. Saggesei, C. Satrianod,1,L. Sattaj, E. Scapparoneb, K. Scholbergi,9, A. Sciubbaj,8, P. Serrae,M. Siolie, G. Sirrie, M. Sittak,6, S. Sondergaardi, P. Spinellih,M. Spinettij, M. Spurioe, S. Staliog,i, R. Steinbergo, J.L. Stonen,

L.R. Sulakn, A. Surdoc, G. Tarl"ep, V. Togoe, M. Vakilir,16, C. Valierie,C.W. Walteri,n, R. Webbr, N. Zaccheob,i

aDipartimento di Fisica dell’Universit "a di Napoli and INFN, 80125 Napoli, ItalybLaboratori Nazionali del Gran Sasso dell’INFN, 67010 Assergi (L’Aquila), Italy

cDipartimento di Fisica dell’Universit "a di Lecce and INFN, via per Arnesano, 73100 Lecce, ItalydDipartimento di Fisica dell’Universit "a di Roma ‘‘La Sapienza’’ and INFN, 00185 Roma, Italy

eDipartimento di Fisica dell’Universit "a di Bologna and INFN, 40126 Bologna, ItalyfL.P.T.P., Faculty of Sciences, University Mohamed I, B.P. 524 Oujda, Morocco

gDipartimento di Fisica dell’Universit "a di Pisa and INFN, 56010 Pisa,ItalyhDipartimento di Fisica dell’Universit "a di Bari and INFN, 70126 Bari, Italy

iCalifornia Institute of Technology, Pasadena, CA 91125, USAjLaboratori Nazionali di Frascati dell’INFN, 00044 Frascati (Roma), Italy

kDipartimento di Fisica Sperimentale dell’Universit "a di Torino and INFN, 10125 Torino, ItalylDepts. of Physics and of Astronomy, Indiana University, Bloomington, IN 47405, USA

mDipartimento di Fisica dell’Universit "a di Roma Tre and INFN Sezione Roma Tre, 00146 Roma, ItalynPhysics Department, Boston University, Boston, MA 02215, USA

oDepartment of Physics, Drexel University, Philadelphia, PA 19104, USApDepartment of Physics, University of Michigan, Ann Arbor, MI 48109, USA

qDipartimento di Fisica dell’Universit "a dell’Aquila and INFN, 67100 L’Aquila, ItalyrPhysics Department, Texas A&M University, College Station, TX 77843, USA

The MACRO collaboration

Received 6 August 2001; received in revised form 26 October 2001; accepted 30 October 2001

Abstract

MACRO was an experiment that ran in the Laboratori Nazionali del Gran Sasso from 1988 to 2000. Its principal

goal was to observe magnetic monopoles or set significantly lower experimental flux limits than had been previously

available in the velocity range from about b ¼ 10�4 to unity. In addition it made a variety of other observations.

Examples are: setting flux limits on other so far unobserved particles such as nuclearites and lightly ionizing particles,

searching for WIMP annihilations in the Earth and the Sun and for neutrino bursts from stellar collapses in or near our

Galaxy, and making measurements relevant to high energy muon and neutrino astronomy and of the flux of up-going

muons as a function of nadir angle showing evidence for neutrino oscillations.

The apparatus consisted of three principal types of detectors: liquid scintillator counters, limited streamer tubes, and

nuclear track etch detectors. In addition, over part of its area it contained a transition radiation detector. The general

design philosophy emphasized redundancy and complementarity. This paper describes the technical aspects of the

15Also at Institute for Space Sciences, 76900 Bucharest, Romania.16Also at Resonance Photonics, Markham, Ont., Canada.

M. Ambrosio et al. / Nuclear Instruments and Methods in Physics Research A 486 (2002) 663–707664

complete MACRO detector, its operational performance, and the techniques used to calibrate it and verify its proper

operation. It supplements a previously published paper which described the first portion of the detector that was built

and operated. r 2002 Elsevier Science B.V. All rights reserved.

PACS: 29; 29.40.Gx; 29.40.Mc; 14.80.Hv

Keywords: Astroparticle physics; Underground detectors

1. Introduction

MACRO was a large area multipurpose under-ground detector located at the Laboratori Nazio-nali del Gran Sasso (LNGS) in Italy. It wasdesigned to search for rare events in the cosmicradiation and it was particularly optimized forsensitivity to the supermassive magnetic mono-poles predicted by Grand Unified Theories of theelectroweak and strong interactions. Over theyears, the collaboration published the results ofseveral monopole searches [1–5]. The experimentcould also perform many other interesting searchesand precision measurements in various other areasof particle physics, cosmic ray physics, andastrophysics. These included: study [6,7] of bothhigh energy ð %EnB100 GeVÞ and lower energyð %EnB4 GeVÞ atmospheric neutrinos showing evi-dence of flavor oscillations [8]; high energyneutrino astronomy [9] and muon astronomy[10,11]; searches for WIMP annihilations in theEarth and Sun [12]; search for low energyðEn\7 MeVÞ stellar collapse neutrinos and anti-neutrinos [13,14]; studies of various aspects of thehigh energy underground muons that provide anindirect tool to study the primary cosmic raycomposition, origin, and interactions [15,16]; andsearches for various conjectured types of matter, asyet unseen, that may exist in the cosmic radiationand contribute to cold dark matter such asfractionally charged particles [17,18] and nuclearites[19]. Over the years the experiment publishedresults on these topics, helping to fulfill the promiseinherent in the development of the unique, largeexperimental halls deep underground at the GranSasso Laboratory. Early descriptions of the experi-ment and its apparatus are given in Refs. [20–22].At this laboratory, the minimum rock over-

burden is 3150 hg cm�2: A muon needs a mini-

mum energy of C1:3 TeV at the surface topenetrate this heavy shield and reach the under-ground laboratory. The cosmic radiation results ina muon flux ofB1 m�2 h�1 in the laboratory withan average muon residual energy of B310 GeV[16].The apparatus was built and equipped with

electronics starting in 1988 and was completed inNovember 1995. Some data were taken with thepartially completed detector between 1989 and1995, but from November 1995 until May 2000 itran in its completed, full configuration; from May2000 to December 2000, a subset of the detector’selectronics was used primarily to measure the fluxof upgoing muons but the search for slowmonopoles was turned off. A previous paper [20],referred to in the following as ‘‘I’’, described thefirst 1/12 of MACRO and plans for the fullapparatus. This is a supplement to that paper, andso reference to it will be made for unchangeddetails of the apparatus; the emphasis in this paperis on differences between the 1992 description andthe actual final detector as well as issues relating tocalibration and performance verification.A glossary of the many acronyms used in this

paper is given at its end.

2. General configuration of the detector

The detector had a modular structure: it wasdivided into six 12:6� 12� 9:3 m3 sections re-ferred to as supermodules (SM), each with separatestructure and electronics readout. The detector’sglobal dimensions were 76:5� 12� 9:3 m3; itstotal mass was B5300 t; and its total acceptanceto an isotropic flux of particles wasB10; 000 m2 sr:

M. Ambrosio et al. / Nuclear Instruments and Methods in Physics Research A 486 (2002) 663–707 665

Redundancy and complementarity were theprimary goals in designing the experiment. Sincewe could not reasonably expect more than a fewmonopoles during its lifetime, we deemed multiplesignatures and the ability to perform cross checksamong various parts of the apparatus as crucial.To accomplish this, the principal detector con-sisted of three independent sub-detectors: liquidscintillation counters, limited streamer tubes, andnuclear track (‘‘track-etch’’) detectors.Fig. 1 illustrates the general layout of the

experiment. It shows the six SMs as well as theseparation of the detector into a lower half (thelower detector) and an upper half (the ‘‘attico’’).Fig. 2 gives a cross-sectional view of the detector.Each of the six SMs contained a total of 77scintillation counters divided into three horizontalplanes (top, center, and bottom) and two verticalplanes (east and west). In addition, the two endSMs contained vertical planes covering the lowerpart of their north and south faces. The lowerdetector had 10 horizontal planes of streamertubes, separated by either crushed rock absorbersor (at its top and bottom) a plane of scintillatorcounters. The total thickness of the crushed rockwas C420 g cm�2 setting an energy threshold of

B1 GeV for a typical muon to pass through thelower detector. Each of its vertical sides had sixplanes of streamer tubes, separated into groups ofthree, with a plane of scintillator between them.The attico had two side walls of streamer tubesand scintillator planes, similar to that of the lowerdetector. Its top consisted of four horizontalplanes of streamer tubes separated into two groupswith a plane of scintillation counters betweenthem. The attico’s central region was mostly openspace, housing the readout electronics, sometransition radiation detectors and an assortmentof work benches and storage cabinets. The detailsof the attico are described below.Because of the crushed rock absorbers, the

lower detector contained most of the total detectormass (B4700 t). Since there was relatively littlematerial in the attico, it was a good approximationto treat it as empty space when considering thestopping power of ionizing particles. Within asupermodule, the horizontal planes of scintillatorsand streamer tubes were approximately hermetic,with neighboring counters in each SM planepushed together, but, between the SMs, a verticalgap was necessary for structural support providedmostly by steel I-beams. On the roof of the attico,

Fig. 1. General layout of the detector installed in Hall B of the LNGS. Overall dimensions of the active part were 76:5 m long, 12 m

wide, and 9:3 m high.


however, the horizontal layers of streamer tubesand scintillation counters were contiguous with nogaps between SMs. Also, the vertical planes ofstreamer tubes and scintillators had gaps (about15% of the total area) for structural support. Allcomponents of the detector, including rock absor-ber, liquid scintillator, plastic in scintillator tanksand streamer tube walls and structural steel, havebeen included in the calculation of the totaldetector mass and have been reproduced in detailin simulations of the detector’s response toionizing particles.The limited Streamer Tube (ST) system con-

sisted of B50; 000 tubes, each having a 3� 3 cm2

cross-section and a length of 12 m: The tubes werefilled with a gas mixture of 73% He and 27% n-pentane. Signals were detected on anode wires onall planes. On horizontal planes a second coordi-nate readout was provided by pick-up strips belowthe chambers at an angle of 26:51: The intrinsicposition and tracking resolutions of the streamertube system were B1 cm on the wire view and

B0:21 on the strip view, but the overall angularresolution for cosmic ray muons was limited toB11 by the multiple Coulomb scattering in therock around the detector.The nuclear track-etch detector was deployed in

three planes, horizontally in the center of the lowerdetector and vertically on the east and north faces.The detector was divided into 18,126 individualstacks which could be individually extracted,etched, and replaced as needed. Each stack haddimensions 24:5� 24:5� 0:65 cm3 and was com-posed of three layers of CR39, three layers ofLexan and 1 mm of aluminum absorber to stopnuclear fragments.Details of the scintillator system, streamer tube

system, and track-etch are given in Sections 3, 4and 5 respectively.In addition to the three detection elements

already described, a Transition Radiation Detec-tor (TRD) was installed in part of the attico,directly above the central horizontal plane of thedetector. It was composed of three individual

Fig. 2. A cross-sectional view of the detector showing the layout of streamer tubes, scintillation counters, track-etch detectors, and

rock absorbers. In addition (not shown), there were vertical planes of track-etch detector just outside the main apparatus on one side

and one end, and transition radiation detectors in part of the attico.


modules (overall dimensions 6� 6� 2 m3), madeof 10 cm thick polyethylene foam radiators andproportional counters. Each TRD counter mea-sured 6� 6� 600 cm3 and was filled with 90% Arand 10% CO2: The TRD provided a measurementof the muon energy in the range of100 GeVoEo930 GeV: Its energy measurementsaturated at B930 GeV but muons of higherenergies were still detected and identified. Detailsof this system are given in Section 6.Since the attico’s construction was somewhat

different from that of the lower detector and it wasnot described in I, its details are given here. Itconsisted of two vertical walls and a top coverimmediately above the lower detector; its dimen-sions were 76:5� 12� 3:5 m3: As noted above, theroof of the attico, being contiguous, was morehermetic than the lower detector at the cost of lessmodularity for the detector. For data acquisitionand triggering the modularity was enlarged to aDouble Supermodule (DSM) that corresponded toone mVax computer in the data acquisition system.Nonetheless, some vestiges remained of the lowerdetector’s segmentation into supermodules. As inthe ST planes of the lower detector, the atticodetectors were divided into vertical and horizontalplanes. The vertical faces were equipped with sixST planes and one plane of scintillator counters, asin the lower detector. The attico roof wascomposed of four ST planes and one scintillation

counter plane as shown in Fig. 3. A 5 cmpolystyrene foam spacer was inserted betweeneach pair of horizontal ST planes to improve thetracking capability of the attico roof. As in thelower detector, all attico ST horizontal planes wereequipped with pick-up strips allowing the deter-mination of the second coordinate in each plane.Unlike the lower detector, the attico vertical planeswere equipped with 5:5 cm wide strips (pitch6:0 cm) at right angles to the ST wires, while thestrips on the roof horizontal planes were 3:1 cmwide (pitch 3:25 cm) at a 26:51 stereo angle.Although the strip width in the vertical planeswas about twice that used in the horizontal ones,this choice did not degrade the angular resolutionof the tracking system because of its 903 stereoangle.

3. The scintillator system

The scintillation counters and their performanceare fully described in I and so only a shortdescription, for completeness, will be given here.The discussion will describe the final waveformdigitizing system (WFD) used in the experiment,the various triggers produced by the scintillatorsystem, and the methods used to verify the properoperation of the system and its calibration.

Fig. 3. End view of one scintillator counter in the horizontal layer of the attico roof showing relative locations of the scintillator, the

streamer tubes, and the strips.


3.1. The scintillation counters

The counters in the horizontal planes had anactive volume of 11:2� 0:73� 0:19 m3; and thosein the vertical planes, 11:1� 0:22� 0:46 m3: Allwere filled with a mixture of 96.4% mineral oiland 3.6% pseudocumene, with an additional1:44 g l�1 of PPO and 1:44 mg l�1 of bis-MSB aswavelength shifters. The horizontal counters wereviewed by two EMI 800 photomultiplier tubes(PMTs) at each end and the vertical countersby one 800 PMT at each end; in the attico,Hamamatsu PMTs were used. Each PMT housingwas equipped with a light collecting mirror. Thetotal number of scintillators was 476 (294horizontal and 182 vertical) with a total activemass of 560 t: The scintillator counter construc-tion for the whole of MACRO was the same as inthe first SM.The signals from the PMTs were reproduced in

fast fan-out circuits and distributed to a collectionof triggering and measuring electronics describedbelow. Some triggers were formed from singlecounter signals and others were developed fromsignals summed over the PMT outputs of collec-tions of geometrically related counters. Thisprovided redundancy in the triggering system,and also allowed specialized triggers to selectevents firing multiple counters. Measurementswere made of the total charge and relative timeof occurrence of the signals from each end of eachcounter in two independent systems, the EnergyResponse Processor (ERP) and the Pulse HeightRecorder and Synchronous Encoder (PHRASE).Both of these systems are fully described in I and,except for the addition of a subsystem to measurethe time between the TDC stop signals for theERPs in different SMs, the systems finally usedwere essentially unchanged from those descrip-tions; they will not be further detailed here. Inaddition, however, for almost all events, the fullwaveforms of the PMT signals were measured with5 ns sampling for amplitudes below a saturationlimit set by either the PMT or the fan-out circuit.This measuring system is described below inSection 3.2. In addition the portion of the PMTwaveform below 255 mV was also recorded in thePHRASE electronics.

A minimum ionizing muon crossing a horizontalcounter vertically had a most probable energy lossof B34 MeV in 19 cm of scintillator; if it passedequidistant from the ends, it produced about 600photoelectrons in each. No deterioration in thescintillator performance was observed in about 10years of operation. Overall, the scintillator sys-tem’s timing and longitudinal position measure-ments on single particles in a counter hadresolutions of B0:5 ns and B10 cm respectively.

3.2. Waveform digitizer

One of the most powerful signatures of a rareevent in a scintillator counter is the waveform ofthe PMT signals it produces. Thus, in a search forrare events in a large scintillator system, digitiza-tion of the PMT waveforms with high resolutionand good fidelity is essential. It allows backgroundrejection in searches for monopoles, nuclearites,and lightly ionizing particles, and it also provides apowerful alternative to ADCs and TDCs forcharge and time measurement.The experiment utilized a custom built, com-

mon-stop waveform digitizer (WFD) to recordthese signals. Amplitude digitizations were madeand stored continuously until a stop signal,suitably delayed after an event, interrupted theprocess, freezing the memory. Its most importantintended use was to observe the transit of a slowmonopole crossing a scintillator. For example, amonopole moving vertically with velocity b ¼v=c ¼ 0:7� 10�3 yields a square pulse 1 ms longwith photoelectron statistical fluctuations super-imposed. An LED simulation of such a signalcontaining the expected number of photoelectronsfor a magnetic monopole (about 5000) is shown inFig. 4(b). This WFD replaced a small number ofLeCroy 2261 CCD-based digitizers used duringthe experiment’s early stages.The WFD had a flow-through design. An input

analog PMT signal fed both a voltage discrimi-nator and a non-linear amplifier. The discrimina-tor triggered when the input was outside the range72:5 mV and the amplifier output was digitizedby a flash ADC (FADC) at a 200 MHz rate. Thebipolar sensitivity of the trigger was very impor-tant to avoid misinterpreting small, bipolar,


oscillatory noise bursts as trains of negative pulses,mimicking sequences of single photoelectronpulses. Examples of both an interesting signaland bipolar noise are shown in Fig. 5(c) and (d)respectively. Further, voltage measurements weremade within a 20 ns range of a trigger so parts of apulse not satisfying the trigger condition were alsocaptured; this is shown by the dotted parts ofFig. 5(c). When the recording process was stoppedby external signals, the RAM was readout throughthe VME backplane by the data acquisition system.Fig. 6 shows an overview of one WFD board.The important features of the design and their

motivating requirements are as follows:200 MHz clock speed. The narrow pulse width

(15 to 20 ns) of the scintillation counters required

a system that had a sampling rate of about 5 ns:Very slow monopoles are expected to registerdistinctive trains of single photoelectron pulses,typically 12 ns FWHM. Fig. 4(c) shows a sequenceof single photoelectron signals.

64 kbyte/channel memory. Because monopolesof interest could take up to 1 ms to traverse thedetector, a deep memory buffer with a maximumsize of 64 kbyte was used to store the waveformdata. This allowed full reconstruction of events inwhich a GUT monopole catalyzed nucleon decayas it approached the detector; with a shorterrecording time, relativistic nucleon decay productscould pretrigger the detector and cut off theinformation about the passage of the monopole.Because of software limitations, we actually used

1 2 3 4

1 2 3 4

(c)

µsec

12

8

4

0

-4

-8

-12

-16

6

4

2

0

-2

-4

-6

-80 0.5 1.00 0.5 1.0

PM

T S

igna

l (m

V)

(a)

(d)

(b)

Fig. 5. This figure shows two examples of waveform outputs for low level signals, one an interesting signal (c) produced by an LED

fired just at the SMT threshold, and the other, (d), electrical noise.

0 1 2 3 4 5 6 7 80 1 20 1 2 3

0.0

-1.0

-2.0

-3.0

-4.0

PM

T S

igna

l (V

)

0.0

-0.1

-0.2

-0.3

-0.4

0.0

-0.01

-0.02

-0.03

-0.04

(a) (b) (c)

µsec

Fig. 4. LED simulations of monopole signals for velocities of b ¼ 2� 10�3(a), 7� 10�4(b), and 1� 10�4(c). The systematic change in

the photoelectron rate is an LED artifact and would not occur for a monopole.


only 50 kbyte of the available buffer, and someinitial running was with even smaller buffer sizes.In monopole searches, these limitations were takeninto account. The memory used allowed up to128 ms of PMT signal history to be recorded;typically only B15% of this was used.

Hardware zero suppression. Hardware zerosuppression was implemented so that only thewaveform data for which a discriminator had beentriggered within 20 ns were recorded. This limitedthe amount of data to a manageable size, and alsoallowed the measurement of waveform details forshort times both before and after a discriminatorhad fired, as shown in Fig. 5(c).

Customized multiplexing. We implemented mul-tiplexing of four PMT signals into one digitizerchannel. However, each PMT signal was indivi-dually discriminated and 4-bits were associatedwith each sample of the recorded waveform toidentify the channel(s) contributing to it, as shownin Fig. 5(a) and (b).

Non-linear front-end amplifier. Since singlephotoelectron pulses had an average amplitudeof �4 mV; a high-gain input amplifier wasrequired. On the other hand, the PMTs couldproduce signals as large as �10 V; requiringattenuation. This wide dynamic range of about4000 was handled by a non-linear front-endamplifier, with a small-signal differential gain of10 and a large-signal differential attenuation of 10,

smoothly mapping the input range into the 8-bitlinear range of the FADC.

VME packaging. The backplane interface wasdesigned to the IEEE 1014-1987 version of theVME bus standard to allow for rapid transfer ofthe memory buffer to the data acquisition system.We used the 9U package to provide sufficientboard space for the many components.There were three external control signals:

CLOCK, STOP, and START. A 200 MHzCLOCK was produced in a central location andfanned out to each WFD card. The WFD StopMaster (WSM), which generated the STOP signal,received inputs from almost all the triggers withina supermodule as well as the ORs of some triggersfrom adjacent SMs as described below. To allowmeasurement of both the entry and exit pulses of aslow monopole, the WSM normally produced theSTOP signal after a delay of 1 ms from the earliestoccurring trigger. In addition, when more than onesupermodule produced its own WFD-stoppingtrigger, the WSM also measured the relative timeof all of them with a 10 MHz clock. VME readouttook place after the WFD was stopped. Finally,after readout, the data acquisition system pro-duced a START signal to put the system back intorecording operation.For each digitization channel, four PMT signal

inputs were processed by one front-end amplifierand summed into a single analog output signal.Each of the four input signals was separatelydiscriminated. These logic signals identified whichof the inputs contributed to the summed analogsignal and they also controlled zero suppression.The summed signal was digitized by an AD9028 8-bit FADC ð8 mV bit�1Þ clocked at 200 MHz:Since, as noted above, the non-linear input circuitamplified small signals by 10 times and attenuatedlarge signals by the same factor, the resolution was0.8 and 80 mV bit�1 for small and large signalsrespectively. The response of the circuit is illu-strated in Fig. 7. Tests showed that this responsewas stable to about 2% at an input of �2 V:Each channel was equipped with a custom

Vitesse Application Specific Integrated Circuit(ASIC) which assembled the digital data, appear-ing once every 5 ns from the FADC and thediscriminators, into parallel format and stored

Disc

Disc

Disc

VME

INTERFACE

VME

DATA

POWER

ASIC

ASIC

ASIC Memory64 K

ASIC

4inputs per

channel

Memory64 K

Memory64 K

Memory64 K

START

STOP

CLOCK

Disc

Sum

Amp

Sum

Amp

Sum

Amp

Sum

AmpFADC

FADC

FADC

FADC

Fig. 6. A diagrammatic side view of one WFD board.


four such pairs in registers. If any of thediscriminator bits in these four samples had beenset, the data in the registers together with a 16 bittime word were transferred into the RAM memoryat an address generated by the ASIC. Thus somerecorded voltage measurements were within the72:5 mV range of the discriminators. This processcontinued until receipt of a STOP from the WSMwhich caused all of the waveform channels in asupermodule to be simultaneously stopped. Sub-sequently, selective readout of WFD channels wascontrolled by software. Based upon the contents oftrigger registers which were readout by the dataacquisition system, a complete list of hit scintilla-tors was generated to control the WFD readoutroutines.The support hardware consisted of six 9U VME

crates each holding 10 or 11 WFD boards andserving one SM. The configuration of the boardwas software controlled. It was possible to enableor disable zero suppression and to set thediscriminator thresholds to values different fromthe normal 72:5 mV: In addition, software couldwrite test patterns to memory to verify properoperation of the read-outs.

3.3. Scintillator triggers

The electronics for the scintillation countersprovided specific triggers for various rare particles,

minimum ionizing particles, and gravitationalcollapse neutrinos. Each SM had its own comple-ment of triggers as described below and auxiliarycircuits were provided to connect adjacent SMswhen necessary.MACRO’s general purpose trigger circuit was

the ERP which used TDCs and ADCs to providetriggering, timing, and charge measurement forPMT pulses corresponding to energy losses inthe scintillator as low as 7 MeV: Furthermore itcould integrate the PMT pulse from a monopoleas slow as 10�2c: Finally, it had provisionsfor detecting a burst of gravitational collapseneutrinos. This circuit was fully described in I andso its design and implementation are not discussedhere.An independent system, the Pulse Height

Recorder and Synchronous Encoder (PHRASE),was particularly adapted for the detection ofgravitational collapse neutrinos but it also hadfunctional capabilities similar to the ERP. Again,this circuit and its operation were fully describedin I and are not repeated here.With its special emphasis on the search for

slowly moving magnetic monopoles, the experi-ment employed a specially designed Slow Mono-pole Trigger (SMT), sensitive to monopoles withvelocities from about 10�4 to 10�2c: As with othertriggering circuits, it was described in I but,because implementation changes were made forthe attico, it is discussed here in Section 3.3.1.For monopoles in the velocity range from about

5� 10�3 to 5� 10�2c; the PMT signals were easilydetected by conventional voltage discriminatorsand these were used to produce a Fast MonopoleTrigger (FMT). Fig. 4(a) shows an example of anLED simulated signal of a monopole with velocityb ¼ 2� 10�3: It contains about B17; 000 photo-electrons, and also shows an afterpulse about 1 msafter the primary signal. The FMT and otherspecialized triggers are discussed in Sections 3.3.2–3.3.5.Finally, after the first supermodule was built,

MACRO was equipped with a special triggerfor Lightly Ionizing Particles (LIP). It wassensitive to fractionally charged particles downto about e=5 and is described in detail in Section3.3.6 below.

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

0 1 2 3 4 5 6 7 8 9 10

Out

put V

olta

ge (

to F

AD

C)

- Input Voltage (from PMT)

calculated modelmeasured response

Fig. 7. The measured response of the WFD non-linear input

amplifier compared with a mathematical model for the circuit.


3.3.1. The Slow Monopole Trigger (SMT)

Magnetic monopoles can exist in nature over awide spectrum of velocities with respect to Earthand, correspondingly, they can have a wide rangeof transit times and light yields in scintillators. Formonopoles traversing the detector with typicalgalactic (10�3c) or solar system (10�4c) velocities,scintillator signals will be a wide pulse of smallamplitude or a train of single photoelectron pulseslasting several microseconds. An LED simulationof the expected signal from a monopole with b ¼1� 10�4 is shown in Fig. 4(c); the single photo-electron train contains only B70 photoelectrons.In order to trigger on such a signature, the SMTwas developed for MACRO. Its philosophy andimplementation for the first supermodule arediscussed in detail in I. Briefly, it consisted oftwo parts:

* an analog circuit which translates an inputPMT pulse into a TTL pulse whose durationwas nominally the time that the input pulseamplitude exceeded half its maximum. It wascalled the Time-Over-Half-Maximum (TOHM)circuit.

* a digital circuit consisting of a scaler whichcounted up at a fixed frequency (66 MHz) whenthe TOHM output was high and at a lowerfrequency, software settable, when the TOHMoutput was low. It was called the LeakyIntegrator (LI) circuit.

An SMT pretrigger occurred when the LI scalerfor the signal coming from one end of a scintillatorcounter reached a predetermined count. Thistypically was 11, corresponding to a signal withwidth X180 ns: A coincidence within 20 ms be-tween pretriggers from the two ends of onescintillator produced an SMT trigger.Each PMT signal in the detector had a dedicated

TOHM and LI channel, but there were twoversions of hardware. The older version was builtwith DIP silicon components and was used in allof the lower detector; there were 72 TOHM NIMmodules and 276 LI CAMAC modules accommo-dating the 552 PMT signals. Its installation wascompleted in December 1992. The 400 scintillatorsignals from the attico and the north and southfaces used a version of the SMT hardware

constructed from surface mount silicon compo-nents and field programmable gate arrays. Packingdensity was increased by more than a factor of fiveso that only 14 CAMAC modules accommodatedthe 400 signals. The installation of this version ofthe SMT hardware was finished in November1995. Both versions of the SMT hardware featuredselectable PMT threshold (settable by CAMACcommand in the attico circuits and manually in theothers) and CAMAC adjustable triggering criteriafor the LI. The threshold was normally set to�2 mV:The SMT circuit for a SM triggered its WSM to

stop its WFD system and also identified thetriggering scintillators to direct the data acquisi-tion system’s readout circuits. Further, SMTsignals were supplied to adjacent SMs so thattheir WSMs could provide auxiliary timing in-formation connecting monopole-like signals indifferent SMs, if there were any.

3.3.2. The Fast Particle Trigger (FPT)

The FPT provided a trigger for particles withb > 3� 10�2 and its principal function was toproduce a veto for the Fast Monopole Trigger (seebelow). For this trigger, eight horizontal or sevenvertical scintillators were treated as a singlesupercounter. The several PMT signals from eachend of a supercounter were analog summed andamplitude discriminated at a level correspondingto about 1/6 of a typical muon signal. Thediscriminator outputs from each end of a super-counter were then put into 100 ns coincidence. Thedigital outputs from all the end-to-end coinci-dences of supercounters belonging to one face of asupermodule were ORed together and used toproduce a 1 ms face pulse. Finally the overlap oftwo or more such face pulses from a single SM or apair of adjacent ones produced the FPT. Inaddition, this system provided the data acquisitionsystem with a register which recorded the super-counters producing the coincidence. It did not

directly trigger the WSM and so it did not forcethe production of waveform information. In fact,however, the vast majority of events selected bythe FPT also triggered the ERP system which bothtriggered the WSM and provided ADC/TDCinformation.


3.3.3. The Fast Monopole Trigger (FMT)

The FMT was a conventional Time Of Flight(TOF) trigger for the detection of magneticmonopoles of intermediate velocities. Except forits output pulses, it used the same hardware as theFPT muon trigger just described. The FMTcircuitry produced B10 ms output pulses ratherthan 1 ms ones for producing two-face coinci-dences. When an FPT trigger occurred, thesesignals were vetoed. Thus the FMT selectedparticles with TOF within the detector between 1and 10 ms; for a nominal pathlength of 9 m; thiscorresponds to a b range of 3� 10�2 to 3� 10�3:The FMT was input to the WSM and so it causedwaveforms to be recorded.

3.3.4. The Highly Ionizing Particle Trigger

(HIPT)

The most energetic events triggering either theFPT or the FMT were automatically flagged forthe data acquisition system by the Highly IonizingParticle Trigger (HIPT). This circuit attenuatedthe FPT supercounter signals by a factor of 10.They were then discriminated at levels correspond-ing B1:7 times the typical minimum ionizingsignal (0:2 V for the horizontals and 0:14 V forthe verticals). Except that its output pulse was10 ms long, the subsequent logic was then the sameas that for the FPT. Like the FMT, the HIPTtriggered the WSM and so waveform informationwas collected for the supercounters involved.

3.3.5. The High Charge Trigger (HCT)

Because of residual gas in the PMTs, theirnormal signals were followed after about 1 ms by arelatively large afterpulse, as can be seen inFig. 4(a), followed by a train of single photoelec-tron afterpulses. The initial rate of these singlephotoelectrons was roughly proportional to thecharge in the initiating PMT signal and it decayedto a quiescent value over times of hundreds ofmicroseconds. This phenomenon occurs even withthe intrinsic residual gas in a new PMT, but isexacerbated by helium buildup as a tube ages. Thehelium used by the ST system diffused through thestreamer tubes’ thin plastic walls, and so theatmosphere in the MACRO experimental hall hada higher helium concentration than the normal

atmosphere. This gas subsequently diffusedthrough the PMT glass envelopes, increasing thePMT residual pressure and the size and durationof the afterpulsing. The afterpulse trains followingvery high-charge PMT signals, such as thoseexpected from nuclearites or fast monopoles, couldsaturate the finite-depth WFD memory and thiswould result in the loss of the originating pulse.Furthermore, abnormally large inputs producedcapacitive coupling overshoots exceeding theWFD þ2:5 mV threshold and these also couldcause memory saturation. As noted above, theWFD stop was normally produced about 1 msfollowing the occurrence of a trigger but for largesignals this time had to be reduced to avoidcompletely filling the memory.The HCT was produced by integrating the FPT

supercounter signals and triggering at a 10 V msthreshold; this corresponded to a charge of B100times that of a muon. To prevent loss of the WFDdata from the originating pulse, the HCT pro-duced a WFD STOP signal after a delay of just120 ms: This time window was just below themaximum that completely filled the 50 kbyte ofwaveform memory when a WFD digitized con-tinuously. Since all WFD channels within a singlesupermodule were stopped simultaneously, aparticle with bo10�4; satisfying this trigger, didnot have the waveform in the scintillator fromwhich it exited the detector recorded by the WFD.This was a necessary compromise; it was essentialthat the WFDs record single photoelectron pulsesand the PMTs themselves suffered from thisintrinsic afterpulsing property. However, boththe waveform from the entering face and theTOF information from the scintillator system wererecorded.

3.3.6. The Low Ionizing Particle (LIP) trigger

The LIP trigger was designed to be sensitive topenetrating particles whose energy loss in thescintillators was as small as 1/25 that of aminimum ionizing muon. Its general philosophywas to be sensitive to very small signals from thescintillators and to maintain the overall triggerrate at an acceptable level by requiring a multiplecoincidence including streamer tubes.


For scintillator inputs, this trigger used signalsfrom the lowest energy-based discriminators avail-able in the experiment. These were in thePHRASE system which suitably combined signalsfrom both ends of a scintillator to correct for lightattenuation in the scintillator; they triggered on anenergy loss greater than B1:2 MeV anywhere in ascintillator. To make these available to the LIPcircuits, the PHRASE units were modified relativeto the configuration originally used in the first SM.The LIP circuit itself was a digital, clocked,

synchronous system. It operated at 5 MHz and wasbased upon two XILINX field-programmable gatearray chips, an XC3090 and a smaller XC3020.These were controlled by an inexpensive 16KPROM, the XILINX XC1765. Its general logicconfiguration is shown schematically in Fig. 8.After the inputs were synchronized to the LIP

clock, they were ORed into three logical faces foreach SM: top (including all of the attico scintilla-tors), center (the horizontal layer between theattico and the lower detector), and bottom (all theother scintillators in the lower detector). Inaddition, to allow the LIP to trigger on particlespassing between SMs, the OR of half of each layerwas passed to the LIP circuits in the adjacent SMs.

A scintillator pretrigger for LIP was a coin-cidence with resolution of B400 ns either betweenall three logical layers in a SM or between twolayers in one SM and one in another. In this lattercase, a pretrigger was formed in both SMsindependently. Pretriggers latched scintillator fir-ing information into a register for subsequentreadout if all the trigger conditions were satisfied.At the low discrimination level used, the triggeringrate from a single scintillator, due primarily toradioactivity, wasB2–3 kHz; yielding a pretriggerrate of several per second per supermodule. If thishad been recorded, it would have saturated boththe data acquisition and the storage systems. Thus,to reduce the rate to an acceptable level, acoincidence with a signal from the ST system wasrequired. To allow time for ion propagation in thestreamer tubes, the LIP logic waited 6:4 ms for anST trigger. If one occurred, the LIP trigger wasproduced and fed to the supermodule’s WSM;otherwise, the latches were cleared and the circuitonce again became sensitive to scintillator coin-cidences.In data analysis, the charge and relative time of

the scintillator signals that produced the triggerwere obtained from the recorded waveforms rather

Streamer Tube Inputs

Face Scintillator

Coincidence

Scintillator

CoincidenceStreamer Tube

If STCoin

Control/Bus Lines

LIP

Trigger

Latch

Scint

Clear

Latch

CoincidenceIf No ST

TT

L C

onverters for Scintillator Discrim

inator Signals

Latches &

Digital P

ulse Generators

CAMAC Interface & Latch Readout

Fig. 8. A schematic of the LIP trigger circuit logic.


than from a traditional ADC/TDC system. Thistechnique strongly suppresses an especiallydangerous source of accidental background in aLIP search to which some ADC/TDC systemsare susceptible: a small radioactivity pulse canstart an ADC/TDC gate just prior to the arrivalof a normal minimum ionizing signal. If theintegration gate closes during the normal signal,this can occasionally result in the reconstructionof an event with a much smaller energy thanactually deposited in the scintillator. Use of thedetailed waveforms in the analysis avoided thisproblem.

3.4. Performance verification and calibrations

With its large complement of triggers andmeasuring electronics, the scintillator systemneeded convenient equipment and procedures todetect malfunctions, allowing timely repairs, andto regularly provide parameters for convertingmeasured quantities (e.g., ADC or TDC counts) tophysical quantities (e.g., MeV of energy loss ornanoseconds relative delay). Three differentsources were used to provide this information:nitrogen lasers which simultaneously pulsed lightinto the longitudinal center of each scintillator in agroup via optical fibers, LEDs mounted near eachPMT which could be excited individually, and thesignals produced by penetrating muons them-selves. For details on the laser system and theLEDs, see I. The procedure used to convert theraw ADC and TDC signals produced by the ERPsto physics quantities is described in Sections 3.4.1and 3.4.2. It depended upon linearity of the PMTs,the fanouts, and the ADC/TDC channels and thiswas regularly verified by special calibration runs inwhich the laser and LEDs were excited withparameters systematically varied. The relative lasersignal amplitude was controlled with a precisionoptical attenuator and was also measured withspecial monitor PMTs and electronics. Similarly,the LEDs were pulsed with precision time delaysrelative to a measured common trigger pulse, andwith precisely set amplitudes and pulse widths.Thus it was possible to test the SMT with goodfacsimiles of the light signals expected frommagnetic monopoles of any velocity within our

range of interest. We also created facsimiles of theextraordinary events which should trigger theFMT and HCT. In addition, we used the LEDsto test the capability of the ERP and PHRASEGC triggers to detect a neutrino burst from astellar gravitational collapse by producing ascatter of many low-energy events distributed overthe whole apparatus during a suitable timeinterval. Approximately once a month, everyscintillator in the system was excited by each ofthese test patterns and the data analyzed immedi-ately. Thus we regularly verified that essentially allof the triggering system was functioning asdesigned and, for the relatively small portions ofthe system in which problems arose (a fewcounters typically), we took corrective actions.This system proved to be highly effective forverifying and maintaining the proper operation ofthe experiment.

3.4.1. ERP energy calibrations

A principal task for the ERP system was tomeasure the energy deposited in each scintillatorby physics events and of particular interest was theenergy range of neutrinos from a stellar gravita-tional collapse (B5–B50 MeV). The procedureused for energy reconstruction depended upon aset of nine constants for each counter: a gain Gi

and a pedestal pi for each end of the counter, i ¼0; 1; and a five-parameter function describing lightattenuation as a function of position in thescintillator; as described below, these parameterswere determined regularly for each counter. Forlight levels up to those produced by muons, it wasshown in I that the PMT charge, measured by theERP ADC circuit, was proportional to the numberof photoelectrons to a good approximation. Thusfor an event j; we reconstructed a quantity wj;i thatwas proportional to the number of photoelectronsfrom end i:

wj;i ¼ GiðADCj;i � piÞ ð1Þ

where ADCj;i is the number of counts measured bythe ADC, pi is the pedestal, and Gi is a gain factorinitially set at a convenient value and subsequentlyadjusted as described below to track variations inthe system gain for each channel. The value of wj;i

was then converted into an energy measurement


according to

Ej;i ¼ wj;i=RðZj;iÞ ¼Gi

RðZj;iÞ

� �ðADCj;i � piÞ ð2Þ

where RðZj;iÞ gives the light attenuation factor forthe event j occurring a distance Zj;i from end i; itwas scaled to convert w-values into energy. Thisresponse function for a scintillator was approxi-mated by

RðZÞ ¼ A expð�Z=BÞ þ C expð�Z=DÞ þ E=Z2 ð3Þ

where the constants A;B;C;D;E were counter-dependent and taken to be the same for both ends.Fig. 9 shows typical examples from the horizontaland vertical scintillators. Finally, a weightedaverage of the energy measurements from thetwo ends was taken as the reconstructed energy:17

E ¼E0

ffiffiffiffiffiffiw0

pþ E1


pffiffiffiffiffiffiw0

pþ


p : ð4Þ

The first step in the calibration procedure was todetermine the pedestals pi: The laser systeminjected light at the center of each scintillator atseveral different levels determined by a variablelight attenuator in the laser beam. Approximately100 pulses were fired at each light level starting ator below the ERP trigger threshold (correspondingtoB5–6 MeV) and extending up to the muon level(B34 MeV). One monitor PMT viewed a constantlight signal from the laser to produce triggers, andanother, the reference PMT, measured the atte-nuated light so that its pulse charge was propor-tional to the amount of light sent to the counters.This charge was digitized by a linear referenceADC, whose pedestal was determined indepen-dently. For each side of each scintillator and foreach attenuator setting, the pedestal-subtractedreference ADC values and ERP ADC values wereaveraged. The ERP averages were then linearly fit

0 200 400 600 800 1000Z - distance from tank end (cm)

500

0

1000

1500

2000Vertical Scintillator

0 200 400 600 800 1000Z - distance from tank end (cm)

500

0

1000

1500

2000

Ave

rage

λ-v

alue

Ave

rage

λ-v

alue

Horizontal Scintillator

Fig. 9. ERP muon response data and their fit to the form of Eq. (3). In this figure, the points shown are averages of groups of four of

the directly determined l-values.

17A statistically optimal energy resolution is obtained with a

weighted average E ¼ ðE0m0 þ E1m1Þ=ðm0 þ m1Þ; where mi is the

mean number of photoelectrons reaching end i for an ensemble

of events. However, the resolution improvement with respect to

formula 4 is not significant in this experiment. In MACRO

software the formula 4 was used for historical reasons.


to the reference averages. The intercept of this fitdetermined the ERP pedestal. Fig. 10 shows anexample of a laser pedestal fit for one side of onecounter.A carefully selected sample of muon events, the

muon calibration sample, was used to determineboth the gain parameters Gi and the parameters inthe light attenuation function in Eq. (3). Using thestreamer tube system, we chose events containingjust one, non-showering muon with a welldetermined trajectory. From the muon’s trajec-tory, its Z position in a counter was determinedand the values wj;i were calculated using the newlydetermined pedestals and the values of Gi from theprevious calibration. Initially, values of Gi wereselected so that the average w-value for muonspassing through the longitudinal center of acounter was 600, approximately the number ofphotoelectrons produced by these signals. Eachcounter was divided into 5 cm longitudinal bins,and the three parameters N; a; l (normalization,width, and location) in the function NLðaðw � lÞÞwere least-squares fit to the distribution of w-values observed in each bin; LðxÞ is the well-knownLandau function,

LðxÞ ¼1

2pi

Z cþiN

c�iNepxþp log p dp; cX0: ð5Þ

Finally, the function lðZÞ was fit using Eq. (3) toobtain the five response function parameters foreach counter; the same function was used for bothends of a counter. This procedure was adoptedsince this fitting function well approximates the

shape of the observed w-distribution and itslocation measure, l; is statistically robust.Lastly, the gain constants Gi were adjusted to set

the energy scale. For each event, a first approx-imation to the energy loss in each counter wascalculated from Eq. (2) using Gi values from theprevious calibration and the newly determinedresponse function RðZÞ: Each muon’s pathlengththrough scintillator oil in each counter wascalculated from the ST trajectory, and finally,values of the gains Gi were adjusted so that thedistribution of the fully corrected energy dividedby pathlength, fitted to the Landau function asdescribed above (with w replaced by energy/pathlength), produced a l value of 1:8 MeV=cmto within 72%: An example of such a fit is shownin Fig. 11.This procedure and energy normalization are

not rigorously correct since the actual distributionof the energy loss per unit pathlength for ourscintillators is the Landau distribution (withtheoretically determined values of the parametersa and l) folded into photoelectron statistics andpathlength uncertainties. However, allowing thewidth-setting parameter a to be adjustable is aconvenient way to approximate the distributionbroadening due to these two effects. Similarly,fixing the value of l is a good approximation to thetrue situation in which its value weakly depends onthe number of photoelectrons and the pathlength.Detailed studies which used the full Landau theoryof the energy-loss distribution folded with thesebroadening effects showed that the procedure usedgave an energy normalization systematically toolarge byB5% with an overall uncertainty of aboutthe same size. Since no physics using thesecalibrations was sensitive to the energy loss in acounter at the 5% level, no correction wasincluded to accommodate this small systematicerror.The gains and pedestals for the ERP attenuated-

ADC values were determined in the same way asfor the unattenuated values. The ERP attenuated-ADC input signal was reduced by a nominal factorof 10 with respect to the unattenuated value andwas used for pulses with very large charge. Overmuch of the range of the attenuated-ADC, thesystem response was non-linear, and when ERP

0 200 400 600 800 1000 1200 1400 1600Pedestal-subtracted Reference ADC Counts

2000

1800

1600

1400

1200

1000

800

600

400

200

0

ER

P A

DC

Cou

nts

Fig. 10. Typical linear fit of the ERP ADC counts to the

reference ADC counts for determining the ADC pedestals.


measurements based upon them were used, specialcalibrations were necessary. These are described inRefs. [2,5]. The ERP attenuated-ADCs were neverused in searches for stellar gravitational collapseevents.

3.4.2. ERP timing calibrations

Several of the physics analyses that could beperformed with MACRO required precise timingin the scintillator system. Some examples are:

* the position of an event along the longitudinalaxis of the scintillator could be determined fromthe relative timing of signals at its two ends.Neutrinos from supernovas should be distrib-uted uniformly throughout the scintillatorvolume, so the gravitational collapse analysiscompared the position of events determined bytiming to a uniform distribution to rule outanomalies of either the GC trigger or otherhardware.

* the TOF of particles between two scintillators.This was the only way the neutrino-inducedupgoing muon signal (B0:5 event per day)could be distinguished from the downgoingmuon background (B104 events per day).

* additional tracking information for muonevents which stop or originate in the detectortriggering very few planes of the ST system.

The ERP was a common-stop system with fixedvoltage-level amplitude discriminators to marktimes. A constant current was gated into anintegrator, starting when a PMT pulse exceededthe amplitude discriminator level. The gate wasclosed by a stop pulse, which was common for allof the scintillators in one SM. Subsequently, thevoltage on the integrator was converted into thedigital TDC value.For each scintillator, the time tj;i at which the

light pulse for event j arrives at scintillator end i

relative to an arbitrary event time, unique for eachSM, was reconstructed as

tj;i ¼ toffi � ri TDCj;i þaiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiADCj;i

p þbi

ADCj;i

!:

ð6Þ

In this equation, toffi ; ri; ai; and bi are parametersfor each end of each scintillator. An offset time,governed by PMT delays, cable lengths, andelectronic delays, was accommodated by toffi andthe conversion factor from TDC counts to time

100

80

60

40

20

00 1 2 3 4 5 6 7 8 9 10

Energy per cm (MeV)

Cou

nts

Fig. 11. Typical distribution of energy loss divided by pathlength from a muon calibration sample. Also shown is the Landau function

with least-square adjusted normalization, width, and location parameters as described in the text.


was ri: The terms in ai and bi provide anapproximate timewalk correction to account forthe fact that the TDCs measured time from themoment that the input signal exceeded a fixedthreshold value rather than from the moment apulse starts. Thus, for two pulses starting at thesame time, the one with the smaller amplitude willcross threshold later, starting the TDC later.The reconstructed times tj;i can also be expressed

in terms of the time tj at which the scintillationoccurred in the scintillator and the distance of thescintillation from the center of the counter, zj :

tj;0 ¼ tj þL

2þ zj

� �n

c; ð7Þ

tj;1 ¼ tj þL

2� zj

� �n

c: ð8Þ

The two parameters, L and c=n; in these equationsare the length of a scintillator and the effectivepropagation velocity of light along its longitudinalaxis, respectively; they are the same for both endsof a counter. The lengths of the scintillators aretaken as known but the light velocity parameter n

was regularly determined to accommodate thepossibility of slow chemical changes in thescintillator oil (none was observed). These lasttwo equations are trivially solved to yield

zj ¼1

2ðtj;0 � tj;1Þ

c

n; ð9Þ

tj ¼1

2ðtj;0 þ tj;1Þ �

L

2

n

c: ð10Þ

The procedure to determine the several para-meters in the above equations utilized the same setof laser pulses described above for finding the ERPADC pedestals, the same muon calibration sam-ple, and data from specialized LED calibrationruns.The laser system was configured so that the laser

pulse, suitably delayed, created the ERP commonstop and so measuring the change in TDC valuesas the phototube pulse height was varied allowedthe parameters ai and bi to be determined.The special LED calibration runs used a

programmable pulser which allowed setting preci-sion delays between the LED excitation signalsand a trigger pulse; the trigger pulse, again suitably

delayed, provided the ERP common stop time.These LED data were used to determine the ri

parameters.The muon calibration sample was used to

determine the time offsets, toffi and the lightvelocity parameter, n: First, n and ðtoff0 � toff1 Þ foreach counter were adjusted so that a linear fit ofthe zj values determined from Eq. (9) to thosedetermined by the streamer tube system yielded aunit slope and zero intercept. Then, the relativeoffsets between scintillators were determined bymaking a second pass through the muon calibra-tion sample, reconstructing the TOF of particlesbetween scintillators using the ST tracking data,and assuming all the particles were downgoingand relativistic. An iterative process was used;starting with the offset times determined in themost recent previous calibration, the averagemeasured value of 1=b for particles striking aparticular scintillator was determined. If this wasnot unity to within the statistical precision of themean (about one part in 20,000), the offsets wereadjusted. This change, of course, affected the TOFmeasured between this scintillator and otherscintillators, and so new b determinations werethen made, and the process was continued until itconverged.The ERP common stop was different for each

SM. To determine the TOF for particles thatcrossed supermodule boundaries, a stand-aloneTDC was used to measure the time betweencommon stops on different SMs. The slopes andoffsets of this TDC were determined in a multi-parameter fit which minimized the differencebetween expected TOF and measured TOF, againutilizing events in the muon calibration sample.One way to measure the timing resolution of the

scintillator system for muons is to compare thelongitudinal position in a scintillator reconstructedfrom timing to that reconstructed by the streamertube system. Summed over all scintillators in theapparatus, the central part of the distribution ofthe difference is almost a Gaussian with a standarddeviation of 10:4 cm and a mean of less than 1 cm:This implies that a scintillator’s timing resolutionwas about 0:5 ns which is consistent with thecentral Gaussian peak in the TOF measurementsof muons.


3.4.3. The WFD system

Since the WFD system was based upon a highlynon-linear amplifier, it was necessary to carefullydemonstrate its performance in all of the energyranges in which it was used. To do this, severalcalibrations and checks on this system wereperformed. Low intensity LED pulses were usedto demonstrate its sensitivity to single photoelec-tron pulses; ambient radioactivity was used toconfirm and calibrate its sensitivity at low energies;cosmic ray muons were used to calibrate itsresponse in the medium energy range by compar-ison against the ERP; and high light level LEDpulses were used to calibrate its sensitivity to largepulses at and above the trigger point of the HCT.For slow monopole detection, in which the

expected PMT response was a sequence of singlephotoelectron pulses, the most important role ofthe WFD was pattern recognition. Accurateenergy reconstruction for such an event was notrequired; the waveforms were simply used todistinguish a single photoelectron sequence frombipolar electronic noise, and to get an estimate ofthe TOF across a single scintillator. Thus toconfirm WFD sensitivity to slow monopoles itwas sufficient to generate low light events usingLEDs; examples are given in Figs. 4 and 5 above.This was done weekly when the efficiency of theslow monopole scintillator trigger was measured.For the fractionally charged particle search

more refined calibrations were required. Forenergies below 5 MeV we used the natural back-ground radiation to calibrate the WFD response.Since 1 ms of data were stored in the WFDs formuon events, which typically leave a PMT activefor a few microseconds at most, there was anabundance of radioactivity waveforms during ournormal running for use in this kind of calibration.Care was taken to avoid using data contaminatedby PMT afterpulsing. An energy spectrum wasobtained for each counter and the most prominentpeaks were assumed to originate from the potas-sium line at 1:46 MeV and the thallium line at2:61 MeV: By comparing these lines to measuredbackground spectra in the MACRO hall, and toMonte Carlo simulation of the scintillator re-sponse for background radioactivity events, lowenergy calibrations were obtained for most scin-

tillator counters as shown in Fig. 12. For furtherinformation see Refs. [17,18].For moderate energy events near 40 MeV;

cosmic ray muons were the primary tool forverifying WFD response. Fig. 13 shows twoexamples of muon signals. As for the ERP, themuon calibration sample was regularly used tocalibrate the WFD data. By computing the WFD-determined energy loss per unit path length as afunction of the longitudinal position of the particlein the scintillator, we verified that we reproducedthe known scintillator response function for eachscintillator. This method provided another cali-brated energy point for the LIP analysis. Simplycomparing the WFD integrated charge to ERPADC value on an event by event basis for eachscintillator end gave another verification of theWFD response in the muon energy range. Byverifying linearity between the WFD and the ERPresponses as shown in Fig. 14 we confirmed thatthe WFDs were behaving as expected for muons.Since large amounts of deposited energy are

expected from fast monopoles, dyons, and nucle-arites, it was also important to confirm that theWFD system was sensitive to very large signals.When the WFD system was initially installed, itwas found that large signals could exceed itsmemory capacity because of small positive over-shoots due to capacitive coupling triggering the

Energy (MeV)

Cou

nts

1

10

102

103

0 1 2 3 4 5

1.3 MeV

2.5 MeV

Fig. 12. A typical low energy spectrum observed in the

horizontal counters with the WFD. The two peaks correspond

to the 1:46 MeV potassium line and the 2:61 MeV thallium line.

The peaks are shifted slightly because of smearing from other

decay components in the spectra.


discriminator and because of the long duration ofPMT activity following a highly ionizing event.Most occurrences of the overshoot problem werecorrected by suitable circuit modifications and theresidual, by the HCT trigger circuit described inSection 3.3.5. Subsequently we routinely verifiedthat pulses at the PMT saturation level wereretained by the WFD system. An example of asaturating muon pulse is given in Fig. 13(b).Together all of these calibrations and checks

confirmed that the WFD responded well to alllight levels of interest in this experiment and weverified that it was an effective tool for seeking

both low energy rare events and highly ionizingevents such as those expected from fast mono-poles, dyons, and nuclearites.

3.4.4. SMT performance verification

In order to investigate the overall performanceof the SMT, a detector-wide test of the triggeringefficiency for each individual SMT circuit was firstperformed in February 1996 and subsequentlybecame part of the normal apparatus performanceverification. The detector’s LED system was usedto perform this test. The LED at one end of ascintillator was pulsed for durations varying from350 ns to 6:5 ms and amplitudes were adjusted sothat the PMT nearer the LED always receivedenough light to give an SMT pretrigger and thePMT at the other end was excited by a few singlephotoelectrons. Figs. 4 and 5(a) discussed aboveshow examples of the LED simulated monopolesignals used.Varying the LED amplitude varies the number

of single photoelectrons at the far PMT and soallowed the pretrigger efficiency of the far scintil-lator end to be mapped from zero to 100%. Eachtime the LED was fired, the WFD was read outfrom both ends of the scintillator whether or not apretrigger occurred. The waveforms from eachevent were analyzed and a simulation in softwareof the function of both the TOHM and LI parts ofthe SMT circuitry was performed. This allowed an

PM

T S

igna

l (V

)

0

-0.2

-0.4

-0.6

-0.8

-1.00 200 400 600

nsec

(a)

0

-2.0

-4.0

-6.0

-8.0

-10.00 200 400 600

(b)

Fig. 13. Two examples of the system’s response to muons as shown by the WFDs. In (a) a muon passed through the counter far from

the PMT, while in (b) one passed very near the PMT, showing the hard saturation of the analog system at 10 V:

ERP attenuated ADC counts0 100 200 300 400 500 600 700 800

0

50

100

150

200

250

300

350

WF

D C

harg

e (V

ns)

Fig. 14. Comparison of WFD charge to ERP attenuated-ADC

for muons passing through a scintillator. The line is to guide the

eye.


a priori determination (i.e., independent of theSMT) of the number of times the SMT circuitshould have triggered. The ratio of the observednumber of SMT triggers to the number determinedby the waveform analysis yielded the SMTefficiency for each scintillator counter. A typicalefficiency graph obtained in this way is shown inFig. 15, in which we plot the efficiency of the SMTcircuit as a function of the number of LI counts inthe waveform analysis.Using the clock rates of the LI, we translate the

99% efficiency point into the light yield necessaryto trigger the SMT as a function of particle

velocity. Fig. 16 shows the measured light yield(normalized to that of a minimum ionizingparticle) required for the SMT to achieve 99%triggering efficiency in a typical scintillator as afunction of particle velocity. In the same plot, thelight yield for bare monopoles with a single Diraccharge [1] is also shown. These results are foralmost all the scintillators in the detector (forvarious reasons, measurements could not be madeon B5% of the scintillators).

3.4.5. FPT/FMT/HIPT/HCT verifications

Roughly once a month, the LED calibrationsystem was used to check the performance of theFPT, FMT and HIPT circuitry. LED firings indifferent detector planes with variable delays(from 0 ns to 11 ms) were used to verify the designperformance of the triggers. Moreover, variableLED pulse height settings were used to createPMT pulses ranging from below threshold all theway up to saturation for verifying the thresholdsettings of the triggers. This calibration programinsured that the sensitivity of the MACROexperiment was always maintained at its designvalue.

4. The streamer tube system

The description of the plastic streamer tubes(STs) as the basis of the tracking system wasextensively reported in Refs. [20,22], and refer-ences therein and so we give only a briefdescription here. The basic unit of the tube systemwas a 3:2 cm� 25 cm� 12 m chamber containingeight individual cells, each of dimension 2:9�2:7 cm2: The silvered Be–Cu anode wire in thecenter of the cell had a diameter of 100 mm: Threesides of the cell were coated with low-resistivitygraphite ðo1 kO square�1Þ to perform the cathodefunction by the electrode-less shaping principle[23]. This structure was inserted inside an uncoatedPVC envelope (1:5 mm thick) and closed by twoplastic end caps. The full detector had 6192chambers in operation, for a total number of49,536 wires.A gas mixture of He (73%) and n-pentane

(27%) was used in order to exploit the

LI Count from Waveform Analysis0 5 10 15 20 25 30

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

SMT

Eff

icie

ncy

Fig. 15. Efficiency plot of a typical SMT circuit as a function of

the number of the LI counts produced by the waveform

analysis. This SMT circuit demonstrates 100% efficiency above

its nominal threshold which was 10 LI counts corresponding to

a signal duration of X180 ns:

Ligh

t Yie

ld /

Muo

n Li

ght Y

ield

10

1

0.1

0.0110 10- 4 - 3

β

99% Efficiency Curve

Expected MonopoleLight Yield

Fig. 16. SMT sensitivity curve as measured with the LED

method described in the text. A particle with light yield above

the efficiency curve has more than a 99% probability to trigger

the SMT.


Drell–Penning effect [24] for slow monopoledetection. The gas volume of the active elementswas 465 m3 and the gas distribution systemnormally recirculated the gas mixture to minimizeconsumption. The tightness of the chambers wastested under low vacuum ð1=3 atmÞ [25] and only asmall diffusion occurred for light gases (for 1 atmof helium the measured diffusion rate wasB5� 10�6 atm cm�2 s�1). The gas circulationwas regulated at a rate of one change of the wholevolume per 5 days and the pressure inside thechambers was maintained at a few mbar aboveatmospheric pressure.Each wire behaved as a transmission line with a

characteristic impedance of 330 O and a propaga-tion time of B3:3 ns m�1: Each wire provided adigital readout for tracking; in addition, analogsignals from groups of adjacent wires (32 for thehorizontal planes and 16 for the vertical) wereadded and the summed charge was measured andreadout. The plastic STs with resistive cathodesallowed a two-coordinate readout [26]: one viewwas provided by the wires and the other byexternal pick-up strips. Two different strip systemswere used in MACRO as described above.MACRO STs operated for many years which

allows us to draw sound conclusions about theirlong term performance and reliability and this isdetailed in Section 4.1. In the attico, the char-acteristic impedance of all the strips was optimallymatched to the input resistance of the digitalreadout electronics, as described in Section 4.2.The full detector had a total of 118,848 digitalreadout channels for the wires and strips, plus1548 analog readout channels for the wires.Developments and improvements in the ST read-out are described in Section 4.3 and a descriptionof the streamer tube trigger systems is given inSection 4.4.

4.1. Operation, reliability and lifetime of the

streamer tubes

In MACRO a large number of wires operatedfor a considerable time period. The need toguarantee a long lifetime for the experiment (itstarted taking data in 1989 and was active to theend of 2000) necessitated particular care in the

fabrication, conditioning and qualification testingof the active detector elements. After the finalinstallation, only 17 out of the 49,536 wires(0.03%) gave no signal because of poor solderingor bad connections in the end caps of thechambers. This careful construction, in additionto the favorable low background rates, allowed theimplementation of simple controls and protec-tions. In particular, the experiment reliably oper-ated with a current limit of 10 mA on eachindividual HV channel supplying 192 wires. Themost important remaining factors for providingstable and safe operations were therefore thestability of the gas system and the prevention ofgas leaks. Experience showed that discharges,which increased current and wire stress, wereassociated with air contamination. Thus specialcare was taken to ensure the tightness of thechambers, including the vacuum tests [25] men-tioned above.Regarding maintenance during the lifetime of

the experiment, we replaced a significant fractionof the sections of short PVC gas tubing connectingone chamber to the next due to evident aging.During MACRO’s lifetime, only a further 84 wiresstarted to exhibit permanent discharging behaviorand so they had to be disconnected. Most of thesewires became noisy within one year from thebeginning of operation and, since 1995, only 6wires per year were disconnected. This timeconstant is characteristic of some tiny residualfabrication defect, practically impossible to ana-lyze. The final fraction of dead channels was1:70� 10�3: These dead channels were almostevenly distributed over the whole detector,although there was some excess in the first SMwhich had been put into operation in 1988: 21disconnections in 12 years, 13 of which occurred inthe first 3 years of operation.The most practical method to check the

detector’s gas mixture and its relative HV workingpoint was to monitor streamer charge and theplateau of the singles counting rate. This was donein a short chamber inserted in series in the gas flowline. Because of the very low radioactivity back-ground, this measurement was performed with asmall b source ð90SrÞ kept in a fixed position overone of the 8 wires of the test chamber. Since the


test chamber was continuously held at the samevoltage as the tubes in the full system, the timeevolution of the plateau length as a function of theintegrated charge allowed us to study the agingeffects on the streamer tubes following the sameconcepts described in Ref. [27]. The b sourceprovided a counting rate of about 90 Hz over 2–3 cm of wire. Each streamer pulse contributed acharge of 200 pC; summing both electron andpositive ion components. We therefore integratedabout 1:56 mC per day over a couple centimetersof wire. Fig. 17 shows the time evolution of thesingles counting rate plateau of a single wire testchamber [28]. The charge and current monitoringin the same test setup has shown that noanomalous effects, like dark current, could beobserved at the operating voltage as long as areasonable counting rate plateau could be ob-served. Since the tubes were normally operated200 V above the knee of the plateau, we see thataging with the He+n-pentane gas mixture made atube inoperable after the integration of about

0:3 C cm�1; i.e., approximately one year for thetest chamber. Applied to the underground opera-tion at the Gran Sasso, this aging rate impliescompletely safe operation over a time much longerthan the lifetime of the experiment. In fact, theaverage counting rate due to the natural back-ground as measured by the STs was B40 Hz m�2

of ST plane, corresponding to 12 mHz cm�1 ofwire. Therefore, the expected lifetime of tubes wasat least 3000 times that of the test chamber. Inaddition to the monitoring of this test setup, thecurrent drawn by all the different groups of wireon the detector has been monitored for the wholeoperation period of the apparatus. No appreciablechanges in the average values could be observed,as expected.

4.2. The attico strips

In the attico all the STs, both horizontal andvertical, were equipped with pick-up strips. This isa significant improvement compared to the lower

0

20

40

60

80

100

120

140

160

180

3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2 5.4 5.6

21-Mar-1995 03-May-1995 (0.03 C/cm)22-Aug-1995 12-Jan-1996 06-Feb-1996

HV (kV)

Sing

les

/ sec

(Start)

(0.12 C/cm)(0.24 C/cm)(0.27 C/cm)

Fig. 17. Singles rate plateau as measured on the test chamber for different values of integrated charge. A localized b source spreadingover about 2 cm of wire was used for the measurement.


detector described in Refs. [20,22] which had stripsonly for the horizontal STs. This enhancementpermitted full tracking capability using the verticalSTs.The pick-up strips for the horizontal planes of

the attico were made using 4 mm thick honey-combed polypropylene foil. The strips were 30 mmwide and the spacing between adjacent ones was2:5 mm: As in the lower part of MACRO, thepick-up strips were grouped in bands of 16elements. Laboratory measurements of the proper-ties of these strips showed that the signal attenua-tion was 0:038 dB m�1; the characteristicimpedance of a single line was 4071 O; and thecross talk between neighboring strips, 14:1 m long,was 5%: Each strip was terminated with 39 Oimpedance to reduce the dependence of pulseamplitude on the streamer location.The 40 O characteristic impedance of these

strips permitted an increase in the threshold ofthe front end electronics by about a factor 2.5compared to what was used with the lowerdetector strips which had a characteristic impe-dance of only 8 O: The efficiency of the attico stripsystem was 3% higher than the system used in thelower detector.

4.3. Attico wire and strip readout

The acquisition system for attico wire and stripsignals used a new and improved version of theSTAS (ST Acquisition System) model C267 and ofthe Splitter Board model SY 480, both producedby CAEN [29]. Because the clock signals issued bya STAS were sent back to the same STAS, this newread-out system was insensitive both to the lengthof the connection cable to the read-out chain andto the delay introduced by the electronics. Theserial output from the chains was acquired fromthe STAS on the high-to-low transition ofacknowledgment signals, thereby synchronizingthe data read-out with the STAS clock.In this new STAS, each channel had its own 512

word FIFO memory and so, in contrast to themodel C187 used in the lower detector, noarbitration logic was necessary. This allowed aclock frequency of up to 4 MHz resulting in fasterdata acquisition. Further, all connections to the

Splitter Board were made optically to avoidground loop problems.In the lower detector, each read-out chain

serviced just one detector plane. In the attico,however, each such chain typically serviced severalplanes to better exploit the STAS module memory.For triggering purposes each plane of counters

sent a FAST–OR signal to the trigger modules. Allthe 23 chains used to read out a DSM in the atticofit into three 8-channel Splitter Board units.

4.4. The ST monopole trigger

The streamer tubes provided a monopole triggercompletely independent of the scintillator system.To achieve this, helium and n-pentane were used inthe gas mixture to exploit the Drell–Penning effect[24]. A monopole passing through a helium atomleaves it in an excited metastable state (Drelleffect). Excited helium atoms can collide with an n-pentane molecule and ionize it (the Penning effect).This sequence provides an efficient process fordetecting monopoles in a streamer tube in thevelocity range of 10�4pbp10�3: For monopolesof bX10�3; the standard excitation–ionizationmechanisms are effective [30].For bp10�3; the analysis strategy was based on

TOF measurements. Due to the large monopoleionization power [30] in the high b region, thelimited proportionality of STs [31] allowed us toreject the muon background by measuring andanalyzing the charge released in the ST system [32](see Section 4.5.4). Here we give a generaldescription of the ST magnetic monopole triggersystem.

4.4.1. The trigger system

Large temporal windows (hundreds of ms) wereneeded in order to search for monopoles (or otherexotic particles) with b as small as B3� 10�5:This, in turn, implied the need for good rejectionpower against the natural radioactivity back-ground (B40 Hz m�2 of ST plane). To achievethis we required alignment of hits in both spaceand time, as described below.Two different triggers were implemented, one

using signals from the STs of the horizontal planesand the other using those from the vertical planes.


Monte Carlo simulations showed that this triggerconfiguration provided a large acceptance of about8500 m2 sr for an isotropic particle flux.

4.4.2. The horizontal streamer tube trigger

A full description of this trigger (which tookdata from 1992) was given in Ref. [33]. Because oftheB3 m gap between the lowest horizontal planeof the attico and the highest horizontal plane ofthe lower detector, a trigger using only the lowerdetector had a better signal to noise ratio than onewhich incorporated the attico signals also.A supermassive particle should cross the appa-

ratus with no significant change in its initialdirection and speed, thus producing a straight linein the z-time plane, where z is the vertical positionin the detector. The trigger algorithm was designedto be sensitive to these space–time tracks. Itoperated as the OR of 320 delayed coincidences(b-slices) with a resolution of B2 ms each. Half ofthese b-slices were arranged to detect upgoingparticles and the other half, downgoing ones. Inthis way a rough TOF measurement was alsoperformed.The trigger circuit achieved this performance

using a serial-in/parallel-out, 480 cell shift registerdriven by a 1 MHz clock, so creating a 480 msmemory depth. Every microsecond, a search wasperformed among the b-slices for at least onespace–time alignment (see Fig. 18) over a mini-mum number of planes with signals, normally 7.When successful this produced the Low THRtrigger. There was also another, more selectivetrigger that required particles to cross at least 8planes with TOFX3 ms; it produced the HighTHR signal. Its rate was a factor of about 100smaller than that of Low THR, which enabled it totrigger the scintillator system WSM, and so forceacquisition of scintillator waveforms.There were 11 trigger units creating parallel

signals coming from groups of two adjacentcentral modules. This reduced the accidentalcoincidence rate to the tolerable level of B0:2 Hzwith negligible loss of acceptance.For sufficiently slow particles (TOFX3 ms cor-

responding to bp5� 10�3), the circuit gavedifferent outputs for upgoing and downgoing

cases. As described below, this was used to provideas much information as possible on each event.

4.4.3. The vertical streamer tube trigger

Each of 26 trigger units dealt with signalscoming from a single vertical module. Thetemporal tracking algorithm described abovecould not be used efficiently in the vertical triggersince a monopole with bB10�4 would takeB30 ms to cross a vertical module. Since themaximum jitter in an ST was C600 ns; its timeresolution was insufficient to exploit time align-ment in the velocity range of interest. Therefore,only spatial alignment of hits was required with asegmentation of 50 cm along the vertical axis,corresponding to the OR of 16 wires.The logic required that at least 4 out of 6 planes

showed hits within a time window of 45 ms: Astricter coincidence of 9:5 ms was required forsignals coming from the inner or the outer planes

10

9

8

7

6

5

4

3

21

MAJORITIES for the 7 β-sliceth

TRIGGER OUTPUTHigh THR Low THR

Fig. 18. Simplified scheme of the shift register used by the

horizontal streamer tube trigger. Each shift register chain gets

signals from one horizontal streamer tube plane. In particular

the definition of the 7th b-slice is shown; other b-slices weresimilarly defined.


separately. In this way, the radioactivity triggerrate was reduced to B0:1 Hz in the wholeapparatus.If the TOF between inner and outer planes was

greater than 600 ns; it was also possible by timingto discriminate between particles entering andleaving the apparatus and so produce a corre-sponding trigger output. Otherwise a fast mono-

pole (bX3� 10�3) output was produced. Therewas also a high threshold condition in this circuit.It required at least 5 out of 6 active planes to be hitin coincidence with the same timing as describedabove. The slow monopole output of this logic wasused to trigger the WSM, forcing WFD acquisi-tion from the scintillators.

4.4.4. Trigger output management

Management of all the trigger outputs wasperformed by a suitable streamer monopole super-visor whose function was to store as muchinformation as possible for each individual event.For example, some monopole trajectories couldsatisfy the trigger conditions before the particlefinished crossing the whole apparatus. In this caseseveral hits could have been lost, but this wasavoided by using both the B1 ms memory depthof the front-end electronics and the trigger’s abilityto recognize the direction of slow particles. Properdelay of the data acquisition allowed the collectionof all possible information from the event.The design of this trigger system paid particular

attention to the possibility that monopoles mightcatalyze nucleon decay [34]. By triggering any fastparticle trigger, decay products could have causeddead time leading to loss of some, or possibly all,monopole information. To avoid this, the start ofevent acquisition was delayed by 480 ms to allowtime for the monopole to exit the apparatus.Summarizing, the magnetic monopole trigger of

the ST system in the MACRO experiment wassensitive to GUT monopoles with b as small asB10�4: Because of this large b range, strongrejection of the hits caused by natural radioactivitywas needed and it was obtained by requiring bothspatial and temporal tracks. In some circum-stances discrimination on the direction of theparticle was also possible, allowing preservation ofthe maximum amount of information, even in the

case of a monopole-catalyzed nucleon decay. Twoindependent triggers were used, one for thehorizontal streamer tubes and the other for thevertical ones. Both this configuration and the sizeof the apparatus allowed a geometrical acceptanceof about 8600 m2 sr for an isotropic particle flux.

4.5. Track reconstruction and performance of the

ST system

Large numbers of cosmic ray muons werestudied in an effort to assess the overall perfor-mance of the ST system. The tracking algorithms,geometrical survey, tracking angular resolution,absolute pointing accuracy, and overall detectionsensitivity for these muons are discussed below.

4.5.1. Tracking

The muon tracking procedure was mainly basedon the streamer hits of the horizontal planes(central tracking). The track was reconstructed intwo views. In the X–Z view (the x axis was that ofthe gallery) it was reconstructed by means of thestreamer tube hits. The inductive pick-up stripspermitted the tracking in the D–Z view, where theaxis was normal to the strips. With respect to thedescription in I the tracking package has beenfurther developed. The hits of the North andSouth frontal streamer tubes, which are parallel tothe horizontal ones, have been inserted in the X–Z

view tracking. Furthermore, new tracking algo-rithms have been implemented in other views. TheY–Z view tracking (lateral tracking) was based onthe hits of the East and West lateral streamer tubes(perpendicular to the horizontal ones) and on acombination of horizontal tube and strip hits. Atleast 5 hit clusters were required: 3 have to be onthe lateral planes, the other 2 were either lateralclusters, horizontal clusters, or both. This trackingalgorithm enlarged the acceptance in the zenithregion with jcos y jo0:2 by a factor B1:4: InFig. 19 an event is shown in which the lateraltracking permitted us to recover the attico hitsbelonging to the track. The first two plots arereferred to X–Z and D–Z views, the last one to theY–Z view. The track is properly reconstructed ineach view.


The X–Y view tracking (attico strips) was basedon the attico lateral pick-up strips. Because of thegeometry of the apparatus, this algorithm wassimilar to the one used for lateral tracking. This

X–Y tracking almost doubled the acceptance formuons with jcos yjo0:2:The reconstruction of the track in space was

mainly based on central track matching. Wire and

X-Z VIEW (WIRES + SCINTILLATORS)

D-Z VIEW (STRIPS)

500

1000

500 1000 1500 2000 2500

500

1000

500 1000 1500 2000 2500 3000

Y-Z VIEW (S.T. FAST HITS + SCINTILLATORS)

250

500

750

1000

200 400 600 800 1000 1200

Fig. 19. Display of a neutrino-induced event (upward-going muon). The track is reconstructed by means of the horizontal streamer

planes (the ‘‘X–Z View’’ for the wires and scintillators, and the ‘‘D–Z View’’ for the strips) and of the vertical streamer planes (the ‘‘Y–

Z View’’). In this event the central tracks are reconstructed only in the lower part of the apparatus. The lateral track recovers also

lateral streamer hits in the attico. The boxes represent the scintillator hits.


strip tracks in the horizontal plane views wereassociated by geometrical criteria (containment ofassociated hits in the detector active volume) andby comparison of the hit plane pattern. Also atticostrip and lateral tracks could be matched to oneanother and to the central tracks. In this case thematching was based on the identification of acommon subset of horizontal plane hits.When only a Y–Z track resulted from the initial

reconstruction, a spatial reconstruction was stillsometimes possible. We were able to produce aspace track when there were at least three pairs ofhorizontal tube and strip hits belonging to thetrack, but not otherwise. Typically, when therewere more than three such tube-strip hit pairs, theinitial reconstruction successfully generated a fullspace track.When tracks in different views could be

associated, we were able to improve the qualityof track reconstruction by using a chi-squaredminimization procedure. An example of associatedtracks was shown in Fig. 19. This procedure ishighly efficient for low multiplicity events (nomore than 3 muons). For high multiplicity events,a different technique was used [35].

4.5.2. Geometry survey and tracking correction

The geometry of MACRO was precisely sur-veyed by standard methods and the positions ofstreamer tubes and strips were checked by muontracking. Minor correction were introduced (lessthan 0:5 cm for horizontal planes, up to somecentimeters for lateral planes). Finally muontracking in the streamer tubes permitted a crosscheck of the scintillator geometry.

4.5.3. Tracking angular resolution and absolute

pointing

One of the most important characteristics of theMACRO detector was its excellent angular resolu-tion combined with its large acceptance. Conse-quently the experiment was sensitive to very faintcosmic point sources which could produce under-ground muons. It was, therefore, important tofully understand the detector’s actual angularresolution for through going cosmic ray muons.This angular resolution was dominated by multi-ple Coulomb scattering in the rock overburden but

it also included small components from the spatialresolution of the STs and the average error in theirtrack reconstruction.The space angle between double muons, as

shown in Fig. 20, was used to infer the overallangular resolution of the apparatus. The datashown were taken after the attico was installed andit shows that 68.3% of the events have a spaceangle yp0:91 between the two muons. If weassume that each muon was scattered indepen-dently, then the overall scattering angle for a singlemuon is 0:91=

ffiffiffi2

pC0:641; better than our previous

published value of 1:31=ffiffiffi2

pC1:01 [10].

Using the well known shadow in the cosmic raysof the moon [36–39], we independently estimatedthe actual angular resolution of the apparatus. Byobserving this shadow (see Fig. 21), we alsoverified the absolute pointing accuracy of theapparatus. MACRO’s observation of the moon’sshadow was the first such detection underground,and its analysis is detailed in a separate publication[40], in which it is shown that the angularresolution found in this way is in agreement withwhat is obtained using double muons.

4.5.4. Timing performances and charge

measurements

The timing capabilities and charge sensitivity ofthe ST system and of the related electronics haveplayed a central role in the ability to search formagnetic monopoles in the 10�4pbp5� 10�3

velocity range.

2000

4000

6000

8000

10000

1.250

2000

4000

6000

8000

10000

0 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00Degrees

Eve

nts

Fig. 20. Space angle distribution for double muons. The shaded

area, covering angles less than 0:91; contains 68.3% of the total

area.


Charge and time of the streamer pulses wererecorded by the QTP system fully described in Iand [41]; here we give only a brief discussion of theoverall system performance extracted from thestudy of single muon events.Time information from the QTPs was used to

search for a time alignment of hits from differentdetector planes (time track). Time track para-meters were used to compute the time of flight ofparticles in the lower part of MACRO. Fig. 22shows the distribution obtained for a sample ofB7� 105 muons as reconstructed by the analysisused in the search for magnetic monopoles in thestreamer tube system. The distribution is nicelyGaussian with a mean value of B15 ns corre-sponding to the flight time of relativistic particlesthrough theB4:8 m height of the lower MACRO.Its standard deviation isB120 ns; fully compatiblewith the expectations due to the drift time jitter ofionization electrons in the tubes. This resolutionallowed us to effectively search for slowly movingmonopoles (bp5� 10�3), rejecting relativisticparticles by a time of flight cut.

The logarithmic response of the streamer chargeto high ionization is used to search for fastermonopoles [31]. Although the energy loss (QM)ofbX5� 10�3 monopoles exceeds B102 times that(Qm) of relativistic muons, the logarithmic depen-dence of the streamer charge to ionization, and theresponse of the front end electronics compress theminimum QM=Qm ratio to B4: As a consequence,

Fig. 21. Shadow of the moon measured as w2 levels (scale at right). The superimposed disk corresponds to the moon in its expectedposition without taking the geomagnetic displacement into account. The region centered on the moon position has a w2 ¼ 32

corresponding to a signal of statistical significance of about 5:6 s:

0-250-500-750-1000 250 500 750 1000

Streamer Tubes TOF (ns)

200

0

400

600

800

1000

1200

1400

1600

Cou

nts

per

bin

Fig. 22. ST measurement of time of flight of muons crossing

lower MACRO.


we made careful corrections to the measuredstreamer charge accounting for geometrical effects[32] and small, long term variations in the gascomposition. As an example, the streamer chargemeasured for single muons as a function of thetrack projection along the wire direction is shownin Fig. 23 for tracks firing one and two adjacenttubes in one plane, respectively.

4.5.5. Streamer tube sensitivity

Over its lifetime, the ST system detected cosmicray muons with uniform efficiency. This wasdemonstrated by examining its sensitivity to verysmall variations (at the percent level) in theabsolute muon rate over a long period of time.In particular, the seasonal variations of theunderground muon flux were studied.Data collected with all 6 lower SMs from 1991

to 1997 were used to search for seasonal variationsin the observed muon rate. These rate variationsare due to the density variations in the upperatmosphere caused by seasonal temperaturechanges. Density variations in the interactionregion for primary cosmic rays affect the under-ground muon rate by changing the ratio of nuclearinteraction to decay of secondary pions and kaonsin the induced air showers. A complete descriptionof this analysis is given in Ref. [16].The data were analyzed by using the expression:

DRm

Rm¼ aT

DTeff

Teff

ð11Þ

where Rm is the muon rate, aT is the temperaturecoefficient, and Teff is the effective temperature,defined as a weighted average of temperaturesmeasured at different isobaric levels. It is given by

Teff ¼P

i TðxiÞ expð�xi=lÞPi expð�xi=lÞ

ð12Þ

where l is the effective mean free path for nuclearinteractions of secondary pions and kaons and xi

is the atmospheric depth at the ith isobaric level.The expected size of this effect is aT ¼ 0:96 [42],

and although this implies that the rate fluctuationsare at the percent level, we nevertheless measuredthem and determined the temperature coefficientto be 0:8370:13: In Fig. 24, the observed relativerate variations are superimposed on the observedrelative temperature variations. The maximumvariations from the mean are seen to be of orderof 71%; and the excellent correlation is clear.

5. The nuclear track detector

The main purpose of the nuclear track-etchdetector was to perform combined and indepen-dent searches for magnetic monopoles and otherrare massive particles (i.e. nuclearites). It wasplaced horizontally in the middle of the lowerdetector, and vertically on the east (lower and

0 1 2 3 4 5

Lproj (cm)

CLS = 1CLS = 2

0

50

100

150

200

Q (p

C)

Fig. 23. Dependence of the streamer charge on the projected

length of the track along the wire for tracks firing 1 tube

ðCLS ¼ 1Þ and two adjacent tubes ðCLS ¼ 2Þ:

-3

-2

-1

0

1

2

3

0 10 20 30 40 50 60 70 80

Months since January 1991

Mo

nth

ly V

aria

tio

n

( %

)

muons

Temperature

1991 1992 1993 1994 1995 1996 1997

Fig. 24. The superposition of the percentage fluctuations in the

muon rate, DRm=Rm; and the percentage fluctuations of the

effective temperature, DTeff=Teff ; since January 1991.


upper) and north (lower) walls. It was organized instacks (‘‘wagons’’), consisting of three layers ofCR39, each 1:4 mm thick, three layers of Lexan,each 0:2 mm thick, and an aluminum absorber1 mm thick. Each wagon was placed in analuminized Mylar bag filled with dry air and eachcould be inserted or removed by sliding it overrails.The effective surface of the horizontal detector

was 812 m2 (13536 ‘‘wagons’’, 24:5� 24:5 cm2

each). On the north vertical surface there were 816wagons of 24� 24 cm2 area, corresponding to47 m2: On the east wall there were 7008 wagons,corresponding to 404 m2: The total nuclear trackdetector surface was thus 1263 m2; its totalacceptance was 7100 m2 sr for an isotropic fluxof fast magnetic monopoles.

5.1. The CR39 nuclear track detector

The track-etch subdetector used two types ofnuclear track detectors, CR39 and Lexan. Most ofthe work concerned the development and proces-sing of CR39. Lexan has a much higher thresholdcompared to that of CR39, making it sensitiveonly to relativistic monopoles. Since no magneticmonopole candidates were found in the CR39,there was no need to analyze any of the Lexan inthis experiment.The CR39 was made by the Intercast Europe

Co. (Parma, Italy) in several batches from 1988 to1999. Calibrations were made using both relativis-tic and low velocity ions, and they covered a rangefrom about 25 MeV cm2 g�1 to about 6�103 MeV cm2 g�1 in Restricted Energy Loss(REL, the energy loss within a 10 nm cylinderaround the particle’s trajectory) and from 4� 10�3

to B1 in b ¼ v=c:

5.1.1. Calibrations with relativistic ions

The CR39 batches were calibrated at differenttimes with several species of relativistic heavy ions:(i) oxygen nuclei of 16 GeV nucleon�1 at the BNL-AGS (1988), (ii) neon nuclei of585 MeV nucleon�1 at Berkeley ð1990Þ; (iii) siliconnuclei of 14:5 GeV nucleon�1 at the BNL-AGS(1990), (iv) sulfur nuclei of 200 GeV nucleon�1 atthe CERN-SPS (1990), (v) gold nuclei of

11:3 GeV nucleon�1 at the BNL-AGS (1992), (vi)lead nuclei of 158 GeV nucleon�1 at the CERN-SPS ð1994; 1996Þ; and (vii) in 1999, iron nuclei of1 GeV nucleon�1 at the BNL-AGS [20,43–46].Here we present the most recent calibration resultsobtained with the lead and iron beams.Stacks of CR39 sheets placed before and after a

target were exposed to 158 GeV nucleon�1 Pb82þ

and to 1 GeV nucleon�1 Fe26þ ion beams. Eachstack was composed of CR39 foils from differentbatches; it was then possible to determine theresponse of each batch to the same incident ion.The exposures were carried out at normal incidenceat a density of B400 ions cm�2: When passingthrough the stack, the incident ions producednuclear fragments with Zo82 and Zo26 for thelead and iron beam, respectively. Both the beamand the fragment ions with electric charges abovethe detector threshold were recorded.After exposure to the lead (iron) ions the CR39

foils were etched in a 6-normal NaOH watersolution at 701C for 30 ð45Þ h: Conical etch-pitsare formed on each side of a detector foil along theion path. The etch-pit base areas, their eccentricity,and central brightness were measured with anautomatic image analyzer [47], which also pro-vided the absolute coordinates of the etched cones;this allowed the tracking of the detected nucleithrough the stack.A typical average base area distribution is

shown in Fig. 25. It refers to the CR39 made inNovember 1995 and etched for 30 h in 6-normalNaOH at 701C shortly after exposure to the leadbeam. The averages are made over six measure-ments of the bases of the etched cones in threeCR39 foils. A large number of peaks is evident.Each peak corresponds to a relativistic nuclearfragment with different charge. The lowest de-tected charge is Z ¼ 5:The response of the CR39 is measured by the

reduced etch ratio p ¼ vT=vB; where vT and vB arethe etching rates along the particle track and of thebulk material, respectively. For each detected ion,p was calculated based on the measurements of thesurface areas of the etch-pits and of the bulketching rate vB:In Fig. 26 are shown the calibration points, p

versus REL for the CR39 foils, produced from


1989 to 1999, and exposed to the BNL-AGS Fe26þ

beam of 1 GeV nucleon�1: In general we can saythat each production batch has its own calibrationcurve with no evident connection with theproduction time. The variations are mainly dueto slightly different production conditions.By comparing the results from different expo-

sures we were able to test the existence of possibleaging and fading effects of the MACRO CR39.Details are given in Ref. [46]; within experimentalerrors, no aging or fading effects were observed.

5.1.2. Calibrations with slow ions

Samples of CR39 were exposed to protons anddeuterons from 50 to 200 keV and to berylliumand silicon ions from 100 to 400 keV [44]. Othersamples were irradiated with 4He2þ nuclei of 6 and10 MeV: After exposure, the samples were etchedin 6-normal NaOH water solution at 40:01C fordifferent times, ranging from 1 to 14 h: In Fig. 27the measured p-values are plotted versus REL.The solid circles correspond to relativistic

nuclear fragments with charge 8pZp68 and theother points are for other nuclei and energies asindicated. The reduced etch rate for low velocityions follows the relativistic calibration curve.These results show that the response of the CR39used in this experiment is well described by a singlecurve of the reduced etch rate, p; versus REL bothfor relativistic (bB1) and for low velocity ions,down to b ¼ 4� 10�3: Furthermore, within thelimits of the available statistics and of thesystematic uncertainties, our results are compati-ble with a full contribution of the nuclear stoppingpower to the CR39 response.

5.2. Search for magnetic monopoles with the track-

etch detector

The restricted energy losses of magnetic mono-poles of different magnetic charges in the nuclear

210

310

1

2

3

p =

v /

vT

B

REL ( MeV g cm )-1 2

Apr. 89 Nov. 89 Nov. 91

Nov. 93 Nov. 94 Nov. 95

Nov. 96 Nov. 97 Apr. 99

Fig. 26. Calibrations of CR39 of different production batches

exposed to the 1 GeV=nucleon iron beam (BNL-AGS, 1999).

Z=5Z=7

Z=9

Z=20

Z=82

0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000

Average Area (pixel )2

0

50

100

150

200

250

300

350

Eve

nts

/ 20

pixe

l2

Fig. 25. Number of nuclear fragments from the fragmentation of relativistic Pb ðZ ¼ 82Þ ions in aluminum; the horizontal scale is theaverage etch-pit base area, which is proportional to the charge Z of the nuclear fragment.


track detector CR39 are presented in Fig. 28 as afunction of b:The detection thresholds of two types of CR39

used by the MACRO experiment (EN3 and L6)[48] are also shown.The track-etch detector was used both as a

stand-alone detector and in a ‘‘triggered mode’’with the scintillator and ST systems.Until now 530 m2 of nuclear track detector have

been etched and analyzed; the analysis was limitedto the CR39 sub-detector only. In May 2000 webegan the analysis of the entire nuclear trackdetector, which we plan to complete by April 2002.

5.2.1. Etching and analysis of CR39

After extraction from the apparatus, threefiducial holes of 2 mm diameter are drilled in each‘‘wagon’’ with a precision drilling machine (theposition of the holes is defined to within 100 mmaccuracy). The top CR39 foils (labeled ‘‘L1’’) are‘‘strongly’’ etched in a steel thermostatic tankð24 foils=etching tankÞ in a 8-normal NaOH watersolution at 80–901C until their final thickness isabout 300–400 mm: An estimate of the bulketching rate is obtained by measuring the initial

and final thickness of four foils. After a ‘‘strong’’etching the surface of the sheet is opaline. Thesignal we are looking for is a hole or a biconicaltrack with the two base-cone areas equal withinthe experimental uncertainties. In order to test theexistence of possible frontside–backside systematicdifferences, several stacks from the MACROdetector were exposed in December 1999 to theBNL-AGS iron beam, at 01; 301; and 451 incidentangles. Strong etching was performed on the L1foils and, for each incident angle, the minor axeson the front ðaFÞ and back side ðaBÞ of each sheetwere measured for about 100 tracks. We obtainedaF=aB ¼ 1:0070:02 (3 standard deviations) for allangles, showing the good front–back uniformity ofthe detector response.Once etched the sheets were scanned twice by

different operators with a 3�magnification op-tical lens looking for any possible optical inhomo-geneity. An optical inhomogeneity is usually aparticle track or a defect of the material. Thedouble scan guaranteed an efficiency of 100% forfinding a possible signal. The detected inhomo-geneities were then observed with a stereo 40�microscope: About 10% of them were classified as

βR

EL

( M

eV g

c

m

)-1

2

1110

-1

101

102

103

104

105

10-2

10-3

10-4

10-5

L6EN3

9gD

6gD

3gD2g

D

g = gD

Fig. 28. Restricted Energy Losses versus b for magnetic

monopoles with magnetic charge g ¼ gD; 2gD; 3gD; 6gD and

9gD: in the CR39 nuclear track detector. The detection

thresholds of two types of CR39 used by the MACRO

experiment (EN3 and L6) are also shown.

REL ( MeV g cm )-1 210 2 10 3

Red

uced

Etc

h R

ate

p

1

2

3

4

5

6 Relativistic Ions

Be 100 − 400 keV

Si 200 − 400 keV

H 50 − 200 keV

D 50 − 200 keV

He 6 − 10 MeV

Fig. 27. Measured values of the reduced etch rate p versus REL

for high velocity and low velocity ions; the solid line is the best

fit through the relativistic data points (solid circles).


‘‘tracks’’ and required further analysis. Each trackwas observed under an optical microscope with80� or 160� magnification; the minor axes in thefront ðaFÞ and the back ðaBÞ sides were measured.We required that D ¼ jaF=aB � 1jp3 times theerror on the difference; this condition was satisfiedby about 20% of the tracks.In order to compute the p-values and incident

angles y for the front and back sides, the trackmajor axes were measured too. Finally a track wasdefined as a ‘‘candidate’’ if p and y on the frontand back sides were equal within 15%. For eachcandidate a coincidence in positions, angles andRELs was required in the corresponding CR39bottom layer (labeled L5) and, if necessary, also inthe intermediate layer (labeled L3).The L5 layers were etched in 6-normal NaOH

water solution at 701C: Thickness measurementswere made to determine the bulk etching ratevB ðB1:2 mm h�1Þ: Using the measured incidentdirection of a ‘‘candidate’’ with its uncertainty andthe fiducial marks, we selected an area of about0.5–1 cm2 in the L5 layer. This area was accuratelyscanned under an optical microscope with highmagnification. The number of analyzed L5s was 5–10% of L1 layers. If a two-fold coincidence hadbeen detected, the CR39 L3 layer would also havebeen analyzed in the same way as L5. However, upto now, no two-fold coincidence has been found,

that is, no magnetic monopole, nuclearite orcharged Q-ball candidate has been detected.The 90% C.L. upper limits on the g ¼ gD

monopole flux from this measurement is at thelevel of 2:4� 10�16 cm�2 s�1 sr�1 at bB1; and3:5� 10�16 cm�2 s�1 sr�1 at bB10�4:

6. The TRD system

Transition radiation detectors are commonlyused for fast particle identification both in the newgeneration accelerator experiments [49] and incosmic ray physics [50,51]. Transition radiation(TR) is emitted in the X-ray region whenever anultrarelativistic particle crosses the interface be-tween two materials with different dielectricproperties. Since this radiation increases (over acertain range) as a known function of the incidentparticle’s Lorentz factor g; it is possible to evaluatea particle’s energy E ¼ mgc2 in this range if themass m is known.For MACRO we built three TRD modules with

36 m2 total surface area (see Fig. 29). In the GranSasso laboratory, the distribution of the muonresidual energy is estimated to be nearly flat up to0:1 TeV and to fall rapidly in the TeV region; itsmean value is B0:31 TeV: Thus, the TRD wasdesigned to be sensitive in just this muon energy

Fig. 29. Artist’s view of the three TRD modules mounted on rails on the top of MACRO. The electronics racks, positioned 2 m from

the front of the detectors, are also shown.


range, from 0.1 to 1 TeV: Below this range there isno transition radiation and the detector exhibits asmoothly increasing response versus g resultingfrom the relativistic rise of the muon ionizationenergy loss in the TRDs gas-filled counters. Forenergies greater than 1 TeV; where the ordinarymuon flux is estimated to be about 6% of the total,the transition radiation signal is saturated.The measurement of the energy of muons in the

TeV region, even with modest resolution, opens awider range of experimental opportunities forcosmic ray physics both in underground and insurface laboratories. In fact, this technique allowsdirect measurement of the muon energy spectra inan underground experiment, independent of bothassumptions about zenith angle distribution andthe calculations of the energy loss in the surround-ing rock that are usually needed [52].

6.1. TRD general description

A typical TRD consists of an array of radiatorsand proportional chamber layers. In our case, thechamber consisted of independent 6 m longproportional counters, each with a 6� 6 cm2

square cross-section and 0:9 mm thick polystyrenewalls. These counters were positioned close to-gether between radiator layers thereby reducinginefficient zones [53].The readout of the TRD was based on the

‘‘cluster counting method’’. In this method, thewire signal is sharply differentiated in order todiscriminate the ionization clusters produced byX-ray photoelectrons from the d-ray background.Requiring a minimum number of counts in acluster discriminated fast radiating particles fromslower non-radiating ones.Each of the three TRD modules had an active

volume of 6� 1:92� 1:7 m3 and was based on 10layers of 32 tubes each, interleaved by 10 foamradiator layers 10 cm thick. This radiator materialwas made of polyethylene foam. We selectedEthafoam 220 (about 35 g l�1 density) that ex-hibits cells of approximately 0:9 mm diameter and35 mm wall thickness. This ensured a relativelywide range for the measurement of g; correspond-ing to the desired energy region for muons. Thebottom tube layer was placed on an eleventh

radiator. Thus the detector was symmetric withrespect to upgoing and downgoing muons, and soit could measure the energy of upgoing neutrino-induced muons that passed through it.The counter cross-section, 6� 6 cm2; was a

compromise between efficiently converting thetransition radiation photons in an argon-basedgas mixture and, at the same time, keeping amuon’s ionization energy release at a relatively lowlevel. The counters were built from extruded tubesinternally painted with graphite to a resistivity of10 kO square�1: They were filled with anAr ð90%Þ–CO2 ð10%Þ mixture for safety and lowcost. With this gas mixture and a 2540 V highvoltage supply, the gain was maintained at about104; providing good proportionality for photonsup to 22 keV energy.We measured the uniformity of the signal pulse

height along and across the anode wires with acollimated 55Fe source and found that it wasuniform within 71:5%.Searching for tracks in the TRD was done by

using tracks reconstructed in the ST system. Hitsin TRD tubes were assigned to an ST track whenthey were within 76 cm of it in the viewperpendicular to the TRD anode wires. Thisassumption fully takes into account the localiza-tion uncertainty due to the finite resolution of theSTs and at the same time it eliminates most of thespurious hits.

6.2. TRD readout and trigger electronics

The electronics readout for each counter wire isshown in Ref. [53]. It consisted of an analogsection and a digital section mounted on a 5�5 cm2 electronics card inserted in the tube endcapimmediately after a 1 nF high voltage capacitor.The input stage of the analog section was acommon base transistor which collected the wirecurrent signal through a series resistor of 270 O;the characteristic impedance of the proportionaltube. The output voltage pulse was shaped by apassive filter to a FWHM of about 80 ns andamplified using an NE 592 amplifier. The voltagegain of this stage was adjusted to about 100 bymeans of a variable resistor. The amplified pulsewas sent through a buffer (LH0002) on an 8 m


long coaxial cable to the digital section. Thedynamic range and the bandwidth of this sectionwere 1 V and 20 MHz respectively, linear to within1%. The typical pulse height produced by a5:9 keV X-ray source on the 50 O termination atthe digital section input was about 100 mV:The digital section consisted of a fast compara-

tor (LeCroy MVL 407 with an active delay of450 ns to account for the signal readout triggerdelay), a 3-bit counter (1=2 LS 393), and a shiftregister (LS 165). This section is housed inside a 32channel standard NIM module (the ‘‘DISCAL’’module) that processes one TRD tube plane. Inorder to discriminate clusters predominantlycaused by transition radiation from d-rays, thecomparator threshold was set to 80 mV: Thiscorresponded to about 5 keV energy, where thetransition radiation spectrum starts. Signals abovethis threshold within a 1:4 ms time gate werecounted and transferred in the shift register chain.In addition, the OR of these signals (DIGOROUT) was available on the DISCAL module’sfront panel for monitoring and self-triggeringpurposes. The minimum signal width accepted bythis stage was measured to be 12 ns; while thedouble pulse resolution was slightly higher.The same STAS module used by the lower

detector ST system, the CAEN C187, controlledboth the bit transfer and the acquisition procedureby addressing each non-zero bit pattern. Thismodule is able to process eight shift registerchains, each with 1024 bits maximum length. Fora TRD module readout corresponding to 320channels of 3 bits each, only a single STAS chainwas needed. For each event five words were readout in an average time of 10 ms:The TRD trigger was based on standard NIM

logic modules. The 10 DIGOR signals from the 10proportional tube layers for each TRD moduleopened a corresponding acquisition gate. Thisoperation allowed the tube hits to be stored inscalers used as memory buffers. They were readout whenever a muon trigger occurred within10 ms: The observed rate of events crossing all tenTRD planes was about 120 muons per day perTRD module. Since natural radioactivity pro-duced a counting rate of about 1000 Hz per TRDmodule, the probability for an accidental coin-

cidence was 10�2; corresponding to one accidentalper day on the average.

6.3. Performance of the TRD

To investigate the g dependence of the responseof a TRD suitable for MACRO use, we set up atest beam at the CERN-SPS with a maximummomentum of 5 GeV c�1: It consisted of twothreshold Cherenkov counters to distinguish elec-trons from pions, and a 10 radiation-length lead-glass counter to select electrons. At 5 GeV; weobtained a pion contamination of 10�4 in theelectron beam. We built a prototype TRD oflimited size in order to fit the available space in atest beam area. In this set of measurements weadopted the cluster counting electronic readout asdescribed in Section 6.1 above.The cluster count distributions for each setting

could be easily fit by Poisson distributions withmeans of, at most, about 10 counts.In Fig. 30 the mean number of clusters for

Ethafoam are shown for various values of g andbeam crossing angles. The mean number ofclusters obtained with electrons without radiatorsare shown only for 01 to demonstrate therelativistic rise and the Fermi density effect plateauof the energy loss of a highly relativistic particle ina gas layer. We get similar behavior for the other

1

2

3

4

5

6

7

8

9

10

10 10 2 10 3 10 4

θ = 0˚θ=15˚θ=30˚θ=45˚θ= 0˚ No Radiator

γ Lorentz Factor

Ave

rage

Num

ber

of H

its

Fig. 30. Mean number of clusters for 10 cm Ethafoam radiator

versus the Lorentz factor g for various incident angles y: The‘‘no radiator’’ data show the relativistic rise and Fermi density

effect of the ionization energy loss in the gas of the counters.


angle settings, namely a 10% increase in meancluster number beyond g ¼ 1000:

7. The data acquisition system

In the first period of data taking, which ended in1994, the detector was run with an acquisitionsystem based on a network of MicroVAX-IIs andthe CERN-Fisher CAMAC System Crate [54].Subsequently, we upgraded the system in order toinclude VME front-end electronics, used particu-larly in the new WFDs for the scintillatorcounters. Such a system was described in somedetail in Ref. [55]. The new system was based on anetwork (ETHERNET/DECNET) of KAV30VAX processors and MicroVAX-IIs. TheKAV30 is a VME single-board computer basedon the rtVax 300 real-time processor. It includes aCVAX microprocessor, a floating point coproces-sor, and a second generation Ethernet controllerand it allows VAXELN applications to interfacewith VMEbus devices. VAXELN is a DigitalEquipment software product specific for thedevelopment of dedicated, real-time systems forthe VAX family of computers. The general layoutof the new data acquisition system is shown inFig. 31. Three KAV30 processors controlled threegroups of two SMs, while three additional Micro-VAX-IIs acted as the supervisors for the PHRASEtrigger that looked for neutrino events originatingin a stellar collapse. Each VME crate where aKAV30 processor was located acted as a VMEMaster Crate.The VME Master Crate was able to drive

several VME crates by means of the CESVIC8250 VMV master/slave interface on theVMV bus (Vertical Bus). In our system, thedistances between master and slave crates was onthe order of a few meters.The VIC8250, a one-slot VME module, offers a

transparent interconnection of up to 15 VMEcrates on the VMV bus, each crate being identifiedby a crate number. In such a VMV system, allVME busses are autonomous and completelyindependent. When the CPU located in the VMEMaster Crate accessed a slave device in a remote

VME crate, the two were linked only during thecycle and were independent the rest of the time.A remote device was accessed by mapping a

1 Mbyte page of the addressable memory space inthe remote crate onto the VME Master Cratememory space. This was done by writing alongword, the page descriptor, into a specificlocation of the VIC8250 CSR space. The pagedescriptor contained the remote crate number, theVME address of the device in the remote crate,and the address modifier. For example, if wewanted the CPU to access the remote memoryresource through the VME standard addressXXX, the page descriptor had to be written inthe CSR into an address which depended onaddress XXX. Thus the addressing of the remotephysical memory, once mapped, was completelytransparent to the user.With the KAV30 processor, it took about 2:0 ms

to write a word/longword on a VME Master Crateand about 2:1 ms for the same operation on aremote VME crate via VMV bus. The times forread operations were slightly longer.CAMAC was accessed from VME by means of

one or more CES CBD8210 (CAMAC BranchDriver) interfaces, each one controlling a fullCAMAC Branch. The CBD8210 supported both16-bit and 24-bit data transfer. CAMAC crateswere accessed by mapping a CAMAC address fieldonto the 24-bit address field of VME. Two

KAV30# 1

VMEMasterCrate

# 1

VMESuperMod

1, 2

µVax # 1for StellarCollapse

DEC Bridge

LNGSLAN

To/From WAN

VAX4000/500

Ethernet / DECNET

CAMACSuperMod

1, 2

Fig. 31. General layout of the data acquisition system.


Interrupt Request (IRQ) lines with front-panelexternal inputs were also provided. These IRQswere used in our system for event and alarminterrupts respectively. The extension board CESDMC8211 was also used for DMA block transfersto and from CAMAC crates. Two CAMACparallel branches were connected to each VMEMaster Crate. Single word transfers from aCAMAC crate, using the KAV30 and CBD8210,required about 4.1 and 7:0 ms for 16-bit and 24-bitdata respectively.The tasks performed by each front-end proces-

sor were: electronics readout, event filtering(planned but not used), data reduction, apparatusmonitoring, calibration, and data forwarding to acentral VAX4000/500. This central computer,running under the VAX/VMS operating system,was used as a file and network server and as a userinterface to the acquisition system. It also pro-vided the disks for raw data logging. All thecomputers, including the VAX4000/500, wereconnected under the DECNET protocol viaEthernet as the MACRO Local Area Network(LAN) and this, in turn, was connected to thegeneral LNGS LAN by a DEC Bridge. TheVAX4000/500 was also connected to the generalLNGS LAN under the TCP/IP protocol. TheLNGS LAN was composed of segments linked viaoptical fibers running for about 6 km between theunderground laboratory and the external labora-tory where the general computing facilities werelocated. From the external laboratory there wasaccess to the Wide Area Network (TCP/IP andDECNET at the beginning of data acquisition).Utilities were implemented to help in controlling

and debugging the apparatus. Thus, during dataacquisition, a user at any DECNET node couldrequest a copy of the raw data buffer from a givenfront-end processor or from the central VAX, andcould require the execution of CAMAC operationsby using Remote Procedure Calls. These capabil-ities were very important in maintaining theefficiency of the experiment.

7.1. Acquisition system performances

The main advantages of VAXELN relative tothe VAX/VMS operating system have been

summarized elsewhere [54]. Here we only want topoint out that this software operated in a host andtarget system environment.The VAXELN system for the experiment

consisted of six user jobs running with differentpriorities, and four system jobs (the Ethernetdriver, the console driver, etc.). In each job, manyconcurrent sub-processes could run with differentpriorities and, in our system, more than 15 sub-processes ran concurrently. In particular, eachactivated DECNET connection was handled by adedicated sub-process.All the software was written in EPASCAL, the

PASCAL native language of VAXELN. Theexchange of data between jobs was done usingthe VAXELN message facility which was veryuseful for handling queues of buffers.The VME and CAMAC input/output during

the data taking was done using a list of operationsloaded from a file resident in the VAX4000/500.The basic element of the list was the equipment.This term denotes a logical block of VME andCAMAC operations, that is, a sequence of read,write, or test functions to be performed onelectronics modules. Both single word operationsand DMA transfer operations could be executedby inserting the proper instructions into theequipment definition. It was also possible to selecta subset of operations inside any equipment bydefining a section in the equipment. A set ofequipment formed a trigger, that is, the fullsequence of operations which were to be per-formed at the occurrence of a particular eventtrigger. The list also included an initialization partwhich contained the necessary operations (moduleresets, register zeroing, VIC VME mapping, etc.)executed at the beginning of a run.CAMAC could also be accessed using a library

of standard ESONE CAMAC routines and aset of fast VAXELN routines. This library (writtenin EPASCAL) was explicitly developed for thenew system and it required the use of theCBD8210 VME board as interface betweenthe VME Master Crate and the CAMAC branch.The CAMAC routines could be called eitherdirectly inside VAXELN or from a computer inthe DECNET network using remote access rou-tines.


Some I/O performances which we obtainedusing the list instructions in the acquisition taskfor VME and CAMAC access are reported inTable 1. In the same table, some I/O performancesobtained using the ESONE CAMAC routinelibrary are also given.The performances obtained using the remote

access CAMAC routines are listed in Table 2. Inthis case, performance was limited by the Ethernetthroughput and by DECNET overhead. Ofcourse, the throughput increases if many CAMACI/Os are requested in the same call by the blocktransfer routine (CSGA) or by the DMA routine.This kind of access to CAMAC was very useful fordebugging the electronics since it allowed writing

and running simple programs on a VAX/VMShost requiring neither knowledge of VAXELN normonitoring with slow devices.At system saturation, the maximum global data

throughput we obtained on the network, includingthe VAX4000/500, was about 130 kbytes s�1:In conclusion, the performance of our acquisi-

tion system was well suited to the needs of anexperiment of the size of MACRO, particularlybecause of its modularity and its good hardware/software maintainability.

7.2. On-line monitoring

Several hardware and software systems werebuilt for the on-line monitoring of the dataacquisition of the experiment. The Event Displayand the Diagnostic Monitor are presented in thefollowing.

7.2.1. The event display

A specific software tool written for the on-linedrawing of the detector data was the EventDisplay program (EVD). It is based on the ISOstandard graphics package, GKS, and heavily usesGKS’s picture segmentation characteristics, logicalI/O device concepts, and metafile facilities.The most important features available to the

operators were:

* different 3D and 2D projection views of theevents,

* the possibility of looking at different detectorparts with different magnification,

* the possibility of selecting and displaying onlyparticular event configurations,

* interactive tracking for streamer tubes hits.

The user interface of EVD was simple andintuitive: the program activities were seen aslogically independent by the operator and werecontrolled by means of nested menu sequences.Event input data came from different sources: on-line data directly from front-end microVAX’s,data previously stored on disk files, and networkdata. The raw data were encoded bytes whichneeded to be interpreted and processed accordingto the detector type. The event was accepted if itmatched some requested selection criteria and the

Table 1

VME and CAMAC I/O performances in the acquisition task

Using the list

VME word/longword transfer

Single read in a remote memory 4:8 msCAMAC 16-bit data transfer

Single operation (with Q test) 15:8 msSingle operation (without Q test) 11:3 msBlock transfer (without Q test) 4:6 ms

CAMAC 24-bit data transfer

Single operation (with Q test) 21:9 msCAMAC DMA data transfer

1st word 370 msAdditional 16-bit words 3:0 msAdditional 24-bit words 5:4 ms

Using the ESONE routines

Single 16-bit data transfer (CSSA) 42 msBlock 16-bit data transfer (CSGA)

1st word 54 msAdditional words 22 ms

Table 2

CAMAC I/O performances using remote calls from the

VAX4000/500

Single CAMAC I/O using CSSA 5 ms

I/O of 10 CAMAC words using the ESONE list

routine (CSGA)

7 ms

I/O of 10 CAMAC words using the ESONE

DMA routine (CSUBC)

8 ms

I/O of 5000 CAMAC words using the ESONE

DMA routine (CSUBC)

60 ms


picture was sent to the display surface afterapplying the transformation and view optionspreviously chosen by the operator. The picturestructure was built by creating several segmentscontaining the output primitives which representedthe current event data while a set of preselectedsegments defined the detector parts that had to besent to the display (a portion of a module, one ormore modules, etc.). Thus the large use ofsegmentation allowed us to satisfy the require-ments of speed and interactivity. Moreover theevent data spooling facility of VAX/VMS acquisi-tion system tasks made it simple to monitor theon-line acquisition from the remote nodes of thenetwork also.

7.2.2. The diagnostic monitor

With the aim of monitoring the DAQ opera-tions to ensure continuous reliability of the datataking, an automatic DiaGNostic system (DGN)was built. It was able to collect data in parallelwith the official data acquisition and data storagesystem but independently of it. Thus, it couldquickly signal anomalous behavior in the appara-tus. The digitized data provided by the acquisitionelectronics were collected event by event at thetrigger rate, decoded, analyzed and grouped asstatistical histograms, multidimensional maps, orfunctions of time. At the checking level, thecurrent data were compared with a set showingthe standard equipment status: differences pro-duced a list of deviations and possible locations ofthe failures. At the same time the current statuswas condensed into a few graphs of prominentparameters to provide some display warnings tothe operators.From the hardware point of view, DGN was

composed of several VME modules located in adedicated crate. As stated above, the moduleswhich digitized the detector signals were based onthe CAMAC and VME standard. The VMEinterface in the DAQ Master crate mapped theaddresses of the CAMAC registers containing thecurrent data as VME addresses in a predefinedmemory space. So we provided a ‘‘spy’’ card,located in the Master crate, which identified themapped addresses and latched the data present onthe VMEbus during the event readout. These

parasitic cards, one for each acquisition branch,transmitted the captured pairs of 32-bit address–data to memory cards located in the DGN crate.Each memory card contained 512 kbytes of RAMwhich, with suitable pointers, were handled as acircular buffer. An assembler program running ona 68020 CPU scanned the buffers and createdpointers to sequences of relevant data; a secondCPU in the crate, running under OS9, used thesepointers to send the data blocks to the downstreamsoftware. The DGN crate hardware processor alsocontained VME cards to perform on-line trackingof the ST data. Several tasks ran concurrently on aDIGITAL 4000/90 dedicated workstation to per-form the data checking and to provide the displayinterface. A task-to-task Ethernet link at aneffective rate of about 0:2 Mbytes s�1 connectedthe VME CPU program to the workstation taskwhich read, decoded, and stored the event data. Aspy task performed the most significant part of theanalysis event-by-event, while a display taskgraphically presented the data taking status(trigger rates, wire maps, muon rates, hit clusterdistributions, etc.). Moreover a parallel task couldoperate as a remote display server, allowing theexperiment to be controlled by people outside thetunnel. The DGN system could collect and analyzeevents at a rate up to 10 Hz: Data transfer was lessthan 50 ms event�1 and the tasks on the VAXsta-tion required about 60–80 ms to process an event.

7.3. The UTC clock

The timing of events relative to the internationalstandard time scale was performed by the LNGSclock. This is synchronized with the UniversalTime Coordinated, controlled by the United StatesNaval Observatory (UTC/USNO) standard viathe Global Positioning System (GPS) time signals.As shown in Fig. 32, a Master Clock,18 consistingof various components and located in the externallaboratory, drives several peripheral units, calledSlave Clocks, located near the LNGS undergroundexperiments.Each GPS satellite continuously broadcasts

signals containing UTC/USNO time as well as

18ESAT S.r.l - Torino (Italy).


data on its trajectory, health and location. In orderto calculate its own three-dimensional position, theGPS Receiver uses four channels to track the foursatellites with the highest quality signals at anygiven time. After a sufficient number of positionmeasurements are made to determine the receiver’sposition with an uncertainty of a few meters, it canbe switched to a ‘‘time-only’’ mode, which thenuses the predetermined precise position to convertthe time messages received from the healthiestsatellite to precise time pulses, PPS, synchronizedwith UTC/USNO.Since the GPS Receiver unit itself is not directly

accessible to the underground experiments forevent timing purposes, a local time scale isgenerated by the Master Clock and is aligned tothe UTC/USNO scale in the following way.The GPS Receiver provides a time stamp

message for each one-second roll-over signal,PPS, in the form of a serial string containing time,date and satellite status. This is used by aMicroprocessor to generate the local time scale

based on a 10 MHz frequency signal coming froma Rubidium Oscillator. This oscillator has highstability [56] with a fractional frequency long-termdrift less than 10�11 month�1 and a short-termstability (Allan variance) of less than 3�10�12=100 s:Each second, the Time Interval unit (20 ns

resolution) measures the time difference betweenthe PPS signal from the UTC/USNO and thecorresponding Local PPS signal from the localscale. Sixty of these time differences, correspond-ing to 1 min; are averaged to yield the quantity Dt

whose distribution is shown in Fig. 33. The sourcesof these time drifts can be grouped in two classes:

* ‘‘short term’’ fluctuations, due to intrinsic noiseon the time signal (transmitter, receiver, andsignal propagation instabilities) and/or to the‘‘Selective Availability’’, which is an artificialnoise introduced by USNO in order to preventcivilian users from attaining the system’s max-imum position accuracy;

Antenna

Master Clock

Ethernet display keyboard

UTC Time Base ns Shifter

Laser TX

to Rb Slave Oscillator Clocks

start

Time Interval

stop

Microprocessor

GPSReceiver

x5

RS232 I/O

RS232 out

10 MHz50 MHz

UTC PPS

LocalPPS

Fig. 32. Master Clock.


* ‘‘long term’’ deviations due to frequency driftsof the local oscillator relative to UTC, comingfrom: (a) typical Rubidium Oscillator agingeffects which can cause time drifts of the orderof 1 ms per day, and (b) environmental condi-tions, mainly temperature variations, whichbecome apparent within a few hours and causedeviations of the same order of magnitude asthe ones discussed above.

The short term fluctuations, Dt; shown as thedotted curve in Fig. 33, normally did not exceed100 ns; but we occasionally observed Dt values asbig as few hundred nanoseconds, generally in-dicating non-optimal quality of the GPS signal.Long term frequency drift can be detected by

making a linear least squares fit of Dt to time itself.A slope significantly different from zero in this fitis an indication of long term drift in the localoscillator. It was shown that 180 measurements ofthe Dt values, corresponding to a time interval of3 h; minimized the frequency deviations due to thecombination of the two long term drift effects. Thetime residual error (the difference between localtime and UTC time) was calculated each 60 s onthe basis of the last 180 data points, and, whennecessary, the ns Shifter made smooth corrections,properly shifting the local clock time scale by stepsof 1 ns:

In Fig. 34, the time residual errors, measuredeach 60 s resulting from the above procedure, arecompared with the Dt of the short term fluctua-tions. Distributions of each are shown in Fig. 33.The typicalB12 ns standard deviation of the timeresidual error distribution is a measure of themagnitude of the instantaneous Master Clock timeerror.This procedure provided excellent traceability of

the local time scale to the UTC/USNO scale,which in turn was normally aligned within lessthan 100 ns to the international standard UTCscale.

7.3.1. Slave clocks

Each millisecond, the time/date information anda synchronization pulse were sent by the MasterClock through an 8 km long mono-modal1310 nm wavelength optical fiber cable to a setof six Slave Clocks. Each Slave Clock corrected forthe signal transit time and reproduced the UTC/USNO scale with a resolution of 0:1 ms: Eventtrigger signals coming from the different sectionsof the experimental apparatus latched these clocksand standard time and date (80 bits in total) werethen recorded for each event.

8. Off-line data analysis

The offline analysis program Data Reductionand Event Analysis for MACRO (DREAM) isbriefly described in I. It was used throughout the

3000

2500

2000

1500

1000

500

0

Cou

nts

-250 -200 -150 -100 -50 0 50 100 150 200 250Time deviations (ns)

Fig. 33. The solid curve, with a mean of 0:33 ns and standarddeviation of 11:5 ns; shows a typical distribution of time

residual errors after all corrections have been applied. The

dotted histogram, with a mean of 1:9 ns and standard deviationof 44 ns; is a typical example of the differences, Dt; between theUTC/USNO PPS signals and the Local PPS signals, averaged

over 1 min:

20 40 60 80 100 120 140 160 1800Time (min)

100

75

50

25

0

-25

-50

-75

-100

Tim

e de

viat

ion

(ns)

Fig. 34. The smooth curve shows the time residual error

calculated every minute compared to the corresponding Dt

values. The slope of the smooth curve in this sample is

�4:9 ns=h:


life of the experiment as the primary off-line dataanalysis tool, having been continually upgradedboth technically and functionally. Initially writtenon a VAX/VMS platform, it was subsequentlyported to IBM/VM, IBM, DIGITAL, HP UNIXand Linux machines.The data files produced by the acquisition

system were transferred with a delay of B1 dayto a VAX/VMS computer in the external labora-tory and immediately processed to produce DSTfiles for neutrino and cosmic ray muon analyses.DST production for different analyses (for in-stance, monopole analysis) followed a differentstream. Those files contain some raw quantities(ADC and TDC values) as well as reconstructedquantities as, for example, coordinates of thestreamer tube hits, time and energy deposition ofthe scintillator counters, and parameters of thetracks in both detector systems.DST files were written in VAX native mode.

They contain only integer quantities and can beeasily processed on different platforms. Reading ofDST files can be performed by the same analysisprogram DREAM; in this way a reprocessing ofthe raw data (for example with different calibra-tion constants) can be performed.A procedure that ran on a dedicated ALPHA

machine fully automated the production of DSTfiles. All DSTs produced after the completion of thedetector are stored on disk (total of 90Mblocks ¼ 46 Gbytes), while the older ones arestored on magnetic tapes. Both the procedure andthe monitoring of the DST production were devel-oped and maintained locally at the experiment site.

9. Conclusions

The MACRO detector was proposed in 1984; itwas installed in Hall B of the underground GranSasso Lab in a modular way starting in October1987. It started data taking in February 1989 withthe first lower supermodule, and in 1995 with thecomplete detector. MACRO completed its pro-gram of data taking in December 2000.During the data taking period the detector

performed very reliably, and it did not have anymajor difficulty. The fraction of time the detector

was on, the percentage lifetime, was very high andreached nearly 99% for the gravitational stellarcollapse system.During its lifetime, MACRO collected more

than 60 million high energy downgoing muons,which allowed continuous checks of the apparatusas well as yielding interesting physics and astro-physics results.MACRO obtained and published results on all

the fields proposed and discussed in the 1984proposal. Of particular relevance are:

* the new limits for GUT magnetic monopoles inthe cosmic radiation,

* the results on atmospheric neutrinos and ontheir oscillations based on 3 event categoriesextending from a few GeV up to hundreds ofGeV,

* the results on cosmic ray physics based on highenergy downgoing muons and coincidenceswith the EASTOP detector located at CampoImperatore.

Besides these results, many technical papers werepublished on the scintillator and streamer tubesystems, on their monitoring, on the nuclear trackand transition radiation detectors. Many analysesare still in progress including some new ones.Furthermore our analyses have stimulated greatimprovements in the Monte Carlo calculationsdealing with high energy cosmic rays.MACRO results have been published in 33

papers in international refereed scientific journals,in addition to many papers and invited talks atmany international and national conferences. Thescientific and technical results were used for a verylarge number of internal memos and for over 140Ph.D. and undergraduate theses in Italy, US andMorocco (for a complete list available on thewww, see Ref. [57]).MACRO pioneered the new generation of large

underground detectors for astroparticle physicsand helped to develop the new Gran Sasso Lab.

Acknowledgements

We gratefully acknowledge the support of thedirectors and of the staff of the Laboratori


Nazionali del Gran Sasso and the invaluableassistance of the technical staff of the institutionsparticipating in the experiment. We thank all ourrun coordinators. We thank the Istituto Nazionaledi Fisica Nucleare (INFN), the US Department ofEnergy and the US National Science Foundationfor their generous support of the MACROexperiment. We thank INFN, ICTP (Trieste),NATO and World Lab for providing fellowshipsand grants for non-Italian citizens.

References

[1] S. Ahlen, et al., Phys. Rev. Lett. 72 (1994) 608.

[2] M. Ambrosio, et al., Astropart. Phys. 6 (1997) 113.

[3] M. Ambrosio, et al., Phys. Lett. B 406 (1997) 249.


[5] R. Liu, A search for fast moving magnetic monopoles with

the MACRO detector, Caltech Dissertation, 1995.




[9] F. Ronga, et al., Proc. XXV ICRC, Durban, South Africa,

Vol. 7, p. 45;

M. Ambrosio, et al., Astrophys. J. 546 (2001) 1038.

[10] S. Ahlen, et al., Astrophys. J. 412 (1993) 301.

[11] S. Ahlen, et al., Phys. Lett. B 249 (1990) 149.

[12] M. Ambrosio, et al., Phys. Rev. D 60 (1999) 082002.

[13] S. Ahlen, et al., Astropart. Phys. 1 (1992) 11.


[15] R. Bellotti, M. Aglietta, et al., Phys. Rev. D 42 (1990)

1396;

S. Ahlen, et al., Nucl. Phys. B 370 (1992) 432;

S. Ahlen, et al., Phys. Rev. D 46 (1992) 895;

S. Ahlen, et al., Phys. Rev. D 46 (1992) 4836;

M. Ambrosio, et al., Phys. Rev. D 50 (1994) 3046;

M. Aglietta, et al., Phys. Lett. B 337 (1994) 376;





M. Ambrosio, et al., Astropart. Phys. 10 (1999) 11;


M. Ambrosio, et al., Astropart. Phys. 9 (1998) 105.


[17] C.W. Walter, A search for lightly ionizing particles with

the MACRO Detector, Caltech Dissertation, 1997.


[19] S. Ahlen, et al., Phys. Rev. Lett. 69 (1992) 1860;

M. Ambrosio, et al., Eur. Phys. J. C13 (2000) 453.

[20] S. Ahlen, et al., Nucl. Instr. and Meth. A 324 (1993)

337.

[21] C. De Marzo, et al., Nuovo Cimento 9C (1986) 281;

M. Calicchio, et al., Nucl. Instr. and Meth. Phys. Res. A

264 (1988) 18.

[22] G. Battistoni, Nucl. Instr. and Meth. A 279 (1989) 137.

[23] G. Battistoni, et al., Nucl. Instr. and Meth. 217 (1983) 429;

J. Allison, et al., Nucl. Instr. and Meth. 201 (1982) 341;

C. Becker, et al., Siegen University preprint SI-82-1, 1982.

[24] S.D. Drell, et al., Phys. Rev. Lett. 50 (1983) 644;

V. Patera, Phys. Lett. A 137 (1989) 259.

[25] M. Anelli, et al., Nucl. Instr. and Meth. A 289 (1990) 294.

[26] G. Battistoni, et al., Nucl. Instr. and Meth. 202 (1982) 459

and reference therein.

[27] G. Battistoni, Proceedings of the Second Workshop on

Gas Sampling Calorimetry, Fermilab, 1986, p. 594.

[28] J. Reynoldson, Tesi di Laurea, Univ. dell’Aquila, June

1996.

[29] CAEN Technical Information Manuals for Models C267

and SY408.

[30] S.P. Ahlen, in: Magnetic Monopole, NATO ASI series,

series B: phys., Vol. 2.

[31] G. Battistoni, et al., Nucl. Instr. and Meth. A 270 (1988)

185;

G. Battistoni, et al., Nucl. Instr. and Meth. A 401 (1997)

309.

[32] G. Battistoni, et al., Nucl. Instr. and Meth. A 399 (1997)

244.

[33] G. Auriemma, et al., Nucl. Instr. and Meth. A 263 (1988)

249.

[34] V.A. Rubakov, JETP Lett. 33 (1981) 644;

C.G. Callan Jr., Phys. Rev. D 26 (1982) 2058.


[36] G.W. Clark, Phys. Rev. 108 (1957) 450.

[37] D.E. Alexandreas, et al., Phys. Rev. D 43 (1991) 1735.

[38] M. Amenomori, et al., Phys. Rev. D 47 (1993) 2675.

[39] M. Aglietta, et al., Astropart. Phys. 3 (1995) 1.


[41] Nucl. Instr. and Meth. A 321 (1992) 609.

[42] P.H. Barrett, et al., Rev. Mod. Phys. 24 (1952) 133.

[43] R. Bellotti, et al., 21st ICRC, Adelaide, Vol. 10, 1990, p.

252;;

S. Ahlen, et al., Nucl. Tracks Radiat. Meas. 19 (1991) 641;

S. Cecchini, et al., Nucl. Tracks Radiat. Meas. 22 (1993)

555.

[44] S. Cecchini, et al., Nuovo Cimento 109A (1996) 1119.

[45] G. Giacomelli, et al., Nucl. Instr. and Meth. A 411 (1998)

41.

[46] S. Cecchini, et al., Proceedings of the 20th International

Conference on Nuclear Tracks in Solids, Portoroz,

Slovenia, August 2000.

[47] A. Noll, et al., Nucl. Tracks Radiat. Meas. 15 (1988)

265.

[48] J. Derkaoui, et al., Astropart. Phys. 10 (1999) 339.

[49] B. Dolgoshein, Nucl. Instr. and Meth. A 326 (1993) 434.

[50] K.K. Tang, Astrophys. J. 278 (1984) 881.

[51] E. Barbarito, et al., Nucl. Instr. and Meth. A 313 (1992)

295.

[52] L. Bergamasco, et al., Nuovo Cimento C6 (1983) 569.


[53] E. Barbarito, et al., Nucl. Instr. and Meth. A 365 (1995)

214.

[54] I. D’Antone, et al., IEEE, Trans. Nucl. Sci. NS-36 (1989)

1602.

[55] I. D’Antone, et al., IEEE, Trans. Nucl. Sci. NS-41 (1994)

83.

[56] Technical specifications for Ball-Efratom (Irvine, Califor-

nia) Model MRT-H or MRK-S.

[57] http://www.df.unibo.it/macro/pub1.htm.

Glossary

ADC: Analog to Digital Converter.ASIC: Application Specific Integration Circuit.Attico: The upper part of MACRO.CAEN: A company which supplies electronicsused in particle physics.

CAMAC: An electronics packaging and signal sizestandard.

CES: Name of an electronics company.CR39: A type of plastic useful for track-etchparticle detection.

CSGA: A CAMAC routine for block transfers.CSR: Special memory in the VIC8250 for pagedescriptor information.

CVAX: A microprocessor in the KAV30 single-board computer.

DIGOR: Digital OR.DISCAL: The digital electronics module used bythe TRD system.

DIP: Dual Inline Package.DMA: Direct Memory Access.DSM: Double Super Module.ERP: Energy Reconstruction Processor.ESONE: European Studies on Norms for Electro-nics; an electronics standard.

Ethernet: A protocol for data exchange betweencomputers.

EVD: Event Display program.FADC: Flash ADC.FIFO: First In First Out.FMT: Fast Monopole Trigger.

FPT: Fast Particle Trigger.FWHM: Full Width at Half Max.GPS: Global Positioning System.HCT: High Charge Trigger.HIPT: Highly Ionizing Particle Trigger.HV: High Voltage.KAV30: A Digital Equipment Corp microcompu-ter.

LAN: Local Area Network.LeCroy: A company supplying electronics forparticle physics.

LI: Leaky Integrator.LIP: Lightly Ionizing Particles.LNGS: The Gran Sasso National Laboratory.MACRO: Monopole, Astrophysics, and CosmicRay Observatory.

NIM: An electronics packaging and signal sizestandard.

PHRASE: Pulse Height Recorder and Synchro-nous Encoder.

PMT: Photomultiplier Tube.PPS: Precise Time Pulses from the GPS receiver.QTP: Charge and Time Processor.RAM: Random Access Memory.REL: Restricted Energy Loss.SM: Super Module.SMT: Slow Monopole Trigger.ST: Streamer Tube(s).STAS: Streamer Tube Acquisition System.TDC: Time to Digital Converter.TOF: Time of Flight.TOHM: Time-Over-Half-Maximum.TRD: Transition Radiation Detector.VAX: A Digital Equipment Corp computerproduct.

VAXELN: Real-time software system.VME: An electronics hardware/software standard.VMV: A bus protocol.WFD: Wave Form Digitizer(s).WIMP: Weakly Interacting Massive Particle.WSM: WFD Stop Master.


http://www.df.unibo.it/macro/pub1.htm

Documents

TheMACROdetectoratGranSassokats/publications/pub32_nucl_instrum... · modules(overalldimensions6 6 2m3),made of10cmthickpolyethylenefoamradiatorsand proportionalcounters.EachTRDcountermea-sured6