high-resolution tracking using large capillary bundles with liquid...

*Corresponding author.E-mail: [email protected] (G. Penso).1Now at DESY, Hamburg, Germany.2Now at Johannes Gutenberg-UniversitaK t, Mainz, Germany.

3Now at TaperVision, Inc., Pomfret, CT, USA.4Now at Institute of Physics, Academia Sinica, Taipei,

Taiwan, Republic of China.

Nuclear Instruments and Methods in Physics Research A 449 (2000) 60}80

High-resolution tracking using large capillary bundles "lledwith liquid scintillator

P. Annis!, A. Bay", L. Benussi!, N. Bruski#, S. Buontempo$, C. Currat",N. D'Ambrosio#, R. van Dantzig%, J. Dupraz&, A. Ereditato$, J.P. Fabre&, V. Fanti!,

J. Feyt&, D. Frekers#, A. Frenkel', F. Galeazzi', F. Garu"$, J. Goldberg),S.V. Golovkin*, A.M. Gorin*, G. GreH goire+, K. Harrison', K. Hoepfner),1,K. Holtz#,2, J. Konijn%, E.N. Kozarenko,, I.E. Kreslo', A.E. Kushnirenko*,B. Liberti', G. Martellotti', A.M. Medvedkov*, L. Michel+, P. Migliozzi$,

C. Mommaert!, M.R. Mondardini-, J. Panman&, G. Penso',*, Y.P. Petukhov,,D. Rondeshagen#, W.P. Siegmund.,3, V. Tyukov,, G. Van Beek!,

V.G. Vasil'chenko*, P. Vilain!, J.L. Visschers%, G. Wilquet!, K. Winter-,T. Wol!#, H.J. WoK rtche#, H. Wong&,4, K.V. Zimyn*

!IIHE, ULB-VUB, Bruxelles, Belgium"Universite& de Lausanne, Lausanne, Switzerland

#Westfa( lische Wilhelms-Universita( t, Mu( nster, Germany$Universita% **Federico II++ and INFN, Napoli, Italy

%NIKHEF, Amsterdam, The Netherlands&CERN, Gene%ve, Switzerland

'Universita% **La Sapienza++ and INFN, Roma, Italy)TECHNION, Haifa, Israel*IHEP, Protvino, Russia

+Universite& Catholique de Louvain, Louvain-La-Neuve, Belgium,JINR, Dubna, Russia

-Humboldt-Universita( t, Berlin, Germany.Schott Fiber Optics, Inc., Southbridge, USA

Received 22 September 1999; accepted 17 November 1999

Abstract

We have developed large high-resolution tracking detectors based on glass capillaries "lled with organic liquidscintillator of high refractive index. These liquid-core scintillating optical "bres act simultaneously as detectors ofcharged particles and as image guides. Track images projected onto the readout end of a capillary bundle are visualizedby an optoelectronic chain consisting of a set of image-intensi"er tubes followed by a photosensitive CCD or by

0168-9002/00/$ - see front matter ( 2000 Elsevier Science B.V. All rights reserved.PII: S 0 1 6 8 - 9 0 0 2 ( 9 9 ) 0 1 2 9 5 - 4

5The studies of glass and plastic micro"bre bundles werecarried out by a part of the present authors, in the framework ofthe WA84 experiment at CERN.

an EBCCD camera. Two prototype detectors, each composed of +106 capillaries with 20}25 lm diameter and0.9}1.8 m length, have been tested, and a spatial resolution of the order of 20}40 lm has been attained. A highscintillation e$ciency and a large light-attenuation length, in excess of 3 m, was achieved through special puri"cation ofthe liquid scintillator. Along the tracks of minimum-ionizing particles, the hit densities obtained were &8 hits/mm at thereadout window, and &3 hits/mm at &1 m away. The level of radiation resistance of the prototype detectors is at leastan order of magnitude higher than that of other tracking devices of comparable performance. ( 2000 Elsevier ScienceB.V. All rights reserved.

PACS: 29.40.Gx; 29.40.Mc; 42.79.Ls; 85.60.Gz

Keywords: Tracking and position-sensitive detectors; Scintillation detectors; Scintillating "bres; Optoelectronics; Image intensi"ers;CCD

1. Introduction

Tracking devices with a spatial resolution in therange of microns are of great interest in high-energyphysics as microvertex detectors for the study ofinteraction vertex topologies, for example decaypatterns of extremely short-lived particles like q-leptons or charmed and beauty hadrons. Followingsome attempts to use optical "bres based on scintil-lating glass or plastic [1}13], glass capillaries "lledwith liquid scintillator (LS) have become increas-ingly attractive when considering the variousexperimental requirements. Over the past fewyears, many groups have contributed to the devel-opment of the capillary technique [14}26] forhigh-quality track imaging, making signi"cant pro-gress in many areas. Coherent arrays of thin glasscapillaries (diameter of 20}30 lm) currently allowvisualization of minimum-ionizing particle trajec-tories with several hits per mm of track length, andwith a spatial resolution of the order of 10 lm. Thee$cient light transmission of liquid-core "bresmakes possible the construction of large-volumedetectors. The real-time readout is performed by anoptoelectronic chain, consisting of Image Intensi"-ers (IIs) coupled either to a traditional photo-sensitive Charge-Coupled Device (CCD) or toa newly developed Electron-Bombarded CCD(EBCCD) camera. These components lie in themainstream of the industrial interest for low-light-level imaging technology, and their quality andperformance parameters are continuously improv-ing. Further, the high level of parallelism providedby the optoelectronic readout makes large de-tectors cost e!ective.

In this paper we "rst give a short overview of ourwork on scintillating micro"bres over the past 10years, and we compare the main features of glassand plastic "bres5 with those of capillaries "lledwith LS. We then report on our recent progressusing the capillary technique: bundle manufactur-ing, bundle "lling with LS, studies of light yield,light attenuation and spatial resolution with largecapillary bundles. The readout methods and thedata analyses are described. The tracking capabili-ties of the new devices have been tested in a pilotexperiment, installed upstream of the CHORUSdetector [27] in the CERN wide-band neutrinobeam. Results from this pilot experiment are alsopresented.

Our work has been carried out in the frameworkof the CERN Research and Development projectRD46 [28], the main goal of which has been tostudy the capabilities of capillary-based detectorsfor high-precision tracking and vertex reconstruc-tion in high-energy physics experiments.

2. From glass and plastic 5bres to glass capillaries

In the last 10 years we have performed extensivestudies of scintillating-micro"bre detectors basedon glass "bres, on plastic "bres, and on glass capil-laries "lled with LS.

The glass-"bre detector that we tested [5}7] con-sisted of a 4]4]40 mm3 bundle of parallel "bres.

P. Annis et al. / Nuclear Instruments and Methods in Physics Research A 449 (2000) 60}80 61

Table 1Characteristics of detectors based on scintillating micro"bres with diameters of 20}30 lm. The hit density, which is strongly dependenton the sensitivity of the readout chain, has not been reported

Glass "bres Plastic "bres Glass capillaries(GS1) (polystyrene with ("lled with LS)

SCSN-81T or PMP)

Attenuation length (cm) 2 9}31 90}300Decay time 0.07}350 ls &3 ns 6}8 nsOptical crosstalk Medium High LowRadiation hardness (Mrad) 0.5}5 0.2}2 &100

Each "bre had a 25]25 lm2 core of GS1, a glassdoped with cerium oxide (Ce

2O

3). The cladding of

each "bre was made of clear glass, 2 lm thick.The plastic-"bre detectors [6,7,10}12] consisted

of cylindrical bundles of hexagonal "bres. The di-mensions of the bundles and of each "bre weresimilar to those of the glass-"bre detector. The "brecores were either SCSN-81T scintillator or polysty-rene doped with 1-phenyl-3-mesityl-2-pyrazoline(PMP). In both cases, the "bre cladding was poly-methyl methacrylate, 3 lm thick.

The detectors based on glass capillaries "lled withliquid scintillator are the last type we have studied.Results on this technique have been published bothfor bundles of small cross-section [15}20] and, morerecently, for large-volume bundles [21}25]. The lat-ter have transverse dimensions of 20}40 mm andlengths of 90}180 cm, while the diameter of indi-vidual capillaries is in the range 20}30 lm. Thecapillaries are "lled with newly developed liquidscintillators [26], based on 1-methylnaphthalenedoped with a pyrazoline derivative.

In comparing the performance of glass "bres,plastic "bres and glass capillaries "lled with LS weconsider "ve main characteristics, discussed belowand reported in Table 1.

f Light yield: The light yield is measured as thenumber of photons emerging at the readout endof the bundle when a minimum-ionizing particlecrosses the bundle perpendicularly and close tothe readout window. Taking into account theemission spectrum of the scintillator and thespectral sensitivity of the "rst photocathode ofthe readout chain, this characteristic is usuallyexpressed in terms of the number of hits per unit

track length, each hit corresponding to a photo-electron emitted by the photocathode of the "rstimage intensi"er of the readout chain. Thisquantity is evaluated by extrapolating to d"0the hit density measured on tracks crossing thebundle at di!erent distances d from the readoutend. In our early tests of a detector based on GS1glass "bres, a hit density of &2.1 hits/mm wasmeasured, whilst for a bundle of plastic "bres thecorresponding value was &1.8 hits/mm [7].More recently, a hit density of &8 hits/mm hasbeen obtained [21}23,25] for glass capillaries"lled with the new LS [26]. The much higher hitdensity observed in this latter case is due partlyto a better scintillation e$ciency, and partly tothe use of an improved readout chain. Measure-ments for 60 lm plastic "bres by another group[13] have yielded hit densities that, at shortdistances, are consistent with those given here forthe glass capillaries "lled with LS.

f Attenuation length of light in bundles: The lightoutput depends on the distance (d) betweena particle's crossing point and the readout end ofthe bundle. For bundles a few centimetres long,such as the bundles of glass and plastic "bres,this dependence is nearly exponential. The at-tenuation lengths measured [7] are &2 cm forGS1 glass "bres, &9 cm for SCSN-81T plastic"bre and &31 cm for PMP-doped polystyrene"bres. For long (1}2 m) glass capillaries "lledwith LS, the dependence of the light output ond can be approximately described by the sum oftwo exponentials. The di!erent LSs tested havean attenuation length j(d) in the range 30}70 cmfor smaller d values, and have j5300 cm whend is large [21}23,25].

62 P. Annis et al. / Nuclear Instruments and Methods in Physics Research A 449 (2000) 60}80

6We have tested di!erent capillary bundles produced by threecompanies: Schott Fiber Optics Inc., Southbridge, MA 01550-1960, USA; TaperVision Inc., Pomfret, CT 06258, USA; Geo-sphaera Research Centre, Moscow 117261, Russia.

f Decay time of the scintillation light: For manyscintillators the time dependence of the lightemission is non-exponential, due to the presenceof several components in the emission process. Ifslow components are present, the scintillatormay be unsuitable for high-rate experiments.This is the case for GS1 glass, which, in additionto a main fast (&70 ns) component, has a largetail up to 1 ms [29]. For plastic and liquidscintillators slow components are negligible,their decay times being of the order of a fewnanoseconds [7,30].

f Optical crosstalk: Light that, at a bundle'sreadout end, emerges from a "bre di!erent fromthe "bre in which it was produced by a traversingparticle, contributes to the image noise. Thiscrosstalk between "bres can be strongly reducedby surrounding each "bre with a light-absorbingmaterial known as an Extra-Mural Absorber(EMA). For the "bre diameter we are consider-ing, the EMA thickness cannot be much morethan &1 lm, otherwise the packing fraction ofthe bundle (i.e. the ratio between the useful scin-tillating volume and the bundle's total volume)becomes too small. The problem is to "nd anEMA material that has good light absorption ina thickness of the order of 1 lm and meets themechanical, thermal and chemical propertiesneeded for the bundle manufacturing process.This problem has been solved for glass "bres, butnot for plastic "bres. For capillary bundles, thecrosstalk is strongly reduced by "lling the inter-stices between capillaries with black glass.

In addition to the optical crosstalk, imagenoise can come from the image-intensi"er chainand from the CCD readout system [31].

f Radiation hardness: The radiation hardness of thescintillating "bres is an important characteristicwhen the tracking devices have to operate ina high-activity environment. The absorbed dosecan produce both a lower scintillation e$ciencyand increased light attenuation along the "bres.The overall e!ect on light output depends on theradiation hardness of the scintillator and on thelength of the bundles. For the glass and plastic"bres considered here, the light yield is signi"-cantly reduced [32}36] for an absorbed dose ofthe order of a few Mrad. For the large capillary

bundles "lled with the new LS, the radiationhardness is much higher. Depending on the par-ticular LS used, a dose of the order of&100 Mrad can be absorbed with only a lim-ited reduction in light yield [37].

3. Manufacture of capillary bundles

The manufacture of a large capillary bundle in-volves the solution of many technical problems,and only a few commercial companies6 are current-ly able to build capillary-based detectors. In thefollowing we refer to two large capillary bundlesproduced by Schott, each bundle containing&(0.5!1.4)]106 capillaries with an inner dia-meter of 20}25 lm.

The choice of the type of glass for bundle manu-facture is based on several requirements. First, it isnecessary to have a low refractive index, so as toachieve the highest numerical aperture for theliquid-core "bres. Secondly, to reduce losses due tonon-perfect total re#ection of light, the inner surfa-ces of glass capillaries must be smooth. Finally,good chemical stability is needed. All requirementsare met by the Schott Technical Glass Type 8250,a borosilicate glass normally used for electron-tubeenvelopes. It has a refractive index of 1.49, an ex-pansion coe$cient of 5.0]10~6 K~1, a density of2.28 g/cm3, a softening point of 7203C, and class-3chemical stability.

The two large bundles that we have tested havebeen manufactured with di!erent processes. Forthe "rst bundle, a traditional procedure was ad-opted. Beginning with a preform consisting ofa single glass tube, a multichannel array is obtainedthrough a series of drawing steps alternating withassembly steps. Speci"cally, the manufacturing pro-cess commenced with an 8250-glass tube of 43 mmouter diameter being drawn to a diameter of 1.84mm. These monocapillaries were stacked to forman hexagonal array of 547 pieces and were placedinside a tube of 8250 glass. Black glass rods were


Fig. 1. Schematic view of the tower used in the "nal drawingstage of the "rst manufacturing method described in the text.

7A packing fraction of &80% was achieved for a smallbundle manufactured by Geosphaera.

inserted in the interstices between the monocapilla-ries to act as EMA. The assembly was then drawnto give an hexagonal `multia measuring 1.7 mmbetween opposite #at sides. Finally, 940 multieswere gathered in a close-packed pattern, and wereinserted into a square glass tube, constituting theouter part of the bundle. This assembly,a `multi}multia, was drawn to the desired "naldimensions. In Fig. 1 we show a schematic view ofthe tower used for this last drawing step.

The bundle obtained with this manufacturingprocess (hereafter referred to as the `squarebundlea), was drawn to a "nal cross-section of2]2 cm2 along the "rst 165 cm of its length, whilein the last 20 cm, near the bundle's readout end, thecross-section increased smoothly from 2]2 to3.4]3.4 cm2. In the bundle's narrower part, whichcorresponded to the particle-detector proper, the

multies (Fig. 2) had a transverse dimension of&700 lm, measured between opposite #at sides,while the inner and outer diameters of the capillar-ies were respectively&20 and 25 lm. The bundle'stapered readout end acted as an image guide, andmagni"ed the detected track images by a factorof 1.7. This magni"cation reduced the e!ect ofthe image-intensi"er chain on the overall spatialresolution.

The packing fraction of a bundle obtained withthis "rst technique is about 55%. It is limited bya closing-up of capillaries that occurs in the "naldrawing step, and which is due to the surface ten-sion of the glass.

In order to obtain a higher packing fraction, thesecond bundle that we have tested was manufac-tured with a new procedure. The "rst two steps ofthe procedure were similar to the correspondingsteps of the traditional procedure except that eachmulti was drawn to the "nal bundle length. Theends of each multi were then quickly sealed usinga #ame, taking care not to heat the gas inside themulties. The multies were stacked in an hexagonalglass tube, which eventually formed the outer partof the bundle. The resulting assembly was insertedin a stainless-steel tube, and the remaining spaceinside this tube was "lled with sand. The entiresystem was placed in an horizontal oven, and washeated for several hours following a controlledthermal cycle. During this last step, air inside thecapillaries expanded and forced the multies againstone another, so reducing the dead spaces and secur-ing the multies in the outer glass tube. With this`expansion methoda we obtained a bundle (here-after referred to as the `hexagonal bundlea) thatwas &90 cm long, and had an hexagonal cross-section with a transverse dimension of &40 mm,measured between opposite #at sides. The innerand outer diameters of the capillaries were respec-tively &25 and 30 lm, and the packing fractionwas & 67%,7 signi"cantly larger than that of the"rst bundle.

In Fig. 3 we show photographs taken througha microscope of capillary arrays manufactured withthe two abovementioned procedures.


Fig. 2. Schematic view of (a) a multi and (b) the way in which multies are assembled in the "nal bundle. A microphotograph of a sectionof a typical capillary bundle is shown in (c). The distance between opposite #at sides of a multi is &0.7 (1.4) mm for the bundlemanufactured with the "rst (second) process described in the text.

Fig. 3. Microphotograph of capillary bundles manufactured with the "rst process (left) and with the second process (right), as describedin the text. EMA is inserted in the interstices between capillaries.

Once a bundle was fully assembled, its ends werecut using a "ne-grain diamond saw, under a #ux ofcooling liquid. To prevent small glass particles andcooling liquid entering into the capillaries duringthe cutting, a nitrogen #ux was blown into thebundle through the opposite end. This cutting stageproduced a planar surface suitable for coupling the

bundles to the "rst image intensi"er of the optoelec-tronic readout chain.

The bundles obtained with the two manufactur-ing methods had transverse widths that varied by&$5% along the bundle length. In Fig. 4 weshow the behaviour of this width as a function ofthe distance from the readout end, for the narrow


Fig. 4. Transverse width (distance between opposite #at sides)measured for (a) square bundle and (b) hexagonal bundle, asa function of the distance d from the readout end. The curves aredrawn to guide the eye.

8R6, R39, R45 and 3M-15 are trademarks of the GeosphaeraResearch Centre.

Fig. 5. Absorption and emission spectra of 1MN#R6.

Fig. 6. Light yield as a function of dye concentration, for a scin-tillator based on 1MN doped with R6. A similar behaviour hasbeen observed for all of the scintillators based on 1MN or IPN.

part of the square bundle and for the hexagonalbundle.

In high-energy physics applications, a relevantcharacteristic of a particle detector is its radiationlength. For the quoted packing fraction, and fora LS based on 1-methylnaphthalene [26], the radi-ation lengths of the square bundle and of the hexa-gonal bundle were, respectively, 19 and 24 cm.

4. Liquid scintillators

In the past few years, many new LSs have beenspecially developed [26] with a view to their use intracking detectors based on glass capillaries. Thebest of them [38] are based on 1-methylnaphtha-lene (1MN) or isopropylnaphthalene (IPN) dopedwith pyrazoline derivatives such as8 R6 , R39, R45or 3M-15.

The principal characteristics of these LSs, andthe bundle "lling procedure, are discussed below.

f Emission spectrum and light yield: For the scintil-lators named, the emission spectra peak in the

green region (490}500 nm), as in the example ofFig. 5, and the scintillation e$ciencies are com-parable with, or better than, those of standardplastic scintillators. The light yield depends onthe concentration, C, of the dye in the solvent.Following an initial increase in light yield with C,a maximum is reached around C"2!3 g/l,then at higher concentrations there is a plateauor a slight decrease [26]. As an example,Fig. 6 shows the scintillation e$ciency of 1MNdoped with R6 as a function of the dye concen-tration. A similar behaviour has been observed[26] for all of the LSs based on 1MN or IPN.


Fig. 7. Emission time of liquid scintillator based on 1MN dopedwith R6. The measurements have been performed using thesingle-photoelectron delayed-coincidence counting method. Thedashed line indicates the average rate of accidentals. Similarresults have been obtained for all scintillators based on 1MN orIPN, and doped with R6, R39, R45 or 3M-15.

9Bicron, Newbury, OH 44065, USA.

f Local light emission: Most of the light emissionoccurs within a few microns of the track of anionizing particle, i.e. within a "bre that the par-ticle crosses. This feature is important in allow-ing advantage to be taken of the potentially highspatial resolution of a micro-capillary array. Inorder to obtain a local light emission, non-radi-ative processes must dominate the energy trans-fer from solvent to dye [26,39]. The probability(g) of non-radiative energy transfer increaseswith dye concentration according to the expres-sion [40] g"C/(C#C

0). For the LSs con-

sidered C0

ranges between 1.0 and 1.6 g/l, anddye concentrations larger than these values arenecessary to ensure a high locality of light emis-sion. However, too high a concentration valuewould result in signi"cant light attenuationalong the capillaries, despite the large Stokesshift (Fig. 5) that is a common feature of the LSsconsidered. The best compromise between therequirements of high scintillation e$ciency, locallight emission and low attenuation is reached ata concentration of &3 g/l for all dye-solventcombinations tested.

f High refractive index: A high refractive index isneeded to maximize the fraction of the scintilla-tion light trapped in the capillary in which it isproduced. This trapping fraction depends on therefractive index n

1of the LS (i.e. the "bre core),

on the refractive index n2

of the capillary glass(i.e. the "bre cladding), and on the geometry ofthe "bre cross-section. For an ideal "bre havinga circular cross-section, and assuming that lightis produced isotropically and uniformly in thecore, the trapping fraction in each directionalong the "bre is [7] 1

2(1!n2

2/n2

1). The refractive

indices of LSs based on 1MN and on IPN are,respectively, n

1"1.62 and 1.59, signi"cantly

higher than the refractive index of the capillaryglass, n

2"1.49. In the case of the ideal "bre of

circular cross-section, the resulting trapping frac-tion in each direction of the "bre is 7.7% for LSsbased on 1MN and 6.1% for LSs based on IPN.

f Short decay time: For all considered LSs based on1MN or IPN, about 96% of the emitted lightdecays with a time constant of &6}8 ns, and theremaining 4% of the light has a decay time of&30}40 ns [30]. As an example, the decay

curve of 1MN doped with R6 is shown in Fig. 7.The curve was obtained using the single-photo-electron delayed-coincidence counting method[41].

f High radiation hardness: In Fig. 8 we show thedependence of the scintillation yield on the ab-sorbed dose for 1MN doped with R6, R39 andR45 [37]. With the "rst two dyes the behaviouris quite similar and an absorbed dose as high as200 Mrad reduces the light yield by only&30}40%. For 1MN doped with R45, thesharp decrease of the light yield observed ata dose below 10 Mrad is due to degradation ofthe dye [26,39]. The radiation hardness of LSsbased on IPN can be even better [37] than thatof LSs based on 1MN. For comparison, thebehaviour of two LSs produced by Bicron9 isalso shown in Fig. 8. As well as a!ecting the LS,the dose absorbed by a capillary bundle candarken the capillary glass. Taking into accountthat, at each total internal re#ection, light pene-trates into the glass to a depth of a fewwavelengths (evanescent wave), this glass


Fig. 8. Light yield as a function of the absorbed dose, for 1MNdoped with: (a) R39; (b) R6; (c) R45; and for two LSs produced byBicron; (d) BC-599-13B; (e) BC-599-13G.

Fig. 9. Light attenuation along capillaries that have 500 lmdiameter, and are "lled with IPN#R45, for various degrees ofscintillator purity: (a) before puri"cation; (b) after one puri"ca-tion stage; (c) after two stages; (d) after three stages.

Fig. 10. Light yield of 1MN#R6 scintillator, when placed inatmospheres of di!erent gases. The light yield in air is set to 1.

darkening results in further light attenuation,especially for rays that undergo many re#ectionsbefore reaching the readout end of the bundle.This e!ect has been measured [37] and is foundto be less important than the radiation damageof LSs, even for non-radiation-hard glass.

f Bundle xlling procedure: The scintillation proper-ties of our LSs depend strongly on their purity.In view of this, the LSs were puri"ed before beingused and steps were taken to avoid any contami-nation during the bundle "lling procedure.

Various puri"cation methods [42] have beenset up. These have made possible drastic in-creases in the light yield and transparency of theLSs. As an example, we show in Fig. 9 the e!ect[26] of various degrees of puri"cation on a LSbased on IPN doped with R45. After three stepsof puri"cation the light yield is increased bya factor &1.8 and the light attenuation length,measured in capillaries of 500 lm diameter, isincreased by a factor &9.

The presence of reactive gases, in particularoxygen, dissolved in the LSs, is another cause ofdeterioration [38] of the scintillation properties.This is illustrated in Fig. 10, where we show thevariation in the light yield of 1MN#R6 whenthe scintillator is placed in atmospheres of di!er-ent gases.

To avoid deterioration of the LS propertiesduring the "lling of a capillary bundle, we havedeveloped the set-up represented in Fig. 11. Withthis set-up, the LS was "rst degassed in vacuumfor several hours, and was then introduced intothe bundle by applying a slight argon overpres-sure. A peristaltic pump forced the LS to circu-late through the bundle for many hours, so as to


Fig. 11. Schematic view of the set-up used for "lling capillarybundles with LS: (1) capillary bundle; (2) polyethylene balloon;(3) glass reservoir for LS; (4) soft hoses; (5) peristaltic pump; (6)glass vessel for introduction of Ar; (7) pressure-gauge; (8) glassvessel for initial LS degassing; (9) vacuum pump; (10) lightsource and (11) microscope for monitoring the bundle "lling.

10Thomson, Division Tubes Electroniques, Boulogne-Billan-court, France.

11Varo Inc., Garland, TX 75041, USA.12The magni"cation m of a non-zoomable image intensi"er is

reported as the ratio of the output diameter to the input dia-meter, all diameters in mm.

eliminate gas bubbles that might be present inthe capillaries, and which would cause poor lighttransmission. This procedure was continuouslymonitored by illuminating one end of the bundleand using a microscope to observe the lightemerging from the other end. During the "llingsequence the LS comes into contact with argon,and with glass, metals and te#on, all meticulous-ly cleaned in advance.

Using the procedure described, the squarebundle and the hexagonal bundle were "lled with1MN#3M15 and 1MN#R39, respectively. Inboth cases the dye concentration was 3 g/l.

5. Optoelectronic readout chain

The light output for a minimum-ionizing particlecrossing a capillary bundle perpendicularly to thebundle axis consists, at most, of a few photons percapillary, corresponding on average to less thanone photoelectron. It is therefore necessary to am-plify the output signal by a large factor before trackimages can be detected and stored using a posi-tion-sensitive receiver. The "rst stage of the readoutchain is the most critical for the detection e$ciency.

It must have the highest possible quantum e$cien-cy, including also the transmittance of the "bre-optic entrance window. Precisely engineeredmechanical systems, such as shown in Fig. 12,ensure that there is good optical coupling betweencapillary bundle and "rst readout stage.

The image receiver adopted in the readout chainof the square bundle was a CCD made by Thom-son,10 model TH7896A-(H), composed of1024]1024 pixels, each with dimensions of19]19 lm2. This CCD was preceded by a series of"ve Image Intensi"ers (II

1}II

5, Fig. 12), the "rst

and the last made by Geosphaera, the others pro-duced by Varo.11 These provided an ampli"cationof +105. The LS contained in the capillaries of thesquare bundle was in direct contact with the "bre-optic window of the "rst II. This II was chosenbecause of its high quantum e$ciency (&20% atthe wavelength of maximum sensitivity). Its demag-ni"cation12 (m"25/40K0.62) was compensatedby the second II, which magni"ed the image bya factor m"46/25K1.8. All II's except the fourthwere electrostatically focused. The main image in-tensi"cation was provided by the fourth tube,which was proximity focused and containeda microchannel plate (MCP). The last II demagni-"ed the image to "t the dimensions(19.5]19.5 mm2) of the CCD. The spatial resolu-tion of the II's of the chain, measured ata wavelength of 500 nm, was &24 lp/mm for theMCP tube, and was in the range 29}39 lp/mm forthe electrostatically focused IIs.

The optoelectronic chain could be gated by ap-plying a &200 V pulse to the photocathode of theMCP tube. This made a triggered readout possible.During the time required for the trigger decision(&180 ns) the decay times of the phosphors of the"rst three IIs provided an optical memory, tempor-arily storing track images. Most of the image per-sistence was introduced by the "rst II, which hada K-67 (green) phosphor, with a &10 ls decay


Fig. 12. Optoelectronic readout chains coupled to the square capillary bundle (upper "gure) and to the hexagonal capillary bundle(lower "gure). The magni"cation m of each non-zoomable II is expressed as the ratio of the output diameter to the input diameter (inmm). The coupling to the readout chain is direct for the square bundle and is via a "bre-optic plate (FOP) for the hexagonal bundle.

time. The P47 (blue) phosphors of the second andthird IIs had a decay time of &35 ns, and so didnot contribute signi"cantly to the optical memory.As shown in Fig. 13, the image persistence wassu$ciently long that the wait for the "rst-leveltrigger resulted in the loss of only a small fraction ofthe light. The fourth and "fth IIs had, respectively,a P46 (green) phosphor, with a decay time of&80 ns, and a KO-425 (blue) phosphor, witha decay time of &50 ns. Since these phosphorscame after the MCP, where the trigger was applied,they did not contribute to the optical memory.

The readout chain of the hexagonal bundle(Fig. 12) had an EBCCD camera as image receiver,and was much more compact than the readoutchain described above. The EBCCD [43}45] com-bined the functions of an electrostatically focused II

and a conventional CCD. It was essentially an IIwhere the phosphor screen was replaced by a rever-sed thinned CCD (thickness &10 lm), consistingof 1024]1024 pixels, each with dimensions of13]13 lm2. Electrons emitted at the photo-cathode were accelerated through &15 kV towardthe CCD, where one electron}hole pair was createdfor every 3.6 eV released in the silicon. A gain of&3500 electrons/photoelectron was obtained[43]. Using specially designed low-noise readoutelectronics, a single photoelectron was clearly dis-tinguished from the background. The EBCCD thenprovided all the signal ampli"cation required toallow detection of track images. Nonetheless, theEBCCD was preceded in the readout chain by an IIsimilar to the "rst II of the readout chain used withthe square bundle. The II had the advantage of


Fig. 13. Monte-Carlo calculation of the variation with time ofthe light emitted by the image intensi"er placed in front of theMCP tube (readout chain of the square bundle). The "rst micro-second has been expanded in the insert. The phosphors of the"rst three image intensi"ers act as an optical memory, and onlythe light emitted before the implementation of the "rst-leveltrigger is lost (shaded area).

Fig. 14. Set-up of the capillary bundles in the CHORUS de-tector, for studies of neutrino interactions. The trigger-counterpositions with respect to the capillary bundles are shown in thelower part of the "gure, for square and hexagonal bundles.

a high quantum e$ciency, and the long decay timeof the II phosphor provided the optical memorywhile waiting for a trigger decision. When the trig-ger was satis"ed, the EBCCD was gated by ap-plying a 1.6 kV pulse to the focusing electrode. Theoptical coupling to the readout chain for the hexa-gonal bundle was slightly di!erent from that for thesquare bundle. A "bre-optic plate (Fig. 12) was"xed to the bundle's end, then the entire structurewas coupled to the entrance window of the II. Thismeant that the LS was more securely sealed, but thedetected light was reduced by &40% because oflosses in the "bre-optic plate.

6. Set-up for tests

The performance of the two large capillarybundles has been studied in a pilot experimentlocated in front of the CHORUS detector [27], inthe CERN wideband neutrino beam. In this situ-ation the bundles could detect either beam muonsor neutrino interactions.

For muon tracking, the bundles were placed per-pendicularly to the beam direction. Muons wereselected by a 2.5]5 cm2 trigger counter placed infront of the bundles.

For studies of neutrino interactions (Fig. 14), thesquare bundle was placed vertically, so that it wasperpendicular to the beam (transverse con"gura-tion), while the hexagonal bundle was placed hori-zontally, with its axis parallel to the beam(longitudinal con"guration). In the images re-corded with the transverse con"guration, most ofthe outgoing tracks from neutrino interactions inthe bundle were collimated in the forward direc-tion, within approximately $150 mrad. In thelongitudinal con"guration, by contrast, tracks ofinteraction products were spread over 2p rad, in anazimuthal projection.


Fig. 15. Scheme of the trigger logic for gating the MCP (EBCCD) in the readout chain of the square (hexagonal) bundle. The scintillatorcounters S

1and S

2(S

3and S

4) are in coincidence with the beam spill and are in anticoincidence with a dead time (DT) pulse during the

acquisition time of the CCD (EBCCD). The CCD chips were continuously cleared except during the integration-time control (ITC)pulse. The interrupt (INTR) starts the data acquisition.

Neutrino interactions in the capillary bundleswere detected by two plastic scintillator counterspositioned immediately downstream of thebundle (Fig. 14). Counters S

1and S

2, used

with the square bundle, had dimensions of160]2.5 cm2; counters S

3and S

4, used with the

hexagonal bundle, had dimensions of 20]20 cm2.Tracks that were measured in the capillary bundlesand crossed the CHORUS emulsion target couldbe reconstructed in the various downstreamdetectors.

The trigger logic for detecting neutrino interac-tions is represented in Fig. 15. A "rst-level `localatrigger was satis"ed when there was a coincidencebetween S

1and S

2(or between S

3and S

4for the

hexagonal bundle) during the 6 ms accelerator spill.The "rst-level local trigger occurred &180 ns afterthe physical event. Subsequently, a 1 ls pulseopened the gate of the MCP (or of the EBCCD),allowing intensi"ed images to pass to the CCD (orto the silicon chip in the EBCCD). The gate pulsede"ned the time to wait for a second-level triggerfrom the CHORUS trigger system. This included

a large veto counter, placed in front of the appar-atus (Fig. 14), for rejection of beam muons. If thissecond-level trigger arrived within 1 ls of the "rst-level trigger, the gate of the MCP (or EBCCD) waskept open for an additional 100 ls, a time longenough to enable collection at the CCD chips ofmost of the scintillation light produced in the capil-lary bundle and intensi"ed by the optoelectronicchain.

Since neither the CCD nor the EBCCD hada fast-clear facility, these sensors were cleared bycontinuous readout at a pixel clock frequency of10 MHz. This means that the chips were fullycleared every &110 ms. When a "rst-level triggerarrived, the continuous clearing was interruptedand the image integration at the silicon chip wasenabled by an integration-time control (ITC) pulseof &200 ls. If the "rst-level trigger was con"rmedby the second-level trigger, an interrupt (INTR)signal began the data acquisition, during whicha dead time (DT) prevented further triggers. Thetiming of all pulses relevant to readout is indicatedin Fig. 15.


13The MCP tube was proximity focused, and so did notcontribute signi"cantly to the distortion.

7. Experimental results

7.1. Magnixcation and distortion of the readoutchains

The magni"cation of an electrostatically focusedII usually has a slight dependence on the distancebetween the point considered and the centre of thephotocathode. The result is pincushion distortion:the image of a radial straight line is still straight andradial, but the image af an o!-centre line appearscurved. This e!ect was appreciable for the squarebundle's readout chain, which contained four elec-trostatically focused IIs,13 but was found to benegligible for the readout chain of the hexagonalbundle.

To have quantitative measures of the distortionwith each bundle and readout chain, a referencegrid was placed in contact with the bundle endopposite the readout window. This grid, which con-sisted of 20 lm width transparent lines at 2 mmintervals, was illuminated from behind by an LED.In Fig. 16 we show the image of the grid as seen bythe CCD through the readout chain used with thesquare bundle. The distortion a!ecting the imagemust be taken into account if the directions ofparticles crossing the bundle are to be correctlyreconstructed. The distortion was parameterizedusing the expression R

0"(R

*/m)(1#aR

*#bR2

*),

where R0

and R*are the respective distances from

the optical centre of a point in the object (grid)space and of a point in the image (CCD) space, m isthe `centrala magni"cation of the chain (i.e. themagni"cation of the chain along its axis), and a andb are coe$cients. First, the coordinates of the op-tical centre were found by taking advantage of theradial symmetry of the distorted grid image. Next,the values of m, a and b were determined from theknowledge of the distance between grid lines in theobject space. In the case of the square bundle'sreadout chain we found that the optical axis wasdisplaced with respect to the bundle centre by&2 mm, that the central magni"cation wasmK1.24, close to the nominal value of 1.27, and

that the two parameters describing the distortionhad values aK!2]10~3 mm~1 and bK!3]10~4 mm~2. Once these parameters had beendetermined, all distortion e!ects could be correctedfor. The distortion correction improved the e$cien-cy for track "nding in o!-line analyses, and allowedbetter de"nition of track directions. The width ofthe distribution of hit displacements from "ttedtracks (residuals distribution), which is typicallya few tens of lm, is reduced by &22% when thecorrection is applied.

7.2. Vignetting

Like the magni"cation, the light ampli"cation ofan electrostatically II is usually dependent on thedistance between the point considered and the op-tical axis of the tube. This is essentially a conse-quence of the curvature of the internal surfaces ofthe "bre-optic input and output windows. Thiscurvature implies larger glass thickness, and solower light transmission, at the window edges. This`vignettinga e!ect can introduce gain non-uniform-ity of 20}50% for electrostatically focused IIs. Inproximity focused IIs, where the input and outputwindows are #at, this e!ect is not present and thegain is uniform to within a few percent over the fullphotocathode surface.

The e!ect of vignetting was expected to be im-portant for the square bundle's readout chain,where four electrostatically focused IIs were pres-ent, and has been studied [46] by analyzing beammuons crossing the bundle perpendicularly to itsaxis. The transverse dimension of the beam waslarge enough to ensure that the triggering muonswere uniformly distributed across the bundle. How-ever, not all of them were recorded because only thecentral part of the bundle was seen by the II chainand by the CCD (Fig. 17).

More than a thousand muon tracks were re-corded. In the o%ine analysis of these events a thre-shold was "rst applied to the CCD pixel pulseheight, so as to reduce the background, then muontracks were reconstructed. In Fig. 17a we show, inthe projection perpendicular to the bundle axis, thepositions of all hits lying on the reconstructedtracks. The applied threshold resulted in a loss ofhits near the edges of the readout zone, where the


Fig. 16. Pincushion distortion in the readout chain of the square bundle. On the left: image of the reference grid, as seen at the CCD. Onthe right: outline of the grid reconstructed in the real space of the bundle, after application of the corrections described in the text. Thecrosses indicate the position of the chain's optical centre.

Fig. 17. (a) Distribution, in plane perpendicular to square bundle, of hits associated with muon tracks uniformly crossing the bundle.The dashed circle and square delimit the zones seen by the image-intensi"er chain and by the CCD respectively. (b) Same as in (a) butwith each hit weighted by its pulse height measured by the CCD. Near the edge of the accepted zone, vignetting causes (a) a lower hitdensity and (b) reduced signal light.

light gain of the chain was lowered by the vignet-ting. The e!ect is even more evident in Fig. 17b,where we show the pulse-height distribution mea-sured by the CCD.

7.3. Hit density and light attenuation

To evaluate the tracking capabilities of the twocapillary bundles, images produced by minimum-

ionizing particles crossing perpendicularly to thebundles were analyzed. For each trigger, coordi-nates and pulse-heights were recorded for all pixelswith pulse height above a threshold value. A tuningof the pulse-height threshold was performed as the"rst step of the o%ine analysis. We then formedclusters from contiguous pixels of above-thresholdpulse height, centred on a local maximum (apixel with a larger pulse height than the eight


14The present results are slightly di!erent from those re-ported in Refs. [25,47]. An error in the data analyses reported inthese previous works has now been corrected.

Fig. 18. Hit densities measured along the tracks of minimum-ionizing particles crossing the bundles perpendicularly to theiraxes, as a function of the distance between the tracks and thereadout end of the bundles. Open points relate to the squarebundle and were taken at di!erent times after the bundle "lling.The curve represents the result of a "t of two exponentials to allthe open points. Attenuation lengths near the square bundle'sends are indicated. Closed circles refer to the hexagonal bundle.

surrounding pixels). A track appeared as a suc-cession of approximately aligned clusters. Eachcluster comprised several tens of pixels and corre-sponded to a photoelectron emitted at the photo-cathode of the "rst II of the readout chain.A cluster's centroid was calculated as the weightedaverage of the pixel coordinates, the weights beingthe pixel pulse heights. Each cluster centroid wasidenti"ed as a hit. Taking into account magni"ca-tion and distortion in the readout chain, so thatdistances were referred back to the real space of thedetector, hit positions were reconstructed in a planeorthogonal to the bundle axis. A straight line was"tted to the tracks, using only hits within a road of$150 lm.

Light attenuation along the capillaries resultedin the hit density along a track depending on thedistance, d, between the track and the readout endof the bundle. For the square bundle, measure-ments of hit density were performed just after "ll-ing, in tests in a 5 GeV pion beam. Measurementswere repeated, for muon tracks, at intervals overa two-year period. During this time the LS in thebundle was left unchanged and the entire detector,with its readout chain, was mounted in the cool box(& 53C) of the CHORUS experiment [27].The results of these measurements14 are shown inFig. 18. No ageing e!ect is present over the twoyears. All points are well "tted by the sum of onlytwo exponentials. The light's attenuation length(Fig. 18) varies from jK50 cm near the bundle'sreadout end, to jK400 cm at the opposite end.Extrapolating the "t to d"0 leads to a maximumhit density of +8 hits/mm.

For the hexagonal bundle, measurements havebeen made for only four distances, and do not allowa meaningful determination of the attenuationlength. The lower hit density measured for thisbundle is mostly due to the "bre-optic plate addedbetween the bundle's readout end and the optoelec-tronic chain, as described in Section 5. Other fac-tors, such as the quantum e$ciencies of the "rsttwo photocathodes of each readout chain, and the

quality of the inner surfaces of the capillaries couldalso contribute to the di!erent hit densities.

7.4. Spatial resolution from single-track events

We have measured the spatial resolution of ourdetectors through an analysis of the tracks of min-imum-ionizing particles. As well as being in#uencedby physical e!ects, such as delta-ray productionand Coulomb scattering, the spatial resolution ofa detector based on glass capillaries is dependenton several factors. These include the inner diameterof the capillaries, the uniformity of the capillaryarray, and the resolution and distortion of the op-toelectronic chain.

The spatial resolution can be described in termsof three quantities. These are the `cluster sizea (p

#4),

the `track residuala (p53) and the `two-track resolu-

tiona (p55). The square of the last-mentioned quanti-

ty is approximately the sum of the squares of theother two. All measurements made are with refer-ence to a two-dimensional image projection ona plane perpendicular to the bundle. Unless in-dicated otherwise, values are referred to the realspace of the bundle: image magni"cation, includingthe magni"cation in the tapered section of thesquare bundle, is taken into account.


Fig. 19. Distributions showing: (a) variation of pulse height with projected distance d#4

from a cluster's centroid; (b) variation of numberof hits with distance d

53from "tted track; (c) variation of pulse height with distance d

55from "tted track. The upper and lower

distributions refer, respectively, to the square bundle and to the hexagonal bundle. Each distribution is "tted by a Gaussiansuperimposed on a background. The standard deviations of the "tted Gaussians are reported. All distances are given in the real space ofthe capillary bundles.

f Cluster size: An isolated cluster was due toa single photoelectron emitted at the photo-cathode of the "rst II of the optoelectronic chain.Its size on the CCD was therefore una!ected bythe bundle characteristics and depended only onthe spatial resolution of the components of thereadout chain. The average cluster pro"le hasbeen obtained by superposing the centroid ofmany isolated clusters and summing the pulseheights of the pixels at each position. This pro"lewas then projected onto an axis, chosen parallelto the pixel sides. Resulting distributions for thetwo readout chains are shown in Fig. 19a. Thesedistributions are "tted by a Gaussian superim-posed onto a broad background. The standarddeviations of the Gaussians provide measures ofthe cluster size. The values obtained arep#4K21 lm for the square bundle and

p#4K26 lm for the hexagonal bundle, in rea-

sonable agreement with the values of 20 and22 lm expected from the nominal parameters ofthe readout chains. If, instead of being referred tothe real space of the bundle, the cluster sizes areevaluated at the CCD, the values obtained arepCCD#4

K27 lm with the square bundle andpCCD#4

K14 lm with the hexagonal bundle, re-#ecting the better spatial resolution of thereadout chain composed of only one II followedby the EBCCD tube.

f Track residuals: Track residuals measure hit dis-placements from "tted tracks. Since each hit isassumed to correspond to a single photoelectron,all hit displacements are given the same weight.In Fig. 19b we show distributions of track resid-uals for particles crossing the capillary bundlesat a "xed distance from the readout end. These


Fig. 20. Images of neutrino interactions occurring: (a), (b) in the square bundle placed perpendicularly to the beam, neutrinos arrivingfrom the left; (c), (d) in the hexagonal bundle placed longitudinally to the beam, neutrinos arriving perpendicularly to the "gure. Theinteraction shown in (b) takes place in the tapered part of the bundle, close to the readout window. The hit density observed along thetracks is consequently much higher than for the event shown in (a). In the longitudinal con"guration, the angle between the emittedparticles and the capillary bundle is, on average, lower than in the transverse con"guration. This is the reason for the enhanced hitdensities observed in (c) and (d).

distributions have been "tted by a Gaussiansuperimposed onto a broad background. Thevalues obtained for the standard deviations ofthe Gaussians are p

53K27 lm for the square

bundle and p53K33 lm for the hexagonal

bundle. These values are signi"cantly larger thanexpected for an ideal capillary bundle. A simula-tion that takes into account the capillary dia-meter, the resolution of the image intensi"ersand the "nite CCD pixel size predicts a value ofp53

in the range 12}15 lm for the two bundles.This suggests that a signi"cant contribution tothe spatial resolution comes from the non-uni-

formity of the large-volume capillary bundles.This view is supported by the "nding [47] thatp53

varies by about $10% for tracks crossingthe square bundle at di!erent positions along itslength.

f Two-track resolution: The combined e!ect of thecluster size and of hit displacements is seen in thedistribution of pulse height about "tted tracks.Distributions measured for particles at a "xeddistance from the bundles' readout end areshown in Fig. 19c. The distributions have againbeen "tted by a Gaussian superimposed ontoa broad background. The standard deviations of


Fig. 21. Neutrino interaction observed in the square capillary bundle and reconstructed in CHORUS. The neutrino arrives from the left.A decay vertex is identi"ed through a small kink, con"rmed by a change in the hit density (see text). A l`, a l~ and a third chargedparticle (labelled `3a) are detected in the CHORUS apparatus.

the Gaussians represent the detectors' two-trackresolution. The values obtained are p

55K34 lm

for the square bundle and p55K42 lm for the

hexagonal bundle. Like p53, p

55has been found

[47] to vary along the length of the square bundle,the variation in this case being about $5%.

A comparison of the two distributions ofFig. 19c shows that the detector based on thesquare bundle had a better two-track resolution

than the detector based on the hexagonal bundle.The magni"cation introduced by the tapered endof the square bundle compensates for the worseresolution of the bundle's readout chain.

7.5. Reconstruction of many-track events

The good tracking capability of the capillarytechnique is illustrated in Fig. 20, where we show


four examples of neutrino interactions, two in thesquare bundle placed perpendicularly to the beam,and two in the hexagonal bundle placed longitudi-nally to the beam.

For events recorded with the transverse con"g-uration (square bundle), various methods allowingdetermination of the resolution on vertex positionand on track direction have been developed [48].After an approximate determination of the vertexposition, all hit combinations are tested to formtrack candidates roughly consistent with an originat the vertex. A circular projection algorithm isused to perform the "nal track selection. Each trackis then represented by a straight line "tted to hitswithin a road of $150 lm. Hits lying in the con-fused region very close to the vertex are not con-sidered. Tracks are "tted to a single vertex, and theerror matrix for the vertex parameters allows def-inition of the resolution on vertex position. Usingthis automatic "tting procedure, several tensof neutrino interactions have been analyzed. Theresolution on vertex position was found to be&30 lm in the direction orthogonal to the beamand &90 lm along the beam. The perpendiculardistance between a track and the vertex de"nes thetrack's impact parameter. A track with an impactparameter signi"cantly di!erent from zero may in-dicate the presence of a short-lived particle amongthe interaction products. The distribution of impactparameters is nearly exponential, with a meanvalue of &43 lm.

Tools have also been developed to allow thereconstruction described above to be performedunder the guidance of an operator interacting witha graphics display. With the interactive "t, thereconstruction e$ciency is increased by &14%.By selecting well-de"ned, distortion-free track seg-ments, a mean impact parameter of &23 lm hasbeen obtained.

The potential of the capillary target for detectionof short-lived particles is illustrated by the eventshown in Fig. 21. This event features "ve chargedparticles emerging from a neutrino interaction inthe capillary bundle. Three of the particles weredetected in the CHORUS apparatus, while theother two were emitted outside its angular accept-ance. Two of the three particles detected inCHORUS were identi"ed in the muon spectrom-

eter as opposite-sign muons, while the third particlewas detected in the "bre tracker. Whereas thel~ comes from the neutrino interaction vertex,close inspection of the l` track shows that it comesfrom the decay of a short-lived particle. The projec-ted decay kink is very small, but the decay can beidenti"ed from a change in the apparent hit densityat the decay point, this change resulting from thedi!erence in angle between the decaying particleand the l`. The short black track emerging fromthe primary vertex could be either a nuclear frag-ment or a track emitted almost parallel to thecapillaries. The energy of the l` (50 GeV) and theprojected #ight path of the decaying particle areconsistent with the decay D`PK1 0l`m

l, the

K1 0 showing up as a &70 GeV shower in thehadronic calorimeter.

8. Conclusions

We have developed high-resolution tracking de-tectors based on glass capillaries "lled with organicliquid scintillators of high refractive index. Capil-lary detectors are an evolution of detectors basedon glass or plastic scintillating micro"bres and o!eran improved performance. Such tracking devicescan be of great interest in high-energy physics asmicrovertex detectors for the study of vertextopologies and decays of short-lived particles.

Two large-size prototype detectors have beenconstructed. The two capillary bundles, each com-posed of +106 capillaries, were manufactured withtwo di!erent procedures, and were operated withdi!erent readout chains. One of the readout chainswas based on Image Intensi"ers read by a standardCCD, while the other took advantage of the newlydeveloped EBCCD tube. In both systems, imageswere detected by &106 pixels.

A high scintillation e$ciency and a light-attenu-ation length in excess of 3 m has been achievedthrough special puri"cation of the liquid scintil-lators used. Along the tracks of minimum-ionizingparticles, the hit densities obtained were&8 hits/mm at the readout window, and&3 hits/mm at a distance of &1 m.

The two capillary-based detectors have been op-erated in front of the CHORUS detector, in the


CERN wide-band neutrino beam. Observation ofmuon tracks and neutrino interactions has allowedevaluation of the track and vertex reconstructioncapabilities. Typical spatial resolutions are in therange 20}40 lm.

Acknowledgements

This work has been carried out in the frameworkof the CERN Research and Development pro-gramme RD46. Financial support was providedthrough the European Union's TMR NetworkContract ERBFMRX-CT98-0196.

References

[1] R. Ruchti et al., IEEE Trans. Nucl. Sci. NS-32 (1985) 590.[2] R. Ruchti et al., IEEE Trans. Nucl. Sci. NS-33 (1986) 151.[3] M. Atkinson et al., Nucl. Instr. and Meth. A 254 (1987) 500.[4] M. Atkinson et al., Nucl. Instr. and Meth. A 263 (1988) 333.[5] C. Angelini et al., Nucl. Instr. and Meth. A 277 (1989) 132.[6] C. Angelini et al., Nucl. Instr. and Meth. A 289 (1990) 342.[7] C. Angelini et al., Nucl. Instr. and Meth. A 295 (1990) 299.[8] H. Blumenfeld et al., Nucl. Instr. and Meth. A 278 (1989)

619.[9] C. D'Ambrosio et al., Nucl. Instr. and Meth. A 325 (1993)

161.[10] M. Adinol" et al., in: R. Nahnhauer (Ed.), Proceedings of

the Workshop on Application of Scintillating Fibers inParticles Physics, Blossin, Germany, September 1990.

[11] M. Adinol" et al., Nucl. Instr. and Meth. A 310 (1991)485.

[12] M. Adinol" et al., Nucl. Instr. and Meth. A 315 (1992) 67.[13] H. Leutz, Nucl. Instr. and Meth. A 364 (1995) 422, and

references therein.[14] A.G. Denisov et al., Nucl. Instr. and Meth. A 310 (1991)

479.[15] A. Artamonov et al., Nucl. Instr. and Meth. A 300 (1991)

53.[16] N.I. Bozhko et al., Nucl. Instr. and Meth. A 317 (1992) 97.[17] M. Adinol" et al., Nucl. Instr. and Meth. A 311 (1992) 91.

[18] M. Adinol" et al., Nucl. Instr. and Meth. A 315 (1992) 177.[19] C. Cianfarani et al., Nucl. Instr. and Meth. A 339 (1994)

449.[20] A. Cardini et al., Nucl. Instr. and Meth. A 346 (1994) 163.[21] S. Buontempo et al., Nucl. Instr. and Meth. A 360 (1995) 7.[22] P. Annis et al., Nucl. Instr. and Meth. A 367 (1995) 377.[23] G. Martellotti et al., Proc. Int. Soc. Opt. Eng., SPIE 2551

(1995) 53.[24] P. Annis et al., Nucl. Phys. B (Proc. Suppl.) 54B (1997) 86.[25] P. Annis et al., Nucl. Phys. B (Proc. Suppl.) 61B (1998) 390.[26] V.G. Vasil'chenko et al., Instr. Exp. Tech. 40 (1997) 175.[27] E. Eskut et al. (CHORUS collaboration), Nucl. Instr. and

Meth. A 401 (1997) 7.[28] R. van Dantzig et al., RD46 Proposal, CERN/LHCC 95-7,

P60/LDRB, February 1995.[29] C. Angelini et al., Nucl. Instr. and Meth. A 281 (1989) 50.[30] G.I. Britvich et al., Nucl. Instr. and Meth. A 425 (1999) 498.[31] C. Angelini et al., Nucl. Instr. and Meth. A 289 (1990) 356.[32] M. Atkinson et al., Nucl. Instr. and Meth. A 254 (1987) 500.[33] P. Pavan et al., Nucl. Instr. and Meth. B 61 (1991) 487.[34] S. Majewski et al., Nucl. Instr. and Meth. A 281 (1989) 497.[35] G.I. Britvich et al., Nucl. Instr. and Meth. A 326 (1993) 483.[36] S. Majewski et al., Nucl. Instr. and Meth. A 281 (1989) 500.[37] S.V. Golovkin et al., Nucl. Instr. and Meth. A 362 (1995)

283.[38] S. Buontempo et al., Nucl. Instr. and Meth. A 425 (1999)

494.[39] S.V. Golovkin et al., Nucl. Instr. and Meth. A 305 (1991)

385.[40] J.B. Birks, The Theory and Practice of Scintillation Count-

ing, Pergamon Press, Oxford, 1964.[41] L.M. Bolinger, G.E. Thomos, Rev. Sci. Instr. 32 (1961)

1044.[42] V.G. Lapshin et al., IHEP Report no. 93-25, Protvino,

1993.[43] S. Buontempo et al., Nucl. Instr. and Meth. A 413 (1998)

255.[44] P. Annis et al., in: A.D. Bross, R.C. Ruchti, M.R. Wayne,

(Eds.), Proceedings of the Workshop on Scintillating FiberDetectors, Notre Dame, USA, 1997, p. 509.

[45] S.V. Golovkin et al., Proc. Int. Soc. Opt. Eng., SPIE 2551(1995) 118.

[46] K. Holtz, Thesis, WestfaK lische Wilhelms-UniversitaK t,MuK nster, 1998.

[47] K. Hoepfner, Thesis, Humboldt-UniversitaK t, Berlin, 1996.[48] K. Hoepfner et al., Nucl. Instr. and Meth. A 406 (1998) 195.


high-resolution tracking using large capillary bundles with liquid...

Documents