Date post: | 22-Nov-2023 |
Category: |
Documents |
Upload: | independent |
View: | 0 times |
Download: | 0 times |
ARTICLE IN PRESS
Nuclear Instruments and Methods in Physics Research A 525 (2004) 485–495
*Corresp
051-209526
maximilian
0168-9002/$
doi:10.1016
Atmospheric muon flux measurements at the external site ofthe Gran Sasso Lab
E. Barbutoa, C. Bozzaa, M. Cozzib, N. D’Ambrosioc, G. De Lellisc, G. De Rosac,M. De Seriod, L.S. Espositob, G. Giacomellib, M. Giorginib, G. Grellaa, M. Ievad,
G. Mandriolib, P. Migliozzic, M.T. Muciacciad, L. Patriziib, C. Pistilloc,P. Righinie, G. Rosae,*, S. Simoned, M. Siolib,*, C. Sirignanoa, G. Sirrib,
G. Sorrentinoc, V. Tioukovc
aDip. di Fisica dell’Univ. di Salerno and INFN, Salerno I-84081, ItalybDip. di Fisica dell’Univ. Sezione di Bologna and INFN, Viale B. Pichat 6/2, Bologna I-40127, Italy
cDip. di Fisica dell’Univ. di Napoli and INFN, Napoli I-80125, ItalydDip. di Fisica dell’Univ. di Bari and INFN, Bari I-70126, Italy
eDip. di Fisica dell’Univ. di Roma ‘‘La Sapienza’’ and INFN, Roma I-00185, Italy
Received 12 December 2003; accepted 15 January 2004
Abstract
Measurements of the atmospheric flux of muons and other charged particles have been performed at the external site
of the Gran Sasso Laboratory (LNGS, Italy, 1000 m a:s:l:), in view of the OPERA experiment. Scintillation counters
and nuclear emulsion plates have been used. Rates and angular distributions at several depths of Fe shielding are
presented and compared with Monte Carlo simulations.
r 2004 Elsevier B.V. All rights reserved.
PACS: 29.30.Ep; 29.40.Mc; 29.40.Rg; 96.40.Tv
Keywords: Emulsion cloud chamber; Scintillation counters; Cosmic ray muons
1. Introduction
In the framework of the CNGS project [1], theOPERA experiment aims at the detection of nts inan almost pure nm beam [2]. OPERA is based onthe nuclear emulsion technique, used to study
onding authors. Tel.: +39-051-2095226; fax: +39-
9.
addresses: [email protected] (G. Rosa),
[email protected] (M. Sioli).
- see front matter r 2004 Elsevier B.V. All rights reserve
/j.nima.2004.01.078
cosmic rays since their discovery and in manyother high-energy physics experiments. In the lastfew years, the technique was applied to neutrinoexperiments on unprecedented scale and substan-tially upgraded [3,4]. Further developments are inprogress.The OPERA target is highly segmented in basic
units called ‘‘bricks’’, about 12:8 cm� 10:3 cmwide, 8:0 cm thick along the beam direction, eachconsisting of 57 double-coated emulsion filmsinterleaved with 56 lead foils, 1 mm thick. The
d.
ARTICLE IN PRESS
Fig. 1. Layout of the iron structure used in the test. Three main
blocks of 25 cm of iron have been used.
E. Barbuto et al. / Nuclear Instruments and Methods in Physics Research A 525 (2004) 485–495486
physics program of the experiment requires analignment between emulsion sheets to within a fewmm; more accurate than the alignment provided bythe packing procedure. This goal can be accom-plished by exposing bricks, selected as containingneutrino interactions induced by the CNGS beam,to an external source of penetrating particles,namely cosmic ray muons. A few particles/mm2
have to be collected.For this purpose, a dedicated pit will be
excavated at the external site of the LNGS,suitably shielded by an iron cover. Bricks will beexposed to cosmic rays at the bottom of the pit,and then extracted and disassembled. The emul-sion sheets will be promptly developed and sent toscanning laboratories. The thickness of the ironcover must be such that (a) effective shielding isprovided against cosmic ray e.m. showers, uselessfor calibration and harmful for the particle(electron) identification methods applied to neu-trino events, (b) a suitable muon flux can becollected within a reasonable exposure time (1 or 2days), (c) to some extent the energy spectrum ofmuons is hardened, (d) they are focussed at narrowdip (zenith) angle. In order to study these items indetail, we performed a test experiment at theexternal site of the LNGS.An iron tower structure was built, allowing to
host detectors in three distinct slots under differentand variable iron thickness. To measure the flux ofcosmic ray muons and the corresponding angulardistribution, both relevant for intercalibrationpurposes, we made separate and independentmeasurements inserting emulsion sheets or scintil-lation counters inside the iron tower. Fluxeswere measured in both cases as a function of theiron thickness. Emulsions also allowed angularmeasurements for passing-through particles.Scintillators allowed some study of particle corre-lation in the time gate. A Monte Carlo (MC)simulation was performed, whose reliability wasvalidated and cross-checked with experimentaldata. Notice that the lower-energy tail of thecosmic radiation, more challenging for simulationcodes, but negligible in other experimentalconditions, will also play a role in the OPERAbrick-calibration case, and thus was carefullytaken into account.
In Section 2 the experimental setup is described.Details are provided about the scintillation coun-ter system as well as the emulsions and theirhandling. The automatic scanning procedure isalso described, with details about the sheetintercalibration and the raw data handling afterscanning. In Section 3, the MC shower generationand the methods adopted to simulate the particlepropagation and the detector response are pre-sented. In Section 4, the experimental results arepresented and compared to the expectations fromthe MC simulation. Conclusions are given inSection 5.Besides the relevance for the OPERA experi-
ment, these measurements are of some interest bythemselves, considering the lack of muon data atvarious atmospheric depths, especially at the lowerend of the energy spectrum.
2. Experimental setup
The experimental setup is shown in Fig. 1. Itconsists of three blocks of iron, 25 cm thick, withcross-sections 63� 63; 42� 42 and 25� 25 cm2;respectively, arranged in an inverse pyramidstructure, separated by air gaps of 10 cm wherescintillation counters or emulsion sheets arelocated. Additional iron slabs, 2.5 and 5 cm thick,could be added to the 25, 50 and 75 cm fixedblocks in order to obtain a finer binning. Thesupport skeleton is made of iron bars, weldedtogether and fixed on the ground. The whole
ARTICLE IN PRESS
E. Barbuto et al. / Nuclear Instruments and Methods in Physics Research A 525 (2004) 485–495 487
structure is hosted inside a hut, at the LNGSsurface laboratory, 1000 m a:s:l: The equipmentwas installed far from any other existing building,at the site where the cosmic-ray pit for OPERAwill be excavated.
2.1. Scintillation counters
We used four plastic scintillators BICRON BC-408, 11 cm� 11 cm� 1 cm; of the same type usedin the AMS experiment [5]. Three PMTs THORNEMI 9214B and one PMT PHILIPS XP 2020 werecoupled to the scintillators by means of fouroptical guides. The electronic consisted of a four-channel CAEN high-voltage supply, a scaler, twocoincidence units, and two discriminators, theCAEN 96 and the LeCroy 821B. The HV workingpoints were fixed each in the middle of thecorresponding counting plateau.Electron signals were studied and discriminated
from muon signals moving the four scintillatorsoutside the iron structure: three scintillators werepiled up one above the other, spaced 20 cm: Thefourth, used as a veto, was placed at the bottomsurrounded by 10 cm of lead. The discriminator’sthresholds were chosen to separate well theelectron signal from the muon signal.A similar setup was used to measure single
counter efficiencies. Three counters were piled up,nearly in contact. The fourth counter, shieldedby 10 cm of lead, was used to trigger through-going muons. In order to obtain the efficiencyof the second scintillator, we computed the ratiobetween four-fold and three-fold coincidences.The results are respectively, e1 ¼ ð98:970:3Þ%;e2 ¼ ð99:470:3Þ%; e3 ¼ ð99:470:3Þ%; e4 ¼
Table 1
Scintillation counter data
Test No. Iron (cm) Livetime (s) Counts/s (d
2 25 1:58� 105 ð1:4470:012 50 9:52� 105 ð3:7770:021 65 7:93� 104 ð1:7170:052 75 1:11� 106 ð1:0570:011 90 7:93� 104 ð0:6170:03
Only statistical errors are quoted; see text for the estimate of the syst
ð98:170:3Þ%; for the four counters (statisticalerrors only). By a simple MC simulation (seeSection 3), we checked that the small deviationsfrom the full efficiency are compatible with sub-millimeter alignment errors. Therefore, in thefollowing, we will assume intrinsic 100% efficiencyfor all the counters.A 50 ns gate defined the coincidence: this is long
enough to include the muon signal, but it alsoincludes two-fold spurious coincidences. There-fore, we required at least three-fold coincidencesfor each measurement.We performed two sets of measurements. The
first set in summer 2001, with four scintillatorsdisposed to form a ‘‘telescope’’ as in Fig. 1. In thesecond period (autumn 2002) we left two scintilla-tors nearly in contact in the upper part of thetower, and we iteratively moved a third scintillatorin the three slots; in this setup, each counter wassurrounded by 10 cm of lead in order to preventpossible lateral leakage of e.m. showers. Datataking parameters and counting statistics are givenin Section 4 (Table 1).
2.2. Nuclear emulsions
Since nuclear emulsions have continuous sensi-tivity, the latent images of tracks are accumulatedsince the plates are produced and until thephotographic development and fixation occur.Thus, in order to collect a truly local flux ofcosmic rays under fixed shielding conditions,plates should be freshly prepared on-site andprocessed soon after the exposure, as in theemulsion handling procedure reported in nextsubsection. Alternative approaches (e.g. forced
ata) Counts/s (MC) Ratio (data/MC)
Þ � 10�1 ð1:3970:01Þ � 10�1 1:0370:01Þ � 10�2 ð3:7570:02Þ � 10�2 1:0070:01Þ � 10�2 ð1:6770:01Þ � 10�2 1:0270:03Þ � 10�2 ð1:0970:01Þ � 10�2 0:9670:01Þ � 10�2 ð0:6470:01Þ � 10�2 0:9570:06
ematic uncertainties.
ARTICLE IN PRESS
E. Barbuto et al. / Nuclear Instruments and Methods in Physics Research A 525 (2004) 485–495488
fading of the latent images after transportation,flux subtraction based on plate rearrangement onsite, etc.), would be suitable for industriallyproduced emulsion plates.
2.2.1. Emulsion handling
Emulsion processing facilities were installed inthe LNGS underground lab. A hut of about14 m� 2 m was arranged into two independentsafe-light darkroom units. The equipment, toolsand procedures were inspired by successful appli-cations to large hybrid experiments at CERN [3,6].For the two separate exposure periods, plates
were prepared in the Gran Sasso cavern bypouring at 40�C the new sensitive gel producedby the Fuji company1 onto the two sides of aplastic base, 300 mm thick, precoated with gelatine.This new kind of high-quality ‘‘diluted’’ and ‘‘fine-grained’’ gel, suitable for the industrial productionof large quantities of nuclear emulsion plates, hasbeen made in cooperation with the NagoyaUniversity for the OPERA experiment. Aftersetting and drying, the thickness of each of thetwo emulsion layers ranged from 100 to 150 mm:Plates were cut to pieces of about 4� 6 cm2:Emulsions were then assembled into stacks of 3
sheets enveloped in light-tight bags of aluminatedpaper and placed in the exposure slots at theoutside iron structure. Two exposures were per-formed, lasting 48 and 74 days. The iron shieldingabove the stacks was 25, 50 and 75 cm in the firstexposure, in summer 2001, and 0, 5, 40, 70 and100 cm in the second exposure in the winter 2001–2002.After exposure, the plates were transported back
down to the cavern and developed in the dark-room inside the hut. A pattern of fiducial marks(grid) was printed on one side of each plate.The contamination from cosmics outside the
controlled exposure conditions can be neglected,since the transportations took less than 150 outsidethe coverage of the Gran Sasso rock and the plateswere kept vertically inside a Pb shielded box.After processing, the plates were found of
excellent photographic standard. However, thepacking procedure did not use, for this test, of
1Fuji Film, Minamiashigara, 250-0193, Japan.
any vacuum packing tightness. Moreover, despitethe temperature control of the outside exposurefacility, mechanical stresses from thermal varia-tions could not be avoided. This implies a looserplate-to-plate alignment, compared to past experi-ments and to what will be required in OPERA.
2.2.2. Automatic scanning of emulsion plates
After the important progress in the field ofautomatic scanning of nuclear emulsion plates[7,8], it is presently affordable to collect statisticsof several 105 charged particle tracks by fullyautomated optical microscopes. New-generation,very fast ðB20 cm2=hÞ microscopes are in pre-paration in Japan and in Europe for OPERA.The plates exposed to cosmic rays at LNGS
were scanned by an automated Nikon microscopeat the University of Rome La Sapienza, equippedwith a 30 Hz 1024� 1024 pixels CCD camera byHamamatsu and a MATROX Genesis framegrabber. A 50�magnification oil-immersion ob-jective and an optical adapter 0:6� were em-ployed, resulting in an effective field of view ofabout 250� 250 mm:The software for real-time two- and three-
dimensional pattern recognition [8], was providedby the Salerno Group [9]. It was designed in theframework of the CHORUS experiment. The on-line output consists of track segments detected oneach of the two emulsion sides, already correctedfor distortion and shrinkage. These corrections areneeded in order to refer to the original plategeometry at the exposure configuration. All tracksup to a dip angle of about 0:7 rad around thevertical direction (zenith angle) were recorded.Side-to-side connections are provided by quasi-on-line tasks at the end of data taking. Plate-to-plateconnections are based on off-line intercalibrationprocedures.Parameters for focus control, speed in depth and
stage displacements, and gray-level discriminationwere set at suitable values on each side, allowingfor a sampling of 20–30 layers across the emulsionin both cases. A fine-tuning was required to takeinto account unavoidable plate-to-plate opticalvariations. Rather large areas were acquired,ranging from 1.8 to 2:7 cm2: In these conditions,
ARTICLE IN PRESS
E. Barbuto et al. / Nuclear Instruments and Methods in Physics Research A 525 (2004) 485–495 489
the scanning speed was in the range of 0.5–1:0 cm2
per hour per side.To scan a set of test plates, of much smaller size
compared to the 35� 70 cm2 CHORUS plates,the large vacuum tight plate holder used inCHORUS was modified. However, with suchsmall plates, several focus faults and localizedinefficiencies on relevant fractions of the scannedareas were recorded, due to the imperfect flatness,the insufficient vacuum tightness and oil leakagesat the edges. Thus, occasionally the data takingwas iterated, overwriting low-quality data, and aneffective area was defined for the final data setafter the off-line analysis.The data handling started with an off-line
filtering of raw data, to eliminate low-qualitytrack segments (small number of grains, highdispersion). Then, a sequence of in-plate and plate-to-plate calibration steps was performed, to re-cover the original geometry of the stack at theexposure time. It turned out that a simpletranslation and rotation was not sufficient toproperly join corresponding areas of adjacentplates. Local numerical corrections were indeedrequired, mapped on 1 mm� 1 mm cells. Afterthat, a typical spread of less than 5 mm betweenextrapolated and measured track impact wasachieved. Also, the slopes measured on each side(segment slopes) were corrected globally and thanlocally by numerical offsets in order to obtain agood average agreement with the slopes measuredacross-the-base (base-track slopes). After the finalalignment, the angular resolution of base tracks,comparing corresponding tracks on two adjacentplates, was reduced to a few mrad (r.m.s.).After stack intercalibration, impact and slopes
of segments relocated on up to six surfaces wereavailable, as well as the global slopes of eachmulti-segment track. Only the tracks connected atleast across the base (base tracks) of one of thethree plates were retained as a clean subsample,and propagated across the next plates within strictacceptance cuts in impact and angle, whose effectswere also studied by MC. The available statisticsrange from 20 000 to 40 000 base tracks for eachexposure (including one-plate only, two- andthree-plate coincidences). For each stack, indivi-dual plate efficiencies were evaluated as a function
of the zenith angle, in order to obtain the final datagiven in Section 4.Better performances are expected for the forth-
coming tests at Gran Sasso, under a standardpacking procedure ensuring better mechanicalalignment of the plates and with the exploitationof the fast small-stage microscopes now underdevelopment, giving improved plate support andfocus control.
3. MC simulation
The simulation of the experimental setup wascoded in two steps, i.e. the generation of eventsand the propagation of particles through detec-tors. For the first step, we used two different eventgenerators: a full shower simulation code, whichfollows the shower development in the atmospheredown to the detection level, and a parametrizedgenerator, which produces only muons at sea level.The particle kinematics from the event generatorwas used as input of the second step, a detectorresponse simulation code based on GEANT3 [10].Some details are given below.
3.1. Full simulation
The COSMOS shower simulation code [11] wasused. This omni-purpose code for the generationand propagation of cosmic rays in a wide range ofenergies was extensively used for neutrino fluxcomputation [12].The package provides rigidity cutoff tables and
automatic solar activity, and geomagnetic fieldcomputations. Various hadronic interaction mod-els are used in different energy regions (Nucrin,Hadrin, Fritiof, DPM). Observation levels, fromsea level up to the top of the atmosphere, can beselected.We ran the code with the following parameters:
observation level at 1000 m a:s:l:; rigidity cut-offand geomagnetic field for LNGS site, solar activityof July 2001, shower generation for zenith angleyo70�; Middle Italy atmosphere. We fed thegenerator with a primary cosmic ray compositionresulting from the fit presented in [13].
ARTICLE IN PRESS
E. Barbuto et al. / Nuclear Instruments and Methods in Physics Research A 525 (2004) 485–495490
The number of primaries to be generated N canbe computed as a function of the livetime T andthe sampling surface S according to the equation
N ¼ STX
i
ZIiðE; y;fÞRiðE; y;fÞ cos y dO dE
ð1Þ
where IiðE; y;fÞ is the differential primary cosmicray flux for species i (i ¼ p; He, CNO, etc.) andRiðE; y;fÞ is the rigidity cutoff function (rangingbetween 0 and 1).We performed two distinct productions, without
and with iron shielding.In the first production, without iron shielding,
about 5:3� 105 primaries were generated, corre-sponding to an S � T factor of C300 m2 s: Allgenerated particles, with the exception of neutri-nos, were followed down to a threshold energy ofECOSMOSth ¼ 1 MeV: In Fig. 2, the differential
fluxes of the main components generated areshown. These results are in agreement withexperimental flux measurements found in theliterature (for a comprehensive data collectionsee [14]). For each particle, the positions wererandomly sampled at the emulsion surface,
10 -8
10 -7
10 -6
10 -5
10 -4
10 -3
10 -2
10 -1
1
10
1 10 102 103 104 105 10610-8
10-7
10-6
10-5
10-4
10-3
10-2
10-1
1
10
1 10 102 103 104 105 106
Kinetic Energy (MeV)
Ver
tica
l In
ten
sity
(cm
-2 s
-1 s
r-1 G
eV-1
)
neutrons
µ+,µ-
photons
e+,e-
Fig. 2. Differential fluxes of the main secondary components
generated by COSMOS. The break of the muon curve below
10 MeV is due to lack of statistics.
disregarding the shower event to which the particlebelong.The simulation of the exposures under the iron
tower shielding was the input both for emulsionstudies and for scintillator counter analysis. In thelatter case, the relative arrival time and position ofparticles in the same shower are relevant. Thus, wefollowed a more detailed sampling procedure. Inprinciple, the shower axis should be sampled in alarge area around the detector in order to includealso particles in the tail of the lateral distribution.This method however, turns out to be veryinefficient, since the lateral distribution of particleswith respect to the shower axis can be spread overseveral hundreds meters, while the detector cross-section is of the order of 1 m2: To overcome thisproblem, a variance reduction technique wasapplied. A grid of ‘‘detector replicas’’ was placedon the horizontal surface, while the shower axisposition was sampled on a central detector. Foreach event, all the particles were traced down andtheir intersections with other replicas were re-corded.Overall, about 1:7� 107 primaries were gener-
ated, corresponding to a ST value ofB10 000 m2 s (7 h of livetime). The thresholdenergy for secondary particles was set toECOSMOSth ¼ 100 MeV; the minimum energy re-
quired to cross the first iron slab.
3.2. Parametrized muon flux
A cosmic ray muon generator based on theinterpolation of the muon flux tables by the Bartolgroup [15] was used. These data start from 1 GeV:Thus, to extrapolate to the range 0.2–1 GeV; wetuned the shape of the energy spectrum with theone obtained with COSMOS. Data sampling wasperformed by attaching muon trajectories to a boxsurrounding the detector. Muons with energieslarger than 100 MeV and angles smaller than 70�
were generated. A correction for altitude depen-dence was applied according to Shibata’s fit.This code was used to cross-check the full MC
simulation and to estimate systematic uncertaintiesconnected with misalignment of the scintillationcounters.
ARTICLE IN PRESS
E. Barbuto et al. / Nuclear Instruments and Methods in Physics Research A 525 (2004) 485–495 491
3.3. Particle propagation and detector response
The two emulsion and scintillation exposures,including the lead surrounding, were simulated.All primary and secondary particles were followeddown to EGEANT
th ¼ 1 MeV; except neutrons,which were followed down to 10 MeV (theprobability to have a detectable signal below thisenergy is negligible, both for scintillation countersand emulsions).In the scintillator simulation, a linear relation
between energy loss and signal amplitude wasassumed. Thus, to reproduce the real setup, allcharged particles were recorded releasing anenergy larger than 20% of the muon signal(1:8 MeV in our counters). Coincidences weredefined as simultaneous crossing of particles insidea 50 ns window.As far as the emulsion simulation is concerned,
all charged particles crossing a stack wererecorded. For each particle, the six intercepts,above and below the bases of the three plates, werestored, to define three base track slopes. For plate-to-plate connections, we applied on these tracksthe same cuts of the real data analysis stream:base-tracks from different plates are connectedonly if Dslope p0:010þ 0:100�slope and Dimpactp20 mm: These cuts are effective for low-energyparticles, which suffer large scatterings. In addi-tion, a cut was applied on the particle energy lossdE=dx; in order to simulate the response ofemulsions to highly ionizing particles (e.g. non-relativistic protons). In the approximation of agrain density proportional to the energy loss, werequired a maximum grain density two timeslarger than a m.i.p.In the following we will refer to slope, position
and energy loss cuts as ‘‘Emulsion cuts’’.
4. Experimental results
Data collected with scintillation counters andwith emulsion plates are presented below, andseparately compared with corresponding MCsimulations. As explained in the previous sections,the geometry, acceptance, alignment procedureand so on were different in the two cases and they
have to be discussed separately. However, bothdata sub-sets fairly agree with the ‘‘same’’ MC,and make us confident about the simulation codeand the practical conclusions of the present test forOPERA purposes.
4.1. Scintillator data analysis
The scintillator data, in the two setups describedabove, are given in Table 1, together with MCpredictions. From the last column of the table, weobserve that the maximum deviation between dataand MC is about 5%.To estimate the error due to misalignment of
scintillators, large statistics were producedðB200 000 muonsÞ with the parametrized muonflux code. It turned out that a B1 cm shift in theðx; yÞ plane or in the vertical position z corre-sponds to counting differences of the order of 4%.In the first test, with the 90 cm configuration,
four-fold coincidences were used; the three-foldcoincidences give a B20% higher counting rate.The MC simulation reproduces this effect andshows that it is mainly due to muons that miss oneof the central counters because of multipleCoulomb scattering. In the second test, alignmenterrors are negligible, since two scintillators arenearly in contact. However, counting rates areaffected by neutrons: in the 25 cm configurationB35% of counts are due to proton recoils fromneutrons of B100 MeV: Moreover, simulationsindicate that no coincidences are due to multiplemuon events coming from the sides.In summary, MC simulations reproduce all the
configurations at B5% level, when componentsother than muons are accounted for. Multiplescattering and neutrons are correctly taken intoaccount, making us confident to use the simulationto interpret emulsion data. We partially explainthe remaining inconsistency between data and MCas due to alignment errors; atmospheric andseasonal effects could also play a role in the muoncounting statistics [16].
4.2. Emulsion data analysis
After plate-to-plate intercalibration, the emul-sion data consist of tri-dimensional penetrating
ARTICLE IN PRESS
E. Barbuto et al. / Nuclear Instruments and Methods in Physics Research A 525 (2004) 485–495492
tracks detected across plates, whose sides wereoriented along the South–North and the West–East directions. Thus, differential distributions asa function of the zenith and azimuthal angles areavailable, to be normalized to the effectivescanning area and to the exposure time. Afterefficiency corrections, these distributions, as wellas the integral flux up to some suitable zenithangle, can be compared with MC.The emulsion exposures and scanning results in the
zenith angular range of 20–300 mrad are summarizedand compared to Monte Carlo in Table 2.The emulsion fluxes were corrected for efficiency
as follows. Base-track counts were collected in20 mrad bins and in 0:5� 0:5 mm2 cells. If weconsider pairs of two adjacent plates, only cellswith large enough statistics of base-tracks were‘‘validated’’, as contributing to the effective scan-ning area. The counts in the validated cells weresummed-up for each angular bin. If N1 ðN2Þ is thenumber of tracks in the first (second) plate and NT
is the number of true tracks that crossed theemulsions, then N1ð2Þ ¼ e1ð2ÞNT; where e1ð2Þ is theefficiencies of the first (second) plate. On the otherhand, the number of tracks in both plates is N12 ¼e1e2NT; from which it follows that the number of‘‘true’’ tracks in each bin:
NT ¼N1N2
N12: ð2Þ
The ‘‘true’’ differential angular distributions andthe integral flux in a given range were thus
Table 2
Emulsion exposure parameters and scanning results in the zenith ran
Iron thickness (cm) Exposure time (day) Base tracks (k) (2
plate)
E
ðm
0 74 21.5 16
5 74 14.0 14
25 48 15.1 21
40 74 11.1 14
50 48 9.0 14
70 74 20.8 17
75 48 10.9 16
100 74 16.4 16
Base-track statistics refer to one-plate counts. Effective scanning area
mm2 � day are given in the last three columns (statistical errors on la
MC fluxes surviving the ‘‘emulsion cuts’’ are reported both for all ch
obtained. For each exposure, two differentestimates of the integral flux were available,for the first two plates out of three and forthe last two, respectively. The two computationsagreed to better than 10% (within 1% forthe stacks covered by 70, 75 and 100 cm iron).In the following, we report fluxes as weightedaverage of these two estimates, assigninga weight proportional to the effective scanningarea, and we assume a conservative 10% systema-tic error.The zenith angle region for integral flux studies
was confined to the interval 20–300 mrad becausethe pyramid structure provided limited angularcoverage in the lowermost positions, and the trackfinding and connection efficiency was poor above300 mrad. The cut below 20 mrad was introducedto remove systematic uncertainties about the‘‘truly vertical’’ direction (both at exposure andat scanning stages), and because of some opticaldefects faking small angle segments by instrumen-tal gray-level pile-up.Figs. 3 and 4 show the dip angle distributions
for different iron coverage thickness. Only statis-tical errors are indicated. A reduction of theparticle flux as a function of the iron thickness isobserved at wider dip angles. The OPERAintercalibration procedure is expected to takeadvantage of this ‘‘focussing’’ effect, since thesame number of useful tracks could be collectedwith a lower number of accompanying wide-angletracks contributing to the background.
ge 20–300 mrad
ffective scan. area
m2ÞMeas. flux (eff.
corr.) ðmm�2 d�1ÞMC flux with cuts
for all part (m only)
ðmm�2 d�1Þ
6 3.16 3.34 (2.49)
1 2.73 3.33 (2.45)
6 2.06 2.41 (2.11)
8 2.12 1.96 (1.82)
1 1.60 1.80 (1.68)
8 1.71 1.60 (1.50)
3 1.71 1.65 (1.49)
6 1.46 1.31 (1.19)
s correspond to all the well-mapped local areas. Integral fluxes/
st digit). Emulsion counts are corrected for efficiency (see text).
arged particles and for muons only.
ARTICLE IN PRESS
00.10.20.30.40.50.60.70.80.9
1
100 200 300Angle (mrad)
Par
ticl
es/m
m2 /d
ay
0 cm of Iron
00.10.20.30.40.50.60.70.80.9
1
100 200 300Angle (mrad)
Par
ticl
es/m
m2 /d
ay
5 cm of Iron
00.10.20.30.40.50.60.70.80.9
1
100 200 300Angle (mrad)
Par
ticl
es/m
m2 /d
ay
25 cm of Iron
00.10.20.30.40.50.60.70.80.9
1
100 200 300Angle (mrad)
Par
ticl
es/m
m2 /d
ay
40 cm of Iron
Fig. 3. Experimental angular distributions obtained with
emulsions (black circles) compared to MC simulation (white
circles) for 0, 5, 25 and 40 cm of iron coverage.
00.10.20.30.40.50.60.70.80.9
1
100 200 300Angle (mrad)
Par
ticl
es/m
m2 /d
ay 50 cm of Iron
00.10.20.30.40.50.60.70.80.9
1
100 200 300Angle (mrad)
Par
ticl
es/m
m2 /d
ay 70 cm of Iron
00.10.20.30.40.50.60.70.80.9
1
100 200 300Angle (mrad)
Par
ticl
es/m
m2 /d
ay 75 cm of Iron
00.10.20.30.40.50.60.70.80.9
1
100 200 300Angle (mrad)
Par
ticl
es/m
m2 /d
ay 100 cm of Iron
Fig. 4. Experimental angular distributions obtained with
emulsions (black circles) compared to MC simulation (white
circles) for 50, 70, 75 and 100 cm of iron coverage.
0
0.5
1
1.5
2
2.5
3
3.5
4
0 20 40 60 80 100
Iron thickness (cm)
Par
ticl
es/m
m2 /d
ay
Fig. 5. Experimental fluxes obtained with emulsions (black
circles), and simulated ones. Measurements are integrated in the
range 20–300 mrad: Statistical errors are quadratically summedwith systematical ones (10% for the data, 5% for the MC
predictions).
E. Barbuto et al. / Nuclear Instruments and Methods in Physics Research A 525 (2004) 485–495 493
Fig. 5 shows experimental and MC fluxes as afunction of the iron thickness. Statistical andsystematic errors are summed quadratically. Theoverall agreement between emulsion data and MCis fairly good.
To better appreciate the effects that emulsioncuts produce on different particles, we present inFig. 6 contributions of electrons, muons andproton recoils, with and without cuts. As expected,the cuts are particularly effective for low energyelectrons (at small iron coverage). A 40 cm of ironare the optimal choice: the muon flux is1:8 muons=mm2=day; and the eþe� contaminationis reduced to about 12%, with an average energy/Eeþe�SC8 MeV: When emulsion cuts are ap-plied, a residual eþe� flux of 0.02 eþe�=mm2=daysurvives. Above 40 cm; the muon flux continues todecrease almost linearly with the iron thickness,while the eþe� flux starts to increase. This effect ismainly due to eþe� regeneration from muonradiative processes inside the iron.We conclude that about 40 cm of iron is the best
choice for a pure iron shielding for OPERA.Different configurations with mixed (high- andlow-Z) materials will also be studied by MC in thenear future.Unfortunately, no ‘‘hardening’’ of the muon
energy spectrum is obtained by the iron shielding.This can be seen in Fig. 7, where we present the
ARTICLE IN PRESS
0
0.5
1
1.5
2
2.5
3
3.5
4
0 20 40 60 80 100
Iron thickness (cm)
Par
ticl
es/m
m2 /d
ay
µ+,µ- (no cuts)
µ+,µ- (cuts)
e+,e- (no cuts)
e+,e- (cuts)
p recoils (no cuts)
p recoils (cuts)
Fig. 6. Contribution of different particle species to the integral
distribution (MC simulation). Electron, muon and proton
components are shown, with and without emulsion cuts. Notice
that for Fe thicknesses above 40 cm the electron component
starts to increase.
10-8
10-7
10-6
10-5
10-4
10-3
10-2
10-1
1
10
1 10 102 103 104 105 10610
-8
10-7
10-6
10-5
10-4
10-3
10-2
10-1
1
10
1 10 102 103 104 105 106
Kinetic Energy (MeV)
Ver
tica
l In
ten
sity
(cm
-2 s
-1 s
r-1 G
eV-1
)
µ+,µ-
e+,e-
Fig. 7. Energy spectra of electron and muon components with
and without iron coverage. We show the spectra without any
coverage (thin lines), with 40 cm of iron (bold lines) and with
40 cm of iron and with emulsion cuts (bold dashed lines).
Fig. 8. Azimuthal distributions of emulsion data under
different iron coverages (units are counts mm�2 d�1).
E. Barbuto et al. / Nuclear Instruments and Methods in Physics Research A 525 (2004) 485–495494
original spectra of electrons and muons (thin lines)and the spectra after 40 cm of iron (bold lines).The average energy ð/EmSCEmedianC2 GeVÞ ofthe muon spectrum is unchanged. It is worth tonote that almost all electrons under 50 cm of ironare due to regeneration phenomena.Finally, emulsion data as function of the
azimuthal angle are shown in Fig. 8. A sinusoidalshape is well visible for the lower Fe thicknesses (0and 5 cm). The effect seems still visible at 25 cm;and vanishes at 50 cm and above (hard muons). Apossible interpretation of this effect is the influenceof Earth magnetic field on very low-energy muons.The ratio of the number of events with Nðf >0Þ=Nðfo0Þ should reflect the charge ratio whichdeviates from 1 as energy increases.
5. Conclusions
For the purposes of the optimal choice of the Feshielding for OPERA, it is clear that (a) a few cmFe allow the absorption of soft e.m. showerparticles, (b) the dependence of the cosmic ray
ARTICLE IN PRESS
E. Barbuto et al. / Nuclear Instruments and Methods in Physics Research A 525 (2004) 485–495 495
flux on the Fe shielding is not linear in the wholerange, and it is steeper until some very softcomponent is absorbed, (c) after about 40 cm;there is no real gain to increase the thickness, sinceaccording to MC no hardening of the muonspectrum is induced, at the expenses of a lowerflux requiring a longer exposure time. Moreover,eþe� contamination has a minimum at 40 cm ofiron, above which regeneration processes starts todominate. Thus, the Fe shielding can be fixed at40 cm: The observed focussing toward narrow dipangles will be beneficial for OPERA.Proton recoils from B100 MeV cosmic ray
neutrons is a non-negligible component. We planto perform some fine-tuning studies in order toexplore the performances of a mixed-materialshielding (e.g. iron followed by aluminum) andto optimize the cosmic-ray pit geometry.In perspective, better and more complete results
could be obtained by exploiting the OPERA ECC(full bricks) technique (e.g. low-energy muonspectra, electron identification, etc.) and up-to-date advanced microscopy. Some of these studiesare planned for the near future, and will be basedon the reliable MC code developed for thepurposes of the present work.
Acknowledgements
The authors wish to warmly thank the staff ofTechnicians from the INFN Sections of Bari(P. Di Pinto, V. Di Pinto and A. Andriani),Bologna (L. Degli Esposti, V. Togo, D. Di
Ferdinando, C. Valieri) and Rome (R. Diotallevi)for their excellent contribution to the setting-up ofthe multi-purpose facilities underground and atsurface, and the exploitation of the experimentalactivity reported in this paper. The help andsupport by the LNGS services and workshops isgratefully acknowledged. We also acknowledge S.Cecchini for fruitful discussions about cosmic rayphysics.
References
[1] G. Acquistapace, et al., CERN 98-02, INFN/AE-98/05,
1998.
[2] M. Guler, et al., The OPERA Collaboration, CERN-
SPSC-2000-028, CERN-SPSC-P-318, LNGS-P25-00, 2000;
CERN-SPSC-2001-025, CERN-SPSC-M-668, LNGS-
EXP-30-2001-ADD-1, 2001.
[3] S. Aoki, et al., Nucl. Instr. and Meth. A 447 (2000) 361.
[4] K. Kodama, et al., Phys. Lett. B 504 (2001) 218.
[5] S.P. Ahlen, et al., Nucl. Instr. and Meth. A 350 (1994) 351.
[6] K. Hoshino, G. Rosa, Nucl. Tracks 12 (1986) 477.
[7] T. Nakano, Ph.D. Thesis, Nagoya University, 1997.
[8] G. Rosa, et al., Nucl. Instr. and Meth. A 401 (1997) 7.
[9] C. Bozza, Ph.D. Thesis, Salerno University, 2000.
[10] R. Brun, et al., Detector Description and Simulation Tool,
V. 3.21, CERN Program Library, 1994, unpublished.
[11] http://eweb.b6.kanagawa-u.ac.jp/kasahara/
ResearchHome/cosmosHome/index.html.
[12] M. Honda, et al., Phys. Rev. D 52 (1995) 4985.
[13] T.K. Gaisser, et al., Primary spectrum to 1 TeV and
beyond, Proceedings of 27th ICRC, Hamburg, 2001.
[14] P.K.F. Grieder, Cosmic Rays at Earth, Elsevier Science,
Amsterdam, 2001.
[15] V. Agrawal, et al., Phys. Rev. D 53 (1996) 1314.
[16] O.C. Allkofer, H. Jokisch, Nuovo Cimento A 15 (1973)
371.