High statistics measurement of the underground muon pair separation at Gran Sasso The MACRO Collaboration M. Ambrosio12, R. Antolini7, C. Aramo7,n, G. Auriemma14,a, A. Baldini13, 9 9 9 G. C. Barbarino12, B. C. Barish4, G. Battistoni6,b, R. Bellotti1, C. Bemporad 13, 1 n E. Bernardini 7, P. Bernardini 10, H. Bilokon 6, V. Bisi 16, C. Bloise 6, C. Bower 8, a J S. Bussino14, F. Cafagna1, M. Calicchio1, D. Campana 12, M. Carboni6, M. Castellano1, 8 1 S. Cecchini2,c, F. Cei 11,13, V. Chiarella6, B. C. Choudhary4, S. Coutu 11,o, G. De Cataldo1, 1 v H. Dekhissi 2,17, C. De Marzo1, I. De Mitri9, J. Derkaoui 2,17, M. De Vincenzi14,e, 7 2 A. Di Credico 7, O. Erriquez1, C. Favuzzi1, C. Forti6, P. Fusco1, G. Giacomelli 2, 0 1 G. Giannini13,f, N. Giglietto1, M. Giorgini 2, M. Grassi 13, L. Gray 4,7, A. Grillo 7, 0 9 9 F. Guarino 12, C. Gustavino 7, A. Habig 3, K. Hanson 11, R. Heinz 8, Y. Huang4, E. Iarocci / x e 6,g, E. Katsavounidis4, I. Katsavounidis4, E. Kearns 3, H. Kim4, S. Kyriazopoulou4, - p E. Lamanna 14, C. Lane 5, T. Lari 7, D. S. Levin 11, P. Lipari 14, N. P. Longley 4,l, e h : M. J. Longo 11, F. Loparco 7, F. Maaroufi 2,17, G. Mancarella 10, G. Mandrioli 2, v i X S. Manzoor 2,m, A. Margiotta Neri 2, A. Marini 6, D. Martello 10, A. Marzari-Chiesa 16, r a M. N. Mazziotta1, C. Mazzotta 10, D. G. Michael4, S. Mikheyev 4,7,h, L. Miller 8, P. Monacelli 9, T. Montaruli1, M. Monteno 16, S. Mufson 8, J. Musser 8, D. Nicol´o13,d, C. Orth 3, G. Osteria 12, M. Ouchrif 2,17, O. Palamara 7, V. Patera 6,g, L. Patrizii 2, R. Pazzi 13, C. W. Peck4, S. Petrera 9, P. Pistilli 14,e, V. Popa 2,i, A. Rain`o1, A. Rastelli7, J. Reynoldson7, F. Ronga 6, U. Rubizzo 12, C. Satriano 14,a, L. Satta 6,g, E. Scapparone 7, K. Scholberg 3, A. Sciubba 6,g, P. Serra-Lugaresi 2, M. Severi 14, M. Sioli 2, M. Sitta 16, P. Spinelli1, M. Spinetti 6, M. Spurio 2, R. Steinberg5, J. L. Stone 3, L. R. Sulak 3, A. Surdo 10, G. Tarl`e11, V. Togo 2, D. Ugolotti 2, M. Vakili 15, C. W. Walter 3 , and R. Webb 15. 1. Dipartimento di Fisica dell’Universit`a di Bari and INFN, 70126 Bari, Italy 1 2. Dipartimento di Fisica dell’Universit`a di Bologna and INFN, 40126 Bologna,Italy 3. Physics Department, Boston University, Boston, MA 02215,USA 4. California Institute of Technology, Pasadena,CA 91125, USA 5. Department of Physics, Drexel University, Philadelphia, PA 19104, USA 6. LaboratoriNazionali di Frascati dell’INFN, 00044 Frascati (Roma), Italy 7. LaboratoriNazionali del Gran Sasso dell’INFN, 67010 Assergi (L’Aquila), Italy 8. Depts. of Physics and of Astronomy, Indiana University, Bloomington, IN 47405,USA 9. Dipartimento di Fisica dell’Universit`a dell’Aquila and INFN, 67100 L’Aquila, Italy 10. Dipartimento di Fisica dell’Universit`a di Lecce and INFN, 73100 Lecce, Italy 11. Department of Physics, University of Michigan, Ann Arbor, MI 48109,USA 12. Dipartimento di Fisica dell’Universit`a di Napoli and INFN, 80125 Napoli, Italy 13. Dipartimento di Fisica dell’Universit`a di Pisa and INFN, 56010 Pisa, Italy 14. Dipartimento di Fisica dell’Universit`a di Roma “La Sapienza” and INFN, 00185 Roma, Italy 15. Physics Department, Texas A&M University, College Station, TX 77843, USA 16. Dipartimento di Fisica Sperimentale dell’Universit`a di Torino and INFN, 10125 Torino, Italy 17. L.P.T.P., Faculty of Sciences, University Mohamed I, B.P. 524 Oujda, Morocco a Also Universit`a della Basilicata, 85100 Potenza, Italy b Also INFN Milano, 20133 Milano, Italy c Also Istituto TESRE/CNR, 40129 Bologna, Italy d Also Scuola Normale Superiore di Pisa, 56010 Pisa, Italy e Also Dipartimento di Fisica, Universit`a di Roma Tre, Roma, Italy f Also Universit`a di Trieste and INFN, 34100 Trieste, Italy g Also Dipartimento di Energetica, Universit`a di Roma, 00185 Roma, Italy h Also Institute for Nuclear Research, Russian Academy of Science, 117312 Moscow, Russia i Also Institute for Space Sciences, 76900 Bucharest, Romania l Swarthmore College, Swarthmore, PA 19081, USA m RPD, PINSTECH, P.O. Nilore, Islamabad, Pakistan n Also INFN Catania, 95129 Catania, Italy 2 o Also Department of Physics, Pennsylvania State University, University Park,PA 16801,USA Abstract We present a measurement of the underground decoherence function using multi-muon events observed in the MACRO detector at GranSasso at an average depth of 3800 hg/cm 2. Muon pair separations up to 70 m have been measured, corresponding to parent mesons with P 1 2 GeV/c. Improved selection criteria are used to reduce detector effects ⊥ ≤ ÷ mainly in the low distance separation region of muon pairs. Special care is given to a new unfolding procedure designed to minimize systematic errors in the numerical algorithm. The accuracy of the measurement is such that the possible contribution of rare processes, such as µ± +N µ± +N +µ+ +µ−, can be experimentally studied. → The measured decoherence function is compared with the predictions of the hadronic interaction model of the HEMAS Monte Carlo code. Good agreement is obtained. We interpret this agreement to indicate that no anomalous P components in soft hadron- ⊥ Nucleus and Nucleus–Nucleus collisions are required by the MACRO experimental data. Preliminary comparisons with other Monte Carlo codes point out that the uncertainties associated with the hadronic interaction model may be as large as 20%, depending on the energy. MACRO data can be used as a benchmark for future work on the discrimination of shower models in the primary energy region around and below the knee of the spectrum. I. INTRODUCTION The knowledge of hadronic interaction processes plays a fundamental role in studies of cosmic rays in the VHE–UHE range (1012 eV E 1017 eV). In particular, the interpre- ≤ ≤ tation of indirect measurements intended to determine the features of primary cosmic rays, such as spectra and composition, depends on the choice of the hadronic interaction model adopted in the description of the atmospheric shower development. For instance, muons 1 observed by deep underground experiments are the decay products of mesons originating mostly in kinematic regions (high rapidity and high √s) not completely covered by existing collider data. The problem is particularly important for nucleus-nucleus interactions for which available data extend only to a few hundreds of GeV in the laboratory frame. It is therefore crucial to find physical observables which are primarily sensitive to the assumed interaction model rather than to the energy spectra and chemical composition of primary cosmic rays. The shape of the muon lateral distribution is well-suited for this purpose. In particular it allows the study of the transverse structure of hadronic interactions, which is one of the most relevant sources of uncertainties in the models [1]. In fact, different aspects of the interactions contribute to the lateral distribution. We can qualitatively understand this by simple arguments, valid in a first order approximation. Let us consider a single interaction of a primary nucleon of total energy E , producing mesons of energy Eπ,K with transverse 0 momentum P , at a slant height H , which eventually decay into muons. Calling r the ⊥ prod separation of a high energy muon (i.e. moving along a straight line) from the shower axis, we have: P ⊥ r H . (1.1) ∼ Eπ,K prod Inthissimplifieddescriptionweareneglectingthetransverse momentumintheparentdecay. The previous expression can be written in a more instructive way, considering that at high energy, apart from terms of the order of (m /E )2, the longitudinal c.m. variable x is T 0 F approximately equal to the laboratory energy fraction: P P r ⊥ H ⊥ logσinel +const. . (1.2) ∼ xπ,KE prod ∝ xπ,KE n−Air F 0 F 0 (cid:16) (cid:17) The assumption of an exponential atmosphere has been used in the last expression. It can be seen how the transverse and longitudinal components of the interaction, as well as 2 the inclusive and total cross sections, convolve together (with different weights) to yield the lateral separation. The role of P remains a dominant one in determining the relative ⊥ separation of the muon component by introducing a loss of collinearity (“decoherence”) with respect to the direction of the shower axis. A qualitative extension to the case of nuclear projectiles can be made within the frame- work of the superposition model, where each nucleon of the projectile of mass number A is assumed to interact independently with energy E /A. Further refinements are needed to 0 account for modifications in the P and x distributions deriving from the nuclear struc- ⊥ F ture of projectile and target, as will be discussed later. A reliable evaluation of the lateral distribution function can be obtained only by Monte Carlo methods. Deep underground experiments are capable of selecting atmospheric muons in the TeV range produced in the initial stages of the extensive air shower (EAS) development. They can perform a measurement of muon separation which is highly correlated to the lateral dis- tribution. Since the shower axis position is not usually known, the distribution of muon pair separation in multimuon events is studied. Muons associated with the same events, coming in general from different parent and shower generations, are grouped together. Further- more, a wide range of primary energy is integrated in the same distribution. It is generally assumed, and supported by many simulations, that the shape of this distribution is only slightly affected by the mass composition of primaries [2], thus preserving the sensitivity to the interaction features. As an example, in Fig. 1 we show the dependence of the average pair separation, as detected at the depth of the underground Gran Sasso laboratory, with respect to the P of the parent mesons and to their production slant height in the at- ⊥ h i mosphere. These have been calculated by means of the HEMAS Monte Carlo code [3] for a mixed primary composition [10]. This code employs an interaction model based on the results of the experiments at hadron colliders. The decoherence function as measured in an underground experiment is also affected by 3 multiple scattering in the rock and, to some extent, geomagnetic deflection. For a detector with geometrical acceptance A(θ,φ), for zenith and azimuthal angles θ and φ, respectively, we define the decoherence function as the distribution of the distance between muon pairs in a bundle: dN 1 1 d2N(D,θ,φ) = dΩ, (1.3) dD ΩT A(θ,φ) dDdΩ Z where N(D,θ,φ) is the number of muon pairs with a separation D in the direction (θ,φ), Ω is the total solid angle covered by the apparatus and T is the total exposure time of the experiment. A muon bundle event of multiplicity N will contribute with a number of µ independent pairs N = N (N 1)/2. µ µ − In principle, a decoherence study can be performed without a single large area detector, and in early attempts the muon lateral separation was studied via coincidences between two separate movable detectors [4]. The advantage offered by a single large area detector is the ability to study the features inherent in the same multi-muon event, such as higher order moments of the decoherence distribution [5]. The large area MACRO detector [6] has horizontal surface area of 1000 m2 at an ∼ average depth of 3800 hg/cm2 of standard rock (E 1.3 TeV) and is naturally suited µ ≥ for this kind of measurement. An analysis of the muon decoherence has already been per- formed [7,8,9,10]. The bulk of multiple muon events in MACRO corresponds to a selection of primary energies between a few tens to a few thousands of TeV/nucleon. Hadronic in- teractions and shower development in the atmosphere were simulated with the previously noted HEMAS code. In particular, a weak dependence on primary mass composition was confirmed for two extreme cases: the “heavy” and “light” composition models [11]. The MACRO analysis was designed to unfold the true muon decoherence function from the measured one by properly considering the geometrical containment and track resolution ef- ficiencies. This procedure permits a direct comparison between measurements performed by 4 different detectors at the same depth, and, more importantly, whenever new Monte Carlo simulations are available, allows a fast comparison between predictions and data without the need to reproduce all the details of detector response. The first attempt, obtained while the detector was still under construction, and there- fore with a limited size, was presented in [7]. The same analysis, with a larger sample based on the full lower detector, was extended in [8]. With respect to the HEMAS Monte Carlo expectations, these results indicated a possible excess in the observed distribution at large separations. In Ref. [10] we presented the decoherence distribution without the unfolding procedure; the claimed excesses were not confirmed. In order to reach more definitive con- clusions, a more careful analysis of the systematics associated with the unfolding procedure was considered necessary. A detailed discussion of this item will be addressed in Section 4. A more careful discussion of the Monte Carlo simulation is also necessary. The bulk of the muon bundles collected by MACRO are low multiplicity events, coming from parent mesons in the far forward region of UHE interactions, not easily accessible with collider experiments. This requires an extrapolation to the highest energies and rapidity regions, introducing possible systematic uncertainties. For instance, some doubts have been raised [1] concerning the treatment of meson P in HEMAS. In the HEMAS hadronic interaction ⊥ code, secondary particle P depends upon three different contributions: ⊥ P increases with energy, as required by collider data in the central region; ⊥ • h i P increases in p–Nucleus and Nucleus–Nucleus interactions, relative to that for pp ⊥ • h i collisions, according to the “Cronin effect” [12]; P varies with x , according to the so called “seagull effect” [13]. ⊥ F • h i The sum of these effects yields some doubt about a possible overestimate of P for energetic ⊥ secondary particles, an hypothesis recently restated in [14]. It is therefore crucial to perform 5 ahighprecisiontestofthetransversestructureofthismodel,sinceitaffectsthecalculationof containment probability for multiple muon events and, consequently, the analysis of primary cosmic ray composition [9,10]. Inthispaper, anewanalysisoftheunfoldeddecoherencefunctionispresented, performed with improved methods up to 70 m. The present work enlarges and completes the data analysis presented in [9,10]. Preliminary results of this unfolding procedure [15] showed an improved agreement between experimental data and Monte Carlo predictions. Particular attention is paid to the small-separation (D 1 m) region of the decoherence ≤ curve, in which processes such as muon-induced hadron production can produce a back- ground to the high energy muon analysis. At the energies involved in the present analysis (E 1 TeV), moreover, muon-induced muon pair production in the rock overburden could µ ≥ yield an excess of events with small separation, as suggested in [16]. This process is usu- ally neglected in Monte Carlo models commonly adopted for high energy muon transport [3,17,18,19]. Section 2 is devoted to the description of the detector and of data analysis, with a focus on new event selection criteria. In Section 3 the features of the Monte Carlo simulation are presented together with the comparison between experimental and simulated data in the MACRO detector, while Section 4 is dedicated to the unfolding procedure. A comparative discussion ofthefeaturesofdifferent hadronicinteractionmodels issummarized inSection 5. InSection6,theproblemexisting inthefirstbinsofthedecoherencedistributionispresented in detail, testing new hypotheses on its origin. Conclusions follow in Section 7. II. DETECTOR DESCRIPTION AND DATA ANALYSIS The MACRO detector [6], located in hall B of the Gran Sasso Laboratory, is a large area detector equipped with streamer tube chambers, liquid scintillation counters and nuclear 6 track detectors arranged in a modular structure of six “supermodules”. Each of these is 12 m 12 m 9 m in size and consists of a 4.8 m high lower level and a 4.2 m upper “attico”. × × In this paper only data from the lower level of the apparatus are included; therefore only the lower detector will be described further. Tracking is performed by means of limited streamer tubes, which are distributed in ten horizontal planes separated by 60 g cm−2 of CaCO (limestone rock) absorber, and in six 3 ∼ planes along each vertical wall. The streamer tubes have a square cross section of 3 3 cm2, × and are 12 m long. From each plane two coordinates are provided, the wire (perpendicular to the long detector dimension) and strip views. The latter employs 3 cm wide aluminum strips at 26.5◦ to the wire view. The average efficiencies of the streamer tube and strip systems were 94.9% and 88.2% respectively, in the period of this analysis. The spatial resolution achieved with this configuration depends on the granularity of the projective views. The average width of a cluster, defined as a group of contiguous muon ”hits,” is 4.5 cm and 8.96 cm for the wire and strip views, respectively. Muon track recognition is performed by an algorithm which requires a minimum number of aligned clusters (usually 4) through which a straight line is fit. The differences between the cluster centers and the fit determine a spatial resolution of σ =1.1 cm for the wire view and W σ =1.6 cm for the strip view. These resolutions correspond to an intrinsic angular resolution S of 0.20 for tracks crossing ten horizontal planes. In reconstructing the best bundle configuration, the tracking package flags track pairs as parallel, overlapping, or independent and not parallel. This is achieved in two steps, in each projective view: two tracks are defined as parallel if their slopes coincide within 2 σ or if their angular • separation is less than 30 (60 if the tracks contain clusters whose widths exceed 30 cm). Otherwise, the track pair is flagged as independent and not parallel if its distance 7 separation is larger than 100 cm. tracks at short relative distance are labelled as overlapping if their intercepts with the • detector bottom level coincide within 3.2 σ (2 σ if their angular separation is < 1.50). The routine chooses the most likely bundle as the set having the largest number of parallel tracks and the largest number of points per track. Subsequently, tracks flagged as not parallel are considered in order to include fake muon tracks originated primarily by hadrons or δ-rays in the surrounding rock or inside the detector. A two-track separation of the order of 5 cm is achieved on each projective view. However, this capability can be substantially worsened in case of very large, but rare, catastrophic energy losses of muons in the detector. Only tracks with a unique association in the two views can be reconstructed in three dimensional space. At this level, pattern recognition is used to require a complete matching between tracks belonging to different projective views. This is automatically achieved when two tracks pass through separate detector modules. When they are in the same module, matching of hit wires and strips on the same detector plane is accomplished by taking advantage of the stereo angle of the strips with respect to the wires. In some cases the track pattern correspondence between the two views is also used. The possibility to analyse muon decoherence in three dimensional space is important to have an unbiased decoherence distribution. However, the unambiguous association of muon tracks from the two projective views cannot be accomplished for high multiplicity events because, in events characterized by a high muon density, the tracking algorithm is not able to resolve the real muon pattern without ambiguities, especially when tracks are superimposed. In Ref. [9,10] we presented the muon decoherence function in the wire view alone, which allowed the extension of the analysis to higher multiplicities. We have analyzed about 3.4 105 events, corresponding to a 7732 hr live time for the · lower part of the apparatus. These events were submitted to the following selection criteria: 8