Direct measurement of galactic cosmic ray fluxes with the orbital detector AMS-02

Transcription

1 UNIVERSITÀ DEGLI STUDI DI BOLOGNA FACOLTÀ DI SCIENZE MATEMATICHE FISICHE E NATURALI DOTTORATO DI RICERCA IN FISICA, XV CICLO PhD Thesis Direct measurement of galactic cosmic ray fluxes with the orbital detector AMS-02 DIEGO CASADEI Advisor: Chiar.mo Prof. ANDREA CONTIN PhD Coordinator: Chiar.mo Prof. GIOVANNI VENTURI Bologna, Italy, March 10, 2003

2

3 Introduction The Alpha Magnetic Spectrometer (AMS) experiment is a high energy particle detector developed to measure cosmic ray fluxes outside the Earth atmosphere. The first version of the detector, called AMS-01, successfully flew aboard of the shuttle Discovery on June 2 12, 1998 (NASA STS-91 mission), collecting over one hundred million events. The next version of the detector, called AMS-02, will be installed on the International Space Station (ISS) Alpha at the end of 2005, where it will operate for at least three years. Cosmic rays (CR) are charged (and neutral) particles coming from the outer space that span a very wide energy range (from few MeV up to ev), and are believed to be accelerated by supernova explosions in our Galaxy for energy below ev per nucleon. For higher energies their acceleration sites may be other galactic objects (like pulsars), whereas for the highest energies an extragalactic origin seems to be necessary to explain the uniform distribution of the CR incoming direction. Their global spectrum is described in chapter 1, where the model of charged particle acceleration by supersonic shock waves is sketched. In addition, this chapter describes the particles diffusion inside the Galaxy and inside the eliosphere, where they may finally reach the Earth. The cosmic rays composition is very rich: in addition to protons, that constitute about 90% of the total number of particles, He nuclei (about 10%), electrons (about 1%), positrons and all other stable and instable nuclei are present, with the exception of the isotopes with a very short life time. Chapter 2 shows that the relative abundances with respect to solar system and galactic averages may be explained by the different volatility of the elements, being favored the acceleration of less volatile elements. In addition, this chapter reviews the CR electron and positron Physics, emphasizing the local nature of the direct measurements in contrast with proton measurements, that are more representative of a galactic average. The main motivation to support the AMS experiment is the search of primary iii

4 INTRODUCTION antimatter in cosmic rays. The experimental tests carried on with accelerators affirm that every time we create new fermions we must create the same amount of anti-fermions (more precisely, each reaction must conserve separately the baryonic and leptonic numbers), as reviewed in chapter 3. On the other hand, astronomic measurements show that we live in a homogeneus domain composed entirely of matter, at least up to the galaxy clusters scale, immersed in a thermal bath of photons (the cosmic microwave background). These photons are considered the relics of the matter-antimatter annihilation epoch, where the radiation decoupled from the matter. An important question is whether the cosmic baryogenesis was matter-antimatter symmetric, or some form of symmetry breaking happened, violating the laws of the standard model. In addition to large homogeneus antimatter domains, that could have been created as fluctuations over a globally symmetric scenario in the same way as our matter domain, primordial antimatter could have survived in the form of globular clusters of antistars, that may even be present in our Galaxy. If the homogeneus domains are not too distant from the Milky Way, or if antistars exist in the Galaxy, we could detect antinuclei in the solar system (the probability to create them during CR collision with the interstellar medium is negligible). The most probable antinucleus is antihelium, but also anticarbon, antinitrogen or antioxygen could be produced by antistars: their detection would be considered the proof that antimatter domains do exist indeed. Antiprotons and positrons are commonly created during the cosmic ray interactions with the interstellar medium, hence they are not good candidates for the search of primordial antimatter. Howewer, their purely secondary spectrum can be accurately predicted, and experimental data can be checked to see if exotic sources exist. For example, proton-antiproton and electron-positron pairs can be created by the annihilation of the dark matter candidates, like supersymmetric particles. The dark matter is not directly visible but has important gravitational effects: the energy density of the universe is due for 70% to the vacuum constant and for 30% to the matter, while the radiation gives a negligible contribution. On the other hand, the baryonic fraction of the matter component in the universe can not be higher than few percent, in order to be consistent with the primordial nucleosynthesis model. Hence, the majority of the matter component is due to non baryonic (i.e. not strongly interacting) dark (i.e. not charged) particles. Collider data exclude candidates with mass lower than roughly hundred GeV/c 2, hence their annihilations may produce a bump in the spectrum of antiprotons and positrons at high energies, where no data are presently available. The AMS-02 detector would iv

5 INTRODUCTION be able to make a precise measurement up to 1 2 TeV, thus having the potential to discover the first signatures of new particles. Chapter 4 illustrates the working principles of AMS and describes the AMS- 01 and AMS-02 detectors, built by an international Collaboration where the Italian contribution is very important: the time of flight (TOF) system is completely developed and built in the INFN laboratories in Bologna, that is also partecipating in the proximity focusing ring imaging Čerenkov (RICH) detector; the tracker is mainly developed by the INFN Perugia group; the electromagnetic calorimeter (ECAL) is mainly developed by the Pisa and Siena INFN groups; the Roma 1 INFN people partecipate in the development of the transition radiation detector (TRD) electronics. I have been working on AMS in the Bologna group since 1996, when I started following the development and construction of the AMS-01 TOF system. During the last three years I worked on the AMS-02 TOF and RICH systems, and on the analysis of AMS-01 data. In particular, I collaborated with the mechanical engineers for the design and positioning of the light guides and for the choice of the shape of the scintillator counters of TOF, following the constraints imposed by the high magnetic field in the photomultipliers zone, by the geometrical acceptance and the weight budget, as described in chapter 5. In parallel, I worked on the conical mirror of the RICH subdetector, starting with samples of different types and finally studying the reflectivity of the mandrel that is used to build the mirror. In addition I partecipated to the calibration of the multi-pixel phototubes used by the RICH, and worked on the problem of the magnetic shielding of these photomultipliers, as described in chapter 6. Recently (October 2002) I organized the ion beam test at CERN SPS for the TOF group. The test was requested by the RICH Collaboration, and the RICH prototype was the main experiment. Howewer, parasitic detectors could be accomodated in the same area, allowing for a combined test of a tracker prototype and a couple of TOF counters. Preliminar results of this ion beam test are presented in chapter 7. Presently I organize the work of the development of the S-crate electronics, that contains the front-end boards of TOF and ACC (the anticoincidence system), the high voltage power supplies, and the slow control boards. Bologna, March 2003 Diego Casadei v

6 vi INTRODUCTION

7 Contents Introduction iii 1 Cosmic rays below 1 TeV High energy particles from space Cosmic ray acceleration Shock waves Particle acceleration by shock waves Interstellar propagation Propagation inside the eliosphere Solar modulation Geomagnetic cutoff Cosmic ray composition Cosmic ray ions CR source composition Propagation models Measurements of CR ions spectra Cosmic ray electrons Energy losses Cosmic antimatter Antimatter CPT theorem Anti-systems Cosmic antimatter Can total annihilation be avoided? Antimatter domains vii

8 CONTENTS 3.3 Can we detect cosmic antimatter? Indirect ways Direct detection Antimatter cosmic rays Positrons and antiprotons Antihelium and antinuclei The AMS experiment Particle identification Rigidity measurement Charge measurement Velocity measurement The AMS-01 detector The magnet The tracking system The time of flight system The anticoincidence system The threshold Čerenkov counter The AMS-01 trigger Results of the STS-91 mission Primary CR spectra Secondary spectra The AMS-02 detector The magnet The tracking system The time of flight system The anticoincidence system The RICH detector The trasition radiation detector The electromagnetic calorimeter The AMS-02 TOF system The time of flight system of AMS Magnetic field Mechanical design The TOF planes The TOF counters viii

9 CONTENTS 5.4 Electronics Data acquisition Slow control Powering The beam test Selection cuts Charge peaks Time resolution The AMS-02 RICH subdetector The Čerenkov emission Particle energy loss in the radiator Čerenkov radiation The RICH design Large acceptance proximity focusing RICH detector Radiator Mirror Photomultipliers and light guides Magnetic shielding The beam test The RICH prototype The radiators Ion beam test at CERN SPS Experimental setup RICH prototype Tracker prototype TOF counters Preliminary results A Abbreviations 163 B TOF PMT positioning and field 165 Bibliography 178 ix

10 x CONTENTS

11 Chapter 1 Cosmic rays below 1 TeV 1.1 High energy particles from space At the beginning of the XX century, it was discovered that a lot of energetic particles come to the Earth from all directions. Such cosmic rays (CR) are consisting mainly of protons (about 90%), but Helium nuclei ( 10%), electrons ( 1%) and all isotopes are present, spanning a very wide energy range (from few MeV up to few ev) (for a review see for example [1], [2], [3]). In addition to charged particles, also a detectable flux of energetic photons and neutrinos is present. Figure 1.1 shows the differential flux of all primary cosmic ray charged particles, from about hundred MeV to the highest measured energies (few ev). With the exception of solar particles, whose energies are usually below 10 GeV (but they can overcome this limit during violent solar flares), the CR particles in this broad energy range have origin outside the solar system. Below few hundred MeV a different component emerges that is mainly composed of hydrogen and helium, called anomalous cosmic rays, but we will not describe it further because it is below the range of the AMS experiment (for a review see for example [4]). A visual inspection of figure 1.1 makes it clear that the CR spectrum is fastly decreasing with increasing energy following a power law with very few structures. The first evident deviation from a power law is at the low energy corner: here the spectrum flattens and the flux reaches the maximum (some thousand particles per square meter per steradian per second). The flux of particles with lower energies is dumped by the solar wind, and the maximum oscillates following the 11 year solar half-cycles: the energy spectrum of galactic cosmic rays is influenced by the solar activity. This process is known as the solar modulation, and will be 1

12 COSMIC RAYS BELOW 1 TEV Flux ( m 2 sr s GeV ) 1 Flux of Cosmic Rays 2 (1 particle per m per second) Knee 2 (1 particle per m per year) 2nd knee Ankle 2 (1 particle per km per year) Energy ( ev) Figure 1.1: All particle cosmic ray spectrum, adapted from [2] and [3]. described below ( 1.4). Starting from about 10 GeV to about 4 PeV (= ev) the spectrum has a smooth shape, as is better seen in figure 1.2, that shows the differential flux of the most important CR nuclei through a compilation of several measurements done before 1998 [5]. For kinetic energy E higher than few GeV per nucleon, the 2

13 1.1 High energy particles from space He Differential flux (m 2 sr s MeV/nucleon) C Fe H Kinetic energy ( MeV / nucleon ) Figure 1.2: Compilation of several experimental measurements of the differential cosmic ray flux of hydrogen, helium, carbon, and iron nuclei [5]. energy spectra are well described by a power law: Φ i (E < E 0 ) = E0 0 N(E) de = E0 0 k i E γ i de, (1.1) where Φ i (E < E 0 ) is the integral flux of CR particles of species i (usually ex- 3

14 COSMIC RAYS BELOW 1 TEV E 2.7 dn/de (cm 2 sr 1 s 1 GeV 1.7 ) E (ev/nucleus) Figure 1.3: All cosmic ray particles energy spectrum [5], multiplied by E 2.7 in order to better show the spectral index changes. pressed in particles m 2 sr 1 s 1 ), γ i is the spectral index of the power law, and k i is the normalization constant. The spectral indices are around 2.7 for all nuclei below the spectrum knee (roughly at 4 PeV), when the spectrum abruptly steepens (the spectral index becomes equal to 3; see figure 1.3). Above the knee, the experimental techniques used insofar are not able to measure the elemental composition of the cosmic rays: at most the analysis of atmospheric showers can give us the logarithm of the atomic mass of the incoming particles [6]. Recently it was pointed out (see for example Hörandel [3] and references therein) that the spectrum shows a slight change in the spectral index starting from about 400 PeV (the second knee ): the spectrum smoothly increases the slope up to few ev (the ankle ), when it flattens again, reaching a spectral index of about 2.5 (figure 1.3). At the highest energies ever detected (few ev) the measured flux is limited by the very low statistics of such rare events: only a total of few tens of events were discovered by the various esperiments in the last 30 years. Such particles have a curvature radius larger than the Galaxy disk thickness, hence they follow 4

15 1.1 High energy particles from space pratically a straight line between the acceleration site and the Earth. At present, the measured ultra-high energy cosmic rays (UHECR) incoming direction distribution is nearly uniform (with the exception of an ambiguous possible clustering along the line of sight of the Galaxy center [7] [8]), indicating an extragalactic origin. The various changes in the spectral index of the CR spectrum reflect the different origin and the propagation history of cosmic rays with different energy: below the knee their curvature radius is smaller than the galactic disk thickness, hence their sources must belong to our Galaxy, where CR propagate by diffusion ( 1.3). The curvature radius of a particle with charge ze in a uniform magnetic field is: r = p zeb 10 E (15) pc, (1.2) ze B ( 6) where E (15) is the particle energy in PeV and B ( 6) is the field intensity in µg. Above the knee, the curvature radius become greater than the disk thickness, and CR may escape into the galactic halo, where the density is very low and the magnetic field is weaker than the disk one. Only a fraction of the particles that are diffusing through the halo can re-enter the disk, going into a zone with stronger magnetic field. This leakage of CR induces an increase of the spectral index because the escape probability is greater for higher energy particles. The power required to mantain the observed CR energy density w CR 1 ev cm 3 [1] can be provided by supernova explosions, whose rate in our Galaxy is about (30 y) 1, if a mechanism with typical efficiency of order 10% is found. The preferred theory is diffusive acceleration applied to the strong shock waves originated by supernova explosions ( 1.2). This mechanism is based on the repeated crossing of the shock front by the ambient particles of charge ze, that on the average gain energy on each encounter with the supersonic wave (the so called first order Fermi mechanism ), and it is effective up to energies of z ev ( 1.2.2). Hence different species will show different cut-off limits, when their spectra must change slope [3]. This may be the cause of the shape of the all-particle spectrum between the knee and the second knee. Actually, the results from KASKADE [6] show that the knee can be explained as the effect of this cut-off energy for protons and helium, the most abundant species in cosmic rays. On the other hand, the propagation itself could be responsible for the second knee [6], due to the charge dependence of the escape time τ esc from the Galaxy. Extrapolation of τ esc from the GeV range to higher energies gives at 3 5

16 COSMIC RAYS BELOW 1 TEV PeV cτ esc 300 pc for protons, comparable with the disk thickness. Hence one should expect increasing anisotropies with larger energies, and a cut-off energy again proportional to the particle charge. The origin of the structure between the knee and the second knee in the CR spectrum is still debated, and other possible explanations have been suggested [6]. During supernova explosion, the expanding shock wave is able to accelerate charged particles up to the knee range, hence the existence of measurable CR fluxes at higher energies requires a new kind of engine in our Galaxy. The spectrum above the knee is smoothly connected to the flux below the cut-off energies, and the most natural way to obtain this effect is to imagine that the second acceleration process acts on the particles accelerated by supernovae. This reacceleration mechanism can take place in the vicinity of pulsars [9], where the rapidly rotating magnetic field is a powerful astrophysical dynamo, that is able to accelerate charged particles up to ultra high energies. At the ankle even the most heavy elements have energies beyond the supernovae (and may be also the pulsar) acceleration limit, hence a new component is required to explain the observed flux [3]. In addition, above the ankle the cosmic ray energy is so big that their trajectory is not bended very much even by the disk interstellar magnetic field, and can be considered a straight line. Because of the quite uniform distribution of the incident directions over the whole sky, it is probable that such particles were originated outside our Galaxy. In the following, we will focus on charged cosmic rays in the energy range above few hundred MeV and below few TeV, that are the target of the AMS experiment (chapter 4). Such particles are originated by galactic sources and their spectra give us informations about the sources, the interstellar medium, the galactic magnetic field, and the eliosphere. Thus, the study of the cosmic ray flux of protons, electrons, helium and all other nuclei, when combined with informations about their reaction cross sections, energy loss and interactions with magnetic fields, is a fundamental probe for the knowledge of our Galaxy. 1.2 Cosmic ray acceleration While both hydrogen and helium can be of primordial origin, the latter and all heavier cosmic ray nuclei up to iron are produced in stellar cores and are injected in the Galaxy by stellar winds and supernova explosions. Supernovae are also responsible for the formation of heavier nuclei up to the most massive elements, 6

17 1.2 Cosmic ray acceleration and are thought to be the main engine for the cosmic ray acceleration below the knee. Following Schlickeiser [10], we can generally say that energetic charged particles can get accelerated both by momentum diffusion due to cyclotron and/or transit-time damping of the electric fields of the ambient electromagnetic fluctuations, and by momentum convection due to compression of the CR scattering centers. This can be seen in the diffusion approximation to CR dynamics, based on the observation that energetic particles have usually a nearly isotropic pitch angle distribution relative to the ambient plasma. One can thus interpret this isotropy as the result of the scattering of CR by the low-frequency magneto-hydrodynamic turbolence, for example by Alfvén waves and fast magnetosonic waves, both with velocity much less than the light speed (hence with magnetic component much greater than the electric one: δb = (c/v) δe is the relation between the two kinds of fluctuation). Due to the existence of such e.m. fluctuations, that act as scattering centers for energetic charged particles, it is possible to describe the evolution of the isotropic part of the distribution function of the particles of species a in the phase-space M a (x, p, t) (averaged over all momentum directions) with a diffusion-convection equation for non-relativistic bulk speeds of the background medium [10], in the mixed comoving coordinate system, where position is referred to the observer reference system while momentum is measured in the rest frame of the streaming plasma: M a t S a (x, p, t) = z + 1 p 2 p [ M a κ zz z ] V M a ( ) p 2 M a A 2 p + p3 V 3 z M a p 2 ṗm a M a T c (z, p) (1.3) where the ambient magnetic field defines the z direction, κ i j is the spatial diffusion tensor, U c is the non-relativistic bulk speed of the plasma and V U + 1 (p 2 va 1 ) 4p 2 p (1.4) is the effective cosmic ray bulk speed; A 1 is the rate of adiabatic deceleration, and A 2 is the momentum diffusion coefficient. S a (x, p, t) is the injection function (including sources and sinks) and T c (z, p) is the momentum loss time, that depends on the interactions made by the particle with the surrounding medium. 7

18 COSMIC RAYS BELOW 1 TEV The transport equation (1.3) contains spatial diffusion along the field lines and spatial convection (first and second term inside square brackets, respectively), momentum diffusion and convection (first and second term inside parentheses, respectively), continuous momentum losses (last term inside parentheses), and catastrophic momentum losses (last term) due to point-like interactions. Most of existing models of CR propagation and acceleration make use of this equation (usually with simplifications). The total number density of CR particles of type a at position x is N a (x, t) = 4π while the corresponding total injection rate is Q a (x, t) = 4π 0 0 M a (x, p, t)p 2 dp, (1.5) S a (x, p, t)p 2 dp. (1.6) Integrating equation (1.3) over momentum from 0 to after having multiplied it for 4πp 2 dp gives 1 : N a Q a (x, t) = [ ] N a κ zz t z z VN a N a (1.7) T c where the momentum-averaged spatial diffusion coefficient and catastrophic loss time are κ zz 4π 0 T c (z) 4π 0 κ zz M a p 2 dp N a N a dp p 2 M a /T c (z, p) (1.8a) (1.8b) respectively. By integrating equation (1.7) over the confinement volume V for the cosmic rays, using the Gauss theorem to convert a volume integral into a surface integral, we get: N a t = Q a (t) + G a N a T c, (1.9) 1 This integration over the momentum variations (terms in equation (1.3) inside parentheses) is obviously evaluated to zero, because the population is unchanged. 8

19 1.2 Cosmic ray acceleration where N a (t) = Q a (t) = G a (t) = V V S N a (x, t) d 3 x, Q a (x, t) d 3 x, [ ] N a dσ κ zz z VN a (1.10a) (1.10b). (1.10c) Here G a is the number of CR particles escaping from the surface S of the confining region. The balance equation (1.9) says that the number of cosmic ray particles in the confinement region can only change due to new particles being injected by the sources with rate Q a ; particles destroied by spallation, annihilation or natural decay (the N a /T c term); particles escaping the region through its boundary with rate G a Shock waves Both momentum diffusion and momentum convection, the possible causes of particle acceleration [10], occur near cosmic shock waves. These are direct consequences of violent dynamical phenomena in the universe, as high-velocity stellar winds and supernova expanding shells impinging on the surrounding medium. In such cases a discontinuity is formed as the separation between the region dominated by the expanding medium and the surrounding plasma. The mass flux through this discontinuity is G ρu x (1.11) where ρ is the plasma density and u x is its velocity component normal to the discontinuity surface (u y is the trasversal component). With stationary discontinuities G = 0, while G 0 with shocks. If we mark with index 1 the upstream values and with index 2 the downstream values, and use the notation [A] A 1 A 2 (1.12) for the variation of any physical quantity A after crossing the shock front, we can write the fundamental system of equations of discontinuities in magnetic fluid 9

20 COSMIC RAYS BELOW 1 TEV dynamics [10] in the following form: [P] + G 2 [V] + [B 2 y]/8π = 0 (1.13a) G [u y ] = B x [B y ]/4π (1.13b) B x [u y ] = G [VB y ] (1.13c) G [w + G 2 V 2 /2 + u 2 y /2 + VB2 Y /4π] = B x [u y B y ]/4π (1.13d) where P is the gas pressure, V = 1/ρ is the specific volume of the fluid, B x and B y are the normal and trasversal components of the magnetic field respectively, and w = e + P/ρ (1.14) is the enthalpy per unit fluid mass, where e is its internal energy. In ideal gasses w = γ PV, (1.15) γ 1 where γ is the ratio of specific heats. We will consider charged particles acceleration by a fast, but non-relativistic, super-alfvénic parallel shock front [10], where the upstream Alfvén waves have a Kolmogorov spectrum I(k) Θ(k k 0 )k q and the spectral index is q = 2. Θ(k k 0 ) is the step function and k 0 is the minimum wavenumber. Parallel shock waves. A parallel shock wave is one whose normal is parallel to the magnetic field B. In this case equations (1.13a) (1.13d) reduce to that can be conbined into a single equation: [P + G 2 V] = [P + Gu x ] = 0 (1.16a) [w + G 2 V 2 /2] = [w + u 2 x/2] = 0 (1.16b) w 1 w (V 1 + V 2 )(P 2 P 1 ) = 0. (1.17) Using equation (1.15) and the ideal gas law PV T we get the temperature ratio between the particles that have crossed the shock front and those that have not: T 2 = (γ + 1)P 1 + (γ 1)P 2 P 2 = r P 2, (1.18) T 1 (γ 1)P 1 + (γ + 1)P 2 P 1 P 1 10

21 1.2 Cosmic ray acceleration where the compression ratio has been introduced: From the last equation (1.19) it follows: r ρ 2 ρ 1 = GV 1 GV 2 = (γ + 1)P 1 + (γ 1)P 2 (γ 1)P 1 + (γ + 1)P 2. (1.19) P 2 P 1 = r(γ + 1) (γ 1) γ + 1 r(γ 1). (1.20) Introducing the Alfvénic Mach number of the flow M A,1 u x,1 V A,1 (1.21) in terms of the upstream normal gas velocity component u x,1 and the gas Alfvén velocity V A B ( B ) ( = n ) 1/2 e, (1.22) 4π(mp + m e )n e 1 G 1 cm 3 and the sound Mach number of the flow M s,1 u x,1 v s,1 = G2 V 1 γp 1 = 1 (P 2/P 1 ) γ[(v 2 /V 1 ) 1], (1.23) one gets the compression ratio expressed as where r = (γ + 1)M2 s,1 (γ 1)M 2 s,1 + 2 = (γ + 1)MA,1 2, (1.24) (γ 1)M 2 A,1 + 2β β = ( MA,1 M s,1 ) 2 = ( vs,1 V A,1 ) 2 = 4πγ P 1 B 2 x. (1.25) For strong shocks M A,1 1 and the compression ratio approaches the limit r(m A,1 1) γ + 1 γ 1. (1.26) 11

22 COSMIC RAYS BELOW 1 TEV Particle acceleration by shock waves The steady-state upstream transport equation for relativistic charged particles, in the case of a fast parallel shock with Alfvénic waves with a power spectrum q = 2, can be written in the following way [10]: x ( κ 1 F 1 x ) + V 1 F 1 x = Q 1(x, p) (x > 0), (1.27) where the index 1 is used to mark upstream quantities, F(x, p, t) is the isotropic part of the gyrophase-averaged particle phase space density, κ(x) is the spatial diffusion coefficient (it is idependent of momentum for relativistic particles), Q(x, p) represents sources and sinks, and V 1 = u 1 + H c1 V A,1 is the CR bulk speed. H c [ 1, +1] is the normalized cross helicity state: H c = (δb f) 2 (δb b ) 2 (δb f ) 2 + (δb b ) 2 (1.28) ( f represents forward particles, moving in the same direction as the fluid, b is for backward particles, with opposite direction). On the other hand, the steady-state downstream transport equation is: x ( κ 2 F 2 x ) F 2 + V 2 x + ϑ p 2 p V 2 A,1 p4 F 2 rκ 2 p = Q 2(x, p) (x < 0), (1.29) where the downstream CR bulk velocity V 2 = u 2 + H c2,k V A,2, H c2,k is the downstream cross helicity at constant wavenumber k, ϑ is a parameter that depends on the spectral index q = 2, the magnetic helicity state of the upstream waves, and the ration between reflected and transmitted coefficiens [10]. The term with square brackets represents momentum diffusion, and contains the compression ratio r. To solve equations (1.27) and (1.29) we impose the following boundary and continuity conditions [10]: 1. far upstream (x ) we require no accelerated particles: F 1 (+, p) = 0; 2. no damping of Alfvén waves: these act as infinite source of energy. In order to avoid that particles may reach any energy, we impose a finite size L for the downstream region: F 2 ( L, p) = 0; 3. Q 1 = Q 2 = 0: particle injection at the position of the shock only, with rate q 0 (p); 12

23 1.2 Cosmic ray acceleration 4. the phase space density must be continue at the shock front: F 1 (0, p) = F 2 (0, p). The same must be true for the current density: [ ] [ ] F 1 κ 1 x + V F 2 1F 1 κ 2 (x 0 + ) x + V 2F 2 (x 0 ) + V (1.30) 1 V 2 3p 2 p (p3 F 1 ) + q 0 = 0. Multiplying (1.27) with κ 1 (x) and introducing the new vari- Upstream solution. able µ 1 (x) = the solution in the upstream region is Downstream solution. variable x 0 dx κ 1 (x ), (1.31) F 1 (x > 0, p) = F 0 (p) exp( V 1 µ 1 ). (1.32) Multiplying (1.29) with κ 2 (x) and introducing the new µ(x) = x 0 dx κ 2 (x ), (1.33) the downstream equation (1.29) can be written in the form: 2 F 2 µ + V F p + A ( p 4 F ) 2 = 0 (1.34) p 2 p p with A ϑva,1 2 /r. It can be shown [10] that the solution of this equation can be written in the following form: F 2 (µ, p) = 0 q 0 (p) G(µ, p, t) dt, (1.35) where the Green s function has the following complicated expression: G(µ, p, t) = 2e V 2µ/2 ( ) p ωm [ ] 1 [ ( D S (ωm ) 1 + µ )], (1.36) tv 2 t ω 2 η m=1 ω=ω m sin where ω is the Fourier conjugated variable of t, η µ( L), S (ω) is a complex function that depends on the compression ratio, the parameters η, ϑ and H c2,k, the upstream Alfvénic Mach number, and the downstream bulk velocity. The complex function D(ω) depends on S (ω) and the values ω m are the zeroes of D(ω). 13

24 COSMIC RAYS BELOW 1 TEV Zero momentum diffusion. The very complicated analytical solution (1.36) admits a quite simple representation for the case of zero momentum diffusion, when calculated at the position of the shock fornt (µ = 0) with injection q 0 (p) = δ(p p 0 ): F 2 (0, p) = 3(p/p 0) s (1.37) p 0 V 2 (R 1) where R = V 1 /V 2 is the scattering center compression ratio and s is the only nonvanishing zero of D(ω): s = ω = 3 R + (e2n 1) 1 R 1. (1.38) The Peclet number N = ηv 2 /2 measures the importance of the acceleration by the downstream Alfvén wave field. In case of strong acceleration while in the opposite case s(n 1) s(n 1) 3R R 1, (1.39) 3 2N(R 1). (1.40) With large Peclet numbers, the spectral index of the power law of the accelerated particles depends only upon the scattering center compression ratio R = V 1 /V 2. The differential intensity of CR particles is: dφ de p2 F 2 (0, p) p Γ, (1.41) with Γ = s 2 = R + 2 R 1. (1.42) If the scattering center compression ratio R equals the gas compression ratio r, as in the infinite Mach number limit, the spectral index Γ is Γ gas 2, since 1 < r 4 for adiabatic shocks with γ = 5/3. Finite momentum diffusion. For finite momentum diffusion, the solutions are mathematically much more complex: equation (1.36) shows that they can be written as infinite sums of power laws whose spectral indices follow from a trascendental eigenvalue equation [10]. Each individual power law component is weighted 14

25 1.3 Interstellar propagation by coefficients that depend on the actual downstream position. In addition, they and the eigenvalues ω m depend on the scattering center compression ratio, the shock wave Peclet number and the level of momentum diffusion. Usually the general solution (1.36) is simplified imposing additional constraints, and computed with numerical methods. Hence it can be used in complex simulations, where specific models for the source of the shock wave are assumed, and specific functional forms may be used for the injection spectrum q 0 (p) and the diffusion coefficients κ i (x). Maximum particle energy. An important assumption for the previous treatment to be correct [10] is that the cosmic ray gyroradius R L is less than the maximum wavelength λ max of the plasma waves: R L /λ max 1. As upper bound for λ max one can use the characteristic size L of the system, in order to get a limit on the maximum achievable particle rigidity 2 : R pc ze BL, (1.43) where p is the relativistic momentum, c is the speed of light, ze is the particle charge, and B is the ambient magnetic field. A useful way to write the same thing is: ( ) ( ) B L E/z 1µ G 1 pc ev. (1.44) Supernova shock waves have L 1 pc and B (10 100) µg, hence they cannot accelerate cosmic rays to rigidities much above V. This is the energy range where the spectrum knee shows itself: the spectral index change can be caused by the cut-off of the protons, that are the most abundant CR fraction. Other elements reach higher and higher cut-off energies, and the second knee could correspond to the maximum energy achievable by Fe nuclei [3]. 1.3 Interstellar propagation Up to the knee range of the cosmic rays spectrum, the supernovae can be the sites of the CR acceleration and the resulting spectrum can be well represented by a power law in momentum. Then accelerated particles start to diffuse through the 2 R has the dimensions of an energy divided by a charge, i.e. of an electric potential: when p is measured in GeV/c, R is measured in giga-volt (GV). 15

26 COSMIC RAYS BELOW 1 TEV Galaxy, where they travel on the average for several million years before reaching the Earth. Due to their different interaction cross sections, the elements have a slightly different propagation history, and the measured abundances on the Earth can be used to infer their characteristic diffusion time (about 20 My) and the matter depth (5 10 g cm 2 ) traversed during propagation. Before reaching the detector, the cosmic rays must diffuse inside the eliosphere, where they loose energy due to the interactions with the magnetic field transported by the solar wind, and finally they interact with the geomagnetic field ( 1.4), that is like a filter that forbids very low energy particles to reach the Earth atmosphere (but at the same time all soft secondary particles are trapped inside the magnetosphere). The simplest galactic propagation model is the leaky box one: the galactic disk is assumed to be a cylinder with radius of order 10 kpc and height of about 1 kpc, and cosmic rays are assumed to travel freely inside the disk, where a magnetic field of (3 6) µg exists. When CR reach the cylinder boundary they bounce elastically, but there is a finite probability (increasing with the particle momentum) that they cross the boundary and escape from the Galaxy. This model is analytically solvable and suggests that the CR spectrum remains a power law, with a higher spectral index than at the source, but it is far too simplistic to be able to explain the spectra of different elements. As further steps, one can introduce a bigger cylinder fully containing the disk but with weaker magnetic field, as a model for the galactic halo. One may then consider a diffusion process within a medium with two different densities (for disk and halo), then add other informations about the spiral arms, the galactic bulge, the magnetic field configurations, and so on. The problem will become very complicate and in general only numerical methods can be used to sketch the solution, for different configurations in the parameters space. The diffusion steady state equation for element j can be written [11]: (K j N j V c N j ) E [ Vc 3 E k ( 2m + Ek ) N j ] + m + E k E (b jn j ) E (d jn 2 j ) + Γ j N j = q j + Γ k j N k m k >m j (1.45) The first terms represent diffusion (K j is the diffusion coefficient) and convection 16

27 1.4 Propagation inside the eliosphere (V c is the convection velocity). The divergence of V c is connected to the energy loss due to adiabatic expansion of cosmic rays. The next term represent the ionization and Coulomb energy losses, plus the first term of the reacceleration, all included in b j [12]: b j (E) = de + dt ion de + dt Coul de + dt adiab de dt reac. (1.46) The next is the second order term in reacceleration (d j is the energy diffusion coefficient), and the inelastic collision term with the ISM. On the right side, q j is the source term, while the last sum is over the spallation reactions producing nuclei of kind j. This equation may be used to determine the source composition starting from the measured abundances at Earth (chapter 2). 1.4 Propagation inside the eliosphere Measuring cosmic rays from the Earth obviously means considering a sample of particles diffusing inside the eliosphere and reaching at least the distance 1 AU from the Sun (1 AU = km). Hence the measured spectrum is not the same as the average galactic spectrum, neither it is the same as the local interstellar spectrum (LIS, i.e. the spectrum outside the eliosphere but not too distant from the Sun). Cosmic rays entering the eliosphere (that has a boundary at a distance of order 100 AU from the Sun) must diffuse inside a zone where a continuous outflow of highly conductive plasma exists. This solar wind is composed mostly by hydrogen and helium isotopes, but electrons, positrons and the isotopes of several light elements are also present [1]. The solar wind particle kinetic energy is usually below few GeV per nucleon, in solar quiet periods, but during solar flares it can reach 100 GeV per nucleon. The magnetic field lines are frozen inside this plasma, that is expanding. Due to the rotation of the Sun around its axis these lines tend to assume a spiral shape, but the magnetic field configuration inside the eliosphere is complicated by the polarity reversal of the solar field every 11 years. Hence the energy spectrum of the measured particles is affected by the solar activity and shows periodical behavior (solar modulation), at least below few GeV per nucleon. At low energy it is necessary also to take into consideration the Earth magnetic field, that is not a perfect dipole (figure 1.4). The typical energy of the geomagnetic field is lower than the eliomagnetic field at 1 AU, and can be neglected down 17

28 COSMIC RAYS BELOW 1 TEV Shock front or bow shock solar wind Van Allen radiation belt magnetopause Figure 1.4: A sketch of the structure of the Earth magnetosphere. to few Earth radii. Using a detailed model of the Earth magnetic field it is possible to trace incident particles (and to back-trace the measured particles [13, 14]), to study the geomagnetic effect Solar modulation The simplest model of solar modulation [15] uses a spherical approximation of the eliosphere and has as free parameters the solar wind velocity V(r, t) and the radial diffusion coefficient k(r, E k, t). Charged particles with kinetic energy E k propagate by diffusion and convection and adiabatically exchange energy with the expanding solar wind. Their differential density U(r, E k, t) at distance r from the Sun and time t can be found by solving the equation: U t = k U (UV) V E k (α(e k )E k U), (1.47) where the left side term is the temporal variation of the cosmic ray density, that has a small periodic fluctuation following the 11 years half solar cycle, and in the right hand side the three terms describe: 1) the particle diffusion inside the magnetic field generated by the Sun, 2) convection, and 3) the adiabatic deceleration. 18

29 1.4 Propagation inside the eliosphere The diffusion coefficient in this model has the form k = λv/3, where v is the particle velocity and λ = λ(r, R, t) is the mean free path in the magnetic field, that depends on the time t, the radial distance r and the particle rigidity R. In the spherical approximation one can write [15]: k = β R k 1 (r, t), (1.48) where β = v/c and k 1 (r, t) is the diagonal component of the diffusion tensor. The last term of equation (1.47) contains the term α(e k ) = E k + 2E 0 E k + E 0 (1.49) where E 0 = Mc 2 is the total particle energy at rest. Gleeson and Axford [15] showed that the solution of equation (1.47) corresponding to the stationary case with constant k and V, can be expressed in the following form: J(r, E, t) = E 2 E 2 0 J(, E + Φ(t)), (1.50) (Φ(t) + E) 2 E 2 0 where E = E k + E 0 is the total particle energy. Hence the particle flux J(r, E, t) inside the eliosphere at time t and distance r from the Sun can be related to the stationary interstellar flux J(, E + Φ(t)) through an expression that depends on the energy Φ(t) lost by the particle reaching this radial distance from infinity. Neglecting any charge sign asymmetry one can write Φ(t) = z eφ(t), (1.51) where the solar modulation parameter φ = r b drv/(3k r 1 ) is usually directly inferred by the experimental data. This can be done basicaly in two ways [16]: by relating the total cosmic ray flux on the top of the atmosphere (extrapolated by neutron counting facilities) to the solar activity (for example the number of sun spots), or by unfolding a reference spectrum (usually the proton one) using the relation (1.50) Geomagnetic cutoff The Earth magnetosphere has a long tail, produced by the interaction of the solar wind with the geomagnetic field (see figure 1.4). Between Sun and Earth, at about 19

30 COSMIC RAYS BELOW 1 TEV Figure 1.5: Isocountours of the geomagnetic field in a low orbit altitude (400 km above sea level). 15 Earth radii from our planet, the solar wind produces a shock front beacuse it is supersonic with respect to the Earth atmosphere. This bow shock [17] is a shield against the energetic particles coming from the Sun, that can penetrate only near the polar caps. On the opposite side from Sun, along the tail of the magnetosphere, the geomagnetic field joins the solar field forming a complicate structure in which there are accessible and forbidden zones for charged particles. Low energy cosmic rays are affected by this structure, and their flux on the Earth depends on the geomagnetic coordinates (figure 1.5). In particular, there is a zone, called South Atlantic Anomaly (SAA) where the lower Van Allen belt, usually beyond 600 km above sea level, reaches about 250 km a.s.l. and is crossed by space-crafts flying at low orbit altitudes. The SAA is above the Atlantic Ocean, near the Brazil coast 3, and is filled by low energy (down to 10 MeV) protons, electrons and few light nuclei. 3 See spaceweb/textbook/radbelts.html for example. 20

31 1.4 Propagation inside the eliosphere geomagnetic cp min E k min latit. θ ( ) (GeV) (GeV) Table 1.1: Geomagnetic cut-off for radially incident protons [1]. The geomagnetic field acts as a filter for low energy particles, that can penetrate only regions that depend on their charge, momentum and incident direction. For example, the table 1.1 shows the minimum momentum or energy for radially incident protons at three different geomagnetic latitudes. On the other hand, soft secondaries produced in the atmosphere will likely be trapped, if they have energy below the cut-off. For protons with momentum direction at angle ω with respect to the normal to the geomagnetic meridian plane, one has [18]: [ RT cp min = 59.4 R cos 2 θ cos ω cos 3 θ ] 2 (GeV), (1.52) where R T = km is the Earth radius and R is the orbital radius. 21

32 22 COSMIC RAYS BELOW 1 TEV

33 Chapter 2 Cosmic ray composition Cosmic rays consist of protons, α particles (He nuclei), electrons and positrons, and nuclei of all the isotopes. Even though their positive charge is high, during the propagation through the low-density interstellar medium (ISM) the high energy nuclei are not able to capture electrons, hence they are usually fully ionized when detected above the Earth atmosphere. At low energy (about 10 MeV per nucleon) there is a population of partially ionized nuclei that are called anomalous cosmic rays (ACR). They are thought to be neutral atoms ionized by the solar wind that are accelerated by shocks in the solar system (for example on the eliopause) [4]. We will not consider ACR here because their energy is below the threshold of AMS. The relative abundances of the different elements are related to the composition at the source and the propagation history of cosmic rays. The isotopic spectra of few elements are expecially important because they are of pure secondary origin (like B, for example) or radioactive (like 10 Be and 26 Al), thus allowing for an estimation of the matter thickness traversed by CR and of the propagation time between sources and detection. Electrons (and positrons) have very rapid energy losses through electromagnetic processes, while heavy particles mainly degrade their energy by ionizing the interstellar medium. Hence e and e + diffuse in smaller volumes than ions and their measured spectra are mostly determined by the last recent and near supernovae. Positrons are thought to be produced during the CR propagation in the ISM. For this reason they are also very useful for the fine tuning of the parameters of propagation models. Protons and antiprotons, and electrons and positrons, are also important for two reasons. First, particle and antiparticle are pratically equivalent for propa- 23

34 COSMIC RAY COMPOSITION gation models, because they differ only by the sign of the electric charge. This difference can be used to study the charge-dependent effects of the solar modulation. Second, the antiparticle spectra may show distinctive features that could be interpreted as signatures of the annihilation of the exotic particles that consitute the dark matter in our Galaxy (as in the rest of the universe). 2.1 Cosmic ray ions The primordial nucleosynthesis produced the protons and the bulk of the helium nuclei, but only a negligible part of heavier elements ( 7 Li is the only detectable fraction) [19]. The rest of the baryonic matter was (and is being) produced by stellar nucleosynthesis (up to Fe) and SN explosions CR source composition Although supernova (SN) shocks are generally believed to be the engines of the cosmic rays acceleration, at least below the spectrum knee, the sources of the particles that are accelerated are still debatable [20]. The most efficient site for particle acceleration is the low-density hot ISM, where the energy that the particles gained interacting with the shock front is not efficiently dissipated. Howewer, the accelerated particles may come from the cooler and partially ionized phase of the medium, or they could be pre-accelerated by stellar coronae in a manner depending on the first ionization potential (FIP) (see [20] and references therein). Recently, a different scenario emerged that is based on the interstellar grain acceleration by SN shock waves, where the different volatility of the elements is the important quantity, determining the relative abundances in the cosmic rays [21]. Figure 2.1 shows the results from the model of Ellison et al. [21] about the source abundances of the different elements, obtained requiring that after the propagation from the source to the Earth the relative abundances best fit the experimental results (as of 1997). The elements are divided into refractory, semivolatile, volatile, and highly volatile groups, based on decreasing condensation temperature. The refractories are essentially completely locked in grains in the ISM, while the highly volatile are gaseous. Meyer et al. [22] showed that the galactic CR composition data are ordered in terms of the following behaviors: the abundances strongly increase with the mass number A for volatile elements (hydrogen is an exception); the refractory 24

35 2.1 Cosmic ray ions Figure 2.1: Galactic CR source aboundances versus atomic mass number, obtained with the model by Ellison et al. [21]. elements are all enhanced with respect to the volatile elements, without any mass dependence. Hence the acceleration process is more efficient for those elements locked in grains. Diffusive acceleration naturally leads to acceleration efficiencies that increase as the particle rigidity, i.e. as the A/Q ratio. Dust grains have great mass and small charge, produced by ionization caused by collisions with plasma electrons or other grains, and by photoelectric effect caused by UV photons [21]. The result is a very high A/Q ratio, i.e. a very high acceleration efficiency. However grains suffer energy losses and can escape from the region more easily than protons and nuclei, hence they cannot be accelerated above 100 kev per nucleon. Gain erosion by sputtering produces free ions with low effective charge Q 3e (usually they are not fully ionized), and the elements with high ionization potential have high probability to escape as neutral particles from the acceleration site. The ions have high probability to be scattered back to the shock by the 25

36 COSMIC RAY COMPOSITION magnetic irregularities of the ambient plasma, becoming the seeds of the usual diffusive acceleration process. But they have on the average higher energy of the ambient volatile elements, hence they can be accelerated to high energy (with E/A max Q/A ev) more easily [21]. This scenario predicts that all the gaseous elements accelerated out of the ISM or circumstellar matter reach the abundances that lie between the dotted and dashdotted lines in figure 2.1. Motivated by the observed 22 Ne excess, in addition to the average interstellar medium a source of heavy elements (as a 12 C-, 16 O- and 22 Ne-enriched Wolf-Rayet stellar wind) seems to be necessary [21]. This explains why they are above the predicted level in figure 2.1. No significant amount of SN ejecta material is accelerated by the shock wave: supernovae are the engines of the CR acceleration, but they act upon a medium that is composed by their past stellar wind elements, the surrounding material and the additional component of heavy elements. Thus, the similarities between the solar abundances (FIP biased) and the CR source composition (volatility biased) seems to be purely coincidental [21]. Recently it was pointed out that the various models of CR acceleration, when normalized to the observed Fe spectrum, unpredict the Be data by a large factor, suggesting a scenario where CR metals are accelerated out of SN ejecta in super bubbles [20]. In this case, a given supernova acts upon the material ejected by nearby and recent supernovae, whose explosions were triggered by other explosions, producing a depleted dominion rich of heavy elements. Due to the fact that supernova progenitors form mostly in O-B associations and have short lifetimes, the great majority of the core-collapse supernovae explode in hot, low-density superbubbles that are hundreds of parsecs large and can last for tens of megayears. C and O nuclei accelerated by these successive SN shocks interact with ambient H and He, producing the bulk of the beryllium in the Galaxy Propagation models The relative abundances of the different species in cosmic rays give important informations on the propagation details and on the source composition. For example the instable nuclei can be used to infer the propagation time of CR in the Galaxy, while pure secondary nuclei constrain the matter thickness (usually in g cm 2 ) traversed during propagation. The simplest model is the leaky box, where the CR are freely propagating in a given volume and have finite probability to escape its boundaries. In stationary 26

37 2.1 Cosmic ray ions conditions, the sources produce particles that may be lost due to inelastic interactions or to escape, and the number density variation due to these processes can be written in a very simple form: N/τ i + N/τ e, where τ i is the mean interaction time and τ e is the mean escape time. When the secondary production is considered, it is possible to write down a set of differential equations that connect the measured relative abundances to the source composition. It comes out that the propagation time (inferred from instable nuclei) is t 20 My, while the matter traversed by cosmic rays is x 9 g cm 2. We can infer the average density of the medium n = x/(ctm H ) 0.3 cm 3. As the disk density is about 1 cm 3, it follows that cosmic rays must spend a fraction of the time in an empty region, called the diffusion halo. The leaky box model is able to work with stable primary and secondary elements, but it fails with unstable nuclei. In addition, the parameters of the model have no direct physical meaning. Because the charged particles are scattered by magnetic irregularities, the correct propagation model is a diffusion model: the steady state equation for element j is formula (1.45) of section 1.3 [11], that we write again: (K j N j V c N j ) [ Vc E 3 E k ( 2m + Ek ) N j ] + m + E k E (b jn j ) E (d jn 2 j ) + Γ j N j = q j + Γ k j N k m k >m j In the leaky box model all quantities are spatially averaged (so that convection has no meaning), and the diffusion term is replaced by the escape term: (K j N j ) N j /τ e. Since this model is homogeneus, the mean time spent by CR in the Galaxy and the matter thicnkess x are correlated. This is not true in models keeping into account the halo. As far as stable species are concerned, leaky box and diffusion models are equivalent [11]. This can be seen in the weighted slab technique : the general solution is written in the form: N j (r) = 0 Ñ j (x) G(r, x) dx (2.1) then equation (1.45) decouples into two indipendent equations. The first one involves G(r, x) and depends only on the geometry of the problem and on the chosen 27

38 COSMIC RAY COMPOSITION diffusion scheme, but not on the species j. The other equation involves Ñ j (x) and contains the physics aspects of the propagation. Different models (leaky box, diffusion) correspond to different G(r, x), that is the path length distribution (the probability distribution that a particle j crosses the matter x). The leaky box model gives G(x) = exp( x/λ)/λ, where λ = x is the average quantity of matter traversed by CR. In diffusion models, for a wide class of geometries, the function G is given by an infinite series of exponentials involving the diffusion coefficients. The first exponential is sufficient (i.e. the leaky box is recovered) only for small values of x Measurements of CR ions spectra Direct measurements of the cosmic ray spectra in the solar system have been done by balloon and satelite experiments. The most important are HEAO-3 [23], ACE [24], Voyager 1 and 2 [25] and Ulysses [26]. These measurements can be used to constrain the parameter space of propagation models, as it was done by several groups. Recently, Donato et al. [27] and independently Moskalenko et al. [28] emphasized that, in order to have a good fit at the same time for the B/C and sub- Fe/Fe ratios, it is necessary to consider the effect of the local bubble (LB) where the Sun is collocated. In particular, the study of the radioactive nuclei (the most important are 10 Be, 14 C, 26 Al, 36 Cl, 60 Fe) shows that in first approximation the Sun is within a shell of radius 50 pc with density 0.1 cm 3, surrounded by a shell with almost null density extending up to radii 200 pc. Beyond this second shell, the average density is 1 cm 3 [27]. This is consistent with the picture of the local interstellar medium consisting of an asymmetric bubble with radius of pc filled by hot (T K) and low-density (n cm 3 ) gas, surrounded by a dense neutral gas boundary ( hydrogen wall ), as derived from spettroscopic measurements [29, 30, 27, 31, 32, 33]. The presence of this low density zone around the solar system does not affect the diffusion of stable ions, but the secondary nuclei abundances, produced by spallation, are perturbed. For radioactive species the effect is very important, because they sample a relatively small fraction of the volume in which stable nuclei propagate: the effect is well represented by an exponential attenuation exp( r hole /l rad ), where r hole is the radius of the underdense region around the Sun, and l rad = K(E)γτ 0 is the radial distance from the source that a particle that is 28

39 2.1 Cosmic ray ions B/C ratio Φ = 450 MV 0.3 Sc+Ti+V/Fe ratio LIS Voyager Ulysses ACE HEAO 3 Chapell,Webber 1981 Dwyer 1978 Maehl et al LIS ACE Ulysses Voyager ISEE 3 HEAO 3 Sanriku 450 MV 800 MV Kinetic energy, GeV/nucleon Kinetic energy, GeV/nucleon Figure 2.2: Boron to carbon (left panel) and sub-fe to iron (right panel) ratios: measurement can be fit at the same time by a diffusive propagation model that takes into account CR reacceleration and convection, and a fresh contribution of C and O nuclei in the local bubble. The computed local interstellar spectrum is shown for comparison. Details about different parameters can be found in Moskalenko et al. [28]. following a random walk reaches on the average. K(E) is the diffusion coefficient and γ is the Lorenz factor of the nucleus with lifetime τ 0 at rest [27]. Values 60 pc r hole 80 pc are suggested by 10 Be/ 9 Be and 36 Cl/Cl ratios measured by ACE, while the 26 Al/ 27 Al ratio seems not compatible. By enlarging the r hole range to 100 pc it is possible to fit at least the Ulysses and Voyager 26 Al/ 27 Al data [27]. The need for a local structure is motivated by Moskalenko et al. [28] in a different way. They stress the fact that propagation models fail to fit simultaneously the B/C and sub-fe/fe ratios, and the antiproton flux, even in presence of reacceleration. However, their numerical simulation (called GALPROP 1 ) can fit everything if one postulates that the local bubble (LB) contains fresh unprocessed component at low energy (figures 2.2 and 2.3). The presence of this 12 C and 16 O component would lower the local B/C ratio, because B is produced by spallation, that is not an important effect in the neighborhood of the Sun. This is consistent with the idea that the local bubble [30] was probably pro

40 COSMIC RAY COMPOSITION Flux, m 2 s 1 sr 1 GeV LIS Φ = 550 MV Tertiary Antiprotons BESS BESS 98 MASS91 CAPRICE Kinetic energy, GeV Figure 2.3: Cosmic ray antiproton (left panel) and ions spectra (right panel) measured by different experiments are compatible with a diffusion/reacceleration propagation model with contribution from the local bubble [28]. The measured spectra are sensibly different from the local interstellar spectra below 10 GeV per nucleon. E 2 Flux, GeV/nucleon m 2 s 1 sr LIS LIS LIS LIS ACE HEAO MV 800 MV Boron Carbon Oxygen Iron Kinetic energy, GeV/nucleon duced in a series of supernova explosions whose progenitor was an O-B star association 2. The LB age is 10 My and it was produced by SN explosions, with the last SN 1 2 My ago or 3 SN in the last 5 My. There is also evidence in favor of a nearby recent SN (at about 30 pc from the Sun) [28]. Thus it is very probable that particles coming directly from SN remnants still influence the local spectra and abundances of cosmic rays. Table 2.1 shows the results obtained by Moskalenko et al. [28] about the relative abundances to Si computed for the local bubble component and the galactic sources, compared to the measured solar system abundances. 2 An O-B association is a large, very loose form of an open star cluster consisting of young spectral type O and B stars. They cover large volumes of space, are loosely held together by gravity and have short lifetimes (a few million years) as a distinct object. 30

41 2.2 Cosmic ray electrons Z Solar LB Galactic Z Solar LB Galactic system sources sources system sources sources * * * * *) Upper limit. Table 2.1: Elemental abundances normalized to Si given by the GALPROP program [28]. 2.2 Cosmic ray electrons Cosmic ray electrons are probably accelerated by the same engines that accelerate CR protons and nuclei (supernova explosions), but they differ significantly from hadrons for what concerns the energy lost during the propagation through the interstellar medium. Because of their small mass and lack of strong interactions, electrons suffer large energy losses due to electromagnetic processes as synchrotron radiation, inverse Compton scattering, and bremsstrahlung. These losses effectively limit the volume that can be pervaded by the elctrons that were emitted by a given source: the measured spectrum must carry information on smaller scales than the stable elements. Even though there are conceivable sources of primary positrons, like pulsars, primordial black holes or supersymmetric particles annihilation, the measured fluxes are compatible with the simple hypothesis of complete secondary origin. In fact, the secondary production of e + and e by pion decay (the pions are produced in the CR proton interactions with the interstellar medium) yields almost the same amount of electrons and positrons, whereas the measured e + /e fraction is about 10%. Thus electrons are mostly of primary origin, while the positrons have secondary origin and can be used to set constrains on the parameters of CR propagation. 31

42 COSMIC RAY COMPOSITION Energy losses Bremsstrahlung is significant when electrons propagate through HII regions (radio emission) or near X binaries (X-rays), and the emission power due to bremsstrahlung is proportional to the product of the ion and electron densities of the plasma, and to the square of the average ion charge: ( ) de = 4N e N ion Z(Z + 1.3)r dt eαcge 2 (2.2) brem where r e is the classical electron radius, g is the Gaunt factor, and E = γmc 2 is the total energy. In the cases of interest: ( ) de dt brem γ ln γ γ fully ionized H neutral H (2.3) If the electrons have a power law distribution in momentum with spectral index η, the bremsstrahlung radiation has a power law spectrum with the same spectral index η [1]. Synchrotron radiation is the dominant process when CR electrons propagate through regions where a magnetic field of intensity H exists, the emission power being proportional to U mag H 2 : ( ) de = 4 dt sync 3 σ T cu mag γ 2 (2.4) where σ T = re 28π/3 is the Thomson cross-section (r e = e 2 /(4πε 0 m e c 2 ) = m is the classical electron radius) and U mag is the magnetic field energy density. For an electron power law distribution in momentum with spectral index η, the synchrotron radiation has a spetrum that depends on the average magnetic field intensity and a spectral index equal to (η 1)/2 [1]. Inverse Compton scattering is the dominant process in photon rich environments, like the jets emitted by active galactic nuclei, and in case of low magnetic field and matter density, when the electrons have to interact at least with the cosmic microwave background photons. If the energy density of the photon field is U rad : ( ) de = 4 dt i.c. 3 σ T cu rad γ 2 (2.5) 32

43 2.2 Cosmic ray electrons Measurement Year Sun e /e + E min E max Ref. pol. sep. (GeV) (GeV) MASS Y [35] CAPRICE Y [36] HEAT Y [37] HEAT Y [37] Nishimura , N [38] BETS , Y [39] AMS Y [40] Table 2.2: Cosmic ray electrons measurements. In addition to AMS-01, only experiments which published data tables are reported. Positive and negative Sun polarities refer to epochs when the magnetic field emerging from the North Pole of the Sun points outward and inward, respectively [41]. A power law spectrum with spectral index η for the electrons produces a power spectrum in frequency due to inverse Compton scattering with spectral index equal to (η 1)/2 [1]. In conclusion, electrons suffer total energy losses ( ) de = ae 2 + be + c ln E, (2.6) dt tot while electromagnetic losses different by ionization are usually not important for CR protons and nuclei. In the leaky box jargon, this can be translated into the effective lifetime τ loss 300 (E/1 GeV) 1 My. This corresponds to a mean diffusive radial distance r loss 1 kpc (E/1 GeV) 1/2 (K/0.03 kpc 2 My 1 ) 1/2 from the source [12]. Hence the measured CR electron spectrum gives information only about a rather small volume around the solar system. For example, electrons with E > 10 GeV should be very sensitive to the local bubble, while protons are sampling a large part of the galactic disc (up to the Galaxy core [34]) and the halo. On the other hand, electrons can be traced through the Galaxy thanks to their electromagnetic emission, and their interstellar density can be inferred from the measurements of the synchrotron emission in the radio band. In this way we can infer the average electron density in our and in different galaxies. Figure 2.4 shows the inferred local interstellar spectrum (LIS) of the cosmic ray electrons, obtained demodulating the spectra measured by recent experiments 33

44 COSMIC RAY COMPOSITION E 3 F (m -2 sr -1 s -1 GeV -1 ) AMS-01 e - BETS97+98 (e - + e + ) Nishimura 2000 (e - + e + ) HEAT95 e - HEAT94 e - CAPRICE94 e - MASS91 e - Nishimura 1980 (e - + e + ) E k (GeV) Figure 2.4: Local interstellar spectrum of cosmic ray electrons measured by recent experiments between about 1 GeV and 2 TeV [16]. [16] (table 2.2). The spread of the data points could be due to systematic uncertainties arising from the correction for the residual atmosphere or for the detector response. If these effects are not strongly energy dependent, it is possible to renormalize the spectra to the same value at fixed energy, in order to minimize the spread. Figure 2.5 shows that it is possible to fit of the renormalized data set with a 34

45 2.2 Cosmic ray electrons F (m -2 sr -1 s -1 GeV -1 ) AMS-01 e - BETS97+98 (e - + e + ) Nishimura 2000 (e - + e + ) HEAT95 e - HEAT94 e - CAPRICE94 e - MASS91 e - Nishimura 1980 (e - + e + ) / 80 (3 ± 1) < E k /GeV < (2070 ± 750) γ LIS = 3.42 ± 0.02 N LIS = (340 ± 18) m -2 sr -1 s -1 GeV E k (GeV) Figure 2.5: Local interstellar spectrum of cosmic ray electrons. Data taken by different experiments were renormalized to the AMS-01 flux at 20 GeV, with the exception of CAPRICE94 [16]. single power law in kinetic energy, obtaining a spectral index of 3.4 between 3 GeV and 2 TeV. Whereas several authors (see for example [42]) emphasize the idea that the electron LIS should have slope changes due to the interplay between propagation by diffusion, energy losses and source distribution, figure 2.5 shows that the overall data set does not strongly suggest any spectral index change. In addition, the overall LIS has the same slope as the electron LIS inferred from AMS-01 alone, that extends down to 1.5 GeV, and the low energy flattening of different experiments may be explained with a charge dependent solar modu- 35

46 COSMIC RAY COMPOSITION lation effect [16]. Thus, the spectral change at 10 GeV foreseen by Moskalenko and Strong [43] is not compatible with the inferred electron LIS obtained with the most recent experiments. Instead, the electron LIS may suggest that the measured cosmic ray spectra are due to a recent and nearby supernova (or perhaps a few SNe), whose remnant may even contain the solar system. This would also affect the spectra of the CR ions, that would have a galactic and a local component. Future high statistics measurements of the ions spectra made by AMS-02 (see chapter 4) and other experiments should be able to determine the relative importance of the two components. 36

47 Chapter 3 Cosmic antimatter The discovery of antiparticle solutions of the Dirac s equation (1929) was soon followed by the experimental discovery of the positron by Blackett and Occhialini (1932). Since then, it was experimentally established that whenever we create new particles in laboratory, they come in two different forms that are well balanced, generically called matter and antimatter. In particular, the creation (and the annihilation) of fermions is governed by few conservation laws, the baryonic and leptonic numbers conservation being the most important. These laws say that if we create few fermions, each with a positive baryonic (or leptonic) number, in the same reaction other fermions will be created, each with negative baryonic (or leptonic) number, in order to keep the total baryonic and leptonic numbers constant. In the simplest case, this means that we cannot create a single fermion: we must create a couple of particle and antiparticle at least. When we think about the creation of the present universe, our first attempt would lead naturally to a symmetric cosmology in which matter and antimatter are present in the same amount, but astrophysical measurements say that we live in a big domain that seems to be completely made of matter. Of course, the possibility exists that the universe is symmetric on average, but may consist of a collection of homogeneus domains separated by walls filled with radiation only. An important point is the estimation of the domain size. If they are of the scale of galaxies or galaxy clusters, we may detect antimatter cosmic rays (CR) coming from the nearest domain. On the other hand, if some antistar exists in our Galaxy we may even detect antinuclei with Z > 2. Nevertheless, antimatter cosmic rays do exist and are detected. The two species already measured are antiprotons and positrons, both of secondary origin: they are 37

48 COSMIC ANTIMATTER produced by the CR interactions in the interstellar medium (ISM) or in the Earth atmosphere, and by the annihilation of exotic particles, if any. 3.1 Antimatter What the word antimatter means is not simple, thus we start from its constituents: the antiparticles. If all the characteristics of an elementary particle (i.e. a particle without any internal structure), like its mass, charge, spin are known, then its associated antiparticle is like its specular image: it has the same mass, but opposite charge and spin (and all the other quantum numbers) CPT theorem In modern Physics, particles (and antiparticles) are described by the Relativistic Quantum Theory of Fields, in which a couple of field operators are capable of destroying or creating one copy of each kind of particles in every given state. One can switch between a particle and the associated antiparticle using the charge conjungation operator C: C : ψ(r, t) ψ(r, t) where ψ(r, t) represents a particle quantum field and ψ(r, t) is the corresponding antiparticle. There are two other important discrete operators: the time reversal operator T and the parity operator P (the spatial inversion): T : t t P : r r. Even though in general there is no exact symmetry with respect to these operators, the CPT theorem says that every quantum field theory, relativistically covariant, that admits a minimum energy state and obeys the principle of microcausality (requiring that independent measures can always be done on two spacetime points which are outside each other s light cone), is invariant under the action of C, P and T together, without any dependence from the order they are applied. The strict correspondence between particles and antiparticles is a result of the CPT symmetry. In particular, the fact that their masses are exactly equal is due to the commutative property between CPT and the Hamiltonian operator. In 38

49 3.1 Antimatter addition the CPT composite operator is antiunitary: it relates the S-matrix 1 for an arbitrary process to the S-matrix of the inverse process with all spin threecomponents reversed and particles replaced with antiparticles (quoted from [44], p. 183). This means that the following two probability amplitudes are equal (an overline denoting antiparticles): A(a 1 + a b 1 + b ) = A(b 1 + b a 1 + a ) (the demonstration of this theorem can be found in [45] and [44]) Anti-systems The existence of antiparticles, obtained making C acting upon particles, does not guarantee the existence of bound systems made with antiparticles. In other words, the presence of the C symmetry alone does not imply that our system can simply be replaced by another system with antiparticles in place of particles: to obtain the anti-system we do need the more complex CPT symmetry, that involves also the spatial and temporal reflections, changing indeed the dynamics of the system (not only its composition). The antimatter is composed by compound systems like anti-atoms, that are made of a cloud of positrons surrounding a nucleus containing antiprotons and antineutrons. Due to the C invariance of the electromagnetic interactions, all chemical interactions would be the same as ordinary matter, allowing for macroscopic agglomerates. Howewer, one important point is that the fundamental interactions (the electroweak interaction at least) appear neither to be symmetric with respect to the C and P operators, nor to the composite operator CP, as found by Cronin and Fitch in 1964 (for the references to original works, see [44] or [46] and references therein). The experimental discovery of the antimatter, in the sense of bound systems of antiparticles, was done in 1965 by A. Zichichi and his collaborators at CERN and by S. Ting and his collaborators at Brookhaven [46]. Recently (1996), the 1 In Quantum Field Theory, the scattering process is modeled as a very short and intense interaction, that is able to change the particle state from the initial free motion to a generally different final state (again a free wave). The S-matrix contains the probability amplitude of the transition from any initial state to all the final states, thus representing the most complete description of the scattering process itself. 39

50 COSMIC ANTIMATTER simplest antiatom, the antihydrogen, was obtained and studied at CERN and at FERMILAB (see [47] et references therein). Thus we see that the antimatter is actually a very complex system, even if the CPT theorem guarantees that its behavior is exactly the same as the common matter: the hypothetical humans and things made completely of antimatter would behave, in their anti-world, exaclty as the normal humans and things. 3.2 Cosmic antimatter As far as high energy physics experiments are concerned, matter and antimatter are created in the same amount. The CPT symmetry holds up to a level of 10 12, electronic lepton number violations may occur only below (whereas the limit on the total lepton number, including all families, is below ), and the experimental tests on baryon conservation imply that violations (if any) must be below 10 6 [19]. What about the matter we see in the universe? If we trust that the conservation laws hold true during all the history of the universe, we must admit that somewhere there should be the antimatter that balances the matter we are consisting of. As an alternative, we could state that the baryonic and leptonic numbers were not conserved in the past, i.e. that unknown physical processes may have happened during the first phases of the universe evolution. If the conservation laws were valid for the whole universe life, then it is important to explain the presence of non annihilated matter. Today we can evaluate the ratio between the cosmic background radiation (CBR) photons and the nucleons (plus antinucleons) number as η [19]. The CBR is the result of the annihilation of particles and antiparticles (in nearly thermal equilibrium in the very early phase of the universe) when the temperature became low enough to break the coupling between radiation and fermions. If all the matter in the universe is of a kind only, then η is the relative difference between the amount of matter and antimatter at the epoch in which the radiation uncoupled from the fermions. On the other hand, if the symmetry is conserved, η is the value of the local fluctuation in the quantity of matter and antimatter in that epoch. 40

51 3.2 Cosmic antimatter Can total annihilation be avoided? The non zero value of η is a problem: if the universe is made of matter only (plus the radiation) we have to understand the mechanism that produced such a little asymmetry; if the universe is symmetric, we have to understand how fluctuations could survive and generate the observable structures. The first attempt to answer to the latter question was due to Alfvén [48] (1965). He considered an ambiplasma (a fully ionized plasma consisting of protons, antiprotons, electrons and positrons) and the possible formation of homogeneus cells separated by the radiation emitted from leidenfrost 2 leyers in which all the annihilations happen. Omnès [49] (1971) found that this layer is stable when the magnetic field is negligible: the annihilations cannot disrupt the separation walls. Unno and Fujimoto [50] (1974) applied these results to a specific case, trying to explain quasars as super massive stars composed of matter and antimatter. If the antimatter is a little fraction of the total mass, they showed that it would constitute a domain surrounded by the matter and separated by a leidenfrost layer, like a bubble inside a star. To explain how the leidenfrost layer can be able to keep separate matter and antimatter, Aly [51] (1978) computed the annihilation rate at the boudary in 3 cases: radiative and plasma eras of Big Bang model, strong magnetic field, two intergalactic hot domains. In the meanwhile, Lehnert [52, 53] (1977, 1978) showed that at interstellar densities no well defined boundary can form in presence of neutral gas or in a unmagnetized fully ionized plasma. Only a magnetized ambiplasma could produce a leidenfrost layer, whose thickness is proportional to BT 1/2 : for 10 5 < T < 10 8 K and B = 10 8 T, the wall thickness is about 10 7 m. Thus the walls are very thin (of the order of magnitude of the Earth diameter) and pratically invisible for distant observers. In analogy with the geomagnetic field structure, that has well defined zones separated by current and neutral sheets that can be detected only by spacecrafts traversing them, Alfvén [54] (1979) pointed out that the whole space is likely subdivided in cells. Howewer, Dolgov [55] pointed out that the behavior of the domain walls could be different from the leidenfrost process, or more precisely it could be the opposite: instead of repulsion, the layer could attract matter and antimatter towards the annihilation region. Actually, because electrons, positrons and neutrinos have larger mean free paths than the domain walls thickness, the energy and pressure 2 Leidenfrost is the German term used to indicate the process in which the water drops bounce upon a hot layer without touching it, sustained by their own vapor pressure. 41

52 COSMIC ANTIMATTER density of the annihilation region decreases, increasing the diffusion of matter and antimatter towards each other and amplifying the efficiency of the annihilation. In the framework of the inflation the problem may even be ill-posed, because the size of any conceivable antimatter domain would have been enlarged so much that the domain boundaries are well beyond the radius of the visible universe Antimatter domains It seems possible that mechanisms exist that could produce the formation of separated homogeneous domains during the cosmic evolution, thus allowing the universe to be matter-antimatter globally symmetric. This is an important point, because it does not require any conservation law breaking, but we need an estimation of the domain dimensions. In addition the possibility exists to have antimatter confined into condensed bodies like antistars in our Galaxy. Uniform domains. Steigman [56] (1976) considered homogeneous and uniform domains filled with matter or antimatter and concluded that they should be at least of the scale of galaxy clusters, due to the constraints coming from the measured gamma rays flux. The diffuse gamma ray background flux with energy E > 100 MeV, of the order of 10 5 cm 1 sr 1 s 1, was used by Steigman to infer upper limits for the antimatter fraction in the Local Cluster. For the hot H II intergalactic medium, his upper limit is 10 7, while inside our Galaxy the limits are more stringent: for galactic clouds and inter-cloud medium, in the halo. In a recent paper, Dolgov [55] emphasizes that, even if the baryon/antibaryon asymmetry may be non-uniform in space, allowing for large antimatter domains to exist in the universe, still there is no definite theory: neither the size nor the distance of the domains can be predicted with any certainty. If Steigman [56] estimated that the minimum distance has to be 10 Mpc, Cohen et al. [57] imposed a much stronger limit, under the assumption of a baryo-symmetric universe, concluding that the antimatter domains have to be very distant from us, at least few Gpc. A key point is the smoothness of the cosmic microwave background (CMB) radiation, that requires density fluctuations below 10 4 at scales larger than 15 Mpc, thus implying that, if existing, matter and antimatter domains must be in close contact. But the annihilations products will carry away very efficiently the energy from the contact region, because electrons, positrons and neutrinos have larger mean free paths than the domain walls thickness. Hence the energy and 42

53 3.2 Cosmic antimatter pressure density of the annihilation region decreases, increasing the diffusion of matter and antimatter towards each other and amplifying the efficiency of the annihilation (in contrast with the leidenfrost process) [55]. Had this processes happened after the hydrogen recombination, the domain walls regions would have been strong gamma ray sources. Non observation of this background means that any antimatter region must be near or beyond the horizon. If the annihilation took place before the hydrogen recombination, there would be some effect on the CMB energy spectrum, but no interesting limit can presently be inferred from this effect. Dolgov [55] reviewed few different models, amongst wich it appears that a viable model that could account for the existence of domains smaller than the visible universe is based on isocurvature fluctuations : the initial baryon/antibaryon asymmetry was zero and started to rise only relatively late, due to fluctuations in the baryons density. Baryon (antibaryon) rich regions cooled faster and diffusing photons from hotter regions had the effect to drag those regions away, thus providing a way to get separated matter and antimatter domains. In this model the annihilations could be weak enough to create a universe consisting of possibly large domains separated by thin baryon and antibaryon voids. Again, how large are those domains? Condensed bodies. The diffuse cosmic gamma rays background cannot put an equally stringent limit to the amount of condensed antimatter bodies, like antistars or anti-planetoids. Steigman [56] inferred for the antistars number an upper limit of 10 7 in our Galaxy (i.e of the total stars number). This less stringent limit is due to the fact that if antimatter is confined into compact structures like antistars, it is well separated from the matter environment and is able to survive longer than in gas clouds. An antistar is not expected to be a strong gamma ray emitter, at least if it does not cross a galactic cloud neither it impacts on other condensed bodies. Dudarewicz and Wolfendale [58] (1994) gave as lower limit on the distance of the nearest antistar about 30 pc, and give a 10 3 upper limit on the fraction of antistars in M31. Howewer, they emphasize that the fraction could be of order unity at the Hubble radius, having superclusters and anti-supercluters sufficiently well separated, in order to restore the matter-antimatter global symmetry, even though they conclude that a perfect symmetry appears impossible. Recently, Khlopov [59] suggested the possibility that antimatter stars could have survived since the beginning of galaxy formation: they should be searched 43

54 COSMIC ANTIMATTER for in the globular clusters. In fact, condensation of an antimatter domain cannot form an astronomical abject smaller than a globular cluster, and isolated antistars formation in the surrounding matter is impossible, since the necessary thermal instability would finally favor the total annihilation. Thus antistars can form in an antimatter domain only, and they must constitute today a whole antimatter globular cluster at least. In this case, the antistars in the Galaxy could be found in the roughly spherical globular clusters halo around the Galactic center. The number of globular clusters is 10 2 and each cluster contains about 10 6 old stars. In addition, the antistars are well separated from the rest of the Galaxy, and the upper limit of 10 7 calculated by Steigman under the assumption of a unifomr distribution may be well underestimated. For example Khlopov [60] states that the maximum antimatter wind compatible with the observed γ-ray flux is about 10 5 solar masses, and the flux at the Earth is somewhat lower of the present limits (of order of 10 6 ). 3.3 Can we detect cosmic antimatter? There are few ways in which cosmic antimatter may show itself. Indirect ways are based on the detection of annihilation radiation, or on the measurement of the helicity of photons and neutrinos emitted during non CP invariant processes. The direct way is the detection of antinuclei among cosmic rays Indirect ways Matter-antimatter bodies collisions. Sofia and Van Horn [61] (1974) considered the collision between a star and antimatter chunks (m kg) and found that the annihilations due to the stellar wind are not important and that the annihilation rate is limited by the rate at which the matter is swept out by the chunk due to the stellar radiation. Thus the impact with the star cannot be avoided and the chunk penetrates into the star for 10 6 m before eventually evaporating completely. The chunk would become a hot and expanding antimatter bubble that will return to the stellar surface due to buoyancy in 10 2 s. Annihilations produce charged and neutral pions, and they decay to electrons/positrons of MeV and gamma rays of 70 MeV ( prompt photons). These photons suffer 10 scatterings prior to escape, degrading to energies of hundreds of kev. 44

55 3.3 Can we detect cosmic antimatter? The inverse Compton scattering of those e + /e in the stellar atmosphere ( 10 4 K) will produce a number of 60 kev photons with time constant of 10 3 s ( delayed photons). The signature of such a collision would then be a precursor burst emission line at 500 kev (e + /e annihilation) with a 70 MeV continuum lasting s, followed by the main annihilation burst at 100 kev ( s) and by the inverse Compton photons (E 100 kev, τ 10 3 s). This signature is not very different from the time evolution of many gamma-ray bursts (GRB). The cross section for the chunk capture with relative velocity v at infinite by a star with mass M and radius R is σ π(2gmr)/v 2. A star like the Sun, for v 10 4 m s 1 has σ m 2, and during 1 year sweeps up a volume V = σvt m 3. In the Galactic disk, a sphere of radius 100 pc (V = m 3 ) contains about 10 5 stars, then there will be a collision every years (much greater that the Univerge age). A similar way was followed by Sofia and Wilson [62] (1976), who considered the collision between antimatter asteroids and the Sun, while Alfvén [54] (1979) considered star-antistar collisions, possibly ending in ambistars, i.e. stars with matter and antimatter whose annihilations contribute with the thermonuclear fusion processes to the total emitted power. The collision with small (r < 10 km) bodies in the Solar system and the encounter of clouds with antimatter clouds were considered by Rogers and Thompson [63, 64] (1980, 1982). They found that very small antimatter objects in the Solar system would produce a gamma ray flux of the order of cm 2 s 1, too low to be detectable. In addition, different clouds will not merge. Instead a thin leidenfrost layer will form ( 10 9 m, compared to m scale length for clouds), and annihilation will burn only a very small ( ) fraction of the total mass, resulting in less severe constraints for gamma rays emission than those considered by Steigman [56] (but this argument may not be valid, as Dolgov [55] recently pointed out). Very recently, Fargion and Khlopov [65] considered antimatter meteorites in the solar system, obtaining a limit of on the antimatter to matter ratio. Atually all the interactions between antistars and the matter in our Galaxy are very weak until they remain in a bound system like a globular cluster (this is indeed the reason why they could have survived until now). Thus it make sense [66] to consider the possibility that antistars escape from their cluster, wander through the Galaxy and possibly interact with the galactic matter. Following Binney and Tremaine [67], a star (or antistar) can escape from a 45

56 COSMIC ANTIMATTER cluster in two ways: ejection, in which the escape speed is gained in a single close encounter with another star, and evaporation, in which several distant encounters produce a gradual velocity increase. The former process is negligible when compared to the latter, whose characteristic time can be roughly estimated as t ev 100 t re, where the mean relaxation time t re = years for a globular cluster. Thus the number of stars in a cluster is: N(t) = N 0 exp ( t/t ev ) (3.1) and the time t 1 elapsed before one star can escape is found by solving the equation N 0 N(t) = 1, that is: t 1 = t ev ln[(n 0 1)/N 0 ]. (3.2) The number of stars in a globular cluster is N , thus we get t 1 = years. Because this time is much shorter than the age of the Galaxy, it is very likely that, if at least one of the galactic globular clusters is made of antimatter, there are many (possibly thousands) antistars wandering near the roughly spherical volume in the center of the Galaxy occupied by the globular clusters. Those antistars may interact wih a matter cloud, star or smaller compact body. An important effect that has to be considered when gaseous material is accreting into an antistar is that the equilibrium between the gravitational and radiation pressure is reached at higher power than the Eddington luminosity L Edd = 4πGMm pc σ T M M erg s 1 (3.3) (M = kg is the solar mass) because when annihilation photons are considered, the Thomson cross section σ T has to be substituted with the relativistic Klein-Nishina formula. This cross section, for photon energies much higher than the electron rest mass energy (m e c 2 = 511 kev), can be approximated by σ KN πr2 e ɛ ( ln 2ɛ + 1 ), ɛ 1 (3.4) 2 where r e = m is the classical electron radius and ɛ = ( ω)/(m e c 2 ) is the ratio between the photon energy and the electron rest mass energy. For (50 70) MeV photons, typically produced by the decay chains of the charged 46

57 3.3 Can we detect cosmic antimatter? and neutral pions arising from the nucleon/antinucleon annihilation, the Klein- Nishina formula gives for the cross section smaller values than the classical Thomson value (σ T = 8πr 2 e/3 = m 2 = barn). Thus, the annihilation photons pratically do not contribute to the total radiation pressure, which still depends on those photons coming from standard processes such as thermal radiation from the stellar surface, e.m. emission from the accretion disk or diffused photons coming from the nucleosynthesis in the stellar core. Instead, the annihilation photons may escape almost freely, taking away a considerable fraction of the total emitted energy: the net effect is that the power emitted by a matter-accreting antimatter-star can become much greater than the usual accretion case, especially in the gamma-ray regime, where normal stars have negligible emission. Hence, a possible search for matter-antimatter accretion systems could be carried on by comparing optical, X-ray and γ-ray luminosities of Galactic sources: the signature would be an excess of emitted power in the γ-ray range [68]. On the other hand, the rare star/antistar head-on collisions would produce an intense energy release for few seconds, due to the surface annihilations, before merging and reaching a (probably super-eddington) stationary luminosity [66]. These close encounters may appear as GRB. Polarization of e.m. emission. With a completely different approach, Cramer and Braithwaite [69] (1977) stressed out that in addition to direct annihilation, antistars may be distinguishable by the polarization properties of their electromagnetic emission. In fact the ordinary thermonuclear reactions which occur in stars systematically convert protons into neutrons through the weak-interaction process of β + decay and electron capture. When positrons (β + ) are emitted, they are preferentially in a right elicity state of strength v/c. Their bremsstrahlung emission is then right-circularly polarized. The same is true also for the forward going annihilation photon. In antistars, antiprotons are converted into antineutrons, producing electrons in a left elicity state. The photons produced by those electrons are then left-circularly polarized. During normal star processes, the photons take roughly 10 6 years to diffuse out of the star and they loose the initial polarization state, but during supernova explosion the photons produced by the 56 Ni decay chain could be detectable. 56 Ni decays by electron capture to 56 Co, which decays by electron capture or positron decay to 56 Fe. The emitted positrons will radiate through bremsstrahlung polarized photons at the surface of the ejected material. These gamma rays may then 47

58 COSMIC ANTIMATTER escape and be detectable. Hence, a measurement of their degree of polarization could tell us the nature of their origin. Supernovae neutrinos. Finally, Barnes et al. [70] (1987) suggested that the initial neutrino bursts from a supernova could reveal whether the source is made of matter or antimatter. In the first (2 10) ms the neutronization reaction e + p ν e + n produces a erg burst of 10 MeV neutrinos, whose flux cuts off abruptly when the infalling matter achieves sufficient density to trap them. This dense infalling matter comes to thermal equilibrium, in which all neutrino flavors are produced. Neutrinos and antineutrinos, approximatively in the same number, carry away 99% of the binding energy of the newly formed neutron star. The electron neutrinos (and antineutrinos) suffer more scatterings than muon and tau neutrinos, and escape with a mean energy of 10 MeV, roughly half than the muon and tau neutrinos mean energy. On the other hand, the produced ν e (and ν e ) number is roughly twice the ν µ or ν τ numbers. The net effect is that the energy of thermal neutrinos is equally divided amongst the three flavors. In water Čerenkov detectors, like (Super)Kamiokande and IMB, all neutrino flavors may interact by ν e scattering, while electron antineutrinos have an additional channel, the inverse β-decay on the hydrogen nuclei (the interaction cross section for oxygen is negligible): ν e + p e + + n. The ratio between the ν e emitted during the burst phase and the number expected from the thermal phase is r = and the expected counting rate for the 10 MeV electron neutrinos and the 20 MeV muon and tau neutrinos follows the proportion: ν e p : (all thermal ν, ν)e : (burst ν e )e = 10 : 1.1 : 3.3r. (3.5) If the progenitor star is made of antimatter, an important difference arises with this picture: the initial burst is due to the antineutronization reaction e + +p n+ν e and the burst contains electron antineutrinos rather than neutrinos. The ν e cross section in water is 18 times higher than the ν e one, and the proportion (3.5) has to be replaced with: ν e p : (all thermal ν, ν)e : (burst ν e )p = 10 : 1.1 : 60r. (3.6) Thus (6 20)% of all oberved events from an antimatter supernova are expected to occur within the first few milliseconds. In addition, the (ν e p) reaction produces electrons with nearly isotropic cross section, while the elastic scattering (ν e e) is 48

59 3.3 Can we detect cosmic antimatter? peaked forward. Hence the expected signature for an antimatter source is an initial burst in a water Čerenkov detector with isotropic distribution. From supernova SN 1987A, located in the Large Magellanic Cloud at 55 kpc from Earth, 11 and 8 events were registered by Kamiokande II and IMB respectively. Due the too low statistics, it is impossible to distinguish between the expected 2 (ν e p) events in case of antimatter star and the expected 0.1 (ν e e) events corresponding to a matter progenitor star. The first event registered by Kamiokande is forward peaked and if it is attributed to the burst it may prove that SN 1987A was produced by a matter progenitor star (see [70] and references therein) Direct detection Direct detection of CR antihelium nuclei would be a very strong indication of the existence of cosmic antimatter: He could be of primordial origin or even be produced by the antiproton fusion in the core of an antistar. Antihydrogen is of course expected as the most abundant element of antimatter domains, but secondary p production in CR interactions with ISM is an overwhelming source of background for any conceivable cosmic antimatter search. The measurements of positrons are even less significative for this search, because positrons (and electrons) are commonly produced during the CR propagation in the ISM, and in addition they loose energy very rapidly, making impossible to probe distances of cosmological interest. Recently, Khlopov [59] suggested the possibility that antimatter globular clusters could have survived since the beginning of galaxies formation. The idea that one antimatter globular cluster may be present in our Galaxy refreshed the interest into the possible observation of cosmic antimatter effects. There are several possible ways in which such an antimatter globular cluster could manifest itself: its e.m. emission may show anomalous circular polarization at all wavelengths, unrelated to any linear polarization which may be present (see Cramer and Braithwaite [69]); their antistar wind would hardly produce detectable reactions with the galactic ISM but they may interact with matter clouds, stars [54] or smaller bodies [61]. But the most important effect may be the detection of antinuclei with Z > 2, that were produced only in negligible quantities during the primordial nucleosynthesis. 49

60 COSMIC ANTIMATTER 3.4 Antimatter cosmic rays If a non zero amount of antimatter did survive the primordial annihilation, it is reasonable to expect that its composition will be similar to that of ordinary matter. Hence, we may think about antimatter domains as composed by protons, positrons, antihelium nuclei, few isotopes of antihydrogen and antihelium, very few heavier nuclei (most of which would be 7 Li). Antistars could have formed inside antimatter domains exactly in the same way as ordinary stars formed in matter domains. Thus we may expect that nuclear reactions happen inside antistars, of the same kind (apart from photon polarization and antineutrino production, as seen in 3.3.1) of normal reactions: protonproton and C-N-O chains. As for matter domains which contains stars and galaxies, antimatter cosmic rays would be produced and accelerated inside antimatter domains, and a fraction of them (that depends on particle momentum and distance from us) could escape from those domains and reach our Galaxy, where they would continue to diffuse for a long time before annihilation can happen, because the interaction length ( 60 g/cm 2 for protons and antiprotons) is greater than the escape length ( 5 g/cm 2 ). Thus, a finite probability exists that cosmic ray detectors in the Solar system may reveal cosmic antimatter. Actually, such instruments would certainly detect antiparticles produced by the interactions of cosmic rays with the interstellar medium. This background can be completely overwhelming for certain kinds of cosmic antiparticles, but this is not the case for antihelium and heavier antinuclei Positrons and antiprotons Protons are the most abundant particles amongst cosmic rays, and CR electrons are about 1% of protons. Very likely their antiparticles would be the most abundant species in antimatter domains, and we may expect that they would constitute the greatest antimatter fraction among cosmic rays detected on the Earth. Other sources of antiprotons and positrons are the reactions of cosmic rays with the interstellar medium. In fact, among the secondary particles produced by energetic inelastic scatterings between two protons (the most abundant species both in CR and ISM) or a proton and a nucleus, the most abundant ones are mesons, like pions and kaons, and antiprotons. In addition, while neutral pions dacay into energetic photons (E γ = 70 MeV in the CMS), charged pions decay 50

61 3.4 Antimatter cosmic rays into muons and electron-positron pairs (also produced by muon decays), so that the secondary production of antiprotons and positrons is a quite common process. Like electrons, positrons have short radiation length and suffer heavy energy losses during propagation in the ISM, hence there is no possibility that CR positrons be of cosmic origin: they are produced by the interactions of cosmic rays with the interstellar medium. On the other hand, antiproton production is hardly disfavoured for energies below 2 GeV for kinematical reasons, so that the secondary antiproton spectrum should have a characteristic peak around 2 GeV (for higher energies, it is the primary proton spectrum that goes down as E 2.8, while the p production yield is almost constant). Thus, cosmic antiprotons (and antiprotons from exotic sources as dark matter particle annihilation [71]) may be searched at low energies or above the secondary peak. Figure 3.1 shows the experimental results Antihelium and antinuclei After hydrogen, helium is the most abundant species in the universe (about 25% of the total baryonic mass, and about 20% of CR particles are He nuclei. 4 He nuclei are of cosmic origin (produced during primordial nucleosynthesis) and of stellar origin (result of the proton-proton nuclear chain). After their acceleration, helium nuclei propagate through the Galaxy for a time similar to the proton propagation time (about years), and may interact with the interstellar medium, producing by spallation the 3 He isotopes. Similarly, we expect that the greatest fraction of CR antinuclei (after antiprotons) is constituted by antihelium isotopes. Actually, the possible detection of He would be a striking demonstration that antimatter plays a cosmic role, as annihilation remnants wandering through the Galaxy or in form of antistars: the secondary production probability of 3 He by cosmic ray interactions with the ISM was estimated to be of order [73] and the probability for secondary 4 He is much lower. While antihelium may be of cosmic or (anti-)stellar origin, the detection of antinuclei could be explained only as a demonstration that antistars do exist in our Galaxy (or may be in some nearby galaxy). Among the possible isotopes, the best candidates for this antimatter search are 12 C, 14 N and 16 O, because they are the most probable production results (after 4 He) of nuclear reactions fueling antistars. Figures 3.2 and 3.3 show the experimental upper limits found by balloon and space experiments (AMS included, see 4.3) on the cosmic ray anti-helium to 51

62 COSMIC ANTIMATTER Figure 3.1: Experimental results on the CR antiproton flux [72]. helium flux ratio and antimatter (i.e. antinuclei) to matter flux ratio, respectively. 52

63 3.4 Antimatter cosmic rays Antihelium/Helium Flux Ratio (95% C.L.) (a) Buffington et al (b) Golden et al (c) Badhwar et al (d) AMS 01 (1998) (e) BESS ( ) (b) (a) (e) (c) (d) (c) Rigidity (GV) Figure 3.2: Experimental results on the CR antihelium-to-helium flux ratio [74, 75, 76, 77, 78]. 53

64 COSMIC ANTIMATTER Antimatter/Matter Flux Ratio (95% C.L.) (a) Aizu et al (b) Greenhill et al (c) Golden et al (d) Smoot et al (e) AMS 01 (1998) (a) (c) (b) (c) (d) (e) (c) (d) (c) Rigidity (GV) Figure 3.3: Experimental results on the CR antimatter-to-matter flux ratio for atomic numbers Z > 1 [79, 80, 81, 82, 83]. 54

65 Chapter 4 The AMS experiment The Alpha Magnetic Spectrometer (AMS) [84] is a particle detector that will be installed on the International Space Station (ISS) in 2005 (NASA shuttle flight UF-4.1) to measure Cosmic Ray (CR) fluxes for at least three years. During the precursor flight aboard of the shuttle Discovery (NASA STS-91 mission, 2 12 June 1998), the test detector AMS-01 was operated for about 180 hours, collecting over one hundred millions CR events [77, 85, 40, 86, 87]. Section 4.2 below describes the AMS-01 detector, while the published Physics results from the STS-91 mission are summarized in section 4.3. Section 4.4 describes the improved version of the detector, called AMS-02, that will operate aboard of the ISS. In addition to a refined silicon tracker and to redesigned time of flight (TOF) and anticoincidence systems, a proximity focusing ring imaging Čerenkov (RICH) detector will substitute the threshold Čerenkov counter of AMS-01, and two additional subdetectors (a transition radiation detector and an electromagnetic calorimeter) will be added to improve the proton to electron separation capability of the instrument. 4.1 Particle identification In order to measure the fluxes of CR particles, the detector has to be able to measure their charge, velocity and rigidity R = pc ze = γβm 0c 2, (4.1) ze where p and ze are the relativistic momentum and the particle charge respectively, m 0 is the particle rest mass, c is the speed of light in vacuo, the particle velocity is 55

66 THE AMS EXPERIMENT v = βc and γ = (1 β 2 ) 1/2 is the Lorentz factor. In principle, when z, β and R are known, it is possible to find the particle rest mass m 0 from the relation (4.1): the particle identification is complete. In real cases, the measurement uncertainties on z, β and R may allow mass discrimination between two given particle species only in some rigidity (or energy) range. In particular, it is easier to distinguish electrons from protons (m p 1837m e ) than protons from deuterons (m d 2m p ), and the latter are better separated than any other nuclear isotopes of the same element Rigidity measurement A particle with charge z and rigidity R, moving through a region where a uniform magnetic field B exists, will follow a helix with curvature radius r = R sin θ, (4.2) Bc where B is the field intesity and θ is the angle between the particle momentum and the magnetic field. If the field is not perfectly homogeneous, like the case of the AMS detector, the trajectory will be more complicated but, in any case, it will depend only on the particle instantaneous rigidity and the local magnetic field. Hence, if the field is known, in order to measure the particle rigidity one has to reconstruct its trajectory, keeping into account the energy eventually lost in the interactions with the detector (that will decrease the instantaneous curvature radius). The tracking system of AMS is a silicon detector with N planes (N = 6 for AMS-01 [ 4.2.2], N = 8 for AMS-02 [ 4.4.2]) placed in the inner bore of the magnet, that are able to measure the (x, y) coordinates of the crossing position and the energy lost by the particle by ionization of the active material. If the spatial resolution of the tracking system is σ pos and the magnetic field strength along the particle trajectory is s mag = B dl, the relative uncertainty on the rigidity is [88]: R R Rσ pos 1 (4.3) s mag N + 4 (that is, the deflection η = 1/R is Gaussian distributed). It is customary to define the maximum detectable rigidity (MDR) the rigidity for which the measurement uncertainty is 100% (that is R/R = 1). The MDR for AMS-01 for protons is 150 GV (see 4.2.2) and AMS-02 will reach 1 TV. 56

67 4.1 Particle identification When the particle rigidity is comparable with the MDR of the instrument (i.e. the deflection tends to zero), it becomes likely that the wrong deflection sign will be attributed to the trajectory ( spillover ). The spillover is a problem for two reasons: first, it produces false antiparticles due to the wrong sign of the measured deflection; second, it leads to a distortion of the measured rigidity spectrum, that will appear steeper due to the disappearance of events from the highest rigidity bins. When the particle rigidity is known, in order to get its momentum it is necessary to measure its charge. The sign of the charge is found by looking at the track curvature and direction in the magnetic field, while the absolute value is given by the energy lost in the active parts of the detector Charge measurement The basic way to measure the particle charge is to measure its energy deposition in the materials that constitute the active detectors. In AMS-01 this is possible with the plastic scintillator counters of the TOF system ( and chapter 5) and the silicon layers of the tracker ( 4.2.2). With AMS-02, in addition to the TOF and tracker measurements, also the electromagnetic calorimeter ( 4.4.7) the transition radiation detector ( 4.4.6) and the RICH ( and chapter 6) will measure the energy lost by the particle traversing the detector. The energy lost by a particle with charge z and velocity βc after a path length dξ = ρ dx inside a medium with density ρ, atomic and mass numbers Z and A respectively, is given by the corrected Bethe-Bloch formula [5]: de dξ = K Z [ z 2 1 A β 2 2 ln 2m ec 2 (βγ) 2 T max β 2 δ ] (4.4) I 2 2 where K = MeV g 1 cm 2, I is the mean ionization energy of the medium, δ is the density effect correction (usually computed using the Sternheimer parametrization [89, 90]), and T max = 2m e c 2 β 2 γ γm e /M + (m e /M) 2 (4.5) (T max 2m e c 2 β 2 γ 2 for all CR particles but electrons). The shell correction to (4.4) is omitted here because its small effect is sensible only at energies below the AMS range [5]. 57

68 THE AMS EXPERIMENT For electrons and positrons the argument of the logarithm of formula (4.4) in the ultrarelatvistic case becomes proportional to γ 3/2 [91], but the dominant process is the bremsstrahlung. The energy loss is described by equation (2.2), that we rewrite here in a simpler form: ( de ) = NE 0 Φ rad, (4.6) dx el where N is the atomic density (in cm 3 ) of the material, E 0 is the initial electron (positron) energy and Φ rad is a function of the material only [92]. Formula (4.4) can be used to infer the particle charge z from the energy deposition measurements of the TOF and tracker layers, and inside the calorimeter, if the particle velocity β is known. In order to measure the charge sign, it is necessary to have a measurement both of the particle direction and curvature. The curvature is measured by the tracker and its direction can be found by the TOF system Velocity measurement The particle velocity can be measured in two ways by AMS-02: through the time of flight measurement and the Čerenkov cone opening angle. The first method is used by the TOF system: a particle with velocity v = βc takes a time t = l/v to go along the path l = L/ cos θ between upper and lower TOF planes (L is their distance and θ is the trajectory colatitude angle). Hence, the time of flight is: L t = (4.7) βc cos θ and its uncertainty σ t will be Gaussian. The uncertainty on β will be: σ 2 β = L2 σ 2 t c 2 t 4 cos 2 θ + σ2 θ sin2 θ L2 σ 2 t t 2 cos 4 θ c 2 t 4 cos 2 θ, (4.8) because the second term inside the parentheses can be safely neglected both in AMS-01 and AMS-02 thanks to the very good angular resolution of the tracker. The time resolution of the TOF system is of order of 0.1 ns, hence the time measurement can be used to infer the particle velocity up to about β On the other hand, a direct measurement of β can be done by the RICH detector of AMS-02, that will have β/β 0.1%. 58

69 4.2 The AMS-01 detector The Čerenkov radiation, produced by particles crossing a medium with velocity greater than the light speed in that medium, is emitted at an angle α with the particle momentum that depends only on the particle velocity and the medium refractive index ( 6.1): cos α = 1 βn. (4.9) Hence the threshold below which there is no Čerenkov emission at all is β min = 1/n, and the measurement uncertainty is: σ 2 β = σ 2 n n 4 cos 2 α + σ α sin 2 α n 2 cos 4 α. (4.10) At very high energies, the particle velocity could in principle be measured by means of the transition radiation opening angle χ 1/γ = (1 β 2 ) 1/2, hence: σ 2 β = χ2 1 χ 2 σ2 χ or σ γ = σ χ χ 2. Howewer, this angle is very small and in pratical applications it is very difficult to be measured. Instead, a common practice is to measure the energy deposition in the detector, that has different shapes for particles above and below the threshold [93] (the transition radiation photons have energy that depends on the βγ of the incident particle and contribute to the detected signal in addition to ionization energy losses). 4.2 The AMS-01 detector Figure 4.1 shows the detector AMS-1 that was flown aboard of the shuttle Discovery on 2 12 June The core of the instrument is a permanent Nd-Fe-B magnet enclosing the tracking system (six silicon planes) and an anticoincidence counter (ACC) system. Above and below the magnet, four layers of scintillator counters give the fast trigger to the experiment and measure the time of flight (TOF) of the particles traversing it, while an aerogel threshold Čerenkov (ATC) counter allows to discriminate between protons and electrons up to a rigidity of 3.5 GV. 59

70 THE AMS EXPERIMENT Figure 4.1: The AMS detector for the STS-91 mission (AMS-01) The magnet The AMS-01 permanent magnet is a Nd-Fe-B cylinder whose axis defines the detector z-axis. It produces a nearly uniform field along a transversal direction, and the magnetic vector B defines the AMS x axis. The y axis is defined by completing the right-handed unit vectors triplet. The cylindrical magnet shell has a magnetization vector with constant modulus and angular direction α in the (x, y) plane given by: α = 2φ + π 2 (4.11) where φ is the longitude angle of cylindrical coordinates [94]. This configuration produces an internal field of intensity: B = B r ln r i r o (4.12) 60

71 4.2 The AMS-01 detector Glue Layers NdFeB Block X Y Z B Z Layers ϕ r 1 r 2 Figure 4.2: The AMS-01 magnet block numbering (left) and field configuration (right). B (G) x Z (cm) Figure 4.3: The AMS-01 magnetic field B x component as function of z. where B r is the residual magnetic flux density of the ring and the inner and outer radii are r i = 55.7 cm and r o = 64.9 cm, respectively. In order to build the magnet structure, the first step was to produce 2 cm thick Nd-Fe-B layers with uniform magnetization M. Then a series of small cylider 61

72 THE AMS EXPERIMENT sections ( 5 5 cm 2 ) were cut away with 64 different orientations with respect to M. Each block has then a uniform magnetization M i and can be glued to blocks i 1 and i + 1 in order to produce a ring with magnetization directions α i given by equation (4.11). The whole magnet is made up of several glued rings (figure 4.2), reaching the height of 80 cm. The magnetic field produced in the bore of the magnet is roughly constant (B x 1.6 kg) within ±30 cm along z axis, and decreases rapidly outside the bore (for z > 40 cm), reaching the level of 200 G for z 65 cm (see figure 4.3). The nominal bending power of the magnet is BL 2 = 0.14 T m 2, when considering the average B x over z < 71.4 cm [94]. Its geometrical acceptance is 0.82 m 2 sr and its weigth is 1998 kg The tracking system The silicon tracker is composed of 300 µm thick double-sided micro-strip sensors of area mm 2, based on the design used by the micro-vertex detectors of ALEPH and L3 experiments at CERN, that were successfully operated with the Large Electron-Positron (LEP) collider (see [95] and [94]). The readout strip pitches are 110 µm along the y direction (p-side) and 208 µm along the x direction (n-side), respectively the bending and non-bending directions. The silicon sensors are grouped together in ladders (7 to 15 sensors) of different length, assembled on 6 ultra-light honeycomb planes (see figure 4.4). For vertical tracks, the total amount of matter of the 6 planes is about radiation lengths. The tracker was not fully assembled for the test flight in 1998: in total there were 28 ladders on the inner planes and 34 ladders on the outer ones. The ladders were read by 168 electronic boards featuring charge measurements via a sampleand-hold technique with 64-channels readout chips whose linearity was good up to 75 MIP signals. Analog output signals from such boards were sent to 12-bit lowpower fast ADCs installed on the electronic boards in charge of data reduction and pedestal subtraction. The particle crossing position is determined by a clustering algorithm that finds the center of gravity of the energy loss measured by a set of strips. For example, the x coordinate of the crossing position in a given plane is computed using the formula: x = Σ ia i x i Σ i A i, (4.13) 62

73 4.2 The AMS-01 detector Figure 4.4: The AMS-01 tracker. where A i is the amplitude of the signal recordered by the strip at the position x i. In order to select only small sets of strips for this weighted mean, the tracker data reduction (TDR) electronics perform a search of seed strips using a threshold of 3.5σ ped, then consider all neighboring strips with signal greater than 1σ ped. The n-side strips have a noise level about 50% higher than the p-side level, resulting in a signal-to-noise ratio for the non-bending direction (x) approximatively half of that of the bending direction (y). The rigidity resolution of the tracker (shown in figure 4.5) is limited by multiple scattering at low energies (up to about 10 GeV for protons), while at high energies it is limited by the magnet bending power and the tracker spatial resolution. The momentum resolution is at the 7% level between 1 and 10 GeV/c, while the MDR is about 150 GV for protons. 63

74 THE AMS EXPERIMENT R/R Proton beam test Data (Z=1) MC R (GV) 10 2 Figure 4.5: The AMS-01 tracker rigidity resolution predicted by Monte Carlo simulations, compared to the measurements done with a proton beam and with CR protons The time of flight system The TOF system [96] was completely designed and built in the INFN Laboratories in Bologna. Its main goals are to provide the fast trigger to AMS readout electronics, and to measure the particle direction, velocity (β), position and charge. In addition, it had to operate in space with severe limits for weight and power consuption. Each TOF plane consists of 14 counters (Bicron BC408 plastic scintillator) 1 cm thick covering a roughly circular area of 1.6 m 2. The scintillation light is guided to 3 Hamamatsu R5900 photomultiplier tubes (PMT) per side, whose signals are summed together to provide a good redundancy and light collection efficiency. The total power consumption of the system (112 channels, 336 PMTs) was 150 W, while its weight (support structure included) was 250 kg. 64

75 4.2 The AMS-01 detector Figure 4.6: The upper two planes of the AMS-01 time of flight system. Time of flight resolution. The single channel time resolution is [96]: σ 2 1 σ(x) = N + σ2 2 x2 N + σ2 3, (4.14) where x is the distance of the particle crossing point from the PMT, N is the number of photons which convert into electrons ( photoelectrons for brevity) on the PMT photocathode, σ 1 depends upon the PMT signal shape and the trigger electronics, σ 2 takes into account the dispersion in the photon path lengths and the constant term σ 3 depends on the electronic noise at the low threshold discriminator input and on the reference time dispersion on each channel. The overall time resolution of a plane can be determined by measuring the time of flight of ultrarelativistic (β 1) particles between two given planes, after correcting for the track length. The time dispersion is expected to decrease with the nuclear charge Z, due to the large number of photoelectrons produced by nuclei with high atomic number, until it reaches the minimum value σ 3. Figure

76 THE AMS EXPERIMENT 125 σ pl (ps) Charge Figure 4.7: AMS-01 TOF: mean single plane time resolution as function of the particle charge derived from STS-91 data [97]. shows the average single plane time resolution as function of the particle charge, using data from the STS-91 flight: the horizontal line show that the limiting level σ 3 is 88 ps [97]. The velocity resolution of the TOF system, σ(β)/β 3%, allows to discriminate p/e + and p/e up to a rigidity of 1.5 GV. Particle separation. One of the main purpose of the TOF system is the measurement of the time of flight of the particles traversing the detector with a resolution sufficient to distinguish upward from downward going particles: an upwardgoing helium nucleus wrongly labelled downward-going would be interpreted as an downward-going anti-he nucleus. The average time of flight of the particles which traverse the detector is of the order of 5 ns, while the time measurement has a resolution σ t 120 ps, independent from the rigidity. Hence the distribution of β 1 for all STS-91 events (figure 4.8) shows two populations peaked at ±1, with Gaussian profiles towards zero (where no physical particle is expected). No ambiguous event was detected. Thus the probability to mistake the particle direction due to non-gaussian effects is well below 10 8, the level reachable with the STS-91 statistics. In addition to the capability to separate downward-going from upward-going particles, one goal of the TOF system was to provide a special flag for ions at the trigger level. Accordingly, it was designed to distinguish in a fast and efficient way cosmic ray protons from other nuclei. 66

77 4.2 The AMS-01 detector Figure 4.8: The distribution of β 1 for all (unshaded) and selected He nuclei for the antimatter search analysis (shaded) [98]. The TOF system provides a measurement of the absolute charge of the crossing particle in addition to the tracker, even if, due to the strong constraints about power consuption, the TOF front-end electronics was not optimized for energy deposition measurements. The charge measurement was realized through a time-over-threshold method, whose response is proportional to the logarithm of the deposited charge. This method results in a good separating power ( ) between singly and doubly charged particles but has a poor charge resolution for Z > 2. The stability of the charge measurement was very good for all the 112 TOF channels, but five channels, as shown in figure 4.9 [97] The anticoincidence system The AMS-01 anticoincidence system has 16 plastic scintillator counters (of the same material as the TOF ones) whose shape is a cylinder section, placed along the inner wall of the magnet, containing the 4 inner tracker planes (figure 4.10). Each paddle is 1 cm thick, 20 cm large and 80 cm high. The counters are read by one photomultiplier in each side of the same type 67

78 THE AMS EXPERIMENT plane 1 plane 2 plane 3 plane 4 Charge dispersion (rms, %) TOF channel Figure 4.9: TOF charge measurement stability during the STS-91 flight [97]. of the TOF ones (Hamamatsu R5600U), connected to the scintillator through a bundle of optical fibers. Their signal is used as veto by the level-1 trigger logics, in order to avoid events produced by the interactions of CR particles in the magnet structure (giving secondaries of opposite charges that are deflected towards upper and lower TOF planes). As side effect, the ACC veto rejects also normal events whose interactions with the TOF or tracker planes produced δ-rays hard enough to reach the ACC counters. This introduces a detector inefficiency that depends upon the particle charge and crossing positions. The maximum allowed energy for δ-ray electrons or positrons is E max = 2β 2 γ 2 m e c 2 but their production spectrum is very steep (P(E) de E 2 de), so high energy δ-rays are rare [88]. Due to their small curvature radius (3 mm for a 10 kev electron in a 1 T magnetic field), δ-rays emitted at angles not too large with respect to the magnetic field direction will follow the field lines untill reaching a mirror point or a material where they stop or annihilate. Their large range is thus a problem, because they can decrease the detection efficiency in two ways: by hitting one of the ACC counters (producing a veto at the trigger level), or by hitting one or two tracker planes. The latter case may produce events that are not well fitted with a single trajectory, hence they may be discarded by the track quality cuts during the off-line analysis. 68

79 4.2 The AMS-01 detector Figure 4.10: Part of the AMS-01 anticoincidence system The threshold Čerenkov counter The threshold Čerenkov counter mounted below the magnet of AMS-01 (figure 4.11) consists of two layer of 8 10 (upper) and 8 11 (lower) aerogel cells with refractive index n = 1.035, directly attached to the universal support structure (USS, figure 4.12), not to the magnet like the other subdetectors. Each cm 3 cell consists of eight 1 cm thick slabs, it is wrapped with three 250 µm reflecting Teflon foils and it is viewed by one Hamamatsu R5900 PMT placed right below (figure 4.13). A wavelength shifter is used to convert 300 nm photons to 420 nm photons matching the maximum efficiency range of the phototubes. The upper plane cells are placed over the separation between the lower cells, in order to reach a good mechanical rigidity and to avoid holes in the ATC acceptance. The ATC is used to extend the separation capability between protons (antiprotons) and positrons (electrons) up to 3.5 GeV. The number of photons produced 69

80 THE AMS EXPERIMENT Figure 4.11: The AMS-01 aerogel threshold Čerenkov counter layers (only half of them is displayed). by a particle with charge Z traveling for a distance L aero inside the aerogel (with velocity β above the Čerenkov threshold) is: ( N ph L aero Z ) n 2 β 2, (4.15) where n is the refractive index of the radiator. Aerogel was chosen because its refractive index n = 1.035±0.001 allows the discrimination of protons and electrons up to 3.5 GeV/c, still yielding an acceptable number of produced photons. A calibration carried on with p 15 GeV/c protons (β 0.99) gave the mean values of the number of detected photoelectrons: N phel = 3.51 ± 0.02 for the upper ATC layer and N phel = 4.02 ± 0.02 for the lower layer. The channel to channel dispersion is (10 15)%. 70

81 4.2 The AMS-01 detector Figure 4.12: The AMS-01 detector inside the Universal Support Structure The AMS-01 trigger The trigger is the digital signal that starts the data acquisition (DAQ) chain. By extension, this terms is often used to represent the logic conditions required for it to be generated by the trigger electronics. In high energy Physics experiments it is customary to define three types of trigger logics, depending on the type of data they act on: the first level trigger (including a pre-trigger called fast trigger) is generated imposing conditions on fast signals only (i.e. on logic signals coming from discriminators), the second level trigger is generated after the first level trigger if digitized data of any single subdetector satisfy the required conditions, and the third level trigger follows the second level trigger signal when the digitized data of the whole detector are checked against the desired set of conditions. The AMS-01 trigger has no second level logics. The fast trigger (FT) logics process the scintillator data and provide, in about 50 ns, the time zero for the 71

82 THE AMS EXPERIMENT L aero Wavelength Shifter (PMP) d PM dpmt Teflon Layers (3) Light Guide PMT PMT Figure 4.13: The AMS-01 ATC cell. time-of-flight measurement. Then the first level trigger rejects events with hits on the anticoincidence counter system and enhances the fraction of particles crossing the tracker planes through the analysis of the pattern of hit counters in the upper and lower TOF planes ( matrix condition). Finally, at the last level (the third one because global digitized data are checked) the trigger logics suppress spurious fast triggers and find preliminary good tracks on the silicon tracker. In order to characterize the effect of every trigger condition, during the STS- 91 flight AMS-01 collected one event requiring the fast trigger alone every 1000 normal triggered events. This sample (called prescaled events data set) can be considered unbiased with respect to the FT efficiency, to be measured in a different way, and was used to measure the effect of all other trigger conditions on cosmic ray protons, helium nuclei and electrons 1 (figure 4.14). The FT signal is generated when at least one counter side in 3 (over 4) different TOF planes produces a signal above a threshold corresponding to 40% of a minimum ionizing particle (MIP) crossing the counters center. The efficiency of the fast trigger could be measured with the same data taken during the STS-91 mission using two different ways: either by looking at events triggered by a given set of 3 TOF planes and searching for a signal in the other ( spectator ) plane that would pass the required threshold, or using as unbiased sample the signals 1 For ions with Z > 2 the prescaled events gave too low statistics. 72

83 4.3 Results of the STS-91 mission Prescaled events (100%) all triggers (99.5%) (85.1%) (35.7%) (32.0%) (14.0%) 4/4 TOF planes (fast trigger OK) matrix condition no anticoincidence veto level 3 TOF OK level 3 tracker OK Figure 4.14: The AMS-01 trigger conditions applied off-line on prescaled events. produced by particles that did not produce the trigger. The latter can be done by exploiting the characteristics of the TOF electronics, sensitive to all particles impinging on the detector in an interval of about 16 µs around the trigger signal. Up to eight hits can be registered by each channel with a time resolution of 1 ns and a full charge measurement. The analysis of these unbiased data provided the instantaneous rate of particles, the dead time and accidental rate, in addition to the total FT efficiency [99]. The background can be estimated by checking the consistency of the TOF data and the trigger mask, and comes out to be about 0.5% of the fast triggers (due to electronics noise). This background is completely eliminated in the last level trigger by requiring the coincidence of both sides of the same counter. 4.3 Results of the STS-91 mission AMS-01 was successfully flown aboard of the space shuttle Discovery on June 1998 for a ten days mission. The instrument was operating for about 180 hours and collected over hundred millions events. The Collaboration has published the results of the analysis about the cosmic antimatter limits [77] and the flux in the 73

84 THE AMS EXPERIMENT low orbit environment of protons [85] [86], electrons and positrons [40], and helium [87]. Results are summarized in ref. [100] Primary CR spectra The primary cosmic rays are dominated by the proton component, as seen in the figure 4.15, that shows the primary spectra of protons, He nuclei, electrons and positrons measured by AMS-01 [86, 87, 40]. The protons differential spectrum measured by AMS-01 can be fit with a power law in rigidity for R > 10 GV with the form φ(r) = φ 0 R γ ( m 2 s 1 sr 1 MV 1 ) (4.16) with [85]: γ (p) = 2.79 ± (fit) ± (sys) φ (p) 0 = 16.9 ± 0.2 (fit) ± 1.3 (sys) ± 1.5 (γ) ( GV 2.79 m 2 s 1 sr 1 MV 1 ). (4.17) In addition, the He spectrum for 20 < R < 200 GV can be fit by a power law with [87]: γ (He) = ± (stat) ± (sys) φ (He) 0 = 2.52 ± 0.09 (stat) ± 0.13 (sys) ± 0.14 (γ) ( GV 2.74 m 2 s 1 sr 1 MV 1 ). (4.18) Figure 4.16 shows the comparison between the AMS-01 protons spectrum and few recent balloon measurements. The agreement is very good between AMS-01 and BESS (the most recent measurement), but elder instruments gave different spectra, with experimental data points that are generally lower (even by a factor 2) than those of AMS-01 and BESS [86]. The left plot of figure 4.17 shows the number of collected Z = 2 particles as function of the rigidity: no antihelium has been found up to 150 GV. The corresponding upper limit on the fraction of cosmic ray antihelium is shown in the right part. Figure 4.18 shows the measured electron and positron spectra, with the simulated background from protons. While this background makes impossible to say something about the positron spectrum above 1.5 GeV, the measurement of the electron spectrum is very good up to about 30 GeV. AMS-01 accumulated a big amount of data, and the spectra of electrons and positrons have unprecedented small error bars. 74

85 4.3 Results of the STS-91 mission F (m 2 sr 1 s 1 GeV 1 A) AMS 01 protons AMS 01 He nuclei AMS 01 electrons AMS 01 positrons E k /A (GeV) Figure 4.15: Primary fluxes of CR protons, He nuclei, electrons and positrons measured by AMS-01 [86, 87, 40]. The e and e + spectra are shown again in figure 4.19, where it appears clearly how the fraction of positrons depends on the energy. 75

86 THE AMS EXPERIMENT 7 6 This experiment CAPRICE [12] LEAP [13] Flux * E K a) This experiment BESS [14] IMAX [15] b) E K (GeV) Figure 4.16: AMS-01 proton flux compared with balloon experiments [86]. 76

87 4.3 Results of the STS-91 mission Events He 0 events AMS STS - 91 He events Sign Rigidity GV Antihelium/Helium Upper Limit Excluded AMS STS - 91 Z = R GV max Figure 4.17: The AMS-01 z = 2 sample [77] (left), and 95% C.L. upper limit to the relative flux of antihelium to helium [77] (right). 2-1 Flux (m sec sr MeV) e - + Background Background e + + Background Background E k (GeV) Figure 4.18: AMS-01 electron and positron fluxes compared to the expected background [40]. 77

88 THE AMS EXPERIMENT 2-1 ) Flux (m sec sr MeV) a) b) 1 10 E (GeV) k e - + e +e + /(e 0.3 e E (GeV) k Figure 4.19: AMS-01 primary fluxes of electrons and positrons (a) and relative abundances (b) [40]. 78

89 4.3 Results of the STS-91 mission Secondary spectra The wide range of geomagnetic coordinates sampled by AMS-01 during the STS- 91 shuttle flight (θ M < 75 ) allowed for a precise measurement of the dependence of the secondary cosmic ray component on the geomagnetic field. Figure 4.20 shows the measured upward and downward proton flux at different geomagnetic latitudes. It is evident that for small latitudes (i.e. in the equatorial region) the primary component is damped at low energy by the Earth magnetic field, that reflects back the incoming particles with rigidity below a cut-off level that is near 10 GV. This level decreases for increasing latitudes. Howewer, the spectrum shows a dip and starts to rise again below few GV. In addition, this second part of the spectrum is the same for upward and downward protons, indicating that it is due to particles that are looping in the geomagnetic field. Indeed below the geomagnetic cutoff, the secondary fluxes agree in the range 0 Θ M 0.8 [85]. It was possible to back-trace those particles to see if they were coming from the atmosphere (hence secondary products) or from outside (primary component): they are particles of secondary origin, trapped in the geomagnetic field. Figure 4.21 shows the measured helium flux at different geomagnetic latitudes. It is again evident the damping below about 10 GV in the equatorial region, with a cut-off level that is decreasing with increasing latitudes. Also the He shows a second spectrum of secondary origin, as can be inferred from the fact that most of the second component nuclei are 3 He isotopes (figure 4.22). Figures 4.23 and 4.24 show a similar effect for electrons and positrons: below a geomagnetic latitude cut-off in rigidity, a second spectrum emerges that is due to secondary particles looping in the Earth magnetic field. 79

90 Downward 0.2 < θ M < < θ M < 0.2 Downward 0.7 <θ M < <θ M < <θ M < <θ M < <θ M <0.4 Downward 1 < θ M < < θ M < < θ M < 0.9 THE AMS EXPERIMENT Flux (m 2 sec sr MeV) a) Upward 0.2 < θ M < < θ M < 0.2 b) Upward 0.7 < θ M < < θ M < < θ M < < θ M < < θ M < 0.4 c) Upward 0.9 < θ M < < θ M < d) e) f) Kinetic Energy (GeV) Figure 4.20: AMS-01 measured proton spectra at different geomagnetic latitudes [85]. 80

91 4.3 Results of the STS-91 mission Flux (m 2 sec sr GV) < Θ M < < Θ M < < Θ M Rigidity (GV) Figure 4.21: AMS-01 measurements of helium flux at different geomagnetic latitudes [87]. 81

92 THE AMS EXPERIMENT β Primary Second 3 He 4 He Rigidity (GV) Figure 4.22: AMS-01 3 He and 4 He measurements show that the secondary helium flux is dominated by 3 He [87]. 82

93 4.3 Results of the STS-91 mission < Θ M < < Θ M < < Θ M < < Θ M < < Θ M < < Θ M Flux (m 2 sec sr MeV) a) e - b) e < Θ M < < Θ M < < Θ M < < Θ M < < Θ M < < Θ M c) e + d) e E k (GeV) Figure 4.23: AMS-01 measured downward fluxes of electrons (a,b) and positrons (c,d) [40]. 83

94 THE AMS EXPERIMENT < Θ M < < Θ M < < Θ M < Flux (m 2 sec sr MeV) a) e - b) e - c) e < Θ M < < Θ M < < Θ M < d) e + e) e + f) e E k (GeV) Figure 4.24: AMS-01 measurements at different geomagnetic latitudes of downward (full circles) and upward (open circles) going electrons (a,b,c) and positrons (d,e,f) [40]. 84

95 4.4 The AMS-02 detector Figure 4.25: The detector that will be installed on the ISS (AMS-02). 4.4 The AMS-02 detector The AMS-02 detector (figure 4.25) will be installed aboard of the ISS in 2005 (NASA shuttle flight UF-4.1), where it will operate for at least 3 years. Scientific goals of AMS-02 are: improved measurements of the antimatter fraction in cosmic rays; perform a search of exotic particle annihilation signatures over the wide energy range ( ) GeV in protons, electrons and γ-rays spectra; refine experimental results concerning CR spectra with a large acceptance, long duration mission. This detector will have a very long exposure time (at least 3 years), it will traversed by a significant sample of high energy particles, up to few TeV per 85

96 THE AMS EXPERIMENT nucleon. Hence its rigidity resolution must be improved with respect to AMS- 01, in order to reach a MDR high enough to cover the range where a detectable flux of primary cosmic rays is expected in the whole time window. The AMS-02 detector is based on a superconducting magnet generating a higher magnetic field than the AMS-01 permanent magnet. In addition the silicon tracker will have a larger area, with 8 indipendent measurements instead of 6 and improved spatial resolution. The TOF system will be similar to AMS-01: the main differences are the number of counters, that has been decreased to save power and weight, and the PMT model, that must operate in a very high magnetic field. In addition, the shape of the light guides was adjusted on each counter in order to minimize the angle between the residual field and the PMT axes. The same kind of phototubes will be installed also on the ACC counters. The ATC will be substituted by a RICH counter, in order to increase the velocity and charge resolution, and two completely new detectors will be added: a TRD on the top of AMS-02 and an electromagnetic calorimeter below the RICH will improve the detector sensitivity to high energy electrons and γ-rays The magnet The AMS-02 superconducting magnet [101] consists of a pair of large racetrack shaped Helmoltz coils that generate the majority of the dipolar magnetic field, plus 12 smaller racetrack coils circumferentially distributed (at angles ±60, ±72, ±84, ±96, ±108, ±120 degrees) whose aim is to increase the magnetic field intensity and to reduce the stray field, in order to limit the torque resulting from the interaction with the geomagnetic field (figure 4.26). All coils, built in the United Kingdom, are situated inside a vacuum tank and operate at 1.8 K with superfluid helium. The free bore of the system has a diameter of 1.1 m, while the external diameter of the vacuum tank is 2.7 m, with height of 1.55 m. The coils are electrically connected in series, and operate with a current of 459 A in the persistent mode : once a constant current has been established, a superconducting switch will cut away the power supply and the current will circulate with zero energy dissipation. At this stage no power is required to mantain the current loop. Howewer, power will be required to turn on/off the current and to mantain the system at low pressure. If any part of the superconducting material is heated 86

97 4.4 The AMS-02 detector Figure 4.26: 3D artist view of the AMS-02 magnet system [101]. above its critical temperature, the current energy will be dissipated by this portion of normal conducting material: a quench occurs. The dissipation will heat the other parts and all the energy will be very soon converted into head, potentially causing damages to the structure. The system has been designed to be able to start operation again after a quench with a delay of three days. The conductor is the same for all coils: a NbTi/Cu superconducting wire embedded in a high purity Al stabilizer. A total of 55 km of strand is required for the whole system. The operating temperature is below 10 K, and it is reached with 2500 liters of liquid helium (the He vessel is filled with superfluid He at 1.8 K before the launch). The coils are thermally connected to the He tank through pipes filled with pressurized superfluid He: this material has the highest heat conductivity among all materials, and high density. Helium will circulate also through the cooling circuit, and finally will be vented into space. The large volume of 2500 liters is needed to be able to operate three years without refilling. 87

98 THE AMS EXPERIMENT The tracking system The main differences between the AMS-01 and AMS-02 tracker are the number of support planes (6 and 5 respectively) and read-out planes (6 and 8 respectively), the geometrical acceptance (from about 0.15 to 0.4 m 2 sr), and the production of the Si modules. The decrement of the number of supporting planes results in a diminution of the matter traversed by the particles, hence of the probability that multiple scattering affects the trajectory fit. This is the main problem with momentum measurements at low rigidities. The increment of the number of read-out (x, y) planes allows for a better measurement of the energy loss and an improvement of the trajectory fit. In addition to a refinement in the ladders production and to the stronger magnetic field, the net result is a better measurement of the particle momentum. The MDR of AMS-02 will be in the range 1 2 TV. The new technique for the construction of the silicon ladders was motivated by the poor performances of the n-side read-out of the AMS-01 tracker [95]. Instead of using four p + blocking strips between two consecutive n + read-out strips, the new ladders have only three blocking strips on the n-side, resulting in a better signal-to-noise ratio. The p-side of the ladders is built along the same lines as the AMS-01 tracker The time of flight system The time of flight system of AMS-02, developed in the laboratories of INFN Bologna, will be similar to the TOF system of AMS-01. Two scintillator planes will be placed above the magnet, and other two planes will be placed below. The counters of adjacent planes are orthogonal, in order to guarantee a certain granularity at the trigger level. Due to the strong residual magnetic field in the phototubes zone, the new TOF will adopt different PMTs (Hamamatsu R5946) and will have curved light guides in order to minimize the angle between the field and the PMT axis. This will produce a worse time resolution than AMS-01, but the new front-end electronics will improve the charge resolution. More details in chapter 5. 88

99 4.4 The AMS-02 detector Figure 4.27: The structure of an ACC module [102] The anticoincidence system The Anti-Coincidence Counter (ACC) system, built in Aachen, form a barrel around the Si tracker of AMS-02 [102]. Its purpose is to flag events produced by particles crossing the detector from the side, by δ-rays or even showers produced by triggered particles, and by back-scattering of the electromagnetic calorimeter. In all cases, the detector would record informations giving bad track fit, and charge and velocity resolution. The ACC modules (figure 4.27) are plastic scintillators (Bicron BSC414) with emission in the range nm, coupled with wavelength shifting fibers (absorption at nm, emission at nm). White glass optical fibers are used to bridge the wavelength shifting ones to the phototubes, that are 1 2 m far 89

100 THE AMS EXPERIMENT from the scintillators. The 16 plastic paddles have a height of 83.2 cm and form a cylinder with an inner diameter of cm. Their thickness is 1 cm and their signal is read by 16 PMTs, by both ends: each phototube sees two adjacent paddles, and the PMT couples are interleaved. Due to the high magnetic field intensity ( 0.16 T), fine mesh phototubes has been chosen that are able to operate up to 0.3 T, provided that the angle between their longitudinal axis and the magnetic field is less than 30 degrees. These cylindrical PMTs are of the same kind as those used by the TOF system (Hamamatsu R5946), and need to be installed in the region of the magnet vessel where the field is parallel to the x axis, i.e. near the y axis. They will be oriented in order to align their longitudinal axis to the field direction. The powering system and the read-out electronics are very similar for the TOF and ACC phototubes: the high voltage will be produced by an elevator placed in the electronics crate, and each channel will be regulated from the top (2300 V) down to the desired tension. The ACC anodes will be sent to the scintillator frontend anticoincidence (SFEA2) board, that will send the discriminated signal to the trigger electronics. Because the ACC signals will be used also to check for backscattering from the calorimeter, they will be used at the first level trigger stage in conjunction with the signals coming from the TOF and ECAL systems The RICH detector The AMS-02 detector will have a proximity focusing Čerenkov counter whose aim is to improve the velocity resolution with respect to the TOF system, and to provide the AMS Collaboration with an additional charge measurement. The RICH is developed by a team of researchers coming from institutions of six different countries (INFN Bologna, ISN Grenoble, LIP Lisbon, CIEMAT Madrid, University of Maryland, and IFUNAM Mexico). The instrument is placed below the lower TOF planes, whose support structure is keeping the RICH radiator. Particles crossing this layer with energy above the Čerenkov threshold will cause the emission of optical photons (peaked at about 430 nm), eventually reaching the pixel plane (directly or after being reflected by the conical mirror), placed about half meter below the radiator. In the middle of the pixel plane there is a square hole, right above the calorimeter. More details in chapter 6. 90

101 4.4 The AMS-02 detector Figure 4.28: The structure of a TRD module The trasition radiation detector The transition radiation detector (TRD), built in Aachen, is a gas detector that measures the energy deposited by the X-rays produced by high energy charged particles crossing the radiator [93] (in addition to ionization losses). When these particles cross the boundary between two media with dielectric constants ε 1 ε 2, they have a small ( 10 2 ) probability to emit transition radiation photons. The AMS-02 TRD modules have 16 straw tubes with diameter 0.6 cm that follow a 2 cm thick radiator. The latter is a fleece of 10 µm fibers with 0.06 g cm 3 density (figure 4.28). The gas mixture is Xe CO 2 at 80%/20%, and the gold-plated sense wires are operated at 1600 V, reaching a 50% probability to detect with one module the transition radiation produced by the particle. The TRD consists of 20 layers of straw modules interleaved with fibers and arranged in a conical octagonal structure. Outside this structure a grid of carbon fiber tubes makes support for gas tubes and cabling (figure 4.29). The top and bottom 4 layers are parallel to the x axis, while the middle 12 layers are parallel to the y axis. In order to operate in space for 3 years, the TRD is equipped with a reservoir of 50 kg of gas, corresponding to 8100 liters of Xe and 2000 liters of CO 2 at 1 atm. 91

102 THE AMS EXPERIMENT Figure 4.29: The AMS-02 transition radiation detector The electromagnetic calorimeter The electromagnetic calorimeter (ECAL), constructed by LAPP (Annecy, France), BISEE (Beijing, China), and INFN (Pisa and Siena, Italy), will identify, amongst all cosmic rays, electrons, positrons and γ-rays in the high energy range. The ECAL is an imaging calorimeter consisting of 9 modules made of Pb and scintillating fibers (figure 4.30), whose area is cm 2 and the depth is 1.8 cm ( 1.8 radiation lengths). Two adjacent modules are rotated by 90 and their fibers follow x or y directions [103]. Fibers are read only at one end, by Hamamatsu R M4 photomultipliers placed alternatively on each side. Each PMT window has 4 pixels, hence the elementary cell of the calorimeter has mm 3 (or mm 3 ) size, corresponding to about 1 radiation length along z and 1 Molière radius along y (or x). A charged particle impinging vertically on the ECAL will cross 16 radiation length and the shower longitudinal profile will be sampled by 18 independent measurements. Depending on the primary particle energy, the signal collected by PMT ranges from few photoelectrons for a MIP particle to 10 5 photoelectrons for 1 TeV electrons. The HV divider chosen by the collaboration saturates above 92

103 4.4 The AMS-02 detector Figure 4.30: The ECAL structure pc of collected charge per pixel, corresponding to about 100 photoelectrons per pixel [103]. 93

104 94 THE AMS EXPERIMENT

105 Chapter 5 The AMS-02 TOF system The long duration of the AMS-02 mission (at least three years) and the detector large acceptance make it possible to collect a high statistics sample of particles with energies of the order of 1 TeV per nucleon. For this reason, the permanent magnet of AMS-01 has been substituted with a more powerful superconducting magnet: this allows AMS-02 to have a maximum detectable rigidity for protons of 1 2 TeV. The higher magnetic field produces also a higher fringing field in the zone where the photomultipliers are placed. While the use of a shielding material makes it possible for RICH and ECAL to use the same kind of PMT that was used by ACC and TOF in AMS-01, this would not be possible for the AMS-02 ACC and TOF systems because the shielding boxes would be too heavy. Instead, a different model of phototube was chosen, that can operate with high magnetic fields provided that its longitudinal axis is almost parallel to the field direction. In order to minimize the angle between the PMT axis and the magnetic field, the ACC phototubes (connected to the scintillator paddles through optical fibers) were positioned in the two zones where they are approximatively parallel to the field. On the other hand, the TOF system does not make use of any optical fiber to avoid photon losses, but each scintillator has four plastic light guides (LG) as short as possible. Howewer, these LG have to be bended in order to minimize the magnetic field effects, at least for the PMTs of the inner TOF planes. Due to the strict packing of the phototubes, the LG bending has been a complicated mechanical problem. Another problem was to design the scintillator system in order to minimize its weight (the allotted budget was decreased by 20% in 2002) still keeping the required large acceptance and trigger efficiency for a precise antimatter measurement. 95

106 THE AMS-02 TOF SYSTEM The complicated geometry and the new kind of PMT make it difficult to obtain similar performances to the AMS-01 TOF. Howewer, the ion beam test carried on during October 2002 at CERN showed that in absence of strong magnetic fields the time resolution of the new counters is almost the same of the AMS-01 TOF, while the charge resolution will be improved thanks to a new design of the frontend electronics. The magnetic field mostly affects the time resolution, while it is not expected to induce strong variations on the charge measurement. 5.1 The time of flight system of AMS-02 The time of flight (TOF) system of the AMS-02 detector has the following essential tasks. First, it has to send to the trigger box the signals used to create the fast trigger signal: this is the very first step of the data acquisition (DAQ) electronics. This signal is used by the TOF electronics as the time zero, which the scintillator time signals are referred to. Second, the particle time of flight is used to measured its velocity β = v/c, when the crossing positions (hence the track length) are known. The TOF counters themselves should be precise enough to give, through the time difference of their two edge signals, the longitudinal position within few centimeters, allowing for the β measurement without the need for information coming from other subdetectors. This spatial information is related to the particle producing the trigger, and is very useful to cross check the tracker and TRD track reconstruction, expecially when secondary hits are recorded by the latter subdetectors. Third, the time of flight is used to distinguish between upward and downward going particles. This is fundamental in order to separate particles from antiparticles. The sign of the particle charge is in fact given by two distinct measures: the track bending and its direction. After the TOF measurement of the particle direction, the tracker measurement of the trajectory curvature is sufficient to distinguish between negative and positive electrical charges, that are deflected in opposite directions by the magnetic field. Finally, the energy loss measurement is used by the TOF system to send to the trigger box a special flag for ions events, that can be used at the first level trigger to disable the anticoincidence counters veto, that would suppress more strongly higher charges (whose flux is low). In addition, the TOF charge measurement is used (with the corresponding measurements by the other subdetectors) during the offline analysis to separate light ions from protons and He nuclei. 96

107 5.1 The time of flight system of AMS-02 Figure 5.1: The Hamamatsu fine mesh R5946 photomultiplier tube. While all the previous items are equally fundamental both for AMS-01 and AMS-02, the new TOF system design is different from the first one, due to more severe constraints given by the stronger magnetic field in the photomultiplier tubes (PMT) zone and by the reduced allotted weight budget. The strong magnetic field, whose intensity reaches 0.2 T for some PMT, makes impossible to use the same PMT model of AMS-01 (Hamamatsu R5900, venetian blind dynodes), and forces us to adopt bended light guides in order to minimize the angle between the field direction and the PMT axis. The new PMT model (Hamamatsu R5946, fine mesh dynodes, figure 5.1) is bigger and heavier than the old device, thus the new counters have two (instead of three) PMT per side. In addition, it has a higher working voltage (2000 V instead of 800 V), making it necessary to develop a different high voltage (HV) scheme. The reduced weight budget (about 238 kg for the whole TOF system) imposes a different number of scintillator counters per plane with respect to the old 4 14 scheme. In addition, to reach a geometrical aperture of 0.4 m 2 sr, the external counters of each plane have a trapezoidal shape, and one additional PMT per side in planes 1 and 4, where the magnetic field constraint is less severe. The effect of the intense and spatially variable magnetic field in the PMT zone (figure 5.2), that forced the choice of a different PMT model and of bended light 97

108 THE AMS-02 TOF SYSTEM y (cm) kG x (cm) Figure 5.2: AMS-02 field map in the TOF PMT horizontal plane. The ring shows where PMTs are positioned. guides, combined with the reduced weight (lesser number of counters and different shape), will give a worse time resolution with respect to AMS-01, due to the increase of the transit time 1 of the phototubes [104] and of its jitter. On the other hand, the TOF charge resolution will be better due to a refined design of the front-end (FE) electronics ( 5.4.1). 1 The transit time is the time elapsed between the arrival of photons on the photo-cathode and the anode current generation. It is related to the multiplication process between the dynodes. 98

109 5.2 Magnetic field 5.2 Magnetic field The phototubes used by the TOF system of AMS-01, the venetian blind Hamamatsu R5900 model, are small and light, and need a relatively low voltage to work (below 900 V). Thus they are well suited for space applications. The maximum magnetic field intesity in the PMT zone of AMS-01 was about 300 G, then it was decided to use a ferromagnetic material to shield them: the PMTs were enclosed by 1 mm thick Permalloy boxes, whose effect was to reduce the field intensity down to few Gauss, a level at which the R5900 model can work without problem. On the other hand, the superconducting magnet of AMS-02 ( 4.4.1) is about 6 times stronger than the permanent magnet of AMS-01 ( 4.2.1). In addition, the new TOF planes are nearer to the magnet than the AMS-01 ones. The result is that the fringing field in the PMT zone is much stronger than in AMS-01, reaching values of the order of 2.5 kg (ten times higher), with a complicate direction distribution (figure 5.2). In order to shield such an intense field, 1 mm thick boxes are not sufficient, because they would saturate. Thicker ferromagnetic boxes are not allowed by the small weight budget allotted for the TOF system. Thus one has to find phototubes of different technology with respect to the old PMTs, in order to allow them to operate in regions with strong magnetic fields. The new model selected for the TOF (and ACC) system of AMS-02 is the cylindric Hamamatsu R5946 fine mesh photomultiplier (figure 5.1). Its dynodes have a mesh shape and are tightly packed. In addition it operates at quite high voltages (about 2000 V), in order to reduce its sensibility to the magnetic field. This PMT can indeed work with the high intensity fields of AMS-02, but it shows a strong dependence on the angle between the field direction and the PMT axis. For example, figure 5.3 shows the measured PMT single photoelectron response for different values of the magnetic field intensity and direction. In general, measurements carried on in Bologna with magnetic field up to 0.4 T, about the single photoelectron response, the time resolution, the charge peak position and its distribution, show that one should avoid angles larger than 20 degrees as much as possible: for higher values, even for lower field intensities than those expected for AMS-02, the PMT single photoelectron response and the PMT transit time jitter rapidly become no more acceptable [104]. On the other hand, the gain and the charge resolution are not strongly affected for angles below 20 degrees, that is the situation of planes 1 and 4, and of many phototubes of the other two planes. 99

110 THE AMS-02 TOF SYSTEM (a) PM 9386 = s /s 1 B = 0 G B = 500 G B = 1000 G B = 1500 G B = 2000 G B = 2500 G B = 3000 G B = 3500 G = s /s angle (degrees) (b) PM 9385 B = 0 G B = 500 G B = 1000 G B = 1500 G B = 2000 G B = 2500 G B = 3000 G B = 3500 G = s /s angle (degrees) (c) PM 9381 B = 0 G B = 500 G B = 1000 G B = 1500 G B = 2000 G B = 2500 G B = 3000 G B = 3500 G angle (degrees) Figure 5.3: Measurement of the single photoelectron response of Hamamatsu R5946 phototubes as function of the intensity and direction of the magnetic field [104]. In order to minimize the angles between the PMT axes and the field direction, that has strong spatial variations all around the scintillator planes, bended light guides has been adopted, whose shape has been tuned keeping into account not only the magnetic field, but also the tight packing between the TRD and the TOF mechanics (figure 5.4), that are both fixed to the same honeycomb layer. Starting from the simulated field map in a three-dimensional spatial grid with 5 cm step, with all Cartesian coordinates in the interval [0, 150] cm, I wrote a FORTRAN program to compute the mean field intensity and direction inside any given PMT, whose front window and rear panel centers are known. In order to be able to work with positive and negative coordinates, the program 100

111 5.3 Mechanical design Figure 5.4: The TRD and the upper two TOF planes of AMS-02 share the support structure. first applies a symmetry transformation to the field components (the field is dipolar and symmetric around the x axis). Then it finds the nearest 8 grid points around the PMT window center and computes the weighted average of the field components (the weight is the inverse of the distance to the grid point). Finally, it computes the value of the acute angle formed by the field vector and the PMT axis. The result is an ASCII file (see appendix B) with a table containing one row per PMT, where the columns are: the 3 coordinates of the PMT window center (in centimeters), the 3 coordinates of the PMT rear center (in centimeters), the magnetic field intensity and its 3 components (in Gauss), the angle between the field vector and the PMT axis (in degrees). 5.3 Mechanical design The simulated field intensities and directions in the PMT positions were used to constrain the mechanical design: the engineers proposed several different schemes for accomodating all the PMTs in the small volumes that they were allowed to fit into, and I checked these schemes against possible problems due to the magnetic field. After several iterations (that involved also the TRD support structure design), the following final scheme of the AMS-02 TOF has emerged. 101

112 310p2 310p1 309p2 309p1 308p2 308p1 307p2 307p1 306p2 306p1 305p2 305p1 304p2 304p1 303p2 303p1 302p2 302p1 301p2 301p1 208p2 208p1 207p2 207p1 206p2 206p1 205p2 205p1 204p2 204p1 203p2 203p1 202p2 202p1 201p2 201p1 310n2 310n1 309n2 309n1 308n2 308n1 307n2 307n1 306n2 306n1 305n2 305n1 304n2 304n1 303n2 303n1 302n2 302n1 301n2 301n1 208n2 208n1 207n2 207n1 206n2 206n1 205n2 205n1 204n2 204n1 203n2 203n1 202n2 202n1 201n2 201n1 THE AMS-02 TOF SYSTEM 108n3 108n2 108n1 107n2 107n1 106n2 106n1 105n2 105n1 104n2 104n1 103n2 103n1 102n2 102n1 101n3 101n2 101n1 TOF-2: Counters & PMTs numbering scheme D. Casadei, 22/04/ p3 408n3 108p2 408n2 108p1 408n p2 107p1 106p2 106p1 105p2 105p1 104p2 104p1 103p2 103p1 102p2 102p1 101p3 101p2 101p1 407n2 407n1 406n2 406n1 405n2 405n1 404n2 404n1 403n2 403n1 402n2 402n1 401n3 401n2 401n Figure 5.5: The AMS-02 TOF numbering scheme for counters and phtototubes. 408p3 408p2 408p1 407p2 407p1 406p2 406p1 405p2 405p1 404p2 404p1 403p2 403p1 402p2 402p1 401p3 401p2 401p1 102

113 5.3 Mechanical design The TOF planes The time of flight system has 4 planes made of 1 cm thick plastic scintillator paddles of different shapes: the outermost counters have a trapezoidal (x, y) projection, while all other counters have a rectangular shape. Figure 5.5 shows the top-bottom view of all counters. The shape of the external counters was chosen to match the desired geometrical aperture still satisfying the severe weight constraints (20% mass reduction was decided at the beginning of 2002). The four planes actually form two structures, as shown in figure 5.5, placed above and below the magnet. Each structure has one plane with counters whose longitudinal axis is parallel to the field direction inside the spectrometer (that defines the x axis), and counters along the y direction in the other plane, in order to have a certain granularity at the trigger time. First and fourth planes have 8 counters along the x axis, while second and third planes have 8 and 10 counters along y, respectively (figure 5.6). Each couple of adjacent TOF planes is enclosed in a carbon fiber light tight envelope, and is connected to the electronics crates thanks to patch-panels installed on the envelope itself, where the powering and read-out cables are plugged. The inner (non trapezoidal) counters are 12 cm large in order to accomodate two Hamamatsu R5946 phototubes per side, and each two adjacent parallel counters have 0.5 cm overlap, in order not to have holes in the acceptance. Hence, at trigger time one has inner matrices of 6 6 and 6 8 of square (x, y) cells with 11.5 cm sides, plus surrounding cells of larger granularity. About 90% of all acceptable tracks will cross the central square cells The TOF counters Figure 5.7 shows a schematic view of a 12 cm large TOF counter. The sensitive material is an organic plastic paddle (polyvinyltoluene) whose scintillator light is internally reflected until it reaches the two edges, where plexyglass light guides bring it to the photomultipliers. The light guides consist of five different parts: a straight extender that prolongs the scintillator paddle is connected to two bended and twisted pieces that end with conical junctions whose the PMTs are fixed to. Actually the extender and the two curved pieces come from the same plexyglass layer, that is first cut along the separation of the bended guides, then it is heated and deformed as needed. After this operation, the whole piece is heated again to avoid optical non- 103

114 THE AMS-02 TOF SYSTEM Figure 5.6: 3D view of the AMS-02 TOF scintillator paddles. uniformities, and its surfaces are finally polished. All the conical parts are equal and they are glued to the bended light guides. The phototube housing (figure 5.8) is a black box divided into two pieces, that enclose the PMT and the edge of the conical light guide. Between the guide and the PMT window a soft transparent pad is placed, that guarantees the needed optical and mechanical couplings. In addition, the printed circuit boards (PCB) hosting the PMT voltage divider are fixed to the rear of the black housing through diamagnetic screws. In order to protect them against low pressure discharges, the PMT pins and the lower PCB are potted with Dow Corning , the same transparent material used to build the optical pads. The rest of the electronics is 104

115 5.3 Mechanical design Figure 5.7: A sketch of a TOF counters of AMS-02. Figure 5.8: The mechanical fixation of the PMT to the conical part of the light guide is realized through the PMT housing. coated with Nusil CV-1152 white material, while the light tightness is obtained using the black CV Nusil product. The scintillator, extenders and light guides are wrapped by a thin mylar foils, that improves reflectivity and protects the surfaces from dust and small debris that may be produced by the enclosing carbon fiber 0.5 mm thick boxes, that provide the needed rigidity. Light tightness is provided by a large carbon fiber envelope 0.7 mm thick that encloses the couple of adjacent planes and their phototubes. 105

116 THE AMS-02 TOF SYSTEM Figure 5.9: The AMS-02 electronic crates placement. S denotes the S-crates. Each couple of TOF planes has to be considered a unique detector piece, whose signals and powering lines are connected to the electronic crates via cables connected to the 4 patch-panels of the enclosing carbon fiber envelope. 5.4 Electronics The time of flight and anticoincidence systems have front-end (FE) electronics boards placed in four crates (called scintillator crates or S-crates ) that are situated in opposite corners (figure 5.9). Each S-crate is doubly redundant: there are two identical copies of each module, connected to different power lines for redundancy. Only one half of the crate is turned on during normal operations. In case of problems, the half with the less severe problems is kept powered on, while the other will not be used. 106

117 5.4 Electronics Low Threshold max 8 ns (clock cicle) SFET timings 250 ns after re formation High Threshold (to pre trigger logics) Fast Trigger min 100 ns max 220 ns no HT Time expansion min 2500 ns max 5500 ns TDC input Figure 5.10: Timings of the SFET module Data acquisition Analog data coming from TOF and ACC phototubes are digitized by the SFEx2 boards and then collected by the scintillator data reduction unit SDR2 through serial links ( TOFwire custom protocol). The SDR2 sends the data via AMSwire LVDS serial links to the 4 boards (called JINJ) on the next higher hierarchical level, where data coming from the four S-crates are put in the same structure as the data from the other subdetectors. The TOF anodes are sent to the scintillator front-end time (SFET2) redundant board, that measures the arrival time of their signals with respect to the fast trigger and provide digital signals to the pre-trigger electronics when the analog signals pass a threshold corresponding to about 0.4 MIP (i.e. 40% of the average energy lost by minimum ionizing particles). The SFET2 is similar to the FE electronics of the AMS-01 TOF: the anode signal is compared to two different thresholds (figure 5.10) and goes into two TDC channels. The first, or low, threshold (LT) starts the time expansion logics, while the second, or high, threshold (HT) is used to produce the fast trigger and it goes in the history TDC channel. After LT, the SFET2 charges a capacitor 107

118 THE AMS-02 TOF SYSTEM and waits at most 250 ns for the FT and then discharges it fastly if this signal does not come in time. In case of FT, the capacitor is discharged over a circuit with greater time constant, and the discharge time is equal to the time between LT and FT multiplied by 25. This information goes to the TDC time expansion channel. A fraction of the anode analog signal is used to measure the charge. The anode channels of the ACC phototubes are connected to the SFEA2 redundant board, that sends digital signals to the trigger boxes (called JLV1A and JLV1B) when ACC anodes pass the 0.4 MIP threshold, and measures their charge. The SFEA2 board receives both ACC and TOF anodes: the latter are processed in the same way of SFET2, whereas the former lack the time expansion logics. In addition, anode charge measurement is carried on directly on SFEA2 for all the input anodes. In AMS-01, the anode and dynode charge was measured by TDC using the time over threshold of a discharging capacitor, in order to reduce power consumption. The read-out dependence from the charge was logarithmic, resulting in a poor particle separation for Z > 3. On the contrary, AMS-02 will use a linear ADC coupled with a sample-and-hold technique, identical to the method employed by the RICH FE electronics. This charge measurement technique is linear over 12 bit, and will be used both for anode and dynode signals, the latter being sent to the SFEC2 boards ( C for charge ). In addition to the threshold used for singly charged particles (the HT), the S-crate electronics provides the trigger boxes also with a special flag for Z > 2 particles (super-high threshold, SHT). The pre-trigger logics that operate on Z 1 and Z > 2 digital signals is hosted by the S-crate itself. For each half TOF plane connected to the crate, three logical levels are sent to the trigger boxes: the logical OR of all Z 1 signals (CP or charged particle flag), the same thing excluding the outermost counters (CT or central charged particle ), and the logical OR of all the Z > 2 levels (BZ or big Z ). These signals are used by the trigger boxes to generate the fast trigger (CP) and the level 1 trigger (CT and BZ) Slow control Slow control commands are foreseen to change the PMT voltage, to turn on/off the S-crate boards, to control and monitor the scintillator power distributor (SPD) box, to read the temperature sensors. Two universal slow control modules (USCM) are placed in each S-crate, connected to different power lines. Commands from the control center (called POCC) will reach the USCM in 108

119 5.5 The beam test the S-crates through a dedicated connection (CAN bus). The USCM will forward any command to the boards it refers to, and will send back status, powering and temperature information (again via CAN bus). The connections of each USCM to the SDR2 and the high voltage controller are realized with one Le Croy bus per USCM (plus a double connection with the SPD controller for redundancy): there is no crossing point in S-crates. The commands that change the powering of SFEx2 boards are sent from USCM to SDR2, then SDR2 forwards them through TOFwire to the desired board Powering The S-crate contains the scintillator high voltage (SHV) redundant module, that consists of two high voltage elevators (HVE) connected to different power lines. Each HVE feeds 24 linear regulators (LR) that can be set at voltages between 2300 V and 1300 V with a 8 bit DAC (hence the minimum step is about 4 V). One, two or three ACC/TOF PMTs can be connected to the same LR. The low voltage DC/DC converters are placed outside the S-crate, in the scintillator power distributor (SPD) box. The SPD receives two indipendent 28 V lines, connected to two separate families of DC/DC converters, providing the S- crate with +5 V, ±5 V, +3.3 V. These LV lines are connected to the S-crate minibackplane through screwed wires. 5.5 The beam test Two AMS-02 TOF counters and one AMS-01 counter were tested at CERN on October 2002 with standard NIM and CAMAC electronics, on the ion beam provided by the SPS facility. The primary Pb beam was directed against a Be target cm long (target T4), producing secondary particles and nuclei with charge spanning a very wide range: Z = 1 82 (see chapter 7). The H8 selection line was tuned to obtain secondaries with A/Z = 2 ( 4 He and almost all stable nuclei up to iron), A/Z = 7/4 (mostly 7 Be), and A/Z = 1 (protons). For what concerns the scintillators, all primary particles were at their minimum ionization plateau (that is at Lorenz factors γ > 3). This section shows first results from the analysis of TOF data, from the runs

120 THE AMS-02 TOF SYSTEM Selection cuts In order to study in a self consistent way the data obtained with the three TOF scintillators tested at SPS, it is necessary to apply a number of cuts to the sample. The three counters (a plastic scintillator from Eljen Technologies, one from Bicron, and a Bicron TOF counter dismounted from AMS-01) are large compared to the beam spot, whose diameter is about 3 cm. Hence they picked up a lot of secondary particles that are off the beam axis. The first effect that can be easily isolated is the deviation of the mean counter time t i = t i,left + t i,right (5.1) 2 from the value that corresponds to the sum of the delays between the TDC start and the arrival of the signals from both ends of the same counter: the beam spot is at the center of the three scintillators, while secondary off-axis particles hit the counters over the whole area. As figure 5.11 shows, the wrong mean time measurements are a small fraction of the total number of events. A Gaussian fit reveals the mean values t i and the standard deviation. As first set of cuts, it is required that events not consistent with t i within one standard deviation are discarded. Then one can consider the half difference between the time measured in both ends of the same counter (omitting the index i): t = (t + x/v) [t + (L x)/v] 2 = x v L 2v (5.2) where t = t i is the particle crossing time for counter i, L is the counter length, v is the effective light speed inside the scintillator, and x is the distance of the crossing position with respect to one of the two sides. Apart from a constant term (L/2v), t i gives the longitudinal coordinate x along the scintillator (v 15 cm ns 1 has been measured in the INFN laboratories in Bologna). The time resolution of the counters ( 5.5.3) is of the order of 0.1 ns, hence the beam particles give a fixed time difference, because the spot radius is approximatively equal to the spatial resolution σ x 1.5 cm of the counters obtained with the half time difference measurement. In general, off-axis particles will give a different value for t, hence they can be discarded by requiring that the half time difference is within one standard deviation from its mean value for each counter (figures 5.12, 5.13 and 5.14). A couple of secondaries that hit the counters in symmetric positions with respect 110

121 5.5 The beam test Entries Events after cut: Cut eff.: 67.84% C1 time (ns) Entries Events after cut: Cut eff.: 69.36% C2 time (ns) Entries Events after cut: Cut eff.: 67.34% C3 time (ns) Figure 5.11: Mean time measured by the three TOF scintillators under ion beam test (runs 2 14). Only events within 1 standard deviation from the Gaussian mean are kept. to the beam spot cannot be rejected by this condition, but they produce a wrong measurement of the mean time, hence they are discarded by the other condition. In order to be able to study the charge resolution of the scintillators, I imposed a constraint also on the relative time of flight between each couple of counters. This condition discards slow secondaries with crossing positions near enough to the beam spot that they cannot be resolved using the half time difference alone. Figures 5.15, 5.16 and 5.17 show that after the mean time and half time differ- 111

122 THE AMS-02 TOF SYSTEM 10 6 Entries C1 time diff. (ns) no cut events within 1 σ (cut eff. = 66.77%) Entries C1 time diff. (ns) mean time cut Figure 5.12: Half difference between the time measurements of both sides of the Eljen scintillator (runs 2 14) without cuts (upper panel) and with the mean time cut (lower panel). Only events within 1 standard deviation from the Gaussian mean are kept. ence conditions the all-particle time of flight has a distribution that can be fitted by a Gaussian with standard deviation of ps. Howewer non Gaussian contaminations are recognizable by eye. Discarding also the events with time of flight measurements outside one standard deviation from the mean values it is possible to obtain a quite clean sample of appreciable statistics (25.6% of the considered events), that will be useful to study the charge resolution. 112

123 5.5 The beam test 10 6 Entries C2 time diff. (ns) no cut events within 1 σ (cut eff. = 64.55%) Entries C2 time diff. (ns) mean time cut Figure 5.13: Half difference between the time measurements of both sides of the Bicron scintillator (runs 2 14) without cuts (upper panel) and with the mean time cut (lower panel). Only events within 1 standard deviation from the Gaussian mean are kept. 113

124 THE AMS-02 TOF SYSTEM 10 6 Entries C3 time diff. (ns) no cut events within 1 σ (cut eff. = 65.15%) Entries C3 time diff. (ns) mean time cut Figure 5.14: Half difference between the time measurements of both sides of the AMS-01 TOF scintillator (runs 2 14) without cuts (upper panel) and with the mean time cut (lower panel). Only events within 1 standard deviation from the Gaussian mean are kept. 114

125 5.5 The beam test Entries C1 C2 (ns) No cut: times cut: TOF cut: (eff. = 25.62%) Entries / 14 P E P E 03 P E C1 C2 (ns) Figure 5.15: Time of flight between the first two counters (runs 2 14) without cuts (upper panel) and with mean time and half time difference cuts (lower panel). Only events within 1 standard deviation from the Gaussian mean are kept. 115

126 THE AMS-02 TOF SYSTEM Entries C1 C3 (ns) No cut: times cut: TOF cut: (eff. = 25.6%) Entries / 13 P E P E 03 P E C1 C3 (ns) Figure 5.16: Time of flight between the first and last counter (runs 2 14) without cuts (upper panel) and with mean time and half time difference cuts (lower panel). Only events within 1 standard deviation from the Gaussian mean are kept. 116

127 5.5 The beam test Entries C2 C3 (ns) No cut: times cut: TOF cut: (eff. = 25.59%) Entries E+05/ 14 P E P E 03 P E C2 C3 (ns) Figure 5.17: Time of flight between the second and third counter (runs 2 14) without cuts (upper panel) and with mean time and half time difference cuts (lower panel). Only events within 1 standard deviation from the Gaussian mean are kept. 117

128 THE AMS-02 TOF SYSTEM Entries ± ± ± ± ± ±1 28.1± ± ± ± ±1.3 40± ± C1LD, sqrt(adc) Entries ± ± ± ± ± ± ± ± ± ± ±0.8 30± ±1 33.3± ± ± ± ± ± ± C1RD, sqrt(adc) Figure 5.18: The charge peaks of the first counter can be seen up to Z = 13 for the left side (upper panel) and Z = 20 for the right side (lower panel) Charge peaks The most important information gained with the beam test are about the charge resolution of the AMS-02 TOF counters. The energy lost de/ dx by a particle with atomic number Z is proportional to Z 2 (see equation (4.4) on page 57), but the emitted scintillation light is proportional to de/ dx only for small values of the energy loss. In general, the signal Q measured by the ADC is the time integral of the PMT current pulse, i.e. it is 118

129 5.5 The beam test 10 4 Entries ± ±0.4 12± ±0.4 15± ± ± ± ± ± ± ± ± C2LD, sqrt(adc) 10 4 Entries ± ± ±0.5 14± ± ± ±0.7 21± ±1 24± ±1 26.7± ± C2RD, sqrt(adc) Figure 5.19: The charge peaks of the second counter can be seen up to Z = 13 for the left side (upper panel) and right side (lower panel). proportional to the emitted scintillation light, and can be written: Q = A de dx 1 + B de + C ( de dx dx ) 2 = az bz 2 + cz 4 (5.3) (the first equality is the Birks formula [105]). Hence a plot of the square root of the ADC signal (after the cuts already discussed) gives the particle charge at the first order, as seen in figures 5.18 and While the gains of the four PMTs mounted on the Bicron scintillator was almost balanced, figure 5.18 shows that this was not the case for the Eljen counter. Actually, the latter was tested with a high- and low-gain couple of PMTs in each side (left to right ratio 3 : 2). Howewer, this is not important for the charge res- 119

130 THE AMS-02 TOF SYSTEM olution (apart from the ADC overflow for Z > 13 of the left side dynodes of the Eljen counter): because the PMTs are similar, the charge resolution depends only upon the light yield of the scintillator. These results strongly supported the choice of the Eljen scintillator (figure 5.18) instead of the Bicron one (figure 5.19), thanks to the Eljen higher photostatistics. The Eljen product will allow to measure the particle charge up to atomic number Z 20, providing important cross check of the tracker, RICH and ECAL charge measurements. In order to correctly measure the particle charge, formula (5.3) has been used to fit the ADC peaks (figures 5.20, and 5.21), obtaining the ADC pedestal and the three parameters of (5.3). Then the relation was inverted to get a plot of the reconstructed particle charge Z (a real number) for each counter side. Finally, the mean of the four measurements has been used as best estimate of the particle charge (again a real number, figure 5.22). 120

131 5.5 The beam test C1LD (ADC ch.) / 9 P P E E 02 P E E 04 P Q = C1LD ped. (ADC ch.) Q = ped. = P 4 P 1 Z P 2 Z 2 + P 3 Z C1RD (ADC ch.) Z / 16 P P E E 02 P E E 05 P Z Figure 5.20: ADC channels correspondence to particle charge for the left (upper panel) and right (lower panel) dynodes of the Eljen counter. Fit with Birks law. 121

132 THE AMS-02 TOF SYSTEM C2LD (ADC ch.) / 9 P P E E 02 P E E 04 P C2RD (ADC ch.) Z / 9 P P E E 02 P E E 04 P Z Figure 5.21: ADC channels correspondence to particle charge for the left (upper panel) and right (lower panel) dynodes of the Bicron counter. Fit with Birks law. 122

133 5.5 The beam test 10 4 Entries Entries Entries C1LD Z Entries C1RD Z C2LD Z Entries Mean charge from C1RD only C1LD, C1RD, C2LD, C2RD C2RD Z TOF Z Figure 5.22: Side charge measurements (upper panel) and mean charge (lower panel). For higher charges than Z = 13 only the right side of the first counter is available. 123

134 THE AMS-02 TOF SYSTEM Entries E+05/ 59 Constant E Mean E-03 Sigma E C1-C2 TOF (ns) Entries E+05/ 49 Constant E Mean E-03 Sigma E C1-C3 TOF (ns) Entries E+05/ 58 Constant E Mean E-03 Sigma E C2-C3 TOF (ns) Figure 5.23: All-particle time of flight distribution for runs The standard deviation is dominated by singly charged particles. The single counter time resolution is ps Time resolution When a criterion is known to find the best estimate of the particle charge, it is possible to study the time of flight resolution of the three counters. First, the cuts on the mean time and the half time difference are applied, then one can study the resolution of the difference between t i and t j with i, j = 1, 2, 3 and i < j. Figures 5.23 and 5.24 show the three possible differences without charge selection, the former without any other cut, the latter with the two following cuts: the time of flight between the first two counters has been plotted after the rejection of events with (t 1 t 3 ) and (t 2 t 3 ) outside one standard deviation from their mean values. The large tails of figure 5.23 disappear when applying this cut. Using 124

135 5.5 The beam test Entries / 10 Constant E Mean E-03 Sigma E C1-C2 (ns) Entries / 11 Constant E Mean E-03 Sigma E C1-C3 (ns) Entries / 11 Constant E Mean E-03 Sigma E C2-C3 (ns) Figure 5.24: All-particle time of flight distribution for runs 2 14, after TOF cuts. The standard deviation is dominated by singly charged particles, and it is of the order of 100 ps. σ 2 i j = σ 2 i + σ 2 j, the resulting single counter resolution is given in the following table: Time without TOF with TOF res. cuts (ps) cuts (ps) σ ± 3 74 ± 1 σ ± 2 79 ± 1 σ 3 90 ± 2 71 ± 1 Figure 5.25 shows the average standard deviations of the Gaussian fit of the time of flight between the three possible couples of TOF counters, as function 125

136 THE AMS-02 TOF SYSTEM of the most probable particle charge. In order to select the particles with atomic number Z, it was required that the four sides of the AMS-02 counters give dynode signals in the Z-th ADC peak within one standard deviation. As foreseen, the time of flight resolution improves with increasing particle charge, reaching a limiting level dominated by the electronic noise. This level was between 50 and 60 ps for the standard electronics used during the ion beam test at CERN. The SFET2 design keeps the value of 50 ps as the maximum allowable electronic jitter. Howewer, the presence of a strong magnetic field, even if its direction is almost parallel to the PMT axis, worsen the time resolution of the Hamamatsu R5946 phototubes: the standard deviation will increase almost at the same rate of the phototube transit time, depending on the PMT position and orientation in the magnetic field ( 5.2). 126

137 5.5 The beam test σ t (ns) / 7 P E E-02 P E E σ t = P1/Z + P Time resolution (ns) Z / 7 P E E 02 P E E 02 σ t = P1/Z + P Z Figure 5.25: Average time resolution of the three counters as function of charge, without (upper panel) and with (lower panel) cuts on TOF. The time resolution gets better with increasing atomic number: the limiting level of ps is the electronics noise. 127

138 128 THE AMS-02 TOF SYSTEM

139 The AMS-02 RICH subdetector Chapter 6 The AMS experiment is able to identify cosmic ray particles through the measurement of the energy lost in active materials, the trajectory curvature and the particle velocity. The energy losses in media are proportional to the square of the charge ( 4.1.2), hence their measurements can be used to determine the atomic number Z of the incident nucleus. From the track curvature one can find the particle rigidity ( 4.1.1) and hence the momentum, in combination with the knowledge of the charge. Finally, the velocity measurement ( 4.1.3) is necessary to find the particle mass from its momentum. The proximity focusing ring imaging Čerenkov counter (RICH) of AMS-02 will be able to measure the cosmic rays velocity β = v/c with a 0.1% uncertainty for β c < β < 1, and the particle charge up to Fe, through the measurement of the Čerenkov opening angle and the number of emitted photons, respectively. These features, in addition to providing redundant charge measurement in combination with the TOF and the tracker, are needed to study the isotopic composition of the cosmic rays. 6.1 The Čerenkov emission When charged particles cross a medium with refractive index n > 1 with velocity v greater than the light speed in that medium (v l = c/n, where c is the light speed in vacuo), they emit Čerenkov radiation. The photons are produced uniformly along the path (if n is constant), and they are emitted at a fixed (if also the variation of v is negligible) angle θ c with respect to the particle momentum direction, while the azimuthal emission is uniform. 129

140 THE AMS-02 RICH SUBDETECTOR The Čerenkov angle θ c is given by the formula: hence it reaches its maximum value cos θ c = 1 nβ, (6.1) θ c = arccos 1 n (6.2) for ultrarelativistic particles (i.e. β 1). On the other hand, the lowest value for the velocity is given by the condition cos θ c = 1 (i.e. θ c = 0): the threshold speed is is β c = 1 n. (6.3) In addition, the mean number of emitted photons by a particle with charge Ze ( N Z = Z 2 N 1 Z 2 L 1 1 ), (6.4) n 2 β 2 where N 1 is the average number of photons emitted by a singly charged particle crossing the same medium, that depends on the path length L = d/ cos τ through the radiator of thicnkess d, on the particle velocity v = βc and azimuthal angle τ, and on the refractive index n of the radiator. Because β 2 1, N 1 is pratically not dependent on the particle velocity Particle energy loss in the radiator The energy lost by a non relativistic particle with charge ze that crosses a medium with atomic number Z and atomic number density N is well described by the formula found by Bethe in 1930 (from Jackson [91], in Gauss units 1 ): ( ] 2γ 2 mv 2 de dx = 4πNZ z2 e 4 mv 2 [ ln ω ) v2 c 2, (6.5) where m and v are the mass and velocity of the particle respectively, and ω is the average electron frequency of the medium, defined by the following geometric mean: Z ln ω = f i ω i, (6.6) 1 In Gauss units, the Maxwell s equations in vacuo are: E = 4πρ, B = (4π/c)J + (1/c) E/ t, E + (1/c) B/ t = 0, B = 0. i 130

141 6.1 The Čerenkov emission where the oscillation intensities f i satisfy N e = f i. For ultra-relativistic particles the Bethe formula overestimates the energy loss, and one needs to add a correction due to the dielectric polarization of the medium ( density effect ): thanks to the relativistic beaming, the charged particles feel the electric field also of distant atoms. This effect was pointed out by Fermi in 1940 and it is more important in the treatment of the scattering with large impact parameter b, while for small b one can still neglect the field produced by distant atoms. The tipical scale between small and large impact parameters (hence between the Bethe approximation and the Fermi correction) is the inter-atomic distance a. For large impact parameters, one can find the total electromagnetic field inside the medium produced by the different atoms by using the Fourier transform of the charge and current densities in the dielectric and solving the Maxwell equations in the Fourier space, with coordinates (ω, k). Then it is possible to compute the energy variation due to a single diffusion process, and finally compute the average value. This calculation is well explained by Jackson [91]. Here we only quote the result (Fermi formula): ( ) de = 2 (ze) 2 dx b>a π v 2 Re 0 ( ) iωλ a K 1 (λ 1 a) K 0 (λa) ε(ω) β2 dω (6.7) where ε(ω) is the dielectric constant, K 0 and K 1 are the modified Bessel functions of order 0 and 1 respectively, and the wavelength λ is defined by the relation λ 2 = ω2 v 2 ( 1 β 2 ε(ω) ). (6.8) The usual density effect correction is found when considering the limit λa 1. In this case, the Fermi formula (6.7) becomes similar to the Bethe expression (6.5), with the exception of the exponent of γ in the logarithm, that is 1 instead of 2. This reduces the relativistic increase (figure 6.1) after the minimum in the energy loss curve as function of βγ = P/mc: ( ) de = (ze)2 ω 2 ( ) p 1.123c ln, (6.9) dx b<a c 2 aω p where ω p = 4πNZe 2 /m is the electronic plasma frequency. If we are interested into the energy released far from the trajectory, we can consider the opposite limit, λa 1. In this case one can use the asymptotic 131

142 THE AMS-02 RICH SUBDETECTOR 50.0 de/dx (MeV g 1 cm 2 ) shell correct. β 2 β 5/3 de/dx β 5/3 de/dx β 2 Radiative effects become important Approx T max de/dx without δ Minimum ionization Complete de/dx π ± on Cu I = 322 ev T cut = 0.5 MeV βγ Figure 6.1: Energy loss of singly charged particles (pions) as function of the relativistic momentum βγ = P/mc [106]. The explanations refer to the energy loss written in the form of equation (4.4). behavior of the Bessel functions and find that the term inside the integral of the Fermi formula will approach the expression: ) λ integral in (6.7) i λ (1 ω 1 exp [ (λ + λ )a]. (6.10) β 2 ε(ω) Usually λ has a positive real part, and the exponential produces a rapid cutoff for large distances: all the energy is deposited along the particle trajectory. Howewer, there can be cases when λ is a pure imaginary number (actually Reλ 0, because a tiny dissipation is always present). In such cases, the asymptotic expression (6.10) becomes independent from a: part of the energy goes to infinity as electromagnetic radiation (Čerenkov radiation). From equation (6.8), we see that this happens if ε(ω) is real and β 2 ε(ω) > 1 (anomalous dispersion, figure 6.2), that is when c v >. (6.11) ε(ω) 132

143 6.1 The Čerenkov emission ε(ω) 1 1 β 2 ω 0 ω Figure 6.2: The shadowed region is the Čerenkov zone. Emitted radiation with frequency ω is possible only if the dielectric constant ɛ(ω) > β 2 (anomalous dispersion) Čerenkov radiation The condition (6.11) states that when a charged particle crosses a material with real dielectric constant ε(ω) with velocity greater than the phase velocity v ph (ω) = c/ ε(ω) of the electromagnetic radiation with frequency ω, the polarization of the medium induces the emission of Čerenkov radiation with the same frequency. In this case we have λ = i λ and the asymptotic expression (6.10) does not depend on a. Hence the Fermi formula now represents the emitted energy per unit path and can be written [91] in the form ( de dx ) rad = (ze)2 c 2 ε(ω)>1/β 2 ω ( 1 ) 1 dω (6.12) β 2 ε(ω) (in Gauss units) given by Frank and Tamm in 1937 as interpretation of the emission found by Čerenkov in The function inside the integral is of course the differential spectrum of the Čerenkov radiation: ( d 2 E dω dx ) rad = (ze)2 c 2 ω ( 1 ) 1 β 2 ε(ω) for all values of ω in the shadowed region of figure 6.2. (6.13) 133

144 THE AMS-02 RICH SUBDETECTOR radiator aerogel thickness = 30 mm 114 mm foil thickness = 1 mm supporting foil 600 mm pitch = 37 mm 468 mm LG height = 31 mm 670 mm ECAL hole 630 mm Figure 6.3: The proximity focusing RICH detector of AMS-02. Usually a medium has more than one region with anomalous dispersion (like the zone in figure 6.2), and the emission is concentrated in frequency bands just below the critical frequencies (ω 0 in figure 6.2). For ultrarelativistic particles β 2 1 and such bands can be very large. The Čerenkov radiation is emitted around the particle line of flight with an angle that depends on its velocity and on the medium refractive index n(ω) = ε(ω): 1 cos θ c (ω) = (6.14) β n(ω) so that the criterium β 2 ε(ω) > 1 is equivalent to the condition that the cosine is real. In addition, the Čerenkov radiation is totally polarized in the plane containing the particle momentum and the propagation direction [91]. Due to the formula (6.14), light of different colors will be emitted with slightly different angles (chromatic dispersion), but this effect is reduced when using media with refractive index very near to one. 6.2 The RICH design The AMS detector is a large acceptance cosmic ray spectrometer, where the measured particles have trajectories with a quite large range of incidence angles and crossing positions. Hence, a traditional ring imaging Čerenkov (RICH) detector would not be adequate, because of the need of almost collinear particles, or a known vertex position. Instead, a proximity focusing RICH counter is used to measure the particle velocity and charge, because of its large angular and spatial acceptance. 134

145 6.2 The RICH design charged particle n>1 radiator n=1 photons pixels plane Figure 6.4: Schematic representation of the proximity focusing RICH detector. In order to reach momenta above 10 GeV/c per nucleon, the solid radiator with the smallest refractive index was chosen: silica aerogel. Central tiles of Na F will be used to cover the momentum gap between the TOF and the aerogel β measurements. In order to increase the geometrical acceptance, a conical mirror will be used to reflect the photons emitted at large angles towards the pixels plane. The PMT plane has a square hole just above the ECAL, in order to reduce the matter thickness in front of this detector. The RICH is shown in figure Large acceptance proximity focusing RICH detector A proximity focusing RICH is a detector where the radiator is a (usually thin) layer of transparent material, and is separated from the photon detector by a region with refractive index very near to one (in AMS this region will be in vacuo, hence n = 1), as sketched in figure 6.4. The Čerenkov photons are emitted only inside the plane of the radiator, from which they escape and propagate until reaching the pixels of the light detector. The refractive index of the radiator determines the opening angle of the Čerenkov cone, that is one of the fundamental parameters that affect the β resolution. The other important parameters are the pixel size and the width of the gap between the radiator and the pixel plane, i.e. the detector angular resolution. Real detectors have few problems, due to the finite number of detected photons, to the approximation n(λ) = n (chromatic dispersion), to the radiator thickness and opacity. The number N det of detected photons that can be used in the data analysis is 135

146 THE AMS-02 RICH SUBDETECTOR given by the number N Z of emitted photons (proportional to the square of the particle charge, equation 6.4), minus the number N i.r. of internally reflected photons inside the radiator, minus the N miss photons missed by the pixels due to geometrical effect or quantum efficiency, minus the number N R.s. of photons scattered at large angles by impurities or microstructures in the radiator (Rayleigh scattering): N det = N Z (τ) N i.r. (τ) N miss (τ) N R.s. (τ). (6.15) The dependence on the particle incidence azimuthal angle τ is due to the path length L = d/ cos τ of the particle through the radiator of thickness d, and to the angular distribution of the photons. The path length is related to the transmission T by the empirical formula [107] T(L, λ) = 0.96 exp ( C L ) λ 4 (6.16) where L is usually expressed in centimeters and λ is in micrometers. The parameter C is called the clarity of the radiator (C is usually expressed in µm 4 cm 1 ). The scattering length L s at wavelength λ is given by the relation: L s (λ) = λ 4 /C. Both the emitted number of photons and the Rayleigh scattering are proportional to L: for materials that are affected by this problem one may improve N det by increasing the thickness d of the radiator only up to a limit given by the Rayleigh scattering. This is the case of the silica aerogel that will be used in AMS-02, whose radiator thickness is limited to 2 3 cm. The photons are emitted at angle of θ c with respect to the particle direction, hence they reach the radiator surface with angles between τ θ c and τ + θ c with respect to the normal, and may be totally reflected in some cases (when the direction is above the total reflection angle). In addition, the photons escaping from the radiator will have angles between τ θ and τ + θ, where θ is the Čerenkov angle in the vacuum gap, given by Snell s law: sin θ = n sin θ c. When reaching the pixels with large incident angles, they may be reflected instead of being detected. Finally, another effect is due to the radiator thickness d and to the reconstruction algorithm, that makes the assumption that all photons are emitted in the mid point of the radiator. This effect, with the chromatic dispersion, may produce a number N bkg of detected photons outside the area where the algorithm is searching them. Such background photons should be subtracted from the number of detected photons given by equation (6.15), because they cannot be used to reconstruct the event. 136

147 6.2 The RICH design θ (degrees) c NaF (n=1.34) Agl (n=1.050) Agl (n=1.030) P (Gev/c/nucleon) Figure 6.5: Čerenkov angle as function of the momentum per nucleon, for different radiators Radiator For better understanding of the cosmic rays propagation inside the Galaxy, it is very important to measure the relative abundances of primary and secondary nuclei as function of the energy per nucleon (as seen in chapter 2). In particular, the knowledge of the relative abundances of secondary to primary elements (like Li, Be, B versus C, for example) is extremely important to study the cosmic rays propagation in the Galaxy. Few balloon measurements of the B/C ratio exist below few GeV per nucleon (see figure 2.2), but with large uncertainties due to the poor statistics. One of the Physics goals of AMS-02 will be the study of this ratio with high statistics from few hundred MeV/c per nucleon up to GeV/c per nucleon, that is the interesting range to put constraints on propagation models. Figure 6.5 shows the Čerenkov angle of photons emitted inside Na F and aerogel, for particle momenta up to 15 GeV/c per nucleon. The Čerenkov threshold is lower for Na F, that has a refractive index of n = 1.34, but the achievable resolution on θ c makes impossible to go beyond 5 GeV/c per nucleon. On the other hand, the aerogel can be used to measure β up to GeV/c per nucleon, but has a threshold at momenta above the TOF useful range. Hence the interesting 137

148 THE AMS-02 RICH SUBDETECTOR A B C AGL NaF AGL θ c Pixels plane ECAL HOLE Pixels plane Figure 6.6: Composite radiator plane: the central Na F square is surroundend by aerogel tiles [109]. In case of no Na F, a particle B entering the ECAL hole would not be detected due to the small Čerenkov cone aperture. Instead, particle C crosses Na F and is detected. momentum window can be covered by using two radiators: sodium fluoride and aerogel with n = 1.03 [108]. The radiator layout will be the following: a central square of Na F 34.5 cm large and 0.5 cm thick, surrounded by aerogel tiles 3 cm thick (figure 6.6). In addition to a larger momentum window, this configuration increases the RICH effective acceptance, because of the ECAL hole in the pixels plane. In fact, particles crossing an aerogel radiator in the middle part and impinging on the calorimeter would produce photons that would not hit any pixel, while the Na F has larger Čerenkov angles, increasing the probability that photons are detected [110] (figure 6.7 shows an example). The β measurement made by RICH will allow AMS-02 to separate positrons from protons and electrons from antiprotons up to GeV/c [111] (figure 6.8), thanks to the low refractive index of the aerogel. This will allow to study with large acceptance the antimatter fraction of cosmic rays over the range where most of the flux is found. For higher energies, ECAL and TRD will become fundamental for the e/p separation. 138

149 6.2 The RICH design AGL30 Na F Beryllium P = 1.2 GeV/ c /nucl. Figure 6.7: Beryllium event producing Čerenkov radiation on aerogel n = 1.03 and Na F n = 1.34 radiators [110]. RICH β Momentum (GeV/c) Figure 6.8: Simulated positron to proton separation for AMS-02, thanks to the RICH β measurement [111]. 139

150 THE AMS-02 RICH SUBDETECTOR Figure 6.9: A picture of the mandrel on the lathe, during surface polishing. Picture taken by the author on September Mirror The geometrical acceptance of AMS-02 is very large (about 0.4 m 2 sr), hence the particles will cross the RICH radiator over all its area with angles ranging from 0 up to about 40 degrees. In order not to loose photons generated at large angles with respect to the z axis, a multi-layer conical mirror will be used to reflect such photons towards the pixel plane. The mirror has the shape of a truncated orthogonal cone, with reflecting inner surface. The carbon fiber structure was designed in Italy by INFN Bologna and Carlo Gavazzi Space SpA, that took care of the realization of the mandrel. Howewer, the final mirror will be built in USA by the CompositeMirrors company, and the conical structure will be assembled starting from 120 degrees sectors. The first phase of the mirror construction is the realization of the mandrel: this is the negative of the final product, that will be the starting point for the deposition of the different layers. The mandrel surface will determine the mirror geometry and reflectivity, because after detaching the mirror from it, only a protecting layer will be deposed on the inner surface of the conical sectors. 140

151 6.2 The RICH design RMS roughness [nm] measurement point Figure 6.10: Measurement of the mandrel surface roughness with three orientations of the incident/reflected light beam plane. The RMS averaged over 90 measurement points is 2.3 nm along the cone directrix (0 deg); for 45 deg it is 8.0 nm; for 90 deg it is 4.4 nm. The highest RMS value at 45 deg is due to the small residual scratch The aluminum mandrel was lathe machined (figure 6.9) to get the desired conical geometry within the allowed tolerances, computed by requiring that the maximum systematic deviation from the ideal situation would produce angular deflection not greater than 3 mrad (i.e. no sensible pixel mismatch on the detecting plane). Its surface was then polished using very fine grained sand and a vibrating rubber head: the whole surface got this treatment many times, progressively decreasing the root mean square deviations from the ideal smooth geometry. Unfortunately, during one of the last passes, a larger sand grain was trapped between the rubber head and the surface, producing a visible rippling scratch over the whole mandrel. Howewer, after the final passes the surface rugosity was decreased below the needed values. The roughness was measured with an optical instrumentation using the dispersion of an incident LASER beam: the measured RMS deviation along three different directions with respect to the cone directrix (figure 6.10) are very good (compared to the Čerenkov wavelength of nm), even though 141

152 THE AMS-02 RICH SUBDETECTOR Figure 6.11: A picture of the RICH PMT assembly [109]. the scratch was not perfectly erased. The carbon fiber structure of the mirror (1 mm thick) will be deposited layer by layer on the mandrel while the latter is rotating ( filament winding technique). In this way, layers with different fiber orientations with constant tension will be glued together. After detaching from the mandrel, aluminum will be deposited on the mirror internal surface, covered by a protection coating with quartz or magnesium fluoride. The mandrel was sent in December 2002 to CompositeMirrors, that should provide us the first 120 deg sample in April 2003, whose optical properties has to be tested in Europe Photomultipliers and light guides The AMS-02 RICH will use 680 Hamamatsu R M16 photomultipliers, the 16 channel version of the venetian blind PMT adopted by the TOF, ACC and ATC systems of AMS-01, for a total of mm 2 pixels. They have a gain G with voltage of V, small enough cross-talk between pixels and quite similar gain and efficiency over the 16 channels. Figure 6.11 shows the assembly of the PMT, enclosed in (half of) the magnetic shielding box, with light guides in front of it and the front-end (FE) electronics on the opposite side. In order to reduce the dead space between pixels, solid light guides (LG) were placed in front of the phototubes. Figure 6.12 shows a schematic view of the RICH light guides, that bring photons to their pixel through internal reflection. 142

153 6.2 The RICH design Figure 6.12: The RICH light guides bring photons to pixels through internal reflection ADC counts Figure 6.13: Typical single photoelectron spectrum [112]. The PMT calibration is carried out by looking at the single photoelectron spectrum, that is the ADC histogram obtained when only few photons reach the cathode (figure 6.13). A pulsed LED is used to generate photons during the data acquisition. This spectrum has a characteristic shape: the high pedestal peak on 143

154 THE AMS-02 RICH SUBDETECTOR the left is followed by an asymmetrical charge distribution. As simple approximation, this spectrum can be considered the Poisson distribution of events with zero, one, two or more photoelectrons. Each photoelectron is then multiplied by the PMT dynodes until a sizeable current pulse is produced by the anode. The PMT response to a single photoelectron can be approximated by a Gaussian charge distribution. Hence, the single photoelectron spectrum can be considered a convolution of the Gaussians corresponding to 1, 2,..., n photoelectrons, where the Poisson distribution determines the relative weights. A fit of the single photoelectron spectrum will reveal the pedestal position p 0, and the mean p 1 and standard deviation σ 1 of the Gaussian corresponding to exactly 1 photoelectron. By definition, the PMT gain G is the (average) number of electrons produced by the anode following the multiplication of one photoelectron [113]: G = (p 1 p 0 )Q ADC (6.17) e where Q ADC is the charge per ADC count and e is the electron charge. Because the expected number of photons per pixel is zero or one for protons, increasing with the square of the incident particle charge up to few hundreds for iron nuclei, one of the most important parameters is the pixel single photoelectron resolution σ 1 φ = (p 1 p 0 ), (6.18) that is related to the charge resolution of the detector (the lesser is φ the better is photon counting, hence the better is the charge measurement). A more complicate statistical treatment of the single photoelectron spectrum [114] brings to the following probability distribution: P(x) e λ δ(x) + (1 e λ ) e λ λ x/s Θ(x) (6.19) S λ Γ(x/S ) where we make use of the Dirac s distribution δ(x), the Euler s Gamma function Γ(x) = e t t x 1 dt and the Heavyside s step function Θ(x). With this approach, 0 the parameter λ and the scale factor S are used to write the gain G S (1 + λ) and the single photoelectron response φ λ/(1 + λ) Magnetic shielding The fringing field of the AMS-02 superconducting magnet is of the order of G in the RICH PMT plane. This is not as high as the field that forced the 144

155 6.2 The RICH design Figure 6.14: Magnetic field (in Gauss units) in the RICH PMT plane. adoption of a different kind of phototube for TOF and ACC: the problem can be solved by putting a shielding material around the PMT. Howewer, the limited weight budget forces a solution where the heavy ferromagnetic shield material is minimized. In practice, shielding boxes of different thickness will be used in the different parts of the PMT plane, according with the simulated field map (figure 6.14). One important effect was discovered making tests with arrow of boxes in solenoidal magnetic fields [115]: while a single box made of soft iron or VA- COFLUX 50 with 0.8 mm thick walls is a perfect shield for a transversal field of 300 G, when 6 boxes are aligned along the field direction a collective effect grows up that lowers the screening of the central boxes. The simulation of this boundary conditioned magnetostatic problem is not an easy task: whereas the external field is very simple and can be computed analytically after defining the currents, the field in the zone where the strictly packed 145

156 THE AMS-02 RICH SUBDETECTOR Data d s Magnetic field (G) at given position (mm) (mm) real sim real sim Table 6.1: Comparison between measurements carried on in Bologna with two rows of 9 boxes inside Helmholtz coils and the ROXIE simulation. Each box has thickness d and it is distant s from the next one. The agreement is within 10 20% (measurement uncertainties are of the order of few Gauss). Figure 6.15: Helmholtz coils used in Bologna to test two rows of 9 shielding boxes, reconstructed with ROXIE. boxes are placed must be computed with a finite element method. With the help of S. Russenschuck, one of the authors of the ROXIE 2 program, and M. Aleksa, both working in the CERN LHC magnet division (AT division, MEL group), I was able to simulate the behavior of 1 3 rows of boxes, and to reproduce the measured effect within 10 20% (table 6.1 shows an example). At the same time, test boxes with 0.6 mm, 0.8 mm, 1.0 mm, 1.2 mm thick

157 6.2 The RICH design Figure 6.16: ROXIE finite element model of a grid of 3 9 ferromagnetic boxes, with horizontal Al bars. Boxes and bars thicnkess: 1 mm. Reflection symmetry with respect to x and y applies. walls were produced and tested in Bologna using two Helmholtz coils (figures 6.15 and 6.16). The mechanical design of the RICH PMT plane was then changed in order to cope with the shielding problem: thin (0.6 mm) boxes are used where the field intensity is not very high, whereas thicker boxes are used in the most critical parts. The worst situation is found in the large rectangular grids that are parallel to the x direction, where the rows have 17 PMTs. Here the central boxes are the thickest ones (d = 1.2 mm), and thin material is used for the outer boxes. In addition, an increased gap between adjacent phototubes has been designed in order to reduce the collective effects (figure 6.17). The superconducting magnet is well simulated in vacuo by ROXIE (figure 6.18), obtaining values that agree with the official field map from the magnet producer within roughly 1%. The presence of the diamagnetic materials that constitute AMS-02 does not sensibly affect the field map. Howewer, the RICH PMT plane and the ECAL phototubes just below the RICH have ferromagnetic shielding boxes that will affect the magnetic field. The nearest PMTs to the magnet are half a meter below the lower TOF plane, hence their shielding will not sensibly affect the field inside the tracker (i.e. no appreciable effect on rigidity resolution). On the other hand, the field map just outside the core of AMS-02 will be someway distorted: this will affect the trajectory of the exiting particles from the spectrometer, and that of back-splash particles from ECAL and RICH. Presently there is no hope to be able to simulate the effect of thousand shielding boxes in a 147

158 THE AMS-02 RICH SUBDETECTOR Figure 6.17: Rectangular grid of shielding boxes, with different thickness along the x axis [109]. Nearly at the center of the structure, a thicker support layer is used, to increase the spacing and to facilitate the shielding effect. Figure 6.18: The AMS-02 superconducting magnet, simulated by ROXIE. 148

159 6.3 The beam test known magnetic field, because of the complexity of the numerical methods. Only real particles will make it possible to reconstruct the field configuration below the spectrometer. 6.3 The beam test The heavy ion beam test carried on at CERN SPS, on October 2002 (see chapter 7), has been very useful to compare the results of Na F and different kinds of silica aerogel tiles. At the same time, it was a very important validation test for the RICH simulation, and a test-bench for the read out electronics. The selection line was tuned to obtain 20 GeV/c per nucleon secondaries with A/Z = 2 ( 4 He and almost all stable nuclei up to iron), A/Z = 7/4 (mostly 7 Be), and A/Z = 1 (protons). Special runs were carried on with protons of lower energies, to study the velocity resolution of the RICH prototype (proton energy: 5, 7, 9, 11, 13 GeV). Figure 6.19 shows the image on the pixel plane of the Čerenkov ring produced by a 20 GeV/A Li nucleus The RICH prototype Figure 6.20 shows the grid of photomultipliers that has been used to assembly the RICH prototype (also known as MiniRICH ). The 96 PMTs are strictly packed (the light guides touch the neighboring ones), and the front-end electronics is the final one. The radiator mechanics is able to support tiles of different geometries, that can be placed from few centimeters to half a meter away from the pixel plane The radiators On the beam test, five different aerogel samples (three from Matsushita, one from Institute of Catalysis, Novosibirsk University) and one sodium fluoride radiator were used by the RICH prototype. The characteristics of the different aerogel samples are summarized in table 6.2. The charge resolution of the RICH is quite good: figure 6.21 shows the reconstructed charge distribution for run 312, using a n = 1.03 sample from Novosibirsk, while figure 6.22 shows the Gaussian fit to the He events (selected using data from scintillators). 149

160 THE AMS-02 RICH SUBDETECTOR Figure 6.19: Li event observed at CERN SPS [109]. Radiator Clarity d N exp 1 σ(z = 2) and n ( µm 4 /cm) (cm) NM ± NM ± OM ± Nov ± Nov ± Table 6.2: Comparison between different radiators with thickness d on He beam. New Matsushita samples ( NM ) with refractive indices 1.05 and 1.03 were tested along with an old Matsushita ( OM ) sample with n = 1.03, and two aerogel tiles from Novosibirsk ( Nov ) with n = 1.03 and n = 1.04 [116]. Nov 1.04 is the only hydrophobic sample, whereas all the other ones are hydrophilic. 150

161 6.3 The beam test 341 mm 279 mm Figure 6.20: MiniRICH PMT plane with light guides, top view. 151

162 THE AMS-02 RICH SUBDETECTOR Z Figure 6.21: Charge distribution seen by the RICH prototype with aerogel from Novosibirsk (n = 1.03) [116] FIT: µ=2.00 σ=0.20 Run 312 Nov charge Figure 6.22: Charge distribution for He nuclei of the Novosibirsk n = 1.03 sample [116]. Dots are real data, the histogram is the simulation. The fits gives the resolution of 0.2 charge units. 152

163 Chapter 7 Ion beam test at CERN SPS In order to understand the behavior of the AMS-02 detector, a very complex simulation has been developed by the Collaboration 1. The computer program is basically divided into two sections: first, the simulation of the analog signals produced by each subdetector and of their digitization by the front-end electronics is followed by the storage of the raw data on disk; second, the raw data are elaborated in order to get both low level (like the energy deposited in a given active part of AMS-02) and high level informations (for example the particle momentum). The user can choose to generate files containing raw data, to read directly real or simulated raw data from disk and process them, and to make both steps for each simulated event. The most delicate part of the simulation is the first one: the physical processes of all particles that may interact with active and passive parts of the detector have to be well understood and reproduced. This is done with the help of the GEANT 3.21 well tested FORTRAN framework [117] or of its updated C++ version Geant4 2. Howewer, extensive tests have to be done with detector prototypes in order to check the consistency of the simulation and to quantify the systematic deviations from the observed beavior. The best way to validate the detector simulation is to use particle beams with known properties, like particle charge, mass and momentum, and to look at the detector response under these well controlled conditions. The use of an external trigger makes possible to obtain absolute efficiencies as function of the particle energy, charge and direction, whereas the data analysis should be able to show the background produced by misidentified particles, bad energy or velocity measure

164 ION BEAM TEST AT CERN SPS ments, etc. In the specific case of AMS, that is a spectrometer with great acceptance and large energy range, the number of degrees of freedom is very high: it is necessary to sample the input parameters space in a systematic way. For example, AMS-01 was tested on a ion beam at GSI with 600 different incident directions [100]. The AMS-02 subdetectors have been tested using particle accelerators during the last few years: a 20 layer TRD prototype was tested in summer 2000 at CERN with GeV/c singly charged particles (electrons, muons, pions and protons) [93]; the ECAL prototype was tested at CERN in 2001 [118]; the RICH prototype was tested at CERN in October 2002 together with a tracker prototype and a couple of TOF counters. This chapter deals with the ion beam test carried on with the SPS accelerator at CERN, on October 14 19, Experimental setup The ion beam test of the RICH prototype was carried on at CERN on October 2002 using a 20 GeV/c per nucleon Pb beam accelerated by the Super Proton Synchrotron (SPS), colliding with a Be target (in T4, see figure 7.1). The secondary fragments (all nuclei up to Pb, mostly with the same momentum per nucleon as the primary beam [119]) were filtered through the H8 selection line, obtaining a 3 cm large ion beam with given A/Z ratio (i.e. defined rigidity). The RICH prototype (figure 7.2) was placed in the NA45 area, together with few parasitic detectors. Among the latter ones, two AMS-02 subdetector prototypes were tested: six tracker ladders in front of the RICH prototype and two TOF counters behind it. The particles of the secondary beam had a spill flat top of 12.7 s over a cycle of 19 s. During the spill on phase, the trigger rate was ranging from hundred to few thousand Hz, depending on the tuning of the selection line, that was set at values A/Z = 2, A/Z = 3/2, A/Z = 7/4, A/Z = 1 for each radiator type RICH prototype The AMS RICH prototype is a proximity focusing Cherenkov imager consisting of 96 units of 16-anode PMTs (1568 readout channels). It has been tested using the secondary fragments beam obtained from a 20 GeV/c per nucleon primary SPS Pb beam. Data were collected during a 4 days run with nuclear elements 154

165 * * 6 6 * Experimental setup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igure 7.1: SPS North Area. 155

166 ION BEAM TEST AT CERN SPS Figure 7.2: RICH prototype inside vacuum chamber, facing up during a radiator change. The radiator support is above the picture (not shown). CERN SPS, NA45 area, October having A/Z = 2 mass to charge ratio, going from deuterium (D) up to the Fe group, including all intermediate isotopes (most notably He, Li, Be, B, C, N, O, Al and Ca). Other beam line settings have been used to select different A/Z values, like 7/4 (to study 7 Be) and 3/2 (for 3 He). In addition to the data taken to study the charge resolution, special runs were carried on with protons between 5 and 20 GeV/c, to study the β resolution. The response of the prototype has been studied for various radiators (chapter 6): Silica aerogels with refractive index 1.03 and 1.05, and sodium fluoride (Na F). Čerenkov rings associated to ions over the covered mass range have been observed, with average number of photons ranging from 4 7 for Z = 1 elements, up to several thousands for high atomic numbers. The charge and mass resolutions are being evaluated from the analysis of the 5 million recorded events. 156