Universita degli studi di Torino

SCUOLA DI SCIENZE DELLA NATURA

Dipartimento di Fisica

A simulation study of the EUSO-SPB

trigger and reconstruction

performances

Tesi di Laurea in Fisica

Relatore:Prof. Mario Edoardo Bertaina

Corelatore:Dott. Francesco Fenu

Controrelatore:Dott. Raffaella Bonino

Candidato:Simone Cambursano

Anno Accademico 2015-2016

A Mamma, Papa, ed Erika

2

Abstract

JEM-EUSO (Extreme Universe Space Observatory on board Japanese Experi-ment Module) is a mission that aims at detecting Extreme Energy Cosmic Rayswith energy E > 5× 1019 eV. It will exploit the fluorescent light emitted in theair shower cascades produced by cosmic particles interacting in atmosphere. It isexpected that the instrument will be housed on the International Space Station(ISS), orbiting the earth every ∼ 90 min at an altitude of 400 km. The telescopehas a field of view of 60◦ and the surface of the Earth covered is approximatelyan area of around 1.7 × 105 km2. The JEM-EUSO telescope is formed by asystem of three Fresnel lenses and a focal surface constituted by ∼ 5000 multi-anode photomultipliers (MAPMT) for a total of ∼ 3×105 pixels. The detectionmethod is the single photoelectron technique; signals are counted in time frameof 2.5 µs. EUSO Super Pressure Balloon (EUSO-SPB) is a scaling prototypeof the JEM-EUSO experiment. This experiment will be the first to measureshowers using the fluorescence technique from the top of the atmosphere. TheEUSO-SPB instrument will be placed on a balloon flying at an altitude of ∼ 40km. The proposed launch site is Wanaka NZ on April 2017. This mission will bethe first extended duration operation of a cosmic ray UV fluorescence detectorat near space altitude, providing an important occasion to test all the tech-nological components of JEM-EUSO. Among the main components, the firsttrigger level algorithm implemented on a FPGA for the acquisition of showersinduced signals will be tested. Such signals must be necessary discriminatedfrom those generated from the atmospheric UV background fluctuations. Aimof this thesis was to evaluate with Monte Carlo simulations the first level triggerperformances of the EUSO-SPB instrument estimating the number of triggeredevents expected for a ultra-long duration flight. More in detail, the response ofthree trigger methods was tested simulating showers in different cloudy cover-age condition. The analysis was performed scanning different values of detectorefficiency and background. A further study was done considering different timeframe acquisition of the detector. The energy reconstruction capability of theinstrument was investigated together with a work aimed to improve the effi-ciency of the reconstruction algorithm. In this thesis is also included the workdeveloped during a period of two-month at the Astroparticule et CosmologieLaboratoire (APC) aimed to the characterization and assembly of the EUSO-SPB instrument. By means of measures carried out on the photomultipliers thatwill compose the detector it was possible to get an efficiency map which wasimplemented in the Euso Simulation and Analysis Framework (ESAF). In thisway fine tuned simulation could be performed giving a more realistic estimationof the number of expected events.

3

Sommario

JEM-EUSO (Extreme Universe Space Observatory on board Japanese Experi-ment Module) e una missione che mira alla rilevazione di raggi cosmici di en-ergia estrema E > 5 × 1019 eV. Per raggiungere tale scopo sfruttera la luce difluorescensa emessa negli sciami prodotti da particelle cosmiche che interagis-cono con l’atmosfera. Si prevede che lo strumento sara ospitato sulla StazioneSpaziale Internazionale (ISS) compiendo un orbita attorno alla terra ogni 90minuti a un’altezza di 400 km. Il telescopio ha un campo visivo di 60◦ e lasuperficie della Terra coperta si estende su un’area di circa 1.7 × 105 km2. Iltelescopio JEM-EUSO e formato da un sistema di tre lenti di Fresnel e da una su-perficie focale costituita da ∼ 5.000 fotomoltiplicatori multianodo per un totaledi ∼ 3×105 pixels. Il metodo di rilevazione e la tecnica del singolo fotoelettrone;i segnali sono contati in finestre temporali di 2.5 µs. EUSO Super Pressure Bal-loon (EUSO-SPB) e un prototipo in scala dell’esperimento JEM-EUSO. Questoesperimento sara il primo a misurare sciami atmosferici attraverso il metododella fluorescenza dal top dell’atmosfera. Il detector di EUSO-SPB verra collo-cato su un pallone che volera a un’altezza di ∼ 40 km. Il sito di lancio propostoe Wanaka (NZ) in Aprile 2017. Questa missione sara la prima operazione diestesa durata di un detector per la rivelazione di fluorescenza ultravioletta in-dotta da raggi cosmici a un’altezza prossima allo spazio, fornendo un’importanteoccasione per un test di tutte le componenti tecnologiche di JEM-EUSO. Trale principali componenti testate vi sara l’algoritmo del primo livello di triggerimplementato su FPGA per l’acquisizione di segnali indotti da sciami. Tali seg-nali devono essere necessariamente discriminati da quelli generati da semplicifluttuazioni del background atmosferico. Scopo di questa tesi e stato valutarecon simulazioni Monte Carlo le performances relative al primo livello di triggerdando una stima del numero di eventi che ci si aspetterebbe nel caso di volo aultra-lunga durata. Nello specifico sono stati testati tre diversi metodi di trig-ger simulando sciami in diverse condizioni di copertura nuvolosa. L’analisi estata effettuata selezionando diversi valori di efficienza complessiva del detec-tor e di backgroud. Un ulteriore studio e stato eseguito considerando diversedurate del time frame di acquisizione del detector. Le capacita di ricostruzionedell’energia di EUSO-SPB sono state investigate unitamente a uno studio final-izzato al miglioramento dell’efficienza di ricostruzione. Nel contesto di questatesi si inserisce inoltre un lavoro svolto durante uno stage di due mesi pressol’Astroparticule et Cosmologie Laboaratoire (APC) volto alla caratterizzazionee all’assemblaggio delle componenti del detector di EUSO-SPB. A seguito dellemisure effettuate sui fotomoltiplicatori che costituiranno il rivelatore e statopossibile ottenere una mappa delle efficienze che e stata successivamente imple-mentata nel software di simulazione Euso Simulation and Analysis Framework

4

(ESAF). In questo modo si sono potute effettuare simulazioni piu realistichedando una stima piu precisa del numero di eventi attesi.

5

Ringraziamenti

Trovo naturale iniziare questa tesi, conclusione di diversi mesi di lavoro ringraziandole persone che lo hanno reso possibile.

Sinceri ringraziamenti al mio relatore Mario Edoardo Bertaina, con pazienzae entusiasmo mi ha seguito in questo affascinante percorso.

Sentiti ringraziamenti al mio corelatore Francesco Fenu. Il suo aiuto ai finidella realizzazione di questo lavoro e stato inestimabile.

Grazie di cuore a Mamma, Papa ed Erika per avermi sostenuto in tutti questianni senza mai avermi fatto mancare il loro appoggio.

Grazie a tutti i miei amici, per essere stati con me in tutti questi anni e concui ho costruito un legame davvero importante.

Grazie a Marce e Cata, gli “Gli Ortolani”, per tutte le risate e le seratepassate insieme.

Grazie a tutti gli amici incontrati a fisica, con cui umanamente ho condivisobellissime esperienze.

6

Acknowledges

A sincere thanks to people of JEM-EUSO collaboration.Real thanks to Etienne Parizot and Guillaume Prevot for their helping in

the organization of my Traineeship in Paris.Thanks to the whole APC crew. Sincere thanks to Simon Bacholle, Pierre

Barrillon, Sylvie Blin, Francesca Chapel, Philippe Gorodetzky, Aera Jung, LechWiktor Piotrowski, Julio Arturo Rabanal Reina.

Special thanks to Hiroko Miyamoto for introducing me to her Tur-Lab ac-tivity of research.

Thanks to the “Paris Family”, Hector Garcia Escandon Miranda, VictorHugo Garcia Rueda, Wojciech Jaworski.

7

Contents

1 JEM-EUSO 201.1 Science objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . 211.2 Requirements and Expected Performances . . . . . . . . . . . . . 211.3 Observational principle . . . . . . . . . . . . . . . . . . . . . . . . 221.4 Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

1.4.1 Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221.4.2 Focal surface . . . . . . . . . . . . . . . . . . . . . . . . . 241.4.3 Focal Surface electronics . . . . . . . . . . . . . . . . . . . 251.4.4 The Atmospheric monitoring system . . . . . . . . . . . . 27

1.5 EUSO-Balloon . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271.5.1 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . 281.5.2 The EUSO-Balloon instrument . . . . . . . . . . . . . . . 281.5.3 The Flight . . . . . . . . . . . . . . . . . . . . . . . . . . 311.5.4 Preliminary Results of EUSO-Balloon first flight . . . . . 31

1.6 EUSO-TA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331.6.1 EUSO-TA telescope . . . . . . . . . . . . . . . . . . . . . 341.6.2 Calibration and observation . . . . . . . . . . . . . . . . . 351.6.3 Laser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351.6.4 Cosmic Ray events . . . . . . . . . . . . . . . . . . . . . . 36

2 Mini-EUSO and EUSO-SPB 372.1 Mini-EUSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

2.1.1 Scientific objectives . . . . . . . . . . . . . . . . . . . . . 372.1.2 Techonological objectives . . . . . . . . . . . . . . . . . . 382.1.3 The Instrument . . . . . . . . . . . . . . . . . . . . . . . . 38

2.2 EUSO-SPB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 412.2.1 The EUSO-SPB Instrument . . . . . . . . . . . . . . . . . 422.2.2 Estimating the EAS event rate . . . . . . . . . . . . . . . 422.2.3 EUSO-SPB expected mission’s conditions . . . . . . . . . 44

3 A simulation study of the trigger performances of EUSO-SPB 463.1 ESAF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463.2 Triggered spectrum calculation . . . . . . . . . . . . . . . . . . . 503.3 Preselection of showers initial parameters . . . . . . . . . . . . . 523.4 The EUSO-SPB persistence tracking trigger . . . . . . . . . . . . 54

3.4.1 PTT : method 1 . . . . . . . . . . . . . . . . . . . . . . . 553.4.2 PTT : methods 2a and 2b . . . . . . . . . . . . . . . . . . 55

8

3.5 Estimation of the background trigger rate for PTT method 1, 2aand 2b . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

3.6 EAS simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 573.6.1 High quality detector performances . . . . . . . . . . . . . 573.6.2 Low quality detector performances . . . . . . . . . . . . . 633.6.3 Variable background analysis . . . . . . . . . . . . . . . . 653.6.4 Variable GTU duration analysis . . . . . . . . . . . . . . 65

3.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

4 Evaluation of the EUSO-SPB reconstruction performances 734.1 The reconstruction algorithm . . . . . . . . . . . . . . . . . . . . 73

4.1.1 Pattern recognition . . . . . . . . . . . . . . . . . . . . . . 734.1.2 Angular and profile reconstruction . . . . . . . . . . . . . 744.1.3 Energy and Xmax reconstruction . . . . . . . . . . . . . . 75

4.2 Reconstruction performances analysis . . . . . . . . . . . . . . . 754.2.1 Energy resolution for fixed event conditions . . . . . . . . 754.2.2 Triggered spectrum against fraction of reconstructed events 774.2.3 Reconstruction algorithm optimization . . . . . . . . . . . 80

4.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

5 Integration and calibration aspects of the EUSO-SPB PDM 905.1 PMT sorting for the EUSO-SPB PDM . . . . . . . . . . . . . . . 90

5.1.1 Photon detection . . . . . . . . . . . . . . . . . . . . . . . 905.1.2 Experimental setup . . . . . . . . . . . . . . . . . . . . . . 925.1.3 Data acquisition . . . . . . . . . . . . . . . . . . . . . . . 92

5.2 Trigger response for a PDM with non uniform efficiency . . . . . 975.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

6 Conclusions 99

9

List of Tables

2.1 Mini-EUSO instrument main parameters . . . . . . . . . . . . . . 392.2 Mini-EUSO instrument main parameters . . . . . . . . . . . . . . 402.3 EUSO-SPB instrument’s upgrades . . . . . . . . . . . . . . . . . 422.4 Expected Cosmic Ray Air Shower detection rates. Photon thresh-

old is the number of photons reaching the instrument optical en-trance aperture in each of 5 consecutive 2.5µs timebins. The darkperiod assumed is listed in table. . . . . . . . . . . . . . . . . . . 45

2.5 Example of expected hours of operation for a dark period at 45◦

S (see text). This dark period includes 18 days, indicated byan asterix, with at least 2.5 hours of moon down darkness for atotal of 118 hours. These hours would be the primary targetedobservation times for EUSO-SPB. . . . . . . . . . . . . . . . . . . 45

3.1 Trigger rates for method 1, 2a and 2b. As expected method 2bgives an higher trigger rate than method 2a because of its lowerthreshold. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

3.2 Number of triggered events by method 1, 2a, and 2b consideringsets of Nsimu,cyl = 990 and 45◦ events for different energies. . . . 57

3.3 Number of events referred to triggered spectra by methods 1,2a and 2b simulating an high performances detector in clear skycondition. A 2 months flight duration and duty cycle of 0.13 areassumed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

3.4 Triggered events by method 1, 2a, and 2b simulating sets ofNsimu,cyl = 990 events, 45◦ inclined at different energies, in dif-ferent cloudy conditions. . . . . . . . . . . . . . . . . . . . . . . . 60

3.5 Number of events referred to triggered spectra obtained withmethods 1, 2a and 2b simulating an high performances detec-tor in different clouds coverage conditions. . . . . . . . . . . . . . 61

3.6 Comparison between triggered spectra performed by methods 1and 2b in clear sky condition, assuming a detector overall effi-ciency of 0.05 (low performance detector) and 0.15 (high perfor-mance detector). . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

3.7 Number of events referred to triggered spectra performed bymethod 1 and 2b in cloudy sky condition, assuming an overallefficiency of the detector of ∼ 0.05. . . . . . . . . . . . . . . . . . 64

3.8 Number of triggered events comparing two different backgroundlevels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

3.9 Sets of parameters used for the variable GTU duration analysis. . 68

10

3.10 Number of events considering different GTU duration with npst

= 1, background = 0.25 counts/µs/pixel and efficiency = 0.05.In case of GTU duration of 1 µs, 2 sets of trigger thresholdsare used in order to avoid rounding effects for low rates of bkg(ncell

th = 8 is for 0.2 counts/GTU/pixel while ncellth = 11 is for 0.3

counts/GTU/pixel). Except for GTU = 1 µs with the nominalbackground of 0.25 counts/µs/pixel rounded to 0.3 counts/µs/pixel,an higher number of events is expected assuming shorter GTUs. 69

3.11 Number of triggered events assuming different GTU durationwith npst = 1, background = 0.75 counts/µs/pixel and efficiency= 0.15. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

3.12 Number of triggered events assuming different GTU durationwith npst = 2, background = 0.25 counts/µs/pixel and efficiency= 0.05. Also in this case two set of thresholds correspondingto two different values of rounded background were simulated inorder to avoid rounding effects occurring for low background rate. 70

3.13 Number of triggered events assuming different GTU durationwith npst = 2, background = 0.75 counts/µs/pixel and efficiency= 0.15. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

4.1 Energy resolution for samples of 1 · 1018 eV, 1.5 · 1018 eV and 5· 1018 eV simulated events. . . . . . . . . . . . . . . . . . . . . . 76

4.2 Number of reconstructed events compared to triggered spectrasimulated in different atmospheric conditions, assuming an over-all efficiency of the detector equal to 0.15. . . . . . . . . . . . . . 80

4.3 Number of reconstructed events compared to triggered spectrasimulated in different atmospheric conditions, assuming an over-all efficiency of the detector equal to 0.05. . . . . . . . . . . . . . 82

4.4 Number of reconstructed events assuming two different values ofthe reconstruction parameters lpixel,ground that corresponds to theaverage dimension of a pixel at ground. The value of 120 m is thebest tuned value providing the highest number of reconstructedevents. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

4.5 Number of reconstructed events with the optimized version of thealgorithm compared to triggered spectra simulated in differentatmospheric conditions. An overall efficiency of the detector of∼ 0.15 is assumed. . . . . . . . . . . . . . . . . . . . . . . . . . . 88

4.6 Number of reconstructed events with the optimized version of thealgorithm compared to triggered spectra simulated in differentatmospheric conditions. An overall efficiency of the detector of∼ 0.05 is assumed. . . . . . . . . . . . . . . . . . . . . . . . . . . 88

4.7 Comparison between fractions of reconstructed events performedwith and without the optimized version of the algorithm. Re-constructed fractions are referred to spectra simulated with anoverall efficiency of the detector of ∼ 0.15. . . . . . . . . . . . . . 88

4.8 Comparison between fractions of reconstructed events performedwith and without the optimized version of the algorithm. Re-constructed fractions are referred to spectra simulated with anoverall efficiency of the detector of ∼ 0.05. . . . . . . . . . . . . . 89

11

5.1 Average gain of a sample of sorted PMTs expressed in unit of 106 975.2 Number of triggered events comparing a non uniform to an uni-

form detector. An average overall efficiency of 0.05 and an aver-aged background of 0.25 counts/µs/pixel are assumed. . . . . . . 98

6.1 Example of expected hours of operation for a dark period at 45◦

S (see text). This dark period includes 18 days, indicated by anasterisk, with at least 2.5 hours of moon down darkness for atotal of 118 hours. . . . . . . . . . . . . . . . . . . . . . . . . . . 100

12

List of Figures

1.1 The observational principle of JEM–EUSO. JEM-EUSO is de-signed to monitor from space the earth’s atmosphere, lookingtowards nadir during night-time, to detect the ultra violet tracks(290-430 nm) generated by the extensive air showers (EAS) prop-agating through the atmosphere. Trough the detection of thefluorescence and Cherenkov photons of the EAS, the energy, ar-rival direction and nature of the primary UHECR particle will bedetermined. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

1.2 Left: Spatial profile of photons from an EAS caused by a 1020 eVproton with zenith angle 60◦. Each square represents a Multi-anode Photomultiplier Tube (MAPMT) with 64 channels, ar-ranged in units of 6 × 6 MAPMTs. Subsections of 6 differentPhoto-Detector Modules (PDMs) are shown. The small squaresshow the number of photons detected by each channel of theMAPMT. This event is crossing diagonally three PDMs, withthe shower starting to develop in the bottom left PDM and con-tinuing in the right top PDM. Right: Time profile of photons,obtained summing all photons of the previous picture detectedon each Gate Time Unit (GTU) (2.5 µs). It is possible to see thecontribution to the signal from the three components (UV light− blue, Cherenkov peak − red, scattered Cherenkov − green). . 23

1.3 Scheme of the JEM-EUSO system: The triple lens structure isshown here. Behind the focal surface the data processing elec-tronic is mounted. Attached to the external part of the detectorboth LIDAR and Infrared Camera are visible. . . . . . . . . . . . 23

1.4 Structure of the Focal Surface. The 2.5 m surface is divided in137 PDM modules. Each PDM contains 36 MAPMTs, each with64 independent channels. The bottom left corner shows a PDMprototype with 36 MAPMTs installed. . . . . . . . . . . . . . . . 24

1.5 Scheme of the signal flow trough the JEM-EUSO focal surfaceelectronics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

1.6 The launch of EUSO-Balloon. . . . . . . . . . . . . . . . . . . . . 281.7 left: EUSO-Ballon, ready for its first flight, Timmins, Ontario

(Ca), August 2014, right: schematic view of the instrument boothand optical bench, without floaters and “crash rings”. . . . . . . 29

1.8 The Photo-Detector Module(PDM). . . . . . . . . . . . . . . . . 29

13

1.9 The intensity map of UV background in logarithmic scale(relativeunits). The red and light blue areas are related to cloud coverage(see Fig 1.10). The displayed values are relative to the meanvalue of UV background intensity over reference area “A” [36]. . 32

1.10 The map of IR radiance (relative units). The map is created byaveraged values for particular positions. The values were chang-ing in time due to movement of clouds and motion of EUSO-Balloon. The displayed values are relative to the mean value ofIR radiance over reference area “A” [36]. . . . . . . . . . . . . . . 33

1.11 Left side: Time fit of laser event, with constant timing error of 0.5GTUs. Right side: Zenith angle reconstruction of the helicopterlaser shots. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

1.12 The design of the EUSO-TA optics. The left and right panelsshow the ray trace and spot diagrams for different incident angles. 34

1.13 Left: an average of ∼ 250 shots of CLF laser; right: an averageof ∼ 150 inclined shots of the Colorado School of Mines laser,located at 40 km from EUSO-TA (62 mJ), the missing part dueto a non-functioning MAPMT in the center of the focal surface.The color scale on both pictures denotes the uncalibrated detectorcounts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

1.14 An UHECR event of ∼ 1018 eV energy traversing at ∼ 2.5 kmdistance from EUSO-TA and therefore visible as a track on asingle GTU, in this case going from top-right to bottom-left (dis-tance and energy are estimated by TA). The color scale denotesuncalibrated counts of the detector. The bottom right PMTs andthe center bottom right PMT were not working at the time of themeasurement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.1 left: Mini-EUSO scheme, right a photo of Zvezda module wheremini-EUSO will be hosted. . . . . . . . . . . . . . . . . . . . . . . 39

2.2 Point spread function of the optical system for light coming par-allel with different inclination (from left to right, 0◦, 5 ◦, 10◦, 15◦

and 19◦). The diameter of the inner white circle is 2.5 mm andthe outer circle is 5 mm. . . . . . . . . . . . . . . . . . . . . . . . 40

2.3 The simplified block diagram of the Mini-EUSO Data Processor. 412.4 Mini-EUSO Data Handling System. . . . . . . . . . . . . . . . . 412.5 The path of the 2015 NASA Super Pressure Balloon engineering

flight launched from Wanaka NZ. The flight duration was 32 days. 422.6 Left: Sample distribution of simulated EAS directions projected

on to a unit hemisphere. Center: EAS core location projected tothe ground. Black points indicate all the events simulated. Redpoints indicated events for which some light entered the EUSO-SPB optical aperture. Right: Similar to the adjacent cent plotwith the additional requirement that at least 300 photons reachEUSO-SPB on at least each of 5 consecutive 2.5 µs time bins.The energies of these simulated events is 1018.75 eV. . . . . . . . 43

2.7 An example simulated event (E=1018.75 eV) passing the selectioncriteria for EUSO-SPB (see text). Left: The projected trackRight: The EAS light curve. . . . . . . . . . . . . . . . . . . . . . 44

14

2.8 Left: The fraction of observable EASs relative the the numbersimulated for different photon thresholds (see text). Right: Theflux of EASs observable by EUSO-SPB. The integral of thesecurves yields the expected event rate. . . . . . . . . . . . . . . . . 44

3.1 a schematic view of the ESAF Simu application structure. Themain application is the so called SimuApplication. The Light-ToEuso application takes care of all the physical process fromshower to detector. The EusoDetector application performs thesimulation of optics and electronics. Black lines represent thecalling hierarchy. Red arrows represent the flux of simulated in-put–output necessary for all the called applications. . . . . . . . 48

3.2 a sketch of the reconstruction framework. The main applicationRecoFramework calls iteratively the MakeModule method whichallocates all the required modules. A vector of pointers to theallocated objects is saved under the name fModules. In the Ex-ecute method the operations of all the modules are performed.All the modules are inheriting from the RecoModule class. Thevirtual methods PreProcess, Process, PostProcess and SaveRoot-Data are called for all the allocated modules. Note that blueboxes represent classes, blue–gray boxes methods, the gray boxis a C++ vector and the circular arrow indicates iterative repe-tition of some method or sequence of methods. . . . . . . . . . . 49

3.3 Logarithmic plot of the measured cosmic ray flux (black markers)reported by the Pierre Auger Collaboration [73] fitted with a 2-degrees polynomial (red curve). . . . . . . . . . . . . . . . . . . . 51

3.4 Representation of the cylinder used as a constraint to generatethe initial showers parameters. The radius of the cylinder is set13 km meaning an area comparable with the balloon field of view.An injecting area radius of 50 km is assumed. Orange arrows rep-resent shower passing trough the cylinder whose parameters willbe processed by ESAF during the simulation, while red arrowsrepresent shower whose parameters are rejected. . . . . . . . . . 53

3.5 Description of the method 1 trigger. Only 3x3 pixel boxes whichdo not belong to different EC can provide a trigger (the grey-

dashed cell is exluded). A set of trigger parameters with npixthr =

3, npst = 1 and ncellthr = 12, corresponding to an average back-

ground of 0.25 counts/µs/pixel, can be considered as a way ofexample. Numbers in red are referred to pixels counting a num-ber of photoelectrons equal or higher npix

thr = 3. To 3 countedphotoelectrons corresponds an excess of 1, to 4 counted photo-electrons corresponds an excess of 2,etc. The trigger condition issatisfies by the 3 × 3 green box where the count of excess photo-electrons, considering an integration time of 1 GTU (npst = 1),amounts to 14. Such as value is higher than the set threshold forthe entire box (ncell

thr = 12) so a trigger can be issued. . . . . . . . 55

15

3.6 Scheme of trigger method 2a and 2b. The EC named, as a wayof example, EC 1 is considered. Boxes including 1 are ideallyGTUs where the trigger condition for method 1, considering nthr

= 1, occurs. Starting from the GTU 9 the excess persists on 3over a bunch of 5 consecutives GTUs (see the blue-line box), soa trigger is issued. . . . . . . . . . . . . . . . . . . . . . . . . . . 56

3.7 Left: Example of a 45◦ event at 1 · 1018 eV triggered only bymethod 2b. Right: Light curve (green plot) representing thenumber of detected photons in each GTU. For low energy events,the lower number of produced and detected photons (comparedfor example to the number of photons associated to the 5·1018 eVevent shown in Fig.3.8 right panel) allows method 2b to performbetter compared to the others because of its lower threshold. . . 58

3.8 Left: Example of 5 · 1018 eV event entirely crossing the field ofview, observed by all the methods. Right: Light curve (greenplot) representing the detected photons. In case of extremelystrong signal such as for this particular class of events, the triggerefficiency is the same independently by the adopted method. . . 58

3.9 Left: Example of Cherenkov like event at 5 ·1018 eV. Right panel:a relatively strong signal due to the reflection of the Cherenkovlight at ground occurs on GTU 22 (green curve), lasting the dura-tion of 1 GTU. Such an event can be triggered only by method 1,which requires a localized excess of signal persisting only 1 GTU,while methods 2a and 2b look for a persistence of the signal onthe EC surface for a few GTUs. . . . . . . . . . . . . . . . . . . . 59

3.10 Left: Plots displaying triggered spectra performed by method 1(red curve), method 2a (green curve), method 2b (blue curve)assuming a 2 months flight duration and a duty cycle of 0.13.An high performances detector corresponding to an overall effi-ciency of 0.15 is assumed. Right: Error bars plot referred to thetriggered spectrum obtained with method 1. The uncertainty onthe total number of triggered events is calculated propagating theuncertainties on each energy bin as shown in formulas 3.4 and 3.6). 60

3.11 Spectra obtained in cloudy coverage conditions assuming an highperformances detector. In each canvas red curves are referred tomethod 1, green curves to method 2a and blue curves to method2b. Top panels: Triggered spectra for thin low altitude clouds(left) and thin middle altitude clouds (right). Bottom panels:Triggered spectra for thick low altitude clouds (left) and thickmiddle altitude clouds (right). . . . . . . . . . . . . . . . . . . . . 62

3.12 Triggered spectra by method 1 (red plot) and 2b (blue plot) inclear sky condition assuming an overall efficiency of 0.05. . . . . 63

3.13 Spectra obtained in cloudy coverage conditions assuming a lowperformances detector. In each canvas red curves are referred tomethod 1 and blue curves to method 2b. Top panels: Triggeredspectra for thin low altitude clouds (left) and thin middle altitudeclouds (right). Bottom panels: Triggered spectra for thick lowaltitude clouds (left) and thick middle altitude clouds (right). . . 64

3.14 Triggered spectrum (method 1) in clear sky condition consideringan efficiency of ∼ 0.05 and average background of 0.5/µs/pixel. . 66

16

3.15 Spectra simulated with method 1 assuming an efficiency of ∼ 0.05and a background of 0.5 counts/µs/pixel. Top panels: Triggeredspectrum for thin low altitude clouds (left) and thin middle alti-tude clouds (right). Bottom panels: Triggered spectrum for thicklow altitude clouds (left) and thick middle altitude clouds (right). 66

3.16 Assuming an average dimension of the 3×3 trigger pixel box of∼ 300 m at an altitude of 5-7 km, it is possible to estimate thepersistence time of the signal in the 3x3 trigger pixel box as afunction of the shower zenith angle. . . . . . . . . . . . . . . . . . 67

3.17 The persistence time of the signal in the 3x3 trigger pixel boxdepends on the zenith angle of the shower . . . . . . . . . . . . . 68

3.18 Triggered spectra for different GTU duration assuming npst = 1.An average background of 0.25 counts/µs and an overall effi-ciency of the detector of 0.05 are considered. Black curves arereferred to the case GTU = 1 µs rounding the background to 0.3counts/µs/pixel (left panel) and 0.2 counts/µs/pixel (right panel). 69

3.19 Triggered spectra for different GTU duration assuming npst = 1.An average background of 0.75 counts/µs/pixel and an overallefficiency of the detector of 0.15 are considered. . . . . . . . . . . 70

3.20 Triggered spectra for different GTU duration assuming npst = 2.An average background of 0.25 counts/µs/pixel and an overallefficiency of the detector of 0.05 are considered. Black curves arereferred to the case GTU = 1 µs rounding the background to 0.3counts/µs/pixel (left panel) and 0.2 counts/µs/pixel (right panel). 70

3.21 Triggered spectra for different GTU duration assuming npst = 2.An average background of 0.75 counts/µs/pixel and an overallefficiency of the detector of 0.15 are considered. . . . . . . . . . . 71

4.1 Top panel. Integrated signal (background and shower track) overthe entire shower development. In principle no shower trackis recognizable. Bottom panel. Event signal generated by theshower. Task of the mentioned pattern recognition algorithm isthe identification of such a pattern. . . . . . . . . . . . . . . . . . 74

4.2 The reconstructed electrons profile is represented by green pointswith black error bars calculated as a consequence of the back-ground subtraction. The flat blue lines identifies the averagebackground level, while the red curve represents the fitting curve,whose parameters are the estimated energy and Xmax. The blackcurve is the effective number of electrons generated in the simu-lation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

4.3 Distribution of α for 45◦ events at 1 · 1018 eV. The energy reso-lution is calculated as the standard deviation of a Gaussian dis-tribution fitting the histogram. . . . . . . . . . . . . . . . . . . . 77

4.4 Distribution of α for 45◦ events at 1.5 · 1018 eV. . . . . . . . . . 784.5 Distribution of α for 45◦ events at 5 · 1018 eV. . . . . . . . . . . 78

17

4.6 a)Distribution of the altitude of the production position of pho-tons for 45◦ events at 5 · 1018 eV including the maximum in thefield of view. The distribution is centered on the altitude valueof ∼ 5000 m. b)Example of light curve (green curve) for an eventwhose maximum is well into the field of view. c)Distribution ofthe altitude of the production position of photons for 45◦ eventsat 5 · 1018 eV triggering on the initial (or final part) of the shower.The distribution shows clearly two peaks centered on ∼ 2000 mand ∼ 9000 m (representing therefore photons coming from loweraltitude associated to the tail of showers and photons comingfrom higher altitude associated to the initial part of showers).d)Example of light curve (green curve) for an event triggering onthe tail of the shower development. . . . . . . . . . . . . . . . . . 79

4.7 Triggered spectrum against number of reconstructed events inclear sky condition, assuming an overall efficiency of the detectorequal to 0.15. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

4.8 Top panels: Triggered spectrum against number of reconstructedevents simulating clouds with h = 2 Km and optical depth =1 (left). Triggered spectrum against number of reconstructedevents simulating clouds with h = 5 Km and optical depth =1 (right). Bottom panels: Triggered spectrum against numberof reconstructed events simulating clouds with h = 2 Km andoptical depth = 5 (left). Triggered spectrum against number ofreconstructed events simulating clouds with h = 5 Km and opticaldepth = 5 (right). An overall efficiency of the detector of 0.15 isassumed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

4.9 Triggered spectrum against number of reconstructed events inclear sky condition, assuming an overall efficiency of the detectorof 0.05. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

4.10 Top panels: Triggered spectrum against number of reconstructedevents simulating clouds with h = 2 Km and optical depth =1 (left). Triggered spectrum against number of reconstructedevents simulating clouds with h = 5 Km and optical depth =1 (right). Bottom panels: Triggered spectrum against numberof reconstructed events simulating clouds with h = 2 Km andoptical depth = 5 (left). Triggered spectrum against number ofreconstructed events simulating clouds with h = 5 Km and opticaldepth = 5 (right). An overall efficiency of the detector of 0.05 isassumed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

4.11 Triggered events distribution (red plot) compared to reconstructedevents distribution (green plot) as a function of the event zenithangle. Going to high values of zenith angle the discrepancy be-tween triggered events and reconstructed events increases, show-ing that for high zenith angle events the capability of the patternrecognition algorithm is highly dependent by the speed of thesignal on the focal surface. . . . . . . . . . . . . . . . . . . . . . . 84

18

4.12 The scheme illustrates the principle of the altitude speed depen-dency of the signal track on the focal surface of the detector.On the left side is shown the state for JEM-EUSO. Due to itsorbit altitude of ∼ 400 km the track speed of an air shower pro-jected on the surface of the detector can be considered constant,independently from the altitude of the shower development in at-mosphere. The track can be in fact seen as a point moving inthe field of view at an infinite distance from the detector. On theright side is shown the case for EUSO-SPB. Considering a flightaltitude of ∼ 40 km, the approximation of infinite distance airshower development is not valid and the altitude of the shower inatmosphere can influence the projected velocity of the signal onthe focal surface. . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

4.13 Signal speed on the detector surface against average altitude ofsignal generating photons for a sample of events at 65◦ (top panel)and 30◦ (bottom panel). The slopes of the red curves fitting thetwo plots identify in both of the cases an increase of ∼ 20%. . . . 87

5.1 An example of a single photoelectron spectrum. The spectrumis shown as a histogram of the number of pulses with a givencharge, in counts, returned by a charge-to-digital converter. Thefirst peak is the pedestal, corresponding to pulses in which nope are collected. The second peak, on the right, corresponds topulses in which one pe is collected. The gain is the differencebetween the means of the zero pe and one pe peaks (shown byred markers), while the efficiency of the PMT is proportional tothe surface of the one pe peak. . . . . . . . . . . . . . . . . . . . 91

5.2 Scheme of the setup used for the PMT sorting. . . . . . . . . . . 925.3 Example of S-curve containing a total amount of 64 curves, one

for each pixel of the PMT. In the region between 250 DAC and450 DAC is present the so called inflection point, corresponding tothe DAC value used to calculate the gain of the PMT. Between500 and 650 DAC one can see the region called plateau whichaverage count rate value is used to extimate the efficiency of thePMT. The right part is dominted by the pedestal. . . . . . . . . 93

5.4 Examples of S-curves related to the two main groups of PMTsfound out from the analysis. Top panel. S-curves taken froma PMT with non uniform plateaus from pixel to pixel. Bottompanel. S-curves taken from a PMT which pixels present uniformefficiency. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

5.5 Distributions of the pixels efficiency of each of the 100 tested PMTs. 965.6 Comparison between triggered spectra simulated assuming a uni-

form and a non uniform detector. . . . . . . . . . . . . . . . . . . 98

19

Chapter 1

JEM-EUSO

JEM-EUSO (Extreme Universe Space Observatory, on board the Japanese Ex-periment Module of the International Space Station) is a space based missionthat aims at study ultra high energy cosmic rays (UHECRs) at E ≥ 4×1019 eV[1], [2]. It will detect at night time the UV radiation (230-490 nm) generatedby UHECRs-induced exstensive air shower (EAS) propagating through the at-mosphere. The instrument will be attached to the ISS, orbiting the Earth every∼ 90 min at an altitude of ∼ 400 km with a speed of ∼ 7 km/s and an orbitalinclination of 51.6◦.

Figure 1.1: The observational principle of JEM–EUSO. JEM-EUSO is designedto monitor from space the earth’s atmosphere, looking towards nadir duringnight-time, to detect the ultra violet tracks (290-430 nm) generated by theextensive air showers (EAS) propagating through the atmosphere. Trough thedetection of the fluorescence and Cherenkov photons of the EAS, the energy,arrival direction and nature of the primary UHECR particle will be determined.

Due to its altitude JEM-EUSO significantly enhances the aperture(6 ÷ 7 ×105 km2 · sr) compared to any existing observatory. Furthermore, thanks to theISS orbit, JEM-EUSO surveys the whole celestial sphere with a rather uniform

20

exposure, minimizing the systematic uncertainties.

1.1 Science objectives

JEM-EUSO is devoted to the investigation of the nature and origin of extremeenergy cosmic rays (EECR) which, at ≥ 4× 1019 eV, constitute the most ener-getic component of the cosmic radiation. The main science goals of the missionare:

• A high statistics measurement above the GZK cut-off, in order to clarifyif the already measured flux suppression by the Pierre Auger Observatory[3] is a consequent result from the GZK mechanism or is due to othermechanisms, e.g. an intrinsic cut-off in the acceleration mechanism at thesources or a local deficit of sources.

• The study of the anisotropies of the EE sky in order to identify individualsources and measuring their energy spectra.

The spectrum of scientific goals of the JEM-EUSO mission also includesexploratory objectives as:

• The discovery of UHE Gamma-rays, whose shower maximum is stronglyaffected by the geomagnetic and LPM effects;

• the study of the UHE neutrino component, which can be achieved bydiscriminating weakly interacting events through the position of the firstinteraction point and the shower maximum;

• the study of the galactic and local extragalactic magnetic fields, throughthe analysis of the magnetic point spread function.

Taking advantage of its atmospheric sounding capabilities, JEM-EUSO willalso observe atmospheric luminous phenomena such as lightning, nightglow, andmeteors which are expected to emit in the UV band.

1.2 Requirements and Expected Performances

The key element to estimate the science potential of JEM- EUSO is its exposure[4]. It will reach ∼ 60 × 103 km2 · sr · yr at 1020 eV that is 9 times Auger.JEM-EUSO will well overlap (about one order of magnitude, starting from 2-3 × 1019 eV) with ground-based experiments to cross-check systematics andperformances. At higher energies JEM-EUSO will be able to accumulate statis-tics at a pace per year at about one order or magnitude higher than currentlyexisting ground based detectors. JEM-EUSO will also operate in tilt mode tofurther increase the exposure at the highest energies (E > × 1020 eV) by afactor of ∼ 3 compared to nadir mode. The total number of events expected atenergies E > 5.5 × 1019 in 5 years of operation ranges between 500 and 1200events depending on the CR flux in input based on TA or Auger results [6]. Thescientific requirements of the mission can be summarized as:

• observational area greater than 1.3 × 105 km2;

21

• direction reconstruction uncertainty ≤ 2.5◦ at E ×1020 eV;

• Energy resolution (expressed in terms of the 68% of the distribution) ≤30% for E > 1020 eV and θ = 60 ◦ [5];

• capability to discriminate between nuclei, gamma ray and neutrinos, whichimplies Xmax determination error ≤ 120 g/cm2 for 60◦ inclined showers atE = 1020 eV;

• measuring the timing properties of luminous atmospheric events with µsresolution.

1.3 Observational principle

JEM-EUSO will measure the UV light (290-430 nm) coming from EAS inducedby UHECR interaction in the Earth’s night atmosphere. The UV light is emittedby fluorescence and Cherenkov radiation in the atmosphere. UV photons fromthe shower are focused through a wide field of view (±30◦) diffractive opticsemploying Fresnel lenses. The light is detected by a 3 ×105 pixel focal planeelectronics which records the track of the EAS with a time resolution of 2.5 µsand an angular resolution of 0.074◦, corresponding to a pixel size of about 0.51km (0.61 km at the edge of FoV) at 400 km. The spatial and temporal profileof the event will allow to determine the energy and direction of the primaryparticles. For a cosmic ray particle penetrating the atmosphere, the numberof generated secondary particles is proportional to the shower energy and islargely dominated by electrons and positrons. The total energy carried by thecharged secondary particles is converted into fluorescence photons through theexcitation of the air nitrogen molecules. The fluorescence light is isotropic andproportional to the number of charged particles in the EAS. The atmosphericshower properties together with the detector response have been implementedin the software package ESAF (EUSO Simulation and Analysis Framework) [7].

1.4 Instrumentation

JEM-EUSO consists of a UV telescope and an atmospheric monitoring system(AMS). The telescope normally operates in single photon counting mode witha frame of 2.5 µs but is capable of switching to charge integration mode in caseof strong illumination. The exploded diagram of the telescope in Fig. 1.3 showshow different components will be assembled. These will be described in detailin the following sections.

1.4.1 Optics

The JEM–EUSO optical system is divided into :

• first lens: the lens facing space(double sided, Fresnel lens with 2.65 mexternal diameter);

• iris: is used to cut out the edge of the lenses;

22

Figure 1.2: Left: Spatial profile of photons from an EAS caused by a 1020 eVproton with zenith angle 60◦. Each square represents a Multi-anode Photomul-tiplier Tube (MAPMT) with 64 channels, arranged in units of 6 × 6 MAPMTs.Subsections of 6 different Photo-Detector Modules (PDMs) are shown. Thesmall squares show the number of photons detected by each channel of theMAPMT. This event is crossing diagonally three PDMs, with the shower start-ing to develop in the bottom left PDM and continuing in the right top PDM.Right: Time profile of photons, obtained summing all photons of the previouspicture detected on each Gate Time Unit (GTU) (2.5 µs). It is possible to seethe contribution to the signal from the three components (UV light − blue,Cherenkov peak − red, scattered Cherenkov − green).

Figure 1.3: Scheme of the JEM-EUSO system: The triple lens structure isshown here. Behind the focal surface the data processing electronic is mounted.Attached to the external part of the detector both LIDAR and Infrared Cameraare visible.

23

• second lens: is a Fresnel and diffractive surface lens, and acts to reducethe vignetting; and chromatic aberration. It is placed between the twomain lenses;

• third lens: the lens focuses the light on the focal surface.

The combination of 3 Poly(MethylMethAcrylate) - PMMA Fresnel lensesallows a full angle FoV of 60◦ with a resolution of 0.075◦, corresponding to apixel with a diagonal of about 550 m on earth [8], [9].

1.4.2 Focal surface

The focal surface (FS) is, in the current baseline, spherical with ∼ 2.5 m cur-vature radius, and 2.3 m diameter. It is covered with about 5.000 MAPMTs(Hamamatsu R11265-M64), each with 64 pixels. The quantum efficiency of thisdevice at peak wave length is about 40%. FS consists of 137 Photo-DetectorModules (PDMs). Each PDM comprises a 3 × 3 set of Elementary Cells (ECs).Each EC is formed by a 2 × 2 array of MAPMTs (see 1.4). The total numberof pixels is 315648. A Cokroft-Walton (CW) circuit handles the HV for eachEC-unit. In order to protect the PMTs from intense light events (TransientLuminous Events, meteors) each HVPS is able to reduce the current drawn bythe PMT in case of strong illumination [10]. The FS detector converts photonsinto electrical pulses, which are accumulated and processed by the electronicsduring a Gate Time Unit (GTU) of 2.5 µs.

Figure 1.4: Structure of the Focal Surface. The 2.5 m surface is divided in 137PDM modules. Each PDM contains 36 MAPMTs, each with 64 independentchannels. The bottom left corner shows a PDM prototype with 36 MAPMTsinstalled.

24

1.4.3 Focal Surface electronics

The 137 PDMs of the Focal Surface are each composed of 36 MAPMTs. Theyare individually read by an ASIC SPACIROC chip. Data coming from eachpixel are sent to an FPGA that reads the whole PDM, hosted in a PDM-board(Fig. 1.5) The PDM-board stores the data coming from the 36 MAPMTs in a128 frame circular buffer, with each frame corresponding to a GTU of 2.5 µs.The trigger logic on the PDM-board is thus continuously active on blocks of320 µs. The PDM trigger logic task is to reduce the background by a factor103. Each PDM has a valid trigger at 7 Hz rate. In case of a trigger, datafrom a sub-group of 8 PDMs are all sent to one of 18 CCBs (Cluster ControlBoards) for further processing. The CCB has the task of performing second-level-trigger, selecting potential events at a rate of 0.1 Hz on the whole FS. Incase of a positive trigger the acquisition of the whole FS is stopped and dataare sent to the CPU for storage.

Figure 1.5: Scheme of the signal flow trough the JEM-EUSO focal surface elec-tronics.

Front-End electronics

The JEM-EUSO Front-End ASIC, named SPACIROC (Spatial PhotomultiplierArray Counting and Integrating ReadOut Chip) [11], [12], [13] has been designedfor the readout of 64 channels MAPMTs. Its basic purpose is to perform singlephoton counting from UHECR showers from a few 1019 eV to 1021 eV (dynamicrange of 100 photons/GTU/pixel to take into account also the background).Additionally, in case of higher intensity light beyond single photon counting ca-pabilities, typical of atmospheric phenomena, it is capable of charge-to-time (Q-to-T) conversion called KI [14], extending the range of phenomena that can beobserved by the telescope, with range of 0-200 pe/GTU/pixel [15]. SPACIROCcircuit has 64 inputs dedicated to each single anode of one MAPMT. Readoutis carried out independently for each one of the 64 channels. The output ofthe photon counting discriminators and the KI comparators are digitized andprocessed by the digital part of the ASIC.

25

EC and PDM board

The PDM board electronics is based on a Virtex-6 FPGA. Referring to Fig1.5 data are sent from FS electronics to EC-ASIC board and from there tothe PDM board. Each EC-ASIC board handles 384 channels equivalent to 6different MAPMT. A PDM board handles 6 EC-ASIC boards. Every 2.5 µs,data (2.304 channels) are stored in a 128 GTU circular memory buffer. On thisdata, the trigger algorithm is continuously applied, looking for signal persistencethat may be due to EAS.

Cluster Control Board

The Cluster Control Board (CCB), part of the Data Processing (DP), receivesdata from 8 PDM boards for further processing. Its core is a Virtex-4 FPGAFX-60. Its task is to apply the second-level-trigger to the data coming from eachPDM board. In case of a valid trigger the acquisition is stopped and data fromrelevant PDMs (possibly also linked to other CCBs) is sent via SpaceWire to theon-board CPU system. Each CCB is expected to have a trigger rate of about10−3 Hz, for a total Focal Surface Trigger rate of 0.1 Hz. CCBs are used at thelast stage of the read-out structure and mainly perform further managementand reduction of the data to 297 kbps for transmission of data from the ISS tothe ground operation center [16].

Clock board, IDAQ board and CPU system

All the fast operations on data handling and time synchronization are performedthrough two boards, the clock (CLK) board (synchronization) and the IDAQboard (data handling). The CLK board [17] is the time synchronization systemof the apparatus. It generates and distributes the system clock (40 MHz), theGTU clock (400 kHz) and the synchronization signal to all the devices of the FSelectronics. The board generates and receives all the signals needed to controlthe timing of data acquisition. It is connected to the GPS system of the payloadand this allows it to measure the absolute arrival time of the events with aprecision of few microseconds. Another important task of the CLK board is themanagement of trigger signals. The board receives 2nd level triggers from CCBsand forwards to CCBs any triggers coming from CPU. When the board receivesa 2nd level trigger, it goes in a busy state and sends a busy signal to CCB. Inthis way, the CLK Board can perform live-time and dead-time measurements.The board communicates with CPU through the IDAQ board. The IDAQ boardcollects the data from the (18) CCB boards and transmits them to the CPUthrough a PCI bus. Main CPU tasks are:

• power on/off of all subsystems,

• perform periodic calibrations,

• acquire observation data from the FS detector and atmospheric monitor,

• define trigger mode acquisition,

• read Housekeeping (HK) data related to the mission system,

• take care of real time contingency planning,

26

• perform periodic Download/Downlink,

• handle slow control 1553 commands.

Housekeeping module

The purpose of Housekeeping module (HK) is to monitor and to relay controlcommands to the several subsystems that constitute the JEM-EUSO instru-ment. HK tasks include:

• sensor monitoring of different subsystems,

• generation of alarms for the CPU,

• distribution of telecommands to subsystems,

• telemetry acquisition from subsystems,

• monitoring of the status of subsystems.

1.4.4 The Atmospheric monitoring system

The amount of both fluorescence and Cherenkov signals reaching JEM-EUSOdepends on the extinction and scattering properties of the atmosphere. Ex-tinction leads to an overall reduction of the UV light intensity detectable bythe telescope, while scattering properties of the atmosphere at the EAS loca-tion determine the amount of Cherenkov light which is re-directed toward thetelescope. A correct reconstruction of UHECR energy and of the type of theprimary cosmic ray particle requires, therefore, information about absorptionand scattering of the UV light. Also, the presence of clouds and aerosol layerswill alter the physical properties of the atmosphere. Uncertainties of extinctionand scattering coefficients related to the variable meteorological conditions in-troduce distortions of the UV signal from EAS leading to systematic errors inthe determination of the properties of UHECR from the UV light profiles [18],[19]. The Atmospheric Monitoring (AM) system of JEM-EUSO [20] will provideinformation on cloud coverage and the optical properties of cloud/aerosol layersat the time and location of the EAS. The capability to reconstruct the prop-erties of the primary cosmic ray depends on the accurate measurement of theatmospheric conditions in the region of EAS development. The AtmosphericMonitoring (AM) system of JEM-EUSO will host a LIDAR, operating in theUV band, and an Infrared camera to monitor the cloud cover in the JEM-EUSOField of View, in order to be sensitive to clouds with an optical depth τ ≥ 0.15and to measure the cloud top with an altitude resolution of 500 m.

1.5 EUSO-Balloon

EUSO-Balloon is a JEM-EUSO pathfinder experiment. As JEM-EUSO is de-signed to observe UHECR-induced EAS by detecting their ultraviolet lighttracks “from above”, EUSO-Balloon is a nadir-pointing UV telescope as well.With its Fresnel Optics and Photo-Detector Module, the instrument monitorsa surface area of ∼ 50 km2, collecting series of images at a rate of 400000frames/s. On August 25, 2014, EUSO-Balloon was launched from Timmins

27

Stratospheric Balloon Base (Ontario, Canada) by the balloon division of theFrench Space Agency CNES. From a floating altitude of 38 km, the instrumentoperated during the entire astronomical night, observing UV-light from a vari-ety of ground-covers, and artificial EAS, produced by flashers and a laser duringa two-hour helicopter under flight.

Figure 1.6: The launch of EUSO-Balloon.

1.5.1 Objectives

The objectives of the Balloon flight were threefold:

• perform a full end-to-end test of JEM-EUSO prototype consisting of allthe main subsystems of the space experiment;

• measure the effective terrestrial UV background, with a spatial and tem-poral resolution relevant for JEM-EUSO;

• detect tracks of ultraviolet light “from above” for the first time.

1.5.2 The EUSO-Balloon instrument

The general layout of EUSO-Balloon is shown in Fig. 1.7. Its main componentsare the optical bench and the instrument booth. Besides the focal plane detector(PDM), the instrument booth houses the telemetry system (SIREN), CNESspecific instrumentation (ICDV,Hub) and two battery-packs. The design of allcomponents and sub-assemblies is based on similar JEM-EUSO components andsub-assemblies.

The Photo-Detector Module (PDM)

The PDM is composed of 36 MAPMTs, associated front-end electronics, High-Voltage Power supplies and trigger logic. Each MAPMT (Hamamatsu R11265-103-M64) has 8 × 8 pixels of 2.9 × 2.9 mm2. A UV glass filter (SCHOTTBG3 with anti-reflection coating) is bonded to the window of the MAPMTand transmits the UV light in a band between 290 and 430 nm. The full-timeanalog readout is performed by six EC-ASIC boards which detect the individual

28

Figure 1.7: left: EUSO-Ballon, ready for its first flight, Timmins, Ontario (Ca),August 2014, right: schematic view of the instrument booth and optical bench,without floaters and “crash rings”.

photons in each PMT channel, count them per unit of time (GTU), and performthe analog to digital conversion. The High-Voltage Power Supply consists of aminiaturized Cockroft-Walton generator [23]. In order to protect the photo-detectors against highly luminous events (lightnings, etc.), custom made High-Voltage switches are capable of reducing the gain in a few microseconds to givea full dynamic range of 6 orders of magnitude in photon counts. Behind the EC-ASIC stack, the PDM board handles the interfaces between the boards in thePDM and the Data Processor. Its core is an industrial Virtex6XC6VLX240TFPGA which has to absorb a data volume of 1 GBytes per second from the EC-Units and transmitting a part of the data to the Central Cluster Board (CCB)of the Data Processor (DP).

Figure 1.8: The Photo-Detector Module(PDM).

29

The Data Processor (DP)

The sub-assemblies of the DP collect PDM data, process them (trigger, timeand position tagging), handle their onboard storage, and send a subset to thetelemetry system. The DP includes also the housekeeping system. The Con-trol Cluster Board (CCB) including an FPGA Xilinx Virtex-4 FX-60, processes,classifies and performs a second level trigger filtering on the collected data fromthe PDM board. The Clock-Board (CLKB) hosts the interface with the GPSreceiver. It tags the events with their arrival time and payload position. Italso measures the up-time and dead-time of the instrument and provides signalsfor time syncronization of the event. As the on-board triggers were not imple-mented on the 2014 flight, the CLKB also generated a “fake trigger signal ”,enabling the acquisition of the events. The CPU, including a Atom N270 1.6GHz processor, collects data from the CCB and CLKB. It manages the MassMemory and handles the interfaces with the telecommand/telemetry system.The mass storage is composed of two Solid-State Drives (SSD), each one with512 GB capacity operating in fault-tolerant mode RAID 1 discs. One acquiredevent has a size of about 330 kB. Since only a limited data rate can be trans-mitted to the ground through CNES’ new NOSYCA telemetry system, all dataare stored on board [26].

The optics

The optical bench contains two Fresnel lenses with a front surface of 1 × 1 m2

each. The EUSO-Balloon optics is dimensioned to reproduce background ratesper pixel comparable to the one for JEM-EUSO (i.e 1 photoelectrons per pixelper GTU). The two lenses L1 and L3 are aspherical Fresnel Lenses with focallengths of 258.56 cm and 60.02 cm, respectively (focal lengths are referencevalues only, single lenses are not producing stigmatic images). The positionof L1 can be adjusted along the optical z-axis within the optical bench, thefocal distance of PDM is adjusted by a traslation stage in the instrument boot.Together with the 16.7 cm × 16.7 cm2 PDM the optics provides a field of viewof about ±5.5◦ [24] and [25].

The Infrared Camera (IRcam)

In order to monitor the cloud condition (coverage and height), EUSO-Balloon isequipped with an IRcam observing the field of view of the main instrument. Thecamera provides images with a resolution of 640 × 480 pixels, in two wavelengthbands centered in 10.8 µm and 12 µm. The data from IRcam, along withauxiliary data (temperature,pressure and humidity), are stored in a RAID1configuration of 2 SSD of 32 GB of capacity [27], [28].

The gondola

The structure consists of two main modules, the optical bench and the in-strument booth (see Fig 1.7). Inside the self-contained, watertight instrumentbooth, all electronic equipment is mounted on a system of aluminum ”shelves”that provides a thermal link to the 1.2 × 1.2 m2 aluminum backplate (alsocalled radiator). EUSO-Balloon is equipped with an ensemble of aluminum”crash-rings”, designed to absorb the kinetic energy on ground impact through

30

inelastic distortion (Fig 1.7). Due to the coincidence of a science requirementand a safety constraint, EUSO-Balloon has deliberately been designed to protectall sensitive equipment in the event of a water-landing.

1.5.3 The Flight

During more than five hours of operation, EUSO-Balloon recorded and storedon board a total of 258.592 data-packets, corresponding to roughly 33 millionframes (GTUs). A variety of ground covers have been over flown, includingdifferent types of soil and vegetation, wetlands, open water, urban and industrialareas to characterize the background intensity in several conditions. During itsflight EUSO-Balloon crossed areas characterized by scattered and broken cloudsat low altitudes (around 700 - 800 hPa) and ice thick clouds at higher altitudes(around 200 - 300 hPa). A detailed analysis of the atmospheric conditions isreported in [29], [30]. All this variety of situations turned out to be an idealcase to test the conditions that JEM-EUSO is expected to view during its orbitson the ISS. Moreover, to calibrate and reproduce artificially air shower-tracks asystem consisting of a pulsed UV laser and two UV flashers (LED and Xe) flewon a Bell 212 helicopter below EUSO-Balloon. The wavelengths of these sourceswere chosen to mirror the fluorescence emission of electrons in air. The nominallaser energy was equivalent to the light emitted by a 100 EeV EAS [9]. Duringthis time the sources were fired ∼ 150.000 times with two energy settings (see[31], [32] for details). The laser energy was changed every two minutes between15 mJ and 10 mJ and was fired at a rate of 19 Hz. This repetition rate waschosen to guarantee random coincidences between the readout of the balloon(20 Hz) and the laser shots. This was necessary because there was no timesynchronization between the two systems. Around 300 tracks were found inEUSO-Balloon data. The light sequence was synchronized in the following way.First, a UV led signal was shot for 12 GTUs with increasing luminosity toachieve a projected number of photoelectrons at PDM level raising from ∼ 1 to∼ 50 counts. The sequence of led intensities was kept constant during the entireflight. This light-signal appears on the FS as a static source and can be used todetermine the position of the helicopter in the FoV. About 25 GTUs after theend of the led signal, a ∼ 5 mJ laser shot lasting 7 ns was fired. The laser eventtook at maximum 10 GTUs to cross the entire FoV of the telescope. A Xe-flasher lamp was finally fired ∼ 5 µs after the laser shot for a 8 GTU duration.The variable light intensity of the Xe-flashers was reaching a maximum afterthe first 3 GTUs and then decreasing for the remaining time. Four differentabsolute intensities were used to mimic different energies of vertical showers.

1.5.4 Preliminary Results of EUSO-Balloon first flight

Technological Aspects

The main objective of the first flight was to operate a full scale end-to-end test ofall the key technologies and instrumentation of JEM-EUSO detectors. The ac-quisition system performed rather well with an integrated data taking of 15.300s of the total time at float 18.900 s. 258.592 events were recorded, each of onecomposed by 128 frames (Gate Time Units, GTU) of 2.5 µs each, for a total of∼ 83 s of acquisition time, distributed uniformly along the flight, allowing to get

31

a detailed temporal evolution of the light intensity on the various locations. Thefront-end electronics and the multi-anode photo-multipliers (MAPMT) behavedrather well. Only 1 over 9 Elementary Units (EC) had a failure. In total only 5over 36 MAPMTs could not be used for data analysis. The optics performance(global efficiency and point spread function) was calibrated before and afterflight at IRAP Toulouse. The point spread function was verified during flightby using led and flasher images. Details are reported in [33]. The bi-spectralInfrared camera operated as a stand-alone device during the flight to obtain theCloud Top Height (CTH) [34].

UV background measurement

The main scientific objective of the EUSO-Balloon flight is the absolute mea-surement of the UV background intensity. This is relevant for JEM-EUSO asit is one of the key parameters to estimate the exposure curve as a function ofenergy [35]. Due to its refractive system with very fine spatial and temporal res-olutions, EUSO-Balloon was able to determine the space and time variations ofthe UV intensity much better of previous experiments. A detailed description ofthe analysis and results to infer the UV background intensity of EUSO-Balloonis reported in [36]. The relative intensity map of UV background in logarithmicscale is shown in Fig 1.9. The bright areas with high intensities represent arti-ficial light in the city of Timmins and its neighbourhoods, mines, and airport.The dark blue areas indicate the lowest values of UV background.

Figure 1.9: The intensity map of UV background in logarithmic scale(relativeunits). The red and light blue areas are related to cloud coverage (see Fig 1.10).The displayed values are relative to the mean value of UV background intensityover reference area “A” [36].

Analysis of the clear sky region showed that there are no significant variationsin the UV background intensity from different ground surfaces, such as forestand lakes. The intensities of UV emission from pixels with different surfaces arethe same within the measurement uncertainties. In general, there is an anti-correlation between the UV flux from a given direction and the IR radiance fromthe same direction in presence of clouds, where the UV intensity can rise till bya factor of two, while this effect is not present in case of clear-sky conditions.Qualitative explanation for the anti-correlation is that clouds with higher opticaldepth are more efficient in scattering the UV radiation and producing an albedowhich increases the overall intensity of the UV background in the cloudy pixels.UV radiation is absorbed in the atmosphere and higher altitude clouds havehigher albedo (at equal optical depth). Higher clouds are also colder and produce

32

Figure 1.10: The map of IR radiance (relative units). The map is created byaveraged values for particular positions. The values were changing in time dueto movement of clouds and motion of EUSO-Balloon. The displayed values arerelative to the mean value of IR radiance over reference area “A” [36].

lower IR radiance. In general, combination of the measurement of IR emissionand UV albedo of the clouds provides a tool for characterization of the clouds,which should improve the quality of reconstruction of EAS occurring in thecloudy sky. In presence of urban areas the UV light intensity rises even higherthan 10 times compared to green areas. This relative behavior is essentially inagreement with measurements performed by Baby [39].

Helicopter events, IR data and other analyses

The helicopter events revealed to be extremely useful to understand the system’sperformance (optics, photo-detector and front-end electronics) and to test thecapability of EUSO-Balloon to detect and reconstruct EAS-like events. Lasertracks are used to test the reconstruction algorithms [37]. The typical timefit of a laser event and the direction reconstruction are shown in Fig 1.11. Itis important to remember that the read-out period of 2.5 µs is optimized forJEM-EUSO, which is expected to detect EAS at ∼ 400 km distance, insteadof ∼ 35 km as in case of EUSO-Balloon. The fact that EAS-like tracks can bereconstructed also in EUSO-Balloon is quite promising in view of JEM-EUSO.The EUSO-Ballon data have been analysed to test the performance of the FirstLevel Trigger of JEM-EUSO [38]. Around 300 laser tracks were detected. Thesystem showed to be flexible enough to adapt its response to the very variablebackground conditions during the night. The rate of triggers on UV background,clouds, cities, etc. satisfies JEM-EUSO’s requirements. No EAS events wererecognised so far among the triggers. However, this is expected as the dataacquisitions were based on a synchronised clock.

1.6 EUSO-TA

EUSO-TA is a fully functional prototype of the JEM-EUSO telescope. It islocated at Black Rock Mesa, Utah, at the site of one of the fluorescence lightdetectors of the Telescope Array (TA) experiment [41]. From there it observes –simultaneously with TA – artificial light and cosmic ray events, allowing for testsof the technology, calibration of the detector and reduction of the systematicuncertainties of the measurements. EUSO-TA consists of a refractive opticalsystem with two 1 m2 squared Fresnel lenses, focusing the light in a 11◦ ×

33

Figure 1.11: Left side: Time fit of laser event, with constant timing error of 0.5GTUs. Right side: Zenith angle reconstruction of the helicopter laser shots.

11◦ field of view on one Photo-Detector Module (PDM). The electronics cantrigger asynchronously, or using an external trigger. The external trigger canbe provided by the TA Fluorescence Detector (TAFD) allowing for simultaneousobservations of E ≥ 1018 eV cosmic ray showers.

1.6.1 EUSO-TA telescope

The baseline design of the optics is shown in Fig 1.12 with simulated spotdiagrams for incident angles of 0◦, 2◦, 4◦ and 6◦

Figure 1.12: The design of the EUSO-TA optics. The left and right panels showthe ray trace and spot diagrams for different incident angles.

The 17 cm × 17 cm PDM, located in the focal plane, is composed of 36MAPMTs [42] each containing 64 anodes, for a total of 2304 pixels. The PDMcontains 36 front-end ASICs for the readout of the tubes [43], a 1st-level triggerFPGA board, High Voltage (HV) and HV switches. The PDM is controlled bythe Data Processing (DP) unit, consisting of a 2nd-level trigger board [44], CPUboard, Clock board, GPS board, house keeping board and low voltage powersupply [45].

34

1.6.2 Calibration and observation

An on-site calibration can be performed measuring reference stars, which pro-vide a photon flux that is stable with a known UV spectrum at different el-evation angles, which allows correcting for atmospheric attenuation. In thisconfiguration it is feasible to compare the night sky background measurementsperformed with EUSO-TA to the corresponding flux values obtained from theanalysis of TAFD data. The absolute night sky background flux observed bythe two instruments looking toward the same sky direction (normalized to equalsolid angle) provides a valuable tool for cross-calibration. Indeed, an overlap oflight curves from the diffuse component measured by two independent telescopesdemonstrates the accuracy of the pixel-to-pixel variation correction. EUSO-TAacquires data while running simultaneously with TA but without causing anyinterference with TAFD operations. This offers the possibility to monitor thegain of individual camera pixels in realistic working condition, in which back-ground flux can induce gain shifts. This monitoring can take place during thewhole acquisition period, thus allowing detection of any eventual gain drift aswell as verification of the linearity of pixel response in the range of variation ofthe night sky background.

1.6.3 Laser

To study the EUSO-TA response to a known light source the light coming fromthe TA Central Laser Facility (CLF), distant from EUSO-TA by about 21 kmhas been used. The CLF shoots vertically laser pulses of 355 nm in front of thedetectors [47]. During standard observation nights the CLF shoots every halfan hour for 30 s with 10 Hz shooting frequency. The scattered light of the ∼3mJ beam was clearly visible traversing through the EUSO-TA field of view (seeFig 1.13, left). The registered light intensity is dependent on the atmosphericconditions – obstruction of the light path due to clouds – compatible with thefluctuations of the emitted light, showing the good reconstruction capabilitiesof EUSO-TA. In addition to the CLF have been also performed measurementsof a mobile UV laser of Colorado School of Mines. The laser could be shotwith energies in the range of about 1-86 mJ, with pointing adjustable in twodirections. The mechanics featured automatic changing of the pointing, allowingfor easy ”swipes” through the field of view (see Fig 1.13, right). EUSO-TA wasable to detect a few shots of 1 mJ energy shot vertically from 34 km distance.This corresponds to a UHECR with E = 1019 eV , suggesting that the sensitivityis better than expected. The laser was shot at growing distances (34, 40, 60, 100km) with preliminary analysis showing detection for a 85mJ shot at distance of100 km (see Fig1.13, left). This is very promising for space detection of UEHCR(ISS is at 400 km), considering the non-optimal atmospheric conditions, the highoptical thickness (since the observation was almost horizontal) and the smaller(by a factor 4) lens area. Furthermore, the EUSO trigger has been successfullytested off-line using the same algorithm to be implemented in space on the rawdata files.

35

Figure 1.13: Left: an average of ∼ 250 shots of CLF laser; right: an average of∼ 150 inclined shots of the Colorado School of Mines laser, located at 40 kmfrom EUSO-TA (62 mJ), the missing part due to a non-functioning MAPMTin the center of the focal surface. The color scale on both pictures denotes theuncalibrated detector counts.

1.6.4 Cosmic Ray events

The still ongoing analysis has detected so far few events, seen in coincidence withTAFD, one of the is discussed in the following. The proximity of the event –around 2.5 km – makes it appear as a E ∼ 1018 eV track going through the wholeFOV on a single GTU (one frame of 2.5 µs), as shown in Fig 1.14. Therefore,its parameters have to be derived from TAFD which, thanks to larger FOV andhigher time-resolution, could see the shower movement. In future is expectedto see also more energetic events further away from the detector, moving onseveral consecutive GTUs through the field of view. However, such events aremore sporadic, requiring significantly longer observation time.

Figure 1.14: An UHECR event of ∼ 1018 eV energy traversing at ∼ 2.5 kmdistance from EUSO-TA and therefore visible as a track on a single GTU, inthis case going from top-right to bottom-left (distance and energy are estimatedby TA). The color scale denotes uncalibrated counts of the detector. The bottomright PMTs and the center bottom right PMT were not working at the time ofthe measurement.

36

Chapter 2

Mini-EUSO and EUSO-SPB

Mini-EUSO, aiming at the observation and measurement of the UV night emis-sion from the Earth, is developed as a next step pathfinder and a precursor ofthe JEM-EUSO mission. On the other hand the EUSO super pressure balloon(EUSO-SPB) experiment will be the first pathfinder of JEM-EUSO to measureEASs produced by EECRs from the top of the atmosphere using the fluorescencesignal.

2.1 Mini-EUSO

The measurement and characterization of the UV background produced in theatmosphere of the Earth is a key element for any experiment aiming at the ob-servation of Ultra High Energy Cosmic Rays (UHECR) from space. Such mea-surements, achievable with small-size, cost-effective devices, besides the intrinsicscientific relevance, can provide useful indications on the design, optimizationand performance of full-scale instruments to be placed on board orbiting spacestations or free-flier satellites.

2.1.1 Scientific objectives

The scientific objectives of Mini-EUSO range from cosmic ray to atmosphericscience.

Main objective

Balloon experiments, such as NIGHTGLOW [48], [49] and EUSO- Balloon [50]provide relevant information on the UV background reflected from ground, butnot a direct detection of the airglow emission which is located around 100 kmaltitude. Mini-EUSO, with its spatial resolution of ∼ 5 km and a time frameof 2.5 µs will be able to characterize in a much detailed way the intensity andvariation of the UV radiance, airglow included, as a function of time and positionof the ISS. These data will be of very much relevance in the proper estimation ofthe exposure curve of space-based experiments such as JEM-EUSO [15]. Despiteits very high energy threshold for cosmic ray detection (Ethr ∼ 5 × 1020 eV,with its annual exposure of ∼ 15000 km2 · yr · sr), Mini-EUSO will provide a

37

significant contribution in estimating an absolute limit on the cosmic ray fluxabove those energies for a null detection.

Additional objectives

Mini-EUSO has the potential to become one of the first operational space-basedplatforms for meteor observation. In comparison to the observation of extremelyenergetic cosmic ray events, meteor phenomena are very slow, since their typicalspeeds are of the order of a few tens of km/sec. The observing strategy developedto detect meteors may also be applied to the detection of nuclearites [51], whichhave higher velocities, a wider range of possible trajectories, but move wellbelow the speed of light and can therefore be considered as slow events aswell. Furthermore Transient Luminous Events (TLE), such as red sprites, elvesand blue jets may affect UHECR measurements and must be carefully studiedbefore the main mission Another relevant issue would be the observation ofspace debris. Since their orbital velocities are very high, collisions can involverelative impact velocities of the order of 10 km/s, even the fragments withgreater than MJ kinetic energies may cause a severe or catastrophic damage onfunctioning satellites such as the ISS. As described in [52] Mini-EUSO could beused as a prototype system for tracking such space debris. A super-wide field-of-view telescope (such as JEM-EUSO) and a novel high efficiency fibre-basedlaser system (CAN) could constitute a very useful orbiting debris remediationsystem. Other scientific objectives of Mini-EUSO include the bio-luminescenceobservation above the sea from space and atmospheric science by coupling UVmeasurement together with simultaneously taken IR and VIS images.

2.1.2 Techonological objectives

Mini-EUSO addresses also some important issues by the technological point ofview that are summarized as follows:

• First use of Fresnel lenses in Space (meant at an altitude higher than thetop of atmosphere ∼ 40 km).

• Optimization and validation of JEM-EUSO observational scheme.

• Increase of the Technological Readiness Level (TRL) of JEM-EUSO in-strumentation, a typical parameter in the development of devices to bequalified and certified for space.

• Test and R&D of advanced solutions for future space missions, such asstudies on development of SiPM (Silicon Photomultiplier) based photo-sensors for space applications [53].

2.1.3 The Instrument

The Mini-EUSO instrument, whose scheme is shown in 2.1, is composed of onesingle element of the basic JEM-EUSO detection unit, the Photo Detector Mod-ule (PDM), consisting of 36 Hamamatsu MAPMTs M64, 64 pixels each, for atotal of 2304 pixels. A system of two Fresnel lenses, 25 cm diameter each, formsthe optical system. The front-end electronics is based, opportunely rescaled, onthe same ASIC and FPGA boards being developed for JEM-EUSO; the Data

38

Acquisition System and Processing unit is a compact, low-power consumptionsystem . Infrared (IR) and Visual (VIS) cameras are foreseen to be installedas ancillary instruments employed to complement the UV observations of theMini-EUSO focal surface. The compact, Short Wavelength Infrared Camera(SWIR Cam), working as a standalone system, will be attached to the instru-ment with the main task of performing atmospheric monitoring measurements(as the thermodynamic phase detection of clouds). Data from the cameras willhelp in the measurements of the emissions of the Earth and the study of transientphenomena. The Mini-EUSO instrument will be closed in a specific container,a mechanical box providing also all the needed interfaces to the transparent,nadir looking UV window of the Russian module Zvezda of the ISS. The mainparameters of the instrument, in its flight configuration are listed in table 2.1.

Figure 2.1: left: Mini-EUSO scheme, right a photo of Zvezda module wheremini-EUSO will be hosted.

Parameter ValueMass,kg 30Size,mm 350×350×600Power,W 30Voltage,V 28

Wavelength,nm 300-400

Table 2.1: Mini-EUSO instrument main parameters

Optical System

The optical system designed for Mini-EUSO consists of two double sided flatFresnel lenses and the focal surface. The diameter of both Fresnel lenses is 250mm. The Mini- EUSO optics has a low focal number F# 1.04, and the effectivefocal length is 260 mm. Its Field of View (FoV) is ±19◦ at r = 900 mm onthe focal surface. The thickness of the lenses, 5 mm, reduces the mass of theoptics resulting in a light system (∼ 0.3 kg/lens). The main characteristics of

39

the Fresnel lenses are shown in table 2.2, and the point spread function of theoptical system is shown in Fig. 2.2. The photon collection efficiency (PCE),calculated from a raytrace simulation of the active optical system by a codespecifically developed, is ∼ 60% which takes into account several loss factorsas surface reflection, material absorption, surface roughness, Fresnel facet backcut and support structure obscuration.

Figure 2.2: Point spread function of the optical system for light coming parallelwith different inclination (from left to right, 0◦, 5 ◦, 10◦, 15◦ and 19◦). Thediameter of the inner white circle is 2.5 mm and the outer circle is 5 mm.

Lens surface Groove pitch Number of grooves

Front lensFront 1.91∼21.4 mm 33Back 11.75∼56.59 mm 5

Rear lensFront 0.01∼32.04 mm 11Back 1.63∼21.37 mm 36

Table 2.2: Mini-EUSO instrument main parameters

Data Processor and Data Acquisition System

The Data Processor (DP) is the sub-system of the Mini-EUSO apparatus whichperforms the control of the instrument and the data management and storage.The main functional requirements of the DP are:

• configure the front-end electronics;

• manage the mass memory for data storage (removable disks);

• measure live and dead time of the instrument;

• provide signals for time synchronization of individual events;

• perform housekeeping monitoring;

• control and steer the HVPS (High Voltage Power Supply) system;

• commute from ”day” to ”night” operating mode and vice versa;

• acquire data from IR and VIS camera;

• provide access to a small part of data which can be transmitted to ground.

The required DP functionality, can be obtained by connecting several sub-systems performing the various tasks:

• CPU,

40

Figure 2.3: The simplified block diagram of the Mini-EUSO Data Processor.

Figure 2.4: Mini-EUSO Data Handling System.

• Data storage (DST),

• Clock Cluster Control Board (CLK CCBB),

• Housekeeping system (HK),

• Data Processor Power Supply (DP-LVPS).

A simplified block diagram of the DP is shown in Fig. 2.3. The requirementson the mass, the power as well as the volume for the DP are 5 kg, 15W and(15x15x15) cm3 respectively, For the Mini-EUSO DP the PCIe/104 embeddedcomputer standard has been selected. The standard defines both form factorsand computer bus, is modular, and allows to stack together boards to producea customized embedded system. The form factor is somewhat smaller than adesktop PC. The DP is then a stack of six PCIe/104 boards, as shown in 2.4

2.2 EUSO-SPB

The EUSO super pressure balloon (EUSO-SPB) experiment [54] will be the firstto measure EASs produced by EECRs from above. The proposed launch siteis Wanaka NZ. For reference, the path of NASA’s first super pressure balloonflight that was launched from this site (March 2015) is shown in Fig. 2.5.Flying the EUSO-SPB instrument on an ultra-long duration balloon flight ofat least 20 nights centered on the new moon will provide the exposure thatis required to achieve this important milestone. This mission will be the firstextended duration operation of a cosmic ray UV fluorescence detector at nearspace altitude and provide an important end-to-end test of JEM-EUSO. TheEUSO-SPB mission will also afford the possibility to detect other faint andfast UV light flashes that could be a background for EAS observations or be ofinterest in their own right with some discovery potential.

41

Figure 2.5: The path of the 2015 NASA Super Pressure Balloon engineeringflight launched from Wanaka NZ. The flight duration was 32 days.

2.2.1 The EUSO-SPB Instrument

The EUSO-SPB instrument is an updated version of the first EUSO-Ballooninstrument that was flown successfully from the Timmins Stratospheric Bal-loon Facility on August 24/25 2014 (see Chapter 1 for the description of theEUSO-Balloon instrument). Briefly, the instrument is a 1m UV telescope withrefractive optics and a 12◦ × 12◦ FoV. The focal surface is instrumented witha photon detector module (PDM) that uses the design and components thatare planned for the 137 PDM focal surface of JEM-EUSO. The PDM is coveredwith a UV filter assembly that passes light between 290 nm and 430 nm to spanthe UV fluorescence spectrum of EASs. The PDM is divided into 2304 pixels,each with a FoV of 0.25◦ × 0.25◦. The 2304 pixel outputs are digitized in 2.5µs bins. The system is operated in single photoelectron (PE) counting mode.An event contains the digitized PE counts of all pixels for 128 consecutive timebins. This corresponds to a readout window of 320 µs duration. The EUSO-SPBinstrument will included the upgrades listed below.

Item DescriptionMAPMTs Higher quantum efficiency multi-anode photomultiplier tubes

(Hamamatsu model R11265-113-M65 MOD2)Solar Power Solar Panels and updated power system. Panels will be mounted

on the 4 sides of the gondola for redundancy.Optics Include a third fresnel lens that was in the original design to

provide chromatic correction over the EAS UV spectrum.HV Updated HV distribution systemTrigger Upgraded trigger for EASsControl Update system control programming for a 50 day missionTelemetry Interface to the Columbia Scientific Balloon Facility’s Support

Instrument Package (SIP)

Table 2.3: EUSO-SPB instrument’s upgrades

2.2.2 Estimating the EAS event rate

The work developed in this thesis is mainly aimed to evaluate the trigger perfor-mances of EUSO-SPB estimating the EAS event rate. An event reconstructionanalysis is also performed. Several Monte Carlo simulation analysis have been

42

done before this work in order to extimate the number and the energy rangeof EASs that EUSO-SPB will record. The current work aims to give a moredetailed overview taking into account several aspects never treated in the previ-ous analysis. Details are presented in Chapters 3, 4, 5. A first study presentedin [55], using the JEM-EUSO Off line software framework [56], is carried outon EASs for which a minimum number of UV scintillation photons reached theoptical aperture (reached lens within the FoV) of EUSO-SPB. Sets of 60000proton induced EASs were simulated between 0◦ and 80◦ in zenith angle. TheEASs were randomized in azimuth angle and core location over a 150 km radiuscircle (Fig. 2.6). Eight simulated data sets were generated and randomizedin this manner corresponding to 8 energy steps from 1017.75 eV to 1019.25 eV.For each EAS in each set, the UV optical emission in the atmosphere was thensimulated. The EASs for which a minimum number of UV

Figure 2.6: Left: Sample distribution of simulated EAS directions projectedon to a unit hemisphere. Center: EAS core location projected to the ground.Black points indicate all the events simulated. Red points indicated events forwhich some light entered the EUSO-SPB optical aperture. Right: Similar tothe adjacent cent plot with the additional requirement that at least 300 photonsreach EUSO-SPB on at least each of 5 consecutive 2.5 µs time bins. The energiesof these simulated events is 1018.75 eV.

photons reached the optical aperture (reached lens within the FoV) of EUSO-SPB were identified. A further selection required a minimum number of photonsin each of 5 consecutive time bins. An example event track and light curve aredisplayed in Fig. 2.7 with selection cut minimum of 300 photons. For an overallefficiency factor of 0.40 (optics transmission) × 0.25 (quantum efficiency) × 0.25(fraction of point spread function on one pixel), this corresponds to at least 7.5PEs per time bin on the brightest pixels. This level is expected to be well abovethe UV background rates that are expected to be about 1 PE per pixel per timebin. The EAS detection rate was calculated as N =

∫J(E)·Asim·f(E)dE where

N is the number of EASs detected per unit time, J(E) is the measured cosmicray flux as a function of energy E as reported in [57], Asim is the simulatedaperture of the 150 km radius circle over the 0◦ to 80◦ zenith angle range (A=3.7 × 105 km2 sr), and f(E) is fraction,for a particular energy, of the simulatedEASs that would be observed by EUSO-SPB. The simulated f(E) data and theproduct J(E)·Asim·f(E) are plotted in Fig. 2.8 for various photon minima. Theestimated aperture, Asim·f(E=1019.25 eV) approaches 1000 km2sr for EASs thatexpected to be clearly observable. The minimum energy is estimated to be about

43

1018.25 eV. The detection rates for three different photon thresholds are listedin table 2.4. The nominal rate is about 0.2 per hour. This estimate is consistentwith the results of a similar study [58] performed with the ESAF [59] simulationframework. where the number of obvious EAS events expected is about 10-15for a 20 night mission that spans a dark period (∼ 1.2 events per 8-hour night).

Figure 2.7: An example simulated event (E=1018.75 eV) passing the selectioncriteria for EUSO-SPB (see text). Left: The projected track Right: The EASlight curve.

Figure 2.8: Left: The fraction of observable EASs relative the the numbersimulated for different photon thresholds (see text). Right: The flux of EASsobservable by EUSO-SPB. The integral of these curves yields the expected eventrate.

2.2.3 EUSO-SPB expected mission’s conditions

The proposed launch date for EUSO-SPB is March 2017. The 2017 March/Aprildark period at the latitude of Wanaka, NZ includes 18 nights with at least 2.5hours of moon down dark time for a total of 118 hours (table ). The averagemoon down dark time per night is 6.5 hours. The start times and end timesand, to a lesser extent, the total amount of darkness will vary depending on thetrajectory of the balloon. In practice, the dark times will be reduced slightlybecause a net eastward drift of the balloon will lead to slightly earlier moonriseand sunrise times. This reduction may be offset by collecting additional data

44

PhotonThreshold

Events/hour Events/darkperiod

200 0.42 50300 0.18 21400 0.09 11

Table 2.4: Expected Cosmic Ray Air Shower detection rates. Photon thresholdis the number of photons reaching the instrument optical entrance aperture ineach of 5 consecutive 2.5µs timebins. The dark period assumed is listed in table.

during low moon periods. The expected data telemetry transfer rate from theballoon to the ground is 180 kbits/s using two Tracking and Data Relay SatelliteSystem (TDRSS) data links. On nights with the longest dark period (9 h), thedata collected at an average event rate of 0.25 Hz could be downloaded over a24 hour period. A separate Irridium satellite link will provide a low rate channelfor sending simple commands and queries to the instrument control computer.At the start of the mission a UV LED and a pulsed 355 nm laser will be flownin an aircraft under the balloon for several hours to provide a set of calibrationflashes and tracks for the instrument. The aircraft will fly below 10000 ft sothat the UV laser beam can be fired out an open window. The intention is toperform these tests while there is still line of site communications to the balloonso that data can be downloaded at a reasonably high rate.

date UTC hrs dark start UTc stop UTC2017 Mar 19 2:08 8:38 (twilight) 10:46 (moon 62%)2017 Mar 20 2:52* 8:36 (twilight) 11:28 (moon 52%)2017 Mar 21 3:40* 8:34 (twilight) 12:15 (moon 42%)2017 Mar 22 4:34* 8:32 (twilight) 13:07 (moon 32%)2017 Mar 23 5:34* 8:30 (twilight) 14:04 (moon 23%)2017 Mar 24 6:38* 8:28 (twilight) 15:06 (moon 14%)2017 Mar 25 7:46* 8:26 (twilight) 16:12 (moon 7%)2017 Mar 26 8:51* 8:24 (twilight) 17:15 (twilight)2017 Mar 27 8:54* 8:22 (twilight) 17:16 (twilight)2017 Mar 28 8:57* 8:20 (twilight) 17:18 (twilight)2017 Mar 29 9:01* 8:18 (twilight) 17:19 (twilight)2017 Mar 30 9:04* 8:26 (twilight) 17:20 (twilight)2017 Mar 31 8:40* 8:41 (moon 14%) 17:22 (twilight)2017 Apr 1 7:57* 9:25 (moon 24%) 17:23 (twilight)2017 Apr 2 7:08* 10:16 (moon 35%) 17:24 (twilight)2017 Apr 3 6:12* 11:13 (moon 46%) 17:25 (twilight)2017 Apr 4 5:12* 12:14 (moon 58%) 17:27 (twilight)2017 Apr 5 4:09* 13:19 (moon 69%) 17:28 (twilight)2017 Apr 6 3:04* 114:24 (moon 79%) 17:29 (twilight)2017 Apr 7 2:00* 15:30 (moon 87%) 17:31 (twilight)

Table 2.5: Example of expected hours of operation for a dark period at 45◦ S(see text). This dark period includes 18 days, indicated by an asterix, with atleast 2.5 hours of moon down darkness for a total of 118 hours. These hourswould be the primary targeted observation times for EUSO-SPB.

45

Chapter 3

A simulation study of thetrigger performances ofEUSO-SPB

In this chapter a study of the trigger performances of EUSO-SPB based onMonte Carlo simulations is presented. The ESAF software is introduced, fol-lowed by a description of the different trigger logics implemented in such simu-lation framework and tested during the analysis. Aim of this work is to estimatethe response of the 1st level trigger of EUSO-SPB comparing spectra obtainedvarying different simulation parameters (atmospheric cloud coverage, detectoroverall efficiency, average UV background rate, GTU duration). The total num-ber of expected triggered events in 1 month of mission duration, taking intoconsideration the detector response in different cloudy situation, is estimated.

3.1 ESAF

The ESAF package is a simulation software specifically designed for the perfor-mance assessment of space based cosmic ray observatories. This software hasbeen written mainly in C++ and makes use of the ROOT routines of CERN [60].The entire software has been developed following an object oriented approachand is structured in a modular way. The compilation of the ESAF software pro-duces two distinct executable files called respectively Simu and Reco. The firstone performs the simulation of the entire physical process from shower to teleme-try. In this context, several air shower generators like SLAST [61], CONEX [62][63], CORSIKA [64] and others are available for use. An atmospheric model-ing according to the 1976 Standard US Atmosphere [65] is implemented as wellas different parameterizations for Fluorescence and Cherenkov yield. Both theNagano [66] and Kakimoto [67] studies have been implemented in the software.The standard Cherenkov theory is used in the ESAF modeling. The Rayleighscattering and ozone absorption processes are simulated in ESAF by means ofthe LOWTRAN 7 [68] atmosphere software. Several different versions of theoptics Monte Carlo simulator, developed at RIKEN, (RIKEN ray trace code)have been interfaced with ESAF. Beside that, a GEANT 4 optics interface [69]

46

and a parametric optics simulator are implemented. Both the PMT and theEC electric signal treatment is performed in a parametric way. The last partof the simulation chain consists of the trigger sequence. A multiple stage trig-ger scheme has been therefore developed in order to maximize the ratio of realevents to background [70]. Once the trigger sequence has been applied theSimu executable stops producing an output .root file. The simulation code isstructured in several independent modules the higher of which is the so calledSimuApplication. An instance of such a class is created in the simu main.cc filewhere the method SimuApplication::DoAll() is called. Such a method performsthe iterative call of the SimuApplication::DoEvent() method which takes careof the entire physical process on a single event basis. Such a method will cre-ate an instance of the LightToEuso class which executes the entire process fromprimary particle to photons on pupil. Several choices are available on which sim-ulator is to be used but the default option is the so called StandardLightToEusoclass. By calling the StandardLightToEuso::Get(), the virtual Get() methodsof the shower generator, of the light production and transport will be called.Each one of the mentioned Get() methods will deliver output objects describingthe shower profile, photons in atmosphere and photons on pupil. The choiceof the object oriented approach shows its power here where the call of severalpolymorphic Get() methods allows great flexibility. Always inside the SimuAp-plication::DoEvent() method the virtual Detector::Get() method will be called.This method takes care of the entire detector simulation. Several choices areopen at this stage between various detector configurations. The most importantof them can be considered to be the EusoDetector, (activating the RIKEN raytrace code), the G4Detector (activating the Geant 4 optics) and other testingor debugging detector simulators. Calling one of the above described methodswill activate both optics and electronics simulators. As final output of the entireprocedure a Telemetry object is produced.

The reconstruction procedure is activated in the reco main.cc file. Here aninstance of the RecoFramework class is created and the method RecoFrame-work::Execute() is called. While in the constructor function RecoFramework::RecoFramework() the module chain is built, the RecoFramework::Execute() methodperforms the entire sequence of calls to reconstruct the event. In fact, themodule sequence is first initialized through an iterative call of the Module-Factory::MakeModule() method which allocates all the RecoModule objects re-quested by parameter files. Moreover, a vector named fModules with all thepointers to the created RecoModule objects is created. In the RecoFrame-work::Execute() method all the modules (which are inheriting from RecoMod-ule) are initialized, called and cleared. Eventually all the output data are savedin the root file. For performing all the mentioned operations, the polymorphicmethods RecoModule::PreProcess(), RecoModule::Process(),RecoModule::PostProcess()and RecoModule::SaveRootData() are declared in each module. Each of suchmodules has a specific function which can be either pattern recognition, direc-tion fitting, profile reconstruction or Xmax and energy reconstruction. Severalmodules have been implemented in the course of the years but the most actualand currently updated are the LTTPreClustering and PWISE for the patternrecognition, the TrackDirection2 for the direction reconstruction and the Pmt-ToShowerReco for the energy reconstruction. A schematic view of the abovementioned structure is shown in Fig.3.1 and Fig.3.2.

Another extremely important issue is the availability of the output informa-

47

Figure 3.1: a schematic view of the ESAF Simu application structure. The mainapplication is the so called SimuApplication. The LightToEuso application takescare of all the physical process from shower to detector. The EusoDetector ap-plication performs the simulation of optics and electronics. Black lines representthe calling hierarchy. Red arrows represent the flux of simulated input–outputnecessary for all the called applications.

tion. The Simu executable produces a .root output file which is available fordata analysis. Every user can develop specific root macros to open and read suchfiles. Each simulation root file contains a root TTree object called etree. Such anobject can be retrieved after opening the root file, by means of the TFile::Get()method. Thanks to the ESAF EEvent class, the retrieval of all the TBranchesobjects of the tree is possible. By calling the method EEvent::EEvent() thememory segment for the retrieval of all the branches is allocated. By calling theEEvent::SetBranches() the addresses for all the data objects is set. With theTTree::GetEntry() method the access of the single events is possible. This isby itself a very standard procedure to access root data. Thanks to the retrievalmethods of the EEvent class it is possible to access all the pointers to the avail-able branches and consequently to the saved data. A small resume of the codelines necessary access the true primary energy is given here:

TFile ∗ fsimu = new TFile(”rootFile.root”);TTree ∗ etree = (TTree∗)fsimu → Get(”etre”);EEvent ∗ evs = new EEvent();evs → SetBranches(etree);etree → GetEntry(eventNumber);evsreco → GetTruth().GetEnergy();

The same concept is valid for the Reco output file. In this case the output filewould end with the .reco.root extension. Also here, after opening the file, it willbe necessary to retrieve the TTree object called recotree always by means of theTFile::Get() method. Analogously to what has been done for the simulation

48

Figure 3.2: a sketch of the reconstruction framework. The main applicationRecoFramework calls iteratively the MakeModule method which allocates allthe required modules. A vector of pointers to the allocated objects is savedunder the name fModules. In the Execute method the operations of all themodules are performed. All the modules are inheriting from the RecoModuleclass. The virtual methods PreProcess, Process, PostProcess and SaveRootDataare called for all the allocated modules. Note that blue boxes represent classes,blue–gray boxes methods, the gray box is a C++ vector and the circular arrowindicates iterative repetition of some method or sequence of methods.

49

it will be necessary to create a RecoRootEvent object and set the addressesof all the branches with the method TTree::SetBranchAddress(). The methodTTree::GetEntry() will provide direct access to the single event data. Thanksto the RecoRootEvent methods then, the pointers to all the branches are madeavailable. A small resume of the code lines necessary access the reconstructedenergy is given here:

TFile ∗ freco = new TFile(”rootFile.reco.root”);TTree ∗ recotree = (TTree∗)f reco → Get(”recotree”);RecoRootEvent ∗ evsreco = new RecoRootEvent();recotree → SetBranchAddress(”events”, &evsreco);recotree → GetEntry(eventN umber);evsreco → GetRecoPmtToShower().GetEnergy();

Of absolute importance is also the presence of all the available parameters underthe config directory. Such parameters can be passed as command line parame-ters (by using the syntax - - Class.P arameter = value) or grouped in a singleuser–defined config file. In such a case the file should be identified by meansof the - - usrcf g = fileName command line expression. Generally it must beconsidered that the priority order would be as follows: command line parame-ters first, then user file parameters and finally config directory parameters. Ismandatory to mention the fact that ESAF consists also of a collection of macrosto perform several auxiliary operations. Between them the opticsresponse.C, theMakePixelMapPhotonFile.C and the EEventViewer.C can be mentioned. Aimof these macros is respectively, the production of the OpticsResponse map, ofthe so called PixelAngleMap and the management of a graphical interface for theevent visualization. Several versions of ESAF adapted to the specific needs ofeach JEM-EUSO pathfinder has been implemented. For the current simulationstudy the EUSO-SPB version of ESAF has been used.

3.2 Triggered spectrum calculation

The number of triggered events collected in a certain energy bin of the spectrumfor a given duration of the mission was calculated by means the convolutionbetween the detector exposition Ψ(E) and the differential flux calculated with

a curve fitting the measured differential cosmic rays flux dΦ(E)dE reported by the

Pierre Auger Collaboration in [73] (see Fig.3.3):

∆N(E) = Ψ(E) · dΦ(E)

dE·∆E (3.1)

where the exposition Ψ(E) is defined as:

Ψ(E) = Ageo(E) · t ·DC (3.2)

with:

• Ageo: the geometrical aperture of the detector as function of the energy,

• t: the mission duration,

• DC: the observational duty cycle as the fraction of time in which the skyis dark enough to measure EAS.

50

log10(E/eV)17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2 19.4

]-1

eV

-1sr

-1 s-2

log10(F

lux)

[m

34−

33−

32−

31−

30−

29−

Auger Flux (ICRC 2013)

Figure 3.3: Logarithmic plot of the measured cosmic ray flux (black markers) re-ported by the Pierre Auger Collaboration [73] fitted with a 2-degrees polynomial(red curve).

The above mentioned Ageo is:

Ageo =

∫S

dS

∫ 2π

0

dφ

∫ π2

0

dθ cos θ · sin θ · εdetector(E,Bkg, S) (3.3)

whereas:

• θ and φ: the arrival direction angles,

• S: a test area sufficiently larger than the visible field of view surface,

• εdetector(E,Bkg, S): the average efficiency of the detector.

The uncertainty δN(E) on the number of detected events for a certain energywas calculated assuming a binomial error on the probability to have a trigger:

δN(E) ∝

√P (E) · (1− P (E))

N simu(E)(3.4)

with the probability P(E) extimated as the trigger efficiency for that partic-ular energy value:

P (E) ≡ εdetector(E,Bkg, S) =N trigg(E)

N simu(E)(3.5)

whereas:

• Nsimu(E,S): the number of events generated in a simulation as a functionof the energy, injecting area. Event directions are sampled on the full solidangle.

• Ntrigg(E,S,Bkg): the number of events triggered by the trigger algorithm,as a function of the energy, injecting area and considered background.

Energy was simulated sampling during the simulation run time a differentialspectrum ∝ E−1, which means generating a constant number of events consid-ering logarithmic scale energy bins. The procedure to generate other shower

51

parameters such as ground impinging point coordinates, zenith angle and az-imuthal required more details and will be discussed in the following. The erroron the total number of triggered events composing the spectrum was obtainedpropagating the uncertainties on the number of events for every energy bin:

δN(E)tot2

=∑i

δN(E)i2

(3.6)

The basic simulations parameters used to calculate spectra were:

• energy range: from 2 · 1017 eV to 1 · 1019 eV;

• mission duration: 2 months

• Duty Cycle = 0.13, which is the value estimated for JEM-EUSO.

It must be noticed that calculated spectra must be considered relative.Aim of this part of the analysis was indeed to provide relative comparisonsamong spectra obtained changing different simulation parameters such asdetector efficiency, background level and cloudy coverage. To do that, aprecise estimation of the DC for EUSO-SPB was therefore not necessary.The absolute number of expected triggered events, given a certain amountof detector operational time, will be estimated in the final part of this work.

3.3 Preselection of showers initial parameters

An extremely important macro called ParametersGeneretor.C was implementedin order to generate externally to the ESAF framework shower parameterssuch as ground impinging point coordinates, zenith and azimuthal angles. Thepreselection of such as parameters (unlike the energy) was a mandatory approachif one considers the geometry of the simulation. Indeed, to avoid bias effectsconcerning the averaged trigger efficiency estimation (see formula 3.5) the EASshowers randomization needed to be performed over a circular area which wassupposed to be much wider compared to the detector field of view (∼ 64 km2).As a consequence a huge fraction (> 90%) of events does not have any chanceto enter the field of view. For this reason most of the ESAF iterations arewasted for such non relevant events. The adopted strategy to solve this issueconsisted to generate with the so-called ParametersGenerator.C for each eventthe shower parameters out of ESAF. More in detail, the code implemented inthis macro, given a certain amount of simulated events (Ntot) and a certaininjecting area, samples randomly shower directions covering the full solid angle.A screening is carried out on all those events crossing an hypothetical cylinderwhose dimensions are compatible with the detector field of view. In this waythe number of events very far from the field of view is considerably reduced.The total amount of such selected events is Ncyl. The generated sample ofNcyl events is saved in a .txt file and passed to ESAF to be processed. Duringthe simulation each event is assigned an energy as mentioned in the previousparagraph.

This procedure allows a consistent reduction of simulation time since onlyshowers relatively closed to the detector field of view are processed. On theother hand, the preselection of initial parameters has consequences concerning

52

8 km

13 km 50 km

Figure 3.4: Representation of the cylinder used as a constraint to generate theinitial showers parameters. The radius of the cylinder is set 13 km meaning anarea comparable with the balloon field of view. An injecting area radius of 50km is assumed. Orange arrows represent shower passing trough the cylinderwhose parameters will be processed by ESAF during the simulation, while redarrows represent shower whose parameters are rejected.

53

the definition of the efficiency of the detector. The averaged trigger efficiencyreported in formula 3.5 must be redefined as :

εEUSO-SPB(E,Bkg, S) =N trigg(E) · kN simu,cyl(E)

(3.7)

with:

k =N cyl

N tot(3.8)

whereas:

• Ncyl: the number of simulated events passing trough the cylinder;

• Ntot: the number of simulated events, given a particular injection area,sampling the entire solid angle;

In order to obtain Ncyl events with the standard ESAF procedure we wouldhave needed Ntot iterations. Instead we only simulate events Ncyl saving afactor ∼ 1-k time. It must be noted that the preselection of initial parametersof showers, increases the probability of a single events to generate a detectablesignal, but does not guarantee the considered event meets the requirements tobe triggered. Considering an injecting area of 2500 km2, k is ∼ 0.086 whichmeans a gain of ∼ 90 % simulated time. A huge reduction of simulation timehas been a necessary step, without which such simulation study could not havebeen performed.

3.4 The EUSO-SPB persistence tracking trigger

The trigger rate for the first level trigger 1, named Persistence Tracking Trigger(PTT), is required to be in the current configuration of ∼ 0.2 Hz/EC. Since toa triggered event corresponds a dead time of the detector of 50 µs, this meanstotal amount of ∼ 10% of dead time. Such a rate needs to go ∼ 0.02 Hz/ECto satisfy the telemetry budget available for EUSO-SBP. This reduction will becarried out by the CPU analyzing the data stored on the on board memory.The work developed in this thesis focused on the PTT. The PTT is performedby the FPGA and operates at EC level, which is the basic unit of the front-end electronics. The main task of 1st trigger level is to reject most of eventsdue to background fluctuations by requiring a locally persistent signal abovethe average background lasting a few GTUs, ensuring a trigger rate compatiblewith the value of ∼ 0.2 Hz/EC. In this trigger level, pixels are grouped in cellsof 3 × 3 pixels. Each inner pixel of a PMT (except for the pixels at the edgesof the MAPMT) belongs to 9 different cells as it can be the center of a cell orbelong to the edges. To save memory, cells do not overlap on two PMTs. Itwas demonstrated through simulations that by accounting for them, the triggerefficiency would only increase marginally.

1even if for EUSO-SPB a second trigger level is until now not predicted, it will be use thisnotation as an heritage as conceived for JEM-EUSO.

54

3.4.1 PTT : method 1

A trigger is issued if for a certain number of consecutive GTUs, npst, there isat least one pixel in the EC with an activity (counts) equal to or higher than a

preset threshold, npixthr, and the total number of excess photoelectrons in the 3 ×

3 cell is higher than a preset value ncellthr . npix

thr and ncellthr are set as a function of

the UV background average rate value, while npst in the current simulation istuned on 1 GTUs but in principle it could be set on different values. From nowon it will be used to refer to this specific trigger logic as method 1. A conceptualexample of trigger issued with method 1 is explained in Fig.3.5.

Figure 3.5: Description of the method 1 trigger. Only 3x3 pixel boxes whichdo not belong to different EC can provide a trigger (the grey-dashed cell is

exluded). A set of trigger parameters with npixthr = 3, npst = 1 and ncell

thr =12, corresponding to an average background of 0.25 counts/µs/pixel, can beconsidered as a way of example. Numbers in red are referred to pixels counting anumber of photoelectrons equal or higher npix

thr = 3. To 3 counted photoelectronscorresponds an excess of 1, to 4 counted photoelectrons corresponds an excess of2,etc. The trigger condition is satisfies by the 3 × 3 green box where the countof excess photoelectrons, considering an integration time of 1 GTU (npst = 1),amounts to 14. Such as value is higher than the set threshold for the entire box(ncell

thr = 12) so a trigger can be issued.

3.4.2 PTT : methods 2a and 2b

A second 1st level-trigger logic was introduced in the development of this work.Aim of this new logic was to reduce the threshold on the number of excess pho-toelectrons in the trigger pixel cell nthr,cell, allowing in principle the acquisitionof lower-energy events compared to method 1. However, even considering lowerthresholds, the logic must be able to keep the background events rate under

55

control. In this case a trigger is issued when nthr,cell is satisfied, but requiringan excess on any 3 × 3 pixel cell on the entire EC (not the same cell as formethod 1) during at least 3 (not necessarily consecutive) GTUs over a bunchof 5 consecutive GTUs. From here it will used to refer to this logic as method2a and method 2b, depending respectively by selecting ncell

thr = 7 or ncellthr = 6.

This particular values were chosen referring to an average background of 0.25counts/µs/pixel, which is the result obtained from the Timmins Balloon Flightdata analysis in case of no clouds occurring in the detector field of view [36].Therefore the basic difference between the two methods is only about the thresh-old on the integral of excess photoelectrons in the 3 × 3 pixel cell. A conceptualscheme of the trigger method 2a and 2b is displayed in Fig.4.9.

Figure 3.6: Scheme of trigger method 2a and 2b. The EC named, as a way ofexample, EC 1 is considered. Boxes including 1 are ideally GTUs where thetrigger condition for method 1, considering nthr = 1, occurs. Starting from theGTU 9 the excess persists on 3 over a bunch of 5 consecutives GTUs (see theblue-line box), so a trigger is issued.

3.5 Estimation of the background trigger ratefor PTT method 1, 2a and 2b

All trigger methods were tested performing simulation on the UV poissonianbackground. An average rate of 0.25 counts/µs/pixel was assumed for the reasonexpressed above. Trigger rates obtained simulating 9 seconds of background fora single EC are showed in table 3.1.

Trigger method npixthr npst ncell

thr Trigger rate (Hz/EC)1 3 1 12 0.12a 3 1 7 0.12b 3 1 6 0.3

Table 3.1: Trigger rates for method 1, 2a and 2b. As expected method 2b givesan higher trigger rate than method 2a because of its lower threshold.

Referring to table 3.1 method 1 and method 2a provided trigger rates lowerthan the imposed constraint of 0.2 Hz/EC, while the expected trigger rate formethod 2b (0.3 Hz/EC) was higher. Despite of that, such method was taken intoconsideration during the analysis in order to see how the trigger efficiency could

56

improve in case of EAS simulation, even having a slightly higher backgroundtrigger rate.

3.6 EAS simulation

In this section results about the study of the performance of the above intro-duced methods for EAS simulations are shown. Handled points can be summa-rized as :

• High quality detector performances, considering different atmospheric con-dition;

• Low quality detector performances, considering different atmospheric con-dition;

• tests on different average background level;

• estimation of weighted spectra assuming different cloud types

• tests on variable GTUs duration.

3.6.1 High quality detector performances

An evaluation of the trigger response assuming different values of the detectoroverall efficiency was performed since the development of the optics instru-mentation is ongoing and the global efficiency of the detector is nowadays stillunknown. A first set of simulations have been done considering a guess of highperformances detector, assuming an overall efficiency of ∼ 0.15.

Clear sky atmosphere

The efficiency of the three trigger methods was first tested in clear sky conditionssimulating 45◦ showers at the energies of 1018 eV, 1.5 · 1018 eV and 5 · 1018 eV.Data sets of Nsimu,cyl = 990 events were simulated for each energy. Aim of thisfixed-conditions events simulation was to identify possible peculiarities occurringin any of the three trigger methods considering the most populated bin of thezenith angle distribution. The number of events triggered by the three methodsfor each energy is reported in table 3.2.

Energy(eV) Method 1 Method 2a Method 2b1018 23 31 54

1.5 · 1018 77 78 825 · 1018 168 152 154

Table 3.2: Number of triggered events by method 1, 2a, and 2b considering setsof Nsimu,cyl = 990 and 45◦ events for different energies.

Numbers listed in table 3.2 can be summarized as:

• method 2b had an higher efficiency compared to the others at 1 · 1018 eV.This was expected due to its lower value for the ncell

thr trigger parameterallowing a better detection of low energy events (see Fig.3.7 as exampleof an event triggered only by method 3).

57

• three methods seemed to be comparable at 1.5 · 1018 eV.

• method 1 was more efficient at 5 · 1018 eV. An event by event analysisrevealed that the higher number of triggered events was due to Cherenkov-like events, whose particular signal morphology could be observed only bythis method. Such as signal occurs when the Cherenkov component ofthe EAS enters in the field of view and after few GTUs is immediatelyreflected from the earth’s surface causing a relatively strong short-spottedflash lasting just the duration comparable with a single GTU. (see examplein Fig. 3.9). Only method 1, because of its logic construction whichrequires a localized excess of signal persisting no more than 1 GTU wasable to identify this class of events. On the other hand, events with highluminosity well into the field of view (see Fig.3.8) are equally seen by allthe three methods.

Gtus 0-24 Hits on screen: 178

X [mm]200− 100− 0 100 200 300

Y [m

m]

200−

150−

100−

50−

0

50

100

150

Cou

nts

0

5

10

15

20

25

30

Figure 3.7: Left: Example of a 45◦ event at 1 ·1018 eV triggered only by method2b. Right: Light curve (green plot) representing the number of detected photonsin each GTU. For low energy events, the lower number of produced and detectedphotons (compared for example to the number of photons associated to the5 · 1018 eV event shown in Fig.3.8 right panel) allows method 2b to performbetter compared to the others because of its lower threshold.

Gtus 0-27 Hits on screen: 871

X [mm]300− 200− 100− 0 100 200

Y [m

m]

150−

100−

50−

0

50

100

150

Cou

nts

0

5

10

15

20

25

30

Figure 3.8: Left: Example of 5 ·1018 eV event entirely crossing the field of view,observed by all the methods. Right: Light curve (green plot) representing thedetected photons. In case of extremely strong signal such as for this particularclass of events, the trigger efficiency is the same independently by the adoptedmethod.

According to formula 3.3 the number of events for the spectra triggered bythe three methods is reported in table 3.3. Method 2b, in case of clouds absence,

58

Gtus 0-25 Hits on screen: 101

X [mm]150− 100− 50− 0 50 100 150 200 250

Y [m

m]

50−

0

50

100

150

200

Cou

nts

0

5

10

15

20

25

30

Figure 3.9: Left: Example of Cherenkov like event at 5 · 1018 eV. Right panel:a relatively strong signal due to the reflection of the Cherenkov light at groundoccurs on GTU 22 (green curve), lasting the duration of 1 GTU. Such an eventcan be triggered only by method 1, which requires a localized excess of signalpersisting only 1 GTU, while methods 2a and 2b look for a persistence of thesignal on the EC surface for a few GTUs.

resulted to perform better than others. Furthermore, by simulating with spectraand not at fixed energies, it turned out that method 2b was more efficient atlower energies, while all three methods looked equivalent at higher energies (seeFig.3.10).

Triggered Eventsmethod 1 99 ± 7method 2a 90 ± 6method 2b 126 ± 9

Table 3.3: Number of events referred to triggered spectra by methods 1, 2a and2b simulating an high performances detector in clear sky condition. A 2 monthsflight duration and duty cycle of 0.13 are assumed.

Cloudy atmosphere

Further simulations were performed considering atmosphere with clouds, parametrizedinside ESAF in terms of altitude and optical depth (OD). Different value of theseparameters were scanned during the analysis:

• h = 2 km for low altitude clouds,

• h = 5 km for middle altitude clouds,

• OD = 1 for thin clouds,

• OD = 5 for thick clouds.

As for the clouds absence case, a first run of simulations was carried outconsidering sets of 990 events, 45◦ inclined at different energies. Results arelisted in table 3.4.

Results in table 3.4 can be summarized as:

59

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

5

10

15

20

25

30

Balloon triggered spectrum

method 1

method 2a

method 2b

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.20

2

4

6

8

10

12

14

16

18

20

Balloon triggered spectrum

Figure 3.10: Left: Plots displaying triggered spectra performed by method 1(red curve), method 2a (green curve), method 2b (blue curve) assuming a 2months flight duration and a duty cycle of 0.13. An high performances detectorcorresponding to an overall efficiency of 0.15 is assumed. Right: Error bars plotreferred to the triggered spectrum obtained with method 1. The uncertainty onthe total number of triggered events is calculated propagating the uncertaintieson each energy bin as shown in formulas 3.4 and 3.6).

Energy (eV) Hcloud (Km) O D Triggered Events Triggered Events Triggered Events(Method 1) (Method 2a) (Method 2b)

5 · 1017 2 1 0 0 05 · 1017 2 5 19 0 21 · 1018 / / 23 31 541 · 1018 2 1 16 22 421 · 1018 2 5 79 15 405 · 1018 / / 168 152 1545 · 1018 5 1 157 127 1285 · 1018 5 5 103 117 125

Table 3.4: Triggered events by method 1, 2a, and 2b simulating sets of Nsimu,cyl

= 990 events, 45◦ inclined at different energies, in different cloudy conditions.

60

• method 1 allowed to trigger events at 5 · 1017 eV in case of thick clouds,while methods 2a and 2b in this condition were inefficient. A more detailedanalysis revealed this was due to Cherenkov like events. The Cherenkovcomponent of the shower is expected in fact to be reflected on the topof the thick cloud generating an intense short-spotted flash lasting lessthan 1 GTU, threfore, detectable only by method 1. On the other handthresholds of method 2a and 2b, even if lower, don’t allow to detect suchclass of events.

• method 2b at 1 · 1018 eV in case of thin clouds was more efficient. In thiscase the showers reflection on the top of the clouds is spread on severalGTUs and is not as intense as for thick clouds because of the lower value ofOD. EASs can be therefore easily triggered by methods 2a and 2b, whoselogic basically searches a weaker but larger signal persistence (thresholdsare indeed lower compared to method 1 lasting few GTUs. On the otherhand for thick clouds method 1 was more efficient, always due to EASsthat enter the field of view and after few GTUs reach the top of the cloudgenerating a Cherenkov flash.

• 5 ·1018 eV events were better seen by method 1 in case of thick-low clouds,even this time due to Chrenkov flashes. However considering high clouds(h = 5 Km) and the same optical depth the number of triggered eventsdecreased. In this particular condition it can happen that the altitudeof the cloud does not allow the shower to achieve the maximum duringits atmospheric development and consequently the Cerenkov flash is notstrong enough to be detected by method 1. Methods 2a and 2b are moreefficient because, even if the maximum of the shower is not in the FoV,their lower thresholds allowed to trigger anyway during the early stagesof the EAS development.

The same procedure adopted for the clouds absence analysis was used toextimate triggered spectra in case of cloudy conditions. Results are reported intable 3.5.

Hcloud (Km) O D Triggered Events Triggered Events Triggered Events(Method 1) (Method 2a) (Method 2b)

/ / 99 ± 7 90 ± 6 126 ± 92 1 88 ± 6 78 ± 7 111 ± 102 5 249 ± 21 84 ± 8 103 ± 95 1 83 ± 7 86 ± 9 111 ± 105 5 247 ± 24 76 ± 6 109 ± 9

Table 3.5: Number of events referred to triggered spectra obtained with meth-ods 1, 2a and 2b simulating an high performances detector in different cloudscoverage conditions.

Referring to table 3.5 method 1 was better suited, especially at lower ener-gies, due to the Cherenkov flash occurring in case of thick-clouds. Method 2bwas still the most efficient in case of thin clouds (regardless the altitude) sinceits lower thresholds allowed to catch a lower signal spread on few GTUs (seeplots in Fig.3.11).

61

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

5

10

15

20

25

Balloon triggered spectrum

method 1

method 2a

method 2b

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

5

10

15

20

25

Balloon triggered spectrum

method 1

method 2a

method 2b

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

5

10

15

20

25

30

35

40

45

Balloon triggered spectrum

method 1

method 2a

method 2b

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

10

20

30

40

50

Balloon triggered spectrum

method 1

method 2a

method 2b

Figure 3.11: Spectra obtained in cloudy coverage conditions assuming an highperformances detector. In each canvas red curves are referred to method 1,green curves to method 2a and blue curves to method 2b. Top panels: Triggeredspectra for thin low altitude clouds (left) and thin middle altitude clouds (right).Bottom panels: Triggered spectra for thick low altitude clouds (left) and thickmiddle altitude clouds (right).

62

3.6.2 Low quality detector performances

An overall efficiency of the detector of ∼ 0.05 was assumed, according with theglobal efficiency value estimated as for the first balloon mission. This can beconsidered a conservative case since efforts of the entire collaboration are aimedto provide a better instrument for the EUSO-SPB flight. Triggered spectraperformed with method 1 and method 2b were compared both in clouds absenceand different cloudy coverage conditions.

Clear atmosphere

Detector Triggered events Triggered eventsEfficiency (Method 1) (Method 2b)

0.15 99 ± 7 126 ± 90.05 9.9 ± 0.8 14.4 ± 1

Table 3.6: Comparison between triggered spectra performed by methods 1 and2b in clear sky condition, assuming a detector overall efficiency of 0.05 (lowperformance detector) and 0.15 (high performance detector).

Referring to table 3.6, method 2b, as expected by previous simulation, wasfound out more efficient in clear sky condition. It must be noticed that thenumber of triggered events dropped a factor ∼ 10 changing from an efficiencyof ∼ 0.15 to an efficiency of ∼ 0.05. However, it should be reminded that thebackground was kept constant, while it should have been changed by a factor3 as well. In other words the number of events collected with the assumptionof detection efficiency of 0.15, assume a background level three times smallerthan what should have been extrapolated from the Timmins flight, taking intoaccount the different detection efficiency of the two detectors. However, as thenightglow background is largely variable, the assumed one is no unrealistic, eventhough not much probable.

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

0.5

1

1.5

2

2.5

3

3.5

Balloon triggered spectrum

method 1

method 2b

Figure 3.12: Triggered spectra by method 1 (red plot) and 2b (blue plot) inclear sky condition assuming an overall efficiency of 0.05.

63

Hcloud (Km) O D Triggered Events Triggered Events(Method 1) (Method 2b)

/ / 9.9 ± 0.8 14.4 ± 1.02 1 10.2 ± 1.1 10.9 ± 0.92 5 32.4 ± 2.7 12.7 ± 1.15 1 8.9 ± 0.7 11.4 ± 1.15 5 26.1 ± 2.8 11.2 ± 0.9

Table 3.7: Number of events referred to triggered spectra performed by method1 and 2b in cloudy sky condition, assuming an overall efficiency of the detectorof ∼ 0.05.

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

0.5

1

1.5

2

2.5

Balloon triggered spectrum

method 1

method 2b

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

2.2

2.4meethod 1

method 2b

Balloon triggered spectrum

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

1

2

3

4

5

6

Balloon triggered spectrum

method 1

method 2b

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

Balloon triggered spectrum

method 1

method 2b

Figure 3.13: Spectra obtained in cloudy coverage conditions assuming a lowperformances detector. In each canvas red curves are referred to method 1 andblue curves to method 2b. Top panels: Triggered spectra for thin low altitudeclouds (left) and thin middle altitude clouds (right). Bottom panels: Triggeredspectra for thick low altitude clouds (left) and thick middle altitude clouds(right).

64

Cloudy conditions

Even considering cloudy coverage the number of triggered events was suppressedby a factor ∼ 10 comparing spectra obtained with efficiency ∼ 0.15 to spectrasimulated with efficiency ∼ 0.05 (see table 3.7). Furthermore method 1 wasstill the most efficient in case of thick clouds allowing to trigger ∼ 0.5 eventper day. An average optical depth of 4.6 ± 1.1 along the first SPB trajectoryin 2015 was estimated using ISCCP data [75]. The analysis revealed that thefraction of time with clear atmosphere conditions was only (23 ± 7)%. Suchresults, with the possibility to detect events by mean Cherenkov reflection (thisis impossible for ground experiments), make the method 1 the most suitable forthe upcoming mission.

3.6.3 Variable background analysis

In the following simulations a background of 0.5 counts/µs/pixel was assumed.Such as value corresponds to the measured background during the first balloonflight as occurred in case of clouds coverage. Trigger parameters such as npix

thr

and ncellthr were consequently adapted. It must be noticed that in ESAF only

simulations with a fixed average background can be done so time dependingbackground analysis are until now not feasible. Besides the aim of these sim-ulations was just to have an understanding how trigger performances scaledtaking into account a two times background level. An efficiency of 0.05 wasalso assumed in order to have a conservative point of view in terms of detectorperformances. Simulations were performed considering trigger method 1 bothin clear sky and in clouds coverage condition. Results listed in table 3.8 showan average background of 0.5 counts/µs/pixel causing a reduction of a factor ∼2-3 of the number of triggered events.

Hcloud (Km) O D Triggered Events Triggered EventsBkg = 0.5 counts/µs/pixel Bkg = 0.25 counts/µs/pixel

(Method 1) (Method 1)/ / 3.2 ± 0.3 9.9 ± 0.82 1 3.6 ± 0.3 10.2 ± 1.12 5 17.0 ± 1.5 32.4 ± 2.75 1 3.9 ± 0.4 8.9 ± 0.75 5 17.5 ± 2.1 26.1 ± 2.810 1 0 010 5 0 0

Table 3.8: Number of triggered events comparing two different background lev-els.

3.6.4 Variable GTU duration analysis

A study scanning different GTU durations were done taking into considerationthat for EUSO-SPB the smaller average dimension of a pixel at ground comparedto JEM-EUSO could mean a shorter persistence time of the signal in the 3x3trigger pixel box. In other words the so far adopted GTU duration of 2.5 µs isa heritage of the JEM-EUSO concept, but not optimized for a balloon flight.

65

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Balloon triggered spectrum

method 1

Figure 3.14: Triggered spectrum (method 1) in clear sky condition consideringan efficiency of ∼ 0.05 and average background of 0.5/µs/pixel.

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Balloon triggered spectrum

method 1

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Balloon triggered spectrum

method 1

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

0.5

1

1.5

2

2.5

3

3.5

4

Balloon triggered spectrum

method 1

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

0.5

1

1.5

2

2.5

3

Balloon triggered spectrum

method 1

Figure 3.15: Spectra simulated with method 1 assuming an efficiency of ∼ 0.05and a background of 0.5 counts/µs/pixel. Top panels: Triggered spectrum forthin low altitude clouds (left) and thin middle altitude clouds (right). Bottompanels: Triggered spectrum for thick low altitude clouds (left) and thick middlealtitude clouds (right).

66

Referring to Fig.3.16 the persistence time of the signal in the 3x3 trigger pixelcan be calculated as:

tAC =l

c · tan θ2

(3.9)

with:

• l: the average dimension of the projection of the 3×3 trigger box at an analtitude shower development of ∼ 5-7 km (∼ 300 m),

• θ: the EAS zenith angle,

• c: the speed of light in the vacuum.

Figure 3.16: Assuming an average dimension of the 3×3 trigger pixel box of ∼300 m at an altitude of 5-7 km, it is possible to estimate the persistence time ofthe signal in the 3x3 trigger pixel box as a function of the shower zenith angle.

As shown in the plot represented in Fig.3.17, EAS with θ > 45◦ stay for atime < 1 GTU of 2.5 µs in the 3x3 pixel-box. In this condition a data acquisitionperformed with an higher temporal resolution adopting shorter GTUs impliesa better noise-signal ratio, leading in principle to a better recognition of EAStracks. Method 1 was tested in clear sky condition simulating different sets ofnpst, overall efficiency and background (see table 3.9).

Referring to table 3.9 it must be noticed that to an overall efficiency of 0.05corresponds a 0.25 counts/µs/pixel background. A three times efficiency of 0.15was consequently coupled with a three times background of 0.75 counts/µs/pixelto be more consistent. Spectra calculated assuming as simulation parameters

67

Figure 3.17: The persistence time of the signal in the 3x3 trigger pixel boxdepends on the zenith angle of the shower

Set npst detector efficiency Background (counts/µs/pixel)1 1 0.05 0.252 1 0.15 0.753 2 0.05 0.254 2 0.15 0.75

Table 3.9: Sets of parameters used for the variable GTU duration analysis.

sets 1, 2, 3 and 4, varying the GTU duration, are reported respectively in table3.10, 3.11, 3.12 and 3.13.

Analysis showed that for each set of simulated parameters GTU of 2.5 µswas the one in general performing worse as expected, while shorter GTUs werefind out to increase the number of triggered events. Using shorter GTU in-volves several consideration from the hardware point of view. For example anincreasing of the power consumption is expected as a consequence of the fact ofintroducing a faster clock signal in order to have shorter GTUs. More resourcesare required also considering the trigger algorithm performing faster calcula-tions. On the other hand a feedback from the collaboration people working onthe acquisition system (ASIC) reveals the impossibility to read data using tem-poral step shorter than 1.6 µs. All this factors contribute to keep unchangedthe GTU duration of 2.5 µs even though it might be checked in future till whichduration the GTU could be reduced without major impacts from the hardwareand trigger calculation point of view.

3.7 Conclusions

An overview of the trigger performances of the EUSO-SPB PTT system wasgiven considering several aspects. Three different trigger methods (method 1,method 2a and method 2b) were tested in several cloud coverage conditions, sim-ulating the detector response assuming two different values of overall efficiency.From an high performances to a low performances detector, the number of trig-gered events dropped by a factor ∼ 10 both in clear sky and in cloud coverageconditions. According to the cloud coverage analysis reported in [75], method

68

npixth npst ncell

th GTU duration(µs) Triggered events(Method 1)

2 1 11 1 2.6 ± 0.22 1 8 1 12.2 ± 1.73 1 7 1.5 15.7 ± 1.63 1 9 2 12.2 ± 0.93 1 12 2.5 9.9 ± 0.8

Table 3.10: Number of events considering different GTU duration with npst =1, background = 0.25 counts/µs/pixel and efficiency = 0.05. In case of GTUduration of 1 µs, 2 sets of trigger thresholds are used in order to avoid roundingeffects for low rates of bkg (ncell

th = 8 is for 0.2 counts/GTU/pixel while ncellth

= 11 is for 0.3 counts/GTU/pixel). Except for GTU = 1 µs with the nominalbackground of 0.25 counts/µs/pixel rounded to 0.3 counts/µs/pixel, an highernumber of events is expected assuming shorter GTUs.

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

0.5

1

1.5

2

2.5

3

Balloon triggered spectrum

sµ1 sµ1.5

sµ2 sµ2.5

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

0.5

1

1.5

2

2.5

3

Balloon triggered spectrum

sµ1 sµ1.5

sµ2 sµ2.5

Figure 3.18: Triggered spectra for different GTU duration assuming npst =1. An average background of 0.25 counts/µs and an overall efficiency of thedetector of 0.05 are considered. Black curves are referred to the case GTU= 1 µs rounding the background to 0.3 counts/µs/pixel (left panel) and 0.2counts/µs/pixel (right panel).

npixth npst ncell

th GTU duration(µs) Triggered events(Method 1)

4 1 8 1 40.1 ± 4.14 1 14 1.5 20.3 ± 1.54 1 19 2 19.2 ± 1.64 1 25 2.5 21.4 ± 1.7

Table 3.11: Number of triggered events assuming different GTU duration withnpst = 1, background = 0.75 counts/µs/pixel and efficiency = 0.15.

69

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

1

2

3

4

5

6

7

8

9

Balloon triggered spectrum

sµ1 sµ1.5

sµ2 sµ2.5

Figure 3.19: Triggered spectra for different GTU duration assuming npst = 1.An average background of 0.75 counts/µs/pixel and an overall efficiency of thedetector of 0.15 are considered.

npixth npst ncell

th GTU duration(µs) Triggered events(Method 1)

2 2 15 1 4.3 ± 0.42 2 11 1 13.4 ± 1.33 2 10 1.5 10.8 ± 1.03 2 12 2 11.8 ± 0.93 2 17 2.5 7.1 ± 1.6

Table 3.12: Number of triggered events assuming different GTU duration withnpst = 2, background = 0.25 counts/µs/pixel and efficiency = 0.05. Also inthis case two set of thresholds corresponding to two different values of roundedbackground were simulated in order to avoid rounding effects occurring for lowbackground rate.

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

0.5

1

1.5

2

2.5

3

Balloon triggered spectrum

sµ1 sµ1.5

sµ2 sµ2.5

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

0.5

1

1.5

2

2.5

3

Balloon triggered spectrum

sµ1 sµ1.5

sµ2 sµ2.5

Figure 3.20: Triggered spectra for different GTU duration assuming npst = 2.An average background of 0.25 counts/µs/pixel and an overall efficiency of thedetector of 0.05 are considered. Black curves are referred to the case GTU= 1 µs rounding the background to 0.3 counts/µs/pixel (left panel) and 0.2counts/µs/pixel (right panel).

70

npixth npst ncell

th GTU duration(µs) Triggered events(Method 1)

4 2 11 1 35.0 ± 2.64 2 20 1.5 28.9 ± 2.14 2 27 2 23.3 ± 1.74 2 36 2.5 16.9 ± 1.3

Table 3.13: Number of triggered events assuming different GTU duration withnpst = 2, background = 0.75 counts/µs/pixel and efficiency = 0.15.

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2

0

1

2

3

4

5

6

7

8

9

Balloon triggered spectrum

sµ1 sµ1.5

sµ2 sµ2.5

Figure 3.21: Triggered spectra for different GTU duration assuming npst = 2.An average background of 0.75 counts/µs/pixel and an overall efficiency of thedetector of 0.15 are considered.

71

1 was found out to be the best candidate because it was the most suitable totrigger events in thick-cloud conditions which is the most common situationthat will occur on the EUSO-SPB trajectory. On the other hand method 1allowed the detection of Cherenkov-like events reflected on the top of the cloudsthat can not be carried out from ground based experiments. An analysis formethod 1, assuming a low efficiency detector and a two times background inpresence of clouds as the one measured during the first balloon flight. In thiscase, the higher thresholds set for the trigger algorithm caused the number ofevents dropping a factor ∼ 2-3 with an expected number of triggered events of0.5 per day in case of thick cloud condition. Further tests simulating severalsets of npst, detector overall efficiency and background were performed scanningdifferent GTU duration. The GTU duration of 2.5 µs was in general the oneperforming worse as expected. However, with the current electronics it is notpossible to make major improvements which would require a reduction of theGTU duration by a factor of ∼ 2.

72

Chapter 4

Evaluation of theEUSO-SPB reconstructionperformances

Aim of this chapter is the description of the event reconstruction for EUSO-SPB performed using the ESAF software. An overview of the reconstructionalgorithms is presented and then reconstruction performances are evaluated interms of number of reconstructed events compared to the number of triggeredevents. The work about the optimization of the reconstruction algorithm is alsoshown.

4.1 The reconstruction algorithm

4.1.1 Pattern recognition

After the trigger is issued all data are sent to Earth for the reconstruction ofthe event properties. Data can be thought as a table of pixel–GTUs carryingthe information of how many counts are associated to each pixel for each time.The first necessary step would be the recognition of the signal inside the data.Several algorithms have been developed for this purpose. In the present studythe pattern recognition algorithms used to identify the real signal inside the datasent at ground is the so called LTTPreClusteringModule developed by Bertainaand Gorgi [76]. In few words it looks for high density tracks of signal on the focalsurface. The signal is in fact characterized by a spot moving on the focal surfacewith a speed equal to the projected speed of light. The pattern recognitionalgorithms will start from an initial condition which is presently defined as thepixel–GTUs with the maximum counts density. Such a frame is defined asthe 3 × 3 box which realizes the highest count integration within 5 consecutiveGTUs. This is done in order to reduce the probability of a random fluctuation ofbackground. In fact the persistence over such a long time is very likely due to thepresence of a signal spot. Following the initial condition selection an integrationis started along several predefined directions. The integration box will have 3× 3 size and will be moved for the length corresponding to several GTUs. Thespeed of the integration box will be decided according to the projection on the

73

focal surface of a set of predefined shower directions. The track which maximizesthe integration is chosen as the center for the selection. All the pixels inside apreset distance will be therefore selected. Such a predefined distance is set fromparameter and is expressed in terms of pixel size units. A visual example ofthe above described algorithms can be given in Fig.4.1. Here one single protonshower (θ = 45◦) has been simulated and reconstructed. The top side plotrepresent the integrated signal over the entire shower development.

Gtus 0-23 Hits on screen: 35178

X [mm]300− 200− 100− 0 100 200

Y [m

m]

150−

100−

50−

0

50

100

150

Cou

nts

0

5

10

15

20

25

30

Gtus 0-23 Hits on screen: 219

X [mm]300− 200− 100− 0 100 200

Y [m

m]

150−

100−

50−

0

50

100

150

Cou

nts

0

5

10

15

20

25

30

Figure 4.1: Top panel. Integrated signal (background and shower track) overthe entire shower development. In principle no shower track is recognizable.Bottom panel. Event signal generated by the shower. Task of the mentionedpattern recognition algorithm is the identification of such a pattern.

As can be seen the background saturates the image and at least by eye notrack recognition is possible. On the bottom plot just the signal pixels can beseen. This information is of course not available for any measurements sincesignal is not directly distinguishable by background.

4.1.2 Angular and profile reconstruction

In order to perform any angular reconstruction at least two informations arenecessary. The first is the already mentioned reconstructed track. The secondfundamental information is a map which associates to each pixel a directionin the FOV. In ESAF such a map is called PixelAngleMap and associates toeach pixel ID a θ and a φ angle. Associated with such an angle are also therelative sigmas expressing the width of the FOV associated with each pixel.

74

The angular reconstruction algorithm will first determine the so called TrackDetector Plane which represents the plane where both shower and detector arelocated. Once this plane has been found the shower timing information is usedto determine the inclination of the shower itself. Through all the steps, theapplication of fit algorithms will be necessary both for the TDP and inclinationdetermination. Between them the least square, median, the linear and thehough fitting procedures can be mentioned. An iterative procedure combiningthe informations carried by the zenith angle of the shower and the reconstructedtrack is then used to reconstruct the shower profile (core position, maximumposition and maximum altitude)[77].

4.1.3 Energy and Xmax reconstruction

The reconstruction chain continues with the calculation of the shower pfofile (i.ethe number of electrons for each step) using a parameterization of the atmo-spheric transmittance and the position of the shower at each time according tothe shower reconstructed. The final electron profile is then fitted with a showerparameterization function to obtain the energy and Xmax parameters. An essen-tial step of the reconstruction is the background subtraction. This amounts tothe number of pixels selected in the GTU multiplied by the average backgroundon each pixel-GTU. Before subtracting the background as fixed amount for eachpixel-GTU the error bars associated with each GTU will be calculated as theroot square of the amplitude value. The absolute error bars will be preservedafter the background subtraction. For all the following steps the error bars mustbe scaled according to the conversion factors applied in each step. In Fig.4.2 anexample of reconstructed luminosity curve of an 1 · 1019 eV event at 45◦ is shown(see the green points plot with error bars). The red-line and black-line plot rep-resent respectively the fitting curve whose parameters are used to estimated theenergy and the Xmax, and the luminosity curve based on the Monte Carlo truth.The blue-flat line corresponds to the averaged background. An event, to be re-constructed, must satisfy some constraints. More in detail, a minimum numberof five reconstructed points over the average estimated background with a chisquare per point lower than three was required. These requirements were doneas compromise taking into account that the dead zones on the focal surface ofthe detector could lead to a loss of signal, reducing the amount of points withsignal large enough to be counted.

4.2 Reconstruction performances analysis

4.2.1 Energy resolution for fixed event conditions

It must be said that, given the relatively low number of expected triggeredevents, providing a reconstructed spectrum is not a goal of the mission. Takingthis into consideration an estimation of the reconstruction quality expressed interms of energy resolution was performed. As first, three data sets of 45◦ eventsrespectively at 1 · 1018 eV, 1.5 · 1018 eV and 5 · 1018 eV were simulated. Thereconstruction efficiency for each sample could be defined as:

εEUSO-SPB,Rec(E) =N rec(E)

N simu,cyl(E)(4.1)

75

)2atmosphere thickness(g/cm200 400 600 800 1000 1200

Ne

0

1000

2000

3000

4000

5000

6000

7000610×

Figure 4.2: The reconstructed electrons profile is represented by green pointswith black error bars calculated as a consequence of the background subtraction.The flat blue lines identifies the average background level, while the red curverepresents the fitting curve, whose parameters are the estimated energy andXmax. The black curve is the effective number of electrons generated in thesimulation.

whereas:

• Nsimu,cyl(E): the number of simulated events for a certain energy, gener-ated as shown in chapter 3, section 3.3;

• Nrec(E): the number of reconstructed events for that particular energy;

The energy resolution was calculated as σ of the Gaussian fit to the α dis-tribution, whereas:

α =Erec − Esimu

Esimu(4.2)

with:

• Erec(E): the reconstructed energy of the event

• Nsimu(E): the simulated energy of the event

Results about such as samples are shown in table 4.1.

Energy(eV) Energy resolutionErec−Esimu

Esimu

1 · 1018 0.261.5 · 1018 0.275 · 1018 0.37

Table 4.1: Energy resolution for samples of 1 · 1018 eV, 1.5 · 1018 eV and 5 ·1018 eV simulated events.

Referring to table 4.1, moving to higher energy the resolution got worse. Anexplanation of this non intuitive behavior can be understood looking at Fig.4.5.

76

In the histogram on the panel two different populations of events can be easilyobserved, one identified by the green dashed line and the other by the blueline. This could not be observed for events at 1 · 1018 eV and 1.5 · 1018 eV(see respectively Fig.4.3 and Fig.4.4). The blue line peak corresponds to thoseshowers whose maximum is well into the field of view (see green plot in Fig.4.6(b)as a way of example). In Fig.4.6(a) is shown the distribution of the altitude atwhich each photon was generated. Having the events the same zenith angle(45◦), the maximum of the shower is around 5000 m altitude. Whenever themaximum is visible most of the photons distribute around that altitude. Thereconstruction procedure is therefore reconstructing events around α = 0 (upto some systematic still under study). In cases where only the initial (or final)part of the shower is visible (see green plot in Fig. 4.6(d)) only photons comingfrom higher-lower altitude can be detected. Fig.4.6(c) displays the distributionof the altitude of the photon production of the green population of Fig.4.5.This tells that for the higher energies also the tails of the shower are triggering(and therefore only photons coming high or low altitudes are seen). In suchcases however the the reconstructed energy is underestimated. Only for eventsin such energy range is possible in fact to see this double peaked distributionwhile at the lower energies the tails cannot trigger. Anyway this effect, beingand underestimation, does not contribute to the spillover an to the deformationof the flux shape. Moreover, such events can be easily recognized on the lightcurve and on the image of the shower. The resolution at 5 · 1018 eV, not takinginto consideration events without maximum improved from 0.37 to 0.26.

α1− 0.8− 0.6− 0.4− 0.2− 0 0.2 0.4 0.6 0.8 10

1

2

3

4

5

6

7

Energy Resolution

Figure 4.3: Distribution of α for 45◦ events at 1 · 1018 eV. The energy resolutionis calculated as the standard deviation of a Gaussian distribution fitting thehistogram.

4.2.2 Triggered spectrum against fraction of reconstructedevents

An evaluation of the fraction of reconstructed events, compared to the totalamount of triggered events collected in spectra obtained in chapter 3 was carriedout. Considered spectra were those obtained with method 1 both in clear skyand cloudy conditions, simulated using the following sets of parameters:

• average background = 0.25 counts/µs/pixel;

77

α1− 0.8− 0.6− 0.4− 0.2− 0 0.2 0.4 0.6 0.8 10

2

4

6

8

10

12

14

16

Energy Resolution

Figure 4.4: Distribution of α for 45◦ events at 1.5 · 1018 eV.

α1− 0.8− 0.6− 0.4− 0.2− 0 0.2 0.4 0.6 0.8 10

5

10

15

20

25

Energy Resolution

Figure 4.5: Distribution of α for 45◦ events at 5 · 1018 eV.

78

altitude(m)0 2000 4000 6000 8000 10000 12000 140000

0.001

0.002

0.003

0.004

0.005

Photons altitude distribution

(a) (b)

altitude(m)0 2000 4000 6000 8000 10000 12000 140000

0.001

0.002

0.003

0.004

0.005

0.006

0.007

Photons altitude distribution

(c) (d)

Figure 4.6: a)Distribution of the altitude of the production position of photonsfor 45◦ events at 5 · 1018 eV including the maximum in the field of view. Thedistribution is centered on the altitude value of ∼ 5000 m. b)Example of lightcurve (green curve) for an event whose maximum is well into the field of view.c)Distribution of the altitude of the production position of photons for 45◦

events at 5 · 1018 eV triggering on the initial (or final part) of the shower.The distribution shows clearly two peaks centered on ∼ 2000 m and ∼ 9000m (representing therefore photons coming from lower altitude associated to thetail of showers and photons coming from higher altitude associated to the initialpart of showers). d)Example of light curve (green curve) for an event triggeringon the tail of the shower development.

79

• detector overall efficiency 0.05 and 0.15.

The number of reconstructed events for a certain energy bin of the spectrumwas estimated using a similar procedure adopted for the calculation of the num-ber of trigger events as for the triggered spectra (see 3, section 3.2). In this casethe measured flux weighted a reconstruction efficiency as defined in formula 4.1.

The number of reconstructed events compared to the total amount of trig-gered events considering an efficiency of 0.15 is reported in table 4.2. As one cansee reconstructed fraction in clear sky conditions and considering clouds withOD = 1 (low and middle altitude) was ∼ 38% of the total amount of triggeredevents. For low altitude thick clouds the fraction of reconstructed events cor-responded to ∼ 25% and dropped to ∼ 13% considering middle altitude thickclouds. One must notice that even if these percentages are lower, the abso-lute number of reconstructed events for low altitude thick clouds is comparablewith other remaining cases and is even higher considering middle altitude thickclouds.

For computational reasons the ESAF spectrum was E−1. However a realisticmeasurement of the spectrum steepness gave E−3.1 (see fit on Auger ICRC2013 Infill Data in chapter 3). The resolution function alpha must be thereforecalculated on a weighted spectrum in order to take into account the real spectralshape. All the events were therefore weighted by a factor E−3.1, producing thealpha distribution and calculating the RMS (see tables 4.2 and 4.3).

One must notice that for cloudy conditions there was not a significant wors-ening of the resolution. Moreover for a lower efficiency one got a slightly worseresolution than the 0.15 case.

Hcloud (Km) Optical depth Triggered events Reconstructed events Energy Resolution(Method 1) (Method 1)

/ / 99 ± 7 37 ± 4 0.362 1 88 ± 6 36 ± 4 0.405 1 83 ± 7 31 ± 4 0.302 5 249 ± 9 63 ± 21 0.365 5 247 ± 24 32 ± 6 0.40

Table 4.2: Number of reconstructed events compared to triggered spectra sim-ulated in different atmospheric conditions, assuming an overall efficiency of thedetector equal to 0.15.

The amount of reconstructed events for efficiency 0.05 is reported in table4.3. Reconstructed fractions for each case are comparable with those obtainedwith an efficiency of 0.15.

The main evidence reveals that an average of ∼ 29% of triggered eventscould be reconstructed.

4.2.3 Reconstruction algorithm optimization

In the following is reported the work done to improve the performances of thereconstruction algorithm. As mentioned in paragraph 4.1.1 the EAS signal canbe considered as a spot moving every GTU on the focal surface of the detector ontop of the background. The key element of the pattern recognition algorithm,

80

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.20

2

4

6

8

10

12

14

16

Fraction of reconstructed events

Triggered events

Reconstructed events

Figure 4.7: Triggered spectrum against number of reconstructed events in clearsky condition, assuming an overall efficiency of the detector equal to 0.15.

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.20

2

4

6

8

10

12

14

16

18

20

22

Fraction of reconstructed events

Triggered events

Reconstructed events

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.20

2

4

6

8

10

12

14

Fraction of reconstructed events

Triggered events

Reconstructed events

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.20

5

10

15

20

25

30

35

40

45

Fraction of reconstructed events

Triggered events

Reconstructed events

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.20

10

20

30

40

50

Fraction of reconstructed events

Triggered events

Reconstructed events

Figure 4.8: Top panels: Triggered spectrum against number of reconstructedevents simulating clouds with h = 2 Km and optical depth = 1 (left). Triggeredspectrum against number of reconstructed events simulating clouds with h = 5Km and optical depth = 1 (right). Bottom panels: Triggered spectrum againstnumber of reconstructed events simulating clouds with h = 2 Km and opticaldepth = 5 (left). Triggered spectrum against number of reconstructed eventssimulating clouds with h = 5 Km and optical depth = 5 (right). An overallefficiency of the detector of 0.15 is assumed.

81

Hcloud (Km) Optical depth Triggered events Reconstructed events Energy resolution(Method 1) (Method 1)

/ / 9.9 ± 0.8 3.6 ± 0.4 0.382 1 10.2 ± 1.1 3.7 ± 0.5 0.485 1 8.9 ± 0.7 2.4 ± 0.3 0.392 5 32.4 ± 2.7 8.2 ± 1.0 0.395 5 26.1 ± 2.8 2.8 ± 0.5 0.34

Table 4.3: Number of reconstructed events compared to triggered spectra sim-ulated in different atmospheric conditions, assuming an overall efficiency of thedetector equal to 0.05.

log10(E/eV)17.8 18 18.2 18.4 18.6 18.8 19 19.2 19.40

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

Fraction of reconstructed events

Triggered events

Reconstructed events

Figure 4.9: Triggered spectrum against number of reconstructed events in clearsky condition, assuming an overall efficiency of the detector of 0.05.

82

log10(E/eV)17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.2 19.40

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

Fraction of reconstructed events

Triggered events

Reconstructed events

log10(E/eV)17.8 18 18.2 18.4 18.6 18.8 19 19.2 19.40

0.5

1

1.5

2

2.5

Fraction of reconstructed events

Triggered events

Reconstructed events

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.20

1

2

3

4

5

6

Fraction of reconstructed events

Triggered events

Reconstructed events

log10(E/eV)17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19 19.20

0.5

1

1.5

2

2.5

3

3.5

4

4.5

Fraction of reconstructed events

Triggered events

Reconstructed events

Figure 4.10: Top panels: Triggered spectrum against number of reconstructedevents simulating clouds with h = 2 Km and optical depth = 1 (left). Triggeredspectrum against number of reconstructed events simulating clouds with h = 5Km and optical depth = 1 (right). Bottom panels: Triggered spectrum againstnumber of reconstructed events simulating clouds with h = 2 Km and opticaldepth = 5 (left). Triggered spectrum against number of reconstructed eventssimulating clouds with h = 5 Km and optical depth = 5 (right). An overallefficiency of the detector of 0.05 is assumed.

83

aimed to the identification of the signal track inside the data, is the 3 × 3pixels integration box, moving, after a selection of the initial condition, on theentire focal surface along several directions. The integration is then performedchoosing the projected speed on the focal surface from a set of predefined showerdirections which maximizes the integration. Comparing the number of triggeredto the number of reconstructed events, a first hypothesis about such a losswas made taking into consideration that the signal could not have properlyfollowed the movement of the integration box. Such a guess was confirmedconsidering the triggered events histogram compared to the reconstructed eventsone, plotted as a function of the shower zenith angle. As shown in Fig.4.11 theloss of reconstructed events increases moving to high values of zenith angle.This means that the speed of the signal on the focal surface, especially for highvalues of zenith angle, is not correctly recognized by the 3 × 3 pixel box. Thesignal is probably faster than the integration box.

theta0 10 20 30 40 50 60 70 80 90

0

5

10

15

20

25

30

35triggered theta

rec theta

trigg and rec vs theta

Figure 4.11: Triggered events distribution (red plot) compared to reconstructedevents distribution (green plot) as a function of the event zenith angle. Goingto high values of zenith angle the discrepancy between triggered events andreconstructed events increases, showing that for high zenith angle events thecapability of the pattern recognition algorithm is highly dependent by the speedof the signal on the focal surface.

A first trial to improve the number of reconstructed events was done workingon the distance traveled on the focal surface by the 3 × 3 pixel box. Once thealgorithm selected the speed of the 3 × 3 pixel box is immediate to calculatethe projected distance at ground traveled by the box as :

∆Sbox,ground = V box ·∆t (4.3)

with ∆t = 1 GTU.The distance of the box traveled on the focal surface in 1 GTU can be simply

deduced as :

∆Sbox,fs =∆Sbox,ground

lpixel,ground(4.4)

84

with lpixel,ground being the dimension of the pixel field of view at ground.The formula 4.4 shows how lpixel,ground is a key parameter in order to have awell tuned ∆S capable to follow GTU after GTU the spot moving on the focalsurface. For the first configuration such a parameter, set originally at 200 m, wasfound out to be far from the effective EUSO-SPB pixel size at ground. A morerealistic value tuned on 120 m allowed to improve the number of reconstructedevents. Such a test was applied on the spectrum in clear sky condition, simulatedusing an overall efficiency of 0.15. In table 4.4 are reported the number ofreconstructed events considering the first value assumed for lpixel,ground and theoptimized value :

Triggered events Reconstructed events Reconstructed events(Method 1) (lpixel,ground = 200 m) (lpixel,ground = 120 m)

99 ± 7 37 ± 4 42 ± 4

Table 4.4: Number of reconstructed events assuming two different values of thereconstruction parameters lpixel,ground that corresponds to the average dimensionof a pixel at ground. The value of 120 m is the best tuned value providing thehighest number of reconstructed events.

Such an improvement confirms that the not proper recognition of the signalfrom the box is a good point to start in order to improve the reconstructionperformances. In this philosophy, another approach considering a fine tuningof the speed of the integration box can be attempted. As already mentioned,the speed of the integration box is decided according to the projection on thefocal surface of a set of predefined shower directions and it remains constantduring the entire event development. This approximation is suitable for JEM-EUSO where the altitude of the orbit is ∼ 400 km so the projection of the trackspeed can be considered the same independently from the shower developmentaltitude. For the EUSO-SPB altitude (∼ 40 km) the approximation is not trueas shown in the picture reported in Fig.4.12. As a consequence an altitudedependent speed of the integration box can be introduced.

In order to have an idea of the impact that the variation of the altitude couldhave on the speed of the signal were considered samples of events with the samezenith angle. Examples of events at 65◦ and at 30◦ are shown in Fig.4.13. Plotswere obtained reporting for each simulated event the speed of the signal on thedetector surface as function of the average altitude of photons generating suchas signal for a given GTU. The signal speed was calculated as :

V signal =

√∆x2 + ∆y2

1GTU(4.5)

whith :

• ∆x : the x-coordinates distance on the detector focal surface of the spotsignal center of mass calculated each GTU

• ∆y : the y-coordinates distance on the detector focal surface of the spotsignal center of mass calculated each GTU

It must be noticed that the spreading of the points on a wide range of speeddepends by the average size of the pixel (∼ 3 mm) which is comparable with the

85

ALTITUDE-SPEED DEPENDENCY

400 km

JEM-EUSO

EUSO-SPB

40 km

Figure 4.12: The scheme illustrates the principle of the altitude speed depen-dency of the signal track on the focal surface of the detector. On the left sideis shown the state for JEM-EUSO. Due to its orbit altitude of ∼ 400 km thetrack speed of an air shower projected on the surface of the detector can be con-sidered constant, independently from the altitude of the shower development inatmosphere. The track can be in fact seen as a point moving in the field of viewat an infinite distance from the detector. On the right side is shown the casefor EUSO-SPB. Considering a flight altitude of ∼ 40 km, the approximationof infinite distance air shower development is not valid and the altitude of theshower in atmosphere can influence the projected velocity of the signal on thefocal surface.

86

distance traveled by the signal every GTU. Fitting the two plots and consideringfor each one the range of altitudes where the points are spread is possible tosee an increase of the speed of speed of ∼ 20% which is not negligible. Thisremains an open statement which can provide further improvement concerningthe reconstruction performances of the algorithm.

altitude(m)2000 4000 6000 8000 10000 12000 14000 16000

spee

d(cm

/GTU

)

0

0.5

1

1.5

2

2.5

Signal speed vs shower averaged altitude

altitude(m)0 1000 2000 3000 4000 5000 6000 7000 8000 9000

spee

c(cm

/GTU

)

0

0.2

0.4

0.6

0.8

1

1.2

Signal speed vs shower averaged altitude

Figure 4.13: Signal speed on the detector surface against average altitude ofsignal generating photons for a sample of events at 65◦ (top panel) and 30◦

(bottom panel). The slopes of the red curves fitting the two plots identify inboth of the cases an increase of ∼ 20%.

The most significant optimization was achieved working on the initial con-dition choice carried out by the pattern recognition algorithm. As said in para-graph 4.1.1 the entire reconstruction chain starts defining the 3 × 3 box whichrealizes the highest count integration within 5 consecutive GTUs. In case of fastsignal such a integration time was found out to be too long, allowing a countintegration not only on the real signal but also on the background. Reducingthe integration time from 5 to 1 GTU it was possible to get an improvementin terms of number of reconstructed events. The reconstruction using the opti-mized algorithm was therefore carried out on triggered spectra simulated with

87

method 1 both in clear sky and cloudy conditions (see tables 4.5 and 4.6 forcases simulated using respectively a detector overall efficiency of ∼ 0.15 and ∼0.05). Tables 4.7 and 4.8 show the comparison between the reconstructed frac-tions with and without the optimized algorithm. It can be seen that the numberof reconstructed events improved for each case without a drastic deteriorationof the resolution.

Hcloud (Km) Optical depth Triggered events Reconstructed events Energy Resolution(Method 1) (Method 1)

/ / 99 ± 7 58 ± 6 0.402 1 88 ± 6 56 ± 5 0.355 1 83 ± 7 45 ± 5 0.292 5 249 ± 9 100 ± 13 0.395 5 247 ± 24 84 ± 14 0.40

Table 4.5: Number of reconstructed events with the optimized version of thealgorithm compared to triggered spectra simulated in different atmospheric con-ditions. An overall efficiency of the detector of ∼ 0.15 is assumed.

Hcloud (Km) Optical depth Triggered events Reconstructed events Energy resolution(Method 1) (Method 1)

/ / 9.9 ± 0.8 6.0 ± 0.6 0.372 1 10.2 ± 1.1 5.2 ± 0.5 0.365 1 8.9 ± 0.7 5.0 ± 0.5 0.312 5 32.4 ± 2.7 13.5 ± 1.6 0.405 5 26.1 ± 2.8 6.6 ± 0.9 0.37

Table 4.6: Number of reconstructed events with the optimized version of thealgorithm compared to triggered spectra simulated in different atmospheric con-ditions. An overall efficiency of the detector of ∼ 0.05 is assumed.

Hcloud Optical Fraction of reconstructed Energy Fraction of reconstructed Energy(Km) depth events resolution events resolution

(non optimized algorithm) (optimized algorithm)/ / 0.37 0.36 0.59 0.402 1 0.40 0.40 0.63 0.355 1 0.37 0.30 0.54 0.292 5 0.25 0.36 0.40 0.395 5 0.13 0.40 0.34 0.40

Table 4.7: Comparison between fractions of reconstructed events performed withand without the optimized version of the algorithm. Reconstructed fractions arereferred to spectra simulated with an overall efficiency of the detector of ∼ 0.15.

This optimization is however not concluded and a future work will most likelyfurther improve the results. For example the speed function of the integrationbox for the pattern recognition offers still some margin of improvement.

88

Optical Fraction of reconstructed Energy Fraction of reconstructed EnergyHcloud (Km) depth events resolution events resolution

(non optimized algorithm) (optimized algorithm)/ / 0.36 0.38 0.61 0.372 1 0.36 0.48 0.51 0.365 1 0.27 0.39 0.56 0.312 5 0.25 0.39 0.42 0.405 5 0.11 0.34 0.25 0.37

Table 4.8: Comparison between fractions of reconstructed events performed withand without the optimized version of the algorithm. Reconstructed fractions arereferred to spectra simulated with an overall efficiency of the detector of ∼ 0.05.

4.3 Conclusions

The reconstruction performances of EUSO-SPB were tested using the ESAFframework. The analysis included tests on the energy resolution consideringsamples of fixed condition events. An unexpected worsening of the resolution,compared to events at 1 · 1018 eV and 1.5 · 1018 eV, was found out for 5 · 1018 eVevents. At high energy, indeed, the signal was strong enough allowing to triggerevents developing in the field of view both on the initial and in the final stagesof the shower. Events triggering on tails lead to a sub population of eventswith a underestimated energy. Excluding such events gave a better resolution.Such events can be recognized inside data and don’t represent a problem inthe analysis phase. The reconstruction was also applied to triggered spectraobtained with method 1 considering an overall efficiencies of the the detector of0.15 and 0.05, and an average value of background of 0.25 counts/µs/pixel. Forlow and middle altitude thick clouds the fraction of reconstructed events waslower compared to the clear sky and thin clouds cases. However, given an higherrelative number of triggered events, the absolute number of reconstructed eventswas comparable with other remaining cases considering low altitude thick clouds,and even higher considering middle altitude thick clouds. An average of 29% oftriggered events could be reconstructed. An improvement of the reconstructionperformances were done working on the part of the reconstruction procedurecarried out by the pattern recognition algorithm. Such as optimization wasfulfilled tuning reconstruction parameters such as the detector pixel dimensionat ground and changing the initial condition concerning the identification of thesignal inside the data. The fractions of reconstructed events were estimatedonce again using the algorithm optimized version. In each case the fraction ofreconstructed events improved without a substantial deterioration of the averageenergy resolution, bringing the final fraction to over 49%, with high margins forfurther improvements.

89

Chapter 5

Integration and calibrationaspects of the EUSO-SPBPDM

As seen in the chapter 3 one of the aspects still to be defined is the globalefficiency of the detector. Therefore, the aim of this chapter is to outline theperformances of a sample of 100 multi-anode PMTs (HAMAMATSU R11265-113-M64 MOD2), delivered and tested at the APC (Astroparticule et Cosmolo-gie Laboratoire). On such sample a procedure of sorting was carried out inorder to find the best candidates. Best performing PMTs will be assembled onthe EUSO-SPB PDM.

5.1 PMT sorting for the EUSO-SPB PDM

With sorting procedure one refers to the analysis operated on the already men-tioned sample of 100 PMTs in order to evaluate the best of them in terms ofefficiency and gain performances. For EUSO-SPB, to have similar efficiencies isimportant, since the trigger will set the thresholds based on the pixel with thehighest count A second level sorting was done choosing between the sample ofthe preselected good-efficiency PMTs the ones with the best average gain.

5.1.1 Photon detection

In this section are shown the main concepts useful to understand the opera-tional use of PMT as done for the following measures and analysis. A generalknowledge of the functioning of a PMT is assumed but for a detailed descrip-tion one can refer to [78]. All tested PMTs operated in single photoelectroncounting, which means that photons arrived at the photocathode at a rate suchthat only one or zero photoelectrons are produced at a time, and so the anodepulse from each collected photoelectron can be separated. In this case the an-ode signal can be sent into a discriminator circuit with a threshold set to rejectnoise, and each count from the discriminator then corresponds to the arrivaland conversion of a photon at the photocathode. To operate in single photo-

90

electron counting mode, the probability that more than two photoelectrons areemitted from the photocathode within the time-resolution of the measurement(called gate) must be negligible. For each gate, that must be referred as a singleevent, the number photoelectrons emitted follows the Poisson distribution andthe condition of single photoelectron can be satisfied when 10 % of total countedevents provide a signal. The charge spectrum in these conditions is known as asingle photoelectron spectrum. An example of a single photoelectron spectrumtaken by integrating the charge during a fixed gate is shown in Fig.5.1. Onthe left hand side of the spectrum is the peak corresponding to events duringwhich no photoelectron was collected. This peak is often called the pedestal. Intheory, the pedestal will be centered on zero charge with noise fluctuations, butin practice the pedestal has a non-zero mean due to low-charge dark current,such as leakage current, and to the readout electronics offset [78]. The peakon the right is the single photoelectron peak. Several information about thePMT like gain and efficiency can be extracted from the spectrum. In additionto measuring single photoelectron spectra of PMTs by taking a charge or pulseheight spectrum, the equivalent information can be obtained in a counting ex-periment. To do this the anode signal of the PMT is sent through an integratingpreamplifier and then through a discriminator circuit. The discriminator givesa output pulse whenever it receives an input pulse with a voltage over a setthreshold. The plot of the count rate vs threshold is known as an S-curve. Asthe single photoelectron spectrum is the charge distribution of the PMT signal,and the S-curve is simply its cumulative distribution function, and so the singlephotoelectron spectrum can be recovered by derivation.

Figure 5.1: An example of a single photoelectron spectrum. The spectrum isshown as a histogram of the number of pulses with a given charge, in counts,returned by a charge-to-digital converter. The first peak is the pedestal, corre-sponding to pulses in which no pe are collected. The second peak, on the right,corresponds to pulses in which one pe is collected. The gain is the differencebetween the means of the zero pe and one pe peaks (shown by red markers),while the efficiency of the PMT is proportional to the surface of the one pe peak.

91

5.1.2 Experimental setup

The measurement environment was an essential element. Working with PMTsrequires that the immediate setup be placed inside a black box to control theexposure of the high-gain PMT to light. This was to both ensure that thePMT was not damaged by over exposure and that the signal-to-backgroundratio of the measurement was good. A scheme of the black box with the mainelements of the setup for the measures is shown in Fig.5.2. In the current setupa continuous light of 398 nm emitted by a LED was sent concurrently to thePMT and a NIST, a photodiode acting as a reference detector. The LED andthe NIST were placed on the lower and the upper port of an integrating sphere,while the light spotting the PMT exit from the left port. The basic function ofthe integrating sphere was to spatially integrate radiant flux acting as a stablewell-characterized splitter. As first approximation, the light could be considereduniform on the whole surface of the PMT. All the 64 PMT pixels were drivenby a Multi-channel High Voltage Power Supply that provided for each PMTstage dynod increasing voltage values. Signals coming from the 64 PMT pixelswere read by an ASIC circuit and analog to digital converted by a DAC hostedon the Asic Board. Such board was connected via USB to a Computer with aLab-View software that could be run for the data acquisition.

LABVIEWSOFTWARE

MULTI CHANNEL HVPS

USB CABLE

PMTNIST

INTEGRATINGSPHERE

LIGHTTEST BOARD

BLACK BOX

LED

VPSAMMETER

Figure 5.2: Scheme of the setup used for the PMT sorting.

5.1.3 Data acquisition

The data acquisition consisted into getting for each PMTs the correspondingS-curve (see an example in Fig.5.3). During the acquisition the threshold ofthe discriminator hosted in the Asic-board was increased by a 10 DAC stepevery 1000 GTUs. The S-curve analysis allowed to extract information aboutthe efficiency and gain of each PMT.

92

Figure 5.3: Example of S-curve containing a total amount of 64 curves, one foreach pixel of the PMT. In the region between 250 DAC and 450 DAC is presentthe so called inflection point, corresponding to the DAC value used to calculatethe gain of the PMT. Between 500 and 650 DAC one can see the region calledplateau which average count rate value is used to extimate the efficiency of thePMT. The right part is dominted by the pedestal.

Efficiency

The efficiency of each pixel has to be considered relative and was given in ar-bitrary units of the height of the plateau at DAC 600, divided by the valueof power of the light spotting the NIST. From the S-curves analysis two maincategories of PMTs were clearly seen, as shown in the two representative exam-ples in Fig.5.4. In the bottom plot is shown an S-curve of a PMT where theplateaus of each pixel (measured at DAC 600) were well grouped, while in thetop plot is shown an S-curve were two groups are clearly visible. As alreadymentioned, the height of the plateu is representative of the efficiency. Since thethresholds of the trigger for each PMT are expected to be chosen referring tothe counting rate of the most efficient pixel, for EUSO-SPB is important to havesimilar efficiencies among pixels. Therefore, as first PMTs where the height ofplateaus were well grouped were selected. In other words looking to all the100 efficiency distribution reported in Fig.5.5 only one-peaked and reasonablynarrow distributions were chosen.

Gain

An amount of about 50 PMTs over the 100 tested was found with similar chan-nels efficiency. Such sub sample was then sorted by PMT average gain. Forthe gain calculation the S-curves were fitted in order to find for each one theDAC value corresponding to the inflection point. Then gains are calculated inunits of 106 using a DAC calibration curve previously calculated, fitted by afourth degree polynomial. The gains obtained are absolute ones. In fact thesame ASIC was used for all 100 PMTs and therefore the pedestals are all at thesame DAC.

93

0 100 200 300 400 500 600 700 800DAC

0

2000

4000

6000

8000

10000

12000

14000

phe/1

000 G

TU

s

S-curves PMT1789

0 100 200 300 400 500 600 700 800DAC

0

2000

4000

6000

8000

10000

12000

14000

phe/1

000 G

TU

s

S-curves PMT1792

Figure 5.4: Examples of S-curves related to the two main groups of PMTs foundout from the analysis. Top panel. S-curves taken from a PMT with non uniformplateaus from pixel to pixel. Bottom panel. S-curves taken from a PMT whichpixels present uniform efficiency.

94

12

34

56

78

9051015202530

1458

.0

12

34

56

78

9051015202530

1459

.0

12

34

56

78

9051015202530

1461

.0

12

34

56

78

9051015202530

1462

.0

12

34

56

78

9051015202530

1463

.0

12

34

56

78

9051015202530

1465

.0

12

34

56

78

9051015202530

1468

.0

12

34

56

78

9051015202530

1469

.0

12

34

56

78

9051015202530

1471

.0

12

34

56

78

9051015202530

1492

.0

12

34

56

78

9051015202530

1494

.0

12

34

56

78

9051015202530

1495

.0

12

34

56

78

9051015202530

1496

.0

12

34

56

78

9051015202530

1500

.0

12

34

56

78

9051015202530

1501

.0

12

34

56

78

9051015202530

1502

.0

12

34

56

78

9051015202530

1503

.0

12

34

56

78

9051015202530

1505

.0

12

34

56

78

9051015202530

1507

.0

12

34

56

78

9051015202530

1508

.0

12

34

56

78

9051015202530

1509

.0

12

34

56

78

9051015202530

1510

.0

12

34

56

78

9051015202530

1511

.0

12

34

56

78

9051015202530

1514

.0

12

34

56

78

9051015202530

1516

.0

12

34

56

78

9051015202530

1517

.0

12

34

56

78

9051015202530

1518

.0

12

34

56

78

9051015202530

1519

.0

12

34

56

78

9051015202530

1532

.0

12

34

56

78

9051015202530

1536

.0

12

34

56

78

9051015202530

1538

.0

12

34

56

78

9051015202530

1541

.0

12

34

56

78

9051015202530

1543

.0

12

34

56

78

9051015202530

1544

.0

12

34

56

78

9051015202530

1547

.0

12

34

56

78

9051015202530

1548

.0

12

34

56

78

9051015202530

1561

.0

12

34

56

78

9051015202530

1562

.0

12

34

56

78

9051015202530

1563

.0

12

34

56

78

9051015202530

1564

.0

12

34

56

78

9051015202530

1565

.0

12

34

56

78

9051015202530

1566

.0

12

34

56

78

9051015202530

1567

.0

12

34

56

78

9051015202530

1593

.0

12

34

56

78

9051015202530

1595

.0

12

34

56

78

9051015202530

1601

.0

12

34

56

78

9051015202530

1604

.0

12

34

56

78

9051015202530

1608

.0

12

34

56

78

9051015202530

1610

.0

12

34

56

78

9051015202530

1612

.0

95

12

34

56

78

9051015202530

1748

.0

12

34

56

78

9051015202530

1749

.0

12

34

56

78

9051015202530

1753

.0

12

34

56

78

9051015202530

1756

.0

12

34

56

78

9051015202530

1757

.0

12

34

56

78

9051015202530

1762

.0

12

34

56

78

9051015202530

1764

.0

12

34

56

78

9051015202530

1765

.0

12

34

56

78

9051015202530

1766

.0

12

34

56

78

9051015202530

1767

.0

12

34

56

78

9051015202530

1768

.0

12

34

56

78

9051015202530

1769

.0

12

34

56

78

9051015202530

1771

.0

12

34

56

78

9051015202530

1772

.0

12

34

56

78

9051015202530

1773

.0

12

34

56

78

9051015202530

1784

.0

12

34

56

78

9051015202530

1785

.0

12

34

56

78

9051015202530

1786

.0

12

34

56

78

9051015202530

1787

.0

12

34

56

78

9051015202530

1788

.0

12

34

56

78

9051015202530

1789

.0

12

34

56

78

9051015202530

1791

.0

12

34

56

78

9051015202530

1792

.0

12

34

56

78

9051015202530

1793

.0

12

34

56

78

9051015202530

1801

.0

12

34

56

78

9051015202530

1803

.0

12

34

56

78

9051015202530

1804

.0

12

34

56

78

9051015202530

1805

.0

12

34

56

78

9051015202530

1806

.0

12

34

56

78

9051015202530

1807

.0

12

34

56

78

9051015202530

1808

.0

12

34

56

78

9051015202530

1809

.0

12

34

56

78

9051015202530

1810

.0

12

34

56

78

9051015202530

1811

.0

12

34

56

78

9051015202530

1812

.0

12

34

56

78

9051015202530

1825

.0

12

34

56

78

9051015202530

1826

.0

12

34

56

78

9051015202530

1827

.0

12

34

56

78

9051015202530

1828

.0

12

34

56

78

9051015202530

1829

.0

12

34

56

78

9051015202530

1830

.0

12

34

56

78

9051015202530

1832

.0

12

34

56

78

9051015202530

1833

.0

12

34

56

78

9051015202530

1834

.0

12

34

56

78

9051015202530

1836

.0

12

34

56

78

9051015202530

1846

.0

12

34

56

78

9051015202530

1847

.0

12

34

56

78

9051015202530

1849

.0

12

34

56

78

9051015202530

1850

.0

12

34

56

78

9051015202530

0.0

Figure 5.5: Distributions of the pixels efficiency of each of the 100 tested PMTs.

96

Table 5.1 shows averaged gains of a randomly selected sub sample of the 100tested PMTs. The gain of differents PMTs could differs for a factor ∼ 3 (forexample comparing the PMT number 1749 to the PMT number 1771). After

PMT number Gain1748 3.921749 4.771753 3.251756 2.291757 2.731762 4.411764 3.251765 4.211766 3.391767 3.491768 2.231769 3.481771 1.941772 2.601773 3.171784 2.631785 4.491786 3.701787 2.831788 2.67

Table 5.1: Average gain of a sample of sorted PMTs expressed in unit of 106

sorting all the PMTs by gain it was possible to define their position on the PDMfocal surface. The best gains PMTs were put in the center, while those withless gain at the corners.

5.2 Trigger response for a PDM with non uni-form efficiency

The relative efficiencies values measured for each pixel of the PMTs selectedfor the PDM integration were then implemented in a map which consideredthe right PMT order on the focal surface as decided using criteria mentionedabove. From this map two further maps of variable background and variablepixel quantum efficiency could be recovered. Such maps were then implementedin ESAF and used to simulate the response of a non uniform efficiency detector(unlike simulations performed in chapter 3), where with non uniform one meanta PDM in which to each pixel is associated its own value of quantum efficiencyand average background rate. The average background rate associated to eachpixel was chosen by matching a specific value of average background to theaverage relative efficiency value calculated taking into consideration all pixelsof the entire PDM. The background of each pixel was consequently scaled withthe pixel relative efficiency. A similar procedure was adopted to calculate thequantum efficiency related to each pixel. A spectrum simulation for a non

97

uniform detector was performed in clear sky condition, assuming an averagedetection efficiency of 0.1 (corresponding to an overall efficiency of the detectorof ∼ 0.05) and an average background rate of 0.25 counts/µs/pixel. The resultwas then compared to the one obtained assuming an uniform detector withthe same parameters (see table 5.2). One can see that having a non uniformdetector, at least for such a set of simulation parameters, does not significantlyaffect the number of triggered events. Triggered spectra for both cases areplotted in Fig.5.6.

Triggered events Triggered events(uniform detector) (non uniform detector)

9.9 ± 0.8 7.1 ± 0.7

Table 5.2: Number of triggered events comparing a non uniform to an uniformdetector. An average overall efficiency of 0.05 and an averaged background of0.25 counts/µs/pixel are assumed.

log10(E/eV)17.8 18 18.2 18.4 18.6 18.8 19 19.20

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

Balloon triggered spectrum

non uniform detector

uniform detector

Figure 5.6: Comparison between triggered spectra simulated assuming a uniformand a non uniform detector.

5.3 Conclusions

A sorting procedure over a sample of 100 multi-anode PMTs was carried out. Afirst sorting was done by selecting a sub sample of PMTs in whose all 64 pixelsshowed the most uniform efficiency. Then a second sorting was performed byconsidering the best gain PMTs of such sub sample. In this way it was possibleto select the 36 PMTs wich will be integrated on the EUSO-SPB PDM definingtheir position on the the focal surface. Such configuration was then implementedin the ESAF software to simulate the effect of a PDM with non uniform response.The number of triggered events in the non uniform configuration was found outnot to change significantly. Simulating a non uniform detector was a crucial stepbecause allowed to evaluate how the number of triggered events could changein case of a more realistic detector configuration.

98

Chapter 6

Conclusions

The work developed in this thesis focused on the assessment, through simula-tions, of the first level trigger and reconstruction performances of EUSO-SPB.In chapter 3, a comparison between three trigger methods (method 1, method2a and method 2b) was carried out. The three methods were tested simulat-ing spectra in different cloudy coverage conditions and considering an averagebackground rate of 0.25 counts/µs/pixel, corresponding to the value measuredduring the first EUSO-Balloon flight in August 2014. Since the developmentof the detector is still ongoing, the overall efficiency of the detector is nowa-days an unknown parameter. Therefore the analysis was performed assumingan overall efficiency of the detector of 0.05 (corresponding to the value of effi-ciency occurred for the first EUSO-Balloon detector) which could be seen as aconservative case, and a more optimistic case corresponding to an overall effi-ciency of the detector of 0.15. Switching from a low performances to an highperformances detector, the number of triggered events increased by a factor ∼10 both in clear sky and in cloud coverage conditions. As a conclusion, method1 was found out to be the best candidate because it was the most suitable totrigger events in thick-cloud conditions which is the most common situationthat will occur on the EUSO-SPB trajectory. Moreover, method 1 allowed thedetection of Cherenkov-like events on the top of clouds that can not be observedby ground based experiments. A further analysis done with a low performancedetector shown that in case of an higher background of 0.5 counts/µs/pixel therelative number of triggered events dropped further by a factor of ∼ 2-3 withan expected rate of 0.5 events per day in case of thick cloud condition. Testswere also performed simulating the detector response considering GTUs withduration less than 2.5 µs. The GTU of 2.5 µs was in general the one performingworse. The comparison carried out between the triggered spectra could be seenas relative, since the aims of this analysis was to show the differences amongthe spectra obtained using different simulation parameters. An estimation ofan absolute number of triggered events is also provided. To do that triggeredspectra obtained in different cloudy conditions were weighted with the estimatedpercentages of different cloudy coverages occurred on the trajectory of the SuperPressure Balloon flown in 2015 [75].

Backgrounds of 0.25 counts/µs/pixel and 0.5 counts/µs/pixel were assumedrespectively for clear sky and cloudy atmosphere cases (independently from theOD). A low performances detector was also considered. Low, middle and high

99

clouds were respectively associated to a simulated altitude of 2, 5 and 10 km.The estimation of the mission duty cycle was obtained referring to table

6.1, taken from [54]. As one can see, the 2017 March/April dark period at thelatitude of Wanaka, NZ includes 18 nights with at least 2.5 hours of moon downdark time for a total of 118 hours. These hours would be the primary targetedobservation times for EUSO-SPB.

date UTC hrs dark start UTc stop UTC2017 Mar 19 2:08 8:38 (twilight) 10:46 (moon 62%)2017 Mar 20 2:52* 8:36 (twilight) 11:28 (moon 52%)2017 Mar 21 3:40* 8:34 (twilight) 12:15 (moon 42%)2017 Mar 22 4:34* 8:32 (twilight) 13:07 (moon 32%)2017 Mar 23 5:34* 8:30 (twilight) 14:04 (moon 23%)2017 Mar 24 6:38* 8:28 (twilight) 15:06 (moon 14%)2017 Mar 25 7:46* 8:26 (twilight) 16:12 (moon 7%)2017 Mar 26 8:51* 8:24 (twilight) 17:15 (twilight)2017 Mar 27 8:54* 8:22 (twilight) 17:16 (twilight)2017 Mar 28 8:57* 8:20 (twilight) 17:18 (twilight)2017 Mar 29 9:01* 8:18 (twilight) 17:19 (twilight)2017 Mar 30 9:04* 8:26 (twilight) 17:20 (twilight)2017 Mar 31 8:40* 8:41 (moon 14%) 17:22 (twilight)2017 Apr 1 7:57* 9:25 (moon 24%) 17:23 (twilight)2017 Apr 2 7:08* 10:16 (moon 35%) 17:24 (twilight)2017 Apr 3 6:12* 11:13 (moon 46%) 17:25 (twilight)2017 Apr 4 5:12* 12:14 (moon 58%) 17:27 (twilight)2017 Apr 5 4:09* 13:19 (moon 69%) 17:28 (twilight)2017 Apr 6 3:04* 114:24 (moon 79%) 17:29 (twilight)2017 Apr 7 2:00* 15:30 (moon 87%) 17:31 (twilight)

Table 6.1: Example of expected hours of operation for a dark period at 45◦ S(see text). This dark period includes 18 days, indicated by an asterisk, with atleast 2.5 hours of moon down darkness for a total of 118 hours.

In the calculation of the number of triggered events it was also assumed afactor 0.9 for the detector operating time, corresponding to the fraction of timein which the detector is operative. Two weighted spectra assuming OD = 5and OD = 1 were calculated. The number of expected triggered events variedbetween ∼ 7.3 (assuming OD = 5) and ∼ 2.6 (assuming OD = 1), showing thateven in conservative conditions, there will be a chance to detect a few events.

In the framework of this thesis was also operated a procedure of PMTs sort-ing over a sample of 100 MAPMTs, performed at the APC (see chapter 5).The best 36 performing PMTs were selected for the integration of the EUSO-SPB PDM. Thanks to this measurement it was possible to implement in thesimulation framework a map of variable detection efficiency and background,simulating the response of a non uniform detector as for the real EUSO-SPBPDM. A simulation performed in clear sky condition, assuming an average back-ground of 0.25 counts/µs/pixel and a low performance detector showed that thenumber of triggered events, compared to the one obtained for a uniform de-tector, was reduced by ∼ 30%. The reconstruction procedure was applied totriggered spectra obtained in chapter 3. Using a non optimized version of the

100

pattern recognition algorithm, an average of ∼ 29% of triggered events could bereconstructed. An improvement of the reconstruction performance was fulfilledtuning parameters such as the detector pixel dimension at ground and changingthe initial condition concerning the identification of the signal inside the datacarried out by the pattern recognition algorithm. The optimized algorithm ver-sion was run on the triggered spectra providing an improvement of the fractionsof reconstructed events up to ∼ 49% without a substantial deterioration of theaverage energy resolution and with high margins for further improvements.

101

Bibliography

[1] Y. Takahashi - JEM-EUSO Collaboration, The JEM-EUSO Mission,NJPh11, 065009/1-21, 2009.

[2] M. Casolino, J.H. Adams, M.E. Bertaina, et al., Detecting ultra-high energycosmic rays from space with unprecedented acceptance: objectives and designof the JEM-EUSO mission, Astrophysics and Space Science Transactions 7,477-482, 2011.

[3] The Pierre Auger Collaboration, Observation of the suppression of the fluxof cosmic rays above 4 × 1019 eV.

[4] M. Bertaina, P. Bobik, F. Fenu, K. Shinozaki on behalf of the JEM-EUSOCollaboration, JEM-EUSO observational technique and exposure.

[5] M. Bertaina et al., Performance and air-shower reconstruction techniquesfor the JEM-EUSO mission.

[6] M. Bertaina et al., Requirements and Expected Performances of the JEM-EUSO mission.

[7] T. Mernik et al., Proc. 33rd ICRC, 2013 (ID777).

[8] Y. Takizawa et al. - JEM-EUSO collaboration, The TA-EUSO and EUSO-Balloon optics designs, Proceedings of this conference (id832).

[9] Y. Hachisu et al. - JEM-EUSO collaboration, Manufacturing of the TA-EUSO and the EUSO-Balloon lenses, Proceedings of this conference (id1040).

[10] S. Bacholle, C. Blaksley, P. Gorodetzky, J. Karczmarczyk, J. Szabelski, ACockcroft-Walton High-Voltage Power Supply for the EUSO Instruments.

[11] Ahmad, S., Barrillon, P., Blin-Bondil, S., Dagoret-Campagne, S., de LaTaille, C., Dulucq, F., Kawasaki, Y., Hirokazu, I.: Spaciroc: a rad-hardfront-end readout chip for the jem-euso telescope. J. Instrum. 5(12), C12012(2010).

[12] Ahmad, S., Barrillon, P., Blin-Bondil, S., Dagoret-Campagne, S., De laTaille, C., Dulucq, F., Martin- Chassard, G., Kawasaki, Y., Miyamoto, H.,Ikeda, H., Iguchi, T., Kajino, F.: Spaciroc: A front-end readout asic forthe jem-euso observatory. In: in Nuclear Science Symposium and MedicalImaging Conference (NSS/MIC), 2011 IEEE, pp. 699–703 (2011).

102

[13] Ahmad, S., Barrillon, P., Blin-Bondil, S., Dagoret-Campagne, S., Taille,C.D., Dulucq, F., Martin-Chassard, G., Kawasaki, Y., Miyamoto, H., Ikeda,H., Iguchi, T., Kajino, F.: Spaciroc: A front-end readout ASIC for jem-eusocosmic ray observatory. Physics Procedia. In: Proceedings of the 2nd Inter-ational Conference on Technology and Instrumentation in Particle Physics(TIPP 2011), vol. 37, no. 0, pp. 1600–1607 (2012).

[14] Ahmad, S.: Developpement et realisation d’un circuit de micro electron-ique pour le detecteur spatial de rayons cosmiques JEM-EUSO. PhD thesis(2012).

[15] Hiroko Miyamoto, performance of a front end ASIC for JEM-EUSO.

[16] J. Bayer et al., Proc. 33rd ICRC, 2013 (ID0432).

[17] Scotti, V., Osteria, G., The JEM-EUSO Collaboration: The JEM-EUSOtime synchronization system. Nucl. Inst. Methods Phys. Res. A. 718, 248–250(2013).

[18] Shinozaki, K., Saez, G., Guzman, A.: For the JEM-EUSO Collaboration.JEM-EUSO observation in cloudy conditions.

[19] Saez - Cano, G., et al.: For the JEM-EUSO collaboration. Observation ofultra-high energy cosmic rays in cloudy conditions by the JEM-EUSO spaceobservatory. In: Proceedings of 32nd International Cosmic Ray Conference(ICRC), vol. 3, pp. 231. Beijing. (Preprint) arXiv:1204.5065 (2011).

[20] Toscano S., Morales de los Rios J. A., Rodriguez Frias M. D., NeronovA., Wada S.: The JEM- EUSO Collaboration: The atmospheric monitoringsystem of the JEM-EUSO instrument. Exp. Astron. (2014).

[21] The JEM-EUSO collaboration, The EUSO-Balloon mission. In: Proceed-ings of 34th International Cosmic Ray Conference (ICRC) #322 (2015).

[22] The JEM-EUSO Collaboration, Preliminary results from the EUSO-Balloon flight. In: Proceedings of 34th International Cosmic Ray Conference(ICRC) #358 (2015).

[23] Blaksley, C., et al., Proc. 34 th ICRC, 2015, A Cockcroft-Walton High-Voltage Power Supply for the EUSO Instruments.

[24] Takizawa, Y., et al., Proc. 33 rd ICRC, 2013 (ID832), The TA-EUSO andEUSO-Balloon optics designs.

[25] Takizawa, Y., et al., Proc. 33 rd ICRC, 2013 (ID1040), Manufacturing ofthe TA-EUSO and the EUSO-Balloon lenses.

[26] Osteria, G. et al., Proc. 34 th ICRC, 2015, The Data Processor System ofEUSO-Balloon: in flight performance.

[27] Rodriguez Firas, M.D., et al., Proc. 34 th ICRC, 2015, The Infrared Cameraon EUSO-Balloon.

[28] Fernandez, J., et al., Proc. 34th ICRC, 2015, Performances of the InfraredCamera on EUSO-Balloon.

103

[29] A. Merino et al. (JEM-EUSO Coll.), Proc. 34th Int. Cosmic Ray Conf.(Den Haag), #0979 (2015).

[30] I. Tabone et al. (JEM-EUSO Coll.), Proc. 34th Int. Cosmic Ray Conf. (DenHaag), #0889 (2015).

[31] J.H. Adams Jr. et al. (JEM-EUSO Coll.), Proc. 34th Int. Cosmic Ray Conf.(Den Haag), #0429 (2015).

[32] J. Eser et al. (JEM-EUSO Coll.), Proc. 34th Int. Cosmic Ray Conf. (DenHaag), #0860 (2015).

[33] C. Catalano et al. (JEM-EUSO Coll.), Proc. 34th Int. Cosmic Ray Conf.(Den Haag), #0717 (2015).

[34] J. Fernandez Soriano et al. (JEM-EUSO Coll.), Proc. 34th Int. Cosmic RayConf. (Den Haag), #1008 (2015).

[35] J.H. Adams Jr. et al. - JEM-EUSO Collaboration, Astroparticle Physics44 (2013) 76.

[36] S. Mackovjak et al. (JEM-EUSO Coll.), Proc. 34th Int. Cosmic Ray Conf.(Den Haag), #1302 (2015).

[37] J. Eser et al. (JEM-EUSO Coll.), Proc. 34th Int. Cosmic Ray Conf. (DenHaag), #0860 (2015).

[38] G. Suino et al. (JEM-EUSO Coll.), Proc. 34th Int. Cosmic Ray Conf. (DenHaag), #0925 (2015).

[39] S. Giarruso et al., Proc. 28th Int. Cosmic Ray Conf. (Tsukuba), 2 (2003)849.

[40] M. Casolino, Marco Ricci, J. Adams, P. Barillon, J. Bayer, J. Belz, D.Bergman, M. Bertaina, F. Bisconti, M.J. Christl, A. Cummings, G. Distratis,A. Ebersoldt, T. Ebisuzaki, J. Eser, F. Fenu, T. Fujii, M. Fukushima, D.Gottschall, A. Haungs, P. Hunt, D. Ikeda, A. Jung, F. Kajino, Y. Kawasaki,O. Larsson, J.N. Matthews, T. Nonaka, S. Ogio, G. Osteria, P. Picozza 1,15, L.W. Piotrowski 1 , Z. Plebaniak 18 , H. Sagawa 11 , N. Sakurai 10 ,V. Scotti, H. Shin, T. Shibata, K. Shinozaki, J.D. Smith, P. Sokolsky, M.Takeda 11 , Y. Takizawa 1 , Y. Tameda 7 , C. Tenzer 5 , S.B. Thomas 6 , G.B.Thomson 6 , H. Tokuno, T. Tomida, Y. Tsunesada, L. Wiencke and the JEM-EUSO collaboration, EUSO-TA, a ground telescope at the Telescope Arrayto test JEM-EUSO detector performance. In: 34th International Cosmic RayConference (ICRC) #636 (2015).

[41] H.Kawai et al.,Nuclear Physics B (Proc. Suppl.) 175-176 (2008) 221-226.

[42] H. Prieto el al., 33rd International Cosmic Ray Conference Proceedings,arxiv:1307.7071, 95.

[43] H. Miyamoto et al., 33rd International Cosmic Ray Conference Proceed-ings, arxiv:1307.7071, 103.

104

[44] J. Bayer et al., 33rd International Cosmic Ray Conference Proceedings,arxiv:1307.7071, 99.

[45] F. Kajino et al., 33rd International Cosmic Ray Conference Proceedings,arxiv:1307.7071, 11.

[46] A. Segreto et al., Proceedings of 32nd International Cosmic Ray Confer-ence.

[47] Y. Takahashi et al., AIP Conf. Proc. 1367, 157 (2011).

[48] L.M. Barbier et al., Astropart. Phys., 22 (2005) 439.

[49] O. Catalano et al., Nucl. Instr. Methods A, 480(2-3) (2002) 547.

[50] M. Bertaina et al. - JEM-EUSO Collaboration, Proc. 34th ICRC(2015).

[51] A. De Rujula and S.L. Glashow, Nature 312 (1984) 734.

[52] T. Ebisuzaki et al., Acta Astronautica 112 (2015) 102.

[53] A. Haungs et al. - JEM-EUSO Collaboration, Proc. 34th ICRC, 2015, theseproceedings.

[54] L. Wiencke et all., EUSO-Balloon mission to record extensive air show-ers from near space, Proc. 34th International Cosmic Ray Conference #631(2015).

[55] J. S. Fenn, Estimating the cosmic ray extensive air shower detection ratefor the EUSO-Super Pressure Balloon mission, MSc Dissertation, ColoradoSchool of Mines (2015). (http://hdl.handle.net/11124/17150).

[56] T. Paul et al., (JEM-EUSO collaboration), A new design for simulationand reconstruction software for the JEM-EUSO mission in Proc. 34th ICRC,(The Hague), (2015).

[57] The Pierre Auger Collaboration, J. Abraham et al., Measurement of theEnergy Spectrum of Cosmic Rays Above 10 18 eV using the Pierre AugerObservatory, Physics Letters B, 685(45):239 – 246, (2010). (arXiv:1002.1975[astro-ph.HE]).

[58] S. Bacholle et al., (JEM-EUSO collaboration), EUSO-Balloon trigger ef-ficiency in preparation of a long duration flight in Proc. 34th ICRC, (TheHague), (2015).

[59] F. Fenu et al., (JEM-EUSO collaboration), The ESAF simulation frame-work for the JEM-EUSO mission in Proc. 32rd ICRC, (Beijing),Vol 3, #0592(2011). (DOI: 10.7529/ICRC 2011/V03/0592).

[60] http://root.cern.ch/.

[61] D. Naumov, SLAST, Shower Light Attenuated to the Space Telescope,EUSO mission internal document (2003).

[62] T. Pierog et al., First Results of Fast One–dimensional Hybrid Simulationof EAS Using CONEX, arXiv:astro–ph/0411260v1 (2004).

105

[63] M. Alekseeva et al., Extensive Air Shower Simulation Program CONEX:Matching Monte Carlo and Numerical Methods, proceeding of the Int. CosmicRay Conference (Lodz) (2009).

[64] D. Heck et al., CORSIKA: A Monte Carlo Code to Simulate Extensive AirShowers, Wissenschaftliche Berichte–Forschungszentrum Karlsruhe (1998).

[65] National Oceanic Atmospheric Administration et al., U.S standard atmo-sphere, 1976, public document (1976).

[66] M. Nagano, New measurement on photon yields from air and the ap-plication to the energy estimation of primary cosmic rays, arXiv:astro–ph/0406474v2 (2004).

[67] F. Kakimoto, A measurement of the air fluorescence yield, Nucl. Instr. andMet. in Phys. Res., vol. A 372, 527–533 (1996).

[68] Air Force Geophysics Laboratory, Users guide to LOWTRAN 7, publicdocument (1988).

[69] S. Bitkemerova, Integration of ESAF and Geant4 for simulation of spacebased telescopes, proceeding of the Int. Cosmic Ray Conference (Lodz) (2009).

[70] M. Casolino, Data Acquisition System of the JEM–EUSO project, prooced-ing of the Int. Cosmic Ray Conference (Beijing) (2011).

[71] S. Bottai, Some algorithm for direction reconstruction in EUSO and opti-mization of pixel size, EUSO mission internal document (2002).

[72] The JEM–EUSO collaboration, ”An evaluation of the exposure in nadirobservation of the JEM–EUSO mission”, Astrop. Phys. Vol. 44–76 (2012).

[73] The Pierre Auger Collaboration: A.Aab et all., Contribution to the 33rdInternational Cosmic Ray Conference (ICRC 2013).

[74] The JEM–EUSO collaboration, ”An evaluation of the exposure in nadirobservation of the JEM–EUSO mission”, Astrop. Phys. Vol. 44–76 (2012).

[75] Andrea Veneziani, results of a ongoing stage at ARPA (Torino,Italy).

[76] M. Bertaina et all, Performance and air-shower reconstruction tecniquesfor the JEM-EUSO mission.

[77] F. Fenu, A Simulation Study of the JEM–EUSO Mission for the Detectionof Ultra–High Energy Cosmic Rays, Phd Thesis.

[78] Carl J. Blaksley, Characterization of Multi-Anode Photomultiplier Tubesfor JEM-EUSO, Phd Thesis

106