Development of a Cosmic Ray Telescope

Bernardo D’Almeida Maurício do Rosário

Thesis to obtain the Master of Science Degree in

Engineering Physics

Supervisor(s): Prof. Pedro Jorge dos Santos Assis

Examination Committee

Chairperson: Prof. Pedro Miguel Felix BrogueiraSupervisor: Prof. Pedro Jorge dos Santos Assis

Member of the Committee: Prof. Fernando José de Carvalho BarãoProf. Bernardo António Neto Gomes Baptista Tomé

April 2016

ii

Nature is not mute, it is Man that is deaf.

Terence McKenna

iv

Acknowledgments

Em primeiro lugar gostava de agradecer ao Pedro Assis por toda a ajuda e orientacao ao longo deste

projecto. Agradeco a todas as pessoas do LIP que, de alguma forma, colaboraram para a conclusao

deste trabalho. Ao Luis pela sua incansavel prestabilidade e simpatia. A todos os amigos e familia pelo

o apoio e amor incondicional. Um obrigado especial a minha avo por me inspirar.

v

vi

Resumo

Esta tese baseou-se no desenvolvimento e estudo de um prototipo de um detector compacto de muoes

(CMuD). Este foi desenvolvido com o intuito de aproximar jovens alunos do campo da fisica de as-

troparticulas. Ao desenvolver um detector simples, robusto, barato e seguro abre-se a possibilidade de

jovens estudantes criarem a suas proprias experiencias de fısica de particulas. Este detector de muoes

e composto por um cintilador de plastico com dimensoes 20× 20 cm2 atravessado por fibras opticas

que guiam a luz para um fotomultiplicador de silicio. Estudaram-se algumas caracterısticas deste fo-

tosensor como o ganho e a corrente escura. De seguida mediu-se a eficencia e a pureza do detector

CMuD. Estas caracterısticas foram tambem medidas com este detector em auto-coincidencia usando

dois fotomultiplicadores para ler um CMuD e em coincidencias de dois CMuDs.

Palavras-chave: Detector de Cintilacao, Fotomultiplicador de Silicio, Eficiencia, Pureza, Co-

incidencias, Auto-Coincidencia.

vii

viii

Abstract

This project was focused on the study and development of the Compact Muon Detector (CMuD) proto-

type. By devising a simple, compact, affordable and safe particle detector we expect to bridge the gap

between students and the field of astroparticle physics. Providing an asset to the academic and personal

formation of students. The CMuD is made of a 20×20 cm2 plastic scintillator bisected by 9 wavelength

shifting fibers that converge to a silicon photomultiplier. A study of the main features of the SiPM was

performed to determine its gain and dark current rate. The measurement of the efficiency and purity

of the muon detector was performed. These characteristics were also determined with the detector in

self-coincidence, using two photomultipliers to read the CMuD signal and in coincidence between two

CMuD.

Keywords: Scintillation Detector, Silicon Photomultipler, Efficiency, Purity, Coincidence, Self-

Coincidence

ix

x

Contents

Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v

Resumo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii

List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv

1 Introduction 1

1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Cosmic Rays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2.1 Short History of Cosmic Rays Detection . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Overview of Scintillation Detectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.3.1 Scintillation Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.3.2 Photo-detectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2 Compact Muon Detector (CMuD) 19

2.1 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

2.2 Scintillator EJ-200 & Y-11(200) Wavelength Shifting Fibers . . . . . . . . . . . . . . . . . 20

2.3 Silicon Photomultiplier - MPPC Hamamatsu S12572-50P . . . . . . . . . . . . . . . . . . 22

2.4 Signal Estimate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.4.1 Energy Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.4.2 Light Yield . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.4.3 Collection Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.4.4 Trapping Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

2.4.5 Photon Detection Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2.4.6 Signal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

3 Silicon Photomultiplier Study 29

3.1 Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.1.1 SiPM Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.1.2 Keithley 6487 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.1.3 Amplification board TB-411 8+ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.1.4 Oscilloscope - Tektronix TDS3032 . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

xi

3.1.5 NIM Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.1.6 PREC Front-end . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

3.1.7 Altera DE2 Development and Education Board . . . . . . . . . . . . . . . . . . . . 33

3.2 Gain Variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

3.2.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

3.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3.3 Dark Current & Photo-electron Determination . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.3.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.3.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

4 CMuD Characterization 43

4.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

4.1.1 Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

4.1.2 Purity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

4.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4.2.1 CMuD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4.2.2 CMuD in Self-Coincidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

4.2.3 Two CMuD’s in Coincidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

5 Conclusions and Outlook 61

Bibliography 63

xii

List of Tables

1.1 Typical light yield, expressed in photons per keV, of deposited energy, wavelength of max-

imum emission, and decay time for scintillating materials usually used in detection appli-

cations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.2 Some solvents and solutes widely used in plastic scintillators. . . . . . . . . . . . . . . . . 12

1.3 Typical characteristics of the PMT, APD and SiPM.[24] . . . . . . . . . . . . . . . . . . . . 17

2.1 EJ-200 plastic scintillator main characteristics. [27] . . . . . . . . . . . . . . . . . . . . . . 20

2.2 Y-11(200) WLS fibers characteristics.[28] . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.3 Characteristics of the MPPC Hamamatsu S12572-50P. . . . . . . . . . . . . . . . . . . . . 22

4.1 Measurement of the event rate of each individual scintillation detector. . . . . . . . . . . . 45

4.2 Count & Rate of detected muons in a 600 s acquisition for two different set-ups. . . . . . . 46

4.3 Efficiency measurement with a fixed threshold of 16ph.e. and different time windows. . . . 50

4.4 Efficiency of a CMuD for different thresholds. . . . . . . . . . . . . . . . . . . . . . . . . . 51

4.5 Purity of a CMuD measured for different thresholds. . . . . . . . . . . . . . . . . . . . . . 52

4.6 Efficiency of a CMuD in self-coincidence. . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.7 Purity of a CMuD in self-coincidence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

4.8 Efficiency of two CMuD’s in coincidence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

4.9 Purity of two CMuD’s in coincidence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

xiii

xiv

List of Figures

1.1 Cosmic rays energy spectrum. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 Particles and radiation produced by extensive air shower. . . . . . . . . . . . . . . . . . . 3

1.3 Example of a charged gold-leaf electroscope. . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.4 Victor Hess in one of his famous balloon experiements. . . . . . . . . . . . . . . . . . . . 4

1.5 Positron tracks photographed by Carl Anderson. The direction of the curve indicates that

this particle has positive charge. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.6 Layout of the Pierre Auger Observatory. The black dots represent the cherenkov detec-

tors, the blue radial lines represent the sectors of fluorescence detector and the red points

sites with specialized equipment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.7 Molecular energy levels transitions in organic scintillators. . . . . . . . . . . . . . . . . . . 10

1.8 Energy band structure of an inorganic scintillator. . . . . . . . . . . . . . . . . . . . . . . . 10

1.9 Scintillation mechanism of plastic scintillator. Approximate fluor concentrations and en-

ergy transfer distances.[17] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1.10 Example of variation of specific fluorescence dL/dx in anthracene with energy loss.[20] . 13

1.11 Scheme of photomultiplier tube. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

1.12 Structure of an APD in a Silicon Photomultiplier. It shows from top to bottom: the quench-

ing resistance (red), the anti-reflective layer (white), the n-p junction (blue and green), the

depletion region (white), the p substrate and finally the metal contact (grey). . . . . . . . . 16

2.1 CMuD prototype. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

2.2 Emission spectrum of the EJ-200 plastic scintillator. . . . . . . . . . . . . . . . . . . . . . 21

2.3 Absorption and emission spectrum of Y-11(200) WLS fibers from Kuraray’s.[28] . . . . . . 21

2.4 Light yield, in photoelectrons, for different combinations of scintillator and fibers. [29] . . . 22

2.5 Muon energy loss in different materials in relation to its momentum. Radiative effects,

relevant for muons are not included. These only become significant for particles with

higher momentum and absorbers with higher Z. . . . . . . . . . . . . . . . . . . . . . . . 24

2.6 Representation of a particle crossing the CMuD.[33] . . . . . . . . . . . . . . . . . . . . . 25

2.7 Angular distribution of cosmic muons at ground level.[35] . . . . . . . . . . . . . . . . . . 25

2.8 Geant4 simulation of photons generated in a scintillator by an ionizing particle.[36] . . . . 26

2.9 Photons propagating inside a wavelength shifting fiber.[36] . . . . . . . . . . . . . . . . . 28

xv

3.1 Arrangement of the cells of the SiPM in parallel with series quenching resistor.[37] . . . . 29

3.2 MPPC Hamamatsu S12572-50P The pulses have well defined amplitudes that are multi-

ple of the number of one ph.e. and proportional to the number of cells that discharged. . . 30

3.3 SiPM circuit used. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.4 TB-411 8+ low noise amplifier board. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.5 Picture of the Front-end board. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

3.6 Altera DE2 Board. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

3.7 Experimental setup used to measure the SiPM gain. . . . . . . . . . . . . . . . . . . . . . 35

3.8 Amplified signal of a 100 pF capacitor discharge, charged by tension of 300 mV. Acquired

by the oscilloscope. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3.9 Variation of the charge in a 100 pF capacitor with amplitude of the supply signal. . . . . . 36

3.10 Example of a 1 ph.e pulse with SiPM biased at 65.55 V. . . . . . . . . . . . . . . . . . . . 36

3.11 Plot of the SiPM gain as a function of the overvoltage and corresponding linear fit. . . . . 37

3.12 SiPM circuit in metal box. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.13 On top is represented a scheme of the main components of the setup. Bellow, the picture

shows the setup used for the measurement of the dark current. The components are:

1-Keithley Voltage Source, 2 - MPPC circuit, 3 - Amplifier TB-411 (Not Used), 4-5 - Front-

end Voltage sources, 6 - Front-end, 7 - Raspberry Pi (used to program the front-end) and

8 - Ethernet output to GPIO of DE2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

3.14 Plot of the dark current rate in logarithmic scale. . . . . . . . . . . . . . . . . . . . . . . . 39

3.15 Plot of the inverted dark current derivative. . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

3.16 Linear fit of the dark current rate in logarithmic scale versus threshold in ph.e. . . . . . . . 41

4.1 New SiPM circuit used to assemble the CMuD. . . . . . . . . . . . . . . . . . . . . . . . . 43

4.2 The two scintillation detectors used to produce a muon trigger. . . . . . . . . . . . . . . . 44

4.3 Triggered event in each individual scintillator and respective discriminator output. . . . . . 45

4.4 SignalTap - Delay between the hodoscope trigger and CMuD signal. From top to bottom,

the GPIO 1[15] corresponds to the muon trigger, the GPIO 1[11] to the CMuD signal and

in the bottom is the monostable variable. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4.5 Dark box used to isolate the set-up. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4.6 Reference scintillation detector used in the purity measurement. . . . . . . . . . . . . . . 48

4.7 Plot of the CMuD efficiency in ph.e scale. . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

4.8 Plot of the purity in ph.e scale. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

4.9 CMuD in self-coincidence. Division of the WLS fibers. . . . . . . . . . . . . . . . . . . . . 53

4.10 Signal of the CMuD in self-coincidence and respective dark current rate. . . . . . . . . . . 54

4.11 Plot of the efficiency in ph.e scale. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.12 Plot of the efficiency in ph.e scale. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

4.13 Measurement of two CMuD’s in coincidence aligned and side by side. . . . . . . . . . . . 57

4.14 Plot of the detection efficiency of two CMuD’s in coincidence in ph.e scale. . . . . . . . . 58

xvi

4.15 Plot of the purity in ph.e scale. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

xvii

xviii

Chapter 1

Introduction

The field of cosmic rays and particle physics is currently pushing the fringe of our knowledge. Trying

to prove the soundness of the standard model while at the same time trying to figure out how our

universe came into existence. It challenges our concepts and imagination yet it seems so out of reach

and disconnected from the common people. The purpose of this project is to create a bridge that

allows students to come into contact with this field. By building a device that is able to measure some

elementary particle we expect to bring students closer to this field while also helping their development.

This thesis will focus on the development and study of a prototype that is able to detect particles of

cosmic origin.

The layout of this thesis consists of an introduction to cosmic rays and scintillation detectors. Fol-

lowed by a study of the silicon photomultiplier used to build the CMuD prototype and finally efficiency

and purity studies of this detector by itself and in coincidence.

1.1 Motivation

It is common to perceive that studies in the field of particle physics are only possible with complex and

vastly expensive equipment while supervised by hundreds of highly qualified scientists and engineers.

Yet that paradigm is changing. With few resources it is already possible to build instruments that can

detect and make measurements of actual elementary particles. Using, as a base, components from

large scale particle detectors.

The goal of this project is to study such a detector. It must be able to be built using few resources in

order to be relatively cheap, safe to handle, so that there is no danger for the users, and robust enough

that it will not break easily. With these characteristics it opens the possibility to study particle physics at

high school level, bringing students closer to this field of research. It provides a teaching tool, that is an

asset to the formation of students that so far is not available.

1

1.2 Cosmic Rays

The field of cosmic rays is more than a century long and, as almost all great scientific discoveries,

it started with a curious phenomenon that could not quite be explained. The study of the apparent

spontaneous discharge of the electroscope gave rise to the notion of cosmic radiation. Apparently Earth

was constantly being bombarded by radiation coming from space. This cosmic radiation is composed

by different types of particles: mainly protons, alpha particles, heavy nuclei, electrons, photons and

neutrinos. Cosmic rays have a very wide energy range, from around 1 GeV to as much as 108 TeV

which is about 40 million times the energy of particles accelerated in the LHC (Large Hadron Collider).

Figure 1.1: Cosmic rays energy spectrum.

As the energy of the cosmic rays increases, the flux of the particles decreases steeply as it can

be seen in figure 1.1. Particles with intermediate energy levels, around what is called the ”knee” of

the spectrum, are called Very High Energy cosmic rays and they occur with a frequency of one per

square meter per year. For the highest energy cosmic rays, above 1019 eV the rate of events falls to

one per square kilometer per century. With such a wide range of energy, flux and composition it is only

natural that very different techniques are used to detect cosmic rays in different parts of the spectrum.

For cosmic rays with energies smaller than 1016 eV and higher fluxes the direct detection is possible

by using high altitude balloons or satellites that intercept this particles before they interact with Earth’s

atmosphere. Above this energy the flux is too low for space experiments. Therefore, indirect methods of

detection are used, exploiting the extensive air showers (EAS) created by these particles.

When a cosmic ray enters Earth’s atmosphere it eventually collides with air molecules, mainly nitro-

gen or oxygen, and, due to the great energy available, this interaction produces a cascade of lighter

2

Figure 1.2: Particles and radiation produced by extensive air shower.

particles as illustrated in figure 1.2. The collision generates secondary particles typically light hadrons

such as kaons and pions. These particles have a very short life time and quickly decay into lighter parti-

cles. The neutral pions decay into gammas, feeding the electromagnetic component via the production

of electron-positron pairs that continue to interact and dissipate energy until they reach their critical en-

ergy. Charged pions and kaons eventually decay or interact to form muons and neutrinos contributing to

the muonic component of the cascade. The number of secondary particles increases until the resulting

secondary particles reach the energy threshold to generate more particles from interactions. Then, the

number of particles is attenuated as their energy is dissipated. While the electromagnetic component

increases and is then absorbed by the atmosphere after a certain depth, the muonic component stays

almost unchanged, since it interacts very little with matter. Although muons are unstable elementary

particles with a relatively short lifetime, the relativistic effects, consequence of their high energy (mean

energy of 4 GeV at ground level [1]) allows them to reach Earth’s surface. These particles are ideal for

the purposes of this project since they are very abundant and easy to detect, their study can be per-

formed with relatively low cost. Also it provides a good opportunity for students to autonomously study

these particles providing the basic concepts of particle and cosmic ray detection.

1.2.1 Short History of Cosmic Rays Detection

Although it was not known at the time, the first instrument used to measure cosmic radiation was the

electroscope. This instrument is an early scientific instrument used to measure the magnitude of electric

charge on a body.

The golden-leaf electroscope consists of a vertical metal rod from which hangs two parallel strips of

flexible gold leaf that when charged spread apart in a ”V” shape (figure 1.3). It can be charged by contact

using a charged rod and touching the top metallic disc on top of the instrument and discharged simply

by touching the disc, i.e. grounding the terminal by providing a connection to earth so the electrons can

balance the charge.

A curious phenomenon was observed in this instrument: the electroscope appeared to sponta-

neously discharge without any contact. One would expect the electroscope to maintain its charge if

3

Figure 1.3: Example of a charged gold-leaf electroscope.

the isolation was perfect but even after improvements of the instrument there was no change. By the

end o the 19th century it was known that the spontaneous discharge was due to the ionization of the

air inside the electroscope bottle. At that time it was believed that this ionization was due to the natural

radiation of the Earth and using the rate of discharge it was possible to quantify it.

In start of the 20th century the source of this radiation was started to be questioned. Some important

experiments on that direction were Father Wulf measurement at location with different heights and more

importantly Dominico Pacini underwater measurements that he concluded: ”that a sizable cause of

ionization exists in the atmosphere, originating from penetrating radiation, independent of the direct

action of radioactive substances in the soil. ”[2].

Figure 1.4: Victor Hess in one of his famous balloon experiements.

Later in 1912, Victor Hess took the electroscope and flew in a hot air balloon (a photograph of the

campaign is shown in figure 1.4) measuring the rate of air ionization at different altitudes. He observed

that this rate decreased from the ground level to approximately 1 km and from there on it increased. The

result seemed to confirm two things. For one the natural radioactivity of the Earth seem to contribute

to the ionization of the air since the ionization rate seem to decrease at lower altitudes. Secondly, the

subsequent increase was attributed to penetrating radiation arriving from Space. This conclusion was

4

disputed at first by scientist like: G. Hoffman (1924) and F. Behounek (1925) who believed the source

was radioactive elements in the atmosphere and R. Millikan (1924) that believed the radiation was of

local origin [3]. But eventually the data was overwhelming and by 1928, Robert Millikan coined the term

Cosmic Rays to describe the penetrating radiation [4], which still remain to this day.

For some time it was believed that the radiation was electromagnetic in nature hence the name

cosmic “rays”. However, in 1933-34 three independent measurements (Alvarez/Compton[5], Johnson[6]

and Rossi[7]) found that the intensity of cosmic rays is greater from the east than west. Thus proven that

cosmic rays must be electrically charged since they are affected by Earth’s magnetic field.

With time, more sophisticated instruments were developed to detect radiation. One of particular

importance is the cloud chamber. This instrument consists of one container filled with supersaturated

alcohol vapor. Because of this, any small disturbance triggers the formation of small droplets. Therefore,

when a charged particle crosses the chamber, a trail of droplets forms along the particle path. The tracks

formed were then recorded through photographic emulsion. This allowed the first observation of cosmic

rays by Dimitry Skobelzyn in 1927. But by itself the cloud chamber can only show that a charged

particle passed by. To distinguish between particles a uniform magnetic field is applied and observing

the characteristics of its path curvature its mass and charge can be derived.

Figure 1.5: Positron tracks photographed by Carl Anderson. The direction of the curve indicates thatthis particle has positive charge.

Using this technique, in 1932, Carl Anderson discovered antimatter in the form of a antielectron,

later called positron[8], one of the photographs is reproduced in figure 1.5. This particle has exactly

the same mass as the electron but opposite charge. This discovery had great implications and gave

birth to the field of particle physics. Soon more particles were being discovered. In 1937, the same

Carl Anderson and Seth Neddermeyer[9] observed a particle that seemed to have negative charge but

curved less sharply than the electron and more than the proton. The muon was initially mistaken for the

meson theorized by Yukawa but eventually proved to have the wrong properties. The real first meson

discovered was the pion, in 1947, this one showing the correct properties to mediate the nuclear force.

The Kaon followed up in the same year and was the first ‘strange’ particle found. All these particles

5

were found from cosmic rays and a few more important discoveries were made before the era of particle

accelerators started.

Pierre Auger, who had positioned particle detectors high in the Alps, noticed that two detectors, lo-

cated many meters apart, both signaled the arrival of particles at exactly the same time. He suggested

that particles arrived in bunches and with the development of coincidence units with better time resolu-

tion this phenomenon was systematically studied. In 1938[10], he was able to establish the occurrence

of extensive air showers. It was a great breakthrough as it allowed the correlation between the sec-

ondary particles detected at ground level with the primary high energy cosmic rays that reach the top

atmosphere. This opened the path for the study of high energy cosmic rays. In 1946, Bruno Rossi in the

United States and Georgi Zatsepin started experiments on the structure of air showers and constructed

the first arrays of detectors. By using an array of correlated detectors that covered a determined area

of the ground surface it was possible to detect the front of the extensive air showers and from there

extrapolate the characteristics of the original particle like the energy and incidence angle.

Two year later Enrico Fermi put forth an explanation for the acceleration of cosmic rays[11]. In Fermi’s

diffusive shock accelerator, protons speed up by bouncing off magnetic clouds in space. Solar flares and

supernovae explosions are believed to act as such accelerator. In the 50’s with the development of the

first particle accelerators the two fields took different paths. Around this time scientists started to turn

away from cosmic rays and focus on the particle accelerators since they allowed the study of certain

fundamental curves (scattering, ionization, range) and were not dependent on nature to provide events.

Even so, the cosmic rays field had the uniqueness study of ultra-high energy particles since the man

made accelerators can not reach such high energies.

In 1962 an air shower with the energy of 1020 eV, was observed in the Volcano Ranch experiment,

New Mexico[12]. It was the first extreme-energy cosmic ray ever detected, such a high energy could not

be explained by the Fermi’s diffusive shock accelerator model. The mystery increased when, around

the same time, it was discovered that the universe is permeated by low-energy photons called cosmic

microwave background that dates back to the epoch of recombination (when electrons and protons

became bound to form hydrogen atoms). It was pointed out that high-energy cosmic rays traveling long

distances would lose energy due to interactions with photons in this background radiation. Therefore,

particles traveling long intergalactic distances could not have energies greater than 5 × 1019 eV, this is

known has the Greisen–Zatsepin–Kuzmin limit (GZK limit).

In the 60’s an alternative method to detect extensive air showers - fluorescence technique - was

developed. When a charged particle from an EAS travels through the atmosphere it excites and ionizes

the nitrogen molecules of the air. Some of this excitation energy is then isotropically emitted in the

form of UV radiation. Since the isotropic fluorescence light is only intense enough to be observed in

very high energetic showers, and because it allows the observation of a large volume of atmosphere,

this technique is particularly well suited for the low flux and ultra high energy cosmic ray detection. It

has the advantage of providing a picture of the longitudinal development of the shower and a more

direct measurement of its energy with respect to ground detectors. On the other hand new sources of

systematical error arise and functioning is restricted to moonless nights. The first attempts to observe

6

EAS with this technique were made by Cornell Cosmic Ray Observatory using 500 photomultiplier tubes

each corresponding to a pixel covering a solid angle of approximately 6 by 6 degrees. The device was

operated for several years but was not sensible enough to detect the light emitted from air showers. The

first successful observation of this fluorescence light[13] would come in 1976 with the installation of a

prototype telescope in Volcano Ranch by physicists from The University of Utah. Volcano Ranch was

chosen for the test because the site also hosted a large ground array which had been operated since

1958 and was located in the remote desert where light pollution would be minimum. The prototype

detectors were able to observe air showers in coincidence with the ground array which confirmed the

results.

After the successful testing the research group began the construction of a full scale observatory

based on the same basic design. Located in the west desert of Utah, in 1981, the Fly’s Eye[14] became

the first cosmic ray experiment employing the fluorescence technique. The experiment was operational

until 1993 and at the time of its shutdown had compiled the world’s largest ultra-high-energy cosmic ray

data set. In 1991 it detected an event with the energy of 3.2 × 1020 eV, the highest energy cosmic ray

ever detected. Two years later the AGASA group in Japan and Yakutsk group in Russia, both using

detector arrays, reported two separate events with energies in the same order of magnitude.

Figure 1.6: Layout of the Pierre Auger Observatory. The black dots represent the cherenkov detec-tors, the blue radial lines represent the sectors of fluorescence detector and the red points sites withspecialized equipment.

To further study the mysteries of ultra high energy cosmic rays, in an international collaboration effort,

a large scale observatory was put together. Located in the vast plains of Pampa Amarilla, Argentina,

The Pierre Auger Observatory, officially completed in 2008, is the first hybrid detector to combine both

the fluorescence and detector array techniques, allowing the cross-calibration of the two methods and

reducing the systematic errors by combining the data. The basic set-up, shown in figure 1.6, is com-

posed by a giant array of 1600 Cherenkov detector tanks covering an area of 3000 square kilometers

along with 27 atmospheric fluorescence detector telescopes. The first results from Auger firmly establish

the existence of a cutoff in the spectrum and the anisotropy of the highest energy cosmic rays, as well

7

as raising very interesting puzzles concerning their nature and their interactions in the atmosphere [15].

So far we focused mainly in the evolution of UHE cosmic rays field but there are other important

areas that also should be mentioned. The field of high energy gamma rays studies photons that reach

Earth’s atmosphere with energies in the GeV -TeV range. These particles originate pure electromagnetic

showers that develop high in atmosphere hence its detection at ground level is difficult. These also can

not be detected by fluorescence detectors since the energy of these particles is too low for the sensitivity

of the detectors. However, the secondary particles from the air shower produce Cherenkov radiation that

propagates along the shower direction producing a signal detectable at ground level. The study of high

energy gamma rays are important because they allow to pin point the source of cosmic rays since they

travel in a straight line unlike charged cosmic rays that have their direction scrambled by magnetic fields.

In the late 1960s the program to study the Cherenkov light emitted by EAS started with the Whipple

observatory. The idea was to record this light by mean of telescopes. The background from charged

cosmic rays was however enormous. The method became more efficient with the introduction of imaging

techniques (Hillas, 1985), in which segmented focal surfaces and shower shape variables allowed to

distinguish showers initiated by energetic gamma rays from the background. In recent years, larger

telescopes and telescope arrays have been built, allowing the discovery of an impressive number of

gamma ray sources in the sky. The most significant results have been provided by large scale detectors

such as: High Energy Stereoscopic System (HESS) located in Namibia and MAGIC telescope (Major

Atmospheric Gamma Imaging Cherenkov Telescopes) located in the Canary Islands.

1.3 Overview of Scintillation Detectors

As the field of cosmic rays and particle physics developed it pushed the necessity to develop new detec-

tion technology to increase the sensitivity of the detectors. Therefore, through the years many different

types of detectors have been developed, each with its particular characteristics directed to a specific

goal or application. This thesis will only focus on scintillation detectors since that is the type of the pro-

totype developed at LIP. With their ease of fabrication, low cost, fast response and the possibility to have

large detection areas it is ideal for muon detection and academic purposes.

The basis of particle detection relies on the knowledge of the interactions occurring when particles

encounter matter. These interactions depend both on the characteristics of the particle and of the

detector material. When a charged particle traverses matter, it excites and ionizes molecules along

its path. Certain types of molecules release a small fraction of this energy as optical photons. This

process, known as scintillation, is specially relevant for organic substances containing aromatic rings,

such as polystyrene (PS) and polyvinyltoluene (PVT) as well as scintillating liquids with toluene and

xylene.

A typical scintillation detector consists of four main parts: a scintillator, a light guide, a photo-detector,

and a read-out system. There are many different types of scintillation materials which can be classified

into inorganic crystals, glasses and gases, and organic compounds such as crystals, plastics and liq-

uids. The light guides are used to couple the scintillation material to the photo-detector. Photomultiplier

8

tubes are the most commonly used photo-detectors with scintillators. Lately, Geiger-mode avalanche

photo-diodes (G-APD), also known as silicon photomultipliers (SiPM), have started being used to detect

scintillation light instead of traditional photomultipliers tubes. The final component of the scintillation

detection system is the signal processing and data acquisition electronics which allows to count and

quantify the amplitude of the signals.

1.3.1 Scintillation Materials

Scintillators convert the energy of impinging particles into optical photons. Table 1.1 lists typical values

of light yield and peak emission wavelength for different types of scintillation materials commonly used

for detection applications. The scintillators can be divided in two main types, organic and inorganic,

according to the process of excitation and de-excitation of electrons in the material.

Material Light Yield (φ/keV) λpeak (nm) Decay Time (s)Glass Scintillator 3 - 5 400 ≈ 10−9

Plastic Scintillator 8 - 10 420 ≈ 10−9

Liquid Scintillator 11 - 13 420 ≈ 10−9

Inorganic Crystal 40 - 90 415 - 480 ≈ 10−7

Table 1.1: Typical light yield, expressed in photons per keV, of deposited energy, wavelength of maximumemission, and decay time for scintillating materials usually used in detection applications.

Organic scintillators use the ionization produced by charged particles to generate optical photons,

usually in the blue to green wavelength region. They are broadly classed into three categories; crystals,

liquids, and plastics. Organic scintillators are made of aromatic hydrocarbon compounds that have

weak interaction between them. In equilibrium practically all electrons are in the fundamental state S0

but in the presence of incident radiation the electrons are excited and populate the vibrational levels

within S1 state (figure 1.7). In this state the electrons are relatively unstable and eventually decay

back to its fundamental state. In this transition a photon is emitted with the energy corresponding to

gap between the excited and fundamental states. S1 can also decay to adjacent triplet levels, with a

significantly lower energy and a much longer decay time. The process then corresponds to either one

of two phenomena, depending on the type of transition and hence the wavelength of the emitted optical

photon - fluorescence or phosphorescence. The absorption and emission processes are spread out

over a wide range of photon energies with an overlap between the two spectra. This means that a

fraction of the emitted light is re-absorbed. This self-absorption leads to a shortened attenuation length.

The wavelength difference between the absorption and emission peaks is called the Stokes’ shift. The

greater this shift is, the smaller the self-absorption. In the ideal case the absorption spectrum would not

overlap with the emission and the scintillators would be transparent to the emission spectrum.

Whereas the scintillation mechanism in organic materials is of molecular nature, inorganic materials

depends on the electronic band structure determined by its crystal lattice. Absorption of energy can

result in the elevation of an electron from its normal position in the valence band across the gap into

the conduction band, leaving a hole in the valence band (figure 1.8). In the case of a pure crystal, the

return of the electron to the valence band with the emission of a photon is a very inefficient process and

9

Figure 1.7: Molecular energy levels transitions in organic scintillators.

the typical gap widths result in photons with an energy too high with respect to the visible range. To

enhance the probability of visible photon emission during de-excitation, small amounts of an impurity,

called activators, are added to inorganic scintillators. These activators create sites in the lattice resulting

in energy states in the forbidden gap through which the electron can de-excite back to the valence band.

Because the energy is lower than that of the full forbidden gap, this transition gives rise to a visible

photon and can be used for the scintillation process.

Figure 1.8: Energy band structure of an inorganic scintillator.

Different scintillating materials have different properties, the material choice is then dependent on

the specifics of application intended. This particular project is interested in a cheap, robust and durable

material with a good light yield. Although inorganic scintillators have typically better light yields and

linearity (good energy resolution), they are relatively slow and the need for a crystalline structure makes

them expensive and hard to grow to large sizes. Therefore, the choice of organic scintillators seemed

more adequate for the goals of this prototype. Organic crystals are very durable, but their response

is anisotropic (which spoils energy resolution when the source is not collimated), and they cannot be

easily machined, nor can they be grown in large sizes; hence they are not very often used. This leaves

us with plastic and liquid scintillators, both have similar characteristics but liquid scintillators are difficult

to handle from the mechanical point of view and, for the purposes of this thesis, only plastics scintillators

10

will be discussed.

Plastic Scintillators

If an organic scintillator is dissolved in a solvent that can then be subsequently polymerized, the equiva-

lent of a solid solution can be produced. A common example is a solvent consisting of styrene monomer

(common solvents are represented in table 1.2) in which an appropriate organic scintillator is dissolved.

The styrene is then polymerized to form a solid plastic (PS). Other plastic matrices can consist of

polyvinyltoluene (PVT) or polymethylmethacrylate (PMMA) [16]. Because of the ease with which they

can be shaped and fabricated, plastics have become an extremely useful form of organic scintillator.

Plastic scintillators are available commercially with a good selection of standard sizes of rods, cylin-

ders, and flat sheets. Because the material is relatively inexpensive, plastics are often the only practical

choice if large-volume solid scintillators are needed. There is also a wide selection of plastic scintillators

available as small diameter fibers. They can be used either as single fibers or grouped together to form

bundles in applications where the particle position must be reconstructed with good spatial resolution.

Plastic scintillators used in detectors consist of binary and ternary solutions of selected fluors in a

plastic base containing aromatic rings. Most of them have a base of either PS or PVT. Figure 1.9 depicts

the scintillation mechanism.

Figure 1.9: Scintillation mechanism of plastic scintillator. Approximate fluor concentrations and energytransfer distances.[17]

Ionization in the plastic base produces UV photons with short attenuation length in the order of a few

millimeters. Longer attenuation lengths can be obtained by dissolving a primary fluor in high concentra-

tion, about 1% in weight, into the base. The fluor is selected to efficiently re-radiate the absorbed energy

at wavelengths where the base is more transparent. The other function of this fluor is to shorten the de-

cay time and to increase the total light yield. The scintillator base materials can have long decay times.

For example, pure PS has a decay time of 16 ns. The addition of a primary fluor in high concentration

can shorten the decay time by an order of magnitude. It is often necessary to add a secondary fluor, at

fraction percent levels, and occasionally a third one in order to shift light to longer wavelengths to better

match the region of maximum sensitivity of the photo-detectors. Light collection from large scintillators

or complex geometries can also be aided through the use of optical elements that employ wavelength

11

shifting techniques. Many liquid or plastic scintillation detectors routinely incorporate wavelength shifting

fibers to their design since as they provide an optical guide for the photons while at the same time can

match their wavelength to the response peak of the photo-detector used.

Solvent Secondary Fluor Tertiary FluorPolystyrene

(PS) p-terphenyl POPOP

Polyvinyltoluene(PVT) DPO TBP

Polymethylmethacrylate(PMMA) PBD BBO

Polyvinylbenzene(PVB) DPS

Table 1.2: Some solvents and solutes widely used in plastic scintillators.

The response of organic/plastic scintillators to charged particles can be described by a relation be-

tween the fluorescent energy emitted per unit path length, dL/dx, and the specific energy loss for the

charged particle dE/dx. A widely used relation first suggested by Birks[18][19] is based on the as-

sumption that a high ionization density along the track of the particle leads to quenching from dam-

aged molecules and a lowering of the scintillation efficiency. If we assume that the density of damaged

molecules along the wake of the particle is directly proportional to the ionization density, we can repre-

sent their density by B(dE/dx), where B is a proportionality constant. Birks assumes that some fraction

k of these will lead to quenching. A further assumption is that, in the absence of quenching, the light

yield is proportional to energy loss:

dL

dx= S

dE

dx(1.1)

where L is the luminescence, E the energy and S is the normal scintillation efficiency. To account for

the probability of quenching, Birks then writes:

dL

dx=

S dEdx

1 + kB dEdx

(1.2)

Equation 1.2 is commonly referred to as Birks formula [20]. As a practical matter the product kB is

treated as an adjustable parameter to fit experimental data for a specific scintillator.

When excited by fast electrons (either directly or from gamma-ray irradiation), dE/dx is small for suf-

ficiently large values of E and Birks’ formula then predicts a linear response from the scintillator. On the

other hand, when dE/dx is very large so that saturation occurs along the track, Birks’ formula becomes

S/kB which is a constant (see figure 1.10). This evidences the non linearity of plastic scintillators for

high energy depositions.

For the vast majority of plastic scintillators, the prompt fluorescence represents most of the observed

scintillation light. A longer-lived component is also observed in many cases, however, corresponding

to phosphorescence (figure 1.7). The composite yield curve can often be represented adequately by

the sum of two exponential decays-called the fast and slow components of the scintillation. Compared

with the prompt decay time of a few nanoseconds, the slow component will typically have a characteristic

12

Figure 1.10: Example of variation of specific fluorescence dL/dx in anthracene with energy loss.[20]

decay time of several hundred nanoseconds. Because the majority of the light yield occurs in the prompt

component, the long-lived tail would not be of great consequence except for one very useful property:

The fraction of light that appears in the slow component often depends on the nature of the exciting

particle. Pulse shape discrimination techniques allow to differentiate between alpha, beta and gamma

radiation.

Synthesizing, when comparing with inorganic scintillators, organic materials have faster time re-

sponses, they are easily shaped, hence cheaper. Organic scintillators allow pulse shape discrimination

and have small dependence on the temperature; on the other hand, have worst light yields, not good

energy resolution due to its saturation at high energies and its low density does not allow the proper

detection of gamma radiation. Plastic scintillators also suffer from radiation damage, they are subject to

aging and can degrade upon exposure to some chemicals.

1.3.2 Photo-detectors

A photo-detector is a device that converts light into an electrical signal. This process is essential in all

scintillation detector since it correlates the light yield generated by a passing particle to an electrical

quantity. There are two main types of photo-detectors used in particle physics, namely scintillation

detectors; the photomultiplier tube and the photodiode.

Photomultiplier Tube

Photomultiplier tubes are extremely good at converting light into an electrical signal; electrical pulses can

be obtained from only one photon with an efficiency of∼10-20%. Photo multiplier tubes (PMTs) are used

to detect photons. When the photons reach the PMT, they create electrons via the photoelectric effect

by impinging on a cathode made of photosensitive material, a photocathode. The electrons are then

gathered by an anode from which an electric signal is received, see Figure 1.11. The signal amplitude,

13

in the quoted operation range, is linear with the number of incoming photons.

Figure 1.11: Scheme of photomultiplier tube.

Photomultiplier tubes first major component is the photocathode (figure 1.11). A light photon may in-

teract in the photocathode to eject a low-energy electron into the vacuum. This process can be thought

to occur in three steps: the absorption of the photon and energy transfer to the electron in the photo-

cathode material, the migration of the photoelectron to the surface of the photocathode and the escape

of the electron from the photocathode surface[21].

The energy available to be transferred from a scintillation photon to an electron is, for blue scintillation

light, approximately 3 eV. However, some of this energy is lost in electron-electron collisions as the

electron migrates to the surface and then the potential barrier at the surface vacuum interface (the

work function) must be overcome in order for the electron to enter the vacuum. There are therefore

energy limitations on the system posed by the potential barrier at the surface/vacuum interface and the

interactions in the material. In addition, the thickness of the material that can generate photoelectrons

into the vacuum is limited, because there is a rate of energy loss as the electron migrates to the surface.

In either case only a thin layer of material contributes photoelectrons into the vacuum.

The cathodes are sometimes designed so that the light is incident on one side, while the electrons are

ejected on the other. The photocathode is therefore extremely thin, because of the problems of electron

interaction in the material, and is therefore semi-transparent to visible light. This means that a large

portion (more than half) of the light passes straight through without interacting with the photocathode.

The quantum efficiency is the ratio between the number of photoelectrons emitted by the number of

incident photon. This ratio is a strong function of wavelength of the incident light. An effort should be

made to match the spectral response of the photocathode to the emission spectrum of the scintillator.

The average quantum efficiency over the emission spectrum of a typical scintillator is about 15-20%

while the peak quantum efficiency is 25-30%, depending of the PMT model.

After the electrons are ejected from the cathode, several stages of multiplication are employed in

the dynodes (figure 1.11) to generate gains in the order of 106. Electrons are focused and accelerated

toward the first dyode by an high electric field, where secondary electrons are emitted. The secondary

electrons are then accelerated again toward a second dynode, where they create secondary electrons,

and so on. This process is repeated ∼ 8-10 times to achieve very high gains. In order to create this

multiplication chain it is necessary to apply a very high voltage to the PMT- 1-2 kV - which is then divided

14

by each individual multiplication stage. By the end of the multiplication stage the resulting electrons are

collected by the anode to generate an electric signal.

Photodiode

A photodiode is a device based on a junction of oppositely doped regions (p-n junction) in a sample of

semiconductor. This creates a region depleted of charge carriers that results in high impedance. The

photodiode functions using an illumination window, which allows the use of light as an external input.

When light enters the device, electrons in the structure become excited. If the energy of the light is

greater than the band gap energy of the material, electrons will move into the conduction band. The

result is creation of holes throughout the device in the valence band where the electrons were originally

located. Electron-hole pairs generated in the depletion region drift to their respective electrodes: n for

electrons and p for holes, resulting in a positive charge build-up in the p layer and a negative charge

build-up in the n layer. The charge is directly proportional to the amount of light falling on the detector.

If an external circuit is connected to the electrodes, current will flow in the circuit. This describes the

photovoltaic mode of operation. The photodiode can be operated in three different modes: photovoltaic,

photoconductive and avalanche.

The photoconductive mode is achieved by applying a reverse bias to the photo-detector. This in-

creases the electric field strength between the electrodes and the depth of the depletion region. The

advantages of this kind of operation are lower capacitance, and hence higher speed, as well as improved

linearity. However, dark current is directly dependent on reverse bias voltage, and thus becomes larger

with increasing bias voltage.

Avalanche photodiodes are photodiodes with a structure optimized for operating in linear mode with

an high reverse bias, approaching the reverse breakdown voltage, typically 100-200 V for silicon pho-

todiodes. It is a p-n device with internal gain due to the high internal field at the junction of positive

and negative doped silicon. In an APD, a photoelectron in this field gains enough energy to create an

electron-hole pair by impact ionization; both the initial electron and the additional electron again undergo

high acceleration and can initiate further electron-hole pairs — thus starting an avalanche. If the electri-

cal field is not too high, the accelerated holes does not gain enough energy to create electron-hole pairs

in addition, otherwise the process runs out of control and a breakdown can occur. The multiplication in

practical APD’s is moderate, between 50 and 200 [22]. A gain of 104 is in principle possible but at values

higher than a few hundred, the environment (e.g. temperature and voltage supply) needs to be highly

regulated because the APD has to be operated extremely close to the breakdown voltage. The size of

APDs is limited due to the production yield to achieve an extremely uniform field distribution over the

sensitive area. The biggest area available commercially is 2.5 cm2 [23].

Silicon Photomultiplier

A silicon photomultiplier is composed by many small (in the µm scale) avalanche photodiodes coupled in

parallel, creating a grid. Each of these APD’s operates in the Geiger-Mode. This is achieved by applying

15

a reverse voltage higher than its breakdown voltage. This means that any photo-generated carrier

in the depletion region can trigger an avalanche multiplication of carriers by impact ionization. Both

electrons and holes contribute with a positive feedback effect, which makes the carrier multiplication

self-sustaining. Unlike what happens in linear mode APD’s, where avalanches develop in only one

direction (from the p-material to the n-material) and stop multiplying when the carriers reach the low field

area of the n-material.

Figure 1.12: Structure of an APD in a Silicon Photomultiplier. It shows from top to bottom: the quenchingresistance (red), the anti-reflective layer (white), the n-p junction (blue and green), the depletion region(white), the p substrate and finally the metal contact (grey).

The essential difference is that secondary avalanches can be triggered from holes or photons in the

p-layer. Therefore, once a Geiger discharge begins it continues as long as the electric field in the APD

is maintained. Therefore, a quenching mechanism is necessary to prevent the formation of secondary

discharges so the next photon can be detected. This can be attained by placing an ohmic resistance

in series, since it produces a drop in the operating voltage, thus stopping avalanche multiplication. On

the other hand, by using a quenching resistor it also imposes a recovery time, which corresponds to the

time it takes for the APD the reach the original operational voltage.

Each pixel of the grid simply works as a digital switch with two states, photon detected or not detected.

This means that each cell can only detect one photon at a time, but since many of them are spread in

a small area the probability of two photons arriving to the same pixel at the same time is very small.

The idea is that each one of these cells work as an independent single photon counter. Each cell can

be interpreted as a charged capacitor whose discharge is triggered by a photon. The resulting signal is

then the sum of the signal of each cell. This produces a device that generates signals with a very well

defined amplitude depending on the number of cells that discharged at one time.

Table 1.3 shows some of the main characteristics of a few types of photo-detector discussed in this

chapter. The characteristics of each photo-detector makes them suitable for different types of appli-

cations. The PMT is particularly well-suited for applications requiring low noise. While the SiPMs are

compact, cheaper, use lower bias voltages and are insensible to magnetic fields.

16

PMT APD SiPMBias Voltage High (1-2 kV) Medium (100-500 V) Low (25-70 V)

Gain ≈ 106 ≈ 102 ≈ 106

Temp. Sensitivity Low High LowMech.Robusteness Low Medium High

Light Exposure No Ok OkSpectral Range UV/Blue Red Green

Readout Electronics Simple Complex SimpleForm Factor Bulky Compact Compact

Large Area Avaliable? Yes No YesMagnetic Field Comp. No No Yes

Rise Time Fast (ps) Slow (ns) Fast (ps)Noise Low Medium High

Complexity/Price High(Vacuum, HV)

Medium(low noise electronic) Relatively low

Table 1.3: Typical characteristics of the PMT, APD and SiPM.[24]

17

18

Chapter 2

Compact Muon Detector (CMuD)

The purpose of this thesis is the development and study of compact muon detector prototype directed

for academic purposes, that can be controlled by students on their own and even be used within high

school lessons. This project was originally inspired by the CosMOS [25] (Cosmic Muon Observer) ex-

periment developed by DESY and other partner institutes within Netzwerk Teilchenwelt. This experiment

is composed by a scintillation counter and a data acquisition card.

The CMuD prototype was developed in LIP, Laboratorio de Instrumentacao e Fısica Experimental de

Partıculas, where the project was envisioned, developed, assembled and tested [26].

Figure 2.1: CMuD prototype.

2.1 Design

The CMuD is composed of three main elements:

• Scintillator

• WLS fiber

• MPPC - Silicon Photomultiplier

19

The CMuD uses a plastic scintillator tile with 20 × 20 × 1.25 cm3. Nine wavelength shifting fibers,

with 1 mm of diameter are placed in parallel shallow grooves along the scintillator, separated 2 cm from

each other. The detector is wrapped in a double layer of Tyvek reflective material to prevent the light

from escaping it. The WLS fibers are used to collect and guide the photons generated by the scintillator

to a multi-pixel photon counter (MPPC), while at the same time changing its wavelength to better match

the photo-detector response peak. A plastic piece with grooves provides the mechanical alignment of

the fibers. The metal box seen in figure 2.1 houses the MPPC along with the front-end electronics that

the decouples the power supply from the signal generated. The first prototype was built using spare

parts available at LIP, which strongly conditioned the selection of the components.

2.2 Scintillator EJ-200 & Y-11(200) Wavelength Shifting Fibers

This project used the combination of the plastic scintillator, EJ-200 from the Eljem Technologies, with

the multi cladding WLS fibers, Y-11(200) from Kuraray.

The EJ-200 scintillator combines the two properties of long optical attenuation length and fast timing,

and is therefore particularly useful for time-of-flight systems using scintillators greater than one meter

long. It combines long attenuation length, high light output, signal uniformity and an emission spectrum

well matched to the common photomultipliers. This material main characteristics can be seen in table

2.1. It is of special relevance its scintillation efficiency and wavelength emission peak. The emission

spectrum of the scintillator should coincide, as much as possible, with the absorption spectrum of the

fibers in order to maximize the detectors light output.

EJ-200 CharacteristicsLigh Yield 64% (% Anthracine)

Scint. Efficiency 10 000 photon/MeV e−

Wavelength Emission Peak (nm) 425Rise Time 0.9 ns

Decay Time 2.1 nsPulse Width ≈ 2.5 ns

Refractive Index 1.58Number Density H 5.17×1022 cm−3

Number Density C 4.64×1022 cm−3

Number Density e− 3.33×1022 cm−3

Density - ρ 1.023 g cm−3

Table 2.1: EJ-200 plastic scintillator main characteristics. [27]

The Y11-200 are a multi cladding fiber, which mean they are compounded by three layers of different

materials. The refractive index is highest in the core material and smallest in the exterior layer. This

helps increasing the light yield and trapping efficiency of the fiber. In table 2.2 are represented some of

the characteristics of the Y-11(200) fiber.

The trapping efficiency is the percentage of photons that are captured inside the fiber after the photon

re-emission. When a photon reaches the fiber it is absorbed and a photon with a higher wavelength is

isotropically emitted. The refraction indexes of the materials that constitute the fiber determine the

20

Figure 2.2: Emission spectrum of the EJ-200 plastic scintillator.

Y-11(200) Characteristicsncore 1.59nclad1 1.49nclad2 1.42

Trapping Efficiency 5.48 %Trapping Angle 26.7o

Attenuation Length (m) >3.5 mEmission Peak (nm) 476 nm

Table 2.2: Y-11(200) WLS fibers characteristics.[28]

maximum angle, in relation to the axis of the fiber, in which the photon can be emitted without escaping

it. From that trapping angle the efficiency can be calculated.

Figure 2.3: Absorption and emission spectrum of Y-11(200) WLS fibers from Kuraray’s.[28]

In figure 2.3 are represented the absorption and emission spectrum of the Y-11(200) fibers. The

absorption peak of the fiber is around the 430 nm wavelength which corresponds to the blue color, while

the emission peak is 476 nm which translates to green. This represents the average wavelength shift

the photons go through.

The relative quality of the combination EJ-200/Y-11(200) can be asserted by comparing its output

with different combinations of scintillator and WLS fibers. Studies found in the literature are presented

in figure 2.3.[29]

The light yield measurement was conducted on test panels constructed from combinations of the five

21

plastic materials and five WLS fibers listed in figure 2.3. Although the EJ-200 is not represented in this

study the BC-408 from Saint Gobain is its commercial equivalent[30], having similar properties with a

lower cost. Each plastic scintillator was wrapped with two layers of TYVEK-1025D reflector. To collect

light from scintillator were chosen to use WLS fibers running along the scintillator in shallow groves.

Such construction avoided bulky light guides, which could introduce dead areas around panel.

Figure 2.4: Light yield, in photoelectrons, for different combinations of scintillator and fibers. [29]

The data seems to show that the use of the Y-11(200) fibers gives the overall best light yield results,

except in the case of the BC-404. The use of the scintillator EJ-200 with the Y-11(200) appears to

provide a decent relative response when comparing with other less cost effective options.

2.3 Silicon Photomultiplier - MPPC Hamamatsu S12572-50P

The silicon photomultiplier’s low price, compact size and robustness made it ideal for the purposes of

this project. Besides, the MPPC has the advantage of requiring low voltages for operation which protects

students from exposures to high-voltage sources. The multi-pixel photon counter S12572-50P has 3 ×

3 mm2 area with 3600 individual cells with a pixel pitch of 50 µm and it uses an operational voltage

of ∼ 68.75 V (see table 2.3) at 25 degrees Celsius. Such voltage can be generated from a 5 V input

using a DC-DC converter, making it possible to make the detector portable. The major downsides of this

device resides in its gain variation with temperature and relatively high noise which makes the separation

between signal and noise quite challenging. Some characteristics of the MPPC can be seen in table

2.3.

Hamamatsu S12572-50PPeak Sensitivity (λp) 450 nm

PDE (λ = λp) 35 %Dark Count (0.5 phe) 1-2 ×106 Hz

Gain 1.25 ×106

Gain Temperature Coeff (∆TM ) 2.7×104 oC−1

Breakdown Voltage (25oC) 64.15 VBias Voltage (25oC) 66.75 V

Terminal Capacitance 320 pFVBias variation with T (∆TV ) 60 mVoC−1

Table 2.3: Characteristics of the MPPC Hamamatsu S12572-50P.

22

2.4 Signal Estimate

With the key elements of the CMuD defined it is now possible to make a crude estimation of the signal

a muon would generate. This signal represents the number of photoelectrons (phe) that will result after

all the steps, from the photons creation to its translation to an electrical signal, are completed. With

this estimate it will be possible to predict the order of magnitude of the threshold that should be applied

to detect these particles. The calculation of the signal takes into account five separate processes: the

energy deposition in the scintillator by the ionizing particle, the conversion of that energy into photons,

the collection of those photons by the WLS fibers, the re-emission of the photons in a new wavelength

inside the fibers and the detection of these photons in the silicon photomultiplier. The signal, Nphe, is the

product of the number of photons generated, Nγ by the collection efficiency , εcoll, trapping efficiency,

εtrap and the SiPM photon detection efficiency, εPDE (see equation 2.1).

Nphe = Nγ εcoll εtrap εPDE (2.1)

2.4.1 Energy Deposition

When a muon crosses a scintillator it transfers some of its energy to it. This energy exchange happens

primarily through ionization and atomic excitation. This energy is used to promote electrons to excited

states and as they return to a more stable orbitals, photons are emitted. Therefore, in order to estimate

the number of photons generated by a passing muon it is necessary to calculate the energy this particle

transfers to the scintillator.

The mean rate of energy loss (stopping power) per distance traveled due to ionization and atomic

excitation, expressed in MeV g−1cm2 is given by the Bethe-Bloch equation[31][32],

−⟨dEdx

⟩= Kz2Z

A

1

β

[1

2ln(2mec

2β2γ2TmaxI2

)− β2 − δ(βγ)

2

]. (2.2)

Where:

• K = 4πNAr2em

2ec

2 = 0.307 MeVg−1cm2,

• z corresponds to the charge of the incident particle,

• Z and A are, respectively, the atomic number and atomic mass of the scintillator,

• β = vc , the particles velocity by the speed of light,

• me represents the electron mass at rest

• γ = 1√1−β2

is the Lorentz factor

• I denotes the mean excitation energy value of the medium atoms,

• Tmax represents the maximum transferable kinetic energy in a single collision and

• δ(βγ) stands for the density effect correlation to ionization energy loss.

23

As equation 2.2 shows, the particle energy loss by ionization is relatively independent of the medium,

only with a factor ZA to represent its influence, which is also constant for many elements. This is con-

firmed in figure 2.5 where the muon energy loss in very different mediums shows only a small variation.

Nevertheless, it is important to mention that the equation 2.2 does not take into account the density

of the medium, only its composition. Therefore, this factor ρ must be later added in the energy loss

calculation.

Figure 2.5: Muon energy loss in different materials in relation to its momentum. Radiative effects,relevant for muons are not included. These only become significant for particles with higher momentumand absorbers with higher Z.

Figure 2.5 shows the Bethe-Bloch equation (2.2) for a muon traversing different mediums. As it can

be observed, independently of the material, there is a minimum energy loss which corresponds to the

so called minimum ionizing particle (MIP). The energy of a MIP seems to be basically constant and

independent of the medium between pµ = 0.3 − 0.4 GeV/c. The energy deposited will then increase

slightly for particles with higher energies. It is usual to treat particles with energies higher than EMIP as

MIPs. Since cosmic muons, as most relativistics particles, are essentially MIPs it is possible to estimate

the mean rate of energy loss in this muon detector. The EJ-200 is a plastic scintillator mainly made of

polymers of carbon and hydrogen compounds. We will consider the medium as pure carbon and use its

MIP value to estimate the energy deposition from there.

1

ρ

⟨dEdx

⟩C≈ 2 MeV/g/cm2 (2.3)

dEloss = 2 ρ dx (2.4)

It is important to mention that this predicts the mean energy loss. This energy deposited variation

is given by the Landau distribution and is particularly relevant for thin layers or low density materials

since it shows large fluctuations towards high losses (Landau tail). The Landau distribution tends to a

24

Gaussian-like distribution when the thickness increases.

In equation 2.4, dx represents the infinitesimal progress of the particle inside the medium and ρ its

density. The density of the EJ-200 scintillator can be found in table 2.1 to correspond to ρ = 1.023

g cm−3. Integrating equation 2.4 along the particles path (see figure 2.6) we obtain:

Eloss = 2 ρa

cosφ(2.5)

Figure 2.6: Representation of a particle crossing the CMuD.[33]

Where a = 1.25 cm is the thickness of the scintillator and φ the incidence angle of the muon. Since

cosmic muons are the result of complex interaction with the atmosphere the incidence angle at which

these particles traverse the scintillator is not fixed. Nevertheless, there are studies performed on the

angular distribution of these particles.

Figure 2.7 shows the muon angular distribution has a dependence of cos2 φ in relation to the zenith[34].

Figure 2.7: Angular distribution of cosmic muons at ground level.[35]

25

Considering φ = 0 the most probable incidence angle we obtain:

Eloss = 2.56 MeV. (2.6)

2.4.2 Light Yield

The scintillation efficiency of the conversion of excitation energy to photons is an intrinsic characteristic

of the scintillator that can be found in table 2.1. The light yield of the EJ-200 is of 10000 photons for each

MeV of deposited energy. Therefore, the light yield is then given by:

Nγ ≈ 25 600 photons. (2.7)

This represents the typical number of photons generated by a cosmic muon that has MIP energy and

a incident angle perpendicular to the detector.

2.4.3 Collection Efficiency

The next step is to estimate the percentage of the photons generated that are collected by the optical

fibers.

Figure 2.8: Geant4 simulation of photons generated in a scintillator by an ionizing particle.[36]

A rough approximation of the collection efficiency was performed by considering the portion of pho-

tons that reach the scintillator’s top surface where the fibers are placed. We assumed that all photons

are generated isotropically along the muon path - an image of a Gent4 simulation is presented in figure

2.8.

It was considered that the photons only have two possible paths to reach the top surface, directly,

when the photons are emitted with the direction of the top surface, or indirectly, when they are directed at

the bottom surface and are reflected to the top surface. The photons are reflected by the Tyvek material

wrapped around the scintillator with an efficiency of R = 90%.

The fraction of photons that are directed to the top and bottom surface is approximated by the ratio

of the surface areas by total area of the scintillator. The total area of the scintillator,AT was calculated

26

by adding the area of each side - AT = 900 cm2. Since the top and bottom surface have the same

dimensions their area is the same and equal to AS = 400 cm2. The portion of photons that reach the

top surface, Γtop is therefore given by:

Γtop =ASAT

+ASAT

R ≈ 0.84% (2.8)

From the photons that reach the top surface, only the ones that reach the surface covered by the WLS

fibers are collected. In total there are nine fibers with 0.1 cm of diameter spread along the scintillator

with two centimeters of interval between them, as showed in figure 2.6. Accordingly, we can calculate

the area occupied by the fibers, Afib = 18 cm2. If the percentage of photons collected by the fibers is

proportional to the area they occupy in the top surface, the collection efficiency can be calculated by:

εcoll ≈ ΓtopAfibAS

≈ 3.8% (2.9)

The method uses rough approximations and should be compared with other efficiencies calculated

by computational simulations. The study [33] uses the G3sim software to simulate a scintillator detector

with an area of 50 × 50 and 18 WLS fibers placed in the same manner as our detector. The fibers have

1 mm diameter and the scintillator a thickness of 2 cm. The collection efficiency obtained in this study

was:

εcollstd = 9.8% (2.10)

The study reveals a collection efficiency in the same order of magnitude as our approximation. Con-

sidering that in this study the ratio between the number of fibers and scintillator surface area is smaller

than in the CMuD it would be expected that our detector had an higher collection efficiency. In any case

our calculation can be considered an underestimation of the collection efficiency and will be used in the

signal calculation.

2.4.4 Trapping Efficiency

The purpose of WLS fibers is to absorb the photons that enter the fibers and then emit a new photon

in a new wavelength, that better matches the efficiency peak of the SiPM. The propagation of a photon

inside these fibers can be see in figure 2.9.

The emission spectrum of the scintillator in figure 2.2 shows photons wavelength distribution, the

maximum seems to be λ = 425nm which correspond to violet color. The energy of the photon is

important to understand how it interacts with the fiber. Since it is a low energy gamma the most likely

interaction is the photoelectric which corresponds to the excitation of one electron and its later fall to the

a stable orbital that emits a different photon, this one with a shifted wavelength. Although the photon

conversion has a one to one ratio, not all photons are collected by the fibers. This is so, as a result of the

isotropic emission of these photons. The trapping efficiency of these photons is an intrinsic characteristic

of the WLS fiber that depend of the refractive index of the materials of each layer. The trapping efficiency

27

Figure 2.9: Photons propagating inside a wavelength shifting fiber.[36]

of the Y-200 can be seen in table 2.2. Transmission and bending losses inside the fiber where ignored.

εtrap = 5.4%. (2.11)

2.4.5 Photon Detection Efficiency

The last step is the conversion of the photons that travel along the fibers onto the MPPC into photo-

electrons. The photon detection efficiency depends on three parameters, them being: the fill form of the

silicon, a geometrical factor that depends on the active area in respect to the total area of the device,

the quantum efficiency that is the probability of a photon to generate a charge carrier, which by itself

depends on the photon wavelength, and finally the probability that this carrier generates an avalanche

which depends on the overvoltage. The PDE considered corresponds to the efficiency peak of the

photo-detector with an overvoltage of 3.5 V.

εPDE = 0.35 (2.12)

2.4.6 Signal

Now we have all the information necessary to estimate the average signal a muon will produce. With

the number of photons generated and the efficiency of all the respective steps onto the electrical signal.

These factors can now be substituted in equation 2.1 to obtain:

Signal = 25 600 ∗ 0.038 ∗ 0.054 ∗ 0.35 ≈ 18 ph.e (2.13)

This represents a rough estimate since not all the factor are being considered such as photon ab-

sorption by the scintillator, fiber losses and connection defects between the components. As well as the

problems stated in the determination in the collection efficiency factor.

28

Chapter 3

Silicon Photomultiplier Study

The silicon photomultiplier consists of an array of APD cells, each operated in Geiger-Mode. Electron-

ically speaking, the cells of the silicon photomultiplier are connected in parallel, sharing the same bias

voltage applied to the whole SiPM, see figure 3.1. Each cell has a series quenching resistor Rq. If the

voltage across the APD becomes larger than the intrinsic breakdown voltage of the diode, the electric

field will be high enough to accelerate electron-hole pairs, created by a triggering photon or thermal

noise, across the junction. The current generated leads to a voltage increase at the quenching resistor,

Rq. Consequently, the voltage across the diode decreases and the electric field is not strong enough to

accelerate charge carriers through the junction anymore. Thus the avalanche process is ‘quenched’. Be-

cause no current flows through the diode at this time, the voltage at the diode increases again, reaching

the bias voltage and the process restarts.

Figure 3.1: Arrangement of the cells of the SiPM in parallel with series quenching resistor.[37]

Having explained the basic working principle of a single SiPM cell it is now possible to describe the

complete SiPM as an array of Geiger Mode - APDs. In this section, the gain variation and the dark

current rate of the SiPM were studied. When talking about the SiPM response it is important to know

that its output signal has well defined amplitudes which is shown in 3.2. This height depends on the

number of cells that discharged and of the gain produced in the avalanche which itself depends on the

applied bias voltage. The smallest pulse a SiPM is able to output is referred to as a one photo-electron

equivalent (1 ph.e.). Larger pulses are integer multiples of this signal with some uncertainty due to

29

statistical processes in the cell.

Figure 3.2: MPPC Hamamatsu S12572-50P The pulses have well defined amplitudes that are multipleof the number of one ph.e. and proportional to the number of cells that discharged.

A useful concept that is often used in the context of SiPMs is the overvoltage, which is defined as,

VOv = VBias − VBd, (3.1)

where VBias denotes the applied bias voltage and VBd is the breakdown voltage above which the cell

enters Geiger-mode. The breakdown voltage is an intrinsic parameter of the device that can be seen in

table 2.3. The overvoltage is defined has the excess bias voltage applied with respect to the breakdown

voltage.

With the increase in temperature the band gap energy in the p-n junction of the SiPM decreases. The

interatomic spacing increases with the thermal energy because the amplitude of the atomic vibrations

also increase. As a consequence the breakdown voltage of the MPPC varies with temperature by

following equation,

VBd(T ) = VBd(T0) + ∆TM(T − T0), (3.2)

where the factor ∆TV = 60 mV oC−1 is given in table 2.3. Many properties of the SiPM like gain,

(correlated) noise effects and PDE depend on the overvoltage. These properties have to be character-

ized at constant temperature or at least at constant overvoltage.

The dominating noise effect on longer time-scales is thermal noise, generated randomly by thermally

induced charge carriers. Each time a pair is produced by thermal effect the silicon will produce a similar

pulse as if it was photon triggered.

Electrons and holes of an avalanche may recombine in an APD cell generating a photon which can be

detected in another cell. This is called optical crosstalk. This effect can be reduced by adding trenches

between individual cells and is currently being introduced in the manufacturing process of SiPMs.

The third effect is entitled afterpulsing. This effect dominates especially in impure silicon and can be

reduced by producing purer silicon. Free carriers are blocked within the silicon due to lattice scattering

30

and can create an avalanche in the same cell shortly after.

3.1 Material

This section introduces the material used in the experimental setups employed to test the characteristics

of the MPPC and CMuD. It is composed by a small introduction and summary of the most important

technical information of each component.

3.1.1 SiPM Circuit

The circuit seen in figure 3.3 was used to bias the device. The bias voltage is represented by HV+ and

the output signal is collected at the node labeled ‘signal’.

Figure 3.3: SiPM circuit used.

The capacitor C1 in with association with resistor R1 creates a low pass filter that prevents sharp

altercations in the MPPC supply voltage, helping to protect the device. The capacitor C2 placed before

the signal output does the opposite, being a high pass filter cuts all the DC component and, therefore,

only quick variation generates a signal (SiPM discharge). The SiPM works as a switch: in the absence

of discharges no current passes through; when it discharges the circuit closes and there is a rapid flow

of electrons that decreases the tension and generates a negative signal output.

3.1.2 Keithley 6487

The Keithley 6487 is a high precision picoammeter/voltage source with 5 digit precision. It has a voltage

resolution of 0.2 mV and can reach up to 500 V. This device was used to generate the bias voltage to

the SiPM circuit.

3.1.3 Amplification board TB-411 8+

The amplifier, MAR-8ASM+, from Mini-Circuits is integrated in a small board, TB-411 8+, and uses

SMA female connectors for the input and output signal (see figure 3.4). The amplifier uses a supply

voltage of 12 V and presents a gain of ≈31 dB at a frequency of 50 MHz. It also inverts the signal upon

amplification.

31

Figure 3.4: TB-411 8+ low noise amplifier board.

3.1.4 Oscilloscope - Tektronix TDS3032

The Tektronix TDS3032 is a dual channel digital phosphor oscilloscope with a bandwidth of 300 MHz

and takes up to 2.5 GS/s with a storage depth of 10 kS. The digitizer has a resolution of 9 bit and ranges

from 2 mV/div to 1 V/div with a total of 8 div on the screen. The horizontal resolution ranges from 2

ns/div to 10 s/div at 10 div on the screen. The TDS3032 also features an Ethernet communication port

and interface that was used to transfer data to a computer.

3.1.5 NIM Instrumentation

The use of modular crate electronics is common in particle physics experiments because they allow

circuits to design and put together very quickly. By dividing the the circuit into individual modules it is

possible to see the physical logical connections but also when the experiment is done, can be removed

and used again. All modular electronics utilized in this thesis use LEMO-type connectors for all inputs

and outputs. A short description of the modules used follows.

LeCroy Octal Discriminator - Model 623 A

This discriminator module generates a logic pulse when the input signal surpasses the defined threshold.

The minimum threshold of the Model 623 A is -30 mV, variable up to -1 V via front- panel screwdriver

adjustment. A monitor point is provided to permit measurement of the threshold level with a voltmeter.

The delay between the input and output signal is of 11 ns.

Canberra - Dual-Timer Counter

The Dual Counter/Timer module can count the number of pulses that reach it. In combination with the

discriminator module it creates a simple signal processing unit able to determine particle rates. This

module can count rates greater than 250 MHz which is more than enough for the applications needed

in this thesis. It also includes programmable timer resolution and counters. All functions, including

threshold settings, are programmable through the instrument’s front panel.

32

LeCroy Model 364AL 4-Fold Logic Unit

This coincidence module is made of two separate coincidence units. Each with the ability to perform

basic logical operations, such as: AND, OR, complement, and more. The output pulse duration is front-

panel switch-selected to be either the time overlap of the input signals or to be a clipped 3.8 ± 0.3 ns

duration with inputs> 5 ns or equal to time overlap. The delay between input and output is approximately

6 ns.

3.1.6 PREC Front-end

The use of the PREC front-end, shown in figure 3.5, allows to eliminate a few of the components men-

tioned before by integrating them all in one board. The front-end incorporates the amplification and

discrimination of the signal in one circuit. The goal is to eliminate as much electronic noise as possible.

The digital signal is then fed to an FPGA board that performs the logical operations intended.

Figure 3.5: Picture of the Front-end board.

The FE board has 8 channels and a control system. Each channel is composed by an input, am-

plification and discrimination stage. A final output stage buffers the signals. The amplification stage

relies on the MAR-8ASM+ amplifier from Mini-Circuits which is the same model as the one used in

TB-411board. The input and output are AC-coupled by capacitors. The discrimination stage uses the

comparator ADCMP604. This is a fast comparator with a LVDS output to minimize pickup noise. The

reference threshold is set by a DAC controlled by I2C. This DAC has a maximum output voltage of 400

mV (maximum threshold) and a 12 bit resolution. The threshold is set by a raspberry pi that uses a

software interface accessible through the local network.

3.1.7 Altera DE2 Development and Education Board

The DE2 board, seen in figure 3.6, is an FPGA development platform designed for university and college

laboratory use. It is suitable for a wide range of exercises in courses on digital logic and computer

organization, from simple tasks that illustrate fundamental concepts to advanced designs. It uses the

FPGA Cyclone II EP2C35F672C6 with EPCS16 16-Mbit serial configuration. A USB Blaster interface is

33

used to configure the board. This board has incorporated two different clocks, one of 27 MHz and of 50

Mhz. This board receives the input from the front-end via the GPIO interface.

Figure 3.6: Altera DE2 Board.

3.2 Gain Variation

Gain is the internal amplification of the SiPM expressed as the average number of charge carriers

produced from a single original pair electron-hole. Its gain depends on the overvoltage voltage applied.

If the electric field in the p-n junction is higher when an electron-hole pair is generated the acceleration

is also higher, producing a larger number of carriers in the avalanche.

3.2.1 Methods

The number of carriers generated in a SiPM discharge is proportional to the charge in the resulting

signal. As seen in figure 3.2, there are well established pulse heights according to the number of cells

that discharged. By integrating the output pulse we can calculate the number of charge carriers that

produced it and since the discharge is generated by a single (or a multiple) ph.e. it is possible to relate

it to the gain of the SiPM.

To observe the SiPM signal it was used the setup shown in figure 3.7. The Keithley was used

to bias the SiPM circuit. The output signal was amplified using the TB-411 board in order to allow the

visualization of the signal in the TDS3032 oscilloscope. The signal acquired was then sent to a computer

via Ethernet connection where the integration was be performed.

This integral result gives the charge of the amplified signal instead of the desired SiPM discharge.

Therefore, to determine the gain in the SiPM we need to find the gain of the operational amplifier. This

was accomplished by measuring the discharge of a regular capacitor.

The same set-up was used, only substituting the SiPM by a capacitor and the Keithley’s DC voltage

source by a rectangular wave. That way the capacitor would continuously charge and discharge. The

charge, Qt, in a capacitor is given by,

34

Figure 3.7: Experimental setup used to measure the SiPM gain.

Qt = C ∆V, (3.3)

where C is the capacitance of the capacitor and ∆V the wave amplitude.

The charge can also be calculated by the integral of the discharge pulse. The ratio between the

theoretical charge, Qt and integral, Qpulse, will give the gain, G, added by the TD-411 amplifier. This

conversion factor can then be used to determine the gain of the SiPM.

G =QpulseQt

=Qpulse∆V C

(3.4)

3.2.2 Results

To determine the conversion factor it was chosen a capacitor with the capacitance in the same order of

magnitude as the SiPM (320 pF, see table 2.3). In this case we used a 100 pF capacitor supplied by a

rectangular signal with a frequency of 5 kHz.

Figure 3.8: Amplified signal of a 100 pF capacitor discharge, charged by tension of 300 mV. Acquiredby the oscilloscope.

The pulse area was then determined for different amplitudes of the rectangular wave. The value

of each measurement was obtained by acquiring 3 different discharges for each wave amplitude and

determining its mean. The error is given by the maximum deviation from it.

35

Figure 3.9: Variation of the charge in a 100 pF capacitor with amplitude of the supply signal.

A linear regression was fitted to obtain the ratio Qpulse

∆V = 4.28 × 10−8, seen in 3.9. Using equation

3.4 we can calculate the gain to be:

G ≈ 428 (3.5)

With the conversion factor determined, the same procedure was applied for the SiPM signal. The

charge was determined by integrating the SiPM output pulse. The oscilloscope cursor was employed to

mark the 1 ph.e. peak (see figure 3.2) and used to select the pulses that resulted from that discharge.

Figure 3.10 shows a SiPM pulse corresponding to 1 ph.e.

Figure 3.10: Example of a 1 ph.e pulse with SiPM biased at 65.55 V.

Three pulses were acquired for each overvoltage value. The resulting charge was divided by the

gain of the amplifier using the conversion factor and expressed in number of electrons. The result is the

number of electrons that were generated in the avalanche inside the SiPM cell. The data was plotted as

a function of the overvoltage in figure 3.11.

The gain of the MPPC shows a linear relation with the overvoltage with a slope 5.28 × 105. To the

recommended bias voltage, VBias = 66.75V , corresponds a gain of, G ≈ 1.0× 106.

36

Figure 3.11: Plot of the SiPM gain as a function of the overvoltage and corresponding linear fit.

3.3 Dark Current & Photo-electron Determination

In the absence of photons, the thermal motion of electrons-holes inside the depletion region can be

enough for it to jump to the conduction band and thus triggering a discharge inside that cell. The events

created by this process contribute to the noise of the photo-detector. This random process occurs

independently inside each cell and it’s rate depends only of the temperature and the size of the depletion

region of the cell. Therefore there is a fixed rate of false events always present in the SiPM. This rate

is called dark current, since it is the signal that the SiPM produces even in the absence of light. The

determination of a photo-electron is important to define a scale that is independent of the amplification

or circuit used to measure the signal.

3.3.1 Methods

The first step to measure the dark current was to seal the SiPM with black duct tape and to place it inside

a metallic box to prevent contamination from external light sources (figure 3.12). In such a way that it is

assured that only the dark current rate will be measured.

Figure 3.12: SiPM circuit in metal box.

For the measurement of the dark current and photo-electron determination the setup shown in figure

3.13 was used. The purpose of this setup is to count the number of dark current events which amplitude

37

surpasses a given threshold. Since the SiPM signal’s amplitude is proportional to the number of cells

that discharged and these discharges occur in a random process, the probability of a number of cells to

discharge at the same time should follow a Poisson distribution. Therefore by varying the threshold we

expect the dark current rate to decay exponentially.

The determination of a photo-electron can be obtained directly from the dark current rate curve. Since

this rate should decrease in, more or less, well defined steps of one photo-electron (see the SiPM signal

in figure 3.2), by measuring the distance between these steps we can obtain the voltage corresponding

to one ph.e.

To be able to identify these steps, the setup used should have a good threshold resolution, providing

several points in the transitions zones. This was achieved by using the Front-end board as a discrimi-

nator. The Front-end board receives the signal through a MMCX connector amplifies and discriminates

it. The resulting signal is then connected to the GPIO input of the DE2 board where a counter is pro-

grammed. This allows to eliminate the external amplification step and some cable connections which

helps to reduce the noise. The setup can be seen in figure 3.13.

Figure 3.13: On top is represented a scheme of the main components of the setup. Bellow, the pictureshows the setup used for the measurement of the dark current. The components are: 1-Keithley VoltageSource, 2 - MPPC circuit, 3 - Amplifier TB-411 (Not Used), 4-5 - Front-end Voltage sources, 6 - Front-end, 7 - Raspberry Pi (used to program the front-end) and 8 - Ethernet output to GPIO of DE2.

The signal from the MPPC is connected to one channel of the front-end where it is amplified and

discriminated. The threshold of the comparator used in the discrimination stage can be set with 12bit

38

resolution which allows to finer threshold resolution. A simple counter software was written and pro-

grammed in the FPGA. The counter used a 50 MHz clock to define a time window and the number of

events were shown in the 7 segment display. The count is initiated and reset with a switch.

For the measurement of the dark current, the discriminator threshold was set to 4 mV and incre-

mented in steps of 0.5 mV. In the counter it was defined a time window, tw, of 10 s. When the counter

is started it counts events until the timer reaches zero and stops. The number of events is shown on its

display. The SiPM is supplied with a VBias = 67.25 V which correspond to an overvoltage of 3.1 V. This

was also the bias voltage used throughout the study of the CMuD.

The resulting dark current rate is obtained by dividing the number of events by the time window.

RDC =n

tw(3.6)

The respective error is given by equation 3.7.

eRDC=

√n

tw(3.7)

3.3.2 Results

In figure 3.14 is represented the data acquired corresponding to the dark current. In the lower thresholds

it is possible to observe a few plateaus that correspond to the voltage gap between each ph.e, which

can be seen in figure 3.2. When the threshold is in this gap the rate decreases much slower since there

are few events with that range of amplitudes. As the threshold increases the steps blend with the main

curve and disappear. This is because there is always some uncertainty in the signal resulting from a

discharge of a cell, that can be seen by the thickness of the peaks shown figure 3.2. Therefore, when

the signal is the sum of many ph.e, the uncertainties are enough to disperse the signal over different

amplitudes.

Figure 3.14: Plot of the dark current rate in logarithmic scale.

39

To determine the ph.e voltage, the threshold distance between each visible step was measured. To

achieve this, the derivative of the dark current rate shown in figure 3.14 was performed. The result is

plotted in 3.15.

Figure 3.15: Plot of the inverted dark current derivative.

The plot shows peaks and valleys corresponding, respectively, to the drops and plateaus of the steps

seen in the dark current measurement. To determine the photo-electron voltage the first four valleys

were used and the distances averaged to obtain:

1 ph.e = 3.3± 0.1mV. (3.8)

This means that the amplitude of the output signal increases about 3.3 mV for each ph.e, i.e., cell

discharge. The step corresponding to 1 ph.e is not represented in the plot because the Front-end

limitations did not allow lower thresholds. From now on the threshold voltage will be expressed in ph.e

scale using this scaling factor.

As expected, the dark current appears to follow a Poisson distribution and decrease exponentially. To

estimate the rate at which it decreases a fit was attempted. To perform this, only the data that represents

the threshold between each ph.e peak was selected. This was done because the amplitude of the

signal corresponding to a certain number of cells discharge has some statistical variations. Therefore,

by using the data point in the middle of the transition zones between ph.e. we ensure to count all the

events corresponding to that number of discharged cells. The first data points were easily taken from the

plateaus corresponding to the gap between ph.e. peaks, seen in figure 3.14. For higher thresholds those

points were estimated by incrementing the previous point by a step of one ph.e, calculated previously

(equation 3.8). The data selected was plotted and fitted in figure 3.16.

The fit gives the approximate rate at which the logarithmic dark current rate decreases with the

threshold in photo-electrons.

log10 [RDC(Th) ] = −0.194Th + 5.81 (3.9)

40

Figure 3.16: Linear fit of the dark current rate in logarithmic scale versus threshold in ph.e.

41

42

Chapter 4

CMuD Characterization

In this chapter some of the characteristics of the CMuD were determined using different setups. In the

first setup the efficiency and purity of the CMuD were studied. These characteristics were also measured

for a CMuD in self-coincidence and with two CMuD in coincidence.

4.1 Methods

To assemble the Compact Muon Detector the first step was to build a new circuit board for the SiPM.

Although the same circuit, seen in figure 3.3, was used the size was greatly reduced to fit into a small

metal box. The result is shown in figure 4.1. The photo-detector was elevated to fit a small square

opening in the lid where the fibers couple with the SiPM.

Figure 4.1: New SiPM circuit used to assemble the CMuD.

4.1.1 Efficiency

The efficiency is a very important characteristic of a detector. It represents the probability of the detector

‘seeing’ a particle which passed through it.

To measure the efficiency a simple hodoscope was implemented. The hodoscope is capable of

signaling muons that are certain to cross the CMuD generating a trigger. The efficiency is given by the

43

number of times the CMuD generated signal, nD in coincidence with the hodoscope trigger over the

number of triggers generated by the hodoscope, nµ:

ε =nDnµ, (4.1)

with an uncertainty given by the binomial probability density function and the Bayes theorem[38]:

σε =

√(nD + 1 )(nD + 2 )

(nµ + 2 )(nµ + 3)− (nD + 1 )2

(nµ + 2)2. (4.2)

Muon Hodoscope

The muon hodoscope was implemented using two small scintillator detectors operated in coincidence.

A trigger is issued when a particle crosses both scintillators. The scintillators must have a smaller area

than the CMuD in order to restrain the solid angle and consequently assure that if a coincidence occurs

that particle had to cross the CMuD. The scintillation detectors used can be seen in figure 4.2.

a)Top scintillator. b) Bottom scintillator

Figure 4.2: The two scintillation detectors used to produce a muon trigger.

The top scintillator is a very simple device, consisting of a 7×7 cm2 square plastic scintillator coupled

directly to a photomultiplier tube (PMT) placed on top. The bottom scintillator has a paddle shape with

an area of 10×20 cm2 and uses an optical guide to carry the photons from the scintillator to the PMT. As

the picture shows both are covered with black tape to prevent the signal contamination by exterior light

sources. After biasing each PMT with a high voltage source of 1500 V an analysis of both signals was

conducted using the TDS3032 oscilloscope to determine the adequate threshold for each scintillator.

This allowed an optimization of the coincidence trigger lowering the rate of possible false events while

maintaining a reasonable efficiency.

Their signal showed a difference of about one order of magnitude in the output signals. With the top

scintillator having a peak value of −250 mV and the bottom −30 mV (see figure 4.3). Since the PMT’s

response is proportional to the number of photons generated in the scintillator, this discrepancy can be

44

credited to photon loss in the optical guides since it is the only significant variant between the detectors.

a)Top scintillator signal. b) Bottom scintillator signal

Figure 4.3: Triggered event in each individual scintillator and respective discriminator output.

The discrimination of the these signals was done by using the Octal Discrimination Module 623A,

discussed in section 3.1.5. The thresholds were set to −200 mV and −30 mV to the top and bottom

reference detectors respectively. In figure 4.3 can be seen the signal before and after the discrimination.

The discriminator output is NIM a digital signal with a fixed time length. This time length represents the

coincidence time window that is around 60 ns. The delay this module adds - ∼ 11 ns - can be seen in

figure 4.3 by the time difference between the original signal and the discriminated output pulse.

The discriminated signal was connected to the Dual-Timer Counter module, discussed in section

3.1.5, in order to make a quick measurement of the detection rate of each reference detector. The

results are presented in table 4.1.

Top Detector Bottom DetectorRate (Hz) 6.0 ± 0.4 5.0 ± 0.4

Table 4.1: Measurement of the event rate of each individual scintillation detector.

This measurement was done with a 40 s acquisition window. It would be expected that the bottom

detector would have the highest rate due to the larger detection area instead they show similar rates.

But it is likely that due to the threshold limitations of the discriminator some events are lost.

It is possible to make an estimation of the false coincidence event rate with the following equation:

Rfake = R1R2 δ (4.3)

where R1 and R2 correspond to the frequency of two independent events and δ the coincidence time

window. This equation gives the probability of coincidence between two uncorrelated events.

Rfake = 1.8× 10−6 Hz (4.4)

Such a low rate of fake events means that when there is a coincidence between the two detectors it

is very likely that it corresponds to a correlated event, i.e. a muon traversing both detectors.

45

The hodoscope was completed by adding the 364AL 4-Fold Logic Unit module, addressed in section

3.1.5, to make the coincidence of the two discriminated signals. This module receives both outputs

from the discriminator and outputs the logical conjunction of them, identical to a logic AND gate. When

a coincidence takes place it produces an output pulse. The output was then connected to the counter

module. The coincidence rate was then measured for two different setups. In the first setup the detectors

were placed one on top of the other, aligning their detection areas. In the second case they were placed

side by side minimizing their alignment.

Aligned Side by SideCount 138 ± 12 2 ± 1

Rate (Hz) 0.23 ± 0.02 0.003 ± 0.002

Table 4.2: Count & Rate of detected muons in a 600 s acquisition for two different set-ups.

The rate of the aligned setup shows that, on average, around one muon passes through the detectors

in every four seconds, far from the 5 Hz rate measured before. Nevertheless, it can be explained by the

decrease of false events due to noise in the PMT, scintillator and electronics and the acceptance angle

that is greatly reduced in the coincidence setup.

According to the estimation, in equation 4.4, the rate of false events should be much smaller than

the one measured (table 4.2. However, that estimate considered the probability of unrelated events to

coincide yet it is possible that these two events are a consequence of a extensive air shower or muon

with a ‘strange’ incidence’ angle. To confirm this idea the measurement was repeated and on all of those

the result was zero. These results give strength to the suspicion that it was a result of a sporadic event

and not of systematic origin. The system is capable of signaling muons crossing the detectors while

minimizing false events.

CMuD Counter Setup

When the hodoscope is triggered by a muon, the signal from the CMuD is tested to see if an event is

also detected or not. This is achieved by making the coincidence of both signals. That coincidence

is performed in the FPGA. To do this, the muon trigger signal (output of the coincidence module) is

connected to one of the channels of the Front-end to invert the signal and feed it to the FPGA. Since this

signal is digital the amplification and discrimination process in the Front-end is not relevant. If the muon

trigger circuit did not add delay in the process the coincidence would be a simple AND gate. However,

each individual module adds some delay to the signal. Delay is also added in the PMT’s and through

the propagation in wires.

While the CMuD signal is directly connected from the detector to the Front-end (see figure 3.13),

the muon trigger goes through: PMT → Discriminator Module → Coincidence Module → Front-end.

By comparing the both circuit and summing the delays in the extra modules and wiring it was possible

to estimate of the delay. The PMT’s used in the scintillation detectors are Hamamatsu’s model H6410

which have a transit time of ∼ 41 ns. The discriminator has a delay of 11 ns and the coincidence module

6 ns. The delay added by the wiring is proportional to the length of the cables used which adds to about

46

20 ns. In total there should be a delay of about 78 ns.

The Quartus II software, used to program the DE2 board, has an internal logic analyser - SignalTap-

that allows to capture data overtime according to a trigger. This tool was used to see the delay between

both signals.

In figure 4.4 it is shown the signal of the hodoscope trigger, labeled GPIO 1[15], and the signal from

the CMuD, labeled GPIO 1[11]. It is also shown an auxiliary signal for the signal tap trigger that is

generated using a monostable module with a time constant of 100 ns. The data is acquired in steps of

20 ns. From the data it can be estimated that the hodoscope trigger arrives between 60-80 ns after the

CMuD signal.

Figure 4.4: SignalTap - Delay between the hodoscope trigger and CMuD signal. From top to bottom, theGPIO 1[15] corresponds to the muon trigger, the GPIO 1[11] to the CMuD signal and in the bottom isthe monostable variable.

A delay module was added to the code to shift the CMuD signal by 60 ns. This delayed signal

and hodoscope trigger were then fed to a counter implemented in the FPGA. Whenever a hodoscope

trigger is received its counter is incremented. If the CMuD also has a signal in coincidence within a

programmed time window the corresponding counter is also incremented. With the two counter and

coincidence system in place it was possible to measure the efficiency.

Finally, before start the acquiring, all the detectors were placed inside the dark box shown in figure

4.5, to avoid light contamination from outside sources in the scintillators and/or fibers.

Figure 4.5: Dark box used to isolate the set-up.

The cables used to feed the PMT’s and the SiPM were slipped to the inside of the box through the

opening on its right side that is covered with dark and thick cloth to prevent the intrusion of light. The

scintillators were placed vertically aligned with each other and with the CMuD in between.

47

4.1.2 Purity

The purity of a detector determines the probability of a detected event to correspond to a particle. The

purer the detector is the less false events are counted. The setup used to measure the purity is similar

to the one used in the efficiency measurement with some fundamental differences. Firstly the role of

the reference scintillators is not to generate a muon trigger but instead to certify, or not, if each event

the CMuD sees corresponds to a muon. This means that every particle that crosses the CMuD must

also cross and be detected by both reference scintillators. Hence, in this measurement, the reference

detectors must have a detection area larger than CMuD.

For the measurement of the purity two large paddle shape scintillation detector were used and are

shown figure 4.6. The top is about 150 × 50 cm2 in dimensions while the bottom is 40 × 25 cm. Both

scintillators use a PMT biased at 1.5 kV as a photo-detector.

a) Top scintillation detector.

b) Bottom scintillation detector.

Figure 4.6: Reference scintillation detector used in the purity measurement.

If every muon that crosses the CMuD triggered a coincidence in these detectors then we would

have a perfect method to determine the purity. Just by determining the percentage of total number of

events detected by the CMuD that matched a coincidence in the scintillation detectors. But, in reality,

the inefficiency of the scintillators requires the introduction of a correction factor. This factor corresponds

to the probability of a muon that crosses both reference scintillators to actually be detected. Assuming

48

these are independent events:

εcc = εtop εbot, (4.5)

where εcc is the correction factor, εtop and εbot are the efficiencies of the top and bottom scintillators

respectively.

To measure the efficiency of the scintillators it was used the same method as for the CMuD, described

in section 4.1.1. The threshold of test scintillator was set to −30 mV, the minimum the discriminator

module allows. The hodoscope was setup and the number of events of the test scintillator in coincidence

with the hodoscope was counted. The efficiency was calculated by equation 4.1:

εtop = 22.5± 0.8 %

εbot = 40.6± 2.3 %

Therefore, the correction factor of the purity is:

εcc = 9.1± 0.6 %. (4.6)

This correction factor means that only about 9% of the muons that cross the CMuD and the two

reference detectors generate a coincidence trigger.

The purity was given by the following equation,

P ∗ =nµnT

(4.7)

where number of events recorded by the reference scintillators, nµ, was divided by the total number

of events seen by the CMuD, nT . The error of this measurement is similar to error of the efficiency given

by equation 4.10.

σP ∗ =

√(nµ + 1 )(nµ + 2 )

(nT + 2 )(nT + 3)− (nµ + 1 )2

(nT + 2)2. (4.8)

The correction factor, εcc, due to the inefficiency of the reference detectors was added to obtain the

corrected purity:

P =P ∗

εcc. (4.9)

The error of the corrected purity was given by the non-linear combination of these variables:

σP =

√(σP ∗

εcc

)2

+

(P ∗ σεccε2cc

)2

. (4.10)

For the measurement of the purity the setup was moved to LIP’s the dark room to minimize ambient

49

light contamination. The detectors were also wrapped in thick dark plastic to ensure no light contami-

nates the measurements.

4.2 Results

In this section the results for three different configurations are presented. The efficiency and purity were

measured for each one of these setups. In the first setup the characteristics intrinsic to the CMuD were

studied. Then the CMuD was mounted in self-coincidence by dividing the 9 WLS fibers by two SiPM’s

and make the coincidence of their signals. In the third case a second CMuD was used to test the

performance in the telescope configuration.

Before the measurement it is necessary to define the time window that it will be used. The time

window is produced by the delay module programmed in the FPGA. The average delay was already

observed to be around 60 ns (see figure 4.4) but this value may vary. The light propagation time in the

scintillator, from generation to collection, varies with the position at which the particle crossed in the

scintillator. Hence, the delay may fluctuate by some nanoseconds that could be enough to miss the

clock’s rising edge. A set of efficiency measurements were done, with fixed threshold and different time

windows, to compare the results with different time windows. The CMuD was biased with 67.25V at the

temperature of 28oC.

Time Window (ns) Counters nDnµ

Efficiency (%)

20 4107± 6469.0± 0.6

5940± 77

40 6248± 7995.8± 0.3

6540± 81

60 7336± 8696.5± 0.2

7598± 87

Table 4.3: Efficiency measurement with a fixed threshold of 16ph.e. and different time windows.

The time of each acquisition varied which explains the significant difference in the total number of

triggers counted (nµ) between each measurement. The results seem to show a big jump from 20 to 40

ns time window. This could be explained by the delay fluctuations due to . There is also an increase from

40 to 60 but it is much smaller. This can be attributed to the increase of false coincidence probability due

to the increase of the time window. The time window was fixed to 40 ns for the following measurements.

4.2.1 CMuD

In this section it was studied the characteristics of the Compact Muon Detector. The experimental

procedures are detailed in section 4.1. All the results were expressed in photo-electron scale using the

scaling factor shown in equation 3.8.

50

Efficiency

The efficiency was calculated by the ratio seen in equation 4.1. Although the acquisition time varied,

each measurement took at least 24 hours, in order to obtain large counts and minimize the error. The

error of the efficiency is given by equation 4.10. The detector was biased with 67.25 V.

The events counted and efficiencies for different threshold values are presented in table 4.4.

Threshold (ph.e) Counters nDnµ

Efficiency (%)

13 6296± 7998.4± 0.2

6399± 80

16 6248± 7995.8± 0.3

6540± 81

19 6248± 7995.8± 0.4

6540± 81

22 4241± 6583.7± 0.5

5067± 71

25 4294± 6666.6± 0.6

6447± 80

28 3923± 6258.1± 0.6

6748± 82

Table 4.4: Efficiency of a CMuD for different thresholds.

For thresholds close to 13 ph.e. the probability of detection is almost 100%, which means that all

muons are detected. If there is no influence of false events this would mean basically every muon

generates a signal of at least 13 ph.e. in the CMuD.

To verify that the number of false events do not have influence in this result, the false coincidence

rate due to the dark current was calculated using equation 4.3. The dark current is greater in the lower

thresholds, therefore, the threshold of the smallest measurement was selected and calculated by the

Poisson distribution shown in equation 3.9 to be:

RDC(13) ≈ 1240Hz. (4.11)

Using equation 4.3, the fake coincidence rate between the hodoscope trigger, in table 4.2, and the

dark current rate noise was calculated:

Rfake(13ph.e) ≈ 1× 10−5Hz. (4.12)

With such a small rate the contribution of false events for the efficiency can be ignored.

The efficiency data is plotted in figure 4.7 as a function of threshold in photo-electrons. As it was

expected the efficiency decreases with the increase in the threshold. It seems to slowly decrease in the

smaller thresholds and to have a more pronounced drop near 20 ph.e.

51

Figure 4.7: Plot of the CMuD efficiency in ph.e scale.

Purity

The purity was determined using the expression in equation 4.9. In this measurement a flag was set

in the DE2 board that stopped the acquisition when the number of events seen by the CMuD reached

260 000. The CMuD was biased at 67.25 V.

Threshold (ph.e) Counters nµnT

Purity (%)

16 127± 110.54± 0.06

260 000

19 342± 191.5± 0.1

260 000

22 781± 793.3± 0.3

260 000

25 1813± 657.7± 0.5.

260 000

28 4922± 6620.8± 1.4

260 000

30 5718± 6224.2± 1.6

260 000

31 7677± 6232.5± 2.2

260 000

Table 4.5: Purity of a CMuD measured for different thresholds.

The data acquired was represented in table 4.5 and plotted in figure 4.8. As it would be expected the

higher the threshold the higher is the purity of the detector. This is a consequence to the decrease in

the dark current rate. If the minimum number of ph.e. generated by a muon in the CMuD is around 15

ph.e, it would be desirable for the purity at that threshold to be high. This would mean that this detector

would be able to detect basically all the particle that cross it with very little false counts.

Instead the purity seems to be very low at that threshold, only increasing for about ∼ 28 ph.e. This is

because the SiPM dark count rate at 15 ph.e. is still about 794 Hz - determined by equation 3.9. From

the same equation it is possible to estimate when the dark current of the SiPM becomes small enough

to allow an increase in purity. By calculating the zero of the equation 3.9 we can estimate when the dark

current becomes only 1 Hz.

52

Figure 4.8: Plot of the purity in ph.e scale.

RDC = 0→ Th0 =≈ 30ph.e (4.13)

The dark current becomes 1 Hz at the threshold of 30 ph.e. That is also around the double of the

minimum number photo-electrons produced by a muon, according to the efficiency measurement in

figure 4.7, which corresponds to a detection efficiency of about 50%. On of the best ways to increase

the purity while maintaining high overall efficiencies is through the use of coincidence. In the following

sections two method of coincidence were implemented and analyzed.

4.2.2 CMuD in Self-Coincidence

In this setup the CMuD was placed in self-coincidence. This was achieved by dividing the 9 WLS fibers

that make the detector into two different SiPM’s. This means that there are two signals coming from the

same detector. A coincidence of these two signals was performed and the resulting signal was taken as

the output of the CMuD. In figure 4.9 it can be seen the the fibers divided by two different SiPM circuits

to create the self-coincidence system.

Figure 4.9: CMuD in self-coincidence. Division of the WLS fibers.

The top circuit corresponds to the one used in the previous measurements and it received 4 of the

fibers. the circuit was replicated and placed on the bottom receiving the other 5 fibers. It is important

53

that the fibers are consecutively divided to the bottom and top circuit because the fibers that collect more

photons are the ones closer to the place where the muon crossed the scintillator. This division ensures

that each those two fibers are always going to a different SiPM’s. This helps to make the resulting two

signals identical, which is essential if the intent is to create a coincidence.

The coincidence of the SiPM’s is done in the FPGA with a logical AND with a time window of 40 ns.

The signal was measured for different thresholds. Two different acquisitions were done: one where the

fibers were divided by the SiPM’s as in figure 4.9 and another where the SiPM’s were not connected

to the detectors at all. The second measurement will give the dark rate of this setup. The difference

between the two measurements should provide an insight about the signal the CMuD in self-coincidence

generates.

Figure 4.10: Signal of the CMuD in self-coincidence and respective dark current rate.

In figure 4.10 is represented the signal resulting from the self-coincidence of the CMuD and from

the coincidence dark rate. When comparing the data with the dark current rate measurement from the

SiPM, shown in figure 3.14, it can be seen a decrease in about two orders of magnitude in the count

rate. This is because, in this case, for a count to occur both SiPM’s must generate a signal within the

coincidence time window. And if this process is random, like in the dark current, the new rate is given by

equation 4.3. By substituting for the dark current at 5 mV we obtain:

RccDC = R2DC δ = (5.5× 105)2 40× 10−9 ≈ 3.5× 103 Hz (4.14)

which seems to correspond to the rate obtained. Even though the dark current rate is greatly reduced,

the self-coincidence signal seems to never detach from it. This means that is likely that the signal

produced by a muon is now too small to be separated from the noise. If that was not the case, when the

noise rate goes to low frequencies there should be a plateau in the self coincidence signal. Which is not

observed. In any case the efficiency and purity were measured.

54

Efficiency

Taking into account the expected decrease in the signal, the thresholds used were significantly reduced.

The center of the reference scintillators was, as best as possible, placed in the space between two

fibers in an attempt that the muons seen by the scintillators generate a signal in both SiPM’s. The data

collected is represented in table 4.6.

Threshold (ph.e) Counters nDnµ

Efficiency (%)

4 204± 1414.8± 1.0

1380± 37

5 321± 1814.2± 0.7

2257± 48

6 177± 138.6± 0.6

2050± 45

7 824± 294.9± 0.2

16970± 130

9 54± 71.8± 0.2

2937± 54

11 24± 51.8± 0.5

1342± 36

Table 4.6: Efficiency of a CMuD in self-coincidence.

The dispersion in the number of events counted for each threshold is because it was not used a fixed

acquisition time. The results show a great decrease in the efficiency being only about 15% for 4 ph.e.

threshold and rapidly decreasing to about 2%. This can be explained by the division the signal goes

through with the separation of the fibers by the SiPM’s. The data is plotted in figure 4.11.

Figure 4.11: Plot of the efficiency in ph.e scale.

According to the CMuD measurements, done in section 4.2.1, each muon generates, at least, around

15 ph.e. If these photo-electrons originated evenly from the fibers closer to the path of the muon it is

probable that the self-coincidence would work, generating a signal of about 6-7 ph.e. (∼ 25 mV) in each

SiPM. But it seems likely that the signal generated is asymmetric and thus creating a significantly higher

signal in one SiPM and a lower in the other making it difficult see a coincidence.

55

Purity

Analyzing the self-coincidence signal, in figure 4.10, it can be expected for the purity to be very low since

the signal basically coincides with the dark current in all spectrum. In any case the threshold was set

to 8 ph.e. (∼ 27 mV) and increased from there. The data acquired is shown in table 4.7 and plotted in

figure 4.12. A few data point shows a large error because rate of events was so small that a very long

acquisition time was necessary to make a representative measurement.

Threshold (ph.e) Counters nµnT

Purity (%)

8 60± 80.5± 0.1

13 6250± 369

10 35± 60.7± 0.1

54 125± 233

11 38± 62.0± 0.4

20 113± 142

13 36± 63.7± 0.7

10 615± 103

14 10± 33.7± 1.2

2950± 54

15 156± 125.1± 0.5

33 298± 54

17 12± 89.4± 2.8

1398± 37

Table 4.7: Purity of a CMuD in self-coincidence.

The measurement shows a very low purity even for relatively high thresholds where the dark current

ratio is close to zero. The data confirms that the signal a muon generates in a CMuD in self-coincidence

is too small to be decoupled from the noise. An increase in the number of fibers, e.g. 9+9, could increase

the signal by increasing the collection efficiency of the CMuD hence improving the self-coincidence

efficiency and purity.

Figure 4.12: Plot of the efficiency in ph.e scale.

56

4.2.3 Two CMuD’s in Coincidence

In this section two CMuD’s were setup in coincidence. The coincidence of the two signals was performed

with a logical AND gate in the FPGA. With this setup it is possible to reduce the dark rate, similarly to

the reduction in self-coincidence, while maintaining the signal generated by a muon in the CMuD. The

SiPM’s were biased with 67.25 V and the coincidence signal was measured with the detectors vertically

aligned and side by side. The difference between these measurements should correspond to the muon

rate. The result is plotted in figure 4.13.

Figure 4.13: Measurement of two CMuD’s in coincidence aligned and side by side.

As was expected, the order of magnitude of the coincidence rate resembles the self-coincidence

signal plotted in figure 4.10. However, unlike the self-coincidence measurement, there is a clear sepa-

ration between the signal and the dark current at around 7 ph.e. (∼ 23 mV). For higher thresholds the

signal seems to maintain a stable value, of about , Rµ = 4 Hz, while the dark current keeps decreasing.

This plateau should correspond to the muon flux, Rµ, crossing the CMuD. To verify this assertion, the

expected muon flux was calculated and compared with our measurement.

The muon flux at sea level can be approximated as 1 muon per square centimeter, per minute, this

approximation does not take into consideration the incidence angle of the muons[39]. Multiplying by the

area of the CMuD, As = 400 cm2, and converting to hertz, we obtain:

Φµ ≈ 6.7 Hz. (4.15)

Before comparing the rates it is necessary to correct the measured signal rate,, Rµ, with efficiency

of the CMuD’s in coincidence. The efficiency was determined in the following section.

Efficiency

For the efficiency measurement, the two CMUD’s were vertically aligned and placed between the two

reference scintillators. The hodoscope signal was coincided with the signal of the two CMuD’s to produce

57

a count. The data collected is represented in table 4.8 and plotted in figure 4.14.

Threshold (ph.e) Counters nDnµ

Efficiency (%)

10 3 272± 5779.6± 0.6

4 112± 64

11 3 927± 6380.9± 0.6

4 854± 70

13 74 274± 27379.7± 0.1

93 178± 305

15 17 166± 13180.7± 0.3

21 265± 146

17 3 372± 5874.3± 0.3

4 538± 67

19 768± 2868.3± 1.4

1 126± 34

21 748± 2760.1± 1.4

1 244± 35

22 688± 2656.0± 1.4

1 228± 35

Table 4.8: Efficiency of two CMuD’s in coincidence.

The efficiency of two CMuD’s in coincidence seems to stay constant for low thresholds, maintaining

a value of about 80%. Around 15 ph.e the efficiency starts to steadily decrease. This behavior is similar

to the one observed in the measurement of the CMuD efficiency, shown in figure 4.7. This seems to

confirm that the minimum signal a muon generates in a CMuD is around 15 photo-electrons.

Figure 4.14: Plot of the detection efficiency of two CMuD’s in coincidence in ph.e scale.

With the efficiency determined, it was possible to correct the muon flux, , R − µ, shown in figure

4.13 to obtain the average rate that a muon crosses the CMuD. Since the signal measurements were

performed for thresholds lower that 15 ph.e (∼ 49.5 mV), the efficiency was considered to be 80%. This

value was used to correct the muon flux, to obtain:

Rµ = 5.25 Hz (4.16)

The muon flux measured renders a similar value to the estimation,Φµ, shown in equation 4.15. There-

58

fore, it seems plausible that the plateau seen in the higher thresholds of figure 4.13 corresponds to the

muon rate.

Purity

In this section, the purity of two CMuD’s in coincidence was measured. The first threshold was set to

10 ph.e (∼ 33 mV) where, according to the signal measured in figure 4.13, appears to be a significant

separation between signal and noise. The threshold was incremented in steps of 1 or 2 ph.e. The data

acquired is shown in table 4.9 and plotted in figure 4.15.

Threshold (ph.e) Counters nµnT

Purity (%)

10 1 545± 3952.2± 1.3

32 519± 180

11 623± 2568.4± 2.7

10 012± 100

13 2 385± 4992.8± 1.9

28 242± 168

14 2 912± 5496.7± 1.8

33 086± 182

15 789± 2898.0± 1.9

8 847± 94

17 761± 2897.8± 2.1

8 555± 92

19 563± 2498.2± 2.1

6 300± 79

21 10 460± 10298.3± 1.1

116 935± 342

Table 4.9: Purity of two CMuD’s in coincidence.

Figure 4.15: Plot of the purity in ph.e scale.

The purity measurements shows a great increase from the 10 ph.e. to the 13 ph.e threshold, going

from around 50% to about 90% purity. This threshold range, corresponding to about 33-43 mV, seems

to match the transition of the dark rate from a 1-2 hertz to about 0.2 Hz in the signal measurement in

figure 4.13. In this transition the dark rate goes from having the same order of magnitude of the muon

59

flux, shown in equation 4.16, to being one order of magnitude lower. This explains the increase in purity

observed. For thresholds higher than 13 ph.e, the purity remains stable value of about 98%.

The data shows that two CMuD’s in coincidence can produce a detector with a very high purity and

efficiency. Comparing the plots of the efficiency, in figure 4.14 and purity, in figure 4.15, it is possible to

determine that the threshold that maximizes the combination of these characteristics is between 14-15

ph.e.

60

Chapter 5

Conclusions and Outlook

The purpose of this work was the study of a Compact Muon Detector (CMuD). This detector was de-

signed to be cheap, robust and safe to handle making it accessible for high school level students. The

CMuD is composed by a scintillator traversed by WLS fibers converging to a silicon photomultiplier.

With the design and material of the CMuD selected it was conducted an estimation of the signal a

muon would generate when crossing the detector. Considering the energy deposition of a muon, light

yield of the scintillator, collection and trapping efficiency of the fibers and photon detection efficiency of

the SiPM, it was estimated that a passing muon generates a 18 ph.e. signal.

The study of the silicon photomultiplier yielded the relationship of its gain with the overvoltage used

to bias the device. The gain increases linearly by 5.28± 0.15 × 105 electrons, for every volt added. From

the measurement of the SiPM, dark current rate for different thresholds, it was possible to determine the

value of 1 ph.e for the setup that was used throughout this thesis. The value of 1 ph.e. was calculated

to be, 3.3± 0.1 mV, this scaling factor was used to convert threshold to a scale that is independent of the

experimental setup used and temperature. The error of this measurement could be reduced by measur-

ing the dark current with a finer threshold increment to attain a better resolution in the transition zones.

As expected, the dark current was shown to decreases exponentially following a Poisson distribution,

which was confirmed by fitting the data to an exponential curve. The result was used to estimate the

dark current rate of the CMuD at higher thresholds.

The characteristics of the CMuD were studied in three different setups. In the first setup, the intrinsic

characteristics of the CMuD were evaluated. The results show that this detector has an efficiency of

about 100% for thresholds lower than 14 ph.e. From there the efficiency decreases slowly, having a

more pronounced drop at around 20 ph.e. The decrease in efficiency can be associated with the signal

produced by a passing muon. If at the threshold of 14 ph.e basically all the particles are detected, this

means that every muon produces a signal of, at least, these many photo-electrons in the CMuD. The

signal measured seems to be in agreement with the estimated signal. On the other hand, the purity

measurement revealed to be relatively low, being about 5% and only starting to increase at 25 ph.e.

This can be attributed to the high dark current rate that still is present at these thresholds. To reduce

the dark current rate and achieve a better efficiency/purity combination, two coincidence setups were

61

implemented and tested. In the first coincidence setup a CMuD was placed in self-coincidence. To

realize this setup, the fibers were divided by two SiPM’s and their signal was coincided. However, the

efficiency and purity measurement show very poor results. This is probably due to the asymmetry in the

division of the signal to each SiPM, making it very difficult to see a coincidence. In the final setup, two

CMuD’s were placed in coincidence to create a telescope. With this setup we expected to maintain the

signal unchanged while reducing the dark current rate. This was confirmed by the results of the efficiency

and purity measurements. The efficiency seems to maintain a stable value of 80% for thresholds up to

15 ph.e and from there it starts decreasing steadily. This behavior is similar to the efficiency of a single

CMuD, which seems to confirm that the minimum signal produced by a muon is around 14 to 15 ph.e.

With the coincidence setup it was possible to obtain a very high purity signal. The measurements show

the purity reaching almost 100% for thresholds above 13 ph.e. The conclusion can be drawn that the

last setup provides the best efficiency/purity compromise with its maximization happening when their

plateaus coincidence - at 14-15 ph.e threshold.

This concluded the study of the Compact Muon Detector performed in this thesis. The prototype

seems to meet the requirements for this project and shows overall good characteristics, specially when

used in coincidence. In the future a DAQ card will be integrated to bias the detector and perform

the amplification, discrimination and coincidence of the signal, replacing the experimental apparatus

necessary this study. This will simplify the use of the detector and allow its distribution by local high

schools for testing.

62

Bibliography

[1] E. J. Weinberg. Particles and fields. Physical Review D54, 1996.

[2] A. De Angelis. Atmospheric ionization and cosmic rays: studies and measurements before 1912.

Astroparticle Physics, 53:19–26, 2014.

[3] M. Bertolotti. The discovery: Victor f. hess and the balloon ascents. In Celestial Messengers, pages

33–44. Springer, 2013.

[4] R. A. Millikan and G. H. Cameron. The origin of the cosmic rays. Physical Review, 32(4):533, 1928.

[5] L. Alvarez and A. H. Compton. A positively charged component of cosmic rays. Physical Review,

43(10):835, 1933.

[6] T. H. Johnson. Progress of the directional survey of cosmic-ray intensities and its application to the

analysis of the primary cosmic radiation. Physical Review, 48(4):287, 1935.

[7] B. Rossi. Directional measurements on the cosmic rays near the geomagnetic equator. Physical

Review, 45(3):212, 1934.

[8] C. D. Anderson. The positive electron. Physical Review, 43(6):491, 1933.

[9] S. H. Neddermeyer and C. D. Anderson. Cosmic-ray particles of intermediate mass. Physical

Review, 54(1):88, 1938.

[10] P. Auger, P. Ehrenfest, R. Maze, J. Daudin, and R. A. Freon. Extensive cosmic-ray showers. Re-

views of Modern Physics, 11(3-4):288, 1939.

[11] E. Fermi. On the origin of the cosmic radiation. Physical Review, 75(8):1169, 1949.

[12] J. Linsley. Evidence for a primary cosmic-ray particle with energy 10 20 ev. Physical Review Letters,

10(4):146, 1963.

[13] H. Bergeson, G. Cassiday, T.-W. Chiu, D. Cooper, J. Elbert, E. Loh, D. Steck, W. West, J. Linsley,

and G. Mason. Measurement of light emission from remote cosmic-ray air showers. Physical

Review Letters, 39(13):847, 1977.

[14] R. Baltrusaitis, R. Cady, G. Cassiday, R. Cooperv, J. Elbert, P. Gerhardy, S. Ko, E. Loh, M. Salamon,

D. Steck, et al. The utah fly’s eye detector. Nuclear Instruments and Methods in Physics Research

63

Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 240(2):410–428,

1985.

[15] P. J. S. Assis. Data acquisition and performance studies in cosmic ray experiments. volume 1,

pages 1–180. Instituto Superior Tecnico, Jan. 2011.

[16] A. Mapelli. Scintillation particle detectors based on plastic optical fibres and microfluidics. volume 1,

pages 1–193. ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE, Jan. 2009.

[17] K. Nakamura and P. D. Group. Review of particle physics. Journal of Physics G: Nuclear and Particle

Physics, 37(7A):075021, 2010. URL http://stacks.iop.org/0954-3899/37/i=7A/a=075021.

[18] J. BIRKS. Chapter 9 - organic plastic scintillators. In J. BIRKS, editor, The Theory and Prac-

tice of Scintillation Counting, International Series of Monographs in Electronics and Instrumen-

tation, pages 321 – 353. Pergamon, 1964. ISBN 978-0-08-010472-0. doi: http://dx.doi.org/10.

1016/B978-0-08-010472-0.50014-8. URL http://www.sciencedirect.com/science/article/

pii/B9780080104720500148.

[19] Y. Koba, H. Iwamoto, K. Kiyohara, T. Nagasaki, G. Wakabayashi, Y. Uozumi, and N. Matsufuji.

Scintillation efficiency of inorganic scintillators for intermediate-energy charged particles. Prog.

Nucl. Sci. Technol., 1, 2011.

[20] J. R. Winckler. The theory and practice of scintillation counting. j. b. birks. pergamon. lon-

don; macmillan, new york, 1964. xx + 662 pp. illus. 17.50. Science, 148(3667):217–217, 1965.

ISSN 0036-8075. doi: 10.1126/science.148.3667.217. URL http://science.sciencemag.org/

content/148/3667/217.1.

[21] G. Knoll. Radiation detection and measurement. John Wiley & Sons, 3rd edition, 1999.

[22] A. S. Huntington, M. A. Compton, and G. M. Williams. Linear-mode single-photon apd detectors. In

Optics East 2007, pages 67710Q–67710Q. International Society for Optics and Photonics, 2007.

[23] D. Renker and E. Lorenz. Advances in solid state photon detectors. Journal of Instrumentation, 4

(04):P04004, 2009. URL http://stacks.iop.org/1748-0221/4/i=04/a=P04004.

[24] sensL. An introduction to the silicon photomultiplier. Technical Note, 1(1):1–8, Oct. 2011.

[25] R. Franke, M. Holler, B. Kaminsky, T. Karg, H. Prokoph, A. Schonwald, C. Schwerdt, A. Stoßl,

and M. Walter. Cosmo-a cosmic muon observer experiment for students. arXiv preprint

arXiv:1309.3391, 2013.

[26] F. Barao. Design and construction of a portable cosmic ray muon detector. 2014. URL http://

fcomp.ist.utl.pt/LRC.web/actividades/actividades.2013-14.portable-telescope.html.

[27] E. Technologies. Ej-200 datasheet. Technical report, Eljen Technologies, .

[28] Kuraray. Kurarays scintillation materials datasheet. Technical report, Kuraray.

64

[29] W. Bugg, Y. Efremenko, and S. Vasilyev. Large plastic scintillator panels with wls fiber readout:

Optimization of components. Nuclear Instruments and Methods in Physics Research Section A:

Accelerators, Spectrometers, Detectors and Associated Equipment, 758:91–96, 2014.

[30] E. Technologies. Eljen scintillators products commercial equivalence. Technical report, . URL

http://www.eljentechnology.com/index.php/products/plastic-scintillators.

[31] T. K. energy MeV and I. Mean excitation energy eV. 30. passage of particles through matter.

[32] E. Garutti. The physics particle detectors. 2012.

[33] P. Mohanty, S. Dugad, and S. Gupta. Monte carlo code g3sim for simulation of plastic scintillator

detectors with wavelength shifter fiber readout. Review of Scientific Instruments, 83(4):043301,

2012.

[34] P. N. Dinh, N. T. Dung, B. D. Hieu, P. T. Phuong, P. Darriulat, N. T. Thao, D. Q. Thieu, and

V. Van Thuan. Measurement of the zenith angle distribution of the cosmic muon flux in hanoi.

Nuclear Physics B, 661(1):3–16, 2003.

[35] C. Y. E. Ho. Cosmic ray muon detection using nai detectors and plastic scintillators. 2010.

[36] R. O. Birger Vaksdal. Gent4 simulation of a portable muon detector. 2014.

[37] AdvanSiD. introduction to sipms. Technical report. URL http://advansid.com/attachment/get/

up891411030571.pdf.

[38] T. Ullrich and Z. Xu. Treatment of errors in efficiency calculations. arXiv preprint physics/0701199,

139, 2007.

[39] H. Portella. Muon spectra at sea level. In International Cosmic Ray Conference, volume 2, page

143, 1999.

65

66