Detectors for the ATLAS High-Granularity Timing Detector

(1)

IN

DEGREE PROJECT ENGINEERING PHYSICS, SECOND CYCLE, 30 CREDITS

STOCKHOLM SWEDEN 2019,

Auto-triggering studies of Low Gain Avalanche Detectors for the ATLAS High-Granularity Timing Detector

FREDRIK SJÖSTRÖM

KTH ROYAL INSTITUTE OF TECHNOLOGY SCHOOL OF ENGINEERING SCIENCES

(2)

(3)

Auto-triggering studies of Low Gain Avalanche

Detectors for the ATLAS High-Granularity Timing Detector

FREDRIK SJÖSTRÖM

Master in Engineering Physics Date: June 18, 2019

Supervisor: Jonas Strandberg Examiner: Mark Pearce

School of Engineering Sciences

Swedish title: Auto-triggering beteende hos Low Gain Avalanche Detectors för ATLAS High-Granularity Timing Detector

(4)

(5)

iii

Abstract

To enhance the reach of its physics program, the Large Hadron Collider (LHC) at CERN is being upgraded. The upgrade, known as High Luminosity LHC (HL-LHC) aims to drastically increase the collision rate which also raises the number of pile-up interactions. Pile-up interactions are the proton-proton collisions occuring in the same bunch crossing as the interaction of interest.To be able to fully take advantage of the upgrade, ATLAS, one of the particle detectors at the LHC, needs to be upgraded as well. One part of this upgrade is the High-Granularity Timing Detector (HGTD). It is a new type of detector which introduces a way to separate collisions occurring close in space but distinct in time. The new detector will be able to reduce the contamination from pile-up interactions to the desired level. The HGTD is based on Low Gain Avalanche Detector (LGAD) technology.

This thesis investigates auto-triggering behavior in the sensors planned to be used in the HGTD. Auto-triggering is the problem of noise reaching high enough levels to be falsely registered as a signal by a sensor. The study investigates the auto-triggering ability as a function of bias voltage of 24 different LGADs. The study concludes that 3 of the 24 tested LGADs indicate severe auto-triggering problems for the highest applied bias voltage for each sensor.

(6)

iv

Sammanfattning

För att öka möjligheterna till upptäckter hos fysikprogrammet vid Large Hadron Collider (LHC) vid CERN håller acceleratorn på att uppgraderas. Uppgrade- ringen går under namnet High Luminosity LHC (HL-LHC) och ämnar att dras- tiskt öka antalet kollisioner i acceleratorn. Detta kommer även att öka mäng- den pile-ups", det vill säga oönskad produktion av partiklar som kan komma att kontaminera intressant data. För att fullständigt kunna ta tillvara på uppgra- deringen behöver även ATLAS, en detektor kopplad till LHC, förbättras. En del av denna uppgradering är high granularity timing-detektorn (HGTD). Det- ta är en ny typ av detektor som möjliggör distinktion mellan kollisioner som ligger nära i rummet men som skett vid olika tidpunkter. Syftet med detektorn är att reducera den negativapåverkan från pile-ups till en godkänd nivå. HGTD är baserad på Low gain Avalanche detektor-teknik (LGADs).

Detta arbete undersöker autotriggande beteende hos sensorer som planeras att användas i HGTD. Autotriggande är ett problem som uppstår när elektroniskt brus når så pass höga nivåer att en sensor förväxlar det med en riktig signal.

Denna studie undersöker autotriggande hos 25 olika LGADs. Av de 25 senso- rerna var det endast möjligt att slutföra en analys på 24 stycken. Studien fastslår att 3 av de 24 testade LGADs uppvisade allvarliga problem med autotriggande vid den högsta testade spänningen för respektive sensor.

(7)

List of Figures

2.1 Components of the Standard Model grouped into matter (fermions) and carriers of force (bosons) [4]. . . 4 2.2 Feynman diagram of the β⁻decay mode. Time flows vertically. 7 2.3 Force between two nucleons as a function of the distance be-

tween them, computed from the Reid potential [6]. . . 7 2.4 Core process of electromagnetical interactions. . . 8 3.1 Overview of the Large Hadron Collider and its pre-accelerators.

The only detector shown is the ATLAS experiments. . . 9 3.2 Integrated luminosity in the LHC for different years [8]. . . 11 3.3 Graphical representation of jets originating from hard scatter

interactions and pile-up interactions [9]. . . 12 4.1 Schematic overview of the ATLAS detector [10]. . . 13 4.2 Schematic overview of the Inner Detector of the ATLAS ex-

periment [12]. . . 14 4.3 A small sector of the accordion structure of the barrel calorime-

ter in a plane transverse to the beam-axis [17]. . . 17 5.1 Plan for the HL-LHC upgrade of the LHC. Here LS stands

for Long Shutdown and EYETS stands for Extended Year-End Technical stop [23]. . . 20 5.2 Difference in detector activity between LHC and HL-LHC [25]. 21 5.3 Positioning of the HGTD inside the ATLAS detector [26]. . . 23 5.4 Visualisation of z-t plane of an event with hard scatter (red)

and approximately 200 pile-up interactions (blue). The reconstructed vertices are represented by the dashed vertical lines [27]. 25 5.5 Example of an event with noise and a pulse. . . 25 5.6 Schematic drawing of the overlap between modules on the

cooling plate. The horizontal line represent r = 320 mm [26]. 27

vii

(10)

viii LIST OF FIGURES

5.7 Number of hits in the HGTD as a function of the position. The overlap at R < 320 mm is 80%, while the number for R >

320 mm is 20% [27]. . . 28 5.8 Cross Section of an LGAD sensor [29]. . . 28 5.9 Occupancy vs. radius for various pixel sizes at a pile-up of

< µ > = 200 [27]. . . 29 5.10 Orientation of readout rows for the first layer, last layer, and

the overlay of the two. The layers are rotated 15^◦ in opposite directions [27]. . . 29 5.11 Nominal neutron-equivalent fluence and ionising dose as func-

tions of the radius in the outermost sensor layer of the HGTD for 4000 fb⁻¹, i.e. before including safety factors. The contribution from charged hadrons is included in "Others" [27]. . . . 30 5.12 Slew rate based on the thickness of a sensor and the gain [27]. 31 5.13 A comparison between constant threshold triggering (left) and

constant fraction triggering (right) of two pulses with the same max time, but different amplitudes [30]. . . 33 6.1 Overview of the data acquisition setup [31]. Up to two oscillo-

scopes can be used to test more than three DUTs simultaneously. 35 6.2 Mimosa26 board inside an aluminium housing [33]. . . 37 7.1 Representative time distribution maximum peak times between

an LGAD sensor and the SiPM plate. . . 42 7.2 Tracking data with hits inside and outside the area sensor. . . . 43 7.3 Auto-triggering results from April 2018 (a) and a close-up on

the result (b). . . 45 7.4 Auto-triggering results from June 2018 test beam. . . 46 7.5 Auto-triggering results from October 2018 (a) and a close-up

on the result (b). . . 47 7.6 Placement plots for the problematic sensor batch 505 (a) and

605 (b). As shown, no clear placement of the sensor can be made in neither of the two batches. The rectangles that can be visualized on both plots are the regions of interest. . . 48 7.7 ∆t for sensor W4S207 for the two tested batches. Comparing

to how it should look, shown in Figure 7.1, there is a clear indication of a problem. . . 49

(11)

List of Tables

5.1 Main parameters of the HGTD [27]. . . 24

5.2 Schematic overview of sources of auto-triggering for irradiated and unirradiated sensors. . . 26

6.1 Sensors tested in the April 2018 test beam session. . . 36

6.2 Sensors tested in the June 2018 test beam session. . . 36

6.3 Sensors tested in the October 2018 test beam session. . . 37

6.4 Names and short descriptions of the variables in the processed ROOT files in the ntuple data. . . 39

7.1 A table with all the tested sensors and the maximum ratio of autotriggers, along with its error, voltage, and fluence. . . 44

A.1 Results of the auto-triggering analysis using April 2018 test beam data. . . 58

A.2 Results of the auto-triggering analysis using October 2018 test beam data. . . 59

A.3 Results of the auto-triggering analysis using June 2018 test beam data. . . 59

ix

(12)

Chapter 1 Introduction

One way to achieve new knowledge of the structures and mechanics govern- ing the world is through conducting experiments. Throughout the history of natural science, experiments have been used to gain insight on how everything from the largest stars to the tiniest quarks behave. As time progresses and more mysteries are being unraveled higher demands are put on the machines used in the tests.

The Large Hadron Collider (LHC) is an accelerator at CERN (The European Organization for Nuclear Research) accelerating beams of particles to nearly the speed of light [1]. The accelerated beams are set to collide at four specific points in the accelerator. When the two beams have collided, the exchange of energy makes it possible to create new particles. These byproducts from the collisions are tracked by several detectors connected to the collider. One of these experiments is the ATLAS experiment. Since the end of 2009, the experiment has been gathering data from collisions. It was one of the two detectors involved in finding experimental evidence of the Higgs boson in 2012 [2]. The High-Luminosity Large Hadron Collider (HL-LHC) is a new phase of the LHC planned to begin its operations in 2026, with aim to greatly improve the number of colliding particles in each passing particle bunch crossing [3].

In addition to the new phase, the ATLAS experiment is being upgraded as well.

The High-Granularity Testing Detectors (HGTD) is a new device proposed for the the upgrade with the purpose of distinguishing between collisions taking place close in space but separated in time. However, before the installment, the components of the detector need to be thoroughly tested to verify them ful-

1

(13)

2 CHAPTER 1. INTRODUCTION

filling the requirements regarding performance, reliability, and construction.

Without proper testing it is possible that parts of the upgrade do not work as effectively as desired, causing problems in future data gatherings.

In this thesis, data from three HGTD test beam runs in 2018 is used to evaluate different sensors probability to register and pass on false triggers. It is important to be able to safely assume that the data recorded represent actual activity and not a false positive reaction the sensors produce themselves.

1.1 Author’s Contribution

The author’s code is based upon studies and software written by Antek Szadaj as his master’s thesis at KTH. The software used for the analysis performed in this thesis has been written by the author unless otherwise stated. Additionally, the author has participated in weekly HGTD test beam meetings discussing his and the other members’ test beam studies. Finally, the author has assisted the test beam group by taking shifts during the two test beam data acquisition sessions at the Prévessin site in September and October 2018.

(14)

Chapter 2 Standard Model

The Standard Model is the theory in particle physics classifying the known elementary particles in the universe. The theory also describes the interactions of the particles through three of the four fundamental forces. These forces are the electromagnetic, weak, and strong interactions. The fourth force, the grav- itational force, is not included in the model. Twelve of the elementary particles have spin ¹₂ and are therefore classified as fermions, while five of them have integer spin and consequently classifies as bosons. A schematic overview of the fermions and the bosons are presented in Figure 2.1. Even though the first usage of the term Standard Model dates to 1975, the confirmation has taken time. For example, the tau neutrino was confirmed in 2000 and the Higgs boson as recent as 2012.

2.1 Fermions

The fermions of the model can be seen as the building blocks of all matter around us and are categorized into two different groups with six particles each.

These two groups are the quarks and the leptons. Each fermion also belongs to a generation. In total, there are three different generations where each generation contains two quarks and two leptons. The generations are ordered in such a way so a particle of a higher generation has a greater mass than the corresponding particle of the previous generation. A heavy particle is less stable than a light particle and will eventually decay into a lighter, and more stable particle. This process continues until the particle decays to a particle of the first generation. The stable universe is made up of particles belonging to the first generation, with the possible exception of the neutrinos.

3

(15)

4 CHAPTER 2. STANDARD MODEL

Figure 2.1: Components of the Standard Model grouped into matter (fermions) and carriers of force (bosons) [4].

Studying Figure 2.1 it is possible to see that the three generations of quarks and leptons vary a lot in masses, ranging from the electron neutrino which is lighter than 2.2 eV/c² to the top quark with its 173.1 GeV/c². This mass difference between the quarks and lepton generations is one of the unanswered questions along with what dark matter is and what happened to the antimatter after the big bang, indicating that the Standard Model is incomplete and need further studying.

2.2 Quarks

One of the two groups of fermions is the group of quarks. There are six flavours of quarks (q), named up (u), down (d), charm (c), strange (s), top (t), and bottom (b). The up and down quark make up the first generation, the charm and strange quark the second generation, and the top and bottom quark the third generation of quarks. Each generation contain one particle with an electric charge of −¹₃e and one particle with the electric charge +²₃e. Along with the quarks, there exist antiquarks which have the same properties as their corresponding quarks, except for any charge which is opposite. The antiquarks are separated from the quarks by using a bar over its quark version (¯q).

(16)

CHAPTER 2. STANDARD MODEL 5

The quarks have an intrinsic property called color. Color exists in three different charges; red, green, and blue, and their corresponding anticolors; antired, antigreen, and antiblue. This attribute has nothing to do with what is tradi- tionally referred to as color and is more similar to the familiar electric charge.

Multiple quarks can be combined to create more complex particles. The particles created by quarks are named hadrons. In order to do this, the resulting particle must be colorless. A hadron is deemed to be colorless if it is made of either a color-anticolor quark pair or three quarks with different colors. A particle constituted by a color-anticolor pair is known as a meson. Since the quarks have a spin of ¹₂ the spin of a meson is either 0 or 1 and are therefore classified as a boson. The lightest meson is the pion which can be either electrically charged or neutral. The positive pion π⁺ is made up of ud, while the¯ negative is its antiparticle ¯ud. The neutral pion π⁰is constructed of^u¯^{u−d ¯}^√₂^d, and is its own antiparticle. All mesons are unstable and decay. The most stable meson has a mean lifetime of a few hundredths of a microsecond.

If the quarks are combined into a group of three quarks instead of two, the resulting particle is known as a baryon. Since the baryons are made up of an odd number of quarks, their spin is half-integer and is therefore classified as fermions. As with the charged mesons, every baryon has a corresponding antiparticle, known as antibaryon, which is created replacing the quarks with antiquarks and vice versa. In addition to charge and spin, each baryon has an intrinsic property known as baryon number. The baryon number is conserved during an interaction or decay. The proton (with its quark content uud), is the lightest particle with non-zero baryon number and therefore can not decay into something without breaking the conservation of the baryon number. Because of this, the proton is stable. The neutron (udd) is another example of a baryon.

The neutron, together with the proton, makes up the atomic nucleus.

2.3 Leptons

The second group of fundamental fermions is leptons. The six leptons are the electron (e), muon (µ), and tau (τ ) particles, and their corresponding neutrinos: electron neutrino (νe), muon neutrino (νµ), and tau neutrino (ντ). Every lepton and neutrino has an antiparticle with the same mass as its counter particle, but with opposite charges. In contrast to the quarks, the leptons are colorless. The electron and its corresponding neutrino make up the first generation of the leptons, muon and its neutrino the second, and tau and the tau neutrino

(17)

the third generation. While the electron, muon, and tau particles all have an electric charge of -e and distinct rest masses, the neutrinos are electrically neutral and much lighter than the charged leptons. For a long time the neutrinos were believed to be massless, but it has been disproven [5]. The masses of the neutrinos are not unique to a flavor. The mass of each flavor is compiled of a quantum superposition of three distinct mass states. Due to this, the neutrinos demonstrate a property known as neutrino oscillation - a neutrino measured with a specific flavor can later be measured having a different flavor. As with the baryons, each lepton family is given a lepton number which must be conserved generation-vise during interactions. If a lepton of a certain generation is created or destroyed in an interaction, the antilepton of the same generation must also be created or destroyed.

2.4 Forces

Apart from the matter particles, the Standard Model also describes three of the four fundamental forces namely the weak force, the strong force, and the electromagnetic force. The fourth force, gravitation, is not explained. Each of the forces explained by the Standard Model is being mediated with different particles. Every force carrier has spin 1 and is categorized as bosons accordingly.

The carriers are presented in Figure 2.1. The weak force is being mediated by the W and Z bosons and ranges approximately 10⁻¹⁸m. It acts on all particles except gluons. The W boson comes with either a negative (W⁻) or a positive (W⁺) electric charge of ±e and is each other’s antiparticles, while the Z boson is neutral and its own antiparticle. The weak interaction is in charge of chang- ing flavors of the quarks. This change of flavor is mediated by the W bosons and enables processes such as the beta decays. In beta minus decay, a down quark within a neutron converts to an up quark by releasing a W⁻boson. This alternates the neutron to a proton and causes the nucleus to emit an electron and an electron antineutrino. A representation of this procedure is given in Figure 2.2.

(18)

CHAPTER 2. STANDARD MODEL 7

Figure 2.2: Feynman diagram of the β⁻decay mode. Time flows vertically.

The strong force is the force responsible for binding the quarks together in the mesons and baryons. It is also responsible for binding the protons and neutron together in the atomic nucleus. Without this force, the protons inside a nucleus would not stay confined due to the massive repulsion caused by the same charged protons in a close space. The force acts repulsive for particles closer than a distance of 0.8 fm and is as most attractive at a distance of approximately 1 fm. The force falls off rapidly for particles further away. The behavior of the force is presented in Figure 2.3.

Figure 2.3: Force between two nucleons as a function of the distance between them, computed from the Reid potential [6].

(19)

The last force to be presented is the electromagnetic force. In a similar way as gravity the force is infinite. Even though the electromagnetic force is much stronger than gravity large objects are usually unaffected of the force simply because the objects usually have their charges canceling each other out resulting in charge neutrality. The force carrier for the electromagnetic force is the photon, a massless boson with spin 1. This force is the cause of both electrical and magnetical phenomena. Atoms and molecules are mainly held together by the electromagnetic force. Essentially, all electromagnetic interactions boil down to the following interaction, as shown in Figure 2.4.

Figure 2.4: Core process of electromagnetical interactions.

(20)

Chapter 3 Large Hadron Collider

The Large Hadron Collider (LHC) is a circular particle collider built beneath the France-Switzerland border, near Geneva. With its circumference of 27- kilometre and a total collision energy of 13 TeV, it is the world’s largest and most powerful particle accelerator [7]. Since the machine began operation in 2008, it has been used as a research tool for investigating particle physics. The LHC accelerate two beams of particles to close to the speed of light and col- lides at four crossing points. Around these four crossing points detectors are positioned, each with different designs. It is important to have detectors with independent design for cross-confirmation of new discoveries made [7]. An overview of the complex making up the Large Hadron Collider is presented in Figure 3.1.

Figure 3.1: Overview of the Large Hadron Collider and its pre-accelerators.

The only detector shown is the ATLAS experiments.

9

(21)

10 CHAPTER 3. LARGE HADRON COLLIDER

Even though the LHC sometimes uses beams of heavy ions, beams of protons are primarily used. The protons to be accelerated through the accelerator are created by stripping electrons from hydrogen atoms. These protons are injected into the Booster at an energy of 50 MeV. In this booster, the particles are accelerated to an energy of 1.4 GeV. When the protons reach this energy, the beam is put through the Proton Synchrotron (PS). This portion of the complex accelerates the beam to an energy of 25 GeV. The protons are then sent to the Super Proton Synchrotron (SPS), which accelerates the beam to 450 GeV. Finally, when the beam reaches this energy, it is transferred to the Large Hadron Collider in both clockwise and counterclockwise direction.

Here the beams are accelerated to the nominal energy.

3.1 Luminosity

Luminosity, L, is a concept in particle physics and accelerator physics measuring how many particles are passing through a given space at a given space.

It is proportional to the number of collisions at a specific time and space and can therefore be used as a measurement on how much data the accelerator can produce at a given time. There are different methods of increasing luminosity in a particle collider. One way is to focus the beam more precisely on the collision spots, which is done by refocusing magnets and redesigning optics in the accelerator. Other ways of increasing the value of L is by increasing the number of particle bunches in the beam or by increasing the number of colliding particles in each bunch. These operations lead to some difficulties however, such as increasing the number of pile-ups. The pile-up problem is discussed more thoroughly in the following chapter.

By integrating the luminosity with respect to the time it is possible to quantify the size of the collected data by the detector. The integrated luminosity is usually given in fb⁻¹. A graph on the integrated luminosity at LHC is shown in Figure 3.2.

(22)

CHAPTER 3. LARGE HADRON COLLIDER 11

Figure 3.2: Integrated luminosity in the LHC for different years [8].

3.2 Pile-up

When the proton beams are colliding in the LHC interactions occurs. Typi- cally only one interaction is interesting in a bunch crossing. This interaction is referred to as the hard-scatter interaction. Additional proton-proton interactions happening in the same bunch crossing as the interaction of interest are referred to as pile-up interactions. These pile-up interactions create a problem when trying to measure particles or jets originating from the interaction of interest. The pile-up interactions produce uninteresting jets which can be hard to separate from the hard-scatter jets. In addition, particles originating from the pile-up interactions might end up in a hard-scatter jet and contaminate it.

A third possibility is stray particles from different pile-up interactions might mistakenly be identified as a jet even though the origins of the particles differ.

All of the three problems are presented in Figure 3.3. At low jet transverse momentum, pile-up jets are primarily produced by random combinations of particles originating from multiple vertices.

(23)

12 CHAPTER 3. LARGE HADRON COLLIDER

Figure 3.3: Graphical representation of jets originating from hard scatter interactions and pile-up interactions [9].

(24)

Chapter 4 The ATLAS Detector

The ATLAS (A Toroidal LHC Apparatus) is one of the two general-purpose detectors at the LHC. The experiment is used for a variety of physics, from the- ories including extra dimensions to searching for candidate particles making up dark matter. Data gathered by the detector was also used in the observation of the Higgs boson [2]. A schematic overview of the detector is presented in Figure 4.1.

Figure 4.1: Schematic overview of the ATLAS detector [10].

The coordinate system of ATLAS is defined as the x-axis pointing towards the center of the collider, the y-axis pointing up, and the z-axis pointing in such a way that the system becomes right-handed [11]. Along with this, a polar coordinate system is used for describing the transverse plane. As usual, r denotes the radial coordinate which relates to the traditional Cartesian coordinates as

13

(25)

14 CHAPTER 4. THE ATLAS DETECTOR

r = px²+ y². In practice, this is equal to the distance from the beam line.

The azimuthal angle φ is defined so that the direction given by φ = 0 coincides with the x-axis and therefore points at the center of the LHC tunnel. θ is the angle between the positive direction of the beam axis and the three-momentum p of a particle.

Pseudorapidity is a commonly occurring measurement used in detector physics.

Usually noted η, the pseudorapidity describes the angle of a particle relative to a beam axis. The pseudorapidity is given by

η ≡ − ln

tan θ 2

, (4.0.1)

The ATLAS detector contains four major parts: The inner detector, the calorimeter, the muon spectrometer, and the magnetic system.

4.1 Inner Detector

The Inner Detector (ID) is cylindrical with a length of 6.2 meters and a radius ranging from a few centimeters from the proton beam axis to r = 1.2 meters. The direction, momentum, and charge of every electrically charged particle produced in the proton-proton collisions are tracked in this part of the detector. The inner detector itself is made up of three major components: the pixel detector, the semiconductor tracker and the transition radiation tracker.

A cut-through view of the inner detector is shown in Figure 4.2.

Figure 4.2: Schematic overview of the Inner Detector of the ATLAS experiment [12].

(26)

CHAPTER 4. THE ATLAS DETECTOR 15

4.1.1 Pixel Detector

The pixel detector is the component positioned closest to the interaction point.

Its main task is to provide measurements of the created particles with high precision and granularity. The system is made up of three barrel layers and three disks on each end-cap. The barrel layers are placed on average radii of 5, 9, and 12 cm, while the disks span the radii between 9 and 15 cm away from the beam line. The three barrels have a total of 1456 modules, while the disks have a combined number of 288 modules. Each module has 46 080 read-out channels, serving an array of 18 by 160 pixels making the total number of pixels 80 million. The read-out chips are constructed with a requirement to withstand up to 300 kGy of ionizing radiation and over 5 × 10¹⁴ neutron hits per squared centimeter over the 10 operational years. The resolution of the pixel detector is 12 µm in r-φ and 70 µm in the z directions [13].

4.1.2 Semiconductor Tracker

The SemiConductor Tracker (SCT) is situated outside the pixel detector. The read-out channels in the semiconductor tracker are strip-shaped. It consists of 60 m² of silicon, distributed over 4 cylindrical barrel layers and 18 end-cap disks. 4 088 two-sided modules and over 6 million read-out channels make it possible to get the position to the accuracy of 17 µm per barrel layer.

4.1.3 Transition Radiation Tracker

The final component of the inner detector is the Transition Radiation Tracker (TRT). It is composed of approximately 300 000 drift tubes, each 4 mm in di- ameter. Each tube is made of a multi-layer film working as a cathode, while a gold-plated tungsten wire is positioned in the center of each tube and serves as an anode. The tubes are filled with a gas mixture of Xenon, CO², and O which ionizes when a charged particle passes through a tube. An electric field is created by keeping the wall of the tubes at a high negative voltage. The field accelerates the electrons created in the ionization process towards the wire anode, liberating even more electrons in the gas. These electrons make up a current signal inside the tube. The wires are split and read out at both ends of the tube. This doubles the number of electronic channels and also reduces the occupancy [14]. A low occupancy reduces the risk of the same tube getting hit multiple times simultaneously which might cause problems with the data recording and track reconstruction.

(27)

In the barrel, the tubes are positioned parallel to the beam-line and cover from 560 to 1080 mm in radius, and |z| < 720 mm, while the two endcaps have the tubes positioned perpendicular to the beam-line and cover from 617 to 1106 mm in radius and 827 < |z| < 2774 mm. Due to the larger radii, it is too expensive to cover the areas required with silicon detectors [15]. That is why the final part of the inner detector is using another technique than the two innermost.

4.2 Calorimeter

The purpose of the calorimeters is to measure the energy of a particle. In order to measure this, the particle needs to be fully absorbed. Therefore most calorimeters use a disruptive measurement method when measuring - it termi- nates the movement of the particle it examines. The calorimeter can measure both charged and neutral particles, it can stop neither muons nor neutrinos however. The calorimetry system is made up of two components: The electromagnetic Liquid Argon (LAr) calorimeter and the Tile Hadronic Calorimeter (TileCal). Together the two calorimeters provide coverage of particles with a pseudorapidity of |η| < 4.9.

4.2.1 Liquid Argon Calorimeter

The Liquid Argon (LAr) Calorimeter is made up of an electromagnetic barrel, two end-cap calorimeters and one forward calorimeter on each side of the inner detector. One of the end-caps is electromagnetic and the other one is a hadronic end-cap. The components are all using LAr as the substance the passing particles ionize. The energy of the particle can be inferred from the ions in the substance. The electromagnetic part of the calorimeter covers |η| < 3.2, the hadronic end-cap covers 1.5 < |η| < 3.2, while the forward calorimeter covers 3.1 < |η| < 4.9. In order to make the installation symmetric in φ, the barrel is structured in an accordion-shape [16]. An illustration of this structure is presented in Figure 4.3.

The forward calorimeters (FCal) is composed of three cylindrical modules.

The modules are ordered sequentially, the module closest to the interaction point is made of copper and deals mainly with electromagnetic measurements.

The two subsequent modules are, instead of copper, primarily of tungsten and deals with hadronic measurements.

(28)

CHAPTER 4. THE ATLAS DETECTOR 17

Figure 4.3: A small sector of the accordion structure of the barrel calorimeter in a plane transverse to the beam-axis [17].

4.2.2 Hadronic Tile Calorimeter

The Hadronic Tile Calorimeter (TileCal) surrounds the LAr calorimeter in the barrel region, i.e. |η| < 1.7. It is divided into three parts, one fixed central barrel, and two moveable extended barrels. The barrels are each made up of 64 modules covering an azimuthal angle of φ = 0.1. The TileCal is a sampling calorimeter, meaning it uses alternating layers of passive absorbers and active detectors. The absorbing plates are made of iron while the active detector is scintillating tiles. The scintillating tiles extend in planes perpendicular to the beamlines. Each side of a tile has wave-length shifting fibers which function as read-outs of the tile. These read-outs are sent to photomultiplier tubes. The entire structure contains approximately 5 000 cells and 10 000 tubes [18].

4.3 Muon Spectrometer

Outside the calorimeters is the largest component of the detector, namely the muon spectrometer. The purpose of the spectrometer is to detect and measure the momentum of muons exiting the calorimeters. This is done by using a system of magnets and accompanying trigger and tracking chambers.

4.4 Trigger and Data Aquisition

The ATLAS experiment is designed to be able to observe up to 1.7 billion proton-proton collisions per second, which yields a combined volume of data of more than 60 million megabytes per second [19]. Since not all of the observed events contain characteristics important to the physicists, the detector needs to be able to thin out uninteresting collisions. This is done by the trigger system.

(29)

The trigger system selects interesting events in a three-step process. The first step, the level-1 hardware trigger, works on a subset of information from the muon detectors and the calorimeter. If the data from an event is decided to be kept, it is retrieved from a storage buffer. The decision is made less than two-and-half microseconds after the event occurs. In total, the level-1 trigger can select up to approximately 100 000 events every second. The selected data is being sent to the following two steps of the process, namely the level-2 and event filter. The two steps are together named High-Level Trigger (HLT). The high-level trigger is software based, and analyses the data saved from step 1 more thoroughly. Around 1000 events every second are selected by the HLT.

These events are fully assembled into an event record, which is passed on and saved to a data storage system.

4.5 Jet selection

Pile-up is primarily dealt with by accurately associating the created jets with their reconstructed interaction points. By doing so it is possible to reduce the pile-up problem in the detector. It is important to note the pile-up problem is not dealt with by reducing the number of collisions, but by opting out uninteresting tracks. One important factor to study when solving the pile-up problem is the RpT jet variable. The variable is a measurement of how much transverse momentum in a jet is from the interaction of interest. For a moderate amount of pile-ups, small values of R^pT links to jets that have a small amount of charged particle transverse momentum originating from the interaction of interest. These jets are likely to be pile-up jets. The problems of pile-up jets can be somewhat neutralized by, in addition to the spatial information, also include timing information of interactions.

(30)

Chapter 5 Upgrade

5.1 Physics

The observation of the Higgs Boson in 2012 and the following studies of the boson’s properties are some of the highlights of the physics program at the LHC so far. Decay modes of this new particle to the electroweak bosons of the Standard Model have been observed with more than 5 standard deviation significance [20].

Interesting physics means rare processes. It is possible to continue making discoveries and fine-tuning the already made revelations with the same luminosity and operational parameters as earlier, but it will take a long time. One way of reducing the wait is by improving the luminosity of the collider. An improvement of luminosity leads to more collisions and therefore more data available for analyzing. This increases the significance of an analysis, which goes as

σ ∼ S

√B, (5.1.1)

where S is the number of events from the sought process, and B from the background processes. Both of these variables are linearly dependent on the luminosity. An increase in the luminosity of 4 yields, by inserting in Eq. 5.1.1, a statistical improvement of 2. Therefore it is essential that the luminosity of the LHC is improved.

19

(31)

20 CHAPTER 5. UPGRADE

5.2 High-Luminosity Large Hadron Collider

As time progresses new technology arises and older instruments and machines grow obsolete. The LHC has been in operation since March 2010, producing proton-proton collisions. First with a center-of-mass energy of 7 and 8 TeV and, after a 2 year long hiatus in 2015, with a center-of-mass energy of 13 TeV [21]. For further advances in the physics program at the LHC, an upgrade is needed. The upgrade project is named High-Luminosity Large Hadron Col- lider (HL-LHC). The main goals of the upgrade is to achieve a peak luminosity of 5 × 10³⁴cm⁻²s⁻¹ and consequently an integrated luminosity of 250 fb⁻¹ per year to a total of 3000 fb⁻¹ in the active years of the collider [3]. As can be read in the installment plan given in Figure 5.1, the installment will start taking place in the middle of the coming decennium. The installment is set to coincide with a time when many critical components of the accelerator reach the end of their life cycles. Because of this, the purpose of the upgrade is not solely to acquire data faster and fully explore the potential of the collider, but it is also a necessity for operation beyond 2025 [22].

Figure 5.1: Plan for the HL-LHC upgrade of the LHC. Here LS stands for Long Shutdown and EYETS stands for Extended Year-End Technical stop [23].

The desire to raise the instantaneous luminosity to 5 × 10³⁴cm²s⁻¹at the end of HL-LHC operation puts pressure on the components of the collider. Some systems must be changed and possibly improved to avoid becoming a bottle- neck, aging too fast or even breaking down completely. The affected systems are primarily the inner triplet magnets, the cryogenics, the collimation, the dis- persion suppressor regions, radiation to electronics, and the superconducting links [3].

(32)

CHAPTER 5. UPGRADE 21

5.3 ATLAS Upgrade

The increase of luminosity caused by the upgraded LHC poses not only a chal- lenge to the parts of the LHC but also to the experiments connected to it. The upgrade of the LHC experiment results in 5 to 10 times higher occupancy of the detector [24]. With an increase of luminosity comes a higher demand on the detectors to successfully manage the larger amount of collisions. A simulation of the differences between a typical event at the current LHC luminosity level and HL-LHC luminosity level is presented in Figure 5.2. A greate chal- lenge is to adequately assign the right production vertices for charged particles created in the interactions.

(a) Simulation of 5 simultaneous collisions, corresponding to luminosities achievable by LHC.

(b) Simulation of 400 simultaneous collisions, corresponding to luminosities achievable by HL-LHC.

Figure 5.2: Difference in detector activity between LHC and HL-LHC [25].

The increase of occupancy of the detector is especially present in the systems of the detector at low radii and large η, such as the inner detector, the forward calorimeter, and the forward muon wheels. In order to successfully operate these systems need to be either notable upgraded or even completely replaced.

The plan is the swap the tracking detectors with new, radiation-hard ones that have a higher granularity and bandwidth. As a complement, the front-end electronics of the calorimeter will be replaced with radiation-hard versions as well. The capacity of the trigger and data acquisition system need to be ex- panded to be able to fully operate in this era of higher event rates and sizes.

The upgrades of the ATLAS detector is divided into two phases, scheduled to be completed in long shutdown 2 and long shutdown 3, respectively. Part of the phase-I upgrade consists of installing new schemes of triggering. These schemes contain multiple different trackers and improved granularity for the

(33)

first level calorimeters. Another part is the installation of new muon small wheels which will guarantee efficient tracking through a high particle rate.

After this penultimate round of upgrades of the ATLAS detector, the data- taking will be resumed.

The phase-II upgrade of the ATLAS detector coincides with the high luminosity upgrade of the LHC. The plan is to achieve a instantaneous luminosity of 5 × 10³⁴cm²s⁻¹up to a maximal luminosity of 7.5 × 10³⁴cm²s⁻¹. To handle the increase of luminosity the trigger and calorimeters are further upgraded.

But the core of this upgrade is to replace the entire Inner Detector with a new silicon-based Inner Tracker, ITk. The inner tracker will consist of one inner and one outer system. The innermost system will be pixel detector-based, while the surrounding system will be based upon strip detector technology.

A pixel support tube will separate the two volumes. The new inner tracker will be accompanied by a new High-Granularity Timing Detector (HGTD).

This timing detector will be positioned outside the inner tracker and in front of the end-cap and forward calorimeters approximately ±3.5 m away from the interaction point. It is more thoroughly described in chapter 5.5.

5.4 Radiation Hardness

The environment of a particle detector’s components is tough. Secondary particles produced by the collision of beams generate radiation damage on the sensors and the read-out electronics. This is especially true to the parts closest to the beamline, which will be subjected to more radiation than the ones radially further away. It is important that all of the parts of the detector still work as desired even after exposure to radiation. One measure of radiation exposure is the total ionizing dose effect. Effects on the device proportional to the accumulated ionizing dose on the device are for example creation of new electron-hole pairs in the gate insulation layers, which can activate defects in the material and thus degrade the device. Another problem caused by radiation exposure occurs when neutrons interact with the semiconductor lattice.

This interaction results in, amongst other things, a rearrangement of the atoms in the lattice, leading to an increase of recombination centers, and a depletion of minority carriers. In conclusion, radiation may create everlasting damages which might affect the efficiency of the detector.

(34)

5.5 High-Granularity Timing Detector

The High-Granularity Timing Detector (HGTD) is a proposed solution to the pile-up problem caused by the LHC upgrade. The detector has been developed to function under a total integrated luminosity of 4000 fb⁻¹and with 200 in- elastic collisions per bunch crossing. The HGTD will have an active area from r = 120 mm to r = 640 mm. Figure 5.3 shows a see-through of the ATLAS detector and points out the placement of the new detector. Table 5.1 gives a list of the main parameters of the HGTD.

Figure 5.3: Positioning of the HGTD inside the ATLAS detector [26].

The major improvements in the ATLAS detector resulting from the HGTD can be summarised as

• Pile-up reduction up to |η| < 4.

• Timing information for essentially all primary vertices.

• Timing determination for charged particles in the area 2.4 < |η| < 4.0.

• Measurements of luminosity for individual bunch crossing.

The HGTD and ITk will work in symbiosis. The ITk will track the spatial components of vertices and tracks while HGTD will complement with timing information. Because of this combination of detector parts, it is possible to reproduce events in both time and space. This allows for separation of objects that are spatially close but separated in time. An example is presented in Fig- ure 5.4. The figure visualizes the z-t plane of collision vertices within a single event with 200 interactions. Each ellipse indicates the resolutions of a truth

(35)

Pseudo-rapidity coverage 2.4 < |η| < 4.0

Thickness in z 75 mm (+ 50 mm moderator)

Position of active layers in z z = ±3.5 m

Radial extension:

Total 110 mm < r < 1000 mm

Active area 120 mm < r < 640 mm

Pad size 1.3 mm × 1.3 mm

Active sensor thickness 50 µm

Number of channels 3.59 M

Active area 6.4 m²

Average number of hits per track

2.4 < |η| < 3.1 ≈ 2

3.1 < |η| < 4.0 ≈ 3

Collected charge > 2.5 fC

Average time resolution per hit (start and end of operational lifetime)

2.4 < |η| < 3.1 ≈ 40 ps (start) ≈ 70 ps (end) 3.1 < |η| < 4.0 ≈ 40 ps (start) ≈ 85 ps (end) Average time resolution per track (start and end of operational lifetime) ≈ 30 ps (start) ≈ 50 ps (end)

Table 5.1: Main parameters of the HGTD [27].

vertex, i.e. a vertex of a true proton-proton interaction. The ellipses are 30 ps long in time and 1 mm wide in the z-direction. The dotted lines correspond to the z position of the reconstructed vertices in the event. The interesting vertex of the event is marked with red, while the blue ellipses indicate pile-up interactions. Since the tracker projects the event to the z-axis it has a problem separating vertices close in space, even though they are distinct in time.

The HGTD will be able to give a time-coordinate to the interactions during an event. It is therefore possible to extend the one-dimensional projection on the z-axis with an additional time axis, as shown in Figure 5.4. Adding this second axis makes it possible to distinguish the vertices close in space but separated in time and therefore improves the track-to-vertex association. The spread of collision times in the nominal operating scheme of the HL-LHC is expected to be 175 ps, and the time resolution for the proposed HGTD is approximately 30 ps per track for MIPs (minimum-ionizing particles) [27]. Time resolution of 30 ps per track increases the track-to-vertex association capabilities by a factor 6.

(36)

Figure 5.4: Visualisation of z-t plane of an event with hard scatter (red) and approximately 200 pile-up interactions (blue). The reconstructed vertices are represented by the dashed vertical lines [27].

5.5.1 Noise

In a perfect world, it would be possible to gather information using electronic devices and get data with no unwanted disturbances at all. Unfortunately, in reality, this is not the case. In electronics, noise is defined as being undesired disorders of a signal. The noise in the data describing events in this analysis can be illustrated by plotting the recorded voltage against the time, see Figure 5.5.

0 20 40 60 80 100

Time [ns]

−10 0 10 20 30 40 50

Voltage [mV]

An event with a pulse

Figure 5.5: Example of an event with noise and a pulse.

(37)

For most of the time, the noise while having no passing particle is kept at a notable lower amplitude than the amplitude from a particle passing through the sensor. As long as this is the case it is not difficult to separate the events with a pulse registered from the events only containing noise. However, there are some instances when the sensor is generating a pulse large enough to be registered as an actual pulse, even though no signal is passing through at that time. In this thesis, the phenomena with false pulses the read-out electronics misclassify as actual signals will be called auto-triggering. Table 5.2 shows the different sources of auto-triggers for irradiated and unirradiated sensors.

Table 5.2: Schematic overview of sources of auto-triggering for irradiated and unirradiated sensors.

Irradiated Unirradiated

Self-activation Yes No

Amplified eh-pairs Yes Yes

Delayed charge Yes No

A more thorough description of the three different causes of auto-triggering sensors is listed below.

1. Self-activation in neutron irradiated sensors.

2. Thermal activation from electron-hole pairs produced in high electric field-regions.

3. Delayed charge in the sensor caused by trapping and instabilities in it.

5.6 Low Gain Avalanche Detector

The HGTD upgrade will use a type of silicon detector known as Low Gain Avalanche Detector (LGAD) to meet the desired timing resolution. The detector is a planar detector with an internal gain first developed by Centro Nacional de Microelectrónica (CNM).

A sensor will be divided into arrays of pixels (pads). Each sensor will be bump- bonded to a custom Application Specific Integrated Circuit (ASIC) with the purpose of reaching the desired time resolution and radiation hardness. In addition, the ASIC will also provide the ability to count the number of registered hits in the sensor. The sensor and the ASIC will be connected to a flexible

(38)

cable, together forming a module. These modules will be mounted on both the front and the back of a common cooling disk. The modules are mounted to overlap in such a way so the number of hits exceeds the number of disks in the detector, see Figure 5.6. The optimal overlap between modules on the two sides of a disk has been found to be 80% for r < 320 mm, and 20% for r > 320 mm. In the radial direction, both the spacing and the placement of the modules are chosen so that a passing charged particle will, on average, result in three hits for r < 320 mm, and two for r > 320 mm. This result is deduced from the preliminary simulation shown in Figure 5.7. The higher demand on average hits for the inner region is required in order to successfully meet the desired per-track time resolution when the sensors have been irradiated.

Figure 5.6: Schematic drawing of the overlap between modules on the cooling plate. The horizontal line represent r = 320 mm [26].

The sensors are primarily composed of n-on-p doped silicon semiconductors.

An additional highly doped p-layer is inserted just below the n⁺ layer, see Figure 5.8. This extra p-layer greatly increases the doping concentration by the junction [28]. When a charged particle travels through the detector, a current is created from the electrons and holes in the material. this current is produced below the amplification region and after creation, the electrons travel to the top part of the device where the collecting electrode is positioned. When the current of electrons reaches the region of amplification, even more electron- hole pairs arise. This causes a large internal gain.

The size of the pixels is important when it comes to optimization of the detector. This is largely determined by occupancy. The occupancy of a pixel is defined as the percentage of pixels registering a hit. Since the timing information might be lost for pixels with more than one hit, smaller sizes are preferred. A smaller pixel size also reduces noise caused by electronics. However, smaller pixels result in more pixels in the detector and therefore an increase of both

(39)

Figure 5.7: Number of hits in the HGTD as a function of the position. The overlap at R < 320 mm is 80%, while the number for R > 320 mm is 20%

[27].

Figure 5.8: Cross Section of an LGAD sensor [29].

inactive zones between the pixels and channels to be instrumented. Figure 5.9 shows the occupancy as a function of radius for three different read-out cell sizes. As expected, the occupancy is negatively correlated to the radius and positively correlated to the pixel size. The required maximum occupancy level for the upgrade is 10%, which is obtained with a pixel size of 1.3 × 1.3 mm². Each end-cap of the HGTD contains two double-sided layers mounted on two cooling disks that also work as support. Sets of modules with bundled cables form read-out rows. Each cooling disk is separated into four quadrants where the rows are alternated horizontally and vertically between each segment. The layers of the cooling disks are rotated by 15^◦ in opposite directions, shown in Figure 5.10.

(40)

Figure 5.9: Occupancy vs. radius for various pixel sizes at a pile-up of < µ >

= 200 [27].

Figure 5.10: Orientation of readout rows for the first layer, last layer, and the overlay of the two. The layers are rotated 15^◦ in opposite directions [27].

5.7 Radiation hardness of the HGTD

By the end of HL-LHC, when the integrated luminosity reaches a total of 4000 fb⁻¹ the neutron-equivalent fluence at r = 120 mm reaches 6.8 × 10¹⁵ neq cm⁻². The total ionizing dose exposing the parts will be approximately 4.2 MGy. Both of these numbers are presented in Figure 5.11, which shows the particle fluence and total ionizing dose as a function of radius and at z

= ±3.5 m. Due to the importance of keeping the sensors and the electronics functional a safety factor of 1.5 is applied to both of these values. In addition to this, another safety factor of 1.5 is enforced to the total ionizing dose to secure against any uncertainty in the behavior of the electronics after exposure to irra- diation. Because of this, the sensors will have a total safety factor of 1.5 while the electronics will have a safety factor of 2.25. The sensors are more volatile to the particle fluence and therefore determines the limit of neutron-equivalent

(41)

fluence as 10.2 × 10¹⁵neqcm⁻², while the electronics is more sensitive to the total ionizing dose and therefore sets the limit for it to 9.5 MGy.

(a) Nominal neutron-equivalent fluence for HL-LHC.

(b) Nominal ionising dose for HL-LHC.

Figure 5.11: Nominal neutron-equivalent fluence and ionising dose as func- tions of the radius in the outermost sensor layer of the HGTD for 4000 fb⁻¹, i.e. before including safety factors. The contribution from charged hadrons is included in "Others" [27].

Due to the large amount of radiation damage the innermost sensors and electronics are subjected to it is suggested that the innermost part of the HGTD should be replaced halfway through the HL-LHC program. The proposal is to replace both the sensors and read-out electronics up to a radius of about 320 mm (3.1 < |η| < 4.0). In total, this corresponds to 32% of the modules. Be- cause of this replacement, the maximum neutron-equivalent fluence effecting the devices will be 5.1 × 10¹⁵neq cm⁻², and a total ionising dose of 4.7 MGy.

5.8 Time resolution

To successfully improve the track-to-vertex association ability of the detector, a time resolution of 30 ps per track is proposed. This proposed time resolution will improve the track-to-vertex association ability of the ATLAS detector with a factor 6. The main contributions to the time resolution for a detector element are given by the equation

(42)

σ²_total = σ_L²+ σ_clock² + σ_elec² , (5.8.1) where σ_L²is the Landau fluctuations, σ_clock² is the contribution coming from the clock, and σ_elec² comes from the read-out electronics [27]. The first term of the right hand side in Eq. (5.8.1) is dependent on the thickness of the sensor.

Thinner silicon gives a lower impact of the fluctuations on the time resolution.

For an LGAD sensor with a thickness of 50 µm, σL ≈ 25 ps [26]. The contribution from the clock, σclockis demanded to be reduced to below 10 ps. The third term of the right hand side of in Eq. (5.8.1) contributes as

σ_elec² = σ_TW² + σ_TDC² + σ_Jitter² . (5.8.2) In Eq. (5.8.2), σ_TW² is the contribution from the time walk. Time walk is the effect when the trigger time is dependent on the signal’s peak height. For the data in the test beam, a constant fraction discriminator (CFD) with a fraction of 50% of the amplitude of the peak has been used for correcting the time walk. The impact from the time walk can be set to zero as long as an offline correction is used. The second contribution to Eq. (5.8.2) is the contribution from the Time-to-Digital Converter, σ²_TDC. The purpose of a time-to-digital is to digitalize the Time Of Arrival (TOA) and Time Over Threshold (TOT) measurements for an event. The time the conversion takes is dependent on the bin size of the converter. The final contribution is the jitter contribution, σ_Jitter² , given by the time jitter equation

σ_Jitter = N

(dV/dt) ' t_rise

S/N, (5.8.3)

Figure 5.12: Slew rate based on the thickness of a sensor and the gain [27].

(43)

where S is equal to the signal, N is the noise, dV/dT is the slope of the signal, and triseis the rise time [27]. The dependence of the signal slope by the thickness of the device and the gain is presented in Figure 5.12. By examining the figure it is possible to see that a thin sensor with a large gain gives the smallest contribution of σJitter to the time resolution. For an LGAD with a thickness of 50 µm, and a gain of 20 gives a slew rate of approximately 130 [27]. σTW

should be smaller than 10 ps. Finally, assuming a TDC bin of size 20 ps, the effect of the time resolution of σTDCapproximately 7 ps and is neglected. The total contribution per LGAD becomes

σ_LGAD =√

25²+ 10²+ 25² ≈ 37 ps. (5.8.4) This gives a track time resolution of

σ_Track= σ_LGAD

√

Number of sensor hits

, (5.8.5)

which fulfills the criteria of maintaining the track resolution to a maximum of 30 ps for the two areas with an average of two and three hits per track.

5.9 Time reconstruction

As mentioned in chapter 5.8, it is important for the upgrade of the detector that the LGAD sensors time resolution reduces below a specific threshold. In theory, the optimal time resolution is achieved by performing complex time reconstruction techniques using all the information of the shape of the incom- ing pulse. In practice, the data bandwidth for the HGTD is limited to only three different methods: Constant Threshold Discriminator (CTD), Constant Fraction Discriminator (CFD), and Zero-Crossing Discriminator (ZCD) [28].

Constant Threshold Discriminator is the most straightforward method, defining the time of arrival as the time when the signal surpasses a constant threshold. The time is interpolated linearly between the first sample point before and the next sample point after the threshold. The method suffers from time walk effects. Its influence can be fended if the signal amplitude is known, which in most cases, it is not. A more complex approach is using a fraction of the pulse height instead of a specific value, and defining the time of arrival as the time when the pulse reaches that specific threshold. This is done using the Constant Fraction Discriminator method. This method minimizes the effect of the time walk since it defines the time of arrival as the time when the signal crosses a constant fraction of the maximum amplitude. A problem with the

(44)

CFD method is that it relies on a threshold depending on the maximum amplitude which is unknown until after the actual pulse is finished. Because of this, it is impossible to implement the CFD method in the read-out electronics. Computation of the time of arrival between the two mentioned methods is presented in Figure 5.13.

Figure 5.13: A comparison between constant threshold triggering (left) and constant fraction triggering (right) of two pulses with the same max time, but different amplitudes [30].

The final method, the Zero-Crossing Discriminator method, is the most complex one. It copies the signal, delays the copy by dZCD and weakens it by a factor fZCD. The zero-crossing time of the difference between the original and the mended copy is independent of the signal amplitude and therefore does not share the read-out problem CFD method has. As long as the shape remains the same for the original and the copied pulse it will, by construction, be independent of the amplitude. It is this method, along with the CTD method combined with a time walk correction, that is under consideration for the front-end electronics in the HGTD.

(45)

Chapter 6 Test Beam Setup

In order to get information on how the LGAD sensors will behave in the HGTD, they must be tested in conditions similar to the environment they will be used in. This is done using a test beam of charged pions. This chapter describes the experimental setup of the test beam, the components used, and ends with a description of data files acquired from the tests.

6.1 Method of Gathering Data

The data used in this thesis was gathered at CERN’s site in Prévessin, on the French side of the border between Switzerland and France. The data was collected using a beam telescope, a setup used to focus a beam and track reconstruction to track the charged particles originating from the beam. A test beam is ideal to use for developing and testing new sensors since it portrays realistic usage of the detector, also allowing an easy change of measurement parameters. There exist many different telescopes. The one used in this experiment was AIDA, which is a EUDET-type of beam telescope [31]. A setup scheme of the EUDET beam telescope is presented in Figure 6.1.

A charged pion ray, with particles with an energy of 120 GeV was used in the experiment. The machine used for testing is a telescope consisting of six aluminum-housed sensors put perpendicular to the beam. These telescope sensors are of the Mimosa26 type. The first three sensors, numbered 0, 1, and 2, make up the upstream part of the set-up. After the first three mimosa planes, the beam reaches the Silicon Photomultiplier (SiPM) and the LGAD sensors.

The LGADs are positioned on a support table between the two mechanical arms holding the mimosa planes. The support table makes it possible to further

34

(46)

CHAPTER 6. TEST BEAM SETUP 35

Figure 6.1: Overview of the data acquisition setup [31]. Up to two oscilloscopes can be used to test more than three DUTs simultaneously.

detail the placement of the LGADs by freely moving it around in the active area of the telescope. For each oscilloscope, one of the connected boards needs to be a SiPM sensor, which is used for reference. Up to eight boards can be sequentially mounted and connected to two oscilloscopes, making it possible to test six LGADs simultaneously. The following three mimosa planes make up the part called downstream with plates labeled 3, 4, and 5. In this setup, an FE-I4 chip is placed between mimosa plane 3 and 4 and is used as a reference plane, spanning a Region Of Interest (ROI). After the downstream plates, a final scintillator completes the experimental setup of the telescope. Both the FE-I4 and the scintillator are used for triggering. Only signals hitting both the FE-I4 and the scintillator can be recorded. A Trigger Logic Unit (TLU) receives signals from the two planes and sends a signal to the oscilloscope and National Instruments (NI) crate to store data.

6.2 Device Under Test

The Devices Under Test (DUTs) in the set up are the LGAD sensors. All the CNM single-padded LGADs studied during the three test beams in 2018 are presented in Tables 6.1, 6.2, and 6.3. The single-padded LGADs are only housing one diode, in comparison the array which contains multiple pads. The sensors have been implanted using gallium. The radiation in the tables is ex- pressed in neutron equivalent fluence per square cm. For read-out, the detectors are equipped with ATLAS LGAD Timing Integrated Read-Out Chips (ALTIROCs).

(47)

36 CHAPTER 6. TEST BEAM SETUP

Table 6.1: Sensors tested in the April 2018 test beam session.

April 2018

Sensor Radiation [neq/cm²] Sensor Radiation [neq/cm²]

W4S1064 6 × 10¹⁵ W4S1068 1 × 10¹⁵

W4S1016 6 × 10¹⁴ W4S1058 1 × 10¹⁵

W4S1017 3 × 10¹⁵ W4S1067 1 × 10¹⁴

W4S1095 1 × 10¹⁴ W4S1030 0

W4S1102 6 × 10¹⁵

Table 6.2: Sensors tested in the June 2018 test beam session.

June 2018

Sensor Radiation [neq/cm²] Sensor Radiation [neq/cm²]

W5S1013 0 W5S1038 10¹⁴

W4S1021 6 × 10¹⁵ W5S1078P 10 × 10¹⁴

W5S1115 1 × 10¹⁵ W5S1039P 30 × 10¹⁴

W5S212 0 W4S207 6 × 10¹⁴

6.3 Mimosa26

Tracking in the test beam setup is done by Mimosa26 sensors. The sensors are wrapped up in aluminum housing, and are of the size 13.7 × 21.5 mm². In total, each plate is composed of approximately 700 000 pixels ordered in a 576 rows × 1152 columns matrix. With a pitch of 18.4 µm, these pixels cover an active area of 10.6 × 21.2 mm². The sensors are Monolithic Active Pixel Sensors (MAPS) and have a fast binary read-out, making it possible to obtain single-point spatial resolution in the LGADs with a precision of order 2 µm.

A complete read-out cycle for the Mimosa26 takes 115.2 µs. The average fake hit rate for the Mimosa 26 sensor is 10⁻⁴per pixel [32].

6.4 Silicon Photomultiplier

The LGADs are not equipped with any time reference system. Therefore a Silicon PhotoMultiplier (SiPM) is needed in order to get a time resolution of the tested LGADs. The SiPMs have a known time resolution of < 20 ps and work as reference sensors for the LGADs [34]. The SiPMs have active sizes of 3 × 3 mm² and are single cell mounted on small boards. The sensors are fastened on baseboards with amplifiers and enclosed in 3D printed boxes.

Detectors for the ATLAS High-Granularity Timing Detector

Auto-triggering studies of Low Gain Avalanche Detectors for the ATLAS High-Granularity Timing Detector

FREDRIK SJÖSTRÖM

Auto-triggering studies of Low Gain Avalanche

Detectors for the ATLAS High-Granularity Timing Detector

FREDRIK SJÖSTRÖM

Abstract

Sammanfattning

Contents

List of Figures

List of Tables

Chapter 1 Introduction

1.1 Author’s Contribution

Chapter 2

Standard Model

2.1 Fermions

2.2 Quarks

2.3 Leptons

2.4 Forces

Chapter 3

Large Hadron Collider

3.1 Luminosity

3.2 Pile-up

Chapter 4

The ATLAS Detector

4.1 Inner Detector

4.1.1 Pixel Detector

4.1.2 Semiconductor Tracker

4.1.3 Transition Radiation Tracker

4.2 Calorimeter

4.2.1 Liquid Argon Calorimeter

4.2.2 Hadronic Tile Calorimeter

4.3 Muon Spectrometer

4.4 Trigger and Data Aquisition

4.5 Jet selection

Chapter 5 Upgrade

5.1 Physics

5.2 High-Luminosity Large Hadron Collider

5.3 ATLAS Upgrade

5.4 Radiation Hardness

5.5 High-Granularity Timing Detector

5.5.1 Noise

5.6 Low Gain Avalanche Detector

5.7 Radiation hardness of the HGTD

5.8 Time resolution

5.9 Time reconstruction

Chapter 6

Test Beam Setup

6.1 Method of Gathering Data

6.2 Device Under Test

6.3 Mimosa26

6.4 Silicon Photomultiplier