Analysis of Test Beam Data for Sensors in the High ...1117693/FULLTEXT01.pdf · A test beam study...

Analysis of Test Beam Data for Sensors in the HighGranularity Timing Detector

Author:Filip Backman (910516-5550)

[email protected]

Department of PhysicsRoyal Institute of Technology (KTH)

Supervisor: Bengt Lund-JensenCo-supervisor: Alex Kastanas

June 2, 2017

Typeset in LATEX

ISRN KTH/FYS/––17:35—SEISSN 0280-316XTRITA-FYS 2017:35©Filip Backman, 2017

Abstract

A test beam study of 7 different Low Gain Avalanche Diode (LGAD) sensors is presented.The study is made for a proposed upgrade with an additional subdetector system in theATLAS experiment at CERN called the High Granularity Timing Detector (HGTD).The HGTD proposal is briefly presented, both as a trigger, mitigating pile-up, and as apotential luminometer. The thesis includes a presentation of the alignment and trackingprocedure using an EUDET-type beam telescope. The study of the LGADs is focusedon the particle detection efficiency, the gain uniformity and the time resolution of thesensors.

Sammanfattning

En teststrale-studie pa 7 olika ”Low Gain Avalanche Diode” (LGAD)-sensorer presen-teras. Studien gors for en foreslagen uppgradering av ett ytterligare subdetektorsystemi ATLAS experimentet pa CERN som kallas ”the High Granularity Timing Detector”(HGTD). Forslaget presenteras kortfattat, bade som en utlosare for att separera ”pile-up”, samt som en potentiell luminositetsmatare. Tesen inkluderar en presentation av lin-jering och sparning med hjalp av ett EUDET-stralteleskop. Studien av LGAD-sensorernafokuserar pa effektiviteten for partikelupptackt, jamnheten for gain, samt tidsupplosnin-gen for sensorerna.

Contents

1 Introduction 31.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.2 Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.3 Outline of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.4 Author’s Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Background 52.1 Large Hadron Collider . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52.2 The ATLAS experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.2.1 Inner Detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.2.2 The Calorimeters . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.2.3 The Muon Spectrometer . . . . . . . . . . . . . . . . . . . . . . . 11

2.3 Upgrade of the ATLAS Detector . . . . . . . . . . . . . . . . . . . . . . . 122.4 The High Granularity Timing Detector . . . . . . . . . . . . . . . . . . . 122.5 HGTD for Measuring Luminosity . . . . . . . . . . . . . . . . . . . . . . 142.6 Low Gain Avalanche Diodes . . . . . . . . . . . . . . . . . . . . . . . . . 17

3 Experimental Test Beam Setup 193.1 Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193.2 The Telescope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193.3 The LGADs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

4 Analysis Method and Tracking 224.1 Alignment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224.2 Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234.3 Analysis of the LGADs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

4.3.1 Hit Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244.3.2 Gain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254.3.3 Time Resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

5 Results 285.1 Alignment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285.2 The Single Pad Run . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285.3 The Array Run 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355.4 The Array Run 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

1

6 Discussion and Conclusions 426.1 Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 426.2 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

7 Outlook 44

8 Acknowledgments 45

A MIMOSA plots for the SPR 51

B MIMOSA plots for the AR1 56

C MIMOSA plots for the AR2 61

D Support information for Analysis 66

2

Chapter 1

Introduction

1.1 Introduction

The ATLAS experiment [5] is located at CERNi studying proton-proton collisions fromthe Large Hadron Collider (LHC) [7]. In search of new physics, CERN has to enhancethe performance of the LHC and upgrade accordingly. The detectors have to keep upwith new higher luminosity. The High Granularity Timing Detector (HGTD), which isthe focus of this thesis, is one of the upgrades proposed to keep up with the evolution ofthe rest of the experimental site.

Before a new subdetector system can be accepted in the system, it’s impact on thephysics studies through it’s performance, construction and reliability needs to be proven.For the HGTD this includes tests to determine the specifics of sensors, providing infor-mation for the construction of the detector. This is often done in several steps withdifferent detectors to be able to choose the most suitable setup for the new subdetectorpurpose.

This thesis will present the process of analyzing the data from the HGTD test beamof October 2016, and propose some enhancements for the upcoming tests in 2017. Theresults of the processed data is presented.

1.2 Aim

An analysis to provide data for an informed decision on whether or not to build theHGTD is made. The further tests, which will be done in the summer and early autumn2017, will also determine which sensors are the most suitable to use for the HGTD ifapproved. The aim of this thesis is to analyze the particle detection efficiency, the gainand the time resolution as functions of position in the sensors studied. The particledetection efficiency provides information on how well the sensors register hits of passingparticles. The gain is of interest for the uniformity of the signal, and the time resolutionprovides information on performance as a timing device.

iThe European Organization for Nuclear Research.

3

1.3 Outline of the Thesis

The thesis will begin by providing a background in chapter 2. A technical specificationof the beam telescope and the setup will be provided in chapter 3. The method usedto extract data is provided in chapter 4, before presenting the results in chapter 5. Adiscussion with conclusions follows in chapter 6 and at last an outlook is presented inchapter 7.

1.4 Author’s Contribution

The author has collaborated with Alex Kastanas and Edvin Sidebo of KTH for theanalyses presented in the thesis. The results are produced by the author unless else isspecified.

4

Chapter 2

Background

2.1 Large Hadron Collider

The Large Hadron Collider (LHC) is a circular accelerator located at CERN, which issituated at the Swiss-French border outside of Geneva, Switzerland. The LHC is placed100 meters below ground and has a circumference of approximately 27 km. In the LHC,both heavy ions and protons are brought to collide. This thesis will only describe theproton-proton collision case, since it is the main objective of the ATLAS experiment.Before protons are inserted into the LHC, they need to be accelerated in several pre-accelerators. These accelerators are: the Linac2, the Proton Synchrotron Booster (PSB),the Proton Synchrotron (PS) and the Super Proton Synchrotron (SPS).

The protons that are used for collisions are not uniformly separated in the beam pipe,but grouped up in so-called bunches. The bunches are separated by 25 ns and containmore than 1011 protons in a width of only 64 µm at the interaction point. The amountof protons in each bunch has increased over time and will continue to do so. This willlead to a higher luminosity, which is the number of collisions per bunch crossing. Graphsof the evolution of luminosity at LHC is shown in Figure 2.1.

Figure 2.1: The expected evolution of the LHC in the upcoming runs [8].

5

The first accelerator, the Linac2, is a linear accelerator used to give the particlesa kinetic energy of 50 MeV. It uses cylindrical conductors around the proton beamline, which are alternatively positively and negatively charged. This is done so thatthe conductors ahead are charged to attract the proton, pulling it forward, whereasthe conductors behind will repel the proton, pushing it forward. The protons are theninjected into the first circular accelerator, the PSB, which continues to accelerate theparticles to 1.4 GeV. The beam of particles in circular accelerators are contained by amagnetic field produced by dipole magnets along the beam pipe. In the acceleratorsthere are radiofrequency (RF) cavities which are placed in intervals along the pipe. TheRF cavities work in a similar fashion to the linear accelerator with alternating fieldsaccelerating the protons. The main advantage of circular accelerators is that they canprogressively raise the kinetic energy of the particles for every lap they take. The PSaccelerates the particles to 25 GeV and injects them into the last pre-accelerator, theSPS, where their energy is increased to 450 GeV. The particles are at this point injectedinto the LHC, and accelerated up to an energy of 6.5 TeV. A schematic layout of all theaccelerators connected to the LHC can be seen in Figure 2.2.

Figure 2.2: A schematic of the LHC, with all the collision points and the pre-acceleratorsdisplayed. The CMS, ALICE, ATLAS and LHCb experiments are located at the collisionpoints. Courtesy of CERN.

When the protons have been accelerated in the LHC, they are brought to collide inthe different experiments shown in Figure 2.2. ATLAS is the largest of these detectorsystem and is the only one that will be discussed in this thesis.

2.2 The ATLAS experiment

Detailed information about this design and performance can be found in [5] and is onlysummarized in this section.

The ATLAS experiment is the largest detector ever constructed for a particle collider.ATLAS has the shape of a cylinder with a diameter of 25 m, and a length of 44 m. It

6

is symmetric in the z-direction of the beam from the beam spot and it weighs approxi-mately 7000 metric tonnes. The ATLAS detector system has several different subdetectorregions, each designed for a specific purpose. The main regions are: The inner detector,the calorimeter and the muon spectrometer. Their positioning can be seen in Figure 2.3.

In ATLAS it is very common to refer to a spatial coordinate called pseudorapidity(η). It is a way of defining the angle that a particle has taken relative to the beam line.The definition of η is formulated in Equation 2.1, where θ is the angle between the beamline and the particle.

η = ln

(tan

(θ

2

))(2.1)

The ATLAS detector can be separated in two main areas; the barrel and the end-caps.The barrel is the cylindrical area which surrounds the beam line radially, whereas theend-caps are the front and back end of the detector. To separate these spatially one oftenrefers to the end-caps being in the high |η| range (or the forward region), whereas thebarrel is in the low |η| range (or the central region). A cross-section of the ATLAS in thebarrel region is illustrated in Figure 2.4, where also the characteristic tracks of differentparticles are shown.

Figure 2.3: A schematic of the full ATLAS detector, showing the different subdetector systems.Courtesy of ATLAS.

2.2.1 Inner Detector

The inner detector is aimed at tracking charged particles in a magnetic field. It is the firstdetector region that a particle from the interaction will travel through. It is enveloped in asolenoid magnet of 2 T, which will bend the charged particles and allowing measurementsof their momentum and charge. It consists of three different subdetector regions. ThePixel Detector, the Semi-Conductor Tracker (SCT) and the Transition Radiation Tracker(TRT). A schematic of the inner detector can be seen in Figure 2.5

7

Figure 2.4: A cross-section of the central part of ATLAS, showing how different particles willproduce tracks. Courtesy of the ATLAS experiment.

The inner detector uses track points in the different subdetectors to measure thedirection and momentum of all charged particles that traverse it. The subdetectors willbe discussed below.

Figure 2.5: A schematic of the inner detector. Courtesy of CERN.

Pixel Detector

The Pixel Detector is a high granularity semi-conductor. The sensors have anodes andcathodes attached to a semi-conductor material. When a charged particle traverses it, it

8

will free electrons/holes which will travel toward the anode/cathode. The current is thenread out and a signal is produced. It has 4 layers of pixel sensors. Altogether these fourlayers have 80 million pixels with a size of 50x400 µm. The innermost layer is placed 3.5cm away from the beam spot and the outermost at a radius of 12.2 cm. It also has 3end-cap discs on each side having 6.6 million pixels.

This part of the detector gives an output which corresponds to the energy deposited inthe detector by the particle which traverses it, and gives signals above a noise threshold.

SCT

The SCT consists of silicon microstrips, that are placed in a cylindrical fashion envelopingthe pixel detector. The microstrips themselves are placed parallel to the beam line in thebarrel and in the end-caps they are directed radially. It is the second closest subdetectorfrom the beam line. The SCT consists of 4 barrel layers and 18 planar end-cap discs. Ithas roughly 8 million channels and the readout strips have a pitch of 80 µm and a lengthof 6 cm. This gives it a positional measurement accuracy of 17 µm × 580 µm, where thelatter is obtained due to a 40 mrad stereo angle. This detector only registers hits abovethreshold or no hit.

TRT

The transition radiation tracker is the third part of the inner detector. The entire moduleconsists of 50000 tubes in the barrel and 250000 tubes along the end-caps. The straws inthe barrel are 144 cm long whereas those in the end-caps are 39 cm long. The straw tubesare cylindrical with a radius of 2 mm. They are filled with a Xenon gas mixture and havegold-plated tungsten wires with a radius of 0.3 mm placed in the center. When a chargedparticle traverses the gas, it gets ionized. Due to an electric field within the straw, thefreed electrons will travel towards the wire in the center. As the electrons travel towardsthe center they will accelerate which creates an avalanche effect freeing more electrons.The electrons will then create a signal inside the wire, which will be read out. The strawsare surrounded by a plastic material, which is used to induce transitional radiation. Theamount of radiation is proportional to the Lorentz boost factor (E

m), thus making this

detector able to separate heavier hadrons from electrons due to the difference in mass.

2.2.2 The Calorimeters

The calorimeters measure the energy of the incoming particles. They are created suchthat the particles will deposit their energy and be fully absorbed. Both calorimetersof ATLAS are so called sampling calorimeters. This means that they use two differentmaterials. One absorber which induces showers from the incoming particles, and oneactive material that will measure the number of particles in the shower, which is propor-tional to the energy deposited. These two materials are placed in an alternating fashionthroughout the calorimeter. A schematic of ATLAS is shown in Figure 2.6, where thecalorimeters are highlighted.

The calorimeters of ATLAS can be divided into two subcategories, the electromagnetic(EM) calorimeter and the hadronic calorimeter. The EM calorimeter will contain theentire charge deposited by both photons and electrons, whereas the hadrons that traverse

9

will deposit some energy within it. The hadrons will continue to shower in the hadroniccalorimeter in which they will be fully absorbed.

Figure 2.6: A schematic of the ATLAS calorimeters and their positions inside the ATLAS.Courtesy of CERN.

The Electromagnetic Calorimeters

The EM calorimeters consists of a barrel and two end-caps, using Liquid Argon (LAr) asthe active substance. In all of the EM calorimeters, lead is used as absorption material.

The Argon needs to be at a temperature of roughly −183C to be in a liquid state.The radially placed lead absorbers are accordion shaped, and the space between is filledwith LAr. The accordion folds are parallel to the beam in the barrel part, while theyare radial in the end caps. The barrel is 6.4 m long and has a thickness of 53 cm. Theend-caps have a radius of 2.077 m and a thickness of 0.632 m.

The Hadronic Calorimeters

The hadronic calorimeter system consists of a central and two extended barrel tilecalorimeters and two LAr end-caps on each side. The tile calorimeters have iron ab-sorbers and plastic scintillators as the active material. The hadronic calorimeters areplaced directly outside the EM calorimeters. The LAr hadronic end-caps work in a simi-lar way as the LAr EM calorimeter, but has planar copper plates as the absorber materialinstead.

The tile barrel together with the extended barrels have a total mass of 2600 metrictonnes, which makes it one of the more massive pieces in ATLAS.

The Forward Calorimeter

The LAr forward calorimeter (FCAL) covers the region 3.1 < |η| < 4.9. It consistsof three layers, the first being the EM part, and the two further out being hadroniccalorimeters. The absorber of the EM calorimeter is made of copper, and the two outer

10

are made of tungsten. The detector has a limited depth, which makes the only optionfor this part to be very dense, so that the particles are completely absorbed. The lengthof this detector is around 10 nuclear interaction lengths.

2.2.3 The Muon Spectrometer

After the calorimeters, only muons and neutrinos are left. The neutrinos are not measuredin the ATLAS detector as they only interact weakly. Their energy is instead estimatedby calculating the missing transverse energy. The muons however are detected in the lastlayer called the muon spectrometer, separating them from the neutrinos. It is envelopedby a large toroidal magnetic system, which bends the path of the particles and makesthe muons measurable in the spectrometer. The magnetic system consists of 8 toroidalcoils placed around the barrel and 8 in each end-cap. A schematic of ATLAS where thesedifferent parts are highlighted is shown in Figure 2.7.

The muon spectrometer consists of 3 different detector types. The monitored drifttubes, the cathode strip chambers and the trigger chambers.

Figure 2.7: A schematic of ATLAS, showing where the different parts of the muon spectrom-eter are placed. Courtesy of the ATLAS experiment.

Monitored Drift Tubes

The monitored drift tubes (MDTs) are placed in the |η| < 2.0 region and use a similartechnology as the TRT, which gives it high precision tracking. It consists of around350000 tubes with a resolution of 80 µm. The active material is Ar-CO2 and it has atungsten wire centered in the tube.

Cathode Strip Chambers

The cathode strip chambers are placed in the end cap regions of ATLAS. It consists oftwo discs with 8 chambers each on them. They replace the MDTs for the innermostlayer in the region 2.0 < |η| < 2.7 where the particle flux is too high for the MDTs.

11

It uses the same gas, but has a multi-wire configuration with radially directed anodes.The cathodes are segmented into strips where some are parallel to the wire and some areperpendicular, which gives a spatial resolution of 60 µm.

Triggering Chambers

The trigger chambers give a rapid measurement of the muons passing by. There are twodifferent categories of trigger chambers. Thin Gap Chambers (TGCs) and Resistive PlateChambers (RPCs). The RPCs are placed in the central region of the barrel, and consistof a resistive plate configuration with gas in the gaps between. The muons that traverseinduce a current in the resistive plates, which create an avalanche of electrons who’scharge is read out. The TGCs are placed in the high η region and has a similar setup asthe CSC. This is needed to measure the muons in this region since their momentum isgenerally higher and thus bent less.

2.3 Upgrade of the ATLAS Detector

To achieve the physics goals of the LHC experiments in the search of signs of new physics,as large number of collisions as possible is required. Rare processes and the fact that onlya tiny fraction of the proton-proton interactions involves the full energy of the beams,necessitates an upgrade of the LHC to increase the interaction rate by a factor 10. Theaim is to achieve this in the mid 2020’s. The results of a simulation to show the differencebetween events at the increased luminosity at the High Luminosity (HL) LHC comparedto the LHC at present time is shown in Figure 2.8.

The factor 10 increase in luminosity in the upgraded HL LHC will result in an averageof 200 collisions per bunch crossing. This has large implications for the experiments thathave to cope with so-called pile-up, which is when particles from several collisions givesignals in the detectors at the same time. To mitigate pile-up it is of importance tohave a high timing resolution and granularity as this will separate collisions in time andprevent saturation of the detectors.

The ATLAS experiment will among other improvements replace the inner detectorto be completely based on semi-conductor technologies. The main focus of this thesis isa proposed High Granularity Timing Detector (HGTD). The HGTD will, through highgranularity, good timing resolution and quick read out, be proficient in separating pile-upevents. In order to efficiently trigger only on the interesting collisions, a possibility wouldbe to use the HGTD information.

A further motivation for the High granularity timing detector is it’s possibility tobe used for fast luminosity measurement. This will be discussed more thoroughly insection 2.5.

2.4 The High Granularity Timing Detector

The HGTD is a proposed upgrade for the ATLAS experiment, and a decision on whetherto incorporate it into the detector will be made in September 2017. The HGTD is foreseento be placed in front of the FCAL and the end-cap EM calorimeter, covering the region

12

(a) Simulation of the current LHC. (b) Simulation of the HL LHC.

Figure 2.8: The difference in typical detector activity between HL LHC and the LHC dis-played. Courtesy of ATLAS (http://atlas.web.cern.ch/Atlas/GROUPS/UPGRADES/).

2.4 < |η| < 4.2. The area which it covers will be approximately 20000 cm2 in a diskshape (Figure 2.9).

Figure 2.9: Proposed HGTD as of January 2017. Red depicts the beam line, the yellow areashows where the detectors with finer granularity are placed, and the blue area shows the sensorswhich are 9 times as large. ATLAS public plots [1].

The main purpose of the HGTD is to mitigate pile-up by being able to separatedifferent collisions in a bunch crossing by measuring the collision time and the particletime of flight. To be able to achieve this, a high timing resolution is needed. To avoidsaturation of the sensors, high granularity is also of importance, with an aim of getting

13

http://atlas.web.cern.ch/Atlas/GROUPS/UPGRADES/

an occupancy below 10% at high luminosity. [10]The studied proposal contains pads in two different sizes. The sensor pads at high

|η| are required to be smaller than the outer pads, as the particle flux is higher there(Figure 2.10). The proposed sizes are 3× 3 mm for the region 2.4 < |η| < 3.2 and 1× 1mm for 3.2 < |η| < 4.2.

There are two proposed designs of HGTD, and both use Low Gain Avalanche Diodes(LGAD), which are fast silicon sensors (section 2.6) in four active layers (Figure 2.12).One of the designs contains interleaved tungsten layers between the active sensors toinduce showers in the layers behind. The tungsten would cover only the region 2.4 <|η| < 3.2, to reduce the degradation effect it will have on the FCAL energy resolution [10].The shower induction from the tungsten layers raises the occupancy in the sensors, asshown in Figure 2.11.

Figure 2.10: The occupancy of the HGTD design with SiW config as a function of radiusbased on MC simulations. An alternative x-axis is displayed above the image. ATLAS publicplots [1].

2.5 HGTD for Measuring Luminosity

Luminosity measurements can be made by calculating the number of hits in the forwardregion of ATLAS as most events have some products which travels through this region.Soft scatteringi events dominate the collisions in ATLAS and this type of interactiongenerate charged particles in this region. When hard scattering events occur, there isoften more activity in the central region but some rest products will travel along thebeam line and will also be picked up by the HGTD.

iEvents where only one parton from each proton interact.

14

(a) The occupancy in all |η|-ranges.(b) The occupancy in the |η|-rangesproposed.

Figure 2.11: A comparison between the Si configuration and the SiW configuration in termsof occupancy in the third layer based on MC simulations. ATLAS public plots [1].

An obstacle to consider that creates background signals in luminosity measurements isso-called afterglow from previous bunch crossings. This includes effects of when particlesfrom previous bunch crossings have irradiated a detector part, which will then releaseneutrons that travel through the detector and give rise to detector signals. These particlestravel at a low speed, and create background noise.

The HGTD’s intrinsic properties with low occupancy and high time resolution aredesigned to solve this. The low occupancy means that the sensors will rarely be saturatedby the particles traveling through. The high time resolution means that afterglow effectscan be efficiently separated from particles that are the result of a collision as their timingis independent of the bunch crossing time.

A work currently in progress is to show that the number of hits will have a linearrelation to the luminosity. This is done by making Monte Carlo (MC) simulations ofbunch crossings with a variation of 〈µ〉ii and showing that the output from the proposedHGTD model would show this behavior.

iiThe number of average collisions per bunch crossing.

15

Carbon Fiber: 1mm

Cooling : 3mm

Si-Sensor : 0.15mm

HV Kapton : 0.15mm

Glue : 0.2mm

PCB : 1.5mm

Chip height : 1.2mm

Ratios : x 10:1y 1:1

Space: 1mm between absorber (including cooling) and HV kapton 1mm between layers Total width : 42.8mm

Aerogel: 5mm

(a) Silicon only

(b) Silicon - Tungsten

Figure 2.12: Cross-sectional view of the two different proposed designs for HGTD. ATLASpublic plots [1].

16

2.6 Low Gain Avalanche Diodes

Low Gain Avalanche Diodes (LGADs), seen in Figure 2.13, are a type of silicon detectorthat has an extra doping layer, where a high electric field generates avalanche effects bymultiplication of the signal charge. Fast silicon sensors for high-rate operation need tobe thin since the saturation of the drift velocity (Figure 2.14) limits the charge collectiontime. The time resolution however depends on the signal-to-noise ratio (SNR) of thesensor. Using charge multiplication to increase the charge yield could help to achievegood time resolution for thin silicon sensors [11].

The purpose behind LGADs is to improve the speed of the signal. The current in thereceptors is proportional to; the charge, the speed of the carriers (holes/electrons), andthe electric weighting field.

Figure 2.13: A schematic of an LGAD [6].

Maximizing the current, while keeping the entire pad as uniform as possible leads tomaximizing the parameters, and also to making sure the entire pad will have a uniformsignal strength.

The velocity v is dependent on the electric field but has a saturation value, where therise of the electric field no longer effects the velocity. This can be seen in Figure 2.14.One aim is to operate in the saturated area to maximize the carrier velocity and thushave a stronger signal.

Figure 2.14: The carrier velocity as a function of the electric field [2].

17

Maximizing the charge is done by introducing the doping layer that induces anavalanche. The avalanche will free more electrons and in such a way create more carriers,that create more charge in the signal.

The last variable to consider is the weighting field, which is determined by the electricfield covering the pad. This field is what effects the uniformity the most. Uniformity isachieved by having large pads. The reason why this makes the field uniform is that theanodes, which create this field, have a finite extent and thus the field will be non-uniformat the edges. Therefore larger pads will have less edges in total for the entire subdetectorsystem.

18

Chapter 3

Experimental Test Beam Setup

3.1 Setup

The test beam telescope setup used is called AIDA and is an EUDET-type beam telescope[3]. The information presented about the sensors used in the telescope is largely basedon the specifications of its sister telescope DATURA [9]. For more information on thedifferent telescopes and their components, see [3].

The telescope, seen in Figure 3.1, is built out of 6 light MIMOSA silicon detectors.These were used to get a spatial pointing resolution in the LGADs during the tests. Thetrigger logic unit for this setup was in the form of an FE-I4 plane placed between the 4thand 5th MIMOSA plane in the so-called downstream part. The MIMOSAs were separatedinto sets of three detectors which are referred to as the upstream (the ones closest to thebeam origin) and the downstream (the three which are behind the LGADs). The beam isa fixed target pion beam line with an energy of 120 GeV, which produces parallel beamsperpendicular to the detector planes.

3.2 The Telescope

The MIMOSA silicon detectors have a thickness of 50 µm. For reference, this is a thirdof the thickness of the studied LGADs. They have a pixel size of 18.5 µm × 18.5 µm.The pixels are placed in 1152 columns and 576 rows, which covers an area of 21.312 mm× 10.656 mm. The MIMOSA are read out column-wise over an integration time of 115.2µs. Hits that pass a certain charge deposition threshold are read out, but without pulseheight information. This method removes noise below the threshold effectively, but doesnot provide collected charge information for a charge barycenter approximation of thecluster center. Clustering is needed for the tracking as the trace of the particles don’tnecessarily fire in only one pixel, but may trigger several adjacent.

The FE-I4 is extremely fast, which makes it possible to trigger on single hits. Forevery hit in the FE-I4, an average of 300 hits are registered in each of the MIMOSAplanes. The FE-I4 has a pixel size of 250 µm × 50 µm and contains 80 columns and 336rows. This provides an active area of 20 mm × 16.8 mm. A masking of the FE-I4 isapplied, which means that the active area of the plane is limited only to trigger in thearea of interest for analysis of the LGADs.

19

Figure 3.1: Detailed view of the setup, containing distances between the different detectors.The positions of the LGADs are referred to as 0 to 3, from left to right. Figure adapted andadjusted from image provided by Richard Polifka.

3.3 The LGADs

For the October 2016 test beam data taking a detailed figure of the setup is shown inFigure 3.1. Two different sets of tests were recorded during this period. In the first, padsof four sensors were placed in 2-by-2 arrays with 4 readout channels each. In this type oftests, which will be referred to as the array runs (AR), only the position 0 and 2 shownin Figure 3.1 were used. The ARs provide us with spatial information and a look at thegap between sensors. They also allow analysis of the crosstalk between the sensor pads.They do however not contain a Silicon Photo Multiplier (SiPM), which is needed as atime reference for time resolution studies. The other type of tests had LGADs placedat position 0, 1 and 2, whereas position 3 contained a SiPM. The SiPM, which has aknown time resolution of 15 ps, is used as a reference to get the time resolution of theLGADs in the tests. These types of tests will be referred to as the single pad runs (SPR).Several different LGADs are tested with different bias voltage. For the three sets of dataused in this thesis, the specifics of the sensors are listed in Table 3.1. For the readoutof the LGADs, two oscilloscopes were used in the ARs and only one for the SPR. Theoscilloscopes used had 4 channels, each coupled to an LGAD used in the tests and onefor the SiPM in the SPR, whereas each of the oscilloscopes was used for the readout ofthe different arrays in the ARs.

20

RunPosition 0Sensor, Resistance(bias Voltage / scale)

Position 1Sensor, Resistance(bias Voltage / scale)



1364 & 1367 (SPR)W5 LGA31, 2 kΩ(220V / 100mV)

W5 LGA33, 470 Ω(220V / 100mV)

W5 LGA34, 470 Ω(80V / 100mV)

SiPM3(27.5V / 100mV)

1662, 1663 & 1664 (AR1)IN2P3, 360 Ω(200 V / 50mV)

-W7 HG22, 1070 Ω(180 V / 50 mV)

-

1707 & 1708 (AR2)W11 HG11, 1070 Ω(80 V / 50mV)

-W7 HG22, 1070 Ω(180 V / 50 mV)

-

Table 3.1: The specifics of the sensors tested corresponding to the run.

21

Chapter 4

Analysis Method and Tracking

The purpose of the study presented is to analyze three properties of the LGAD sensorsstudied as functions of position; the particle detection efficiency (or hit efficiency), thegain and the time resolution. In order to get good spatial resolution, reconstruction ofthe tracks in the test beam is required.

The reconstruction of tracks from the test beam raw data is done in several steps. Ina first step the alignment of the beam telescope planes were studied. By reconstructingtracks it was possible to get a position of the particle in the global coordinate system, inwhich the LGAD positions are included.

The LGADs were then analyzed to measure the hit efficiency, the gain uniformity andthe time resolution as functions of x and y positions given by the beam telescope.

4.1 Alignment

The alignment procedure is split into two different parts, an approximate alignment anda track-based alignment. A selection was made to use only events which had exactlyone FE-I4 hit, which removed up to 10% of the events recorded. Using the alignmentconstants, the local coordinates of each pixel are translated to global coordinates.

A first approximate alignment was made where all the mimosas were approximatelyaligned with respect to the FE-I4. This was done by taking the FE-I4 hit and requiringhits from as many mimosa planes as possible within the triggered FE-I4 pixel area plusa padding of 200 µm. This method was done by using a brute force function, testing allpossible alignments, and reducing the step size during several iterations until a maximumnumber of hits within the area of acceptance was acquired. This approximate alignmentprovides a setting in which a track-based alignment is functional.

After the first approximate alignment, tracks were reconstructed to allow track-basedalignment. The number of compatible hits from each plane is presented in Figure 4.1.Whenever a hit was found, a clustering was made where all adjacent pixels were testedto see if they also registered a hit. If two or more hits were discovered in neighboringpixels, a calculation of the average position was made and used for the hit data. A first fitwas made with all hits that fulfilled an initial demand of being in range of the triggeredFE-I4 pixel with a padding of 200 µm. A second test was then made to look for newmatching hits along the track of the first fit. If the new set of hits were superior to theprevious, these were chosen instead. This process was repeated until an optimized trackwas found.

22

Figure 4.1: Number of compatible hits before clustering (run 1662 of AR1).

When the optimized tracks had been constructed, the tracking alignment started,where the mimosas were aligned to minimize the residuals between the registered MI-MOSA hits and the tracks. A residual smaller than half the mimosa pixel was acceptedas a perfect hit, as this is the spatial extent of the pixel size in the MIMOSA planes. Thisprocedure considered 1000 events per iteration. The alignment of the planes was donefor the layers closest to the FE-I4 first and then continued outwards to the other planes.This order lowered the constraints on the outer layers which had a worse alignment dueto multiple scattering.

4.2 Tracking

Following the alignment, new tracks were constructed with slightly different requirements.When producing the tracks hits in at least 4 MIMOSA planes were required, with a limiton the slope of 3 × 10−4. Requiring hits in at least 4 MIMOSA planes for these trackswas due to the LGADs being displaced by up to 1 mm in the ARs. This prevented thefirst two layers of the MIMOSA to cover the area of interest (Figure B.7, Figure B.8,Figure C.7 and Figure C.8). One of the array runs (AR2) also had an issue with a large

23

dead area in the center of the LGADs for MIMOSA plane 4 (Figure C.11), and thusthe demand was a criterion to even register hits in this area. All of the plots for theMIMOSA can be seen in Appendix A, Appendix B and Appendix C. The limit on theslope is imposed because the residuals of the tracks depended on the slope. This canbe seen in Figure 4.2, and is explained by multiple scattering processes in the LGADs.The multiple scattering bends the beams to non-linear tracks throughout the telescope,thus causing a larger slope than the tracks from particles that did not undergo multiplescattering. In Figure 4.3, one can see that the majority of the tracks have 6 planes hit.These tracks are later used for spatial positioning of the LGADs. The resolution of thetracks was determined by the residuals from the track to the center of the fired cell of theFE-I4. Gaussian fits were made for this in both the x- and y-directions of the FE-I4. Thepixel size of the FE-I4 is so large that it can only be approximated by two half-gaussiansalong the edges.

Figure 4.2: The total residual of the tracks as a function of the slope of the track (for run1662 of AR1).

4.3 Analysis of the LGADs

The hit efficiency and gain uniformity are studied as functions of spatial coordinates forall tests, whereas in the single pad tests the time resolution was also extracted. Thethresholds for the output of the LGADs are selected by studying graphs of their outputamplitude (Figure 4.4). There is a steady noise in the pads, which is removed by applyingthis threshold.

4.3.1 Hit Efficiency

The hit efficiency of the LGADs is found by looking at all tracks and seeing if any of theLGADs were activated, the thresholds for the different sensors are listed in Table 4.1.For every track produced we look at which of the channels give a signal above threshold

24

Figure 4.3: Number of MIMOSA hits per track constructed (for run 1662 of AR1).

and attribute a hit to the coordinate of the track. If the charged particle correspondingto the track gives no signal in the LGADs, no hit was attributed at the track position onthe sensor. After all the events were analyzed, the hit efficiency, defined as the ratio ofhits compared to the total number of passing particles, was calculated for each bin.

RunSensor at Position 0Threshold [V]

Sensor at Position 1Threshold [V]



SPRLGA31-0.05

LGA33-0.04

LGA34-0.03

SiPM-0.06

AR1W7 HG22-0.04

-IN2P3-0.02

-

AR2W7 HG22-0.04

-W11 HG11-0.045

-

Table 4.1: The determined thresholds for the different LGADs.

4.3.2 Gain

The gain is the amplification factor of signal for the charge deposited within a sensor.For all the registered hits in the LGAD sensor, which are selected in the same way asfor the hit efficiency plots, the total charge deposited in the sensor is proportional to theintegral of the entire pulse. The total charge is filled in a histogram. A Landau fit is madeto extract the most probable charge deposition in the sensor. After the fit is done, thecharge is divided by a factor which is calculated by multiplying the board transimpedance,given in Table 3.1, with a normalization factor and the expected collected charge from aminimum ionizing particle in the absence of amplification, which is 0.46 fC in this case.

25

Figure 4.4: Voltage output in different channels of the LGADs. Threshold put at -0.04 forthe first 4 and -0.02 on the last 4 (for run 1662 of AR1).

4.3.3 Time Resolution

To calculate the time resolution the time difference between the SiPM and the LGADswere studied for each spatial coordinate. For this value, the standard deviation, σTOT,was used as the width, which is the time resolution of the entire system. This can bewritten on the form shown in Equation 4.1, which could be solved for σLGAD since thetime resolution of the SiPM is known.

σTOT = σ2SiPM + σ2

LGAD (4.1)

For it to be possible to calculate the time resolution with statistical significance, one

26

needs to ensure enough statistics. When this was calculated, a demand was set thatevery pixel needed to be filled with at least 10 particles.

27

Chapter 5

Results

This chapter presents the results of the alignment as well as the measured efficiency, gainand time resolution of the tested LGADs. More information on the studied LGADs canbe found in [4].

5.1 Alignment

The alignment constants used for all the MIMOSA planes in tracking are shown inTable 5.1. These constants translate all MIMOSA pads to the coordinate system ofthe FE-I4. The differences in alignment constants between the different runs is in themagnitude of µm, thus showing the stability of the setup.

RunAlignment [µm]Layer 0

Alignment [µm]Layer 1





SPR(-4244.9,6283.2)

(-4786.9,6072.2)

(-5283.9,5614.2)

(-6406.9,4531.2)

(-6640.9,4058.2)

(-6370.9,4250.2)

AR1(-4248.7,6291.1)

(-4815.7,6081.1)

(-5296.7,5628.1)

(-6415.7,4545.1)

(-6657.7,4076.1)

(-6393.7,4287.1)

AR2(-4248.9,6283.2)

(-4812.9,6073.2)

(-5293.9,5625.2)

(-6422.9,4539.2)

(-6657.9,4076.2)

(-6399.9,4291.2)

Table 5.1: The alignment constants for the MIMOSA of the different runs.

The residuals and hits for the MIMOSA planes can be seen in Appendix A, Ap-pendix B and Appendix C.

5.2 The Single Pad Run

The single pad run consists of two different runs (1364 and 1367). A run is defined as aconsecutive set of testing under stable conditions. The LGADs studied in this test arethe LGA31, LGA33 and LGA34, where all have a size of 1× 1 mm. They have the samesetup with the exception of the FE-I4 mask. Run 1364 has a large active area, givingless information for the same number of triggers, and run 1367 has an active area whichcuts the LGADs. The effect of this is visible in Appendix A and Figure D.8.

The difference between the position pointed to by the beam telescope and the centerof the FE-I4 pixel hit in x and y directions is shown in Figure 5.1. For a perfect pointingwith the beam telescope, this would have the shape of a step function. Thus the slope

28

at the edges of the distributions are caused by the track pointing resolution. Fittinghalf-gaussian functions to the edges thus determines the tracking accuracy. The FE-I4is chosen instead of a MIMOSA plane for determining the accuracy, since it is unbiasedto the track-based alignment procedure.

(a) σ = 8.64, 8.69 for the differentedges.

(b) σ = 8.78, 9.05 for the differentedges.

Figure 5.1: Track residuals in x and y for the FE-I4 in the SPR. The corresponding σ to thefits are presented beneath the plots.

The hit efficiency of the LGA31, LGA33, and LGA34 sensors as a function of positioncan be seen in Figure 5.2, Figure 5.3 and Figure 5.4 respectively. The thresholds chosento determine whether or not they fired was selected from distributions similar to thoseshown in Figure 4.4. The thresholds selected for the sensors in the SPR can be seen inTable 4.1. The position bin sizes chosen are 10× 10 µm. The choice is made to have atleast 5 particles in each bin. The number of hits per bin is presented in Figure D.8.

Figure 5.2: Efficiency of the LGA31 sensor. The region to the left is effected by the edge ofthe active area of the FE-I4.

A larger bin size of 50 × 50 µm is chosen for both the gain and time resolution.This selection is made due to them both being more sensitive to statistical fluctuations.

29

Figure 5.3: Efficiency of the LGA33 sensor. The region to the right is effected by the edge ofthe active area of the FE-I4.

Figure 5.4: Efficiency of the LGA34 sensor. The region to the right is effected by the edge ofthe active area of the FE-I4.

Landau fits for an arbitrarily chosen representative bin are shown in Figure 5.5 for thethree studied sensors. A fit for every bin was produced and the results are presented inFigure 5.6, Figure 5.7 and Figure 5.8 for the three sensors. A minimum of 10 hits perbin is required for the fit to remove low statistics bins at the edges of the active area.

The time resolution, as defined in subsection 4.3.3, of the different LGADs in the SPRare shown in Figure 5.9, Figure 5.10 and Figure 5.11. Only the data from run 1364 wasused, due to a lack of timing information from the oscilloscope for run 1367, lowering thestatistics of the bins. The number of hits per bin can be seen in Figure D.11, Figure D.12,

30

Figure 5.5: Landau fits for the pulse registered multiplied by the gain factor described in sub-section 4.3.2 in an arbitrarily chosen bin (5000, 13000 in Figure 5.6, Figure 5.7 and Figure 5.8)for the SPR.

Figure 5.6: Gain of the LGA31 sensor. The recorded number of hits for each bin can be seenin Figure D.1.

and Figure D.13.

31

Figure 5.7: Gain of the LGA33 sensor. The recorded number of hits for each bin can be seenin Figure D.2

Figure 5.8: Gain of the LGA34 sensor. The recorded number of hits for each bin can be seenin Figure D.3

32

Figure 5.9: The time resolution of LGA31 as a function of position.


33


34

5.3 The Array Run 1

The array run 1 contains the runs 1662, 1663 and 1664. They have two boards, eachholding 4 LGADs. The LGADs studied in this test are the IN2P3 and the W7 HG22which have the sizes 3× 3 mm and 2× 2 mm respectively.

The track pointing resolution is calculated in the same way as for the SPR, and ispresented in Figure 5.12. The difference in the pointing resolution between the AR1 andthe SPR is due to hot cells in MIMOSA plane 1 in combination with a misalignment ofthe planes 0 and 1. This can be seen in Figure B.7 and Figure B.8 on page 58.



Figure 5.12: Track residuals in x and y for the FE-I4 in the AR1. The corresponding σ tothe fits are presented beneath the plots.

The hit efficiency for the IN2P3 and W7 HG22 sensors can be seen in Figure 5.13.The thresholds selected for determining if it is firing or not are presented in Figure 4.4and Table 4.1. The bins have a size of 80 × 80 µm. This bin size is chosen to have atleast 100 particles per bin.

Figure 5.13: Efficiency of the IN2P3 sensor.

35

Figure 5.14: Efficiency of the W7 HG22 sensor.

The Landau fits to the pulse height made for the LGADs in an arbitrarily chosenrepresentative bin are shown in Figure 5.15. The same method as for the SPR is used.The bin size is set to 133× 133 µm to ensure enough statistics for a fit.

Figure 5.15: Landau fits for the pulse registered multiplied by the gain factor described insubsection 4.3.2 in an arbitrarily chosen bin (4200, 8700 in Figure 5.16 and Figure 5.17) for theAR1.

The gain for all bins are shown in Figure 5.16 and Figure 5.17.

36

Figure 5.16: Gain of the IN2P3 sensor. The recorded number of hits for each bin can be seenin Figure D.4

Figure 5.17: Gain of the W7 HG22 sensor. The recorded number of hits for each bin can beseen in Figure D.5

37

5.4 The Array Run 2

The array run 2 contains the data taken from the runs 1707 and 1708. These havearray sensors similar to AR1, but the boards instead hold the W11 HG11 and W7 HG22LGADs. These LGADs have the sizes 3× 3 mm and 2× 2 mm.

The resolution of the track pointing for the AR2 is shown in Figure 5.18. The pointingresolution for this test is affected by a dead area in MIMOSA plane 4 (Figure C.11) as wellas from the misalignment in MIMOSA plane 0 and 1, previously mentioned in section 5.3.All the MIMOSAs hit information for AR2 can be seen in Appendix C.



Figure 5.18: Track residuals in x and y for the FE-I4 in the AR2. The corresponding σ tothe fits are presented beneath the plots.

The hit efficiencies of the W11 HG11 and W7 HG22 can be seen in Figure 5.19 andFigure 5.20. The threshold determining whether or not the sensor registered a hit wasselected in the same way as for both the SPR and the AR1 and can be seen in Table 4.1.The bin sizes in all plots of the AR2 are the same as for AR1, i.e. 80 × 80 µm for theefficiency plots and 133× 133 µm for the gain plots.

An example of the Landau fits made for each of the different LGADs are shown inFigure 5.21.

The gain calculated for both different LGADs in the AR2 are found in Figure 5.22and Figure 5.23.

38



39

Figure 5.21: Landau fits for the pulse registered multiplied by the gain factor described insubsection 4.3.2 in an arbitrarily chosen bin (4200, 8700 in Figure 5.22 and Figure 5.23) for theAR2.


40


41

Chapter 6

Discussion and Conclusions

6.1 Tracking

The tracking method presented in this thesis reconstructs tracks from hits in all planesand maximizes the number of hit planes used for the fit within the constraints mentionedin section 4.2. Another method available is the so-called seeding method, where tracksare initially reconstructed using only two planes and then extrapolated to pick up hitsin the surrounding planes. The motivation for using the method of this thesis is that itdepends less on the data from a few selected planes. For example a seeding procedureusing MIMOSA plane 3 and 4 would be unstable for the AR2, since the MIMOSA plane4 has a large dead area (Figure C.11 on page 65). The seeding method was done as across check by another group at the ATLAS experiment.

Using brute force, testing all combinations, for the first stage alignment is a decisionthat was made as the hits in each layer are non-correlated, so approximating it as anyfunction would be impossible, and local minima would make the alignment diverge.

Multiple scattering is the largest problem throughout the tracking, since this will bendthe path of the particles and thus create a non-linear trace. The path of most scatteredparticles deviate only slightly from the original path, which lets us reconstruct relativelyprecise tracks even though they are approximated as straight lines. The effect is furtherlimited by the method of setting constraints on the slope of the tracks. The multiplescattering in the MIMOSA planes is negligible due to the thickness of the planes, thusimplying that it mainly occurs in the LGADs, which make scattering occur primarilybetween the upstream and downstream MIMOSA planes.

6.2 Analysis

The alignment constants seen in Table 5.1 show that the setup is similar in all threetests that were analyzed. The largest difference is between the SPR and both the ARs,which is due to them being separated the most in time. Several other beam tests of otherequipment, which are not presented in this thesis, were done between these runs, leadingto people working on the beam telescope setup effecting the alignment constants betweenthe tests.

Using two separate gaussian fit for the FE-I4 resolution is necessary due to the plateaubeing too broad for a single fit. This is however not optimal as these fits are separated

42

from each other. The resolutions gathered from the two different fits are however similar.Hot cells effect the results and can be seen in both AR1 and AR2 (Figure B.8 &

Figure C.8). One might also notice that all four of these plots (Figure B.7, Figure B.8,Figure C.7 & Figure C.8) are shifted slightly in the y-direction, which means that theiredge cuts the area of the studied LGADs. The hot cells in combination with the missinginformation from the planes result in the hot cells drawing tracks toward it’s position.This effect is due to the procedure presented in this thesis prioritizing tracks with hits in 5MIMOSA planes to those with only 4. The effects are clearly visible in both Figure D.9and Figure D.10. The issue of the large dead area in Figure C.11 does not effect thetracking as much since it is not coupled with hot cells.

The track residuals of the MIMOSAs can be seen in the Appendix A, Appendix Band Appendix C. Here one may notice that the SPR has the best residuals. This is alsobelieved to be an effect of the missing information in the AR1 and the AR2, as well asthe hot cells. This effect is also seen in the FE-I4 for all tests (Figure 5.1, Figure 5.12 &Figure 5.18) where the track pointing resolution is calculated.

For the SPR, the active area of the FE-I4 was changed between the two runs. For run1364, the selected active area was larger than the area covered by the LGADs studied.This provided information in areas of no particular interest for the tests done. For run1367, however the active area of the FE-I4 was narrowed down with the purpose of onlyselecting events of interest for the analysis of the LGADs. Thus, for a given number ofevents run 1367 has more events of interest than run 1364.

The efficiency of both the ARs and the SPR are high as expected. Silicon pads ingeneral have an efficiency of 99% and above, which is supported by these results. For theSPR the binning is smaller than for the ARs, which provides it with much less statisticsper bin and thus a larger error. It does however provide more information on the edgesas an effect of this, since the bins are less smeared. This trade-off was considered whenchoosing the bin sizes. The difference in bin sizes between them is motivated by twofactors. The efficiency being higher for the SPR, and thus making it possible to reducethe samples without loss of information. Also, the track pointing resolution of the SPRis better, providing a more accurate position of the particles traversing the LGADs.

The gain of the pads is fairly uniform, the difference between the sensors seen in thearray runs may suggest a larger issue. This difference may be due to a difference in theproduction of the sensors and the uncertainty in the board transimpedance.

The time resolution of the SPR is non-uniform, but seems to be below 50 ps for allpads. The non-uniformity may be due to errors in the calculation since it is based on therelation between multiple sensors, the LGADs and the SiPM, where there are errors inboth. The statistics for the time resolution is also lower due to it only using one of theruns (1364) in the SPR, which further creates errors.

43

Chapter 7

Outlook

For upcoming LGAD tests it would be interesting to get the time resolution for allLGADs. The array runs didn’t have any way of analyzing this, because of the lack of areference pad. A suggestion would be to include a SiPM for all runs, this can be achievedby using data acquisition systems with more channels.

The size of the active area needed in the SPR is much smaller than that of the ARs.This is due to two factors. One is that only one sensor is tested in each plane, the other isthat the sensors tested in the SPR have a finer granularity (1×1 mm) whereas the sensorsused in the ARs are larger (2× 2 mm and 3× 3 mm). A smaller area of interest makes afiner selection of the FE-I4 possible. This leads to more statistics for the area of interestper triggering event. A suggestion would be to swap these, so that the smaller sensorsare used in ARs and the larger ones in SPRs to make the active area approximately thesame size for all runs. This would create more similar statistics in the different tests.

The uncertainty regarding the board transimpedance in the array runs is somethingthat not only effects the gain estimations for the different pads, this also has implicationson the time resolution, in a way explained in section 2.6. The uniformity of the pads arenot effected by this, but this leads to normalisation issues since the transimpedance is amultiplicative factor to the gain.

There are some suggestions on how to enhance the software used. The trackingalignment could be done over larger sample sizes. Instead of aligning with regard to 1000events at a time, one could use 10000 events per iteration. This limitation was set bythe computational power of the machine used. The tracking can also be enhanced byremoving the hot cells as found in MIMOSA plane 1.

For the next run, a suggestion to rotate the LGADs with a slight angle to align withthe FE-I4 may reduce the error caused by the binning. For the result plots of the AR,the tilt is clearly visible, where a slight step is found on the edges of the sensors. Theseedges have less statistics, and are thus more prone to producing erroneous results.

A final suggestion for the next run is to align the MIMOSA carefully to the LGADs,to avoid the misalignment of planes in the beam telescope and thus making it possibleto create tracks with a higher accuracy.

44

Chapter 8

Acknowledgments

I would like to thank the entire ATLAS group working at KTH for useful discussions,which have provided me with insights into both particle physics and research in general.Special thanks go out to my collaborator Edvin Sidebo for his work on the software,especially in the time resolution segment where he was the main contributer, and toProf. Bengt Lund-Jensen for providing support throughout my thesis project. And lastbut not least - I owe the greatest of gratitudes to Alex Kastanas who has held my handthroughout the project, providing code for all parts of the analysis and for taking timeout of his busy schedule to provide me with useful insights regarding the subject. Hismood never failed him, not even when we had to stay late at night to ”nuke the code”(His words... and also mine). A thanks go out to everyone working in the Particle-and Astroparticle group at KTH for the warm reception and interesting conversationtopics. Finally, I would also like to thank my family and my girlfriend for supporting methroughout this project.

45

List of Figures

2.1 The expected evolution of the LHC in the upcoming runs [8]. . . . . . . . 52.2 A schematic of the LHC, with all the collision points and the pre-accelerators

displayed. The CMS, ALICE, ATLAS and LHCb experiments are locatedat the collision points. Courtesy of CERN. . . . . . . . . . . . . . . . . . 6

2.3 A schematic of the full ATLAS detector, showing the different subdetectorsystems. Courtesy of ATLAS. . . . . . . . . . . . . . . . . . . . . . . . . 7

2.4 A cross-section of the central part of ATLAS, showing how different par-ticles will produce tracks. Courtesy of the ATLAS experiment. . . . . . . 8

2.5 A schematic of the inner detector. Courtesy of CERN. . . . . . . . . . . 82.6 A schematic of the ATLAS calorimeters and their positions inside the

ATLAS. Courtesy of CERN. . . . . . . . . . . . . . . . . . . . . . . . . . 102.7 A schematic of ATLAS, showing where the different parts of the muon

spectrometer are placed. Courtesy of the ATLAS experiment. . . . . . . 112.8 The difference in typical detector activity between HL LHC and the LHC

displayed. Courtesy of ATLAS (http://atlas.web.cern.ch/Atlas/GROUPS/UPGRADES/). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.9 Proposed HGTD as of January 2017. Red depicts the beam line, the yellowarea shows where the detectors with finer granularity are placed, and theblue area shows the sensors which are 9 times as large. ATLAS publicplots [1]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.10 The occupancy of the HGTD design with SiW config as a function ofradius based on MC simulations. An alternative x-axis is displayed abovethe image. ATLAS public plots [1]. . . . . . . . . . . . . . . . . . . . . . 14

2.11 A comparison between the Si configuration and the SiW configuration interms of occupancy in the third layer based on MC simulations. ATLASpublic plots [1]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.12 Cross-sectional view of the two different proposed designs for HGTD. AT-LAS public plots [1]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.13 A schematic of an LGAD [6]. . . . . . . . . . . . . . . . . . . . . . . . . 172.14 The carrier velocity as a function of the electric field [2]. . . . . . . . . . 17

3.1 Detailed view of the setup, containing distances between the different de-tectors. The positions of the LGADs are referred to as 0 to 3, from left toright. Figure adapted and adjusted from image provided by Richard Polifka. 20

4.1 Number of compatible hits before clustering (run 1662 of AR1). . . . . . 234.2 The total residual of the tracks as a function of the slope of the track (for

run 1662 of AR1). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

46



4.3 Number of MIMOSA hits per track constructed (for run 1662 of AR1). . 254.4 Voltage output in different channels of the LGADs. Threshold put at -0.04

for the first 4 and -0.02 on the last 4 (for run 1662 of AR1). . . . . . . . 26

5.1 Track residuals in x and y for the FE-I4 in the SPR. The corresponding σto the fits are presented beneath the plots. . . . . . . . . . . . . . . . . . 29

5.2 Efficiency of the LGA31 sensor. The region to the left is effected by theedge of the active area of the FE-I4. . . . . . . . . . . . . . . . . . . . . . 29

5.3 Efficiency of the LGA33 sensor. The region to the right is effected by theedge of the active area of the FE-I4. . . . . . . . . . . . . . . . . . . . . . 30

5.4 Efficiency of the LGA34 sensor. The region to the right is effected by theedge of the active area of the FE-I4. . . . . . . . . . . . . . . . . . . . . . 30

5.5 Landau fits for the pulse registered multiplied by the gain factor describedin subsection 4.3.2 in an arbitrarily chosen bin (5000, 13000 in Figure 5.6,Figure 5.7 and Figure 5.8) for the SPR. . . . . . . . . . . . . . . . . . . . 31

5.6 Gain of the LGA31 sensor. The recorded number of hits for each bin canbe seen in Figure D.1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

5.7 Gain of the LGA33 sensor. The recorded number of hits for each bin canbe seen in Figure D.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

5.8 Gain of the LGA34 sensor. The recorded number of hits for each bin canbe seen in Figure D.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

5.9 The time resolution of LGA31 as a function of position. . . . . . . . . . . 335.10 The time resolution of LGA33 as a function of position. . . . . . . . . . . 335.11 The time resolution of LGA34 as a function of position. . . . . . . . . . . 345.12 Track residuals in x and y for the FE-I4 in the AR1. The corresponding

σ to the fits are presented beneath the plots. . . . . . . . . . . . . . . . . 355.13 Efficiency of the IN2P3 sensor. . . . . . . . . . . . . . . . . . . . . . . . . 355.14 Efficiency of the W7 HG22 sensor. . . . . . . . . . . . . . . . . . . . . . . 365.15 Landau fits for the pulse registered multiplied by the gain factor described

in subsection 4.3.2 in an arbitrarily chosen bin (4200, 8700 in Figure 5.16and Figure 5.17) for the AR1. . . . . . . . . . . . . . . . . . . . . . . . . 36

5.16 Gain of the IN2P3 sensor. The recorded number of hits for each bin canbe seen in Figure D.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

5.17 Gain of the W7 HG22 sensor. The recorded number of hits for each bincan be seen in Figure D.5 . . . . . . . . . . . . . . . . . . . . . . . . . . 37

5.18 Track residuals in x and y for the FE-I4 in the AR2. The correspondingσ to the fits are presented beneath the plots. . . . . . . . . . . . . . . . . 38

5.19 Efficiency of the W11 HG11 sensor. . . . . . . . . . . . . . . . . . . . . . 395.20 Efficiency of the W7 HG22 sensor. . . . . . . . . . . . . . . . . . . . . . . 395.21 Landau fits for the pulse registered multiplied by the gain factor described

in subsection 4.3.2 in an arbitrarily chosen bin (4200, 8700 in Figure 5.22and Figure 5.23) for the AR2. . . . . . . . . . . . . . . . . . . . . . . . . 40



A.1 The residuals from the hits to the tracks in MIMOSA plane 0 for SPR . . 51

47

A.2 The residuals from the hits to the tracks in MIMOSA plane 1 for SPR . . 51A.3 The residuals from the hits to the tracks in MIMOSA plane 2 for SPR . . 52A.4 The residuals from the hits to the tracks in MIMOSA plane 3 for SPR . . 52A.5 The residuals from the hits to the tracks in MIMOSA plane 4 for SPR . . 52A.6 The residuals from the hits to the tracks in MIMOSA plane 5 for SPR . . 52A.7 The hits in MIMOSA plane 0 used for the tracks in the SPR. . . . . . . . 53A.8 The hits in MIMOSA plane 1 used for the tracks in the SPR. . . . . . . . 53A.9 The hits in MIMOSA plane 2 used for the tracks in the SPR. . . . . . . . 54A.10 The hits in MIMOSA plane 3 used for the tracks in the SPR. . . . . . . . 54A.11 The hits in MIMOSA plane 4 used for the tracks in the SPR. . . . . . . . 55A.12 The hits in MIMOSA plane 5 used for the tracks in the SPR. . . . . . . . 55

B.1 The residuals of the tracks in MIMOSA plane 0 . . . . . . . . . . . . . . 56B.2 The residuals of the tracks in MIMOSA plane 1 . . . . . . . . . . . . . . 56B.3 The residuals of the tracks in MIMOSA plane 2 . . . . . . . . . . . . . . 57B.4 The residuals of the tracks in MIMOSA plane 3 . . . . . . . . . . . . . . 57B.5 The residuals of the tracks in MIMOSA plane 4 . . . . . . . . . . . . . . 57B.6 The residuals of the tracks in MIMOSA plane 5 . . . . . . . . . . . . . . 57B.7 The hits in MIMOSA plane 0 used for the tracks in the AR1. . . . . . . . 58B.8 The hits in MIMOSA plane 1 used for the tracks in the AR1. . . . . . . . 58B.9 The hits in MIMOSA plane 2 used for the tracks in the AR1. . . . . . . . 59B.10 The hits in MIMOSA plane 3 used for the tracks in the AR1. . . . . . . . 59B.11 The hits in MIMOSA plane 4 used for the tracks in the AR1. . . . . . . . 60B.12 The hits in MIMOSA plane 5 used for the tracks in the AR1. . . . . . . . 60

C.1 The residuals of the tracks in MIMOSA plane 0 . . . . . . . . . . . . . . 61C.2 The residuals of the tracks in MIMOSA plane 1 . . . . . . . . . . . . . . 61C.3 The residuals of the tracks in MIMOSA plane 2 . . . . . . . . . . . . . . 62C.4 The residuals of the tracks in MIMOSA plane 3 . . . . . . . . . . . . . . 62C.5 The residuals of the tracks in MIMOSA plane 4 . . . . . . . . . . . . . . 62C.6 The residuals of the tracks in MIMOSA plane 5 . . . . . . . . . . . . . . 62C.7 The hits in MIMOSA plane 0 used for the tracks in the AR2. . . . . . . . 63C.8 The hits in MIMOSA plane 1 used for the tracks in the AR2. . . . . . . . 63C.9 The hits in MIMOSA plane 2 used for the tracks in the AR2. . . . . . . . 64C.10 The hits in MIMOSA plane 3 used for the tracks in the AR2. . . . . . . . 64C.11 The hits in MIMOSA plane 4 used for the tracks in the AR2. . . . . . . . 65C.12 The hits in MIMOSA plane 5 used for the tracks in the AR2. . . . . . . . 65

D.1 Number of hits per bin used in the gain fit for the LGA31 (SPR). . . . . 66D.2 Number of hits per bin used in the gain fit for the LGA33 (SPR). . . . . 67D.3 Number of hits per bin used in the gain fit for the LGA34 (SPR). . . . . 67D.4 Number of hits per bin used in the gain fit for the IP3N2 (AR1). . . . . . 68D.5 Number of hits per bin used in the gain fit for the W7 HG22 (AR1). . . . 68D.6 Number of hits per bin used in gain fit for the W11 HG11 (AR1). . . . . 69D.7 Number of hits per bin used in the gain fit for the W7 HG22 (AR2). . . . 69D.8 Number of hits per bin used for the hit efficiency of the SPR. . . . . . . . 70D.9 Number of hits per bin used for the hit efficiency of the AR1. . . . . . . . 70D.10 Number of hits per bin used for the hit efficiency of the AR2. . . . . . . . 71

48

D.11 Number of hits per bin used for the time resolution of the LGA31. . . . . 71D.12 Number of hits per bin used for the time resolution of the LGA33. . . . . 72D.13 Number of hits per bin used for the time resolution of the LGA34. . . . . 72

49

List of Tables

3.1 The specifics of the sensors tested corresponding to the run. . . . . . . . 21

4.1 The determined thresholds for the different LGADs. . . . . . . . . . . . . 25

5.1 The alignment constants for the MIMOSA of the different runs. . . . . . 28

50

Appendix A

MIMOSA plots for the SPR

The plots shown in Figure A.1 - Figure A.6, are the unbiased residuals between the tracksand the hit registered in the MIMOSA plane considered. The unbiased residual meansthat a track from a fit of all the other 5 planes was made and compared to the positionof the hit in the MIMOSA layer.

Figure A.1: The residuals from the hits to the tracks in MIMOSA plane 0 for SPR


The figures Figure A.7 - Figure A.12 show the distribution of the hits in the differentMIMOSA layers which are used for the tracks.

51





52

Figure A.7: The hits in MIMOSA plane 0 used for the tracks in the SPR.


53



54



55

Appendix B

MIMOSA plots for the AR1

The plots shown in Figure B.1 - Figure B.6, are the unbiased residuals between the tracksand the hit registered in the MIMOSA plane considered. The unbiased residual meansthat a track from a fit of all the other 5 planes was made and compared to the positionof the hit in the MIMOSA layer.

Figure B.1: The residuals of the tracks in MIMOSA plane 0


The figures Figure B.7 - Figure B.12 show the distribution of the hits in the differentMIMOSA layers which are used for the tracks.

56





57

Figure B.7: The hits in MIMOSA plane 0 used for the tracks in the AR1.


58



59



60

Appendix C

MIMOSA plots for the AR2

The plots shown in Figure C.1 - Figure C.6, are the unbiased residuals between the tracksand the hit registered in the MIMOSA plane considered. The unbiased residual meansthat a track from a fit of all the other 5 planes was made and compared to the positionof the hit in the MIMOSA layer.

Figure C.1: The residuals of the tracks in MIMOSA plane 0


The figures Figure C.7 - Figure C.12 show the distribution of the hits in the differentMIMOSA layers which are used for the tracks.

61





62

Figure C.7: The hits in MIMOSA plane 0 used for the tracks in the AR2.


63



64



65

Appendix D

Support information for Analysis

The hits used per bin for the gain plots are shown in Figure D.1 - Figure D.7. Theseplots are made before the criterion of 5 hits per pixel is used, which cleans up the hitsoutside of the sensor area.

Figure D.1: Number of hits per bin used in the gain fit for the LGA31 (SPR).

All the positions of the tracks used for the efficiency plots are shown in Figure D.8,Figure D.9, and Figure D.10.

The number of hits for each bin in the time resolution plots of the SPR are shown inFigure D.11, Figure D.12 and Figure D.13.

66



67

Figure D.4: Number of hits per bin used in the gain fit for the IP3N2 (AR1).

Figure D.5: Number of hits per bin used in the gain fit for the W7 HG22 (AR1).

68

Figure D.6: Number of hits per bin used in gain fit for the W11 HG11 (AR1).

Figure D.7: Number of hits per bin used in the gain fit for the W7 HG22 (AR2).

69

Figure D.8: Number of hits per bin used for the hit efficiency of the SPR.

Figure D.9: Number of hits per bin used for the hit efficiency of the AR1.

70

Figure D.10: Number of hits per bin used for the hit efficiency of the AR2.

Figure D.11: Number of hits per bin used for the time resolution of the LGA31.

71



72

Bibliography

[1] ATLAS public plots. https://twiki.cern.ch/twiki/bin/view/AtlasPublic/

LArHGTDPublicPlots. Accessed: 2017-05-26.

[2] Charge collection. http://www.hephy.at/user/friedl/diss/html/node11.html.Accessed: 2017-05-22, Updated: 2001-07-14.

[3] EUDET-type beam telescopes. https://telescopes.desy.de. Accessed: 2017-05-23.

[4] RD50 - radiation hard semiconductor devices for very high luminosity colliders.Accessed: 2017-05-23.

[5] G. Aad et al. The ATLAS Experiment at the CERN Large Hadron Collider. JINST,3:S08003, 2008.

[6] N. Cartiglia et al. Performance of Ultra-Fast Silicon Detectors. JINST, 9:C02001,2014.

[7] L. Evans and P. Bryant. Lhc machine. Journal of Instrumentation, 3(08):S08001,2008.

[8] A. Hoecker. Physics at the LHC Run-2 and Beyond. In 2016 European School ofHigh-Energy Physics (ESHEP 2016) Skeikampen, Norway, June 15-28, 2016, 2016.

[9] H. Jansen et al. Performance of the eudet-type beam telescopes. EPJ Techniquesand Instrumentation, 3(1):7, 2016.

[10] T. McCarthy. Upgrade of the ATLAS Liquid Argon Calorimeters for the High-Luminosity LHC. In IEEE Nuclear Science Symposium and Medical Imaging Con-ference - 2016 Strasbourg, France, October 29-November 6, 2016, 2016.

[11] G. Pellegrini et al. Technology developments and first measurements of Low GainAvalanche Detectors (LGAD) for high energy physics applications. Nuclear Instru-ments and Methods in Physics Research A, 765:12–16, November 2014.

73

https://twiki.cern.ch/twiki/bin/view/AtlasPublic/LArHGTDPublicPlots

https://twiki.cern.ch/twiki/bin/view/AtlasPublic/LArHGTDPublicPlots

http://www.hephy.at/user/friedl/diss/html/node11.html

https://telescopes.desy.de

Analysis of Test Beam Data for Sensors in the High ...1117693/FULLTEXT01.pdf · A test beam study...

Documents

Transcript of Analysis of Test Beam Data for Sensors in the High ...1117693/FULLTEXT01.pdf · A test beam study...