Kirti Final Complete PHD Thesis

LARGE PT PARTICLE PRODUCTION IN p-p
COLLISIONS AT LHC ENERGIES
THESIS SUBMITTED TO THE UNIVERSITY OF DELHI

FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
KIRTI RANJAN
SUPERVISOR: PROF. R. K. SHIVPURI
DEPARTMENT OF PHYSICS AND ASTROPHYSICS

UNIVERSITY OF DELHI
DELHI 110007
INDIA
2002
The scientist does not study nature because it is
useful; he studies it because he delights in it,
and he delights in it because it is beautiful. If
nature were not beautiful, it would not be
worth knowing, and if nature were not worth
knowing, life would not be worth living.
Henri Poincaré
Dedicated to…
my loving Parents and Sisters,
Alka and Shalini
Contents
Acknowledgements i
List of Publications iv
Chapters
Chapter 1: Motivation and Framework: The CMS Experiment at LHC 1
1.1. Introduction 1
1.2. CERN 2
1.3. The Large Hadron Collider (LHC) Machine 4
1.3.1. Some interesting Features 4
1.3.2. LHC Experiments 6
1.4. The Compact Muon Solenoid (CMS) Experiment 7
1.4.1. CMS Detector Overview 7
1.4.1.1. Magnet 8
1.4.1.2. Central Tracking System 10
1.4.1.3. Muon System 10
1.4.1.4. Hadron Calorimeter 10
1.4.1.5. Electromagnetic Calorimeter 11
1.4.2. The Physics Potential of CMS / LHC 12
1.5. Preshower Detector: raison d’être 16
1.5.1. Need of Preshower: Physics Goal 16
1.5.2. Design Criterion Imposed by the Physics Goal 18
1.5.3. Expected Response to Single and Double Photons 20
1.6. Motivation for the Present Work 21
1.7. Thesis Organization 23
Chapter 2: Physics of Silicon Detectors – A Résumé 26

2.1. Silicon Detector: An Introduction 26
2.2. Semiconductors: General properties 28
2.3. Reverse Biased p - n Junction 30
2.3.1. Abrupt p - n junction 30
2.3.2. Expressions of Some Useful Physical Quantities for
Abrupt p-n Junction 32
2.3.3. Reverse Leakage Current 34
2.3.4. Breakdown Voltage 38
2.4. Principle of Silicon Detector Operation 45
2.5. Silicon Microstrip Detector for CMS Preshower 46
2.5.1. Silicon Microstrip Detector 47
2.5.2. Wafer Parameters 47
2.5.3. Sensor Design / Geometry 47
2.5.4. Fabrication 48
2.5.5. Acceptance Criterion 50
Chapter 3: Semiconductor Device Simulation 51

3.1. Why Simulation ? 51
3.2. TMA MEDICI - A Device Simulation Program 53
3.2.1 Physical Description 53
3.2.1.1. Drift-Diffusion Model 53
3.2.1.2. Physical Models 54
3.2.1.3. Boundary Conditions 54
3.2.2 Numerical Methods 55
3.2.2.1. Discretization 55
3.2.2.2. Nonlinear System Solutions 56
3.2.2.3. Simulation Grid 56
3.2.3 MEDICI Program Description 57
3.3. Validation – An Absolute Requirement 60
Chapter 4: Effect of Metal-Overhang on Silicon Strip Detectors 61

4.1. High-Voltage Si Detectors: Techniques to
Improve Breakdown Voltage 62
4.2. Modulation of Electric Field by Extended
Electrode: Origin of Metal-Overhang Technique 63
4.3. Device Structure Used in Simulation 66
4.4. Influence of Various Parameters on Silicon Strip
Detector Equipped with Metal-Overhang 68
4.4.1. Comparison between the Structures without
and with metal-overhang 68
4.4.2. Effect of Field-Oxide Thickness 73
4.4.3. Effect of Junction Depth 78
4.4.4. Effect of the Width of Metal-Overhang 82
4.4.5. Effect of Substrate Parameters: Device-Depth
and Substrate Doping Concentration 85
4.4.6. Effect of Surface Charges 86
4.5. Comparison with Experimental Work 94
4.6. Conclusions 95
Chapter 5: Effect of Passivation on Breakdown Performance of

Metal-Overhang Equipped Si Sensors 97
5.1. Passivation in Si Detectors 98
5.2. Radiation Damage in Si Sensors 100
5.3. Device Structure & Simulation Technique 101
5.3.1. Modeling of Radiation Damage 102
5.4. Comparison between Semi-Insulator vs. Dielectric Passivation 104
5.4.1. Effect of Field-Oxide Thickness and Junction Depth 104
5.4.2. Effect of Metal-Overhang Width 112
5.4.3. Effect of Passivation Layer Thickness 114
5.4.5. Effect of Device Depth and Substrate Doping Concentration 117
5.4.6. Effect of Bulk Damage on Full Depletion and
Breakdown Voltage 119
5.6. Static Measurements on Irradiated Si Sensors 128
5.6.1. Irradiation Facility 128
5.6.2. Measurement Set-ups 128
5.6.3. Measurement Results on Irradiated Sensors 130
5.7. Conclusions 133
Chapter 6: Comparison of Junction Termination Techniques for High-Voltage

Si Sensors: Metal-Overhang vs. Field Limiting Ring 135
6.1. FLR Structure 135
6.2. Device Model 137
6.3. Comparison of FLR and MO Structures 138
6.3.1. Effect of Guard Ring Spacing (for FLR structure) and
Field-Oxide Thickness (for MO structure) 138
6.3.2. Effect of Junction Depth 140
6.3.3. Effect of Relative Permittivity of Passivant 143
6.3.5. Effect of the Width of Guard Ring (GW; for FLR
structures) and Metal-Extension (WMO; for MO structures) 147
Chapter 7: Large Transverse Momentum (pT) Direct Photon Production

at LHC 151
7.1.An Introduction to Standard Model and QCD 151
7.2.QCD Phenomenology of High pT Inclusive processes 153
7.3.Direct Photons in the QCD framework 155
7.3.1. Contributions to Direct Photons 155
7.3.2. Backgrounds to Direct Photons 157
7.3.3. Motivation 159
7.4. Theoretical Formalism 161
7.4.1. Scale Sensitivity 161
7.4.2. Pseudorapidity Dependence 162
7.4.3. Isolation Technique 163
7.4.4. KT Smearing 164
7.5. Monte Carlo Simulation 165
7.6.Direct Photon Production at Tevatron: Comparison of Data
With Theory at s =1.8 TeV & s =630 GeV 165
7.7. Expectations for Direct Photons at LHC 168
7.7.1. Leading Order (LO) Cross section 168
7.7.2. Next-to-Leading Order Cross Section 169
7.7.3. K-factor 170
7.7.4. Scale Dependence of Inclusive Cross Sections 171
7.7.5. Sensitivity to Gluon Distributions 171
7.7.5.1. The pT Spectrum 171
7.7.5.2. The η Spectrum 172
7.7.6. Pseudorapidity Dependence 174
7.7.7. Cone Size Dependence 175
Bibliography 177
Acknowledgments
Nobody lives in complete isolation, and we accomplish nothing without the input
and encouragement of those around us. Although I have asserted that the work presented
in this thesis is entirely my own, it is a delight to acknowledge here the important
contributions that many people have made over the past few years which have allowed
me to get where I am today.
I would like to express my largest gratitude to my supervisor, Prof. R. K. Shivpuri
for his constant support, continuous encouragement and guidance throughout this
investigation, and the outstanding working conditions at the Laboratory. He is the best
advisor and teacher I could have wished for, actively involved in the work of all his
students, and clearly always has their best interest in mind. Time after time, his vast
knowledge, deep insights, tremendous experience and easy grasp of physics at its most
fundamental level helped me in the struggle for my own understanding. It was both a
privilege and honour to work with him. I have learned various things, from him, such as
the way of thinking, and the way of proceeding in research, and so on. I am indebted to
him for connecting me with CMS experiment, thereby opening avenues for me to gain
experience in doing physics analysis as well as hardware work. I am also obliged to the
Head of the Department of Physics, Prof. K. C. Tripathi for providing the necessary
facilities in the Department.
In the Lab, I was surrounded by knowledgeable and friendly people who helped
me daily. They helped to create an informal and congenial atmosphere that, in my
opinion, makes thesis writing a much less onerous task than it otherwise might be. First
and foremost, I gladly acknowledge my debt to my colleague Ashutosh Bhardwaj.
Without his constant friendship, encouragement and advice, I would never have reached
here. I can never forget the long teatime discussions with him and also the wonderful
time he has provided me as a friend and colleague during our visits to various research
institutes and in particular CERN, Geneva. Thanks a lot, Ashutosh! I would also like to
thank Namrata for providing invaluable comments, ideas, and general assistance. Thanks
are due to Sudeep Chatterji and Ajay K. Srivastava, with whom I had many productive
scientific and general discussions. I would also express my gratitude to other members of
i
my group, Manoj Kumar, Shailesh Kumar, Bani Mitra and Bipul Bhuyan. My junior
colleague Ashish Kumar deserves a special mention as he helped me a lot with the
PYTHIA software and direct photon physics.
I would like to express my gratitude to Dr. Philippe Bloch and Dr. Anna Peisert
for providing wonderful working conditions at CERN, and also for giving me freedom to
work anytime at the Detector Lab, CERN. I owe a lot to them regarding the installation
and working of I-V and C-V systems. I also thank Dr. Dave Barney who helped me with
my online presentation at CERN and also spared his precious time to show me the
wonderful CMS site and snow-covered Zura mountains. Special thanks are due to Apollo
Go who introduced me to LabVIEW programming language and CRISTAL Database. I
also wish to thank Dr. Gagan Mohanti, Dr. Swagato Banerjee and Dr. Pratibha Vikas for
giving me a good company and providing homely environment at CERN.
I also express my sincere thanks to our collaborators from B.A.R.C., Dr. S. K.
Kataria, Mr. M. D. Ghodgaonkar, Mr. V. B. Chandratre, Dr. Anita Topkar, Mr. M. Y.
Dixit and Mr. Vijay Mishra for their assistance, cooperation and valuable annotations at
various India-CMS meetings. Thanks are also due to a dear friend, Vishal D. Srivastava,
BARC for many useful discussions carried on long-distance via e-mail. I am very
thankful to him as he alongwith Mr. V. B. Chandratre introduced me to the nitty-gritty of
the Process and Device Simulation. I also extend my gratitude to our other collaborators
from T.I.F.R., Prof. S. N. Ganguli, Prof. Atul Gurtu and Prof. Sunanda Banerjee and from
Panjab University, Prof. J. M. Kohli, Prof. Suman Beri, Dr. Manjeet Kaur and Dr. J. B.
Singh for their encouragement in accomplishing this work. I extend my appreciation to
Dr. O. P. wadhwan and Dr. G. S. Virdi, CEERI, Pilani and Mr. Subhash Chandran, Mr.
Prabhakar Rao and Mr. Shanker Narayan, BEL, Bangalore for providing me with detailed
information about the detector fabrication
I wish to thank Mr. P. C. Gupta for his support and general assistance. I also
appreciate the continuous assistance I received from Mr. Rajendra Mishra and Mr.
Mohammad Yunus.
This work was financially supported by the Council for Scientific and Industrial
Research, India through the Junior and the Senior Research Fellowships. I gratefully
acknowledge their generosity.
ii
In addition to the people in University, I am lucky enough to have the support of
many good friends. Life would not have been the same without them. There are too many
people to mention individually, but some names stand out. I wish to thank my friends in
high school (Hitesh Dighe, Rhitu Parn, Sandeep Tewatia and Brijesh Rawat), my friends
as an undergraduate (Devendar Nahar and Sai Bhushan), and my friend as a graduate
student (Abhinav Kranti), for helping me get through the difficult times, and for all the
emotional support, camaraderie, entertainment, and caring they provided.
It would, of course, be completely amiss for me to end my acknowledgements
without recognizing the immense contribution that my family has made to my work.
Their love and support has been a major stabilizing force over these past years. Their
unquestioning faith in me and my abilities has helped to make all this possible and for
that, and everything else, I dedicate this thesis to them.
KIRTI RANJAN
iii
List Of Publications
1. “Analysis and Optimal Design of Si Microstrip Detector with Overhanging Metal
Electrode”, K. Ranjan et al., Semicond. Sci. Technol. 16, 635 (2001).
2. “Analysis and Comparison of the breakdown performance of Semi-insulator and

dielectric passivated Si Strip detectors”, K. Ranjan et al., CMS-NOTE
(CERN,Geneva), 2002/014, also accepted for publication in Nuclear
Instruments and Methods in Physical Research A.
3. “Performance characteristics of semi-insulator and dielectric passivated Si strip

detectors”, K. Ranjan et al., Physica Status Solidi (a) 191(2), 658 (2002).
4. “High-voltage planar Si detector for high-energy physics experiment: comparison

between metal-overhang and field-limiting ring techniques”, K. Ranjan et al.,
communicated to Journal of Applied Physics.
Paper presented
1. “Influence of electrode geometry on electric field distribution within silicon

microstrip detector”, K. Ranjan et al., Workshop on CMS at LHC held on 11-15
Dec., 2000 at TIFR, Mumbai, India.
2. “Comparison of the passivants on the Breakdown Performance of Si strip

detectors”, K. Ranjan et al., National Seminar on Physics of Materials for
Electronic and Optoelectronic Devices held on 25-27 Feb. 2002 at J. N. Vyas
University, Jodhpur, Rajasthan, India.
iv
Chapter 1
----------------------------------------------------------------
Motivation and Framework: The CMS
Experiment at LHC
----------------------------------------------------------------
1.1 Introduction
From the time humans began to ask questions about themselves and their world,
they have wondered what the world is made of and how it behaves. Over and over, in
different ages and by different methods, people have tried to answer the questions, “What
is the smallest possible piece of matter? What are the fundamental forces of nature?”.
Today, the fundamental science of high-energy, or particle physics continues to pursue
answers to these most ancient, and most modern questions.
To a great degree, the progress of particle physics has followed from progress in
accelerator science and instrumentation. There is no substitute for experiment, and
experimental techniques require inventions in both hardware and software and continuous
innovation in analysis. Currently, the large experimental particle physics programs (CDF
and D0) have undergone major upgrades and are taking data right now at the Tevatron,
Fermi National Accelerator Lab, Chicago, USA. The confirmation with very high
precision of the Standard Electroweak and Quantum Chromodynamics (QCD) models by
these experiments has prompted new physics questions which can only be answered by
the construction of new specialized accelerators, and associated detectors. One such
accelerator, the Large Hadron Collider (LHC) is now being constructed at CERN,
Geneva and will become operational from the year 2007. This proton-proton (pp) collider
will have a peak luminosity of L = 1034 cm-2s-1 and a collision energy of s =14 TeV,
about seven times larger than at the Tevatron collider at Fermilab. The primary goal of
the LHC machine will be the search for the Higgs boson, which results from spontaneous
1
symmetry breaking mechanism in the Standard Model. The confirmation of the existence
or non-existence of the Higgs boson will be the greatest achievement of the LHC. In
addition, some other interesting topics, like search for various Minimum Supersymmetric
Model Higgs bosons, gluino and squark, new massive vector bosons, CP- violation
measurements in B sector, quark gluon plasma etc. will also be addressed.
To exploit the physics capability of the LHC, four large detectors are being
constructed. These detectors are significantly more complex than their present generation
counterparts because of physics and operational requirements. The Compact Muon
Solenoid (CMS) experiment is a general-purpose detector designed to exploit the physics
of pp collisions over the full range of luminosities expected at the LHC. One of the CMS
design objectives has been to construct a very high performance Electromagnetic
Calorimeter (ECAL), which will play a significant role in the detection of Standard
Model Higgs boson in low mass range (mH < 140 GeV/c2) through the two-photon decay
mode. A Preshower silicon strip detector has been included in the endcaps of ECAL to
reduce the background to this Higgs channel by facilitating γ−π0 separation. Since this
exciting physics programme at CMS will be carried out in an extremely difficult
experimental environment, hence it imposes very challenging requirements on the silicon
detector specifications.
This chapter provides an overview of the CERN LHC machine and CMS
detector. Since most of the work reported in this thesis is on the detailed analysis of the
breakdown performance of silicon strip sensors to be used in Preshower at CMS, hence a
more detailed discussion on the Preshower detector is presented. Finally, the motivation
behind the work and the thesis organization is described.
1.2 CERN
CERN (Conseil Europeen pour la Recherche Nucleaire) is the European

Laboratory for Particle Physics situated in Geneva (at the border of Switzerland and
France), and is the world’s largest particle research center. The creation of this laboratory
was recommended at the UNESCO meeting in Florence in 1950 as the only way forward
for front-line particle physics research in Europe.
2
Accelerators at CERN: CERN's accelerator complex is the most versatile in the
world and represents a considerable investment. It includes particle accelerators and
colliders, can handle beams of electrons, positrons, protons, antiprotons, and heavy ions.
Each type of particle is produced in a different way, but then passes through a similar
succession of acceleration stages, moving from one machine to another. The first steps
are usually provided by linear accelerators, followed by larger circular machines. CERN
has 10 accelerators altogether, so far the biggest being the Large Electron Positron
collider (LEP). CERN's first operating accelerator, the Synchro-Cyclotron, was built in
1954, in parallel with the Proton Synchrotron (PS).
Fig.1.1: Accelerators at CERN.
The PS is today the backbone of CERN's particle beam factory, feeding other
accelerators with different types of particles. The 1970s saw the construction of the Super
Proton Synchrotron (SPS), at which Nobel-prize winning work was done in the 1980s.
The SPS continues to provide beams for experiments and is also the final link in the chain
of accelerators providing beams for the 27 km LEP machine. CERN's next big machine,
due to start operating in 2007, is the Large Hadron Collider (LHC). A more clear view of
the accelerators can be seen in Fig.1.1.
3
1.3 The Large Hadron Collider (LHC) Machine
The LHC machine being constructed at CERN in the LEP tunnel would be
operational in 2007. The LHC will collide counter-rotating beams of protons with total
center of mass energy ( s ) of 14 TeV. In order to maintain an equally effective physics
programme at a higher energy E, the luminosity of a collider (a quantity proportional to
the number of collisions per second) should increase in proportion to E2, since the cross
section of the particle production decreases like 1/E2. Whereas in past and present
colliders the luminosity culminates around L = 1032 cm-2 s-1, in the LHC it will reach L =
1034 cm-2 s-1. It is expected that the LHC machine would reach one-tenth of the peak
luminosity during the first year of operation, and one-third & two-third of the peak
luminosity respectively in the following two years (referred to as low luminosity period)
[1.1]. LHC will operate at its full limit from the fourth year onward. This will be
achieved by filling each of the two rings with 2835 bunches of 1011 particles each. The
separation between the two bunches would be 25 ns (~ 7.5 m) with ~ 23 pp interactions
for each bunch crossing. The LHC energy and luminosity would be about an order of
magnitude higher than the present collider machines, which would allow for the search of
new massive particles produced with small cross-sections and probe the structure of
matter at extremely small distances.
1.3.1 Some Interesting Features

Several interesting features of the LHC machine, as shown in Fig.1.2, have
enormous consequences for the detector design, which are described below.
• The bunch spacing of 25 ns means that interactions of one bunch crossing occur
before all particles from interactions of a previous bunch crossing have traversed
the detector. In order to prevent the pile-up of interactions over several bunch
crossings, a fast detector signal response, small detector dead time, and extensive
signal pipelining prior to initial trigger decisions is required.
4
Fig.1.2: Schematic layout of the collisions at LHC.
• To achieve the design luminosity, the colliding proton bunches are short (~ 7.5 m)
and intense (~ 1011 protons per bunch). The total inelastic, non-diffractive cross-
section at the LHC energies is expected to be 80 mb, corresponding to an
interaction rate of 109 Hz. Fig.1.3 shows the cross-section for several Standard
Model processes as a function of s [1.2]. Typically, in case of the Higgs particle
with mass = 500 GeV, about 17 K events are expected per operating year (107 s) at
design LHC luminosity, compared to a total of 1.7 x 1016 events from inelastic
interactions. The LHC experiments must identify rare processes at this level.
• At the design luminosity, ~ 23 interactions occur in each bunch crossing. This
results in ~ 104 tracks in the detector each 100 ns, the typical duration of a pulse in
the detectors. The individual detector element must therefore be highly granular in
order to minimize the contribution of the pile-up in a given detector cell.
5
Fig.1.3: Energy dependence of some characteristic cross-sections at pp colliders.
• The high flux of particles from pp interactions places the detectors and associated
electronics in a high-radiation environment. Only radiation resistant detectors and
read-out electronics have to be used.
1.3.2 LHC Experiments

Among the four experiments (CMS [1.3], ATLAS [1.4], ALICE [1.5] and LHC-B
[1.6]) being constructed for operations at LHC (Fig.1.4), the design philosophy has been
determined by the physics goals of the experiments and the relevant available technology
able to meet the experimental requirements. In the case of the LHC, the LHC-B and
ALICE detectors are optimized to their specific roles – a respective study of the b-quark
physics sector, and the study of heavy ion collisions. The CMS and ATLAS are general-
6
purpose detectors, and differences in their design reflect choices of the collaborations
with technologies able to meet the design requirements.
Fig.1.4: Schematic layout of the LHC.
1.4 The Compact Muon Solenoid (CMS) Experiment
The main design goals of CMS detector are a highly performant muon system, the
best possible electromagnetic calorimeter, a high quality central tracking system and a
hermetic hadron calorimeter. The CMS detector is designed to measure the energy and
momentum of photons, electrons, muons, and other charged particles with high precision,
resulting in an excellent mass resolution for many new particles ranging from the Higgs
boson up to a possible heavy Z ′ in the multi-TeV mass range. Fig.1.5 shows the CMS
detector which has an overall length of 21.6 m, with a calorimeter coverage to a
pseudorapidity of η = 5 (θ ~ 0.80), a radius of 7.5 m, and a total weight of 12500 tonnes
[1.7].
1.4.1 CMS Detector Overview

CMS consists of a powerful inner tracking system based on fine-grained silicon
microstrip and pixel detectors, a scintillating crystal calorimeter followed by a sampling
hadron calorimeter made of plastic scintillator tiles inserted between copper absorber
7
plates, and a high-magnetic-field (4T) superconducting solenoid coupled with a
multilayer muon chamber (Fig.1.5). Figures 1.6(a) and 1.6(b) show the transverse and the
longitudinal view of the CMS detector respectively. The key elements of the CMS
detector (Fig.1.5) are described briefly in the subsequent sections.
1.4.1.1 Magnet
The choice of magnet system was the starting point for the CMS detector design.
The requirement for a compact design led to the choice of solenoidal magnet system of
length 13 m and inner diameter 5.9 m, which can generate a strong magnetic field of 4 T,
which guarantees a good momentum resolution for high momentum (~ 1 TeV) muons up
to rapidities of 2.5 without strong demands on the chamber space resolution. The
magnetic flux is returned through a 1.8 m thick saturated iron yoke instrumented with
four layers of muon chambers.
Fig.1.5: Three-dimensional layout of the CMS detector showing internal detectors.
8
Fig.1.6(a): Transverse view of the CMS detector.
Fig.1.6(b): Longitudinal view of the CMS detector.
9
1.4.1.2 Central Tracking System
The CMS tracking system is designed to reconstruct high-pT muons, isolated
electrons and hadrons with high momentum resolution and an efficiency better than 98%
in the range |η| < 2.5. It is also designed to allow the identification of tracks coming from
detached vertices. The momentum resolution required for isolated charged leptons in the
central rapidity region is ∆pT / pT ~ 0.1 pT (pT in TeV).
In the new “full silicon” design [1.8], the tracker is constituted by an innermost
region, instrumented with silicon pixel sensors and an external region, instrumented with
silicon strip sensors, that extends up to the region once covered by gas chambers [1.9].
1.4.1.3 Muon System

Muons are expected to provide clean signatures for a wide range of physics
processes. The task of the muon system is to identify muons and provide, in association
with the tracker, a precise measurement of their momentum. In addition, the system
provides fast information for triggering purposes - a challenging problem at the LHC. At
the LHC, the efficient detection of muons from Higgs bosons, W, Z and tt decays
requires coverage over large rapidity interval.
The muon detectors, placed behind the calorimeters and the coil, consist of four
muon stations interleaved with the iron return yoke plates. They are arranged in
concentric cylinders around the beam line in the barrel region, and in disks perpendicular
to the beam line in the endcaps. CMS will use three types of gaseous particle detectors
for muon identification: Drift Tubes (DT) in the central barrel region, Cathode Strip
Chambers (CSC) in the endcap region and Resistive Parallel Plate Chambers (RPC) in
both the barrel and endcaps.
1.4.1.4 Hadron Calorimeter

The Hadronic Calorimeter (HCAL) plays an essential role in the identification
and measurement of quarks, gluons, and neutrinos by measuring the energy and direction
of jets and of missing transverse energy flow in events. Missing energy forms a crucial
signature of new particles, like the supersymmetric partners of quarks and gluons. For
good missing energy resolution, a hermetic calorimetry coverage to |η| = 5 is required.
10
The HCAL will also aid in the identification of electrons, photons and muons in
conjunction with the tracker, electromagnetic calorimeter, and muon systems.
Barrel & Endcap: The hadron barrel (HB) and hadron endcap (HE) calorimeters
are sampling calorimeters with 50 mm thick copper absorber plates which are interleaved
with 4 mm thick plastic scintillator sheets. The HB is constructed of two half-barrels each
of 4.3 m length. The HE consists of two large structures, situated at each end of the barrel
detector and within the region of high magnetic field. Because the barrel HCAL inside
the coil is not sufficiently thick to contain all the energy of high-energy showers,
additional scintillation layers (HO) are placed just outside the magnet coil.
Forward: There are two hadronic forward (HF) calorimeters, one located at each
end of the CMS detector, which complete the HCAL coverage to |η| = 5. The HF is built
of steel absorber plates since steel suffers less activation under irradiation than copper.
Hadronic showers are sampled at various depths by radiation-resistant quartz fibers, of
selected lengths, which are inserted into the absorber plates.
1.4.1.5 Electromagnetic Calorimeter

The physics process that imposes the strictest performance requirements on the
Electromagnetic Calorimeter (ECAL) is the light mass Higgs decaying into two photons.
Thus the benchmark against which the performance of the ECAL is measured is the di-
photon mass resolution. CMS has chosen lead tungstate (PbWO4) crystals (over 80000)
which have high density, a small Molière radius and a short radiation length allowing for
a very compact calorimeter system. A high-resolution crystal calorimeter enhances the H
→ γγ discovery potential at the initially lower luminosities at the LHC. The crystal will
project tower geometry each 230 mm in length and 22 mm x 22 mm in cross-section. The
time-constant of scintillation light coming from crystals is only 10 ns. In CMS light will
be detected by Si-avalanche photodiodes which can provide gain of 50 even in high
magnetic field environment. In order to facilitate γ/π0 separation in the forward-backward
region ( 1.653 ≤ η ≤ 2.61 ) a Preshower silicon strip detector is included in the baseline
CMS design.
11
1.4.2 The Physics Potential of CMS / LHC
The CMS detector has been designed to provide answers to the most important
open questions in High-Energy Physics (HEP). LHC will address one of the key
questions in particle physics namely that of the origin of the Spontaneous Symmetry
Breaking (SSB) mechanism in the electroweak sector of the Standard Model. Amongst
others, some questions involve the following specific challenges [1.10]:
• Standard Model (SM) Higgs boson search at masses above the maximum reach of
LEP and Tevatron, order of 100 GeV – 1 TeV.
• Search for Minimal Supersymmetric Standard Model (MSSM) Higgs bosons up
to masses of 2.5 TeV. In this framework, five Higgs bosons are expected.
• Search for SUper SYmmetric (SUSY) partners of quarks and gluons – squark and
gluino up to masses of 2.5 TeV.
• Study of CP – violation in the B sector and time dependent mixing of b – mesons.
• Search for new heavy gauge bosons ( W ′, Z ′ ) up to masses 4.5 TeV.
• Detailed studies of production and decays of top quark.
• Search for composite structures of quarks and leptons.
• Search for quark – gluon plasma (QGP) in heavy ion collisions.
In the subsequent sections a brief discussion on the aforementioned topics would
be given.
Search for Standard Model Higgs Boson: In the framework of the Standard
Model particles acquire mass through their interaction with Higgs field. This implies the
existence of a new particle: the Higgs boson (H). The theory does not predict the mass of
the H, but it does predict its production rate and decay modes as a function of its mass.
With beam energies of the 102 GeV for the CERN LEPII accelerator, the LEP
experiments exclude mH < 108 GeV/c2 [1.11].
Depending on the Higgs mass, the detection of a SM Higgs involves several
different signatures (Fig.1.7).
12
Fig.1.7: Expected observability of SM Higgs boson and mass range in CMS.
The most promising channel for a SM Higgs with mass between the expected
LEPII limit (~ 108 GeV) and 140 GeV is the decay H → γγ , which has the cleanest
signal but suffers from the small branching ratio (1000 times smaller than most other
channels). This signal has to be detected above a large background from continuum γγ
events. The CMS detector has excellent electromagnetic calorimetry, which can measure
γ energy and direction to 1% accuracy.
In the mass range 130 < mH < 700 GeV the most promising decay channel is
H → ZZ / ZZ * → l + l − l + l − (4l ± , gold plated channel ) . The detection relies on the
excellent performance of the muon chambers, the tracker and the electromagnetic
calorimeter. The reconstruction of the l+l- makes it practically background free and has
very significant branching ratio. However, for Higgs mass between 130 GeV – 200 GeV
this channel suffers from very low branching ratio. Another important channel in this
range is H → WW → 2l ± 2ν , having highest branching ratio (six times larger than the
miss
4l ± channel). Also, it has a distinct signature, two high pT leptons and high ET .
For large Higgs mass, 500 GeV < mH < 1 TeV, the natural width is already quite
large and the mass resolution becomes less important. Therefore, in this case the decay
13
channels H → WW → l ±ν jj , H → ZZ → l + l −νν , and H → ZZ → l + l − jj are expected
to provide more favourable signals. In these channels the biggest background comes from
single W (or Z) production along with QCD jets. The CMS simulation studies show that
double forward jet tagging can control the background and extend the Higgs search up to
1 TeV [1.3]. Fig.1.8 shows the explorable mass range of Higgs at LHC.
Fig.1.8: Explorable mass range of Higgs at LHC.
SUSY Higgs and SUSY Particles Searches: Supersymmetry (SUSY) is

believed to be the most promising and elegant extension of the Standard Model. The
Minimal Super Symmetric Model (MSSM) requires the existence of two Higgs doublet,
resulting in five physical Higgs bosons: h 0 , H 0 , A 0 and H ± decaying through a variety
of decay modes to γ, e±, µ±, τ± and jets in final states. Higgs bosons masses and couplings
can be expressed in terms of two parameters, mA and tanβ. Within this model the lightest
neutral Higgs boson h0 has a mass smaller than 130 GeV. The masses of the other four
Higgs bosons are under lesser constraints.
At the LHC, the strongly interacting gluinos ( g ) and squark ( q ) dominate the
SUSY particle production. With increasing luminosity, the LHC will allow the search for
these particles in the TeV range.
Charge-Parity (CP) Violation: The LHC would be working as a B-factory,
producing 1012 – 1013 bb per year, which would open the way for precise measurements
of parameters characterizing the CP violation effects in the B-system. The CMS detector
14
having very good tracking, calorimetry and muon system, would be able to detect most
promising decay channels which have small branching ratios like, Bd → J /ψ K S , and
0 0
Bd → π +π − ; with J /ψ → µ + µ − and K S → π + π − . These channels would measure

0 0
two β and α angles of the unitary triangle. The first reaction has a relatively high
branching ratio and clear signature. It is easier to trigger on multi-muon with low pT or a
single muon. With an integrated luminosity of 104 pb-1 in CMS, an accuracy in
measurements of sin2α = 0.057 and sin2β = 0.05 can be achieved. Both LHC machine
and CMS detector are very favourable to study this aspect of physics. Furthermore by
observing the time development of BS − BS oscillations, the mixing parameter χ S can
0 0
be measured for values up to 20 [1.3].

Quark-Gluon Plasma Search: At the LHC, collisions of high-energy Pb ions
can be studied at center of mass energy of 6.3 TeV per nucleon pair and a luminosity of
1.8 x 1027 cm-2s-1 [1.1]. The heavy ion beams at LHC will provide collision energy
densities well above the threshold for the formation of quark gluon plasma (QGP). In this
new state of matter all heavy quark bound states, except for Υ , are suppressed by color
screening [1.3]. The CMS detector has got one of the best muon system. So measurement
of µµ pairs rate coming from Υ family can be made to examine the suppression of Υ ′
and Υ ′′ relative to Υ in different heavy ion collisions and relative to pp collisions [1.12].
The CMS detector will be used to detect low momentum muons produced in the heavy
ion collisions and reconstruction of Υ , Υ ′ and Υ ′′ mesons. Another probe of QGP
formation is the jet production, where jet quenching is expected [1.13]. The CMS
calorimetry also allows this to be tested.
Top Quark Physics: The top quark has been identified at Tevatron and a few
hundred events have been produced. The mass has been measured with an accuracy of ~
± 5 GeV/c2, and the top pair production cross-section has been measured to be ~ 6 pb. At
the LHC, the pair production cross-section will be largely from the process ( gg → tt )
rather than from ( qq → tt ) and will be σ tt ~ 800 pb. Assuming an integrated luminosity
of only 1040 cm-2year-1, a total of ~ 8 x 109 pairs will be produced. This will allow a
measurement of the top mass with a precision of ± 1 - 2 GeV/c2, limits on new physics
such as resonant tt production, and limits or even the discovery of rare decays.
15
Heavy Flavour Production: Heavy flavour production is potentially a vast
domain of physics that can be studied at low luminosity. The cross-sections are very large
at the LHC, typically ~ 1mb for beauty and ~ 1 nb for top.
1.5 Preshower Detector: raison d’être
The CMS ECAL consists of one crystal barrel (EB), two crystal endcaps (EE),
and two Preshower endcaps (SE). The EE and SE form a combined system for the ECAL
endcaps. The Preshower consists of two self-contained disc-shaped structures situated
between the tracker and the EE. These two structures are known as SE+ and SE-, one for
each endcap of CMS. The Preshower detector coverage is the rapidity interval
1.653 ≤ η ≤ 2.61 as shown in Fig.1.9 [1.14].
Fig.1.9: Longitudinal view showing the Preshower coverage.
1.5.1 Need of Preshower: Physics Goal

One of the principal objectives of CMS is the discovery of the postulated Higgs
boson – the “missing link” in the Standard Model of particle physics. Current theoretical
predictions, and measurements from LEP, point to a relatively low mass Higgs; around
130 GeV/c2. If this is the case, the most promising channel will be via its decay into two
photons ( H → γγ ), even though this channel has a tiny probability of occurrence (< 1%).
The two photons from the decay will have rather high transverse momenta (pT), due to
16
the mass of the Higgs and, in general, will be isolated from each other and from other
particles. Fig.1.10 shows a Feynman diagram depicting the decay of Higgs boson into
two photons. However, high pT photons can also originate from other sources, forming
background to H → γγ , shown in Fig.1.11.
Fig.1.10: The two-photon decay of the standard model Higgs boson. It is important to
note that on right hand side of the diagram there are two photons traveling in different
directions.
Fig.1.11: Background to the H → γ γ channel at LHC. The relevant observation is that

in each case there are two single-photon-like objects on the right hand side (the
“squiggles”) which could be interpreted as originating from the decay of Higgs boson.
Two types of background are irreducible – real isolated pairs of photons are
produced at the interaction point, due to either quark annihilation or gluon-gluon fusion.
It is impossible, on the basis of the detection of two photons alone, to distinguish between
this type of event and a H → γγ . The remaining two types of background are reducible.
The first of these results in one photon close in space to other particles: requiring the
photon to be greater than a given distance from any neighbouring particles can reduce
this background. The second of the reducible backgrounds is due to neutral pions (π0s) in
jets: the π0s decay at the interaction point to two closely spaced photons; these two
photons may be so close together that a coarse-grain calorimeter may not be able to
resolve them. A typical π0 faking a Higgs photon may have a pT of 60 GeV/c. In the
barrel of CMS the typical separation of the two photons from the π0 decay will be around
1 cm, whereas in the endcaps it is just a few mm. In the barrel the crystal calorimeter is
sufficiently granular to be able to distinguish energy deposits from single photons and
17
closely-spaced double photons. However, this is not the case for the endcaps, where the
photon separation is smaller and the granularity of the crystal is lower, and hence the
need of preshower detector. Thus, the principal purpose of the CMS Preshower detector
is to reject neutral pions in the endcaps.
1.5.2 Design Criterion Imposed by the Physics Goal

The main requirement of the Preshower detector is the incorporation of fine-grain
sensors responsive to photons. CMS has chosen to construct the Preshower as a sampling
calorimeter: lead absorbers initiate electromagnetic showers from incoming photons and
silicon strip sensors placed immediately afterwards measure the energy deposited due to
charged particles within the showers. The transverse shower shape seen by the silicon
sensors differ for incident single photons and closely spaced double photons.
In the Preshower, the absorber should be thick enough to initiate photon showers,
however, a too thick absorber can degrade the energy resolution of EE. The optimum
thickness of absorber was found to be around 3 radiation lengths (3X0). Thus, the initial
design of the CMS Preshower comprised a single 3 X0 absorber followed by two
orthogonal layers of silicon sensors [1.3]. However, the performance of this setup, in
terms of π0 rejection, was found to be less than expected [1.15]. To improve the
performance, silicon sensors are required to be placed as close as possible to the absorber
[1.15]. A repercussion of this is that it is necessary to have two separate absorber layers,
each of which is followed by a plane of silicon sensors. Indeed this is the approach which
has been adopted for the CMS Preshower [1.14]. A transverse view of the Preshower is
shown in Fig.1.12. It is formed with two orthogonal planes of silicon detectors, each
plane is preceded by a thin absorber of 2 X0 and 1 X0 respectively.
18
Fig.1.12: The final schematic structure of the Preshower.
Each Preshower disk is segmented longitudinally (along the beam direction) into
passive and active parts. Passive parts include aluminium honeycomb “windows” filled
with paraffin wax, cooling planes and “absorber” planes (used to initiate and develop
electromagnetic showers). The total thickness of the Preshower is 20 cm.
The active part of the Preshower comprises of two orthogonal planes of silicon
strip sensors, built from a large number of identical micromodules. Each micromodule
comprises an aluminium tile that allows detector overlap in one direction (along the
strips), onto which a thin ceramic support is glued. A 6.3 x 6.3 cm2, 320 ± 20 µm thick,
silicon sensor, divided into 32 strips at 1.9 mm pitch, is then glued to the ceramic. The
hybrid containing the analogue front-end electronics (PACE) is also glued to the ceramic
and bonded to the sensor (Figures 1.13(a) and 1.13(b)). The micromodules are assembled
on ladders containing two adjacent columns of detectors (Fig.1.13(c)), which are attached
to the absorbers to form an X-Y grid of detectors. There are a total of 4288 micromodules
in the complete Preshower.
19
Fig.1.13: (a) A silicon detector and a single micromodule used for preshower, (b) real
micromodule used in beam tests, and (c) Number of micromodules assembled on a base-
plate. These modules together with the motherboard (containing digital electronics etc.)
constitute a “ladder”.

1.5.3 Expected Response to Single and Double Photons
Detailed Monte-Carlo simulations have been performed in order to examine the
performance of the Preshower in terms of π0 rejection and energy resolution. These
simulations have enabled to tune the Preshower design in both coarse and subtle ways.
Figures 1.14(a) and 1.14(b) show two simulated events in the SE + EE due to a single
photon and a double photon (from π0 decay) respectively.
Fig.1.14: (a) a single 30 GeV pT photon (dotted line) incident on the SE and EE, and (b)
two closely spaced photons (in this case around 1cm – event chosen for clarity) from the
decay of a 30 GeV pT π˚ incident on the SE and EE. Note that for electromagnetic shower
only charged particles are shown.
20
The histograms associated with each figure (dark area) show the approximate
pulse height seen in a group of silicon strips for the second plane in the SE due to the
passage of particles produced in the electromagnetic showers.
1.6 Motivation for the Present Work
Silicon microstrip sensors (along with the silicon pixel sensors) are the most
precise electronic tracking detectors for charged particles in the present generation of
high-energy physics experiments. However, the first and the most important issue, which
has to be considered in the unprecedented radiation environment at LHC, is the long-term
operation of the silicon detectors. The radiation damage, among other things, results in
the change of the effective doping concentration (Neff). It is found that the n-type
detectors become progressively less n-type with increasing fluence until they invert to
become effectively p-type (type-inversion). Beyond inversion the device continues to
become more p-type under further irradiation, apparently without limit. The depletion
voltage required to operate a silicon detector is directly proportional to Neff, hence at
higher fluences Neff can be such that the required operating voltage exceeds the
breakdown voltage of the device and efficient operation is no longer possible. For
Preshower, the integrated fluence of charged hadrons and neutrons above 100 keV in the
center of the Preshower detector will reach ~ 2 x 1014 particles/cm2 after 10 years of
operation with 85% of them being neutrons and the operational voltage of the silicon
detectors is expected to be about 400 V for the highest fluence [1.16].
This important issue of bulk damage in silicon sensors can be addressed in two
ways. The first one is trying to produce detectors on silicon substrates which are less
sensitive to the radiation damage. For this purpose collaboration between groups working
in this field, namely the CERN RD-48 (ROSE) collaboration is carrying out studies on
the use of oxygenated silicon as starting material [1.17]. However, its benefit is found to
be limited to charged hadron environment.
The other solution is to increase the detector bias voltage progressively so that full
depletion can be eventually attained anyway; thus potentially leading to the occurrence of
the early micro-discharges and avalanche breakdown. Hence, one of the main aims in the
silicon detector development is to solve this problem by fabricating high-voltage planar
21
junctions with careful design strategies. In particular, the adoption of Metal-Overhang
(MO) technique [1.18] has been suggested as an effective method to reduce breakdown
risks. The principal goal of this thesis is to analyze the influence of various physical and
geometrical parameters on the breakdown performance of silicon strip detectors equipped
with metal-overhang. Such a complete analysis would be of great importance for
optimization purposes. This analysis is performed using two-dimensional device
simulation program, TMA-MEDICI [1.19]. The adoption of device simulation programs
allows for the prediction of the characteristics of silicon detectors which helps in
evaluating the device sensitivity to various design parameters, thus aid in optimizing the
final design.
It would be worth mentioning here that the metal-overhang technique is actually
borrowed from the power device technology (where it is commonly referred to as “field
plate” technique), but due to the completely different process parameters and the
radiation environment, the optimization performed in the present work is an attempt for
the first time.
In recent years, high-energy physicists have arrived at a picture of the
microscopic physical universe, called "The Standard Model", which unifies the nuclear,
electromagnetic, and weak forces and enumerates the fundamental building blocks of the
universe. One particular ingredient of the Standard Model that requires further
quantitative investigation is Quantum Chromodynamics (QCD), which has emerged as a
viable theory of strong interactions over the last two decades. QCD has been a successful
theory in describing the interactions between the fundamental building blocks inside
hadrons–quarks and gluons. In addition, it is in good agreement with experimental data
collected both at fixed target and colliding beam experiments. However, the features of
QCD, although qualitatively verified, are far from being completely understood. It is then
crucial to investigate the properties of QCD at hadron colliders to probe the inner
structure of the hadrons from the standpoint of perturbative QCD (pQCD) techniques and
the parton model of strongly interacting particles.
Study of direct photon production in high-energy hadronic collisions provides a
clean tool for testing the essential validity of perturbative QCD predictions as well as for
constraining the gluon distribution of nucleons. The motivation to work on direct photons
at LHC energies stems from the fact that global QCD analysis of direct-photon
22
production processes show good agreement between data and QCD at high transverse
momentum and thus provides an essential validity of the QCD predictions. This
encouraged us to test QCD predictions at LHC energies. An understanding of high mass
di-photon production will also be essential in the search for the light mass Higgs Boson
(mH < 140 GeV/c2) decay via two-photons. We have used recent parton distribution
functions and the latest version of event simulation program PYTHIA (version 6.2) to
predict the direct-photon cross-section at LHC energies.
1.7 Thesis Organization
The thesis is organized in seven chapters, a brief description of each chapter is

given below.
The present chapter, i.e., Chapter 1 provided an overview of the CERN LHC
machine and CMS detector. Since the main emphasis of this thesis is on the detailed
analysis of the breakdown performance of silicon strip sensors to be used in Preshower at
CMS, hence a more detailed discussion on the Preshower detector is presented.
The fundamental structure of most of the silicon detectors is the p-n junction, thus,
an intuitive, non-rigorous description of the properties of p-n junctions can provide some
grasp on the main aspects of detector operation. In Chapter 2, general properties of
semiconductors (with emphasis on silicon) and reverse biased p-n junction are reviewed.
The chapter also describes the principle of operation of silicon sensors, and the
specifications/acceptance-criterion of the silicon strip detectors to be used in CMS
Preshower.
Because of the geometrical complexity of the metal-overhang structures, numerical
simulations that aid design optimization are expected to be unwieldy. In the present work,
two-dimensional device simulation tool, TMA-MEDICI is sought to analyze the
breakdown performance of silicon detectors. Chapter 3 describes the methodology and
approach of the simulation program. Various important aspects of the program including
mesh generation, boundary conditions, physical models description, and numerical
solution of the equations are briefly discussed. The values of the breakdown voltage
obtained using simulation depend critically on the correct choice of the physical models,
boundary conditions and most importantly on ionization coefficients.
23
We next address in Chapter 4 the breakdown voltage analysis of planar silicon
microstrip detectors equipped with metal-overhang. The chapter traces some significant
developments in the evolution of the overhang technique. The results reported in
literature clearly suggest that this technique is simple and versatile. However, little data is
available on the analysis and design of this technique. In this chapter, a physical
interpretation of the beneficial aspects of the overhang structure in terms of interaction of
charges in the space charge region and in metal-overhang is provided. The influence of
the salient design parameters, namely, field-oxide thickness, junction depth, overhang
width, device depth, substrate doping concentration and surface charges on the
breakdown voltage of metal-overhang structures is studied. Such an exhaustive analysis
of the structure equipped with metal-overhang correlating the breakdown voltage to these
parameters is of great importance for device optimization.
One of the primary objectives of the detector research in the high-energy physics
experiments is to stabilize the long-term behaviour of silicon strip detectors and it is of
utmost importance to protect the sensitive detector surfaces against moisture and other
adverse atmospheric environment. This is achieved by depositing the final passivation
layer over oxide of the silicon detector. Chapter 5 presents the influence of the relative
permittivity of the passivant on the breakdown performance of the Si detectors using
computer simulations. The semi-insulator and the dielectric passivated metal-overhang
structures are then compared under optimal conditions. By analyzing simulation results,
influences of all the salient physical and geometrical parameters on these structures have
been elaborated. Another important factor, which can significantly affect the long-term
functionality of the Si sensors, is the radiation damage and hence a crucial issue for the
detectors at LHC is their stability at high operating voltages. The effect of bulk damage
caused by hadron environment in the passivated Si detectors is simulated by varying
effective carrier concentration and minority carrier lifetime. Static measurement results
on some of the irradiated Si sensors, performed at CERN, Geneva along with the
irradiation facility and measurement set-ups is also described in this chapter.
In Chapter 6, we compare the two most commonly used termination techniques
employed in silicon detector technology namely, metal-overhang and floating field
limiting ring (or guard ring), under identical conditions with the aim of defining layouts
and technological solutions suitable for the use of silicon detectors in harsh radiation
24
environment. The results demonstrate the superiority of metal-overhang technique over
field limiting ring technique for planar shallow-junction high-voltage silicon detectors
used in high-energy physics experiments.
Chapter 7 starts with a brief review of the Standard Model and QCD, followed
by the study of direct photon physics. The chapter describes the study of direct photons in
the kinematical regions accessible at LHC energy. For standardization purposes, the
Fermilab Tevatron data on direct photons using latest version of event simulation tool
PYTHIA (version 6.2) with recent parton distribution function CTEQ5M1 is presented.
In order to explain the discrepancy between data and theory in the low transverse
momentum of photons (pT), the effect of parton transverse momentum (kT) prior to hard
scattering on the direct photon cross section is also investigated. It is found that the Next-
to-Leading Order (NLO) theory supplemented with kT correction accounts to a great
extent the low pT differences between data and theory. Predictions for direct photon cross
section at s =14 TeV for LHC along with various theoretical uncertainties is described.
It is found that the direct photons can be used to probe gluons at very low values of
momentum fraction (x) and at very high values of momentum transfer between two
partons (Q2).
25
Chapter 2
----------------------------------------------------------------
Physics of Silicon Detectors: A Résumé
----------------------------------------------------------------
2.1 Silicon Detector: An Introduction
Silicon detectors are used in almost all the High-Energy Physics (HEP)
experiments built in the last fifteen years, from fixed target to large Collider experiments,
and also in many specialized applications like spectrometers for space sciences or
detectors for medical diagnostics. Some of the characteristics, which are the basis of the
success of the silicon detectors and make them excellent devices for measurements,
include (a) excellent speed (~ 10 ns), (b) spatial resolution of 10 µm, (c) compactness, (d)
linearity of the response vs. deposited energy, (e) good resolution in the deposited energy
(3.6 eV is needed to create an electron-hole pair as against 30 eV in a gas detector), (f)
excellent mechanical properties, and (g) tolerance to high radiation doses up to ~ 10
Mrad.
The very first semi-conductor detectors were built in the early 1950s and
consisted of rectifying p-n junctions on crystalline germanium [2.1]. Germanium required
substantial cooling for good energy resolution because of its low intrinsic resistivity and
was rapidly overtaken by surface barrier silicon devices [2.2], which could be operated
satisfactorily at room temperature. These detectors are also used in low energy
spectroscopy. Due to the large electron-hole yield and low leakage currents, an energy
resolution below 1 keV is routinely achieved. However, they were not commonly
employed for particle detection. From the late 1950s through to early 1970s, nuclear
emulsion and bubble chamber dominated the fixed target experiments. The situation,
however, changed somewhat in the mid 1970s, after the discovery of the J/Ψ meson, a
bound quark-antiquark pair state with a new quantum number called charm. Since charm
26
events are produced rarely, strategies had to be developed which allow the suppression of
non-charm events. Only silicon detectors provided the excellent spatial resolution
necessary to distinguish between non-charm events containing only tracks that originated
from the interaction point, and the tracks originating from the charm decays which occur
at a certain distance from the interaction point due to the small but finite lifetime of the
charm particles. From there onwards, silicon detectors have been used extensively in
vertex determination both in fixed target and colliding beam experiments. The first vertex
detectors based on silicon strip technology were used successfully in MARK II
experiment at SLAC (for e+-e- experiments) and in CDF at FNAL (for hadron collider).
In a recently concluded Large Electron Positron (LEP) collider experiment at CERN, all
four detector systems had vertex detectors using silicon strip detector technology [2.3]. A
shift in paradigm occurred with the development of tracking detectors for hadron
colliders, from D0 (FNAL) to CMS (CERN). The new emphasis is not only on vertexing
but also on full tracking including charge and momentum determination in the magnetic
field. Silicon detectors are best suited to meet the challenges of high accuracy and
efficient track measurements because of the high level of possible segmentation into
strips and pixels. Their compactness and possibility to integrate the front-end electronics
on the same chip has also proved to be extremely advantageous. With an achieved
position resolution of a few microns, silicon detectors have already contributed
significantly to the study of τ−leptons, heavy quarks like charm and beauty and last but
not least to the discovery of the top quark at FNAL.
Future challenges: At present, the Si detector technology is well established and
optimized for purposes of present generation HEP experiments like those at Tevatron
(FNAL). Nevertheless, using these devices in the experiments such as those foreseen at
the LHC implies solving problems which are quite new to this scenario. Although other
materials such as gallium arsenide or diamond may be more radiation hard than silicon,
however, silicon is still an attractive choice because the fabrication process along with the
necessary electronics is developed to a high standard and reliability.
The fundamental structure of most of the silicon detectors is the p-n junction,
thus, an intuitive, non-rigorous description of the properties of p-n junctions can provide
some grasp on the main aspects of detector operation. In the following sections, general
27
properties of semiconductors (with emphasis on silicon) and reverse biased p-n junction
are reviewed. The chapter also describes the principle of operation of silicon sensors and
the specification/acceptance-criterion of the silicon strip detector to be used in CMS
Preshower.
2.2 Semiconductors: General Properties
A. Band structure in solids: The most important result of the application of

quantum mechanics to the description of electrons in the periodic lattice of crystalline
materials is the formation of allowed energy levels grouped into bands. The energy of
any electron within the pure material must be confined to one of these energy bands. The
lower allowed energy band, called the valence band, corresponds to those electrons that
are bound to specific lattice sites within the crystal. The next higher-lying allowed band
is called the conduction band and represents electrons that are free to migrate through the
crystal. These allowed energy levels are separated by gaps or ranges of forbidden
energies that the electrons in a solid cannot possess, called forbidden gap or band gap.
The band gap energy (Eg) is the energy difference between the top of the valence band
and the bottom of the conduction band. In perfect semiconductor there are no electron
energy levels in this band gap. For insulators, Eg is usually 5 eV or more, whereas for
semiconductors, it is considerably small ~ 1 eV (for silicon, Eg ~ 1.12 eV at room
temperature).
B. Charge carriers: The phenomenon of conduction is of principal interest in the
study of semiconductor physics. In the absence of thermal excitation (at 00 K), the
semiconductors have a configuration in which the valence band is completely filled and
the conduction band is completely empty, and hence no free electrons are available for
electrical conduction. However, at any non-zero temperature, some bonds are broken and
it is possible for a valence electron to be excited out of the covalent bond and elevated
into the conduction band. The excitation process not only creates an electron in the
otherwise empty conduction band, but it also leaves a vacancy (called a hole) in the
otherwise full valence band. Both of these charges, electron in the conduction band and
hole in the valence band, can conduct electricity under the influence of an applied electric
field and hence contributes to the observed conductivity of the material.
28
At low and moderate electric fields, the drift velocity of electrons (ve) and holes
(vh) is proportional to the applied field (E) and is given by
ve = µ e E & vh = µ h E (2.1)
where the proportionality constant µe and µh are the respective mobilities of

electrons and holes. The electron and hole mobilities in a crystal are not equal, but are of
the same order of magnitude unlike the case of electrons and positive ions in gases where
the two values differ greatly. At higher electric fields, the drift velocity increases more
slowly with the field. Eventually, a saturation velocity is reached ~ 107 cm-s-1 which
becomes independent of further increase in electric field.
C. Effect of impurities: Intrinsic semiconductors - In a completely pure
semiconductor, all the electrons in the conduction band and all the holes in the valence
band would be caused by thermal excitation. Since under these conditions each electron
must leave a hole behind, the number of electrons in the conduction band (n) exactly
equals the number of holes in the valence band (p), i.e.,
n = p = ni , (2.2)
where ni is called the intrinsic carrier concentration and its value in silicon at room
temperature is 1.45 x 1010 cm-3. This intrinsic carrier dictates the lower limit of the
leakage current in a reverse-biased diode.
Doped semiconductors – The electrical properties of a semiconductor changes
drastically with the addition of small concentration of impurities. We first consider that
the impurity is pentavalent (like phosphorous). Because there are five valence electrons,
surrounding the impurity atom, there is one left over electron after all covalent bonds are
formed. This fifth electron is very loosely bound, its binding energy being about 2% of
the Eg. In all practical situations due to thermal agitations, a large fraction of the
impurities are ionized and the concentration of electron in the conduction band (n) can be
as high as impurity doping density (ND). The impurity atom, which donates this free
electron, is called a donor impurity, and the semiconductor is called the n-type
semiconductor. In a similar manner, a trivalent impurity atom (like boron) can accept
with a very small expenditure of energy an electron from the valence band. This leaves
behind a hole in the valence band, which can move freely through the crystal. This type
of impurity is known as acceptor impurity and the semiconductor is referred to as p-type
29
semiconductor. In this case the number of holes (p) can be as high as the acceptor density
(NA). In real crystals, however, both donor and acceptor impurity atoms are present and
the two atoms compensate each other. Thus the quantity of interest is always the net
ionized impurity concentration (ND - NA): ND > NA for n-type semiconductor and ND <
NA for p-type semiconductor.
The added concentration of electrons in the conduction band for the case of n-type
crystal increases the rate of recombination, shifting the equilibrium between the electrons
and holes in a manner given by mass-action law
Eg Eg
− −
2
n. p = ni = N c N v e k BT
∝T e
3 k BT
(2.3)
where Nv and Nc are the effective density of states in the valence band and
conduction band respectively, kB is the Boltzmann constant and T is the absolute
temperature. Therefore, the net effect of doping is to increase the concentration of one
type of carrier and reduce correspondingly the concentration of opposite type. Thus, in n-
type semiconductor the electrons are called the majority carriers, and holes the minority
carriers.
One measure of the impurity level in the semiconductors is the electrical
conductivity, or its inverse, resistivity which is given by
1 1
ρ= = for n >> p (2.4)
q( µ n n + µ p p) qµ n n
For intrinsic silicon at room temperature ρ is equal to 227 kΩ-cm (µn = 1350
cm2V-1s-1, µp = 480 cm2V-1s-1). The resistivity of doped materials is lower, because of a
high carrier density.
2.3 Reverse Biased p - n Junction
2.3.1 Abrupt p - n Junction

When the doping concentration changes abruptly from a surplus of acceptors NA
on the p-side to a surplus of donor ND on the n-side one obtains an abrupt p-n junction.
Fig.2.1 represents an abrupt p-n junction in thermal equilibrium.
The strong gradient of carriers across the junction causes diffusion of electrons in
the p-type region and holes in the n-type region. The unbalanced fixed charges left
30
behind by diffusing carriers constitute the space charge region (SCR): negative ions in
the p-region and positive ions in the n-region. An electric field appears across the
junction due to the potential drop in the SCR. This results in a flow of drift current in the
direction opposite to the diffusion current. The equilibrium is reached when the potential
drop prevents further charge diffusion across the junction. This potential is called
diffusion or built in potential (Vbi), and it is about 0.6 V for silicon. Since SCR is
depleted of the mobile carriers, it is also known as depletion region.
In thermal equilibrium, with no applied voltage, the net flow of both electron and
hole currents is zero. The diffusion potential (Vbi) is equal to (from Fig.2.1)
qVbi = E g − (qVn + qV p ) , (2.5)
where q is the electron charge and Vn & Vp are the electrical potentials as shown in
Fig.2.1.
Fig.2.1: Abrupt p-n junction in thermal equilibrium. (a) Space-charge distribution. The
dashed lines indicate the majority carrier distribution tails. (b) Electric field
distribution. (c) Potential variation with distance where Vbi is the built-in
potential. (d) Energy-band diagram.
31
2.3.2 Expressions of Some Useful Physical Quantities for Abrupt p-n
Junction
The potential and electric field distribution in the depletion region can be
calculated by solving the one dimensional Poisson equation:
− d 2V ∂E ρ ( x)
dx 2
=
∂x
=
ε si
=
q
ε si
[p ( x) − n( x) + N D+ ( x) − N A− ( x) ] (2.6)
which for the abrupt junction under the depletion layer approximation becomes
(assuming that all impurity ions are ionized)
− d 2V q
≈− NA for − x p < x ≤ 0 (2.7)
dx 2
ε Si
q
≈ ND for 0 < x ≤ x n
ε Si
The electric field is obtained by integrating equation (2.7) to yield,
− qN A
E ( x) = (x + x p ) for − x p ≤ x < 0 (2.8)
ε Si
qN D
= ( x − xn ) for 0 < x ≤ xn
ε Si
The electric field reaches its maximum value Em at the junction, i.e., at x = 0,
qN D x n qN A x p
E m = E ( x = 0) = = (2.9)
ε Si ε Si
Potential distribution along with the built in potential can be obtained by integrating
equation (2.8) once again
 x2 
V ( x) = E m  x −  (2.10)
 2W 
1
Vbi = V ( x = W ) = E mW , (2.11)
2
where W = xn + xp is the total width of the depletion region.
Eliminating Em from the equations (2.9) and (2.11) yields for abrupt p-n junction,
2ε si  NA + ND 
W=   Vbi (2.12)
q  NAND 
32
Asymmetric junctions are often used in detector technology, which is obtained if
the doping density in one side is very large as compared to the density in the other side.
In particular if NA >> ND, one obtains one-sided abrupt p+-n junction, for which the
extension of the depletion region in the p+ side can be neglected, i.e., xp<<xn ~W.
Therefore, for this case the depletion width and electric field are given by,
2ε Si
W ≈ xn = Vbi , (2.13)
qN B
qN B
E ( x) = (x − W ) (2.14)
ε Si
where NB = ND, i.e., doping concentration of the less doped region.
A more accurate result for the depletion layer width can be obtained from
equation (2.6) by including the correction factor coming from the Debye tail of the
majority carrier distribution at the edges of the depletion region (shown as dashed line in
Fig.2.1(a)), each of which introduces a term kBT/q. The depletion width is essentially the
same as given by equation (2.13), except that Vbi is replaced by (Vbi – 2kBT/q) [2.4]. The
depletion-layer width at thermal equilibrium for a one-sided abrupt junction becomes
2ε Si  2k T 
W=  Vbi − B  (2.15)
qN B  q 
p – n junction under reverse bias: The width of the SCR can be changed if an
external voltage is applied to the diode. For an applied reverse bias (V), it is given as,
2ε Si  2k T 
W (V ) = Vbi − B + V  (2.16)
qN B  q 
 2k T 
For reverse voltage larger in comparison with  Vbi − B  (Vbi ~ 0.6 V and
 q 
kBT/q ~ 25 mV at room temperature), a simplified expression can be used for the
depletion of the SCR width on the applied voltage:
2ε siV
W (V ) = (2.17)
qN B
The minimum voltage needed for full depletion is called the full depletion voltage
(VFD), and can be deduced from equation (2.17), replacing W(V) by the detector
thickness d,
33
qN B d 2 d2
VFD = = (2.18)
2ε si ρµ n ε si
Equation (2.17) implies that the depletion thickness is inversely proportional to
the square root of substrate doping concentration NB. A silicon detector is normally
fabricated from a lightly doped n-type wafer so that a large active volume can be
obtained at small voltages.
The depletion region can be regarded as a parallel plate capacitor because of the
build up of the space charge. The junction or depletion layer capacitance per unit area is
defined as [2.4]
dQc d (qN BW ) ε
Cj = = = Si (2.19)
dV qN W (V )
d [( B )W 2 ]
2ε Si
where dQc is the incremental increase in charge per unit area upon an incremental
change of the applied voltage dV.
By substituting W from equation (2.17), the above relation for the reverse biased
diode can be written as:
qε Si N B
Cj = for V < V FD
2V (2.20)
ε
= Si for V > V FD
d
The junction capacitance therefore decreases with the applied voltage till the
whole volume is completely depleted; after which it remains constant reaching a value
consistent with the geometrical one. It is clear from equation (2.20) that by plotting 1/C2
vs. V a straight line should result for a one sided abrupt junction till full depletion is
attained and then the curve saturates. The slope of the curve gives the impurity
concentration of the substrate (NB).
2.3.3 Reverse Leakage Current

Reverse leakage current together with the breakdown voltage constitute two most
important characteristics of the silicon detectors. In order to reduce noise below the
acceptable levels, small values of leakage currents are desirable in silicon detectors.
34
Recombination-generation (R-G): Recombination-generation is the basic
mechanism giving rise to all type of currents observed in p-n junctions. There are three
basic R-G processes [2.4]: Radiative, Auger and Shockley-Read-Hall (SRH). The
transition of an electron from the conduction band to valence band is made possible by
emission of a photon (radiative or band-to-band process) or by transfer of energy to
another free electron or hole (Auger process). The third process is the recombination via
trapping centers (or R-G centers) present in the forbidden energy gap due to the presence
of impurity atoms or crystal defects (SRH or multiphonon process). Each of these
recombination processes has a generation counterpart. Direct optical transition (or
photogeneration) is the counterpart of radiative process and impact-ionization is that of
Auger process. The inverse of SRH recombination is the thermal electron-hole (e-h) pair
generation.
In reverse biased p-n junctions, there is a paucity of charge carriers in the
depletion region and e-h pair once generated, get separated under the influence of electric
field and hence their probability of recombination is diminished. Thus, generation
mechanism is chiefly responsible for the current-flow in the reverse biased silicon
detectors. Optical generation is important in direct band-gap materials (like gallium
arsenide) but plays little role in indirect band-gap semiconductors (like silicon). Impact-
ionization plays a significant role only at very high electric fields. Thus, under dark and
low field condition, optical generation and impact ionization are almost negligible in
silicon detectors. SRH generation takes place whenever there are impurities or defects.
Since semiconductors always contain some impurities, this mechanism is always active
and is particularly important for silicon diodes.
For single level generation process in which only one trapping energy level (Et) is
involved, the net generation rate (U) due to SRH process is given by [2.5]
ni2 − np
U = , (2.21)
(n1 + n)τ P + ( p1 + p)τ n
where ni is the intrinsic concentration, n and p are the electron and hole concentrations
and p1 and n1 are the equilibrium value of the hole and electron concentrations if the
fermi level (EF) were at the trap level (Et), i.e.,
35
 E − Ei   E − Et 
n1 = ni exp t  ; p1 = ni exp i  , (2.22)
 k BT   k BT 
Ei is the intrinsic energy level, τp and τn in equation (2.21) are minority carrier lifetimes
for hole and electron respectively and are defined as,
1 1
τp = & τ = , (2.23)
σ P vth N t σ n vth N t
n
where Nt is the density of trapping centers, σn and σp are the respective capture cross-
sections for the electron and holes respectively, and vth is the carrier thermal velocity.
Leakage current in Silicon detectors: Currents in p+-n silicon detectors due to e-h pair
generation at some place in the device are shown in Fig.2.2 [2.6] (we will assume ideal
junctions free of pinholes, defects etc.). Total current (I) can be written as the sum of the
currents in each separate region,
I = I I + I II + I III + I IV (2.24)
where, II = Due to R-G centres in the space-charge region
III = Due to interface traps at the SiO2/Si interface
IIII = Due to R-G centres in the undepleted (quasi-neutral) bulk
IIV = Due to the high-low (n-n+) junction at the back surface
Fig.2.2: Subdivision of the diode structure into four regions, according to which the
leakage current is calculated. W is the width of the depleted region and t ′ is the width of
the undepleted region.
Region I: In Region I, current-flow depends upon the generation rate of e-h pairs
in the depletion region and is given as [2.5]:
36
I I = q | U | WA (2.25)
where q is the electron charge, W is the depletion width, A is the cross-sectional area of
the p-n junction and U is the net generation rate as given in equation (2.21). In this
region, both the electron and hole concentration can be neglected (n, p = 0) and assuming
U to be constant over the entire layer [2.7], equation (2.21) reduces to
ni2 n n
U = ≡ i ≡ i , (2.26)
n1τ P + p1τ n τ g 2τ 0
n1τ p + p1τ n  E − Ei   E − Ei 
and τg = = τ p exp t  + τ n exp − t  (2.27)
ni  k BT   k BT 
τg is the SRH generation lifetime and for particular case when Et = Ei and σn = σp, it
reduces to τ g = 2τ 0 , where τ0 is defined as the effective lifetime within a reverse biased
depletion region.
Substituting U in equation (2.25), the current in region I is given as
qniWA
II = (2.28)
2τ 0
The presence of W in this equation implies that this component of the leakage
current increases with increase in reverse bias ( V dependence) and saturates at full
depletion voltage [2.4, 2.5, 2.7].
Region II: The surface generation component in Region II is given as [2.5],

I II = qU s As (2.29)
where As is the junction area at the surface and Us is the surface generation rate per unit
area at the oxide-silicon interface. For the completely depleted surface [2.5],
1
Us = ni s o (2.30)
2
where so is the surface recombination velocity and is defined as s o = σv th N st for centers
with energy levels Et = Ei., and Nst is the surface density of R-G centers. Thus,
qni s 0 As
I II = (2.31)
2
The complicated geometry and field distribution at the surface region of a silicon
detector makes it difficult to predict the depleted interface area As. The amount of
37
interface states contributing to the surface recombination velocity s0 is a characteristic of
the quality of the SiO2 processing during fabrication. Surface current is an unavoidable
component of microstrip detector leakage current, however, the inclusion of guard rings,
which are the diodes surrounding the detector, can help in minimizing this component.
Regions III & IV: Regions III and IV are undepleted and thus the diffusion of
minority carriers (holes in our case) determines the contribution of these regions to the
leakage current. The combined contribution is given by [2.6]
q ni2 Dn A
I III + I IV = , (2.32)
N B Ln ,eff
where the diffusion length Ln,eff is an effective diffusion length that couples the bulk
diffusion length Ln (region III) with the surface generation velocity at the back surface sc
(region IV). For simplicity the SCR width at the surface is assumed to be identical to that
in the bulk as shown in Fig.2.2. A small contribution to the diffusion current also comes
from the undepleted p+ region (electrons being the minority carrier). Although diffusion
component is the dominant current source at room temperature for semiconductors with
high ni like germanium (because of ni2 dependence of diffusion current), but for silicon it
is less important. Moreover, for a fully depleted silicon detector, this component is almost
negligible for operating temperatures below 100°C [2.5].
Thus, the total reverse current in silicon detector where the depletion region
extends mainly in the n-side, can be expressed as the sum of the generation currents in
region I and region II.
qniWA qni s 0 AS
I= + (2.33)
2τ 0 2
2.3.4 Breakdown Voltage

In the reverse bias p-n junction, the breakdown is defined as the steep rise in
current when the reverse bias goes above a certain limit (VBD). The high current flowing
in the breakdown regime makes silicon detector unusable and could destroy the device if
the bias is not rapidly decreased below VBD. Therefore, it is important to keep VBD of the
silicon sensors as high as possible. Three basic breakdown mechanisms [2.4] in the
38
reverse biased p-n junctions are: thermal instability, tunneling or Zener breakdown and
avalanche multiplication.
A. Thermal instability: At high reverse voltages, the heat dissipation caused by the
reverse current increases the junction temperature, which in turn, further increases the
reverse current and causes thermal runaway. However, this effect is particularly
important for semiconductors with relatively small band gaps (for example, germanium)
and not so important in silicon.
B. Tunneling or Zener breakdown: Due to the occurrence of very high fields within
the depletion region at high reverse voltages, some of the covalent bonds between
neighbouring atoms are “torn” apart, resulting in the generation of conduction electrons
and holes. This corresponds to the tunneling of electrons from the valence band edge of
p-type to the conduction band edge on the n-side. Zener breakdown occurs in p-n
junctions that are heavily doped on both sides of the metallurgical junction and hence is
relatively unimportant for silicon detectors, where the substrate is almost intrinsic.
C. Avalanche multiplication: The most common mode of breakdown mechanism in
silicon detectors is the avalanche breakdown. When a semiconductor is subjected to an
increasing electric field, a point is reached when mobile carriers in the depletion region
attain saturation drift velocity. With further increase in electric field, the velocity of
individual carriers exceeds their thermal velocity, i.e., they become “hot” carriers. At a
critical electric field, these carriers gain sufficient kinetic energy so that their collisions
with the lattice atoms can knock-on the valence electrons, leaving a hole behind. This
process of generation of electron-hole (e-h) pairs is called impact ionization (Fig.2.3).
Each newly generated carrier also gets involved in the ionization of further e-h pairs.
Consequently, impact ionization is a multiplicative phenomena and the device is
considered to undergo avalanche breakdown when the process attains an infinite rate.
Ionization Coefficients: To characterize the avalanche process it is useful to
define ionization coefficients for holes (αp) & electrons (αn) [2.8]:
αp : number of e-h pairs produced by a hole traversing 1cm through the depletion
layer along the direction of electric field.
αn : number of e-h pairs produced by an electron traversing 1cm through the
depletion layer along the direction of electric field.
39
Extensive measurements of these coefficients for silicon have been conducted
[2.9] and are found to vary with electric field as:
−b p / Ε
α n = a n e −b
n /Ε
& α p = ape (2.34)
where an = 7 x 105 cm-1, bn = 1.23 x 106 Vcm-1, ap = 1.6 x 105 cm-1, bp =2 x 106 Vcm-1
[2.10]. These expressions are found to be applicable for electric fields ranging from 1.75
x 105 to 6 x 105 Vcm-1 [2.10].
In many cases an approximation of the ionization coefficients is found to be
useful to obtain a closed-form expression for breakdown voltage and is given by [2.11]
α n ≅ α p ≅ α i ≅ 1.8 × 10 −35 E 7 (2.35)
Breakdown condition: Assume that an e-h pair is generated within the

depletion region at a distance x from the junction. In p+-n junctions under reverse bias,
the hole will be swept towards the junction side (p+) and the electron toward the depletion
layer edge. When traversing a distance dx, the hole will create αpdx e-h pairs and the
electron will produce αndx e-h pairs [2.8]. The average total number of e-h pairs created
in the depletion layer due to single e-h pair initially generated at x (called Multiplication
coefficient) is given as [2.9]
x W
M ( x) = 1 + ∫ α p M ( x )dx +' '
∫α n M ( x ' )dx ' (2.36)
0 x
Fig.2.3: The process of impact ionization and avalanche multiplication, from [2.5].
A solution of this equation is:

x 
M ( x) = M (0) exp  ∫ (α p − α n ) dx '  (2.37)
0 
40
Eliminating M(0) from equations (2.37) using (2.36), we get:
x 
exp  ∫ (α p − α n ) dx ' 
0 
M ( x) = (2.38)
 
α n exp  ∫ (α p − α n ) dx '  dx
W x
1− ∫
0 0 
Avalanche breakdown occurs when M(x) tends to infinity, i.e., when
W'
x
∫ α n exp  ∫ (α p − α n ) dx  dx = 1
'
(2.39)
0 0 
where W ′ is the depletion layer width at breakdown. This is the general condition of
avalanche breakdown in reverse biased p-n junctions. This equation can be simplified
using equation (2.35)
W'
∫α
0
i dx = 1 (2.40)
One-sided plane-parallel abrupt junction (non-punch through case): A closed

form analytical expression for the VBD using electric field distribution given by equation
(2.14) and ionization coefficient approximation (equation (2.35)) can be obtained using
equation (2.40)
W' 7
 qNB 
∫ 1.8 × 10
− 35
 ( x − W )  dx = 1 (2.41)
0  ε Si 
Using this equation, the depletion layer width at breakdown for the parallel-plane
junction can be obtained [2.12]:
′ = 2 . 67 × 10 10 N B− 7 / 8
W PP (2.42)
where subscript “PP” means plane-parallel junction and NB is expressed in cm-3. Critical
′ in the
electric field at breakdown ( E cPP ) can be found ([2.12]) replacing xn by WPP
maximum electric field expression (equation (2.9)) and using equation (2.42)
qN BWPP′
E cPP = = 4010 N B1 / 8 (2.43)
ε Si
where the field is expressed in Vcm-1.
41
Similarly breakdown voltage ( V BDPP ) for the one sided abrupt plane parallel
′ in equation
junction can be obtained ([2.12]) replacing Em by E cPP , and W by WPP
(2.11), and using equation (2.42)
V BDpp =
1
2
(E cPP WPP)
′ = 5.34 ×1013 N B−3 / 4 (2.44)
i.e., the breakdown voltage is given by the area under E vs. x curve.
One-sided plane-parallel abrupt junction (punch through case): It was assumed

hitherto that the lightly doped side of the junction extends beyond the edge of the
depletion layer under avalanche breakdown. This is not always true, and in particular in
case of radiation detector which has to be operated in full depletion mode, the device
depth (d) is smaller than W ′ and such a diode is called punch-through (PT) diode.
Assuming that breakdown of each device occurs for the same critical electric field Ec, let
E1 be the electric field at x = d when the field at the junction reaches its critical value
(Ec), (Fig.2.4).
Fig.2.4: A comparison of the non-punch through (NPT) and punch-through (PT) diode.
The breakdown voltage for PT diodes is given by:

V BDPT = ( E c + E1 )d / 2 (2.45)
Using equation (2.44), it can be shown that [2.4]
42
2
V BDPT  d   d 
≅ 2  −  (2.46)
V BDPP W ′  W ′ 
Thus, the breakdown voltage of the punched through diode is always less than
that of its normal counterpart. The punch through usually occurs when doping
concentration NB is sufficiently low. An interesting aspect of the plane-parallel PT diodes
is that the NB has a negligible influence on its breakdown voltage [2.13].
Planar diffuse junction termination (edge effect): In the previous sections, we

assumed an infinite extension of the junctions without any edges. But in real situation,
when the junction is fabricated by diffusing the dopants through the mask windows (as in
planar technique), the impurities also diffuse laterally at the edges of the mask window
(Fig.2.5(a)). For purposes of breakdown analysis, the lateral diffusion can be considered
to be equal to the junction depth to a good approximation. This results in the formation of
cylindrical junctions inside the diffusion windows and spherical junctions at the edges for
the actual diodes (Fig.2.5(b)). For actual devices therefore, it becomes imperative to
consider edge effects (cylindrical and spherical regions) since the curvature limits the
breakdown voltage to values much below the ‘ideal’ limits set by the plane-parallel
junction [2.4].
(a) (b)
Fig2.5: (a) Planar diffusion process which forms junction curvature near the edges of
the diffusion mask. rj is the radius of curvature, and (b) The formation of approximately
cylindrical and spherical regions by diffusion through a rectangular mask.
Edge effect - Cylindrical junction: Solving the Poisson equation in cylindrical

coordinates, electric field distribution is given by
43
qΝ B  rd 2 − r 2 
Ε( r ) =   (2.47)
2ε Si  r 
 
where rd is the radius of curvature of the depletion layer edge.
Examination of equation (2.47) shows that for cylindrical junctions, the high field
region is largely confined to small values of r near the boundary of the metallurgical
junction. This allows an approximation for the electric field distribution E(r) = K/r (K
being some constant independent of r) [2.9, 2.12] and substituting this in the ionization
integral equation (2.40), the ratio of Ec and VBD for cylindrical and plane-parallel junction
are given as [2.12]:
1/ 7
Ec ,CYL  3W ′ 
=  (2.48)
E c , PP  4r 
 j 
  r  2  r j   
6/7
 W ′    r j  
8/7 6/7
V BDCYL
 1 
= 
j
 + 2   ln 1 + 2  −   (2.49)
V BDPP  2  W ′  W ′  r   W ′  
       j     
Here, rj is the junction depth.
Edge effect – spherical junction: Following the same approach as used in the
previous section, and with an approximation of electric field distribution E(r) = K/r2 [2.9,
2.12] for calculating the ionization integral for spherical junction, the expressions
corresponding to equations (2.47), (2.48), and (2.49) are given by [2.12]
qN B  rd 3 − r 3 
E (r ) =   (2.50)
3ε Si  r2 
 
1/ 7
E c , SP  13W ′ 
=  (2.51)
E c , PP  8r 
 j 
 r 2 6/7
  r j  3 13 / 7 2 / 3
 
VBDSP    r   r 
=    + 2.14   −   + 3    
j j j
(2.52)
VBDPP   W ′   W ′    W ′   W ′   
  
Fig.2.6 shows the plot of normalized breakdown voltage against the normalized
radius of curvature both for cylindrical and spherical junction. It can be seen that the
breakdown voltage for the spherical junction is less than that of the cylindrical junction,
and that of cylindrical junction is less than the breakdown voltage of plane parallel
44
junction. However, it is also clear that the breakdown voltage of curved junctions (both
cylindrical and spherical) increases with increase in the radius of curvature and begins to
approach the parallel plane case for large values of junction depth.
Fig.2.6: Normalized breakdown voltage as a function of the normalized radius of

curvature both for cylindrical and spherical junctions [2.4].
2.4 Principle of Silicon Detector Operation
A silicon detector is essentially a reverse biased diode with the depleted zone
acting as a solid-state ionization chamber. When charged particles pass through a silicon
detector, many e-h pairs get produced along the path of the particle. Average energy
required to create a single e-h pair is about 3.6 eV for silicon. The energy loss in silicon
can be measured by “counting” the total number of created pairs. Under the application
of reverse bias, electrons drift towards the n+ side and holes to the p+ side. This charge
migration induces a current pulse on the read out electrodes and constitutes the basic
electrical signal. Integration of this current equals the total charge and hence is
proportional to the energy loss of the particle. The high mobility of electron and holes
enables this signal charge to be collected very quickly. It should be pointed out here that
only the charge released in the depletion region can be collected, whereas the charge
created in the neutral, non-depleted zone recombines with the free carriers and is lost. For
this reason silicon detectors usually operate with an applied voltage sufficient to fully
45
deplete all the crystal volume. Fig.2.7 shows the principle of operation of silicon
microstrip detector.
A minimum ionizing particle traversing a <111> oriented Si layer of depth 300
µm deposits the most probable energy of about 90 keV and produces about 25000 e-h
pairs (~ 4f C).
Fig.2.7: Principle of operation of silicon strip detector.
2.5 Silicon Microstrip Detector for CMS Preshower

India’s participation in Preshower detector of the Compact Muon Solenoid (CMS)
experiment was agreed upon in early 1997 and under this agreement, Delhi University
(DU) and Bhabha Atomic Research Centre (BARC) are responsible for the development
of silicon strip detectors. Three other groups, EHEP – Tata Institute of Fundamental
Physics (TIFR) (Mumbai), HECR – TIFR (Mumbai) and Panjab University have
contributed to the fabrication of the Outer Hadron Calorimeter (HO–B) of CMS. All five
groups have agreed to work together and formed an India–CMS collaboration. The R&D
of the detector technology has been jointly carried out by scientists at DU, BARC, Bharat
Electronics Limited (BEL), Bangalore and Central Electronics Engineering Research
Institute (CEERI), Pilani. Initially the technology was developed at CEERI on 2-inch
wafer and later on the prototypes were fabricated using 4-inch wafers at BEL. Each
silicon detector is a square of 63 x 63 mm2 divided into 32 strips.
46
2.5.1 Silicon Microstrip Detector
The silicon microstrip detector is by far the most widely used semiconductor
particle tracking device. At its simplest, the microstrip detector consists of a series of
reverse biased diodes constructed on a single silicon wafer. Microstrip detectors, thus,
provide the measurement of one coordinate of the particle’s crossing point with high
precision. Using very low noise readout electronics, the measurement of the centroid of
the signal over more than one strip further improves the precision. Usually strips are p+
implants to provide p+-n junction in the n-bulk, which may be DC or AC coupled to the
read out electronics. On the ohmic side, n+ is implanted to provide the ohmic contact to
bias the detector.
2.5.2 Wafer Parameters [2.14]

The detectors have been fabricated on float zone n-type, <111> oriented, 2.5 - 4
KΩ-cm wafers supplied by TOPSIL or Wacker. The quality of wafers is a critical factor
for detector fabrication, as even a single defect occurring over the detector area, which is
quite large, would result in a bad strip giving non-acceptable performance. Hence wafers
with zero defect density and high life time of the order of milliseconds have been used for
detector fabrication. Since Preshower strip capacitance is dominated by back plane
capacitance, there is no particular advantage to use <100> orientation, and a more
classical <111> orientation has been chosen which has the largest number of available
bonds per unit area. Single-sided polished wafer is considered because of cost
effectiveness and availability. Also it is found that it is the thickness of the back plane
implant and not the quality of the back surface which influences the detector
performance. Resistivity of 2.5 - 4 kΩ-cm is chosen to achieve low leakage current, high
breakdown voltage and satisfactory performance after bulk inversion due to radiation
damage. A wafer thickness of 320 ± 20 µm is chosen to achieve a satisfactory S/N ratio
and desired operating voltage.
2.5.3 Sensor Design / Geometry [2.14]

Detector dimension of 63 x 63 mm2 is the maximum available area possible on a
4-inch wafer. Strip-pitch of 1.9 mm is governed by the photon separation (from π0s) and
transverse shower spread. Strip-width of 1.78 mm is the optimum compromise between
47
reduction in interstrip capacitance and good charge collection for irradiated sensors. The
metal-overhang of 10 µm on each side of the strip improves the breakdown performance.
A good ohmic contact between metal and silicon at the backside is achieved by using n+
implant of suitable thickness [2.14] and very high doping. Table 2.1 provides the
specifications for the wafer & geometry of the Preshower silicon sensor.
Table 2.1: Preliminary specifications (wafer/geometry) of the Preshower silicon sensor.

Parameter Value
Wafer size 4"
Thickness 320 ± 20 µm
Resistivity 2.5 KΩ cm − 4 KΩ cm
Polishing Single-sided
n+ layer thickness > 2.5 µm
Total area 63 x 63 mm2
Number of strips 32
Strip pitch 1.9 mm
p+ strip width 1.78 mm
Al strip width 1.8 mm
2.5.4 Fabrication
Planar technology is the principal method of fabricating modern semiconductor
devices [2.15] (Fig.2.8). In total, four masks are used including passivation. For cost
effectiveness, the p+ strips are Directly Coupled (DC) to the Aluminium readout lines and
the front-end electronics includes a leakage current compensation mechanism [2.16].
Figures 2.9(a), 2.9(b), and 2.9(c) show the layout, cross-section and the complete 63 x 63
mm2 CMS Preshower silicon strip detector.
The basic planar process used at CEERI, Pilani and BEL, Bangalore to fabricate
the silicon microstrip detectors involve the following steps:
48
1. Initial Oxidation
2. p+ lithography
3. Oxidation for screen oxide
4. Re-expose p+ mask
5. Implantation of Boron for p+ strips
and guard rings
6. n+ implant at the backside
7. Implant anneal and redistribution
8. Contact lithography
9. Front Metallization
10. Metal lithography
11. Metal sintering at 450 ˚C
12. Passivation
Fig.2.8: Main steps in the planar
fabrication process of detectors [2.15].
Fig.2.9: (a) Layout of silicon preshower sensor, (b) Cross-section of a preshower

silicon microstrip detector, and (c) complete 63 x 63 mm2 CMS Preshower Si detector
(with front-end electronics).
49
2.5.5 Acceptance Criterion [2.17]
The silicon sensors to be used in Preshower detector have to meet the
specifications as listed in Table 2.2, specified by Preshower group at CERN. The VBD >
500 V is required for sensors to be placed in the central ring, close to the beam pipe
where the expected radiation flux is maximum, whereas 300 V detectors will be placed in
the periphery where the flux level will be less than by an order of magnitude.
Table 2.2: Acceptance criterion for the Preshower silicon sensor [2.17].
Test Criterion
Width (W) W ≤ 63 mm
Mechanical
tolerances Thickness (t) 300 µm ≤ t ≤ 340 µm
I ≤ 5 µA at VFD
I-V I ≤ 10µA at 300 V

Global VBD ≥ 300 V for category 1
measurements
VBD ≥ 500 V for category 2
C-V VFD ≤ 100 V
Max. 1 strip with I ≥ 1 µA at VFD

I
Strip-by-strip Max. 1 strip with I ≥ 5 µA at 300 V
measurements No strips connected to the neighbour or
C
guard ring
In India, the silicon detectors are being fabricated at BEL, Bangalore to meet the
desired specifications. Theoretically, breakdown voltage of 500 volt can be obtained for
high resistivity substrates used in Preshower, however, due to the large area, junction
curvature, and surface charges, it becomes very difficult to obtain high yield. As
discussed earlier, metal-overhang of 10 µm is chosen in the PSD baseline design to
improve the breakdown voltage. However, in order to effectively utilize the advantages
of metal-overhang, effect of various geometrical and physical parameters on its design is
a prerequisite. In Chapters 4 and 5, we have studied and analyzed the influence of these
parameters on overhang structures using computer simulation.
50
Chapter 3
----------------------------------------------------------------
Semiconductor Device Simulation
----------------------------------------------------------------
Since most of the work done in this thesis is based on the results obtained from
device simulation, it is imperative to discuss the various aspects of a simulation program.
This chapter describes the methodology and approach of the simulation program, TMA-
MEDICI [3.1], used in the present work.
3.1 Why Simulation ?
The reduction in active device dimensions to micron and submicron geometries

has resulted in an intimate coupling of the process conditions and device behaviour to a
degree unknown a few years ago. It becomes more and more difficult to develop new
processes due to the inherent complexity of semiconductor device fabrication. The use of
Computer-Aided Design (CAD) tools has emerged as a very elegant mechanism to aid
process and device engineers in their task of finding an optimum process and hence
proven to be invaluable in the development of new technologies.
Traditionally, a new technology development has been guided by an experimental
“trial-and-error” approach. Starting with an existing process, certain steps in the process
are changed, together with the structural dimensions. The modified process is then used
to fabricate several lots of a device. However, this approach requires many iterations to
optimize a new process, and fabricating one lot in a modern process can cost considerable
amount of money and consume weeks or even months of effort. The use of accurate
simulation tools in the proper computing environment, on the other hand, allows for
comparatively inexpensive and time-saving “computer experiments”. Reduction of both
51
optimization time and prototypization expenses is therefore expected from the adoption
of device simulation.
A Boon for Silicon Detector Development: Technology-CAD tools are routinely
used in IC production and development environment; however, their diffusion within the
HEP community is relatively recent. Some of the advantages of using device simulation
for the development of silicon sensors are as follows:
• The development of silicon detectors for LHC demands large resources, both in terms
of time and money. Device simulation programs may help in the prediction of the
characteristics of silicon detectors, depending upon their geometry and fabrication
parameters.
• Numerical simulation gives an opportunity to look into the internal device mechanism
of silicon detectors by examining quantities like electric field and carrier distribution,
thus, allowing for the physical interpretation of several interesting experimental
findings and phenomena.
• Simulations make it possible to find a connection between non-measurable physical
quantities (like electric field, impact ionization generation rate etc.) and measurable
terminal parameters (like terminal voltage and current).
• In order to optimize the breakdown performance of the detector design, it is of utmost
importance to identify regions at which leakage current may preferentially develop,
and to correlate the threshold of such phenomena to geometrical and physical device
characteristics. The adoption of CAD tools allows for evaluating the actual field
distribution within the device and makes it possible to identify critical regions.
• Simulation helps in evaluating the device sensitivity to various design parameters and
thus aid in optimizing the final design.
• Because of the large number of silicon sensors and the difficulty to replace defective
devices in the Preshower detector system, the operation time of these sensors has to
be at least 10 years in the CMS experiment. In this period an equivalent irradiation
dose of more than 1014 cm-2 hadrons is expected, which has never been obtained
before in any HEP experiment. This requires a good prediction of the effect of
radiation-induced changes in silicon detectors. Since it is not possible to actually
52
predict the changes in bulk characteristics in advance, a good estimate of the device
behaviour after radiation damage can be obtained using device simulation.
3.2 TMA-MEDICI – A Device Simulation Program
TMA-MEDICI is a powerful two-dimensional device simulation program that can

be used to simulate the behaviour of p-n junctions, MOS & bipolar transistors, and other
semiconductor devices. The program can be used to predict the electrical characteristics
of a device for arbitrary bias conditions. MEDICI can also perform AC small signal
analysis in addition to DC steady state and transient analysis. The different aspects of the
program used in this work, which includes the physical description and the numerical
methods along with a sample program is briefly described in the following sections.
3.2.1 Physical Description
3.2.1.1 Drift-Diffusion Model

The primary function of MEDICI is to solve three semiconductor partial
differential equations (PDEs) self-consistently, Poisson’s equation for the electrostatic
potential (Ψ) and two current continuity equations for the electron and hole
concentrations, n and p respectively. These equations describe the electrical behaviour of
semiconductor devices.
+ −
Poisson equation: ∈ ∇ 2ψ = − q( p − n + N D − N A ) − ρ S (3.1)
∂n 1
Continuity equation for electrons: = ∇.J n − U n (3.2)
∂t q
∂p 1
Continuity equation for holes: = − ∇.J p − U p (3.3)
∂t q
where ND+ and NA- are the ionized impurity concentrations, ρS is the surface charge
density, Jn and Jp are the respective current densities, and Un & Up are the respective net
recombination rate for electrons and holes. MEDICI incorporates both Boltzmann and
Fermi-Dirac statistics, including the incomplete ionization of impurities.
53
3.2.1.2 Physical Models
A number of models are incorporated into the program for accurate simulations.
MEDICI supports Shockley-Read-Hall (SRH) (including tunneling in strong electric
fields), Auger & direct recombination, and also an additional recombination component
at insulator-semiconductor interfaces (described by surface recombination velocity). Both
concentration and temperature dependent lifetimes are included.
MEDICI provides several mobility model choices, which accounts for scattering
mechanism in electrical transport. For low electric fields, several mobility choices are
available which includes the effect of local impurity concentration, temperature and
carrier-carrier scattering. Along insulator-semiconductor interfaces, the carrier mobilities
can be substantially lower than in bulk due to surface scattering. An enhanced surface
mobility model [3.2] has been included to consider the phonon scattering, surface
roughness scattering and charged impurity scattering. For considering the effects due to
high field in the direction of current flow, field-dependent mobility model based on the
Caughey-Thomas [3.3] expressions is also available.
In order to predict the breakdown voltage accurately, generated carriers due to
impact-ionization can be included self-consistently in the solution of the device
equations. This method is particularly advantageous in predicting the avalanche-induced
breakdown of a reverse biased silicon detector. To simulate the effect of junction
curvature (as in case of planar process), lateral doping extension can also be specified.
Since ionization coefficients play a significant role in predicting the breakdown voltage
accurately, a correct choice of these coefficients is very essential. In MEDICI, the
dependence of ionization coefficients on the local electric field and temperature is based
on the Selberherr Model [3.4]. The values of the coefficients are found to be best
described by Overstraeten and DeMan data [3.5] for the breakdown voltage analysis in
the present work.
3.2.1.3 Boundary Conditions

MEDICI supports four types of basic boundary conditions: Ohmic contacts,
Schottky contacts, contacts to insulators and Neumann (reflective) boundaries. Ohmic
contacts are implemented as simple Dirichlet boundary conditions, in which the surface
potential and electron & hole concentrations (ψs, ns, ps) are fixed. The minority and
54
majority carrier quasi-Fermi potentials are equal and are set to the applied bias of that
electrode (φn = φp = VBias). The surface potential is fixed to a value consistent with zero
- +
space charge, i.e., n s + N A = p s + N D . Schottky contacts to the semiconductor are
defined by a work function of the electrode metal and optional surface recombination
velocity. Contacts to insulators generally have a work function, dictating a value for
surface potential similar to that used in ohmic contacts. The electron and hole
concentrations within the insulator and at the contact are forced to be zero. Along the
outer (noncontacted) edges of the devices, homogeneous Neumann boundary conditions
are imposed to make sure that the current only flows out of the device through contacts.
Additionally, in the absence of surface charges along such edges, the normal component
of the electric field goes to zero and current is not permitted to flow from the
semiconductor into an insulating region. In general, at the interface between the two
different materials, the difference between the normalized components of the respective
electric displacements must be equal to the surface charge density present along the
interface.

ε 1 nˆ.∇ψ 1 − ε 2 nˆ .∇ψ 2 = σ s (3.4)
As discussed in the subsequent chapters, surface charge density (σS) is a very important
physical parameter in determining the breakdown performance of silicon detectors. It
arises due to fabrication process itself and also due to ionizing radiation damage and is
unavoidable. MEDICI incorporates both fixed and trapped charges during simulation.
3.2.2 Numerical Methods

In order to solve the semiconductor PDEs described by equations (3.1) - (3.3),
these equations are discretized in a simulation grid. The resulting set of algebric
equations is coupled and nonlinear, and is solved using non-linear iteration methods.
3.2.2.1 Discretization
To solve the device equations on a computer, they must be discretized on a
simulation grid. The continuous functions of the PDEs are represented by vectors of
function values at the nodes, and the differential operators are replaced by suitable
difference operators. Thus, instead of solving 3 unknown functions (ψ, n, p), MEDICI
55
solves for 3N real numbers, where N is the number of grid points. The key to discretizing
the differential operators is the Box Method [3.6]. Each equation is integrated over a
small volume enclosing each node, yielding 3N nonlinear algebric equations. The
integration equates the incoming flux with the sources and sinks inside it. The integrals
involved are performed on an element-to-element basis, leading to a simple and elegant
way of handling general surfaces and boundary conditions.
3.2.2.2 Nonlinear System Solutions

As already mentioned, discretization gives rise to a set of coupled nonlinear
equations, which must be solved by nonlinear iteration method. Two approaches are
commonly used: Decoupled solutions (Gummel’s method) and coupled solutions
(Newton’s method). Newton’s approach, with Gaussian elimination method is by far the
most stable solution method. Full Newton is the method of choice for two carrier
simulations at high currents. For low current solutions, the Gummel method offers an
attractive alternative, in which the PDEs are solved sequentially. MEDICI contains a
powerful continuation method (used with two-carrier Newton method) for the automatic
tracing of I-V characteristics. This method automatically selects the bias step and
boundary conditions appropriately for the bias conditions and is particularly helpful in
predicting breakdown voltages.
3.2.2.3 Simulation Grid

Grid (or Mesh) plays a very important role in device simulation and its correct
allocation is absolutely essential for obtaining accurate results. MEDICI uses a non-
uniform triangular simulation grid and can model arbitrary device geometries with both
planar and non-planar surface topographies. The primary goal of the grid generation is to
achieve accurate simulation results with the least amount of simulation time. A coarse
mesh implies less simulation time but less accurate results whereas a fine grid increases
accuracy at the expense of time. This requires a suitable trade-off between the two.
Generally, the grid should have most node points where the gradients of the physical
quantities (like doping, potential etc.) are the highest and a less dense mesh in a uniform
region. User specification is thus, the most difficult aspect of general grid structure. To
minimize this effort, MEDICI provides a regridding mechanism that automatically
56
refines an initial grid wherever a key variable varies by more than a specified tolerance. It
is seen that distorting a rectangular mesh unavoidably introduces a large number of very
obtuse triangles. MEDICI also provides mesh-smoothing procedures to deal with them.
An example of a MEDICI mesh is shown in Fig.3.1.
P+
N-
Fig.3.1: A typical MEDICI mesh (regridding is performed around the junction region).
3.2.3 MEDICI Program Description

In this work, TMA-MEDICI version 2000.4 is used to analyze the breakdown
performance of silicon detectors equipped with metal-overhang. For that, firstly a suitable
grid is generated in accordance with the device structure and then the models and the
solution algorithms are specified to simulate the electrical characteristics. A command
input file needed for the simulation generally has the specific structure/organization
ordered in four groups: (a) Structure specification, (b) Coefficient and Material
parameters, (c) Solution specification and (d) Input / Output statements. The MEDICI
program structure with important command statements and comments is given in Table
3.1. An example of a typical MEDICI program is given in Fig.3.2.
57
Table 3.1: The structure and related command statements of the MEDICI.
Command Related Comments
Group Statements
Structure Mesh Initiates the mesh generation or to read a previously
Specification generated device structure from a data file.
Region Specifies the location of material regions in the structure.
Electrode Specifies the location of electrode in the structure.
Profile Specifies impurity profile for the structure.
Regrid Allows refinement of coarse mesh.
Stitch Append the generated structure to the simulation mesh.
Coefficients Material Specifies material properties.
& Material Mobility Specifies parameters associated with mobility models.
Parameters
Impurity Specifies parameters associated with impurities.
Contact Specifies parameters associated with electrodes; specifies
special boundary conditions.
Interface Specifies interface parameters for the structure.
Solution Models Enables the use of physical models during solution.
Specification Symbolic Performs a symbolic factorization.
Method Sets parameters associated with solution algorithms.
Solve Generates solutions for specified biases.
Input/Output Extract Extracts selected data over device cross-section.
Statements Plot.1d Plots a quantity along a line through the structure; plots
terminal characteristics from data in a log file.
Plot.2d Plots device boundaries, junctions, and depletion edges
in two dimensions.
Plot.3d Initiates three-dimensional plots.
Contour Plots two-dimensional contours of a quantity.
Log Specifies files for storing terminal and user-defined data.
Load Reads a solution stored in a file.
Save Writes solution or mesh information to a file.
58
$ An example of a typical MEDICI program
$ Create an initial Simulation Mesh and save the mesh file
MESH OUT.F=<filename>
X.MESH X.MAX= H1= H2= H3=
Y.MESH N= location= ratio=
Y.MESH Y.MAX= H1= H2= H3=
$
$ Region Definition
REGION NAME= <name>
$
$ Electrode Definition
ELECTR NAME=<name> x.min= x.max= y.min= y.max=
$
$ Specify Impurity Profiles
PROFILE N-TYPE N.PEAK= <value> x.min= x.max= y.min= y.max=
$
$ Specify Surface Oxide Charge Density at the Si-SiO2 Interface (say 3x1011 cm-2)
INTERFAC QF=3E11
$
$ Material Specifications, All parameters are set by default values, but I want to show
$ how to specify user defined values. If I want to change the band-gap of Silicon from
1.08eV
$ (default) to another value, say 1.12eV, we may write
MATERIAL REGION=<name> EG300 =1.12
$ Define Models, example Shockley Read Hall recombination with concentration
$ dependent lifetimes, concentration dependent mobility and impact ionization model
MODEL CONSRH CONMOB IMPACT.I
$
$ Solution for zero bias using Gummel algorithm
SYMB GUMMEL CARRIERS=0
METHOD ITLIMIT=20
SOLVE v<electrode>=0
$
$ Switch to Newton for high biases with two carriers
SYMB NEWTON CARRIERS=2
METHOD ITLIMIT=20
$ Save log file for I-V plot
LOG OUT.F=<filename>
$
$ Now start the solutions and save the solution file using continuation method
SOLVE CONTINUE electrod=<name> c.vstep= c.vmax= c.imax=
SAVE OUT.F=<filename>
$ END
Fig. 3.1: Example of a MEDICI program.
59
3.3 Validation: An Absolute Requirement
A major drawback with computer models is the way they are used. It is not difficult
to get wrong results from simulation. Using the wrong input parameters, a mesh which is
too coarse, or using a model outside its range of validity can each lead to incorrect
answers. Thus, the models of the physical systems, which we choose to use, cannot be
trusted without extensive validation and it is essential to subject the results of simulations
to very rigorous scrutiny. The only way to test the validity of the simulation results is by
comparison with experiments. In order to support the simulation analysis in this work,
TMA-MEDICI has been calibrated against the experimental data reported in literature.
Since a part of this thesis deals with the simulation of metal-overhang detectors (both
passivated and unpassivated), experimental data available on such structures are reported
and simulated in subsequent chapters. A very good agreement between the experiments
and simulations is found, thus validating the present effort.
60
Chapter 4
----------------------------------------------------------------
Effect of Metal-Overhang on Silicon Strip
Detectors
----------------------------------------------------------------
The contribution of Si microstrip detectors to the past and present generation

High-Energy Physics (HEP) experiments can hardly be overemphasized. Their
exploitation in future, high luminosity colliders, like LHC where the unprecedented
luminosity levels will be unleashed, however, requires some serious issues concerning
radiation hardness to be carefully considered. For example, progressive radiation damage
suffered by operating detectors influences their performance in many respects: most
notably bulk defects are introduced, acting as deep-level traps, and this eventually results
in the “type-inversion” phenomena [4.1]. This, inter-alia, lowers the Charge Collection
Efficiency (CCE) of the working detectors [4.2]. In order to compensate for the
performance degradation induced by the radiation, the detector bias voltage needs to be
progressively increased so that full depletion can be eventually attained anyway [4.3],
thus potentially leading to the occurrence of the early micro-discharges and avalanche
breakdown. Hence, one of the main aim in the development of Si detectors is to solve this
problem. Such phenomena, however, are inherently “localized” at some device critical
regions, strip-edge or surface, so that careful design strategies may help in pushing the
device operating limit farther away. In particular, with reference to Si microstrip
detectors, the adoption of overhanging metal contacts has been suggested as an effective
mean to reduce junction breakdown risks [4.4].
In this chapter, we will first briefly discuss about the techniques used in Si
detectors to improve the breakdown performance and then a report on the investigations
done so far on the Metal-Overhang (MO) analysis is presented. A computer based
analysis of Si microstrip detectors will then be discussed with the aim of investigating the
61
influence of various physical and geometrical parameters on the breakdown performance
of detectors equipped with metal-overhang.
4.1 High-Voltage Si Detectors: Techniques to Improve

Breakdown Voltage
The introduction of low-noise planar technology to the Si detectors fabrication by

Kemmer [4.5] has certainly helped in adapting Si sensors to the vastly different needs of
HEP experiments, for example, in vertex determination, in electromagnetic calorimeter
and in relativistic particle detection. However, in the planar technology, the plane-portion
of the junction is terminated by the curved regions. Electric field crowding at these
curved regions severely limits the breakdown voltage [4.6]. Several junction termination
techniques have been developed, both for power devices and for detector development.
In power devices and in IC technology, floating Field Limiting Ring (FLR) (or the
guard ring) [4.7], Junction Termination Extension (JTE) [4.8], REduced SURface Field
(RESURF) [4.9], Variation in Lateral Doping (VLD) [4.10] and Field Plate (FP) [4.11] (a
more common name of “Metal-Overhang” (MO) in power devices) are found to be
suitable.
However, for detector-grade planar Si technology, only FLR and MO have been
found to be attractive as these techniques are simple in fabrication and suitable for
vertical current flow devices. The FLR technology has already been studied extensively
by many investigators [4.12 - 4.15] and this has resulted in a considerable improvement
in the design of detectors. Multiguard structures, however, suffer from instabilities caused
by high electric fields and surface charges [4.16]. An alternative strategy to minimize the
instabilities caused by the oxide charges and improve the breakdown performance of Si
detectors is the implementation of the “Metal-Overhang” (MO) technique, proposed by
Ohsugi et al. [4.4]. The adoption of “overhanging” metal contacts helps in distributing
the electric field, reducing corner effects and thus minimizing breakdown risks. This
technique is attractive for Si strip detectors also because of its small area and minimized
dead wafer space.
62
4.2 Modulation of Electric Field by Extended Electrode: Origin
of Metal-Overhang Technique
Ohsugi et al. [4.4, 4.17, 4.18] investigated the early micro-discharge phenomena
(and also the avalanche breakdown) due to the high fields that occur along the junction-
strip edge inside Si bulk, which is potentially a serious problem in operating the Si
sensors in a high radiation environment. They found that there are three major sources of
the high fields along the strips: 1.) Bias effect - reverse bias applied to the sensor to
achieve full depletion, 2.) Metal-Oxide-Semiconductor (MOS) effect – due to the potential
difference between the external readout electrode and the implant-strip, which is a
specific problem to the AC coupled readout sensors, and 3.) Oxide charge effect – due to
the charge trapped at an interface between SiO2/Si or at defects inside the SiO2. On the
basis of this understanding of the causes of micro-discharge, they proposed some new
ideas of improved geometry. The first one was a simple idea to weaken the field strength
around the strip-edge by rounding off the implant-strip edge (by increasing the junction
depth). The second idea was to “relax” the corner field of the implant-strip by adding an
extended electrode, which was placed on the SiO2, and is called “Metal-Overhang”
(MO), having the same potential as the implant-strip. A definite advantage of the
overhang structure is also seen as reducing the effect of the oxide charge trapped in the
Si/SiO2 interface. In fact, except for the MOS effect, all these arguments were also found
to be applicable to the single-sided and/or direct-coupled sensors. The idea of the
extended-electrode was also encouraged to be applied in conjunction with the FLR
technique to suppress both the micro discharge and the junction breakdown. The work
was well supported by the experimental results on the sensors irradiated with γ−rays.
The physical explanation of this observation, in terms of potential and electric
field distribution using 2-D device simulation, was provided by Passeri et al. [4.19]
(simulated structure is shown in Fig.4.1). They studied the influence of MO width on the
strip-pitch (P) and strip-width (W). A significant increase in the breakdown voltage was
also obtained with overhanging electrodes for heavily irradiated structures, and this
behaviour was confirmed by performing experimental measurements. The usefulness of
MO equipped detector in improving the breakdown performance of heavily irradiated
63
structure was also experimentally verified by Demaria et al. [4.20]. In fact, for these
reasons, the adoption of MO has recently been assumed as a baseline design requirement
for detectors to be installed at the inner tracker [4.21] and also in the Preshower Detector
of CERN CMS experiment [4.22].
Fig.4.1: Cross-sectional schematic of the MO equipped structure used by Passeri et al.

[4.19].
Beck et al. [4.23] showed the robustness of this technique to the oxide charges,
which for detectors is relatively important and becomes more so as a result of ionizing
radiation damage. The importance of MO in improving the breakdown voltage of the Si
detectors is also boosted by the fact that it can be used in conjunction with other field
termination technique, for example with field limiting ring as reported by some authors
[4.24, 4.25].
It is worth mentioning here that the use of MO (or “field plate”) has earlier been
proposed in Ref. [4.26, 4.27] in HEP experiments. However, in the former [4.26], it is
considered as a tool to study the effect of surface charges as MO offers the possibility of
changing the surface charge density during operation of the detectors by simply changing
its potential. In [4.27], it is used for a different Si detector technology, the surface barrier
detectors (Schottky effect) in H1/HERA experiment. None of them, however, studied the
effect of MO on junction termination. Many contributions on the study of MO can also be
found in the technical literature together with analytical works for the conventional
diodes [4.11, 4.28 – 4.31]. In fact, the use of “field plate effect” in power devices dates
back to 1967, when Grove & Fitzgerald [4.28] showed that the breakdown voltage of
either p+/n or n+/p junctions can be modulated over a wide range by the application of an
64
external surface field. F. Conti & M. Conti [4.11] made a detailed analysis of the
breakdown voltage by means of an analytical mode and computer simulation for field-
plate junctions. Since then, many workers have investigated and analyzed the field-plate
technique for application in power devices and IC technology [4.29 – 4.33]. However, the
optimization procedure developed in these reports needs to be reconsidered when MO is
applied to the Si sensors in HEP experiments for the following reasons:
1.) These studies were developed over ten years ago and they were mainly committed
to the power device and IC technology. When MO is applied to the Si detectors,
changes in process parameters like substrate resistivity, implant profile,
passivation layer material and thickness etc. must be taken into account. Although
some of the works on power devices are very detailed but they suffer from the
fact that every result strongly depends on the technology used. Moreover for IC
fabrication, both n-type and p-type silicon are used, doping concentration used are
much higher and the life time of minority carriers are some orders of magnitude
lower. For silicon detector fabrication, p+-n junctions are obtained on high
resistivity substrates (NB ~ 1011 - 1012 /cm3), typically a few hundred microns
thick on n-type silicon of <111> orientation are used. Minority carrier lifetimes
are extremely high to have low leakage current.
2.) The irradiation induces modifications of the electrical behaviour of the Si sensors
and hence the optimization must take care of the effects of the irradiation also.
3.) All the breakdown voltage analysis of the junction without MO & with MO were
based on the integration of the ionization coefficients along a radial path, on the
assumption that the field is cylindrically symmetric at the edge of the planar
junction, and good agreement with the measurement has been demonstrated in the
bulk doping range (NB) ~ 1014 - 1016 /cm3 & for junction depth (XJ) > 1.0 µm.
However, it should be expected that in the practical planar structure of interest,
the presence of the planar region results in a larger depletion radius and reduced
field strength at the termination compared to the pure cylindrical junction. The
radial field approximation should consequently become less accurate as either
substrate doping concentration and/or junction depth is reduced [4.23].
65
Experimental work reported in literature clearly suggests that the MO technique is
simple and versatile and therefore, it is attractive for wide range of blocking voltages in
Si strip detectors. However, almost all the work done on the Si detectors with MO [4.4,
4.17, 4.18, 4.19, 4.23] are more concerned about the overhang design dependence on the
strip-width/strip-pitch and oxide charges. Thus, despite the significant developments in
the performance of MO technique, none of the investigations done so far have reported
the influence of the various geometrical and physical parameters taken altogether, like
oxide thickness and junction depth on the breakdown performance of MO equipped
detectors. In particular it is found that a simple, general and widely applicable design
consideration is lacking even for structures having uniform field oxide under the MO. A
complete analysis of the structure equipped with metal-overhang correlating the
breakdown voltage to these parameters is of great importance for optimization purpose.
In the present work, effect of the various parameters on the breakdown performance of
the Si microstrip detector is analyzed to achieve design optimization. The parameters are:
1) tOX, thickness of oxide below metal-overhang, 2) XJ, radius of cylindrical junction, 3)
WN, thickness of n-layer below the field oxide, 4) NB, substrate doping concentration, 5)
WMO, width of metal-overhang and 6) QF, surface charge density. A judicious choice of
these parameters is required to achieve the maximum realizable breakdown voltage. The
optimization of the various parameters is performed using two-dimensional device
simulation program, TMA-MEDICI, version 2000.4 [4.34].
4.3 Device Structure Used in Simulation
The detailed knowledge of the applied detector technology is an essential input

for correct device simulations. As already mentioned, the fabrication of detectors used in
the present study is currently in progress at Bharat Electronics Ltd. (BEL), India, to be
used in the Preshower detector of CMS at LHC, CERN and hence the parameters used in
the simulation are assumed on the basis of the technological process characteristics.
A cross-section of the simulated device structure analyzed in the present work
(Fig.4.2), consists of a two-strip subset of a single-side micro-strip array with metal-
overhang. The structure is symmetric around the center of the device so only one half of
66
the cross-section of each of the strip had to be taken into account. Such a detector is built
on an n-type, 4 KΩ-cm (NB ~1x1012 /cm3), 300 µm thick and <111> oriented Si wafer.
The p+ strips doping profile is Gaussian with a peak surface concentration of 5x1019 /cm3.
The p+-n junction is assumed to be cylindrical at its edge with the lateral curvature equal
to 0.8 times the vertical junction depth. Depletion is achieved by positively biasing the
back ohmic contact. Al electrode extending over the thicker oxide, which covers the
interstrip gap, acts as a metal-overhang. For simulating the effect of surface charge, it is
assumed that all the trapped charges are located at the Si-SiO2 interface. This typically
results in an equivalent surface charge density (QF) of the order of 3x1011 /cm2 for the
non-irradiated detector with moderately good oxides, whereas the amount of trapped
charge is expected to saturate at 1.0x1012 /cm2, even under heavy irradiation condition,
for the <111> Si orientation used in detector fabrication [4.35].
Fig.4.2: Cross-sectional schematic of two-strip subset of a Si strip detector with metal-

overhang.
67
4.4 Influence of Various Parameters on Silicon Strip Detector
Equipped with Metal-Overhang
As mentioned earlier, the parameters which determine the breakdown voltage

(VBD) of the metal-overhang structure are:
1.) tOX : thickness of oxide below metal overhang,
2.) XJ : radius of cylindrical junction,
3.) WN : thickness of n-layer below the field oxide
4.) NB : substrate doping concentration
5.) WMO : width of metal-overhang and
6.) QF : density of surface charge
These parameters can be classified into two categories: Substrate parameters and
process parameters. The substrate parameters WN and NB depend on the silicon substrate
chosen; a careful and judicious choice of this is necessary. The process parameters tOX,
XJ, QF and WMO can be tailored during process to obtain desired results. Therefore, in
order to achieve design optimization, the effects of these parameters are presented in
subsequent sections.
4.4.1 Comparison between the structures without and with metal-

overhang
A first simulation result is shown in figures 4.3 - 4.6, which compare the 2-D
equipotential distribution, surface electric field, 3-D avalanche generation and the I-V
plot for a device featuring no overhang and a 30 µm wide overhang respectively. Due to
the junction curvature the Space Charge Region (SCR) also extends laterally, however, at
the surface the electron accumulation layer underneath the oxide in structures without
MO prevents this extension. Thus, the applied voltage drops across a shorter distance
(Fig.4.3(a)). This results in the potential crowding and hence a local increase in electric
field (Fig.4.4) at the junction curvature. Hence the avalanche generation initiating
breakdown occurs close to the Si-SiO2 interface. When the overhanging contact is
present, it acts there as a p-channel MOS gate, which under the given bias conditions,
suppresses the accumulation layer and tends to deplete the underlying region. This
68
“pushes” the potential gradient away from the implant corner (Fig.4.3(b)) and hence
distributes the potential more uniformly as compared to a structure without metal-
overhang. Thus, metal-overhang reduces the crowding of the electric field within the Si
substrate by distributing it at two points: One situated near the curved portion of the
junction (‘A’ in Fig.4.2) and the second near the metal-overhang edge (‘B’ in Fig.4.2). In
this way it decreases the maximum electric field within the silicon substrate (Fig.4.4) and
thus, at a given bias, impact-ionization carrier generation is less for a MO structure
(Fig.4.5) and hence its breakdown voltage (VBD) is considerably large as compared to a
structure without MO (Fig.4.6).
Fig.4.3: 2-D potential distribution within the device for a structure (a) without meta-
overhang & (b) with metal-overhang of width 30 µm at VBIAS = 500 volt. Values of the
parameters used for simulation are: XJ = 1.0 µm, tOX =1.6 µm, WN = 300 µm,
NΒ=1x1012/cm3 & QF=0.
69
Fig.4.4: Surface electric field for a structure (a) without metal-overhang & (b) with
metal-overhang of width 30 µm at VBIAS = 500 volt. Values of the parameters used for
simulation are: XJ = 1.0 µm, tOX =1.6 µm, WN = 300 µm, NΒ=1x1012/cm3 & QF=0.
Fig.4.5(a): 3-D impact-ionization carrier generation rate for a structure without metal-
overhang at VBIAS = 500 volt. Values of the parameters used for simulation are: XJ = 1.0
µm, tOX =1.6 µm, WN = 300 µm, NΒ=1x1012/cm3 & QF=0.
70
Fig.4.5(b): 3-D impact-ionization carrier generation rate for a structure with metal-
overhang of width 30 µm at VBIAS = 500 volt. Values of the parameters used for
simulation are: XJ = 1.0 µm, tOX =1.6 µm, WN = 300 µm, NΒ=1x1012/cm3 & QF=0.
Fig.4.6: I-V characteristics for a structure without metal-overhang & with metal-
overhang of width 30 µm. Values of the parameters used for simulation are: XJ = 1.0 µm,
tOX =1.6 µm, WN = 300 µm, NΒ=1x1012/cm3 & QF=0.
71
Qualitatively, the effect of metal-overhang can be very well explained in terms of
the charge interaction that take place between (1) the positive space charge in the n-side
depletion region, (2) the negative space charge in the p+ side depletion region and (3) the
negative charge in the metal overhang, as shown in Fig.4.7.
Fig.4.7: Cross-sectional schematic of a metal-overhang structure showing the charge

distribution in the space charge region and in the metal-overhang.
For convenience, the positive space charge in the n-side depletion region is
divided into 4 parts:
(i) QPP: in the plane-parallel part of the n-side depletion region beneath the p+ junction.
(ii) QCY: in the curved portion of the n-side depletion region around the junction
curvature.
(iii) QPP-MO: in the plane-parallel part of the n-side depletion region under the metal-
overhang.
(iv) QCY-MO: in the curved portion of the n-side depletion region around the MO edge.
In the planar junction termination without MO, only the first two components (i)
and (ii) are present, whereas in the structures equipped with MO, (iii) & (iv) are also
present, which strongly modify the potential and electric field distribution.
In a planar cylindrical junction without MO, the electric flux lines emanating
from the positive QCY terminate on the negative space charge in the curved portion of the
p+ side depletion region. Since the electric flux per unit area crossing the curved portion
of the junction is more than that at the plane portion of the junction, there is a field
crowding at the curved portion. This field crowding limits the VBD of such junctions to
values much lower than that of an ideal plane-parallel junction.
72
In the planar junctions equipped with MO, the electric flux that emanates from
positive QCY terminates not only on the negative charge on the cylindrical portion but
also on the negative charge on the metal-overhang. As a consequence, the potential is
now distributed between the junction curvature and the MO edge, which in turn reduces
the field crowding and increases the breakdown voltage. However, in this case it should
be noted that the capacitive coupling between the QCY-MO and negative charge on the
metal-overhang also takes place, and hence the breakdown of this structure can take place
either at the junction curvature or at the MO edge.
In short, the space charge region on the lightly doped side (n) of the junction is
primarily responsible for causing a severe field crowding at the curved portion of the
junction when the overhang is absent. However, in the presence of overhang, this charge
interacts very strongly with the charge on the metal overhang thereby leading to the
dramatic reduction in the field crowding at the junction curvature.
4.4.2 Effect of Field-Oxide Thickness

Due to the qualitative similarity between the field crowding at the MO edge and at
the junction curvature, the thickness of the oxide (tOX) in the MO structure plays a role
similar to the junction depth. Fig.4.8 shows the plot of VBD vs tOX for XJ =1.0 µm. It can
be seen that VBD increases with increase in tOX, attains a maximum value corresponding
to a certain optimum oxide thickness tOX(OPT) and then decreases for further increase in
tOX. Thus, tOX(OPT) apparently divides the whole plot into two regions, i.e., tOX < tOX(OPT)
and tOX > tOX(OPT). In order to understand this behaviour, maximum electric field within
the device (both at the MO edge and at the junction curvature) vs. oxide thickness is
plotted in Fig.4.9.
In the region tOX > tOX(OPT), and for very thick oxides (For example, tOX = 12 µm)
the space-charge in the substrate does not interact appreciably with the charges on metal-
overhang and hence the potential crowding at the junction curvature remains same as it
would be without metal-overhang. Thus, the dominant peak electric field (Fig.4.10(a))
and the avalanche breakdown occurs at the junction curvature only.
73
4400
tOX < tOX(OPT) tOX > tOX(OPT)
4200
4000 tOX(OPT)
3800
Breakdown voltage(volt)
3600
3400
3200 XJ=1.0 micron

12 3
NB=1x10 /cm
3000
WMO = 30 microns
WN =300 microns
2800
QF = 0
2600
2400
2200
-2 0 2 4 6 8 10 12 14 16
Oxide thickness (microns)
Fig.4.8: Breakdown voltage vs. field-oxide thickness.
1.4E+05
at junction curvature
1.2E+05 ('A' in Fig.4.2)
Maximum elcetric field (volt/cm)
1.0E+05 tOX(OPT)
8.0E+04 XJ=1.0 micron

NB=1x1012 /cm3
6.0E+04 tOX < tOX(OPT) tOX > tOX(OPT) WMO = 30 microns
WN =300 microns
QF = 0
4.0E+04
2.0E+04
at MO edge
('B' in Fig.4.2)
0.0E+00
-2 0 2 4 6 8 10 12 14
oxide thickness (microns)
Fig.4.9: Maximum electric field (at the junction curvature and at the metal-overhang
edge) vs. field-oxide thickness at VBIAS = 500 volt.
74
However, as the field-oxide thickness is reduced (For example, tOX = 4 µm), the
coupling between the positive QCY and the negative charge on the metal-overhang
increases, thus “pushing” the potential crowding away from the junction curvature to the
overhang edge. This reduces the peak surface electric field at the junction (Fig.4.9).
However, at the same time, the capacitive coupling between the positive QCY-MO and the
negative charge on the MO increases, and hence an increase in the peak surface field at
the MO edge is also observed (Fig.4.9). Thus, as shown in Fig. 4.9, in the region tOX >
tOX(OPT) as tOX is reduced, there is a systematic decrease in the peak electric field at the
junction curvature with the simultaneous increase in the field at the overhang edge. Due
to this distribution, the overall field distribution becomes milder and the maximum
electric field within the Si bulk decreases, hence the breakdown voltage increases.
At optimal oxide thickness, i.e., for tOX = tOX(OPT) the field distribution is such that
the maximum electric field within the device is almost same at the overhang edge and the
junction curvature (Fig.4.10(b)), and the impact-ionization simultaneously occurs at the
two edges. Since the maximum electric field within the device (either at the junction
curvature or at the MO edge) for this case (tOX = tOX(OPT)) is less than that for any other
value of the oxide thickness (Fig.4.9), we get the maximum breakdown voltage.
For further reduction in tOX, i.e. in the region tOX < tOX(OPT), the coupling between
the QCY and the negative charge on the metal increases continuously, which “pushes” the
potential crowding further away from the junction edge, so the peak surface electric field
at the junction edge is reduced. However, at the same time the interaction between the
QCY-MO and metal charge becomes so strong that a dominant electric field now appears at
the MO edge (Fig.4.10(c)). Thus, in this region, as tOX is reduced the peak electric field at
the MO edge (and hence in the device) increases (Fig.4.9) and hence VBD decreases.
75
Fig.4.10: Surface electric field plot for a structure with metal-overhang of width 30 µm
at VBIAS = 500 volt for (a) tOX > tOX(OPT), (b) tOX = tOX(OPT), and (c) tOX < tOX(OPT).
76
This picture can also be understood by looking at the impact-ionization carrier
generation rate at VBIAS = 500 volt as shown in Figures 4.11(a), 4.11(b) and 4.11(c). It
can be seen that for the region tOX > tOX(OPT), VBD takes place at the junction curvature
(Fig.4.11(a)) and for tOX < tOX(OPT) it occurs beneath the MO edge (Fig.4.11(c)). For tOX =
tOX(OPT) (Fig.4.11(b)), it simultaneously takes place at the two edges (carrier generation is
almost same along the two paths). The field distribution is such that the generation rate is
minimum for tOX = tOX(OPT) and hence the maximum breakdown voltage is obtained for
this case.
Fig.4.11: 3-D impact-ionization carrier generation rate plot for a structure with metal-
overhang of width 30 µm at VBIAS = 500 volt for (a) tOX > tOX(OPT) & (b) tOX = tOX(OPT).
Values of the parameters used for simulation are: XJ = 1.0 µm, tOX =1.6 µm, WN = 300
µm, NΒ=1x1012/cm3 & QF=0.
77
Fig.4.11(c): 3-D impact-ionization carrier generation rate plot for a structure with
metal-overhang of width 30 µm at VBIAS = 500 volt for tOX < tOX(OPT). Values of the
parameters used for simulation are: XJ = 1.0 µm, tOX =1.6 µm, WN = 300 µm,
NΒ=1x1012/cm3 & QF=0.
4.4.3 Effect of Junction Depth

Fig.4.12 shows the plot of VBD vs tOX with junction depth (XJ) as a running
parameter. The qualitative behaviour of the VBD is same for all values of XJ. An
interesting point to note from the figure is that for the region tOX < tOX(OPT), VBD is almost
same for all values of XJ, i.e., VBD is independent of XJ whereas in the region tOX >
tOX(OPT), VBD is heavily dependent on XJ.
4000
3600
Breakdown voltage (volt)
3200
XJ=3.5 micron
2800 XJ=2.0 micron
2400 NB=1x1012/cm3
XJ=1.0 micron
WMO = 30 microns
XJ=0.5 micron
WN = 300 microns
2000
QF = 0 XJ=0.2 micron
1600
0 2 4 6 8 10 12 14
Fig.4.12: Breakdown voltage vs. field-oxide thickness with junction depth as a running
parameter.
78
To understand this nature, we have plotted the surface electric field for three
different values of XJ in figures 4.13(a) and 4.13(b). From Fig.4.13(a), it is clear that in
the region tOX < tOX(OPT) the peak surface electric field value within the value (which
occurs at the MO edge) is almost same for all values of XJ. The magnitude of the surface
electric field and the MO edge breakdown in this region, hence, depends only on the tOX
and is independent of XJ. For tOX > tOX(OPT), the avalanche takes place at the junction
curvature and hence in this situation XJ plays an important role in determining VBD. It is
known that when the value of XJ is increased, the electric flux per unit area at the curved
portion decreases and therefore the field crowding reduces. This is also featured in the
Fig.4.13(b), where it can be seen that as XJ increases, the peak surface electric field at the
junction decreases substantially. Thus, in this region VBD increases with increase in XJ.
Fig.4.13: Surface electric field plot for a structure with metal-overhang of width 30 µm
at VBIAS = 500 volt for different values of junction depth (a) tOX < tOX(OPT) and (b) tOX >
tOX(OPT)).
79
Another important feature, which is conspicuous from Fig.4.12, is that in the
region tOX > tOX(OPT), a given variation in the field oxide thickness effects a larger change
in the VBD for a shallow junction than for a deep junction. For instance, when the oxide
thickness is reduced from 12 µm to 1.6 µm, VBD increases from 2090 volt to 3660 volt (~
75%) for XJ = 0.2 µm, whereas it increases from 3160 volt to 3900 volt (~ 23%) for XJ =
3.5 µm. This is because the capacitive coupling between the positive space charge and
the negative charge on the MO is stronger for a shallow junction than for a deep junction
as the distance separating these two charges is smaller in a shallow junction. Also for a
given value of metal-overhang, tOX(OPT) required for achieving VBD depends upon the
value of XJ (Table 4.1).
Table 4.1: Breakdown voltages for different junction depths corresponding to optimized
oxide thickness for WMO=30 microns.
XJ (micron) tOX(opt) (micron) VBD(volt)
0.2 1.2 3800
0.5 1.2 3900
1.0 1.6 4015
2.0 1.6 3950
3.5 2.0 3900
Fig.4.14 shows the variation in the maximum breakdown voltage vs. junction
depth for a structure with metal overhang under the optimal conditions. For comparison,
the breakdown voltage of the structure without MO is also shown. In the absence of
metal-overhang, breakdown voltage increases when the junction depth increases. For MO
equipped structures, it can be seen that the VBD achieved for tOX(OPT) goes through a
maximum for certain optimum junction depth XJ(OPT). For XJ < XJ(OPT), the maximum
VBD increases with increase in XJ due to junction curvature effect and for XJ > XJ(OPT),
VBD decreases due to reduction in n-layer thickness below the junction. However, the
maximum VBD obtained under the optimal condition does not appreciably vary with XJ.
From Fig.4.14, it can again be seen that the beneficial effects of the MO structure
are more pronounced for the shallow junctions (as the difference in VBD between the two
80
curves is more for a shallow junction than for a deep junction). This is of great help for
HEP experiments where shallow junctions are desired in order to have small dead layer.
4500
tOX = tOX(OPT)
4000
Maximum breakdown voltage (volt)
3500
3000
without metal-overhang
2500
2000
1500
NB = 1x1012/cm3
1000
WN = 300 microns
WMO = 30 microns
500 QF = 0
0
0 2 4 6 8 10 12 14 16
Junction depth (microns)
Fig.4.14: Breakdown voltage vs. junction depth without metal-overhang and with metal
overhang of 30 µm (for optimum oxide thickness).
However, it must be pointed out that if the metal-overhang does not extend for
sufficiently large distance beyond the edge of the junction, VBD no longer remains sensitive
to tOX and the effect of metal-overhang for such structures is practically negligible. This is
evident from Table 4.2, which shows the values of VBD for different values of tOX for a
structure with XJ = 15 µm. It is clear that for XJ = 15 µm and WMO = 30 µm, VBD shows
only marginal increase (~ 7%), as the values of tOX increase by over an order of magnitude.
This is because such a small width of metal-overhang is not sufficient to provide the
flattening of equipotential lines for the large values of junction depth and consequently has
no effect on VBD.
81
Table 4.2: Variation of breakdown voltage vs. oxide thickness for XJ=15 microns and
WMO=30 microns.
tOX(microns) VBD (V)
0.4 3600
0.8 3650
1.2 3750
1.6 3750
2.0 3800
4.0 3850
12.0 3850
4.4.4 Effect of the Width of Metal-Overhang

Fig.4.15 shows the variation of VBD with oxide thickness for different values of
MO width (WMO). Again the VBD of the structure without metal-overhang is also shown
for comparison. It can be seen that VBD increases with the increase in WMO, which is due
to the flattening of the equipotential lines near the junction edge. The limiting value of
VBD is approached as soon as the equipotential lines spread out to the maximum possible
extent and then VBD becomes almost constant for further increase in WMO. However, for
microstrip detectors, strip-pitch becomes a limiting factor for deciding WMO. Interstrip
capacitance (Cint) also increases with the increase in WMO, which in turn results in
increasing noise. Hence it is necessary to optimize WMO depending upon the specification
needed for a particular detector operation. In Si Preshower detector at CMS with large
strip-pitch and device depth, back-plane capacitance plays a dominant role in the total-
detector capacitance. Hence, the increase in VBD due to increase in WMO can be taken as a
guiding factor in deciding the width of metal-overhang.
82
4000
3500 XJ=0.2 microns

NB = 1x1012 /cm 3
3000 WN=300 microns
Breakdown Voltage(volt)
QF = 0
2500
WMO = 30 microns
2000
WMO = 20 microns
WMO = 10 microns
1500
1000 Without metal-overhang
500
0
0 2 4 6 8 10 12 14 16
Oxide thickness(micron)
Fig.4.15: Breakdown voltage vs. field-oxide thickness with width of metal-overhang as

a running parameter.
The influence of the increasing the width of the metal-overhang can be perhaps
more easily understood by looking at the electric field distribution within the device near
the junction curvature and the metal-overhang edge (Fig.4.16): the action of the increase
in the width of metal overhang, which “pushes” the potential gradient and hence the field
crowding away from the implant corner can be straightforwardly appreciated there. For a
simple case of no-overhanging structures, a field peak is located at the p+ edge (Fig.
4.16(a)). A systematic decrease of electric field amplitude within the silicon substrate is
observed as WMO is increased, clearly showing the benefits of increasing the extension of
overhang. It should be noted that the effect of increasing the extension of the metal
contact significantly enhances the electric field amplitude within the SiO2. Nevertheless,
due to the much higher breakdown critical fields, discharge phenomena within the oxide
are not of practical concern, at least within the usual operating voltage range of silicon
detectors.
83
Fig.4.16: 2-D electric field distribution plot at VBIAS = 500 volt for tOX = 0.4 µm & XJ =
0.2 µm for structure with (a) no-overhang, (b) WMO = 10 µm, (c) WMO = 20 µm, (d)
WMO = 30 µm, and (e) WMO = 40 µm.
84
Also, for a given XJ, tOX(OPT) takes different values for different values of WMO as
listed in Table 4.3.
Table 4.3: Breakdown voltages for different metal-overhang widths corresponding to

optimized oxide thickness for XJ=0.2 microns.
Wmo(micron) tOX(opt) (micron) VBD(volt)
10 0.8 2660
20 1.2 3370
30 1.2 3800
4.4.5 Effect of Substrate Parameters: Device-Depth and Substrate Doping

Concentration
It is well known that for the punch through diodes, device depth (WN) primarily
determines the breakdown voltage of p+-n- -n+ junction. Fig.4.17 shows that the VBD
increases with the increase in WN for all values of tOX. The maximum VBD occurs at the
same value of tOX(OPT) for all WN.
4500
XJ=1.0 microns
4000
12 3
NB=1x10 /cm
3500 WMO=30 microns
Breakdown Voltage(volt)
QF = 0
3000
WN=300 microns
2500
2000 WN=200 microns
1500
WN=100 microns
1000
500
0
0 2 4 6 8 10 12 14
Fig.4.17: Breakdown voltage vs. field-oxide thickness with device-depth as a running

parameter.
85
Also it is known that for Punch Through (PT) diodes, substrate doping
concentration (NB) has a negligible influence on the breakdown voltage of the plane-
parallel junction [4.36]. However, in the planar devices with cylindrical termination, NB
affects the optimal conditions and the breakdown voltage. To study the effect of NB on
the planar MO terminated junctions, we have varied the doping concentration from
5x1011 /cm3 to 5x1012 /cm3. These concentrations approximately correspond to the
resistivity of about 1 – 10 KΩ-cm, which are close to the resistivity of the silicon wafers
used for Si detector fabrication. Table 4.4 lists the variation in the optimal oxide
thickness and maximum breakdown voltage as a function of NB. It can be seen that the
maximum VBD decreases only marginally and the optimal oxide thickness increases with
an increase in NB.
Table 4.4: Breakdown voltages for different values of substrate doping concentration
corresponding to optimized oxide thickness for XJ=1.0 microns (WMO = 30 µm).
NB(/cm3) tOX(opt) (micron) Maximum VBD(volt)
5x1011 1.1 4080
7.5x1011 1.3 4050
1x1012 1.6 4015
2.5x1012 1.7 3950
5x1012 1.9 3900
4.4.6 Effect of Surface Charges

Till now we have been considering the ideal condition for Si/SiO2 interface in
simulation, where the interface was assumed to be charge free. However, in practical Si
detectors, interface traps and oxide charges exist, that, in one way or another, affects the
ideal interface characteristics. The basic classification of these traps and charges are
shown in Fig.4.18 [4.37], and described below.
(1.) Interface trapped charges, located at the Si/SiO2 interface with energy states in the
Si forbidden bandgap, can exchange charges with Si in short time. However, most
of the interface charge can be neutralized by low-temperature hydrogen annealing
[4.37].
86
Fig.4.18: Charges associated with the thermally oxidized silicon [4.37]. Value of x in
SiOx lies between 1 & 2.
(2.) Oxide charges include the oxide fixed charge (QF), the oxide trapped charge (Qot),
and the mobile ionic charge (Qm). The fixed oxide charge cannot be charged or
discharged over a wide variation of surface potential and hence is “fixed”. It is
generally located within the order of 10 Å of the Si/SiO2 interface. It is generally
positive and its density depends on the oxidation and annealing conditions, and on
the Si orientation. The passage of ionizing radiation in the oxide causes the built up
of trapped charge in the oxide layers of the detector by breaking Si-O bonds. The
electron-hole pairs created in the oxide either recombine or move in the oxide
electric field: the electrons toward the SiO2/Si interface, the holes toward the
metallic contact. The electrons are considerably more mobile than the holes and are
injected into the Si bulk whereas the less mobile holes drift much more slowly and
trapped in the oxide. The trapped holes at the SiO2-Si interface constitute the
radiation induced positive oxide trapped charge (Qot). These trapped holes may also
be responsible for the increased interface trap density.
87
In simulation, surface damage can be taken into account by properly characterizing
the oxide-trapped charge and the surface recombination centers. To a first order
approximation, the damage caused by ionizing radiation can also be taken into account by
increasing the oxide charge density [4.19, 4.24]. For a p+/n Si detector, these charges are
responsible for the dense surface accumulation layer, which in turn results in narrower
depletion region along the surface compared to the situation in the planar bulk area or if
oxide charges were not present (as shown in Fig.4.19). With a constant bias applied
across the depletion region, the contraction of the depletion region at the surface leads to
an increase in the electric field in the silicon close to the Si/SiO2 interface and hence
results in the premature breakdown of the device.
Fig.4.19: Cross section of a p+-n junction showing that positive oxide charges can
increase the lateral field in the depletion layer of a p+-n junction: (a) no oxide charges
(b) oxide charges pinching the depletion layer near the Si-SiO2 interface.
Surface avalanche breakdown is considered the most common breakdown

mechanism for standard silicon detectors. In summary, the breakdown performance of the
Si detectors is greatly affected by the presence of surface charges. This prompts us to
investigate the effect of oxide charges on the optimal conditions of the MO terminated
junctions.
Fig.4.20 shows the plot of breakdown voltage as a function of surface charge
density (QF) with tOX as running parameter. It can be seen that, as expected, VBD
decreases as Qf increases for all values of tOX. However, for tOX > tOX(OPT), the
88
deterioration in VBD is comparatively sharp even for very small values of Qf. In order to
understand this behaviour, we have plotted the 2-D potential contours in Fig. 4.21(a, b &
c) and Fig. 4.22 (a, b & c) for tOX < tOX(OPT) and tOX > tOX(OPT) respectively.
4500
XJ = 1.0 micron
4000
tOX = 1.6 micron NB = 1x1012/cm3
WN = 300 microns
3500 (tOX = tOX(OPT))
WMO=30 microns
3000
2500 tOX = 0.4 microns

(tOX < tOX(OPT))
2000
1500
1000 tOX = 12.0 microns

(tOX > tOX(OPT))
500
without metal-overhang
0
0 2E+11 4E+11 6E+11 8E+11 1E+12 1.2E+12
2
Fixed oxide charge density (/cm )
Fig.4.20: Breakdown voltage vs. fixed oxide charge density with tOX as a running
parameter.
In the region tOX < tOX(OPT), and in the absence of oxide charge the avalanche is
located under the MO edge as shown by the potential crowding in Fig.4.21(a). However,
this situation is strongly modified by the presence of the oxide charge. As QF is
increased, potential crowding increases both at the junction and under the MO edge
(figures 4.21(b), 4.21(c) and 4.21(d)). At QF ~ 5x1011/cm2, simulation results indicate that
the crowding of equipotential lines at the junction curvature becomes so strong
(Fig.4.21(c)) that electric field at the junction edge exceeds its value at the MO edge, and
hence the breakdown takes place at the junction curvature. Although, a large fraction of
the reverse bias is still sustained beyond the MO edge, as is evident from Fig.4.21(c), a
89
further increase in QF causes a rapid fall towards the unguarded junction value. In
conclusion, the MO is screened from the Si bulk as the oxide charge density increases.
For tOX > tOX(OPT), breakdown takes place at the junction curvature even for QF =0
(Fig.4.22(a)). Hence the effect of increasing oxide charges results in further dense
crowding of the equipotential lines at the junction edge (figures 4.22(b), 4.22(c) &
4.22(d)). In this case the Si bulk is already partially screened from the MO by the thick
oxide, and due to the presence of oxide charges, the advantage of using MO is completely
lost. In fact the value of VBD for tOX =12.0 µm approaches that of an unguarded junction
for large values of QF as indicated in Fig.4.20.
The importance of optimization can perhaps be best appreciated from Fig.4.20. It
can be seen that for structures without MO, and for MO equipped structures with very
thick oxides, the value of VBD reduces to meager ~ 140 volt for QF = 1 x 1012 /cm2 and
the silicon sensor would no longer be able to sustain full depletion voltage (VFD).
However, optimization of field-oxide thickness presents an attractive alternative, as the
VBD for the optimized structure remains 1600 volt, well above the VFD. In fact it can be
seen from Fig.4.20 that the VBD obtained for the optimized MO structure is always
maximum.
90
Fig.4.21: 2-D equipotential distribution plot at VBIAS = 50 volt in the region tOX < tOX(OPT)
with XJ = 1.0 µm and WMO = 30 µm for (a) QF=0, (b) QF=2.5x1011 /cm2, (c)
QF=5x1011/cm2, and (d) QF = 1x1012/cm2.
91
Fig.4.22: 2-D equipotential distribution plot at VBIAS = 50 volt in the region tOX > tOX(OPT)
with XJ = 1.0 µm and WMO = 30 µm for (a) QF=0, (b) QF=2.5x1011 /cm2, (c) QF=5x1011
/cm2, and (d) QF = 1x1012 /cm2.
92
The effect of oxide charge can also be understood from Fig.4.23 in which the plot
of maximum electric field within the device (both at the surface near the junction
curvature and at the MO edge) vs. QF is shown for two different values of tOX, i.e., for tOX
< tOX(OPT) and tOX > tOX(OPT). It can be seen that as QF increases, maximum electric field at
the surface near the junction edge (ES) and at the overhang edge (EMO) increases for both
the cases. Due to the formation of accumulation layer, however, increase in ES is more as
compared to EMO. Also, it is found that increase in ES is more rapid for tOX = 12 µm than
for tOX = 0.4 µm. For instance, as QF is increased from 0 to 2.5x1011 /cm2, ES increases
from 1.6 x 104 volt/cm to 16.6 x 104 volt/cm for tOX = 12 microns, whereas it increases
from 0.9 x 104 volt/cm to 6.7 x 104 volt/cm for tOX = 0.4 microns. This indicates that for
thick oxides, MO is almost completely screened by the oxide charges even for QF =
2.5x1011 /cm2. However, the difference between the values of ES for different tOX goes on
decreasing with increased charge level, thus implying that oxide charges play an
important role in determining the breakdown voltage in the MO equipped junctions for
large values of QF. It must be pointed out that even for QF=1x1012 /cm2, the value of ES
for tOX = 0.4 µm is less than that for tOX = 12 µm, indicating that MO still helps in
improving the breakdown voltage.
3.5E+05
3.0E+05 tOX = 12 microns (tOX > tOX(OPT))

Maximum electric field (volt/cm)
tOX = 0.4 microns (tOX < tOX(OPT))

2.5E+05
2.0E+05
1.5E+05
at the surface near the
junction edge (ES)
1.0E+05
5.1E+04
at the MO edge (EMO)
1.0E+03
0.0E+00 2.0E+11 4.0E+11 6.0E+11 8.0E+11 1.0E+12 1.2E+12
2
Oxide charge density (/cm )
Fig.4.23: Maximum electric field value in the Si substrate vs. field-oxide thickness for
two different values of tOX at VBIAS = 50 volt.
93
Another important effect of the QF on VBD is conspicuous from Fig.4.24, where
the plot of maximum VBD vs. XJ is shown for two different values of QF. It can be seen
that the maximum breakdown voltage of the shallow junctions are more affected by QF as
compared to the deep junctions.
3650
3450
QF = 0
3250
3050 QF = 3x1011 /cm2
2850
2650
NB = 1x1012/cm3
WN = 300 microns
2450 WMO = 20 microns
2250
0 2 4 6 8 10 12 14 16
Fig.4.24: Maximum breakdown voltage vs. junction depth for two different values of QF .
4.5 Comparison with Experimental Work
In order to verify the simulated results and to validate the numerical accuracy of
the computer program used in this work, experimental data available in the literature
[4.11] and [4.31] on the field-plate diodes were simulated. These results are given in
Table 4.5 along with information on the salient parameters of the device structure whose
experimentally measured breakdown voltage data are compared with the values estimated
in the present work. It can be seen that there is in general a good agreement between the
present simulation and experiments, thus validating the present effort.
94
Table 4.5: Comparison of the simulation results with the experimental work.
NB(1014/cm3) XJ tOX VBD(volt) VBD(volt) (P-E)/P

(micron) (micron) Present Experiment (%)
work (P) (E)
0.30 127 137 -8
0.41 142 160 -13

8.5
0.56 183 190 -4
Ref.[4.31] 5.0
0.90 215 225 -5
1.01 4.0 2.15 553 570 -3

Ref.[4.11] 2.65 599 615 -3
4.6 Conclusions
Due to high luminosities of the future HEP colliders, Si detectors are required to
sustain very high voltage operation well exceeding the bias voltage needed to fully
deplete them. Because of its definite advantages over other termination schemes, the
“overhanging” metal contact is an attractive technique for improving the breakdown
performance of these detectors. In this chapter, punch through planar Si-microstrip
detector with metal-overhang is analyzed taking all the salient physical and geometrical
parameters into account, using a 2-D computer simulation programs. It can be concluded
that this technique provides highly effective means of increasing breakdown voltage, but
necessarily require the optimization of the various parameters.
The design of a Si microstrip detector equipped with metal-overhang involves a
proper choice of the parameters: the junction depth, oxide thickness, width of metal-
overhang, device depth, substrate doping concentration and surface charges so as to
obtain the VBD close to the maximum achievable value. It is shown that for structures
equipped with metal-overhang, maximum VBD occurs for a given set of substrate
parameters and XJ when the oxide thickness is optimized to tOX(OPT). Breakdown voltages
of most planar junction termination techniques like field limiting ring are very sensitive
95
to the variation in junction depth. A definite advantage of the metal-overhang is observed
here, as maximum breakdown voltage of the detectors equipped with metal-overhang
remains almost constant for a wide variation in junction depth in the absence of surface
charge. This feature can help in fabricating the Si sensors with shallow junctions and very
high breakdown voltages and thus minimizing the dead wafer space. However, for very
small junction depths, VBD is very sensitive to the variation in tOX, thus demanding a
critical process control.
Although, the surface-state oxide charge should be controlled in the fabrication
processes, the increase in the level of oxide-trapped charge is clearly of concern for
detectors in a radiation environment. Breakdown voltage of the silicon sensors decreases
sharply with increase in surface charge density, however, the maximum breakdown
voltage can still be obtained for the structure with the optimal oxide thickness. It is also
found that the breakdown voltage increases with increase in metal-overhang width due to
the flattening of the equipotential lines near the junction edge.
Comparison of the present computed results with several sets of experimental data
has shown a good agreement validating the present analysis and verifying accuracy of the
computer program.
96
Chapter 5
----------------------------------------------------------------
Effect of Passivation on Breakdown Performance
of Metal-Overhang Equipped Si Sensors
----------------------------------------------------------------
One of the main aims of the detector research in the High-Energy Physics (HEP)
experiments is to stabilize the long-term behaviour of Si strip detectors. However, normal
operating conditions for Si detectors in HEP experiments are in most cases not as
favourable as for experiments in nuclear physics. In HEP experiments the detector may
be exposed to moisture and other adverse atmospheric environment. It is therefore of
utmost importance to protect the sensitive surfaces against such poisonous effects. These
instabilities can be nearly eliminated and the performance of Si detectors can be
remarkably improved by implementing suitably passivated detectors. Dielectric is
commonly used as a surface passivant for this purpose both in power devices and in Si
detector technology. Since its introduction, however, semi-insulators have gained a lot of
interest in view of their application as a surface passivation material for high-voltage
metal-overhang (MO) equipped structures in power devices. Although well appreciated in
the power devices field, the use of semi-insulating film in the Si sensors for HEP
experiments has not been investigated.
In the previous chapter, we have analyzed the breakdown performance of Si
detector equipped with MO in detail. It is very interesting to compare the effects of the
two type of passivation films (dielectric vs. semi-insulator) on the breakdown
performance and the long term stability of the MO terminated Si detectors in HEP
experiments. This chapter presents the results on the effect of relative permittivity of the
passivant on the breakdown performance of the Si detectors using computer simulations.
The semi-insulator and the dielectric passivated MO structures are then compared under
97
optimal conditions. Influence of the salient design parameters such as field oxide
thickness, junction depth, metal-overhang width, device depth, substrate doping
concentration and the surface charge on the breakdown performance of these structures
are systematically analyzed, thus providing a comprehensive picture of the behaviour of
MO structures and helping in the detector optimization task.
Another important factor, which can significantly affect the long-term
functionality of the Si sensors, is the radiation damage and hence a crucial issue for the
detectors at LHC is their stability at high operating voltages. Recently an interesting
experimental result was reported by Bloch et al. [5.1] in which the breakdown voltage of
the Si detectors was found to be increased after neutron and proton irradiation. A similar
result was also found earlier by Albergo et al. [5.2]. The results of the I-V and C-V
measurements taken on some Si sensors for Preshower detector, presented in this chapter,
also verifies this result. This inspired us to investigate the effect of radiation damage on
the passivated Si sensors. Although interaction of the radiation with Si is a complex
phenomena and its detailed analysis is not yet possible, however, the simulation results
obtained in references [5.3] and [5.4] showed us a simple but effective way to realize it.
These results were also supported by experimental observations. Using the same
methodology the effect of bulk damage caused by hadron environment in the passivated
Si detectors is also simulated in this chapter by varying effective carrier concentration
(calculated using Hamburg Model [5.5]) and minority carrier lifetime (using Kraner
[5.6]).
Static measurement results on some of the irradiated Si sensors, performed at
CERN, Geneva along with the irradiation facility and measurement set-ups is also
described in this chapter.
5.1 Passivation in Si Detectors

The electrical behaviour of a Si microstrip detector can be influenced during
operation by surface contamination, high humidity and surface moisture [5.7]. Working
with Si detectors for long periods of time has shown that the reverse current and
breakdown voltage change drastically, sometimes hours after biasing a seemingly good
detector. A severe contamination of the surface (say by touching it with fingers) may lead
98
to increase of surface leakage current by orders of magnitude [5.8]. These instabilities are
believed to be caused mainly by charge density variations on the oxide surface. The
surface resistivity of the oxide changes dramatically with humidity or in dry conditions,
and this can cause potential spreading on the oxide surface [5.9]. For example, for
unprotected p+-n junction devices, under dry or vacuum condition, positive surface
charges at the Si-SiO2 interface are fully effective and as already mentioned in the
previous chapter, an accumulation layer under the oxide is created. This may lead to very
high electric field densities at the edge of the p+ implantation (junction edge) and to a
breakdown at voltages much lower than needed for full depletion of the detector. In the
other extreme case, when the negative charges collect on the oxide surface due to the
exposure of uncovered oxide surface to high humidity, an inversion layer under the oxide
is developed. This leads to an extension of the depletion zone towards the edge of the
detector; the reverse current can increase by an order of magnitude due to electron-hole
pair generation [5.10]. It is thus, crucial to control surface contamination for long term
stabilization of fully depleted p+/ n- /n+ Si detector. This is achieved by depositing the
final passivation layer over oxide of the Si detector. The need for the final passivation in
planar structures was also well emphasized by Kemmer [5.8] while introducing the planar
technology for low noise detector fabrication and suggested the use of Si3N4 as
passivation layer. Since then in the detector grade technology, dielectrics are generally
chosen as passivant. Instabilities due to the drift of sodium ions in SiO2 film can be
suppressed by depositing dielectrics (like Si3N4) [5.11] or Phospho-Silicate-Glass (PSG)
[5.12] over the oxide film.
A similar situation was also faced in high voltage power devices and IC
technology where stability in terms of breakdown voltage is essential. There also,
dielectrics were commonly used as a surface passivant for this purpose [5.13]. However,
it has been found by Matsushita et al. [5.14] that there are still some problems in the
planar process with dielectrics as passivation. They introduced a novel passivation
scheme, the Semi-Insulating Polycrystalline Silicon (SIPOS) as a surface passivant which
solved many of the problems as the film was semi-insulating and almost electrically
neutral. This technique proved to be particularly beneficial for field plate (or MO)
terminated high voltage planar junctions. Jaume et al. [5.15] showed that since semi-
99
insulator linearizes the potential at the MO edge, thus it reduces the peak surface electric
field over there and hence the edge breakdown can be largely suppressed with the use of
SIPOS. Since then semi-insulator films are being largely used in the fabrication of high
voltage planar junctions terminated with MO in power devices and IC technology [5.14-
5.17].
5.2 Radiation Damage in Si Detectors

Irradiation damage in Si detectors can be broadly categorized into surface damage
and bulk damage.
Surface damage: The passage of an ionizing radiation in Si detectors causes
accumulation of positive trapped charge and produces traps at the Si-SiO2 interface called
interface states. This results in the formation of electron accumulation layer beneath the
surface resulting in the contraction of the depletion region over there and hence causing a
premature breakdown of the device. This effect is already discussed in detail in the
previous chapter. Another major consequence of surface damage is the increase in
interstrip capacitance (Cint) and decrease in interstrip isolation. However, it has been
found that with careful design, the change in Cint can be limited to 10-20% at operational
frequencies of LHC electronics [5.18-5.20].
Bulk damage: There is a general consensus that in the hadron environment, the
viability of long-term operation and performance of Si detectors is affected mainly due to
bulk damage [5.5 & 5.21]. Bulk damage in Si detectors by hadrons is caused mainly by
Non-Ionizing Energy Loss (NIEL) interaction of primary particle with a lattice Si atom
displacing a Primary Knock-on Atom (PKA) out of its lattice site resulting in a Si
interstitial (I) and a left over vacancy (V) (Frenkel pair). These vacancies and interstitials
migrate through the Si lattice and undergo numerous reactions with each other and the
impurity atoms existing in the Si to form stable complexes. The major macroscopic
effects expected from bulk damage are [5.5, 5.22]: (a) increase in the leakage current
since the defects act as centers to increase the generation bulk current; (b) deterioration of
charge-collection efficiency (CCE) as the defects also act as the trapping centers; (c)
change in the effective carrier concentration (Neff) due to the removal of donor levels and
creation of acceptor like states, leading to the type-inversion.
100
Reverse leakage current is strongly temperature dependent, given by [5.23]:
I (T ) ∝ T 2 exp(− E g / 2k B T ) (5.1)
where T is the operating temperature in Kelvin, Eg the band-gap energy and kB the
Boltzmann constant. Thus, reverse current can be largely reduced by operating the Si
sensors at low temperature (-5 °C for CMS), and hence is not a fundamental problem to
the long-term operation. The main obstacle to the operation of Si detectors is the change
in the effective dopant concentration, which results in an increased full depletion voltage
at high hadron fluences. In order to ensure a good charge collection efficiency even after
irradiation, one of the recently developed approach is the use of oxygenated Si as starting
material, which helps in reducing full depletion voltage at high fluence. The ROSE
collaboration (Research & Development On Silicon for future Experiments) has done
extensive research in this direction [5.5, 5.21, 5.22, 5.24-5.27]. However, such an
improvement is only observed for charged hadron irradiation whereas for neutron-
induced damage the diffusion oxygenated float zone (DOFZ) silicon leads only to
benefits in connection with low resistivity Si [5.5, 5.21, 5.22, 5.24-5.27]. An alternative,
and more conventional approach of improving CCE is to increase the detector bias
voltage progressively so that the full depletion can be eventually attained anyway [5.28].
However, operating the detectors at high biases is constrained by the breakdown
phenomena. Thus, high breakdown voltage is imperative for the operation of detectors at
high neutron fluences.
5.3 Device Structure & Simulation Technique
The device structure and the parameters used in the present simulation are same
as used in the previous chapter. However, an additional passivation layer is also
incorporated in the structure, which can be either a dielectric or a semi-insulator as shown
in Fig.5.1. The semi-insulator effect as a passivant is simulated with a linear potential
distribution along the field oxide-passivation layer interface. Therefore, the solution to
the Laplace equation in the passivation layer becomes redundant. In contrast, for the
metal-overhang structure passivated with dielectric, the Laplace equation is solved within
the passivation layer also. In order to simulate a more realistic situation, the value of
101
surface charge density (QF) is kept fixed at 3x1011/cm2 (unless otherwise specified),
which corresponds to QF for the non-irradiated <111> oriented Si detector with
moderately good oxides.
Fig.5.1: Cross-sectional schematic of two-strip subset of a Si strip detector with metal-

overhang.
5.3.1 Modeling of Bulk Damage

The generation of electrically active defects and the donor removal effect result in
a fairly complicated picture of the radiation damage defects: the implementation of the
defect levels within the present device simulator is presently infeasible. Considering that
the dominant macroscopic parameter is the effective doping concentration, in Ref. [5.3]
Li et al. pursued a simplified approach; the effect of radiation damage was simulated by
simply varying Neff and the results were supported by the experimental observation. In
another work by Richter et al [5.4], this simulation approach has been successfully used;
in which, in addition to Neff, the minority carrier lifetime (τ) was also varied to take into
account the changes in reverse leakage current. The simulated results were again verified
experimentally indicating that this approach is indeed helpful in predicting the electric
102
field and the junction breakdown of the irradiated silicon detectors. In the present work
we have followed similar approach to analyze the electric field distribution and
breakdown phenomena after irradiation. Considering a detector being under continuous
neutron irradiation as will be the case for PSD at CMS, we have simulated a pre-
irradiation condition characterized by bulk doping concentration Neff = 1 x 1012 cm-2 and
τ = 1 x 10-3 s, and further damage is taken into account by varying Neff and τ.
In this work Neff is parameterized using Hamburg model [5.5, 5.25], which
includes the self-annealing effect during long periods of operation as projected in the
actual experiment.
∆N eff = N eff , 0 − N eff (φ eq , t (Ta )) = N A (φ eq , t (Ta )) + N c (φ eq ) + N Y (φ eq , t (Ta )) (5.2)
∆N eff consists of three components, a short term beneficial annealing NA, a stable damage
part Nc, and the reverse annealing component NY, which are given as [5.5, 5.25]
N c (φ eq ) = N c 0 (1 − exp(−cφ eq )) + g cφ eq (5.3)
t
N A = φ eq g a exp(− ) (5.4)
τa
1
N Y = g yφ eq (1 − ) (5.5)
1+ t /τ y
The values of different parameters used in the simulation for calculation of Neff
are given in Table 5.1 [5.5, 5.29]. This model is chosen for simulation because it best
describes actual operating scenario of LHC environment.
To incorporate the effect of increase in leakage current with fluence, we have
changed the minority carrier lifetime (τ) in our simulation package using the definition of
Kraner [5.6] as:
1 1
= + k φ eq , (5.6)
τ τ0
where τ0 is the minority carrier lifetime of the initial wafer, Φeq is the integrated fluence,
and k is the damage constant. The initial minority carrier lifetime used is 0.1 ms and
value of k used is 4 x 10-8 cm2 s-1 as given by Kraner [5.6] for a minimum ionizing
particle.
103
Table 5.1: Damage parameters used for calculation [5.5, 5.29]:
Parameters Values
Neff0 1x1012 cm-3 (for 4.2 kΩ cm), 1.6x1012 cm-3 (for 2.5 kΩ cm)
Nc0 0.7 x Neff0
c 2.5x10-14 cm-2
gc 1.5x10-2 cm-1
ga 1.8x10-2 cm-1
ta 55 hrs. at 200C, 3587 hrs. at –50C
gy 5.2x10-2 cm-1
ty 480 days at 200C, 64760 days at –50C
5.4 Comparison between Semi-Insulator vs. Dielectric

Passivation
5.4.1 Effect of Field-Oxide Thickness and Junction Depth

In the two-dimensional computer simulation studies on high voltage Si strip
detectors involving metal-overhang structure reported in literature [5.30] and also in the
previous chapter, the dielectric medium was invariably taken to be air. However, in
practice, a high-voltage device equipped with metal-overhang is protected by a suitable
passivant. To study the influence of different passivants on VBD, Fig.5.2 shows the plot of
breakdown voltage as a function of field-oxide thickness (tOX) for different passivants:
two dielectrics (εdie=3.9 and εdie=7.5), a semi-insulator and also when the device is
unpassivated (εdie=1). The qualitative nature of the curve is same in all the cases; VBD
increases with increasing tOX, attains a maximum value corresponding to certain tOX(OPT)
and then decreases for further increase in tOX. For the region tOX < tOX(OPT), breakdown
occurs at the metal-overhang edge, whereas for tOX > tOX(OPT), it takes place at the
104
junction curvature. For tOX = tOX(OPT), electric field distribution is such that the breakdown
simultaneously occurs at the two edges, and we get the maximum breakdown voltage.
4100
3800 sem i-insulator XJ=0.2 microns
3500
3200 NB = 1x1012 /cm 3

εdie=7.5
Breakdown voltage (V)
WN = 300 m icrons
2900 WMO=20 m icrons
2600 QF=3x1011/cm 2
2300
2000
1700
εdie=3.9
1400
1100 εdie=1.0
800
0 1 2 3 4 5 6 7 8 9 10 11 12 13
Oxide thickness (m icrons)
Fig.5.2: VBD vs. tOX with different passivants for XJ = 0.2 µm.
It can be seen from Fig.5.2 that VBD in the region tOX < tOX(OPT) is very sensitive to
the variation in εdie, implying that metal-overhang edge breakdown is strongly affected by
the presence of dielectric passivation layer. However, in the region tOX > tOX(OPT), VBD is
almost insensitive to changes in εdie indicating that junction breakdown is practically
independent of the dielectric passivation layer. In the region tOX < tOX(OPT), VBD
corresponding to εdie=3.9 is considerably higher than that for εdie=1.0, and VBD for
εdie=7.5 is greater than VBD for εdie=3.9, thus demonstrating that the breakdown voltage
increases with increasing εdie.
It is also clear from Fig.5.2 that the breakdown voltage obtained for the semi-
insulator passivated structure is significantly greater than that achieved using the
dielectric passivation for all values of tOX. In order to understand this behaviour, the
equipotential contours for the different cases are plotted in Fig.5.3(a)-(d). A careful
observation of the potential distribution reveals that the equipotential contours spread out
more in the dielectric medium when the εdie is larger, thus, relaxing the potential
105
crowding near the metal-overhang. Consequently, for a given voltage the electric field in
Si bulk decreases resulting in higher VBD for large εdie. Comparing the potential contours
of the dielectric and semi-insulator passivated structures, it can be seen that the semi-
insulator layer linearizes the potential inside the field oxide, alleviating the potential
crowding at the metal-overhang edge. Thus, the peak electric field at the metal-overhang
edge is tremendously reduced and the voltage handling capability of the device is
significantly improved. This can be better appreciated by looking at the electric field
distribution plot within the Si substrate (Fig.5.4 (a)-(d)). It is clear that at a given bias, the
peak electric field amplitude at the metal-overhang edge decreases as the εdie of the
dielectric passivant layer increases. In fact, for the semi-insulator passivated structure,
field crowding at the metal-overhang is completely eliminated due to the efficient
potential contours spreading and hence the breakdown voltage obtained for semi-
insulator passivated structure is maximum. In summary, junction curvature electric field
effects are reduced by the presence of MO, however, this results in the field crowding at
the MO edge. Semi-insulator layer spreads out the equipotential lines at this edge and
hence the complementary functions of these two (MO and the semi-insulator layer) can
be exploited to improve the breakdown performance of Si sensors.
Fig.5.3(a): Potential distribution near the surface at breakdown for XJ=0.2µm in the
region tOX < tOX(OPT) for εdie=1.0.
106
Fig.5.3: Potential distribution near the surface at breakdown for XJ=0.2 µm in the region
tOX < tOX(OPT) for (b) εdie=3.9, (c) εdie=7.5, and (d) semi-insulator passivated structure.
107
Fig.5.4: Electric field distribution near the surface at Vbias=500 volt for XJ=0.2 µm in the
region tOX < tOX(OPT) for (a) εdie=1.0, (b) εdie=3.9, (c) εdie=7.5, and (d) semi-insulator
passivated structure.
108
Further, it is clear from Fig.5.2 that the tOX(OPT), required to accomplish maximum
breakdown voltage, is lower if εdie of the passivant layer is larger. Thus, higher values of
εdie allows for reduction in tOX required for attaining a given breakdown voltage. It is seen
from Fig.5.2 that tOX(OPT) is still lower for the semi-insulator passivated structure than for
the dielectric passivated one. Thus, it is clear that a given VBD can be achieved at a lesser
tOX if larger εdie is used, and at a still lower tOX if a semi-insulator is used.
To study the effect of junction depth (XJ) on the breakdown voltage for different
passivants, we show in Fig.5.2 and figures 5.5(a), 5.5(b) and 5.5(c) the variation of VBD
vs. tOX for different values of XJ. It is known that when XJ increases, the electric field
crowding at the junction curvature decreases due to decrease in flux per unit area leading
to increase in VBD. A qualitatively similar behaviour is observed when the junction depth
is increased, as shown in the plots (Fig.5.5 (a)-(c)). Increasing the junction depth results,
however, in an increase in the breakdown voltage for all the passivants.
4100
(a) XJ=1.0 microns
3800 semi-insulator
3500
εdie=7.5
12 3
3200 NB = 1x10 /cm
WN = 300 microns
2900 WMO=20 microns
11 2
QF=3x10 /cm
2600
2300 εdie=3.9
2000
1700
εdie=1.0
1400
1100
800
0 1 2 3 4 5 6 7 8 9 10 11 12 13
Fig.5.5(a): Breakdown voltage vs. field oxide thickness with different passivants for
XJ=1.0µm.
109
4100
semi-insulator (b) XJ=3.5 microns
3800
εdie=7.5
3500
3200
2900
2600
2300 εdie=3.9
NB = 1x1012/cm3
2000 WN=300 microns
WMO=20 microns
1700 QF=3x1011 /cm2
1400 εdie=1.0
1100
800
0 1 2 3 4 5 6 7 8 9 10 11 12 13
4000
semi-insulator (c) XJ=15.0 microns
3800
3600
3400
εdie=7.5
3200
3000
εdie=3.9
2800 NB = 1x1012 /cm3
WN=300 microns
2600 WMO=20 microns
QF=3x1011 /cm2
2400
2200
εdie=1.0
2000
0 1 2 3 4 5 6 7 8 9 10 11 12 13
Fig.5.5: Breakdown voltage vs. field oxide thickness with different passivants for (b)
XJ=3.5 µm, and (c) XJ=15.0 µm.
110
An important characteristic of the semi-insulator passivated structure, which is
conspicuous from Fig.5.2 and figures 5.5(a), (b) & (c) is that in the region tOX < tOX(OPT),
the oxide thickness is not a very critical parameter, as VBD varies only marginally for all
junction depths. This is contrary to the MO with dielectric passivated devices where the
edge breakdown voltage changes rapidly with change in oxide thickness in the same
region.
Figures 5.6(a) & 5.6(b) depict another important feature of the semi-insulator
passivated structure. As can be seen that the optimal oxide thickness (Fig.5.6(a)) and the
maximum breakdown voltage (Fig.5.6(b)) increases with increase in junction depth for
dielectric passivated structure, whereas the maximum breakdown voltage obtained under
the optimal conditions for the semi-insulator passivated structure is nearly constant over a
wide range of junction depth (Fig.5.6(b)). Specifically, VBD changes by a marginal 4%
for a semi-insulator passivated structure, whereas it changes by as much as 27% for a
dielectric passivated structure (εdie = 7.5), when the junction depth is increased from
1.0 µm to 15 µm. This aspect of the semi-insulator makes it an extremely important
passivant for developing high-voltage Si detector with relatively shallow junctions. Also,
it is clear from Fig.5.6(a) that for all junction depths, the optimal oxide thickness for the
semi-insulator passivated structure is less than that of the dielectric passivated structure.
2.4
εdie=1.0
Optimal oxide thickness (microns)
2
εdie=3.9
εdie=7.5
1.6
(a)
1.2
semi-insulator
0.8
NB = 1x1012 /cm3
WN=300 microns
0.4
WMO=20 microns
QF=3x1011 /cm2
0
0 2 4 6 8 10 12 14 16
Fig.5.6(a): Optimal oxide thickness vs. junction depth vs. junction depth for the semi-
insulator and dielectric passivated structure.
111
4000
sem i-insulator
3800
Max. V BD (for t OX=tOX(OPT)) (volt)
εdie=7.5
3600
εdie=3.9
3400 εdie=1.0
(b)
3200
NB = 1x1012 /cm 3
3000 WN=300 m icrons
WMO=20 m icrons
2800
QF=3x1011 /cm 2
2600
2400
0 2 4 6 8 10 12 14 16
Junction depth (m icrons)
Fig.5.6(b): Maximum breakdown voltage (for tOX = tOX(OPT)) vs. junction depth for the
semi-insulator and dielectric passivated structure.
5.4.2 Effect of Metal-Overhang Width

The influence of the metal-overhang extension (WMO) on the VBD of the semi-
insulator and dielectric passivated metal-overhang structure is compared in Fig.5.7.
4000 WMO=40 microns ( semi-insulator)

WMO=20 microns ( semi-insulator)
3500
3000
12 3
NB = 1x10 / cm
WN = 300 microns
2500
XJ=0.2 microns
11 2
QF=3x10 /cm
2000
WMO=40 microns ( εdie=3.9)
1500
1000
WMO=20 microns ( εdie=3.9)
500
0 1 2 3 4 5 6 7 8 9 10 11 12 13
Fig.5.7: Breakdown voltage vs. field oxide thickness with different metal-overhang
widths for the semi-insulator and dielectric (εdie=3.9) passivated structure.
112
At a first glance, the effect of increasing the extension of the metal contact
enhances the VBD for both the cases due to the flattening of the equipotential lines near
the junction edge. However, the increase in VBD with WMO for dielectric passivated
structure is large as compared to semi-insulator passivated structure for all values of tOX.
The figure also shows that for any given value of the metal-overhang width, the VBD for
the semi-insulator passivated structure is greater than that for the dielectric passivated
structure.
It can be seen from Fig.5.8(a) that for the dielectric passivated Si strip detector,
the optimum oxide thickness increases with increase in metal-overhang extension, attains
a maximum value and then decreases with further increase in WMO.
The maximum VBD (Fig.5.8(b)) obtained under the optimal condition increases
continuously with increase in WMO. Ideally, an infinite metal-overhang extension is
required to maximize its advantage for high-voltage devices. However, in Si strip
detectors, strip-pitch becomes a limiting factor for extending the overhang beyond a
certain value and in fact, noise associated to the strip capacitance also increases with an
increase in overhang width. Here, the semi-insulator passivated structure offers a decisive
advantage over its dielectric counterpart because for such structures the maximum
breakdown voltage obtained and also the optimal oxide thickness required for
accomplishing it remain practically constant over a wide variation in metal-overhang.
1.4
NB = 1x1012 /cm 3
1.2 W N=300 microns
X J=0.2 microns
1 QF=3x1011 /cm 2
tOX(OPT) (microns)
0.8
εdie=1.0
0.6 εdie=3.9
εdie=7.5
0.4
semi-insulator
0.2
0
5 10 15 20 25 30 35 40 45
Width of metal-overhang (microns)
Fig. 5.8(a): Optimal oxide thickness vs. metal-overhang width for the semi-insulator and
dielectric passivated structure.
113
3900
semi-insulator
3600
Max. VBD (for tOX=tOX(OPT) ) (V)
3300
3000
2700 12 3
NB = 1x10 /cm
WN=300 microns
2400
XJ=0.2 microns
11 2
2100 QF=3x10 /cm
εdie=7.5
1800 εdie=3.9
εdie=1.0
1500
5 10 15 20 25 30 35 40 45
Width of metal-overhang (microns)
Fig.5.8(b): Maximum breakdown voltage (for tOX = tOX(OPT)) vs. metal-overhang width
for the semi-insulator and dielectric passivated structure.
5.4.3 Effect of Passivation Layer Thickness

Fig.5.9 shows the variation of VBD with passivation layer thickness (tpass) for
different passivants.
3850
semi-insulator
NB = 1x1012 /cm3
3350 XJ=1.0 micron
tOX=0.4microns
WMO=20 microns
WN = 300 microns
2850
VBD (volt)
QF=3x1011/cm2
εdie=7.5
2350
εdie=3.9
1850
Unpassivated
1350
0 1 2 3 4 5 6 7 8 9 10
passivation layer thickness (microns)
Fig.5.9: Breakdown voltage vs. passivation layer thickness for different passivants.
114
It can be seen that for dielectric passivated structure, VBD increases with increase
in tpass, however, this increase is gradual for higher values of tpass. The effect of increasing
tpass significantly increases the spreading of equipotential lines along the surface of the
device, reduces the electric field crowding and hence increases the VBD. However, the
semi-insulator effect is based on the linearization of the potential at the oxide/passivation
layer interface and hence VBD remains independent of tpass for semi-insulator passivated
structures.

As already mentioned in the previous chapter, the SiO2-embedded positive charge
layer causes a thin electron accumulation layer to build up at the Si-SiO2 surface. This
results in a reduction of the depletion width, increasing the electric field in the Si close to
the Si-SiO2 interface, eventually causing an avalanche breakdown in that region at much
smaller voltages. Results for the breakdown voltage as a function of surface charge
density are plotted in Fig.5.10(a)-(c) for three cases, i.e., tOX < tOX(OPT), tOX = tOX(OPT) and
tOX > tOX(OPT) respectively.
(a) tOX < tOX(OPT)
4000
semi-insulator
3500
3000
ε die=7.5 12
NB = 1x10 /cm
3
2500
ε die=3.9 WN =300 microns
XJ=0.2 microns
2000 ε die=1.0
WMO=30 microns
1500
1000
0.0E+00 2.0E+11 4.0E+11 6.0E+11 8.0E+11 1.0E+12 1.2E+12
2
Surface charge density (/cm )
Fig.5.10(a): Breakdown voltage as a function of surface charge density with different

passivants for tOX < tOX(OPT).
115
(b) tOX = tOX(OPT)
3800
sem i-insulator
3300
2800 NB = 1x1012 /cm 3

WN =300 m icrons
2300 XJ=0.2 m icrons
WMO=30 m icrons
1800
εdie=7.5
1300 εdie=3.9
εdie=1.0
800
0.00E+00 2.00E+11 4.00E+11 6.00E+11 8.00E+11 1.00E+12 1.20E+12
Surface charge density (/cm 2)
(c) tOX > tOX(OPT)

2000
semi-insulator
1750
1500
NB = 1x1012
WN =300 microns
1250
dielectrics XJ=0.2 microns
1000
WMO=30 microns
750
500
250
0
0.00E+00 2.00E+11 4.00E+11 6.00E+11 8.00E+11 1.00E+12 1.20E+12
Surface charge density (/cm2)
Fig.5.10: Breakdown voltage as a function of surface charge density with different

passivants for (b) tOX = tOX(OPT), and (c) tOX > tOX(OPT).
In the region tOX < tOX(OPT), VBD of dielectric passivated structure decreases with
increase in QF. However, in the same region, Fig.5.10(a) shows that the semi-insulator
passivated structure remains almost insensitive to the surface charge for the same
variation of QF, thus showing its superiority over the dielectric passivated structure. A
similar behaviour is also observed for the case when tOX = tOX(OPT) (Fig.5.10(b)). Here
also, only a marginal decrease in the value of the computed VBD is found with increasing
oxide charge, thus indicating that the semi-insulator structure offers nearly total-
116
immunity against the surface charge for tOX ≤ tOX(OPT), which is again an outstanding
attribute of semi-insulator passivated structures in adverse radiation conditions.
However, the situation is different for tOX > tOX(OPT) (Fig.5.10(c)), when the
breakdown occurs at the junction edge, the simulation results indicate that the breakdown
voltage assumes almost the same value after QF = 4.0x1011 /cm2 for all the dielectric
passivated structures. In fact, the VBD for the semi-insulator passivated structure also
overlaps with that of the dielectric passivated structure after QF = 6.5x1011 /cm2. Thus, in
the region tOX > tOX(OPT) and for large values of QF, VBD of the device is mainly governed
by QF irrespective of the type of passivant used. From the above discussion, it is clear
that in order to make full use of the passivant properties of the semi-insulator, the field
oxide thickness used in the fabrication of the Si strip detectors, should be kept less than
or equal to its optimal value.
5.4.5 Effect of Device Depth and Substrate Doping Concentration

We have also investigated the dependence of VBD on the depletion layer width
(which is almost equal to device depth (WN) for punch-through (PT) structures) and
substrate doping concentration (NB).
4000
3500 NB = 1x1012 /cm3 semi-insulator

XJ=1.0 micron
εdie = 7.5
3000 tOX=0.4 microns εdie = 3.9
WMO=20 microns
2500 QF=3x1011/cm2 εdie = 1.0
2000
1500
1000
500
50 100 150 200 250 300 350
WN (microns)
Fig. 5.11: Maximum breakdown voltage vs. device depth for different passivants.
117
It can be seen from Fig.5.11 that maximum breakdown voltage obtained for semi-
insulator structure is greater than that achieved for dielectric passivated structures for all
values of WN.
To study the effect of NB on the optimal conditions, we have varied the doping
concentration from 1 x 1012 to 1 x 1013 /cm3. Fig.5.12(a) and Fig.5.12(b) show the
variation in the optimal oxide thickness and breakdown voltage as a function of NB
respectively for the dielectric and semi-insulator passivated structures. It can be seen that
the variation of both the optimal oxide thickness and the breakdown voltage is smaller for
semi-insulator passivated structure. This shows that the semi-insulator passivated
structures can be employed in a wide range of Si detectors (different resistivities and
device depth) used in the HEP experiments.
1.7 XJ=1.0 micron

WN = 300 microns
WMO=20 microns
11
QF=3x10 /cm
2
εdie=7.5
optimal oxide thickness (microns)
1.5
(a)
1.3
semi-insulator
1.1
0.9
0.7
0.0E+00 2.0E+12 4.0E+12 6.0E+12 8.0E+12 1.0E+13 1.2E+13
3
Substrate doping concentration ( /cm )
Fig.5.12(a): Optimal oxide thickness voltage vs. substrate doping concentration for two
passivants (a dielectric and a semi-insulator).
118
4050
3800 semi-insulator
Maximum VBD (volt)
3550 (b)
XJ=1.0 micron
3300
WN = 300 microns
WMO=20 microns
3050 11 2
QF=3x10 /cm
2800
2550 εdie=7.5
2300
0.00E+00 2.00E+12 4.00E+12 6.00E+12 8.00E+12 1.00E+13 1.20E+13
3
Substrate doping concentration (/cm )
Fig.5.12(b): Maximum breakdown voltage vs. substrate doping concentration for two
different passivants (a dielectric and a semi-insulator).
5.4.6 Effect of Bulk Damage on Full Depletion and Breakdown Voltage

The depletion voltage required to operate a Si detector is directly proportional to
Neff, and is given as:
2
| N eff | q WN
VFD = (5.7)
2ε S i
For Preshower, the fluence profile per year integrated with time [5.29] along with
the minority carrier lifetime (using Kraner’s definition [5.6]) is given in Table 5.2. In
Fig.5.13, we have plotted variation of Neff and VFD as a function of time and fluence
during the 10 years of LHC operation for two initial resistivities of 4.0 KΩ cm (Neff =1.0
x 1012 cm-3) and 2.5 KΩ cm (Neff =1.68 x 1012 cm-3) using equations (5.2) and (5.7). The
values of Neff considered here correspond to the various levels of the fluence expected
over the full LHC operation. It is considered that the Si detectors will operate at –5oC for
10 years of operation with the exception of 2 days/year at room temperature needed for
detector maintenance [5.29]. From Fig.5.13 it can be seen that the Si detector with initial
resistivity of 4.0 KΩ−cm becomes intrinsic after about 3 and a 1/2 year (after which type-
inversion starts), whereas for 2.5 KΩ−cm resistivity wafer the type inversion is slightly
119
delayed (intrinsic after about 4 years). Also, it is clear that VFD approaches 244 volt for
4.0 KΩ−cm wafer, and is marginally lower for 2.5 KΩ−cm wafer after 10 years.
Table 5.2: Fluence profile (along with Neff and τ) of neutrons expected for PSD detectors.
Year Fluence (each ear) x Integrated fluence Neff x 1011 Minority carrier
1013 (n/cm2) x 1013 (n/cm2) (/cm3) lifetime (ms)
1 0.2 0.2 9.16 0.01111
2 0.6 0.8 7.19 0.00303
3 1.2 2.0 3.68 0.00123
4 2.5 4.5 -2.62 0.00055
5 2.5 7.0 - 8.18 0.00036
6 2.5 9.5 - 13.5 0.00026
7 2.5 12.0 - 18.8 0.00021
8 2.5 14.5 - 24.3 0.00017
9 2.5 17.0 - 29.9 0.00014
10 2.5 19.5 - 35.6 0.00013
Fig.5.13: Variation of effective carrier concentration (Neff) and full depletion voltage as
a function of time and fluence during the 10 years of LHC operation.
120
Fig.5.14(a) and Fig.5.14(b) show the plot of 2-D electric field distribution for the
dielectric passivated structure. It is clear that before type-inversion the depletion region
spreads from the front side however, after type-inversion it grows from the rear side
indicating that the main junction has shifted from the front side to the backside. Also,
after type-inversion, another dominant electric field also develops around the curvature of
the front p+/p junction (Fig.5.14(b)). Thus, two dominant peak electric fields are obtained
after type-inversion, one at the back junction (p-/n+) and other at the front junction (p+/p-).
Fig.5.14: Simulated electric field distribution within the detectors at an applied bias of
35V for two values of Neff : (a) before type inversion (n-type), Neff =1 x 1012 cm-3, and (b)
after type inversion (p-type) Neff = -8.18 x 1011 cm-3.
121
Fig.5.15 shows the plot of breakdown voltage as a function of Neff for a dielectric
(εdie =7.5) and semi-insulator passivated structures. It can be seen that breakdown voltage
continuously increases with increasing fluence. In order to understand this behavior,
figures 5.16(a)-(d) show the three-dimensional electric field distribution within the device
for progressive radiation.
p-type (after type-inversion) n-type (before type-inversion)

4050
semi-insulator
3800

3550
3300
εdie=7.5
3050
XJ = 1.0 micron
WMO=20 microns 2800
WN = 300 microns
11 2
QF=3x10 /cm
2550
2300
-4.0E+12 -3.0E+12 -2.0E+12 -1.0E+12 0.0E+00 1.0E+12 2.0E+12
3
Effective doping concentration (/cm )
Fig.5.15: Breakdown voltage vs. Neff for different passivants.
In the p+/n/n+ detector, the electric field before type-inversion is strongest at the
front p+ strip junction and breakdown occurs over there (either at the junction curvature
or at the metal-overhang edge). It can be seen that as Neff is decreased from 1 x 1012 cm-3
(Fig.5.16(a)) to 1.5 x 1010 cm-3 (Fig.5.16(b)), the peak electric field at the front junction
decreases due to the effective spreading of the potential at the junction curvature. Thus
VBD increases, as device becomes less n-type before type-inversion.
122
Fig.5.16: 3-D electric field distribution within the detectors at an applied bias of 500V
for different values of Neff: (a) before type inversion (n-type), Neff =1 x 1012 cm-3, (b)
intrinsic, Neff =1.5 x 1010 cm-3 & (c) after type inversion (p-type), Neff = -8.18 x 1011 cm-3.
123
Fig.5.16(d): 3-D electric field distribution within the detectors at an applied bias of
500V for Neff = -3.56 x 1012 cm-3 after type inversion (p-type).
As the device is inverted to p-type (Fig.5.16(c)), the depletion layer spreads from
the rear side. As a consequence of homogeneous irradiation, the field is higher near the
back junction and since the back junction is plane-parallel the field is very uniform also.
This, in turn, results in the smooth down of local field peak at curvature of front junction.
An important feature of Fig.5.16(c) is that after type-inversion, although the main
junction is at the rear side, the maximum electric field and hence the avalanche
breakdown still occurs at the front side. This is because of the difference in the nature of
two junctions: front junction has a curvature whereas back junction is plane parallel and
for a given bias, electric field at the curved junction is greater than the electric field at the
plane parallel junction. Thus, bulk inversion results in the reduction in the potential
crowding at the front junction (and hence within the silicon bulk), which in effect leads to
an improvement in the breakdown performance.
This is also clear from the maximum electric field vs. Neff plot as shown in
Fig.5.17, wherein it can be seen that as the detector is inverted and becomes
progressively p-type, the peak electric field at the front junction (and hence in the Si
substrate) decreases at a given bias, thus improving the breakdown performance of Si
detectors with radiation.
124
p-type (after type-inversion) n-type (before type-inversion)
1.6E+05
Maximum electric field (volt/cm)

1.1E+05 XJ = 1.0 micron
at front junction
WMO = 20 microns
WN = 300 microns
11 2
QF=3x10 /cm
6.0E+04
at back junction
1.0E+04
-4.0E+12 -3.0E+12 -2.0E+12 -1.0E+12 0.0E+00 1.0E+12 2.0E+12
3
Effective doping concentration (/cm )
Fig.5.17: Variation of maximum electric field at the front and back junction with Neff at
an applied bias of 500 V.
It can also be seen from Fig.5.15 that VBD of semi-insulator passivated structure is
less sensitive to Neff as compared to dielectric passivated detectors. This can be attributed
to the fact that in semi-insulator structures, the potential distribution is more uniform
before type-inversion also and the field crowding at the front junction edge beneath the
metal-overhang is already alleviated. Another important result is that for all values of
Neff, maximum VBD for semi-insulator structure is greater than that of the dielectric
passivated ones.
In order to support the simulation analysis performed for the passivated structures,
the simulator has been calibrated against the experimental data [5.16 & 5.31]. For clarity,
cross-sections of the structures given in these references are shown in Fig.5.18(a) [5.16]
and Fig.5.18(b) [5.31]. The results are given in Table 5.3 along with the information on
125
salient device parameters. A very good agreement between the experiments and
simulations is found, thus validating our present effort.
Fig.5.18: Cross-sections of the structures simulated for experimental verification. (a)

corresponds to structure given in Ref. [5.16] and (b) is for Ref. [5.31]. Figures are not to
scale.
126
Table 5.3: Comparison of the simulation results with the experimental work
Resistiv Collector-base tOX Field plate- Present Exp.data (E) (P-E)/P
ity junction (µm) stop channel simulation (V)
(ρ) distance (∆L) result (P)
(Ω-cm) (µm) (V)
14 µm 60 830 880 -6%
60 (QF=3x1011 1.25 110 1005 1045 -4%
[5.16] /cm2) 210 1080 1065 1%
14 µm 80 1170 1150 2%
75 (QF=3x1011 1.25 110 1200 1245 -4%
[5.16] /cm2) 210 1540 1500 3%
ρ XJ Semi-insulator Nitride
(Ω- (µm) (SIPOS) passivated passivated
cm)
QF VBD VBD (P-E)/P QF VBD VBD (P-E)/P
(x1011) (P) (E) (x1011) (P) (E)
(/cm2) (V) (V) (/cm2) (V) (V)
70 0.5 1260 1200 5% 0.75 1115 1190 -6%
50 (tOX=1 1.0 1305 1390 -6% 2 1440 1420 1%
[5.31] µm) 4.5 1530 1500 2% 3.4 1560 1630 -4%
127
5.6 Static Measurements on Irradiated Si Sensors
The static measurements (Current-Voltage (I-V) and Capacitance-Voltage (C-V))

were performed at CERN, Geneva, on five irradiated Si microstrip detectors of different
designs and coming from four manufacturers. Measurements on these detectors were
carried out earlier, both before and after irradiation, by Dr. Anna Peisert, CERN. We
have compared our measurements with the earlier set of measurements to analyze the
improvement in I-V characteristics and study the variation in full depletion and
breakdown voltage.
5.6.1 Irradiation Facility

The Irradiation of all the detectors were carried out at the CERN Proton
Synchrotron (PS). The damage factor for this beam is 0.5 to 0.6, i.e., the equivalent 1-
MeV-neutron fluence is about half the 24 GeV-proton fluence. The period needed to
achieve the fluences between 2.8x1014 p/cm2 and 3.2 x 1014 p/cm2 (error ~ 6% [5.32])
was about 6-10 days. During irradiation, the detectors were kept under realistic operating
conditions, cooled to a temperature of about –7°C, with an applied bias of 150 volt.
Sensors were stored at temperatures below –3°C after the exposure and taken out into
ambient conditions only for measurements.
5.6.2 Measurement Set-ups
Set-up used for measuring the static characteristics (I-V and C-V) of the sensors is
shown in Fig.5.19. Total current and capacitance measurements were performed with High-
Voltage (HV) source (Keithley 237), which also acts as the current meter and LCR meter
(HP4284A) (along with an isolation box) connected to a PC running LabVIEW program.
128
Fig.5.19: I-V & C-V setup at CERN
Total detector current (I) was measured by connecting all the 32 strips together as
shown in Fig.5.20(a). Keithley 237, which can supply up to 1100 volt, biases the detector
and also measures the current. Non-irradiated sensors were measured at room
temperature and the irradiated ones at temperature below 00C. For irradiated detectors
current is calibrated at -5°C using the relation
I = a ebT,
where values of a and b were already calculated using current vs temperature
measurement for different fluences.
For total capacitance, the measurements were carried out using HP4284 LCR Meter
using Keithley 237 HV source to bias the detector (Fig.5.20(b)). An HV isolation box was
used to decouple the HV from the LCR meter. For the measurement, AC frequency of 100
kHz was used for non-irradiated detectors and 5kHz for the irradiated ones with 30 mV AC
amplitude.
129
Fig.5.20: Test set-up for measurement of (a) total current and (b) total capacitance.
5.6.3 Measurement Results on Irradiated Sensors

The detectors used in the study along with device depth and irradiation fluence
are listed in Table 5.4.
Table 5.4: Tested detectors, their depths and the irradiation flux to which they were
exposed.
Detector Depth (µm) Fluence (p/cm2)
India 2000-1 300 3 x 1014
Elma RTS 12-24-21 282 3 x 1014
Taiwan 1-3 310 2.8 x 1014
Greece N4 384 2.8 x 1014
Taiwan H-8044-4 325 3.16 x 1014
Figures 5.21(a) & 5.21(b) show an example of the I-V & C-V characteristics of
one of the sensors fabricated at Bharat Electronics Limited (BEL). It can be seen that
before irradiation, detector has a very low leakage current at small voltages and the
breakdown occurs at about 60-70 volt. Here, we define the breakdown voltage as a
voltage at which the leakage current shows a sharp increase, accompanied by a rapid
130
increase in the capacitance (or a decrease of 1/C2 Fig.5.21(b)). When the measurement
was performed after irradiation, the leakage current increases manifold, by an order of 3-
4. However, an interesting fact to notice is that the current behaviour is now stable up to
400 volt without any signs of breakdown. In the subsequent measurements, when the
detectors were kept at cold temperature, the improvement both in terms of leakage
current and breakdown voltage is observed. The value of full depletion voltage is
deduced from the 1/C2 vs. VBias plot. It can be seen that immediately after irradiation, VFD
increases from 50 volt (before irradiation) to 210 volt indicating the occurrence of bulk-
inversion. The decrease in the value of VFD in the next measurements after irradiation
signifies that beneficial annealing has taken place. In fact, its value has come down to
145 volt in the measurement taken on 2nd May 2001.
Fig.5.21(a): Leakage current of an Indian detector as a function of applied bias.
131
Fig.5.21(b): Inverse square capacitance of an Indian detector as a function of applied
bias. The measured full depletion voltage is also indicated.
Similar I-V behaviour is also observed for the detectors (figures 5.22(a)-(c)) from
other foundries. These results, thus, further validate the improvement in the breakdown
performance of Si sensors after irradiation.
Fig. 5.22(a): Leakage current of a Taiwan detector as a function of applied bias.
132
Fig. 5.22(b): Leakage current of the 2nd Taiwan detector as a function of applied bias.
Fig. 5.22(c): Leakage current of another Indian detector as a function of applied bias.
5.7 Conclusions
In this chapter, the application of the 2-D device simulation to the analysis and
comparison of the dielectric and semi-insulator passivated metal-overhang structure has
been described. By analyzing simulation results, influences of all the salient physical and
geometrical parameters on these structures have been elaborated.
It is demonstrated that higher values of relative permittivity (εdie) of the passivant
dielectric play an important role in determining VBD, and results in an increase in
breakdown voltage as compared to the unpassivated detector. Also, VBD increases with an
133
increase in εdie in the region tOX < tOX(OPT) for dielectric passivated structures. However,
semi-insulator passivated structure results in still higher values of VBD for all values of
tOX due to the better distribution of equipotential lines under the same conditions. It is
found that the optimal oxide thickness decreases with increase in εdie for dielectric
passivated structure and for the semi-insulator passivated structure, the decrease is still
greater. For semi-insulator passivated structure, the maximum breakdown voltage
achieved for the optimal field oxide thickness remains fairly constant over a wide
variation in the junction depth. Thus, the present study shows that the semi-insulator
passivated structures allow for a design of Si strip detectors with shallow junctions and
thinner oxides, reducing dead layer and making the detectors more suitable for high-
energy physics experiments.
Also, for semi-insulator passivated structures, the maximum VBD obtained under
the optimal conditions is found to be independent of the overhang width. The small
values of metal-overhang would help in the design of high voltage and low noise Si strip
detector, since increasing the metal extension results in higher interstrip capacitance and
hence the noise associated with it. VBD of semi-insulator passivated structure is also
found to be independent of tpass, whereas that of dielectric passivated structure increases
with increase in tpass.
Effect of the bulk damage in Si sensors shows that the increase in fluence results in
the smooth down of the local field peak at the front side of the single sided p+/n/n+
detector (where the detector breakdown occurs) and hence an improvement in the
breakdown performance is observed.
Another very important feature of the semi-insulator passivated structure is the
nearly constant breakdown voltage for a wide variation in QF for tOX ≤ tOX(OPT), this
establishes its supremacy and offers an extremely important design flexibility when
realizing high-voltage junctions for Si strip detector. Thus, the present study shows that
semi-insulator passivated structures are attractive for achieving high breakdown voltages
of Si strip detectors.
The measurements performed on irradiated detectors show an improvement in
terms of breakdown performance after irradiation. After irradiation, the depletion voltage
and total leakage current of irradiated detectors are decreasing with time over the period.
134
Chapter 6
----------------------------------------------------------------
Comparison of Junction Termination Techniques
for High- Voltage Si Sensors: Metal-Overhang vs.
Field Limiting Ring
----------------------------------------------------------------
In Chapters 4 and 5, we have analyzed the influence of salient design parameters

on the breakdown performance of silicon strip detectors equipped with metal-overhang
(MO). Another attractive junction termination technique (Chapter 4), which is widely
employed in Si detectors, is the floating Field Limiting Ring (FLR) (or guard ring)
technique. Since high-voltage planar Si junctions are of great importance in High-Energy
Physics (HEP) experiments, it is very interesting to compare these two techniques for
achieving the maximum breakdown voltage under optimal conditions. Although both the
structures have been investigated separately, yet none of the earlier works have reported
the comparative analysis of these structures under similar physical and geometrical
conditions. The purpose of this chapter is to compare these two termination techniques
under identical conditions with the aim of defining layouts and technological solutions
suitable for the use of Si detectors in harsh radiation environment. The results
demonstrate the superiority of metal-overhang technique over field limiting ring
technique for planar shallow-junction high-voltage Si detectors used in high-energy
physics experiments. Before discussing the results, however, a brief description of the
FLR technique is provided.
6.1 FLR structure
The effect of crowding of the field lines in the proximity of the junction
termination can be reduced by means of a floating field limiting ring (FLR) [6.1- 6.6].
135
The FLR (or guard ring) structure produces a shaping of the lateral spread of space
charge region, reduces the potential gradient along the silicon surface and thus limits the
field intensity. It consists of diffused region (called guard ring) that is isolated from the
main junction but sufficiently close to it. The guard ring is left floating and is free to
adopt any potential during device operation. In Fig.6.1 the effect of a floating guard ring
on the electric field crowding is exhibited. When a reverse bias is applied, the depletion
layer is initially associated with the main junction and as the bias is increased it extends
outward around the guard ring. The spacing between the main junction and the guard ring
is such that the punch through occurs before the avalanche breakdown of the cylindrical
junction associated with the main junction. Thus, the maximum electric field is limited;
any further increase in the reverse voltage is taken up by the ring.
Fig.6.1: Comparison of the electric field crowding for a planar junction (a) without
and (b) with a floating field limiting ring.
The effect of FLR, however, becomes insignificant when the spacing between the
main junction and the guard ring is too large or too small [6.7], hence, the FLR spacing
must be optimized to get maximum breakdown voltage. A simple, but effective criterion
to optimize the design of complex multiple floating guard ring structure is proposed in
Ref. [6.7]. The criterion is very robust in the sense that it is applicable to a wide variation
of various physical and geometrical parameters. However, it has been observed that the
accumulated surface charge due to radiation damage affect the long-term stable
behaviour of the FLR structure [6.8]. Also, as mentioned in chapter 5, the surface of the
oxide layer is generally exposed to changing environmental conditions, resulting in the
gradual build-up of charges and thus, changing the potential distribution in the guard
structure with time.
136
6.2 Device Model
The simulations are performed on devices of smaller dimensions leading to reduced

simulation time, still allowing for the study of desired effects. The FLR and MO
structures investigated in the present work are shown in figures 6.2(a) and 6.2(b)
respectively. In the FLR structure (Fig.6.2(a)), GS is the spacing between the main
junction and the guard ring, and GW is the width of the field ring window. In the MO
structure (Fig.6.2(b)), tOX is the field oxide thickness and WMO is the width of the metal
extension.
Fig.6.2: Cross-sectional schematic of a Si detector (a) equipped with FLR, and (b)
equipped with MO. Points marked as ‘A’ and ‘B’ in the figures corresponds to the peak
electric field within the Si substrate for the two structures. Figures are not to the scale.
137
In the present simulation, the values of GW (for FLR structure) and WMO (for MO
structure) are kept fixed at 30µm (unless otherwise specified). Such a detector is built on
an n-type, 300µm thick and <111> oriented Si wafer. In both the structures, p+ region
doping profiles are taken to be Gaussian with a peak concentration of 5x1019/cm3 at the
surface and the substrate doping NB is taken to be uniform. It is assumed that lateral
diffusion depth at the curvature of the p-region is equal to 0.8 times the vertical junction
depth (XJ). These profiles are assumed to be identical for both the structures. In all the
simulations, we consider the ideal condition where the detector is free from localized
defects, pin-holes etc. and the breakdown occurs either at the junction curvature or at the
metal-overhang edge.
6.3 Comparison of FLR and MO Structures

To optimize the design of Si strip detectors, various physical and geometrical
parameters have to be taken into consideration, viz., substrate doping concentration (NB),
junction depth (XJ), and oxide charge density (QF). The optimal design of a FLR structure
in addition depends on ring-to-ring spacing, ring doping profiles & width of each ring
and the optimal design of MO structure in addition depends on the width of the metal
extension and oxide thickness. All of these parameters strongly affect the electric field,
leading to a complex design through many trial geometries. For the sake of present
comparative work, the FLR is optimized with respect to guard ring spacing (GS) as given
in [6.7], and field oxide thickness is optimized for MO structure as in Chapter 4. For both
the structures under optimal condition, effect of the various physical and geometrical
parameters on the breakdown voltage is also analyzed.
6.3.1 Effect of Guard Ring Spacing (for FLR structure) and Field-Oxide
Thickness (for MO structure)
To optimize the guard ring spacing (GS) in the FLR structure, the spacing is
varied between 10 µm and 80 µm in steps of about 10 µm. Fig.6.3 shows the plot of
breakdown voltage (VBD) as a function of guard ring spacing (GS) for FLR structure.
It can be seen that VBD increases with the increase in GS, attains a maximum
value corresponding to certain GS(OPT) and then decreases for further increase in GS. For
138
GS < GS(OPT), avalanche occurs at the curvature of the guard ring (‘B’ in Fig.6.2(a)) and
for GS > GS(OPT), it takes place at the curvature of the main junction (‘A’ in Fig.6.2(a)).
At optimal condition, i.e., GS = GS(OPT), impact ionization simultaneously occurs at the
curvature of the two junctions (at the main junction and the guard ring).
Fig.6.3: Breakdown voltage vs. guard ring spacing (for FLR structure) and field oxide
thickness (for MO structure).
Fig.6.3 also shows the variation of VBD with field oxide thickness (tOX) for MO
structure. The qualitative nature of the behaviour of VBD is similar as in case of FLR.
Here the optimized value of tOX, i.e., tOX(OPT) corresponds to the simultaneous occurrence
of breakdown at the junction curvature (‘A’ in Fig.6.2(b)) and the metal-overhang edge
(‘B’ in Fig6.2(b)). It can be seen from Fig.6.3 that in the region tOX < tOX(OPT), the VBD
changes more sharply with change in tOX as compared to the variation of VBD with GS for
GS < GS(OPT), thus demanding critical process control in MO structures. However, it is
clear from the comparison that the maximum breakdown voltage obtained under the
optimal condition is greater for MO structure than the FLR one. This can be better
appreciated by looking at the two-dimensional electric field distribution plot shown in
figures 6.4(a) and 6.4(b) for the FLR and MO structures respectively. These distributions
are obtained under optimal condition at VBIAS = 500 volt. It can be seen that at a given
bias, the peak electric field amplitude (Emax) within the device for the optimized FLR
structure is greater than the Emax for the optimized MO structure, and hence the observed
behaviour of breakdown voltage for the two structures.
139
Fig.6.4: Two-dimensional electric field distribution plot within the device at VBIAS = 500
volt for (a) optimized FLR structure, and (b) optimized MO structure.
6.3.2 Effect of Junction depth

Fig.6.5(a) shows the plot of VBD vs. GS for FLR case and Fig.6.5(b) shows the
plot of VBD vs. tOX for MO structure for different values of junction depth (XJ). For FLR
structure, VBD increases with increase in XJ for all values of GS. This is because the
electric flux per unit area at the junction curvature decreases with increase in XJ thus,
increasing the VBD. However for MO technique, the metal-overhang edge breakdown
140
which takes place for tOX < tOX(OPT) is almost insensitive to the variation in XJ, whereas
the breakdown voltage for tOX > tOX(OPT) depends on the changes in XJ. For each value of
junction depth, values of GS and tOX are optimized.
1600
For FLR structure
1400
1200
XJ = 15 microns
1000
WN = 300 microns
800 NB = 1x1012/cm3
GW = 30 microns
600 QF = 2x1011/cm2
400
XJ = 3.5 microns
200 XJ = 1.0 microns
XJ = 0.2 microns
0
0 10 20 30 40 50 60 70 80 90 100
Guard ring spacing (microns)
Fig.6.5(a): Breakdown voltage vs. guard ring spacing (for FLR structure) with junction
depth as a running parameter.
1600
For MO structure
1400
XJ = 15 microns
1200
1000
XJ = 3.5 microns
600
XJ = 0.2 microns
WN = 300 microns
400 NB = 1x1012/cm3
WMO = 30 microns
200 QF = 2x1011/cm2
0
0 2 4 6 8 10 12
Fig.6.5(b): Breakdown voltage vs. oxide thickness (for MO structure) with junction
depth as a running parameter.
141
Figures 6.6(a) and 6.6(b) show an important feature of the metal-overhang
structure in which we have plotted GS(OPT) & tOX(OPT) vs. XJ and maximum VBD vs. XJ
respectively.
70 7
Optimal guard ring spacing (microns)

60 WN = 300 microns 6
NB = 1x1012/cm3
50 QF = 2x1011/cm2 5
40 4
Optimal oxide thickness
( for MO structure; WMO=30
30 3
microns )
20 2
10 1
Optimal guard ring spacing
( for FLR structure; GW=30 microns)
0 0
0 2 4 6 8 10 12 14 16
Fig.6.6(a): Optimal oxide thickness and optimum guard ring spacing vs. junction depth
for the MO and FLR structure respectively.
1600
1400
For MO structure
1200
1000
800 WN = 300 microns

For FLR structure
NB = 1x1012/cm3
600
QF = 2x1011 /cm2
400
200
0
0 2 4 6 8 10 12 14 16
Fig.6.6(b): Maximum breakdown voltage obtained under optimal conditions vs. junction
depth for the FLR and MO structure.
142
It can be seen from Fig.6.6(a) that the variation of tOX(OPT) with XJ is smaller than
that of GS(OPT) with XJ. In fact GS(OPT) increases continuously with the increase in XJ.
Also, it is clear from Fig.6.6(b) that the maximum VBD of the MO technique is only
weakly dependent on the junction depth as compared to the FLR structure under the
optimal conditions. Specifically, maximum VBD increases from 950 V to 1360 V (30%)
for the MO structure whereas it increases from 200 V to 1350 V (85%) for FLR
technique as XJ is varied from 0.2 µm to 15 µm. Also, for shallow junctions, the MO
technique offers a higher breakdown voltage than the FLR. This aspect of MO makes it
an extremely useful technique for developing high-voltage Si detectors as this would
permit the realization of high breakdown voltage with relatively shallow junctions, thus
offering the much desired process compatibility. This inherent merit of MO design can be
further appreciated by noting that the FLR and most other junction termination
techniques vitally depend on increasing the junction depth for improving the breakdown
voltage. However, for large values of XJ such as XJ=15 µm, the VBD of the two structures
assume almost the same value.
6.3.3 Effect of Relative Permittivity of Passivant

To see the influence of relative permittivity (εdie) of passivant dielectric on VBD of
the FLR and MO structures, maximum VBD and optimal values of GS and tOX are plotted
against εdie in figures 6.7(a) & 6.7(b) respectively. The structure is simulated for two
passivants: SiO2 (εdie=3.9), Si3N4 (εdie=7.5), and also when the device is unpassivated
(εdie=1). The qualitative nature of the variation of VBD with GS and tOX is same in all the
cases.
It can be seen from Fig.6.7(a) that for both FLR and MO structures, the VBD
corresponding to εdie=3.9 is higher than that for εdie=1.0, and the VBD for εdie=7.5 is
greater than VBD for εdie=3.9, thus demonstrating that εdie plays an important role in
determining VBD of cylindrical junction. However, the increase in the maximum VBD for
MO structure is greater than that of FLR design.
143
1600
For MO structure
(WMO=30 microns)
Maximum breakdown voltage (V)
1400
1200
WN = 300 microns
NB = 1x1012/cm3
XJ = 3.5 microns
1000 QF = 2x1011/cm2
For FLR structure (GW=30

microns)
800
600
0 1 2 3 4 5 6 7 8
Relative permittivity of passivant
Fig.6.7(a): Maximum breakdown voltage obtained under optimal conditions vs. relative
permittivity of the passivant for the FLR and MO structure.
35 4.5
For FLR structure (GW=30 microns)
4
30

3.5
25
3
20 2.5
WN = 300 microns For MO structure

15 (WMO=30 microns) 2
12 3
NB = 1x10 /cm
XJ = 3.5 microns 1.5
10 QF = 2x1011/cm2
1
5
0.5
0 0
0 1 2 3 4 5 6 7 8
Relative permittivity of passivant
Fig.6.7(b): Optimal oxide thickness and optimum guard ring spacing vs. relative
permittivity of the passivant for the FLR and MO structure.
144
Another very important property of MO design is also clear from Fig.6.7(b) that
the tOX(OPT), required to accomplish maximum breakdown voltage, is lower if εdie of the
passivant layer is larger. Higher values of εdie allows for reduction in tOX required for
attaining a given breakdown voltage in MO structures, thus reducing the dead area from
the detector. Whereas for FLR structure, the value of GS(OPT) is independent of the εdie of
the passivant dielectric.

In practice, a positive oxide fixed charge adversely affects the breakdown voltage
of p+-n junctions, which are usually preferred over n+-p junctions for blocking high
voltage. A termination technique that renders the VBD insensitive to this charge is
attractive for high-voltage devices. Therefore, the influence of oxide fixed charge on the
optimal conditions & maximum breakdown voltage of both FLR and MO structures are
shown in figures 6.8(a) and 6.8(b).
35 4.5
4
30

For MO structure 3.5
(WMO=30 microns)
25
3
20 2.5
For FLR structure (GW=30
microns)
15 2
WN = 300 microns
NB = 1x1012/cm3 1.5
10 XJ = 3.5 microns
1
5
0.5
0 0
1.0E+11 3.0E+11 5.0E+11 7.0E+11 9.0E+11 1.1E+12
Fixed oxide charge density (/cm2)
Fig.6.8(a): Optimal oxide thickness and optimum guard ring spacing vs. fixed oxide
charge density for the FLR and MO structure.
145
1400
1200
WN = 300 microns MO structure

Maximum breakdown voltage (V)
1000 NB = 1x1012/cm3 ( Optimal oxide thickness, same for all QF)

XJ = 3.5 microns
800 FLR structure

( Optimal guard ring spacing for all QF)
600
400
FLR structure
200 ( Optimal guard ring spacing for QF=2x1011/cm2)
0
1.0E+11 3.0E+11 5.0E+11 7.0E+11 9.0E+11 1.1E+12
2
fixed oxide charge density (/cm )
Fig.6.8(b): Maximum breakdown voltage obtained under optimal conditions vs. fixed
oxide charge density for the FLR and MO structure.
It is clear from Fig.6.8(a) that the optimum oxide thickness (tOX(OPT)) for MO
design is independent of oxide fixed charge; in contrast, the optimum field ring spacing
for FLR structure is extremely sensitive to QF. In other words, the optimal MO design is
applicable over a wide range of QF, whereas the optimal FLR design is confined to the
specific value of QF, thus further revealing the performance leverages of the MO
technique over the FLR technique.
Also, it is evident from Fig.6.8(b) that the maximum VBD of the FLR structure is
more sensitive to oxide fixed charge than the MO structure, thus clearly illustrating the
superiority of the MO structure. Further, the problem of extending the optimal FLR
design, valid for a particular value of QF to other values is also demonstrated. Here the
optimum ring spacing obtained at QF = 2x1011/cm2 has been used. It is clear that VBD
decreases more sharply if GS is optimized only for a single value of QF. Since QF is a
physical parameter, which increases with increase in ionizing radiation, it is not possible
to optimize GS for all values of QF once envisaged in the device structure. The above
146
danger never exists in the design of the MO structure, where tOX(OPT) is almost insensitive
to the variation in QF.
Further, a multiple FLR structure occupies larger area than that required for a
single FLR structure. In contrast, a multi-step MO structure occupies an area comparable
to that of a single-step MO structure. Hence, a multi-step MO structure appears to be
more compact than a multiple FLR structure.
6.3.5 Effect of the Width of Guard Ring (GW; for FLR structures) and
Metal-Extension (WMO; for MO structures)
Fig.6.9(a) shows the plot of VBD as a function of GS with GW as a running
parameter and Fig.6.9(b) shows the plot of VBD vs. tOX with WMO as a running parameter.
It can be seen from Fig.6.9(a) that for FLR structures, the maximum breakdown voltage
increases around 70 volt as the GW is increased from 30 µm to 70 µm. For MO structures
the corresponding increase in VBD is ~ 175 volt as WMO is increased from 30 µm to 70
µm. The reason for this increase in VBD is that as the width (GW for FLR structures and
WMO for MO structures) increases, the depletion region is pushed further outwards,
distributing the potential over a larger distance and making the overall picture more like a
planar junction. However, it should be pointed out that after reaching a certain width,
maximum VBD saturates due to the sufficient flattening of equipotential lines, and further
increase in GW and WMO would no longer increase the VBD.
500
450 For FLR structure
400
350
300 GW=70 microns
250
GW=50 microns
200
WN=300 microns GW=30 microns
150
NB=1x1012/cm3
(a)
50 QF = 2x1011/cm2
0
0 10 20 30 40 50 60
Guard ring spacing (microns)
Fig.6.9(a): Breakdown voltage vs. guard ring spacing with GW as running parameter.
147
1600
For MO structure
1400 WMO=70 microns
1200
WMO=50 microns
1000
800 WMO=30 microns
600
WN=300 microns
400 NB=1x1012/cm3 (b)
XJ = 1.0 microns
200 QF = 2x1011/cm2
0
0 1 2 3 4 5 6 7
Fig.6.9(b): Breakdown voltage vs. field oxide thickness with WMO as running parameter.
It is also to be noted that optimized guard ring spacing (for FLR) and optimized
field-oxide thickness (for MO) increases with increase in guard ring width (GW). In order
to explain this behavior, we have plotted in figures 6.10(a) & 6.10(b), the plot of 2-D
potential distribution around the main junction and the floating ring for GW = 30 µm and
GW = 70 µm at fixed GS = 30 µm. It can be seen that the potential, to which a wide
guard floats (Fig.6.10(b)) with respect to the inner biased region, is smaller than that for a
narrower guard (Fig.6.10(a)). To illustrate, the 100 volt curve touching the outer edge of
the floating guard for GW = 70 µm (Fig.6.10(b)) lies very much within the main diode
and the guard ring for GW = 30 µm (Fig.6.10(a)). Thus, for GS = 30 µm, a wider guard
has to support a large potential drop resulting in crowding of equipotential lines near the
outer junction of the guard ring. Similar behaviour is also observed for metal-overhang
structure.
148
Fig.6.10: 2-dimensional equipotential contours in step of 20 V at GS=30 µm for (a)
GW = 30 µm and (b) GW = 70 µm for FLR structure.

In our previous chapters, the simulation results have been verified for the metal-
overhang structures. In order to validate the simulated results for the guard ring structure,
data available in the literature [6.9] was simulated. The results are given in Table 6.1. A
good agreement between the simulation and experimental results is observed.
149
Table 6.1: Comparison of simulations with experimental results.
NB XJ tOX QF GS VBD(volt) VBD(volt) (P-E)/P
3 2
(/cm ) (micron) (micron) (/cm ) (micron) Present Experime (%)
work (P) nt (E)
10 212 ~220 -4
5 x 1011 1.2 0.85 7.5 x 1011 20 295 ~310 -5
[6.9] 40 223 ~230 -3
6.5 Conclusions
Numerical comparisons of the breakdown voltage of metal-overhang and field-

limiting ring techniques, presented in this chapter, have demonstrated the superiority of
the MO design. For shallow junctions, the MO technique has higher breakdown voltage
than the FLR structure. The MO structure can be used to improve the VBD of planar
junctions without greatly increasing the junction depth, thus the MO technique is very
important for achieving high breakdown voltages in Si strip detectors used in HEP
experiments. In addition, the design layout of a MO structure is more flexible than the
FLR structure. It is demonstrated that higher values of relative permittivity (εdie) of the
passivant dielectric play an important role in determining breakdown voltage, and results
in an increase in breakdown voltage as compared to the unpassivated detector. The
optimal guard ring spacing is insensitive to the variation in εdie, whereas optimal oxide
thickness decreases with increase in εdie for dielectric passivated structure. These results
indicate that the MO structure is attractive for planar shallow-junction high-voltage
devices. Effect of increasing the width of guard ring (for GR structure) and metal-
extension (for MO structure) shows that the breakdown voltage and the optimized
spacing increase with increase in GW and WMO.
Further, the breakdown voltage of a MO structure is practically immune to fixed
oxide charge density (QF), and its optimal design is independent of QF. Thus, the present
study shows that the MO structures allow for a design of Si strip detectors with shallow
junctions and thinner oxides, reducing dead layer and making the detectors more suitable
for HEP experiments.
150
Chapter 7
----------------------------------------------------------------
Large Transverse Momentum (pT) Direct Photon
Production at LHC
----------------------------------------------------------------
Study of direct photon production in high-energy hadronic collisions provides a

clean tool for testing the validity of perturbative Quantum Chromodynamics (QCD)
predictions as well as for constraining the gluon distribution of nucleons. This chapter
starts with a brief review of the Standard Model and QCD, followed by our simulation
work on direct photon physics. The present analysis describes the study of direct photons
in the kinematical regions accessible at LHC energy. After presenting the theoretical
description of the Tevatron direct photon data, predictions for direct photon cross section
at s =14 TeV along with various theoretical uncertainties is described.
7.1 An Introduction to Standard Model and QCD
In recent years, high-energy physicists have arrived at a picture of the

microscopic physical universe, called "The Standard Model", which unifies the nuclear,
electromagnetic, and weak forces and enumerates the fundamental building blocks of the
universe (as shown in Table 7.1). The Standard Model encompasses two families of
subatomic particles that build up matter and that have spins of one-half unit (fermions).
These particles are the quarks and the leptons, and there are six varieties, or "flavours," of
each, related in pairs in three "generations" (Table 7.1). In the Standard Model, the forces
are communicated between particles by the exchange of quanta which behave like
particles of spin 1 (bosons) (Table 7.2). The Standard Model has proved a highly
successful framework for predicting the interactions of quarks and leptons with great
accuracy.
151
Table 7.1: The fundamental particles in the Standard Model.
FERMIONS (matter constituents)
Leptons (spin = ½) Quarks (spin = ½)
Flavour Mass Electric Flavour Appr. Mass Electric
(GeV/c2) Charge (GeV/c2) Charge
1st gener. νe < 7x10-9 0 u 0.005 2/3
e 0.000511 -1 d 0.01 -1/3
2nd gener. νµ < 0.0003 0 c 1.5 2/3
µ 0.106 -1 s 0.2 -1/3
3rd gener. ντ <0.03 0 t 175 2/3
τ 1.7771 -1 b 4.7 -1/3
Table 7.2: The fundamental interactions along with the force carriers.
FOUR INTERACTIONS
Gauge Bosons as Force Carriers
Interaction Gauge Mass Electric Spin-parity Coupling
2
Bosons (GeV/c ) Charge Constant
Strong g (gluon) 0 0 1- αs ~ 1
Unified γ (photon) 0 0 1- αem = 1/137
Electroweak W+ 80.33 +1 1- 1.02 x 10-5
-
W 80.33 -1 1- 1.02 x 10-5
Z0 91.187 0 1+ 1.02 x 10-5
Gravity graviton 0 0 2+ 0.53 x 10-38
HIGGS BOSONS
Breaking EW symmetry
H (higgs) > 105 0 0 ?
Quantum Chromodynamics (QCD): One particular ingredient of the Standard

Model that requires further quantitative investigation is the Quantum Chromodynamics
(QCD), which has emerged as a viable theory of strong interactions over the last two
decades. In this theory, the force acts between color charges carried by quarks and
gluons. The magnitude of the force between two color charges is proportional to the
product of the charges. The intrinsic strength of the interaction is defined by a
dimensionless running coupling constant (αs). There are eight massless gluons in QCD,
152
which not only transmit the strong force but also change the color of the quarks. Quarks
and gluons are collectively referred to as partons.
The asymptotic freedom [7.1] in QCD allows one to apply the perturbative
technique for calculating cross-section at high energies for processes that are dominated
by short distance interactions. The QCD coupling constant, which is a function of
momentum transfer between two partons (Q2), is given to the leading log in Q2 as [7.2]:
12 π
α s (Q 2 ) = (7.1)
(33 − 2n f ) ln(Q 2 Λ2 )
where Λ is the characteristic scale parameter, required by the theory and sets the scale for
Q2 dependence and is of the order of several hundred MeV, and nf is the number of quark
flavors kinematically available at the collision energy. The above equation shows that as
Q2 increases, αs(Q2) decreases and as Q2 → ∞ (distance → 0), αs(Q2) →0. This means
that at high momentum transfer, quark confinement [7.3] can be neglected, that the
constituents are essentially free in the hadrons and the hadron-hadron interaction may be
simply considered as a parton-parton interaction.
7.2 QCD Phenomenology of High pT Inclusive processes
Parton – parton interaction can be well understood by large transverse momentum

(pT) inclusive particle production. Particles produced at large pT are well described by
QCD. In parton – parton collisions, high pT mesons are produced by the fragmentation of
partons into hadrons. These mesons come out as a jet in the parent parton direction. To
study the characteristics of these hadrons one will have to rely on parton – parton
scattering, as well as on parton fragmentation function. An alternative way to study the
elementary process is to investigate the production of direct photons at large pT.
Consider a basic diagram for hadronic interactions A + B → C + X as shown in
Fig.7.1. A and B are interacting hadrons, C is the hadron with large pT, and X represents
all other particles in the final state. The incoming particles A and B contain partons ‘a’
and ‘b’ respectively, which scatter, producing partons labeled ‘c’ and ‘d’ which have a
large transverse momentum component qt. Subsequently hadron C is produced from
parton ‘c’ via the fragmentation process. Since qt is the conjugate variable to the impact
variable of the parton scattering process, large qt implies that partons have scattered at a
153
small distance where αs is small and Q2 is large. Hence perturbation theory can be
applied.
Fig.7.1: Schematic diagram of a two-body reaction, A+B C+X, which has been
factorized according to the prescriptions of perturbative QCD.
One can express the invariant cross section for A+B→C+X as the weighted sum
of the differential cross-section, of all possible parton scatterings that can contribute
[7.2],
dσ  sˆ
EC 3
d pC
( AB → C + X ) = ∑ ∫
abcd 
2 2
dx a dxb dz c Ga A ( x a , Q )Gb B ( xb , Q ) 2
zc π
 (7.2)
dσ 
× (ab → cd ) DC / c ( z c )δ ( sˆ + tˆ + uˆ ) 
dtˆ 
where EC and pC are the energy and momentum of the final state hadron C. sˆ, tˆ and uˆ
are the Mandelstam variables for the massless parton subprocess, defined as
sˆ = ( p a + p b ) 2 ; tˆ = ( p a − p c ) 2 ; uˆ = ( p a − p d ) 2 (7.3)
xa and xb are the fractions of the longitudinal momentum carried by partons ‘a’ and ‘b’ of
hadrons A and B, zc is the fraction of the ‘c’ parton’s longitudinal momentum carried by
hadron C. The probability of finding a parton in the interval between xa and xa+dxa is
denoted by the parton distribution function (PDF) Ga A ( x a , Q 2 ) . DC c ( z c ) is the
probability that hadron C carries a momentum fraction between zc and zc+dzc of the
parent parton’s (c) momentum and is referred to as the fragmentation function. Structure
and fragmentation functions cannot be calculated using perturbation theory and must be
obtained using data from various hard scattering processes. One can predict the cross-
sections if the structure functions, fragmentation functions and the cross sections for all
the parton subprocesses are known. In equation (7.2), the partons have been assumed to
154
be massless and the initial and the final state partons are collinear with the corresponding
hadrons, i.e., the partons have no intrinsic transverse momentum (kT).
7.3 Direct Photons in the QCD Framework
QCD has been a successful theory in describing the interactions between the
fundamental building blocks inside hadrons – quarks and gluons. In addition, it is in good
agreement with experimental data collected both at fixed target and colliding beam
experiments. The analysis of enormous amounts of data led us to a deeper understanding
of the properties of the fundamental interactions, and also revealed the inner structure of
the hadrons. However, the features of QCD, although qualitatively verified, are far from
being completely understood. It is then crucial to investigate the properties of QCD at
hadron colliders to probe the inner structure of the hadrons from the standpoint of
perturbative QCD (pQCD) techniques and the parton model of strongly interacting
particles.
The production of high transverse momentum direct (or prompt) photons [7.2,
7.4] from the parton–parton interactions [7.5] at the Fermilab Tevatron pp collider
experiments and future LHC pp collider offers a good general testing ground for the
validity of perturbative QCD and for an understanding of the contribution to the hard
scatterings from the gluons, in particular. Here, direct means that the photons are
generated in the hard scattering process, and not from secondary decays or as radiation
product of initial or final state partons.
7.3.1 Contributions to Direct Photons

Leading-Order (LO) Contributions: The major contributing physics processes
for direct photon production to leading order of the strong coupling constant αs at large
pT, where a perturbative expansion in QCD is expected to be valid, are the two Compton
processes and the two annihilation processes that are shown in Fig.7.2 [7.2, 7.4]. Their
respective contribution to the direct photon yield is clearly a function of the nature of the
constituent distributions within any hadron.
155
Fig.7.2: The leading order diagrams for direct photon production.
Next–to–Leading-Order (NLO) Contributions: In recent years, NLO pQCD

calculations have been performed for the direct photon production process, with a few
diagrams shown in Fig.7.3. These are processes with extra gluons radiated from the initial
or final state partons, or processes with gluon loops as correction to the original LO
process.
Fig.7.3: A few higher order direct photon processes.
Direct Photon Pair Production: Occasionally two direct photons are produced
in the hard scatter. At lowest order these photons are produced in quark-antiquark
annihilation (Fig. 7.4(a)) with a production rate of ~1% of that for single direct photon.
One higher order diagram that can contribute a significant portion of the cross section
[7.6] is the quark box diagram (gluon-fusion sub process gg → γγ ) in Fig.7.4(b).
156
Although this process is of the order αs2, the cross section is comparable to Born
contribution and becomes even larger at high center of mass energies due to large gluon
density at small value of photon momentum fraction x.
Fig.7.4: The leading contributions to the production of two prompt photons.
7.3.2 Background to Direct Photons

The experimental candidate photon samples are always contaminated by
substantial backgrounds, which greatly complicate the analysis of the direct photon
signal. The dominant background to the prompt photon events comes from jets. While
most jets consist of many particles, and are thus easily distinguishable from a single
photon, a small fraction (one in 103-104) fragments in such a way that a single particle
gains most of the energy of the parent parton. If that particle is a neutral meson, like π0 or
η that can decay to two photons, the decay product may be indistinguishable from a
single photon since at high energies the two photon showers coalesce into a single cluster
in the calorimeter. The isolation criteria rejects bulk of these jets leaving about 0.1% of
them which fragment this way and mimic a true photon signal [7.7]. While only one in
103-104 jets fragments this way, the dijet cross section is 103-104 times larger than that of
photon cross section. Therefore, the rate at which single particle jets are produced is
similar to the rate at which prompt photons are produced, thus contributing a severe
background to the direct photon sample. It is therefore very important to understand the
background evaluation and extraction as precisely as possible. Its precise knowledge is
also crucial to pin down the existence of new particles such as H → γγ or any breaking
down of symmetry in the Standard Model. The decay of π0 mesons into two photons
forms the largest contribution [7.8] to background, since π0’s are most commonly
produced.
Anomalous Contributions: As shown in Fig.7.5, there is another source of
single photons: bremsstrahlung from an outgoing quark in a dijet event [7.9]. In this case
the photon is not produced directly from the interaction vertex and is therefore not really
157
a direct photon. However, the existence of this production mechanism affects the way in
which direct photons are measured and modeled theoretically. A careful choice of few
selection cuts can usually minimize these contributions. Table 7.3 summarizes the
processes that give rise to direct photons.
Fig.7.5: Examples of anomalous processes.
Table 7.3: Various processes (order in αs and αem) contributing to the production of
single and double direct photons (along with the bremsstrahlung contributions).
Description Order Subprocess
Annihilation αemαs qq → γg
Compton αemαs qg → γq
Single bremsstrahlung αemαs qq → q(q → γ)
gq → g(q → γ)
qg → q(g → γ)
gg → g(g → γ)
QCD-induced gγ coupling αemαs2 gg → γg
QED annihilation αem2 qq → γγ
Single bremsstrahlung αem2αs qg → γ (q → γ)

Double bremsstrahlung αem2αs2 qq → (q → γ)(q → γ)
gq → (g → γ)(q → γ)
gg → (g → γ)(g → γ)
Quark box αem2αs2 gg → γγ
158
7.3.3 Motivation
Direct photon production as a subject has been studied extensively on both the
theoretical and experimental level. The reasons for the continuing interest in the study of
direct photon physics are:
1.) Simplicity: The importance of direct photon production arises from the well-
understood electromagnetic coupling of a photon to a quark and the consequent
anchor that this process can provide in helping unfold the underlying quark-gluon
dynamics and hadron structure. The great advantage of using direct photons is
that they emerge from the collisions as free particles, carrying the full momentum
of the partonic collisions and consequently provide pristine information about the
hard scattering. In contrast, gluons or quarks must fragment into hadrons (of
reduced pT), and the hadrons must first be associated with their respective partons
before the extraction of the physics of constituent interaction can take place.
2.) Fewer Number of Subprocesses: As photons do not carry electric charge they
cannot interact directly with each other. This greatly reduces the number of
subprocesses which contribute to the direct photon production process. As shown
in Fig.7.2, Compton and annihilation diagrams are the only two subprocesses
which contribute to the first order direct photon production. There are only 18
diagrams for three quark subprocesses, compared to 127 separate two body
scattering diagrams for three quarks flavour for single hadron production [7.2].
Also, because the photons do not carry charge, unlike gluons, they do not
hadronize removing the inherent ambiguities present in the case of jets which can
be either due to quarks or gluons. Experimentally, the photons can be clearly
identified and their energy and direction can be measured precisely, unlike jets,
which are messy due to fragmentation and can only be defined given a certain
reconstruction algorithm.
3.) Gluon Physics: As shown in Fig.7.2, gluons are involved in both the Compton
and annihilation diagrams. In the Compton diagram, the gluon is involved in the
initial state whereas in the annihilation diagram, the gluon appears in the final
state. By separating the contributions from the Compton diagram to the direct
159
photon production process one can measure the gluon structure function of the
colliding hadrons. The gluon fragmentation function can be measured by isolating
the contributions from the annihilation diagram. The direct photon cross section is
very sensitive to the gluon content of proton because of the dominant contribution
from the quark-gluon hard scatterings at the LO in pp and pp collisions. This is
in contrast with deep inelastic scattering (DIS) experiments where the quarks are
the major participants and gluons enter only as second order effects. Thus prompt
photon cross section constitutes a classical tool for constraining the gluon density
[7.5] in conjunction with DIS processes especially at large longitudinal
momentum fraction x (beyond x ~0.15), where there is large uncertainty.
4.) QCD Tests: The next to leading log (NLL) [7.10] predictions for direct photon
production are available. The theory can be more reliably compared with the data
over a wide kinematic range.
5.) Window to a New World: Other than being a good testing ground of recent NLO
QCD calculations, the direct photons observed at the collider experiments can
help us in searching for exotic phenomena like the excited quark states,
q + g → q* → q + γ
On the other hand, in the search for the Standard Model Higgs particle, photons
play an important role since the di-photon signal is a unique signature of the
neutral Higgs decay process in the light mass range H → γγ.
The two major problems with using direct photons as probes are the following [7.4]:
1.) The yield is greatly reduced relative to jet production (for analogous graphs, this
is a factor of ~ 30 at the pT values of interest; however, as indicated previously,
far more graphs contribute to hadron jets than to the yield of direct photons and
consequently the overall γ/jet production ratio is ≤ 0.001);
2.) There is substantial background from the decays of π0s and η0s mesons that make
the extraction of direct photon signals challenging. Fortunately, because the
photon carries away the entire pT in the elementary collision, whereas π0s or η0s
are fragments of the constituents (typically, with small momentum fractions) it is
expected that for fixed angle in the center of mass the γ/hadron production ratio
will increase with pT, and the γ yield will eventually surpass that of π0s .
160
7.4 Theoretical Formalism
In the framework of QCD perturbation theory, the differential cross section for
the inclusive single prompt photon production, h1 h2 → γ X , in transverse momentum
p T and rapidity η can be written in a factorized form as
dσ dσ dir dσ brem
= +
dpT dη dpT dη dpT dη
where we have distinguished the “direct” component σdir from the “bremsstrahlung”
component σbrem. Each of these terms is known in the next-to-leading logarithm
approximation [7.10] in QCD, i.e., we have
dσ dir α ( µ ) dσˆ ij α ( µ ) dir ½

= ¦ ³ dx1 dx 2 Gi / h1 ( x1 , M ) G j / h 2 ( x 2 , M ) s ® + s K ij ( µ , M , M F )¾
dpT dη i , j =q , g 2π ¯ dpT dη 2π ¿
dσ brem dz α (µ ) 2
= ¦ ³ dx1 dx 2 2 Gi / h1 ( x1 , M )G j / h 2 ( x 2 , M ) Dγ / k ( z , M F ) ( s )
dpT dη i , j ,k = q , g z 2π
° dσˆ ij k α (µ ) ½°
K ij ,k ( µ , M , M F )¾
brem
×® + s
°̄ dpT dη 2π °¿
where the parton densities in the initial hadrons Gi/h1 and Gj/h2 and the parton to photon
fragmentation function Dγ/k have been convoluted with the partonic cross sections of the
hard scattering subprocesses, x being the parton’s momentum fraction and z being the
longitudinal momentum fraction of parent parton carried by the bremsstrahlung photon.
Here we have neglected the transverse motion of partons ( kT ) prior to hard scattering.
The higher order correction terms to the direct and bremsstrahlung cross sections are
represented by Kijdir and Kij,kbrem respectively. The parton distribution functions [7.11] and
fragmentation functions [7.12] are extracted via global analysis of experimental data
particularly from deep inelastic lepton-proton scattering.
7.4.1 Scale Sensitivity

The intrinsic uncertainties of the NLO QCD predictions are related to the choice
of three arbitrary scales: the renormalization scale µ which appears in the evolution of
161
strong coupling constant αs, the factorization scale M associated with the initial state
collinear singularities and the fragmentation scale MF related to the collinear
fragmentation of a parton into a photon. Roughly speaking, these are the parameters
which control how much of the higher effects are resummed in αs(µ), Gi/h and Dγ/k
respectively and how much is treated perturbatively in Kijdir and Kij,kbrem. As these scales
are unphysical, the theory can be considered reliable only in the region of the phase space
where the predictions are stable with respect to the scale variations [7.13]. At the leading
logarithm level, the photon cross section depends sensitively on the specific choice used
for scales. When the next-to-leading logarithm terms are included, it makes the theory
more complete and less sensitive to the choice of scales. Current NLO QCD calculations
for direct photon cross section have been performed both analytically [7.10, 7.14] and in
Monte Carlo framework [7.15] which conventionally choose all the three scales to be
equal to the photon transverse momentum p T .
7.4.2 Pseudorapidity Dependence

Previous theoretical analysis [7.16] has shown that the direct photon cross section
has a pseudorapidity (η) dependence which is sensitive to the parameterization of the
gluon distribution functions. This sensitivity is even more dramatic in the lower
transverse momentum or in the forward regions of the detector. Since earlier direct
photon experiments with the exception of DØ & CDF have concentrated on the central
region, the forward direct photon detection capability of the CMS detector at LHC allows
us a new kinematical region for investigating the pseudorapidity dependence where the
gluon distribution within hadrons can be constrained. This motivation is based on the fact
2 pT
that the transverse momentum fraction probed by the photons is xT = , which is
s
related to the momentum fraction of the partons ( x ), as sˆ = x1 x 2 s . If the initial partons
are of nearly equal momenta, the photon-jet system will retain its center-of-mass back-to-
back nature in the laboratory frame. Then in the central region (η = 0), xT = x . Now, if
one parton is of much greater momentum than the other, then the system is boosted as the
more energetic parton overwhelms the softer ones, and the final state objects tend to be
on the same side of the event. Since gluons typically carry much less of the momentum of
162
the proton than do quarks i.e. x g << x q , one expects that in direct photon production
large momentum imbalances will dominate, and the final state will tend to be boosted in
the direction of incoming quark. In other words, both the photon and the jet will tend to
be produced at small angles, either both forward (η > 0) or both backward.
7.4.3 Isolation Technique

Because the bremsstrahlung photons tend to be collinear with the quark, and
therefore the jet, from which it is radiated, an isolation criterion is used routinely in the
collider regime to suppress such events. This method is based on the expectation that
direct photons are fairly isolated in the detector while photons from anomalous
contributions usually have quite a few hadrons in their vicinity coming from
fragmentation products of the outgoing parton. The isolation requirement is typically
implemented by measuring the amount of energy in the calorimeter inside a cone of
radius R ( R = ∆η 2 + ∆φ 2 : typically R = 0.4 - 1.0) centred on the photon candidate and

requiring that the hadronic energy be smaller than a certain amount. This strongly
discriminates against production of photons from bremsstrahlung process, but the
backgrounds that mimic this process are too large to allow a direct measurement. This
requirement, unfortunately, can do nothing to remove bremsstrahlung photons that are
radiated at large angles with respect to jets. Such an isolation cut suppresses but does not
totally remove this component.
Theoretical calculations involve the non-perturbative fragmentation functions to
account for bremsstrahlung contribution, which is partially removed by the isolation cut
matching that of the experiment. At first guess, one might expect that bremsstrahlung (for
example, qq → qqγ or qg → qgγ etc.) could be of the order O(αemαs2) and perhaps
negligible in most regions of phase space. This however, is not the case entirely because
the fragmentation function of a constituent into a photon scales as αem/αs and the cross
section for bremsstrahlung component is of the same order O(αemαs) [7.16] as the two
leading order fundamental QCD subprocesses. The bremsstrahlung photons contribute a
large fraction (> 50%) of the total non-isolated photons at 1.8 TeV at low xT which is
reduced to 15-20% after the isolation cut is invoked [7.17].
163
7.4.4 kT Smearing
The global QCD analysis [7.18] of the direct-photon production process from
both fixed target and collider experiments spanning over a wide range of parton x-values
(0.01 to 0.6) has been puzzling. The transverse momentum (pT) distribution of the
measured inclusive direct-photon cross-section agrees qualitatively well with the next-to-
leading-order (NLO) QCD predictions for conventional choices of scales in the high pT
region. But most data sets show deviations from the NLO QCD calculations and have a
steeper pT distribution in the low pT region. Neither global fits with new parton
distribution functions nor improved photon fragmentation functions can resolve this
discrepancy since the deviation occurs at different x-values for experiments at different
energies. The obvious source of uncertainty due to choice of scale can also not be
responsible for the discrepancy since it provides a small normalization shift with no
change in slope. The suspected origin of the disagreements is from effects of initial-state
soft-gluon radiation which generates transverse components of initial-state parton
momenta, referred to as kT effects.
Current NLO QCD predictions assume that the interacting partons are collinear
with the beam and the partons emerging after the hard scattering are produced back-to-
back with equal pT. However, in the hadron-hadron centre of mass frame, the colliding
partons may no longer be collinear; i.e., they can have some transverse momentum kT
with respect to each other, which gives a boost in the direction of one of the outgoing
particles. It has been suggested that the smearing of transverse momentum of initial-state
partons can probably explain the low pT discrepancy since any uniform smearing on a
steeply falling pT distribution enhances significantly only the low pT end of the spectrum
[7.19].
Such kT can arise from several sources. There is a primordial kT due to
confinement of partons within hadron (~ 0.5 fermi in size) which is approximately 0.3 –
0.4 GeV/c. The majority of such transverse momentum can, however, be attributed to the
emission of multiple soft gluon by the partons prior to the hard scatter. Evidence of
significant kT has long been observed in measurements of dimuon, diphoton, and dijet
production [7.19]. Studies of high mass pairs of particles such as direct photons and π0’s
can be used to extract information about the parton kT. The kinematic distributions of
164
these high mass pairs illustrate the evidence for significant kT which increase almost
logarithmically with increasing centre of mass energies, from <kT> ~ 1GeV/c at fixed-
target energies increasing to 3-4 GeV/c at the Tevatron collider.
Fully resummed pQCD calculations for single direct photon production are
anticipated shortly. Since current NLO QCD calculations do not account for the effects of
multiple soft-gluon emissions, we employed a phenomenological model to incorporate kT
effects in the NLO calculations of direct photons. We use LO pQCD (PYTHIA which
incorporates kT effects using a Gaussian smearing technique) to create kT-enhancement
factors as a function of pT for inclusive cross sections and then apply these factors to the
NLO calculations. We recognize that this procedure involves a risk of double-counting
since some of the kT-enhancement may already be contained in the NLO calculation.
However, we expect the effect of such double-counting to be small [7.19].
7.5 Monte Carlo Simulation
A Monte Carlo (MC) simulation based on PYTHIA 6.2 code [7.20] is used to
calculate the direct photon cross section by generating N = 105 pp events at the center-of
mass energy of 14 TeV. The parton level subprocesses employed to simulate γ-jet events
were: qg → qγ , qq → qγ and gg → gγ . PYTHIA describes the hard scattering
between hadrons via leading order perturbative QCD matrix elements. The parton
distribution functions used in the analysis were those given within PYTHIA.
7.6 Direct Photon Production at Tevatron: Comparison of Data

with Theory at s =1.8 TeV & s =630 GeV
In Fig.7.6, we compare the measurement of the cross section for production of

isolated prompt photons in proton-antiproton collisions at Tevatron [7.21] at s =1.8
TeV and s = 630 GeV by CDF & DØ Collaboration with the corresponding theoretical
calculations. The data used here was collected during Run 1B [7.22]. The NLO QCD
calculations [7.14] are derived using the latest PDF, CTEQ5M1 [7.23] with the
renormalization, factorization and fragmentation scales set at pT. For CDF data, this
165
calculation imposes an isolation criterion, which rejects events with a jet of ET > 1 GeV
in a cone of radius 0.4 around the photon. For DØ data, the total transverse energy near
any photon candidate cluster must satisfy an isolation requirement
R ≤ 0.4 R ≤ 0.2
ET − ET < 2.0 GeV , where R = ∆η 2 + ∆φ 2 is the distance from the cluster
center.
Fig.7.6: The inclusive photon cross sections at center-of-mass energies 1.8 TeV and 630
GeV measured by the CDF & D0 collaborations compared to the NLO QCD predictions
[7.24].
Fig.7.7: A comparison of the Run 1b data at 1.8 TeV and 630 GeV data to NLO QCD
calculations [7.24] as a function of photon pT.
We see from Fig.7.6 that the NLO QCD predictions agree qualitatively with
the measurements over a wide range of pT. The visual comparison between data and
theory is aided by plotting (data-theory)/theory on a linear scale (Fig.7.7) which shows an
166
excess of photon cross section over theory in the low pT region. We notice that the CDF
and DØ data sets at 630 GeV and 1.8 TeV are consistent with each other.
Fig.7.8 shows the effect of different parameters on the 1.8 TeV data. The change
of renormalization scale from µ = pT to µ = pT/2 or µ = 2pT changes the predicted cross
sections by < 10% thus producing a small normalization shift throughout with almost no
change in slope. Simultaneous variations of all the theoretical scales (renormalization
scale µ, factorization scale M and fragmentation scale MF) [7.25] independently also
produces a small change in the shape of the predictions, but does not reproduce the shape
(Fig.7.7) of measured cross sections. However, one should not worry too much about the
large pT regime keeping in mind that data have a 14 % normalization uncertainty, and
that changing scales in the theory also produces roughly the same normalization shift.
Fig.7.8: A comparison of the CDF & DØ Run 1B data at 1.8 TeV to theories with NLO
QCD using different choices of the renormalization, factorization and fragmentation
scales [7.24].
The low pT excess of data over theory is consistent with previous observations
[7.18, 7.19] at collider and fixed-target energies. This excess may originate in additional
multiple soft-gluon radiations (which could give a recoil effect to the photon+jet system)
[7.19] beyond that included in the QCD calculations, or reflect inadequacies in the parton
distribution functions [7.26] and fragmentation contributions. It has been suggested [7.18,
167
7.19] that the smearing of transverse momentum of initial state partons (kT kick) can
probably explain this low pT discrepancy since any uniform smearing on a steeply falling
pT distribution enhances significantly only the low pT end of the spectrum. Higher order
QCD calculations including soft-gluon effects through resummation technique are
becoming available [7.27] but are not currently ready for detailed comparisons. To
explore qualitatively the effect of kT on the comparisons, we have added a simplified
Gaussian smearing in the NLO QCD calculations to see if the measurements could be
sensitive to these effects. Fig.7.9 shows a comparison of the CDF and DØ data at 630
GeV to NLO QCD calculation using CTEQ5M1 with the addition of a 3 GeV kT
correction (a value obtained from the diphoton measurement in CDF). We see that the
NLO theory supplemented with kT correction accounts to a great extent the low pT
discrepancy between data and theory.
Fig 7.9: The comparison of the CDF & D0 data at 630 GeV to NLO QCD calculations
with 3 GeV kT correction [7.24].
7.7 Expectations for Direct Photons at LHC
7.7.1 Leading Order (LO) Cross Section

Fig.7.10 shows the transverse momentum (pT) distribution of direct photons for
various LO subprocesses normalized to the total rate at LHC energy in the kinematical
range 20 GeV < pT < 400 GeV and pseudorapidity interval –3.0 < η < 3.0. The results
were generated by simulating direct photon events using PYTHIA with the CTEQ5M1
parton distribution function and with the renormalization scale µ = pT.
168
Fig.7.10: Contributions of various subprocesses for direct photon production at the
leading-order normalized to the total rate at LHC energy.
The results shown in Fig.7.10 reveal that the Compton scattering provides the
dominant mode of direct photon production in the entire kinematical region which is
indicative of the fact that the direct photon data from LHC would be handy in providing
constraints on the gluon distribution in global fits of parton distributions in the high pT
range. The annihilation scattering provides relatively small contribution in the low and
intermediate pT regions, but its contribution increases with increase in pT. Also, the
gluon-gluon initiated processes are not expected to play a significant role over the pT
range shown.
7.7.2 Next-to-Leading Order (NLO) Cross Section

Fig.7.11 shows the pT spectrum of NLO QCD predictions [7.14] for direct photon
cross section at LHC along with the LO (PYTHIA) estimates, evaluated with the
CTEQ5M1 parton distribution function and renormalization scale µ = pT in the same
pseudorapidity interval –3.0< η <3.0. The NLO calculations use the same isolation cut as
that of CDF. In comparison to the cross-section at Tevatron energy (Fig.7.6), this
distribution extends to greater than three times than at the Tevatron. We see that the NLO
QCD contribution is higher than the LO in the whole pT range under analysis.
169
Fig.7.11: LO & NLO QCD predictions for direct photon cross-section at LHC.
7.7.3 K - factor
All PYTHIA cross-section estimates are based primarily on leading-order (LO)
calculations. Often these LO cross-sections differ significantly from the theoretical NLO
QCD calculations. The ratio of σ(NLO)/ σ(LO) defines the so-called K-factor. Fig.7.12
shows how the NLO results of direct photon cross-section at the LHC differ from the LO
cross-section as a function of pT. Numerical PYTHIA “K-factors” [7.28] are derived for
three PDF’s. K-factors of up to 2 have been plotted for CTEQ5M1 in Fig.7.12. We see
that NLO contribution to the cross section decreases with rise in pT and considerably
depends on the choice of PDFs. This is mainly due to considerable decrease in the higher-
order soft-gluon corrections as pT increases. The theoretical predictions have greatly
improved with precise PDF's.
Fig.7.12: Variation of relative contributions of LO & NLO contributions to direct photon

cross-section at LHC as a function of pT.
170
7.7.4 Scale Dependence of Inclusive Cross Sections
To see the renormalization scale (µ) dependence of the theoretical predictions for
direct photon inclusive cross section we compare the LO and NLO QCD results with
CTEQ5M1 parton distribution function. We choose the central rapidity region, η = 0, for
scales µ = pT/2 and µ = 2pT normalized to the conventional scale µ = pT and the
predictions are shown in Fig.7.13. We see that the LO calculations shows strong scale
dependence at low pT. At pT = 20 GeV, the variation of scale between pT /2 and 2pT leads
to a normalization uncertainty of ~25%. Again, the scale dependence gains more and
more significance in the high pT region and at pT = 400 GeV, the variation of scale
between the above limits changes the cross section by ~25%. This variation of LO QCD
calculations with scale implies the need for incorporating higher order correction factors.
As expected, we notice from Fig.7.13 that the NLO QCD calculations [7.24] are less
sensitive to the choice of scale. The variation of scale between pT /2 and 2 pT leads to a
normalization uncertainty of at most 14% over the whole pT range under consideration,
thus showing the reliability of perturbative QCD predictions.
Fig.7.13: Ratio of LO & NLO QCD cross sections for direct photons at LHC for different
choices of µ (µ = p T /2 and 2 p T ) normalized to that for conventional choice of µ = p T .
7.7.5 Sensitivity to Gluon Distributions

7.7.5.1 The p T spectrum
As an illustration of sensitivity of direct photon production to the different
parameterizations of gluon distribution, we compare the pT spectrum of NLO QCD
predictions for direct photon cross section averaged in the pseudorapidity region │η│< 3
171
due to different choices of parton distribution functions (PDFs): CTEQ3M, CTEQ4M,
CTEQ5M, CTEQ5Hj, MRS99 [7.29] and GRV94M [7.30] normalized to that of
CTEQ5M1 PDF (Fig.7.14). In general, theoretical uncertainties are greatly reduced for
the ratio of cross sections. As can be seen from the Fig.7.14, the ratio of cross sections is
almost insensitive to the choice of PDF at pT > 300 GeV, corresponding to xT > 0.05, but
exhibits more and more sensitivity as we move to the low pT region.
Fig.7.14: Transverse momentum distribution of direct photon cross section at LHC for
different parton distribution functions.
The ratio of the recent PDFs (CTEQ5M, CTEQ5Hj) and CTEQ5M1 is consistent
with unity within at most 4% excess over the pT range under consideration. The MRS99
PDF coincides with the CTEQ5M1 at high pT and shows a deficit at low pT of at most
5%. GRV94M and CTEQ3M are significantly lower at low pT by at most ~10% and 20%
respectively. Thus we notice that the pT spectrum of prompt photons is sensitive to the
small-pT behaviour of gluon distribution.
7.7.5.2 The η spectrum

Low pT Region: Fig.7.15 shows the pseudorapidity distribution of NLO QCD
cross section for direct photons with their transverse momenta, 20 GeV < pT < 50 GeV,
for different parton distribution functions. We note that production of photons is fairly
high in the central rapidity region. We also see that η spectrum is quite sensitive to the
parton distribution function, particularly in the central region. This is more explicitly
exhibited from the η spectrum of the ratio of cross section for different PDFs normalized
172
to that of CTEQ5M1. Thus η spectrum of direct photons is more helpful in obtaining
information about the gluon distribution for small-x gluons.
Fig.7.15: Pseudorapidity spectrum of direct photon cross section at LHC for different
PDF’s for transverse momentum of photons, 20 GeV < p T < 50 GeV.
Large pT Region : Fig.7.16 shows the pseudorapidity distribution of NLO QCD

predictions for direct photon cross section with different PDFs for high transverse
momentum of photons, 300 GeV < pT < 400 GeV. We see that from the η spectrum of
direct photons it is very difficult to distinguish between the different parameterizations of
the gluon distribution. Thus, the η distribution of direct photon cross section is almost
insensitive to the large-x behaviour of gluons.
Fig.7.16: The Pseudorapidity spectrum of direct photon cross section at LHC for
different PDFs for transverse momentum of photons, 300 GeV < p T < 400GeV.
173
7.7.6 Pseudorapidity Dependence
Fig.7.17 shows the pT distributions of LO & NLO QCD predictions for integrated
direct photon cross section in different pseudorapidity windows. As expected, cross
section is more for larger η interval.
Fig.7.17: LO and NLO QCD predictions for direct photon cross section at LHC
integrated in different pseudorapidity intervals.
Fig.7.18 compares the averaged differential cross sections for direct photons in
different pseudorapidity bins normalized to the cross section for η ~ 0. We see that direct
photons are produced fairly copiously in the central region. The production rate decreases
in the high η domain, particularly at high pT.
Fig.7.18: Ratio of direct photon cross section in different pseudorapidity bins normalized
to the cross section in the central rapidity region.
174
7.7.7 Cone Size Dependence
Fig.7.19 illustrates the cone size dependence of the NLO QCD predictions for
direct photon cross section as a function of pT in the pseudorapidity bin (-3 < η < 3) using
CTEQ5M1 parton distribution function and the renormalization scale µ = pT, wherein the
pT spectrum of the ratio of cross sections with different cone sizes to that of cone size =
0.7 is shown. As can be seen from the figure, the cross section decreases as cone size
increases. At pT = 50 GeV, changing the cone size from 0.1 to 0.4 and 0.7 reduces the
cross section by 13% and 38% respectively. This behaviour is expected because the
isolation criterion excludes events with a certain hadronic energy E0 inside a cone of size
R. Now, keeping E0 fixed and increasing R means that we are not even allowing such
events in a large cone, so it is a stricter criterion (keeping in mind that the jet cross
section increases considerably with cone size [7.31]), and hence the cross section must
decrease. We see from Fig.7.19 that cross section decreases almost uniformly over the
whole pT region except at low pT for cone size = 0.1 where it shows some shape
variation.
Fig.7.19: Ratio of cross sections for different cone sizes to that of the cross section for
0.7 cone size, from NLO QCD, with CTEQ5M1 and evaluated at µ = p T .
7.8 Conclusions
Direct photon production continues to be an interesting arena to test modern
perturbative QCD calculations. In this chapter, we have compared the isolated prompt
photon data measured by CDF at s =1.8 TeV with the NLO predictions using the latest
175
parton distribution function. The data show a clear excess over theory for pT < 25 GeV. It
suggests that a more complete theoretical understanding of processes that contribute to
low pT behaviour of the photon cross section is needed which have been addressed by
resuming higher order contributions. It is found that the NLO theory supplemented with
kT correction accounts to a great extent the low pT discrepancy between data and theory.
The LHC run with greatly extended kinematical range and high statistical
precision of data will offer tremendous opportunities to refine our understanding of the
production of photons in hard scattering processes. It is found that the rate of prompt
photon production is expected to be very high at LHC compared to that at Tevatron.
PYTHIA results indicate that the Compton scattering will dominate the production
mechanism in the entire kinematical range considered in the analysis. The pT spectrum of
the relative contributions of LO and NLO cross section shows that the higher order
contribution dominates in the low pT region but decreases in importance considerably at
high pT. The NLO QCD predictions depend only marginally on the choice of scale. The
pT distribution of direct photon cross section is almost insensitive to the different
parameterizations of gluon distributions in the high pT region (pT > 300 GeV), but shows
quite a bit of sensitivity at small pT values. We also see that at low values of pT, the shape
of the rapidity dependence of the photon cross section is very sensitive to the small-x
behaviour of gluon distribution. It means that the η spectrum can be used to constrain the
gluon distributions. It is found that direct photons are produced fairly copiously in the
central rapidity region. Its production cross section depends strongly on the cone size
(used in the isolation cut) and is found to decrease with increasing cone size.
176
----------------------------------------------------------------
Bibliography
----------------------------------------------------------------
Chapter 1
[1.1] The CMS Collaboration: Letter of Intent, CERN/LHCC 92-3 (1992).
[1.2] D. Denegri, “Standard model physics at LHC (pp collisions)”, CERN-PPE/90-181
(1990) 56-117.
[1.3] The CMS Collaboration: Technical Proposal, CERN/LHCC 94-38 (1994) and
references there in.
[1.4] The ATLAS Collaboration: Technical Proposal, CERN/LHCC 94-39 (1994).
[1.5] The ALICE Collaboration: Letter of Intent, CERN/LHCC 94-16 (1993).
[1.6] The LHCB Collaboration: Letter of Intent, CERN/LHCC 95-5 (1995).
[1.7] The CMS Collaboration: Technical Design Report, The Electromagnetic
Calorimeter Project, CERN/LHCC 97-33 (1997).
[1.8] The CMS Collaboration: Addendum to the CMS Tracker TDR, CERN/LHCC 2000-
016 (2000).
[1.9] The CMS Collaboration: Technical Design Report, CMS Tracker, CERN/LHCC 98-
6 (1998).
[1.10] J. B. Singh, Pramana – journal of physics, 54 (2000) 519.
[1.11] A. G. Clark, in Instrumentation in elementary particle physics (VIII ICFA School,
Istanbul, Turkey), edited by S. Kartal, Melville, New York (2000) 115.
[1.12] T. Matsui and H. Satz, Phys. Lett. B178 (1986) 416.
[1.13] H. Satz, ECFA LHC Workshop Proceedings, CERN 90-10, ECFA 90-133 vol.I
(1990) 188.
[1.14] The CMS Collaboration: Engineering Design Review, CMS ECAL Preshower and
Endcap (2000).
[1.15] D. Barney and P. Bloch, CMS TN 1995/114 (1995).
[1.16] A. Peisert and N. Zamiatin, CMS NOTE 2000-061 (2000).
177
[1.17] RD 48 3rd Status Report, CERN/LHCC 2000-009 (Dec.1999).
[1.18] T. Ohsugi et al., Nucl. Instr. and Meth. A383 (1996) 116.
[1.19] TMA, MEDICI V. 2000.4, Users manual, February 2000.
Chapter 2
[2.1] K. G. McKay, Phys. Rev. 84 (1951) 829.
[2.2] G. Dearnaley & A. B. Whitehead, Nucl. Instr. and Meth. 12 (1961) 205.
[2.3] A. S. Schwarz, Nucl. Instr. and Meth. A 342 (1994) 218.
[2.4] S. M. Sze, Physics of Semiconductor Devices, John Wiley and Sons, New York
(1981).
[2.5] A. S. Grove, Physics and Technology of Semiconductor Devices, John Wiley and
Sons, New York (1967).
[2.6] D. K. Schroder, IEEE Trans. El. Dev. 44 (1997) 160.
[2.7] C. T. Sah et al., Proc. Ire 45 (1957) 1228.
[2.8] J. M. Moll, Physics of Semiconductors, McGraw Hill Book Co., New York (1964).
[2.9] S. K. Ghandhi, Semiconductor Power Devices, New York NY: Wiley (1977).
[2.10] R. Van Overstraeten and H. DeMan, Solid State Electron. 13 (1970) 583.
[2.11] W. Fulop, Solid-State Elec. 10 (1967) 39.
[2.12] B. J. Baliga, Modern Power Devices, John Wiley and Sons, New York (1987).
[2.13] V. Anantharam and K. N. Bhat, IEEE Trans. El. Dev. 27 (1980) 939.
[2.14] A. Peisert, Results of R&D and specifications, presented at Preshower Silicon
PRR meeting CERN, Geneva (July 2001).
[2.15] J. Kemmer, Nucl. Instr. and Meth. A169 (1980) 499.
[2.16] A. Peisert and N. Zamiatin, CMS NOTE 2000-061 (2000).
[2.17] A. Peisert, Silicon sensors tests, presented at Preshower Silicon PRR meeting
CERN, Geneva (July 2001).
Chapter 3
178
[3.2] J. T. Watts, “Surface Mobility Modelling”, presented at Computer-Aided Design of
I. C. Fabrication Processes, Stanford University (Aug.3, 1988).
[3.3] D. M. Caughy and R. E. Thomas, Proc. IEEE 55 (1967) 2192.
[3.4] S. Selberher, Analysis and Simulation of Semiconductor Devices, Wien-New York:
Springer Verlag (1984).
[3.5] R. Van Overstraeten and H. DeMan, Solid State Electron. 13 (1970) 583.
[3.6] R. S. Varga, Matrix iterative analysis, Englewood Cliffs, NJ: Prentice Hall (1962).
Chapter 4
[4.1] For example H. Ziock et al., Nucl. Instr. and Meth. A342 (1994) 96; E.Fretwurst et
al., Nucl. Instr. and Meth. A342 (1994) 119.
[4.2] D. Passeri, P.Ciampolini and G. M. Bilei, IEEE Trans. Nucl. Sci. 46(3) (1999) 260.
[4.3] The CMS Collaboration: Technical Proposal, CERN/LHCC 94-38 (1994).
[4.5] J. Kemmer, Nucl. Instr. Methods 226 (1984) 89.
[4.6] S. M. Sze and G. Gibbons, Solid-State Elec. 9 (1966) 831.
[4.7] Y.C. Kao and E. D. Wolley, Proc. IEEE 55 (1967) 1409.
[4.8] V. A. K. Temple, IEDM Tech. Digest 1977 (1977) 432.
[4.9] J. A. Apples and H. M. J. Vaes, IEDM Tech. Digest (1979) 238.
[4.10] R. Stengl et al., IEEE Trans. El. Dev. 33 (1986) 426.
[4.11] F. Conti and M. Conti, Solid-State Elec. 15 (1972) 93.
[4.12] A. Bishoff et al., Nucl. Instr. And Meth. A326 (1993) 27.
[4.13] L. Evensen et al., Nucl. Instr. And Meth. A337 (1993) 44.
[4.14] M. da Rold et al., IEEE Trans. El. Dev. 44 (1997) 721.
[4.15] K. H. Wyllie, Nucl. Instr. And Meth. A409 (1998) 27.
[4.16] M. M. DeSouza et al., Solid-State Electron. 44 (2000) 1381.
[4.19] D. Passeri et al., IEEE Trans. Nucl. Sci. 47 (2000) 1468.
[4.20] N. Demaria, 8th International Workshop on Vertex Detectors, Texel, Netherlands,
20-25 1999.
179
[4.21] N. Demaria, CMS Silicon Tracker Week Meeting, CERN, Geneva, Switzerland, 15
October 1999.
[4.22] A. Peisert, Results of R&D and specifications, presented at Preshower Silicon PRR
meeting CERN, Geneva (July 2001).
[4.23] G. A. Beck, Solid-State Elec., 45 (2001) 183.
[4.24] M. Da Rold et al., IEEE Trans. Nucl. Sci. 46 (1999) 1215.
[4.25] B. S. Avset and L. Evensen, Nucl. Instr. and Meth. A377 (1996) 397.
[4.26] W. R. Th. Ten Kate, Nucl. Instr. and Meth. A253 (1987) 333.
[4.27] E. Fretwurst et al., Nucl. Instr. and Meth. A288 (1990) 1.
[4.28] A. S. Grove, Physics and Technology of Semiconductor Devices, John Wiley and
Sons, New York (1967).
[4.29] A. Rusu and C. Bulucea, IEEE Trans. El. Dev. 26 (1979) 201.
[4.30] C. B. Goud and K. N. Bhat, IEEE Trans. El. Dev. 38 (1991) 1497.
[4.31] V. P. O’Neil and P. G. Alonas, IEEE Trans. El. Dev. 26 (1979) 1098.
[4.32] W. Feiler et al., IEEE Trans. El. Dev. 35 (1988) 684.
[4.33] K. P. Brieger et al., IEEE Trans. El. Dev. 35 (1988) 684.
[4.35] R. Wunstorf, IEEE Trans. Nucl. Sci. 44 (1997) 806.
[4.36] V. Anantharam and K. N. Bhat, IEEE Trans. El. Dev. 27 (1980) 939.
(1981).
Chapter 5
[5.1] P. Bloch et al., IEEE Trans. Nucl. Sci. 47 (2000) 414.
[5.2] S. Albergo et al., IL Nuovo Cimento 112A(11) (1999) 1271.
[5.3] Z. Li et al., Nucl. Instr. and Meth. A409 (1998) 180.
[5.4] R. H. Richter et al., Nucl. Instr. and Meth. A377 (1996) 412.
[5.5] RD 48 3rd Status Report, CERN/LHCC 2000-009 (Dec.1999).
[5.6] H. W. Kraner, Nucl. Instr. and Meth. 225 (1984) 615.
[5.7] L. Evensen et al., Nucl. Instr. Methods A337 (1993) 44.
[5.8] J. Kemmer, Nucl. Instr. Methods 226 (1984) 89.
180
[5.9] A. Longoni et al., Nucl. Instr. Methods A288 (1990) 35.
[5.10] L. Borrello et al., CMS Conference Report 1998/014 (1998).
[5.11] T. E. Burgess et al., J. Electrochem. Soc., Solid State Science 116(7) (1969) 1005.
[5.12] D. R. Kerr et al., IBM Journal of Research and Development 8(4) (1967) 376.
[5.13] D. S. Zoroglu and L. E. Clark, IEEE Trans. El. Dev. 19 (1972) 4.
[5.14] T. Matsushita et al., IEEE Trans. El. Dev. 23 (1976) 826.
[5.15] D. Jaume et al., IEDM Tech. Dig. 1990 (1990) 803.
[5.16] D. Jaume et al., IEEE Trans. El. Dev. 38 (1991) 1681.
[5.17] J. W. Osenbach and W. R. Knolle, IEEE Trans. El. Dev. 37 (1990) 1522.
[5.18] RD-20 Status Report, CERN/DRDC 94-39 (2000) and references there in.
[5.19] T. Ohsugi et al, Nucl. Instr. and Meth. A342 (1994) 22.
[5.20] R. Wunstorf et al, Nucl. Instr. and Meth. A377 (1996) 290.
[5.21] G. Lindstorm et al., Nucl. Instr. and Meth. A426 (1999) 1.
[5.22] G. Lindstorm et al., Nucl. Instr. and Meth. A466 (2001) 308.
(1981).
[5.24] A. Ruzin et al., Nucl. Instr. and Meth. A426 (1999) 94.
[5.25] M. Moll et al., Nucl. Instr. and Meth. A439 (2000) 282.
[5.26] G. Lindstorm et al, Nucl. Instr. and Meth. A465 (2001) 60.
[5.27] M. Moll et al, Nucl. Instr. and Meth. B186 (2002) 100.
[5.28] N. Bacchetta et al., Nucl. Instr. and Meth. A426 (1999) 16.
[5.29] P. Bloch, CMS Internal Note 2000/047 (2000).
[5.30] D. Passeri et al., IEEE Trans. Nucl. Sci. 47 (2000) 1468.
[5.31] E. P. Burte and G. H. Schulze, IEEE Trans. El. Dev. 38 (1991) 1505.
[5.32] E. León-Florián et al., Report CERN-ECP 95-15 (1995).
Chapter 6
[6.1] Y.C. Kao and E. D. Wolley, Proc. IEEE 55 (1967) 1409.
[6.2] M. S. Adler et al., IEEE Trans. El. Dev. 24 (1977) 107.
[6.3] H. Yilmaz, IEEE Trans. El. Dev. 38 (1991)1666.
[6.4] A. Bishoff et al., Nucl. Instr. And Meth. A326 (1993) 27.
[6.5] L. Evensen et al., Nucl. Instr. And Meth. A337(1993) 44.
181
[6.6] M. da Rold et al., IEEE Trans. El. Dev. 44 (1997) 721.
[6.7] A. Bhardwaj et al., Semicond. Sci. and Technol. 16 (2001) 849.
[6.8] M. M. DeSouza et al., Solid-State Electron. 44 (2000) 1381.
[6.9] M. Da Rold et al., IEEE Trans. Nucl. Sci. 46 (1999) 1215.
Chapter 7
[7.1] D. J. Gross and F. Wilczekl, Phys. Rev. Lett. 30 (1973) 1334.
[7.2] J. F. Owens, Rev. Mod. Phys. 59 (1987) 465.
[7.3] F. Halzen and A. Martin, Quarks & Leptons, John Wiley & Sons (1984).
[7.4] Tom Ferbel and W. R. Molzon, Revs. Mod. Phys 56 (1984) 181.
[7.5] CTEQ Collaboration, G. Sterman et al., Handbook of Perturbative QCD, Rev.
Mod. Phys. 67 (1995) 157.
[7.6] E. L. Berger et al., Nucl. Phys. B 239 (1984) 52.
[7.7] E. C. Reid and H. F. Heath, CMS NOTE 2000/063 (2000).
[7.8] G. Ballochi et al., Phys.Lett. B 436 (1998) 222.
[7.9] P. Aurenche et al., Phys. Rev. D 42 (1990) 1440.
[7.10] P. Aurenche et al., Phys. Lett. B 140 (1984) 87 and P.Aurenche et al., hep-
ph/9811382 (1998).
[7.11] CTEQ Collaboration, H. L. Lai et al., Phys. Rev. D 55 (1997) 1280.
[7.12] M. Gluck, E. Reya and A. Vogt, Phys. Rev. D 48 (1993) 116.
[7.13] P. M. Stevenson & H. D. Politzer, Nucl.Phys. B 277 (1986) 756.
[7.14] M. Gluck, E. Reya and A. Vogt, Eur. Phys. J. C5 (1998) 461.
[7.15] H. Baer, J. Ohnemus, and J. F. Owens, Phys. Rev. D 42 (1990) 61
[7.16] P. Aurenche et al., Phys. Rev. D 42 (1990) 1440.
[7.17] H. Baer and M. Rano, Phys. Rev. D 54 (1996) 2017.
[7.18] J. Huston et. al., Phys. Rev. D 51 (1995) 6139.
[7.19] L. Apanasevich et. al., Phys. Rev. Lett., 81, 2642, (1998) and L. Apanasevich et.
al., Phys. Rev. D59, 074007, (1999).
[7.20] T.Sjostrand, L.Lonnblad, and S.Mrenna, hep-ph/0108264, LU TP 01-21, Aug
2001.
[7.21] U. Baur et. al., hep-ph/0005226, May 2000.
[7.22] D. Acosta et. al., Phys. Rev. D 65 (2002) 112003.
182
[7.23] J. Huston et al. (CTEQ Collaboration), Phys. Rev. D 58 (1998) 114034.
[7.24] M.Gluck, LE.Gordon, E.Reya, W.Vogelsang, Phys. Rev. Lett., 73 (1994) 388 &
W.Vogelsang , Private Communications.
[7.25] Dana Sarah Partos, Ph.D.Thesis, Brandeis University, 2001.
[7.26] H. L. Lai et al. (CTEQ Collaboration), Phys. Rev. D 55 (1997) 1280.
[7.27] E. Laenen, G. Sterman and W. Vogelsang, Phys. Rev. Lett. 84 (2000) 4296.
[7.28] M. Spira and M. Dittmar, CMS NOTE 1997/080 (1997).
[7.29] A. D. Martin et al., Eur. Phys. J. C4 (1998) 463 & Nucl. Phys. Proc. Suppl. 79,
(1999) 210.
[7.30] M. Gluck, E. Reya, A. Vogt, Z. Phys. C 67 (1995) 433.
[7.31] M. Bhattacharjee , Ph.D. Thesis, Delhi University, 1997.
183

Kirti Final Complete PHD Thesis

Uploaded by

Copyright:

Available Formats

Kirti Final Complete PHD Thesis

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Kirti Final Complete PHD Thesis

Uploaded by

Copyright:

Available Formats

LARGE PT PARTICLE PRODUCTION IN p-p

COLLISIONS AT LHC ENERGIES

THESIS SUBMITTED TO THE UNIVERSITY OF DELHI

SUPERVISOR: PROF. R. K. SHIVPURI

DEPARTMENT OF PHYSICS AND ASTROPHYSICS

Chapter 2: Physics of Silicon Detectors – A Résumé 26

Chapter 3: Semiconductor Device Simulation 51

Chapter 4: Effect of Metal-Overhang on Silicon Strip Detectors 61

Chapter 5: Effect of Passivation on Breakdown Performance of

Chapter 6: Comparison of Junction Termination Techniques for High-Voltage

Chapter 7: Large Transverse Momentum (pT) Direct Photon Production

2. “Analysis and Comparison of the breakdown performance of Semi-insulator and

3. “Performance characteristics of semi-insulator and dielectric passivated Si strip

4. “High-voltage planar Si detector for high-energy physics experiment: comparison

1. “Influence of electrode geometry on electric field distribution within silicon

2. “Comparison of the passivants on the Breakdown Performance of Si strip

CERN (Conseil Europeen pour la Recherche Nucleaire) is the European

Fig.1.1: Accelerators at CERN.

1.3.1 Some Interesting Features

1.3.2 LHC Experiments

Fig.1.4: Schematic layout of the LHC.

1.4 The Compact Muon Solenoid (CMS) Experiment

1.4.1 CMS Detector Overview

Fig.1.5: Three-dimensional layout of the CMS detector showing internal detectors.

Fig.1.6(b): Longitudinal view of the CMS detector.

1.4.1.3 Muon System

1.4.1.4 Hadron Calorimeter

1.4.1.5 Electromagnetic Calorimeter

Fig.1.8: Explorable mass range of Higgs at LHC.

SUSY Higgs and SUSY Particles Searches: Supersymmetry (SUSY) is

Bd → π +π − ; with J /ψ → µ + µ − and K S → π + π − . These channels would measure

be measured for values up to 20 [1.3].

1.5 Preshower Detector: raison d’être

Fig.1.9: Longitudinal view showing the Preshower coverage.

1.5.1 Need of Preshower: Physics Goal

Fig.1.11: Background to the H → γ γ channel at LHC. The relevant observation is that

1.5.2 Design Criterion Imposed by the Physics Goal

1.6 Motivation for the Present Work

1.7 Thesis Organization

The thesis is organized in seven chapters, a brief description of each chapter is

2.1 Silicon Detector: An Introduction

2.2 Semiconductors: General Properties

A. Band structure in solids: The most important result of the application of

where the proportionality constant µe and µh are the respective mobilities of

2.3 Reverse Biased p - n Junction

2.3.1 Abrupt p - n Junction

2.3.3 Reverse Leakage Current

depletion voltage [2.4, 2.5, 2.7].

Region II: The surface generation component in Region II is given as [2.5],

2.3.4 Breakdown Voltage

Breakdown condition: Assume that an e-h pair is generated within the

A solution of this equation is:

One-sided plane-parallel abrupt junction (non-punch through case): A closed

(2.11), and using equation (2.42)

One-sided plane-parallel abrupt junction (punch through case): It was assumed

The breakdown voltage for PT diodes is given by:

Using equation (2.44), it can be shown that [2.4]

Planar diffuse junction termination (edge effect): In the previous sections, we

Edge effect - Cylindrical junction: Solving the Poisson equation in cylindrical