Università degli Studi di Milano

Facoltà di Scienze Matematiche,

Fisiche e Naturali
Dottorato di Ricerca in Fisica, Astrofisica e
Fisica Applicata. Ciclo XVIII

A new micro-strip tracker for the

new generation of experiments at
hadron colliders

Tutore: Dott. Luigi MORONI

Cotutore: Prof. Stefano FORTE

Tesi di Dottorato di:

Dott. Mauro Emanuele DINARDO

Anno Accademico 2004-2005

ii
Io stimo piú il trovare un vero, benché di cosa
leggiera, che’l disputar lungamente delle massime
questioni senza conseguir veritá nissuna.

Galileo Galilei, Opere. Ristampa dell’Edizione

Nazionale, Firenze 1968, Vol. IV, p. 738.

To Debora, Francesco, Giulia,

Alessandro and to the baby who
is coming to life.
iv
Acknowledgements

I wish to acknowledge Luigi Moroni, together with Daniele Pedrini, Dario

Menasce and Silvano Sala, for giving me the opportunity to accomplish my
Ph.D studies in their group (even after having flood the basement of their
house); I really believe this Ph.D. was one of the best chance I have ever
had. I want to thank in particular Luigi for having taught me so much and
having accurately read and correct this thesis; he was a very careful tutor.
I thank Dario since he was, together with Stefano Magni, my mentor in the
software development. I also thank Silvano, with whom was very pleasant
to work at Fermilab and, besides, for having taught me the backhand at
tennis.
The persons more close to me while I was at Fermilab were Teresa and
Pasquale D’Angelo. I really want to thank Pasquale, together with Silvano
and Gianluca Alimonti, for helping me to perform some of the most inter-
esting measurements reported in this thesis, and Teresa who was so friendly
and helped me to feel at home even being so far from Italy.
I thank Sandra Malvezzi for every single chat we had, she always gave
me the right suggestions and I believe she will always do.
I also thank Gabriele Chiodini who came with Pasquale, Silvano and
me at IUCF and helped us to irradiate the Silicon micro-strip sensors; but
even more I thank Gabriele for having brought me to visit some of the most
interesting places of Indiana . . .
With Laura, Marco and Massimo I shared the three years of the Ph.D.
We had lots of interesting discussions and I hope to work with them in the
future.
I thank Francesca who worked with me during the last year of my Ph.D.
vi

She was always so willing to learn; without her some of the extraordinary
measurements presented in this thesis wouldn’t be possible.
The last but, for sure, not the least I want to thank Lorenzo and Claudia.
With them I spent the most agreeable moments at Fermilab; indeed they
prepare the best barbeque I have ever eaten. In particular I thank Lorenzo
for being extremely helpful at work and to be a very good friend.
Contents

Introduction 1

1 The BTeV experiment 5

1.1 Physics motivation . . . . . . . . . . . . . . . . . . . . . . . . 6
1.1.1 B production at Tevatron . . . . . . . . . . . . . . . . 6
1.1.2 Detector requirements . . . . . . . . . . . . . . . . . . 9
1.2 Detached vertex trigger . . . . . . . . . . . . . . . . . . . . . 12
1.3 Pixel detector . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4 Forward tracker . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.4.1 Silicon strip detector . . . . . . . . . . . . . . . . . . . 16
1.4.2 Straw detector . . . . . . . . . . . . . . . . . . . . . . 18
1.5 RICH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.6 EM-cal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.7 Muon detector . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.8 The Level 1 trigger algorithm . . . . . . . . . . . . . . . . . . 27
1.8.1 Segment finding . . . . . . . . . . . . . . . . . . . . . 27
1.8.2 Track and vertex finding . . . . . . . . . . . . . . . . . 28
1.8.3 Level 1 trigger performance . . . . . . . . . . . . . . . 29
1.9 Sensitivities . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.9.1 Experiment comparison . . . . . . . . . . . . . . . . . 32
1.10 Flavour tagging . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2 Theory of radiation damage in Silicon sensors 37

2.1 Radiation induced crystal damages . . . . . . . . . . . . . . . 37
2.2 Macroscopic effects after irradiation . . . . . . . . . . . . . . 41
viii CONTENTS

2.3 Evolution with time of the effective doping concentration . . 43

3 Silicon sensors radiation damage characterization 45

3.1 The Silicon micro-strip sensor and readout electronics . . . . 46
3.2 Sensor irradiation and measurements . . . . . . . . . . . . . . 49
3.3 Laser setup and measurements . . . . . . . . . . . . . . . . . 53
3.4 Collected charge model . . . . . . . . . . . . . . . . . . . . . . 60
3.4.1 Transport . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.4.2 Induction . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.5 Model accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.6 Results and discussion . . . . . . . . . . . . . . . . . . . . . . 68
3.6.1 Errors treatment . . . . . . . . . . . . . . . . . . . . . 68
3.6.2 Fit results . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.6.3 Remarks on fits . . . . . . . . . . . . . . . . . . . . . . 73
3.6.4 Results on full-depletion voltage and carriers
trapping-time . . . . . . . . . . . . . . . . . . . . . . . 76
3.6.5 From trapping-time to fluence . . . . . . . . . . . . . . 78
3.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

4 Micro-strip detector prototype 83

4.1 The readout chip . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.1.1 The analog section . . . . . . . . . . . . . . . . . . . . 84
4.1.2 Measurements . . . . . . . . . . . . . . . . . . . . . . 87
4.1.3 The digital section . . . . . . . . . . . . . . . . . . . . 90
4.1.4 Efficiency simulations . . . . . . . . . . . . . . . . . . 97
4.2 Hybrid module performance . . . . . . . . . . . . . . . . . . . 98
4.3 Detector performance . . . . . . . . . . . . . . . . . . . . . . 102
4.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

Conclusions 109

A CP violation 117
A.1 Theoretical account for CP violation . . . . . . . . . . . . . . 117
A.2 The CKM matrix . . . . . . . . . . . . . . . . . . . . . . . . . 120
A.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 120
CONTENTS ix

A.2.2 Unitarity triangles . . . . . . . . . . . . . . . . . . . . 121

A.3 CP violation in meson decays . . . . . . . . . . . . . . . . . . 125
A.3.1 Notations and formalism . . . . . . . . . . . . . . . . . 125
A.4 The three types of CP violation in meson decays . . . . . . . 129
A.4.1 CP violation in mixing . . . . . . . . . . . . . . . . . . 129
A.4.2 CP violation in decay . . . . . . . . . . . . . . . . . . 130
A.4.3 CP violation in the interference between decays with
and without mixing . . . . . . . . . . . . . . . . . . . 131
A.5 B decays classification . . . . . . . . . . . . . . . . . . . . . . 132
A.6 Techniques for determining β . . . . . . . . . . . . . . . . . . 133
A.6.1 Results on sin(2β) . . . . . . . . . . . . . . . . . . . . 135
A.6.2 Other modes for measuring sin(2β) . . . . . . . . . . . 136
A.7 Techniques for determining α . . . . . . . . . . . . . . . . . . 136
A.8 Techniques for determining γ . . . . . . . . . . . . . . . . . . 138
A.9 Summary of crucial measurements for CKM physics . . . . . 140
A.10 Rare decays as probes beyond the Standard Model . . . . . . 141
x CONTENTS
List of Figures

1.1 B-meson event. . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.2 βγ of the B’s versus η. . . . . . . . . . . . . . . . . . . . . . . 7
1.3 b − b̄ production angle correlation. . . . . . . . . . . . . . . . 8
1.4 Layout of the BTeV spectrometer. . . . . . . . . . . . . . . . . 11
1.5 BTeV three-level trigger architecture. . . . . . . . . . . . . . . 12
1.6 Pixel detector elements and dimensions. . . . . . . . . . . . . 14
1.7 Sketch of a Silicon tracker plane. . . . . . . . . . . . . . . . . 16
1.8 Sketch of the mechanical support of a Silicon tracker plane. . 17
1.9 Sketch of the main RICH components. . . . . . . . . . . . . . 20
1.10 Cherenkov angle as a function of particle momentum. . . . . 21
1.11 Measurements of the EM-cal resolution. . . . . . . . . . . . . 22
1.12 Perspective view of the muon detector. . . . . . . . . . . . . . 25
1.13 Muon system octants. . . . . . . . . . . . . . . . . . . . . . . 26
1.14 Level 1 trigger background rejection and efficiency. . . . . . . 30

2.1 Point defects in a simple lattice. . . . . . . . . . . . . . . . . 38

2.2 Distribution of vacancies produced by protons and neutrons. . 39
2.3 Dependence of Nef f on the accumulated radiation. . . . . . . 42
2.4 Fluence dependence of leakage current. . . . . . . . . . . . . . 43
2.5 Typical annealing behaviour of N ef f . . . . . . . . . . . . . . . 44

3.1 Sketch of the Silicon micro-strip sensor. . . . . . . . . . . . . 46

3.2 CMS IB2 Silicon micro-strip sensors cross-section. . . . . . . 47
3.3 Layout of the CMS IB2 Silicon micro-strip sensors. . . . . . . 48
3.4 Irradiation as a function of position in the forward tracker. . 49
3.5 IUCF irradiation geometry. . . . . . . . . . . . . . . . . . . . 50
xii LIST OF FIGURES

3.6 IUCF proton beam shape. . . . . . . . . . . . . . . . . . . . . 50

3.7 I-V characteristics for Mod1&Mod2 . . . . . . . . . . . . . . 51
3.8 Leakage current at different fluences. . . . . . . . . . . . . . . 51
3.9 Histogram of the noise for Mod1&Mod2. . . . . . . . . . . . . 52
3.10 Setup apparatus for local measurement of Q-V characteristics. 54
3.11 Sensor cooling system. . . . . . . . . . . . . . . . . . . . . . . 55
3.12 Temperature dependence of the collected charge. . . . . . . . . 55
3.13 Sketch of the laser spot. . . . . . . . . . . . . . . . . . . . . . 56
3.14 Gain and pedestal calibrations. . . . . . . . . . . . . . . . . . 56
3.15 Collected charge per strip after averaging ∼1000 pulses. . . . 57
3.16 Typical signal from an Am241 γ-source. . . . . . . . . . . . . 58
3.17 Typical pedestal peak spectrum. . . . . . . . . . . . . . . . . . 58
3.18 Mod1 Q-V characteristics. . . . . . . . . . . . . . . . . . . . . 59
3.19 Positive charge trapped in the oxide layer. . . . . . . . . . . . 64
3.20 Two-slope linear electric field. . . . . . . . . . . . . . . . . . . 65
3.21 Fit to the two-slope linear electric field simulation. . . . . . . 66
3.22 Parabolic electric field. . . . . . . . . . . . . . . . . . . . . . . 67
3.23 Fit to the parabolic electric field simulation. . . . . . . . . . . 67
3.24 NDF-normalized χ2 distribution. . . . . . . . . . . . . . . . . 69
3.25 Sketch of the overall measurement points on Mod1. . . . . . . 70
3.26 Fit to measurement points at 10 mm from chip side. . . . . . 71
3.27 Fit to measurement points at 40 mm from chip side. . . . . . 71
3.28 Fit to measurement points at 60 mm from chip side. . . . . . 72
3.29 Fit to measurement points at 80 mm from chip side. . . . . . 72
3.30 Fit to measurement points at 90 mm from chip side. . . . . . 73
3.31 Fit to measurement points at 100 mm from chip side. . . . . 73
3.32 Fit to measurement points at 110 mm from chip side. . . . . 74
3.33 Full-depletion voltage along strips. . . . . . . . . . . . . . . . 76
3.34 Trapping-time along strips. . . . . . . . . . . . . . . . . . . . 77
3.35 Laser multiple reflections inside the micro-strip Silicon sensor. 78
3.36 2D-reconstruction of the fluence profile. . . . . . . . . . . . . 79
3.37 Contour plot of the fluence profile. . . . . . . . . . . . . . . . 80
3.38 Reconstructed fluence profile cross-sections. . . . . . . . . . . 80
LIST OF FIGURES xiii

4.1 Layout of the Fermilab Silicon Strips Readout chip (FSSR). . 84

4.2 FSSR analog channel block diagram. . . . . . . . . . . . . . . 85
4.3 FSSR programmable inject mask circuit. . . . . . . . . . . . . 85
4.4 Preamplifier selectable feedback capacitances. . . . . . . . . . . 86
4.5 FSSR flash-ADC. . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.6 FSSR single chip test board. . . . . . . . . . . . . . . . . . . . 88
4.7 Linearity of the DAC of threshold 0. . . . . . . . . . . . . . . 88
4.8 Linearity of the internal pulser DAC. . . . . . . . . . . . . . . 89
4.9 Analog signal after shaper and baseline restorer. . . . . . . . . 89
4.10 Analog signal at different shaping times . . . . . . . . . . . . 90
4.11 Analog signal at different preamplifier gains. . . . . . . . . . . 90
4.12 Analog signal at different injected charges. . . . . . . . . . . . 91
4.13 FSSR chip block diagram. . . . . . . . . . . . . . . . . . . . . 92
4.14 Schematic principle of LVDS signals. . . . . . . . . . . . . . . 93
4.15 The FSSR core organization. . . . . . . . . . . . . . . . . . . 94
4.16 The End-of-column Logic block’s state diagrams. . . . . . . . 95
4.17 The timing diagram of a single pixel hit in a single column. . 95
4.18 Readout efficiency of the FSSR digital section. . . . . . . . . 97
4.19 FSSR hybrid module picture. . . . . . . . . . . . . . . . . . . 98
4.20 Typical threshold curves with sensor disconnected. . . . . . . . 99
4.21 Histograms of the noise. . . . . . . . . . . . . . . . . . . . . . 99
4.22 Dispersion of the thresholds. . . . . . . . . . . . . . . . . . . . 100
4.23 FSSR pure electronic noise and threshold dispersion. . . . . . 101
4.24 FSSR global analog section gain. . . . . . . . . . . . . . . . . 102
4.25 FSSR detector sketch . . . . . . . . . . . . . . . . . . . . . . . 103
4.26 FSSR detector pictures. . . . . . . . . . . . . . . . . . . . . . 103
4.27 Typical threshold curves with sensor connected. . . . . . . . . 104
4.28 Calibrations of the flash-ADC thresholds. . . . . . . . . . . . 105
4.29 Spectrum of the Am241 expressed in mV units. . . . . . . . . 105
4.30 Noise and internal pulser spectra. . . . . . . . . . . . . . . . . 106
4.31 Spectrum of the Am241 expressed in KeV units. . . . . . . . . 107

A.1 The six CKM triangles. . . . . . . . . . . . . . . . . . . . . . 122

A.2 Unitarity triangle. . . . . . . . . . . . . . . . . . . . . . . . . 122
xiv LIST OF FIGURES

A.3 Up to date constraints to CKM parameters. . . . . . . . . . . 124

A.4 CP violation in mixing. . . . . . . . . . . . . . . . . . . . . . 129
A.5 Decay diagram at the tree level for B o → J/ψKS . . . . . . . . 134
A.6 Bd mixing diagrams. . . . . . . . . . . . . . . . . . . . . . . . 134
A.7 Penguin diagram for B o → π + π − . . . . . . . . . . . . . . . . 137
A.8 Two diagrams for B̄so → Ds± K ∓ . . . . . . . . . . . . . . . . . 139
A.9 Diagrams for the two interfering processes B − → K − π + . . . 139
A.10 Loop or “Penguin” diagram for a b → s or b → d transition. . 142
List of Tables

1 Collision time and number of charged tracks per crossing. . . 3

2 Date-rate and number of channels per experiment. . . . . . . 4

1.1 Tevatron main characteristics. . . . . . . . . . . . . . . . . . 6

1.2 Some crucial measurements and detector requirements. . . . . 10
1.3 Pixel detector properties. . . . . . . . . . . . . . . . . . . . . . 15
1.4 Forward Silicon tracker detector properties. . . . . . . . . . . 18
1.5 Forward straw tracker detector properties. . . . . . . . . . . . 19
1.6 RICH detector properties. . . . . . . . . . . . . . . . . . . . . 21
1.7 EM-cal detector properties. . . . . . . . . . . . . . . . . . . . 23
1.8 Muon detector properties. . . . . . . . . . . . . . . . . . . . . 27
1.9 Level 1 trigger cuts. . . . . . . . . . . . . . . . . . . . . . . . 29
1.10 Level 1 trigger efficiencies and background rejection. . . . . . 30
1.11 Yearly sensitivities to CP violating angles. . . . . . . . . . . . 31
1.12 Comparison of the CP reach of several experiments . . . . . . 33
1.13 Results on D2 for four algorithms for Bs decays. . . . . . . . 34
1.14 Results on D2 for four algorithms for B o decays. . . . . . . . 35

3.1 Main characteristics of the Silicon micro-strip sensors. . . . . 47

3.2 Laser main characteristics. . . . . . . . . . . . . . . . . . . . 53
3.3 Fit deviation from two-slope linear electric field simulation . . 66
3.4 Fit deviation from parabolic electric field simulation. . . . . . 68
3.5 Biases on returned fit parameters. . . . . . . . . . . . . . . . . 68
3.6 Fit results on full-depletion voltage. . . . . . . . . . . . . . . . 75
3.7 Fit results on trapping-time. . . . . . . . . . . . . . . . . . . . 75
3.8 Additional measurement points in the highest irradiated region. 76
xvi LIST OF TABLES

3.9 Radiation beam properties. . . . . . . . . . . . . . . . . . . . . 79

4.1 FSSR nominal gains expected from simulations. . . . . . . . . 86

A.1 The three families of quarks. . . . . . . . . . . . . . . . . . . . 118

A.2 The three types of CP violation in meson decays. . . . . . . . 132
A.3 Classification of some B decays. . . . . . . . . . . . . . . . . 133
A.4 Measurements of sin(2β). . . . . . . . . . . . . . . . . . . . . 135
A.5 Other modes useful for cross-checking sin(2β). . . . . . . . . . 136
A.6 B o → ρπ decay modes. . . . . . . . . . . . . . . . . . . . . . . 138
A.7 Required CKM measurements for b’s. . . . . . . . . . . . . . . 140
Introduction

This thesis concerns the development and characterization of a prototype

Silicon micro-strip detector that can be used in the forward (high rapidity)
region of a hadron collider. These detectors must operate in a high radiation
environment without any important degradation of their performance. The
innovative feature of these detectors is the readout electronics, which, being
completely data-driven, allows for the direct use of the detector information
at the lowest level of the trigger. All the particle hits on the detector can be
readout in real-time without any external trigger and any particular limita-
tion due to dead-time. In this way, all the detector information is available
to elaborate a very selective trigger decision based on a fast reconstruction
of tracks and vertex topology. These detectors, together with the new ap-
proach to the trigger, have been developed in the context of the BTeV R&D
program; our aim was to define the features and the design parameters of
an optimal experiment for heavy flavour physics at hadron colliders.

Application of these detectors goes well beyond the BTeV project and, in
particular, involves the future upgrades of experiments at hadron colliders,
such as Atlas, CMS and LHCb. These experiments, indeed, are already
considering for their future high-intensity runs a new trigger strategy a la
BTeV. Their aim is to select directly at trigger level events containing B-
hadrons, which, on several cases, come from the decay of Higgs bosons, Z o ’s
or W ± ’s; the track information can also help on improving the performance
of the electron and muon selection at the trigger level. For this reason, they
are going to develop new detectors with practically the same characteristics
as those of BTeV. To this extent, the work accomplished in this thesis could
serve as guide-line for those upgrades.
2 Introduction

I contributed to all the phases of this development, from the initial design
to the final assembly of the prototypes and their tests. This included stud-
ies of the behaviour of the micro-strip sensors when exposed to the highly
non-uniform irradiation expected in the forward region of a collider experi-
ment. I also developed a method for their full characterization. The method
allows for a complete bidimensional mapping, point to point, of the sensor
characteristics over its entire active area. Information is gathered through
the Q-V characteristic, measured scanning the sensor with an infra-red laser
source. I built the setup that I used to perform the measurements on the
Silicon micro-strip sensors, which I previously irradiated non-uniformly up
to a peak fluence of ∼1014 1 MeV equivalent neutrons per cm2 . I developed
an analytical model to fit the Q-V characteristics and extract the local full-
depletion voltage and the local carriers trapping-time. With this method
one can even obtain the profile of the absorbed dose. The description of the
method is the subject of chapter 3.
I contributed to the definition of the design requirements for the readout
chip of the detector and I certified, with simulations, the readout efficiency
capabilities of the digital section at different luminosities. These measure-
ments are reported in chapter 4.
Subsequently I prepared a test-stand at Fermilab and Milan by which
I performed a complete characterization of the chip prototypes. The mea-
surements I performed on the first version of the chip helped the electronics
engineers to tune the design parameters for the second and last version.
The characterization procedure was based on the measurement of the ana-
log section linearity, the test of all the digital section functionalities, and the
determination of the global performance of the chip in terms of gain, noise
and threshold dispersion. These measurements are reported in chapter 4.
After having fully characterized the sensors and the chip, separately,
and having understood their behaviour and performance, I assembled them
into the Silicon micro-strip detector prototype which I characterized and
calibrated with an Am241 γ-source. These measurements are reported in
chapter 4.
The scientific community has already demonstrated interest for this de-
tector design. Indeed, the collaborations of CBM (Compressed Baryonic
 3

Matter experiment, at GSI) and PHENIX (Pioneering High Energy Nuclear

Interaction eXperiment, at BNL) are considering the possibility of using it
for their Silicon tracker. Both of them are experiments of heavy-ion physics
that need a performant trigger based on tracking and vertexing, that could
efficiently detect particular processes in a high-multiplicity environment.
The detector I studied provides exactly those capabilities.
In chapter 1 I describe the BTeV experiment stressing its innovative
features and presenting its extraordinary sensitivities. In chapter 2, as in-
troduction to my study on radiation hardness, I report the most up to date
knowledge on radiation induced damages in Silicon detectors. Concluding
remarks on my work are summarized in the last chapter.

Challenges to the new generation of experiments

The increasing average number of charged tracks generated at each bunch

crossing and the decreasing collision time, see Tab. 1, results in unprecedent
data rates, see Tab. 2, and track multiplicities. At the same time the gran-
ularity of tracking detectors in HEP experiments has improved, as can be
gauged by the very large number of channels which constitute the majority
of the total number of channels of a HEP experiment, see Tab. 2.

Accelerator Type Luminosity Collision # charged

[1030 cm−2 s−1 ] time [ns] tracks
LEP e →← ē 65 2200 ∼20
Tevatron p →← p̄ 200 396 ∼600
LHC p →← p 10000 25 ∼800

Table 1: Luminosity, collision time and average number of charged tracks per
each bunch crossing at the Large Electron Positron Collider (LEP), Tevatron
and Large Hadron Collider (LHC) [1, 2, 3, 4].

In the new generation of experiments at hadron colliders, highly sophis-

ticated selection criteria, before storing data on archive, are required. Here
the tracking system plays a crucial role. Tracking detectors, besides han-
4 Introduction

dling an ever increasing number of tracks per unit time, are exposed to very
high radiation levels.

Experiment Accelerator # channels Data rate

Input Output
Aleph, Opal LEP ∼250−500 K ∼45 KHz ∼15 Hz
L3, Delphi
CDF, D0 Tevatron ∼1 M ∼7 MHz ∼50 Hz
BTeV Tevatron ∼23 M ∼2.5 MHz ∼2.5 KHz
Atlas, CMS LHC ∼100 M ∼500 MHz ∼100 Hz

Table 2: Date-rate and number of channels per experiment.

Information provided by the tracking detectors is usually used just in

the off-line analysis programs, since, up to few years ago, it was practically
impossible to perform track reconstruction and vertex finding in real-time
at trigger level. Currently, with the improvement of the Application Spe-
cific Integrated Circuit (ASIC) technology, Field Programmable Gate Array
(FPGA) and increased speed of networks it is possible to fully exploit this
information at trigger level. The main physics motivations to use tracker
information at the lowest level of the trigger are:

1. identify special decay topology: some particles produced during

collisions, specifically D-hadrons and B-hadrons, have life-times of the
order of picoseconds. In particular B-hadrons selection at trigger level
is very important to tag the presence of several much heavier particles
such as Higgs bosons, Z o ’s and W ± ’s. Therefore a trigger based on
impact parameter detection is a powerful tool for event selection

2. improve calorimetric and muon measurements: the recon-

structed tracks matched with electromagnetic-calorimeter clusters can
improve electron identification, as well as stubs in the muon system
can ameliorate muon identification and momentum resolution

The R&D program carried out by the BTeV collaboration has addressed
and resolved the crucial issues highlighted above.
Chapter 1

The BTeV experiment

The BTeV experiment was designed to challenge the Standard Model (SM)
explanation of CP violation, mixing and rare decays of beauty and charm
quark states (see appendix A) and find out what lies beyond the SM. In
doing so, the BTeV results would also shed light on phenomena associated
with the early universe such as why the universe is made up of matter and
not antimatter. The interest on flavour physics and CP violation arises from:

• the observed baryon asymmetry of the Universe requires CP violation

beyond the SM

• almost all extensions of the SM contain new sources of CP and flavour

violation

• flavour physics and CP violation may help to distinguish between dif-

ferent models

Therefore the BTeV aim is not only to measure the CKM elements, but to
over constrain the SM by many “redundant” measurements in order allow
for detailed cross-checks.
The b sector is particularly interesting for the following reasons:

• all the three angles of the unitarity triangle involving the beauty quark
are of the same magnitude, this leads to large CP violating effects

• some of the hadronic physics is understood model-independently

(mb  ΛQCD )
6 The BTeV experiment

• top quark loops are neither GIM nor CKM suppressed (large mixing,
rare decays)

1.1 Physics motivation

BTeV is designed to run at the Tevatron in order to exploit the huge poten-
tials on B production offered by hadron colliders. In the next sections will
be described the physics motivation for the BTeV technical design.

1.1.1 B production at Tevatron

The Tevatron running at luminosity of 2×10 32 cm−2 s−1 , produces 4×1011 b-

hadrons per year of operation. Differently from b-factories that run at the b−
b̄ resonance Υ(4S), at Tevatron all B species, B u , Bd , Bc , Bs and b-baryons
are produced at the same time. The total cross-section is 75 mb, while the
b − b̄ production cross-section is 1/500 of the total cross-section [5, 6, 7],
to be compared with 1 nb at b-factories. In Tab. 1.1 are summarized the
Tevatron main characteristics.

Tevatron characteristics Values

Luminosity 2×1032
Interactions per second 15×106
B − B̄ each 107 s 3×1011
B events per background 1/500 (only 1/500000
event are “interesting” B decays)
Bunch spacing 396 ns
Luminous region length Z = 30 cm (r.m.s.)
Luminous region radius σx ' σy ' 30 µm
Interactions per beam crossing <6>

Table 1.1: Tevatron main characteristics.

The number of interactions per second results in a data volume that is

too large to be stored on archives for analysis. The trigger must select from
this huge rate the approximately 1000 events per second that contain B-
1.1 Physics motivation 7

Figure 1.1: Event containing a B-meson, that eventually decays into two
particles, showing the significantly detached secondary vertex from the pri-
mary vertex.

hadrons and enter the spectrometer. A very sophisticated trigger is needed

to reject the huge number of typical interactions which involve only light
quarks. To form a trigger, we must exploit properties of events with B-
hadrons that differentiate them from the much larger number of ordinary,
or “minimum bias,” events. Figure 1.1 illustrates the key characteristic
that distinguishes B events. The B’s produced in the interaction travel a

Figure 1.2: βγ of the B’s versus η.

short distance, between zero and many millimeters from the point of the
interaction (with a most probable decay length of 3 mm) and then decay
8 The BTeV experiment

into two or more (typically 5) particles. The presence of these “detached

vertices” or “secondary vertices” is the signature of a B event. However, this
requires the trigger to reconstruct tracks and assemble them into vertices
to find those with evidence of detached vertices. This task must be done in
quasi-real time so that a decision can be made on average every 396 ns. This
represents a formidable challenge that has not been achieved yet in particle
physics.

Figure 1.3: The production angle (in degrees) for the hadron containing the
b-quark plotted versus the production angle for the hadron containing the
b̄-quark in the same event, from the Pythia Monte Carlo generator. Zero
degrees represents the direction of the incident proton and 180 o the incident
antiproton.

BTeV is designed to cover the forward region of the p − p̄ interaction

from 10 mrad to 300 mrad (1.9 < η < 4.5). There are several advantages
to operate in this region and they come from the peculiar features of the B
production at hadron colliders:

• the B-hadrons are much faster than in the central region around η = 0.
1.1 Physics motivation 9

This means that their decay products are much less deviated from
their trajectories by multiple Coulomb scattering and, therefore, can
be reconstructed with better precision

• there is a strong angular correlation between the b and b̄ directions,

which greatly facilitates the flavour-tagging since both the produced
B-hadrons are likely to fall within the spectrometer acceptance

Figure 1.2 shows the behaviour of βγ of the B-hadrons as a function of η

√
obtained by the Pythia Monte Carlo generator at s = 2 TeV. It can be
clearly seen that, near η = 0, βγ is at a minimum value around 1, while,
at higher |η|, βγ can easily reach values around 6. The pseudo-rapidity η is
defined as:
η = − ln (tan (θ/2)) (1.1)

θ is the angle of the particle with respect to the beam direction. In Fig. 1.3
the production angle of the hadron containing the b-quark is plotted versus
that of the hadron containing the b̄-quark. There is a very strong correlation
peak near the proton and the antiproton direction. Both these features can
be easily understood recalling that the dominant mechanism for b-quark
production at this energy is gluon-gluon fusion, i.e. a two-body process.
Whenever the two gluons have strongly unbalanced Feynman-x, the center
of mass of the produced b− b̄ pair is boosted along the direction of the higher
momentum gluon.

1.1.2 Detector requirements

The spectrum of detector requirements following from a choice of a general

menu of physics measurements related to CP violation in B decays, see ap-
pendix A, are reported in Tab. 1.2. The considerations on vertex separation,
K and π separation, photon detection and lepton identification, summarized
in this table, leads to the following BTeV design requirements [8]:

pixel detector: based on several pixel planes placed normally to the beam
direction along the interaction region, it is used to trigger on detached
heavy quark decay vertices at the first trigger level. Fine-segmented
10 The BTeV experiment

Physics Decay modes Detector properties

quantities V K/π γ τ L
√ √ √
sin(2α) B o → ρπ → π + π − π o
√ √ √
cos(2α) B o → ρπ → π + π − π o
√ √ √
sign(sin(2α)) B o → ρπ & B o → π + π −
√ √ √
sin(γ) Bs → Ds± K ∓
√ √
sin(γ) B − → D̄ 0 K −
√ √ √
sin(γ) B o → π + π − & Bs → K + K −
√ √ √ √ √
sin(2χ) Bs → J/ψη 0 , J/ψη
√
sin(2β) B o → J/ψKS
√ √ √
cos(2β) B o → J/ψK o , K o → π`ν
√
cos(2β) B o → J/ψK ∗o & Bs → J/ψφ
√ √ √
xs Bs → Ds+ π −
√ √ √ √
∆Γ for Bs Bs → J/ψη 0 , Ds+ π − , K + K −

Table 1.2: Some crucial measurements and the relative detector require-
ments. V: vertex separation; K/π: K and π separation; γ: photon detection;
τ : superb decay time resolution; L: lepton identification.

pixel detectors are crucial since they provide precise and unambiguous
three-dimensional space points for track reconstruction

dipole magnet: centered at the interaction region, it places a magnetic

field of 1.6 T on the pixel detector thus allowing the use of momentum
determination in the trigger

Silicon micro-strips & straws tubes: for precision tracking. This sys-
tem, when coupled with pixels, provides excellent momentum and mass
resolution out to 300 mrad

Ring Imaging CHerenkov detector (RICH): for excellent charged

particle identification. The RICH provides hadron identification from
3−70 GeV and lepton identification from 3−20 GeV, out to the full
aperture of 300 mrad

ElectroMagnetic calorimeter (EM-cal): for reconstructing final states

1.1 Physics motivation 11

with single photons, π o ’s, η’s or η 0 ’s and identifying electrons. It’s

made with a high quality PbWO4 crystals with excellent energy reso-
lution, position resolution and segmentation

muon detector: for muon identification. It consists of a dedicated de-

tector with steel toroids instrumented with proportional tubes. This
system has the ability to both identify single muons above momenta
of ∼10 GeV/c and supply a di-muon trigger

detached vertex trigger at Level 1: using the pixel detector, detects

characteristic B decays topology. It makes BTeV efficient for most
final states, including purely hadronic modes. The trigger eliminates
from its calculations tracks of very low momentum particles. These
particles can badly scatter and their tracks can appear to be erro-
neously detached from the primary vertex, resulting in “fake” triggers

Figure 1.4: Layout of the BTeV spectrometer.

Figure 1.4 shows the layout of the BTeV spectrometer. I will start describing
BTeV from its most innovative and challenging part, the detached vertex
12 The BTeV experiment

trigger.

1.2 Detached vertex trigger

The challenge for the BTeV trigger and data acquisition system is to re-
construct particle tracks and interaction vertices for every interaction that
occurs in the detector, and to select preferably interactions with B decays.
The trigger performs this task using 3 stages, see Fig. 1.5, referred to as
Levels 1, 2 and 3:

L1: looks at every interaction, and rejects at least 98% of background based
on full track and vertex reconstruction using a Silicon pixel detector

L2: uses L1 results and performs more refined analyses for data selection

L3: performs a complete analysis using all of the data available for an in-
teraction

500 GB/s
2.5 MHz
200 kB/event BTeV detector
7
Front-end electronics > 2 x 10 channels

m PIX m Level-1

L1 muon Global
Level-1

Level 1 Rate Reduction : ~50x L1 vertex

Level-1 Buffers
Information Transfer
Control Hardware
GL1 accept

Req. data for ITCH

crossing #N
Crossing #N
50 kHz 12.5 GB/s Level 2/3 Crossing Switch RDY

250 kB/event Level-2/3 Processor

#1
Farm
#2
Level 2/3 Rate Reduction : ~20x #m-1
#m
Level-2/3 Buffers

Level-3 accept
Data Logging
2.5 kHz 200 MB/s
250 kB / 3.125 = 80 kB/event

Figure 1.5: BTeV three-level trigger architecture. A data-compression al-

gorithm is used before storing data on archive, giving the reduction factor
3.125 in Level 2/3.

The total effect of the trigger is to reject > 99.8% of background and keep
> 50% of B events that are actually used for physics analysis. The Data
Acquisition System (DAQ) [10] saves all of the detector data in memory
for as long as is necessary for Level 1 to analyze each interaction (0.5 ms
1.2 Detached vertex trigger 13

on average for L1) and moves data to L2/3 processing units and archival
storage for selected interactions.
The trigger system is data-driven since it actually deals with “beam
crossings,” treating each crossing as a separate computing problem and try-
ing to determine whether any of the interactions are B events. Since the
crossings have a variable number of interactions and the individual inter-
actions have varying complexity, the time it takes to compute for an indi-
vidual crossing is highly variable. In order to keep all processing elements
busy, BTeV’s trigger and DAQ have no fixed latency at any level. Decisions
are taken in variable amounts of time and transmitted as soon as they are
known, like an asynchronous system 1 , without any requirement of time or-
dering. This in turn requires massive amounts of buffering throughout the
system. To limit the amount of that needs to be buffered, the front end
electronics needs to sparsify the data by transferring only channels above
a certain threshold.
The Level 1 trigger consists of:

pixel trigger: reconstructs tracks and vertices for every bunch-crossing. It

is able to carry out track and vertex reconstruction at the bunch-
crossing rate because of the very high-quality, low-noise, three-
dimension tracking information provided by the pixel detector

muon trigger: looks for tracks in the muon detector consistent with muons
originating in the interaction region. The main jobs of this system is
to identify di-muons events, in our quest to collect a large sample of B
decays containing J/ψ’s, and to help calibrate the Level 1 pixel trigger

In the next sections I will describe the detectors that constitute the BTeV
spectrometer. I will start by its core, the pixel detector.

1
Current modern collider experiments (CDF, CMS, etc . . . ) operate with a pipelined,
synchronous Level 1 trigger system: all trigger data move in lockstep with the beam
crossing clock through the trigger decision chain, therefore no time markers are required
for the data. The synchronicity must be maintained throughout the entire trigger system
with hundreds of boards.
14 The BTeV experiment

1.3 Pixel detector

The key features of the pixel detector are excellent spatial resolution, ease
of tracking pattern recognition, radiation hardness, material thinness and
readout of data fast enough for use in the Level 1 BTeV trigger system.
In particular, pixel detectors have very few noise hits, this, together with
remarkable radiation hardness properties, enables the detector elements to
be placed very close to the beam (in the same vacuum ∼10 −8 torr, outside
several wires used for RF shielding) minimizing track extrapolation errors.

127.5 cm
4.25 cm
Half stations

Y
... X ...
Beam
Z
... ...

10 cm Station
2.125 cm

Two views of 10 cm 1.2 cm

an half station
Beam hole
Pixels orientation

Figure 1.6: Pixel detector elements and dimensions.

The pixel sensors are made of Silicon doped n + /n/p+ type, with pixel
dimensions of 50×400 µm2 and 250 µm thick; they are organized in 22
columns and 128 rows. Each sensor is bump-bonded to the Fermilab PIXel
readout chip (FPIX), made with 0.25 µm CMOS technology. FPIX performs
a data-driven sparsified readout and provides the following information:

• channel up threshold (row-column coordinate)

• 8 bit BCO counter (the time stamp)

• 3 bit analog information

1.3 Pixel detector 15

Properties Values
Total station radiation length 3%
(incl. RF shielding)
Total pixels ∼2.3×107
Readout analog (3 bits, i.e. 8 thresholds)
Trigger data-driven
(signals are used in Level 1)
Rate requirements time between beam crossings: 396 ns,
132 ns BCO also fully supported
Noise requirement desired: < 10−6 per channel/crossing
required: < 10−5 per channel/crossing
Resolution better than 9 µm
Angular resolution better than 0.1 mrad
Radiation tolerance > 6 × 1014 particles per cm2
(10 years of BTeV operation)
Power per pixel ∼60 µW
Occupancy ∼10−4 (at 396 ns BCO)
Vertex separation resolution 138 µm (from simulations)
Proper time resolution 46 fs (from simulations)

Table 1.3: Pixel detector properties.

The pixels detector is composed of thirty stations; each station is divided

in two half L-shaped stations and each half station has two views, one mea-
suring X with high precision and Y with lower precision and the second
measuring Y with high precision and X with lower precision (see Fig. 1.6).
In order to reduce the noise and increase the sensors life-time the pixel de-
tector has to operate at −5o C. During beam refill the half stations of the
detector are moved away to ∼ ±2 cm from the beam, using a system of
actuators and motion sensors. When the beam is stable the detectors are
moved close to the beam for data taking, with a reproducibility < 50 µm.
Other important properties of the pixel detector are reported in Tab. 1.3.
16 The BTeV experiment

1.4 Forward tracker

The BTeV forward tracking system is based on seven stations of straw and
Silicon strip planes, which cover an acceptance of ∼300 mrad in the forward
region. Three stations are placed in the dipole magnet, three stations in the
field-free region just upstream of the RICH and one station just downstream
of the RICH. The major functions of the forward tracking system are to
provide high precision momentum measurements for tracks found in the
pixel system, to reconstruct and measure all parameters for tracks which do
not pass through the vertex detector (such as K S and Λo daughter tracks)
and to project tracks into the RICH counters, EM-cal and muon detectors.

1.4.1 Silicon strip detector

Silicon strip planes are placed in the innermost region, around the beam
pipe, where the particle fluence is very high, and cover the acceptance from
FE electronics

FE electronics
FE electronics

FE electronics
cm
4

30.6 cm
5.

FE electronics
FE electronics
FE electronics

FE electronics

31.6 cm

Figure 1.7: Sketch of a forward Silicon tracker plane. The two pairs of
sensors on each ladder are readout separately by the front-end electronic
chips placed at the two ends of the same ladder. There is some overlap
between adjacent ladders to ensure good efficiency over the entire plane.

the beam pipe to the inner edge of the forward straw system, which starts
at 13 cm. The design consists of stations with three planes of 320 µm
thick single-sided silicon strip detectors with 100 µm pitch (providing a
√
σ = 100 µm/ 12 = 29 µm resolution, adequate for the physics goals). The
1.4 Forward tracker 17

Figure 1.8: Sketch of the mechanical support of a forward Silicon tracker

half-plane. A cooling duct runs through the structure and reaches the regions
where the readout electronics is located and the heat load is concentrated.

Silicon sensors, having an area of ∼7.9×7.9 cm 2 , are arranged in ladders of

2 daisy-chained sensors each, in such a way that four adjacent ladders form
a plane, as illustrated in Fig. 1.7.
The ladders are mounted on a low-mass carbon fiber support (see
Fig. 1.8) that can be stacked and properly rotated to provide three views in
each station: X, U and V . The two stereo views, U and V , are at ±11.3 o
around the Y bend coordinate. Each plane contains 6144 readout channels;
the entire system of seven stations has 129024 channels in total.
The Silicon sensors are the standard p/n type and are produced with the
same technology developed by the CMS collaboration for their IB2 detectors.
A description of these sensors and their radiation hardness performance are
the subject of chapter 3.
The sensors are readout by the Fermilab Silicon micro-Strip Readout
chip (FSSR), made with 0.25 µm CMOS technology. FSSR performs a data-
driven sparsified readout and provides the same information as the pixel
readout chip since it has the same digital back-end section. A description of
the FSSR chip and its performance are the subject of chapter 4. The front-
end electronics is distributed along the two opposite edges of each plane
and is cooled to ∼ −5o C by a fluid circulating in a duct embedded in the
18 The BTeV experiment

Properties Values
Total station radiation length 0.4%
Total channels ∼1.3×105
Readout analog (3 bits, i.e. 8 thresholds)
Trigger data-driven
(signals are used in Level 2/3)
Rate requirements time between beam crossings: 396 ns,
132 ns BCO also fully supported
Noise requirements < 10−3 per channel/crossing
Radiation tolerance > 1.6 × 1014 particles per cm2
(10 years of BTeV operation)
Power per channel ∼4 mW
Occupancy ∼2.4% (at 396 ns BCO)

Table 1.4: Forward Silicon tracker detector properties.

support structure around the periphery of the plane (see Fig. 1.8). Other
important properties of the forward Silicon tracker detector are reported in
Tab. 1.4.

1.4.2 Straw detector

The outer region of the forward tracker is instrumented using straw tube
drift chambers. Straws have been chosen because they can be used to make
large chambers with small cell size and because they can be built to surround
the beam pipe without requiring a massive frame near the beam. The basic
element in the construction of a detector station is the “module” consisting
of 48 straws of 4 mm diameter, arranged in 3 rows. The gas used is Argon-
CO2 with a drift time, in the 4 mm straw section, of ∼50 ns. In order to keep
the straws at a constant humidity, dry nitrogen is flown at room temperature
in the volume surrounding the straws. Other important properties of the
straw detector are reported in Tab. 1.5.
1.5 RICH 19

Properties Value
Total station radiation length 0.9%
Total channels ∼5.3×104
Trigger data-driven
(signals are used in Level 2/3)
Resolution better than 100 µm
Occupancy ∼0.2% (at 396 ns BCO)

Table 1.5: Forward straw tracker detector properties.

1.5 RICH

Charged particle identification is an absolute requirement for an experiment

designed to study the decays of b and c quarks. The forward geometry is
well suited for a Ring Imaging CHerenkov detector (RICH), that provides
powerful particle identification capabilities over a broad range of momentum.
Even with the excellent mass resolution of BTeV, there are kinematic regions
where signals from one final state could overlap those of another final state.
For example, Bs → Ds K − signal must be distinguished from B s → Ds π −
background in order to measure the CKM phase γ (see appendix A). These
ambiguities can be eliminated almost entirely by an effective particle identi-
fier. The RICH detector consists of two independent systems, whose signals
are used in Level 2/3 trigger:

gas radiator: has a 3 m long C4 F8 O gas volume. Charged particles ra-

diate Cherenkov light in this medium. The light is focused with a
segmented mirror onto an array of photo-detectors sensitive to light
between 280−600 nm. The photo-detectors are Multi-Anode Photo-
Multiplier tubes (MAPMT)

liquid radiator: used mainly for separating kaons and protons below
10 GeV/c, consists of a liquid C5 F12 radiator, approximately 1 cm
thick, placed in front of the gas volume. Cherenkov photons generated
in this medium exit the sides of the gas tank and are detected in an
array of Photo-Multiplier Tubes (PMT)
20 The BTeV experiment

Figure 1.9: Sketch of the main RICH components.

The momentum range over which excellent hadron identification is re-

quired is between 3 and 70 GeV/c. The low momentum hadron identifica-
tion optimizes flavour tagging, whereas the high momentum range enables
to separate π’s and K’s from two body B-meson decays. Besides provid-
ing excellent hadron identification, the RICH detector is also an integral
part of the lepton identification system in the solid angle between 200 and
300 mrad. It is the only detector element available to distinguish e, µ and
hadron species within that solid angle, as the muon detector and the elec-
tromagnetic calorimeter have smaller solid angle coverage.
The separation of charged hadrons into different species is accomplished
in the data analysis by characterizing each charged track with a set of proba-
bilities for being an electron, muon, pion, kaon or proton. From a knowledge
of the distribution of Cherenkov angular resolutions per track such proba-
bilities can be derived. For example, the difference in emission angle of
Cherenkov photons from pions and kaons at 70 GeV/c is 0.44 mrad, so
achieving a resolution per track of 0.11 mrad would give a separation of 4σ.
1.5 RICH 21

600 C5F12

Cherenkov angle (mrad)

K
400 p

200

C4F8O

0
0 10 20 30
Momentum (GeV/c)

Figure 1.10: Cherenkov angle for various particle species as a function of

particle momentum for the gaz C4 F8 O (n=1.00138) and the liquid C5 F12
(n=1.24) radiators.

Separation improves dramatically as momentum decreases. Furthermore,

the average Cherenkov resolution per track can be understood in terms of
the average Cherenkov resolution per photon and the number of photons.
A separation of at least 4σ for π, K and p in the momentum range of
3−70 GeV/c is required. Other important properties of the RICH detector
are reported in Tab. 1.6.

Radiator type Angle Cherenkov ring Photons

resolution thickness (r.m.s.) per track
Gas 0.12 mrad < 0.85 mrad > 50
Liquid 1.9 mrad < 6.5 mrad > 12

Table 1.6: RICH detector properties.

22 The BTeV experiment

1.6 EM-cal

A thorough investigation of B decays requires not only the ability to track

and identify charged particles, but also the ability to reconstruct photons.
To address many of the B-physics issues, one needs to disentangle various
isospin components of the decays. This inevitably involves decay modes
containing π o ’s. Detection of neutral pions is critical, for example, in ex-
tracting the α angle of the unitary triangle using B → ρπ → π + π − π o or
B → ρ+ ρ− → π + π − π o π o . It is also crucial to detect η(0 )’s and isolated
photons. The decay mode Bs → J/ψη(0 ) used for the determination of the
angle χ involves either η → γγ, η 0 → π + π − η or η 0 → ργ. Other important
decays involving direct photons are B → γK ∗ (ρ or ω) (see appendix A).

σX,mm

3
Fit: σ x = a ⊕ b/ √ E
2.5 Exp: a = 0.16 ± 0.07, b = 2.80 ± 0.08
MC : a = 0.17 ± 0.01, b = 2.77 ± 0.01
2

1.5

0.5

0 5 10 15 20 25 30 35 40 45
BeamEnergy,GeV

Figure 1.11: Measurements of the EM-cal resolution. On the left: energy

resolution vs. beam energy. On the right: position resolution vs. beam
energy.

Total absorption shower counters made of scintillating crystals have su-

perb energy and spatial resolutions. The crystals act as both the shower de-
velopment media and scintillation light emitter. Since the entire calorimeter
is used to measure the energies of photons, the resolution can be excellent.
Lead tungstate (PbWO4 ) crystals are distinguished by their high density
(8.3 g/cm3 ), short radiation length (0.89 cm), small Moliere radius (2.2 cm)
and short relaxation time (15 ns for the major component) as well as their
1.6 EM-cal 23

high tolerance to radiation. The light output, of 10 photoelectrons per

MeV into 2 inch Photo-Multiplier Tube (PMT) with a standard bi-alkali
photocathode, is modest compared to many other scintillation crystals, but
adequate. For experiments operating in a very high particle density envi-
ronment, it is very important that signals from two particles do not overlap
in space and time very often. The smaller the Moliere radius, the more
compact are the showers created by photons and the less frequently two
of them do overlap in space. The shorter the scintillation signal, the less
likely that two of them overlap in time. The dense nature of the PbWO 4
crystals makes it possible to construct a compact calorimeter. The shorter
calorimeter gives hadron showers less room to spread out when their parent
hadrons interact in the calorimeter, making them less likely to overlap with
photon signals.

Properties Values
PWO crystal radiation length 0.89 cm
Interaction lengths 22.4 cm
Trigger signals are used in Level 2/3
Refractive index 2.3
Maximum of emission 440 nm
Radiation tolerance after 10 Mrad energy and position
resolution should not deteriorate by
more than a factor of 2
Working temperature 15−20o C
Temperature stability 0.1o C, since the light output depends
strongly on the temperature: 2.3%/ o C

Table 1.7: EM-cal detector properties.

The EM-cal consists of ∼10000 crystals, each having a 28×28 mm 2 cross-

section in the back and 220 mm length. They are slightly tapered in shape
so that they can be arranged in a projective geometry where all the crystals
are pointing to a place near the interaction point. The projective geometry
secures better resolutions, in particular, position resolution, especially in the
24 The BTeV experiment

outermost area of the calorimeter. In order to avoid lining up gaps between

crystals with potential paths of photons, the convergence point is displaced
from the interaction point by 10 cm both in the horizontal and vertical
directions.
The first plot of Fig. 1.11 shows the measured energy resolution for a
5×5 array of crystals, superimposed with the fit function:
r
b2 c2 √
σE /E = a2 + + 2 = a ⊕ b/ E ⊕ c/E (1.2)
E E
a: arises from calibration errors, shower longitudinal leakage and non-
uniformity of the light collection efficiency along the length of the
crystals (expected 0.35%)

b: arises from photon statistics variations and the transverse leakage of

shower outside the array (expected 1.68%)

c: arises from the momentum measurement errors due to multiple scattering

of the electrons in the beam line upstream the prototype

The second plot of Fig. 1.11 shows the measured position resolution with
superimposed the fit function:
r
b2 √
σx = a2 + = a ⊕ b/ E (1.3)
E
Other important properties of the EM-cal detector are reported in Tab. 1.7.

1.7 Muon detector

The muon detector has two primary functions:

muon identification: many of the experiment’s physics goals rely on effi-

cient muon identification with excellent background rejection. Muon
identification is important for rare decay searches, CP violation studies
which require tagging, studies of beauty mixing, semileptonic decays
and searches for charm mixing

J/ψ and prompt muon trigger: besides selecting interesting physics (in-
cluding J/ψ final states of B decays, direct J/ψ production and rare
1.7 Muon detector 25

decays), this trigger performs an important service role by selecting

a large enough sample of b events on which the more aggressive and
technically challenging vertex trigger can be debugged and evaluated

µ3
µ2
A µ1
D

B C

Figure 1.12: Perspective view of the muon detector. (A): smooth faceplate
located on each side of each toroid. (B): coils which wrap around both toroids.
(C): main steel for the first toroid. (D): shielding wall located in front of
the last muon station. The locations of the three muon detector stations are
labeled µ1 , µ2 and µ3 .

The muon detector design combines a toroidal magnet with fine-grained

tracking elements. This design permits a “stand-alone” trigger: i.e. a
di-muon trigger based solely on information from the muon detector. In
addition, improved background rejection is possible by comparing this mea-
surement with momentum and tracking information from the rest of the
spectrometer.
Two toroids, approximately 1 m long with 1.5 T field, provide the bend-
ing power and filtering of non-muons. There are three stations of detectors,
one between the two toroids and two behind the toroids (farther from the
interaction point in z), as shown in Fig. 1.12. The momentum of tracks
can be measured using the two, well shielded, downstream stations and the
nominal beam constraint. The station between the two toroids provides an
important confirming hit for the rejection of fake tracks.
26 The BTeV experiment

Figure 1.13: On the left: beams-eye view of one muon station (eight overlap-
ping octants arranged in two layers). On the right: arrangement of planks
to form each of the four views in an octant (r view is repeated).

The basic element in the construction of a detector station is a “plank”

of stainless steel proportional tubes. There are 32 tubes in each plank, ar-
ranged in two rows of 16, offset by half a tube diameter. In order to avoid
ghost tracks in the system, the minimum requirement is that all hits from
one beam crossing be collected before the next beam crossing. The same gas
as used for the straw detectors, a mixture of Argon-CO 2 , meets this goal. To
minimize occupancy at small radii, twelve planks of increasing length are ar-
ranged into pie shaped octants. To minimize pattern recognition confusion,
three arrangement of planks (r, u, or v) are used. The r views are radial.
The u and v views are rotated ±22.5o with respect to the radial views and
measure the azimuthal angle, φ. A schematic of this arrangement is shown
in Fig. 1.13. In order to provide redundancy in the most important view in
terms of pattern recognition for the trigger and momentum measurement,
the r view is repeated. Other important properties of the muon detector are
reported in Tab. 1.8.

In the next sections I will present the extraordinary sensitivities and

flavour tagging performance of BTeV. I will first start by showing the back-
ground rejection capabilities and efficiencies of the Level 1 trigger algorithm.
1.8 The Level 1 trigger algorithm 27

Properties Values
Total channels 36864
Spatial resolution 1.4 mm
p
Momentum resolution better than σp /p = 0.252 + (0.01p)2
Min. bias rejection 500
Di-muon efficiency 80%
Trigger signals are used in Level 1

Table 1.8: Muon detector properties.

1.8 The Level 1 trigger algorithm

The Level 1 trigger algorithm has two major stages:

1. segment finding

2. track and vertex finding

1.8.1 Segment finding

In order to speed the track finding, pixel hits from three neighbouring sta-
tions are used to find the beginning and ending segments of tracks. These
three station segments are called “triplets.” An “inner triplet” is associated
with a track as it enters the pixel detector from the interaction region and
represents the start of the track. Since nearly all tracks traversing the pixel
detector and entering the forward spectrometer have a hit in the first cen-
timeter of the pixel detector, only that limited region is used to “seed” or
initiate searches for triplets. This greatly reduces the number of calcula-
tions that have to be performed. Similarly, an “outer triplet” is associated
with a track as it leaves the pixel detector, either through the side or the
front or rear faces. An “outer triplet” represents the end of the track in
the pixel detector. Again, nearly all outer triplets start very close to the
detector boundary so only a limited region is used to seed the search for
“outer triplets.” The segment finding algorithm works as follows:

1. starting with a seed hit in the “inner region” of plane N − 1, one

28 The BTeV experiment

projects a cone onto plane N that corresponds to a range of legiti-

mate and interesting tracks that would fall within the pixel detector
acceptance

2. for each hit, “IN ,” within this range, one projects from this hit and
the seed back to the Z position of plane N − 2. If the projection falls
within pixel plane N − 2, then the seed is not the first point on an
inner segment with hit IN . One advances to the next hit in plane N

3. if the projection falls inside the beam hole in the pixels at station N −2
instead, then one projects the seed and hit I N into pixel plane N + 1;
if a confirming hit “J” is found, this seed, I N , and JN +1 are an “inner
segment”

The opportunities for parallelism are evident. Outer segment finding is done
in the same way and in parallel. In the bend view, both inner and outer
segments are found (research for strait tracks over three stations in the bend
view guarantees a momentum cut of p > 3 GeV/c). These are eventually
matched and the difference in directions between an inner segment and its
outer matching segment gives a measurement of the momentum. In the
non-bend view, segment finding is done in parallel with the bend view, but
only inner segments are searched for since they provide enough information
to measure the track horizontal position and angle to extrapolate it back to
the interaction vertex.

1.8.2 Track and vertex finding

The next stage involves delivering all the segments associated with a sin-
gle beam crossing to one CPU in the track/vertex finding processor farm.
The processor then does segment matching to form tracks and applies an
algorithm to find “primary interaction vertices.” Vertex finding constitutes
projecting found tracks back into the interaction region and clustering them.
Since tracks from B decays tend to have somewhat higher transverse mo-
mentum relative to the beam direction than tracks from the main interaction
vertex, a requirement is placed on the tracks used in the clustering that they
be below a certain transverse momentum. Typically several interaction ver-
1.8 The Level 1 trigger algorithm 29

tices are found in each crossing, but they are usually quite well separated
due to the length of the Tevatron luminous region. Each track not falling
into these clusters and whose transverse momentum is above some value
(typically 300 MeV/c) is extrapolated back to the nearest interaction vertex
and its impact parameter b, relative that vertex and its associated uncer-
tainty σb , are calculated. The quantity b/σb is used to evaluate detachment.
The primary Level 1 trigger requires two tracks detached with respect to
the same primary vertex to meet the criteria for a “Level 1 accept.”

Cuts Values
Momentum p > 3 GeV/c (implicit in
the reconstruction)
Transverse momentum pt < 1.2 GeV/c (for track
making the primaries)
> 0.5 GeV/c (for candidate heavy
flavour daughter tracks)
Impact parameter b < 2 mm
b/σb > 2.8
Primary multiplicity > 3 tracks
Detached tracks ≥2

Table 1.9: Level 1 trigger cuts.

1.8.3 Level 1 trigger performance

Figure 1.14 shows an example of simulated Level 1 trigger background re-

jection and efficiency, respectively. To avoid bandwidth saturation less than
1% of crossings should pass the trigger; this is attained with the condition
to have 2 detached tracks with b/σb > 2.8. With these criteria, the trigger
efficiencies are large. For example, the process 2 Bs → Ds+ K − has an 80%
acceptance (efficiency here means the fraction of the off-line reconstructable
events that pass the trigger).
In Tab. 1.9 are reported the Level 1 trigger cuts used to compute the
2
Ds+ → φπ + , K ∗0 K + .
30 The BTeV experiment

Figure 1.14: Level 1 trigger background rejection and efficiency. On the

left: trigger response for minimum bias events for a crossing time of 132 ns
and an average of 2 interactions per bunch crossing. On the right: trigger
efficiency for Bs → Ds+ K − events with a crossing time of 132 ns and an
average of 2 interactions per bunch crossing.

efficiencies reported in Tab. 1.10, for a crossing time of 132 ns and an average
of 2 interactions per bunch crossing.

Processes Efficiencies
Minimum bias 1%
Bs → Ds+ K − 80%
Bo → J/ψKS 65%
B o → φKS 74%
Bo → 2-body modes 80%
(π + π − , K + π − , K + K − )

Table 1.10: Level 1 trigger efficiencies for (first entry) crossings containing
only minimum bias events and (remaining entries) crossings also containing
B decays.
1.9 Sensitivities 31

1.9 Sensitivities

The sensitivities to CP violating angles and related quantities are summa-

rized in Tab. 1.11. In the simulations a luminosity of 2×10 32 cm−2 s−1 and
107 seconds of running time (corresponding to one year) have been consid-
ered. All of the measurements listed here to determine χ, α, β, γ and x s ,
except for the determination of γ using B → Kπ modes, are completely free
of hadronic uncertainties or errors due to theoretical models. Ultimately, at
the level of errors below 1o for α, β and γ (and a much lower number for
χ), the effects of other diagrams or higher order terms in the Wolfenstein
approximation need to be considered.

Reaction B×10−6 # event S/B Parameter Error

Bs → Ds+ K − 300 7500 7 γ − 2χ 8o
Bs → Ds+ π − 3000 59000 3 xs 75
B o → J/ψKS 445 168000 10 sin(2β) 0.017
Bo → J/ψK o
K o → π ± l∓ µ 7 250 2.3 cos(2β) ∼0.5
Bo → π+π− 4.5 14600 3 Asymmetry 0.03
Bs → K + K − 17 18900 6.6 Asymmetry 0.02
B− →
o
D (K + π − )K − 0.17 170 1
o
D (K + K − )K − 1.1 1000 > 10 γ 13o
B − → KS π − 12.1 4600 1 < 4o +
B − → K +π− 18.8 62100 20 γ theory
B− → ρ+ π − 28 5400 4.1
B − → ρo π o 5 780 0.3 α ∼4o
B − → J/ψη 330 2800 15
B − → J/ψη 0 670 9800 30 sin(2χ) 0.024

Table 1.11: Yearly sensitivities to CP violating angles and related quantities.

Reactions between lines are used together.

Some comments on these measurements:

32 The BTeV experiment

• sin(2β) is obtained originally by fitting the time distribution, which

results in a 20% improvement in the error relative to that of the time-
integrated symmetry measurement

• to determine α we use the method originally proposed by Snyder and

Quinn B o → ρπ → π + π − π o [11]; ∼500 effective flavour tagged ρ ± π ∓
events and ∼75 ρo π o per year are expected

• although the B → Kπ modes provide the smallest experimental error

in determining γ, there are model dependent errors associated with
this method. On the other hand, two other methods, which use B s →
o
Ds± K ∓ and B − → D K − , provide model independent results and can
be averaged. The interplay of the three methods can be used to resolve
ambiguities

• the measurement of sin(2χ) can take few years, if it is in the SM range.

Including Bs → J/ψφ can reduce the time

• the asymmetry in B o → π + π − may be useful to gain insight into the

value of α with theoretical input or combined with B s → K + K − and
theory to obtain γ

• the sign of cos(2β) can be determined in few years without any theo-
retical assumptions using B o → J/ψK o , with K o → π ± l∓ ν, allowing
the removal of two of the ambiguities in β. This reconstruction can
also be used for CPT tests

As a result to these studies it’s evident that BTeV would have had outstand-
ing performance in carry on its flavour physics and CP violation program.

1.9.1 Experiment comparison

A comparison of BTeV, LHCb, BABAR and BELLE in 2005, and e + e−

machines with a luminosity of 1035 and 1036 cm−2 s−1 , is given in Tab. 1.12
for several states of importance to the study of CP violation in B decays.
This study indeed demonstrates that it takes a 10 36 cm−2 s−1 e+ e−
collider operating at Υ(4S) to match the performance of BTeV on B o and
1.10 Flavour tagging 33

Parameter BTeV LHCb BABAR e + e− e+ e− e+ e− at

107 s 107 s BELLE 1035 1036 1036 vs
(2005) 107 s 107 s hadron
sin(2β) 0.017 0.02 0.037 0.026 0.008 Equal
sin(2α) 0.05 0.05 0.14 0.1 0.032 Equal
γ[Bs (Ds K)] ∼11.5o Hadron
γ[B(DK)] ∼13.2o ∼20o 12o Equal
sin(2χ) 0.024 0.04 Hadron
B(B → πo πo ) ∼20% 14% 6% e + e−
Vub 2.3% ∼1% ∼1% e + e−
(sys) (sys)

Table 1.12: Comparison of the CP reach of several experiments. The last

column is a prediction of which kind of facility would make the dominant
contribution to each physics measurement [12].

B ± mesons, while for Bs or other b-flavoured hadrons, BTeV would have

given the dominant contribution to the measurements.

1.10 Flavour tagging

Flavour tagging, determination of the flavour of the signal B-hadron at the

time of its production, is an essential component of the study of mixing
and CP violation in B decays. We define the tagging “power” D 2 ( is the
efficiency and D is the dilution) as:

NR + N W NR − N W
= , D= (1.4)
T otal NR + N W

NR and NW are the number of correct and incorrect tags, respectively and
T otal refers to the sample of fully reconstructed neutral B-mesons in the
mode of interest. The determination of the flavour tagging D 2 takes advan-
tage of the precision tracking provided by the pixel detector and excellent
particle identification afforded by the RICH.
There are three quasi-independent flavour tagging methods:
34 The BTeV experiment

Sample Independent tag Overlaps removed

D2 (%) D2 (%)
Away-side kaon tag 5.8 5.8
Away-side muon tag 1.3 1.3
Same-side kaon tag 5.7 4.5
Jet charge tag 4.5 0.4
Sum - 12.1
Electrons + Likelihood fit - 0.9
BTeV expected - 13

Table 1.13: Results on D2 for four algorithms for Bs decays.

same-side tagging: (kaon for Bs and pion for B o ) utilizes the correla-
tion which emerge between a neutral B-meson, and the charge nearby
tracks produced in the fragmentation chain. Since these tracks are
produced in the fragmentation of the b-quark (not the decay), same-
side candidates emerge from the primary interaction vertex. When a
Bs (b̄s) forms, an extra s̄ is available to form a K-meson. About half
of the time it will produce a K + (the other half being a K o , which
is not useful) which is 100% correlated with the flavour of the B s at
production. For B o mesons the same analysis holds, except the parti-
cle tag is a charged pion. This tagging method is not affected by the
mixing of the tagging b

away-side tagging: exploits the fact that, in strong interactions, b-quarks

are produced in pairs and therefore the second (away-side) b-quark in
the event must have opposite flavour. Generally this is done by exam-
ining the charge of kaons or leptons from the away-side b-hadron decay
or the charge of its associated jet. Since these particles come from the
decay of the away-side b-hadron, away-side flavour tag algorithms usu-
ally exclude tracks which point back to the primary vertex. Because we
examine the decay products, this method of flavour tagging is affected
by mixing of the away-side b
1.10 Flavour tagging 35

jet charge tagging: reconstructs the location and decay products of the
b → cW − vertex. The particles associated with the W − decay have a
charge which is 100% correlated with the flavour of the parent b-hadron

Sample Independent tag Overlaps removed

D2 (%) D2 (%)
Away-side kaon tag 6.0 6.0
Away-side muon tag 1.2 0.8
Jet charge tag 4.8 1.4
Same-side pion tag 1.8 1.0
Sum - 9.2
Electrons + Likelihood fit - 0.8
BTeV expected - 10

Table 1.14: Results on D2 for four algorithms for B o decays.

In the simulations a simple approach to combine the tagging algorithms

have been used. The algorithms are ranked in order of decreasing dilution
and the algorithm highest on the list determines the flavour tag for a given
event. If an event is not tagged by the first algorithm, the second is check
and so on. In this way the flavour determination comes from a single tag
algorithm, the one with the highest dilution. The final results for D 2 for Bs
are shown in Tab. 1.13 and for B o in Tab. 1.14. The second column shows
the results when the taggers are treated independently and the third column
shows the results when overlaps are removed. It is expected to achieve a
very remarkable tagging powers, D 2 , of 13% for Bs decays and 10% for B o
decays3 .

3
The eway side electron tags are not explicitly treated and we might expect to get
another 0.7% in D2 .
36 The BTeV experiment
Chapter 2

Theory of radiation damage

in Silicon sensors

In these years many studies have been carried out in order to get a better
understanding of the radiation induced damages in semiconductors [13, 14];
even though the Research and development on Silicon for future Experiments
(ROSE) collaboration has been established, the underlying physics processes
are still not yet entirely understood. In this chapter I will present the most
reliable model for the description of this phenomenon.

2.1 Radiation induced crystal damages

The radiation damage process, affecting sensors exposed to high fluences of

incident particles, proceeds through the following steps:

1. the radiation displaces a Silicon atom from its lattice location (primary
knock-on atom recoil)

2. the primary knock-on atom generates a cluster defect at the end of its
recoil, where the atom loose its last amount of energy and the elastic
scattering cross-section increases by several order of magnitude

Radiation generates basically three different kinds of defects in semiconduc-

tor crystals:
38 Theory of radiation damage in Silicon sensors

point defects: they are isolated defects as shown in Fig. 2.1. In order to
create this kind of defect the energy provided to the recoil atom must
be higher than 25 eV, otherwise it degrades to phonon excitations.
They can be of three types:

vacancy: empty lattice site

interstitial: additional atom between regular lattice sites

Frenkel: complex of an interstitial next to a vacancy

complexes of defects: they are formed by reactions of vacancies and in-

terstitials with Silicon dopant impurities, like Phosphorous for n-type
Silicon, and non-dopant impurities, typically Oxygen and Carbon

clusters of defects: local high concentration of point defects that are ex-
pected to behave somewhat differently from uniformly distributed
point defects. Such defect clusters are capable of producing local
charge concentrations and thus, depending on the type, of attract-
ing or repelling charge carriers from the surrounding region. In order
to create cluster defects the energy provided to the recoil atom must
be between 2 KeV and 12 KeV. Above 12 KeV several clusters will be
produced

Silicon atoms Silicon interstitial

Impurity in interstitial site

Vacancy

Impurity in substitutional site Frenkel defect

Figure 2.1: Point defects in a simple lattice.

2.1 Radiation induced crystal damages 39

Defects are not necessary immobile objects in the crystal. Point defects will
typically anneal either by an interstitial filling a vacancy or by diffusing out
to the surface. Depending on temperature, they may also diffuse and on
their way form defect complexes with other defects or impurities. Similarly,
defect complexes and clusters may break up at elevated temperatures and
diffuse until they form different complexes. They may, however, also interact
with another defect and form a new type of defect complex that becomes
immobile at room temperature.
The probability for creating a primary knock-on atom, as well as its en-
ergy distribution, depends on the type and energy of the radiation. Indeed,
charged particles, such as protons, scatter by electrostatic interaction with
the (partially screened) nucleus, while neutral particles, such as neutrons,
scatter by strong interaction with the nucleus only. The kinematics and, in
particular, the energy transferred to the Silicon atom are strongly dependent
on the mass of the impinging radiation. In order to better appreciate the
radiation damage on Silicon, I report in Fig. 2.2 the results of a simulation
of the defects induced by protons and neutrons at different energies.

36824 vacancies 4145 vacancies 8870 vacancies

1
y (µm)

0.8

0.6

0.4

0.2

0
0 0.5 1 0 0.5 1 0 0.5 1
x (µm) x (µm) x (µm)

Figure 2.2: Distribution of vacancies produced by 10 MeV protons (left),

24 GeV protons (middle) and 1 MeV neutrons (right). The plots are projec-
tions over 1 µm of depth and correspond to a fluence of 10 14 cm−2 (figure
from [15]).

Consequence to radiation induced defects is a change in the macroscopic

properties of the Silicon. Defect complexes may change the doping concen-
tration and, after enough dose, n-type Silicon becomes p-type. The following
40 Theory of radiation damage in Silicon sensors

defect complexes are those considered in literature to be the main cause of

change in Silicon doping concentration:

donor removal complexes: examples of reactions for n-type Silicon, that

lead to these kind of complexes, are V+P s →VP and Ci +Ps →Ci Ps .
Where Ci (Pi ) stands for Carbon (Phosphorus) atom interstitial and
Cs (Ps ) stands for Carbon (Phosphorus) atom substitutional (at a
regular lattice site), whereas V stands for Vacancy. Since these are
stable complexes with new electrical properties, the Phosphorous atom
no longer influences on the effective doping concentration

acceptor-like complexes: an example of a reaction that leads to these

kind of complexes is V+O→VO+V→V2 O, where O stands for Oxygen.
This is a stable complex that behaves as an acceptor state

It is worth noting that to limit the formation of these complexes oxygenated

Silicon can be used. Since V and Ci do also react with Oxygen, in presence
of a high Oxygen concentration, the Phosphorus removal is thus suppressed.
Moreover, since the formation of V2 O is in competition with the reaction
V+O→VO, the acceptor-like complexes creation is also suppressed [16].
Other macroscopic consequences to radiation induced defects are the
increase of leakage current, since they create energy states close to the middle
of the band gap, which are very effective in creation of thermal electron-hole
pairs, and the reduction of the carriers trapping-time. The changes, in the
macroscopic properties of the Silicon, on the sensor detecting performance
will be discussed in the next section.
Although the primary interaction of radiation with Silicon is strongly
dependent on the type and energy of the radiation, this dependence is to a
large extent smoothed out by the secondary interaction of primary knock-on
atoms. It is therefore customary to scale measurements of radiation damage
from one type of radiation and energy to another. As the interaction of ra-
diation with low energy electrons produces ionization but no crystal defects,
the quantity used for scaling is the Non-Ionizing Energy Loss (NIEL) [17].
The NIEL scaling can be reduced to the definition of the 1 MeV equivalent
1M eV ) which produces the same damage as an arbitrary
neutron fluence (Φeq.n.
2.2 Macroscopic effects after irradiation 41

beam with a certain spectral distribution and fluence Φ:

1M eV
Φeq.n. = kΦ (2.1)

k is called the hardness factor.

2.2 Macroscopic effects after irradiation

Four main macroscopic effects are seen in sensors as a consequence of ener-
getic hadron irradiation. The first effect is the change in the doping concen-
tration with severe consequences for the operating voltage needed for total
depletion. In this case we can define an effective doping concentration, N ef f ,
using the following relation:
q0
Vd ≈ |Nef f |d2 (2.2)
20 r
q0 is the electron charge, 0 r is the dielectric constant of Silicon and d is
the bulk thickness. The sign of the effective doping concentration term,
Nef f , depends on donor removal and generation of acceptor-like defects:
Nef f = Nd − Na , with Nd being the donor concentration and N a that of
acceptors. Figure 2.3 shows the dependence of N ef f on the accumulated
1 MeV equivalent neutron fluence for standard and oxygen enriched Float
Zone (FZ) [18] Silicon irradiated with neutrons, protons and pions. The
minima of the curves correspond to the fluence at which the bulk effective
doping concentration becomes nearly intrinsic and, for this reason, it is
called “type inversion.” The second effect is the deterioration of charge
collection efficiency due to charge-carriers trapping, causing a reduction of
particle detection efficiency. These first two effects are the most severe, as
the operation voltage cannot be increased to very high values because of the
break-down, and trapping could drastically limit the collected charge. It is
worth noting that a large operating voltage, together with a large reverse
current, leads also to an increase of dissipated power, which may affect the
cooling system and, if not uniformly distributed, can induce thermal run-
away effects.
The third effect is the fluence-proportional increase in the leakage cur-
rent, I, caused by creation of generation-recombination centers. The in-
42 Theory of radiation damage in Silicon sensors

7
standard FZ
6 400
neutrons
pions
5

Vdep [V] (300µm)

oxygen rich FZ

|Neff| [1012 cm-3]

protons
neutrons 300
4 pions
protons
3 200
2
100
1

0 0.5 1 1.5 2 2.5 3 3.5

Φ eq [1014cm-2]

Figure 2.3: Dependence of the bulk effective doping concentration (N ef f ) on

the accumulated 1 MeV equivalent neutron fluence for standard and oxygen
enriched FZ Silicon irradiated with neutrons, 23 GeV protons and 192 MeV
pions (figure from [14]).

crease of the reverse current due to bulk damage, exhibits a simple depen-
dence on particle fluence as shown in Fig. 2.4. Its measurement, at full-
depletion voltage, is proportional to the active volume, V , of the bulk and
1M eV , to which the sensor was exposed:
to the equivalent fluence, Φeq.n.

1M eV
I = αΦeq.n. V (2.3)

This effect has consequences on the signal-to-noise ratio for particle detec-
tion. However, this can be largely reduced by operating the sensors at a
moderately low temperature, such as ∼ −10 o C.

The last effect is the increase of the positive charge trapped in the oxide
layer on the surface of the sensor. This could increase the electrical field
in proximity of the electrodes and eventually cause voltage break-down.
This effect can be drastically reduced by a proper design of the electrodes
geometry.
2.3 Evolution with time of the effective doping concentration 43

10-1
n-type FZ - 7 to 25 KΩcm
-2 n-type FZ - 7 KΩcm
10 n-type FZ - 4 KΩcm

∆I / V [A/cm3]
n-type FZ - 3 KΩcm
p-type EPI - 2 and 4 KΩcm
10-3

n-type FZ - 780 Ωcm

10-4 n-type FZ - 410 Ωcm
n-type FZ - 130 Ωcm
10-5 n-type FZ - 110 Ωcm
n-type CZ - 140 Ω cm
p-type EPI - 380 Ωcm
10-6 11
10 1012 1013 1014 1015
-2
Φeq [cm ]

Figure 2.4: Fluence dependence of leakage current for different sensors with
different resistivity (figure from [14]).

2.3 Evolution with time of the effective doping

concentration

The effective doping concentration, N ef f , changes with time (t) and temper-
ature (T ), and could be described by an empirical model called “Hamburg
model” [19]. In this model, measured data points (see Fig. 2.5) are described
by two exponential terms and a constant term, which does not depend either
on time nor on temperature. The two exponentials account for the following
two process:

1M eV , T, t): describes the process that leads

beneficial annealing, NA (Φeq.n.
to a reduction of acceptors, Na , and to an increase of donors, Nd .
This process has a typical time constant of ∼1.5 hours at 60 o C (∼5
days at room temperature)

1M eV , T, t): describes the process that leads to

reverse annealing, NY (Φeq.n.
a reduction of donors, Nd and to an increase of acceptors, Na . This
process has a typical time-scale of ∼8 days at 60 o C (∼1.5 years at
room temperature)
44 Theory of radiation damage in Silicon sensors

10

∆ Neff [1011cm-3]
NA = ga Φeq NY,∞ = gY Φeq
6

4
NC
2
gC Φeq
NC0
0
1 10 100 1000 10000
annealing time at 60oC [min]

Figure 2.5: Typical annealing behaviour of the radiation-induced changes in

the effective doping concentration at a temperature of 60 o C after irradiation
up to a fluence of 1.4×1013 cm−2 (figure from [14]).

Figure 2.5 shows the typical time dependence of the effective doping concen-
tration. There is a clear evidence for the three contributions to the evolution
in time. From these considerations it follows that, in general, the sensor
life-time can benefit from low temperature. However, periodic annealing at
higher temperature, short enough to let the beneficial annealing act, can
improve the life-time because of the partial restoration of the crystal lattice
order.
Chapter 3

Silicon sensors radiation

damage characterization

One of main issues in employing Silicon detectors in HEP experiments is

to certify through dedicated measurements their radiation tolerance (see
chapter 2). This is particularly crucial for detectors used at hadron colliders,
where the total fluence after ten years of operation can reach values well in
excess of 1014 1 MeV equivalent neutrons per cm2 . The picture is even
more complicated in presence of highly non-uniform irradiation since the
performance of the sensors can dramatically vary over the active area. In
these cases a local characterization of the sensors is required.
For these reasons I developed a method for their full characterization.
The method allows for a complete bidimensional mapping, point to point, of
the sensor characteristics over its entire active area. Information is gathered
through the Q-V characteristic, measured scanning the sensor with an infra-
red laser source. I performed the measurements on the Silicon micro-strip
sensors that I previously irradiated non-uniformly up to a peak fluence of
∼1014 200 MeV protons per cm2 . I developed an analytical model to fit the
Q-V characteristics and extract the local full-depletion voltage and the local
carriers trapping-time.
As expected, I observed a continuous evolution of the bulk behaviour
moving along a single strip toward the highest irradiated region, a smooth
transition from the original p-type to the inverted type behaviour. From
46 Silicon sensors radiation damage characterization

the measurement of the carriers trapping-time I was able to reconstruct the

absorbed dose profile and compare it with the actual one.

3.1 The Silicon micro-strip sensor and readout

electronics
The Silicon micro-strip sensors, used to carry out the radiation tolerance
studies, are the same as those used for the second layer of the Inner Barrel
(IB2) for the CMS experiment [20] developed by Hamamatsu.
pitch 120 µm

bulk n metallization Al

SiO2
Si3N4

30 µm 40 µm
320 µm

implant p+

backplane n+
backplane Al

Figure 3.1: Sketch of the Silicon micro-strip sensor.

The sensors are made in standard FZ [18] Silicon p/n type, 320 µm
thick, with crystal orientation < 1 0 0 > and resistivity ∼2.5 KΩ cm −1
(i.e. ∼150 V full-depletion voltage). They have 512 strips, 116 mm long and
with a 120 µm pitch. The active area is ∼ 61×116 mm 2 . The total inter-strip
capacitance, as seen by the readout electronics, is ∼1.3 pF cm −1 at depletion
voltage. Each strip has a p+ implant width of 30 µm, AC-coupled to an
Aluminum metallization with an overhang of 5 µm on both implant sides,
see Fig. 3.1. The overhang of the strip metallization is necessary in order
to prevent the break-down due to accumulation of electric field in proximity
of the p+ implant after irradiation. The sensors main characteristics are
summarized in Tab. 3.1 and the mask design, together with details on their
3.1 The Silicon micro-strip sensor and readout electronics 47

dimensions, is presented in Fig. 3.3. A cross-section of the sensor showing its

Sensor characteristics Values

Sensors type p/n
Active area dimensions ∼ 61 × 116 mm2
Strips 512
Pitch 120 µm
Bulk thickness 320 µm
Full depletion voltage ∼150 V
Break-down voltage > 500 V
Coupling to readout electronics AC

Table 3.1: Main characteristics of the Silicon micro-strip sensors.

biasing circuit, the guard-ring and the AC (DC) pad connection is reported
in Fig. 3.2. The high voltage is provided through a bias-ring to which each
strip is connected by a polysilicon resistor of ∼1.5 MΩ, while the ground is
provided through the Aluminum back plane of the sensor.

Figure 3.2: CMS IB2 Silicon micro-strip sensors cross-section in proximity

of the edge of the strip. Shown are the bias circuit, the guard-ring and the
AC (DC) pad connection.

The readout chip employed is the TAA1 produced by Ideas [21]. TAA1
chips are made with 0.8 µm CMOS technology and have a low power con-
sumption of just 0.29 mW per channel. Each chip has 128 channels and
from each channel the analog signal can be readout. The main characteris-
tics of the TAA1 chips are the 3 µs shaping time, that guarantees a very low
48 Silicon sensors radiation damage characterization

readout noise of ∼150 e− per channel, and the possibility to readout either
positive or negative input signals. The VA-DAQ data acquisition system,
comprised of a LabVIEW readout software and a board to interface the PC
to the TAA1 chips, is as well provided by Ideas [22]. One of the main fea-
tures of the VA-DAQ board is to digitize, in two’s complement with a 14 bit
ADC, the analog signal coming from the TAA1 chips; moreover it provides
an internal pulser used to characterize each single channel of the chips.

Figure 3.3: Layout of the CMS IB2 Silicon micro-strip sensors.

3.2 Sensor irradiation and measurements 49

3.2 Sensor irradiation and measurements

Spectrometers like BTeV, instrumented in the very high η region (1.9 <
η < 4.5), present a typical non-uniform irradiation that scales as ∼ 1/r 2 ,
where r is the transverse distance from the beam line. The non-uniform
irradiation causes a non-uniform effective doping profile in Silicon sensors
(see chapter 2), with the consequence of having regions with different full-
depletion voltages. The Silicon sensors performance can thus dramatically
vary over the active area. Figure 3.4 shows a simulation of the radiation
environment, of the BTeV experiment, for the first micro-strip station; the

Figure 3.4: Radiation dose as a function of the position in the first station
of the forward Silicon tracker. The horizontal magnetic field concentrates
more particles above and below the central beam hole than on either side.

radiation dose peak expected is ∼1.6×10 13 particles per cm2 per year, corre-
sponding to ∼0.8×1013 1 MeV equivalent neutrons per cm2 per year. Since
we are dealing with highly energetic particles, the fluence has been rescaled
according to Eq. (2.1) using a hardness factor k = 0.5 [17]. Given the
previous considerations, I decided to irradiate our micro-strip sensors non-
uniformly. The irradiation was performed at Indiana University Cyclotron
Facility (IUCF) [23], with a proton beam of 200 MeV energy. The beam was
approximately centered on the edge of the sensors opposite to the readout
50 Silicon sensors radiation damage characterization

Figure 3.5: IUCF irradiation geome-

try. Shown are the proton beam Gaus-
sian shape and the approximated posi-
tion of the beam peak which was cen-
tered on the edge of the sensor, oppo-
site to the readout electronics.

electronics, as shown in Fig. 3.5. The beam shape was measured with the
wire scanner technique and the data, together with a Gaussian fit, are pre-
sented in Fig. 3.6. I irradiated two sensors: Mod1 up to a peak fluence of

Figure 3.6: IUCF proton beam shape measured with the wire scanner tech-
nique. Gaussian fit σ ' 19 mm.

∼0.8×1014 1 MeV equivalent neutrons per cm2 , corresponding to 10 years

of BTeV operation; and Mod2 up to a peak fluence of ∼3.5×10 13 1 MeV
equivalent neutrons per cm2 , corresponding about to the dose at the type
inversion (see Fig. 2.3). Since low energetic protons, like those of IUCF,
cause the same damage as neutrons of 1 MeV energy, I used a hardness
3.2 Sensor irradiation and measurements 51

factor k = 1 [17] for fluence rescaling.

Figure 3.7: I-V characteristics for Mod1 (on the left) & Mod2 (on the right)
measured during irradiation at room temperature. In the inset, in the upper
right corner, is shown the I-V characteristic before irradiation.

Figure 3.7 shows the I-V characteristics measured at different fluences

during irradiation at room temperature. It’s worth noting that the break-

Figure 3.8: Mod1&Mod2 leakage current measured biasing the sensors at

400 V as a function of the peak fluence. The straight line is the expected
empirical curve taken from literature. The measurements were performed at
room temperature.

down voltage still remains higher than 500 V. After irradiation the sensors
were always kept at a temperature lower than −12 o C without applying any
52 Silicon sensors radiation damage characterization

annealing cycle. Figure 3.8 reports the leakage current, measured biasing
the sensors at 400 V, as a function of the peak fluence. Since the irradiation
was non-uniform, I calculated the total absorbed dose by integrating the
two-dimensional Gaussian shape of the radiation beam, Φ 1M eV
eq.n. (x, y), over
the sensor surface. Thus, Eq. (2.3) should be modified as follow:
Z Z
1M eV 1M eV
I = α Φeq.n. (x, y)dV = αd Φeq.n. (x, y)dxdy (3.1)

d is the sensor bulk thickness and α = 4.56×10 −7 A cm−1 was taken from
literature [24]. The expected behaviour of the leakage current with the
fluence is confirmed by my measurements.

Figure 3.9: Histogram of the noise for Mod1 (on the left) & Mod2 (on the
right), measured at different conditions: (1) pure readout electronics noise,
sensor disconnected; (2) measured noise on strips before the irradiation; and
(3) after the irradiation.

Figure 3.9 shows the noise value measured on several channels of Mod1
and Mod2 for three different conditions:

1. pure readout electronics noise, sensor disconnected

2. noise measured on strips before irradiation

3. noise measured on strips after irradiation

I observed an appreciable increase of the noise due to the irradiation, which,

for the highest irradiated sensor, goes up from ∼985 e − to ∼1200 e− .
3.3 Laser setup and measurements 53

3.3 Laser setup and measurements

The characterization method that I developed is based on local measure-

ment of the Q-V characteristic. I used an infra-red laser source [25], whose
main characteristics are reported in Tab. 3.2, to locally generate electron-
hole pairs uniformly all along the bulk thickness of the sensor.

Laser characteristics Values

Wavelength 1064 nm
Spectral band width 0.1 nm
Pulse duration 1 ns
Beam power 5 mW
(at the tip of the optical fiber)
Absorption length in Silicon ∼1.1 mm
at 25o C [26]

Table 3.2: Laser main characteristics.

In order to perform a bidimensional scan of the sensor with the laser

source, I placed the detector on an X − Y moving table, whose step-motors
were controlled by a personal computer that guaranteed very precise and
reproducible movements. The laser pulses were carried through a monomode
fiber to the final focusing lens, which was mounted directly on the Z-axis
of the table. The sensor was kept at −12 o C by an underlying Aluminum-
Copper plate cooled by a Nestlab CB-80 water-glycol refrigeration system.
The atmosphere around the detector was filled with Nitrogen to prevent
condensation. The whole setup apparatus is sketched in Fig. 3.10, while
Fig. 3.11 shows a picture of the actual setup, where the Nitrogen bag and
the Aluminum-Copper thermal exchanger are visible. A pulse generator
provided the very fast signal to the laser diode current-driver; the same
generator provided also a synchronous NIM signal, which was used to trigger
the Ideas readout electronics.
The stability of the laser wavelength was guaranteed by a temperature
control system operating with Peltier cells. The system constantly moni-
tored the temperature of the laser diode with a thermistor and consequently
54 Silicon sensors radiation damage characterization

ALIGNMENT MICROSCOPE
PC VIDEO
NITROGEN−ATMOSPHERE TEMPERATURE
ENCLOSURE CONTROLLER

BLACK BOX BEAM

SPLITTER
PULSE
LASER
GENERATOR





































 OPTICAL TIMING

ELECTRONICS 












AMPLIFIER UNIT

IDEAS 

























MICRO−STRIP

SENSOR
















































CHIPS
IDEAS
DAQ PC



























OSCILLOSCOPE
TRIGGER
SIGNAL Y
Pump
X MOVING TABLE
M

REFRIGERATOR

THERMAL EXCHANGER RTD THERMOCOUPLE

Figure 3.10: Setup apparatus for local measurement of Q-V characteristics

on irradiated Silicon micro-strip sensors. It consists of an X − Y moving
table controlled by a personal computer. The sensor is kept at −12 o C by
an underlying Aluminum-Copper plate cooled by a water-glycol refrigeration
system. The atmosphere around the detector is filled with Nitrogen to avoid
condensation. The moving table is kept in a dark box.

regulated the current in the Peltier cells that cooled the laser diode. The
laser pulse amplitude was monitored through an oscilloscope; indeed an op-
tical beam splitter sent half of the beam to the laser optics and half of the
beam to an Opto Electronics [27] photo detector connected to the oscillo-
scope.
It is worth noting that the temperature stability is crucial for these mea-
surements; precisely, the laser absorption in Silicon, in the infra-red region
of the spectrum, critically depends not only on the laser diode temperature
but also on the sensor temperature. Figure 3.12 reports two sets of mea-
surements of the collected charge performed on a non-irradiated sensor at
different temperatures, 27o C and −13o C. From these measurements the
dependence of the collected charge on the temperature can be estimated; in
the simplified hypothesis of linear dependence, it turns out to be ∼17500 e −
per degree Celsius, which means a variation of the collected charge of ∼2.2%
3.3 Laser setup and measurements 55

Figure 3.11: Sensor cooling system. A water-glycol mixture was flowed

through ducts placed inside the Aluminum plate, over which the sensor was
placed.

Figure 3.12: Temperature dependence of the collected charge for a non-

irradiated sensor. The collected charge was measured at positions transversal
to strips at two different temperatures, 27 o C and −13o C. The measure-
ments were performed biasing the sensor at 160 V.

per degree Celsius at −13o C. Therefore I constantly monitored the temper-

ature with a thermocouple placed on the Aluminum-Copper cooling plate
and a thermistor placed close to the sensor. I regulated the Nitrogen flux
and the refrigerator pump in order to keep the temperature stable within
1o C.
I illuminated the sensor on the strip side. In this situation, only the
56 Silicon sensors radiation damage characterization

fraction of the laser light impinging on the sensor between adjacent strip
metallizations could reach the bulk of the sensor. To reduce to a negligible
level the dependence of the quantity of absorbed light on the position of
its source, I illuminated a wide region, ∼20 strips, by placing the sensor
out of the focal plane of the lens, as shown in Fig. 3.13. Given the ratio of
metallization to pitch, 40 µm/120 µm, only 2/3 of the laser light penetrates
the bulk.

IR−laser

Silicon micro−strip sensor

Figure 3.13: Sketch of the laser spot and the relative position of the sensor
with respect to the laser focusing plane.

For each position of the laser beam on the sensor and for different bias
voltages, I sent ∼1000 laser pulses and I measured the collected charge. The
Gain [mV/fC]

Pedestal [mV]

80 150

70

60 100

50

40 50

30

20
0
10

0
-50
0 100 200 300 400 500 0 100 200 300 400 500
# Strip # Strip

Figure 3.14: Gain and pedestal calibrations.

final value of the total collected charge was obtained by the following data
handling procedure:

1. I previously calibrated the system in the same experimental condi-

3.3 Laser setup and measurements 57

tions, of temperature and bias voltage, as the laser measurements. I

subtracted the pedestal and converted the scale from millivolt to elec-
trons for all the ∼1000 measurements, channel to channel. A typical
calibration, measured at −12o C and 350 V, is shown in Fig. 3.14

2. calculated and subtracted the common-mode for each of the ∼1000

measurements

3. summed all the ∼1000 measurements channel to channel, and divided

the result by the number of pulses; obtaining one single pulse as shown
in Fig. 3.15

4. fitted with a Gaussian the laser peak. This was done in order to smooth
the charge collection difference between strips due to dead channels

5. finally the total collected charge was calculated integrating the Gaus-
sian fit

χ 2 / ndf 3206 / 16
9.896e+04 ± 127
×103
Gaussian Const
Gaussian Mean 207.6 ± 0.0
Intensity [ e - ]
Intensity [mV]

Gaussian Sigma 4.133 ± 0.004

100
1000

80
800

600 60

400 40

200 20

0
0
200 250 200 250
# Strip # Strip

Figure 3.15: Collected charge per strip after averaging ∼1000 pulses. On
the left: row data. On the right: data after equalization and common-mode
subtraction with the Gaussian fit superimposed. This measurement was taken
biasing the sensor at 350 V.

Figure 3.15 shows a typical laser pulse averaged over ∼1000 pulses, before
and after equalization and common-mode subtraction.
An Am241 γ-source was used to provide an absolute calibration of the
apparatus. Am241 emits photons of 59.6 KeV, corresponding to ∼2.6 fC in
58 Silicon sensors radiation damage characterization

χ2 / ndf 63.77 / 63
31.72 ± 1.20

Entries
80 Const-S
Mean-S 367.1 ± 3.3
70 Sigma-S 45.86 ± 2.21

60

50

40

30

20

10

0
-200 -100 0 100 200 300 400 500 600
ADC

Figure 3.16: Typical signal from an Am 241 γ-source. At higher ADC values
the 59.6 KeV signal peak is visible above the noise. Both peaks are fit with
Gaussians. The spectrum was shifted to the left positioning the pedestal to
zero.

Silicon. A typical spectrum of the γ-source seen by a single strip is shown in

Fig. 3.16. Here the spectrum has been shifted so the pedestal peaks at zero.
The peak corresponding to 59.6 KeV is clearly visible at higher ADC values;
while at lower ADC values is visible the pedestal peak. The position of the
signal, as well as that of the pedestal, was measured by Gaussian fit to the
59.6 KeV peak and pedestal peak, in absence of signal, respectively. The
noise spectrum is reported in Fig. 3.17. From the signal fit I obtained a value
of 367.1±3.3 ADC counts; that corresponds to ∼134.4 mV. Therefore the
gain was ∼52 mV fC−1 , whereas the internal calibration gave ∼57 mV fC −1 .

χ2 / ndf 124.6 / 93
70
Entries

Constant 52.15 ± 1.61

Mean 32.77 ± 0.97
60 Sigma 38.82 ± 0.76

50

40

30

20

10

0
-200 -100 0 100 200 300
ADC

Figure 3.17: Typical pedestal peak spectrum with superimposed the Gaussian
fit for the determination of its position.
3.3 Laser setup and measurements 59

×10
3
×10
3

Q[e ]
Q[e ]

-
-
1000
1000

800
800

600
600

400 Legend 400 Legend

Data Data
Fit Fit
200 200
Holes Electrons
Electrons Holes
00 00 100 200 300 400 500
100 200 300 400 500
Bias [ V ] Bias [ V ]

Figure 3.18: Mod1 Q-V characteristics superimposed with fits to the model.
On the left: data from the opposite side of the radiation beam peak; ρ 0 =
3302 ± 12 e− µm−1 , τ = 1000 ns (fixed parameter), Vd0 = 101.0 ± 1.4 V,
Vof f = −0.1 ± 0.1 V. On the right: data from the radiation beam peak
side; ρ0 = 3341 ± 23 e− µm−1 , τ = 29.3 ± 1.4 ns, Vd0 = 25.3 ± 1.5 V,
Vof f = 6.0 ± 0.2 V.

Figure 3.18 presents two typical sets of measurements superimposed with

fits to the model that will be described in the next section. In blue and red
are also reported the two contributions, due to electrons and holes respec-
tively, to the total induced signal. These measurements were performed
in two particular positions of Mod1 sensor, the first plot refers to measure-
ments taken on the non-irradiated region, the second refers to measurements
taken on the highest irradiated region. It is worth noting the change of the
shape between the two sets of measurements and also the change of the rel-
ative contribution of electrons and holes to the total induced signal. In the
next section I will describe the model that I developed to extract physical
information from the shape of the Q-V characteristic.

An important conclusion can already be made; these sensors, non uni-

formly irradiated up to a fluence corresponding to ∼10 years of BTeV op-
eration, are able to collect more than the 95% of the total charge if biased
at 350 V at least. Which makes them well suited to be used in the BTeV
experiment.
60 Silicon sensors radiation damage characterization

3.4 Collected charge model

I approximated the sensor as a simple p/n diode with infinite plane elec-
trodes. To the extent of describing the total collected charge by all the
strips, the validity of this approximation is only limited by the different
electric field configuration near the p-implant. In this region the transport
of the carriers is different and, thus, could in principle affect the evaluation
of the carriers trapping-time. On the other hand, the extension of this region
is negligibly small, and therefore no appreciable biases will be introduced by
my model [28]. Interestingly enough, for long carriers trapping-time as for
non-irradiated sensors, my model reproduces the data with good accuracy,
as shown in Fig. 3.18. This means that the progression of the depleted re-
gion with the bias voltage is practically identical even in proximity of the
p-implant1 . The main advantage of the present approach is that one can
apply the simplest version of the Ramo theorem and obtain an analytical
expression for the total collected charge, which is extremely advantageous
for these kind of applications. My main goal is to characterize the sensor
performances for their application in HEP experiments and certainly not
to perform a very precise measurement of the physical parameters involved.
All in all, this model provides an approximate description of the local Q-V
characteristics, i.e. the total collected charge as a function of the applied
bias voltage in a particular position of a sensor. It is employed to get an
estimate of the sensor main physical parameters and, hence, to analyze the
damage induced by irradiation.

I will begin by describing the transport of the carriers during their mi-
gration toward the electrodes and then I will introduce the Ramo theorem
to calculate the total induced signal.

1
Obviously to claim this I have to assume an uniform generation of charge in the bulk
of the sensor. On the other hand, this seems to be the most reasonable assumption,
since, otherwise, I should advocate a singular conspiration of effects. “Pluralitas non est
ponenda sine necessitate,” W. Ockham (1280−1347).
3.4 Collected charge model 61

3.4.1 Transport

In the case of a p/n diode with infinite plane electrodes the problem one has
to deal with becomes monodimensional, the only coordinate surviving being
x along the electric field lines. Any charge density ρ(x), either electrons or
holes, generated at x in the depleted region, drifts toward the proper elec-
trode along the electric field lines and evolves with the following equation:
dx
ρ(x + dx) − ρ(x) = − ρ(x) (3.2)
λ(x)

λ(x) is the mean free path for trapping and is given by:

λ(x) = τ v(E(x))
1
(3.3)
with: τ = σvth Nt

E is the electric field, v the drift velocity, τ the carrier trapping-time, σ

the trapping cross-section, vth the thermal velocity and Nt the effective trap
concentration. σ, vth , Nt and τ do not vary with the electric field. The
drift velocity, on the other hand, depends on the electric field through the
relation [29]:
µ0 E
v(E) = (3.4)
1 + (µ0 /vs )E
where µ0 is the mobility and vs is the saturation velocity (88 µm ns −1
for holes and 116 µm ns−1 for electrons). The mobilities I used were
rescaled from the quoted values, at 300 o K (507 cm2 V−1 s−1 for holes
and 1590 cm2 V−1 s−1 for electrons), according to the well known formulae:
µh ∝ T −2.2 and µe ∝ T −2.4 [30]. There is no evidence that the mobilities
and saturation velocities change with irradiation [31].
The solution of the Poisson equation for a constant charge density is a
linear electric field:
(
Ax if Vb ≤ Vd
E(x) = (3.5)
Ax + B if Vb > Vd

It is possible to express the electric field in terms of the bias voltage V b , the
full-depletion voltage Vd and the bulk width w (320 µm). In this case A and
2Vd Vb −Vd
B can be written as A = w2
and B = w . This solution is rigorously
correct for a constant charge density, but, as stated in section 3.4, there
62 Silicon sensors radiation damage characterization

might be slight changes in case of irradiation. I will discuss in section 3.5

the effect of this approximation.
By integrating Eq. (3.2) from x1 to x2 one can obtain the residual charge
density at x2 , for a charge density which was originally generated at x 1 and
has migrated subjected to trapping up to x 2 .
Rx
− 2 dx
ρ(x1 , x2 ) = ρ(x1 )e x1 λ(x)
(3.6)

Equation (3.6) can be explicitly calculated by using Eq. (3.3) and Eq. (3.4):
  −1
E(x2 ) τ µ0 A x2 −x1
ρ(x1 , x2 ) = ρ(x1 ) e− τ vs (3.7)
E(x1 )

In order to obtain a final analytical expression of the collected charge, I

approximated at first order the last term of Eq. (3.7):
  −1  
E(x2 ) τ µ0 A x2 − x 1
ρ(x1 , x2 ) = ρ(x1 ) 1− (3.8)
E(x1 ) τ vs

This approximation is justified as long as (x 2 −x1 )  (τ vs ), which is satisfied

for τ  4 ns, since the highest value that x 2 − x1 can assume is the bulk
thickness, 320 µm, and the lowest saturation velocity is 88 µm ns −1 for
holes.

3.4.2 Induction

Now I apply the Ramo theorem, i.e. a charge Q moving in a potential

difference dV , induces on the electrodes a charge dQ I given by:

dQI Vb = QdV (3.9)

Vb is the electric potential applied between the electrodes. Putting Eq. (3.9)
in an integral form and using the transport equation, Eq. (3.8), one can
obtain:
X X
1
Z Z
QI = ρ(x, x0 )E(x)dxdx0 (3.10)
Vb 0 x0
q
Vb
X =w Vd is the depth of the depleted region. Finally, integrating explicitly
Eq. (3.10), the following expressions for the signal induced by holes and
3.4 Collected charge model 63

electrons can be obtained:


ρ0 2 − Rh
QIh = E(X)3 +
Vb A2 (2 − Rh ) 3(1 + Rh )
E(x)3
 
E(X)
− −X +
(1 + Rh )τ vs A(2 + Rh )
E(X)2−Rh B 1+Rh B3
  
B
−1 + +
1 + Rh A(2 + Rh )τ vs 3
E(X)3 E(X)
  
ρ0
−X +
Vb A2 (2 − Rh )τ vs 1 + Rh 4A
E(X)2−Rh B 1+Rh B4
  
E(X)
− −X + (3.11)
1 + Rh A(3 − Rh ) 4A(3 − Rh )

E(X)3

ρ0
QIe = +
Vb A2 (2 + Re ) 3
E(X)1−Re B 2+Re
  
1 E(X)
− 1+ −X +
1 − Re τ vs A(2 − Re )
B4

3 2 + Re
B + +
3(1 − Re ) (1 − Re )(2 − Re )τ vs
E(X)4 E(X)1−Re B 3+Re

ρ0
− + +
Vb A2 (2 + Re )τ vs 4A(3 + Re ) A(1 − Re )(3 + Re )
B4

− (3.12)
4A(1 − Re )
−1 −1
Rh is defined as Rh = τ h µh A for holes and Re = τ e µe A for electrons. It is
worth noting that in doing these calculations I had tacitly assumed a con-
stant charge, electron-hole pairs, density, ρ 0 , generated by the laser pulse
through the whole depleted region. My data seem to confirm this assump-
tion (see the comments at the end of section 3.6.4 and at the beginning of
section 3.4).
So far this model has four free parameters:

• ρ0 : the initial charge density in the bulk, generated by the laser pulse.
It can be expressed either in [e− µm−1 ] or [h µm−1 ]

• τe : the electrons trapping-time [ns]

• τh : the holes trapping-time [ns]

• Vd : the full-depletion voltage [V]

64 Silicon sensors radiation damage characterization

The number of free parameters can be reduced from four to three using the
empirical relation [32]:
1 1M eV
= γe,h Φeq.n. (3.13)
τe,h
which relates the trapping-times of electrons and holes to the radiation flu-
ence. The coefficients γe,h are known from literature and are in units of
ns−1 × 1014 1 MeV eq.n.×cm2 , and do not depend on the sensor type. For
instance, at −10o C, γe = 0.042 ± 0.003 for electrons and γh = 0.061 ± 0.003
for holes [33]. Thus:
γe
τh = τe (3.14)
γh
This model can be easily extended to accommodate effects due to non-
linearities of the electric field which are typically observed on irradiated
sensors [34, 35, 36, 37]. To this extent I allow for an additional fit parameter,
metallization Al
SiO2
+++++++++++++++++++++++++++++++++++++
n Si3N4

implant p+
Depleted






















































































region

















































































backplane n+

































Neutral 
region
backplane Al

Figure 3.19: Positive charge trapped in the oxide layer. It changes the elec-
tric field in proximity of the electrode.

Vof f . This defines a kind of an effective bias voltage, V b0 = Vb − Vof f , which

should replace Vb in my previous equations. In the case of an excess of
charge close to the p-implant, which is typically caused by positive charge
trapped in the oxide layer (see Fig. 3.19), V of f acts as an initial offset voltage,
which serves to saturate the charge excess; for V b values higher than Vof f ,
the sensor quickly reaches the canonical asymptotic behaviour in V b0 . This
requires that the first measurement points, up to voltages of the order of
Vof f , should be eliminated from the fit. In an analogous way, V of f is even
capable to allow for huge deviations from linearity of the electric field, as
those encountered at very high absorbed doses, which could be explained by
trapping of generation currents in acceptor and donor traps [38] (cfr. the
next section).
3.5 Model accuracy 65

At the end, this model has four free parameters, namely: ρ 0 , τe , Vd0 and
Voff .

3.5 Model accuracy

I will discuss the accuracy of the model in the context of the main approxi-
mations that have been introduced:

• linear electric field inside the bulk

• first order approximation of a term in the transport equation, Eq. (3.7)

I studied in details the biases, that these assumptions might introduce in

the extraction of the main physical parameters, by fitting with my model
different sets of Q-V characteristics, which were accurately simulated.
Electric field [ V µ m-1 ]

Electric field [ V µm-1 ]

Electric field [ V µm-1 ]

0.5
0.4
0.8
0.35
0.4
0.3 0.7

0.25 0.3
0.6
0.2
0.2
0.15 0.5
0.1
0.1
0.05 0.4

00 50 100 150 200 250 300 00 50 100 150 200 250 300 0 50 100 150 200 250 300
Bulk [ µ m ] Bulk [ µm ] Bulk [ µm ]

Figure 3.20: Two-slope linear electric field. First figure: under depletion
(20 V). Second figure: at depletion (40 V). Third figure: over depletion
(150 V). The excess of positive charge at the electrode side is ten times
greater than in the bulk and its extension is 60 µm.

To study the effect of the positive charge trapped in the oxide layer, I
approximated the excess of charge near the electrode with a step function
having a depth of 60 µm and magnitude ten times greater than the fixed
charge in the bulk. The electric field generated by this charge configuration
has two slopes, as shown in Fig. 3.20. In this case, the fit to the simulated
Q-V characteristic is shown in Fig. 3.21. The parameters used for the sim-
ulation were those typically measured on my two sensors. The deviation of
the fit parameters from the actual values are reported in Tab. 3.3. The only
sizable deviation, 4%, is on the full-depletion voltage, which results from the
sum of Vd0 = 31.9 V and Vof f = 9.7 V.
66 Silicon sensors radiation damage characterization

×10
3

Q[e ]
-
1000

800

600

400

Legend
200 Fit
Two-slope linear field

00 100 200 300 400 500

Bias [ V ]

Figure 3.21: Fit to the two-slope linear electric field simulation. The pa-
rameter values are: ρ0 = 3300 e− µm−1 , τe = 60 ns, Vd = 40 V and the
extension of the excess of positive charge at the electrode side is 60 µm.

Fit parameters Actual values Fit deviations

Initial charge density 3300 e− µm−1 < 0.1%
Electron trapping-time 60 ns 1%
Full-depletion voltage 40 V 4%

Table 3.3: Deviation of the fit parameters from two-slope linear electric field
simulation.

Some recent measurements [35, 36, 37] have shown that, at high flu-
ences, of the order of (5−10)×1014 1 MeV equivalent neutrons per cm2 ,
the electric field inside the bulk becomes highly non-linear and assumes a
quasi-parabolic shape. The charge density configuration, that generates this
field, seems to be due to the two generation currents, of holes and electrons,
trapped by acceptor and donor traps respectively. I approximated this be-
haviour with a linear charge density that generates a parabolic electric field,
as shown in Fig. 3.22. I tested my model at the limit of its validity, which
corresponds to a trapping-time of 5 ns (or equivalently to a radiation fluence
of ∼5×1014 1 MeV equivalent neutrons per cm2 ). Fit to the Q-V character-
istic is shown in Fig. 3.23. From the last five points of the simulation the
model was able to recover the actual values of the parameters with small
deviations, as reported in Tab. 3.4.
3.5 Model accuracy 67

2.4

Electric field [ V µ m-1 ]

Electric field [ V µ m-1 ]

Electric field [ V µm-1 ]

1.6
2.2 2.5
1.4
2
1.2 1.8 2
1.6
1
1.4
1.5
0.8 1.2
1
0.6 1
0.8
0.4 0.6
0.4 0.5
0.2
0.2
00 50 100 150 200 250 300 00 50 100 150 200 250 300 00 50 100 150 200 250 300
Bulk [ µ m ] Bulk [ µm ] Bulk [ µ m ]

Figure 3.22: Parabolic electric field. First figure: under depletion (100 V).
Second figure: at depletion (250 V). Third figure: over depletion (350 V).

×10
3
Q[e ]
-

700

600

500

400

300

200
Legend
100 Fit
Quadratic field
00 100 200 300 400 500
Bias [ V ]

Figure 3.23: Fit to the parabolic electric field simulation. The parameter
values are: ρ0 = 3300 e− µm−1 , τe = 5 ns and Vd = 250 V.

I also estimated the effects introduced by the fact that the γ e,h coeffi-
cients are not well known. I run the simulation with the coefficient ratio of
Eq. (3.14) displaced from its central value by ±σ, where σ is the error on
the γe /γh ratio. In both cases, of two-slope linear electric field and parabolic
electric field, the deviation of τe from its actual value is < 6%, which is of
the same order of the γe error. The effect on the other fit parameters is
negligible but for the full-depletion voltage in the case of parabolic field, for
which the deviation is ∼6%.

From these studies and many other, performed for different field config-
urations and absorbed doses, I can conservatively estimate the biases on the
fit parameters as quoted in Tab. 3.5. I don’t report any error on the biases
since they have been obtained by fitting the simulation that has infinite
statistic.
68 Silicon sensors radiation damage characterization

Fit parameters Actual values Fit deviations

Initial charge density 3300 e− µm−1 2%
Electron trapping-time 5 ns 4%
Full-depletion voltage 250 V 4%

Table 3.4: Deviation of the fit parameters from parabolic electric field simu-
lation.

Fit parameters Biases

Initial charge density 2%
Electron trapping-time 6%
Full-depletion voltage 6%

Table 3.5: Biases on returned fit parameters.

3.6 Results and discussion

With the model just described I characterized the highest irradiated sensor,
Mod1. Before showing the results of the fit I will briefly discuss the statistical
and systematic errors I applied to the measured points.

3.6.1 Errors treatment

It is easy to demonstrate that in my case the statistical measurement error is

negligible with respect to the systematic error due to the relative instability
of the employed setup. Indeed, the measured values of the collected charge
at different voltages were obtained by averaging over ∼1000 measurements
of the same quantity. The statistical error of a single measurement has three
components:

• fluctuations of the generated electron-hole pairs, assumed poissonian,

√
ne,h

• electronic noise of the readout channels, σ noise , and

• measured fluctuation of common-mode, σ CM

3.6 Results and discussion 69

Taking into account the fact that each measurement comes from the sum of
the charge collected by 80 strips, N strips , the total statistical error becomes:
2
σstat 2
= ne,h + Nstrips (σnoise 2 ). The second term of the previous ex-
+ σCM
pression, which constitutes the electronic contribution to the error, can be
computed assuming conservatively that σ CM ∼ σnoise = 1200 e− and thus
√
obtaining 2 · 80 · 12002 ' 15175 e− . The total error will then be practi-
cally independent on the collected charge since it varies from ∼15180 e − , for
the smallest value of collected charge in my measurement (∼150000 e − ), to
∼15210 e− for the highest value (∼106 e− ). This means that the statistical
error obtained averaging 1000 measurements is ∼480 e − , which corresponds
to ∼0.32% of the signal in the worst case.

This figure of statistical error is negligible compared to the systematic

error introduced by my setup, which can be a priori estimated around 1%
taking into account the laser system stability and temperature fluctuations.
Entries

2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Chi-square/NDF

Figure 3.24: NDF-normalized χ2 distribution for the 12 fits to the measured

Q-V characteristics well away from the irradiated region.

Actually I cross-checked this guess by estimating the fractional system-

atic error required to obtain an average χ 2 value equal to the number of
degrees of freedom for the fits to the data taken in the non-irradiated region
of the sensor, where my model is expected to closely reproduce the measure-
ments. The required systematic error turned out to be ∼1.2%. This is the
value I used for my measurements. The NDF-normalized χ 2 distribution for
the 12 fits to the measured Q-V characteristics well away from the irradiated
region (the first 12 points on the left of Fig. 3.25) is reported in Fig. 3.24.
70 Silicon sensors radiation damage characterization

3.6.2 Fit results

As one can see from the overall measurement points sketched in Fig. 3.25,
I performed a detailed map of the Mod1 sensor active area. The reference
point for the coordinate along the strips is the edge of the sensor on the
chip-side, in the non-irradiated region, while the strip number corresponds
to the strip on which the laser spot was centered. The results of the fits

Silicon sensor
strip 470

Most Irradiated Zone

strip 344
Readout electronics

strip 311
strip 281

strip 208
strip 174
strip 145

strip 110

strip 77

10 mm
40 mm
60 mm
80 mm
90 mm
100 mm
110 mm

Figure 3.25: Sketch of the overall measurement points on Mod1.

to the measured Q-V characteristics are reported in Figs. 3.26, 3.27, 3.28,
3.29, 3.30, 3.31 and 3.32. The electrons and holes contributions to the total
induced signal are also reported in blue and red respectively.
Since the first twelve points, on the left of Fig. 3.25, belong to a non-
irradiated region, the carriers trapping-time are high enough to allow that
practically all the electron-hole pairs generated in the bulk reach the elec-
trodes. I therefore performed the first twelve fits fixing the electron trapping-
time to a value for which the exponential term of Eq. (3.7) could be consid-
ered equal to 1, thus I set τe = 1000 ns.
I adopted the following criteria or guidelines to fit each Q-V characteristic:

1. I started using all data points in the fit

2. if the NDF-normalized χ2 value of the fit was greater than 2, I dis-

regarded the first measured point and restart the fit. This was done
until I reached a normalized χ2 < 2. In any case this process must
3.6 Results and discussion 71

×10
3
×10 ×10
3 3

Q[e ]

Q[e ]

Q[e ]
-

-
1000
1000 1000

800 800
800

600 600 600

400 Legend 400 Legend 400 Legend

Data Data Data
Fit Fit Fit
200 200 200
Holes Holes Holes
Electrons Electrons Electrons
00 100 200 300 400 500 00 00
100 200 300 400 500 100 200 300 400 500
Bias [ V ] Bias [ V ] Bias [ V ]

×10
3
×10
3
×10
3

Q[e ]
Q[e ]
Q[e ]

-
-
-

1000 1000
1000

800 800
800

600 600 600

400 Legend 400 Legend 400 Legend

Data Data Data
Fit Fit Fit
200 Holes
200 Holes 200 Holes
Electrons Electrons Electrons
00 00 00 100 200 300 400 500
100 200 300 400 500 100 200 300 400 500
Bias [ V ] Bias [ V ] Bias [ V ]

Figure 3.26: Fit to measurement points at 10 mm from chip side. Strips 77,
145, 208 ,281, 344 and 470 from the upper left to the lower right.

×10 ×10
3 3
×10
3
Q[e ]

Q[e ]
Q[e ]
-

-
-

1000 1000 1000

800 800 800

600 600 600

400 Legend 400 Legend 400 Legend

Data Data Data
Fit Fit Fit
200 Holes 200 Holes 200 Holes
Electrons Electrons Electrons
00 100 200 300 400 500 00 00 100 200 300 400 500
100 200 300 400 500
Bias [ V ] Bias [ V ] Bias [ V ]

×10
3
×10
3
×10
3
Q[e ]
Q[e ]

Q[e ]

-
-

1000 1000 1000

800 800 800

600 600 600

400 Legend 400 Legend 400 Legend

Data Data Data
Fit Fit Fit
200 200 Holes
200
Holes Holes
Electrons Electrons Electrons
00 00 00 100 200 300 400 500
100 200 300 400 500 100 200 300 400 500
Bias [ V ] Bias [ V ] Bias [ V ]

Figure 3.27: Fit to measurement points at 40 mm from chip side. Strips 77,
145, 208 ,281, 344 and 470 from the upper left to the lower right.

stop before the knee of the Q-V characteristic, which corresponds to

the full-depletion voltage

3. if the fit wasn’t able to reach adequate convergence, the role of elec-
trons and holes was exchanged in the model and the fit procedure was
restarted from step 1

It could happen, as it did in one case, that convergence can be achieved both
for a choice of electrons and holes as well for the reverse choice. In this case,
72 Silicon sensors radiation damage characterization

×10 ×10 ×10

3 3 3

Q[e ]

Q[e ]

Q[e ]
-

-
1000 1000 1000

800 800 800

600 600 600

400 Legend 400 Legend 400 Legend

Data Data Data
Fit Fit Fit
200 200 200
Holes Holes Holes
Electrons Electrons Electrons
00 100 200 300 400 500 00 100 200 300 400 500 00 100 200 300 400 500
Bias [ V ] Bias [ V ] Bias [ V ]

×10 ×10
3 3
×10
3
Q[e ]

Q[e ]
Q[e ]
-

-
-
1000 1000 1000

800 800 800

600 600 600

400 Legend 400 Legend 400 Legend

Data Data Data
Fit Fit Fit
200 Holes 200 200
Holes Holes
Electrons Electrons Electrons

00 100 200 300 400 500 00 00 50 100 150 200 250 300 350 400 450
100 200 300 400 500
Bias [ V ] Bias [ V ] Bias [ V ]

Figure 3.28: Fit to measurement points at 60 mm from chip side. Strips 77,
145, 208 ,281, 344 and 470 from the upper left to the lower right.

×10
3
×10 ×10
3 3
Q[e ]

Q[e ]

Q[e ]
-

-
1000 1000 1000

800 800 800

600 600 600

400 400 Legend 400 Legend

Legend
Data Data Data
Fit Fit Fit
200 200 Holes 200 Holes
Holes
Electrons Electrons Electrons
00 100 200 300 400 500 00 00
100 200 300 400 500 100 200 300 400 500
Bias [ V ] Bias [ V ] Bias [ V ]

×10 ×10
3 3
×10
3
Q[e ]

Q[e ]
Q[e ]

-
-

1000 1000 1000

800 800 800

600 600 600

400 Legend 400 Legend 400 Legend

Data Data Data
Fit Fit Fit
200 Holes 200 Holes 200 Holes
Electrons Electrons Electrons
00 00 100 200 300 400 500 00 100 200 300 400 500
100 200 300 400 500
Bias [ V ] Bias [ V ] Bias [ V ]

Figure 3.29: Fit to measurement points at 80 mm from chip side. Strips 77,
145, 208 ,281, 344 and 470 from the upper left to the lower right.

the final decision was guided by the behaviour of adjacent measurement

points, i.e. imposing a consistency principle.
I report in Tabs. 3.6 and 3.7 the results of the fits on full-depletion
voltage and carriers trapping-time respectively; moreover additional mea-
surement points, performed in the highest irradiated region, are reported in
Tab. 3.8. I don’t report the same table for the ρ 0 parameter since all values
are consistent with 3300 e− within 2%.
3.6 Results and discussion 73

×10 ×10 ×10

3 3 3

Q[e ]

Q[e ]

Q[e ]
-

-
1000 1000 1000

800 800 800

600 600 600

400 Legend 400 Legend 400 Legend

Data Data Data
Fit Fit Fit
200 Holes 200 Holes 200 Holes
Electrons Electrons Electrons
00 100 200 300 400 500 00 100 200 300 400 500 00 100 200 300 400 500
Bias [ V ] Bias [ V ] Bias [ V ]

×10 ×10
3 3
×10
3

Q[e ]

Q[e ]
Q[e ]

-
-

1000 1000 1000

800 800 800

600 600 600

400 Legend 400 Legend Legend

400
Data Data Data
Fit Fit Fit
200 Holes 200 Holes 200 Holes
Electrons Electrons Electrons
00 00 100 200 300 400 500 00 100 200 300 400 500
100 200 300 400 500
Bias [ V ] Bias [ V ] Bias [ V ]

Figure 3.30: Fit to measurement points at 90 mm from chip side. Strips 77,
145, 208 ,281, 344 and 470 from the upper left to the lower right.

×10
3
×10
3
×10
3
Q[e ]

Q[e ]
Q[e ]
-

-
-

1000 1000 1000

800 800 800

600 600 600

400 Legend 400 Legend 400 Legend

Data Data Data
Fit Fit Fit
200 Holes 200 Electrons 200 Electrons
Electrons Holes Holes
00 100 200 300 400 500 00 00
100 200 300 400 500 100 200 300 400 500
Bias [ V ] Bias [ V ] Bias [ V ]

×10 ×10 ×10

3 3 3
Q[e ]

Q[e ]

Q[e ]
-

1000 1000 1000

800 800 800

600 600 600

400 Legend 400 Legend 400 Legend

Data Data Data
Fit Fit Fit
200 Electrons 200 Electrons 200 Holes
Holes Holes Electrons
00 100 200 300 400 500 00 100 200 300 400 500 00 100 200 300 400 500
Bias [ V ] Bias [ V ] Bias [ V ]

Figure 3.31: Fit to measurement points at 100 mm from chip side. Strips
77, 145, 208 ,281, 344 and 470 from the upper left to the lower right.

I will now discuss the relative contribution that electrons and holes give
to the total induced signal.

3.6.3 Remarks on fits

Holes are the carriers that give the highest contribution to the total in-
duced signal for n-type bulk. This fact is explained reminding that, for non-
irradiated sensors, holes drift toward higher electric field magnitude, thus,
74 Silicon sensors radiation damage characterization

×10 ×10 ×10

3 3 3

Q[e ]

Q[e ]

Q[e ]
-

-
1000 1000 1000

800 800 800

600 600 600

400 Legend 400 Legend 400 Legend

Data Data Data
Fit Fit Fit
200 Electrons 200 Electrons 200 Electrons
Holes Holes Holes
00 100 200 300 400 500 00 100 200 300 400 500 00 100 200 300 400 500
Bias [ V ] Bias [ V ] Bias [ V ]

×10
3
×10
3
×10
3

Q[e ]

Q[e ]
Q[e ]

-
1000

-
-

1000
1000

800 800
800

600 600
600

400 Legend 400 Legend 400 Legend

Data Data Data
Fit Fit Fit
200 200 200
Electrons Electrons Electrons
Holes Holes Holes
00 00 100 200 300 400 500 00
100 200 300 400 500 100 200 300 400 500
Bias [ V ] Bias [ V ] Bias [ V ]

×10
3
×10 ×10
3 3

Q[e ]
Q[e ]

Q[e ]
1000

-
-

-
1000 1000

800 800 800

600 600 600

400 Legend 400 Legend

Legend
400
Data Data Data
Fit Fit Fit
200 200 200 Holes
Electrons Electrons
Holes Holes Electrons
00 00 00 100 200 300 400 500
100 200 300 400 500 100 200 300 400 500
Bias [ V ] Bias [ V ] Bias [ V ]

Figure 3.32: Fit to measurement points at 110 mm from chip side. Strips
77, 110, 145, 174, 208 ,281, 311, 344 and 470 from the upper left to the
lower right.

the integral of the work done by the electric field on this type of carrier is
greater than the work done on electrons. Indeed, in the fits of Figs. 3.26
and 3.27 performed in the non-irradiated region of Mod1, the total induced
signal by holes is always higher than the signal induced by electrons.
In case of moderate irradiation, before type inversion, the carriers
trapping-time is decreased and both electrons and holes might be trapped
before reaching the electrodes. In this situation, due to the higher mobility
of electrons with respect to holes, the difference of the two contributions to
the total induced signal is reduced. Indeed, we can also foresee a situation
in which, even before type inversion, electrons give the highest contribution.
The first plot of the second row of Fig. 3.30 shows exactly this behaviour, in
which, to fit that data set, I didn’t exchange the role of electrons and holes
with respect to a non-irradiated sensor.
In case of high irradiation the bulk becomes of p-type and the junction
moves to the back-plane side. In this case the carriers going toward the
3.6 Results and discussion 75

Full-depletion voltage
D # strip
[mm] 77 145 208 281 344 470
10 98.4±1.6 103.9±1.5 100.9±1.4 107.4±1.5 102.2±1.4 101.0±1.4
40 123.9±1.6 128.0±1.7 120.3±1.6 123.5±1.6 111.7±1.4 131.2±1.6
60 124±2.0 117.0±1.8 140.5±2.9 115.2±3.3 134.7±1.7 150.4±1.2
80 145.8±2.6 106.6±3.9 80.5±3.6 86.8±5.2 102.0±3.7 149.0±2.4
90 102.5±3.9 75.2±3.9 38.0±7.5 15.0±7.1 50.5±2.9 109.2±3.8
100 39.6±3.0 23.0±1.3 22.6±1.4 24.1±1.5 21.0±1.2 58.9±2.7
110 25.5±1.2 21.7±1.5 31.3±1.5 41.7±1.7 30.8±1.4 28.6±4.8

Table 3.6: Fit results on full-depletion voltage. The data reported are the
sum of the effective full-depletion voltage, V d0 , and the Vof f parameter. The
errors reported are the sum in quadrature of the errors returned by the fit
on the two parameters. An error of 6%, coming from the bias of the model,
should be added as reported in Tab. 3.5. D: distance from chip side.

Trapping-time
D # strip
[mm] 77 145 208 281 344 470
10 1000 1000 1000 1000 1000 1000
40 1000 1000 1000 1000 1000 1000
60 1058±1282 1962±3157 1237±681 930±579 2768±3404 800±688
80 242±98 162±84 74±10 68±11 70±13 1665±1555
90 84.3±18.4 72.3±14.0 59.8±6.1 67.6±4.3 62.7±7.5 77.1±14.6
100 67.4±6.5 40.1±1.6 38.0±1.6 37.6±1.6 44.2±1.9 70.3±9.0
110 50.8±2.7 35.7±1.5 29.3±1.4 25.5±1.5 29.2±1.3 61.3±5.0

Table 3.7: Fit results on trapping-time. The errors reported are those re-
turned by the fit. An error of 6%, coming from the bias of the model, should
be added as reported in Tab. 3.5. D: distance from chip side.

higher electric field magnitude are electrons, which are also the carriers with
the highest mobility; therefore they will give always the highest contribution
to the total induced signal. The plots for which I had to exchange the role
of electrons and holes are in Figs. 3.31 and 3.32, for instance: (strip 145,
100 mm), (strip 208, 100 mm), (strip 281, 100 mm), (strip 344, 100 mm),
76 Silicon sensors radiation damage characterization

Vd and τ for strips 77, 110 and 174 at 110 mm

Parameter # strip
110 174 311
Vd 16.1±1.3 19.8±1.7 31.5±1.7
τ 45.5±2.2 30.3±1.2 26.2±1.3

Table 3.8: Additional measurement points in the highest irradiated region

for strips 110, 174 and 311, at 110 mm with respect to the non-irradiated
side. The errors reported are those returned by the fit. An error of 6%,
coming from the bias of the model, should be added as reported in Tab. 3.5.

(strip 77, 110 mm), (strip 110, 110 mm), (strip 145, 110 mm), (strip 174,
110 mm), (strip 208, 110 mm), (strip 281, 110 mm), (strip 311, 110 mm)
and (strip 344, 110 mm).

3.6.4 Results on full-depletion voltage and carriers trapping-

time

I will now discuss the progression of full-depletion voltage and carriers

trapping-time along the strips considering three significant examples of my
measurements: laser beam centered on lateral strips (#77, #470) and laser
beam centered on a central strip (#281). Figure 3.33 reports the full-
Full-depletion Voltage [V]

Full-depletion Voltage [V]

Full-depletion Voltage [V]

160 160 160

140 140 140
120 120 120
100 100 100
80 80 80
60 60 60
40 Legend 40 Legend 40 Legend
20 Strip 77 20
Strip 281 20 Strip 470
0 00 00
0 20 40 60 80 100 20 40 60 80 100 20 40 60 80 100
Coord. Along Strips [mm] Coord. Along Strips [mm] Coord. Along Strips [mm]

Figure 3.33: Full-depletion voltage along strips. Strips 77 and 470 are lateral
strips, whereas strip 281 is a central strip, which presents a minimum at
∼90 mm, as expected for intrinsic Silicon.

depletion voltage along the three strips. All the strips present a big drop in
the measured full-depletion voltage in correspondence of the highest irradi-
ated side (from 80 mm on). For the central strip the full-depletion voltage
3.6 Results and discussion 77

reaches a minimum at ∼90 mm, then it starts raising again.

300 300
Trapping-time [ns]

300

Trapping-time [ns]
Trapping-time [ns]
Legend Legend Legend
250 Strip 77 250 250
Strip 281 Strip 470
200 200 200

150 150 150

100 100 100

50 50 50

75 80 85 90 95 100 105 110 115 75 80 85 90 95 100 105 110 115 75 80 85 90 95 100 105 110 115
Coord. Along Strips [mm] Coord. Along Strips [mm] Coord. Along Strips [mm]

Figure 3.34: Trapping-time along strips. Strips 77 and 470 are lateral strips,
whereas strip 281 is a central strip, which presents the lower trapping-time
value, correlated to a higher concentration of crystal damages.

Figure 3.34 reports the carriers trapping-time along the three strips. All
the strips present a decrease of the trapping-time in correspondence of the
highest irradiated side. The 110 mm lateral strips present a trapping-time
greater than 50 ns, whereas for the central strip, and in correspondence to
a region of higher concentration of crystal damages, it drops to ∼20 ns.
The data presented on the full-depletion voltage and carriers trapping-
time are consistent with the picture in which the irradiation beam was cen-
tered on the edge of the sensor, opposite to the readout electronics, as ex-
pected. Indeed, the decrease of the full-depletion voltage along the central
strip 281, or else the decrease of the bulk effective doping concentration that
originally was of n-type, is related to an increment of the absorbed fluence.
At ∼90 mm, and in the central region, the effective doping concentration
of the bulk becomes intrinsic. Going forward, toward the edge of the sen-
sor, the absorbed dose is even higher and the effective doping concentration
of the bulk becomes of p-type, causing the increment of the full-depletion
voltage (see Fig. 2.3 for comparison).
The model also correctly estimates the initial amount of electron-hole
pairs generated by the laser. The value deduced by the model is: Q M =
ρ0 · w = 3300 · 320 ' (1.06 ± 0.02)×106 e− per each laser pulse, which is
in very good agreement with the laser power at the tip of the optical fiber.
In fact from the laser characteristics (see Tab. 3.2) I calculated the energy
5 mW·1 ns
impinging on the sensor surface, 2·1.6×10−19 C
= 15.6 × 106 eV, where the
factor 2 is introduced by the presence of the beam splitter. This value is
78 Silicon sensors radiation damage characterization

120 µm
E0 1/3E0 2/3(1−α)E’ 2/3(1−α)E’’ 2/3(1−α)E’’’
40 µ m

Al strips
E’ (1−α)E’ E’’ (1−α)E’’ E’’’ (1−α)E’’’ E’’’’

320 µm
θ

Bulk
Al back plane

Figure 3.35: Laser multiple reflections inside the micro-strip Silicon sensor.
E0 is the impinging laser energy. (1 − α)E is the fraction of energy not
absorbed by the Silicon bulk.

dumped by a factor that takes into account the multiple reflections due to
strips and back-plane metallization. Let’s define α = 1 − e −2w/x0 as the
fraction of energy released in crossing two times the Silicon bulk thickness
as shown in Fig. 3.35; in the hypothesis of small incident angle θ, the total
energy, Ein , released inside the bulk is:

Ein = E0 2/3α + E0 2/3(1 − α)1/3α + E0 2/3(1 − α)2 1/9α + · · · =

∞ 
1−α n

X 2α
= E0 2/3α = E0 (3.15)
3 2+α
n=0

The total electron-hole pairs generated inside the sensor by each single laser
pulse is QC = Ein /3.65, where 3.65 is the energy needed, in elettronvolt, to
generate an electron-hole pair in Silicon at −12 o C. The calculated QC is
equal to the measured QM if the laser absorption length in Silicon (see the
laser characteristics in Tab. 3.2), at −12 o C, is x0 = 1920 µm, which is a very
reasonably assumption as one can see from I. Abt et. al. [26], S.M. Sze [29]
and S. Gadomski et. al. [39].

3.6.5 From trapping-time to fluence

A very interesting result is related to the reconstructed radiation fluence.

From the carriers trapping-time I easily recovered the radiation fluence
3.6 Results and discussion 79

through the empirical relation expressed by Eq. (3.13). The sensor fluence-

0.8

0.6

0.4

0.2

0
60
Co 50
ord 40
. T 30
ran 100
sv. 20 60
80
[mm]
Str 10
ips 0 40
long Strips
oord. A
20
[m 0 C
m]

Figure 3.36: 2D-reconstruction of the fluence profile.

map was then fitted with a 2D-Gaussian shape giving the radiation beam
profile shown in Figs. 3.36 and 3.37. My measurements are in good agree-
ment with the beam properties measured at IUCF (see Fig. 3.6) as reported
in Tab. 3.9. This constitutes an important cross-check for the reliability of

Beam properties IUCF measurement My measurement

Peak 0.8×1014 (1.01±0.12)×1014
[1 MeV eq.n. per cm2 ]
Width σx = σy ' 19 σx = 21.9 ± 3.0
[mm] σy = 33.4 ± 1.5

Table 3.9: Radiation beam properties. Comparison between irradiation pro-

file gathered through my model and the beam characteristics measured at
IUCF with the wire scanner technique.

my model. Reconstructed fluence cross-sections, fitted with a 2D-Gaussian

at different positions, referred as (A) and (B) in Fig. 3.37, are also reported
in Fig. 3.38.
Another good result is that the minimum at 90 mm for the central strip in
Fig. 3.33 is in correspondence of a region that received a radiation fluence of
∼0.35×1014 1 MeV equivalent neutrons per cm2 , which is about the expected
fluence needed to obtain type inversion for this kind of sensors, as one can
see from Fig. 2.3.
80 Silicon sensors radiation damage characterization

Coord. Transv. Strips [mm]

80 0.9

(A) (B) 0.8

60 0.7

0.6
40
0.5

20 0.4

0.3
0
0.2

-20 0.1

0
0 20 40 60 80 100 120 140 160 180 200 220
Coord. Along Strips [mm]

Figure 3.37: Contour plot of the fluence profile (colour scale is in unit
of 1014 1 MeV equivalent neutrons per cm2 ). The fit parameters are:
(1.01±0.12)×1014 1 MeV equivalent neutrons per cm2 , mean along strips
120±6 mm, σ along strips 21.9±3.0 mm, mean transversal to strips
33.4±1.5 mm, σ transversal to strips 19.0±0.8 mm. Dashed line: sensor
position. (A) and (B) refer to the fluence profile cross-sections reported in
Fig. 3.38. Fluence [1014 1MeV eq.n] [cm]-2
Fluence [1014 1MeV eq.n] [cm]-2

1 χ2 / ndf 5.294 / 3 1
Coef 0.3769 ± 0.0408
Mean 29.61 ± 1.10
0.8 Sigma 11.76 ± 0.74 0.8

0.6 0.6

0.4 0.4 χ2 / ndf 2.053 / 6

Coef 0.9098 ± 0.0390
Mean 32.32 ± 0.77
0.2 0.2 Sigma 18.69 ± 1.02

00 0
10 20 30 40 50 60 0 10 20 30 40 50 60
Coord. Transv. Strips [mm] Coord. Transv. Strips [mm]

Figure 3.38: Reconstructed fluence profile cross-sections. On the left: cross-

section at 80 mm from chip side; cfr. (A) in Fig. 3.37. On the right: cross-
section at 110 mm from chip side, corresponding to the highest irradiated
region; cfr. (B) in Fig. 3.37.

3.7 Conclusions

With a very simple method based on an analytical model for the Q-V char-
acteristics I was able to extract all the physical information required to char-
acterize a Silicon sensor for use in HEP experiments. The method allows for
3.7 Conclusions 81

a local (point to point) characterization of the sensor, which is particularly

important in case of a highly non-uniform irradiation. The analytical model
has four free parameters only and it is capable to extract reliable informa-
tion even in case of highly radiation-damaged sensors. This method has also
been proven to provide a very good estimate of the dose locally absorbed by
the sensor. Thanks to this method, I can conclude that the sensors I tested
are well suited for use in the forward region of hadron collider experiments.
82 Silicon sensors radiation damage characterization
Chapter 4

Micro-strip detector
prototype

The trigger system of BTeV is conceived to be asynchronous with the Bunch-

CrOssing (BCO); this implies that no trigger signal is available to initiate
readout of micro-strip data. Furthermore, in order to be able to readout
all the information about an event at each BCO, the amount of data that
needs to be buffered must be limited. All hit data must be readout in a zero
suppressed format and spurious hit data must be minimized. The read-
out electronics must therefore perform a data-driven sparsified readout,
which means that every time a channel is above a settable threshold, a hit
data must be sent out. The Fermilab Silicon Strips Readout chip (FSSR)
is a custom Integrated Circuit (IC) that has been designed to meet these
requirements.

4.1 The readout chip

The IC [40] is made with standard deep-submicron 0.25 µm CMOS technol-

ogy, which has been proved to have good radiation hardness performance.
It uses enclosed geometry transistors [41] to meet the radiation tolerance re-
quirement, which is expected to be (2−8)×10 13 1 MeV equivalent neutrons
per cm2 , in 10 years of BTeV operation. FSSR has 128 channels, each con-
nected to a detector strip. The signals from the strips, after amplification
84 Micro-strip detector prototype

and shaping, are compared to a preset threshold. To achieve the required

position resolution, the channels have to provide only a binary information
(hit / no hit), generating a logical 1 at the output if a signal exceeding the
threshold is detected. An additional 3 bit analog information, provided by a
flash-ADC, together with an internal square-wave pulse generator are used
for calibration and monitoring purposes.
The dimensions of the readout IC are ∼7.5×4.5 mm 2 and its layout is
presented in Fig. 4.1. The IC needs two power supply voltages, +2.5 V
analog and +2.5 V digital, and two separate grounds for the analog and
digital section respectively; the nominal power dissipation per channel is
∼4 mW.

Figure 4.1: Layout of the Fermilab Silicon Strips Readout chip (FSSR).

For each channel with a signal above threshold the strip number, the
analog information and the related BCO number are readout and transmit-
ted. Two versions of FSSR have been made with only slight differences. I
will refer to the second version.

4.1.1 The analog section

The analog section of the FSSR core [42] consists of 128 channels, each
including (see Fig. 4.2):

charge preamplifier: integrates the input charge generated in the active

volume of the sensor, the Gm transconductance is needed to discharge
4.1 The readout chip 85

Detector Gm

CF Comparator
CAC
Baseline
T(s)
Restorer
Qδ
CD
Preamplifier Integrator +
Shaper Vth

Figure 4.2: FSSR analog channel block diagram.

the feedback capacitance CF

integrator and shaper: used to ameliorate the signal-noise ratio. Its

transfer function is CR-(RC)2 -type with a programmable peaking time
of: 65 ns, 85 ns, 100 ns and 125 ns

comparator: provides the binary information (hit / no hit) to the digital

section

BaseLine Restorer (BLR): included to achieve baseline shift suppres-

sion. It can be bypassed by a programmable switch

inj0 inj1 inj127

InjectIn

40 fF
in0
Preamplfier in1
Input

in127

Figure 4.3: FSSR programmable inject mask circuit.

The electronic calibration can be performed either by the internal square-

wave pulse generator or by an external pulser, providing voltage steps on the
integrated inject capacitance of 40 fF. The injected channels can be selected
by a programmable inject mask as shown in Fig. 4.3.
86 Micro-strip detector prototype

Shaping time Feedback capacitance Gain

[ns] [fF] [mV fC− 1]
125 150 93
85 150 96
65 150 101
125 100 137
85 100 142
65 100 149

Table 4.1: FSSR nominal gains expected from simulations.

The gain of the preamplifier is also programmable by changing the value

of its feedback capacitance, as shown in Fig. 4.24. The nominal gain values
expected from simulations, as a function of the shaping time and feedback
capacitance values, are reported in Tab. 4.1.

Cf2=50 fF
Switch

Preamplifier Preamplifier
Input Output

Cf1=100 fF

Figure 4.4: Preamplifier selectable feedback capacitances.

Figure 4.5 shows the block diagram of the flash-ADC; there are eight pro-
grammable 8 bit-DACs that generate thresholds common to all the channels.
A buffer has been introduced in order to reduce the capacitive load to the
shaper due to the 1−7 DACs, thus avoiding unwanted changing of the trans-
fer function of the analog chain. The flash-ADC generates a thermometric
code which is then converted to a 3 bit binary code.

It is also possible to monitor the output of all the analog blocks of the
128th channel and consequently check the correctness of the analog section
working-point.
4.1 The readout chip 87

Threshold
Circuit 7 Comparator 7
Single−ended/ −
Differential
Conversion +

DAC 7

Threshold
Circuit 6 Comparator 6

Single−ended/ −
Differential
Conversion +

DAC 6
ADC
Output

Buffer Threshold
Circuit 1 Comparator 1

− Single−ended/ −
Differential
+ Conversion +

DAC 1

Threshold
Circuit 0 Comparator 0

BLR/Shaper Single−ended/ − Binary

Differential
Output Conversion + Information

DAC 0

Figure 4.5: FSSR flash-ADC. The analog signal is digitized using three bits.
The eight thresholds are common to all the channels.

4.1.2 Measurements

A special board, shown in Fig. 4.6, was built for testing the FSSR chip
without any load. The board allows to have access, through lemo connectors
and special pins, to the main analog signals of the chip.
The amplitude of the threshold for the number 0 comparator is available
on one pad of the chip. The linearity of the DAC that generates this thresh-
old was measured, the data and the relative linear fit are reported in Fig. 4.7.
The DAC of threshold 0 presents a remarkable linear behaviour. The DAC
circuits that generates the other thresholds are rigorously identical to the
DAC of threshold 0.
The amplitude of the internal pulser is also available on one pad of the
chip. Figure 4.8 reports the measured amplitudes for different DAC values.
88 Micro-strip detector prototype

Figure 4.6: FSSR single chip test board. The chip is under the black cover
and is wire-bonded to pads on the board. The connector of the digital signals
is visible on the left side of the board. The other lemo connectors and special
pins are used for power supply and diagnostic purposes.

Figure 4.7: Linearity of the DAC that generates the threshold for the number
0 comparator.

As one can observe from the linear fit, also the internal pulser DAC shows
a good linear behaviour.
I will now present measurements about the analog section for different
settings of the shaping time and preamplifier gain. Figure 4.9 shows the
shape of the analog signal measured after the shaper and the baseline re-
storer. These measurements were performed with a shaping time of 125 ns,
low preamplifier gain and injecting 1 fC of charge. It’s evident the effect of
the BLR to restore the signal to zero.
Figure 4.10 shows the analog signal for the four different shaping times:
4.1 The readout chip 89

Figure 4.8: Linearity of the DAC that generates the internal pulser ampli-
tude.

Figure 4.9: Analog signal after shaper and baseline restorer. Shaping time:
125 ns; injected charge: 1 fC; low preamplifier gain.

65 ns, 85 ns, 100 ns and 125 ns. The measurements were taken after the
baseline restorer, with low preamplifier gain and injecting 1 fC of charge.
In Fig. 4.11 the signals measured after the baseline restorer are compared
for the two different preamplifier gain values. These measurements were
performed with a shaping time of 125 ns and injecting 1 fC of charge.
I also measured the linearity of the analog channel by correlating the
amplitude of the signal peak after the baseline restorer with different injected
charges1 . The measurements were performed with a shaping time of 125 ns
1
The most probable quantity of charge generated in 320 µm of Silicon by a minimum
90 Micro-strip detector prototype

Figure 4.10: Analog signal at different shaping times. The signal is measured
after the baseline restorer. Injected charge: 1 fC; low preamplifier gain.

Figure 4.11: Analog signal at different preamplifier gains. The signal is

measured after the baseline restorer. Shaping time: 125 ns; injected charge:
1 fC.

and with low preamplifier gain. The data and the relative linear fit reported
in Fig. 4.12 show the remarkable linear behaviour of the analog channel.

4.1.3 The digital section

The architecture of the digital back-end is called pseudo-Pixel. It is based

on the BTeV pixel readout chip, FPIX2 [43]. The I/O protocols for the two

ionizing particle is ∼3.6 fC.

4.1 The readout chip 91

Figure 4.12: Analog signal at different injected charges. The signal is mea-
sured after the baseline restorer. Shaping time: 125 ns; low preamplifier
gain. On the left: analog signals. On the right: linear correlation between
the injected charge and the peak of the analog signal.

chips are identical. The 128 channels are subdivided into 16 columns of 8
rows, FSSR will look like an FPIX2 of 16×8 array of channels. The FSSR
architecture can be described as including four logic sections, as shown in
Fig. 4.13:

• the core consists of 128 analog readout channels, the End-of-column

logic (16 blocks, one for each column of front-end channels) and the
Core Logic, which controls the data flow from the core to the data
output interface

• the data output interface accepts data from the core, serializes the
data and transmits them off-chip, using a point-to-point protocol

• the programming interface accepts commands and data from a serial

input bus and, in response to a command, provides data on a serial
output bus. The user can read back the contents of the main pro-
grammable registers

• the programmable registers are used to store input values for DACs
that provide currents and voltages required by the core, for instance
the threshold levels for the discriminators and the bias currents of the
92 Micro-strip detector prototype

Microstrip analog readout electronics

(128 channels)

End−Of−Column Logic

CORE
(16 blocks)

Core Logic

PROGRAMMABLE Clock Next

DACs Control Block

DATA OUTPUT
REGISTERS

INTERFACE
Logic Word

Word Serializer
PROGRAMMING INTERFACE
Steering Logic

BCO CLK I/O MCLKA / MCLKB High Speed

(Readout Clock) Output

Figure 4.13: FSSR chip block diagram. Arrows represent control and data
flow.

analog section. They also have additional functions, such as controlling

data output speed and selecting the pattern for charge injection tests

All I/O (except the test signal injection) is differential and is fed by means of
Low Voltage Differential Signaling (LVDS) to reduce data corruption. The
functioning principle of LVDS signals is illustrated in Fig. 4.14.
Let me now go more in detail about the FSSR Core. FSSR Core is a
column-based architecture that uses an indirect addressing scheme to as-
sociate pixel hits with a time stamp. It is best understood as consisting
of three mutually dependent functional blocks as shown in Fig. 4.15. These
three blocks are the Core Logic, the End-of-column Logic and the Pixel
Cell:

Core Logic: understands time. At the rising edge of the bunch-crossing

clock, it stamps every time slice with an 8 bits BCO number. This
number is broadcast to all End-of-column Logic blocks. The Core
Logic also contains a very simple state machine that knows if the Core
is Talking or Silent. The Core is Silent until the End-of-column Logic
blocks indicate that they have data to output. When this happens,
at the next rising edge of the readout clock, the Core will switch to
4.1 The readout chip 93

1.4 V
100 Ω

A−T Ab−F

A−F Ab−T

4 mA

Figure 4.14: Schematic principle of the Low Voltage Differential Signaling

concept. When the signal being driven is TRUE, switches A-T and Ab-T
are closed, and switches A-F and Ab-F are open. The voltage on the non-
inverting amplifier input is 1.4 V and 4 mA flows through the 100 Ω resistor,
making the voltage of the inverting terminal 1 V. When the signal is FALSE,
switches A-F and Ab-F are closed, A-T and Ab-T are open and the voltages
and direction of the current flow are reversed.

the Talking state, and output can begin. The Core Logic does not
wait for any kind of chip token or validation from the DAQ. While in
the Talking state, the Core starts passing a Horizontal Token across
the End-of-column Logic blocks to arbitrate rights to the output bus.
When the Horizontal Token drops out of the other side of the End-of-
column Logic blocks, readout is done. The Core switches back to the
Silent state at the next rising edge of the readout clock

End-of-column Logic blocks: also understand time, in that whenever

there is a hit, they store the BCO numbers broadcast by the Core
Logic. Obviously, they also understand hits, which are driven to them
from the Pixel Cells via the HFastOR signal. The End-of-column Logic
blocks also understand the existence of the Pixel Cells because they
communicate to those pixels through a series of command and tokens.
Finally, the End-of-column Logic blocks understand output. When
the Core is in the Talking state and when a particular End-of-column
Logic block has the Horizontal Token and when that End-of-column
94 Micro-strip detector prototype

Figure 4.15: The FSSR core organization showing the three major logical
blocks: the Pixel Cell, the End-of-column Logic blocks and the Core Logic.

Logic block has hit pixels to output, then it outputs those pixels

Pixel Cells: themselves know nothing of time. They only understand hits
and commands from their End-of-column Logic block. These com-
mands are Idle (do nothing), Listen (listen for new hits), Reset (reset
your contents), and Output (output your contents). There are four sets
of such commands coming from the End-of-column Logic block. If a
pixel is Empty and it receives a hit, it associates itself with whichever
command set is issuing the Listen command. From that point and
until the Pixel Cell is reset, it obeys only the commands from the as-
sociated command set. Rights to the column output bus are arbitrated
by a Column Token issued by the End-of-column Logic block

The End-of-column has two type of state machines: the Column State
Machine, which operates at the rising edge of the readout clock; four Com-
mand State Machines that operate on the rising edge of the BCO clock,
see Fig. 4.16. In the Empty State, the Command State Machines issue the
Idle Command to the Pixel Cells. In the Listen State, they issue the Listen
Command. Once a hit is received, several things happen:

1. the BCO number currently being broadcast by the Core Logic is stored
in the register associated with the Command State Machine currently
in the Listen State
4.1 The readout chip 95

Figure 4.16: The End-of-column Logic block’s state diagrams.

2. that state machine makes the transition to the Full State where it once
again issues the Idle command

3. the state machine picked by the Hit Priority Encoder as next to Listen
moves to the Listen State

Figure 4.17: The timing diagram of a single pixel hit in a single column.

Figure 4.17 reports the timing diagram of the Core operations when a “hit”
occurs. The Column State Machine starts in the “Nothing to Say” or Noth-
ing State where it remains until it sees an Output command issued by any
of the Command State Machines. At the next rising edge of the readout
96 Micro-strip detector prototype

clock, the Column State Machine makes the transition to the “Something to
Say” or Something State. At this point, the Core logic is alerted to the fact
that there is data to output, and the Column Tokens are sent up to the first
Pixel Cell that needs to be output. The state machine then waits for the
arrival of the Horizontal Token from the Core Logic. When the token ar-
rives, the Column State Machine makes the transition to the Talking state,
and it releases the readout clock to the Pixel Cells enabling them to out-
put their data. Simultaneously, the stored BCO number (which associates
the hit pixels with the time they were hit) is driven onto the bus. At the
next rising edge of the readout clock, the Column State Machine makes the
transition to the Silent state. When the entire array has been read out, the
Horizontal Token drops out of the last End-of-column Logic Block, and the
Core Logic makes its transition from Talking to Silent. This signals to all
the Column State Machines that they can make their own transition back
to the Nothing State. The elapsed time between the occurrences of an hit
and the output of the data is the sum of latency time for BCO clock period
completion, plus one BCO clock period, plus three readout clock periods.
FSSR is designed to work at 396 ns BCO (2.5 MHz) as well at 132 ns
(7.5 MHz) with a readout clock designed to be 70 MHz. The number of
output lines, through which data are sent out, is selectable and can be
1, 2, 4 or 6. This means that each single data word can be readout as
a single stream of 24 bits, or two streams of 12 bits and so on. Data are
readout on both edges of the readout clock for a maximum data transmission
rate of 840 Mbit sec−1 . The 24 bits of the data word carry the following
information: bit 0, word marker = 1; bits 1−3, channel pulse-height;
bits 4−11, BCO number; bits 12−16, column number; bits 17−20, row
number; bits 21−23, not used, set to 0. Data sent out from the chip are
not time-ordered. The BCO number is used off-line to build events. Since
the data output lines are point-to-point connected, no chip ID information
is needed. The same lines are used for synchronization and monitoring
purposes too.
A bus line, shared by all the chips of the same detector, is used to
program the chips. The command word has 13 bits organized as: bits 0−2,
instruction (write, read, set ,reset or default); bits 3−7, register number;
4.1 The readout chip 97

bits 8−12, chip ID.

The other main programmable quantities are the polarization currents
of the analog section, used to set the working point of the circuit, and the
128 bits of the kill and inject masks.

4.1.4 Efficiency simulations

Simulations of the digital section in Verilog language have been carried out
to check the performance in terms of readout efficiency. The input to the
Verilog program is a file generated by the simulation framework of BTeV.
The file is a list of channels that have been hit by a particle. In the simulation

110 99.6% 95.6%

Readout efficiency [%]

100

90

80

70 Legend
Dig. lines : 1
60
Dig. lines : 2

50 Dig. lines : 4
Dig. lines : 6
40
4 6 8 10 12 14 16 18 20
< Int. per beam crossing >

Figure 4.18: Readout efficiency of the FSSR digital section. The simulation
was run at 396 ns of BCO and 70 MHz of readout clock with 6, 12 and 18
average interactions per BCO. The four cases correspond to 1, 2, 4 and 6
digital output lines.

I considered only the chip with the highest occupancy corresponding to the
innermost detector of the first station. The chip analog section is considered
just as a delay in the signal processing chain. I set the FSSR readout clock of
the simulation at its nominal value of 70 MHz and the BCO period to 396 ns.
The readout efficiency is calculated for three different average number of
interactions per BCO: 6, 12, and 18, and for different number of output lines.
As one can see from Fig. 4.18, using 6 output lines the readout efficiency
goes from 99.6% to 95.6% for 12 and 18 average number of interactions per
BCO respectively. It is worth noting that the number of interactions per
98 Micro-strip detector prototype

BCO for BTeV is 6 at 2 TeV and a luminosity of 2×10 32 cm−2 s−1 .

This performance ensures that the detector system operates with ade-
quate readout efficiency in all the Tevatron-like accelerator environments.
In particular for BTeV a readout efficiency > 95% was required.

4.2 Hybrid module performance

In this section I will present the results on gain, noise and threshold disper-
sion of FSSR chips assembled on a “hybrid” module. A hybrid module is
a circuit board that has been designed to interface FSSR chips to a Silicon
micro-strip sensor. It hosts six FSSR chips with the relative passive compo-
nents for power supply filtering. The module accommodates also pads for
diagnostic purposes and the high voltage filtering network for sensor bias.
A picture of a hybrid module is shown in Fig. 4.19.

Figure 4.19: FSSR hybrid module picture. The hybrid module with the six
chips together with the “pig-tail” carrying the connector to the flex circuit,
are visible.

All the measurements, here reported, have been performed with the in-
ternal square-wave pulse generator. The threshold curves were measured
setting the comparator thresholds at fixed values and changing the pulser
amplitude. Each point of a threshold curve represents the percentage of
times that the comparator fires for a certain value of injected charge. 255
pulses were sent for each point. The measurement error has been calculated
p
as σ = (1 − )/255 with 0 <  < 1. The conversion from mV to elec-
trons was performed considering a nominal value for the inject capacitance
4.2 Hybrid module performance 99

of 40 fF.

χ2 / ndf 6.365 / 15 χ2 / ndf 1.239 / 13

21.6 ± 0.0 15.33 ± 0.05

Norm. Entries

Norm. Entries
Mean Mean
1 1
Sigma 1.017 ± 0.045 Sigma 1.079 ± 0.044

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

010 15 20 25 30 35 05 10 15 20 25 30
Pulser Ampli. [mV] Pulser Ampli. [mV]
χ2/ ndf 12.13 / 18 χ2 / ndf 39.64 / 28
20.03 ± 0.06 14.13 ± 0.07
Norm. Entries

Norm. Entries
Mean Mean
1 1
Sigma 1.629 ± 0.063 Sigma 2.262 ± 0.057

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

010 15 20 25 30 35 05 10 15 20 25 30
Pulser Ampli. [mV] Pulser Ampli. [mV]

Figure 4.20: Typical threshold curves with sensor disconnected at different

combinations of shaping times and preamplifier gains. From the upper left to
the lower right: 125 ns shaping, low gain; 125 ns shaping, high gain; 65 ns
shaping, low gain; 65 ns shaping, high gain.

χ2 / ndf 7.511 / 7 χ2 / ndf 8.915 / 7

24.67 ± 3.52 22.55 ± 2.82
Entries

Entries

Constant Constant
30 Mean 229.8 ± 1.8 Mean 214.5 ± 2.0
Sigma 17.4 ± 2.0 25 Sigma 18.58 ± 1.54
25
20
20
15
15

10 10

5 5

0
100 150 200 250 300 350 400 0
100 150 200 250 300 350 400
- -
Noise [ e ] Noise [ e ]
χ2 / ndf 15.35 / 10 χ2 / ndf 15.83 / 16
16.72 ± 2.23 12.37 ± 1.67
Entries

Constant
Entries

Constant
25 Mean 390.8 ± 2.8
18 Mean 484.8 ± 3.5
Sigma 25.5 ± 2.5 16 Sigma 33.49 ± 3.23

20 14
12
15
10
8
10
6
5 4
2
0 0
350 400 450 500 550 600 650
300 350 400 450 500 550 600
- -
Noise [ e ] Noise [ e ]

Figure 4.21: Histograms of the noise at different combinations of shaping

times and preamplifier gains. From the upper left to the lower right: 125 ns
shaping, low gain; 125 ns shaping, high gain; 65 ns shaping, low gain; 65 ns
shaping, high gain.
100 Micro-strip detector prototype

Figure 4.20 shows four typical threshold curves measured at different

settings of the shaping time and preamplifier gain for the threshold of the
number 0 comparator. The relative fit function, erf(x; µσ), is also presented.
The resulting error-function parameters have been histogrammed in or-
der to obtain a figure that can describe the global behaviour of the circuit.
The σ-parameter of the error function corresponds to the noise of the chan-
nel. Figure 4.21 presents the dispersion of this parameter on a chip for the
threshold of the number 0 comparator. For comparison, the Gaussian fits
are also given. The four histograms refer to different settings of the shaping
time and preamplifier gain.
χ2 / ndf 8.557 / 8 χ2 / ndf 6.004 / 4
19.5 ± 2.7 28.2 ± 4.1
Entries

Entries
Constant Constant
22 35
Mean 4793 ± 47.2 Mean 3349 ± 33.4
20 Sigma 437.1 ± 48.1 Sigma 310 ± 39.4
30
18
16 25
14
12 20
10
15
8
6 10
4
5
2
0 0
3000 3500 4000 4500 5000 5500 6000 2000 2500 3000 3500 4000 4500 5000
- -
Threshold [ e ] Threshold [ e ]
χ2/ ndf 9.777 / 9 χ2/ ndf 10.65 / 7
22.57 ± 2.87 31.08 ± 4.03
Entries

Constant
Entries

Constant
30 3130 ± 26.9
Mean 4472 ± 38.8 35 Mean
Sigma 394.2 ± 34.2 Sigma 278.3 ± 24.5
25 30

20 25

20
15
15
10
10
5 5

0 0
3000 3500 4000 4500 5000 5500 6000 2000 2500 3000 3500 4000 4500 5000
- -
Threshold [ e ] Threshold [ e ]

Figure 4.22: Dispersion of the thresholds of a single chip at different com-

binations of shaping times and preamplifier gains. From the upper left to
the lower right: 125 ns shaping, low gain; 125 ns shaping, high gain; 65 ns
shaping, low gain; 65 ns shaping, high gain.

Since the thresholds of the comparators are common to all the channels,
another important parameter to measure is the threshold dispersion among
channels of the same chip. Figure 4.22 reports the histograms of the µ-
parameter of the error function with Gaussian fits for the threshold of the
number 0 comparator. The four histograms refer to different settings of the
shaping time and preamplifier gain.
The same measurements described above have been performed also for all
4.2 Hybrid module performance 101

Figure 4.23: FSSR pure electronic noise and threshold dispersion measured
at different combinations of shaping times and preamplifier gains: 125 ns
shaping, low gain; 125 ns shaping, high gain; 65 ns shaping, low gain; 65 ns
shaping, high gain. On the left: noise per threshold. On the right: threshold
dispersion per threshold DAC.

the eight DACs that generate the comparator thresholds and the results are
summarized in Fig. 4.23. For the higher shaping time the noise is ∼200 e −
with both preamplifier gains, consistent with what expected, whereas for the
lower shaping time the noise rises to ∼400−500 e − with the low preamplifier
gain and to ∼500−600 e− with the high preamplifier gain. The expected
noise for low shaping time should be just tens electrons higher than the
noise for high shaping time; the discrepancy I measured must be attributed
to the fact that, at 65 ns shaping time, the circuit is much more sensitive
to ground connections, that I didn’t fully optimize in my setup. From the
second plot of Fig. 4.23 it’s evident the effect of the gain on the threshold
dispersion, since the higher the preamplifier gain, the lower the threshold
dispersion. Globally the threshold dispersion is ∼300−500 e − , as expected.

The global gain was measured by a linear fit to the plot of the threshold
set values in mV versus the corresponding charge at the input, averaged
over all the chip channels. The results are reported in Fig. 4.24; the four
characteristics refer to different settings of the shaping time and preampli-
fier gain. The measured gains are: 96 mV fC −1 , for 125 ns shaping time
and low preamplifier gain; ∼110 mV fC −1 , for 65 ns shaping time and low
102 Micro-strip detector prototype

Figure 4.24: FSSR global analog section gain measured at different combi-
nations of shaping times and preamplifier gains.

preamplifier gain; 149 mV fC−1 , for 125 ns shaping time and high preampli-
fier gain; ∼151 mV fC−1 , for 65 ns shaping time and high preamplifier gain.
The measured values are consistent with the nominal values obtained from
simulations and presented in Tab. 4.1.
All the measurements presented so far show that the FSSR analog section
performance is excellent both in noise and threshold dispersion. In the next
section I will present the final measurements on a detector prototype.

4.3 Detector performance

The detector is basically composed of an hybrid module and a Silicon micro-

strip sensor. To build the prototypes I used the same sensors as those de-
scribed in chapter 3 of which I have a deep knowledge. A sketch of the detec-
tor is presented in Fig. 4.25, while pictures of the actual setup are reported
in Fig. 4.26. The entire detector was placed in an Aluminum Faraday-cage
and, in order to avoid overheating of the electronics, I used three Peltier
cells placed underneath the hybrid 2 to facilitate the heat transfer from the
chips to the Aluminum Faraday-cage.
2
The support has a window just in correspondence to the hybrid, allowing the direct
contact between Peltier cells and the hybrid itself.
4.3 Detector performance 103

512 channels
Hybrid Chips Pitch Adapter Sensor
Wire−bonds

Figure 4.25: FSSR detector sketch. A pitch adapter is needed to adapt the
50 µm FSSR pad pitch to the 120 µm sensor pad pitch.

Figure 4.26: FSSR detector pictures. Are evident the blue and red capacitors
for high voltage filtering and the Copper sticky tape that brings the high
voltage to the back-plane of the sensor. In the second picture are evident
the heat-exchanger and the fans used to dissipate the heat subtracted to the
chips by three Peltier cells placed underneath the hybrig.

The detector was characterized with the same methods described in the
previous section. Figure 4.27 shows a typical threshold curve of the number
0 comparator, measured at shaping time of 125 ns and low preamplifier
gain. The noise is ∼4.6 mV, equivalent to ∼1150 e − considering a nominal
value for the inject capacitance of 40 fF, corresponding to a signal-noise
ratio of ∼20 for a minimum ionizing particle. The expected total noise,
104 Micro-strip detector prototype

χ2 / ndf 18.24 / 25
µ 61.6 ± 0.2

Norm. Entries
1 σ 4.592 ± 0.128

0.8

0.6

0.4

0.2

030 40 50 60 70 80 90 100
Pulser Ampli. [mV]

Figure 4.27: Typical threshold curves with sensor connected, measured with
a shaping time of 125 ns and low preamplifier gain.

NT , as a function of the shaping time, τ , can be derived from the following

equation [42]:
r
Aw
NT = (CD + CIN ) + B1/f = αCD + β (4.1)
τ
CD is the load capacitance while CIN is the capacitance of the preamplifier
input, Aw is a coefficient of the white-noise while B 1/f is the coefficient
of the 1/f -noise. The measurements performed with the single chip test
board and with different load capacitances are well fitted by Eq. (4.1), giving
α = 24.8 e− pF−1 and β = 200 e− . Since the nominal capacitance per cm
for this type of sensors is ∼1.3 pF cm −1 , the expected total noise for a pure
capacitive load value of 1.3 · 12 = 15.6 pF is N T = 24.8 · 15.6 + 200 ' 600 e− .
The discrepancy between this value and the total noise measured with the
sensor connected is due to the fact that, since a sensor acts also like an
antenna, the non-fully-optimized ground connections and shielding allow
for noise pickup.
I performed an absolute calibration with a radioactive source of Am 241
having an intensity of 1 mCi, whose dominant emission is γ at the energy of
59.6 KeV, corresponding to ∼16550 e − in Silicon. In order to measure the
spectrum one must set the flash-ADC thresholds in a way to define equal-
size bins. This was done calibrating each of the eight thresholds of the flash-
ADC; the measurements and the linear fits are reported in Fig. 4.28. I set the
bin size to 4 mV and the 0 threshold to 43 mV, which, in the hypothesis of
4.3 Detector performance 105

χ2 / ndf 10.72 / 1 χ2 / ndf 27.85 / 1 χ2 / ndf 31.7 / 1

100 100 100

Volt on Inject Cap. [mV]

Volt on Inject Cap. [mV]

Volt on Inject Cap. [mV]
offset 2.82 ± 0.40 offset 5.317 ± 0.453 offset 4.936 ± 0.467
slope 0.9621 ± 0.0063 slope 0.9666 ± 0.0067 slope 0.9539 ± 0.0065
90 90 90
80 80 80
70 70 70
60 60 60

50 50 50

40 40 40

3030 40 50 60 70 80 90 100 30 30
30 40 50 60 70 80 90 100 30 40 50 60 70 80 90 100
Comparator th [DAC] Comparator th1 [DAC] Comparator th2 [DAC]
0
χ2 / ndf 18 / 1 χ2 / ndf 0.004442 / 1 χ2 / ndf 6.174 / 1
110 66 120

Volt on Inject Cap. [mV]

Volt on Inject Cap. [mV]

Volt on Inject Cap. [mV]

offset 9.398 ± 0.506 offset 1.578 ± 8.396 offset 9.043 ± 0.583
slope 0.9395 ± 0.0064 slope 0.9839 ± 0.1340 slope 0.9173 ± 0.0069
100 65 110
100
90 64
90
80 63
80
70 62
70
60 61 60
5050 55 60 65 70 75 80 85 90 95 100 6060 61 62 63 64 65 66 50
50 60 70 80 90 100 110 120
Comparator th [DAC] Comparator th4 [DAC] Comparator th [DAC]
3 5
χ2 / ndf 1.872 / 1 χ2 / ndf 0.5373 / 1
130 130
Volt on Inject Cap. [mV]

Volt on Inject Cap. [mV]

offset 12.41 ± 0.64 offset 12.92 ± 0.68

slope 0.8967 ± 0.0071 slope 0.899 ± 0.007
120 120
110 110

100 100

90 90

80 80

70 70

60 60
60 70 80 90 100 110 120 60 70 80 90 100 110 120
Comparator th [DAC] Comparator th7 [DAC]
6

Figure 4.28: Calibrations of the flash-ADC thresholds. From the upper left
to the lower middle: threshold 0 to threshold 7.

χ2 / ndf 20.62 / 1
8817 ± 75.6
Entries

Entries

9000 Constant
9000 Mean 62.54 ± 0.07
8000 Sigma 6.878 ± 0.098
8000
7000
7000
6000
6000
5000
5000
4000 4000

3000 3000

2000 2000
1000 1000
0 0
35 40 45 50 55 60 65 70 75 80 35 40 45 50 55 60 65 70 75 80
Thresholds [mV] Thresholds [mV]

Figure 4.29: Spectrum of the Am241 expressed in mV units. The measure-

ment was performed with a shaping time of 125 ns and low preamplifier gain.
I set to ∼4 mV the bin spacing, starting from ∼43 mV. On the left: raw
data. On the right: after weighting the bin contents.

40 fF inject capacitance, should cover a range from ∼10750 e − to ∼17750 e− .

Since the threshold value, set in the DAC, must be an integer between 0 and
106 Micro-strip detector prototype

255 and the slope of the calibrations is ∼1 mV DAC −1 , the error on the bin
spacing is at most 2·0.5/4 = 25%. Therefore, after measuring the Americium
spectrum, I recalibrated the thresholds and I weighted the bin contents of
the spectrum with respect to the measured actual bin spacing. The Am 241
spectra before and after bin weighting are reported in Fig. 4.29.
χ2 / ndf 14.86 / 3
Constant 931.2 ± 24.9
Entries

Entries
6 900 Mean 62.51 ± 0.10
Sigma 4.734 ± 0.084
800
5
700
4 600

3 500
400
2
300
1 200
100
0
0
35 40 45 50 55 60 65 70 75 80 35 40 45 50 55 60 65 70 75 80
Thresholds [mV] Thresholds [mV]

Figure 4.30: Noise and internal pulser spectra. The measurements were
performed with a shaping time of 125 ns and low preamplifier gain. I set to
∼4 mV the bin spacing, starting from ∼43 mV. On the left: spectrum of the
noise. On the right: spectrum of the internal pulser whose amplitude was
set to the same energy as the Am241 γ-emission.

I measured the noise spectrum for 20 minutes, the same acquisition time
as for the Americium. As one can see from Fig. 4.30 on the left, the number
of noise hits is negligible. I also compared the Americium peak with the
spectrum of the internal pulser whose amplitude was set to the same energy
as the Am241 γ-emission. In Fig. 4.30 on the right is presented a typical
spectrum of the internal pulser. The σ of the Gaussian fit is ∼1175 e − ,
which is the approximately the noise measured with the threshold curve of
Fig. 4.27. The mean values of the Gaussian fit to the Americium spectrum
and to the internal pulser spectrum are identical with a precision of one
tenth of millivolt. The absolute calibration, performed with the γ-source, is
completely consistent with the calibration performed with the internal pulse
generator, considering a nominal value for the inject capacitance of 40 fF.
Indeed, even if the mismatch between actual and nominal value of integrated
capacitances is ∼20%, the relative deviation among capacitances on the same
chip is just ∼1%. Thus, since the relation between the input, V in , and the
4.4 Conclusions 107

χ2 / ndf 20.62 / 1
Constant 8817 ± 75.6

Entries
9000
Mean 59.6 ± 0.1
8000 Sigma 6.554 ± 0.093

7000
6000
5000
4000
3000
2000
1000
0
35 40 45 50 55 60 65 70 75 80
Energy [KeV]

Figure 4.31: Spectrum of the Am241 expressed in KeV units. The measure-
ment was performed with a shaping time of 125 ns and low preamplifier
gain.

output, Vout , of the preamplifier is Vout = Vin ·CIN J /CF (CIN J and CF being
the inject and feedback capacitances respectively) the expected deviation on
Vout is ∼1%. Figure 4.31 shows the Am241 spectrum expressed in KeV units.

4.4 Conclusions
The measurements performed on the FSSR demonstrate that the chip fulfills
the design requirements and certify its the remarkable qualities in terms of
noise, ∼200 e− without sensor and ∼1150 e− with sensor, and threshold
dispersion, ∼400 e− . Silicon micro-strip detectors equipped with FSSR chips
are well suited to be employed in all the HEP experiments that require data-
driven sparsified readout, allowing the use of tracker information as soon as
possible for trigger decision at the lowest level.
108 Micro-strip detector prototype
Conclusions

The aim of the present thesis was the design, construction and characteriza-
tion of a new Silicon micro-strip tracker to be employed in the forward re-
gion of the new generation of HEP experiments at hadron colliders. Among
other important characteristics, these detectors should in particular feature
a data-driven readout electronics in such a way that their information could
be available as soon as possible for making a trigger decision at the lowest
level. Our goals have been successfully reached.
In particular, my measurements on radiation tolerance of the sensors
have certified that they can easily resist to doses well in excess of 10 14 1 MeV
equivalent neutrons per cm2 , even in presence of highly non-uniform irra-
diation. To perform these studies I developed a method that allows for a
full, point to point, characterization of the sensors using an infrared laser
source. With this method one can even derive the radiation dose absorbed
by a small portion of the sensor. The two-dimensional dose profile I mea-
sured on the prototype is completely consistent with that provided by the
irradiation facility personnel. This fact constitutes a strong cross-check for
my measurements and guarantees the reliability of my method.
A custom chip was developed to perform a low-noise, data-driven and
sparsified readout of the detectors. During the design phase of the chip, I
simulated, in Verilog language, the performance of its digital section. At a
luminosity of 2×1032 cm−2 s−1 and energy of 2 TeV the readout efficiency
is remarkably high, > 99%.
I successfully assembled and tested some detector prototypes. The mea-
sured performance fulfills the design requirements. In particular the noise
is ∼1150 e− , corresponding to a signal-noise ratio of ∼20 for a minimum
110 Conclusions

ionizing particle, and the threshold dispersion is ∼400 e − .

The success of this work is also demonstrated by the interest that sev-
eral people of our community have already expressed for the detectors and
their readout electronics. The CBM [44] and PHENIX [45] collaborations
are already considering the possibility to use this detector for their Silicon
trackers.
Bibliography

[1] LEP design report: CERN-LEP/84-01.

[2] Fermilab accelerator division web site: http://www-bd.fnal.gov

[3] LHC design report: http://ab-div.web.cern.ch/ab-div/Publications/LHC-

DesignReport.html

[4] H.B. Prosper and M. Danilov, “Techniques and Concepts of High-

Energy Physics”, ed. by Springer.

[5] F. Abe et al., “Measurement of the B Meson Differential Cross-Section

√
in pp̄ collisions at s = 1.8 TeV”, CDF / PUB / BOTTOM / PUBLIC
/ 3759 submitted to ICHEP ’96 and references therein; F. Abe et al.,
Phys. Rev. Lett. 75, 1451 (1995).

[6] R. Abbott et al., “The bb̄ Production Cross Section and Angular Corre-
√
lations in pp̄ collisions at s = 1.8 TeV”, FERMILAB-Pub-99/144-E;
S. Abachi et al., Phys. Rev. Lett. 74, 3548 (1995).

[7] D. Fein, “Tevatron Results on b-Quark Cross Sections and Correla-

tions”, presented at Hadron Collider Physics (Bombay), January 1999.

[8] BTeV web site: http://www-btev.fnal.gov

[9] BTeV Technical Design Report, Chapter 11, “The BTeV Trigger”.

[10] BTeV Technical Design Report, Chapter 12, “Event Read out and Con-
trol System”.

[11] A.E. Snyder and H.R. Quinn, Phys. Rev. D 48, 2139 (1993).
112 BIBLIOGRAPHY

[12] Z. Zhao et al., “Report of Snowmass 2001 Working Group E2: Electron-
Positron Colliders from the φ to the Z”, to appear in the proceeding
(hep-es/0201047).

[13] G. Lutz, “Semiconductor Radiation Detectors”, ed. by Springer, Ger-

many 1999.

[14] ROSE collaboration, Nucl. Instrum. Methods A 466, 308 (2001).

[15] ROSE collaboration, Nucl. Instrum. Methods A 491, 194 (2002).

[16] ROSE collaboration, Nucl. Instrum. Methods A 439, 282 (2000).

[17] ROSE/TN/2000-02 June 2000 on http://cern.ch/rd48

[18] More information about Float Zone Silicon can be found at the
web site: http://w4.siemens.de/archiv/en/innovationen/iuk/zonenziehver
fahren.html

[19] E. Fretwurst et al., Nucl. Instrum. Methods A 342, 119 (1994). G. Lind-
ström et al., Nucl. Instrum. Methods A 426, 1 (1999).

[20] CMS tracker sensors web site: http://cms.pi.infn.it/roby/designbarrel/

webpage.html

[21] Ideas readout chip web site: http://www.ideas.no/products/ASICs/

pdf/TAA1.pdf

[22] Ideas DAQ web site: www.hep.princeton.edu/∼tziegler/svd/vadaq/VA

DAQ UserManual.pdf

[23] Indiana University Cyclotron Facility home page: http://www.iucf. in-

diana.edu

[24] ROSE collaboration, Nucl. Instrum. Methods A 479, 487 (2002).

[25] Document Fermilab-TM-1849 on http://lss.fnal.gov/ird/index.html

[26] I. Abt, S. Masciocchi, B. Moshous, T. Perschke, R.H. Richter, K. Riech-

mann and W. Wagner, Nucl. Instrum. Methods A 423, 303 (1999).

[27] Opto Electronics home page: http://www.tempo.textron.com/

BIBLIOGRAPHY 113

[28] G. Kramberger, V. Cindro, I. Mandić, M. Mikuž, M. Zavrtanik, IEEE

Trans. Nucl. Sci. 49(4), 1717 (2002).

[29] S.M. Sze, “Physics of semiconductor devices”, 2nd ed. by Willy, Singa-
pore 1981.

[30] V. Eremin and Z. Li, Nucl. Instrum. Methods A 362, 338 (1995).

[31] T.J. Brodbeck, A. Chilingarov, T. Sloan, E. Fretwurst, M. Kuhnke and

G. Lindstroem, Nucl. Instrum. Methods A 477, 287 (2002).

[32] G. Kramberger, V. Cindro, I. Mandić, M. Mikuž and M. Zavrtanik,

Nucl. Instrum. Methods A 476, 645 (2002).

[33] ROSE/TN/2001-01 January 2001 on http://cern.ch/rd48

[34] G. Casse, M. Glaser and E. Grigoriev, Nucl. Instrum. Methods A 426,

140 (1999).

[35] “Type inversion in irradiated Silicon: half truth”, arXiv: physics /

0409049 v2 13 Sep 2004

[36] “Simulations of heavy irradiated Silicon pixel sensors and comparison

with test beam measurements”, arXiv: physics / 0411143 v1 16 Nov
2004

[37] “Electric field measurements in heavy irradiated pixel sensors”, arXiv:

physics / 0412036 v1 6 Dec 2004

[38] V. Eremin, E. Verbidskaya and Z. Li, Nucl. Instrum. Methods A 476,

556 (2002).

[39] S. Gadomski, M. Turala, E. Barberis, N. Cartiglia, J. Leslie and H.F.-

W. Sadrozinski, Nucl. Instrum. Methods A 326, 239 (1993).

[40] R. Yarema, J. Hoff, A. Mekkaoui, M. Manghisoni, V. Re et al.,

FERMILAB-PUB-04-260-E, presented at the IEEE Nuclear Science
Symposium, Rome, Italy, 2004.

[41] G. Anelli, M. Campbell, M. Delmastro, F. Faccio, S. Florian et al.,

IEEE Trans. Nucl. Sci. 46(6), 1690 (1997).
114 BIBLIOGRAPHY

[42] V. Re, L. Ratti, M. Manghisoni, G. Speziali and G. Traversi, “Sur-

vey of Noise Performances and Scaling Effects in Deep Submicrons
CMOS Devices from Different Foundries”, presented at the IEEE
Nuclear Science Symposium, Rome, Italy, 2004. Downloadable from:
http://ieeexplore.ieee.org/Xplore/dynhome.jsp

[43] J. Hoff, A. Mekkaoui, D. Christian, S. Zimmerman, G. Cancelo,

P. Kasper and R. Yarema, IEEE Trans. Nucl. Sci. 48(3), 485 (2001).

[44] CBM web site: http://www.gsi.de/fair/experiments/CBM/index e.html

[45] Phenix web site: http://www.phenix.bnl.gov/

[46] N. Cabibbo, Phys. Rev. Lett. 10, 531 (1963); M. Kobayashi and
K. Maskawa, Prog. Theor. Phys. 49, 652 (1973).

[47] L. Wolfenstein, Phys. Rev. Lett. 51, 1945 (1983).

[48] S. Stone, “Prospects For B-Physics In The Next Decade”, in Techniques

and Concepts of High-Energy Physics IX, ed. by T. Ferbel, NATO ASI
Series, Plenum, NY (1996).

[49] C. Dib, I. Dunietz,F.J. Gilman and Y. Nir, Phys. Rev. D 41, 1522
(1990).

[50] H. Harari and Y. Nir, Phys. Lett. B 195, 586 (1987).

[51] CKM fitter web site: http://www.slac.stanford.edu/xorg/ckmfitter/ckm

welcome.html

[52] P.F. Harrison and H.R. Quinn, “The BaBar physics book: Physics at
an asymmetric B factory”, Papers from Workshop on Physics at an
Asymmetric B Factory (BaBar Collaboration Meeting), Rome, Italy,
11-14 Nov. 1996, Princeton, NJ, 17-20 Mar. 1997, Orsay, France, 16-19
Jun. 1997 and Pasadena, CA, 22-24 Sep. 1997.

[53] Lepton Photon conference Upsala 2005 web site: http://lp2005.tsl.

uu.se/∼lp2005/
BIBLIOGRAPHY 115

[54] G.C. Branco, L. Lavoura and J.P. Silva, “CP violation”, ed. by Oxford,
UK Clarendon (1999).

[55] “Review of Particle Physics”, http://pdg.lbl.gov

[56] BABAR collaboration, 18 Jul 2005, arXiv:hep-ex/0507074 v1.

[57] BELLE collaboration, 8 Jul 2005, arXiv:hep-ex/0507037 v1.

[58] Document Fermilab-Conf-00/044-E CDF on http://lss.fnal.gov/ird/ in-

dex.html

[59] Y. Grossman, G. Isidori and M. Worah, Phys. Rev. D 58, 057504 (1998).

[60] S. J. Richichi et al., (CLEO) “Two-Body B Meson Decays to η and η 0

- Observation of B → ηK ∗ ”, Conf. 99-12 (hep-ex/9908019) (1999).

[61] S. Kopp, “Studies of B o Decays for Measuring sin(2β)”, presented at

DPF’99, to appear in proceedings (hep-ex/9904009).

[62] M. Gronau Phys. Rev. Lett. 63, 1451 (1989). M. Gronau and D. London
Phys. Rev. Lett. 65, 3381 (1990).

[63] D. Du, I. Dunietz and Dan-di Wu, Phys. Rev. D 34, 3414 (1986).

[64] R. Aleksan, I. Dunietz and B. Kayser, Z. Phys. C 54, 653 (1992).

[65] R. Aleksan, A. Le Yaouanc, L. Oliver, O. Pène and J.C. Raynal, Z.

Phys. C 67, 251 (1995).

[66] D. Atwood, I. Dunietz and A. Soni, Phys. Rev. Lett. 78, 3257 (1997).

[67] M. Gronau and D. Wyler, Phys. Lett. B 265, 172 (1991).

[68] I. Dunietz, Phys. Lett. B 427, 179 (1998).

[69] Y. Grossman and H.R. Quinn, “Removing Discrete Ambiguities in CP

Asymmetry Measurments”, Phys. Rev. D 56, 7259 (1997).

[70] Another method of measuring |V td | is to measure the branching ratio of

K + → π + ν ν̄. A precise measurement would still be subject to theoret-
ical uncertainties mostly arising from the uncertainty in the charmed
116 BIBLIOGRAPHY

quark mass and |Vcb |. See G. Buchalla, A.J. Buras and M.E. Lauten-
bacher, Rev. Mod. Phys., 68, 1125 (1996). A Brookhaven experiment,
E787, has claimed to see two events and hopes to obtain substantially
more data. See S. Adler et al., E787, Phys. Rev. Lett. 88, 041803 (2002).

[71] K. Lingel, T. Skwarnicki and J.G. Smith, Ann. Rev. Nucl. Part. Sci.
48 169 (1998).

[72] M. Bander, D. Silverman and A. Soni, Phys. Rev. Lett. 43, 242 (1979).
Appendix A

CP violation

A.1 Theoretical account for CP violation

In the Standard Model (SM) of fundamental interactions there are no ex-

plicit mass-terms in the Lagrangian; indeed, had the vector bosons explicit
mass-terms, the theory wouldn’t be gauge-invariant. Particles acquire mass
dynamically, interacting with a particular field called the Higgs field. In
particular the rest-mass of a particle is the result of its interaction with the
Higgs condensate.
The symmetries under which the SM Lagrangian of the electroweak sec-
tor is conceived are (Weinberg-Salam-Glashow model): the Lorentz sym-
metry and the local gauge symmetries SU (2) L ⊗ U (1)Y . The symmetry
U (1)Y corresponds to the invariance to a scalar local phase change of the
fields, while the symmetry SU (2)L deals only with left-handed fields and
corresponds to the invariance under a local phase rotation of the doublets
Ψi = Ψ


ui
Ψdi , the index i = 1, 2, 3 is the flavour (or generation) index. When
L
the Higgs field acquires its Vacuum Expectation Value (VEV) by choosing
a ground state, the gauge group is Spontaneously Broken (SSB), in such a
way that the only symmetry that survives is U (1) EM of electromagnetism
SSB
(SU (2)L ⊗ U (1)Y −→ U (1)EM ); this is done in order to have the corre-
sponding gauge particle (the photon) massless. To realize this aim one has
to introduce a set of scalar fields transforming in a convenient way under
SU (2)L ⊗ U (1)Y . The simplest choice turns out to be a complex represen-
118 CP violation

tation of SU (2) of dimension 2: Higgs doublet. Since the vacuum must

be electrically neutral, one of the components of the doublet must also be
neutral. One can write the Higgs doublet and its conjugate as:

 ¯o 
φ+
 
φ
Φ= Φ̄ = (A.1)
φo −φ−

The three generation of quarks, in the Weinberg-Salam-Glashow, are de-

scribed in Tab. A.1. The Yukawa Lagrangian, describing the interaction

u
c
t

d L s L b L
uR , d R cR , s R tR , b R

Table A.1: The three families of quarks.

between quarks and the Higgs field for the three generation of quarks, is:

3 h i
LqY uk = −
X
GU R̄
ij Ui ( Φ̄ †
L j ) + G D
R̄
ij Di (Φ †
L j ) + h.c. (A.2)
i,j=1

u0 c0 t0
RUi = u0R , c0R , t0R , RDi = d0R , s0R , b0R and Lj =




d0 L , s 0 L , b 0 L . From the
vacuum expectation values of Φ and Φ̄ one can obtain the mass terms for
the up quarks:
 
u0
(u0 , c0 , t0 )R M U 
 
0 
 c  + h.c. (A.3)
t0
L

and for the down quarks:

 
d0
(d0 , s0 , b0 )R M D 
 
0 
 s  + h.c. (A.4)
b0
L

U (D) √ U (D)
Mij = (v/ 2)Gij are non-diagonal matrices. The weak eigenstates,
q 0 , are linear superposition of the mass eigenstates, q, given by the unitarity
A.1 Theoretical account for CP violation 119

transformations:
   
u0 u
   
 c0  = UL,R 
 c 

 
t0 t
L,R L,R
   
d0 d
   
 s0  = DL,R  s 
 (A.5)
  
b 0 b
L,R L,R

U (D)L,R are unitary matrices to preserve the form of the kinetic terms of
the quarks. These matrices diagonalize the mass matrices:
 
mu 0 0
UR−1 M U UL = 
 
 0 mc 0 

0 0 mt
 
md 0 0
−1
M D DL = 
 
DR  0 m s 0 
 (A.6)
0 0 mb

The V-A charged week current will be proportional to:

   
d0 d
0 0 0

0
 †  
(u , c , t )L γµ  s  = (u, c, t) L (UL DL )γµ  s 
  
 (A.7)
b0 b
L L

with the generation mixing of the mass eigenstates, q, described by:

V ≡ UL† DL (A.8)

On the other hand the neutral current will be proportional to:

   
u0 u
  †  
(u0 , c0 , t0 )L γµ  0 
 c  = (u, c, t)L (UL UL )γµ  c 
  (A.9)
t0 t
L L

We can notice that there is no mixing in the neutral sector since the matrix
UL is unitary: UL† UL = 1.
While neutrinos were initially assumed to be massless, recent observa-
tions of neutrino mixing have revealed the existence of neutrino mass leading
to them having an analogous mixing matrix as the quarks.
120 CP violation

The quark mixing, by convention, is restricted to the down quarks:

   
d0 d
   
 s0  = V  s  (A.10)
   
b 0 b
L L

V is the Cabibbo-Kobayashi-Maskawa (CKM) matrix [46], having one free

imaginary phase, being the source of the CP violation, and three free real
angles. Indeed, let’s take the example of the kaon sector. The two mass
eigenstates KS and KL are a superposition of the two flavour eigenstates
K o and K̄ o : KS = K o + K̄ o and KL = K o − K̄ o , with  in general
a complex parameter. If KS and KL were CP eigenstates the following
relations would be satisfied only for  = 1: CP |K S i = |KS i and CP |KL i =
− |KL i.  depends on the diagonal terms of the Hamiltonian that describes
the evolution of the K o −K̄ o system, and since the Hamiltonian itself directly
depends on the CKM matrix,  can’t be equal to 1 because of the presence
of the CKM phase.
Let’s summarize the concepts expressed until now:

• in the week sector of the SM there is a single source of CP violation

coming from the free phase of the CKM matrix

• in the SM, CP violation originates from the flavour structure of the

charged current interactions of quarks

• the CP-violating CKM phase could be eliminated through an appro-

priate unitary transformation of quark fields if any two quarks with
the same charge had the same mass, CP violation is therefore closely
related to the “flavour problem” of mass hierarchy among the three
generations

A.2 The CKM matrix

A.2.1 Introduction

The physical point-like states of nature that have both strong and elec-
troweak interactions, the quarks, are mixtures of base states described by
A.2 The CKM matrix 121

the CKM matrix:

    
d0 Vud Vus Vub d
    
 s0  =  Vcd Vcs Vcb  (A.11)
 s 
 
  
b0 Vtd Vts Vtb b

The Vij ’s are complex numbers that can be represented by four independent
real quantities. These numbers are fundamental constants of nature that
need to be determined from experiment, like any other fundamental constant
such as α or G. In the Wolfenstein approximation the matrix is written
as [47]:
 
1 − λ2 /2 λ Aλ3 (ρ − iη(1 − λ2 /2))
 
VCKM = 
 −λ 1 − λ2 /2 − iηA2 λ4 Aλ2 (1 + iηλ2 ) 

Aλ3 (1 − ρ − iη) −Aλ2 1
(A.12)
This expression is accurate to order λ3 in the real part and λ5 in the imag-
inary part. It is necessary to express the matrix to this order to have a
complete formulation of the physics we wish to pursue. The constants λ
and A have been measured using semileptonic s and b decays [48]; λ ∼ 0.22
and A ∼ 0.8. The η parameter allows for CP violation.

A.2.2 Unitarity triangles

The unitarity of the CKM matrix1 allows us to construct six relationships.

These equations may be thought of as triangles in the complex plane. They
are shown in Fig. A.1. In the bd triangle, the one usually considered, the
angles are all thought to be relatively large. It is described by:

∗ ∗
Vub Vud + Vcb Vcd + Vtb Vtd∗ = 0 (A.13)

To a good approximation:

∗
|Vud | ' |Vtb | ' 1 (A.14)

which implies:
Vub Vtd∗ ∗
+ + Vcd =0 (A.15)
Vcb Vcb
1
Unitarity implies that any pair of rows or columns are orthogonal.
122 CP violation

 "$# 

  
    
(  
 
(


  
      

 

#&%    %'" !
 

 
 ) *
   

     
   
 
+

Figure A.1: The six CKM triangles. The bold labels, i.e ds, refer to the rows
or columns used in the unitarity relationship.

∗ = −λ, we can define a triangle with sides:

Since Vcd

1 (A.16)
Vtd∗ 1 Vtd∗
q
= (1 − ρ)2 + η 2 = (A.17)
Aλ3 λ Vts
Vub p 1 Vub
= ρ2 + η 2 = (A.18)
Aλ3 λ Vcb
This CKM triangle is depicted in Fig. A.2. The rescaled unitarity triangle
η̄
(ρ̄, η̄) 2
ρ̄ = ρ(1 − λ2 )
α 2
η̄ = η(1 − λ2 )
1−
iη̄

ρ̄ −
ρ̄ +

iη̄

γ 1 β
(0, 0) (1, 0) ρ̄

Figure A.2: Unitarity triangle.

is derived from Eq. (A.13) by:

∗ ) is real
1. choosing a phase convention such that (V cb Vcd
∗|
2. dividing the lengths of all sides by |V cb Vcd

3. aligns one side of the triangle with the real axis

A.2 The CKM matrix 123

4. makes the length of this side 1

The form of the triangle is unchanged. Two vertices of the rescaled unitarity
triangle are thus fixed at (0, 0) and (1, 0). The coordinates of the remain-
ing vertex correspond to the Wolfenstein parameters (ρ̄, η̄). The rescaled
triangle is illustrated in Fig. A.2.
The three angles of the unitarity triangle are denoted by α, β and γ [49],
together with the χ0 and χ angles of the ds and sb triangles, have the
following expressions.
 ∗ 
Vcb Vcd Vtb Vtd∗
   ∗ 
Vub Vud
α ≡ arg − ∗ , β ≡ arg − ∗ , γ ≡ arg −
Vub Vud Vcb Vcd Vtb Vtd∗
 (A.19)
Vts Vtb∗ Vcd Vcs∗
  
0
χ ≡ arg − , χ ≡ arg −
Vcs Vcb∗ Vud Vus∗

To make predictions for future measurements of CP violating observables,

we need to find the allowed ranges for the CKM phases. There are three
ways to determine the CKM parameters (see e.g. [50]):

direct measurements: related to SM tree level processes. At present, we

have direct measurements of |Vud |, |Vus |, |Vub |, |Vcd |, |Vcs |, |Vcb | and
|Vtb |
†
CKM Unitarity: (VCKM VCKM = 1) relates the various matrix elements.
At present, these relations are useful to constrain |V td |, |Vts |, |Vtb | and
|Vcs |

indirect measurements: related to SM loop processes. At present, we

constrain in this way |Vtb Vtd | (from ∆mB and ∆mBs ) and η (from εK )

When all recent data are taken into account, the following values of CKM
parameters and unitarity triangle angles arise from a global fit [51].

λ = 0.2262+0.0010
−0.0010 , A = 0.825+0.011
−0.019 , (A.20)
ρ̄ = 0.207+0.036
−0.043 , η̄ = 0.340+0.023
−0.023 , (A.21)
sin(2β) = 0.724+0.018
−0.018 , sin(2α) = −0.28+0.24
−0.20 , γ = 58.6+6.8
−5.9 (A.22)

The full information can be described by allowed regions in the (ρ̄, η̄) as
illustrated in Fig. A.3 or the (sin(2α), sin(2β)) planes (see e.g. [52]). We
124 CP violation

1.5
excluded area has CL > 0.95

ex c
lude
γ ∆md

da
t CL
1

> 0.
sin 2β

95
∆ms & ∆md
0.5

εK α α
γ β
0
η

|Vub/Vcb|

-0.5

εK
α
sol. w/ cos 2β < 0
-1 (excl. at CL > 0.95)

CKM
fitter
γ
EPS 2005

-1.5
-1 -0.5 0 0.5 1 1.5 2

Figure A.3: Up to date constraints to CKM parameters (figure from [51]).

now have measurements not only of β but also of γ and α. The latter from
B o → ρ+ ρ− and the former from Dalitz analysis of B − → D o K − , where
D o → Ks π + π − . See talks ot the Lepton Photon conference Upsala 2005 [53].
Equations (A.21) and (A.22) show yet another important feature of CP
violation in the SM. The fact that η/ρ = O(1) or, equivalently, sin(γ) =
O(1), implies that CP is not an approximate symmetry within the SM. This
is not an obvious fact: after all, the two measured CP violating quantities in
the kaon sector, εK and ε0K , are very small, of the order of 10−3 and 10−6 ,
respectively. While the SM predicts that in some b physics processes the CP
asymmetry is of order one.
A.3 CP violation in meson decays 125

A.3 CP violation in meson decays

In the previous sections have been shown how CP violation arises in the SM.
Now the implications of this theory, for the phenomenology of CP violation
in meson decays in particular for B-meson decays, will be illustrated. There
are three different types of CP violation in meson decays:

mixing: which occurs when the two neutral mass eigenstate admixtures
cannot be chosen to be CP-eigenstates

decay (or direct): which occurs in both charged and neutral decays, when
the amplitude for a decay and its CP-conjugate process have different
magnitudes

interference of decays with and without mixing: which occurs in de-

cays into final states that are common to B o and B¯o

In each case it is useful to identify a particular CP-violating quantity that

is independent of phase conventions and discuss the types of processes that
depend on this quantity. This will be done in the next section.

A.3.1 Notations and formalism

To define the three types of CP violation in meson decays and to discuss their
theoretical calculation and experimental measurement, I will first introduce
some notations and formalism. I will refer specifically to B-meson mixing
and decays. The phase convention for the CP transformation law of the
neutral B-mesons is defined by:

CP |B o i = wB B̄ o , CP B̄ o = wB
∗
|B o i , (|wB | = 1) (A.23)

Physical observables do not depend on the phase factor w B . The time evo-
lution of any linear combination of the neutral B-meson flavour eigenstates:

a |B o i + b B̄ o (A.24)

is governed by the Schrödinger equation:

      
d a a i a
i =H ≡ M− Γ (A.25)
dt b b 2 b
126 CP violation

for which M and Γ are 2×2 Hermitian matrices. CPT invariance guarantees
H11 = H22 , that is M11 = M22 and Γ11 = Γ22 . The off-diagonal terms in
these matrices, M12 and Γ12 , are particularly important in the discussion
of mixing and CP violation. M12 is the dispersive part of the transition
amplitude from B o to B̄ o , while Γ12 is the absorptive part of that amplitude.
In the SM these contributions arise from the box diagrams with two W
exchanges (see Fig. A.6).
The light BL and heavy BH mass eigenstates are given by:

|BL,H i = p |B o i ± q B̄ o (A.26)

and their time evolution is given by:

|BH (t)i = e−iMH t e−ΓH t/2 |BH i
(A.27)
|BL (t)i = e−iML t e−ΓL t/2 |BL i
The complex coefficients q and p obey the normalization condition |q| 2 +
|p|2 = 1. The mass difference and the width difference between the physical
states are defined as follows:

∆m ≡ MH − ML , ∆Γ ≡ ΓH − ΓL (A.28)

Solving the eigenvalue equation gives:

1
(∆m)2 − (∆Γ)2 = (4|M12 |2 − |Γ12 |2 ), ∆m∆Γ = 4<(M12 Γ∗12 ) (A.29)
4
q ∆m − 2i ∆Γ ∗ − iΓ∗
2M12 12
=− =− (A.30)
p 2M12 − iΓ12 ∆m − 2i ∆Γ
The time evolution of an initially pure B o or B̄ o state is:
o q
Bphys (t) = f+ (t) |B o i + f− (t) B̄ o
p
p (A.31)
o o
B̄phys (t) = f+ (t) B̄ + f− (t) |B o i
q
where:
1 h −i(ML − i ΓL )t i
i
f± (t) = e 2 ± e−i(MH − 2 ΓH )t (A.32)
2
In the B system, |Γ12 |  |M12 | (see discussion below), and then, to leading
order in |Γ12 /M12 | Eqs. (A.29) and (A.30) can be written as:

∆mB = 2|M12 |, ∆ΓB = 2<(M12 Γ∗12 )/|M12 | (A.33)

q ∗
M12
=− (A.34)
p |M12 |
A.3 CP violation in meson decays 127

To discuss CP violation in mixing, it is useful to write Eq. (A.30) to first

order in |Γ12 /M12 |:
∗
  
q M12 1 Γ12
=− 1− = (A.35)
p |M12 | 2 M12

To discuss CP violation in decay, we need to consider decay amplitudes.

The CP transformation law for a final state f is:

CP |f i = wf f¯ , CP f¯ = wf∗ |f i , (|wf | = 1) (A.36)

For a final CP eigenstate f = f¯ = fCP , the phase factor wf is replaced

by ηfCP = ±1, the CP eigenvalue of the final state. I define the decay
amplitudes Af and Āf¯ according to:

Af = hf Hd B o i , Āf¯ = f¯ Hd B̄ o (A.37)

Hd is the decay Hamiltonian. CP relates A f and Āf¯. There are two types
of phases that may appear in Af and Āf¯. Complex parameters in any
Lagrangian term that contributes to the amplitude will appear in complex
conjugate form in the CP-conjugate amplitude. Thus their phases appear
in Af and Āf¯ with opposite signs. In the SM these phases only occur in the
CKM matrix which is part of the electroweak sector of the theory, hence
these are often called “weak phases.” The weak phase of any single term
is convention dependent. However the difference between the weak phases
in two different terms in Af is convention independent because the phase
rotations of the initial and final states are the same for every term. A second
type of phase can appear in scattering or decay amplitudes even when the
Lagrangian is real. Such phases do not violate CP and they appear in A f
and Āf¯ with the same sign. Their origin is the possible contribution from
intermediate on-shell states in the decay process, that is the absorptive part
of an amplitude that has contributions from coupled channels. Usually the
dominant rescattering is due to strong interactions and hence the designation
“strong phases” for the phase shifts so induced. Again only the relative
strong phases of different terms in a scattering amplitude have physical
content, an overall phase rotation of the entire amplitude has no physical
consequences. Thus it is useful to write each contribution to A in three
128 CP violation

parts: its magnitude Ai , its weak phase term eiφi and its strong phase term
eiδi . Then, if several amplitudes contribute to B → f we have:

Āf¯ Ai ei(δi −φi )

P
= Pi i(δi +φi )
(A.38)
Af i Ai e

To discuss CP violation in the interference of decays with and without

mixing, we introduce a complex quantity λ f defined by:

q Āf¯
λf = (A.39)
p Af

We further define the CP transformation law for the quark fields in the
Hamiltonian (a careful treatment of CP conventions can be found in [54]):

q → wq q̄, q̄ → wq∗ q, (|wq | = 1) (A.40)

The effective Hamiltonian that is relevant to M 12 is of the form:

∆b=2 ¯ (1 − γ5 )b 2 + e−2iφB b̄γ µ (1 − γ5 )d 2

∝ e+2iφB dγ
 µ   
Hef f (A.41)

2φB is a CP violating (weak) phase. For the B system, where |Γ 12 |  |M12 |,

this leads to:
q/p = wB wb∗ wd e−2iφB (A.42)

To understand the phase structure of decay amplitudes, we take as an ex-

ample the b → q q̄d decay (q = u or c). The decay Hamiltonian is of the
form:

Hd ∝e+iφf [q̄γ µ (1 − γ5 )d] b̄γµ (1 − γ5 )q

 
(A.43)
¯ µ (1 − γ5 )q
+e−iφf [q̄γ µ (1 − γ5 )b] dγ
 

φf is the appropriate weak phase. Then:

∗
Āf¯/Af = wf wB wb wd∗ e−2iφf (A.44)

Eqs. (A.42) and (A.44) together imply that for a final CP eigenstate,

λfCP = ηfCP e−2i(φB +φf ) (A.45)

A.4 The three types of CP violation in meson decays 129

A.4 The three types of CP violation in meson de-

cays

A.4.1 CP violation in mixing

q
6= 1 (A.46)
p

This results from the mass eigenstates being different from the CP eigen-
states. In fact if |q/p| = 1 this term represents a pure phase and due to
the freedom of choosing an arbitrary phase in CP transformation we could
define:
q
CP |B o i = B̄ o (A.47)
p

With the convention of Eq. (A.47) BH and BL become CP eigenstates. But

if Eq. (A.46) holds than there is no legitimate CP transformation, of B o
and B̄ o , that will make the mass eigenstates CP eigenstates. This type of
CP violation requires also a relative phase between M 12 and Γ12 . In fact it
could be demonstrated that BH and BL can be CP eigenstates if and only
if = (M12 Γ∗12 ) = 0, that is when the relative phase between M 12 and Γ12
vanishes. The way this comes about is because there are two ways in which
a B o can mix into a B̄ o : one is the direct way (connected with M 12 , the
dispersive part) and the other is via a common decay mode of the two states
(connected to Γ12 , the absorptive part), as depicted in Fig. A.4. For the

M12
Bo B̄ o

Γ12
f

Figure A.4: CP violation in mixing.

neutral B system, this effect could be observed through the asymmetries in

130 CP violation

semileptonic decays:
   
o + o −
Γ B̄phys (t) → ` νX − Γ Bphys (t) → ` νX
aSL =     (A.48)
o
Γ B̄phys o
(t) → `+ νX + Γ Bphys (t) → `− νX

In terms of q and p:
1 − |q/p|4
aSL = (A.49)
1 + |q/p|4
In the neutral B system, the effect is expected to be small, . O(10 −2 ).
The reason is that one expects that a SL . ∆ΓB /∆mB . The difference in
width is produced by decay channels common to B o and B̄ o . The branching
ratios for such channels are at, or below, the level of 10 −3 . Since various
channels contribute with differing signs, one expects that their sum does not
exceed the individual level. Hence, we can safely assume that ∆Γ B /ΓB =
O(10−2 ), where ΓB = (ΓH + ΓL )/2. On the other hand, it is experimentally
known that x (= ∆mB /ΓB ) for the Bd meson is xd = 0.771 ± 0.012 (while
for the Bs meson is xs > 20.6 at 95% CL) [55]. To calculate a SL , we use
Eqs. (A.49) and (A.35), and get:
Γ12
aSL = =( ) (A.50)
M12
To predict it in a given model, one needs to calculate M 12 and Γ12 . This in-
volves large hadronic uncertainties, in particular in the hadronization models
for Γ12 .

A.4.2 CP violation in decay

|Āf¯/Af | 6= 1 (A.51)

This appears as a result of interference among various terms in the decay

amplitude, and will not occur unless at least two terms have different weak
phases and different strong phases.
Γ(B + → f + ) − Γ(B − → f − )
af ± = (A.52)
Γ(B + → f + ) + Γ(B − → f − )
CP asymmetries in charged B decays are purely an effect of CP violation in
decay. Equation (A.52) in terms of the decay amplitudes:
1 − |Āf − /Af + |2
af ± = (A.53)
1 + |Āf − /Af + |2
A.4 The three types of CP violation in meson decays 131

To calculate af ± we use Eqs. (A.53) and (A.38). For simplicity, we consider

decays with contributions from two weak phases and with A 2  A1 . We
get:
af ± = −2(A2 /A1 ) sin(δ2 − δ1 ) sin(φ2 − φ1 ) (A.54)

The magnitude and strong phase of any amplitude involve long distance
strong interaction physics and our ability to calculate these from first prin-
ciples is limited. Thus quantities that depend only on the weak phases are
cleaner.

A.4.3 CP violation in the interference between decays with

and without mixing

|λfCP | = 1, =(λfCP ) 6= 0 (A.55)

Any λfCP 6= ±1 is a manifestation of CP violation. The special case of

Eq. (A.55) isolates the effects of interest since both CP violation in decay,
Eq. (A.51), and in mixing, Eq. (A.46), lead to |λ fCP | 6= 1. For the neutral
B system this effect can be observed by comparing decays into final CP
eigenstates of a time-evolving neutral B state that begins at time zero as
B o to those of the state that begins as B̄ o :
o
Γ(B̄phys o
(t) → fCP ) − Γ(Bphys (t) → fCP )
afCP = o o (A.56)
Γ(B̄phys (t) → fCP ) + Γ(Bphys (t) → fCP )

This time dependent asymmetry is given by:

1 − |λfCP |2 cos(∆M t) − 2=(λfCP ) sin(∆M t)

afCP = (A.57)
1 + |λfCP |2

and, for |λfCP | = 1, Eq. (A.57) simplifies considerably to:

afCP = −=(λfCP ) sin(∆mB t) (A.58)

It is expected to be an effect of O(1) in various B decays. For such

cases, the contribution from CP violation in mixing is clearly negligible.
For decays that are dominated by a single CP violating phase (for example,
B → J/ψKS and KL → π o ν ν̄), so that the contribution from CP violation
in decay is also negligible, afCP is cleanly interpreted in terms of purely
electroweak parameters. Explicitly, =(λ fCP ) gives the difference between
132 CP violation

the phase of the B − B̄ mixing amplitude, 2φB , and twice the phase of the
relevant decay amplitude, 2φf , (see Eq. (A.45)):

=(λfCP ) = −ηfCP sin(2(φB + φf )) (A.59)

A summary of the main properties of the different types of CP violation

in meson decays is given in Tab. A.2.

Type Exp. Theory Calculation Uncer-

tainties
 
1−|q/p|4 Γ12
mixing aSL 1+|q/p|4
= M12 Γ12 large
1−|Āf − /Af + |2 A2
decay af ± 1+|Āf − /Af + |2
−2 A 1
sin(δ2 − δ1 ) δi , Ai large
sin(φ2 − φ1 )
inter- afCP −=(λfCP ) ηfCP sin(2(φB + φf )) small
ference

Table A.2: The three types of CP violation in meson decays.

A.5 B decays classification

B decay can be classified into four classes. Classes (1) and (2) are expected
to have relatively small CP violation in decay and hence are particularly
interesting for extracting CKM parameters from interference of decays with
and without mixing. In class (3) and (4), CP violation in decay could be
significant and might be observable in charged B decays.

1. decays dominated by a single term (e.g. b → cc̄s and b → ss̄s).

The SM predicts very small CP violation in decay: O(λ 2 ) for b → ss̄s.
Any observation of large CP asymmetries in charged B decays for these
channels would be a clue to physics beyond the SM. The correspond-
ing neutral modes have cleanly predicted relationships between CKM
parameters and the measured asymmetry from interference between
decays with and without mixing

2. decays with a small second term (e.g. b → cc̄d and b → uūd). The
expectation that penguin-only contributions are suppressed compared
A.6 Techniques for determining β 133

to tree contributions suggests that these modes will have small effects
of CP violation in decay, of O(λ2 − λ), and an approximate prediction
of the relationship between measured asymmetries in neutral decays
and CKM phase can be made

3. decays with a suppressed tree contribution (e.g. b → uūs).

The tree amplitude is suppressed by small mixing angles, V ub Vus . The
non-tree term may be comparable or even dominate and give large
interference effects

4. decays with no tree contribution and a small second term (e.g.

b → sūd). Here the interference comes from penguin contributions
with different charge 2/3 quarks in the loop and gives CP violation in
decay that could be as large as 10%

Classes Decays
(1) B → J/ψK
B → φK
(2) B → DD
B → ππ
(3) B → ρK
B → πK
(4) B → KK

Table A.3: Classification of some B decays.

The B decays most relevant to the measurement of the α, β and γ angles

of the unitarity triangle are classified in Tab. A.3. The role played by these
decays in this context is highlighted in detail in the following sections.

A.6 Techniques for determining β

The angle β of the unitarity triangle is defined by:

Vtb Vtd∗
 
β ≡ arg − ∗ (A.60)
Vcb Vcd
134 CP violation

The decay B o → J/ψKS , belonging to class (1) (see section A.5), is the pri-
mary source for measurements of sin(2β). In the common phase convention,
CP violation is expected to arise mostly from the mixing, driven by =(q/p),
while the decay amplitude, =(Āf¯/A), is expected to contribute only a small
part. The final state J/ψKS is a CP eigenstate and its decay is dominated
by only one diagram, shown in Fig. A.5. For this decay mode we have:

b
c
c} Jψ
W-
}K s
d
s
d
Figure A.5: Decay diagram at the tree level for B o → J/ψKS .

Vtb∗ Vtd Vcs∗ Vcb ∗V

   
Vcd cs
λ(Bd → J/ψKS ) = − ⇒
Vtb Vtd∗ Vcs Vcb∗ Vcd Vcs∗ (A.61)
⇒ =(λJ/ψKS ) = sin(2β)

Where the first term comes from Bdo − B̄do mixing, see Fig. A.6, the second
Āf¯
from the ratio A and the third from the K o − K̄ o mixing.

b d b t,c,u d
t,c,u W - t,c,u W
-
d b d b
t,c,u

Figure A.6: The two diagrams for Bd mixing. Although u, c and t quark
exchange are all shown, the t quark plays a dominant role, mainly due to
his mass, since the amplitude of this process grows with the mass of the
exchanged fermion.

In this case we do not get a phase from the decay part because:
Āf¯ (Vcb Vcs∗ )2
= (A.62)
A |Vcb Vcs |2
A.6 Techniques for determining β 135

is real. The final state is a state of negative CP , i.e. CP |J/ψK S i =

− |J/ψKS i. This introduces an additional minus sign in the result for =(λ).
Before finishing the discussion of this final state we need to consider in more
detail the presence of the KS in the final state. Since neutral kaons can mix,
we pick up another mixing phase. This term creates a phase given by
  ∗ V )2
(Vcd
q cs
= (A.63)
p K |Vcd Vcs |2

which is zero. It is necessary to include this term, however, since there are
other formulations of the CKM matrix than Wolfenstein, which have the
phase in a different location, it is important that the physics predictions not
depend on the CKM convention.

A.6.1 Results on sin(2β)

The first statistically significant measurements of CP violation in the B

system were made recently by BABAR and BELLE [56, 57]. This enormous
achievement was accomplished using an asymmetric e + e− collider tuned to
the Υ(4S). The measurements are listed in Tab. A.4, along with other
previous indications [58]. The average value of 0.67±0.15 is taken from

Experiment sin(2β)
BABAR 0.68±0.30±0.04
BELLE 0.65±0.04±0.02
Average 0.736±0.049
+0.41
CDF 0.79−0.44
ALEPH 0.84+0.82
−1.04 ±0.16
OPAL 3.2+1.8
−2.0 ±0.5

Table A.4: Measurements of sin(2β).

BABAR and BELLE only. This value is consistent with what is expected
from the other known constraints on ρ and η. We have:
p
1 ± 1 − sin2 (2β)
η̄ = (1 − ρ̄) (A.64)
sin(2β)
136 CP violation

There is a four fold ambiguity in the translation between sin(2β) and the
linear constraints in the ρ̄ − η̄ plane. These occur at β, π/2 − β, π + β and
3π/2 − β. Two of these constraints are shown in Fig. A.3, but from recent
measurements one of the two ambiguities can be excluded at 95% CL. The
other two can be viewed by extending these to negative η̄. We think η̄ > 0
based only on measurement of 0K in the neutral kaon system. This analysis
clearly shows that current data are consistent with the SM.

A.6.2 Other modes for measuring sin(2β)

New physics can add differently to the phases in different decay modes if it
contributes differently to the relative decay amplitudes Āf¯/A. Therefore it

Decay mode Branching ratio

B o → φK o (8.1+3.2
−2.6 ) × 10
−6

B o → D+D− ∼ 10−3
B o → D ∗+ D − ∼ 10−3
B o → η0 K o (5.8+1.4
−1.3 ) × 10
−5

B o → J/ψπ o (1.94 ± 0.28) × 10−5

Table A.5: Other modes useful for cross-checking sin(2β).

is interesting to measure CP violation in redundant modes. For example,

the decay B o → φKS should also measure sin(2β). If it is different than
that obtained by B o → J/ψKS , that would be a strong indication of new
physics [59]. Other interesting modes to check sin(2β) are listed in Tab. A.5.
The branching ratios listed with errors have been measured [56, 60, 61], while
those without are theoretical estimates.

A.7 Techniques for determining α

The angle α of the unitarity triangle is defined by:
 ∗ 
Vcb Vcd
α ≡ arg − ∗ (A.65)
Vub Vud
Measuring α is more difficult than measuring β in several respects. First
of all, the decay amplitudes are modulated by V ub rather than Vcb , making
A.7 Techniques for determining α 137

the overall rates small. Secondly these decays belong to class (2) or (3) (see
section A.5), thus the gluonic penguin rates, see Fig. A.7, are of the same
order causing difficulties in extracting the weak phase angle. The penguin

-
W

b t
g
d
u}π-
d u
d}π +

Figure A.7: Penguin diagram for B o → π + π − .

diagrams add a third amplitude to the tree level and mixing amplitudes. It
turns out, however, that this complication can be used to remove discrete
ambiguities. The decay B o → π + π − has often been cited as a way to
measure sin(2α). However, the penguin pollution mentioned above, makes
it difficult to extract the angle. Gronau and London [62] have shown that
an isospin analysis using the additional decays B − → π − π o and B o → π o π o
can be used to extract α where a flavour tagged asymmetry measurement
is needed in the π o π o final state. This is extremely difficult and there is
generally no decay vertex in the π o π o final state.
There is however, a theoretically clean method to determine α. The
interference between tree and penguin diagrams can be exploited by mea-
suring the time dependent CP violating effects in the decays B o → ρπ as
shown by Snyder and Quinn [11]. There are three such neutral decay modes,
listed in Tab. A.6, with their respective penguin, denoted by P, and tree am-
plitudes, denoted by T ij , where i lists charge of the ρ and j the charge of the
π. For the ρo π o mode, isospin constraints are used to eliminate T 00 . The
amplitudes for the charged decays are also given. For the ρπ final state, the
ρ decay amplitude can be parameterized as:
cos(θ)Γρ
f (m, θ) = (A.66)
2(mρ − m − i0.5Γρ )
mρ is the ρ mass of 0.77 GeV and Γρ the width of 0.15 GeV. θ is the helicity
138 CP violation

Decay mode Decay amplitudes

√
2A(B + → ρ+ π o ) = S1 = T +0 + 2P1
√
2A(B + → ρo π + ) = S2 = T 0+ − 2P1
A(B o → ρ+ π − ) = S3 = T +− + P1 + P0
A(B o → ρ− π + ) = S4 = T −+ − P1 + P0
2A(B o → ρo π o ) = S5 = T +− + T +− − T +0 − T 0+ − 2P0

Table A.6: B o → ρπ decay modes.

decay angle and the cos(θ) dependence arises because the ρ must be fully
polarized in this decay which starts with a spin-0 B and ends with a spin-1
ρ and spin-0 π. The full decay amplitudes for B o → ρπ → π + π − π o and the
corresponding B̄ o decay are given by:

A(B o ) = f + S3 + f − S4 + f o S5 /2
A(B̄ o ) = f + S̄3 + f − S̄4 + f o S̄5 /2 (A.67)

where the superscript on the f indicates the charge of the ρ. The sum over
the three neutral B decay amplitudes involves only tree amplitudes; the
penguins vanish. The angle between this sum for B o decays (≡ T ) and the
sum for B̄ o (≡ T̄ ) is precisely α. Computing the amplitudes gives a series
of terms which have both sin(∆mt) and cos(∆mt) time dependences and
coefficients which depend on both sin(2α) and cos(2α).

A.8 Techniques for determining γ

The angle γ of the unitarity triangle is defined by:
 ∗ 
Vub Vud
γ ≡ arg − (A.68)
Vtb Vtd∗

The angle γ could in principle be measured using a CP eigenstate of B s

decay that was dominated by the b → u transition. One such decay that
has been suggested is Bs → ρo KS . However, since it belongs to class (3)
(see section A.5), there are the same “penguin pollution” problems as in
B o → π + π − , but they are more difficult to resolve in the vector-pseudoscalar
final state, the pseudoscalar-pseudoscalar final state here is π o KS , which
A.8 Techniques for determining γ 139

does not have a measurable decay vertex. There are two main methods of
measuring γ:

using time-dependent CP violation in B s decays: this method uses

the decays Bs → Ds± K ∓ where a time-dependent CP violation can
result from the interference between the direct decays and the mixing-
induced decays [63, 64, 65]. Figure A.8 shows the two direct decay
processes for B̄so .

Vub Vcs Vcb Vus

W- }
c
Ds- W
-
}
u
K-

{ {
b s b s
o o

} }
Bs s Bs s
u + c
K D+s
s s

Figure A.8: Two diagrams for B̄so → Ds± K ∓ .

a)
W
- u
}K + }K
u +

B {u
b s s
W
c
u (D )
o }π
d
u

b) u

}K
o

B{
b c (D ) u +
W- s
}K
u s W-
u
d
u }π
Figure A.9: Diagrams for the two interfering processes, (a) B − → D o K −
(colour allowed) followed by D o → K + π − (doubly-Cabibbo suppressed) and
(b) B − → D̄ o K − (colour suppressed) followed by D o → K − π + (Cabibbo
allowed).

using carged B decay rates: this method was proposed by Atwood,

Dunietz and Soni [66], who refined a suggestion by Gronau and
140 CP violation

Wyler [67]. A large CP asymmetry can result from the interference

of the decays B − → K − D o , D o → f and B − → K − D̄ o , D̄ o → f ,
where f is a doubly-Cabibbo suppressed decay of the D o , for example
f = K + π − , Kππ, see Fig. A.9. The overall amplitudes for the two
decays are expected to be approximately equal in magnitude, note
that B − → K − D̄ o is colour-suppressed and B − → K − D o is colour-
allowed. The weak phase difference between them is γ. To observe a
CP asymmetry there must also be a non-zero strong phase between
the two amplitudes. It is necessary to measure the branching ratio
B(B − → K − f ) for at least 2 different states f in order to determine
γ up to discrete ambiguities.

A.9 Summary of crucial measurements for CKM

physics

Table. A.7 lists the most important physics quantities and the decay modes
that can be used for their measurement. Other modes which also may turn

Physics quantity Decay mode

sin(2α) B o → ρπ → π + π − π o
cos(2α) B o → ρπ → π + π − π o
sign(sin(2α)) B o → ρπ & B o → π + π −
sin(γ) Bs → Ds± K ∓
sin(γ) B − → D̄ 0 K −
sin(γ) B o → π + π − & Bs → K + K −
sin(2χ) Bs → J/ψη 0 , J/ψη
sin(2β) B o → J/ψKS
cos(2β) B o → J/ψK o , K o → π`ν
cos(2β) B o → J/ψK ∗o & Bs → J/ψφ
xs Bs → Ds+ π −
∆Γ for Bs Bs → J/ψη 0 , Ds+ π − , K + K −

Table A.7: Required CKM measurements for b’s.

A.10 Rare decays as probes beyond the Standard Model 141

out to be useful include B o → D ∗+ π − and its charge-conjugate [68], which

measures sin(−2β − γ) albeit with a small ∼1% predicted asymmetry and
B → Kπ modes which can be used to find γ albeit with theoretical uncer-
tainties. There are three alternative ways to measure γ which serve both
to remove ambiguities and perform checks. It will be much more difficult
to find other modes to check α, however. One approach is to measure the
CP asymmetry in B o → π + π − and use theoretical models to estimate the
effects of penguin pollution. Minimally, a great deal would be learned about
the models. It also turns out that the third ambiguity in α can be removed
by comparing the CP violating asymmetry in π + π − with that found in ρπ
and using some mild theoretical assumptions [69]. After the three angles α,
β and γ have been measured, we need to check if they add up to 180 o . A
discrepancy here would be unexpected. To be sure, this check is not com-
plete if ambiguities have not been removed (even if the angles sum to 180 o ,
new physics could hide).

A.10 Rare decays as probes beyond the Standard

Model

Rare decays are described by the decay diagrams that feature loops, which
makes them sensitive to high mass gauge bosons and fermions; this is par-
ticularly important for decays belonging to class (4) (see section A.5). Thus,
they are sensitive to new physics. However, it must be kept in mind that
any new effect must be consistent with already measured phenomena such
as Bdo mixing and b → sγ. These processes are often called “penguin” pro-
cesses [71]. A Feynman loop diagram is shown in Fig. A.10 that describes
the transition of a b quark into a charged -1/3 s or d quark, which is effec-
tively a neutral current transition. The dominant charged current decays
change the b quark into a charged +2/3 quark, either c or u. The interme-
diate quark inside the loop can be any charge +2/3 quark. The relative size
of the different contributions arises from different quark masses and CKM
elements. For b → s, in terms of the Cabibbo angle (λ=0.22), we have for
t : c : u − λ2 : λ2 : λ4 . The mass dependence favors the t loop, but the
142 CP violation

-
W

b t,c,u s,d
Figure A.10: Loop or “Penguin” diagram for a b → s or b → d transition.

amplitude for c processes can be quite large ∼30%. Moreover, as pointed

out by Bander, Silverman and Soni [72], interference can occur between t,
c and u diagrams and lead to CP violation. In the SM it is not expected
to occur when b → s, due to the lack of a CKM phase difference, but could
occur when b → d. In any case, it is always worth looking for this effect;
all that needs to be done, for example, is to compare the number of K ∗− γ
events with the number of K ∗+ γ events. There are other possibilities for
physics beyond the SM to appear. For example, the W − in the loop can be
replaced by some other charged object such as a Higgs; it is also possible
for a new object to replace the t.