..
12-:
:3...
I .

.

1‘ Hull?
‘n’N .
yv'h‘o .

.1 .l
1 -.v 1.11.)..-

rllzs‘vtud .p
3 II?- III v. '1.
I :11.

. V'I.
'tvlv it?!
‘3'}.{1}
II.

 

. .5115. I.
I’l‘
..
III}

.
\.IlI~)vuv
.VbtaV..p.<.

 

W303

MICHIGAN STATE UN ERS TY BRARIE

u l I")! lll/lllllllllllllllll

3 1293 00899 0321

ll

 

 

 

 

 

 

 

 

 

 

 

 

 

 

This is to certify that the

dissertation entitled

"A Measurement of the eht Ratio Difference
Between Short (250ns) and Long (2.2fls)
Integration Times with the D0 Uranium-
Liquid Argon Central Calorimeter"

presented by

Bo Pi

has been accepted towards fulfillment
of the requirements for

Ph-D- degree in PhYSiCS

 

 

V. C- flaw/Qua

 

Major professor

//~ ML» 92

Date

 

MSU is an Affirmative Action/Equal Opportunity Institution 042771

 

l' w W —'\
LEBRARY
Michigan state

l University

p J

*—

at

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I
MSU I. An Affirmative Action/Equal Opportunity Institution
cMmMmS-DJ

 

A MEASUREMENT OF THE e/tt RATIO DIFFERENCE BETWEEN SHORT (250

ns) AND LONG (2.2 ns) INTEGRATION TIMES WITH THE D0 URANIUM-
LIQUID ARGON CENTRAL CALORIMETER

By
Bo Pi

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Physics and Astronomy

1992

 

Abstract

A Measurement of the e/n Ratio Difference Between Short (250 ns) and Long (2.2 us)
Integration Times with the D0 Uranium-Liquid Argon Central Calorimeter

By
Bo Pi

The difference of the ratios of the high energy electron and pion responses (e/n) in the
D0 Uranium-liquid Argon central calorimeter is measured using the D0 calorimeter
trigger readout (short integration time: 250 ns ) and precision readout (long integration
time: 2.2 us). This measurement found a 5% difference in the e/n ratio between short

and long integration times, with estimated uncertainty of 2.3%.

This dissertation is dedicated to the heroic students and civilians who died
in Beijing on June 4, 1989.

Acknowledgments

There are many people who provided support and help to make this work possible. In
particular, I would like to thank my advisor Maris Abolins, for his wise advise and patient
guidance, throughout many years of my graduate school. I also would like to say special
thanks to Dan Edmunds and Philippe Laurens, for their friendship and caring support.

I would like to thank all my friends at DO test beam: Bob Hirosky, John Borders, Scott
Snyder, Terry Heuring, Terry Geld, Marcel Demarteau, Welanthantri Dharmaratna, Mike
Tartaglia and Dan Owen, whose hard work made this research possible.

Special thanks to Rich Astur, Steve Jerger, Kate Frame and John Borders, who spent
countless hours reading and correcting this dissertation. Special thanks to Carol
Edmunds and Sarah Laurens for those great dinners. I also would like thank Dolores

Foote, Teresa Thomas and Lisa Ruess whose support made my school years easier.

I would like to thank T. D. Lee and the men and women who worked to make the
CU SPEA program possible, and without whom, my graduate study here at Michigan
State University would not have been possible. I would also like to thank the National

Science Foundation for its continued support throughout these years.

Most of all, I would like to thank my parents, to whom I owe my life and wisdom.

Table of Contents
1 Introduction.
2 D0 Calorimeter
2.1 General Calorimeter Principles. .
2.2 High Energy Particle Shower in the Calorimeter.
2.2.1 High Energy Electromagnetic Shower.
2.2.2 High Energy Hadronic Shower. .
2.3 Sampling Fraction.

2.4 The e/h Signal Ratio. .

2.4.1 The Response to High Energy Electrons and Photons (c/mip). .

2.4.2 The Response to Ionizing Hadrons (ion/mip).
2.4.3 The Response to Neutrons (n/mip).
2.4.4 The Response to Nuclear Gammas (y/mip). .
2.4.5 The Experimental Concerns. .

2.5 D0 Uranium-Liquid Argon Sampling Calorimeter. .

2.5.1 The e/h Ratio of Uranium-liquid Argon Calorimeter.

10

12

12

l4

14

15

15

2.5.2 Example Calculation.
2.6 DO Calorimeter Structure.
D0 Calorimeter Electronics
3.1 General Information. .
3.2 The D0 Calorimeter Precision Readout System. .
3.3 Level-one Calorimeter Trigger Electronics. .
3.3.1 Introduction. .
3.2.2 Difficulty of Fast Shaping.
3.2.3 SPICE Simulation of Level-One Calorimeter Trigger.
3.2.4 NYU Test Bench Measurement. .
3.2.5 Single Channel Measurement.
3.2.6 Summary of Trigger Differentiator’s Gain Correction.
3.4 The Test Beam Trigger Resistor Selection.
3.4.1 The Trigger Resistor Calculation.
3.4.2 Selection of The Load-Two Test Beam Trigger Resistor.
D0 Test Beam Experiment
4.0 General Information and History of the Test Beam Experiment.

4.1 Beam Line Magnets and Target Wheels.

l7

19

3O

31

32

32

33

34

36

37

37

38

42

57

4.2 Beam Trigger, Veto and Tags.
4.3 Proportional Wire Chamber Tracking System and Beam
Momentum Calculation. .

4.4 Test Beam Load-One Configuration.

4.5 Test Beam Load-Two Configuration.

Test Beam Measurement

5.1 General Information of the Load-Two Analysis.

5.2 Correction
5.2.1 Non-linearity of Trigger Differential Driver.
5.2.2 Sampling Fraction Correction.
5.2.3 Cross Talk Between Fast Trigger Channels. .
5.2.4 The Trigger Electronics Gain Variation. .
5.2.5 CCEM Between Module Crack Effect.

5.3 Load-Two Electron and Pion Energy Scan Results. .
5.3.1 The Electron Energy Scan Data. .
5.3.2 The Pion Energy Scan Data. .

5.4 ‘ Load-One Energy Scan Results. .

5.5 Load-One and Load-Two Electron Energy Scan Results

57

58

58

59

67

7O

71

74

76

82

84

84

86

86

Comparison.
6. Results and Monte Carlo Simulation
6.1 General Information of the GEANT Monte Carlo Simulation.
6.1.1 Monte-Carlo Simulation of the Electron and Pion Energy
Scan. .
6.1.2 Monte-Carlo Simulation of the Pion Shower Leakage.
6.2 Summarized Results of the Difference in the Load-Two Test Beam
e/rt Ratio Between Different Integration Times. .
6.3 Precision Readout e/rt Ratio Result. .
6.4 Discussion.
6.5 Conclusions. .

Bibliography. .

iv

87

104

105

106

107

111

112

114

126

List of Figures

1.1
2.1

2.2

2.3

2.4

2.5

2.6
2.7

2.8

3.1
3.2

3.3a
3.3b

3.4
3.5

3.6

D0 detector.

Side view of the Monte-Carlo simulation of a 50 GeV electron

hitting the D0 central calorimeter.

Side view of the Monte- Carlo simulation of a 50 GeV pion
hitting the DO central calorimeter. . .

D0 liquid Argon calorimeter.

End view of the DO central calorimeter. The mner ring is made

of 32 EM modules, the middle ring is made of 16 FH modules,

and the outer ring is made of 16 CH modules.

End view of a DO central calorimeter electromagnetic module

D0 central calorimeter gap structure.

End view of a DO central calorimeter fine hadronic module
(CCFH). . .

End view of a D0 central calorimeter coarse hadronic module
(CCCH). . . . . . . . . . . .
Block diagram of the D0 calorimeter data path. .

Conceptual schematic of the D0 level-one calorimeter trigger
electronics.

D0 calorimeter preamplifier and BLS circuits.

D0 calorimeter trigger adder, trigger cable driver and trigger
front-end receiver circuits. .

SPICE simulation of the D0 calorimeter signal pulse shapes.

SPICE simulation of the D0 level-one calorimeter trigger
response vs. input capacitance. .

SPICE simulation of the D0 front-end electronics' transistor
beta effects. . . . . . . . . .

22

23

24

25

26
27

28

29

45

47
48

49

50

3.7

3.8

3.9
3.10
3.11
4.1
4.2
4.3
4.4
4.5
4.6
4.7
5.0

5.1

5.2

5.3

5.4

5.5

5.6
5.7

aloe: diagram of the NYU D0 calorimeter electronics test
nc . .

a) SPICE simulation and NYU test bench measurement of D0
level-one calorimeter trigger response vs. input capacitance.
b) SPICE simulation and test beam load-one single channel
measurement. . . . . . . . . .
D0 calorimeter beam and calibration pulse shapes. .

Trigger adder gain vs. total parallel input resistance.

The test beam load-two trigger summing resistor distribution.
D0 test beam cryostat and transporter.

Fermilab NWA beam line.

Close up of NWA beam line.

D0 test beam scan map.

Load-one test beam module configuration.

Load-two test beam module configuration.

Load-two test beam trigger electronics coverage.

SPICE simulation of the TRG and BLS responses to fast and
slow showers. .

a) Level-one calorimeter trigger response vs. pulse amplitude.
b) Ratio of the TRG and precision BLS vs. electron beam
energy, with and without the trigger driver linearity correction.

Conceptual schematic of D0 calorimeter signal crosstalk.

The excess energy in the level-one calorimeter trigger HD
channel vs. energy in the EM channel. . .

a) The EM trigger tower gain distribution.
b) The HD trigger tower gain distribution.

Energy distributions with difi‘erent gain smearing. a) 100 GeV
pion with 10% gain smearing, b) 100 GeV electron with 10%
gain smearing, c) 100 GeV pion with 20% gain smearing, d)
100 GeV electron with 20% gain smearing. .

Pion energy resolution vs. the gain variation.

Schematic of the CC module crack effect.

51

52
53
54
55

61

62

63

65

89

91

92

93

94
95

5.8

5.9

5.10

5.11

5.12

5.13

5.14

6.0

6.1

6.2

6.3

6.4

6.5

6.6

6.7

6.8

a) Ratio of the TRG and BLS vs. the vertical position (cm) of
the 50 GeV electron beam.

b) Energy response of the BLS vs. the vertical position (cm) of
the 50 GeV electron beam. .

Monte Carlo simulation of the energy distribution of 50 GeV
pion showers in the D0 central calorimeter. . . .

The linearity of electron energy scan at 11:0.05 and d) = 31.6.
The linearity of electron energy scan at 11:0.45 and t) = 31.6.
The linearity of pion energy scan at n=0.05 and ¢= 31.6.

The linearity of pion energy scan at n=0.45 and ¢= 31.6.

The electron and pion energy resolution vs. beam momentum. .

Geometry information for Monte-Carlo simulation of the load-
two test beam. a) standard test beam configuration, b) with
extra modules to study the pion shower leakage.

The Monte Carlo simulation of the electron and pion energy

Monte-Carlo simulation of the load-two test beam energy
resolution.

Load-two e/rt ratio difference between short (TRG) and long
(BLS) readout.

Load-two BLS e/1t ratio. .
Load-one BLS e/rt ratio. .

Load-one and Load-two e/rt ratio comparison with the load-
two data shifted up byO..045 .

The longitudinal energy distribution of 5 GeV electrons in
U/LAr gaps with 10 keV cutoff energy. The top is energy in
all materials (U, LAr and G10), the middle rs energy in LAr
only, the bottom is energy in 610 only. . .

The longitudinal energy distribution of 5 GeV electrons in
U/LAr gaps with 100 keV cutoff energy. The top is energy in
all materials (U, LAr and G10), the middle rs energy in LAr
only, the bottom 1s energy in 610 only... .

97

98

100

101

102
103

116

117

118

119

120

121

122

123

124

6.9

The longitudinal energy distribution of 5 GeV electrons in
U/LAr gaps with 1.0 MeV cutoff energy. The top is energy in
all materials (U, LAr and G10), the middle is energy in LAr
only, the bottom 1s energy in 610 only. .

viii

125

List of Tables

2.1
3.1

3.2
5.1
5.2
6.1

6.2

6.3

6.4

The detectable signals in 100 GeV electron and pion showers. .

The sampling fractions and feedback capacitance in the test
beam calorimeter trigger. . . . . .

The test beam trigger summing resistors.
The trigger and precision sampling fractions.
The test beam electron and pion energy resolution. .

The test beam load-two pion shower energy leakage.

The list of the uncertainties of the test beam e/rt ratio difference
measurement. . .

The combined results of the e/rt ratio difference measurement. .

The load-two BIS e/rt ratio measurement.

18

39
42
72
80
107

108

110

112

Chapter 1
Introduction

Over the last 15 years, there has been a trend by which the high energy physics
experiments have changed from the spectrometer to the calorimeter. Some major
discoveries, e.g. the existence of the intermediate vector boson W and Z, became
possible as a result of this development. The new generation of experiments increases
the emphasis on the calorimeter more than ever, and the performance of the calorimeter
becomes the major factor in discriminating experiments. This dissertation is a study of
one of the most important issues affecting calorimeter performance: the ratio of the
electron and pion responses (e/rt), using the D0 Uranium-liquid Argon central
calorimeter. A short introduction will be given on the Fermi National Accelerator

Laboratory and the DO experiment.

The Fermi National Accelerator Laboratory has the highest energy proton (p) anti-
proton (5) collider in the world“. The accelerator has two operational modes: fixed
target and collider. When the accelerator operates in the fixed target mode, it provides
a 800 GeV proton beam to the fixed target experiments at the rate of 2x1013
protons/second. In the collider mode, the accelerator can accelerate and store both
proton and anti-proton beams at 900 GeV in the Superconducting Tevatron ring, and
steer them to a head-on collision. The collision creates a center of mass energy of 1.8
TeV. The luminosity has reached a maximum of 1.5x1030/s, and is expected to reach

5.0x1030/s in 1992. There are a total of 6 interaction points (A B C D E F) around the

2

accelerator ring. Low beta focus magnets have been installed at two of these points, B0
and D0, in order to get the maximum luminosity. Two big experiments, CDF (BO) and

D0, are installed at these interaction areas.

The D0 experimental (Figure 1.1), designed with very specific physics goals in mind,
utilizes the most recent developments in detector technology. The detector's
performance is greatly improved from previous collider detectors. The D0 detector was
designed to provide optimal measurement of leptons (electrons, muons and their
neutrinos) and hadrons (jets) produced in high-energy collisions of protons and anti-
protons at Fermilab. It is based on a Uranium- liquid Argon calorimeter organized in
projective towers surrounding the interaction region over nearly the full 4n solid angle.
The fine longitudinal segmentation of the calorimeter helps the electron discrimination
and energy measurement in the large background of pions. A tracking system also
covers nearly the full 41: solid angle of the interaction region. A transition radiation
detector (TRD) is built inside the tracking system, to further improve the identification
of electrons. Neutrinos, which escape undetected, reveal their presence in the form of
an imbalance in the transverse momentum summed over the full detector, and the
quality of this measurement is crucially dependent on the good resolution and
uniformity of the calorimeter. Muons, on the other hand, are identified by their
characteristic ease of penetration through the thick calorimeter as well as through the
layers of iron comprising the toroidal magnetic spectrometer outside the calorimeter.
The latter provides a measurement of the muon's transverse momentum over much of

the solid angle, including a determination of the sign of its charge.

The calorimeter is the central part of the detector system. Its performance is critical to
many physics measurements of the experiment. Because of its importance, we have

done three test beam experiments to expose different parts of the calorimeter to high

3

energy electron and pion beams. These test beam experiments gave us opportunities to

study the calorimeter response in great detail.

The design of the D0 calorimeter is based on the most recent knowledge of the hadronic
showers. In the past several years, we have begun to understand the various processes
involved in a hadronic shower at a fairly detailed level. However, in order to quantify
this knowledge, precise experimental results are needed. It has been predicted that the
response of hadron showers in neutron rich material (like Uranium) will increase as a
function of the time over which the calorimeter signal is integrated. The existence of
this phenomenon has been verified by several groups[29’3°'31], and in the case of
Uranium-liquid Argon calorimeters, the HELIOS collaboration is the only experiment
that has done a study on this subjectml. But its measurement are not very precise.
This dissertation is a study of the time dependence of a hadronic shower in the D0
Uranium-liquid Argon calorimeter. The e/rt ratio of the D0 Uranium-liquid Argon
calorimeter has been measured with the fast calorimeter trigger electronics (250ns
integration time) and the precision calorimeter electronics (2.2113 integration time), and

with particle energies ranging between 15 and 150 GeV.

The D0 calorimeter, as well as the properties that affect the e/1t ratio, will be described
in the next Chapter. Chapter 3 is a description of the D0 calorimeter electronics
system. It will include a discussion of the difficulties with the fast calorimeter trigger
electronics, as well as the solution. Chapter 4 is a description of the D0 calorimeter test
beam facility and the configurations of the two test beam runs in the period 1990 to
1992. The analysis of the 1991-1992 (load-two) test beam run data will be presented in
Chapter 5, including discussions of the various corrections applied to the final e/rt ratio
measurement. In Chapter 6, the final result of the e/rt ratio measurement with different
integration times will be presented, as well as a description of the detector Monte Carlo

simulation and a discussion of the results.

 

2.. .s...

7.
C/ -7

z 71
.WQN» $111
111.1. \\ Ea
\ III M\ a].

. was

     

  

  

HHHHH

Figure 1.1 Do detector

Chapter 2
D0 calorimeter

2.1 ' General calorimeter principles

Conceptually, a calorimeter is a block of matter that intercepts a high energy particle
and has sufficient thickness to cause the particle to interact and deposit all its energy
inside the detector volume in a subsequent cascade or "shower" of increasingly lower
energy particles. Figures 2.1 and 2.2 show a GEANTB] Monte Carlo simulation of a
50 GeV electron and pion shower in the DO central calorimeter, and they demonstrate
the complexity of high energy particle showers. Eventually, most of the incident
energy is dissipated and appears in the form of heat. Some (usually a small fraction) of
the deposited energy goes into the production of a more practical signal (e.g.,
scintillation light, ionization charge). It is an important feature of a calorimeter that this
signal be proportional to the initial energy. In principle, the uncertainty in the energy
measurement is governed by statistical fluctuations in the production of the detectable
signals (e.g. ionization charge), and the fractional resolution (015) improves with
increasing energy (E) as w/E . In Chapter 1, it was noted that calorimetric detectors
offer many other attractive capabilities, apart from the energy response. These
properties are listed as followingls]:

A) Calorimeters are sensitive to neutral as well as charged particles.

B) The size of the detector scales logarithmically with particle energy E, whereas

for magnetic spectrometers the size scales with particle momentum «[13.

6

C) With segmented detectors, information on the shower development allows
precision measurements of the position and the angle of the incident particle.

D) The differences in the patterns of energy deposition of electrons, muons, and
hadrons can be exploited for particle identification.

E) The fast time response allows operation at high interaction rates, and the

patterns of energy deposition can be used for real-time event selection.
2.2 High energy particle shower in the calorimeter

In this section, I will briefly summarize the mechanisms of high energy particle

interactions with materials.

2.2.1 High energy electromagnetic shower.

The electromagnetic (EM) interacting particles include photon (y), electron (e‘) and
positron (e"'). At high energy (kinetic energy Ek> a few hundred MeV), photon,
electron and positron showers are very similar except for the first interaction. For the
high energy photon the first interaction will be to produce an electron and positron pair.
For the high energy electron and positron, the first interaction will likely be
bremsstrahlung which produces high energy photons. In general, the photons lose their
energy through the following processes:

A) Pair (e+, 6) production.

3) Compton scattering.

C) Photoelectric effect.
The electrons and positrons lose their energy through:

A) Bremsstrahlung.

B) Ionization.

C) Multiple scattering.

7

The EM shower is a chain of these interactions, and it will keep on going until all the
energy of the initial particle is dissipated. All of these processes are governed by one of
the best established theoretical framework in physics: quantum electrodynamics, and

they can be calculated from theoretical formulas with good precision.
2.2.2 High energy hadronic shower

For hadronic interacting (HD) particles (hadrons: pions. protons, etc.), the interactions
with material are more complicated than electromagnetically interacting particles.
When a high energy hadron hits a block of matter, the hadron will interact with one of
its nuclei. In this process, mesons are usually produced (rt, K, etc.), and these particles
may go on to produce a chain of interactions themselves. Some of the particles
produced in a hadron shower interact exclusively electromagnetically (e.g. no, 1]).
Therefore, hadron showers in general contain a component that propagates
electromagnetically. The fraction of the initial hadron energy converted into the
electromagnetic component may vary strongly from event to event depending on the
detailed processes occurring in the early phase of the shower development. The
hadrons (e.g. proton, n+,1t’) produced in a hadron shower interact with nuclei and
transfer a part of their energy to the nuclei. The excited nuclei will release their energy
by emitting a certain number of nucleons and low-energy gammas (Ys). This chain of
nuclear reactions will keep on going until the energies of the hadrons fall below about 1
GeV. Then hadrons will lose their energy mostly through ionization. However, there is
one exception, that is neutrons. Since a neutron is not a charged particle, it does not
lose energy by ionization. These neutrons, created in the hadron showers, lose their
energy exclusively through nuclear processes, i.e. either through elastic scattering off
nuclei, or through capture by nuclei. The nuclei that capture the neutrons will be in an
excited state, and will lose their energy either through fission or simply by emitting Ys.

Fission could create more low energy neutrons and 18.

8
The calorimeter signal of a hadron shower will come from the contributions of all these
different processes. Some of these nuclear processes take a long time (~ us) to finish.
The theoretical calculations of these strong interaction processes are not as well
understood as electromagnetic processes. Experimental results to further understand

these processes in hadron showers are still needed
2.3 Sampling fraction

In order to be able to measure the energy of an incident particle in a sampling
calorimeter like D0, one has to understand the sampling fraction, or the fraction of
energy deposited in the active layers through the ionization process, as well as the
efficiency of the detection technique. The latter factor is determined by the saturation
effect1 and the amplifier’s efficiency in detecting the signal. For the liquid Argon
calorimeter, the source of the signal is the ionization charges in the liquid Argon gaps.
The first factor is determined by the structure of the sampling calorimeter, the materials
chosen, and the interaction cross-sections of various particles interacting with these

materials. I will discuss these factors in the following sections.

Since the detectable signals in the calorimeter are produced by the ionization of various
charged particles, the calorimeter response of a minimum-ionizing particle (mip) is
used as a scale for the calorimeter signal. The mip is a particle that interacts with
materials exclusively through the ionization process, and its energy loss per unit of
mass is minimuml33]. Thus, the saturation effect is minimized. Many measurements
and calculations have been done for various materials to determine their mip energy

loss (fi), and there is a standard table of a? for various materials. In practice, the

ifs of the active and passive materials are used to determine the sampling fractions of

 

1finsamdmefixtmliquidArgonmnizadmcabrhnaaismusedbyWCMrgemnmein
thedriftingpath. ltcausesthedetectortolosesignal. Itissensitivetotheappliedhighvoltage.

9

a calorimeter. The active layer is the sensitive layer in which energy deposited in the
layer is detected, and the passive layer is the layer of inactive materials. The sampling
fraction (the mip sampling fiaction) is defined as:

d—E[one active layer]
S dx

nip—

 

dE . dE . (2.1)
Z[one actrve layer] + Ex-[one passwe layer]

The low energy muon is the a known particle that behaves like a minimum-ionizing
particle. Most detectors use low energy muon response to verify their sampling
fractions. Throughout this dissertation, the mip sampling fraction is applied to weight

the calorimeter signals unless otherwise noted.
2.4 The e/h signal ratio

For a given calorimeter, hadron showers are generally detected with an energy
resolution that is worse than the electromagnetic showers. There are several reasons for
the poorer hadron energy resolution. First, after the early phase of hadron showers, the
initial energy of the hadrons is converted into the pure electromagnetic component
(EM), and the pure hadronic component (HD). For the HD component of hadron
showers, some fraction of the energy will disappear without contributing to the
calorimeter signal. This is, for example, the energy spent in breaking up the nuclei, and
the energy of some particles (v, u, and a fraction of neutrons) that are created in the
showers which escape from the detector. This will cause the calorimeter response to
the HD component to have a much broader distribution than the EM component, and
the average response to the EM and the HD components to be different. Furthermore,
the fraction of the initial hadron energy converted into the EM component fluctuates
event by event; this and the unequal responses between the EM and 111) components of
the hadron showers further worsen the hadron energy resolution.

10
The calorimeter response difference between the EM and HD components of the hadron

showers is characterized by the e/h ratio, defined 33110];

(response of EM components)
(response of pure hadronic components)

 

5.-
h

In order to understand the % ratio of a calorimeter, I will demonstrate a simple analysis

of the calorimeter response to various secondary particles in a hadron showerll 1]:

(a) . High energy photons, which arise mainly from no decays, and form the EM
shower component.

(b) Ionizing hadrons (charged pions, kaons, protons, etc.).

(c) Soft neutrons.

((1) Soft Ys from the nuclear processes (fission, nuclear de-excitation, etc.)..

Items (b), (c) and (d) are the pure 111) components.

The mip response will be used as a scale. Relating the response of the four mentioned
. . , e . .
types of particles to mip s, the Z ratro can be wntten asm]:

E: e/mip (2.2)
’1 fauion/mipi'f. n/mip+ f7 Y/m‘iv

where f .

f,, and f, are the average fractions of the energy in the pure hadronic
component that are in the forms of ionizing hadrons, neutrons and nuclear Ys. The
calorimeter responses to various terms of Formula 2.2 will be briefly discussed in the

next section. A detailed discussion can be found in [11].

2.4.1 The response to high energy electrons and photons (e/mip)

The detectable signal of the high energy electrons and photons comes from ionization

in the active layers by the electrons and positrons generated in the EM shower

1 1
development. Although one might expect this signal to be equal to the mip signal, in

practice they are not. The reason for this is the following.

For the relativistic charged particle with charge e, the ionization energy loss per unit of

mass is given by the Bethe-Bloch equationl33]:

-15.- 21.1. ZMC’Y’B’ - 2-55.
dx —4rtNAr.m,c A32 [ln(—'—l——— fl 2 (23)

where 13:? 741—52)“.

In this formula, Z, A and I refer to the atomic number, the atomic weight and ionization

constant in the target material, respectively, "1,, and r, are the mass and the classical

radius of the electron, v is the velocity of the incident particle, NA is Avogadro's
number, and 6 is a small density effect correction. Thus the ionization energy loss per

unit mass is proportional to g, with some weak dependence on the other properties of

the material. Because in is relatively constant, (f—‘f- is approximately constant also.

Whereas, in the EM shower there are a large number of soft 1's (EK<1 MeV) produced,
and the dominant process by which these soft 75 to generate a detectable signal is to
transfer their energy to electrons through the photo-electric process. But the cross
section of the photo-electric process with K shell electrons is proportional to 25 14-5]:

2 Z

a" “ Zs(.flfc_.)
her

where the Z is the charge of matter, Mr is the photon energy, and m, is the electron
reset mass. This means that the photo-electric process happens primarily in the high 2

materials (usually passive materials) of a sampling calorimeter. The low energy photo-

electrons have a very short mean free path, so they deposit their energy near where they

12
are produced. Only when these photo-electrons are produced near the surface of the

passive materials will they be able to travel to the active material and create detectable

signals.

Because of this difference in interaction cross-sections, a sampling calorimeter using
materials of very different 2 values in its passive and active layers will show different
electron and mip response. In general, if Zpassive > Zactive, then e/mip < 1.0 , and if

Zpassive < Zactive, then e/mip > ”1

2.4.2 The response to ionizing hadrons (ion/mip)

In the high-energy hadron shower, the average fraction of the energy in the pure

hadronic component deposited in the form of ionization is roughlyll 1]

Z

fion=:’

(2.4)
where Z and A are the charge and atomic weight of the matter.
The simple reason for this formula is that most of the hadrons produced in a high

energy hadron shower are protons and neutrons, and the probability of knocking out

protons in a matter goes approximately as in

For Uranium, about 40% of the energy in the pure hadronic component is deposited in
the form of ionization. The charged hadronic component is dominated by protons (70-
75%)[“]. The remaining contribution comes from charged pions, kaons, etc.. The
ionization energy density deposited by these charged hadrons is greater than mip's, thus
the ion/mip signal ratio of a calorimeter is very dependent on the material of the
calorimeter. For a practical calorimeter with saturation readout material (like liquid

Argon, etc.), the ion/mip signal ratio has a range of 0.85 - 1.0121].

13
2.4.3 The response to the neutrons (tr/mip)

A large portion of the energy in the pure hadronic component is carried by neutrons.
There are more neutrons than any charged hadronic particles produced in hadronic
showers. This is simply because there are more neutrons in most nuclei. Because the
neutron has no charge, it can not directly create a detectable signal through ionization as
the proton does. The processes by which neutrons are produced and lose their

energyllo,l 1] will be examined in the following.

The neutrons in the high energy hadron shower are produced through nuclear processes.
Initially, fast neutrons (ErGeVs) are produced through nuclear processes induced by
high energy hadronic particles (charged pions and protons etc.). These fast neutrons
may develop quite complicated showers themselves and produce more neutrons. These
neutrons will lose their energy by elastic and inelastic scattering off the nucleus. Once
the neutrons slow down to soft neutrons (Ek ~ a few MeV), nuclear fission may
dominate the neutron interaction processes. When the neutrons are further slowed
down (Ek< MeV), nuclear neutron capture processes will dominate. In each of these

processes, there is energy released in the form of soft 1‘s, and more neutrons may be

produced.

In a calorimeter using materials containing hydrogen, the neutrons will lose about half
their energy each time they scatter off a proton (hydrogen nucleus), and generate
detectable signals through proton ionization. As the atomic number of the material
goes up, this form of energy transfer decreases very rapidly. The neutrons take longer
to lose their energy to become soft neutrons. In the case of Uranium-liquid Argon, the
contribution of neutron-nucleus scattering is negligible. However the presence of
Uranium significantly increases the number of neutrons in the hadron showerlloouvm,

and the soft 7s produced from neutron-nucleus processes (fission, neutron capture, etc.)

14
will make significant contributions to the total detectable signal. On the other hand, the

processes of slowing down and capturing the neutrons take a relatively long time, on
the order of usUS]. In order to detect the signals from these contributions, the

calorimeter has to have an integration time on the order of us.

In summary, the n/mip signal ratio depends on the materials of the calorimeter and the

signal integration time.
2.4.4 The response to nuclear gamma's (y/rnips)

The Ys from the nuclear processes induced by charged hadrons and neutrons are in the
energy range of a few MeV. These soft Ys are produced mostly in the passive materials
(Uranium, etc.). Their contribution to the detectable signal is through the same photo-
electric process as the soft Ys in EM showers (Section 2.4.1). If a high 2 material is
used as the passive material and a low Z material is use for the active material, this

form of signal is suppressed as described in Section 2.4.1.
2.4.5 The experimental concerns

Because all hadronic particle showers contain pure electromagnetic and pure hadronic
components. In practice, the calorimeter response of the pure electromagnetic
component "e" is measured with high energy electrons, and the response of the pure
hadronic component "h" is measured with high energy charged pions. For high energy
hadron showers the relation between the db ratio and the e/rt ratio isl1 1]

e e/h

._. 2.5
.(E) 1-(f..(E))[1-e/hl ‘ ’

 

where (fm(E))=0.1lnE(GeV). (2.6)

15
This is a reasonable approximation in the energy range from 10 to 100 GeV. From now

on, I will use the e/rt ratio to represent the e/h ratio.

The deviation of e/rt from unity will cause a calorimeter's hadronic response to have a
non-Gaussian energy distribution, the energy resolution will not scale with E and
will have a large constant tennz, and the calorimeter signal will not be proportional to

the hadron energy (non-linearity)[9'111. With experimental energies becoming larger

. . e .
and larger, the devratron from — =1 becomes one of the most 1mportant factors
' 7t

limiting a calorimeter‘s performance. In today’s p—p hadron collider experiments,
being able to precisely measure the energy of multiparticle jets (including
electromagnetic and hadronic particles) becomes one of the most important factors

influencing collider detector design.
2.5 D0 Uranium-liquid Argon (U/LAr) sampling calorimeter

In the past 10 years, much progress has been made in understanding the factors limiting
calorimeter performance. The D0 calorimeter was designed on the basis of this new
knowledge. The D0 calorimeter is built with Uranium as the absorber and liquid Argon
as the active material. The choice of this combination gives maximum density, ease of
calibration, straight forward projective tower segmentation, and good energy resolution.
From the early studiesll9» 29], the Uranium-liquid Argon combination gives nearly
equal electromagnetic and hadronic energy responses. Also, the high density of
Uranium is vital to the design, because it allows the overall dimensions of the detector

to be kept as small as possible.

 

2 2

s N
2Thecalorimeterenergyresolutioncarbepm'ametrizedas(if) -c2+-E-+?-,whereCisthe
constantterm,Sisthesamplingtenn,andNistheextemalnoiseterm. Anidealcalorimetershouldhave
C=0andN=0.

16

2.5.1 The e/h ratio of Uranium-liquid Argon calorimeter

One of the major reasons that DO chose the Uranium-liquid Argon combination for its

calorimeter is that it gives an e/rt ratio close to one. There are several effects which

cause this ratio to be close to unity:

a) Because of the large Z difference between Uranium and liquid Argon, the
response of the EM particles is suppressed relative to the mip (Section 2.4.1).

Our test beam measurement shows that e/mip ~ 0792].

b) Because of the fission property of Uranium, a large number of neutrons are
created in the hadronic showers, and they make a significant contribution to the

hadronic signalm].

Applying the above theories of high energy particle interaction with materials, I will
estimate the high energy electromagnetic and hadronic particle signals in the D0

calorimeter. The mip signal will used as a scale.
For electromagnetic particles, the total detectable signals are:
70% E, where E is the total energy of the incident particle.

For the hadronic particles, the contributions to the detectable signals are:

1) From Formula 2.6, the EM component is: (f.(E)) = 0.11n E(GeV)

2) The HD component is: (f...) = 1- (fa). From Formula 2.4, the ionization part is

then (fa...) = 40%*(f“). However, the saturation effect of the calorimeter has to be

considered. The estimated detection efficiency for this energy is about 90951211,

l7
3) The signals contributed by neutrons are estimated from theoretical calculationsm]
and experimental measurementsm]. The following estimate of the neutron processes
per GeV of hadronic energy are based on [12, 18];
the average number of fissions caused by neutrons is 10 fissions/GeV,

the average number of neutrons created in the shower is 45/GeV.

The detectable energy in the form of soft ‘Ys released fiom these neutron processes is;
the average fission contribution:

7.5 MeV/fission * 10 fissions/GeV = 75 MeV/GeV,
the average neutron nucleus inelastic scattering contribution:

5 MeV/neutron * 45 neutrons/GeV = 225 MeV/GeV,
the average neutron capture contribution:

7.5 MeV/neutron cap * 45 neutrons/GeV = 338 MeV/GeV.
But the neutron capture largely happens after the neutrons slow down to the keV energy
range. The processes which slow down these neutrons take up to a few us.
Brueckmannlm has calculated that after 100 ns, ~25 % of the neutrons would be
captured, and nearly 100% of the neutrons would be captured after 2 us. In the case of
the D0 calorimeter, we use 2.2 as integration time for the precision readout, and
approximate 250 ns integration time for the fast trigger readout. Following
Brueckmann's calculation, I calculate that after 250 as, about 45% of neutrons would be
captured. The total neutron signals between 250 ns and 2.2 us are

250 as: 452 MeV/GeV,
2.2 us: 638 MeV/GeV.

These signals are in the form of soft 1s, and the detection efficiency of these soft 13 is
only 40%[12]. Therefore the detectable soft 7 signal from neutron processes between
250 ns and 2.5 as integration time is 181 to 255 MeV/GeV.

18

2.5.2 Example calculation

To summarize the above estimates, I will calculate the detectable signals from 100 GeV

electromagnetic particles and hadronic particles. I will use mip's sampling as a scale. It

 

I

The detectable signal in 100 The detectable signal in 100
GeV electron showers (GeV) GeV pion showers (GeV)

  
  
   
  
  
   
   

 

Electromagnetic component 70 35

 

Hadronic ionizin com nent l8

 

Soft 13 from neutrons 9.1 j 12.7(250nsj 2.211s)

 

 

 

 

Total detectable si

Table 2.1

is summarized in Table 2.1.

The electron showers are made up of 100 percent electromagnetic components. The
detection efficiency for the high energy electron shower is only 70% (e/mip=0.7), so
the detectable signal for the electromagnetic particles is 70 GeV.

The electromagnetic component of the pion showers is <fem>=0.llln(E)=50 GeV.
With the same detection efficiency as the high energy electrons (70%), the detectable
signal is 70%‘50 = 35 GeV.

The pure hadronic component of the pion showers is <fhd>=(1-50%)*1OOGeV=SOGeV.
In this component, the fractional energy deposited through ionization is 40%, and the
detection efficiency of this signal is 90%. This yields a detectable signal of
50*40%*90% = 18 GeV. The total energy in the form of soft Ys produced through
neutron-nucleus processes is 0.45*50 = 22.5 GeV with 250 ns integration time, and

0.64*50 = 32 GeV with 2.2 us integration time. The detection efficiency of these soft

19

VS is only 40%. This yields a detectable signal of 40%*22.5 = 9.1 GeV with 250ns
integration time, and 40%*32 = 12.7 GeV with 2.2 us integration time.

I conclude that at 100 GeV,
250 ns: e/rt = 70/62.l = 1.12.
2.5 us: e/rt = 70/65.7 = 1.07.
Therefore, a 5% ch: ratio difference between 250 as integration and 2.2 as integration

times is expected

In the above estimate, there are two parameters that are not known very precisely; the
efficiency of detecting the signal from the ionizing hadronic component, and the
efficiency of detecting the soft ys from neutron processes. But the estimate of the em
ratio difference between short and long integration times is not sensitive to these two

parameters.
2.6 D0 central calorimeter structure

The D0 calorimeter [2] (fig. 2.3) is divided into three parts: one central calorimeter (CC),
and two identical end calorimeters (EC). In this dissertation, I am mainly going to
analyze the calorimeter test beam data taken with the CC only, so I will only describe
the CC structure here. More detailed information about the D0 calorimeter structure

(both CC and EC) can be found in [2].

The central calorimeter is segmented longitudinally and transversely. Longitudinally, it
is constructed with three different modules named by their functions: the
electromagnetic (CCEM) module, the fine hadronic (CCFH) module and the coarse
hadronic (CCCH) module. Both CCEM and CCFH modules are further divided
longitudinally. In the transverse dimension, all modules are subdivided into segments

20
with a size of 21t/64 radians in azimuth and 0.1 pseudorapidity (mm in the Z

dimension, with the exception of the third CCEM sub-layer.

Figure 2.4 shows the central calorimeter consisting of 32 EM modules, 16 FH modules,
and 16 CH modules. The shaded region shows the modules used for the 1990-1991

load-two test beam runs.

A CCEM module is shown in Figure 2.5. It has 20.5 radiation lengths (X0) of material
in 21‘ Uranium-liquid Argon (U/LAr) gaps, with 610 signal boards in the center of the
liquid Argon gaps. There are extra G10 laminated readout boards in some U/LAr gaps.
Figure 2.6 shows the U, LAr and 010 gap structure of CCEM. The Uranium plates are
3 mm thick; the LAr gaps are 0.089 inch yielding a sampling fraction of about 12.0%.
The CCEM is read out in four separate sub-layers, comprised of 2.0, 2.0, 6.8 and 9.7
radiation lengths. This fine longitudinal segmentation in the early part of the array
contributes to the excellent electron and pion discrimination of the calorimeter. Each
CCEM module is divided into two readout segments in azimuth (4)). with A¢=21tl64.
In the z dimension, the readout segments are made to have each tower subtend uniform
An = 0.1 intervals. The exception to the transverse segmentation described above is in
the third CCEM sub-layer, where the typical EM showers are at the peak of their
longitudinal development. Throughout this layer, the transverse segmentation is
increased twofold in both 4) and 11 dimensions. This design improves the resolution of
the EM shower positions and the rejection of photon hadron overlaps which mimic

electron candidates.

A CC fine hadronic (CCFH) module (Figure 2.7) is comprised of 50 U/LAr gaps
totaling 3.24 interaction lengths. The gaps all contain 0.236 inch 1.7% niobium 98.3%
uranium alloy plates and 0.089 inch LAr gaps. This gives a sampling fraction of about
7%. For readout the module is divided longitudinally into three sub-layers of 1.36, 1.04

21
and 0.84 interaction lengths to try to equalize the capacitance of the readout channels.

The transverse segmentation is done the same way as in the normal CCEM layers.

The main purpose of the CC coarse hadronic (CCCH) module (Figure 2.8) is to catch
the longitudinal leakage of hadronic showers. For the average hadronic showers at the
Fermilab collider energy, the longitudinal energy leakage is expected to be around a
few percent. Since the resolution of this measurement will n0t have any great
contribution to the overall energy resolution, the equalization of the electron and hadron
responses is not important. The CCCH is made of 10 gaps of 1.625 inch copper plates
and 0.089 inch LAr gaps. This gives 2.93 interaction lengths and a very coarse

sampling fraction of 1.7%.

22

.eoBEtoEu 35:8 8
2.. misc. E583 >00 an a he coca—25m 25-2.52 he 33> 25 <

Euc—

_.~ 2:3".

 

 

 

23

.BBEtch 35:8 w . \ ..
on 2: 95:: 53 >00 cm a ._e coax—:85 230 2:22 he 32> on: < N N 2: E x

.932“: ..nefifl. ,

Elli! gi/I‘

 

 

 

 

. z . i. . .l.l. l..V l
/ . . ). . H4... . /. «I
.V t C x .. x :2 x .a... x. 4.. x e p... ./ .\ u /././././ x4. ,. /./.
\ \ . .Jx \ x . _ .4» I \ .z / I / ./ II I ./
.\. \x \ \ \E\ O :s \\:.\“.t . ._.! . 2. xx .47.. .z ./ ./ ./ ..‘4
. \ .. \ V \ . . T I. . .. I. . . .l ’
\. \ \ \.~ / ..A\; ». \ ’. _ ..4. I. .\4 7 4 //. I . r I .

24

   
 
  
  
  
   

END CALORIMETER

Outer Hadronic
(Coarse)

Middle Hadronic
(F-"me 81 Coarse)

CENTRAL
CALORIMETEFI
Electromagnetic

Inner Hadronic Fine Hadronic

(fine 8. Coarse) Coarse Hadronic

Figure 2.3 D0 liquid Argon calorimeter

25

 

Figure 2.4 End view of the D0 central calorimeter. The inner ring is
made of 32 EM modules, the middle ring is made of 16 FH modules,
and the outs ring is made of 16 CH modules.

26

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Id jd

 

 

 

 

 

 

 

Figure 2.5 End view of a D0 central calorimeter electromagnetic module (CCEM).

27

W

U plate (3.0mm)\

I
LAr Gap (2.3mm)'-"")

[1

 

 

 

cl

 

 

Signal Board (1.09mm)’y [

 

 

Readout Board (1.67mm)

 

 

           
 

.......-..-.--.-.-----‘~-----._~.---- .............

.\\\\\Qr\\\\ \\\\\\\\\\\\\j
'c\\\\\ \\\\\ \\\\\ .\\\\‘

 
 
      
 

Copper Pads ,0 10

HV apply on Carbon-loaded epoxy coating

WM
Copper Trace A \\‘ G 10

 

Figure 2.6 DO central calorimeter gap structure

28

 

Figure 2.7 End view of a D0 central calorimeter fine hadronic module (CCFH).

29

 

 

 

 

 

 

Figure 2 8 End view of a D0 central calorimeter coarse hadronic module (CCCH).

Chapter 3
D0 Calorimeter electronics

3.1 ' General information

The use of liquid Argon as a unity gain ionization chamber to sample EM and HD shower
development offers unique advantages in maintaining the calibration of the energy
response of the calorimeter. At the same time one is faced with having to measure the
signals over a large dynamic range. Furthermore, the high interaction rate expected at
Tevatron requires new solutions to avoid pile-up in measuring the correct signal
amplitudes. The output signals are best measured in electron charges "e", and range
typically from 50,000 to 500,000 e's for muons in various parts of the calorimeter and up
to 6x108 e's for 500 GeV electromagnetic or hadronic showers. If an accuracy of 10% is
desired for single muon signals, a dynamic range of 105 is required. This dynamic range
can be reduced by a factor of 5 by judicious longitudinal segmentation.

The D0 calorimeter parameters arem:
Electrons collccwd lMeV of energy in liquid Argon: 9450
Fraction of energy in liquid Argon in the EM calorimeter: 0.12
Fraction of energy in liquid Argon in the FH calorimeter: 0.07

EM calorimeter resolution: 0. 15/ JE-
Hadron calorimeter resolution: 0.50/ «IE-
Total number of readout channels: 50,000

30

31

3.2 The D0 calorimeter precision read out system

The D0 calorimeter electronics system is made of three major subsystems (Figure 3.1):
the preamplifier, the base line subtracter (BLS), and the analog to digital converter
(ADC). The preamplifier is a high-gain charge-sensitive amplifier, which converts the
input charge to voltage. Figure 3.3a shows the circuit of the D0 preamplifier. The BLS
performs the following functions. It takes a sample of the preamplifier output voltage
right before the p - ['1 interaction (this is called the base sample), and it takes another
sample of the preamplifier output voltage 2.2 us after the p - 5 interaction (this is called
the peak sample). Then it subtracts the base from the peak, and stores the voltage in an
analog buffer. Finally the two samplers are free to take samples of the next p- p
interaction, and the voltage in the analog buffer is available for the ADC to digitize. The
sampling time of 2.2 as is chosen to optimize the noise and speed required for the 3.5 [is
Tevatron p - p collision cycle. Another feature of the BLS is the fast trigger signal
pickup circuit. It provides fast trigger signals (with shaping constant of 250 ns) for the
level-one calorimeter trigger. Figure 3.3a shows the BLS signal shaping circuit and
trigger pickup circuit. Figure 3.4 shows the SPICE simulation of the signal at the
preamplifier input, the signal at the precision BLS samplers, the signal at the trigger
pickup output, and the signal at the trigger flash ADC (FADC) input. The ADC uses 12
bits and two different amplifiers to provide a 15 bit dynamic range. References [1], [2]

and [3] provide detailed descriptions of the D0 calorimeter electronic design.

Another subsystem integrated within the D0 calorimeter precision electronics is the
precision calibration pulser system. Because of the large number of calorimeter channels,
"off the shelf" components were used for the preamplifiers in order to keep the cost
down. For example, the feedback capacitors have a variation of a few percent, the
transistors used in the preamplifier have a beta value only guaranteed by the supplier to

the range of 200 to 400, and the resistors used in the preamplifier are one percent

32

precision resistors. All these factors cause the gain of the preamplifiers to have a
variation of a few percent. In order to achieve a gain uniformity of less than one percent,
a precision calibration pulser system was built in the D0 calorimeter electronics. The
pulser calibration system has a precision pulse generator that generates a precision
pseudo-Gaussian shaped current pulse. It injects charge into each preamplifier input
through a precision resistor (0.1%). The channels into which the pulser injects charge
are chosen by a computer. The responses of the calorimeter electronics are read out and
analyzed in order to determine the gain of each preamplifier channel. This calibration

system enables us to achieve a uniformity of 0.3% for the precision calorimeter readout.
3 . 3 Level-one calorimeter trigger electronics

3 . 3. 1 Introduction

The DO level-one calorimeter triggerllél is designed to be able to trigger on every p - p’
collision without introducing any dead time. In order for the other detector front-end
electronic systems to be able to latch their data, the level-one calorimeter trigger decision
must be made within 900 as after each p — p collision (not counting cable delay).
Instead of waiting to collect all the charge in the calorimeter liquid Argon gaps (which
takes over 400 ns), a special trigger signal path is designed to provide a fast signal for the
calorimeter trigger electronics. It uses a differentiator with a shaping constant of 250 ns,
applied to the preamplifier output signal, to get the calorimeter energy information 250 as
after the particles hit the calorimeter. The simulated signals are shown in Figure 3.4.

In order to efficiently process the information coming out of the highly segmented
calorimeter, signals from some calorimeter readout 12113 are grouped together to form a
calorimeter "trigger tower". There are two [kinds of trigger towers: the electromagnetic
(EM) tower that consists of the sum of all the layers in the EM calorimeter in two units of

n and 1]) segments, and the hadronic (HD) tower that consists of the sum of all the layers

33

in the fine hadronic (FH) calorimeter (Uranium hadronic calorimeter) in two units of n
and 4) segments. All the channels needed to form a trigger tower (both EM and HD) are
contained on one BLS card. This scheme reduces the fast calorimeter trigger electronics
channels to 2560 from the 50,000 individual preamplifier channels, while still retaining
projective towers with adequate segmentation. The final calorimeter trigger input signals

are 1280 EM and HD trigger towers that have the size of 0.2x0.2 in n and 4) dimensions.

Figure 3.2 shows a conceptual schematic of the D0 level-one calorimeter trigger analog
signal path. The key components which affect the gain of the level-one calorimeter
trigger include the detector readout pad (Cd), the cable connecting the detector readout
pads with the preamplifiers (T1), the charge sensitive preamplifier with integrating
capacitor Cf, the cable connecting preamplifiers with the BLS's (T 2), the trigger pickup
differentiator made of C; and R;, the first trigger tower adder that adds all the layers of
EM or HD calorimeter and the adjacent 41 units, the second trigger tower adder that adds
the adjacent 11 units, the trigger signal differential driver (TrgDrv) that drives the long
cable carrying the signals of the EM and HD trigger towers to the calorimeter trigger
electronics, the calorimeter trigger analog receiver, and the flaSh ADC (FADC). The
digitized data passes through the digital part of the calorimeter trigger electronics to make
the final trigger decision. In this dissertation, I will only discuss the analog components
of the calorimeter trigger electronics, which are directly related to the measurements. The

digital part of the calorimeter trigger electronics is described in [16].
3 . 2 . 2 Difficulty of fast shaping.

Without specific input impedance matching for each preamplifier, the gain of the trigger
differentiator is sensitive to the variation of the preamplifier input impedancem]. It is
necessary to match each channel's gain before they are summed into trigger towers.

Trigger summing resistors Ri and R11 are chosen to correct the gain differences of each

 

 

 

34

channel. In the following sections, we will show how we determine the relation of

trigger gain to input impedance, and some tests we have done to verify this relation.
3 . 2.3 SPICEI7] simulation of level one calorimeter trigger electronics

In order to understand the input impedance effects on the gain of the trigger differentiator,
a detailed simulation of the level-one calorimeter trigger analog circuit (Figure 3.2 and
3.3) is done with SPICE3C1I7] (an electronics circuit simulation program) using
measured parameters of the active elements. The input impedance of the preamplifier is
determined by T1 and Cd. Cf and T2 also affect the trigger differentiator's gain. The total

input capacitance of the preamplifier (Cin) is the sum of Cd and the capacitance of T1. In

the D0 detector system, there are only two kinds of feedback capacitors Cf: 5.5pf and
10.5pf. Cd varies from 1.0 nf to 7.0 nf. T1 is a 50 Q coaxial cable and its length varies

from 10 to 20 feet (it is fixed at 18 feet for the test beam). I used the SPICE simulation to

study the relation of the trigger differentiator’s gain with each of the elements that affect

the input impedance.
a) The Gain vs. Cm:

As shown in Figure 3.5, the gain of the trigger differentiator has a strong dependence on
the Can and Cf. In this figure, I use fixed T1=18 feet, T2=180 feet and two kinds of Cf
(5.5 pf and 10.5 pf), By varying Cd, I can get different Ci“. Under these conditions, the

relation between the trigger differentiator’s gain and input capacitance is:

G,_,,, =0.4709-0.0627C,., +0.0311C; (3.1)
0,0,, = 0.4535 + 0.01575C,, - 0.02247C; + 0. 003370,: - 0.000157c; (3.2)

I will show how to use these Formulas to correct the gain Cin dependence.

b) The Gain vs. Lin

35

I used the SPICE simulation to study the effect of Lin on the gain of the trigger
differentiator. The study shows that there is only a weak dependence. The gain will
change 5% per 1 pH of Lin. The total input inductance Lin is the sum of the contribution
of the pads and traces of the detector module and the cable T1. For the CC, the module
pads and traces make a very small contribution to the Lin (on the order of lOOnH); T1 is
the dominant contributor, its contribution is a fraction of pH. Therefore, only the
inductance of T1 needs to be considered in the trigger differentiator's gain correction. In
the case of the test beam, where the T1 cable had a fixed length, the inductance did not

affect the uniformity of the trigger gain.

c) Resistance effect

The input resistance is generally very small, and its effect on the trigger gain is negligible.
d) Preamplifier transistor's beta effect

Because of the long cable between preamplifiers and BLS electronics (180f at the test
beam), the beta value of the preamplifier output driving transistor Q3904 (Figure 3.3a)
will affect the signal quality at the BLS receiving end. Any pulse shape change will affect
the gain of the trigger differentiator. I have done a set of SPICE simulations to study the
preamplifier transistor's beta effect. Figure 3.6a shows the calorimeter trigger response
(TRG) and precision electronics response (BLS) vs. the transistor's beta values for a
fixed size triangle pulse with different input capacitance. The beta effect on the BLS is
very small, but the effect on the TRG is large, and the effect is dependent on the input
capacitance of the preamplifier. The manufacturer guarantees that the transistor's beta
value is in the range of 200 to 400. With this limit and the SPICE simulation results, I
can estimate that the limits on the trigger gain variation caused by the preamplifier

transistor's beta are :l:5% for the EM channels, and i8% for the HD channels.

36
Assuming a uniform distribution of the beta values, the variation of trigger gain is

vanatr/qg 11m1t , which is about 3% for the EM channels and 5% for the HD channels.

 

e) Trigger driver transistor's beta effect

The trigger cable drivers (TrgDrv) (Figure 3.3b) use the same kind of transistor as the
preamplifier. It is expected that these transistors' beta values will have some effect on the
trigger cable driver's gain. Figure 3.6b shows the SPICE simulation of the trigger cable
driver's gain vs. beta value of one transistor. In this simulation the beta of other
transistors was set to the measured value; only the beta value of transistor 2N3906 was
varied. The trigger cable driver transistor's beta effect is less than 2%, with beta changes
between 200 to 400. This is a much smaller effect than the preamplifier driving
transistor. This is because the fast shaping of the trigger differentiator is more sensitive

to the signal shape than the calorimeter trigger receiver.
3 . 2 . 4 NYU test bench measurement

To test the SPICE model, I measured the Gain vs. Cin relation on the D0 front-end
electronics test bench (Figure 3.7) at NYU (New York University). The test was done
by using a triangle shape pulser, which mimics the charge pulse shape from the real
beam, to inject charge on a capacitor, which was connected to the D0 preamplifier with
the standard D0 T1 cable. By changing the capacitor, I measured the Gain vs. Cin
relation. The test bench measurement and the SPICE simulation are shown in Figure
3.8a. The SPICE results are normalized to match the data at the Cin=3.2nf point. The

maximum deviation between the simulation and the measurement is less than 2%.

37

3.2.5 Single channel measurement

To further test the SPICE model, I used our test beam facility to measure the trigger
response with only a single preamplifier connected to the trigger tower sum. This
insured that the trigger signal came from a preamplifier with known Ci“. The
measurement and SPICE simulation are shown in Figure 3.8b. The data points include
different preamplifier channels with different Cm, Cf and different trigger cable drivers
(EM and HD towers). The SPICE simulation was done with each channel's nominal
specification, and was normalized to match the data at Cin=2.5nf point. The figure
shows that the data of the EM channels (channels with Cin < 3 nf) and simulation agree
well, and the data of two HD channels have about 6% deviation from the simulation.
From the analysis of the preamplifier’s transistor beta effect (Section 3.2.3d), the gain of
the HD channels could have a variation of 5% due to the transistor beta variation.
Furthermore, the single channel test has another limitation, which is the trigger signal
crosstalk between preamplifier channels on the same BLS card (Section 5.2.3). Because
of this crosstalk, the Trg/BLS ratio of the single preamplifier channel is sensitive to how
much energy is in the nearby cells. These factors limit the precision of the single channel

test.
3.2.6 Summary of trigger differentiator's gain correction

Mth a SPICE simulation, the trigger gain with different input impedance can be predicted
to a precision of 2%. But the trigger gain has a variation caused by the individual
amplifier transistor‘ s beta value variation. The trigger gain variation will be investigated

in Chapter 5 by using the test beam data.

On a special note, the precision calibration pulser system is not very useful for the trigger

gain calibration, because the calibration pulser uses a pseudo-Gaussian shaped pulse

38

(Figure 3.9), which is chosen to minimize the systematic error from the cables and print
circuit traces of the precision readout system. The charge pulse from the beam is a sharp
triangle shape (Figure 3.9). Also, the calibration pulser injects the charge pulse at the
input of the preamplifier, whereas the charge pulse caused by the particle showers is
injected at the detector pad. Due to these differences between the calibration pulse and the
beam charge pulse, the trigger differentiator has a different response to these two kinds of
charge pulses. The calibration pulser system can not be used to do the quantitative study
of the trigger gain. But it is still useful for checking dead channels and doing qualitative
studies (like crosstalk).

3.4 The test beam trigger summing resistor selection

For the 1991-1992 test beam, we have CC modules with fixed length T1 (18 feet) and T2
(180 feet) cables. I used the SPICE simulation with these cable parameters to get the
Gain vs. Cm relation (G(Cin,Cf)) for the test beam. This way, the cable effects on the
input impedance are included in the formula. Formulas 3.1 and 3.2 are shown in Figure
3.5.

In order to choose the R; to correct the energy scale of each individual preamplifier
channel before the first adder, a few other factors that affect the energy scales of
individual channels need to be included These factors are the sampling fractions and the
gain of the trigger tower adder.

a) The sampling fractions and feedback capacitors used in calculating the CC trigger

summing resistors are shown in Table 3.1.

39

 

 

 

 

 

 

 

 

 

 

 

EM layer 1 12.7 5 .5
EM layer 2 12.7 5.5
EM layer 3 (all sections) 12.7 10.5
EM layer 4 12.7 10.5
FH layer 1 7.0 5.5
FH layer 2 7.0 5.5
FH layer 3 7.0 5.5
Table 3.1

b) Because the trigger tower adder has a finite open loop gain A (a few x102 instead of a
few x105), its gain not only depends on the summing resistor and feed back resistor, but
also has a weak dependence on the total input resistance, that is, the total parallel
resistance of all the summing resistors. A trigger tower adder adds 6 to 14 channels at
once. The total input resistance due to this parallel resistance is much smaller than the
resistance of the individual summing resistor. Figure 3.10 shows the measured relation
(H(Rpm-)) that is the adder gain normalized to Ra’Ri (ratio of feed back resistor and sum
resistor) vs. total parallel input resistance (kw).

An adder with a finite open loop gain, A, and feedback resistance, Rf, is shown in the

following figure:

 

If A ~ few x102 as in the case of the trigger adder, then we have

and we get

h=+efl.
Vin (1+_£_) Rt.
AR.

' )

The expected relation between the adder's gain H normalized to (Rf/RD, and the total

input resistance, Rm, is:

H=——-—.
Rf
1+
A'l'Rin

The adder we used has Rf= 3.0 K9, and if we fit the data in Figure 3.10 with Formula

(3.1)

 

3.1, we get

H(R 1

where Rpm is the total input resistance calculated from the parallel input resistors R].

c = 13.2 (3-2)

The trigger adder's gain will be corrected according to Formula 3.2.
3 . 4 . l The trigger summing resistor (R3) calculation

For energy E deposited in the calorimeter section, the voltage at the output of the
preamplifier is
v _ M (3 3)

m C! r

where e = 1.6*10"’ (Coulombs) is the electron charge, and

41
Na = 9.451t106 (electrons/GeV) is number of free electrons created per GeV energy
deposited in liquid Argon through the minimum ionizing process.

Va"! = V *G(CwC/)* 3.0KQ*

rm

 

H(R,._,,,) (3.4)
is the output voltage of the first adder, and

3.0K!)
Vedder2=vedderl* R *H(Ra_p) (3-5)

is the output voltage of the second adder, where RUN and Raw are the total parallel
resistance of the first and second trigger adders.

 

The trigger differential driver turns Vamp,"2 into a differential signal V+ and V_ with

Vaduz = V+ +V-.

For the test beam, the goal is to measure the energy of the incident particle (not the
transverse energy as in D0), therefore, all the tenninators of the calorimeter trigger analog

receivers were set to have a constant attenuation factor of 2. The voltage at the trigger

FADC input is: me. = sz-

Iproceed by assuming ideal response of the adders ( 11(12”): 1 ), R1; = 3.0 K9, and a

128 GeV dynamic range for the 8 bit FADC being equivalent to 2V on the FADC input

 

 

e*N *128*Sf 3.0m 3.0m
W: J *G C.,C *
c, l“ f)‘ R, 3.0m
R, = 290,304t-21*G(C,,,c,) (3.6)
I

I have assumed that the energy deposited in the calorimeter is through the minimum
ionizing process. In reality, since the D0 calorimeter has an e/mip ratio of 0.7, the
electron signal is only 70% of the signal I estimated above. The above calculation of R;
will give us a true dynamic range of 180 GeV for the electron signal.

42
Once the R; are determined, 1 can calculate the total parallel resistance (R; _par) of the first

adder input resistors R; and H(R;_pa,). The first adder's gain change due to the small

total parallel input resistance can be compensated for at the second adder to a good
R 1

_il'_.=——. For most of the tri er
3.0m H(R,_,,) gg

approximation by choosing R;; as

channels, R; 4,3, is approximately 50 Q. The R;; will be around 2.4 KS). The second
adder will have a parallel input resistance > 1 K9. From Formula 3.2, the change of the
second adder's gain due to the total input parallel resistance is negligible.

3.4.2 Selection of the load-two test beam trigger summing resistor

I used Formula 3.6 and the measured Cu; to calculate the proper trigger summing
resistors R; of the load-two test beam configuration. The distribution of R; is shown in
Figure 3.11. The Figure shows that the same detector layer has approximately the same
R; value. I chose resistor values as the most probable value of the distribution, and uses
only one kind of resistor for each detector layer. The values I chose for each calorimeter
layers are listed in Table 3.2. The calculated resistor values for the PH layers have the

largest variation, which is about 4%.

 

 

 

 

 

 

EMlaycr1&2 10519
EMlaycr3 6659
EMlaW‘l 509::
FH layer 1, 2 and 3 357 9

Table 3.2

43
After we have chosen the R; resistors, we can use Formula 3.2 to calculate the R;; values

to compensate for the gain of the first adder. The result is: R;; = 2.43 K0.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

45

use—=05» Smut. deflate—«U Nd 233m

    

 

 

B36 938 Bunt- v:-
boi name. 288.

 

as! our; .2:

 

xm —
6332 «223 .535 antenna—a0
"D"
a com U_ _
.l
Em _

 

 

 

 

 

 

 

 

 

   

Us: doom. 5 o»

.. L

(DU

 

 

 

x m
._- lu I: .-
his»...
3 396 z
and—52m
5.3—8.5:... Smut...

can! nos—es: 3: op

:55. 3.... eoueaau

I
3 3.5 one J

Moor. ®

 

 

3001'

 

 

H81

 

 

 

r

 

 

 

 

 

 

 

51
. l "—C -6v
r 0.22uF ,1 '
1 3 3 71‘ 15K}: Z‘K 0 UF 7‘ T 6.7“: ~
--————---—1q -———-—-—o——————<:: '3‘)
1r

 

 

 

 

 

 

 

 

 

 

 

 

"7 0.22141:
INPUT ‘ ._.]"l zsxsssv
weorooor n
59F x m—c ou‘rpur
... LK—’ J
022 F H10 022 F 390‘ 1..
. U
15K ‘SKSE , 15K 15K; u .FTG Go
I
2.2K ] "1" 0.10; 71-: .l 4.7uF
: "'4 6V
or
D0 calorimeter preamp (5p? version)
.13)!
0.1 ”F 5.0K
.IJV
4.0K
‘EIOK
10x 3
D—W—b
1
3391'
7.5x a: new R
' 22095 —-]\
4.3x > W
W; _fi If f
TL074 ‘
53100“ 7: saw EEtSOK
1
1
GD :

 

 

 

 

‘

 

 

BLS signal shaping and trigger pidtcfl'

Figure 3.3a .

47

 

 

 

. .- —— l—GED ... . -
01
hour I: u - an! L ‘pn
7“ ipiien 1:| 3‘ flj I JCurL
¥m ‘ v
m
9111

 

 

 

9: First triggeraddel- L——e———GED ... was

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

E1". (1 in PL. :1, J'L on
a" (e m) g”
- fi (e uv) “ m
l h "-1 .2. >=—.
n... I. 1.1! F“ ’ (- r“) y M
u :91. 2
m 1m- ] ‘Eglh (1 ~ u. l 1 . . a,"
é l 0-
Second trigger adder and trigger cable driver I. " "
Cm
ream m ,5 ”l
- .fl .1 H3 .3 a 4‘ ICJil'Lzl 13?} l 2.13.3;1
- m N L I 3.0!
'- " osv u
v
TRIGGER TOWER —L]\L
SlGNAL FROM vat“ ”2" I); “L
THE BLS *l/f 1....
‘ J '5‘ AH i330 W
- C”) I! In C”
. I 1 lgl
- . J H.” ’ a n high—5% mur i"
«am eel-L ‘3‘" c1-
11 L
" 0411'!
Calorimeter trigger front-end receiver nix
. I” 11 5
«.31 v u! __‘«~,,.l___
1.00 I ‘7

 

Figure 3.3b

Pulse Amplitude

48

 

 

 

 

 

 

 

b
-02 1 llLlllnLillllLilllglllL1411
O

 

 

0 1.010‘6 2.010'6
Tlme(10'6s)

Figure 3.4 SPICE simulation of D0 electronics pulse shapes

3.0 10‘6

Voltage af FADC input

0.45

.0
ts

.0
m
tn

.0
on

0.25

0.2

0.15

49

 

YTTTTYijffrITVYYIVTVV‘TYFTIYTTTITT

   
    
 
    
 
 
 

 

—e—- Cf-5.Sp(
—e— cr-lo.spr

IUUU

  
 
  

 

 

 

Y-MO+M1': 4... mafia” ’

 

 
 

0013757391613
002247142le
1 00033721220717 I

I . ~0.00015728287249 I

 

 

Efi'é'

i

 

Y- M0 4 Ml'x 4 m-x' . M9
M0[ 047091391211

 

 

M1 | -0.062743108247
00031144053432

 

LAAIALLAIAAlLJlllllAlllllJLl

 

IFIIIYIUIIUVfi'UIITYUIUUI'

 

 

Llll‘LlLlllllllllllllllLLllllllllLLLLlL

 

1 2 3 4 5 6 7 8 9
Cin (nf)

Figure 3.5 Trigger FADC input voltage vs. Input capacitance

Relative Gain

Relative Gain

1.1

0.9

0.8

0.7

0.6

1.04-

1.03

1.02

1.01

0.99

0.98

0|
0

50

 

 

 

 

 

 

 

_ l l l ' l ' l T l ' I ' T

P . Cl
.. ‘ .,
- ‘l
_- A A I Q <
_ C] I
I . l
i C] ' I
Z 0 Trg (Cin=5.5nf) :
L- O 81.3 (Cin=5.5nf) “
- ‘ & Trg (Cinal.5nf) <
. A BLS (Gin-1.5M) .
T '1
L- -l
P l l l 1 l l l l 1 .l l l l l l l J 1 l l I l l l l l l l I l l l l l l l

 

100 150 200 250 300 350 400 450
Beta

Figure 3.6a SPICE calculation of preamplifier beta effect,
the differences are normalized to the beta=300 point

 

q

IY—TTITTITTTTIIITTj—rIFIrIITTTTTTTTrITTT

 

0 Trgdrv

 

 

 

All111AALAILLllllllJllllllll

UUUI'UVYIIUTTIIUIU1IUUT'IIt

 

 

Llll[AlllllllllLllllllLllLllJlLLlLlllll

50 100 150 200 250 300 350 400 450
Beta

 

Figure 3.6b SPICE calculation of trigger driver beta effect,
The differences are normalized to the beta = 300 point.

 

 

DO preamplifiers

 

 

T2

 

 

 

D0 BLS system

51

Cd \ Inject charge pulse

Trigger cable

 

 

 

 

Oscilloscope

DO calorimeter trigger
front-end electronics

 

 

 

 

 

 

 

—~
—~

 

 

 

Figure 3.7 New York University DO calorimeter electronics test bench.
We use an oscilloscope to measure the trigger and BLS responses to a
fixed size triangle charge pulse.

0.45

0.4

Trg/BLS

0.3

0.25

0.2

2.5

52

 

 

TIUTlTfiI‘IU‘ITrT‘ITIIIV

.0

I I I I

r f I I T I

 

 

royals. NYU lest
SPICE simulation

 

 

.L l l l l 1 L4 1 l 1 L L L l l

T r I I

l L L i I 1 1 1 l 1 L 1 l

LJIIJLJLIJLALIAILILALIL

 

 

N

3

Figure 3.8a NYU test bench data

4

5
Cin (nf)

6

7

(D

 

 

 

 

 

 

W l I r F T r '7' U U—T T I l I I I T 1
. 4
" O Trg/BLS,TBsignelCI-1data ‘
- 9 o SPICEsimuletion .
Z 8 :
. 4
- .1
I .
I e I
00 .

P e
T e 1
D o o d
. J
'..L;.1L.l... 1 L1...‘

 

 

1.5

2

2. 5
Cin (nf)

3

Figure 3.8b Load one single channel trigger test data

e.

Voltage

-10

-12

-14

53

 

fit
A

 

v W 1 I
l

 

 

 

 

- E . . l -4
H / l
i" 1
1 I
a» —<
4 4
- 4
p 1 L 1 1 1 1 l l L l l 1 1 1* L 1 k L
0 200 400 600 800 1000
Time(ns)

Flgure 3.9 Calorimeter beam pulse and calibration pulserpulse shape

54

l l
1 I ;

TU

—-<l—
—.;
l—LL-L .-

. (armyé up crew) . ; 3
'0.9
0.85
0.8

 

C n 13.2 from lit

SWEET
,4. .
ghgfg
./

 

 

 

Relative Gain

 

[11" I'll TTVY 15" 11" III !

 

 

 

 

 

0.75 3 1

/ °§ rg :
0.7 , X2 RG4 3
05518 +; r465“:—
0.6 : 2 1

 

 

N
O

40 60 80 100 120
R - parallel (ohms)

Figure 3.10 Trigger adder gain vs. toral parallel input resistance.

55

 

 

 

 

 

 

 

 

 

*-
250 .—
240 '—
- EM36650
200 :- /
;
160 -
120 :-
2..
80 f.
_ UnusedCHs 1413570 EM1&21°519
. /L x ammo /
‘p _ \
ohlhhfllll L L11 ALLLJLL4L11 LgLLaitlgt
O 200 4100 6C!) 800 1000 1 200 1 4'00
Resistance (9)

Figure 3.11 Distribution of the load-two trigger summing resistor

Chapter 4
D0 test beam experiment

4.0 General information, history of test beam experiment.

In order to verify the D0 calorimeter performance, test the electronics system, and
provide the calibration for the collider experiment, there is a test beam facility to expose
the D0 calorimeter modules to high energy electron and pion beams. The electronics
system for the test beam is a copy of the D0 electronics system. Figure 4.1 shows the
test beam cryostat, which provides the cryogenic container for the liquid Argon, as a
large vessel on a transporter, which can move and rotate in three dimensions. The test
modules are installed in the cryostat. By moving and rotating the transporter, we can
expose different parts of the detector modules to the beam. There is a cryogenic feed-
through port on the top of the cryostat, which provides a path for the electrical signals
from the calorimeter module cells to the preamplifiers. The preamplifiers are installed
on the upper comer of the cryostat. This layout minimizes the cable length between
detector cells and preamplifiers. The electrical signal path from the detector cells to the

ADC and the data acquisition (DAQ) system is shown in Figure 3.1.

There have been three test beam runs to test the D0 calorimeter module design and
performance. The basic beam line instrumentation is the same for all these test beam
runs, and it will be described in the following sections. This dissertation is a result of

the two most recent test beam runs (1990-1992 load-one and load-two). Figure 4.4

56

57

shows the relative D0 geometric coverage of these two test beam runs. We will show

briefly the module configurations of these two runs.
4.1 Beam line magnets and target wheels

The D0 test beam experiments use the Fennilab's Neutrino West (NW) secondary beam
line. The layout of the NW beam line is shown in Figures 4.2 and 4.3. The primary
target is in enclosure NW4, and the secondary beam is transported to experimental hall
(NWA) through horizontal (east/west) bends in NW6, NW7, and NW9, with focusing
elements in NW4 and NW8. To create a relatively pure electron beam, a sweeping
magnet NW4S is used to remove all the charged particles, in conjunction with the
insertion downstream of NW4 Lead target wheel NW4PB to convert photons.
Alternatively, a relatively pure pion beam can be obtained by disabling the sweeping
magnet and removing NW 4PB target, and inserting Lead converters in NW6 (NW6PB)
and/or NW7 (NW7PB) to preferentially scatter electrons from the beam. The final
bend magnet NW9E is instrumented with a Hall probe to measure the magnetic field

during the beam spill. This gives us the measurement of the beam momentum spread
4.2 Beam trigger, veto and tags

The beam trigger signal is made of a triple coincidence of scintillators, SI, 82 and $3.
The scintillator array has a 5"x5" hole in the center to serve as a halo veto. There are
several other vetoes to inhibit trigger pulses. One is the calorimeter MOVING veto,
which vetoes any trigger if the transporter is moving. It limits any electrical and
mechanical noise that could be generated with such movement. Another is a DEAD-
TIME veto, which holds any further triggers for 6 us if a beam trigger fired. It is used
to avoid multiple Base/Peak sampling. If the event is accepted, there is a 10 ms dead
time veto to stop any action, and wait for the DAQ system to finish reading out all the

data. Along with all the vetoes, there are several tags to monitor the beam (Figure 4.3).

58
The Mips tag is located directly behind the cryostat to tag particle leakage, and the

Muon tag is located behind the steel absorber, to tag muons. Several Cerenkov
counters are recorded as Cerenkov tags for electron identification. These are two
helium Cerenkov counters (NW9CC, NWACCl), and a 3 meter long Nitrogen
Cerenkov counter (NWACC2).

4.3 Proportional wire chamber tracking and Beam momentum

calculation

Several Proportional wire chambers (PWC's) were installed in the beam line to
determine the beam particle trajectories on an event by event basis. Their positions are
shown in figure 4.3. The chambers were Fermilab standard "Fenker" chambers. Each
chamber had one or two planes of 128 wires. All of the chambers we used had 1 mm
wire spacing. There are total of 11 planes: 6 recording horizontal and 5 recording
vertical positions. The chambers were nearly centered on the beam line, and very small
adjustments were needed from beam alignment studies. Event by event momentum
calculation was based upon the bend angle, measured with PWC tracks, and the NW9B
magnetic field, measured from the Hall probe or by relating the NW9E read back
current to the field integral. The accuracy of the momentum is limited by the
measurement of the field integral BxL (B is the magnetic field and L is the length of
NW9E magnet), to about 0.2%; and the precision on an event basis is about 0.25% from
the PWC track resolutionlzz]. The RMS of the beam momentum spread is typically
1.5%, with reasonably Gaussian profilesm].

4.4 Test beam load-one configuration

The load one test beam included an end calorimeter electromagnetic (ECEM) module,
an end calorimeter inner hadronic (ECIH) module and an end calorimeter middle

hadronic (ECMH) module. The ECEM and ECIH modules were arranged in the

59
relative positions that they have in the D0 detector, and the ECMH module was placed

behind the beam hole of ECEM and ECIH modules. Figure 4.5 shows the layout of the
modules in the test beam cryostat. By moving and rotating the cryostat, we could send

beam to the different projective towers of the modules.
4.5 Test beam load-two configuration

The load-two test beam included four central calorimeter electromagnetic (CCEM)
modules, two central calorimeter fine hadronic (CCFH) modules, two central
calorimeter coarse hadronic (CCCH) modules, two ECMH and end calorimeter outer
hadronic (ECOH) modules, and massless gap and inter cryostat detector (1CD)
modules. These modules were arranged as they are in the DO calorimeter. Figure 4.6
shows how the central calorimeter modules were positioned in the test beam cryostat.
This arrangement makes up one eighth of the DO central calorimeter (Figure 2.4) and
the CC/EC transition region. Figure 4.7 shows the (t1, 4)) mapping of this test
configuration with the DO calorimeter (n, ‘1’) coordinates, where n is in pseudo-rapidity
unit, and ‘9 is in the D0 index unit (it is 10%cp, ‘1) is the azimuthal angle in radian). I
will use the same (1], 4)) unit throughout this dissertation. For this study, we only use
the data taken with the central calorimeter, shown in Figure 4.7 as shaded regions, and

only the region marked as "trigger coverage" was instrumented with calorimeter trigger

electronics. Two points where the energy scan data were taken are marked on the map.

 

61

.................... 800 GeV protons
Al forge!

MWS' Sweeper <d

NW4PB ------
MWW ...-

W40 - «-
NW4 V " ’

-------------------- Pions.electrons.others

NWGPB ---------
we: —>

NWW "'

---------------- fewer electrons
MWPB

............. f g
ewer e ectrons

ch \\

 

—-
NW9! p»
-.-
NWV —-
==
mm” ------- momentum selected
NWA

 

 

 

 

Figure 4.2 Fennilab NWA beam line

093.2% usafim 88m =2. 8 m ... 2am

 

63

 

 

LoadZ

 

 

Loadl

       

 

 

Intermion point.

Figure 4.4 II) calorimeter test beam scan region

ARGON EXCLUDER

ECEM

ECIH

LIQUID ARGON

 

SIEEL ABSORBER

  

52

 

 

 

 

 

CRYOSTAI SHELL

  

Figure 4.5 Load-one test beam module configuration.

65

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4.6 Load-two test beam module configuration.

STEEL DUMP
\'

\

 

_ oblfiofifi -

09:98 0:69: 23-98..— 56 223m

 

 

 

 

 

 

 

x2e:— Sm m. P N._. t 6.0 1° O 10- gal N. F-
: ._.. .4n______ ._..
I. _u I: l l .u L1
_ _ Ins _ _
+ T 092950 coma: / "I
I. _I / _ I_1
_ _ , Eu _
+ r m r T
_ _ IE _ L
I_ n . I.
c r em. .1
_ _ _ _
._. 1 . J!
I. _n . Eu -_1
_ _ _ _ _ _ .58 as. a; . It _ _
+ T l_I. + _ . _ I...
|__II_|_w___ _SWI_I
_ _ _ _ _ _ _ _ . _ _ _ .
p — _ _ h _ p _ _ _ - — - —

 

 

 

 

 

an
5%
mm
ha.
2..
2..
NM
n n
2..
mm
mm

:3:—
E.—

Chapter 5
Test beam measurement

We took a large amount of data fiom the two loads of calorimeter modules. But we did
not take useful data for the energy scan study with the trigger electronics when the load-
one data was taken. This dissertation is an analysis of the load-two data. In this
Chapter, I will present the load-two energy scan measurement with both trigger readout
(TRG) and precision readout (BLS). I will also include published results of the load-
one energy scan measurements for comparison. In order to understand the systematic
error of the TRG, special data were taken to study the gain variation of the TRG. I will

show the analysis of these data, and discuss the corrections I need to apply.
5.1 The general information of the load-two analysis.

In order to measure the e/rt ratio, precision energy measurements of both electrons and
pions are needed. The electron and the pion showers are intrinsically different. The
electron showers are physically small, and measured in radiation lengths. The pion
showers are relatively large, and measured in interaction lengths, which are usually 20
times larger than radiation lengths in most materials. The measurements for electrons
and pions are done with different regions of detector and electronics channels. To
precisely measure the e/rt ratio, we have to be able to understand the uniformity of the
detector response and correct for the energy leakage due to the large pion showers.
However, to compare the e/rt ratio difference between the fast trigger readout (TRG)

and precision readout (BLS), only the uncertainties of the electronics need to be

67

68

considered. Because both readout systems are connected to the same detector region,
the uncertainties caused by the detector defects are canceled out by taking the
difference of the e/rt ratios measured with the TRG and BLS. Therefore, we expect that
the e/rt ratio measurement will have an uncertainty dominated by the non-uniformity of
the detector response, and energy leakage, whereas, the e/rt ratio difference between the
BLS and TRG will be insensitive to these factors. The uncertainty of the e/rt ratio
difference will be dominated by the uncertainty in the trigger electronics, since the

uncertainty of the precision electronics is much smaller.

The physical reason for the e/rt ratio difference between the TRG and BLS is that there
is a slow signal component in the hadron shower in the Uranium-liquid Argon
calorimeter. As the electron shower is almost instantaneous (finished within a few ns),
the electron responses of the TRG and BLS should be the same. Using this principle, I
can calibrate the TRG with the BLS data. This eliminates most of the uncertainties in
the fast trigger electronics, such as the gain vs. capacitance correction and the gain
variation of the trigger towers. With calibrated TRG, any difference of the e/rt ratio
measurements with the TRG and BLS can be attributed to the slow signal component of
the pion showers in the Uranium-liquid Argon calorimeter. Figure 5.0 shows the
SPICE simulation of the BLS and TRG responses to electron and pion showers. The
electron shower generates a perfect triangular pulse at preamplifier input, shown with
dashed lines; whereas the pion shower generates a triangular pulse with a tail at the
preamplifier input, shown with solid lines. The input capacitance of the simulation is
2.0nf, and the total charge of the two input pulses is normalized. The Figure shows that
the BLS is not sensitive to the input pulse shapes. Its response is the same regardless of
whether the input pulses have a tail or not, whereas the TRG is different, its response
being very sensitive to the leading edge of the input pulse. There are a few corrections

that I need to apply to the data. These are described in detail in the following sections.

69

All the energy measurements shown in this dissertation are done with the following

methods:

(1). The BLS data are pedestal subtracted and gain corrected. The pedestal of each
channel is determined by analyzing the data taken without the beam, and the gain of
each channel is determined by analyzing the data taken with the precision calibration

pulser.

(2). All the beam events are required to have a good beam track (both X and Y), and
a beam momentum reconstruction. Cuts are applied to the beam track position (X, Y

cuts) on an event by event basis to avoid the cracks of the detector.

(3). The electron cluster is defined as the sum of 6x6 cells in the (11, ¢) dimension
(approximately 45x45 cm2 in the front of the CCEM). The BLS electron energy
measurement includes energies in all the CCEM layers and the first CCFH layer. The
TRG electron energy measurement includes the EM trigger towers and the leakage
energy in the first CCFH layer measured by the BLS. To minimize the systematic
uncertainty, the electron energy measurement for the e/1t ratio study uses the same

cluster as the pion's.

(4). The pion cluster is the sum of 10x8 cells in the (11, 4)) space (approximately
78x60 cm2 in the front of CCEM). The BLS pion energy measurement includes
energies in all the central calorimeter layers. The TRG pion energy measurement
includes energies in both EM and HD trigger towers and leakage energy in CCCH

measured by the BLS. There is a longitudinal leakage cut on the energy in the CCCH
(Ech) that is .ECL S 25%.

told

 

70

(5). The energy in different detector layers is weighted by :é- (Si is the sampling

fraction of the layer i) before it is summed together.
5.2 Corrections
5.2.1 Non-linearity of the trigger cable differential driver.

The fast trigger signals are summed into the trigger towers on the BLS cards, and sent
to the CTFE (calorimeter trigger front-end) cards through the trigger cable differential
drivers (TrgDrv) (shown in Figures 3.2 and 3.3). The TrgDrv is not a perfect linear
amplifier over the large dynamic range required. Using the standard D0 precision
calibration pulser, the linearity of the TrgDrv can be measured with the test beam
electronics. Only one preamplifier channel with small detector capacitance was added
to the trigger tower in order to limit the noise. The result is shown in-Figure 5.1a, and
can be fitted with a simple power law. From this power law relation, the following

correction can be applied to the raw trigger data to eliminate the non-linearity.

_ 0.930
£1364»an - EnG_m

Figure 5.1b shows the TRG/BLS ratio of electron energy scan data before and after this

correction. The simple correction improves the linearity to within :|:1%.

The uncertainty of the exponent in the correction is 10%, derived from the fit. I will

show how this uncertainty affects the e/rt ratio measurement. The following relations
apply among the raw data (X), corrected energy (E), non-linearity deviation ((1), and

uncertainties of these parameters (615,861).

E=X““ and 65--5aEln(X) a~2% %~ro%.

71

Since the e/rt ratio is close to one, if the electron and pion showers share the same
trigger towers, then any uncertainty of the e/rt ratio measurement caused by the
uncertainty of exponent correction is canceled. But in reality, electron showers are
mostly contained in one trigger tower and pion showers are spread over several trigger
towers. I assume a worst case scenario where the pion showers are spread evenly
among 10 trigger towers, the uncertainties of the exponent are the same for all trigger
towers. Here I assume that the exponent uncertainty is systematic rather than random.
Then the uncertainty of the e/1t ratio measurement caused by the uncertainty of the

exponent correction can be calculated as follows.

l-a 10 i F“
E, = x, , E, = 2(x,)

i=1

Since E, a E, and a «1, I have x, ~10x;.

Assuming the exponent correction has a deviation of 6a from the actual correction,

the e/rt ratio measurement caused by the exponent correction deviation can be

calculated

E. . 5' " 50‘5“ “(5') ~ 5(1 + 50:11:00)).

5‘ iii-55) 4.125145%} 5‘

ill

 

Thus, the uncertainty of the e/1t ratio measurement caused by the uncertainty of the

exponent correction is 10%aln(10) ~ 0.46% .
5.2.2 Sampling fraction correction

The sampling fraction of the detector is calculated using Formula 2.1, except for the
first layer of the CCEM. There is a large amount of extra material (the cryostat walls
are equivalent to about 2X0) in front of the CCEM module that needs to be included in

the sampling fraction calculation. The method we used was to include the front

72
material in the first EM layer sampling fraction. Because of the nonlinear characteristic

of the longitudinal distribution“) of the electron showers (the rapidly increasing energy
deposition per radiation length at early part of electron showers), the effect of the extra
front material on the electron energy measurement is nonlinear with energy also. From
an empirical studylzz], we found that by incorporating 60% of the front material into
the first CCEM layer sampling fraction calculation, we can have good linearity and
energy resolution of the electron response. The sampling fractions 1 used for the test

beam are calculated this way.

The sampling fractions I used to calculate the trigger summing resistors were calculated
initially without knowing the detailed geometry of the test beam modules and the front
materials. Since then, updated information of the modules and front materials has
become available. Table 5.1 shows both the sampling fractions used to calculate the
uigger summing resistors and the updated sampling fractions. Because the trigger
summing resistors are calculated with the imprecise sampling fractions, the energy from

the raw trigger data needs to be corrected. I will show how to do this correction.

Detector layer The trigger resistor Correct sampling

EMl 12.7% 9.04 l
EM2 1 11
ll
EM4 l2. 1
1 7
FHZ 7
' 7
l

.4

 

Table 5.1
I define two sets of sampling fraction weights [WiT] (W! = Elf, {Sf} is the sampling

fraction used to calculate the trigger summing resistors, the middle column of Table

 

1 Include the front materials.
2 Module only.

73

5.1) and {W1} (W, = E}, {S, } is the precision sampling fraction, the right hand column

of Table 5.1). For a high energy particle, the real energy deposited in the calorimeter
layer 1' is [e1] and the energy measured by the trigger readout in the calorimeter layeri

is [eff]. I assume that the energy measured by the precision readout (BLS) is the real

energy, and that the energy measured by the trigger readout (TRG) is very close to the
real energy. The TRG output which is actually recorded is Em, = ZWK-ef. The

precision readout measures (ei}, and I know the {W1} and (Win. I seek an expression

which approximates U136: 2W,- -e,T ,the trigger energy if the trigger weights had been
correct, using only know quantities. An approximate expression is
W. . e.
U z = WT . e? . ALL . .1
TRG Q 2 . I [Ewing] (5 )

The term in the parentheses gives the correction factor for using {WiT} instead of {Wi}
for precision readout {ei]. This factor should really give the correction for TRG
readout {eiT}, since the {eiT} and {q} are very close. Let W,’ = W, +a,, |a,|<< |W,|
and e, = e,’ + 8,; |£,| <<|e,|, then

T T
[L]-[Z—;r::)= “(”2 '1 L2; M

when the second order terms are dropped.

If 5- ~ 5% (from the e/rt ration difference measurement, Section 6.2) and 5,1- <10%

(which are mostly true, as in Table 5.1), the error of the Formula 5.1 is less than 0.5%.
Formula 5.1 will be used to correct the sampling fractions of the trigger readout.

74

5.2.3 Crosstalk between fast trigger channels

We have observed crosstalk between trigger channels on the same BLS card. Figure
5 .2 shows that if two channels have a small capacitive (Cc) coupling between them and
if the input resistance of the amplifier is small, there will be a crosstalk signal with a
relatively short time constantlzo]. Since the sampling time of the trigger readout is
much earlier than the precision readout (250ns vs. 2.2us), the trigger readout could
pick- up a large fraction of the crosstalk signal. Using pulser tests I have observed the
crosstalk signal at the input of the first trigger tower adder. So the crosstalk signals are
amplified by the nigger tower adders. Because the trigger response is sensitive to the
pulse shape and injection point, the pulser system cannot be used to study the crosstalk
quantity of the beam signal. The real beam data will be used to estimate the crosstalk

effect on the trigger energy measurement.

I used elecu'on energy scan data, and looked for "excess energy" in the HD section of a
trigger tower, with large amounts of energy deposited in the EM section of the same
trigger tower. The "excess energy" is the energy measured by the trigger readout minus
the energy measured by the precision readout. Since the crosstalk in the precision
readout is negligible, the energy measured by the precision readout is real energy
deposited by the electrons. In Figure 5.3, I plot the "excess energy" in the HD section
of a trigger tower vs. energy in the EM section of the same trigger tower. It shows that

there is a crosstalk of 2.2% signal from the EM section to the HD section.

Because the cross talk reshapes the signal, the crosstalk could effectively change the
trigger energy measurement. In order to estimate the crosstalk effect on the trigger
energy measurement, I will examine two exueme cases. One is that the t0tal signal is
conserved; that is, one channel's gain from crosstalk must be another channel's loss.

Another case is that the total signal is not conserved; that is, because of the crosstalk

75

reshaping the signal and the short trigger sampling time, one channel can pick up extra

signal without the Other channels losing any signal.

In the first case, if the amplification is the same between two crosstalking channels, the
total energy should be conserved. However, the sampling fraction differences between
the EM and the HD channels and the capacitance differences between these channels
cause the amplifications applied to the EM and the HD channels to be different. The
trigger summing resistors chosen for the trigger adder (Table 3.2) are: EMI & EM2 =
10519, EM3 = 6659, EM4 = 5099, and His = 3579. In the case of the crosstalk
picked up in the HD channels from the EM channels, the crosstalk signal is amplified
by a factor of 1.7, and for the crosstalk picked up in the EM channels from the HD
channels, the crosstalk signal is amplified by a factor 1/ 1.7. For the electrons, most of
the energy is deposited in the EM sections, and the test beam data (Figure 5.3) shows
that the crosstalk will give the HD section 2.2% excess energy. This means that the
crosstalk signal picked up before amplification by the trigger adders in the HD section
from the EM section is A£=2.2%E/l.7 = 1.3%E. The energy measured by the trigger
readout is E" =(E—A£)+A£*l.7= E +1%E, which is 1% higher than the real
energy. For the pions, the energy will be shared between the EM and the HD sections.
If I assume that the pion energy is evenly shared between the EM and HD sections, and
the crosstalk factor between the EM and the FH section is the same as measured in the
electron case, then the energy of the pions measured by the trigger readout will be

E,I =E+(1.7-1+1/1.7-1)-A€/2= E+0.2%E, which is 0.2% higher than the real

energy.

In the second case, where the two crosstalk channels do not conserve energy, the

crosstalk signal picked up in one channel will directly contribute to increase the total
energy. Following the above calculation, I get for electrons: Em = E + 2.2%]? , and

for pions: E," = E +1.5%E.

76

The real crosstalk mechanism will likely lie in between these two cases. I will correct
the trigger energy measurement by the average of these two cases; that is, 1.6% for the
electrons and 0.85% for the pions. I can estimate that in the worst case, the maximum

uncertainty on the e/rt ratio caused by the crosstalk effect is less than 0.65% and it is

caused by the uncertainty in the crosstalk mechanisms.
5.2.4 The gain variation of the trigger electronics

From the electronics circuit analysis (Section 3.2), we expect a gain variation at the
individual preamplifier channel level as well as at the trigger tower level due to the
variation in the transistor's beta. The trigger differentiator's gain variation caused by the
beta variation is about 3% for EM channels and 5% for HD channels. The trigger cable
driver's gain variation is less than 1%. These are worst case estimates, assuming a
uniform distribution of beta values. In the following section, I will use the test beam

data to study these variations.
1) The individual trigger tower gain variation

Ideally, the ratio of the trigger readout (TRG) and the precision readout (BLS) would be
a constant over the whole detector, because it is only electronics dependent. So I can
use this ratio to measure the variation of the trigger electronics, provided that the gain
variation of the BLS is negligible. I used 100 GeV electron trigger tower scan data
with beam directed at different trigger towers to measure the ratio of the TRG and BLS.
Figures 5.4a and 5.4b show this ratio distribution for 24 trigger towers with EM and
HD sections. The measurements on the HD section are done by selecting the events in
which electrons go through the cracks of the CCEM modules and deposit a significant
amount of energy in the CCFH modules. Because I only use one kind of trigger
summing resistor for the HD channels, and the HI) channels' capacitance changes

systematically over 11 and 0, I expect the HD channels' gain to have a systematic

77
dependence on the 11 and 0 of the trigger tower. I corrected this systematic (11, 0)

dependence by using FH layer one capacitance data of each tower to calculate a

correction factor for the tower.

The trigger tower scan data (Figure 5.4) show that the distribution is not very Gaussian,
so I will use the RMS to characterize the data. For the EM trigger towers the RMS of
the gain is 5.1%. For the HD trigger towers the RMS of the gain is 8.3%. The
statistical uncertainty of each EM trigger tower gain measurement is less than 1.0%,
and the statistical uncertainty of each HD trigger tower gain measurement is less than

2.0%.

These gain variations of the trigger towers are larger than expected. From the energy
resolution study (next Section) and the SPICE simulation of the beta effectlsmit’n 331,
we expect that the trigger tower gain variations caused by the electronics are less than
3% for the EM towers and 5% for the HD towers. These expected variations are
smaller than the ones shown in Figure 5.4. We found that the problem was in the
trigger cable driver (Figure 3.3). The output coupling capacitors (C7 & C8) of the
trigger cable drivers are defective. These capacitors are unstable, develop large leakage
currents, and have a short life time (months). We overlooked this problem during the
test beam experiment, but later, we had the opportunity to study the long term behavior
of these trigger cable drivers at DO. We observed that the trigger tower's gain changed
in the course of a few weeks, and that the trigger towers with significant gain change
will go totally dead, caused by the shorting of these coupling capacitors. The rate of
these failures is nearly a half percent per week. Normally, this type of capacitor has a
leakage current of a few to tens of M. Using SPICE simulation, we have seen more
than a 10% gain change if the leakage current increases to 2 mA. Based on what we

have observed at D0, we believe that some channels developed large leakage currents

78

which caused the larger than expected trigger tower gain variation observed at the test

beam. The trigger tower gain variation can be expressed as:

I,,,,,,=\[a",,,,.,,, +0“: +02 W,"

 

where am is the fractional gain variation due to the preamplifier transistor beta

variation (more discussion on this can be found in the following section). From the
SPICE simulation, it is about 3% for the EM channels and 5% for the HD channels.
0C”. is the gain variation due to the capacitance correction. After applying tower by
tower capacitance correction for HI) towers, this term is less titan 2% for both EM and
HD towers. 0W," is the gain variation due to the trigger cable driver, which we
believe is mostly caused by the defective capacitors. Using the data from Figure 5.4, I
can conclude that the EM trigger tower variation is dominated by GM", and the large
I-ID trigger tower variation is caused by a combination of the large om term and the

ow," term.

All of the test beam data I used for this dissertation were taken during a one and a half
week period. During this short period, the change of the trigger electronics gain due to
the change of the leakage currents of the coupling capacitors should be very small. I
can use the trigger tower scan data to correct the energy scan data to eliminate the large
trigger tower gain variation. Furthermore, the electron and pion energy scan data at the
same energy were taken consecutively, so any time dependence of the trigger

electronics gain will be canceled in the elrt ratio measurements.
2) Individual preamplifier channel gain variation.

In the previous section, I have discussed the trigger tower gain variation observed using
electron trigger tower scan data. From the SPICE simulation, I expect that the beta of
the prearnplifier’s output driver transistor and the capacitance of the feed back capacitor

79
will affect the trigger gain. But there is no direct measurement of these effects with the

trigger data, because the trigger tower is the sum of many preamplifier channels. I will

use the calorimeter energy resolution to estimate the gain variation due to these factors.

The method is the following. As mentioned early, the calorimeter fractional energy

resolution can be parametrized as

2 2 2
(3) =C’+-S—+-I-V— (5.1)

where o is the variation by fitting the energy distribution with a Gaussian, E is the
energy of the beam, C is the constant contribution from systematic errors such as gain
variation, S is due to the statistical error in the energy sampling, and N represents
energy independent contributions to a such as electronics noise. Assuming these
parameters to be independent, and fitting the resolution vs. energy relation with (5.1), I
can determine the constant term, which gives an upper limit of the gain variation. The
results of the fit on the electron and pion energy scan data of both the BLS and TRG are
shown in Figure 5.14. All the energy measurements are done by using the methods
described in section 5.1. The gain of individual trigger towers is corrected using the
100 GeV electron trigger tower scan data. This eliminates the contribution of the
trigger cable driver's gain variation. The sampling fraction correction (Section 5.2.2) is
applied to the trigger data. The noise term in the fit is fixed using the pedestal data
taken within the beam spill, and the sampling term for the trigger data is fixed to the
precision readout sampling term. The results of the fits are shown in Table 5.2.

80

 

 

 

 

 

 

 

 

 

 

Electron data BLS Electron data Pion data Pion data
TRG BLS TRG
c 0.33% (0.1%)1 1.52% (0.1%) 3.80% (02%) 4.59% (0.3%)
5( , /__GeV) 14.6% (0.2%) 14.6%? 45.5% (1.1%) 455952
N(Ge\03 0.44 1.3 1.7 2.7
12 1.9 2.8 4.0 1.9
Table 5.2

It can be seen that the constant term difference between BLS and TRG readout is rather
small. The larger constant term of the TRG is the result of the larger gain variation of
the TRG. In order to get a quantitative estimate of the gain variation from the constant
term of the resolution fit, one has to understand the relation between the gain variation

and energy resolution.

If I assume that the fractional gain variation is a and that the particle showers

r-ia’
distribute their energy uniformly among n cells, the variation of the energy

0' .
measurement due to gain variation will be £5- : 71:1. For different types of particles,
. n

n will be different. In order to understand how the gain variation contributes to the
energy resolution, it is necessary to know the energy distribution of the particle

showers.

The BLS data were used to study the effect of the gain variation on the energy
resolution, and estimate the effective it of the electron and pion showers. Assuming
that the BLS data is free of gain variation after gain correction using the calibration

pulser data, I artificially introduce gain variation to the BLS data by randomly

generating a set of gains with fractional variation of 0,... relative to the normal gain.

 

l’I‘henumherintheparentlreiresistheerrorfrornthefit.
2Fixedtotlreprecisiorrreadoutsamplingterm.
3Thenoisetermusesthemeasurednoisefromthepedestaldata.

81
Figure 5.5 shows the energy distributions of 100 GeV electrons and pions with artificial

gain variations of 10% and 20%.

For the electron shower, the average transverse shower size is much smaller than the
calorimeter cell size“) and longitudinal shower distribution is nearly constant. In the
CCEM the longitudinal fractional energy distribution is approximately 10%, 15%, 50%
and 25% among CCEM layers 1, 2, 3, and 4. So the majority of the energy from the
electron shower will be deposited in 2 to 3 cells. Because of the small number of cells
and the small fluctuation of the electron shower, the gain variation of cells will cause
the energy disuibution to deviate from a Gaussian distribution. This effect is shown in

Figure 5.5. I choose an effective n s 2.5 for the electron shower.

For the pion shower the situation is more complex, because the average pion transverse
shower size is larger than a few detector cells, the energy distribution is very uneven
and fluctuates greatly from event to event. The simple analysis of the electron shower
would not work for finding an effective n for the pion shower. The effect of the gain

variation on the energy resolution can be parametrized as:

 

a, = J01,“ + 2120’2 . (5.2)

where o,_m is the intrinsic detector pion resolution without gain variation, 0“,, is
the fractional gain variation , a“, is the pion energy resolution with the fractional gain

variation 0

and A is a parameter (A = 71:). Figure 5.6 shows a plot of the relative
n

resolution vs. relative gain variation for the 100 GeV pions. Fitting the plot with (5.2),

Iget n-S.

With the above model, I can extract gain variation information from the energy
resolution fit. For the electron energy scan, the BLS and TRG constant terms are 0.3%

and 1.5%. Thus the EM TRG channel's gain variation is smaller than

82
42.5*1.5% z 2.4%. For the pion energy scan, the TRG constant term is 4.6% and the

BLS constant term is 3.8%. Considering the gain variation as the only source of the

TRG's larger constant term‘, the gain variation of the HD TRG channels will be:

 

45%/0.0402 -O.0382 z 5.8%.

This result is close to the previous estimate using the limits of the transistor's beta value

and the SPICE simulation (Section 3.2.3).

In summary, using the test beam electron trigger tower scan data and the electron and
pion energy scan data, I am able to eliminate the large trigger tower gain variation due
to the defective trigger. cable drivers, and I am able to estimate that the individual
trigger channel's gain variation due to the preamplifier's beta variation is less than 2.4%

for the EM channels and 5.8% for the HD channels.
5.2.5 CCEM between module crack effect

All the CC modules have a region which is about 1cm wide on each side of the modules
with a weak elecuic field. This happens because the high voltage resistive coat on the
signal board does not extend all the way to the side of the module. Figures 2.5 and 2.6
show a CCEM module and the resistive coat structure. The electric field in the liquid
Argon outside the resistive coat covered area will drop very quickly as shown in Figure
5.7. The free electrons in these regions will drift with a slower speed over a longer
path. The slower drift speed and longer path will change the charge pulse shape on the
input of the amplifier as shown in Figure 5.7. This pulse shape change will cause the
fast shaping nigger signal to lose amplitude, as shown in Figure 5.0. This effect has
been observed with the electron data as shown in Figure 5.8, which is a plot of the
TRG/BLS ratio vs. Y (vertical) position of the beam. It shows that the ratio drops 50%

 

1Ihemamotherfacmwhichwouldcausedwuiggerenagyresoludoncmstmumtohelargerthan
theprecisionreadout,suchasalargerehrratioandextrasignallossduetothecrackeffect

83
when the electron is shot at the CCEM crack. The modules are arranged as in Figure

4.6, with the cracks running horizontally. Figure 5.8b confirms this arrangement. It
shows that the energy in the CCEM drops nearly to zero when the incident electrons are

aimed at the crack.

For the TRG electron energy measurement, the cracks have little effect because the
electron transverse shower size is much smaller than the CCEM module sizel41. If one
selects the electrons away from the cracks, there will be only a very small fraction of
energy (<1%) deposited in the crack region. However the pion transverse shower size
is larger than the CCEM or CCFH module sizes, so the trigger energy measurement
could lose signal if there is a significant amount of energy deposited in the CCEM and
CCFH crack regions.

It is difficult to make a good estimate of how much energy is deposited in the CCEM
and CCFH crack regions from pion data, but we can use Monte Carlo simulation to
study this effect. Figure 5.9 shows the average energy distribution of 50 GeV pion
showers in the azimuthal (0) dimension based on the plate level test beam Monte Carlo
simulation (Section 6.1). The figure includes the distribution of live energy (energy
deposited in liquid Argon) in the EM and HD sections, and the distribution of total
energy (energy deposited in all the materials) in the EM and HD sections. The inter-
module cracks show up as valleys in the live energy distributions. If I integrate the
total energy over the CCEM and CCFH crack regions, I find that the energy in the
crack regions is 1.5% of the total energy in the detector. Assuming that the TRG only
detects halfof the BLS signal, I will correct the TRG pion response by 0.7% due to the
crack effect. The uncertainty on this estimate depends on the precision of the detector
Monte Carlo simulation. We have done a transverse shower distribution comparison

between load-one data and the detector Monte Carlo simulation. The estimated

84
uncertainty of the simulation is less than 10%[22]. Therefore, the crack correction has

an uncertainty of less than 0.1%, which is negligible for the e/rt ratio study.

5.3 Load-two electron and pion energy scan results

With the load—two configuration, we have taken energy scan data with both electron
and pion beams at two positions of the CC (Figure 4.7). I will show the results of these
two energy scans. The results shown here include all the cuts and corrections described
in Sections 5.1 and 5.2, in addition to a pedestal correction, using the pedestal data we
collected during the beam spill. We discovered that there is a small difference between
the pedestals we get for calibration pedestal data and the pedestal data taken during the
beam spill (inspill pedestals). This is largely due to the narrow window size used to
calculate the calibration pedestal. Using the inspill pedestals to correct'this small
difference, the small error of the calibration pedestal can be eliminated. The TRG
measurement includes the non-linearity correction, the trigger tower gain correction, the
sampling fraction correction, the EM/HD crosstalk correction and the CC crack

correction.

The main purpose of this dissertation is the study of the e/rt ratio difference. No effort

is made to optimize the detector resolution for a certain type of particle. The cluster

size is chosen to minimize the systematic uncertainty of the e/rt ratio measurement. In

this Chapter, I will show the detector response to both electron and pion beams, and I

will summarize the e/rt ratio difference measurements with all the uncertainties and

corrections in the next Chapter.
5.3.1 The electron energy scan data

Figures 5.10 and 5.11 show electron energy scan linearity plots at (n=0.05, ¢=31.6) and
(11:0.45, 41:31.6), where (n, 4)) is in D0 index unit (Section 4.5). They include the

85

energy measurements using both electron and pion clusters. The trigger data is
normalized to fit in the same scale. The error includes the statistical errors of both
energy measurements and the inspill pedestal, correction. The linearity fits and residue
plots show that the detector generally has a linear response for electrons, but there are a

few points that need discussion.

The BLS data show little dependence on the cluster size. There is a small difference in
the electron response at the two data points. It could be caused by the non-uniformity
of the Uranium plates. It is known that the Uranium plate thickness variation is a few
percent and this could account for the variation of the electron response. This variation

has been studied with a large set of data specially taken for the uniformity studym].

The TRG data show that there is a small difference between energy measurements with
different cluster sizes, with the larger cluster (pion cluster) having higher response.
This effect is due to the crosstalk between trigger towers. I have measured and
corrected the crosstalk between the EM and HD channels in the same trigger tower, but
there are similar crosstalk mechanisms between trigger towers. The coupling
capacitance between trigger towers is smaller, so the crosstalk signal should be smaller.
The results show that it is about 1%. Since I will use the same cluster size for both
electron and pion energy measurements to calculate the e/n ratio, this effect should not

affect the e/rt ratio measurement.

There is a small offset in the linearity fits for both the BLS and TRG data. It is
equivalent to less than 300 MeV for the BLS data and 430 MeV for the TRG data. This
offset is due to the small non-linearity caused by the extra material in front of the
detector. Because of the nonlinear energy dependence of the longitudinal elecuon
shower distribution, the simple correction we made by changing the EM layer one

sampling fraction to compensate for this extra material still leaves us this small offset.

86

However, because this offset is small, its effect on the e/rt ratio measurement is

negligible.
5.3.2 The pion energy scan data

The pion energy scan data is shown in Figures 5.12 and 5.13. The detector response for
pions is generally linear. The pion linearity fits show that there are larger offsets .(~ 1
GeV) for pions than for electrons. The major reason for this larger offset is due to the
detector's nonlinear hadron response, especially at low energies (kinetic energy Bk < 5
GeV). Figures 6.5 and 6.6 show that the e/rt ratio increases as energy decreases, and
that this effect is more pronounced for low energy data (Ek< 5 GeV)[21]. If I assume
the detector response to electrons to be linear, then the detector response to pions must
have a nonlinear relation to the pion energy. The response varies between two data
points. This variation is similar to the variation of the electron response. This is an
indication of the detector's non-uniformity. I will include this as a systematic

uncertainty for the e/rt ratio measurement.
5.4 Load-one energy scan results

With the load-one configuration, we took a large amount of data in order to study the
uniformity, linearity, etc. The results have been published in [22]. We did not take
useful data with calorimeter nigger electronics for the energy scan study. Here I only
show the published energy scan results with the BLS data in order to compare with the
load-two results. A detailed discussion can be found in [22].

Energy scan data for both electrons and pions were taken. The data show good linear

response for both electrons and pions. The linearity fit yields a result;

87
for electrons: E (ADC) = 80.2 + 267. 4P ,
for pions: E (ADC) = -565 + 257.1P ,
where P is the beam momentum in GeV/c.
5.5 Load-one and load-two electron data comparison.

The load-one electron linearity fit shows that the load-one detector has a higher electron
response than load-two. If I average over the two sets of the load-two electron energy

scan data, I get the load-two electron response to be:

255.4(AD%eV).

Comparing this with the load-one electron response:

267.4(AD%eV),

the load-one electron response is about 4.5% higher than the load-two response. This is

larger than any systematic uncertainty I can find.

There is evidence that this difference could be caused by a local hardening effectl15].
In its simple form, the local hardening effect is the following process (referring to
Section 2.4.1 for the calorimeter response to the EM showers). In the electromagnetic
shower development, a large fraction of the energy goes into the production of low
energy photons. For low energy photons (below 1 MeV), the photo-electric effect is the
dominant interaction, and the cross section of the photo-elecuic effect is proportional to
Z5 of the material. As a consequence, the soft photons from the electron showers
interact almost exclusively in the absorber, and most of the soft photons will transfer
their energy to knock out electrons in the absorber. These elecu'ons have very low

energy (~ 10s of keV), and very short range. A fraction of the photo-electric

88

interactions are sufficiently close to the surface for the electrons to escape, so these
electrons contribute to the measured signal. However, if some low Z material (inactive)
covers the surface of the high Z absorber, these electrons would not be able to reach the

detecting media, and thus would not contribute to the measured signal.

In the D0 CCEM module, about half of the Uranium plates in the EM layers 1, 2 and 3
have one surface covered with readout boards (made of 3.6mm G10 plate). On the
average, over 75% of the electron energy is deposited in these three layers. Therefore, I
expect that the electron energy response will be suppressed. In the ECEM calorimeter,
there is no readout board covering the Uranium surface, so I would expect the load-one
module (ECEM) to have higher electron response than the load-two module (CCEM).
There are only a total of three readout boards in the CCFH module (with 50 Uranium
plates), so I do not expect any significant effect on the pion energy measurement caused

by the readout boards.

The local hardening effect has been measured by the SICAPO collaborationlls] with the
U/G10/Si sampling calorimeter. From their measurement, and if I assume that liquid
Argon readout is similar to Si readout for detecting these photo-electrons, I expect
about 19% signal suppression of an electron shower with one surface of every Uranium
plate covered with 5mm G10 plates. I“ assume that 3.6mm GlO plate is about 50% to
100% as effective as the 5mm G10 plate used in the SICAPO calorimeter, then I can
estimate that the suppression of the electron signal will be 3.6% to 7.1%. This rough

estimate agrees with the load-one and load-two electron energy scan data.

Voltage

89

 

0.4

0.3

0.2

0.1

-0.2

 

 

   
   

\

pulse at TRG input ”“1” “ BLS "‘9‘"

 

 

pulse at preamplifia input

 

 

0.55 1.2 1.85

Time (10"s)

Figure 5.0 SPICE simulation of TRG readout and precision
readout (BLS) responses to fast and slow showers

2.5

Trg/BLS

 

 

 

 

 

   

 

 
 

 

 

 

 

 

 

0.11 T I I I rIIII T I I T I IIIl' I I I I I .T
A '- 1
2 t- —O-— Trg(ADC)/(Pulser amplitude (DAQ) .
e - ' - .
0 0.1062 "' -‘
i b 4
5 .
‘5. 01025 ~ —
2 r-
:r _ .
§ - y=0.092l‘x"(0.0201) 4
< 0.09875 - 0 Err:0.001 0.002 Chi2=l.5 ‘
v r . . 4
E 4
E- E «
0.095 1 1 L411111 L 1 LLI 1111 1 1 L 1 L111
10 100 1000 104
PulserAmplitude(DAC)

Figure 5.1a NWA trigger driver linearity study using calibration pulser

 

 

 

 

 

 

 

 

1.1 1 rr 1 v t I 1 r I r I r I r I r r r I r rrTfT—rlfrr
b d
b
1.063 -- o _.
b O
r- . . .4
. . o 0 ' .
1.025 - ..
l- o d
b o O ‘1'me 1
O Tanners“
0.9875 - _
" 4
" J
' n
l l L 1 LL L14 1 1 l L l 1 l l l 11L] 1 I LI 1
(L95 L 1* I 4*

0 20 40 60 80 100 120 140 160
BearnManenturMGeV/c)

Figure 5.1b Electron energy scar, Trg/BLS with & without linearity correction

 

91

. ,5;

1 Cd F1m
T°°
i I RinJ1>-a

 

 

 

Ac,
1

Figure 5.2 Calorimeter electronics crosstalk schematics V

92

 

4000

3000

2000

1000

 

Average of excess energy in HD vs. energy in EM

 

 

I x' 9.595
' m 0.24305-0i t 090905-04
" Y I Pl 0 X

 

 

 

 

 

 

 

-1000 11.11...ML1..ELL..LLEL..1....
0 10000 20000 30000 40000 50000 50000
Figure 5.3 The excess energy in the level-one calorimeter trigger HI)

channel vs. energy in the EM channel.

Num. of Towers

Num. of Towers

 

93

 

 

 

 

0.002 0.0028 0.0036 0.0044 0.0052 0.006
Trg/BLS

Figure 5.4a Electron trigger tower scan. EM trigger tower gain disuibution

 

6 ._ I I I I I I I I I I I I I I I

’_ FH Trg/BLS

    

 

 

0.002 0.003 0.004 0.005 0.006
Trg/BLS

Figure 5.40 Electron trigger tower scan. FH trigger tower gain distribution

4O
.35
30
25
20
15
1O

20
17.5
15
12.5
10
7.5

2.5

94

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I. 0 9|)

: b) Enh'n ur

. . a... mu

7 _. was I vans

I — P 724:

"_ : emu-n no. : 4.007

_ _ I.- m. 1 , 0am

: __ ' 10. 20 t 110021

3. Z 1 . . . 1 . .

200 400 600

ADC/GeV '

_ o m

r 1 40 d) “...? 13}

' I45 I Ill!

'— I 20 {ad-n 13.52 1:27

_ a... mu um

_ 10.13: am

: 1 00

I 80

I 50

3 40

L 20

[L O L A .1 I 1 r

200 400 600 200 400 600
ADC/GeV ADC/GeV

Figure 5. 5 The energy distributions with different gain smearing. a) 100 GeV pion

with 10% gain smearing, b) 100 GeV electron with 10% gain smearing, c) 100 GeV pion

with 20% gain smearing, d) 100 GeV electron with 20% gain smearing.

Fractional resolution

95

 

 

 

 

 

 

 

 

 

 

   

 

   

 

 

0.2 r 1 1 1 I 1 r r—Frr 1 T I I r I 1 I I 1 1 v 1 l v r 1 1 I 1 r r r—rt I r
_ ll y = sqn(0.0036+(ml'rn0)‘(ml‘... ll :
. || Value Em ll .
0 16 _ || 1111 045499737341 0.005794 || _
' _ || Orisq 1.6231234] M H -
L 4
0.12 - _
L- .
0.08 - _
r 1
0 04 1 1 1 1 1 1 r 1 1 1 LLI L1 L 1 1 1 I 1 1 1 1 I 1 1 r 1 1 1 1 1 1 I 1 1 1 L

-0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35

Fractional gain variation

Figure 5.6 Trigger gain variation study with 100 GeV pions.

96

H' vla n rnimla

 

 

 

 

High Voltage
(resistive coat on

Rggggm pad (310 surface)
l...14 /

A/A/Izv

 

 

 

 

 

 

  

\Weaker field and longer path

Normal drift

Figure 5.7 Electric field at CCEM crack and charge pulse shape

97

 

I
I
| 0.005 77
I
I

v r-Y [4 ‘v 214*

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.003 — 1

I

I

I

0.002 - I

‘ I

0.001 - I

°-..‘z.‘1$‘i§§is

Irq/uwv
40000
30000 L
20000 I-
15000 3
10000 -
I
3000 1-
I
o t 1 1 l. J 1 l 1
-5 -2 -1 0 | r J 9 I
Mm
Figure5.8 a) Rauoof the TRGandBLS vs. thevertical position(crn)
oftheSOGeVelecrronbeam.

0) Energy re se of the 31.8 vs. the vertical position
(cm) of the 5 GeV electron beam.

 

98

)1 nt i ti

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0"!
°r
.2 =
& ’fl 1 r
.5 .5
3 e .
U 0
g o
10‘
I L
0 1‘
¢
9“?
”I
I
,2 E
II:
3 g 10!.
P a
i i 'r
.3 1. .4
.. III
I u a It I. i
<9 <1)

Figure 5.9 Monte Carlo simulation of the energy distribution of the 50 GeV pion
showers in the D0 central calorimeter.

(Data - Fin/Data

Energy (ADC) 104

 

 

 

 

 

 

 

 
 
 
  
 

 

    

 

 

 

 

 

 

 

0.05 l I I I I I I g

I 0 BISelectroncluster,residual j,

o 03 _ O BISpIoncluster.m1dual _,

- . A TRGelectronclnster.residual .

- a 'l‘RGpicnclusrer,residual .

0.01 L— . ' g j

.. ii! I i . ‘ . 1

-o 01 - 5 ' - J

I 1

-0.03 _- -;

f I I 1 I l I l ‘

- I I I I I I I .

I 1

3 5 :- _ _ y=-sr.sss+253.3x as 0.99993 -3

:---- ya-87514+253.561Rz0.99994 ;

30 -_ - y=-10.372+256.59XR=0.99999 ;

' I ya-112.73+259.39sz0.99999 :

2.5 :— _:

2.0 :— '5

1.5 : -— G— 31.8mm ‘3

I --.- 31.9mm :

1.0 : - a - 1161.21de 3

: -—-l— 11‘0”ch :.

5.0 : 3

: I I I I I I I d

0.0

0 20 40 60 80 100 120 140. 160

Beam momentum (GeV/c)

Figure 5.10 Electron energy scan, eta=0.05, phi=31.6

(Data - Fin/Data

0.05
0.03
0.01

'-0.01

-0.03

Energy (ADC) 104

3.5
3.0
2.5
2.0
1.5
1.0
5.0

100

 

 

 

 

 

 

 

 

 

 

 

  
 
  
  
  

 

 

_ I I I I I l I

I o BLS electron chum. residual .
- . BLS pion cluster. residual ..
” A TRG electron cluster. residunl ‘
.. A TRG pion cluster. residual

.- 1 ‘1
. i I ' i I ' ' :
‘ 1 I I I I I I ‘
_ I I I I I I I .
L ._. _ y a .5634: + 257.54: a: I —'
- ----y=-42.43+257.51xR-l :
-_ . y s 455.717 + 258.51: R8 0.99999 _:
3 y -.- -9l.063 + 261.37: R: 0.99999 :1
7 “z
: 1
L- — G— 815 elm cluster 2‘
I: - -O - 315mm 1
:- - A - TRG demon cluster 1
I —t— TRG {£de 1
r ‘:
_ I I 1 I I L I ‘

0 20 40 60 80 1 00 1 20 140
Beam momentum (GeV/c)

Figure 5.11 Electron energy scan, eta=0.45, phi=3l.6

160

101

 

005 I I I I I I I

1

U I

 

BLS pion, residual _‘:
TRG pion. residual ..

0.03

' I

.0

 

 

 

0.01

-0.01

(Data - FIG/Data
1—o+—e—1
t—eHo—I

t-o-t-e-I
.1
1004
m1
m
we
1

-0.03

UIIIIUI'YUU'I

1 1 L 1 1 1
I I I I

 

—
q

 
   

3.5
3.0
2.5
2.0
1.5
1.0
5.0

y a «235.19 + 256.93! KB 0.99993
- - - _ y = 479.5 + 252.78! R: 0.999“

 

 

Energy (ADC) 104

 

 

 

 

 

IIII'UIUI'IIIU'TIIIIIIII'UIIIIIIIIIIIII

l l l L l l l

20 40 60 80 100 120 140 160
Beammomentum(GeV/c)

 

0

Figure 5.12 Pion energy scan, eta=0.05, phi=31.6

(Data - Fit)/Data

Energy (ADC) 104

0.05
0.03
0.01
-0.01
-0.03

102

 

 

 

 

 

 

 

 

 

 

 

 

 

 
 
 
  

 

 

__ I I f I I I I q
L 0 ms pion. residual j
1- ; . j
. ii ' 8 .
1' 1
" '1
. d
' g 1 I 1 1 1 I a
.J
' 1
7 1
: y a 235.1 + 263.02: R: 0.99993 1
-_ _ - - - y a 457.97 + 25221: R- o.99997 .‘
5- 3
C :
r I
C’ 1
_ 1 1 1 1 1 1 I
0 20 40 60 80 100 120 140

Beam momentum (GeV/c)

Figure 5.13 Pion energy scan, eta=0.45, phi=31.6

160

Fractional resolution

Fractional resolution

0.2

0.16

0.12

0.08

0.04

0.25

0.2

0.15

0.1

0.05

103

 

 

O BLS. electron cluster

0 Trg. electron cluster

cl s I Nlcmz
. [0.146] 0.4TLI.

 

 

 

 

 

0.0152] 0.146] 1.3] 2.3

 

I I I I I I I I I I U l I I I l 1 U I

 

 

q
LillilllllLlllLllllLllllllLlJll

 

 

 

0 20 40 60 80 100 120 140
Beam momentum (GeV/c)

Figure 5.14a Electron energy resolution, eta=0.05, phi=3l.6

160

 

 

 

 

 

 

 

 

fi'jUIYU'UlfUII'II'UIIVII

 

 

 

LL:lirLlrrglLr+lLrglLijlltillL1
0 20 40 60 80 100 120 140 160
Beam momentum (GeV/c)

Figure 5,14b Pion energy resolution, eta=0.05, phi=3l.6

Chapter 6
Results and Monte Carlo simulation.

The final analysis of the load-two e/tt ratio measurement with both fast trigger (TRG)
and precision (BLS) readouts will be presented in this Chapter. To understand the
detector response to good precision, we have to understand the small effects caused by
the imperfections of the detector. In Section 5.2.5, an example was given on using the
Monte Carlo simulation to study the detector crack effect. In this Chapter, I will
present the detector Monte Carlo simulation and the comparison with test beam data.
Another example will be given on using the Monte Carlo simulation to study the pion
shower leakage in the load-two test beam configuration. Finally, there will be a
discussion of the results and the questions that need to be studied further. Conclusions

will be presented
6.1 General information about the GEANT Monte Carlo simulation

GEANT V3.14l3] is an integrated Monte Carlo package supported by CERN. It is
designed to help simulate the behavior of large complex particle detector systems like
the D0 detector. It includes a geometry package, a memory handling package, and a
physics interaction simulation package. The physics processes package includes
E08403] for the electromagnetic shower simulation, and GHEISHAI24] and
NUCRINm] for the hadronic shower simulation. All the primary and secondary
particles are transported through the detector geometry until their energies fall below

some kinematics cut or they move out of the detector.

104

105
The D0 experiment uses GEANT V3.14 as a framework for its detector simulation.

The simulation includes all the detector subsystems, and it is done with geometric detail
as close to reality as possible. For example, the calorimeter simulation is done with
geometric detail to the individual absorber and readout plate level. Such a detailed
simulation enables us to simulate many geometric features of the detector, e. g. the inter-
module cracks. There have been many testsmtm to verify the detector simulation with
test beam data, and large progress has been made to tune the Monte Carlo simulation to

match the test beam results.

A special test beam version of the calorimeter Monte Carlo program was written to
include all the test beam geometric information, including all beam line materials and
cryogenic materials. The load-two test beam geometry is shown in Figure 6.0a. The
following comparison of the test beam data and Monte Carlo simulation is done with
this test beam version Monte Carlo simulation. The hadronic shower simulation uses
GHEISHA, and the kinematics cuts of all the particles are set to a minimum of 10 keV

for maximum accuracy.
6.1.1 Monte Carlo simulation of electron and pion energy scan

A set of Monte Carlo energy scan data has been generated for both electrons and pions
with different energies, at (11:005. 01:31.6) which is the bench mark position of the
load-two test beam. The Monte Carlo data are analyzed the same way as the beam data.
Figure 6.1 shows the linearity of the Monte Carlo data. In this figure, the Monte Carlo
data are normalized to approximately the same scale as the beam data. Comparing with
Figures 5.10 and 5.12, the Monte Carlo data show similar linearity behavior as the
beam data. For example, both the Monte Carlo data and the beam data show good
linearity for electrons and pions, and the offset for the demon fit is smaller than the

offset for the pion fit. However, there is a discrepancy between the Monte Carlo and

106

beam data on the relative response of electrons and pions. I will discuss this in Section

6.4.

One of the important requirements for the Monte Carlo simulation is the ability to
simulate the random fluctuation of the electron and pion showers. To test this, I will
compare the energy resolution of the Monte Carlo simulation with test beam data for
both electron and pion energy scans. Figure 6.2 shows the energy resolution plot of
both electron and pion energy scans. Since the Monte Carlo simulation does not
include any external noise, the energy resolution from the Monte Carlo data is better
than the energy resolution of the test beam data (Figure 6.2a). This effect is more
visible at the low energy points where the noise dominates the energy resolution. To
include the noise effect, I add the average noise measured from test beam pedestal data
(Table 5.2) to the Monte Carlo data. It shows that the energy resolution of the Monte

Carlo data with noise added is very close to that of the beam data (Figure 6.2b).

6.1.2 Monte Carlo simulation of pion shower leakage

In order to measure the e/1t ratio accurately, a precision measurement of the pion energy
is required. Because of the limited size of the test beam calorimeter, the pion showers
are expected to leak a small percentage of their energy outside the detector. In the
load-two configuration, the CC modules have only 7 interaction lengths and the
transverse size is limited as well. Both longitudinal and transverse energy leakage of
the pion showers needs to be studied. Based on the test beam Monte Carlo geometry,
two more FH modules were added on the sides of existing FH modules and a large
piece of Uranium was added at the back (Figure 6.0b). The leakage energy can be
collected in these extra modules. The fractional pion energy leakage is listed in Table
6.1. The fractional energy leakage is defined as:

107

 

where the Em, is the average total energy (energy in the test beam modules + energy in
the extra modules), and is,“ is the average standard energy (energy in the test beam
modules only). The uncertainty in the table is 10% of the leakage energy. The 10% is
the estimated precision of the Monte Carlo pion shower simulation, deduced by
comparing the pion transverse shower shape of the load-one data with the Monte Carlo

simulation [22].

 

 

 

 

 

 

25

_
50 2.85 0.29
75 3.(X) 0.30
100 2.74 0.27
125 2.64 0.26

 

 

 

 

 

 

Table 6.1

6.2 Summarized results of the load-two e/rt ratio difference at

different integration times.

To determine the e/rt ratio difference between short (TRG) and long (BLS) integration
times, I subtract the BLS eltt ratio from the TRG e/rt ratio. This way, many systematic

uncertainties in the energy measurement (pion shower leakage, non-uniformity, etc.) are

 

1Fractionoftotalenergyinthedetector'.
2nteennroitheenergyleakagedividedhythetotatenergyinthedeteetor.

108
canceled. The remaining systematic uncertainty is dominated by the uncertainty in the
trigger electronics' gain. Figure 6.3a shows the e/1t ratio difference between the TRG
and BLS. All the relevant corrections for the electron and pion energy measurements
have been applied in the calculation of the e/lt ratios. Table 6.2 is a list of all the
statistical and systematic uncertainties (in fractional percentage units) caused by the
various sources. The total uncertainty is the sum of all the uncertainties in quadrature.

I will briefly discuss each of the terms below.

 

GeV

Trigger Drv——
non- -linearity

 

TRG sampling
fraction 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5

Trigger cross-

 

0.65 0.65 0.65 0.65 0.65 0.65 0.65 0.65 0.65

 

variation
CCEM crack
effect

Pion shower
3.3128" 0 0 O 0 0 0 0 0 0

‘ Statistical
uncertainty

Trigger FADC
uncertainty 2.01 1.52 1.22 1.01 0.60 0.40 0.30 0.24 0.20

Total
Wm. 4.0 3.6 3.3 3.3 32 32 32 32 3.1

(3.56)...

 

“5358‘“ 3.0 3.0 3.0 3.0 3.0 3.0 3.0 3.0 3.0

 

 

1.3 0.8 0.6 0.5 0.3 0.3 0.3 0.3 0.2

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 6.2

(1). Trigger non-linearity correction: This is a small correction. The 10%
uncertainty on the correction exponent affects the e/rt ratio measurement by

045% [Section 52.1].

(2). Sampling fraction correction: The difference between the sampling fraction

used to calculate trigger resistors and the precision sampling fraction is less than 10%,

109
and the energy difference between the TRG and BLS is less than 5%, therefore, I

estimate the uncertainty of the sampling fraction correction to be 0.5%lseetjou 53-2].

(3). Trigger crosstalk correction: Since the crosstalk correction is small, the
uncertainty of the crosstalk comes mostly from the uncertainty in the crosstalk
mechanism. I use the difference of the two mechanisms as the uncertainty in the
correction. This difference is 0.6% for the EM and 0.65% for the HD energy
measurement. Since it is a systematic uncertainty I will use the larger one (HD

channel's) for the e/rt uncertainty estimatelsecfion- 53-3].

(4). Trigger gain variation: This is the largest systematic uncertainty for the TRG
energy measurement. From resolution studies, I estimate the worst case gain variation
as 2.4% for the EM, and 5.8% for the HD channels. Because the electron and pion
showers are deposited in many preamplifier channels, this gain variation only affects
the average TRG energy measurement by 1.5% for the EM and 2.6% for the HD5095“-
5~2-4]. I combine both uncertainties in quadrature and get that the total uncertainty of

the e/rt ratio caused by the trigger gain variation is 3.0%.

(5). CCEM crack effect: Because the crack effect correction is very small (less than

1%), the uncertainty of the correction is negligiblelsecfion- 525].

(6). Pion shower leakage: Because both the TRG and BLS share the same leakage
correction, it does not introduce any uncertainty in the difference between the TRG and

BLS e/rt ratio measurement.

(7). Trigger FADC digitizing uncertainty: The FADC used in the calorimeter
trigger has a maximum uncertainty of :l: 0.5 ADC counts on the digitized output data. I
use the worst case of 0.5 ADC count as the systematic uncertainty contributed from

FADC digitization.

110

In Table 6.2, the total uncertainty of (i) -(i) is the uncertainty of the individual
Tr. cu

3 3
measurements. There are two sets of independent measurements of the e/1t ratio
difference at two different detector points (n=0.05 and 11:0.45). Since these two sets of
data are measured with different trigger electronics channels and FADC digitizers, by
combining these two sets of data one can get a better uncertainty estimate. To calculate
the combined uncertainty of two individual measurements, one has to separate the
common uncertainties and independent uncertainties in the two measurements. In our
case, the uncertainties caused by trigger driver non-linearity, trigger sampling fraction
and trigger crosstalk are common uncertainties. They are correlated in the two
measurements, so combining the two measurements will not improve these
uncertainties. However, the uncertainties caused by the trigger gain variation, statistics
and the trigger FADC are independent uncertainties. These uncertainties are not
correlated in the two measurements. Combining the two measurements, I can improve

these uncertainties.

Beam Momentum (GeVIC) 15 20 25 30 50 75 100 125 150 ‘

 

...... 1:1 Ii)
3 m 7' nu 0.027 0.033 0.047 0.045 0.041 0.052 0.057 0.055 0.049
Combined uncertainty

(i)... Ii)...

 

0.028 0.026 0.025 0.024 0.024 0.023 0.023 0.023 0.023

   
  

 

 

 

 

 

 

 

 

 

 

 

     

Table 6.3

Table 6.3 is a list of the averages and uncertainty of the e/rt ratio difference by
combining the two sets of data. Since the uncertainties for the two independent
measurements are the same, the combined result is the average of the two different

results. The combined uncertainty is calculated as:

111

 

arz'rgocin‘l'ogm‘l’alrgmpc

oConb = \lofi'rgDn + a‘lz'rgSnp + 0%qu + 2 °

 

In this formula, crew is the total combined uncertainty. amp", 0mm and 01'qu

are the uncertainties cause by trigger driver non-linearity, trigger sampling fraction and
trigger crosstalk. arm, 05,. and am,“ are the uncertainties caused by the trigger

gain variation, statistics and the trigger FADC. Since these uncertainties are not
correlated in the two measurements, the contributions of the am, 03., and 0mm,c
to the combined uncertainty are a factor of 42 smaller than they are in each individual
measurement. Figure 6.3b is the combined results (Table 6.3) of the two
measurements. The results show that there is about a 5% difference in the e/rt ratio
between the short (TRG) and long (BLS) integration times with particle energy greater
than 25 GeV. At lower energy, the systematic uncertainty increases and the difference

appears to be less significant.

6.3 Precision readout (BLS)e/1t ratio result

The load-two c/rt ratio measurements from the precision readout (BLS) system are
listed in Table 6.4. The result includes Monte Carlo estimated pion shower leakage.
The uncertainty includes statistical uncertainty, pion leakage correction uncertainty and
the uncertainty caused by the detector non-uniformity. The fractional energy
uncertainty caused by the non-uniformity of the detector is estimated to be 1.6% by
analyzing the uniformity scan datam]. This analysis has not been completed, and the
1.6% uncertainty is a preliminary result. Since the uncertainty caused by the detector
non-uniformity is the dominant uncertainty, its effect on the e/rt ratio is preliminary too.
Figure 6.4 shows the load-two e/rt ratio results with two sets of energy scan data The

Monte Carlo result in the figure comes from the data generated at (11:0.05, 41:31.6).

112

The Monte Carlo simulation and test beam data have a relative offset that will be

discussed in the next Section.

 

  
 

 

 

 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

Beam Momentum (GeV/c) 25 125 150
gilt ratio at 11:0.05 1.036 1.016 1.007 1.018 1.022 1.000 1.001 0.980 1.008
uncertainty at 11:0.05 0.021 0.019 0.018 0.018 [0.017 0.017 0.017 0.017 0.017
1: ratio at 11:0.45 1.023 1.021 1.020 1.016 ll.(l)7 0.294 1.000 0.986 0.994
uncertain at =0.45 0.020 0.018 0.018 0.017 0.017 0.017 0.016 0.017 0.017
Table 6.4

The published result from the load-one data is shown in Figure 6.5. It includes the
Monte Carlo simulation results with the load-one detector simulation. The uncertainty
in the load-one measurement includes the statistical uncertainty and the pion leakage
correction uncertainty. The dominant uncertainty is caused by the poor statistics of the
inspill pedestal data. The load-one module has better uniformity, therefore the
uncertainty caused by the detector non-uniformity does not contribute to the load-one

e/n ratio measurement.

6.4 Discussion

There are a few points I want to discuss regarding the Monte Carlo simulation and the

load-two results.
1) Local hardening effect

With particle energy greater than 20 GeV, the Monte-Carlo simulation agrees well with
the load-one BLS e/rt ratio measurementlzzl, but the simulation shows a few percent

(4%-6%) higher e/rt ratio than the load-two BLS measurement. Furthermore, the

113
measured load—one e/rt ratio is a few percent higher than load-two, and there is a
discrepancy between the load-two Monte Carlo simulation and the beam data. There is
evidence that this could be the result of a local hardening effect as I stated previously in
Section 5.5, which is not well simulated by the current GEANT Monte Carlo. The local
hardening effect is mostly caused by the suppression of the signal coming from the
photo-electrons. These photo-electrons produced by soft Ys through the photo-electric
effect have energies of 10's of keV or less with a very short range. GEANT Monte
Carlo only tracks particles with energy above a cutoff energy (10 keV minimum), and
particles with energy below the cutoff energy will deposit their energy locally. To give
an example of this effect, Figures 6.7, 6.8, and 6.9 show the longitudinal energy
distributions of 5 GeV electron showers in a few U/LAr gaps, with cutoff energies of 10
keV, 100 keV and 1.0 MeV.. The longitudinal energy distributions in all materials (U,
LAr and 610), in LAr only and in (310 only are shown in these Figures. The spikes at
the medium boundaries are artifacts of GEANT Monte Carlo, which stops particles near
the medium boundaries and deposits their energies locally. In reality, some of this
energy would be transported to the next medium. The relative size of the spikes
increases with the cutoff energy, and even with a 10 keV cutoff energy, there are still
spikes at each medium boundary. This indicates that GEANT Monte Carlo can not

precisely simulate the very low energy phenomena near the medium surface.

The construction of load-two and load-one modules is essentially the same except for
differences in the Uranium plate thicknesses between CCEM and ECEM (3mm in
CCEM, and 4mm in ECEM) and the extra readout boards in the CCEM. Since the local
hardening effect has not been simulated in the Monte Carlo, the Monte Carlo shows
approximately the same e/rt ratio for b0th load-one and load-two. This is an expected
resultm] without the local hardening effect. However, the load-one modules have no

010 covered Uranium surfaces, so there is no local hardening effect, whereas the local

114
hardening effect in the load-two CCEM modules suppresses the load-two electron
response. This results in a smaller load-two e/rt ratio than in load-one. Figure 6.6

shows that after shifting up load-two ratio by 0.0451 , the e/rt ratio for load-two

becomes similar to that for load-one.
2) The slow signal component in the hadron showers

The GEANT Monte Carlo provides time information of energy deposition. I have done
simulations to study the difference of the energy deposition between a 200ns cutoff
time and a 2.2 Its cutoff time in the 100 GeV pion shower. The Monte Carlo shows
very little difference (< 1%) between the two cutoff times. The measurement of the c/rt
ratio difference between short and long integration readout shows about a 5%
difference. Other experimentsl29~30t31t311 show a similar effect indicating that there is a
slow signal component in hadron showers. These results indicate that the GEANT
Monte Carlo needs improvement in its hadron shower simulation, particularly the
neutron-nucleus interaction simulation, because neutron-nucleus processes are the
major contributors to the slow signal component. In recent years, other groups have
made progress in improving the neutron-nucleus interaction simulation. They have

shown the effect of the slow hadronic shower componentl32].

6.5 Conclusions

This dissertation presents a systematic measurement of the e/rt ratio difference at two
different time scales. The final result indicates that there is a 5% difference in the e/rt
ratio between short (250ns) and long (2.211s) integration time with beam energies
ranging between 25 and 150 GeV. The estimated uncertainties on this measurement are

2.3%. Below 25 GeV however, the systematic uncertainty of the measurement

 

1Itisthedifferenceoftlleload-oneandload-twoelectronresponsesthatwebelievetobeenusetlbythe
localhardeningeffect

115

increases to 3%, and the measured difference decreases to 3%, making the difference
measurement less significant. This measurement confirms the prediction from the
theoretical analysis of hadronic shower processesmtm that there is a slow signal
component in the hadron showers, and that the time scale of this slow signal component
is approximately 1 us. This measurement is an improvement on the previous
experimental result of HELIOSIM], the only previous experiment using a Uranium-
liquid Argon calorimeter, which show a 2% to 5% difference between different

integration times and energies.

This analysis had to overcome difficulties with the fast calorimeter trigger electronics to
achieve a systematic uncertainty sufficiently small for a meaningful measurement.
With more data, the systematic uncertainty could be further reduced. The experience
gained from this analysis benefits the commissioning of the D0 calorimeter trigger.
This analysis shows a good agreement of the measured calorimeter energy resolution of
the load-two data with the GEANT Monte Carlo simulation. It also shows a small
discrepancy between the load-two data and the GEANT Monte Carlo simulation in the

e/tt ratio and ratio difference. This provides a guide to fine tune the Monte Carlo

simulation.

116

 

Figure 6.0 Geometry information for Monte-Carlo simulation of the
load-two test beam. a) standard test beam configuration.
b) with extra modules to study the pion shower leakage.

Energy (lO‘ADC)

4.0

3.5

3.0

2.5

2.0

1.5

1.0

0.5

117

 

 

 

 

 

 

     

 

 

; f I I I I I I I I T I I I I I I I I I I I I f r I I I .1
. 1
I- a
y— —t
t :
L o Electra ‘
.. a Plan 1
l- 4
1- a
'_ .1
E 1
E. _‘
i I
:' 1
. y . 56.293 9» 255.06: R- l .
Z - - - - y - 234.44 + 239.68: a. 0.9999. 1
L’ 1
1- «4
P' d
L J
: 4
b J J l I l L 1 l l 4 l 1 1 L 1 L l L 1 1 l 1 1 1 l L L 1
0 2 0 4 0 6 0 8 0 1 0 O 1 2 0 t 4 0

Beam Momentum (GeV/c)

160

Figure 6.1 Monte Carlo simulation of the load-two electron and pion energy scan

0.18

0.12

0.06

Fractional energy resolution

0.18

0.12

Fractional energy resolution

118

 

 

 

 

 

 

 

 

I I I r I I I I I rj I I I fiI I T I T I r
I.
I- ; O MCelectrcn
.. A MCpion
.. E; o TBelectron
_ A TBpion ..
I I
P- d
. r .
r I a
_ . i —
I
1 , i .
.. . Q .1
t- :0 -l
. O
__ O 0 9 . 1
Ii 1 l l l L l 1 l 1 l l l 1 L41 1 l 1 l L l 1 l L L1
0 20 40 60 80 100 120 140 160

Beam momentum (GeV/C)

Figure 6.3a Monte Carlo and test beam data energy resolution comparision.
No electronics noise in MC data

 

 

 

 

 

 

I O sigma/0+” «
I A Stroller-noise r
I I O Siam/c "
I—n - —I
I I A stone/e ..
..
i -I
I. : .1
- 0 i x I —
c ‘ ~
1- . a
80 .
O a
O

C 9 . ‘

1 J._L l 1 l 1 l A l 1 1 L1 PL 1 l l 1 l l 1 L .L l l l

 

 

0 20 40 60 80 100 120 140 160
Beammomentum(GeV/C)

Figure 6.2b Monte Carlo and test beam data e

whom comparison.

Includes average electronics noise in Monte Carlo

119

 

 

 

 

 

 

 

 

 

 

 

 

0.1 r v I l r I I r I I r r I I r v I I r l l I r r l I v r
- f : -
0.08 - I I I I I -
.. .,. ..
_ T .
0.06 - ._

e c J-

(IL, {:13 : 0 I! H “I j
_ Ii ‘3 o .
0 04 — 0 c: -
" fl 0 “
0.02 - n .. i: 0 I ‘
. 1: ‘- . trg-bls.eu=0.05 ' r
b J. D trg-blr.eu=o.45 '

o r I 1 I I l r 1' r l I I r l r I I I 1 r l I r r I l l r r

0 20 40 60 80 100 120 140 160
Beam momentum (GeV/c)

Figure 6.3a Electron/pion ratio difference between short (T rg)
and long (BLS) integration times.

 

 

 

 

 

0.1 . I I I I I I I I _
0.08 1 —
I -
0.06 1 -
c c ' ‘
(fin-(r)- : ;
0.04 - -
C 9 I
0.02 L- _
f- -l
0‘" I.m.m...lrn.lr.rmmi.1...l...1
0 20 40 60 80 100 120 140 160
BeamMomentum(GcV/c)

Figure 6.3b Average electron/pion ratio difference between
short (Trg) and long (BLS) integration times.

120

 

1.15 I I I I 7 7 Y I T I Y rt I T I I til I I I ITI Y IT! I

 

 

 

 

I J
I g j
e MCWi
1 1 ' :1 dpi. bu. eta-0.05 ‘
' " A e/pi. bis. all-0.45 ‘
1.05 - ' -
I- . at
g 4
'5» - - 4

 

 

 

0.95 C- _‘
h ‘1
0.9 L 1 1 L1 1 4 1 l 1 L Ll . 1 L l l 1 J L 1 J_ L1 1 1 l _L 4 #4
O 20 4O 60 80 100 120 140 160
Beam Momentum (GeV/c)

Figure 6.4 Load-two BLS electron/pion ratio.

 

ch:

121

 

1.20

1.15 ‘

1.10

1.05

 

—'°— Data

. é "'°"' GEANT/GHEISHA

 

A A

 

 

1.00

50 100 150

Beam Momentum (GeV/c)

Figure 6.5 Load-one BLS e/1t ratio.

 

122

 

 

 

 

 

 

 

 

 

 

 

 

1-1_ I fII I I I I I ‘

: o Lie/pi I

I 0 L2 e/pi shift up 0.045 3

: ..I :

1.05— In I _

. I I? . ;

I I m I

i j

1’. :

0.95"..1...1...Im..1...1...I.mmtrim‘
0 20 40 60 80 100 120 140 160

Beam Momentum (GeV/c)

Figure 6.6 Load-one and load-two e/pion ratio comparison. Load-two result
is the average of two sets of energy scan data shifted up by 0.045

123

 

 

5 GeV electron

 

 

 

 

 

I 'I—l TT‘FTTT'ITT TTIWT
I. g
L
I- a

p
(I)
{J‘ .
N
0'
0)
OD

gig JJ .1...
5.8

 

 

 

 

 

 

 

 

 

 

 

6.4 7.2
Zenergy in Ur
0.75 E
0.5 :-
o.25 :— I
OLLLILM+III r 1_Ltr 1Lll 1111
4.8 5.2 5.6 6 6.4 6.8 7.2
Zenergy in lor
0.8 _—
0.4 —
O-LLJIALILm‘AIILIIIII 11L
4.8 5.2 5.6 6 7.2

 

Zenergy in 610

 

Figure 6.7 The longitudinal energy distribution of 5 GeV electrons in
U/LAr gaps with 10 keV cutoff energy. The top is energy in
all materials (U, LAr and (310), the middle is energy in LAr
only, the bottom is energy in 010 only.

 

124

 

 

5 GeV electron

 

 

 

 

 

' fl fl
oLil4 JALgJJ llLll 41‘..ng iliii
5.2 5.6 6

6.4 6.8 7.2

Zenergy in Ur

 

 

 

 

 

 

 

 

 

 

 

F"
I-l
0.4 L I
OL LILWIILI Lilrsz—lthil 1111
4.8 5.2 5.6 6 6.4 6.8 7.2
Zenergy in lor
I.
08 r
0.4 L
OFL LIALLLMIILILLLAII llLilLlLlL
4.8 5.2 5.6 6 6.4 6.8 7.2

Zenergy in 010

 

 

Figure 6.8 The longitudinal energy distribution of 5 GeV electrons in
U/LAr gaps with 100 keV cutoff energy. The top is energy in
all materials (U, LAr and (310), the middle is energy in LAr
only, the bottom is energy in G 10 only.

 

125

 

0.8

0.4

0.8

0.4

5 GeV electron

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I'
u
l.
I
1 1 1 1 14¢! I L l A J L I 1 1 L 141 .1 _l_‘ L Lg 11 fl
4.8 5.2 5.6 6 6.4 6.8 7.2
Zenergy
W l l L m LA; 1 l J l 1 L j l I l 1 l 1
4.8 - 5.2 5.6 6 6.4 6.8 7.2
Zenerqy in lor
t
t'
h A L L_l_ 1 l 1 L m l 1 I. l l l 1 l L l L LI 1 L 1 l L

 

 

 

4.8 5.2 5.6 6 6.4 6.8 7.2

Zenergy in G! O

 

 

Figm'e 6.9 The longitudinal energy distribution of 5 GeV electrons in

U/LAr gaps with 1.0 MeV cutoff energy. The top is energy in
all materials (U, LA: and 610), the middle is energy in LA:
only, the b0ttom is energy in 010 only.

I

Bibliography

[1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

Fermi National Accelerator Laboratory. Design Report Tevatron I Project.

Fermi National Accelerator Laboratory, September, 1984.

M. Abolins, et a1. Design Report: the DO experiment at the Fermilab
Antiproton-proton Collider. Fermi National Accelerator Laboratory, Nov.,

1984.
R. Brun, et a1. GEANT3. CERN. DD/EE/84-l, Sept., 1987.

R. R. Roy, et a1. Interaction of Photons and Leptons with Matter. Academic
Press, 1968.

E. Merzbacher. Quantum Mechanics. John Wiley & Son, 2nd Edition, 1970.

D. Schamberger, et al. D0 Calorimeter Eectronics. Nucl. Phys. B 23A, p191,
1991.

T. L. Quarles. SPICE3 Version 3C1 Users Guide. UCB/ERL M89/46, April,
1989.

CW. Fabjan, et a1. Energy Measurement of Elementary Particles. Rep. Prog.
Phys., 52, p1519, 1989.

126

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

[17]

127

R. Wigmans. Advances in Hadron Calorimetery. Annual Review of Nuclear

and Particle Science, 41:133-85, 1991.

R. Wigmans. 0n the Energy Resolution of Uranium and Other Hadron
Calorimeters. Nucl. Instr. and Meth., A259, p389, 1987.

R. Wigmans. High Resolution Hadron Calorimetry. Nucl. Instr. and Meth.,
A265, p273, 1988.

R. Wigmans, et al. An Experimental Study of the Contribution of Nuclear
Fission to the Signal of Uranium Hadron Calorimeters. Nucl. Instr. and Meth.,
A252, p4, 1986.

V. Radeka, et a1. Speed and Noise Limits in Ionization Chamber Calorimeters.
Nucl. Instr. and Meth., A265, p228-242, 1988.

B. Yu, et al. A Uranium-Liquid Argon Calorimeter for Experiment NA34
(HELIOS) at the CERN SPS. Chapter 1.4, p15, SSC Detector R&D at BNL,
April 1990 Status Report. BNL 52244.

SICAPO collaboration. The Local Hardening Effect on Electromagnetic
Showers: A Way for Signal Equalization in Si/High-Z Hadron Calorimeters.
Phys. Lett, B222, p518, 1989.

M. Abolins, et al. The Fast Trigger for the DO Experiment. Nucl. Instr. and
Meth., A289, p543, 1990.

M. Abolins, et al. Hadron and Electron Response of Uranium Liquid Argon
Calorimeter Modules for the DO Detector. Nucl. Instr. and Meth., A280, p36,
1989.

 

 

[18]

[19]

[20]

[21]

[22]

[23]

[24]

[25]

[26]

[27]

128
H. Brueckmann, Calculation of Hadronic Shower. Proc. of Workshop on

Compensated Calorimetry, Pasadena 1985, CALT-68-1305.

C.W. Fabjan, et al. Uranium Liquid Argon Calorimeter. Nucl. Instr. and Meth.,
141, p61, 1977.

J. P. Richer, et al. Cross Talk in Multielectrode Particle Detectors. IEEE
Trans. Nucl. Sci. NS-31, No 1, Feb., 1984.

R. Hirosky. Private Conversation on the Analysis of D0 Liquid Argon Purity
Test Cell Data and Calorimeter High Voltage Scan Data.

M. Abolins, et al. Hadron and Electron Response of DO Uranium Liquid Argon
End Calorimeter Modules. Submitted to Nucl. Instr. and Meth.

The 15684 code system. SLAC Report 265.

H. Fesefeldt. GHEISHA: The Simulation of Hadronic Showers Physics and
Application PITHA 35/02.

K Hanssgen et al. The Monte Carlo Code NUCRIN to Simulate Inelastic
Hadron-Nucleon Interaction at Laboratory Energies Below 5 GeV. Comp.
Phys. Comm., 39, p53, 1986.

D. Huffman. Precision Calibration of Calorimeter Electronics in the DO Liquid
Argon/Uranium Particle Detector. IEEE Nuclear Science Symposium, Nov 2-

9, 1991.

R, Angstadt. et al. Modeling of the DO Data Acquisition System. IEEE Nuclear
Science Symposium, Nov 2—9, 1991.

 

 

[28]

[29]

[30]

[31]

[32]

[33]

129

W. Dharmaratna. Private Communication on the CC Uniformity Scan Data

Analysis.

M. DeVincenzi, et al. Experimental Study of Uranium-Scintillator and Iron-
Scintillator Calorimeter in the Energy Range 135-350 GeV. Nucl. Instr. and
Meth., A243, p348, 1986.

MG. Catanesi, et al. Hadron, Electron and Muon Response of a Urnium-
Scintillator Calorimeter. Nucl. 1mm and Meth., A260, p43, 1987.

T. Akesson, et al. Performance of the Uranium/Plastic Scintillator Calorimeter
for the HELIOS Experiment at CERN. Nucl. lush: and Meth., A262, p243,
1987.

T. A. Gabriel, et al. CALOR89: Calorimeter Analysis and Benchmarking.
Proceeding of the First International Conference on Calorimetry in High Energy
Physics.

Particle Data Group. Review of Particle Properties, Passage of Particles

Through Matter. Phys. Lett. B239, p11]. 13, April, 1990.