NUCLEAR
INSTRUMENTS
& METHODS
IN PHYSICS
RESEARCH
Nuclear Instruments and Methods in Physics Research A325 (1993) 23-91
North-Holland
Section A
A precision experiment on electrons, photons and muons at LHC
UP Collaboration
Received 9 September 1992
We describe the upgrade of the L3 detector for running at LHC . The principle goals are the precise measurement of electrons,
photons and muons
1. Physics at LHC with the L3P detector
1.1 . Physics objectives
The great efforts made at CERN on the construction of a Large Hadron Collider (LHC) at a center
mass energy of 16 TeV and luminosity of 1 .6 X 10 34
cm -2 s-1 will enable physicists to explore competitively new frontiers of physics.
The plan to use the existing LEP tunnel and much
of the existing CERN-LEP infrastructure implies that
LHC will be constructed in an economical and timely
way.
In this report, we propose to construct a general
purpose detector, which will have a uniquely high
resolution in measuring electrons, photons and muons,
using much of the existing L3 equipment and infrastructure (see figs . 1, 2) . The physics objectives of L3P
are not only to search for particles predicted by current theories, such as Higgs, but more important, to
look for unexpected phenomena. The spirit of this
effort is to propose a detector which is complementary
to other detectors at the SSC, Tevatron and LHC. The
design is based on a realistic assessment of the very
high LHC background situation and on our efforts in
studying e, li, y, including the L3 experiment and the
successful design of L* at the SSC. It is our intention
to ensure that a powerful detector will be available for
physics at LHC turn on .
us confidence in the Standard Model, were all made by
precision experiments on leptonic and photonic final
states . Indeed, one can recall the following examples :
1) The discovery of two neutrinos [1] came from
measuring p, and e final states .
2) The discovery of the J particle [2] shown in fig. 3
was done by an experiment on e + e - final states with
a mass resolution of 0.1% and a hadron background
rejection of 1/10 1° .
3) The r lepton [3] was discovered with a detector
measuring the coincidence of lt, e final states .
4) The discovery of P state by the DASP Collaboration [4] at DORIS in July 1975 from a very clear and
elegant observation of 2 y transition of tf~' is one of the
most important confirmations of the existence of charm
quarks .
1.2 . Physics considerations
Over a period of a quarter of a century, there have
been many fundamentally important discoveries in elementary particle physics. These discoveries, which gave
* For the members of the L3P collaboration, see appendix .
Fig. 1. Schematic view of the L3P detector .
0168-9002/93/$06 .00 © 1993 - Elsevier Science Publishers B.V . All rights reserved
24
L3P Collaboration / Precision experiment at LHC
8m-
Fig. 2.
Cross-sectional view of the detector .
5) The T particle [5] (fig . 4) was discovered by an
experiment with a l.L pair mass resolution of 2% .
6) The proof that the J particle is indeed a bound
state of cc quarks comes from precision inclusive pho80
ton measurements with Nal crystals by the Crystal Ball
group [6] (fig. 5) . The identification that the T particle
is a bound state of bb quarks comes from inclusive
242 Events
70 _ SPECTROMETER
At normal momentum
60
- 10% momentum
50
N
C
iL
40
30
20
10
25
30
Ï
I
r,-fl
rn e+e- (GeV)
Fig. 3 . The discovery of the J particle.
FH
35
Fig. 4 The discovery of the T particle.
L3P Collaboration / Precision experiment at LHC
25
These facts lead us to make the following observations:
- These discoveries were not predicted when the
original accelerators were constructed (the Z and W
excepted).
- None of these discoveries were made by detecting hadronic final states .
- In general, the decay rate of heavy particles into
single leptons, single photons and lepton pairs is much
smaller than the decay rate into hadrons . However,
since the background from large momentum transfer
single leptons or single photons and high mass lepton
pairs is generally very small, if an experiment can be
done cleanly and precisely, with good resolution and
good hadron rejection, one can clearly distinguish the
signals from the background, as shown in figs . 3 to 6.
1 .3. Design considerations
E y (MeV)
Fig. 5. Inclusive photon spectrum at the T from the Crystal
Ball experiment at SLAC, showing essentially all of the charmonium spectrum.
photon measurements by the CUSB [7] and CLEO
groups [8] as well as by the ARGUS and Crystal Ball
groups [9].
7) The discovery of the Z' particle [10] (shown in
fig. 6) was done with a large solid angle detector
measuring e+ e- and li + w - final states .
8) The W t [11] were found by measuring their
large momentum single electron and muon decays .
In high luminosity hadron colliders such as ISR as
well as stationary target experiments with high intensity beam, 1O t°-1012 ppp on target (with equivalent
luminosity of 10 ;' and higher), the background is always much higher than anticipated . Monte Carlo calculations of background can only be used as a lower
order estimation of the background . Part of the reason
is because the yields of photons, electron pairs and
muon pairs are many orders of magnitude less than the
yields of hadrons. In addition, there are always abundant amounts of muons, photons and neutrons traveling along the beam and entering the detectors. It is
often very difficult to trace the origin of these particles.
Fig. 6. The discovery of the Z° particle . The spectrum shows an e + e - correlation from Z° decay at UAI, CERN .
26
L3P Collaboration / Precision experiment at LHC
The following experiments serve as examples for
designing a precision lepton and photon pair experiment at LHC:
1 .3.1 . Test of quantum electrodynamics and study of
leptonic decays of rector mesons at DESY
A series of experiments [12] were done in a high
intensity gamma beam of 10" photons per second . The
detector had a mass resolution of AM/M = I % and a
hadron rejection of ee/hh= 1/10". The key elements
of these experiments were such that :
a) The detectors were far away from the target and
no material was placed between the target and the
detector to prevent the conversion of Tr ° - 2-y
e+ e- .
b) Strong magnetic field between the target and the
first detector elements swept away low energy particles .
c) The detector did not expose to the target directly
and therefore did not expose to the neutral particles.
d) Electrons were measured twice: first the momentum (p) was measured by magnetic spectrometers with
threshold Cherenkov counters and second the energy
(E) was measured by pulse height hodoscopes. The
requirement p = E eliminated most of the backgrounds .
c) The minimum transverse distance between the
detector and the beam line was = 2 m therefore the
beam spray did not enter the detector .
1 .3 .2 . Discovery of the J particle at Brookhaven
This experiment [2] was carried out in a high intensity proton beam of 10 12 protons per second, equivalent to a luminosity of 10 36 cm -2 s-1 .
The key elements of this experiment were :
a) Strict application of the experience learnt at
DESY (see above) by putting the detector far away
from the target to eliminate the beam spray (the minimum distance between the detector and the beam line
being again = 2 m) and by minimizing the material in
front of the detector to reduce Tr O - 2y and knock on
electrons. Indeed, it was the development at CERN of
hydrogen gas Cherenkov counters, with 3 mm spherical
collecting mirrors and 125 Vm Mylar windows at each
end of the counters, that made this experiment possible
b) Development of high rate proportional chambers
to operate at low voltage and orientate the wires 120°
apart with the result that these chambers (size I m2)
could work at 20 MHz and were able to reconstruct the
trajectory without ambiguity.
1 .33 Discovery of the T particle at Fermi National
Laboratom
This experiment [5] had a mass resolution of 2%
using (1 .5-3) x 10 1 ' incident protons per accelerator
cycle. The proton intensity was limited by the requirement that the singles rates at any detector plane not to
exceed 10' counts per second .
The mass resolution of 2% was obtained by measuring muon trajectories in free space. The identification
of muons was done by measuring muons with an accuracy of = 15% (at 10 GeV) using trajectory information in a magnetized iron absorber . The minimum
distance between the first counter and the beam line
was again = 1 meter.
1 .3 .4 Mtion pair study pp - pp at the ISR
This is a 2Tr detector [13] measuring muon pair
production with a luminosity of = 10 32 . This experiment was carried out with the classical method of
measuring muons with magnetized iron sandwiched
with large drift chambers . Whereas the result of this
experiment provided very accurate information in
studying pp - Frw scaling precisely, it also revealed the
great difficulty of aligning large area drift chambers
sandwiched in a closed area between magnetized iron .
We learnt that, in practice, it was difficult to align
large area chambers (3 x 6 m2 ) to better than = 1 mm,
with the result that the momentum resolution was
considerably worse than the original design value.
1 .3.5 . The L3 experiment at LEP
At LEP, nearly all the physicists in this publication
participated in building a 4Tr detector (L3) [14] which
measures a 100 GeV particle decaying into a pair of
muons with a mass resolution of 1 .7%. L3 also measures electrons and photons with a coordinate resolution of = 1 mm, an energy resolution of 5% at 100
MeV and an energy resolution of 1.3% at 2 GeV. The
L3 detector has a good hadron rejection. The momentum of w(e) is measured twice, first in the vertex
chamber with a value of P, and second, by the precision muon chambers with a value Pw , or in the BGO
with a value Pe. The muon energy loss vE is measured
by the sampling calorimeter which also monitors hard
photon radiation. The energy balance, P = ,A E + P or
P = Pe eliminates the backgrounds efficiently.
From the above five examples, we conclude that a
high precision detector to measure e, w, y at LHC
should have the following characteristics:
1) To reduce the beam spray : the transverse distance
for most of the detector elements should be as large
from the beam line as possible.
2) To reduce neutral particle background : all elements should be far away from the intersection point.
3) To reduce charged particle backgrounds : a strong
magnetic field surrounding the intersection region is
necessary.
4) To reduce background on photons : the individual
electromagnetic calorimeter elements should be far
away so as to separate the 2y rays from rr ° decay, and
L3P Collaboration / Precision experiment at LHC
there should be a minimum of material before the
electromagnetic detector .
5) To reduce background on electrons : the momentum of the electron (p) defined by electron trajectory
in a magnetic field and the energy of the electron (E)
defined by pulse height with the individual electromagnetic calorimeter elements should be both measured
with high accuracy, so that p = E can be satisfied to
the 1% level.
6) To reduce background on muons: the momentum
of muons should be measured repeatedly (p l , pz
pn ) . The most precise momentum measurement (p)
can only be done by measuring trajectories in free
space. Other measurements PZ to p to = 15% level
are used to reject background by the equation : p1 =P2
pn .
1.4. Technical considerations
A realistic design of a large experiment must also
take into consideration the ability to build the detector
on time . Therefore, one must take into account the
following:
1) Technical Capability : Many physicists in this collaboration have continued R&D efforts on precision
instrumentation in muon physics, in new crystals and in
mass production of high quality crystals for electron
and photon physics and in calorimetry . In the following
we give some examples :
a) Precision instrumentation on muon physics: Over
a period of 20 years, we have continued intensive
R&D efforts [15-181 to develop precision instrumentation for muon physics. This includes the following:
- Alignment systems: development of high precision alignment systems with UV laser verification (5
lim precision) .
- Chamber construction : development of methods
to build high resolution muon chambers covering an
area of = 1000 m2 with 30 wm precision .
- Supporting structure: development of supporting
structures for precise alignment of the chambers, in
collaboration with the Draper Laboratory (10 p,m
reproducibility) .
- Gas study: selection of the best gas for large area
drift chambers in a magnetic field, and particularly
nonflammable gases.
The success of this program is the main reason why we
were able to build the precision muon system in L3,
and our continued effort on muon R&D gives us
confidence to be able to construct a precision muon
system for L3P.
b) Precision instrumentation on electron and photon physics: Since most of the rare earth materials are
produced in China and there are many excellent centers for crystal studies there (Shanghai Institute of
Ceramics, Shandong University, Zhejiang University
27
and Beijing Glass Institute) as well as an excellent
research center in Russia (Institute of Solid State
Physics, Moscow), we have continued large scale systematic efforts in developing new crystals and on mass
production at low cost of large quantities of radiation
resistant crystals . The Chinese effort, under the coordination of Prof. D.S . Yan, and the Russian effort coordinated by Academician Yuri Ossypian, have been
most successful . We are now confident to be able to
produce large amounts of crystals for the L3P experiment .
c) Silicon calorimetry : A large systematic effort on
the properties of silicon calorimetry, VLSI readout
associated electronics and radiation effects on both
detectors and electronics has been carried out by the
SICAPO group. Currently an intensive R&D program
lead mainly by the INFN (Florence, Milan and Rome)
groups in L3 and by the Dubna group is achieving the
required low cost for mass production on both silicon
detectors and associated VLSI readout electronics, as
well as on silicon detector plane assembly .
2) Engineering Resources. The construction of large
experiments depends on a good collaboration between
engineers and physicists . Over the last ten years, the
L3 collaboration has assembled a large group of engineers and technicians from leading physics institutions
all over the world (member states of CERN, Russia,
China and the U.S .) . It has taken many years to establish mutual understanding and collaboration among
these engineers and technicians who are now familiar
with each others measurement systems and practices.
The ability of this group of engineers and technicians
must be taken into consideration in determining the
complexity of the final detector .
1.5. Financial considerations
Our experience in the construction of L3 and in
designing L* enables us to realize that much effort
must be spent in understanding and controlling the
cost of general purpose detectors at LHC. The enormous costs of the SSC and LHC detectors will ultimately determine their completion date . With limited
financial resources available, it is important for us to
design a detector, taking into account the following:
1) Use as much as possible of the existing L3 equipment and infrastructure .
2) Construct the detector in phases but ensuring
that a powerful detector will be available at LHC
turn-on.
3) However, despite current financial limitations,
the L3P Collaboration intends to construct, based on
their past experience, a complementary detector to all
other detectors, at the SSC. Tevatron and LHC. We
intend to fully explore the potentials of LHC without
compromises.
28
L3P Collaboration / Precision experiment at LHC
1 .6 . The L3P detector
Based on the above considerations, we present the
design of the UP detector in figs . 1 and 2 . It has the
following unique features :
1) The central trackers (C) and (D) are far away
from the colliding proton beams. Most parts of the first
detector (D) (made out of proportional chambers, drift
tubes and gas microstrip detectors) are located at least
1 .65 m away from the proton beam (see section 3).
2) We will use CeF3 crystals from China, specially
made for UP, for the region 1771 < 1 .4 . In the forward-backward region 1 .4 < 1771 < 3 .0, a sampling
calorimeter will be used (see section 4) .
3) The total amount of material in front of the
crystals and the electromagnetic sampling calorimeter
is less than 3% X, .
4) Low energy charged particles will be swept away
by a 2 T magnet with a 7 m inner diameter and a 13 m
length . The low field means that the magnet can be
constructed economically and reliably .
5) Photons are measured by the electromagnetic
crystal calorimeter, located at a minimum 3 m away
from the intersection point. At this large distance, the
knowledge of shower profiles from crystals will yield
accurate information on the photon production angle,
thus eliminating complicated detectors to sample preshower information on individual crystals . This large
distance also implies that a 25 GeV -rro - 2-y will
produce two separate shower peaks and thus can be
rejected. The large distance also implies that the neutral background on the crystal surface will be minimized (see section 4) .
6) Electrons are measured twice, first from momentum information in the trackers (C) and (D) to an
accuracy of 1% (100 GeV), and this will be matched
with energy information obtained from the pulse height
measurement from the CeF3 crystals and the sampling
calorimeter, to an accuracy of _ 1% at 100 GeV (see
section 4).
7) Hadrons are measured by a sampling calorimeter
(total absorption >_ 9.1) which is constructed similarly
to the L3 detector, with read-out towers pointing to the
intersection region (see section 5) . It provides complete
77 coverage up to rl = 4.6 (see section 7) .
8) Muons are measured independently three times:
a) with multi-layer drift tubes (A), existing L3
muon chambers (B), central tracker (C&D), together
with vertex information (E) given by LHC (see section
3 and 6) .
Typically, we have :
(ZAP/01 = 1% .
b) Measurements (A, B, C and E) without inner
tracker (D) yield:
(AP/P)2 = 6% .
c) Measurements with only deflection in the return
yoke (from A, B and E) yield :
(OP1P)3 = 15% .
The requirement p, =P2=p3 eliminates most of the
backgrounds . The muon trigger is done by impact
parameter measurements using information from RPCs
in stations (A) and (B).
9) The detector is designed to use much of the
existing L3 infrastructure, the forward-backward muon
chambers, the central muon chambers MO as well as
the existing L3 magnet frame and toroidal doors.
2. The magnet system
2 .1 . General magnet design considerations
The design of the magnet system follows the principles already used successfully in the L3 experiment,
namely a large tracking volume . The reason for this
choice is based on the principle of sagitta measurement
in a solenoid, where the momentum resolution decreases linearly with the magnetic field strength B but
quadratically with the measurement length L:
AP/P ^' 1/BL2 .
The requirement of a large tracking volume coincides with the requirement for the electromagnetic
calorimeter to be far away from the interaction point.
Since the inner radius of the electromagnetic calorimeter is chosen to be 2.9 m the inner bore radius of the
superconducting coil is consequently 3.5 m, i.e . 2.9 m
plus the thickness of the electromagnetic calorimeter .
The overall magnet length is 13 m. This design provides a homogeneous field distribution in the volume
of the central tracker, as shown in fig. 7. The field
strength was chosen to 2 T in order to achieve a 1
resolution in pr at 100 GeV.
For a large magnet with iron return yoke the construction and assembly of the yoke is the most time
consuming part . In our proposed solution the iron yoke
1 8
s
4
Z[M]
Fig. 7. Field distribution
L3P Collaboration / Precision experiment at LHC
29
Table 1
Main parameters of the SC solenoid
Interaction point
Fig. 8. Magnet system overview.
of the L3 detector is used on a reduced diameter with
slightly increased wall thickness. The experience gained
during the assembly of the L3 iron shield led us to
consider a new design for the poles. This will simplify
the assembly and suppress heavy welding in the underground area . In line with our past experience in heavy
lifting, we have chosen to install a 1000 t temporary
gantry crane in order to handle individual pieces up to
800 t unit weight . An overall view of the magnet system
is given in fig. 8 .
2.2 . Thin superconducting solenoid
The use of superconducting magnets has increased
in recent years both in high energy physics and in
fusion research . For example, ALEPH and DELPHI
magnets are operating successfully at LEP/CERN,
and the T-15 magnets at the Kurchatov Institute, the
LCT coils at ORNL and TORE SUPRA at Saclay are
examples of fusion research
For the cooling of superconducting magnets, liquid
helium bath as well as forced flow and thermosiphon
cooling have been used . For the insert superconducting
magnet we worked out two options, one is a design
with forced flow cooling by supercritical helium and
the second one is a solenoid with thermosiphon cooling, as examples .
In the first solution pressurized helium at a temperature between 4.5 and 5 K will be pressed through the
coil cooling channels by the refrigerator outlet pressure . This solution enables us to design a thin solenoid .
The main parameters of the thin superconducting
solenoid are given in table 1 .
2.2.1 . Superconducting cable
The superconducting cable for the UP central
solenoid is made out of Nb-50%Ti Rutherford type
cable with a copper to superconductor ratio of 1.3 . The
flattened cable, made out of 30 strands, is embedded
Vacuum chamber inside radius [m]
Winding inside radius [m]
Winding outside radius [m]
Winding length [m]
Vac. chamb. outside radius [m]
Vac. chamb. overall length [m]
Solenoid volume [m 3 ]
Number of layers
Number of turns
Total ampère-turns [MA* t]
Central induction [T]
Total thickness [A]
Stored energy [MJ]
Cooling power [W]
Dump Voltage [kV]
3.5
3.7
3.8
12 .6
4.1
13
600
5
1250
23
2
0.8
900
500
1
together with a copper cooling tube into two copper
clad aluminium profiles . The tensile stresses are taken
over by two reinforcement tapes of 1 mm thickness.
The components of the conductor are soft soldered
together .
The aluminium residual resistance ratio (RRR) is at
least 500. The overall current density is 32 A/mm 2.
The nominal current is 19 kA for a magnetic field of 2
T in the center . The conductor is insulated by a 0.15
mm thick half-overlapped fiberglass fabric . Additionally, the support cylinder is insulated by a 1 .0 mm thick
fibre-glass jacket . The superconducting coil is impregnated with an epoxy resin under vacuum . The superconducting cable cross section is presented in fig. 9 and
the main parameters are listed in table 2. Fig. 10 shows
the superconducting solenoid cross section.
2.2 .2. Coil winding
The superconducting solenoid is subdivided into
twelve coils each 1.05 m long . Each coil is wound out
of two conductors of about 1250 m length, which is a
feasible manufacturing length . The gap in the centre
between the two half coils is 20 mm where a cooling
tube with fresh helium is connected to the conductor
ooling tube
Fig. 9. Superconducting cable cross section .
30
L3P Collaboration / Precision experiment at LHC
Table 2
Main parameters of the conductor
Parameters
SC material
SC wires
Diameter
Cu to SC ratio
Number of filaments
Critical current at B = 2.5 T
Rutherford type cable
Numb . of wires
Cross section
Critical current at B = 2.5 T
Operation current
Stabilizer
Material
RRRat42KandB=0
Dimension
Data
Nb50%Ti
[mm]
[kA]
0.85
13
66
1 .0
[mm2]
[kA]
[kA]
30
2X12
30
19
AI 99.99%
500
end on the innermost layer. The coolant outlet and the
current connections to neighbouring coils are always at
the end flanges on the outside layers . The coil conductor is wound onto the outside of an aluminium support
cylinder to form essentially a five layer coil .
2.2.3 . Support structure
The coils are provided with electrically insulated
flanges for supporting axial magnetic forces . The axial
force distribution is shown in fig. 11 . The integrated
axial force over the half length of the solenoid is about
6000 t. The structural ring allows a bolted connection
between the sections to form a rigid cylinder . The
solenoid is suspended by cold mass supports, made up
of fibre-glass tie rods, arranged like a bicycle wheel.
These cold mass supports carry the weight of the
solenoid and radial magnetic forces due to the radial
eccentricity of solenoid and iron return yoke . The
axial, magnetic forces of about 6000 t which try to pull
the two half solenoids against each other are supported
by the structural cylinders itself . For axial stability of
the system there are fiber-glass tie rods installed.
_:Supermsulation-=
Radiation shield4K suoerinsulation=
.He outlet
®He Inlet
4K Superinsulation
Radiation shield®
~Supérinsulation=
'=,(Vacuum vessel inside wall
Fig. 10 . Superconducting solenoid cross section.
i
0
2
Z (m)
4
i
i
i
6
Fig. 11 . Axial gradient force along the half-solenoid.
2.2.4 . Vacuum vessel
The vacuum vessel consists of two coaxial aluminium cylinders of 30 mm and 50 mm wall thickness
respectively . The vacuum vessel flanges which receive
the forces from the cold mass supports as well as from
the electromagnetic calorimeter weight have an adequate rigidity .
The vacuum vessel inside surface is coated with a
thin radiation shield which is cooled by pressurized
helium at 60 K. The gas is circulating in tubes which
are attached to the radiation shield and provide a good
thermal connection . The outside surface of the radiation shield is insulated by multilayer superinsulation .
The vacuum vessel is joined by a transfer line with the
helium refrigerator . The transfer line covers the cold
feed and return lines and the current leads. The vacuum vessel evacuated also through the transfer line.
2 .2.5. Manufacturing and assembly
The proposed subdivision of the superconducting
solenoid into 12 single coils allows manufacturing on a
specialised deliverer site as well as on the CERN site .
The following operations are necessary:
1) Welding of the support cylinder,
2) machining of the support cylinder,
3) insulation of the support cylinder,
4) winding of the superconductor onto the support
cylinder,
5) preassembly of the sections,
6) coil assembly,
7) testing of the coil .
A workshop of proper size has to be installed for
manufacturing and assembly of the superconducting
solenoid . The support cylinder, which are made of an
aluminium-magnesium alloy, are prepared on the production site and then shipped to the CERN site . After
the welding, the support cylinder sections will be heat
L3P Collaboration / Precision experiment at LHC
treated. Then the support cylinders machined by a
rotating milling cutter and insulated with fiberglass
reinforced epoxy before the conductor is wound onto
the support cylinder . After the winding procedure,
hydraulic and electrical connections are attached to
the conductor. At the very end the winding is vacuum
impregnated with a two or more components epoxy
resin. Finally, hydraulic and electric tests are carried
out.
After that, the coils, which make up the solenoid,
are moved into the assembly area and are bolted
together in vertical position . All the electric and hydraulic connections are prepared and tested . After
hydraulic and electrical testing the inside vacuum vessel wall, equipped with the thermal radiation shield,
and placed inside the vacuum vessel is lowered down
into the assembled solenoid . In a next step, the vessel
outside wall with its radiation shield will be put over
the solenoid . Finally, the superconducting insert system will be tested and ready to be lowered down into
the pit .
2.2 .6. Magnet instrumentation
To ensure the reliable operation, the following diagnostic systems are installed:
- Multibridge system for normal zone detection in
each of the subsections and across the whole magnet,
- strain gauges to check stresses,
- set of thermometers, pressure transducers, flow
meters,
- acoustic emission detectors,
- magnetic field sensors .
2.2.7. Cryogenic system
In the present design a straightforward forced flow
cooling system is proposed which circulates single phase
supercritical helium adjacent to the current carrying
superconductor . This forced circulation loop provides
several advantages. First it provides thermal capacity
adjacent to the superconductor . Metals have extremely
low thermal capacity at 4.5 K. Without the higher
thermal capacity of the helium, the conductor would be
extremely sensitive to thermal perturbations associated
with conductor motion and the risk of a quench would
be high . The helium in the conductor increases the
critical energy margin . Second, the forced flow in the
channels is in direct contact with conductor joints, and
the heat produced by resistive connections is cooled
directly by the coolant. Furthermore, in case of a
quench the forced flow loop eliminates any possibility
of hot spots in the conductor. Using forced flow cooling enables us to construct a homogeneous magnet and
reduce the amount of helium in the system to a mininum . Last but not least, radiation heat from both sides
of the winding is taken over by the circulating forced
flow helium .
31
For the refrigerator plant, screw compressors are
installed . They are of proven reliability in helium operation and capable of continuous service over periods
up to 8000 h without scheduled maintenance . The
electric power consumption of the compressor group,
including the helium cleaning system, is minimized. In
order to prevent air infiltration into the helium gas
stream, the suction pressure is kept above atmospheric
pressure . The coolers are of a very reliable design in
order to avoid leaks between the helium and water
circuits . For this reason, the coolers are constructed
using an austenitic stainless steel. The compressor
waste heat is lead off by a closed-circuit wet cooling
tower into the atmosphere .
For the cold process, the plant has a cold box
placed in the main hall over ground level. The distance
between the compressor group and the cold box is
about 120 m. The cold box, together with the necessary
equipment for control, vacuum and purge, is mounted
on a frame or platform which is movable with the hall
overhead crane as a unit .
The refrigerator is equipped with gas bearing turbines. The turbines are appropriate for long term reliable service, with minimum downtime for maintenance.
The turbines are located to be readily accessible for
maintenance and repair .
The whole measuring system is provided with a
complete set of sensors and measurement devices connected to a microprocessor that the refrigerator is able
to operate with high reliability and safety in the various
operation modes like : helium system purge, cool-down,
normal operation with or without users, system warmup and emergency.
A forced warm-up process is foreseen . For this
reason room temperature helium from the compressor
system is blown into the last heat exchanger of the
refrigerator.
2.2.8. Power supply and energy dumping system
The magnet is energized in about 30 min. In case of
emergency fast discharge with series connection of the
coil sections takes about 300 s. The discharge circuit
has an independent discharge resistor of about 0.1 fl .
The do current switches used in the dumping circuits
are commercially available. The very low probability of
a quench is ensured by the high stability margin of the
conductor and the high redundancy of the cooling, a
protective discharge circuit, however, is needed for
unforeseen external perturbations. It is important to
have adequate diagnostic facilities to distinguish transient events like small local disturbances from strong
instabilities which force discharge of the magnet . A
multibridge quench detector will be used to detect a
normal zone appearing in any sub-unit of the solenoid .
Two discharging speeds are envisaged. A slow normal
one, which does not heat up the supporting cylinder,
32
L3P Collaboration / Precision experiment at LHC
and a fast one triggered by a coil signal of 20 mV
during at least one second . In case of an emergency
discharge, the resulting eddy currents in the support
structure will heat the windings almost uniformly up to
70 K.
The superconducting solenoid is connected by liquid helium cooled leads onto an air cooled aluminium
bus bar and furthermore to the power supply which is
located on the floor about 100 m away from the detector. To eliminate strayfield, individual bus-bars of opposite polarity will be alternated . This type of system
allows for a compact construction . The power supply
and the interface of the do modules with the bus-bars
system requires a surface of about 50 mz inside a light
building . Transformers fed directly from the high voltage mains will be installed in an adjacent yard of 50
m2.
2 .3. Transparent solenoid (option 2)
In this option we employ a different cooling method .
The schematics of the second SC solenoid option is
shown in fig. 12 . The design of the coil is based on the
design experience of successfully operating magnets
and aims at a minimum thickness of the solenoid .
A monolayer coil of the solenoid is wound onto an
Al-Mg alloy supporting cylinder using an AI stabilized
conductor. The coil is adhesively bonded to the cylinder . As mentioned above, thermosiphon method is
employed for the cooling of the coil .
The mechanical structure of the vacuum vessel is
not only designed to withstand the atmospheric pressure and the weight of the coil but it must also provide
structural support for the electromagnetic calorimeter
and the central tracking system situated inside the SC
coil .
This is achieved through introduction of seven belts
of inner distant bars of 550 mm (radial supports)
positioned between inner and outer shells of the cryostat and penetrating the supporting cylinder . Such a
design allows the inner and outer shells to be as thin as
axial
support
R
3.78 m
winding
12 .4 m
nitrogen
shield
R4m
I
r- 1
13 m
-_____--__-----
beam line
vacuum
vessel
1
587 m
radial
support
~~ ~~
287 m
R35m
7=,--'i
I
______-
IP
Fig. 13 . Transparent solenoid cross section.
20 mm while an average stress in the shells does not
exeed 23 MPa. Although local stresses in the point of
attachment of the bars to the cylinders can be several
times higher than this value, this is acceptable for the
AI-Mg alloy structure.
To facilitate manufacturing, transportation and assembling of the magnet, the winding is subdivided into
eight sections separated by 300 mm gaps (see fig. 13).
Adjacent sections are fixed to each other with flanges.
The radial distance bars pass through these flanges
without touching them .
As the designed current density in the conductor is
59 A/mm 2, it is necessary to diminish a heat leak to
the superconductor and to avoid strong perturbations
connected with mechanical stresses . The conductor is
shielded against heat radiation by the support cylinder
on the inner side and special high thermal conductive
Al plates from the outside. The stresses in the conductor and the supporting cylinder are reduced to quite
low values (the maximum Treska stresses are 104 MPa
in the cylinder and 60 MPa in the conductor) .
The axial gaps between sections do not spoil the
magnetic field uniformity in the central tracker volume.
The main parameters of the magnet are given in
table 3.
2.3.1 . SC cable
Fig. 12 Transparent solenoid overview .
helium
shield
A Rutherford type cable with strands of 1.3 mm
diameter is soldered between two copper clad pure
aluminium profiles . The critical current of the conductor in 3 .1 T (i .e . the maximum field on a strand taking
into account a current self field) is two times more
than the operating current. This sets a safe SC coil
operation temperature margin of about 1 K even taking into account a heating up of the cylinder by eddy
currents during a 2 h period of energizing of the
magnet .
33
L3P Collaboration / Precision experiment at LHC
Table 3
Main parameters of the option 2 magnet
Vacuum chamber
inside radius [m]
overall outside radius [m]
overall length [m]
material
mass [t]
SC coil
inside radius
of supporting cylinder [m]
material of the
supporting cylinder
winding mean radius [m]
outer radius
of the 4.3 K shield [m]
length [m]
number of turns
total thickness [A]
operating current [kA]
central induction [T]
stored energy [MJ]
SC cable material
Superconductor
Stabilizer
3.5
4.1
13
Al-Mg alloy
60
3.7
AI-Mg alloy
3.8
3.85
12.6
800
0.4
29
2
910
Nb-50%Ti
A] 99 .99%
The design of the conductor makes low resistive
joints between conductor pieces in the winding possible. The conductor manufacturing procedure of joining
SC cables to stabilizing substrata is reliable, inexpensive and is standard throughout of the world. Soldered
Àl stabilized conductors of similar cross section are
produced in Russia routinely and are used in many
magnets. Parameters of the conductor are presented in
table 4.
2.3.2. Final assembly
To insure a high quality of construction, the coil
units as well as the vacuum vessel sections are manufactured on a company site . The final assembly will be
carried out on a CERN site .
Table 4
The main parameters of the conductor
Dimensions [mm]
bare
insulated
SC cable length [km]
Stabilizer
RRR
critical current (3 T, 4.2 K) [kA]
number of strands
Strand
diameter [mm]
Sc : Cu
Number of filaments
12 .2 x40
13 .2 x 41
1.2
Cu clad Al
500
55
25
Nb-50%Ti
1.3
1 :1 .5
8900
7nrm
Fig. 14 . Assembly sequence of the transparent solenoid .
The following vertical assembly procedure is proposed: First, the end flange of the vacuum vessel is
installed onto a special support structure. Then, the
superconducting coil units and the vacuum vessel parts
are put one upon another. Finally, the SC coil units are
connected together . The vacuum vessel parts are
equipped with the radiation shield elements before
their installation . Electric and hydraulic tests of the
windings are carried out after each step (see fig. 14).
At the very end, the vacuum vessel sections will be
welded together and a vacuum test will be performed .
Before installation into the detector, the superconducting solenoid will be cooled down and energized at a
surface hall situated near to a refrigerator .
2.3 .3 . Cryogenic system
A cryogenic system (1 kW refrigerator, LHe vessel
and pipe lines) can operate in different regimes:
- precooling of the winding with the cold He gas,
- cooling by natural convection of He vapour-liquid
mixture in the tubes attached to the supporting
cylinder,
- heating of the winding with warm He gas when
necessary.
The cryogenic heat loads are listed in table 5. The
precooling requires about one week at He gas flow of
0.25 kg/s . This is determined by the admissible levels
Table 5
Cryogenic heat load
Static load at 4.2 K [W]
radiation load
supports
transfer lines
current leads (3 .7 g/s of LHe)
conductor joints
Transient load at 4.5 K
during magnet energizing [W]
120
20
20
300
12
400
34
L3P Collaboration / Precision experiment at L11C
ence between the windings and the cylinder does not
exceed 45 K.
2.4. Magnetic flux return frame
Fig. 15 . General view of central tracking system
of the stresses in the magnet structure. The temperature difference in the coil should not exceed 30 K at
room temperature and 50 K at LN temperature .
To absorb the transient heat load during the energizing of the magnet, an additional amount of 1.5 m3
LHe will be used .
2.3.4 . Power supply and energy dumping system
The magnet is energized with a stabilized semiconducting power supply (15 V/450 kW) during two hours.
The current increase rate is limited by the heating of
the support cylinder with eddy currents . In case of
emergency the discharge takes about 60 s with a maximum voltage of not more than ±600 V. About 350 MJ
is dissipated in two dump resistors (0 .02 SZ each). The
same amount of energy is dissipated in the windings
and about 200 MJ in the support cylinder . To minimize
mechanical stresses the maximum temperature differ-
As mentioned in the introduction, the design of the
iron structure is very similar to the L3 construction .
The flux return frame is octagonal with poles at both
ends . The octogonal "barrel" geometry provides good
mechanical stability for the "barrel" which is fabricated from stacked bars . Each pole consists of a selfsupporting steel crown on the outside and a movable
plug at the centre to give access to the inner detectors.
We have minimised the cost of the structure by
using simple manufacturing techniques, keeping in
mind that the amount of work to be done in the
underground area has to be kept to a minimum. In
particular, considering the past experience of L3, structural welding in the underground area has been minimized. This leads to the necessity of handling, from the
surface fabrication facilities to the underground assembly area, individual pieces up to the full magnet diameter and of large individual weight . The maximum individual weight has been fixed at 800 t, compatible with
the foreseen presence of a temporary 1000 t gantry
crane during the magnet assembly period .
At both ends of the magnet, the already existing L3
octagonal crown is provided as a self supporting structure to take the weight of the roof of the magnetic
barrel . These crowns consist of several independent
plates, 0.3 m thick, stacked side by side . The barrel,
also used from the L3 magnet, is constructed from
individual steel bars arranged in an octagonal geometry. The three bottom octants, laying on an octagonally
shaped concrete cradle, provide a foundation for the
Detector support frame
Carbon fiber Honeycomb structure .
Fig. 16 . Layout of the central tracking system .
L3P Collaboration / Precision experiment at LHC
35
whole assembly . The three top octants rest on the
crown at the extremities . The central hole of the crown
is filled with a movable pole plug which rests on a step
built into the corresponding crown. The plug is optimized to get a constant 1.8 T induction in the iron .
This heavy pole plug can be displaced in the beam
direction, on grease pads, to give access to the inner
part of the magnet to load the superconducting solenoid
system and the inner detectors. The construction of the
movable plug is very similar to the construction of the
crown. The grease pad system foreseen is a straight
extrapolation of that used to support the 675 t L3
magnet doors and the 340 t L3 support tube . This is a
step by step system which allows movement in 1 m
steps. To move the pole plug the load is transmitted to
the grease where the pressure reaches 500 bar and the
friction factor is less than 0.001 .
3. Central tracking system
3.1 . Introduction
The layout of the central tracker system is shown in
figs . 1, 2, 15, and 16 . It consists of a barrel covering
J ,q I < 1.4 and a forward tracker at 1.4 < 1771 < 3.0 . In
accordance with the discussion on design considerations presented in chapter 1 .3, this central detector is
designed to have following properties :
- It provides a resolution :
Op/p = 1% at
pt
= 100 [GeV/c ] .
- It contains a total of 0.03 Xo of material in order
to minimize the photon conversion and to avoid
degrading the resolution of the electromagnetic
crystals .
- The first detector is 1.65 m away from the beam
line, thus it is not reached by most background
particles, which are restricted to lower radius by the
2 T magnetic field.
Fig. 17 shows particles arising from a single bunch
crossing (at a luminosity 10 34 cm -2 s- ') in the UP
detector. The major detector background is due to
low p t charged particles produced in "minimum bias"
interactions (fig. 18). The magnetic field curls the trajectories of these particles, which can induce muliple
hits in the trackers (fig . 17). Because the background is
considerably higher at small 0 (see fig. 19), different
detector technologies are specified for the barrel and
forward regions, as described below.
Fig. 17 . Charged particle tracks from 12 minimum bias events
corresponding to one bunch crossing at luminosity 10 34 cm -z
s -I in 2 T magnetic field.
at the interaction point (IP) by measuring their sagitta
in two concentric superlayers composed of drift tubes
and proportional chambers .
High coordinate accuracy is needed in the bending
(r/o) plane in order to obtain a precision momentum
measurement. This accuracy is achieved by sampling
the track position at several measurement planes within
_4
C
~3
T
(0
3.2. Barrel tracker
The barrel spans the rapidity range 1771 < 1.4 (see
fig. 16). The barrel tracker (shown in fig. 20) is used to
determine the momenta of charged tracks originating
Fig . 18.
pt
of minimum bias event particles in different
rapidity ranges .
L3P Collaboration / Precision experiment at LHC
36
e°
Fig. 19 Background rate in different regions of central
tracker.
the two superlayers . By measuring the bending coordinate N times, the average measurements improves in
accuracy by a factor of v/N ; this multi-sampling technique has been successfully applied in the muon chambers of the L3 experiment [14-18].
The two barrel superlayers are located at average
radii of 1 .65 m and 2.7 m. Although the most sensitive
momentum determination is achieved when the three
measurement sites are equally spaced, the inner superlayer (which effectively measures the sagitta of a track
between the IP and outer layer) is displaced to a
2 Drift Tubes
R=2.7m
125p/d2
0°
120°-120°
-
- Prop . Chamber
3mm/J18
2 Drift Tubes
125p/J2
D
4 Drift Tubes
R = 1 .65
0°
m 120°-_
-120°_
125p/v4
Prop. Chamber
1.5 mm / v18
4 Drift Tubes
125p/A
E
R=0
Fig. 20 . Layout of barrel trackkes .
slightly larger radius . This is performed to gain a
considerable background reduction (i .e . a factor 1 .5)
with less than 4% change in the momentum precision.
The inner superlayer, at R = 1 .65 m, performs a
precise coordinate measurement using 8 planes of drift
(straw) tubes [19], which are grouped into two clusters
(each 4 tubes deep), separated radially by 20 cm.
The radial separation within a superlayer introduces
a local lever arm that helps in rejecting background
tracks, which are mainly at low p, (fig . 18), thus curl
significantly in the magnetic field (fig . 17). In contrast,
the high momenta particles from the events of interest
have almost straight trajectories, hence are better distinguished (see fig. 20).
The straw tubes are made of thin (25 p,m) plasticwall tubes with a diameter of 5 mm . The tubes contain
a CF4 + Ar +'SOC4H to gas mixture and operate in the
proportional mode . This gas mixture is chosen because
it has a large electron drift velocity (100 wm/ns). This
property provides a fast (drift + shaping time < 45 ns)
coordinate measurement in each tube with an accuracy
of 125 P,m [20] .
A measurement in the inner superlayer (containing
8 layers of straw tubes) thus provides an accuracy
125 [Lm/ vr8 = 45 wm . The length of the straw
tubes is bounded by the maximum occupancy (- 1%)
that can be tolerated without significant loss of reconstruction efficiency . Accordingly, the inner superlayer
tubes are each 30 cm long.
In order to aid pattern recognition in the tracker
barrel, a set of proportional chambers are sandwiched
between the two layers of drift tubes. Three successive
planes of these chambers are at 0°, 120°, and -120°,
which causes a legitimate particle track to produce
signals on a triplet of wires that form an equilateral
triangle . A fast validity check can then be performed
(i .e . the wires hit in the first two chamber planes
uniquely determine the wire that must fire in the third,
such that all three wire addresses always sum to the
same constant), thereby resolving ambiguities that can
be expected from neutron albedo and a noisy, high-rate
environment . Chambers of this type were successfully
used to resolve high background, multi-track events in
experiments ran at luminosities above 10 36 cm -2 s -1
[2], where the rate per 1 m2 chamber plane was 20
MHz.
To keep material at a minimum these pattern recognition chambers are supported on very large thin
frames, see fig. 21a, and possess on-chamber sparsification electronics that reduce the required cabling bulk
by at least a factor of 16 . The compresive force of the
wires is supported by Rohacell planes (0.05 g/cm 3)
with inlayed C-fiber supports . The 0° plane is segmented into 1 m long overlapping sections, see fig. 21b.
The frames of the large modules also overlap in r(b, in
order to attain full acceptance .
37
L3P Collaboration / Precision experiment at LHC
48m
a)
0 87m
1
120°
+120'
- 120° b)
v
Rohacell
0°
L
ASICS
Fig 21 . Pattern recognition chamber installed in the outer
barrel layer .
by converting photons arising from-rr' decays . In order
to address these concerns, provisions are taken to
minimize the amount of material used in the mechanical structure of the trackers . Accordingly, the thin
plastic walls of all drift tubes and proportional chambers of the central tracker are designed to introduce as
little as 0.03 Xo .
A special R&D program is dedicated to minimizing
the material needed to supply gas and anchor the
wires. It is important to notice that the double layers of
drift tubes within one superlayer will pinpoint ti conversions in the frames by detecting a track after (but
not before) the pattern recognition chambers .
The main features of the barrel tracker are summarized in table 6.
3 .3. Forward trackers
These chambers are filled with the same gas mixture used by the drift tubes. The introduction of CF4 is
known to lower the aging rate of gas detectors [21] .
This is an important feature, considering the large (and
somewhat unpredictable) background rates expected at
the LHC.
The outer (at R = 2.7 m) and inner superlayers have
similar structure. The outer superlayer, however, contains four layers of drift tubes, thus produces a net
bending-plane coordinate measurement of o" = 125
win/r=62 win.
Since the background expected in the outer tracker
is about a factor of 10 lower than at the inner tracker
(see fig. 19), the length of the anode wires in both the
drift tubes and proportional chambers can be increased
from 30 cm to 1 m, and the anode wire pitch in the
proportional chambers coarsened from 1 .5 mm to 3
mm .
As indicated above, the measurement of particle
momenta can be degraded by multiple scattering in the
tracker structure . Excessive material in the tracker
region can also create additional charged background
Forward (backward) trackers (fig . 16) cover the rapidity range 1.4 < 1771 < 3.0 . As in the barrel, charged
particle momentum is determined by a sagitta measurement between two superlayers . Because of the
projection onto the bending plane, particles penetrating the forward detectors encounter an analyzing length
( f 1) that decreases with polar angle 0 (as opposed to
the barrel arrangement, which results in a f 1
independent of 0). The two forward tracker superlayers are situated along the beam axis, at 3 m and 5 m
from both sides of the interaction point.
A forward superlayer consists of several proportional chamber planes in combination with microstrip
gas chambers (MSGC) [22] installed at small angles .
These components are grouped into two sublayers of
similar structure. A schematic of such a sublayer is
shown in fig. 22 .
The forward chamber arrays are designed as a series of disks centered on the beamline, with anode
wires stretched in the radial direction. Fig. 22 portrays
slices of these disks as arranged in a sublayer (the disks
are also radially segmented, as will be discussed below) .
Table 6
Parameters of the barrel central tracker
Parameter
Inner superlayer
Outer superlayer
Total length [m]
Radius of first drift tube layer [m]
Radius of second drift tube layer [m]
Length of drift tubes [m]
Occupancy/(tube x bunch) at 10 34 cm -2 s -1
Number of layers m drift tube layer
Measurement accuracy m the bending plane [win]
Total number of drift tubes (TDC channels)
Radius of proportional chamber layer [m]
Length of proportional chamber wires [m]
Total number of proportional chamber wires (digital channels)
6
1 .55
1 .75
0 .3
0.006
4
45
331752
1 .65
0 .3
229 940
9 .8
2 .6
2 .8
1 .0
0 .0009
2
62
135 717
2.7
1 .0
166430
L3P Collaboration / Precision experiment at LHC
38
sponding wires m neighboring disks by 1/4 of their 0
pitch. This staggered arrangement of chamber disks
produces an effective sublayer position resolution that
is a factor of 4 better than that achieved with a single
chamber. This superlayer arrangement thus achieves a
net coordinate resolution of: (sectorwidth/(12 X 4)).
The anticipated flux of charged particles through
the forward trackkes is plotted as a function of radius
in fig. 23 (for the inner superlayer) and fig. 24 (for the
LO
ONE"
outer superlayer). The chambers spanning the tracker
disks are thus radially segmented into concentric annu-
lar regions, in order to maintain low occupancy and
keep a close wire spacing (hence preserve measureFig. 22. Schematics of forward tracker sublayer .
A superlayer is composed of two sublayers, each of
which contains four disks. Successive disks in a sublayer are rotated about the beam axis (relative to their
predecessor) through an angle that displaces the corre-
ment granularity) . According to the expected background flux at the chamber location, a tracker disk is
divided in up to four such annuli, which are labeled as
regions 1-4 in figs . 22-24. Proportional chambers cover
the regions at larger radius ; the wires are spaced such
that their minimum separation is 2 mm . The disks
composing the superlayer closest to the IP (situated at
3 m along the beam axis) contain two annular regions
Table 7
Parameters of the forward central tracker
Distance from interaction point [m]
Radial span of MWPCs [m]
1
2
3
Number of layers of MWPCs
Chambers with staggered (b pitch
Pattern recognition chambers
MWPC measurement accuracy in the bending plane [Win]
Number of channels m MWPCs
1
2
3
Total on both sides
Occupancy/(MWPC channel Xbunch) at 10 34 cm -2 s- I [%]
1
3
Radial span of MSGCs [m]
1
2
Number of layers of MSGCs
Number of channels m MSGCs
1
Total on both sides
Occupancy/(MSGC channel X bunch) at 1.6 X 1034 CM -Z s -I [%]
1
2
MSGC measurement accuracy in the bending plane [win]
Total number of channels (digital readout)
First superlayer
Second superlayer
3
5
0.8-1 .1
1 .1-1 .7
09-1 .2
1 .2-2 .0
2.0-28
4x2
2x2
100
4x2
2x2
100
3t200
42 000
30 000
46 800
75 600
304000
146400
1.5
0.8
1 8
0.8
0.2
0.32-0.5
0.5-0 .8
052-0.9
IX2
IX2
20 000
32000
10400
34 000
08
05
40
250400
0.5
68 000
40
372800
L3P Collaboration / Precision experiment at LHC
EU
N 4
Nd
3
Ûc
L 2
â
Radius of Forward Detector (M)
Fig . 23 . Dependence of background in the forward tracker at
3 m.
of proportional chambers (regions 3.4 ; see figs . 22, 23).
The superlayer at larger distance (5 m along the beam
axis) is composed of disks split into three such annular
proportional chambers (regions 2, 3, 4 in fig. 24).
The segments of the forward tracker closest to the
beam are covered with microstrip gas chambers (regions 1 and 2 in figs . 22, 23, and region 1 in fig. 24),
with radially directed strips pitched at a 0.2 mm minimum width. There is one such layer of MSGC contained in each sublayer of the forward trackers (thus
the MSGCs appear in only one of the disks portrayed
in fig. 22).
To measure the radial coordinate of tracks and
facilitate pattern recognition, each sublayer is complemented with two additional proportional chambers,
having anode wires at ± 120° (these are shown at left in
fig. 22).
As in the barrel trackers, a fast, low aging-rate gas
mixture of CE, + Ar +'SOC 4H ,o is used in all the
forward chambers .
6
c
s
U
N 4
m
Q
LU 3
C
2
CDa
_,
0
06
16
24
Radius of Forward Detector (M)
Fig . 24 . Dependence of background in the forward tracker at
5 m.
39
The measurements accuracy of the forward system
will be - 100 p,m at larger angle (where measured by
radial proportional chambers) and - 40 pm near the
beamline (where measured by MSGCs) . The parameters of the forward tracking system are listed in table 7.
3 .4 . Mechanical accuracy and alignment
The large volume of the UP inner tracker, together
with its low density (thus extremely small multiple
scattering) enable a high-precision momentum measurement to be performed. As outlined in the example
given below, a resolution of Ap/p < 1% can be
achieved at p = 100 GeV, provided the mechanical
accuracy is known to 0 = 30 pm .
The trajectory of a 1 TeV muon in a B = 2 T field
over f L = 2.8 m will produce a sagitta of:
s = 0 .3Bf 2 /(8p t ) = 600 p.m .
Scaling from 1% at 100 GeV/c, the expected resolution becomes Op/p = As/s = 10%, implying that
As = 8 2 +,J2 <_ 60 p m, where S = 53 p,m is the resolution of a multi-chamber superlayer . Consequently,
we obtain a requirement on mechanical accuracy :
4 < 30 p,m .
This is a very tight alignment tolerance, which demands:
- A carefully designed and highly reproducible support structure, with accurately predictable deflections under load .
- Special instrumentation for high-accuracy initial
survey and alignment.
- Dynamic monitors that precisely record subsequent deviations caused by temperature, pressure,
magnetic field, and other environmental changes.
Existing collider detectors (with the exception of
L3) realized alignment and survey accuracies that are
an order of magnitude coarser. At UP, however, we
note two essential advantages that will aid significantly
in attaining a 30 p m alignment:
- The inner volume is a large open space, allowing
unobstructed alignment lines-of-sight and a monolithic support structure.
- Application of previous L3 experience and precision instrumentation.
Indeed, 30 p,m accuracies were achieved in all 16
modules of the L3 muon detector [14,15] by initial
alignment with subsequent monitoring . This bias been
verified by measuring the momentum resolution in
Z - p,' p, - decays (see fig. 25).
The total volume of the L3 muon system is 1000 m3.
It is divided into 16 modules, each of which has a
volume of (6 X 6 X 3) m3 and weighs 5 .8 t (the entire
UP central tracker is thus comparable in size and
volume to a single L3 muon module). The top plot of
40
L3P Collaboration / Precision experiment at LHC
280
200
Ê
E
120
08
04
E
m
0
m
â
w -04
ô
-oe 1
0
40
1
1
,
1
t
1
90
180
270
Rotation Angle
1
1
360
20
E
Fig. 25 . Momentum resolution for Z ~IL + W- at
L3 .
0
20
40
80
120
Weeks
fig. 26 shows the maximal predicted and measured
deflection of an L3 module under a 360° rotation,
during which the structure remained stable to within
the measuring accuracy of 6 p.m .
The L3 octants track the position of the three
precision chamber layers that measure the sagitta (i .e .
momentum) with specially developped and inexpensive
"straightness monitors" [23], which are depicted in fig.
26 . Light from an LED mounted on layer 1 is projected
through a lens on layer 2 onto a quadrant diode on
layer 3. As calibrated on an optical bench, these points
align perfectly when all quadrants receive equal light.
When installed on the chambers, the detected illumination imbalance thus provides a precise measure of
the three-point misalignment . The bottom plot of fig.
26 shows the behaviour of an L3 module over two years
of running, as measured by these monitors . Better than
10 p,m absolute and 5 pin relative accuracies were
maintained .
Figs . 27 and 28 shows the L3P tracker being supported by a carbon fiber structure with honeycomb
cylinders. The entire detector is light (i .e. 3 t) . Similar
straightness monitors are used to align the superlayers
to one-another and the IP, as schematically depicted in
fig. 28 . The technologies developed at L3 will thus be
adequate for attaining the required 30 p.m mechanical
alignment.
The tracker structure is self-contained and internally aligned . The tracker is suspended from the
calorimeter and adjusted by three kinematic mounts at
each end. Straightness monitor components are
mounted on the two superlayers (photodiode on the
outer and lens on the inner) . A small, lightweight
cylinder at the center of the tracker (termed the "alignment reference" in fig. 28) carries radiation-hard light
sources that illuminate the straightness monitors, which
are between the corners of the fiducial volume at the
Fig. 26. The L3 alignment system : L3 straightness monitor (a),
Measured and calculated deflection during 0°-360° rotation
(b), Alignment accuracy over two years of operation at L3 (c).
edges of superlayer boundaries (as depicted by dotted
lines in fig. 28). This cylinder is in turn located relative
to the beam pipe (to which the interaction point is
determined) by touchless proximity sensors. If the beam
tube is not sufficiently stable, the beam reference can
be acquired by inserting additional monitors that track
the machine quadrupoles. The beam can also be more
KINEMATIC
MOUNT
/
RODS
Fig. 27 . End view of central trackkes showing support structurc
41
L3P Collaboration / Precision experiment at LHC
Detector support frame
Carbon fiber Honeycomb structure .
Alignment
reference
Fig. 28 . Mechanical structure of central trackkes ; elements of the alignment system are shown.
directly located w.r .t. the reference cylinder by using
beam position monitors (i .e . pickup electrodes or flux
loops fixed to the beam tube). The precision calibration of such devices and their operation in the 2 T
magnetic field will require additional R&D.
The entire tracker thus forms a closed system, with
only one "global" reference required from the outside.
The forward, barrel-left, barrel-right, and backward
modules are attached to a common structure, and are
independently aligned by straightness monitors . Since a
precision of 30 wm is required from the sagitta measurement, the straightness monitor elements are referenced directly to the precision-machined end plates
that locate the drift tube sense wires, thereby incurring
minimal alignment transfer error.
Fig. 27 shows an end-view of the chambers, as
mounted in concentric support cylinders that are held
in position by cross-braced tension rods. Besides resulting in a light structure, this mounting strategy exclusively applies linear forces, insuring that the structure
can be reliably modelled (as was accomplished at L3
[15]). This is not possible if plates and welds introduce
additional tensor forces that can change with temperature gradients. For thermal stability, carbon fiber mate-
rials (< 3 ppm/°C) are used, which have a low radiation length, yet maintain high strength .
To avoid induction shocks from changes in the
magnetic field, the support cylinders are longitudinally
interrupted by an insulating strip.
3.5. Readout electronics
The data acquisition system will accept timing information from about 500k drift tubes for time-to-digital
conversion and about 1000k channels of digital (binary)
information from proportional chambers and gas microstrip detectors (see table 8).
Because of the high channel count, we are planning
to site much of the electronics at or near the detector
(i .e . certainly the preamplifier/ discriminator/ sparsification ASICs, and possibly the drift tube TDCs &
digital pipelines as well).
The drift tube TDCs (fig . 29) will measure the drift
time relative to the most recent beam crossing (i .e .
beam gate) with a 1 ns resolution . Accordingly, only 4
bits of TDC range are required to span the full 15 ns
bunch crossing period . The TDC output is sampled by
a 5-bit digital pipeline, clocked synchronously with the
Table 8
Associated electronics requirements
Signal
Straw drift tubes
Prop . chambers
Gas microstrip
Timing
Hit
Hit
Resolution
[ns]
1 ns
15 ns
15 ns
Dynamic
range [ns]
60 ns
Electronics
Nb . of channels
Disc ., TDC
Disc .
Disc.
500k
1000k
200k
42
L3P Collaboration / Precision experiment at LHC
30
C
~20
Ô
DISC
Fig. 29 . Readout channel for time measurements .
beam gate (four bits are used for the drift time information, with an additional bit to signal the hit occurance) . The digital pipeline is planned to be 128 locations deep, allowing a first level trigger decision delay
of 2 ws . Upon receiving a first level trigger, the contents of the pipeline location corresponding to the
accepted event (plus the contents of the 3 previous
stages) are transferred to a local buffer and propagated
through the readout chain . In this fashion, the 60 ns
history before trigger signal, i.e . the whole range of the
straw tube drift time, will be available for subsequent
analysis .
Nearly identical readout electronics will be used for
both the proportional chambers and gas microstrip
detectors (see fig. 30). Since time information is not
produced, these channels will not contain TDCs, hence
their digital pipeline will be only 1-bit wide . Only a
single pipeline location needs to be transferred upon
receipt of a level 1 trigger, if the proportional chamber
signals arrive within a single beam gate (otherwise
previous pipeline locations must also be included).
3.6 . Performance
3.6 .1 . Momentum resolution
As previously discussed, the central tracker produces a momentum that is derived from the track
sagitta determined from the two superlayers . This approach requires that the interaction point is known to
20 p m RMS in the bending plane and to 5.7 cm RMS
along the beam axis .
LEVEL 1 TRIGGER
DISC .
Fig. 30 Readout channel for proportional chambers .
cc 10
0
20
40
60
80
0°
Fig 31 . Central tracking system momentum resolution at 100
GeV/c.
The momentum resolution of the central tracking
system is shown in fig. 31 for 100 GeV/c particles.
3 .6.2. Pattern recognition
The - 45 ns latency of the drift tubes is significantly longer than the 15 ns bunch-crossing period at
the LHC. As a result, background from several bunch
crossings (see fig. 17) will superimpose, correspondingly increasing the effective occupancy of the inner
trackers . Moreover, since the rotation period of spiraling background particles is rather long (30-60 ns per
cycle), particles from previous bunch crossings will
contribute to the instantaneous background included
with the reconstructed event (this effect has been accounted for in fig. 19).xxxx
We have performed a Monte Carlo study [241 to
examine the efficiency of momentum reconstruction in
the central tracking system for isolated charged particles (1 .e . leptons produced in processes such as H° Z° Z ° - 4 ( or heavy gauge boson decays like Z' f 1 r) .
The leptons were reconstructed in the presence of
"minimum bias" background corresponding to five superimposed bunch crossings (75 ns) at LHC luminosity
(1034 cm -2 s -1 ).
An important part of the pattern recognition and
analysis will incorporate the information from other
L3P subdector systems. For example, reconstruction of
an electron relies on matching the position and energy
of a shower measured in the high-granularity electromagnetic calorimeter (see section 4) with the direction
and momentum of a particle track measured in the
central tracker, which will provide an accurate measurement (^- 1 mm) of the impact coordinates .
Muon momenta are independently measured in the
muon system (see section 6) . In the barrel region of
central tracker, a muon reconstruction algorithm uses
the measurements of the background-free outer muon
system (which estimates the momentum with - 15%
43
L3P Collaboration / Precision experiment at LHC
1 00
1 .00
0
a
U
Û
C
C
0 95
0 .95
W
W
Eff = 0 982 ± 0 003
080
080
1
10
10 2
T
Pk
Fig. 32. Muon momentum reconstruction efficiency at different momenta in barrel trackers .
accuracy) to identify the locations of candidate muons
in the central tracker.
This information defines a fiducial zone in the inner
tracking system within which the muon track must
appear . Combining the corresponding tracker data with
the muon system measurements (see A, B in figs . l and
2), we can obtain a - 6% momentum accuracy using
the outer superlayer of the inner tracker (C). The final
resolution of - 1% at 100 GeV/c is obtained when
incorporating data from the inner superlayer (D). A
momentum reconstruction efficiency close to 100% is
reached for muons traversing the barrel region (fig .
32).
Fig. 33 shows the reconstructed muon resolution at
100 GeV/c with and without a 60-event minimum bias
background, plotted by 1711 < 3 . Fig. 34 shows the
corresponding track reconstruction efficiency . Both the
resolution and efficiency in the forward regions (1 .4 <_
1771 5 3) are different from those in the central region
(77 < 1.4) because of the of high background density.
0°
Fig. 34 . 100 GeV/c muon momentum reconstruction efficiency in the forward trackers.
500k such devices will be needed for UP). Prototype
tubes of 25 wm wall thickness, 5 mm diameter, and 1 m
length have been constructed at ETH Zurich . Particular attention is given to reducing the material introduced through gas and electrical connections .
- A corresponding study of suitable drift gases at
MIT has measured drift velocities and examined the
behavior of several candidate mixtures, leading to the
choice of Ar : CF, : iC,H,; see fig. 35 . Additional tests
are underway to determine the exact time-distance
relationship, and optimize gas performance.
- Designs of the tracker structure must be evaluated . Accordingly, they will be modelled and analyzed
via the NASTRAN software package. This procedure
3 .7. R&D program
- An investigation into the accurate, yet practical
mass production of drift tubes has been started (circa
6
o 0 min bias events
9 60 min bias events
C 4
O
ô
N2
0
0
8
16
24
0°
Fig. 33 . Accuracy of 100 GeV/c muon momentum measurement in the forward trackers.
Fig. 35 . Measured characteristics of candidate gas mixture
(ArCF4 :'C4H 10 (40 :40 :20)).
44
L3P Collaboration / Precision experiment at LHC
was used at L3, and produces estimates of the static,
dynamic, and thermal structural response .
- Although the straightness monitors are now a
sufficiently mature technology for our applications, appropriate proximity sensors and beam position monitors must be researched .
- Cost-effective production of proportional chambers with Rohacell carriers are being studied and prototypes are under design . The chambers will be tested
for their longterm stability. Simple techniques of
mounting the readout ASICs are being studied. The
behavior of these chambers with the CE, gas used in
the drift tubes is being investigated .
- Low-cost, radiation-hard electronics for chamber
readout and data processing is under study by ETH
Zurich and LeCroy Research, Geneva .
- Gas microstrip chambers will be investigated with
CF4-based gases in a magnetic field.
4. Electromagnetic calorimeter
4.1 . Introduction
The high-resolution crystal electromagnetic calorimeter makes it possible for this detector to study electroweak symmetry breaking from the LEP limit up to
the LHC limit. The use of crystals at large radius
enables the detection of the low-mass Higgs, so that
the entire Higgs mass range can be covered without
any missing regions. In addition to good performance
on all of the standard physics topics, the excellent
electromagnetic resolution also makes it an ideal tool
to search for unexpected phenomena.
The major criteria that determine the design of the
electromagnetic calorimeter are:
- The calorimeter should maintain the highest possible electron and photon resolution under all conditions.
- The detector should be able to operate at the
maximum LHC luminosities, and maintain clean detection and trigger.
- The large number of elements requires a system
that is conceptually simple . As much information as
possible should come from the crystals themselves,
without relying on other, specialized subdetectors .
- The large uncertainties that exist in our present
knowledge of cross sections and backgrounds dictate a
detector design that is as insensitive as possible to
increased backgrounds .
In order to achieve these goals, the radial location
of the detector, R, the maximum coverage in pseudorapidity 71, and the operating field, B, should be chosen
to minimize the effects of pileup and vertex uncertainty, while maximizing the resolution of the rest of
the detector and the cleanliness of the trigger and
lepton isolation. Pileup, the deposition of energy in the
detector from underlying minimum bias events, requires particular attention at the LHC, due to the high
luminosities and the high cross section. The = 7 .5 cm
RMS proton bunch length introduces an additional
problem for photon pair measurement, as an uncertainty in the z-vertex location introduces an angular
error into the reconstructed invariant mass . The rationale for choosing R, 77, and B may be summarized as
follows:
- Pileup Background :
Average pileup energy - 1/R'
RMS pileup energy - 1/R.
Charged particle cutoff pr - BR .
Pileup (and other background) increases as 71 increases.
- z-uertex :
Effect increases as 71 increases.
Effect decreases as R increases.
- Granularity :
Cell coverage (071 X 0O) becomes smaller as R
increases (because the transverse dimension is fixed by
the size of the shower).
Increased granularity means cleaner isolation cuts,
better-rr o rejection and lower front-end trigger rate.
- Tracker Resolution :
e and W resolution _ 1 /(BR`) .
Therefore, R should be as large as possible, and 71
should be large enough not to lose acceptance, but not
so large as to introduce excessive background . In addition BR and BR Z should be as large as possible (the
effects of magnetic field are discussed below) .
Pileup causes an effective "noise" in the calorimeter because in any event there is an uncertainty in how
much energy was deposited in a particular region of
the detector . Fig. 36 shows the RMS pileup from single
minimum bias events in a cell of 077 X 0(h = 0.01 X
035
070
11
1 05
1 40
Fig. 36 . RMS pileup from single minimum bras events in one
crystal of A71 X 0q5 = 0.01 X O.01. The lowest curve shows the
pileup resulting only from photons, and the upper curves
show the total pileup with B = 0, 1, 2 T respectively .
L3P Collaboration / Precision experiment at LHC
45
2
57400
N
15
1
2
Radius [m]
3
Fig. 37. RMS pileup vs radius in 5 X 5 crystals at _5° = 1034
cm -2 s-' and 171 I < 1.4 .
Fig. 38. Vertex smearing contribution to 8m/m for a 100
GeV H -> yy vs radius for Qs = 5.7 cm and 1771 < 1.4 .
0.01 . As seen from the figure, the predominant effect
arises from photons, so that the effect of a magnetic
field in reducing the pileup is small. (The mean deposited energy is irrelevant, as it is simply the pedestal ;
and the effective noise, the RMS pileup energy, is the
uncertainty in the pedestal .) The RMS pileup depends
not only on the luminosity, but also on the distance
from the vertex . In the electromagnetic calorimeter,
the transverse size of the cells is fixed by the dimensions of the shower, and is therefore independent of
the size of the calorimeter . The solid angle subtended
by a crystal tower, however, decreases as R 2 , so that
the RMS pileup decreases as R . This effect is seen in
fig. 37, which shows the RMS pileup for a 5 X 5 crystal
sum at y= 10 34 cm -2 s-' #1 .
A large-radius design overcomes the effect of event
vertex uncertainty . An uncertainty in the location of
the z-vertex of the interaction introduces an angular
error in the reconstructed invariant mass for yy states .
As R increases, the magnitude of this effect decreases.
A small-radius detector must therefore be equipped
with additional detectors to determine the entrance
angle of the photons. This photon pointing would be
accomplished either by dividing the crystal into longitudinal segments and using the two center-of-gravity
measurements to determine an angle, or by introducing
a very fine-grained position detector and using the
position detector plus the center-of-gravity measurement to reconstruct the angle.
The addition of such a photon pointing device would
add prohibitive complexity to the detector . Thus our
approach to this problem is to increase the distance
from the vertex in order to reduce the angular error.
Without photon pointing, the angular error introduced
by simply assuming that the center of the detector is
the event vertex, becomes a function of R and the ,9 impact point of the photon . The angular error for a
photon impacting a crystal with nominal coordinates
(uom, kom) from an interaction whose vertex is displaced from the origin by an amount 8z is given by
8,9 - (8Z/R)/(1 - Cot2 0nom - (8z/R) cot 0.0.) . 8,9
is thus proportional to 1/R (although the actual error
decreases as ,7 increases, the increase in energy as ,7
increases causes the angular error to worsen as the
pseudorapidity coverage of the detector increases) . This
effect is illustrated in fig. 38, which shows the contribution to the mass resolution for a 100 GeV H --) yy in a
detector with 1771 < 1.4 as a function of R . The choice
of maximum pseudorapidity coverage of the detector
involves a tradeoff between increased geometrical acceptance and decreased resolution as 7lmnx is increased. Fig. 39 shows the cross section (for the basic
UP design, and with kinematic cuts) for a 100 GeV
H - yy as a function of 7lmax . The effective background to this signal is shown in fig. 40 . For a fixed
71max, a larger radius design will always have better
resolution because the background decreases as R
increases. At a fixed radius, resolution will decrease as
,7 increases, but the exact amount depends strongly on
#1
Depending on the choice of cross section, Monte Carlo
tuning and understanding of luminosity, the RMS pileup
can vary by a factor of almost two. "Tuned" minimum bias
events [25] tend to be harder than the standard PYTHIA
[26] events, and have an inelastic cross section of 80 mb
rather than the frequently quoted number of 60 mb . In
addition, many authors compute the mean of the Poisson
distribution of minimum bias events per bunch crossing at
the LHC via the formula (n) = Yo- St where 6t is the
time between bunch crossings at the LHC :15 ns . As the
LHC has a filling factor of about 80%, the proper formula
is (n) _ Yor/kf, where k = 4725 is the number of
bunches/beam, and f = 11 .246 kHz is the revolution frequency . For a given luminosity, then, (n) is 25% higher
than that calculated by Yo- St . For consistency, in our
calculations we assume that a given luminosity corresponds to the quantity to be used in the formula (n) =
Yo, 8t, and we use tuned minimum bias events with an
inelastic cross section of 80 mb .
46
L3P Collaboration / Precision experiment at LHC
Fig. 39 . (7- x BR for a 100 GeV H - yy with our design and
cuts vs 77maV
quantities which are imprecisely known (true-rr O rejection, vertex smearing, pileup, etc.) so that a prudent
choice, given financial constraints, is to select 77max in a
region where the derivative of significance with respect
to 71mnx is small.
An additional advantage of a large-radius design is
the increased sr 0 rejection power obtained due to fine
granularity, without the need for a specialized-rr o detector . Because the jet cross section is high, extremely
good jet, and thus-rro rejection is required as the
background caused by misidentifying two jets as two
photons must be reduced by at least 7 orders of magni1000
-0
c
0
v,
800
600
U
cû
°~ 400
am
.Û
w
200
Maximum il
Fig. 40. Background for a 100 GeV H -> yy with our design
and cuts vs Amax .
tude in order to be less than the irreducible yy background H -y-y . As explained later in the study of
physics examples (section 10), with our design we reach
a suppression of 8 orders of magnitude for misidentifying two jets as two photons. Since the -rr° rejection
energy scales linearly with R, the exponential decay of
do-(-rr°)/dp T gives tremendous rejection power as R
increases. Thus a large-radius design allows excellent
jet suppression based on simple shower shape analysis
and minimal hadron calorimeter information .
4 .2. Description of detector
The UP electromagnetic calorimeter system consists of a cylindrical barrel of tapered, trapezoidal
crystals similar to the barrel L3 BGO calorimeter [14],
covering the pseudorapidity interval I 1? 1 5 1 .4. The
forward electromagnetic calorimeter covers the region
1 .4 < I Y7 I < 3, and is described below.
Several crystals [27] are good candidates for a precision electromagnetic calorimeter at LHC (see table 9) .
A liquid krypton option will be described in section 9.
For design purposes, we have assumed cerium fluoride
(CeF3) crystals . The crystals are located at an inner
radius of 290 cm from the interaction point, and have a
front face of approximately 3 x 3 cm' and a length of
23-25 X, i. With a pseudorapidity coverage of 1 ,Y71 5
1 .4, this gives 129600 crystals (240 in 77, 540 in 0) with
~
A77 x 0(h 0.01 x 0.01 . CeF3 properties, which have
also been incorporated in the simulations, are summarized in table 10 .
Several features make CeF3 an attractive candidate
for a crystal calorimeter at the LHC:
- High density
By using high density crystals, the calorimeter granularity will be excellent. Fine granularity, along with
large radius, minimizes the contamination of the energy measurement due to pileup . A high density crystal
with a small Molière radius also aids in the rejection of
the numerous rr 0 produced at LHC. CeF3 has a high
density due to the high concentration of Cc 3+ ions in
Table 9
Candidate crystals
Crystal
Density
[gm/cm3]
CeF3
BaF2
GSO :2 .5%Ce
CsI(Br)
ThF4
BaLiF3
LiYbF4
6.16
4.89
6.71
4.55
6.32
5.24
6.09
Radiation
length
[cm]
1.68
2.06
1.39
1.86
118
2.13
1 .56
Molière
radius
[cm]
2.63
3.39
2.42
3.80
2.71
3.13
2.70
Decay
time
[ns]
20-30
1,620
31
10-30
10,30
2
< 25
Emission
peak
[nm]
300,340
210,310
450
310
315,450
435
450
Radiation
hardness
[Gy]
> 104
> 104
> 106
> 106
v
> 105
L3P Collaboration / Precision experiment at LHC
47
Table 10
Properties of CeF3
Density
Radiation length
Molière radius
Decay time of light
Total light (relative)
Emission peak
d(Light)/dT (at 2010
CeF3
6.16
1.68
2.6
20 to 30
=1 /2
340
< 0.1
BGO
7.13
1.12
2.7
340
1
480
-1 .4
Units
g/cm 3
cm
cm
ns
N
C
0
_U
0
nm
%/°C
the lattice (1 .88 x 10 22 em -3 ), which yields a small
radiation length and Molière radius .
- Good scintillation properties
Only crystals with optically allowed transitions (such
as 5d --, 4f for Cc") can give rise to a fast scintillation .
In some cases, more energetic transitions, involving the
core electronic states of the cations, can produce a very
fast (= 1 ns) UV emission, as with BaF2 [28], however
this is always accompanied by a slow component.
The scintillation mechanism of CeF3 is simple and
well understood . It involves the fast, intense and temperature independent transition 5d --, 4f of the Ce 3+
ion . Two broad emission bands centred at 300 nm and
340 nm (see fig . 41) produce a light yield roughly half
that of BGO, and a decay time of 20 to 25 ns (see fig.
42).
- Good uniformity
The energy resolution of the calorimeter, and particularly the constant term, will strongly depend on all
possible sources of non-uniformity . Intrinsic scintillators like CeF3 are therefore preferred to doped materials, because it is difficult to control the uniformity of
doping . The light collection in a pointing geometry will
introduce a non-uniformity due to the focussing effect,
100
80
É
0C:
m
40
20
0
200
300
400
500
600
700
a. (nm)
Fig. 41 . Transmission, excitation and emission spectra of
CeF3.
0
25
50
Time (ns)
Fig. 42 . CeF3 decay curve.
75
which depends on the refractive index of the crystal.
Cerium fluoride crystal has a refractive index around
1.5, which will limit this effect to a much smaller value
than that for BGO (n = 2.15). In addition, the scintillation yield from CeF3 is almost temperature independent (less than 0.1%/°C), eliminating the need for
thermal insulation of the detector .
- Good radiation hardness
Much progress has been made recently in the radiation hardness of CeF3 which has been shown to be
intrinsically resistant up to 1000 Gray (photons) and
10 12 neutron/em2 irradiation [29] . Radiation hardness
depends on the quality of the raw material . In inorganic scintillators, the scintillation efficiency is generally not affected by the damage, but impurities even in
very small quantities, can create colour centres which
absorb a fraction of the emitted light.
Research is in progress to find ways to identify the
specific impurities which are responsible for the damage in CeF3 and to find economical ways to remove
them from the raw material . A promising approach,
from the economical point of view, is specific doping
[30], which has already been proven to be successful for
the L3 BGO [31]. When electric charges are trapped by
crystal defects or impurities, resulting in color centers,
a specific doping can allow charge transfer from the
traps to other states, where a nonradiative relaxation is
possible .
- Good mechanical properties
Due to the good mechanical properties of CeF3, a
high production yield can be expected after mechanical
processing. Nonetheless, mechanical processing will be
one of the dominant cost drivers, and extensive R&D
is in progress to find economical techniques for cutting
and polishing crystals .
- Economical
Cerium is the most abundant of the rare earths,
with China as the world's main producer of cerium
oxide. It is widely used for several industrial applica-
L3P Collaboration / Precision experiment at LHC
48
0.36
w
028
020
10 2
10
10 3
Energy (GeV)
Fig.
43 .
Intrinsic CeF3 resolution for 1~u .
tions. The preliminary results from the Crystal Clear
Collaboration indicate that a purity of 99 .95% is probably good enough to guarantee a good quality crystal.
Cerium fluoride can be grown by the BridgemanStockbarger technique in a graphite crucible . The high
density of CeF3 limits the crystal size to about 25
radiation lengths, which can be grown in furnaces of
reasonable size and at modest cost.
The intrinsic EM resolution expected from a sum of
5 x 5 crystals (125) has been simulated with GEANT,
and is shown in fig . 43 . In order to keep the highest
resolution, we have minimized the material in the
tracker (3% X, total), as well as the thickness of the
supporting structure. This limits any degradation in
resolution by supporting walls between crystals or any
material before the crystals in which showers may
originate. For the purposes of calculation, the crystal
resolution for a 125, with o, in % and E in GeV, may
be parameterized as
Q
.
É
=
A
N
fo
, 'u
7 ®BGCE'G = G
Ge,
E
where A = 0.94% is the stochastic term, B = 0.14% is
the constant term, C = 0.025% and a = 0 .34 represent
the high energy leakage, N < 100 MeV is the electronic
noise in the 125 , aPU is the RMS pileup in the 125, f
is the electronic pileup factor, and a is the calibration
error. At a luminosity of y= 10 34 cm -2 s-1 , the
resolution may be written as
o-
0.94%
E
®
0 .1%
E
® 0.14% ® 0.025%E ° 34
®
0 .1%
E
® 0.5% .
(2)
4.3. Crystal production in China
China has demonstrated several advantages in
large-scale production of crystals : Nationwide collaboration has been organized with the strong support of
the Chinese Government ; Infrastructure and experi-
ence to produce large quantities of crystals are already
available from mass production of the L3 BGO and
R/D on CeF3 and BaF2 ; China is a country rich in
rare earth resources.
A large collaboration, including research institutes,
universities, and industries, exists today in China.
Members of the collaboration and their specialties
include:
- Crystal growth :
Shanghai Institute of Ceramics (SIC), Shanghai . Five
Bridgeman growing furnaces are in operation at SIC,
as well as several cutting and polishing machines, and
mechanical and optical measurement devices.
Beijing Glass Research Institute (BGRI), Beijing.
Nine furnaces are in operation at BGRI, with 8 more
large furnaces to be installed .
- Mechanical processing :
Zhongnan Optical Instrument Factory (ZOIF),
Zhicheng . ZOIF specializes in mass production of
high-quality optical products, as well as machining and
finishing.
- Quality control:
Institute of High Energy Physics QHEP), Beijing.
- Properties study :
Institute of Crystal Materials, Shandong University,
Jinan.
Department of Materials Sciences and Engineering,
Zhejiang University, Hangzhou .
University of Sciences and Technology of China,
Hefei.
The Chinese Academy of Sciences is coordinating
active R&D at all of these institutes concerning the
properties of CeF3 and the means for crystal production . We are investigating the use of raw materials with
99 .95% purity by performing comparative growth experiments using CeF3 provided by different firms, and
processed from different salt materials. These studies
also include composition'analyses of the raw materials,
as well as investigations on the pre-treatment and
refining processes to be used .
Crystals of good transparency have already been
successfully grown. Transmission is over 85% at 310
rim, and is still around 80% after irradiation with 1000
Gy . Radiation hardness tests are performed on samples using a 60 Co source, in cooperation with the
Shanghai Nuclear Research Institute.
Mechanical processing is being studied with five
cutting machines, two polishing machines and a large
lapping machine. Results are verified with computercontrolled dimension measuring devices, and surface
roughness meters . Optical properties are determined
with spectrophotometers, and UV and X-ray fluorescence measurements are part of the standard crystal
measurements .
Rigourous quality control is required after production, before shipment and after receipt of the crystals.
L3P Collaboration / Precision experiment at LHC
We plan to assure quality control in a manner similar
to the L3 BGO, by performing the following steps :
- Visual inspection to ensure that there are no
cracks or scratches on the crystal .
- Dimension measurements to verify the seven different dimensions, described in the next section, for
each of the 120 different crystal types . Each crystal will
be compared to the corresponding reference standard .
The tolerance for each of the seven dimensions, as
compared to the standard, are +0 -200 wm. The
planarity of all six faces is 50 ~Lm . The endfaces of the
crystal are perpendicular to the longitudinal axis of the
crystal (as defined by the intersection of the two bisector planes of each pair of opposite side faces) to within
50 wm .
- Transmission measurements will be performed by
spectrophotometers .
- Light yield measurements will be performed with
137CS sources . Two end points and a mid-point will be
compared against a reference standard .
- Arrival inspection will be performed at CERN by
repeating each of these measurements .
- Uniformity measurements will be performed at
CERN . After the crystals have been coated, they will
be measured with a cosmic ray bench, similar to what
was used for the L3 BGO .
4.4. Crystal production in Czechoslovakia
Czechoslovakia has expertise in the production of
heavy crystal scintillators, laser crystals and heavy
glasses doped with rare elements . Collaborating institutes include :
- Monokrystaly
PRECIOSA/ MONOKRYSTALY (former CRYSTUR) Company, Turnov, specializes in mass production of BGO, YAP and YAG crystals, as well as R&D
on fluoride crystals (BaFZ and CeF3 ), Cerium doped
Silicate (GSO : Ce), YAP : Cc and YAG : Ce. Large volume BaFZ and CeF3 crystals have been tested in beams
of relativistic particles and nuclei at the JINR Dubna
synchrophasotron . Several Bridgeman growing fur-
-_-_-_-----_-_--_-_-_ -
49
naces, polishing and cutting machines are available .
Mechanical, optical and luminiscence properties are
controlled .
- Tesla
TESLA Company, Research Institute of Nuclear
Instrumentation, Premysleni, produces Csl and CsI(TI)
crystals, and is engaged in R&D of crystals for laser
spectroscopy Bridgeman and Czochralski growing furnaces are available .
- State Research Institute of Glass
State Research Institute of Glass, Hradec Kralove,
produces lead and fluoride glasses and performs R&D
of scintillation characteristics of heavy glasses doped
with rare earth elements (Ce, Nd, etc .) .
Production of the crystal and glasses, as well as
R&D at all of these institutes have been coordinated
by the Czech Ministery of Industry . Experimental investigation of radiation damage effects on crystals and
glasses can be performed by using 60 Co sources at the
Institute of radiation Dosimetry at Prague and nuclear
reactors at the Nuclear Research Institute at Rez near
Prague .
4.5. Mechanical support
The crystal electromagnetic calorimeter is a cylindrical array of crystals with pyramidal frustrum shapes
pointing to the interaction point (see fig. 2) . It is
divided into 540 identical 0 slices and each slice contains 240 crystals with mirror symmetry . The total
crystal volume is about 64 m 3 and the corresponding
weight is 400 t .
A very rigid supporting structure is needed to take
this load with tolerable deformations . As crystals cannot take any part of the deformations, clearance between crystals must be foreseen and a compromise
found between incurred solid angle loss and exceeding
dimensions of the supporting structure . From the experience gained in L3 during the construction of the
BGO calorimeter, a similar principle is used to minimize solid angular loss : structural material is optimized
for the rigidity and resistance needed once in place in
0
0
%G~Illlllli11111-111114i
Fig . 44 . Electromagnetic calorimeter division .
50
L3P Collaboration / Precision experiment at LHC
the experiment . Reinforcements are used during assembly and installation maneuvers to compensate the
progressive loading during assembly, the position
changes during transportation, calibration and installation maneuvers, and occasional accelerations due to
shocks .
Once in place, the structural material rigidity is
enhanced by pre-stressing thin walls which would have
no stability otherwise, and exerting balanced efforts
from spokes with radial symmetry . To avoid shadowing
the crystal front faces, most of the structural material
is placed on the calorimeter outside in the form of a
cylindrical support tube (ST) inserted in the superconducting coil vacuum tank aperture and surrounding the
electromagnetic calorimeter active volume . The ST is
fastened at its ends to the tank end flanges, which are
in turn fastened to the inner face of corresponding
hadron calorimeter (HC) rings. The HC rings are supported by the outer supporting tube which is itself
fastened to the magnet iron yoke ends . The ST and
electromagnetic calorimeter barrel have forward-backward (77) and left-right (¢) symmetries .
Crystals of truncated pyramidal shapes point to the
interaction point in the 77 direction, and point tangentially to a cylinder with the interaction point as centre
in the 6 direction, to prevent escaping particles (fig .
44).
Each crystal has a constant front area of 1150 mm 2 ,
and a height of 390 mm . For reasons of economy in the
processing of the crystals, the front face (frustrum
small base) and two side faces are made perpendicular
to each other to define a reference trihedron (fig . 45).
There are therefore 120 different crystal shapes (right
hand) and their 120 symmetrical counterparts (left
hand). Seven dimensions specify each of the 120 crystal
Constant Front area
A = 1150 mm2
Fig. 45 . Crystal geometry .
Fig. 46 . Photodiode and capsule.
types: the height of the crystal, H (= 390 mm), the 0
width B and small and large rl widths A and AS of the
front face, and the corresponding dimensions BP, AP
and ASP of the back face .
As with the L3 BGO calorimeter, every crystal is
equipped with a photodetector located on its back face
with a glued coolant-, gas- and lighttight plastic capsule . The capsule contains :
- The glass envelope of the vacuum photodiode . It
is gas tight and one of its faces is glued with a special
adhesive transparent at 375 rim to the back face of the
crystal.
- A preamplifier board.
- Cables for power supply, readout and control.
- Cooling ducts connected to a small metal heat
exchanger in good thermal contact with the preamplifier board.
The capsule is rigid enough to transmit the force
setting the crystal in correct position and to evenly
share it on the crystal back face to avoid dangerous
stress concentrations (fig . 46).
The electromagnetic calorimeter is divided into 16
rings of 8 different shapes. Rings are two-by-two mirror-symmetric. Each ring is made of 36 modular identical crates .
One crate contains 15 crystals in ¢ and 15 crystals
in -7 . A crate with its 225 crystals weighs about 700 kg
(see fig. 47). Each crate is a box with 5 mm thick
metallised carbon fibre composite or 2 mm thick titanium alloy side walls, with parallel faces in ~b and
tapered faces in 77 .
There is no structural partitioning wall between
each crystal. The bottom of the crate is a light sandwich panel with two carbon composite skins 2 mm
thick and a 25 mm Nomex honeycomb core . Crystals
are held in position by a frame which closes the back of
51
L3P Collaboration / Precision experiment at LHC
Table 11
LHC readout requirements
Requirement
Wide dynamic range
High bandwidth
High speed
Pipelined trigger
High Resolution
Fig. 47 . Crystal crates .
the crate. The frame carries pressing devices pushing
on the crystal capsules so that crystals are accurately
set on the corresponding step-like bottom face of the
crate. The pressing force is twice the crystal weight (i .e .
60 N) so that a crystal will be held in place at any
position of the crate and will not press on its neighbours during further operations .
Some freedom is necessary between crystals after
the optical separation film is inserted, because of the
combined dimensional tolerances of the crate and crystals, and to allow for expected elastic deformation of
the crate during handling operations . The 225 individual forces applied on the crate bottom are reacted by
equivalent 135 kN tension in the side walls, which
enhances their rigidity so that assembled crates are
selfsupporting in any position .
The crate side walls are prolonged at the front and
at the back to form a rim with holes. These holes will
be used to fasten the crates to the ringassembly tooling, to fasten the crates together and to attach them to
the support tube .
The power, test, monitoring and signal cables corresponding to each channel in a module (225 channels in
a module) and the cooling ducts for the preamplifier
cooling system are attached to the crate and are coiled
at its back .
Consequence
Low noise
Low noise
Low electronic pileup, fast shaping
High fidelity acquisition and storage
Jitter-free, low-noise trigger
Excellent linearity and fidelity
information (energy deposited in the calorimeter) associated with a given bunch crossing must be stored in
some fashion for several (tens to hundreds) of bunches
before a trigger decision is available. The combination
of these factors, with the requirement for high resolution, demands considerable attention at each stage of
the readout, as can be seen in table 11 .
Operation of the detector in a magnetic field places
limitations on the choice of a photodetector. There are
several candidate photodevices (avalanche photodiodes, hybrid photodiodes, silicon photodiodes, vacuum
photodetectors, etc.) each with advantages as well as
disadvantages. Large-area silicon photodiodes have
successfully been used as photodetectors for calorimeters operating in the microsecond timing regime . The
largest such system with silicon photodiodes is the L3
BGO electromagnetic calorimeter [14], whose performance specifications are listed in table 12 .
The noise performance of a silicon photodiode system at the LHC may then be roughly estimated as
follows: For a single-component scintillator (like BGO
or CeF3 ), the photocurrent is proportional to
(Q/-r)e-`1r, where Q is the number of photoelectrons/MeV and r is the decay time constant of the
light. If all of the charge can be integrated in a single
bunch crossing (as is the case with BGO at LEP), then
the available signal is simply proportional to Q. At the
LHC, however, one would need some 5 to 10 bunch
crossings to collect most of the CeF 3 charge, so that
the signal available is proportional to Q(T s /r), where
T s is the effective electronic shaping time of the signal
(for a linear system). As the noise for a given preamplifier is proportional to CTS 1/2, where C is the total
4.6. Readout
Electromagnetic calorimeter readout at highluminosity hadron colliders presents unprecedented
challenges in signal acquisition . The dynamic range
requirements equal or exceed those at LEP, with three
orders of magnitude increase in speed going from LEP
to LHC. In addition, the extremely high bunch crossing
frequency necessitates a "pipelined" readout: Because
the time between successive interactions is much less
than the time needed to form a trigger decision, the
Table 12
L3 BGO readout
Crystal back face
Readout
Capacitance
Effective dynamic range
Intrinsic noise
Total noise
Residual on next bunch
9cm2
3 cm2 Si photodiode
75 pF/cm2 +75 pF Preamp
> 18 bits ( < 1 MeV to 250 GeV)
0.8 MeV (= 900 electrons RMS)
1 MeV
< 10-5
52
L3P Collaboration / Precision experiment at LHC
capacitance at the input of the preamplifier, the signalto-noise ratio at the LHC behaves like
S
2
/
_
Q 73
N ~(C) 7
'
so that as the shaping time -r s is made shorter, the
S/N ratio degrades as 7s/2 .
In going from L3 BGO (LEP) to L3P CeF3 (LHC), 1
MeV of noise at LEP will become more than 100 MeV
of noise at the LHC, so that the noise performance of
a silicon photodiode system at the LHC would be
unacceptable for the physics goals. Further, although
there now exist large area silicon photodiodes with
thicker depletion layers than those used on the BGO
(thus with somewhat lower capacitance per cm 2, but
with increased leakage current and susceptibility to
radiation damage) such diodes with fast shaping amplifiers would still produce at least 75 to 100 MeV of
noise per crystal . As 430 MeV of noise would contribute 0.5% to the mass resolution for a 100 GeV
H - -y-y, silicon photodiodes, well suited to crystals in
the g,s regime, will require extensive R&D for use at
LHC.
The crystal calorimeter will therefore most likely be
readout with proximity-focussed vacuum photodiodes .
Some of the properties of vacuum photodiodes, listed
below, illustrate why vacuum devices are best suited to
the physics goals of this detector :
- Capacitance : Vacuum photodiodes have an order
of magnitude less capacitance than silicon photodiodes, hence there is at least an order of magnitude less
noise with vacuum photodiodes .
- Leakage current : Vacuum photodiodes have several orders of magnitude less leakage current than
silicon photodiodes, hence the noise due to leakage
current is negligible with vacuum photodiodes, yet may
be dominant with silicon photodiodes in a high radiation environment .
- Quantum and collection efficiency : With CeF3,
the product of quantum efficiency X geometrical area
matching yields roughly equal efficiencies for vacuum
and silicon photodiodes .
- Rtsetime : Vacuum photodiodes are an order of
magnitude faster than large-area silicon photodiodes,
hence more signal is collected at the LHC, facilitating
lownoise, high-speed readout.
- Radiation hardness : Vacuum photodiodes per se
are radiation hard, as only the envelope may be affected. Silicon photodiodes exhibit performance degradation in high radiation environments which may prove
detrimental for a very high-resolution calorimeter .
- Operation in magnetic fields : Silicon photodiodes
are better adapted to operation in strong magnetic
fields . With proper care, though, vacuum photodiodes
may also be used, as explained below.
Fig. 48 . Schematic of vacuum photodiode orientation relative
to magnetic field axis .
Because the capacitance is lower, the leakage current
is lower and the risetime faster, the noise performance
with vacuum photodiodes is an order of magnitude
superior to silicon photodiodes . Further, the only radiation hardness problem with a vacuum device is the
envelope, whereas silicon photodiodes exhibit gain
changes, leakage current increases and reduced lifetimes in radiation environments . Although additional
gain is possible (phototriode, tetrode, - - - ) for maximum gain stability and ease of operation in a magnetic
field, a photodiode is the favored solution .
Operating a vacuum photodevice in a strong magnetic field requires orienting the tube axis (the electric
field axis) with some angle less than 90° with respect to
the magnetic field axis (see fig. 48). A proximity focussed photodiode can easily operate at fields of 1 T
with only a 20° tilt angle, and no appreciable gain loss
(so that the gain as a function of 77 is constant). Such a
tilt angle has only a small effect on light collection and
mechanics (and, of course, is not required on those
crystals which are already tilted by more than 20 °). A
higher gain device, while providing more signal, requires a much greater tilt angle, and operates with
reduced gain in the field.
Using vacuum photodiodes at 2 T or above may
require an increased tilt angle (obtained by reshaping
the crystal or using prisms). Part of our R/D program
on vacuum photodiode readout is devoted to this issue
in order to finalize the choice .
The electronic readout chain consists of photodiodes with preamplifiers, shaping and gain stages, signal
acquisition, and the higher level readout and trigger
stages . The crystals which form a trigger tower (a 5 X 5
array of crystals) comprise a conceptual readout mod-
L3P Collaboration / Precision experiment at LHC
ule, and share common electrical services . The preamplifiers are located directly behind (or on) the photodetectors, and have short cable runs to shaping and line
driving stages . Because of the high dynamic range and
large bandwidth, the signals are split into two components (a high and low gain channel) directly at the
shaping stage, before proceeding to the signal acquisition electronics.
The data pipeline may be formed either in an
analog fashion, in which case the pipeline consists of a
switched capacitor array with virtual 1st and 2nd level
memories formed by pointers ; or in a digital fashion,
with high speed ADCs and digital 1st and 2nd level
memories . Because of the comparative ease in forming
a 1st level trigger, as well as the higher fidelity of
digitization, we favor the digital pipeline approach .
With the digital pipeline, the multiple range signals
are digitized either by separate ADCs, or multiplexed
into a single ADC. The resulting data, an ADC mantissa, along with identification bits indicating which
range was used, are then stored in digital memory
awaiting trigger decisions. The same data are then
presented to a lookup table (RAM) whose outputs
correspond to a calibrated energy (or transverse energy) to be used in the trigger decision . The trigger
may be formed either as an analog current sum, in
which case the calibrated digital data drive a fast
current-output DAC; or in a digital fashion with a
pipelined adder. Note that the analog sum formed
from digital data differs markedly from an analog sum
formed by adding shaped copies of the raw input signal
in that the data are already calibrated, the effects of
time slewing have been completely eliminated, and the
bandwidth is considerably reduced, so that the resulting signal is much cleaner and more precise than would
be possible with a classical analog sum.
The noise per channel is estimated based upon the
front-end preamplifier noise shunted by the input capacitance. This is the dominant noise term for vacuum
photodiodes as they have negligible leakage currents,
unlike silicon photodiodes . The vacuum photodiode is
coupled to a low-capacitance, charge-integrating
preamplifier with feedback capacitor chosen to launch
a fullscale voltage of 4 V into 50-100 dZ (at these
speeds, proper high speed impedance matching must
be taken into account) . Although we foresee the use of
a high transconductance GaAs front-end, we consider
the conservative estimate of 50 mS transconductance
and 20 pF total capacitance (photodetector + CGS).
The preamplifier signals are received with a polezero
filter, to remove the preamp tail, and a second polezero
or ac coupling stage to remove the CcF3 tail . (Either ac
or do coupling may be employed, but because of the
high rate environment, we prefer a fully dc-coupled, or
almost dc-coupled system for utmost accuracy). A
plethora of subsequent shaping and signal capture ar-
53
rangements are possible (further fast integration or
gated, synchronous integration with reset), but for calculations, we assume the simplest case, a third order
system peaking in 15 ns . In such an arrangement, the
electronic pileup factor, defined below, is 1.15, and the
noise per channel is 4 .2 MeV (for Q = 750 e-/MeV)
or 6.3 MeV (for Q = 500 e-/MeV). The same calculation applied to the L3 BGO system predicts the correct
observed noise. The theoretical noise limit for this
arrangement is lower, because of a factor of 2/3 which
appears in the theoretical noise spectral density. Since
real devices give higher noise densities in practice, we
have not made use of the 2/3 factor. Finally, as the
light produced may vary from crystal-to-crystal, the
final shaping choice may be different, and as the extremely high bandwidth may give rise to correlated
noise (we are concerned with correlated noise entering
the system at the 10 p,V level) for all simulations we
have used an estimate of 20 MeV/crystal.
As the time between bunches at the LHC is short
compared to the time needed to integrate the full
detector charge, the electronic enhancement of pileup
must be included in the overall detector resolution . For
a linear shaping system with voltage sampling, if a,
represents the normalized gain (the height of the pulse
on the eth bunch after "t = 0", normalized to the peak
height) then the increase in the observed pileup due to
the tails of pileup pulses from previous bunches is
given by Qoss =fo pu, where f 2 = Y-a; is the electronic
pileup factor (f >_ 1) . We note that a high-luminosity
machines, it is possible to optimize the choice of f by
observing that as the shaping time increases, f increases, however the noise decreases. Thus, for systems
that are intrinsically noisy with fast shaping, the optimal arrangement may well be to increase the shaping
time, thus reducing the noise while increasing the
pileup, in order to arrive at the lowest value for
(Noise 2 + Pileup2)1/2 . As our design is intrinsically low
noise, we make no attempt to perform such an optimization and will use very fast shaping to allow excellent performance at the highest luminosities.
4.7. Trigger
The designs that are currently favored for calorimetric triggering at high-luminosity hadron colliders
consist of multi-level trigger stages, each stage performing more powerful analysis on the data . The first
stage trigger is generally a fully hardware trigger processor which operates on special trigger data constructed by the Level 1 readout. The final stage is a
fully software trigger processor operating either on the
trigger data or the main readout data .
With such a design, severe requirements are placed
on the Level 1 trigger, which must return a trigger
decision within a few Ws, while accepting input data
54
L3P Collaboration / Precision experiment at LHC
Table 13
Calorimetric trigger
Level
Level 0
Levell
Level 2
Level 3
Description
Timing
Synchronous hardware
processor
Local computation of hits
above thresholds
Synchronous or monotonic
hardware processor
Crude isolation cut and
matching with HCAL towers
Monotonic software processor
Precise isolation cut
Charged energy cut
(with tracker)
Software processors
"Off line" analysis
<< 1 ps
10-100 ws
> 1 ms
10
20
30
40
Trigger Threshold [GeV]
Fig. 50. Trigger rate in the barrel+end caps at _~e, = 1034
cm -2 s -1 for one hit above threshold.
every 15 ns . Such a Level 1 trigger must therefore be
pipelined (each operation of the trigger is clocked by
the beam clock) so that the processing capability of
Level 1 is highly restricted . To overcome this problem,
our calorimetric trigger adds an additional stage, Level
0, which performs simple, local discrimination . The
addition of this level, simplifies construction of the
trigger, and allows one to two orders of magnitude
more computation time for Level 1 . The organization
and functionality of the trigger stages are described in
table 13 . The basic trigger flow is then as follows: An
energy sum, as well as a few bits corresponding to hits
above threshold are formed for each electromagnetic
trigger tower. Level 0 receives the trigger bits, and
forms a trigger decision based on n hits above threshold (n, hits above E,, n 2 hits above E2 , etc.) . If the
Level 0 decision is yes, then the digitized 1st level data
are stored in the local 1st level memory while Level 1
processes the trigger tower-sum data . A Level 1 yes
causes the data in the 1st level memory to be loaded
into the 2nd level buffer, for readout to the Level 2
trigger.
The trigger rates due to background with our
multi-level trigger have been simulated, and are shown
in figs . 49 through 52 . Figs . 49 and 50 show the level 0,
1, and 2 rates as a function of transverse energy threshold requiring a single hit above threshold in the barrel,
and complete detector respectively . Figs . 51 and 52
show the rates when requiring two hits above threshold. These rates assume an effective luminosity of 10 34
em -2 s - I with an inelastic cross section of 80 mb, so
that the mean number of minimum bias events per
crossing is (n) = 12. For most processes, the Level 0
trigger requirements would be
1 Hit ET > 10 GeV
1 Hit ET > 15-20 GeV
in the barrel
OR
1 Hit ET > 15 GeV
1 Hit ET > 15-20 GeV
in the endcaps
OR
1 Hit ET > 10 GeV
1 Hit E, > 15 GeV
10 6
Barrel
0
in the barrel
in the endcaps
OR
{ 1 Hit ET > 40 GeV anywhere .
With these conditions, the trigger rates at y= 10 34
cm -2 s-1 are:
10
20
30
40
Trigger Threshold [GeV]
Fig. 49 . Trigger rate in the barrel at
10 34 cm -2 s -I for
one hit above threshold.
10 kHz from Level 0,
< 1 kHz from Level 1,
10 Hz from Level 2,
with final trigger efficiencies of
> 95% for H - yy
>85% for H -> ZZ > 2e + X, m H >_ 150 GeV.
55
L3P Collaboration / Precision experiment at LHC
106
Barrel
105 ~_
N
104r
40
10
20
30
Trigger Threshold [GeV]
40
10
20
30
Trigger Threshold [GeV]
Fig. 51 . Trigger rate in the barrel at f =1034
two hits above threshold.
CM -Z s -I
for
4 .8. Calibration
Maintaining high resolution with the crystal
calorimeter will require considerable attention to the
crystal calibration and to the design issues which determine calibration stability over the lifetime of the detector. Experience with the L3 BGO calorimeter shows
that several design parameters must be carefully chosen to optimize resolution and calibration stability.
Whereas the BGO was designed to have optimum
resolution at low energies, the CeF3 calorimeter should
have optimum resolution at high energies . Some of the
relevant design parameters are listed in table 14 . Each
of these effects has a major contribution to calibration
stability. Temperature effects, while important for
BGO, are expected to be very small in CeF3. The
nonuniformities and nonlinearities due to geometry,
reflectors, index of refraction and attenuation length
are less critical in CeF3 than BGO. The choice of
calibration energies is also important, because the errors due to nonlinearities depend on the ratio of the
energy being measured to the nearest calibration energy . (A simple gain error, such as mismeasuring the
gain of a crystal by some amount e, introduces a
constant term of c = RMS(e) . A nonlinearity, EMEAS
= a, E +ß,E z introduces a constant term of c =
Fig. 52 . Trigger rate in the barrel+end caps at
cm -2 s - ' for two hits above threshold.
_c/°=1034
RMS(a) (1 - E/ECAL), which then depends on the
distance between the measurement point, E, and the
calibration point ECAL).
The initial calibration and detailed understanding
of the crystal nonlinearities requires excellent incident
momentum and position definition, along with fairly
high statistics . As with the L3 BGO, this initial calibration will be performed in a test beam. The crystal ring
assemblies (described below) are mounted on a rotating table, so that each crystal may be positioned on the
test beam axis. In one 14 s SPS cycle, the two second
beam burst contains a sufficient number of particles to
calibrate the crystal. In the remaining 12 s, the turntable
moves to the next crystal so that < 15 s per crystal are
required for calibration .
After installation, we will perform in situ calibration
monitoring, in order to make small corrections to the
initial calibration constants for any gain drifts . This
calibration monitoring will be done with isolated e±
from W, Z --> e t+ X (in 1771 < 14, more than 90% of
the e } are isolated) . The calibration monitoring makes
use of E/p matching with the tracker, followed by
resolution optimization. (This technique is possible because the original calibration constants are already
known. Such a technique is insufficient for actual calibration ab initio .) With the high resolution of the
Table 14
Calibration stability - CeF3 vs BGO
Effect
dLight/dT
Geometry
Uniformity
BGO
Index of refraction
-1 .4%/°C
R = 55 cm
Wide range used
Optimized low energy
High
Calibration Energy
High energy in situ calibration possible
One, 10 GeV
No
CeF3
< 0.1%/°C
R = 290 cm (angles are smaller)
Small range must be used, and optimized for high energy
Low, better
uniformity
Several, high E
Yes
56
L3P Collaboration / Precision experiment at LHC
tracker, 8p/p = 1% at PT = 100 GeV, the number of
events required to monitor the calibration to an accuracy of s is given by
U )2
s )
2 +(
~l£2 ,
( E )Z +
E
p
(4)
where 8E/E is the crystal resolution (8E/E = 0.5%
for 125, and 8E/E= 1% for 1y) and o-,, is the RMS
pileup at the luminosity where the calibration is being
monitored. To monitor the calibration to an accuracy
of F = 0.3% at 50 GeV thus requires about 10 events/
crystal . The cross section for isolated e } from W,
Z - e ±+ X in -q J =1 .4 is about 7 nb for PT > 30
GeV and about 850 pb for PT = 50 ± 6 GeV. Calibration monitoring to an accuracy of e = 0.3% at E = 50
± 6 GeV can therefore be performed over the entire
detector six times per year (107 s) at y= 10 33 cm -2
s -1 and 65 times per year at J°= 10 34 cm -2 s-1 .
Additional gain monitoring systems, such as light
pulses or electronic test pulses may be employed . Several experiments have used light pulsers to monitor
gain stability, however the problems that will arise at
LHC speeds, due to the vast difference in the light
pulse shape and the scintillation pulse shape, may be
non-trivial to solve. Electronic test pulsing may be
readily performed with the addition of local test pulse
drivers on each channel, however with proper electronic design, electronic gain drift should not pose a
serious problem (the 12000 channel L3 BGO Level 1
readout has maintained electronic gain stability of better than 0.01% over five years of operation) .
4.9. Assembly, installation and calibration mechanics
4.9.1 Calorimeter assembly
The calorimeter consists of modular rings. Each
ring is composed of crates containing 225 crystals . This
modular arrangement facilitates assembly at each stage.
For crystal loading, the open crate is set inside an
assembly container with very rigid walls so that it is
undeformed during insertion of the 700 kg of crystals .
The crate is first positioned to have one (b wall vertical. Crystals equipped with their capsules are installed
one by one in a 0 row and then the crate is rotated to
set the next row to vertical . Optical separations are laid
alternatively. Once the 225 crystals are installed,
preamplifiers and their cooling devices are connected,
and complete tests are performed to check circuit
continuity, insulation and tightness. The closing frame
is fastened to the side walls and mechanical pressure is
applied on the back of the crystal. To ensure that the
pressure is normal to the pressing faces, crystals are
pressed (b row after 0 row, with the tooling being
progressively rotated to have the force exerted vertically. This procedure will be performed iteratively to
compensate the elastic deformations induced by the
forces involved . This procedure was successfully used
for the assembly of the BGO crystals in the L3 endcap
modules.
Before lowering into the experiment, the calorimeter is pre-assembled in 16 rings on the surface. 36
modules are assembled horizontally in a ring on a
circular jig.
There are four identical jigs used two by two in a
pendular way for the assembly of two symmetrical
rings, and for the installation of the two previous
symmetrical ones . Each dig is made of a solid ring on
which a rotatable lifting gear is attached, and of 36
tilting plates which can be set to the required 71 angle
corresponding to each of the 8 ring types. Once the 36
crates of a type are tested and ready for installation,
they are set at the right 77 angle on the tilting plates
and bolted together . The ring assembly is pre-tensioned by pulling radially from the jig on the crate
outer rims with same forces as will be exerted from the
supporting tube, in order to give the ring its stability
and prevent deformation during transports .
Rings weigh about 25 t without jig weight . Once
assembled and tensioned, the ring assembly is rotated
to vertical position and lowered to the pit with the 63
tons crane.
4.9.2. Calorimeter beam calibration mechanics
Once rings are assembled, they are ready for beam
calibration . The jig is fit for mounting on a rotary table
providing the 0 orientation and the 77 inclination (see
fig. 53). The cable and service connections do not allow
a complete 2-rr rotation in 0 during a calibration run.
It is realistic to foresee calibration of the 15 77 layers of
crystals on 270 (b positions in one run, then a 180°
rotation of the assembly including the cables and ser-
Fig. 53 . Beam calibration setup.
L3P Collaboration / Precision experiment at LHC
57
vices attached, and a second run for the remaining half
ring . The table is levelled to let the beam line pass at
the point corresponding to the centre of interaction,
and enter the crystal by its front (small) face . As
described above, the two second extracted beam burst
is used to calibrate one crystal, after which the table is
rotated to the position of the next crystal during the
remaining 12 s of the SPS cycle. The active calibration
time (27 h per ring) is small compared to the total
setup time .
4.9.3. Support tube description
The support tube (ST) is 13 m in length, has an
inner radius of 3390 mm and an outer radius of 3490
mm which leaves a 100 mm radial clearance with
respect to the inner wall of the superconducting coil
vacuum tank .
The ST is made of two concentrical 20 mm thick
shells joined by welded end flanges and regularly spaced
welded distance pieces . Such a construction provides
good rigidity (high moment of inertia) with a minimal
amount of material . Tapped holes at the flange ends
are used to bolt the tube to the tank ends . The distance pieces are fixation points for the electromagnetic
calorimeter modules.
The ST carries the electromagnetic calorimeter load
with the help of tensioning spokes attached to the
distance pieces on the tube side, and to the calorimeter
module walls on the other side . The tension in the
spokes is such that module walls will always be in
tension . To do so, the tension applied to any spoke is
twice the part of the weight it must take . The sum of
all forces in the spokes takes the weight of the
calorimeter . The extra tension results in radially and
axially symmetrical (balanced) compression forces in
the tube .
A high resistance aluminium alloy is preferred to
austenitic (nonmagnetic) steel despite a factor 3 in
deformation, because of similar mechanical resistance
and a factor 1/3 in density. Composed of such material, the ST weighs 30 t.
A verification of deformation and stresses has been
performed by finite element calculation . The maximum
flexural deformation of 1 mm occurs in the middle
plane at the highest point . The maximum sectional
(radial) deformation of 0.3 mm occurs in the vertical
direction (ovalisation). Maximal stress (according to
the von Mises criteria) occurs at the tube ends and is of
the order of 6 MPa (dominantly shear) . This value
should be compared to 60 MPa yield stress guaranteed
by the supplier .
4.9.4. Calorimeter installation
Once the ST is in place, rings may be installed. In
order to balance the pre-stresses which give the
calorimeter its stability in spite of its very light struc-
Fig. 54 . Installation of rings in the support tube .
ture, the rings must be installed in symmetric pairs,
starting with the central ones. This implies that access
is required simultaneously from both sides of the detector (see fig. 54).
Two railways are installed on both sides of the
detector and are continued inside the support-tube
using the available fixation points . The ringjigs are
attached to carriages rolling on the railway. The carriages are counterweighted to cantilever the ring load .
For each pair of rings to be installed, the rails inside
the support-tube are withdrawn to leave the ring length
free .
During the insertion, the ring is slightly offcentered
upwards, to leave space for the rails. As the ring
arrives in position, it is lowered back to centre .
When the two rings are in position they are attached together and spokes are connected to the fixation points of the support-tube . Spokes are tensioned,
and expected deformations must be monitored until
the jigs can be detached from the rings. The spoke
positions on the ring circumference are scattered from
one ring to the next so that rings can be temporily
fastened both to the jigs and to the spokes.
Once a pair of rings is installed, the cables and
hoses are uncoiled and arranged along the support-tube
inner face . There is one bundle coming from each
crate, i.e . 36 from a ring . Each bundle has a width such
that it takes less than 1/9 of the crate outer width (in
R-phi) and such that the complete support-ring inner
perimeter is covered when the last (9th) pair of rings is
installed.
The computed support tube deformations are such
that there will be no need to readjust the spoke tensions of previously installed rings after a ring pair is
added. The spoke tensioning device can therefore be
made simple as no remote action will be required .
4.10. Endcap electromagnetic calorimeter
The endcap electromagnetic calorimeter, which covers the range 1.4 < I t71 < 3.0, consists of a high granularity silicon calorimeter . The energy resolution for this
L3P Collaboration / Precision experiment at LHC
58
1c =7mm
A At
Pb
A,,,Ai2,A13= 10mmPb+10mmCu
Dl
D13= 7mm Cu + Active plane (3 5mm)
S
1
, S2, S3 = Service planes (5mm)
---------------
-------
------------------------
Particles
SignalS
Power +controls
28 2 cm
Fig 55 Cross section of an active plane showing the preamplifier
calorimeter is 20%/ v~E ® 2%, which is appropriate
for the reaction H --~ 4e .
The design of the calorimeter is based on the experience gained by the SICAPO collaboration, and on
the R&D proposal P34 [32] which is developping lowcost silicon production and constructing a full scale
calorimeter module .
The calorimeter consists of 13 sampling layers . The
first 10 layers are 1.5 X0 thick and the last three are
3 .0 X, . The total calorimeter depth is 24 .0 X .
The mechanical structure is organized in 10 rings
supported at the back by a thick plate . All external
connections are routed along this plate as shown in fig.
57 .
Each ring is subdivided in A0 sectors, and covers
A77 = 0.15-0.2 .
Silicon detectors planes (active planes) are located
between copper and lead absorbers. They consist of
silicon diodes 400 win thick, with an active area of
2 x 2 cm', matching the Moli6re radius of the
calorimeter (pin = 1 .8 cm). The diodes are sandwiched
between two printed circuit boards . One board contains the preamplifiers lodged in holes in a sheet of
polyethylene envisaged to moderate neutrons. The second board carries the high voltage. The entire sampling layer is covered by a third board for mechanical
protection, and enclosed in a copper box. A spring
made of Cu-Be foil adapts the 3.0 mm thick active
plane to the 3.5 mm free space inside the box. The
detailed structure is shown in fig. 55 .
The rest of the absorber is Ph, 7 mm thick to the
first ten layers to complement the Cu absorber to 1 .5
X,, and 10 mm thick in the last three layers to complement the 7 mm Cu and the mechanical supporting
structure (10 mm/layer). The total thickness of the
calorimeter is thus 28 .2 cm, including the space for the
three service planes described below (see fig. 56).
The Si diodes are readout by preamplifiers (one for
each diode), whose outputs are summed to form minitowers and buffered in an analog pipeline memory
before to being send outside for further processing .
The charge sensitive preamplifiers are based on the
design developed by the SICAPO collaboration [33]
and adapted for the readout in the P34 R&D proposal. They are very fast (shaping time = 20 ns) and
can match the pipeline analog memories developed at
CERN [34]. Four preamp channels are contained in a
VLSI chip, and dissipate 45 mW/channel with a dy-
Pb
Note
Dimensions
Fig 56 . Cross section of the calorimeter.
tim
59
L3P Collaboration / Precision experiment at LHC
Table 15
Main parameters of the silicon calorimeter
Supporting plate
and interconnections
71 coverage
Total depth
Total geometrical thickness
Sampling layers
Cell size
Total Si surface area
Number of silicon diodes
Number of preamplifier channels
Number of summing channels
Number of pipeline analog memory
Cu weight
Pb weight
3
E
v 2
r1=3
Fig. 57 . Schematic view of the endcap silicon calorimeter
layout .
namic range of 10 5. Hence the power dissipated on the
active plane never exceeds 110 W/m2 .
The copper plate acts as a heat conducting sheet,
allowing heat exchange to the cooling circuit. Model
simulation guaranties that temperature gradients are
less than 2°C with an average value of 10°C . This
temperature is necessary to mantain a low leakage
current after irradiation .
The 13 active planes are longitudinally organized in
three segments of 4, 6, and 3 layers . At the end of each
longitudinal segment, the gap between the absorbers is
increased to allow a service plane carrying the HV
distribution to the planes of the segment, the summing
of the signals from the planes of the segment to form
projective minitowers, the pipelining of their outputs,
the controls and the protection devices. With this organization, the number of interconnections are minimized (bonding only), thus reducing the electronics
costs. To reduce the number of readout channels further 4 adjacent silicon diodes will be summed from
77=1 .4to 71=2 .0 .
70
n-
50
p 40
356 Detectors
80 V
November 1991
Taking into account the moderation effect of the
polyethylene located in the active samplers (fig . 56), in
10 years of operations the estimated neutron fluence
will be less than 5 X 10 13 n/cm Z down to 77 = 2 .5 . This
fluence increases gradually to 2 X 10 13 n/cmZ in one
year at 77 = 3.0 . Radiation damage may therefore require replacement of some of the silicon (up to 20 m2)
after 5 to 10 yr of full intensity and continous running.
The silicon could be supplied through JINR (Dubna)
by Russian Industry . The first lot of detectors produced last year in the framework of an extensive R&D
program, demonstrated sufficiently good characteristics (see fig. 58) for the leakage current distribution .
This R&D program includes extensive radiation
damage tests, study of the organization of a silicon
system to be used in calorimetey, and preparation of a
full scale subsystem prototype [32] .
The main parameters of the endcap silicon
calorimeter (for both endcaps) are summarized in table
15 . The occupancy, with this geometry, in one element
of 2 X 2 cm Z front surface is shown in table 16 .
4.11. Conclusions
The requirements of the detector and readout are
set by the desire to produce the best calorimeter for
photons and electrons.
Our fundamental approach to the design of the UP
electromagnetic calorimeter is conservative . Each parameter has been chosen to ensure that this calorimeTable 16
Occupancy as function of 77 for a 2x2 cm Z area
30
Z3 20
Z
10
1 .4-3 .0
24 Xl,
28 .2 cm
13
2x2cm2
705 m2
1 76 x 10 6
800 k
250 k
250 k
58 .6 t
61 .6 t
~111r~
200 400
600
800
1000
Leakage current, nA
Fig . 58 . Distribution of leakage current of fully depleted
detectors produced by ELMA
71
1 .4
1 .6
20
2.5
3.0
Occupancy
0.04%
0.15%
0.76%
3.00%
4.00%
60
L3P Collaboration / Precision experiment at LHC
ter is able to cope with the challenges of precision
physics at the highest luminosities, even after the
degradation of present-day optimisim to the reality of
final detector performance. The key features of our
design may be summarized as :
- Availability of crystals : A coordinated national
effort is under way in China, with the intent of providing us with the large quantities of crystals needed . Such
efforts have already been proven successful with the
12000 crystal L3 BGO calorimeter .
- Large radius: Background is inherently lower as
the detector becomes farther from the interaction point
- pileup noise and front-end trigger rate are reduced,
and 7r ° rejection is enhanced . A large-radius design
ensures the least sensitivity to errors in present assumptions of the backgrounds, and ensures the best
physics performance at any LHC luminosity .
- Simplicity: The performance of our calorimeter is
achieved solely through the use of crystals, without
reliance on complex photon pointing or -rr o detector
systems. Avoiding split crystals for additional detectors
cases the task of maintaining precision calibration .
- High speed, low noise . More than any other detector, electromagnetic calorimeter readout depends on
the extrapolation of current electronics technology to
future technology . We have emphasized a realistic design, in order to produce a detector that has a performance adequate for the physics goals.
- The design is based on proven L3 experience .
5. The hadron calorimeter
5.1 . Introduction
The design of the calorimeter system including electromagnetic and hadronic sections follows from the
general concept of the L3 upgrade to LHC: the detector is a precision instrument to search for new physics
through, mainly, leptonic and photon channels and, in
addition, quark channels (the measurement of jets).
The calorimeter, therefore, should be multifunctional .
- It provides identification of electrons, muons and
jets and serves as "isolation device" by measuring
energy in the vicinity of a muon track, thus defining
isolated muons.
- It provides capability to separate electrons from
hadrons by measuring shower shape and to separate
muons from other particles by total absorption of
hadronic components .
- It should detect muon bremsstrahlung and provide corresponding corrections to muon momentum
measurements .
The identification and energy measurement of
hadronic jets is performed by a combined calorimetric
Hadron Barrel
Fig. 59 . The L3 hadron calorimeter barrel perspective view .
system : a precision homogeneous electromagnetic device followed by a sampling calorimeter with copper
(brass) absorber and with proportional gas detectors.
Physics of gas calorimeters was extensively studied
[35], and the technique has been used in many experiments [14,36,37] . Its time response as well as radiation
hardness is adequate for the LHC rate environment .
5 .2. Mechanical structure and expected performance
In designing the calorimeter we have largely profited from the success of the L3 calorimetric system
which is similar, and can be considered as the first
"prototype" of the new system . The perspective view
of the L3 hadron calorimeter barrel is shown in fig. 59 .
For the UP detector we use basically the same type
of mechanics: structural rings fixed together and supported by rails inside a support tube as illustrated in
fig. 60 .
Fig. 60 . A ring of the L3 HC barrel, supported on rails inside
support tube .
61
L3P Collaboration / Precision experiment at LHC
Hadron Calorimeter modules
(barrel)
l
Proportional Chambers (15 layers)
7 k int
____________----- ÉMÇ--- -I
Cu-absorber
(14 layers)
Module
of
Hadron
Calorimeter
0 .8 X int
Super
conducting
Magnet
1 .5 X int
EM
Calorimeter
Crystals
Fig . 61 . The general layout of the hadron calorimeter . Sectional view along the beam .
Our experience with the L3 type of hadron
calorimeter mechanics has shown that the construction
was straight forward and thus economical, the assembly can be made fast and precise. We have accepted
therefore the same approach for the UP calorimeter .
The general layout of the hadron calorimeter is
shown in fig. 61 .
The calorimeter consists of a barrel (~ 771 < 1.4),
made of seven structural rings, and of two endcaps
(1 .4 < 1771 < 3.0). As shown in fig . 62 each ring as well
as each end cap "disk" consists of 16 modules.
The total number of modules is 144. The total
number of chambers is 2320, with 1680 in the barrel
and 640 in end caps .
A sectional view of a calorimeter module and of the
absorber/ detector structure is shown in fig. 63 . For
the modules of three central rings the absorber layers
have thickness of 7.5 cm, for the two peripheral rings
the layers are 6.1 cm and 4.8 cm thick correspondingly.
Endcap
Barrel Ring
Fig 62 . Ring assembly of the calorimeter . The barrel consists
of seven such rings . The endcap configuration is also shown .
Fig . 63 . Transverse sectional view of a barrel module assembly.
The total thickness of the hadron calorimeter of
= 9 A,,t at 90° and 11 A,ot in the forward direction
provides 98% containment of 0.5 TeV jets [38] and
reduces hadron punchthrough rates in the muon system to a level below the rate of prompt muons coming
from heavy quark decays . The thickness of the CeF3,
which serves as the front part of the calorimeter system, is 1 .5 A,rit . The thickness of the superconducting
magnet which follows the electromagnetic section is 0.8
A int, and, finally, the sampling hadron calorimeter section of 7 A ,t . As shown in fig. 63, this section is
subdivided into 14 layers of 75 mm brass and 15 layers
of gas proportional chambers inserted into 1 em gaps .
Our choice of proportional chambers was based on
the following considerations .
- The characteristic time of proportional chamber
response is 50-100 ns, and since the LHC bunch
frequency corresponds to 15 ns, one might think that
the proportional chambers are too slow. However for
the calorimetric energy measurement in the presence
of the accidental (pileup) background the background
contribution to the energy resolution is proportional to
square root of the background accumulation time . This
means that by using proportional chambers i.e . accumulating background from 3-4 bunch crossings we
increase its contribution to the energy resolution by up
to factor of 2 compared to the fastest available detection technique allowing for one bunch accumulation .
62
L3P Collaboration / Precision experiment at LHC
4
0
Fig. 64 . Energy resolution for two-bet events measured by L3
hadron calorimeter compared with GEANT Monte Carlo
simulation .
Our calculations show that the pileup contribution into
the energy resolution when proportional chambers are
used will correspond to 1-3%, and since our design
goal is the resolution of some 10%, the contribution
from the pileup is unimportant.
- Our past experience with the L3 calorimeter with
10000 proportional chambers gives us confidence that
the technique can be reliably used in a large scale
experiment with no access to the detectors for many
years. On the base of this experience we are also
convinced that the mass production of proportional
chambers can be organized even in parallel in several
labs in different parts of the world.
- The gas detectors are obviously the most economical solution among all known calorimetric detector
techniques .
The jet energy resolution of calorimeters of different configurations with the crystals and superconducting coil in front of the hadronic section has been
studied by Monte Carlo simulation with both the
GEANT-GHEISHA code [24] and a fast parametriza-
AI profile (1mm)
25 lira wires 12rn ;~1 `
25
50
100
250 5002000
Energy (GeV)
Fig. 65 . The expected jet energy resolution of the proposed
hadron calorimeter with crystals and SC magnet in front.
tion developed in the framework of the L3 hadron
calorimeter studies [39] . The code was optimized with
experimental L3 data on hadronic jets from Z° events.
Fig. 64 illustrates how the Monte Carlo describes the
measured jet energy resolution .
The jet energy resolutions for 1.5 A,nr of CeF3
followed by the superconducting coil and by the 7 A, o,
sampling hadronic section is shown in fig. 65 . The
slight increase of the resolution for jet energy above 1
TeV is due to rear leakage effects. Our Monte Carlo
studies [40] show that the jet energy measurement is
not practically affected by the 2 T field traversed by
jets before entering the calorimeter .
5.2.1 . Gas proportional chambers
The conceptual design of proportional chambers is
shown in fig. 66 .
The chambers are made of open-sided aluminum
extrusions with tungsten wires strung along the profile.
The open sides are used to read out the signals through
cathode pads [42]. The chambers are of equal length
Printed
board
(3mm)
Fig. 66 . Hadron calorimeter wire proportional chamber design . Cathode pads' configuration is also shown.
L3P Collaboration / Precision experiment at LHC
and vary in width as seen from fig. 63 . Since the length
of the wires exceeds two meters the wires are supported and positioned with several spacers distributed
evenly along the wire . The working gas considered is a
known fast mixture of 30% CF, and 70% C2H6. The
drift velosity of this mixture is 10 cm/Ws, and corresponding drift time 25-30 ns . The chambers are sealed
and the gas is supplied in parallel to all inner partitions
via two gas collectors incorporated into the chamber
end pieces (not shown) [41] .
In the modern gas calorimetry one has a choice
either to use streamer mode of the chamber operation
or a somewhat less economical proportional mode . The
following considerations lead us to choose proportional
mode .
- The streamer mode has always some admixture
of nonstreamer, either Geiger or, more often, proportional mode . For most of applications (like muon detectors, or counters) this is not very important, but for
the calorimetry this leads to nonlinearities of response
and unstable operation especially over large periods of
time .
- The particle showers developing in a calorimeter
often produce tracks strongly inclined w.r .t . a chamber
electrodes . For proportional mode such tracks give
proportionally larger signal, however for streamer mode
the multistreamers are produced . This again leads to
nonlinearities and unstable operation.
- Due to its high current density a streamer creates
a dead area of several mm of the anode wire, this area
is insensitive over large period of time : more than 100
ws [43] . This means the streamer mode is difficult to be
used for detection of showers with high particle density, like electro-magnetic showers or jets .
- Since streamer chambers operate at much higher
voltage, as compared to proportional chambers, they
are much more sensitive to mechanical or electrical
imperfections of the chambers . Our L3 experience
show, that one can built thousands of absolutely identical chambers in the sense that at the same voltage they
have the same response within few percent margin .
This is not the case for streamer chambers .
- As compared to streamer mode, proportional
chambers need additional preamplifier, since their current pulse is smaller. However with highly integrated
electronics the addition does not affect the overall cost
in a noticable way.
- We have practical experience of operating a system of 500000 proportional wires with less than a
fraction of a percent of wires lost (broken) over a three
year period .
5 .2.2. Segmentation and readout
The total number of readout pads in the hadron
calorimeter (both barrel and end cap) will be 1 .2 X 10 5,
63
with an average pad size of 24 X 24 cm 2 for barrel and
varying from 24 X 24 cm' to 8X8 cm2 for the end
caps . The pads are connected in depth to form towers
pointing to the interaction point. The number of towers in the barrel then is = 5000, and in both end caps
there are = 6000 towers . Assuming one longitudinal
segment this gives a total number of ADC channels
11000.
Since a pad capacitance is rather high (typically 1
nF) the low impedance low noise preamplifiers are
mounted directly on the pads as schematically shown in
fig. 66 . From the preamps the signals are brought to
the outside surface of the calorimeter where analog
sums of the corresponding pads are made to form
towers pointing to the interaction point. After passing
through bipolar shapers the tower signals are transmitted to ADC.
The speed of the electronics is chosen in such a way
that the pileup contribution (r .m .s .) to energy resolution will be 0.5-1 GeV in the end caps and 100-200
MeV in the barrel . Since the energy loss of a minimum
ionizing particle is of the order of 2 GeV, this pileup
allows to detect minimum ionizing particles in the
barrel and with somewhat less accuracy in the endcaps.
5 .2.3. Calibration
For calibration purposes we plan to use some radiactive gas to make relative interchannel calibration .
Such a calibration system provides adequate control
over the detector uniformity . Our experience with the
natural radioactivity of uranium used for calibration in
the L3 hadron calorimeter [35,44], suggests that the
calibration of the whole calorimeter system can be
accomplished within 30 min, to accuracies of 1 or 2% .
This approach has proved to maintain the uniformity
and the stability of the L3 calorimeter system at a level
better than one percent for several years of operation .
Absolute energy calibration should be performed
with test beam and corrected later with real data . In
order to measure the total energy one has to add up
the energy deposition in each subdetector properly .
The summing algorithm should provide a uniform response with optimum energy resolution over entire
solid angle covered by the calorimeters . For example,
this was done in L3 by expressing the total energy as a
linear combination of the signals recorded by each
subdetector, with each signal weighted by an appropriate energy-independent conversion coefficient . These
coefficients depend on the particular subdetector and
on the energy cluster position in it . The objective of
the energy calibration is to determine the value of the
coefficients which provide uniform energy response
and which minimize the error in the energy determination . The calibration was done by scanning the detector
with jets of known energy and with an event thrust
polar angle 0 within the calorimeter acceptance . Then
64
L3P Collaboration / Precision experiment at LHC
the energy of each event was calculated as a sum over
subdetectors :
Emeas - Y_E,(0)W ( 0 ) .
The value of weighting factors W,
were then ob-
tained by a XZ-minimization procedure. The result of
the calibration is a set of weighting factors W, . This
type of absolute energy calibration of a calorimeter
made
of
two
subdetectors :
electromagnetic crystal
calorimeter and sampling hadron calorimeter with gas
proportional chambers, was performed in L3, and the
result is illustrated in fig. 64 .
The barrel and end cap parameters are summarized
we
shall still need the following R&D before the
calorimeter construction can begin.
5.2.5. Chamber design optimization
The proportional chambers' design should be final-
ized by tests of prototype chambers and module assem-
bly. The issues to be addressed are the following.
- Gas mixture optimization . We need a gas mixture
with maximum drift velocity . It should also be stable
against sparking which implies a strong presence of a
quenching gas . Good quenchers are normally hydrocarbons which accelerate ageing and thus should be
avoided. Therefore a compromise solution should be
in table 17 .
found
5.2.4. Progress to R&D
and due to other chamber construction elements, like
- Ageing both due to gas hydrocarbon component
In spite of the fact that all principal questions of the
proposed design were solved during L3 construction,
glues, needs to be studied, and the materials responsible for the ageing should be avoided.
Table 17
Hadron calorimeter parameters
Barrel
Inner radius
Barrel thickness
Module length
Total barrel length
Absorber
Module "0", "+1", "-1"
Module "+2", "-2"
Module "+3", "-3"
Number of layers
Material thickness
Max. ring weight
Total barrel weight
End caps
Length
Number of layers
Absorber
Layer thickness
Material thickness
Weight of front module
Weight of back module
Total end cap weight
Detectors
Detector gap
Readout
Gas
Drift velocity
Gas gap
Drift time
Granularity 0-7 X 0O
Longitudinal segmentation
Number of ADC channels
Barrel
End cap
4 .100 m
1 .210 m
2.140 m
15 .0 m
Brass
Absorber layer thickness
7.5 cm
6 1 cm
4.8 cm
14
7A
605 t
3569t
1 .70 m
20
Brass
7.5 cm
10A
10 .3 t
18 .5 t
2x459.4 t
Gas prop chambers
1 cm
Segmented cathode (pads)
CF4(30%) + C z H 6(70%)
10 cm/ws
5 mm
25-30 ns
0 .06 X0.06
1
5000
2 x 3000
Module weight
37.8 t
30 .6 t
24.1 t
65
L3P Collaboration / Precision experiment at LHC
- Choice of profile dimensions i.e . gas cell . Minimization of the overall chamber thickness.
- Electronics optimization .
- Mass production technique.
5.2 .6. Radiation damage studies
Since the expected radiation levels in the hadron
calorimeter, shielded by electromagnetic section, are
well below 10 Mrad/yr (with the possible exception of
some regions in the endcaps close to the beam pipe
which need refined design optimization) we conclude
that the radiation hardness of the basic components of
the proposed system should not be of serious concern.
The aim of these studies is to select various components of the calorimeter system : gas, glues, sealants,
etc. as well as to test radiation stability of the complete
chamber assembly which employs the combination of
various construction elements and techniques .
In order to optimize the final design the tests will
continue after the selection of the technique has been
made .
6. The muon system
6.1 . Introduction
The purpose of the UP muon system is to detect
and identify muons from the collision, and to trigger on
muons. It also provides a redundant measurement of
the muon momentum and therefore improves the background rejection by momentum matching with the in-
ner tracker. To achieve these goals, the system should
be able to measure the muon track with adequate
accuracy, perform the pattern recognition properly,
and have fast response in the high luminosity LHC
environment. It is therefore designed with a combination of different technologies, including 1) arrays of
drift tubes and drift chambers measuring the bending
of the muon track in the magnetic field with a resolution matched to the multiple scattering ; 2) multiwire
proportional chambers for pattern recognition, and 3)
resistive plate counters (RPC) for triggering and bunch
identification . These technologies have all been well
developed in the past and are highly reliable and cost
effective.
6.2. Detector layout
Fig. 67 schematically summarizes the detector layout for the UP muon system . The system is separated
into the barrel and the endcap regions. Since the muon
system is further away from the interaction point, the
barrel and the endcap regions are divided at 71 = 1
instead of 71 = 1.4 as in the calorimeters . The barrel
region covers 0 < 1771 < 1, and the endcap region covers 1 < I rl I < 3. Each detector station is designed based
on the particle rate and momentum measurement requirements . Whenever adequate, the existing L3 chambers are used to reduce the construction efforts.
The particle rates in the detector stations are estimated by a Monte Carlo simulation based on the
complete detector geometry . A_total of 100000 minimum bias events and 50000 bb events with 16 TeV
C.M . energy generated by PYTHIA [45] are simulated
ON
WN
"EMEEN
'EMEMENN
E11111
MEN
Am
Z n&
ISO
NEW
'1151
1
1;
RPC-0
L3 BARREL MO CHAMBERS
PATTERN RECOGNITION CHAMBER
RPC-0
w
m
ôa
=
Û~p
SOLENOID
PROP .
CRAM.
F
PROP
CHAH
n-
A
E
0
UP Muon Detectors
Fig. 67 . Layout of the muon detectors.
m
66
L3P Collaboration / Precision experiment at LHC
based on the GEANT [46] program. Particles penetrating to the muon stations are registered and the rates
are calculated assuming a 1 x 10 34 cm -2 s - I luminosity. The muon momentum resolution is also estimated
by full simulation of the muon tracks through the
detector, including the effects of multiple scattering
and energy loss .
6.2 .1 . The barrel muon detectors
The barrel region measures the muon in two stations "A" and "B". Station B is located between the
support tube and the iron return yoke, which holds a
1 .8 T magnetic field. Station A is located outside of the
yoke . The amount of the material before the stations
corresponds to 9 A and 14 A, respectively .
The particle rates in the barrel region are shown in
fig. 68 . The rate is approximately constant along the
z-coordinate, at 0.2 Hz/cm2 in station A and 1 .2
Hz/cm2 in station B, and the integrated rates are 1 .8
MHz and 6 MHz, respectively . There is no need for a
high rate detector in these areas. RPCs, pattern recognition chambers, and the existing L3 barrel chambers
(MO) are used in station B. The L3 MO chambers [14]
are multiwire drift chambers with 10 cm side drift cells
and 5.6 m long wires. The chambers are at the same
locations as in L3, so there is no fiducial volume loss .
Eight wires measure the (b coordinate, which is the
bending direction, with a single wire resolution of 250
wm . Four wires measure the z coordinate with a single
wire resolution of 400 p,m. The region 6 < I z I < 8 m is
covered by 8 layers of proportional chambers with 2
mm pitch. Six layers measure the (h coordinate and 2
layers measure the z coordinate . The whole station B
is covered by a layer of pattern recognition chamber,
which consists of 3 layers of proportional chambers .
The orientations of the wires in the three layers are
rotated by 120° relative to each other . On top and
bottom of the station, RPCs with 1 cm wide and 3 m
Barrel Region
L= 10 34 cm -2 s -1
0e
0
400
Z
(cm)
600
Fig. 68 . Particle rate m the barrel region .
100
300
R (cm)
500
700
Fig. 69 Particle rate in the endcap region
long strips in both (b and z directions are used for
trigger purpose.
The barrel station A is instrumented with 8 layers
of drift tubes for coordinate measurements, and two
layers of RPCs . The drift tubes are of 10 cm diameter,
and have a maximal length of 6 m. Six layers of the
drift tubes measure the (b coordinate, and two layers
measure the z coordinates.
6.2 .2. The endcap muon detectors
The muons are measured three times in the endcap
region . Station "B" is located between the endcap
hadron calorimeter and the magnet return iron; station
"A" is located between the magnet return iron and the
forward toroid magnet; and station "F" is located after
the toroid magnet. The amount of material before each
station is 11 A, 16 A, and 20 A, respectively .
Fig. 69 shows the rates as a function of the radial
distance R from the beam line in the endcap regions.
The particle rate increases rapidly as we approach the
beam line . At q = 3, the rate is - 10 kHz/cm 2 at
station B.
Because of the high particle rates, the endcap stations B and A are further divided at 171 I = 2 into two
regions. In the region 1 < 177 ~ < 2, where the particle
flux is less than 80 Hz /cm2, each station has 8 layers of
6 cm diameter drift tubes, RPCs, and pattern recognition chambers . In station B, the momentum measurement is based on the bending in the inner 2 T solenoid
field. For the 0 measurement 6 layers of (h-measuring
tubes and two layers of R-measuring tubes are installed. Two layers of RPC with (h measuring strips are
used for triggering . Station A measures the bending in
both the solenoid field and the forward toroid field .
Four layers of (h-tubes and four layers of R-tubes, and
RPCs with both R and (b strips are therefore used . Fig.
70 shows a cross section of the endcap station A.
L3P Collaboration / Precision experiment at LHC
Fig. 70 . The cross section of the endcap module A.
In the region where 2 < 1771 < 3, because the particle rate is too high for drift tubes or RPCs to work, two
layers of proportional chambers with one layer of the
pattern recognition chambers are used .
The station F is also divided into two regions. The
region 1.4 < 1771 < 2.35 is covered by the existing L3
forward-backward chambers and RPCs . The small angle region (2.35 < 1711 < 3) is covered by proportional
chambers . Station F measures the bending in the toroid
field only . The existing L3 forward-backward chambers are therefore arranged such that two layers of
drift chambers, with four wires in each layer, measure
the R coordinate . One layer of the same drift chambers measures the 0 coordinate .
Table 18 summarizes the main parameters of the
muon system .
6.2.3. The magnetic field
As is shown in fig. 67, the 2 T solenoid field in the
central region returns through the forward return iron
Table 18
Basic parameters of the muon system
Station
R [cm]
z [cm]
Barrel B
Barrel A
Endcap B
1771 < 2
Endcap A
1711 < 2
Endcap F
1711 < 2.3
Endcap B
1771 > 2
Endcap A
1771 > 2
Endcap F
1771 > 2.3
570
750
-800-800
-950-950
Absorption
length
9A
14A
67
and the yoke . The field inside the iron yoke is 1 .8 T,
not fully saturated. Except for the endcap B region, the
muon stations operates at basicly 0 magnetic field.
Stray fields below 0.1 T exists in the corner between
the magnet door and the yoke . This weak field should
not cause operational difficulties for the drift tubes
and drift chambers, because the magnetic deflection
angle is small, as will be shown below.
The magnetic field in endcap station B is perpendicular to the wires and does not affect the resolution
by magnetic deflection . The field, however, reduces the
drift velocity and results in a longer drift time. For a 2
T field, the expected maximal drift time for the 6 cm
drift tube is 1.1 ws compared to the 0.6 Vs without
field. A good magnetic field map and drift velocity
calibration is critical to achieve the required resolution .
6.3. Performance
6.3.1 . Momentum measurement
The momentum resolution as a function of the
polar angle measured by the muon system A, B, and F
is shown in fig. 71 .
Two independent measurements are performed in
the barrel region . As shown in fig. 72, a muon track
bends in the inner 2 T field, by a track angle 00, which
is related to the muon transverse momentum . For a
100 GeV muon at 0 = 90°, the bending is 0(h = 13
mrad . The resolution of this measurement is limited by
multiple scattering, which is 2.2 mrad for a 100 GeV
muon, corresponding to a 17% momentum resolution .
When the muon passes through the return yoke, the
return field bends the track in the opposite direction
and results in a difference 8¢ between the angles
measured in station A and station B. The resolution is
also limited by multiple scattering to be 23% for a 100
GeV muon .
Max. rate
[Hz/cm 2]
1.0
0.2
220-530
780
11A
80
260-700
940
16A
20
210-600
1120
20A
30
80-220
780
11A
10k
95-260
940
16A
Sk
110-210
1120
20A
lk
Stations A,B,F
032
024
âlCL
PT = 10 GeV
018
008
Solenoid
+
1
1
.4-T--ill
1
1
Toroid
(A+B+F)
-
PT = 100 GeV
Solenoid only
(A+B)
Fig. 71 . Mome tum resolution as a function of B, measured by
the muon stations only .
68
L3P Collaboration / Precision experiment at LHC
The occupancy, O, of each detector element is
defined as :
O =fat,
Fig. 72 . The trajectory of a muon traversing the detectors.
Similar considerations apply to the endcap region,
where angular measurements in stations B and A provide two independent momentum measurements . To
improve the resolution for small angle tracks, the forward iron toroid is used as a third independent measurement . Track bending between stations A and F
yields a momentum resolution of 23% for a 100 GeV
muon .
Since the multiple scattering and the bending in the
magnetic field have the same momentum dependence,
the resolution is approximately independent of the
muon momentum . However, for high momentum
muons, the multiple scattering is small and the detector resolution become more important. We require
that the detectors should not affect significantly the
resolution of a 100 GeV muon, which has a 2.2 mrad
multiple scattering passing through the iron yoke . The
detector resolution requirement is therefore set to 1
mrad for the barrel muon stations . In the endcap
region, a muon with 100 GeV transverse momentum
has an averaged multiple scattering of - 1 mrad, the
detector resolution is therefore required to be better
than 0.5 mrad .
The momentum resolution is also important for the
trigger . RPCs with 1 em strips are used for the first
level muon trigger. The strip width results in an angular resolution of 6 mrad, allowing a trigger threshold
for PT < 20 GeV, see section 6.7 .
6.3.2 . Pattern recognition
The pattern recognition can be affected by mainly
three effects : 1) The high particle rate can cause too
high an occupancy in the detector elements, resulting
in an incorrect drift distance measurement. 2) A high
particle rate also causes too many random matching
between the R-O and the z measurements . 3) Showers
caused by high energy muon bremsstrahlung can leak
into the muon system and confuse the reconstruction .
These effects are discussed in the following.
where f is the rate per unit area, a is the area of a
detector element, and t is the maximal response time
for the detector. We used t = 0.6 p s for the 6 cm drift
tubes, and t = 1 g,s for the 10 cm tubes and the drift
chambers . t = 50 ns and t = 30 ns are assummed for
the proportional chambers and the RPCs, respectively.
The particle rates, f, for the muon detectors are listed
in table 18 . We find that the highest occupancy occurs
in the drift tubes in the endcap B station near 77 = 2,
where f = 80 Hz/cm2, corresponding to 9% occupancy
with 1 g,s maximal drift time . Since the double track
resolution of the drift tubes is 1 cm, - 3% of the hits
will be lost due to overlapped tracks . In the barrel
region, the occupancy is down to 0.2% and causes
negligeable reconstruction problem.
To achieve mechanical stability and easy installation, the muon detectors will be assembled in modules.
The largest module covers an area of 6 x 6 m2. For the
barrel region, based on a particle rate of 1.2 Hz/cm2,
we expect a 43% chance that one extra track occurs in
the same module within 1 [Ls of the primary track. For
the drift tube system with only two wire orientations,
this track causes two extra ambiguous tracks (see fig.
73). In the endcap region, where the particle rate is
higher, these ambiguities occur with higher probability.
it is therefore necessary to have pattern recognition
chambers to resolve the ambiguities . Three layers of
proportional chambers with 120° wire orientation rotated between the layers (star chambers) reduce the
probability of ambiguity by a factor of - 1000 for the
barrel region, and a factor of - 100 for the endcap
region . The tracks can then be clearly identified .
Shower leakage caused by hard muon bremsstrahlung in the hadron calorimeter can reach the
muon chambers and spoil the muon measurement.
Two studies were done to understand the effects . First,
ro)
u
Fig. 73 . (a) The true hit (0) and the ambiguous hits (o) m a
detector with only two wire orientations . (b) The ambiguity is
resolved by an extra wire plane.
L3P Collaboration / Precision experiment at LHC
since the double track resolution of drift chambers and
tubes, and the proportional chambers is - 1 cm, we
looked for showered tracks which arrive within 1 cm of
the primary muon track. Single muons were simulated.
Fig. 74 shows the probability of having an extra track
within 1 cm of the muon track as a function of the
muon momentum . Only 2% of 100 GeV muon tracks
are confused . This simulation has been confirmed by
the L3 running experience, where bremsstrahlung measured in dimuon events from Zo decays was compared
to the simulated results. Good agreement was obtained .
We also studied this effect by visually scanning the
simulated events . For 250 GeV muons, we found that
- 20% of the muon tracks produce additional particles
from the hadron calorimeter. However, only 4% of the
events scanned could actually cause track confusion.
We conclude that, for 100 GeV muons, less than
2% will be incorrectly reconstructed due to ambiguities
or overlapped tracks . The probability of mis-reconstruction increases for higher ration energies, due to
muon bremsstrahlung . It, however, is less than 5% for
muon energies up to 250 GeV.
69
6
4
2
Eo
-2
-4
3
E2
v
w
6.4 . Detail detector description
6.4 .1 . Drift tubes
The drift tubes are based on a design proposed by
the SDC collaboration [47]. Fig. 75a shows schematically the end view of a tube . The tube is made of a 2
mm thick aluminum wall, with two strip electrodes for
field shaping. To simplify the high voltage requirement,
the electrodes are at the same voltage as the signal
wire . In a 10 cm diameter tube with 50 wm wire
diameter, and 3.2 cm wide field shaping strip, an
applied voltage of 6.5 kV results in the drift field
configuration as shown in fig. 75a. The field is very
close to a uniform drift field . Fig. 75b shows the
5
4
0
50
100
150
Pp (GeV)
200
250
Fig. 74 . The probability to find one extra track within 1 cm of
the primary muon .
Fig. 75 . (a) The electric field lines m the tube . (b) The field
strength as a function of the distance to the wire .
variation of field strength along the drift path, we find
an average 1000 V/em drift field with a maximal ±80
V/cm variation in the drift region .
The advantages of this tubes are:
1) Simple construction : only two different tube diameters are used, suitable for mass production .
2) Reliable : the field shaping electrodes are located
far away from the signal wire such that the drift field is
insensitive to small geometrical uncertainties. Wire
support is not necessary.
3) Accurate : the tubes can be rotated such that the
drift field is always perpendicular to the particle path
originated from the interaction point, thus improving
accuracy .
4) Easy maintenance: each tube forms an electrically independent unit, making the maintenance easy
and no signal cross talk.
We have done studies and chosen nonflammable
gas Ar : C0 2 : isobutane 86 : 10 : 4 mixture as the drift
gas. Fig. 76 shows the drift velocity and the magnetic
deflection angle versus the drift field for this gas [15] .
The drift velocity for B = 0 is fully saturated at the
designed drift field, so the operation is not sensitive to
the field variation, as well as the pressure and temperature. The magnetic deflection angle is small, therefore
70
L.3P Collaboration / Precision experiment at LHC
8
6
0
I I~~i
ÎIfÎÜ
~f !I lfl~~ ll
TI -11l'
_'111llll 1 ci i
^y 20
2
v
zi 10
B=0 .51 T
PRECISION
BRIDGE
"-
ENDFRAME
0
Fig. 76
1
E [kV/cm]
2
Measured drift velocity and Lorentz angle of
Ar - CO Z : isobutane (86 :10 :4) mixture.
the measurements is not sensitive to the small stray
magnetic field. Overall, we expect a 250 win single wire
resolution .
To improve the resolution and to have a fast momentum trigger based on the more accurate drift tube
measurement, the tubes are arranged according to the
following criteria (see fig. 70):
1) The drift field is
perpendicular to the path of an infinite momentum
particle from the interaction point, so the slope corrections are minimized and thus achieve the best resolution . 2) The line joining two wires of the 0 layers
points back to the interaction point. The drift distance
from these two layers will then be the same for high
momentum muons (d, = d, in fig . 70). The time differ-
ence of signals from these layers thus measures the
track angle compared to an infinite momentum track,
with a resolution of - 2 mrad . This provides a simple
way to implement an accurate momentum trigger within
2
ws
Fig. 77 Exploded view of the L3 MO chambers.
3
of beam crossing .
6.4.2 . L3 chambers
The L3 detector has been designed such that part of
the muon system can be used in the high rate hadron
the (b coordinate . The signal wire plane at the center
of a cell contains 16 signal wires interleaved with field
shaping wires, with a wire spacing of 9 mm . The cells
are 5.6 m in length, and the maximal drift distance is 5
cm (half the cell width) . The wire planes in the MO
chambers are precisely positioned by optically flat glass
edges. The glass pieces were glued to carbon-fiber bars
to form the wire supporting bridges with very small
thermal expansion (< 3 ppm/°C). The positions of the
bridges are related directly to the external alignment
system and are independent of the aluminum chamber
box, thus eliminating the uncertainties due to thermal
expansions . Position resolution measured with LEP
dimuon events is shown in fig. 78, a 220 win single wire
resolution has been achieved .
For UP operation, only 8 of the 16 signal wires are
read out, since the accuracy requirement is not stringent . The angular resolution is 0.8 mrad, corresponding to the uncertainty caused by the multiple scattering
of a 250 GeV muon .
The chamber is covered on the top and the bottom
with I-beam chambers which measure the z-coordim
c
w
7500
collider as well as in the LEP experiment . The outer
5000
outer forward-backward system (FM, FO) of L3 will
continue to be used in UP .
2500
layer of the barrel muon chambers (MO) [14] and the
Fri
65W=220 pin
As shown in fig. 77, the L3 MO chambers are built
inside an aluminium box which contains the gas vol-
ume. The volume is divided by cathode wire planes
into drift cells that are electrically independent to each
other. The cells are along the z direction and measure
0
2
-1
0
1
(m m)
2
Fig. 78 . Single wire resolution of the L3 barrel chambers
L3P Collaboration / Precision experiment at LHC
71
turn resolution, we use 1 cm wide and 3 m long strips .
The signal picked up by the strips has a fast risetime
(less than 1 ns) with a 1 ns risetime jitter. The signals
are on average - 300 mV on a 50 SZ termination, thus
no preamplifier is necessary . Because of the speed and
the accuracy of its timing the RPC is the ideal trigger
and bunch identification device for the UP detector .
The RPC's function is limited in rate to less than
100 Hz/cm2. Below 100 Hz/cm2, the efficiency of the
RPC is - 97%, mainly due to the spacers inserted in
the gas gap. Double layer RPCs are therefore used to
get 100% efficiency.
MAGNET DOOR
AMPLIFIERS
Fig. 79 . The L3 forward-backward chambers .
nate . These chambers are made of drift cells with 10
cm drift space . Each cell has one signal wire and is
separated from the neighboring cells by I-beams that
functions as the cathode plane as well as the mechanical support . Four layers of the I-beam chamber are
used to measure the z-coordinates with a single wire
resolution of 400 ~tm.
The L3 forward-backward (now called station F)
chambers are made of drift cells of four signal wires, as
shown in fig. 79 . The cells are separated by I-beam
cathodes, which also support the chamber structure .
Guard strips on top and bottom of the cell shape the
drift field such that the field is uniform in the drift
region . Three layers of the same drift cells are built
into one module . Two of the layers measure the radial
coordinate, R, which is the relevant coordinate for
momentum measurement. The cells are shifted by half
the cell size, such that the ambiguity caused by the
left-right symmetry is eliminated . The chamber is also
self calibrated since the sum of the drift time measured
by the two layers should be a constant. The single wire
resolution is measured to be 250 V,m.
Two chamber modules are used in the endcap F
station, with a distance of 1 m between the modules.
The angular resolution is 0.1 mrad, sufficient for the
required accuracy .
The same gas used in the drift tubes is being used in
the L3 drift chambers . The chambers operate at the
saturated velocity of 5 cm/ws, with a maximal drift
time of - l ws .
6.4.3 . RPCs
The RPC [50,51] has been developed for high timing accuracy and fast response . For adequate memen-
6.4.4 . Proportional chambers
The proportional chambers used in Barrel station B
and in the small angle endcap region are of conventional design [52] . The wire pitch is 2 mm, giving a
resolution of 600 wm . In the barrel station B, wire
length of 2 m are used . In the endcap region, the wire
length varies according to the geometrical requirements, and is not longer than 2 m.
6.4.5 . Pattern recognition chambers
The proportional chambers used for pattern recognition are of the same type as that used in the tracker
[53] . However, since the particle rate is considerably
lower than in the central region, short wires and fine
wire pitch are not necessary. Mechanically the proportional chambers are limited to a maximal size of about
2 m2 area. However, to reduce the number of readout
channels, it is desired to link wires in neighboring
chambers to form effectively longer wire lengths. One
can also group neighboring wires into a single readout
circuit and make effectively wider pitches.
In the barrel region, we link wires to 6 m length,
matching the size of one detector module . Furthermore, 5 neighboring wires are grouped into one readout channel, corresponding to a detection element of 1
cm wide and 6 m long . The occupancy for one element
is < 0.01% and causes no ambiguity problem.
In the endcap region, due to the variation in geometrical shape and the particle rates, the wire linking
and grouping cannot be done in a uniform way. Depending on the location, the effective element area
changes from 600 cm 2 at 77 - 1 down to - 20 cm 2 near
T7 -3, keeping the occupancy less than I% . The remaining ambiguity is - 0.1% with this occupancy [541,
and cause negligible reconstruction problems .
6.5. Structure and alignment for the Barrel system
Detectors in the same detection station are modulized for easy construction and installation . As shown
in fig. 80, in the barrel region each detection station is
first divided in 0 into eight identical units, each covering one side of the octagon shape of the UP detector .
L3P Collaboration / Precision experiment at LHC
72
STATION-A
Iron Yoke
Muon
station
Electronic
\bubble level
Gravitational
'down' direction
8
dO
Chamber
central
line
Radial direction
-IP
dx
60 =dO+
!RON YOKE
dx
Fig 80. The arrangement of the detector modules for the
barrel muon system.
Fig. 81 . Principle of alignment to the interaction point.
A unit is further divided in Z into four modules of
sizes determined by the detector requirements and
geometrical limits . Each module contains its own electronics and alignment tools, such that they are independent of each other and can be constructed and
assembled independently. Alignment between modules
will be done optically with a modified version of the
straightness monitor developed for the L3 detector
[18] .
Alignment between stations A and B is done by the
modified L3 straightness monitor, which aligns both
the relative angle and position. As shown in fig. 82, the
system includes multiple IR light source, a lens, and a
four quadrant light sensitive diode. The light emitted
from the IR source is focussed by the lens and measured by the diodes . The movements of the IR source
by dx results in a movement of the image by (d x - s)/l .
By measuring the charge sharing among the 4 diodes,
one can determine the position of the IR source relative to the line defined by the 4 quad . diodes and the
lens. The IR sources can thus be individually aligned.
This system has been used in the L3 experiment, where
s = 1, and yields a coordinate resolution of 6 p,m [49].
For the UP muon system, s/l= 1/3 and we expect a
resolution of 15 win . The angle between the two stations can therefore be aligned to < 0.2 mrad, enough
for the resolution requirement .
6.5 .1 . Alignment methods
Much effort has been spent in L3 to develop accurate and cost-effective alignment tools. The alignment
of the UP muon system is based on these experiences .
Since the momentum measurements rely on the
angular measurements of the particle tracks, the alignment system has to determine the angle of the detector
related to the interaction point, as well as the relative
angle between the two stations .
To align the detectors with the interaction point, an
electronic bubble level system is used . Fig. 81 depicts
the principle of the alignment. The angle do relative
to the gravitational "down" direction is measured by
the bubble level connected to the detector central line
by specially mashined surfaces, to better than 0.1 mrad .
This can be achieved by a relatively simple electronic
bubble level. The impact parameter of the "down"
direction, dx, is determined by survey during the installation of the detectors. The survey measures dx
with an accuracy of 1 mm, and we expect thermal
movement to be less than 1 mm during chamber operation . The angular uncertainty due to dx is therefore, 2
mm/5 m, or 0.4 mrad, corresponding to the multiple
scattering of a 500 GeV muon . The accuracy of this
system is therefore sufficient for the required resolution .
IR sources
Station A
iron L
yoke
lens
Station B
=
dx'
dx
L
dx
4 quadrant
diode
Fig. 82 . The straightness monitor.
S
73
L3P Collaboration / Precision experiment at LHC
li~.
MONITOR
//
IllW'OÀ5
017ÀRWZÀ
MAGNET RETURN YOKE
(Bz =-2T.ESLA )
51 0°
380'
IP
Fig. 83 . The alignment between stations A and B.
Fig. 83 shows the alignment scheme for the barrel
system . The straightness monitor on the end of each
modules aligns the corresponding modules in stations
A and B. Special alignment holes are drilled in the iron
yoke for light path . The same straightness monitor is
used for internal alignment within a module to reduce
the errors caused by gravitational sagging. UV lasers
mounted on station B traverse the chambers through
quartz windows are used to final check the alignment
result .
6 .5.2. Barrel station A
As is shown in fig. 80 the barrel station A is divided
into modules of 6 m wide, 4 or 6 m long, and 1 m thick.
The module consists of drift tubes and RPCs . The
detectors are put inside an aluminum box with strong
end plates . As shown in fig . 84, 8 layers of tubes are
used, with 6 measuring 0 and 2 measuring z. Neighboring layers are separated by the spacer plate to
transmit the sheer force. The tube positions are defined by the precision mashined end plates . The whole
WN/ -/"-""-%"
~ """
"W"
Lm%"
"I"w"I0""5"VWWIMEN-VM/"
,
"F"fE"" LNL"LaV""F""L"" I
~%\`% "`%"%"`%,`%"`%W% "
N1 1IFÀ" __90IlF
ZIl&""
RPC
DRIFT TUBES
SPACER SHEET
PROPORTIONAL
CHAMBER
RPC
Fig . 84. The arrangement of a muon station A module .
assembly is very stiff because the tubes are themself
used as structure elements .
The modules are attached to the iron yoke by
supporting rails. The rails are pre-mounted on the iron
and surveyed before installation of the chambers, and
the modules slide in during installation . The bubble
levels on the end plates monitors the tilt angle of the
module . The position of the outer modules Al and A4
relative to the beam line can be surveyed directly, since
the end of the modules are visible from outside . The
positions of A2 and A3 are related to Al and A4 by
straightness monitors .
6.5 .3. Barrel station B
Since the L3 MO chambers are built as selfsustained units, they form a natural base for a module .
Each module covers half of an octant . The RPCs and
the pattern recognition chambers will be attached to
the surface of the top and bottom Z-chambers. The
end frame of the MO chambers will be modified slightly
so that the wire locations can be related to the straightness monitor and the bubble levels. The whole system
is supported on four corners where flexures attach the
module to the iron yoke .
Since the wire locations in the chambers are fixed
by the precision glass bridge, the thermal expansions
are small. Long term calibration based on high energy
muons is sufficient for the alignment purpose.
The two ends of the station B are covered by the
proportional chambers . The chambers are put in modules of 2 m long, 4.4 m wide, and 50 cm thick. The
modules are made out of a carbon fiber frame which
consists of two end-pieces, on which the chambers are
supported and aligned internally, and four rails that
connect the two plates to form a rigid structure. The
global alignment tools are attached to the end plates
and are related to the chambers through the precision
machining of the plates .
L.3P Collaboration / Precision experiment at LHC
74
MUON
SOLENOID BEND :
TOROIDAL BEND :
X IN A, B
Y IN A, F
Fig. 85 The alignment for the endcap muon stations
6.6. Structure and alignment for the endcap systems
6.6.1 . Alignment for the endcap system
The principle of alignment for the endcap stations
is depicted in fig. 85 . Stations B and A measure the
bending in the solenoid field and in the return field.
The relevant alignment parameter is the rotation along
the radial axis, dO, and the alignment of the module
central line with the interaction point. The alignment
can be achieved by two straightness monitor systems
each measuring both the x and z coordinates of the
modules relative to the beam line . Receivers on the
beam pipe define the line with 0 = 0 and the chambers
are aligned to this line through the straightness monitors .
Stations A and F together measures the bending of
muon tracks in the toroid field. The relevant parameter is the differences of the inclination angle dB . This
angle can be measured by the straightness monitors
similar to the system used for the barrel region . UV
lasers are also used to check the alignment.
6.6.2 . Endcap stations B and A
The endcap stations B and A consist of the drift
tubes, pattern recognition chambers, RPCs, and proportional chambers in the small angle region . As shown
in fig. 86, the stations are divided into "IN" and
"OUT" sections . Each section is divided further in
into eight modules. To improve the coverage, the IN
modules are rotate in 0 by 22 .5° relative to the OUT
modules. The modules are made out of aluminum
boxes with stiff end plates, similar in construction to
the modules in barrel station A. The modules are
attached to a honey comb panel in between the IN and
OUT sections . This structure is very stiff and provides
a simple mechanism for installation .
As shown in fig. 70, the OUT modules contain four
layers of drift tubes, one layer of pattern recognition
chambers, and one layer of RPC. The IN modules
contain four layers of drift tubes and one layer of RPC.
The neighboring layers are separated by spacer sheets
to transmit the sheer. As mentioned previously, the
orientation of the drift tubes is such that the drift field
is always perpendicular to the trajectory of an infinite
momentum ration from the interaction point. The wires
with the same orientation in a module are also lined up
Fig. 86 . The arrangement of the endcap modules.
75
L3P Collaboration / Precision experiment at LHC
with infinite momentum muon tracks . Such arrangements can be achieved by precision machined endplates which fixe the tube orientation and location .
Since the IN and the OUT modules are shifted by
22 .5° relative to each other, the dead space caused by
aligned neighboring tubes is greatly reduced.
(a)
6.6.3. Endcap station F
The existing L3 Forward-Backward chambers are
constructed in trapzoidal modules with 45° opening
angle. These modules are rigid because the 1-beams
provide a strong structure support. The modules are
supported on the toroid magnet by an aluminum structure .
!~ 1 strip displacemen
(b) < 2 strip displacement
(c) <- 3 strip displaceme nL~
I
Fig. 88 . The barrel muon trigger efficiency.
6 .7. Trigger and readout
The muon trigger is based on the momentum measured by the RPC and the drift tubes . A level 0 muon
trigger reduces the event rate to less than - 50 kHz,
where higher level triggers based on the digitized information can be performed. The trigger has to arrive at a
decision within 2 [Ls, during which signals from an
event are hold in digital buffers. The muon trigger
signal is also used for bunch identification, and therefore is based on the RPC signals which have fast
response time with small jitter. In the endcap regions,
the drift tube signals are used to improve the momentum resolution and thus better trigger decision .
A Level 1 trigger based on matched R-0 and z
measurements further improves the trigger selection.
The pattern recognition chamber is needed to perform
the matching reliably . Fast track matching algorithms
using the pattern recognition chamber signals have to
be developped to find the true muon tracks, and
matched with the drift tube and RPC measurements .
Global triggers, e.g . dimuon trigger, can then be performed to reduce the rates down to the 1 kHz level.
6 .7.1 . The barrel trigger
As already stated each barrel station is equipped
with two layers of RPCs readout by 1 cm wide pick-up
strips parallel to the beam direction. Because of the
bending in the solenoid field a muon enters station B
with an angle 80 with respect to an infinite momentum track (see fig. 87). This angle is measured by the
displacement of the outer RPC strip hit relative to the
inner RPC strip hit. A track with a transverse momentum of 30 GeV/c gives a displacement of 8x - 1 cm,
corresponding to the width of one strip. A trigger
selection requiring less or equal to 1 strip displacement
(shaded area in fig. 87) is therefore equivalent to a
transverse momentum cut at - 30 GeV/c. Similarly,
wider displacement requirements lead to lower PT Cut'
To reduce random triggers caused by punch through
particles faking high momentum muons, we further
require a coincidence between the RPC signals from
stations A and B.
Fig. 88 shows the trigger efficiency as a function of
the muon transverse momentum, PT based on the
RPC signals . The curves are obtained by full Monte
Carlo simulation including the complete L3P geometry .
Curves (a), (b) and (c) correspond to the requirements
of outer to inner displacements to be less or equal to 1,
2, and 3 strips, respectively . The curve shapes are
determined by the strip width and the multiple scattering of the muons. Inclusive muon trigger rates based
on these trigger conditions are computed by fully simuTable 19
Muon barrel trigger rates
Fig. 87 . The principle of the muon trigger .
road
rate [kHz]
PT at 90%
_ 1 strip displacement
< 2 strip displacement
< 3 strip displacement
0.5
3
20
37
20
13
efficiency [GeV]
76
L3P Collaboration / Precision experiment at LHC
lating events generated by PYTHIA. Table 19 lists the
trigger rates for y= 10 34 cm -2 s-1 , and the 90%
efficient point of these three trigger conditions . Trigger
with 90% efficiency at 13 GeV results in a 20 kHz rate,
well suitable for level 0 requirements .
Such a trigger can be implemented in hardware by
using fast "programmable gate arrays", which allows a
compact design maintaining the flexibility of software
programmability.
6.7.2. Endcap trigger
A trigger selection based on RPC signals is also
used in the endcap region, Fig. 89 shows the trigger
efficiency as a function of the muon transverse momentum in the endcap region . The difference between figs .
88 and 89 comes from the worsening of the transverse
momentum resolution . In fig. 89 the solid lines show
the efficiency based on the RPCs only, by similar
criteria as for the barrel muon trigger. The nonzero
trigger efficiency at less than 10 GeV PT causes a high
trigger rate because of the rapidly increasingly muon
rates at low transverse momentum .
To improve the trigger performance, the drift tube
timing is used to provide an independent trigger selection with better momentum resolution . As has been
described before, the drift tubes in the endcap stations
A and B are arranged such that the two layers separated by 18 cm measure the same drift time for an
infinite momentum particle from the interaction point.
The difference of the drift time from neighboring layers therefore measures the bending of the muon track,
thus the momentum . With a 250 p,m single wire resolution the angular resolution is 2 mrad, three times
better than the RPC-only measurement.
The dashed line in fig. 89 shows the trigger efficiency by requiring the time difference, bt, between
the two layers to be less than 120 ns, and <_ 2 RPC
strip displacements. The curve maintains a 90% effi-
08
T
U
Q)06
U
W
04
02
RPC only :
(a) <_ 1 strip displacement
(b) <_ 2 strip displacement
(c) <_ 3 strip displacement
RPC + Drift tube
(d) <_ 2 strip displacement
ST< 120 ris
I
30
PT(GeV)
I
40
50
Fig. 89 The endcap muon trigger efficiency .
Table 20
Muon endcap trigger rates
road
rate [kHz]
< 1 strip displacement
< 2 strip displacement
and drift tube timing
< 3 strip displacement
20
46
25
110
PT at 90%
efficiency [GeV]
50
20
20
13
ciency at 20 GeV PT . Below 20 GeV, the rejection
power improves from the RPC-only trigger. The total
trigger rate and the 90% efficiency points of these
trigger conditions are listed in table 20 . The rate improves from 46 kHz to 25 kHz by including the drift
time trigger.
Such a trigger can be realised by the coincidence of
the signals from the corresponding tubes in these two
tube layers . The momentum threshold can be continuously tuned by changing the coincidence time window .
The (h tubes in the endcap station B provide two
independent measurements and reduce the inefficiency
due to dead space.
In summary, the UP muon system can perform at
Level 0 a single muon trigger with a momentum threshold of - 13 GeV (90% efficient) for the barrel and
- 20 GeV for the endcap region, with a total trigger
rate of 45 kHz. Trigger decisions are made using local
detector information and therefore fast . Trigger
thresholds can be adjusted to reduce the rates.
6.7.3. Readout
The ration detectors are readout by the electronics
similar to the tracker system . Pipeline TDCs and hit
detectors with 2 ws buffer are used to digitize the drift
tubes and the proportional chambers, respectively . The
maximal drift time for the drift tubes in the muon
system is 1 ws. The TDC circuit therefore scans for 60
memory locations, corresponding to 1 ws time, at a
triggered event . The hit detectors, including the RPCs
and the proportional chambers, are readout by 1 bit
pipelines as in the tracker system . Since many of the
signals are used for trigger purpose, the discriminated
signals must be split out to be used for trigger decisions.
6.8. Summary
The UP muon system is designed to trigger and
measure the muons outside of the calorimeters . The
tracks found in the ration system are used to identify
the muons from the collision, and to guide the track
finding in the inner tracker system . The system achieves
a combined momentum resolution of 14% for a 100
GeV muon, with less than 2% misreconstruction . Sin-
77
L3P Collaboration / Precision experiment at LHC
_
13 .5mtoI .P .
3
Fig. 90 . The very forward calorimeter (3
< I i7 l
<_ 4.6).
gle muon trigger based on the fast signals from the
RPCs accept muons with a threshold of PT - 20 GeV
and a rate of less than 50 kHz.
Together with the central tracker system, the UP
can identify and measure the muons with 98% efficiency and - 1% resolution at y= 10 3° cm -2 s-1 , as
was shown in figs . 31-34 in section 3.
7. The very forward system
The UP calorimetry extends in the forward direction down to 1771 < 4.6. In order to keep the rates and
the radiation at an acceptable level the forward
calorimeters ( 1771 > 3) are placed at a distance of 13 .5
m from the interaction point (see figs . 2 and 90).
A summary of the geometry of the forward
calorimeters is given in table 21 .
In the hadronic part the energy flow is measured
using the same technique as with the central calorimeter: shower particles are recorded in thin MWPC's .
The charge from the amplification process is read out
through a pad system . Pads (area ^ 150 cm2) in successive planes are grouped to form towers pointing to
the interaction region . Wire rates are typically 200
kHz/cm at y= 1 X 10 33 cm -2 s-1 . At this luminosity
chamber meanlife will exceed 40 yr as can be inferred
from operation of the L3 Hadron Calorimeter . A development program is required to ensure continued
chamber operation at full LHC luminosity . The mechanics is conceived such that chamber planes (area
= 8 m2) may be exchanged easily at regular preventive
maintenance intervals. The transverse segmentation
(A77 X 0¢ = 2 ,rr/32 = 0.196) is sufficiently fine for not
limiting the transverse energy resolution, that is mainly
determined by the sampling frequency and the gas.
Details of the energy resolution are discussed together
with the central hadron calorimeter . For reasons of
radiation hardness (at T7 = 4.6 near the shower maximum = 13 MRad/yr are deposited at integrated luminosities of 1 .0 X 10 4 cm -Z /yr) and rate the electromagnetic part of the forward calorimeter is designed as
a thin gap liquid argon calorimeter . The ionization
signals from the shower particles crossing the 2 X 1 mm
thin gaps are integrated for 45 ns . Linearity and energy
resolution are insensitive to this clipping procedure
[55] . A test-module is under construction at Aachen .
The electromagnetic energy resolution of 28%/
with a o-E < 3% at 100 GeV shows that this simple
device with few (13) active planes will measure the
dominant part of the forward energy flow . The resolution of the combined very forward calorimeters will be
approximately constant (- 10%) for laterally contained
jets at all energies of interest (> 75 GeV) . This result is
based on a Monte Carlo study reproducing the L3
hadron calorimeter results [56] to better than 3% . The
transverse segmentation is the same for both parts of
the very forward calorimeter . In the electromagnetic
part however the outer 2 (out of 8) rings are further
subdivided (by 4) for having sufficiently low capacitances compatible with the signal rise time requirements. A longitudinal segmentation (4A, 5A, 4A) of the
hadronic part is desireable for the analysis of the LHC
beam induced background . Thus in total the very forward calorimetry will have 2432 channels . The main
purpose of the very forward calorimeter is the detection of QCD multi-jet processes as background for
high missing ET physics (e .g . SUSY).
V
Table 21
Forward calorimeter geometry. Dimensions of active planes
e.m . part
had . part
e.m . part
had. part
rm,n (Z ., n) [m]
Zmm [MI
Zmax [m]
13 .570
14.252
13 .939
16 .400
Absorber
type
Sampling
Sampling
thickness [mm]
Absorber
total [A]
Weight [t]
Cu
Cu
13X2 Xo
26x0 .5A
28 .6
75 .5
2.49
13 .00
19
130
0.281
0.295
rmax (Zmax) [m]
1 .391
1 .637
No. of pads
per plane
448
256
78
L3P Collaboration / Precision experiment at LHC
Table 22
UP readout
Subdetector
Central Tracker
Straw Tubes
Proportional Chambers
Gas microstrip Chambers
EM Calorimeter
Had Calorimeter
Muon System
Drift Tubes
Proportional Chambers
RPCs
Readout
Channels
TDC
Hit
Hit
ADC
ADC
500 k
1000 k
100 k
200 k
8k
TDC
Hit
Hit
70 k
180 k
200 k
8. Triggering and data acquisition
The UP detector contains an order of magnitude
more readout channels than the L3 detector . In going
from LEP to LHC, therefore, the trigger and data
acquisition system must be able to contend with a
front-end data rate almost four orders of magnitude
above LEP, yet with comparable final rates to tape .
The readout electronics and trigger logic has been
described in the corresponding Subdetector chapters .
The overall readout is summarized below in table 22 .
All detectors in UP are fully pipelined, with data
clocked into digital pipelines every 15 ns . A multi-layer
trigger system controls the transfer of data out of the
local detector pipeline, and off the detector .
The trigger is organized as shown in fig. 91 . A fast
and simple hardware pre-trigger defined as level 0
67MHz
Central
Tracker
ECAL
HCAL
Muon
System
RPC+Drift Tubes
level-0
pip e lines
Star Chambers
10-100kHz
1-lOkHz
FIFO
FIFO
FIFO
level-3
farm
Data Storage
n i clusters with ET > ETi
and/or
n= clusters with ET > ET2
and/or
m i muons with PT > PTI
and/or
m 2 muons with PT > PT2
As the Level 0 data are prompt, and the logic consists
of simple combinations, the Level 0 result is available
well within the 2 l.Ls detector pipeline delay.
The Level 1 trigger is initiated by a Level 0 accept .
Level 1 consists of a hardware processor which operates on calorimetric tower sums, and makes pattern
recognition based on the star chamber information .
Since the front end rate has been reduced by a factor
of 10'-10' by Level 0, Level 1 has ample processing
time .
A Level 1 accept initiates full detector readout to
Level 2. Level 2, which consists of one or more software processors may consists of single units receiving
the full detector information, or sub-units receiving
information from individual subdetectors . Level 2 is
used to perform software cuts on the high-quality main
readout data . Level 2 refines isolation cuts and momentum measurements, as well as having central
tracker data available. The analysis routines will probably operate on a single sub-detector basis, with simple
combinatorial logic at the end to form a trigger decision . Depending on advances in micro-computing,
however, Level 2 may be able to perform UP-wide
analysis .
The results of the Level 2 analysis are used to
throttle data into Level 3. Level 3 performs full event
reconstruction, and transfers data to archival storage.
9. Calorimeter options
FIFO
level-2 (full readout)
10-10OHz
gates data out of the 0th level pipeline into a 1st level
pipeline . The Level 0 trigger is formed from specially
constructed trigger data consisting of discriminated
cluster energies for the electromagnetic calorimetric
trigger, and RPC and drift tube information for the
muon trigger. Level 0 is a combinatorial trigger. It
triggers on events with
10-100MB/s
Fig. 91 . Trigger and readout organization .
In the following we describe four options for the
calorimeters . The first option concerns a precision
barrel calorimeter using noble liquids (see section 9.1).
The second and third option refer to the e.m . endcap
calorimeters (see sections 9.2 and 9.3). The fourth
option deals with the hadron endcap calorimeter (see
section 9.3).
79
L3P Collaboration / Precision experiment at LHC
Table 23
Relative light outputs
LKr attenuation length : 1 0 ± 0 1 m
(at n = 1 .4)
01
0
10
20
30
Thickness of LKr layer (cm)
Ar
0 .35+0.04
0 .42±0.04
Liquid
Solid
0.53+0.05
1 .3 ±0 .1
Xe
0.9
1.9
NaI(TI)
1.0
The detector can be calibrated using a's in situ . The
attenuation length of scintillation light has been recently measured cf. [57] to be about 1 m by varying the
LKr thickness between an a source and a photo-detector as shown in fig. 92 .
Table 23 shows the scintillation yield of noble liquids [57,61,62], normalized to the light yield of Nal(Tl).
The pure LKr scintillation signals are characterized
by a fast rise time ( < 10 ns) and about 100 ns decay
time . A timing resolution of 0.8 ns has been achieved
in a time-of-flight experiment cf . [57]) using a conic
shaped LKr scintillating detector of the size of 75 1.
The decay time of pure LKr can be reduced mixing
with a few % of xenon. Such wavelength shifting technique has been well established in the case of LAr
doped with a few % Xe [63] with nearly 100% efficiency and has been measured by us for LKr.
The production yield of krypton is ten times larger
and the price then times lower than xenon.
40
Fig. 92. The dependence of the LKr scintillation light signal
on the thickness of LKr layer.
9.1 . Option for a precision barrel EM calorimeter with
liquid krypton
Liquid krypton (LKr) is radiation hard and its scintillation light is intense (3 X 10 7 photons/GeV) [57,59]
comparable with that of Nal (see table 23, taken from
ref. [57]). Its radiation length is 4.6 cm and Molière
radius 4 .8 cm .
Scintillating liquid and solid xenon and krypton
detectors have been studied intensively at LAA-CERN
[58], ITEP [57], MIT [59], and Japan [60] . Recent
research work [57-60] using a, e/ ,rr beams and heavy
ion beams has shown: UV silicon photodiodes and
amplifiers work well inside LKr, with an effective
quantum efficiency > 50% and 10 ns peaking time .
0
Kr
9.1 .1 . Design of the LKr calorimeter
The side view of the proposed detector is shown in
fig. 93 . The UP LKr calorimeter covers the inner
tracker over an 77 range of ± 1 .4 at a radius of 2 .45 m.
b
UP Detector Phase II
O
Fig. 93 . Side view of the LKr EM detector .
80
L3P Collaboration / Precision experiment at LHC
MgF2 Coated
AI Cell Partition
Photo Sensor
(Si Photodiodes)
Inner wall
Fig. 94. Schematic drawing of a single cell .
The geometry can be modified upon further studies.
The depth of the calorimeter is 23 X0 (1 .05 m) resulting in an active volume of 145 m3. A very thin (< 0.1
X0) front window of the liquid container is realized by
suspending it from the stable outer wall .
The LKr is typically operated at 1.2 atm and 119 ±
0.5 K with the system monitored by the scintillation
light from one u source per photodiode, thermal sensors, and pressure gauges . This walls separating detection cells occupy < 1% of the active volume. They are
coated with Al and MgF2 to serve as UV mirrors. Such
mirrors have been shown to be radiation hard for UV
light from LKr/Xe [59] and an average reflectivity of
88% for 170 rim light was achieved [64]. The scintillation light is recorded using thin photodiodes mounted
at the outer radius of the cells as shown in fig. 94 . The
diodes as well as the fast amplifiers are submersed in
LKr with the mirrors which serve as Faraday shields,
providing a stable operating environment .
The active volume is subdivided into 50000 cells,
each covering a 871 _ 80 = 0.02. The cells all point
toward the intersection point. Monte Carlo simulations
show that the transverse shower center of a high energy e/y can be located with 1 mm precision using the
center of gravity method, and the total energy and
transverse shower profile measurements, yield an overall -rr/e suppression better than 10 -° . The expected
intrinsic e/y energy resolution, integrated over a 4 X 4
cells of 25 radiation length LKr calorimeter, is better
than 0.5%, as shown in fig. 95 .
One outstanding R&D issue is the uniformity in
light collection efficiency in a LKr calorimeter. Sufficient uniformity has been demonstrated for LXe by
0
w
w
w
0006
0
0 002
0
10
10 '
E (GeV)
0
10
3
Fig. 95 . The intrinsic energy resolution, predicted by the
GEANT Monte Carlo program, as a function of e/ -y energy.
using 3 diodes/cell . Additional diodes could be installed for the proposed LKr calorimeter to achieve
better uniformity or longitudinal segmentation if required in the future.
9.2. Option 1 f6r the forward /backward EM-calorimeters
The forward electromagnetic calorimeter uses the
mature technology of the total absorption Cherenkov
counter arrays [65] . Cherenkov light is intrinsically fast,
allowing the separation of LHC bunches!
Our design parameters are based on existing endcap calorimeters . We use a heavy transparent liquid as
the Cherenkov radiating medium . The liquid which
could be withdrawn and periodically reprocessed to
deal with the radiation damage. In the case of high
luminosity the purification process could be done in a
closed loop . The ultimate radiation resistance is determined by that of the photodetector window : we assume
that this could be solved with a quartz or magnesium
fluoride window .
There are several candidates for the heavy liquid,
some of them used previously . The most promising is
1,1,2,2 tetrabromoethane which has a density of 2.96
g/em 2, an index of refraction of 1 .64, a radiation
length of 3.6 cm and a Molière radius of 3.2 cm . The
addition of a WLS, POPOP, increases the light yield by
a factor of two for a typical S11 photocathode, and can
be removed by simple filtering. Pure distilled material
is colorless to the eye . The energy resolution measured
at 150 MeV was 0.16. Such counters were generally
replaced by lead glass because of the difficulties of
working with liquids. The ability to exchange the liquid
now has become an advantage. Furthermore since the
cells can be geometrically defined by thin mirrors the
problems of dead space is minimized and there is no
need for precise tolerances as is the case for seperate
solid counters .
We discuss the photoelectron yield and the noise:
We first assume that the liquid medium has the same
physical properties as lead glass and that the number
of photoelectrons is 1800 E (GeV) . An amplifier with
1 .5 microsecond peaking time has about 300 noise
electrons, so that 15 ns has 3000 noise electrons. Summing over 9 counters given 9000 noise electrons. Using
the worst case, only one photodiode, we estimate that
the noise at 100 GeV will be 5% using a photodiode, a
front end amplifier with the same transconduct ance
and the same light yield. This can be vastly improved
by several means. If the phototriode has a gain of 4 at
2 T then this number would decrease to 1 .25% .
One can expect twice as much light per GeV with a
WLS dissolved in the liquid, as experimentally observed in previous tetrabromoethane counters with the
WLS [66].
L3P Collaboration / Precision experiment at LHC
The front end electronics can be improved using
better transconductance devices which are being developed for optical fiber data transmission . At present a
noise equivalent of 1250 electrons could be realized
with commerical components . The development of the
electronics is exactly the same as needed by the crystals
of the barrel and they can be identical.
Combining these techniques we conservatively estimate that the resolution due to noise can be less than
70%/E (GeV) giving a resolution
81
Brass element
WLS fibers
Wire plane
WLS fibers
Brass element
Fig . 96 . Assembly of a sampling plane .
The total volume of Cherenkov medium is 15 m3
per endcap .
One effect of radiation will be the release of
Bromine gas . The effects on mirrors immersed detectors and electronics, and electrical feedthrughs must be
studied as well as the purification plant . Tetrabromoethane does not have any unusual safety hazards
and is relatively inexpensive given that the purification
plant can be used to purify the initial material .
The photodetectors could be the same as those of
the crystals in the barrel resulting in economies of
scale and common R&D costs.
9.3. Option 2 for the forward/backward em-calorimeter
9.3.1 . Introduction
We propose an economical sampling brass/gas
calorimeter . The energy resolution will be 25%/
2% . The 71 coverage ranges from 1 .4 to 3 .0 . All the
parts are radiation hard . The sampling gas detector
provides fast optical signals .
F
9.3.2. Description of the sampling calorimeter
- The calorimeter will be assembled out of selfsupporting modules of moderate weight (typically 200 kg).
Each module in turn consists of easily produced sintered brass absorber pieces, between which thin gaps
allow for the insertion of the detector wire planes (fig.
96) .
- A compact design is obtained by integrating the
absorber into the detector part : The absorber plates
(brass) act as cathode relative to the anode wire planes.
In this way the detection gaps can be kept thin (3 mm)
(fig. 96) . The total depth of the electromagnetic
calorimeter is 25 to 29 Xo .
- The high rate of particles requires very fast signal
response and short dead times in order to avoid timing
problems and pile up : We therefore propose to use
short signal wires (< 25 cm), small wire planes (< 625
cm2 ) and CF4 gas, known to have a fast drift time (100
wm/ns) [67] . As described below we make use of the
fast optical signal from the photons produced in the
avalanche at the wire .
In order to avoid aging we will use CF4 gas, which is
known not to age wire chambers [68] . All other materials (brass, tungsten wires, glassfibre epoxy, Araldite,
Capton, special Bicron wavelength shifting fibers WLS BCF-91 and BCF-98) are sufficiently radiation
hard .
In summary, the two endcaps will be assembled out
of 10 rings, with 910 modules in total . The weight
amounts to about 200 t.
9.3 .3. Sampling detector
The modules are subdivided into absorber- and
detector planes . A sampling of 1/2 Xo is chosen up to
12 X0 's, and for the rest of the module the sampling is
one X0 .
Fig . 96 shows one sampling plane : two absorber
plates (sintered brass pieces with a surface of typically
25 x 25 cm 2 ) from the housing for the HV wire plane
and the 3 mm gas gap . The wires have a diameter of 50
wm and a spacing of 2 mm (the jitter for single particles is of the order of 10 ns) . Wavelenghth shifting
fibers (WLS) forming cells, are placed below and above
of the wire plane (fig . 97) and are fed through longitudinal channels along the module to an optical connector on the top of the module . Readout towers are
formed by grouping the fibers . Via a normal optical
fiber the signal of longitudinally grouped planes is led
to a multiple anode photomultiplier placed outside of
the magnet .
Preliminary tests with a prototype chamber working
with CF, have shown that short optical signals can be
observed in coincidence with the electrical signal from
the wire with an efficiency approaching 100% . Improvements in light collection, in matching the photocathode to the emitted light, and in purification of the
gas are expected to provide a substantial increase in
the signal amplitude . An R&D project has been started
by ETH Zurich at CERN .
The following advantages of this new technique may
be listed here :
- Contrary to organic scintillators, the number of
photons produced can be varied by tuning the high
voltage of the wire plane . A scintillator sheet of 3 mm
82
L3P Collaboration / Precision experiment at LHC
Ire
Fig. 98 . The design of parallel plate chamber.
Fig. 97 . Cell structure of the sampling detector . The top
picture shows a cross section of the detector cells The WLS
fibers act as cell walls They are screened on one side, in
order not to capture light from the neighbouring cell . The
lower picture shows the top view of the cells and indicates
how the fibers capture the light from all four sides of one
particular cell .
thickness typically produces a few thousand photons,
whereas 10000 to 100000 are expected from an
avalanche . CF4 is also known to have a relatively large
light output as compared to other gases [69] . Aging is
also less of a problem.
- The signal of minimum ionizing particles will be
well above the single photoelectron noise.
- The light signal duration is of the order of 1 ns,
whereas the electrical signal is much longer .
- The fast signal collection means that the pileup is
minimal.
- Good signal transport over long distances .
- An easy mechanical mounting (grouping of cells)
is possible .
9.3.4 . Segmentation
The granularity of the sampling calorimeter is determined by the read-out segmentation and matches
the hadron calorimeter subdivisions . The modules are
approximately pointing to the vertex. An inclination of
one degree relative to the radius vector has been
chosen in order to avoid cracks and detector-free regions. The detector cells, perpendicular to the radius
vector, (see fig. 2) are of variable size in order to keep
constant Ovl X 0(~ intervals. The WLS fibers belonging
to the same readout cell are joined together in the
particular optical connector on the back of the module .
With this segmentation the 01j X t1 (b granularity
will be 0.02 X 0.02 in the region 11 = 1.4 to 2.2 and
somewhat larger for 71 values up to 3. The mean rate
of charged particles per cell and bunch crossing at the
nominal luminosity of 1034 CM-2 s -1 will not exceed
0.03/bunch-crossing at -7 = 3.
The total number of readout channels needed is
28000. The PM signal will first be shaped and split.
Large and small amplitude signals are red to wide
dynamic range ADC's in order to record showers as
well as minimum ionizing particles.
9.4 . Option for the hadron endcap calorimeter with parallel plate chambers
9.4.1 . Introduction
As was discussed before (see section 5) the pileup
rates in the hadron calorimeter end caps are high and
make the detection of minimum ionizing particles difficult . The limitation comes from the speed of proportional chambers namely one needs first to wait for
ionization to drift to the anode wire, this creates a
jitter of 25-30 ns . Then the shaping and integration
add some 50 ns, thus the total period of the pile up
background collection becomes = 80 ns . This corresponds to four or five beam crossings .
We plan to reduce the time and integrate effectively
over just one beam crossing . For this we plan to use a
novel technique : parallel plate chambers [70] which are
under study since more than two years [71] . This device
has no drift time jitter and since total measured pulse
length is less than 10 ns one can have completed the
charge integration before the next beam crossing comes.
9.4 .2 . Detector design
The parallel plate chamber (PPC) is a single gap,
gaseous detector with planar electrodes, working in the
avalanche mode . The detector design is schematically
shown in fig. 98 .
50
"e"
Pos ion collection
collection
HV = 5400 V
100
i lo
20
Time p s
30
Fig. 99 . The current pulse from PPC with SS Fe source .
i
L3P Collaboration / Precision experiment at LHC
83
40
40
c
G)
LU 20
;_
O
N
Fig. 102. The time-of-flight measured by two planes of PPCs .
-160
5ns/div
Fig. 100. The fast component of the PPC current pulse (after
amplifier) .
Two planar electrodes are separated by a narrow (1
mm) gas gap, with the help of precision spacers . Since
the size of the electrodes is relatively small (5 em X 5
cm) they can be made flat and parallel to provide the
required accuracy and uniformity of 5 Win for the gap.
Out of such "elementary" cells a mosaique of necessary size can be constructed .
Fig. 99 shows current pulse shape of the chamber,
the pulse has a very sharp peak which reflects the
avalanche development, and a long flat tail corresponding to positive ion collection on the cathode. Fig. 100
shows the sharp peak when the time scale is extended .
The full width at the base is less than 3 ns.
The PPC were tested with intense radioactive
sources and showed no saturation effects up to a
collected charge of 1 pC. We have operated PPC in an
intense beam (over 10 6 particles cm -2 s -2 ), and observed no efficiency loss . The measured detection efficiency for minimum ionizing particles is shown in fig.
101. The efficiency at the plateau (= 90%) corresponds
to the value expected from the number of ion pairs left
by minimum ionizing particle in 1 mm of iC,H lo . The
time jitter distribution obtained at the same time is
shown in fig. 102. It is the distribution of time-of-flight
of relativistic particles measured by two PPC planes.
Extracting the time resolution corresponding to one
80
T
U
40
w
0
5300
5500
5700
Voltage (V)
5900
Fig. 101. The PPC efficiency for minimum ionizing particles as
function of high voltage.
PPC we get: At = 682 ps/ v/2 = 480 ps . This demonstrates that PPC can be used for high precision time
measurements.
9.4 .3. End cap hadron calorimeter design
The machanical structure of the end cap hadron
calorimeter is not changed compared to the one described in section 5. The PPC are put into the gaps
forseen for the detectors. However in this case somewhat faster electronics should be used (the preamplifier rise time about 5 ns). Using different sizes of PPC
cells in the mosaique planes one can keep the pile up
noise level well below the signal from a minimum
ionizing particle . The total number of projective towers
(readout channels) is = 6000 for both endcaps.
10 . Physics examples
10.1 Introduction
One of the main physics goals at the LHC is to solve
the outstanding problem of the electroweak symmetry
breaking mechanism. Present theoretical concepts include a fundamental scalar Higgs boson in the Standard Model (SM) as well as in its supersymmetric
extension. Gauge boson pair production in the presence of a strongly interacting electroweak symmetry
breaking sector could be an alternative possibility.
These processes together with other predictions from
extensions of the SM are used as benchmark processes
to review the physics discovery potential of the L3P
detector .
The Higgs search is an excellent reference physics
process to establish detector requirements adequate
for the solution of the electroweak symmetry breaking
sector . In particular, the search for the intermediatemass Higgs (m z < m H < 2m z ) is known to put rather
challenging conditions on the detector design .
In the following section the expected physics performance of the L3P detector is illustrated through the
discovery potential for the Standard Model Higgs.
84
L3P Collaboration / Precision experiment at LHC
10.2. Higgs detection
10.2.1 . H - yy: from 80 to 150 GeV
The most promising detection mode for an intermediate mass Higgs is via two photon decay . A key design
goal of our detector is to be able to discover the Higgs
via H -> yy within one year's running time at y= 10 34
cm -- s -1 . As the Higgs signal will appear as a small
peak on top of a large background, the ability to
maintain high resolution even under the most extenuating running circumstances is required .
The signal and background processes were simulated with the PYTHIA 5.6 Monte Carlo program [72] .
The signal cross sections were taken from refs . [73,74].
We have considered the following background processes :
- Prompt diphotons from qq - yy,
- prompt diphotons from gg --> -y-y, including a
QCD correction factor of 1.5,
- prompt single photons from qq --, gy and qg
qy, accompanied by another photon from QCD
bremsstrahlung or by a -rr o misidentified as a photon.
- fake diphotons from lets misidentified as photons .
The background contribution from dijets is expected to
be small compared to the irreducible background after
the -7r° rejection, described below, is employed . Additional background can come from electrons misidentified as photons, however the clean tracking at large
radius reduces this background well below the irreducible yy background #z .
A complete simulation of the detector response has
been performed in order to predict the statistical significance obtained for H - yy . Photons were searched
for in the crystal barrel ( 177 1 < 1 .4). The GEANT [75]
simulation includes the intrinsic resolution of the crystals, as well as the effects of pileup at high luminosity,
electronic noise and calibration error . The intrinsic
resolution has been simulated with 23 radiation length
crystals, as well as a 5 cm carbon fiber support in front
of the crystals . In addition we have included the 3% X0
of additional material from the tracker. The reconstructed signal taking into account each of these effects
for a 110 GeV H - yy and an integrated luminosity of
10 5 pb - ' is plotted in the following figures: Fig. 103
shows the deviation from the Higgs mass of the yy
invariant mass spectrum, (m'Yy-m,)/m,, in %, for
the intrinsic resolution only . The additional contributions from noise, pileup, calibration and vertex uncertainty (°z = 5.7 cm) are shown in figs . 104 through 107
#z A veto efficiency of 98 .6% per electron brings this background rate a factor 10 below the irreducible -yy background for m H = m 7 .
240
0
N 160
CD
w
b
z so
-4
0
4
Fig. 103 . (mvy - mH)/ m H (mH=110 GeV), intrinsic resolution only a = 0 .30% .
240
ô
~160
w
ô
z0 80
4
Fig. 104.
(M YY
- rnH)/ m H (m H =110 GeV), adding pileup,
a = 0.32% .
240
ô
M160
cm
w
ô
z eo
-4
Fig. 105 . (m,,y-mH)/mH (m H =110 GeV), adding electronic noise, a = 0 .35% .
85
L3P Collaboration / Precision experiment at LHC
20
240
15
dU
C
N
`-' 10
O
C
O
N 160
cN
5
W
ô
90
Z 80
320
0
N
ô 240
cam
w 160
ô
ô
Z 80
( M Yti -
130
150
Fig . 109 . Significance vs m H for H - ,y-y, with 10 5 pb -1 .
Fig . 106 . (MYY-mH)/mH (m H =110 GeV), adding 0 .5%
calibration error, Q = 0.42% .
Fig . 107 .
110
M H (GeV)
M H)/ mH (m H =110 GeV), adding vertex
uncertainty, v = 0 .77% .
Fig. 108 . yy invariant mass for various values of m H , and 10 5
pb-1 .
(for details see section 4) . The expected yy invariant
mass plot for H -> yy for m H = 80, 90, 100, 110, 120,
130, 140, 150 GeV is shown in fig. 108 for l q, I <
1.4 #3 , and an integrated luminosity of 10 5 pb -1 .
After background subtraction, this yields a statistical
significance (S/ fB) as a function of mH as shown in
fig. 109. This curve demonstrates the capability to
discover the Higgs via H - yy within one year's running time at Y=10 34 cm -2 s - ' for mH z 80 GeV.
Reducible background from QCD jets :
The QCD jet cross section is - 7 orders of magnitude higher than the irreducible prompt yy cross section. Therefore, an excellent rejection ( - 10 8 ) against
jets faking photons is essential. As such an effect arises
mainly from jets containing an energetic 1T0 or r), the
high granularity of the detector allows sharp isolation
cuts and good two--y separation .
The performance of the UP detector was studied
using a full GÉANT simulation . A sample of 10 5 jet
events #4 was generated with PYTHIA . A shower
shape cut on the ratio (E7X7-E3X3)/E3x3, where
E 7X7(E3X3) is the energy deposition in 7 x 7 (3 X 3)
cells, was applied to the EM clusters . This ratio is
insensitive to pileup even at the highest luminosities,
due to the fine granularity of the detector . Requiring
the ratio to be below 7%, the y efficiency remains
above 96%, independent of energy . The resulting factor for jets is 80 (330) for an ETX7 threshold of 25 (40)
GeV. An additional suppression factor of 10 was obtained if the total PT of charged particles measured 1n
the inner tracker was required to be below 4 GeV in
the same Or) X OQ region .
The total suppression factor for jets amounts to
1.1 X 10 8, taking into account the probability of having
simultaneously two EM clusters (ET > 40 GeV, ET >
25 GeV) within the same event. The magnitude of the
#3
#4
p :p
> 40 GeV, pp >- 25 GeV .
PT `d > 40 GeV, 1771 < 0 .5 .
86
L 3P Collaboration / Precision experiment at LHC
background contribution from jets is thus well below
the irreducible yy background .
10.2.2 . H-42 : 140<M H <SOOGeV
In the mass range above 140 GeV and up to about
800 GeV, the channel H - ZZ - 4 leptons offers a
very clean signature.
As an illustration, we select three Higgs masses,
MH = 150, 300 and 800 GeV. The width, FH, of the
resonance depends strongly on the mass . For low mass,
the detector resolution is important, since for instance,
when mH < 200 GeV, FH < 2 GeV. The UP detector
is very well suited in this case since the momentum
resolution for (isolated) leptons is of the order of 1% .
For high masses, the natural mass width obeys
F H (TeV) = 0.5 x mH (m, in TeV), and thus determines the shape of the signal .
The angular acceptance plays an important role in
detecting the signal, particularly at lower Higgs masses .
Because of its capability of detecting electrons and
muons in the region 1771 < 3, the UP detector is well
adapted for these searches .
At high lepton energies, QED radiative corrections
have been considered . The final state radiation produces photons which can broaden the invariant mass
peak of the four leptons. Fig. 110 shows the effect of
radiative corrections in the invariant mass spectra for a
150 GeV Higgs decaying into 4 electrons without radiation (fig . 110a), with radiation (fig . 110b) and after
photons within a cone of R = 0.08 are included (fig .
1100. We sec that for the lowest Higgs masses, considerable improvements in significance might be obtained
after including radiative photons in the reconstruction .
The main background contributions to the H - 4 e
signal are expected from _
- pp - tY - W+bW - b, where the W bosons decay
leptonically, W -ev . Additional semileptonic decays
of the b-quarks, b - e vc, or extra leptons in the jets
can yield four or more leptons in the final state faking
the Higgs signal . (It is however known that a large top
background reduction is achieved when the leptons are
required to be isolated [76]). We use PYTHIA to
generate the top signal assuming mt = 150 GeV. The
cross section is taken from ref. [77] .
- pp -) Z° Z° continuum production . The cross section for this process is calculated for each Higgs mass
with PYTHIA, by restricting the kinematic region of
the Z-bosons such as to generate the phase-space of
interest for a given Higgs mass range.
- pp - Z ° bb with Z° -I'f - and b, b -f vc . We
use the exact calculation for gg -> Zbb with mass terms
implemented in PYTHIA [83] predicting a cross section of 0.3 pb .
For signal and background evaluation a detector
parametrization is employed . In general the results of
detailed GEANT studies on resolution, pattern recog-
M eeee(GeV)
Fig. 110. Radiative corrections for Higgs signal for m H = 150
GeV. (a) without radiation (b) with radiation, (c) including
photons within R = 0.08.
nition, lepton isolation requirements and trigger have
been used for the detector parametrization (see discussions in the different detector sections), pile-up effects
expected at high luminosity are included as well . The
overall reconstruction efficiency for an isolated lepton
is assumed to be = 97% (see section 3) .
The Higgs signal has been generated with PYTHIA,
taking into account the radiation effects described
above. The cross sections and the decay branching
ratios into four leptons are taken from ref. [80] . To
extract a Higgs above background, the following cuts
are applied: all leptons must satisfy PT > 20 GeV. At
least one invariant mass of two leptons must be within
±4o- of the Z° mass m Z (this cut is relaxed for
m H = 150 GeV) . The isolation cut requires the transverse energy deposited within a cone of size OR = 0.2,
excluding the photons produced together with the lep-
87
L3P Collaboration / Precision experiment at LHC
Table 24
H1ggs detection efficiency in the four lepton channel
in [GeV]
150
300
800
H
7
QBr [fb]
15
1.6
Acceptance ( 1 771 < 3)
85%
93%
96%
51%
75%
87%
PT > 20 GeV
Z° mass cut
48%
73%
85%
Isolation
ET <<< 5 GeV in AR < 0.2
43%
70%
82%
(20 min. bias pileup included)
Acceptance X efficiency
42%
63%
74%
ton, to be less than 5 GeV. For very high mass (i .e . 800
GeV), the cut pT +pZ2 > 300 GeV is also used . Table
24 lists the different contributions to the signal detection efficiency .
The reconstructed invariant mass distributions of
the four leptons mrcer together with the total expected
background contributions, after all cuts are applied,
are plotted in figs . 111, 112 and 113 .
Table 25 shows the expected number of events for
the signal and the background after all cuts are applied, and for an integrated luminosity of 10 5 pb -1 . All
the decay modes H --, ZZ - eeee, eel.Ll.L and Irlxlxl.L
are included. The numbers quoted for the signal and
the background are within m H ± 2o-H . The significance
is shown as well . Table 25 demonstrates that the Higgs
can be discovered within one year of running at a
luminosity of 10 34 cm -2 s-1 within the mass range
considered . One can also see that for masses below 400
GeV, with a possible exception around - 170 GeV, a
more modest luminosity could be sufficient to detect
the signal within one year of operation.
10.3. Other physics topics
The ability to experimentally verify the electroweak
symmetry breaking mechanism was used as a bench-
Fig. 112. Higgs signal with background for mH = 300 GeV.
mark process for the UP detector design . As discussed
within the previous section, a discovery of the SM
Higgs from 80 GeV up to < 1 TeV is possible with the
UP detector . The H - yy decay mode is theoretically
interesting, due to its sensitivity to new (unpredicted)
charged particles entering the Higgs decay loop .
Searches for new physics signals at the LHC require
a thorough understanding of Standard Model physics
processes like ti, W and Z, intermediate vector boson
pair, direct y production, etc. in terms of absolute
rates and shapes of distributions.
Possible signals arising from extensions of the Standard Model have been studied as well, examples of
which are briefly summarized .
- Supersymmetry, which provides a natural explanation of symmetry breaking, predicts an entire new
set of particles. Searches for a supersymmetric Higgs as
well as gluino and squark signatures are therefore
particularly interesting . In the supersymmetric version
of the Higgs sector, the lightest neutral Higgs is pre-
60
,n 40
C
U7
w
20
I
750
I '
1000
Metef(GeV)
Fig. 111. Higgs signal with background for m H =150 .
T'TI~
1250
Fig. 113. Higgs signal with background for mH = 800 GeV.
88
L3P Collaboration / Precision experiment at LHC
Table 25
Expected number of Higgs events in the L3P detector and
signal significance
m, [GeV]
Signal
Background
S/F
150
189 ±14
69 ± 9
228±
1 .7
300
863 ±30
270 +16
525± 2 .0
800
40 ±6
15 ± 4
10 .3±1 .6
dicted to have a mass in the vicinity of the Z°. In this
mass region (like in the SM case) a search in the yy
decay channel is required, where the UP detector has
an excellent discovery potential, thus exploring a large
fraction of the available parameter space.
- There are several theoretical ideas for an alterna-
tive symmetry breaking mechanism without the exis-
Acknowledgements
Mr . J.L . Benichou assisted by Mr . F. Limia-Conde
and Mr . R. Loos are thanked for their technical studies
and drawings . The persevering help of our secretaries,
Mrs. L. Barrin, Y. Bernard and R. Decreuse, including
the production of many drawings is acknowledged
warmly .
References
[1]
[2]
[3]
[4]
[5]
[6]
tence of a fundamental Higgs scalar . They involve new
strong forces which could manifest themselves via
W, .Z,. scattering resonances . A signal of such p-like
heavy resonances as well as new heavy vector bosons
(Z') can be established with the UP detector. The
ultimate mass reach for these heavy vector particles is
dictated by the available integrated luminosity and not
by the detector performance.
11 . Summary and conclusion
We have designed a detector to measure e, w, y
precisely, based on our experience in this type of
experiments. We believe these final states will most
likely yield new physics. Indeed, it would be a great
[7]
[8]
[9]
[10]
[11]
[12]
disappointment if all the results of the LHC confirm
the Standard Model.
We have chosen a conservative technology and used
as much as possible of the existing L3 equipment so as
to ensure that the detector will work at LHC turn on
and so as to be able to control the costs accurately .
This is based on the following observations :
There are great differences between Monte Carlo
predictions written on experimental proposals and final
experimental conditions . Very few Monte Carlo studies
have produced realistic backgrounds measured in high
intensity experiments. Detectors entirely based on new
technologies will behave correctly on test beams but
will seldom behave in the same way when put into an
experiment where radiation damage, beam sprays, cabling, alignment and supporting structures all tend to
degrade the detector performances. In addition, the
[13]
[14]
[15]
[l6]
[17]
[18]
[19]
cost of a general purpose LHC detector is a large
fraction of the cost of the accelerator. A 25% error in
the estimate can easily prevent its completion .
[20]
G. Dandy et al , Phys . Rev. Lett. 9 (1962) 36
J.J . Aubert et al ., Phys. Rev. Lett . 33 (1974) 1404 .
M.L . Pert et al ., Phys . Rev. Lett . 35 (1975) 1489 .
W. Braunschweig et al., Phys . Lett. 57B (1975) 407.
S.W . Herb et al ., Phys . Rev. Lett . 39 (1977) 252
D.L. Scharre et al ., Proc. Int. Symp . on Lepton and
Photon Interactions at High Energies, Bonn, August
24-29, 1981, ed . W. Pfeil (Physikalisches Institut, University of Bonn, Germany, 1981) p. 163.
K. Han et al ., Phys . Rev. Lett. 49 (1982) 1612;
C. Klopfenstein et al ., Phys . Rev. Lett 51 (1983) 160.
S. Behrends et al ., Phys. Rev. Lett . 52 (1984) 799.
First report by ARGUS and Crystal Ball groups presented by P.M . Tuts in : Proc . Symp . on Leptons and
Photons Interactions at High Energies, Cornell, 1983,
eds. D.G . Cassel and D.L . Kreinick (F .R . Newman Laboratory of Nuclear Studies, Cornell University, NY, 1983)
p. 284.
G. Armson et al . (UAI), Phys . Lett . 126B (1983) 39;
M. Banner et al . (UA2), Phys. Lett. 122B (1983) 476;
P. Bagnaia et al . (UA2), Phys . Lett. 129B (1983) 130.
G. Arnison et al . (UAI), Phys . Lett . 122B (1983) 10;
P. Bagnaia et al . (UA2), Phys . Lett . 138B (1984) 430;
P. Bagnaia et al . (UA2), Phys . Lett. 139B (1984) 105.
J.G . Asbury et al ., Phys. Rev. Lett . 18 (1967) 5;
J.G. Asbury et al ., Phys. Rev. Lett . 19 (1967) 869 ;
G. Alvensleben et al ., Phys . Rev. Lett 21 (1986) 1501 ;
U.J . Becker et al ., Phys . Rev. Lett . 21 (1968) 1504,
H. Alvensleben et al ., Phys. Rev. Lett 24 (1970) 786;
H Alvensleben et al ., Nucl . Phys . B18 (1970) 333;
H Alvensleben et al ., Phys. Rev. Lett . 25 (1970) 1373 ;
H. Alvensleben et al ., Phys. Rev. Lett . 25 (1970) 1377 ;
H Alvensleben et al ., Phys. Rev. Lett . 27 (1971) 888;
H. Alvensleben et al ., Phys. Rev . Lett . 28 (1972) 66 .
D. Antreasyan et al ., Phys . Rev. Lett . 45 (1980) 863.
B. Adeva et al ., Nucl . Instr. and Meth . A289 (1990) 35 .
U.J . Becker et al ., Nucl. Instr. and Meth . A306 (1990)
194.
Y.H . Chang et al , Nucl . Instr. and Meth . A311 (1992)
490.
U. Becker et al ., Nucl. Instr. and Meth . A315 (1992) 14 .
P Duinker et al ., Nucl . Instr. and Meth . A273 (1988) 814.
P. Baringer et al ., Nucl . Instr. and Meth . A254 (1987)
542;
M. Frautschi et al., Nucl . Instr. and Meth . A307 (1991)
52
Y.H. Chang et al ., submitted to Nucl . Instr. and Meth. ;
S.F. Biagi, Nucl . Instr. and Meth . A310 (1991) 133.
L3P Collaboration / Precision experiment at LHC
[21] R. Openshaw et al ., TRIUMF Preprint TRI PP 90-80
(1990) .
[22] F. Angelini et al ., Proc . ICFA studyweek, Barcelona
Sept. 1989, CERN 89-10, p. 465;
F. Hartjes et al., Ibid p. 455;
F. Angelini et al., Proc. ICFA workshop for LHC, Aachen
Oct. 1990, CERN 90-10, Vol. 3, p. 217;
M. Geijsbets et al ., Preprint NIKHEF-H/91-23, Sept .
1991 .
[23] U. Becker, J . Paradiso, Nucl . Instr. and Meth . 196 (1982)
381;
[24] GÉANT version 3.13 (September 1989); CERN Report
CERN DD/EE/84-1 .
[25] G. Ciappetti, A. DiCiaccio, ECFA-LHC Workshop,
Aachen, CERN 90-10 (1990) II-155 .
[26] H.U . Bengtsson and T. Sjôstrand, Comput . Phys. Commun. 46 (1987) 43 .
[27] P. Lecoq, M. Schussler and M. Schneegans, Nucl . Instr.
and Meth . A315 (1992) 337;
Crystal Clear Collaboration, CERN/DRDC P27/91-15;
W.W . Moses and S.E . Derenzo, IEEE Trans. Nucl . Sci.
NS-36 (1989) 137;
D.F . Anderson, IEEE Trans. Nucl. Sci. NS-36 (1989) 137;
D.F. Anderson, Nucl . Instr. and Meth . A287 (1990) 606;
M. Ishu and M. Kobayashi, KEK 92-10.
[28] Y. Valbis et al ., JETP Lett . 42 (1985) 172
S. Kubota et al ., Nucl . Instr. and Meth . A242 (1986) 291.
[29] Crystal Clear status report . CERN/DRDC/92-02.
[301 A.A. Asseev et al ., Proc. 2nd Int . Conf. on Calorimetry in
High Energy Physics, Capri, October 1991, (World Scientific) 313.
[31] P. Lecoq, P.J . Li, B. Rostaing, Nucl . Instr . and Meth .
A300 (1991) 240;
Z.Y . Wei et al ., Nucl . Instr. and Meth . A297 (1990) 163;
Z.W . Yin et al ., Nucl . Instr. and Meth . A275 (1989) 273;
G.J Britvich et al ., IFVE 91-134 Protvino.
[32] F. Carminati et al ., CERN/DRDC/91-54
[33] A. Gola et al ., Nucl . Instr. and Meth . A292 (1990) 648;
A. Gola et al ., Nucl . Phy. B (Proc. Supp .) 23A (1991) 207.
[34] DR2 Status report to DRDC, 14/1/92.
[35] S. Iwata, Nagoya University report DPNU-3-79 (1979);
U. Amaldi, Physica Scripta 23 (1981) 409;
A. Arefiev et al ., Nucl . Instr. and Meth . A285 (1989) 403;
A. Arefiev et al ., Nucl . Instr. and Meth . A245 (1986) 71
[361 P. Aarnio et al ., Nucl . Instr. and Meth . A303 (1991) 233;
D. Decamp et al ., Nucl . Instr. and Meth . A294 (1990)
121 .
[37] O. Adriani et al ., Nucl . Instr. and Meth . A302 (1991) 53 .
[38] J. Krueger CERN 90-10, ECFA 90-133, vol. III, p. 306.
[39] A Arefiev et al ., Nucl . Instr. and Meth . A288 (1990) 364.
[40] Expression of interest, L3 detector upgrade to LHC,
Towards the LHC Experimental Programme, Proc. of the
General Meeting on LHC Physics and Detectors,
ECFA-CERN, Evian- les-Bains, 5-8 March 1992, ed . S.
Hegarty (CERN) p. 303
[41] A. Arefiev et al ., Nucl . Instr. and Meth . A275 (1989) 71 .
[42] D. Decamp et al ., Nucl. Instr. and Meth . A294 (1990)
121.
[43] G.D . Alexeev et al ., Nucl . Instr. and Meth . 177 (1980)
385.
[44] S. Burov et al., CERN -EP/88-84, 15 July 1988 .
89
[45] T. Sjbstrand, Comput. Phys. Commun . 39 (1987) 367.
[46] R. Brun et al ., CERN DD/83/1 (1983) .
[47] SDC Technical Design Report, SDC group, submitted to
the SSCL.
[49] M. Fabre, thesis, Diss . ETH Nr . 9696, 1992.
[501 R. Santonico and R. Cardarelli, Nucl Instr. and Meth .
187 (1981) 337.
[51] M. Bertino et al ., Nucl . Instr. and Meth . A283 (1989) 654.
[521 P. Schilly et al ., Nucl . Instr. and Meth . 91 (1971) 221 ;
G. Charpak et al ., Nucl . Instr. and Meth . 91 (1971) 377.
[531 J.J . Aubert et al ., Nucl . Instr. and Meth . 89 (1975) 1.
[54] S. Miyamoto, Nucl. Instr. and Meth 30 (1964) 361.
[55] G. Costa et al ., Calorimetry in High Energy Physics Il,
Capri, Italy, 14-18 September 1991, ed . A. Eriditato
(World Scientific, Singapore) p. 237.
[56] J.F . Zhou, PhD Thesis, RWTH Aachen, PITHA 90/06,
O Adriani et al ., Nucl . Instr. and Meth . A302 (1991) 53 .
[57] D.Yu. Akimov et al ., submitted to Nucl. Instr. and Meth .
[581 T Ypsilantis, ECFA/LHC Meeting on Detector R&D,
Aachen, 90 ;
and LAA Progress Report, CERN 90/1 ;
J. Seguinot et al ., CERN LAA 92-004, submitted to Nucl .
Instr. and Meth .
[59] M. Chen et al ., ECFA Workshop, CERN 89-10, 1989 ;
M. Chen et al ., Detectors, submitted to Nucl . Instr. and
Meth .
[60] T. Doke, Portgal Phys . 12 (1981) 9;
K. Masuda et al ., Nucl. Instr. and Meth . A309 (1992) 489;
N. Ishida,
T. Doke et al ., submitted to Nucl . Instr. and Meth . ;
M. Miyajima, Japanese Physical Society Meeting at Osaka, March, 1990 ;
S. Sugimoto, M. Miyajima, S. Sasaki, and H. Tawara,
Osaka, OULNS 90-8, June 30, 1990 .
[61] J.A . Northrop, J.C. Gursky and A.E. Johnsruf, IEEE
Trans. Nucl . Sci NS-5 (1958) 81 .
[62] J.A . Northrop and R.A . Nobles, IEEE Trans. Nucl . Sci.
NS-3 (1956) 59 .
[63] S. Nanba et al , J. Phys . Soc. Jpn. 37 (1974) 1031 .
[64] P. Baillon et al ., Nucl . Instr. and Meth . A276 (89) 492;
and ibid 277 (89) 338.
[65] K. Ahmet et al ., Nucl . Instr. and Meth . A305 (1991) 275;
P. Aarmo et al ., Nucl . Instr. and Meth . A303 (1991) 233;
W. Bruckner et al ., Nucl . Instr, and Meth . A313 (1992)
345;
H. Baumeister et al ., Nucl . Instr. and Meth . A292 (1990)
81 ;
L. Bartoszek et al., Nucl . Instr. and Meth . A301 (1990)
47 ;
M. Àkrawy et al ., Nucl . Instr. and Meth . A290 (1990) 76 .
[66] G. Davidson, Ph .D . thesis, MIT, p. 1959 .
[67] S. Ahlen et al ., Part . World 1 (1990) 168; and
IEEE Trans. Nucl . Sci. NS-36 (1989) 567.
[68] IEEE Trans. Nucl . Sci. NS-37 (1990) 1564 .
[69] J. Handschuh et al ., IEEE Trans. Nucl . Sci. NS-20 (1973)
633; and refs. therein.
[70] Yu . Galaktionov et al ., Nucl. Instr. and Meth. A317
(1992) 116.
[71] A. Arefiev et al ., RD5 Technical note RD5-TN/92-02,
CERN, RD5 project ;
A. Arefiev et al ., RD5 Technical note RD5-TN/92-03,
L3P Collaboration / Precision experiment at LHC
90
[72]
[73]
[74]
[75]
[76]
[77]
CERN, RD5 project,
A Arefiev et al ., RD5 Technical note RD5-TN/92-04
CERN, RD5 project.
FLU Bengtsson and T. Slôstrand, Comput. Phys Commun. 46 (1987) 43 .
Z. Kunst, W.J. Stirling, ECFA-LHC workshop, Aachen,
CERN 90-10/ECFA 90-133, Vol. 11, p. 428.
A Djouadi, M. Spira and P .M . Zerwas, Phys . Lett B264
(1991) 440.
R. Brun et al ., CERN Computer News]. 200 (1990) 13 .
D. Froidevaux et al ., Higgs Physics, ECFA-LHC Workshop, Aachen, CERN 90-10/ECFA 90-133, Vol 11, p.
444.
E Reja et al ., ECFA-LHC workshop, Aachen, CERN
90-10/ECFA 90-133 . Vol. 11, p 296
Appendix
The L3P Collaboration consists of the following members :
T. Aziz, R. Bock, W. Braunschweig, E. Geulig, K.
Hilgcrs, H. Hillemanns, W. Karpinski, O. Kornadt, W.
Krenz, Th . Lehmann, B. Lindemann, K. Lübelsmeyer,
A. Nippe, D. Pandoulas, Y.J . Pei, M. Röhner, D.
Schmitz, M. Schoentag, A. Schulz, V. Dratzig, J.
Schwenke, G . Schwering, R. Siedling, K. Subhani, M.
Toporowsky, W. Wallraff, A. Weber, N. Xiao, Y. Zeng,
J.F . Zhou and S. Zitzen, L Physikalisches Institut,
RWTH, W-5100 Aachen, Germany
J.J . Blaising, G. Coignet, M. Lebeau and M. Schneegans, Laboratoire de Physique des Particules, LAPP,
F-74941 Annecy-le Vieux, France
C.Y . Chien, P.H . Fisher, T. Paul, A. Pevsner and C.
Spartiotis, Johns Hopkins University, Baltimore, MD
21218, USA
H. Li, W.M . Wang, B.T. Yu and S.M . Zhou, Beijing
Glass Institute, Beijing, China
J.Z. Bai, C. Chen, G.M . Chen, H.S . Chen, S.N . Chen,
G.L. Dai, C. Fang, J.T . He, X.G . Hu, B.N . Jin, H.B .
Lan, B.M . Li, H.T . Li, R.B . Li, H.M . Liu, Y.S . Lu, J.M .
Ma, Y.F . Mao, Y.H . Ou, J. Shen, X.W . Tang, K.L .
Tung, F. Wang, J.H . Wang, C.L . Wei, R.J . Wu, P.P .
Xie, J. Yan, C.G . Yang, K.S . Yang, Z.Q . Yu, G.R .
Zheng, G.Y . Zhu and B. Zhang, Institute of High
Energy Physics, IHEP, Beijing, China
W.Z . Chen, Z.M . Chen, Z.Y. Guo, Y.N . Liang, G.L .
Liu, D.H . Nan, J.S . Ou, C.Y . Yin and C.Z . Zhang,
Tsinghua University, Beijing, China
F. Block, D. Boscherini, G. Bruni, P. Bruni, G. Cara
Romeo, M. Chiarini, F. Cindolo, F. Ciralli, S. D'Auria,
C. Del Papa, F. Frasconi, P. Giusti, G. lacobucci, G.
Levi, G. Maccaronc, A. Margotti, T. Massam, R. Nania, F. Palmonari, G. Sartorelli, R. Timellini and S .
Tzamarias, INFN Sezione di Bologna and University of
Bologna, 1-40126 Bologna, Italy
T. Àziz, S. Banerjee, S.R . Chendvankar, P.V . Deshpande, S.N . Ganguli, S.K . Gupta, A. Gurtu, P.K . Malhotra t, K. Mazumdar, R. Raghavan, K. Shankar, K.
Sudhakar and S.C . Tonwar, Tata Institute of Fundamental Research, Bombay 400 005, India
S. Ahlen, A. Marin and B. Zhou, Boston University,
Boston, MA 02215, USA
T. Angelescu, F. Cotorobai, N. Gheordanescu, A. Mihul and D. Pop, University of Bucharest, Bucharest,
Romania
Gy .L. Bencze, T. Cs6rg6, E. Dénes, I. Hosvatti, P.
Levai, M . Priszuyak, J. T6th, L. Urbân and J. Zimanyi,
Central Research Institute for Physics of the Hungarian
Academy of Sciences, H-1525 Budapest 114, Hungary
K.S . Kumar, A. Kunin, I. Scott and K. Strauch, Harrard University, Cambridge, MA 02139, USA
A.L . Anderson, U. Becker, P. Berges, J.D . Burger, M.
Capell, Y.H . Chang, J. Chen, M. Chen, S. Chung, I.
Clare, R. Clare, T.S . Dai, F.J . Eppling, A. Klimentov,
V. Koutsenko, T. Kramcr, A. Lebedev, D. Luckey, A.
Rubbia, M.S . Sarakinos, S . Shotkin, B. Smith, M .
Steuer, S.C.C . Ting, S.M . Ting and B. Wyslouch, Massachusetts Institute of Technology, Cambridge, MA
02139, USA
V. Gantmaher, V. Grazjilis, N. Klassen, Yu . Ossipyan,
I. Schegelev and V . Timofeev, Institute of Solid State
Physics, Chernogolocka, Russian Federation
A. Chen and W.T . Lin, National Central University,
Chung-Li, Taiwan
T. Barillari, M. Schioppa and G. Susinno, INFNGruppo
Collegato di Cosenza and University della Calabria,
Cosenza, Italy
O. Adriani, A.M . Cartacci, G. Ciancaglini, C. Civinini,
R. D'Alessandro, E. Gallo, M. Meschini, V. Pojidaev,
P. Spillantini and Y.F . Wang, INFN Sezione di Firenze
and University of Florence, 1-50125 Florence, Italy
G. Anzivino, S. De Pasquale and L. Votano, INFN
Laboratort Nationaai di Frascati, Frascati, Italy
J. Alcaraz, F. Anselmo, R. Barillère, G. Brugnola, N.
Colino, A. Contra, R. De Salvo, M. Felcini, P. Ford, H.
Gerwig, A. Hervé, V. Innocente, G. La Commare, H.
Larsen, G. Laurenti, P. Lecoq, J.M . LeGoff, M. Marino,
C. Nemoz, M. Pieri, J. Salicio, C. Williams, F. Wittgenstein and A. Zichichi, European Laboratory for Particle
Physics, CERN, CH-1211 Genera 23, Switzerland
M. Bourquin, University of Geneva, CH-1211 Geneva 4,
Switzerland
R. Ayad, J .Z . Bai, B. Bencheick, X.D . Cai, U.K.
Chaturvedi, W.Y . Chen, X.T . Cui, X.Y . Cui, M.T .
Dova, C. Gu, A.H . Hasan, D. Hatzifotiadou, H.L . He,
G. Hu, R.A . Khan, M. Kaur, S. Khokhar, J. Lamas
Valverde, E. Leon Florian, Q. Lin, L.B . Liu, Y. Mi, Y.
Mir, N.E . Moulai, M.A . Niaz, S. Qian, K.N . Qureshi,
Deceased .
L3P Collaboration / Precision experiment at LHC
Z. Ren, H.A . Rizvi, R. Sehgal, Y.M . Shabelski, J.Y .
Sun, L.Z. Sun, A.A . Syed, P. Vikas, U. Vikas, M.
Wadhwa, K.L . Wang, Z.M . Wang, S.X . Wu, X.M . Xia,
C.Y . Yang, G. Yang, C.H . Ye, Q. Ye, Y. Ye, J.M . You,
N . Yunus, G.Y . Zamora, M. Zeng, Z.P . Zhang and M.
Zhao, World Laboratory, Lausanne, Switzerland
Z.S . Ding, M.Z. Ge, G.Y . Han, M. Jin, Y. Tang, R.
Wang, C.W . Yu and X.S . Zhang, Zhejiang University,
Hangzhou, China
Q. An, H.F . Chen, Q. Gao, Z.F . Gong, C.H . gu, Y.
Jiang, C. Li, Z.Y . Lin, Y.Y . Liu, W.Z . Lu, W.G . Ma,
L.Z. Sun, C.R . Wang, Y.F . Wang, Z.M . Wang, J. Wu,
Z .Z . Xu, B.Z. Yang, J.B . Ye, S.X . Ye, Z.P . Zhang and
M.H . Zhou, Chinese University of Science and Technology, USTC, Hefei, Anhui 230 029, China
S. Su, National Chiao Tung University, Hsinchu, Taiwan
Y.Y . Lee, W.T . Ni, Y.C. Yang, S.C . Yeh and H.C .
Yen, National Tsing Hua University, Hsinchu, Taiwan
Y. Takahashi, University of Alabama, Huntsville, USA
H .C. Chen, J.R . Han, M. He, M.H . Jiang, Q. Li, C.R .
Wang, N.J . Zhang and X.Y . Zhang, Shandong University, Jinan, China
Ph . Rosselet, University of Lausanne, CH-1051 Lausanne, Switzerland
P. Lebrun and J.P . Martin, Institut de Physique
Nucléaire de Lyon, IN2P3-CNRS / Université Claude
Bernard, F-69622 Villeurbanne Cedex, France
M. Aguilar-Benitez, P. Arce, J. Berdugo, C. Burgos, M.
Cerrada, D. Fernandez, G. Fernandez, P. Garcia-Abia,
E. Gonzalez, C. Mana, L. Martinez-Laso, F.J . Rodriguez, L. Romero, J .M . Salicio and C. Willmott,
Centro de Investigaciones Energeticas, Medioambientales
y Tecnologicas, CIEMAT, E-28040 Madrid, Spain
A. Baschirotto, M. Bosetti, R. Castello, S . Pensotti,
P.G . Raneoita, M. Rattaggi and G. Terzi, INFN Sezione
di Milano and University of Milan, 1-20133 Milan, Italy
D .Yu. Àkimov, A. Arefiev, A.I . Bolozdynya, D.L. Churakov, Yu . Galaktionov, A. Klimentov, Yu . Kolotaev,
V. Koutsenko, A. Kunin, A. Malinin, V. Plyaskin, V.
Pojidaev, A. Rozjkov, E. Shumilov, V. Shoutko, I.
Vetlitsky and I. Vorobiev, Institute of Theoretical and
Experimental Physics, ITEP, Moscow, Russian Federation
L.A . Bragin, N. Chernoplekov, 1 . Karpushov, E. Klimenko, S. Lelekhov, A. Malofeev, V. Mokhnatuk, S .
Novikov and E. Velikhov, Russia Scientific Center
"Kourchatov Institute", Moscow, Russian Federation
A. Aloisio, M.G . Alviggi, E. Brambilla, G. Carlino, R.
de Asmundis, S. Lanzano, L. Lista, P. Paolucci, S.
Patricelli, D. Piccolo and C. Sciacca, INFN Sezione di
Napoli and University of Naples, 1-80125 Naples, Italy
P. Razis, Department of Natural Sciences, University of
Cyprus, Nicosia, Cyprus
J. Seguinot and T. Ypsilantis, Collige de France, Paris,
France
G . Gratta, M. Gruenewald, D. Kirkby, R. Mount, H.
91
Newman, X.R . Shi, C. Tully, C. Zaccardelli and R.Y .
Zhu, California Institute of Technology, Pasadena, CA
91125, USA
R. Battiston, G.M . Bilei, M . Caria, B. Checcuci, S.
Easo, V. Krastev, M. Pauluzzi, L. Servoli and S. Wang,
INFN Sezione di Perugia and Università Degli Studi di
Perugia, 1-06100 Perugia, Italy
L. Cifarelli, University of Pisa, Pisa, Italy
P. Denes, V. Gupta, P .A . Piroué, H. Stone, D.P . Stickland and D. Wright, Princeton University, Princeton, NJ
08544, USA
L. Barone, B. Borgia, F. Cesaroni, F. DeNotaristefani,
M. Diemoz, C. Dionisi, S. Falciano, E. Leonardi, E.
Longo, C. Luci, L. Luminari, G . Mirabelli, G. Organtini, M. Rescigno and E. Valente, INFN Sezione di
Roma and University of Rome, "La Sapienza", I-00185
Rome, Italy
V. Andreev, G. Alkahazov, A. Bykov, P. Kapinos, V.
Kim, A. Tsaregorodtsev, A. Vorobiev and Yu . Zalite,
Nuclear Physics Institute, St . Petersburg, Russian Federation
H. Shen and W.M . Wu, Fudan University, Shanghai,
China
H. Pan, C. Qian, C. Si, S. Xie, B. Xu, W. Yang, Q. Ye
and S. Zhang, Jiaotong University, Shanghai, China
X.L . Fang, C.D . Feng, X.Q . Feng, H.X . Gao, M. Gao,
P.X . Gu, J .K . Guo, G.Q . Hu, Y.L. Hu, S.K . Hua, P.J .
Li, Z.D . Qi, D.Z . Shen, E.W . Shi, W.T . Su, X.X .
Wang, Z.Y . Wei, Y.Y . Xie, L. Xu, Z.L . Xue, D.S . Yan,
Z.W . Yin, X.L . Yuan, Y.F . Zhang, G.M . Zhao, Y.L .
Zhao, W.Z . Zhong, R.M . Zhou, Shanghai Institute of
Ceramics, SIC, Shanghai, China
A.H . Walenta, University of Siegen, Siegers, Germany
N . Shivarov, Bulgarian Academy of Sciences, Institute of
Mechatronics, BU-1113 Sofia, Bulgaria
M.T. Choi, J.K. Kim, Y.G . Kim, S.C. Kim and D. Son,
Center for High Energy Physics, Taejon, Korea
L.S . Hsu and W.L . Lin, National Normal University,
Taipei, Taiwan
D. DiBitonto, T. Pennington and K. Subhani, University of Alabama, Tuscaloosa, AL 35486, USA
A. Bujak, D.D . Carmony, L.J . Gutay, T. McMahon and
B.C . Springfellow, Purdue University, West Lafayette,
IN 47907, USA
H. Anderhub, F. Behner, J. Behrens, B . Betev, A.
Biland, M . Dhina, G. Faber, K. Freudenreich, M .
Hänsli, H. Hofer, 1. Horvath, M. Jongmanns, P.
Lecomte, P. LeCoultre, M. MacDermott, M. Maolinbay, P. Marchesini, D. McNally, F. Nessi-Tedaldi, C.
Neyer, J . Paradiso, F . Pauss, M. Pohl, G. Rahal-Callot,
D. Ren, N. Scholz, U. R6ser, H. Rykaczewski, H.
Suter, J . Ulbricht, G. Viertel, H.P . Von Gunten, S.
Waldmeier, J. Weber and P. Zemp, Eidgenössische
Technische Hochschule, ETH Ziirich, CH-8093 Ziirich,
Switzerland