464 Nuclear Instruments and Methods in Physics Research A237 (1985) 464 474
North- Holland. Amsterdam
THE ARGUS
ELECTRON-PHOTON CALORIMETER
III. Electron-hadron separation
A. DRESCHER, B. GRAWE, B. HAHN, B. INGELBACH, U. MATTHIESEN, H. SCHECK,
J. SPENGLER and D. WEGENER
lnstitut )~ur PI~vsik, Universit~t Dortmund, Fed. Rep. Germany
Received 2 January 1985
The high granularity of the ARGUS shower counter system allows one to determine the lateral energy deposition. An algorithm
based on the number of shower counters set by an incoming particle, the amount of energy deposited in the shower counters and the
lateral distribution of the energy allows to separate electrons from muons and hadrons. For electrons with momenta larger than 0.7
GeV/c the hadron rejection rate is of the order of 50 at an electron detection efficiency of 80%.
l. Introduction
The ARGUS spectrometer (fig. 1) is a detector for
e + e reactions at DORIS II [1]. It consists of a central
drift chamber [2], which measures the momentum of
charged particles and the rate of their energy loss, a
time of flight system [3] and an electromagnetic shower
detector [4]. These components are all inside the coil of
the magnet, which produces a nearly homogeneous field
of 0.8 T. Muons are detected in two planes by propor-
tional tubes [5], positioned behind absorbers, with cut
off momenta of 0.7 GeV/c and 1.1 GeV/c respectively.
Details of the performance of the shower counters in
a test beam and under experimental conditions in the
ARGUS detector have been previously described [4]. In
this paper we will concentrate on the question of how
electrons (used as a synonym for electrons and positrons)
can be separated from other charged particles hitting
the shower counters. A clean electron sample, identified
with high efficiency and with a small hadron admixture,
is of interest for the physics program performed with
the ARGUS detector, namely the study of semileptonic
B decays. These can be used to determine the branching
ratios BR(b ~ c) and BR(b ~ u), as well as to search for
B°-B ° mixing. For these physics goals, the electron
momentum range of main interest is 1.2 GeV/c < Pe <
2.5 GeV/c.
Contrary to most other experiments [6,7] the sep-
aration of electrons and hadrons has to be based en-
tirely on the differences of the lateral energy deposited
by particles hitting the shower counters, because there is
only one longitudinal sample. Since the lateral width of
the shower counters, r = 10 cm, is large compared with
the radius of a typical electromagnetic shower (Moli6re
radius R M < 5 cm) but is small in comparison with the
0168-9002/85/$03.30 © Elsevier Science Publishers B.V.
(North-Holland Physics Publishing Division)
radius of a hadronic shower (absorption length
•abs •
46
cm), the two kinds of showers can be distinguished. On
the other hand, minimum ionizing particles can be
separated from showering electrons, simply on the basis
of the energy deposited.
The paper is organized as follows: in section 2 we
briefly describe those properties of the ARGUS electron
photon calorimeter, which are of importance for the
present analysis. Further details have been given in
previous publications [4]. In section 3 the pattern of
energy deposition for the different particle species is
discussed.
In
section 4 we describe the algorithm which
exploits the results of section 3 with the aim of electron
hadron separation at different momenta.
2. ARGUS electromagnetic calorimeter
The ARGUS shower counters are of the lead scintil-
lator sandwich type with 5 mm thick scintillator plates
(Altustipe UV). The thickness of the lead plates is 1 mm
in the barrel, and 1.5 mm in the endcap region. The
overall length of each counter corresponds to 12.4 radia-
tion lengths for electrons and 0.8 absorption lengths for
pions in the momentum range of interest. The lead and
scintillator plates are separated by aluminized mylar
foils with 98% reflectivity. The lead scintillator sand-
wich is "read out" by a 3 mm thick wave length shifter
coupled to a 1 inch photo tube (Valvo XP 2008 UB).
The barrel counters consist of 64 plates of each material
with 10.9 cm × 10.3 cm lateral cross section. The end-
cap counters contain 45 plates of each material of
trapezoidal shape with maximum lateral dimensions of
11.6 cm and 10.4 cm.
The shower counters are mounted inside of the mag-
A. Drescher et aL/ Electron-hadron separation in ARGUS
465
!
Fig. 1. Schematic view of the ARGUS detector: (1) drift chamber, (2) time of flight counters, (3) barrel shower counters, (4) end cap
shower counters, (5) magnetic coil, (6) muon chambers, (7) iron yoke.
netic coil. The cylindrical arrangement of the barrel
counters (fig. 1) results in a variation of the angle of
incidence between 0 ° and 45 ° for particles coming from
the interaction point. The endcap counters are arranged
in a plane perpendicular to the primary beam (fig. 1),
the angle of incidence for particles hitting this detector
component varies between 20 ° and 45 ° . Because of the
cylindrical symmetry of the detector counters in the
same polar angle region can be treated in common.
Counters at 0 = 90 ° are labelled with the ring number
1, with decreasing polar angle the ring number increases
to 15 for the counters in the endcap at the smallest
polar angles.
The energy resolution presently obtained under ex-
perimental conditions is
° -- .~ /( 0"08 )2 + (0.068)2
E V\~
where E is measured in GeV.
The angular resolution is 0.7 ° for electrons with
momenta Pe > 1 GeV/c. Further details concerning the
construction and properties of the shower counters are
described in refs. [4b-4d].
3. Pattern of energy deposition for different particle
species
The following information about the energy de-
posited by a particle in the ARGUS shower counters is
466 A. Drescher et al. / Electron-hadron separation in ARGUS
available: the total visible energy deposited, E, the
number of neighbouring counters hit (cluster size, n)
and the visible energy E i detected in each counter of a
cluster:
n
E=y'E,, E I>E 2> ... >E,,.
1
For electrons, muons and interacting and noninter-
acting hadrons, these measurable quantities differ in
their mean and their distribution, therefore they can be
used to separate different kinds of particles. Moreover
the values obtained depend on the momentum of the
particle. Since the ARGUS shower detector has cylin-
drical symmetry, the angle of incidence for a particle
varies with its polar angle with respect to the beam axis
(fig. 1), hence E, E i and n show in addition a depen-
dence on the polar angle 0.
These relationships have been determined for the
different particle species using a data set collected with
the ARGUS detector. A clean sample of electrons at
different energies is available from the reactions
e+e-~ e+e .
e+e ~ e+e -/.
The electron momentum was measured in the drift
chamber. A sample of muons of varying momenta was
available from the processes
e+e - ~ #+/, ,
e+e --,#+tz y.
The muons have been identified with the muon
chambers [5]. The sample of hadron tracks was taken
from the data collected on the T-resonance, since the
admixture of leptons for this data set is smallest. To
suppress leptonic decay channels, events were used with
more than ten neutral and charged particles. A small
admixture of leptons due to weak decays of the hadrons
is unavoidable for this data set, therefore the values for
the separation power given in this paper are lower
limits. For hadron momenta smaller than 0.7 GeV/c, ~r-
and K-mesons can be separated with the time of flight
and the dE/dx systems, while protons and antiprotons
are identified up to momenta of 1.2 GeV/c in the same
manner.
3.1. Energy deposited by electrons in the shower counters
Since electrons deposit their total energy in the
calorimeter, the electron momentum, as measured in the
drift chamber, is a linear function of the energy detected
by the shower counters (fig. 2). The width of the distri-
bution is determined by the resolution of the shower
counters o E and the momentum resolution of the drift
chamber, which has been measured to be
op/p = 0.012 p, (p > 1 GeV/c).
6 ~---,- i :.,
===========================
--!
i ::::::::::::::::::::::::::::::::::::
• :::::::::::::::::::::::::::::::::::::::
v
3 :::::::::::::::::::::::::::::::
2
: : :i:!!!:i : i : : . : :
1 :iiii:i ~i:
o
1
2 3
4
5
6
P (GeV/c)
Fig. 2. Energy E deposited in the shower counters vs momen-
tum for electrons.
Adding these two contributions to the resolution
quadratically
02=02+02 '
a
cut
]E-pcI<No, N=3,4 ....
can be applied.
The dependence of the electron detection efficmncy
as a function of N is shown in fig. 3. For N = 3 one
obtains a detection efficiency of 88%. This is smaller
than the expected value for a normal distribution be-
cause the two distribution functions are only approxi-
mately Gaussian. The detection efficiency for a given N
is approximately independent of momentum [8].
Since the electromagnetic shower has a finite spatial
width and the particles in general traverse more than
one shower counter, the energy deposited is spread over
a cluster of neighbouring shower counters. In the analy-
sis, only counters with an energy deposit larger than 10
I00
v 80
" 60
~a 40
20
~1 D o
ID
Fig. 3. Percentage of electrons passing the cut ]E - Pl < No as
a function of N.
0 I , , ~ I
0 8 4 6 8
N
A. Drescher et al. / Electron-hadron separation in ARGUS
467
MeV are taken into account to suppress background.
For a given type of particle the mean number of shower
counters hit (cluster size n) depends on the particle
momentum and the polar angle of the incoming particle
.=n(p,O)
A typical distribution of cluster size for the electron
momentum interval of interest is shown in fig. 4a. The
distribution has a large width due to the averaging over
many impact directions and the inherent shower
fluctuations. The mean cluster size increases linearly
IZ0
80
40
13
I0
8
6
4
2
0
14
12
l0
8
6
4
2
0
.... l .... i .... ; ....
a
5 l0 15 20
Cluster size
b
3
4 5
L
2
P (GeV/e)
i i i • i I
C
III
i i i L
-, i [
R 4 6 8 I0 1:~ 14
Ring number
Fig. 4. (a) Cluster size distribution for electrons (I
GeV/c <_ p
<_ 2 GeV/c)
hitting the shower counters. (b) Mean value of the
cluster size vs electron momentum, The error bars indicate the
width of the cluster size distribution. (c) Mean cluster size as a
function of the ring number, which is a measure of the polar
angle of the particle.
400
~ BOO
200
e,
~
IO0
.... i .... i .... i .... i .... i .... i ....
a
1 2 3 4 5 6 7
dE/dx (MeV/cm)
o
5
r~
100
.~
4
~ 3
8
0.0
.... i .... i .... i .... i ....
b
o.1
o.a o 3
0.4 05
E (GeV)
i -T--T i T , r
C
0 t i , i i i i t
0 2 4 6 8 10 12 14
Ring number
Fig. 5. (a) Energy loss of muons per unit track length in the
shower counters. (b) Total energy loss of muons in the shower
counters. (c) Cluster size and width of muons, hitting the
shower counters as a function of the ring number.
468 A. Dreseher et aL / Electron hadron separation in ARGUS
with the electron momentum, while its width is ap-
proximately constant (fig. 4b). The dependence of the
cluster size on the polar angle is shown in fig. 4c for
electrons with high momentum. The cluster size in-
creases weakly with decreasing polar angle (fig. 4c). The
decrease observed at polar angles of 0---45 ° (ring 10,
11) is due to the gap between the barrel and the end cap
counters (fig. 1).
less than 500 MeV (fig. 5b). The cluster size as a
function of the polar angle has a shape (fig. 5c) similar
to the one for electrons, but the mean number of
counters hit is smaller than 4 and therefore differs
appreciably from the cluster size for electrons (fig. 4c).
From these observations it follows that a cut on the
visible energy and on the cluster size is a powerful
means of distinguishing between muons and electrons.
3.2. Energy deposition by muons in the shower counter
3.3. Energy' deposition of hadrons in the shower counters
Muons lose energy in the shower counters only by
ionization and excitation processes. Normalizing the
visible energy in the shower counters to the track length
in the scintillator, one gets (fig. 5a) a mean energy loss
per unit length of
(d
E/d
x > = 2.4 MeV/cm,
in good agreement with the expected value. The total
energy deposited by muons in the shower counters is
The energy deposited by hadrons in the shower
counters is due to two processes. Noninteracting hadrons
lose their energy by ionization and excitation. Hadrons
that interact strongly initiate a hadron shower, whose
energy is partially deposited in the shower counters.
These two components can be clearly separated for
particle momenta p > 2 GeV/c (fig. 6a). The visible
energy in the shower counters has a pronounced peak at
the position expected for minimum ionizing particles
150
>,
100
o
50
e~
.... r .... r .... , .... , ....
a
00
0.5 I 0 1.5 2.0 2.5
E
(GeV)
o
c~
150
b
,oo i
0 L , J , L , I , ,
0
1
2 3 4 5 6 7
dE/dx (MeV/cm)
Fig. 6. (a) Energy deposited by hadrons in the shower counters.
(b) Energy per unit track length deposited in the shower
counters.
850
200
150
100
50
.... 1 .... i .... , ....
8
IZ .............
5 10 15 20
Cluster size
250
200
150
100
.... i .... i .... i ....
b
0
0 5 10 15 80
Cluster size
Fig. 7. (a) Clustcr size distribution of noninteracting hadrons in
the shower countcrs. (b) Cluster size distribution of interacting
hadrons in the shower counters.