ARTICLE IN PRESS

Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

The ATHENA antihydrogen apparatus

M. Amorettia, C. Amslerb, G. Bonomic, A. Bouchtac, P.D. Bowed, C. Carraroa,
M. Charltone, M.J.T. Colliere, M. Doserc, V. Filippinif, K.S. Finec, A. Fontanaf,
b
.
M.C. Fujiwarag, R. Funakoshig, P. Genovaf, A. Glauserb, D. Grogler
, J. Hangstd,
R.S. Hayanog, H. Higakig, M.H. Holzscheiterc,1, W. Joffraina, L.V. J^rgensene,*,
. b, E. Lodi-Rizzinii,
V. Lagomarsinoa, R. Landuac, C. Lenz Cesarh, D. Lindelof
M. Macria, N. Madsenb, D. Manuzioa, G. Manuzioa, M. Marchesottic,
P. Montagnaf, H. Pruysb, C. Regenfusb, P. Riedlerc, J. Rochetc, A. Rotondif,
G. Rouleauc,2, G. Testeraa, D. P. van der Werfe, A. Variolaa, T.L. Watsone,
T. Yamazakig, Y. Yamazakig
a
Genoa University & INFN, Genoa, Italy
Institute of Physics, Zurich University, Zurich, Switzerland
c
CERN, Geneva, Switzerland
d
Department of Physics & Astronomy, University of Aarhus, Aarhus, Denmark
e
Department of Physics, University of Wales Swansea, Wales, UK
f
Pavia University & INFN, Pavia, Italy
g
Department of Physics and Institute of Physics, Tokyo University, Tokyo, Japan
h
Fed. Univ. Rio de Janeiro (UFRJ), Rio de Janeiro, Brazil
i
Brescia University & INFN, Brescia, Italy
b

The ATHENA Collaboration

Accepted 16 September 2003

Abstract
The ATHENA apparatus that recently produced and detected the rst cold antihydrogen atoms is described. Its main
features, which are described herein, are: an external positron accumulator, making it possible to accumulate large
numbers of positrons; a separate antiproton catching trap, optimizing the catching, cooling and handling of antiprotons; a
unique high resolution antihydrogen annihilation detector, allowing an clear determination that antihydrogen has been
produced; an open, modular design making variations in the experimental approach possible and a nested Penning trap
situated in a cryogenic, 3T magnetic eld environment used for the mixing of the antiprotons and positrons.
r 2003 Elsevier B.V. All rights reserved.
Keywords: Antihydrogen; Penning trap; Cryogenic detector; Positrons; Antiprotons

*Corresponding author. Tel.: +44-22-767-4856; fax: +41-79-201-3214.

E-mail address: [email protected] (L.V. J^rgensen).
URL: http://athena.web.cern.ch/athena/
1
Visitor from Los Alamos National Laboratory, USA.
2
Now at Los Alamos National Laboratory, USA.
0168-9002/$ - see front matter r 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.nima.2003.09.052

ARTICLE IN PRESS
680

M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

1. Introduction
The ATHENA collaboration recently succeeded
% by
in making the rst cold antihydrogen H
mixing antiprotons with a dense cold positron
plasma [1]. This was an important milestone on the
way to the main goal of the ATHENA experiment,
which is a spectroscopic comparison of the
properties of antihydrogen and hydrogen atoms.
This will allow direct tests of CPT invariance for
leptons and baryons with unprecedented accuracy.
For example, the long lifetime of the metastable 2s
state (122 ms) leads to a natural line width of
about 1 Hz for the 1s2s transition, offering, in
principle, the possibility to compare anti-atoms
with their matter counterparts with high precision,
perhaps reaching 1 part in 1018 : In fact, an
accuracy close to 1014 should be achievable for
antihydrogen if the 2 photon 1s to 2s transition
can be measured with a similar precision to that
achieved for hydrogen in 2000 [2]. Such precision
is several orders of magnitudes better than the
present best direct CPT tests on leptons and
baryons, and comparable to the indirect tests in
the neutral kaon system [3].
The rst observation of atomic antimatter
was made in 1996 at CERN when the formation of antiatoms was reported at LEAR (Low
Energy Antiproton Ring). The internal antiproton
beam with a kinetic energy in the GeV range
traversed a Xe gas jet target and produced e e
pairs. In a very small fraction of these collisions a
positron was captured by the antiprotons resulting
in nine events attributed to antihydrogen [4].
Similar observations were made in a subsequent
experiment at Fermilab [5]. These antihydrogen
atoms were all created at kinetic energies in the 1
6 GeV range corresponding to equilibrium temperatures in the 1013 K range. However, antihydrogen needs to be produced in much larger
quantities and at much lower kinetic energies (i.e.
the so-called cold antihydrogen) to facilitate the
aforementioned precision spectroscopic measurements of its properties. Indeed the highest precision is expected for measurements on the
antihydrogen atom held at rest in a neutral atom
magnetic trap. The recent results from ATHENA
have brought spectroscopy a large step closer by

making copious amounts of cold antihydrogen

available for experiments [1].
In the context of cold antihydrogen formation
two mechanisms are of importance and their
parameter dependence determines the boundary
conditions for the design of any experiment to
produce and manipulate cold antihydrogen. The
spontaneous radiative recombination of positrons
% hn
with antiprotons e p% H
pdepends on
the positron temperature p1= T and the
positron density pn and results predominantly
in atomic levels with low principal quantum
numbers. Three-body recombination e e
p% H% e becomes competitive at cryogenic
temperatures p1=T 4:5 and high positron densities pn2 and mainly populates atomic levels with
high principal quantum numbers. De-excitation to
the ground state is slow and the nascent atom is
susceptible to ionization by electric elds.
In the present paper we describe in detail the
equipment used by ATHENA to produce and
detect cold antihydrogen atoms. A schematic
illustration of the apparatus is shown in Fig. 1.
Low energy antiprotons are extracted from the
CERN Antiproton Decelerator (AD). A superconducting solenoid (3 T) with a cold bore houses
the antiproton capture trap and the antiproton
positron mixing trap. The antiprotons are moderated through a silicon beam-proling counter, a
foil and various windows and are then reected in
the catching trap by a high voltage electrode.
About 500700 ns after the arrival of the pulse a
high voltage potential is raised on the entrance
electrode to capture the antiproton bunch. They
are then cooled inside the catching trap to meV
energies by Coulomb interactions with a preloaded electron cloud. The antiprotons are conned in the radial direction by the magnetic eld of
the solenoid and in the axial direction by an
electrostatic eld produced by 10 cylindrical
electrodes having an inner diameter of 2.5 cm.
The eld conguration is similar to the one used in
Penning traps [6]. Further details can be found in
Sections 2, 4 and 5.
The positrons from a 1:4 GBq 40 mCi 22 Na
source are moderated in solid neon and transferred
into a longitudinal magnetic eld region where
they lose energy by collisions with nitrogen gas

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

681

Fig. 1. Overview of the ATHENA apparatus for the production and detection of antihydrogen. An expanded view of the annihilation
detector is shown below the main apparatus.

and are eventually conned in a system of

cylindrical electrodes, as described in Section 7.
They are then transferred to the superconducting
solenoid where they are held in a trap similar to
the one used to store the antiprotons.
The particles stacked in the two traps are
transported to the mixing trap held at a pressure
below 1012 mbar: The formation of antihydrogen
atoms is studied by observing their annihilation
when they impinge upon the electrodes of the
mixing trap. The annihilation detector, a large
solid angle array of silicon microstrip counters and
CsI crystals, surrounds the mixing trap. It
measures the charged hadron tracks (mostly pions)
emitted by the annihilation of the antiproton, in
temporal and spatial coincidence with the two
back-to-back 511 keV photons from e e annihilation. The detector is described in more detail in
Section 8.
In Section 9 we describe the control and data
acquisition system and in Section 10 the online
software for time-ordering and rapid real time
analysis of the data as it arrives is presented.

Finally in Section 11 the ofine analysis software,

Monte Carlo simulation of the apparatus and
event reconstruction and selection schemes are
described.

2. General experimental parameters

In the design of the ATHENA apparatus
emphasis was placed on an open and exible
system. This has allowed for the accumulation of
large numbers of positrons in an external apparatus with subsequent transfer into the main
apparatus as well as limited freedom in the
approach to antihydrogen formation and the types
of experiments that could be performed. Due to
the temperature dependence of both of the
formation mechanisms mentioned in Section 1 it
is crucial to maintain the lowest possible temperature in the trapping region. In addition to this the
two other main experimental conditions necessary
for making antihydrogen are a very low residual
gas pressure and a high magnetic eld.

ARTICLE IN PRESS
682

M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

2.1. The superconducting magnet and the cryogenic

cold nose
The magnet used for the ATHENA experiment
was taken over from the original PS200T system
[7] and consists of a superconducting solenoid with
a 150 mm diameter room temperature bore and a
homogeneous eld region of 1 m length. The
magnet is capable of reaching 5 T but for the
ATHENA experiments was generally operated at
3 T: The requirements of the ATHENA experiment were signicantly more complex than the
antiproton trapping apparatus developed for
PS200T such that a number of modications to
the original system were necessary.
To allow easy access to the trap without
disturbing the cryogenic system of the magnet, a
separate continuous ow cryostat (cold nose) was
installed in the bore of the magnet to cool the
vacuum shell housing the trap. The cold nose can
be cooled to a few K, improving the cryogenic
pumping speed at the location of the traps. In
addition, since the electrons and positrons cool
very efciently by synchrotron radiation in the 3 T
eld, a low ambient temperature allows production of very cold positron and electron plasmas
and, via sympathetic cooling, also very cold
antiprotons. The cold nose is closed off at the
end toward the AD beamline by the degrader foils
and the silicon beamcounter. At the end toward
the positron accumulator it has an open connection to a room temperature vacuum chamber.
During normal operation the outside of the cold
nose needs to be shielded from room temperature
radiation by lowering the bore temperature of the
main magnet. To allow radial space for mounting
the trap, the trap vacuum system, the cold nose,
and nally the detector, a common nitrogen shield
for both the main magnet system and the cold nose
was designed. Under normal operation a temperature of 130 K was achieved on the bore walls.
2.2. The vacuum system
Electromagnetic traps of the Penning and
PenningMalmberg types are used to store, cool
and manipulate the charged particles required by
ATHENA (positrons, antiprotons and the elec-

trons necessary to cool the antiprotons). These

traps (with the exception of the positron accumulator; see Section 7) are all situated within the
vacuum vessel of the cryogenic cold nose inside the
superconducting solenoid. The trap environmental
temperature is representative of the equilibrium
temperature of the antiproton and electron clouds
at the end of the cooling process. In addition,
cooling the trap region reduces the surface outgassing and transforms all the surfaces of the
apparatus into a cryogenic pump. The gas ow
from the room temperature region is reduced by
the pumping system installed in that region and by
the trap and cold nose surfaces. Most of the walls
in the room temperature section have been coated
with a NEG (Non-Evaporable Getter) material.
After an initial pumpout this material is activated
by raising the temperature of the walls to 200 C
for 24 h: This effectively turns the coated surfaces
into vacuum pumps. In addition to this a special
vacuum chamber lled with NEG coated metal
strips was installed. The NEG coated surfaces
together with a 300 l s1 ionpump constituted the
ultra high vacuum pumping system in the room
temperature region. The pressure can only be
directly measured in the room temperature region,
where 1011 mbar is routinely achieved. The
maximum temperature of the trap is approximately 15 K when the complete apparatus is
installed inside the cold nose, though the walls of
the cold nose itself reaches temperatures in the
range 410 K:

3. Antiproton beam monitoring

Pulses of antiprotons are ejected from the AD
with a momentum of about 100 MeV=c (or a
kinetic energy of around 5:3 MeV) [8]. The p% shots
last for 200 ns and are repeated every 100 s:
Typically, 2  107 antiprotons are injected into
the ATHENA apparatus. The locations of the
beam detectors are sketched in Fig. 2. The
antiproton kinetic energy is degraded by 25 mm
of stainless steel, located 10 cm upstream of the p%
catching trap, and by 130 mm of aluminum on the
entrance electrode.

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

683

VERTICAL EXTERNAL
ANNIHILATION DETECTOR
BARREL
BEAM DETECTOR
FRONT
BEAM DETECTOR

Magnet
coil

Degrader
_

e+

Si beam counter
_

mix. trap e+ res.+trans.

Magnet
cryostat
HORIZONTAL EXTERNAL
ANNIHILATION DETECTORS

Fig. 2. Top view of the p% beam line (not to scale). The dashed rectangles show the locations of the four External Annihilation
Detectors below the apparatus.

The antiproton beam is monitored by three

systems of detectors: (i) a silicon counter provides
the p% trigger for the catching trap high voltage
switch and measures the beam prole; (ii) the
external beam detector, which comprises the front
and barrel plastic scintillators, detects the annihilation in a pulse mode and monitors the beam
intensity and stability; (iii) the horizontal and
vertical external annihilation detectors, consisting
also of plastic scintillators, are used to study the
trapped antiprotons by detecting in a single
counting mode the pions following annihilation
either on the residual gas or on the trap walls.
3.1. The silicon beam counter
The silicon beam counter [9] is a 67 mm thick
silicon diode, 15 mm in diameter, that was
originally designed for the Crystal Barrel experiment [10]. The diode (Fig. 3) is segmented into ve
pads, each connected to an individual signal line.
The voltage required to fully deplete the diode is
4:5 V: The beamcounter is located in front of the
antiproton catching trap (see Section 4.1) and is
operated at temperatures ranging from 10300 K;
in a vacuum of E108 mbar and in a 3 T magnetic
eld. A very thin counter has to be used to allow
maximum transmission of the low energy antiprotons into the catching trap. The average energy
loss of 5:3 MeV antiprotons in silicon was

Fig. 3. Photograph of the 67 mm thick silicon counter mounted

in a PCB frame. The contacts to the individual pads are made
using ultrasonic wire-bonds.

estimated from experimental data [11] to be about

11:4 keV per micrometer of silicon, thus creating
3200 electronhole (eh) pairs, compared to 80 eh
pairs for minimum ionizing particles. Thus,
around 3:2  1010 charge pairs mm of silicon are
generated in the ATHENA silicon beamcounter
for an antiproton beam intensity of 107 p% per spill.
In order to detect this high instantaneous current,
a readout system was developed where the signal
current is read directly across a 100 O protection
resistor and fed into a digital oscilloscope and an
ADC. This also allows a direct measurement of the

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

684

to achieve better timing performances and to

reduce signal losses due to recombination of
charge carriers.
3.2. The external beam detectors

Integrated signal [x10 -6 Vs]

Fig. 4. Signal on the central pad of the silicon beam counter

registered on an oscilloscope. The spill duration of around 200
ns is clearly visible.

Vf d

0
0

20

40

60

80

100

Bias Voltage [V]

Fig. 5. Integrated signal registered on the central pad of the

silicon beamcounter as a function of the bias voltage of the
diode. Vfd indicates the voltage that is required to fully deplete
the diode.

spill duration. The signal from the silicon beam

counter is also used to trigger the p-catching-trap.
%
A typical signal registered on an oscilloscope is
shown in Fig. 4.
Fig. 5 shows the fraction of the total signal
measured on the central pad for different bias
voltages. For each voltage setting the mean
integrated signal of three antiproton spills is
shown. The error bars indicate the standard
deviation. The signal starts to plateau at about
30 V but an operating voltage of 100 V was chosen

The external beam detectors monitor annihilation on the degraders (Fig. 2). They consist of two
types of modules, Front and Barrel, both made
from 1 cm thick Bicron BC408 plastic scintillators
[12]. The two Front detectors are 195  100 mm2
each and the two Barrel detectors are 800 
195 mm2 each. The modules of the Front and
Barrel detectors cover about 0.1% and 3% of 4p;
respectively, when antiprotons annihilate on the
degraders. Due to the high instantaneous rate
C1014 s1 during the p% pulse, these detectors do
not count single particles, but operate in current
mode, measuring the total charge deposited by the
annihilation products. However, the light produced in a 1 cm thick plastic scintillator would
saturate the photomultiplier. We therefore used
proximity focused Hybrid Photo Diodes (HPD)
[1315]. These comprise a vacuum tube in which
the photoelectrons are accelerated toward a silicon
diode by a high voltage applied to the photocathode. A gain of a few thousand can easily be
achieved, low compared to photomultipliers
C107 but sufcient for our application, in which
a large amount of light is generated. We used two
different models of HPDs built by DEP [16]
namely model PP0350F for the Front and model
PP0350D the for the Barrel detector. A detailed
description of their characteristics can be found
elsewhere [12].
In order to test the effect of magnetic eld on
HPD gain, we measured the pulse height produced
by the HPDs in the ATHENA magnetic eld using
a pulsed light emitting diode (LED), held in the
center of the photocathode. Measurements with
and without magnetic eld were consistent within
3%, the difference being mainly due to LED
instabilities.
3.3. Absolute beam calibration
An absolute calibration was performed by
antiproton activation of an aluminum foil [17].

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

Antiprotons impinging on aluminium produce

24
Na (with a yield of 2:170:3%) which then
decays by g emission with energies of 1369 and
2754 keV [18]. The half life of 24 Na is 15 h: We
dumped the antiproton beam in an aluminum
target (made of four foils with a thickness of
110 mm; glued together) located at the entrance of
the ATHENA beam pipe. Using the induced
activity, measured off-line with a calibrated
germanium detector, an average beam intensity
per shot of 1:2070:19  107 p% was obtained [12].
As an independent measure of beam intensity, the
number of antiprotons in the AD ring was
measured before extraction by a low noise
Schottky pickup probe [19,20].
During the antiproton irradiation of the aluminum foil the annihilation signals were detected by
the Front and Barrel external beam detectors and
recorded with ADCs. Using the afore mentioned
knowledge of beam intensity the Front and Barrel
external beam detectors were calibrated as reported in Table 1. Here the measured ADC counts
are converted to collected charge. While the
Schottky probe gives the number of antiprotons
inside the AD ring with high accuracy C1%; it is
measured before electron cooling at 100 MeV=c
and before extraction. Hence, possible cooling
inefciencies and losses in the ejection line might
not be taken into account. However, the results
shown in Table 1 are in excellent agreement with
the activation method.
Fig. 6 shows the correlation between the charge
measured with the silicon beam counter and the
number of antiprotons determined with the
external beam detectors. The former begins to
lose linearity, due to saturation, starting at a ux
of C1  107 p=pulse:
%

Table 1
Calibration factors of the external beam detectors measured by
activation and with the Schottky probe (in fC=p),
% showing a
good agreement between the activation and Schottky methods
Calibration method

Front

Barrel

Activation
Schottky

57:579:5
55:572:1

27:274:5
26:171:0

685

Fig. 6. Correlation between the number of antiprotons per AD

pulse measured with the External Beam Detector and the
charge collected by the silicon beam counter.

3.4. The external annihilation detectors

The solid angle coverage of the external
annihilation detectors is sufciently small (see
below) to allow them to be used in single
count mode, unlike the beam counters. This
means they can be used to detect single
p% annihilations on the residual gas of the traps
or on the trap walls. The external annihilation
detectors consist of six coincidence pairs of
Bicron BC408 plastic scintillators. Each scintillator is 10 mm thick, wrapped with aluminized
mylar sheet, glued on one side to a sh tail
shaped lucite light guide and read out by a
Philips XP2020 photomultiplier. The horizontal
scintillators are 720  300 mm2 and the
vertical scintillators 600  400 mm2 ; covering a
solid angle of B30%  4p with slight variations
depending on the exact position of annihilation
inside the traps. The photomultipliers are
doubly screened from the magnetic eld with a
m-metal shield inside an iron housing and
are mounted in a region where the stray magnetic
eld is small. They are biased to a negative voltage
of (22.5) kV, the gain being adjusted by changing

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

686

the bias voltage, while the appropriate thresholds

are determined using minimum ionizing particles,
both from cosmic rays and from p% annihilation.
Pairwise coincidences allow operation at a lower
threshold, reducing the random noise as well as
suppressing neutral background from gammas and
neutrons.
The discriminated signals are processed via
NIM logic and are recorded with a VME multiscalar module (Struck Instruments System,
SIS3806), which is read out via a MXI bus onto
a PC. The system allows dead-time free monitoring of trapped antiprotons with lifetimes longer
than several hours [21]. Typical background rates
are B20 Hz per coincidence pair, but can be
reduced by further requiring charged particle
multiplicities larger than or equal to two. The
signals can also be used to trigger the antihydrogen
detector readout (see Section 8).
The excellent correlation between the beam
intensity measured by the external beam detector
and the number of annihilations of trapped
antiprotons measured by the external annihilation
detector is illustrated in Fig. 7. The details of
antiproton trapping and cooling are discussed in
Section 4.1.

The trap system is realized by a sequence of

electrodes having 1:25 cm inner radius and various
lengths. Different sections of this system are used
to perform different functions. Proceeding from
the antiproton beam entrance and moving toward
the positron accumulator, the rst 12 electrodes
are used to catch, cool and accumulate antiprotons
(catching trap). The following group of electrodes
is referred to as the mixing trap. This is the region
where the antiprotons are merged with the
positron plasma. The last group of electrodes is
used during the positron transfer and re-capture
procedure.
All the electrodes outside the mixing region are
copper, plated with gold to avoid oxidation.
Below, a 0.51 mm thick palladium sub-layer was
deposited to improve plating performance and
avoid the diffusion of gold. Static or varying
voltages can be applied independently to each
electrode allowing the electric eld inside the trap
to be shaped according to the particular operation
required.
4.1. The catching trap
The catching trap is composed of 12 cylindrical
electrodes (Fig. 8). The two outermost electrodes
(HVL and HVR) supply the high voltage to catch
the incoming AD antiproton bunches. The central
part of the trap, comprised of 10 electrodes, is used
to conne the cold antiprotons. This section
includes seven electrodes designed to produce an
harmonic potential (Penning trap) along the axis
by an appropriate choice of the electrode lengths
and applied voltages [22]. One of the harmonic
region electrodes was split in four in order to allow
the application of the rotating wall technique for
plasma compression [23] or, alternatively, to detect
plasma diocotron modes [24].

3500

3000

Annihilation Counts

4. The trap system

2500

2000

1500

1000

500

0
0

10

15

20

25

4.2. The mixing trap

+6

Number of antiprotons (x 10 )
Fig. 7. Counts measured by the external annihilation detectors
upon the release of trapped antiprotons as a function of
antiproton number in the incoming beam as measured by the
external beam detectors.

The mixing trap is the region where the

interaction between antiprotons and positrons
takes place (see Fig. 9). The electrodes have
been arranged in such a way that the nested

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

687

Fig. 8. The antiproton catching trap. The trap design allows the application of up to 15 kV to the two outermost electrodes. The
central section is a seven electrode design; characterized by the presence of a central electrode (RING) and by three compensation
electrodes on each side. The central electrode is a 4-way split in order to drive and detect azimuthal plasma modes.

conguration [25] can be achieved. In this conguration particles of opposite charge can be
conned simultaneously and made to interact.
During the mixing of the two clouds recombination can take place. In essence the mixing region
consists of three sections marked RW, e W and
LW in Fig. 9; in each of them it is possible to set
up a harmonic potential well as described for the
catching trap. The positrons are normally conned
in the central of these three traps e W: All the
electrodes of the mixing region are 3:25 mm thick
and are made of aluminum as they are located in
the antihydrogen production zone inside the
annihilation detector. The use of aluminum
instead of copper, combined with the thinner wall
thickness, makes it possible to signicantly reduce
multiple scattering of the annihilation products on
their passage to the detector, as well as avoid
conversion of high energy photons to electron
positron pairs, which if they occurred inside the
detector volume would lead to an increased
background. Outside the mixing zone, the electrodes are 4 mm thick. As in the catching trap, one of
the electrodes in the positron trap section has been
split in four azimuthal quadrants to allow
for radial compression using the rotating wall
technique.
4.3. Additional electrodes
To increase the experimental exibility of the
traps a section made up of nine electrodes and
approximately 21 cm in total length was added on

the side of the mixing trap towards the positron

accumulator (R1-9 in Fig. 9). These electrodes are
used during the transfer of positrons to the mixing
trap and can act as additional particle reservoirs
for positrons or electrons, thus perhaps in the
future allowing transfer and stacking of positrons
during the recombination phase. At the end of the
mixing trap a high voltage electrode was added
(HVRR in Fig. 9); it can be biased to a potential of
about 5 kV: In this way, antiprotons can be
trapped directly in the mixing section and cooled
by either electrons or perhaps directly by a
positron cloud produced previously by stacking
several accumulation cycles of positrons (see
Section 7). In this way high densities and particle
numbers of positrons could be utilized to further
optimize the antihydrogen formation process.

5. Antiproton capture, cooling and manipulation

The standard operating procedure for antiprotons in ATHENA consists of their capture in the
catching trap, cooling them by collisions with a
preloaded cloud of electrons and transferring the
cold antiprotons to the mixing trap. The last
aluminum degrading foil is mounted on the trap
entrance electrode HVL. The potential of the exit
electrode (HVR) at the other end of the catching
trap is initially set to VHV : Antiprotons traversing
the foil with axial energy lower than eVHV are
reected from HVR and captured by applying a
voltage VHV to HVL before they return there. The

ARTICLE IN PRESS

Antiprotons from
Catching Trap

0.6

-3

Catching Efficiency (x10 )

Positrons from
the accumulator

HVRR
RW
e+W
LW

Fig. 9. The mixing trap. Antiprotons from the catching trap arrive at the left and positrons from the positron accumulator (Section 7) arrive at the right. The nested
trap is made up of three sections each with the possibility of making a harmonic well. The two outer ones of these (RW and LW) are intended for the antiprotons and the
central larger trap e W is intended for the positrons. The position of the antihydrogen detector (Section 8) is shown. The additional high voltage electrode (HVRR) is
marked.

M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

688

0.5

0.4

0.3

0.2

0.1

5
6
VHV [kV]

Fig. 10. Dependence of the catching efciency on the applied

high voltage, VHV : The antiprotons were released from the trap
1 s after capture. The numbers of captured antiprotons are
normalized to the beam intensity measured with HPD-based
external beam detectors.

antiproton arrival time is determined by the silicon

beam counter (see Section 3.1). The optimum trap
closing time, which depends on bunch duration
and energy and on the trap length, was determined
experimentally.
The number of captured antiprotons is determined by the external annihilation detectors (see
Section 3.4) by lowering the voltage of the
entrance electrode (with a 5 ms time constant),
thus allowing the antiprotons to annihilate on the
degrader. The time distribution from the detector
is related to the total number of captured
antiprotons and their axial energy distribution.
Fig. 10 shows how the number of captured
antiprotons increases with applied trap potential.
Typically around 10,000 antiprotons are captured
at 5 kV for an incident AD ux of 2:5  107 =pulse:
Fig. 11 shows how the number of captured
antiprotons decreases when the trap closing time
is increased. The optimum closing time is 500
700 ns: Note the slow decrease in catching
efciency as a function of closing time, indicating
that a signicant fraction of the antiprotons leave
the degrader with very low energy.

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

Catching Efficiency

10

10

10

10

Here mp and me are the p% and electron masses, e is

their electrical charge and L is given by
 
4pe0 kT 3=2
:
L
e2
ne

-3

-4

-5

-6

689

8 10 12 14 16 18 20 22
Closing Time [s]

Fig. 11. Dependence of the catching efciency on the trap

closing time delay.

5.1. Antiproton cooling

Coulomb collisions between antiprotons and
electrons preloaded in the catching trap can
efciently cool the high energy antiprotons [26].
Although the electrons are heated by this process,
they cool themselves by synchrotron radiation in
the 3 Tesla magnetic eld with a time constant of
about 0:4 s: Ideally, the two species of particles will
reach a nal equilibrium temperature equal to that
of the environment. The cooling process is usually
described by the differential equations [27]
dTp
Tp  Te

tc
dt
dTe np 1
Te  Tt

Tp  Te 
te
dt
ne t c
where Te and Tp are the electron and p% temperatures, Tt is the unperturbated electron temperature, ne and np are the electron and p% densities, te is
the synchrotron and tc the electron cooling time.
The latter is given by


3me mp c3
kTp
kTe 3=2

:
tc
82p1=2 ne e4 lnL mp c2 me c2

The solution of these equations shows that 104

antiprotons having energies in the keV range can
be cooled down to less than a few eV within a few
tenths of a second if they overlap completely with
an electron cloud of density around 107 108 cm3 :
The ATHENA electron source consists of a
barium oxide disc cathode (Kimball Physics Inc.)
mounted on a movable support in the positron
transfer region (Section 7.3). The electrons are
loaded in the p% catching trap before the arrival of
the antiproton pulse. This is achieved by shaping a
narrow, low-voltage potential well (typically a few
tens of volts) in the central region of the catching
trap. The electron primary current traverses the
catching trap and reaches the entrance electrode
HVL where it may be dumped on the degrader foil
or repelled by it depending on the electric eld
conguration. Electrons can be loaded in different
sections of the catching trap by various procedures. They differ from each other by the values of
the voltages applied to the trap electrodes during
the primary current passage and they lead to
clouds having different initial shapes and densities
and distinct evolution dynamics.
The cooling process is studied by dumping the
hot antiprotons from the high voltage well and
then later the cold antiprotons cooled by the
electrons and captured in the narrow internal
electron well. Fig. 12 shows a typical time
distribution taken with the external annihilation
detectors for trapped antiprotons released and
annihilating on the degrader at the entrance of the
trap. The electric potential was rst lowered from
5 kV to 40 V (with a time constant of B20 ms)
releasing the higher energy antiprotons (HV dump
in the gure), and then from 40 to 0 V in about
1 ms; thus releasing the antiprotons which had
been cooled via interaction with preloaded electrons (small trap dump). Fig. 13 shows the fraction
of cold and hot antiprotons as a function of
interaction time. Nearly all antiprotons are cooled
in about 60 s: At the end of the cooling process,
antiprotons and electrons share the same volume.

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

690

40000
160

35000

120

30000

100

HV dump

80

Antiprotons

Annihilation counts

140

60
40

Small trap dump

20
0

20000

20100

20200

20300

20400

Time [ms]

25000
20000
15000
10000

Fig. 12. A typical p% annihilation time spectrum measured by

the external annihilation detector (see text). The clock is started
at injection of the p% beam into the trap.

5000
0

2
3
Number of Shots

Fig. 14. Dependence of the number of cold antiprotons on the

number of stacked AD shots. Each AD shots contains about
2  107 antiprotons.

Normalized p Number

0.8

5.2. Electron plasma characteristics

0.6

0.4

0.2

0
0

10

20

30
40
Cooling Time [s]

50

60

Fig. 13. Measured fraction of cold (circles) and hot antiprotons

(triangles) as a function of their interaction time with electrons.
The dashed lines are to guide the eye.

The electrons can then be ejected from the trap by

applying appropriate electric pulses of about
100 ns duration which do not affect the heavier
antiprotons.
Since the catching process does not inuence the
potentials within the central region where the cold
antiprotons and the electrons are collected, several
AD shots can be stacked in the catching trap. This
is illustrated in Fig. 14 which shows a linear
increase in the number of trapped cold antiprotons
with the number of AD shots.

The high density and low temperature of the

electron cloud makes the Debye length shorter
than the trap cloud extension and means that the
electron cloud is well within the plasma regime.
The evolution of the electron plasma can be
studied using destructive or non-destructive diagnostic systems. The destructive diagnostic simply
consists of dumping the electron cloud onto the
degrader mounted on HVL (see Fig. 8), which also
acts as a Faraday cup, and reading the collected
charge by means of a low-noise high-impedance
amplier. A nondestructive diagnostic based on
monitoring of the axial plasma modes is also
implemented, and will be discussed in Section 6.
The electron storage time is limited by collisions
with the residual gas, asymmetries and imperfections in the trap geometry and in the elds, leading
to radial transport across the magnetic eld and
expansion of the plasma. This is a well documented phenomenon in the eld of nonneutral plasmas
[28]. In ATHENA, electron plasmas of few cm in
length with several 108 electrons are routinely
loaded in the catching trap. The electron storage
time depends on the trap length, the depth of the
potential well and the choice of electrodes. Typical

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

electron storage times (the time necessary for the

particle number to reduce by a factor two ) range
from a few hundred to a few thousand seconds
[29,30].
5.3. Antiproton storage time
Fig. 15 shows the number of cold antiprotons
(i.e. captured in the narrow well) as a function of
time, in the presence of electrons or after having
ejected them. The background pressure as measured in the room temperature region of the
apparatus external to the cryogenic region (see
Section 2.2) was about 1011 mbar and the trap
temperature 15 K: We observed that the electrons
reduce the cold antiproton storage time; without
electrons the half-life of the antiproton cloud is
usually better than 10 h:
5.4. Antiproton transfer
Cold antiprotons are transferred toward the
mixing region by adiabatically moving the electrode voltages along the traps. The transfer
efciency in the case when electrons and antiprotons are moved together are compared with
those when electrons are rst ejected from the
catching trap and only the cold antiprotons are
5

-4

Normalized p Number (x10 )

4.5
4
3.5
3
2.5
2
1.5
1
0.5
0

200

400

600

800 1000 1200 1400 1600 1800

Storage Time [s]

Fig. 15. Storage time for cold antiprotons with (squares) and
without (circles) electrons. The antiproton numbers are normalized to the beam intensity measured with the HPD-based
external beam detectors.The lines are to guide the eye.

691

transferred. Transfer efciencies greater than 90%

are obtained when electrons and antiprotons are
transferred together while a dramatic decrease in
the transfer efciency has been observed if cold
antiprotons are moved alone. In addition, while
the number of cold antiprotons stored in the
catching trap linearly increases with the number of
AD shots (Fig. 14), under the experimental
conditions of 2002 the transfer efciency did not
follow the same behavior. Thus when stacking
several shots in the catching trap prior to transfer,
most of the transferred antiprotons (about 75%)
originated from the last trapped bunch. This
behavior was unexpected and could be related to
possible radial (centrifugal) [31] separation of
antiprotons and electrons leading to instabilities
during the transfer. An improvement in the
electron loading and especially the antiproton
transfer procedure during the ongoing 2003 run
has solved this problem. The normal procedure in
the antihydrogen production runs in 2002 required
the stacking and transferring of 3 AD shots
resulting in about 10,000 cold antiprotons available for recombination in the mixing trap.

6. Plasma modes diagnostic

The thermal equilibrium state of a large number
of positrons or electrons conned in a Penning
trap at low temperature is a rigidly rotating
spheroidal plasma [32] with a sharp boundary.
Models predict that the density is almost constant
within the ellipsoid and that it falls off exponentially with the Debye length at the plasma
boundary.
Knowledge of the characteristics of the plasma
(i.e. dimensions, density, temperature) can be
obtained by means of a nondestructive method
based on measurement of the rst two axial
electrostatic mode frequencies (dipole, quadrupole) [33,34].
The modes are excited by applying sinusoidally
time-varying potentials to one trap electrode, while
the plasma response can be measured by acquiring
the induced current on another electrode (see
Fig. 16). The ratio of the induced current to the
excitation amplitude is measured as a function of

ARTICLE IN PRESS
692

M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

the drive frequency. A narrow stepwise frequency

sweep (4 ms duration per 5 kHz step) is made
across the resonant frequency of each mode. For
each frequency step, the amplitude and phase
(relative to that of the drive signal) of the voltage
induced by the plasma motion is acquired. This
excitedetect process is performed by means of a
network analyzer (HP4395) integrated with suitable attenuation and amplifying circuits. The cross
talk signal between the transmitting and receiving
electrodes is acquired without positrons and
subtracted from the signal measured with the
plasma present.
A detailed and simple analytic theory of the
electrostatic modes in nonneutral plasmas exists
[33]. In the framework of this theory the frequencies of the rst two symmetric axial modes depend
on the plasma size, density, and temperature. The

zero temperature model is used to determine the

plasma density and the so called aspect ratio
(which is the ratio between the axial and radial
extension of the plasma).
Moreover the exact plasma response (see
Fig. 17) can be modeled using a resonant circuit
analogy, where the values of the components are
related to the plasma properties. The diagnostic
also allows, in addition to the previous parameters, the plasma length 2z0 to be obtained. The
radius r0 and the particle number N are then
determined and a complete nondestructive diagnostic system is obtained. Temperature T shifts
produced by the application of a radiofrequency
signal resonant on the (1,0) mode are monitored as
changes in the (2,0) mode frequency. The model
[35] is used to calculate the induced temperature
increase. The diagnostic system as well as the
temperature monitoring and control is described in
more detail elsewhere [36,37].

7. Positron accumulator
7.1. Overview and operation

Fig. 16. Schematic diagram of a Penning trap with the modes

analysis and heating circuit.

The operation of the ATHENA positron

accumulator is based on the buffer gas capture
and cooling of positrons in a PenningMalmberg
trap. The techniques used were pioneered by the

Dipole Mode

Quadrupole Mode
0.003

0.016

0.0025
Amplitude [arb. units]

Amplitude [arb. units]

0.014
0.012
0.01
0.008
0.006

0.002

0.0015

0.001

0.004
0.0005
0.002
0

20

20.05

20.1 20.15
f [MHz]

20.2

20.25

0
31.96

32.2

32.44 32.68
f [MHz]

32.92

33.16

Fig. 17. Measurement of the amplitude of the rst two low-order axial modes as a function of the drive frequency. In this case the
measured plasma parameters are n 6:3  107 cm3 ; z0 2:0 cm; and r0 0:1 cm:

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

693

Fig. 18. Schematic overview of the positron accumulator (see text).

University of California San Diego positron [38

40] and electron [23,41,42] groups. As described
below, up to 2  108 positrons have been trapped
in this accumulator prior to transferring them
across a low eld region to the main ATHENA
recombination trap.
The accumulator traps and cools a continuous
beam of slow positrons. These are generated by
moderating
b
particles
from
a
1:4 GBq 40 mCi 22 Na radioactive source and
guiding them into the trapping region using axial
magnetic eld transport. A cryogenic cold head
capable of reaching 5:5 K cools down the source
and makes it possible to grow a solid neon
moderator directly on the source [4345]. Details
of the design of the source holder and its interface
with the cryogenic section can be found elsewhere
[4648]. Fig. 18 shows the trapping region. A
NaI(Tl) detector is located close to a gate valve
which can be used to isolate the source/moderator
end of the apparatus and the main trapping
region. This facilitates optimisation of moderator
growth, with the closed valve used as a simple
positron annihilation target. The absolute beam
intensity was obtained by cross-calibrating the
NaI(Tl) detector and performing coincidence
measurements with a channeltron detector which
was periodically placed in the beam-line. Moderation efciencies (absolute beam intensity divided
by the total positron activity of the source) of
around 0.4% are routinely achieved such that
beam intensities greater than 5  106 positrons/s
are available.
The trapping scheme utilizes nitrogen buffer gas
to trap and cool the positrons. Initial trapping

occurs during the rst passage of the positron

through the trap electrodes by electronic excitation
of the nitrogen gas. Such a transition is favored in
nitrogen compared to positronium formation,
which is the only other major inelastic channel
open at our kinetic energies. After trapping, axial
connement is provided by applying appropriate
electric potentials to the electrode array, whilst the
radial connement is provided by a 0:14 T axial
magnetic eld. Once trapped the positrons continue to lose energy in collisions with the gas,
nally residing in the potential well formed by the
voltages applied to the large diameter trap
electrodes.
One of the trapping electrodes (see Fig. 18) is
split into six segments to compress the plasma by
applying a rotating electric eld (the so-called
rotating wall technique) [23,42]. In this technique the rotating electric eld transfers torque to
the plasma resulting in radial compression. The
method was recently shown to work well for
positron plasmas [49,50]. In the present apparatus
the electrodes used have a signicantly larger
radius C10 cm than in these earlier experiments,
but results presented below show that the positron
plasma can still be inuenced. Rotating wall
compression leads to heating of the plasma, and
since the magnetic eld in the trap is too low for
efcient re-cooling by synchrotron radiation,
another cooling mechanism has to be used. We
have successfully used the nitrogen buffer gas
already present in the trap to provide this cooling,
despite the fact that this gas has a poor positron
cooling rate [39,51,52]. The presence of the
segmented electrodes in the accumulation trap

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

allow rotating wall compression during positron

accumulation. This reduces positron losses due to
cross-eld transport in the presence of the buffer
gas, leading to a larger number of accumulated
positrons.
The detection system used to monitor the
performance of the positron accumulator has
two main components: (i) a segmented Faraday
cup detector consisting of nine plates to extract
information on the size and position of the plasma.
The Faraday cup has an area of 25 cm2 and it is
situated outside the main magnet in a region where
the magnetic eld is about a quarter of that inside
the trap. This means that the plasma size obtained
directly from the Faraday cup is magnied
compared to its actual size inside the magnet and
this has been taken into account to derive the true
plasma dimensions; (ii) a calibrated CsI-photodiode detector (see Section 7.4) to monitor the
annihilation signal generated when the positrons
strike the Faraday cup.
After trapping, the positrons still need to be
transferred to the main magnet containing the
mixing trap. A transfer section was constructed for
this purpose. This section consists of a vacuum
separation valve, a pumping restriction, a number
of transfer electrodes and a transfer magnet
capable of pulsing from 0 to 1 T in 20 ms and
staying at 1 T for 1 s: The performance of this
system is reported below.
7.2. Optimization and results
An optimization program was undertaken to
tune the performance of the accumulator. We
varied the electrode potentials and buffer gas
pressures and ne-tuned the alignment of the
magnetic eld to the physical axis of the system.
Fig. 19 shows the end result, the accumulation of
more than 108 positrons in a few minutes. When
using a suitable frequency and amplitude for the
rotating wall compression (see below in this
section) and applying this signal for the last 50%
of the accumulation time, the lifetime of the
positrons can be doubled in the presence of the
buffer gas whilst maintaining the same accumulation rate. The data using the rotating wall
compression (Fig. 19) was tted with a lifetime

Accumulated positrons [millions]

694

200

150

100

50

0
0

200

400

600

Accumulation time [sec]

Fig. 19. Accumulation of positrons with (closed circles) and

without (open circles) rotating wall compression. The lines are
ts to the data (see text for details).

of 200 s whilst the data without the rotating wall

gave a lifetime of 95 s: It is important to stress that
these results occur with the buffer gas still present
in the trap. The increase in lifetime with rotating
wall compression shows that annihilation on the
gas is not the dominant loss. This points instead to
plasma loss due to collision-induced cross-eld
drift to the electrodes.
Plasma compression was optimized by mapping
out the properties of the plasma as a function of
the frequency and amplitude of the applied
rotating electric eld. The CsI-photodiode detector
was used to record the total number of stored
positrons, whilst the segmented Faraday cup
monitored their position. Plasma centering could
then be observed by the signal increase on the
central plates of the Faraday cup. The ratios
between the signals on the various Faraday cup
plates yielded positional information and absolute
plasma sizes.
Fig. 20(a) shows measured ratios between
signals from the central region of the Faraday
cup with and without rotating wall compression.
The central region covers about 20% of the total
area. The direction of rotation of the applied
electric eld coincides with that of the natural
~B
~ rotation of the plasma. Fig. 20(b) shows
E
the corresponding total number of positrons
observed by the CsI photodiode detector.
The data in Fig. 20(a) exhibit a broad enhancement in the compression in the frequency range

ARTICLE IN PRESS
695

80
1.2
1.0

0.1 V co-rotating
0.2 V co-rotating
0.5 V co-rotating

0.8
0.6
0.4
0.2

Millions of positrons

Central region divided by

total on Faraday cups without RW

M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

0.0

70
60
50
40
30
20
10

(a)

200

400

600

800 1000 1200 1400

Frequency [kHz]

(b)

200

400

600

800 1000 1200 1400

Frequency [kHz]

Fig. 20. (a) Ratio of positron numbers in the central region of the Faraday cup to the total signal with and without rotating wall
compression. (b) Corresponding signal from the CsI-photodiode detector.

300600 kHz increasing with amplitude. The fact

that the ratio for the central region rises above
unity means that parts of the positron plasma
which initially missed the Faraday cup, e.g. due to
cross-eld transport, have been compressed into
the central region. Above 600 kHz an abrupt falloff occurs which coincides with a similar decrease
in the total number of stored positrons. Up to
600 kHz the total number of positrons is very
stable for all amplitudes.
The data in Fig. 20 indicate a compression of
about 2.5. However, the true compression turns
out to be larger. The central region of the Faraday
cups contained ve individual plates, of which
only three actually recorded a signal when the
rotating wall was used. By examining e.g. ratios
between adjacent plates and comparing to similar
data when the Faraday cup was moved slightly off
axis, it was possible to derive the position and size
of the plasma. The plasma had a width of 15 mm
(FWHM) when no rotating wall was applied,
reduced to 34 mm following compression. Since
the total number of positrons stayed constant, or
in some cases increased, the central density
increased by more than a factor of ten. This
compression ratio is much larger than that
reported previously for N2 [50]. This can possibly
be attributed to the higher gas pressures used in
the present study.
7.3. Positron transfer
After accumulation the positrons are transferred
to the mixing trap inside the main 3 T magnet. The

nitrogen buffer gas is pumped out and, after the

pressure in the positron accumulator has fallen
below 108 mbar; the valve is opened and a pulsed
transfer magnet is energized for 1 s: This transfer
magnet produces a eld of 1 T and thus helps
bridge the low-eld region between the positron
accumulator and the main ATHENA magnet (see
Fig. 1) while also making it possible to separate the
two vacuum systems with a pumping restriction.
The positrons are released by lowering a gate
electrode with a fall time of 1 ms and trapped by
closing another gate electrode in the main magnet,
3:2 ms later. This traps the positrons initially in the
entire length of the mixing trap and the adjacent
positron trapping section. The positron plasma is
then subsequently axially compressed into the
central harmonic region of the mixing trap. This
compression takes a few tens of seconds. The
overall efciency for transfer, recapture and
compression is about 50%. It should be noted
that this efciency was found to be mainly limited
by the electronics, particularly the closing time of
the trapping electrode in the main magnet.
Efciency gains are anticipated when faster
electronic units are installed. The present settings
allow the delivery of about 75 million positrons for
recombination every 5 min: The lifetime of the
compressed positrons in the mixing trap is quite
long as no signicant loss was observed during a
hold period of 4000 s:
We conclude this section by remarking that the
operation of the positron accumulator, including
the rotating wall compression, was found to be
stable and reproducible over a wide range of

ARTICLE IN PRESS
696

M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

operating parameters. Thus the size, density and

position of the positron plasma can be predicted
prior to transferring it to the main recombination
trap.
7.4. The CsI monitor
A calibrated CsI crystal with photodiode readout measures the annihilation intensity when the
positron beam is dumped on the segmented
Faraday cup. The calibration was achieved using
the g-lines at 511 keV and 1:274 MeV from a weak
22
Na source of known activity. Since the 511 keV
line from positron annihilation is used to determine the number of positrons, contributions at
low energies from the 1:274 MeV line (e.g.
Compton scattering) need to be subtracted. This
was achieved using a calibrated 60 Co source which
emits gs at 1.17 and 1:33 MeV; and hence
produces a spectrum in the 511 keV region similar
to the 22 Na 1:274 MeV line.
The sources were placed in turn inside the
positron accumulator at the position of the
segmented Faraday cup so that conditions were
as close as possible to those experienced during the
experiment. In particular the gs produced a
spectrum similar to that of the Compton scattered
photons from the annihilation of the positron
plasma.
Two quantities were extracted from the background-corrected spectra of the 511 keV line: (i)
the average energy per photon and (ii) the overall
efciency of the detector. The average energy per
photon is used to calculate the number of
positrons annihilating on the Faraday cup during
a standard plasma dump. In its operating position,
the CsI photodiode detector had an efciency of
0:04970:006% including solid angle acceptance.
Two further CsI photodiode detectors were used
to monitor the transfer of the positron plasma
from the positron accumulator to the mixing trap.
One was situated at a place from when it
particularly observed any losses in the transfer
section and the other was situated directly outside
the main magnet (see Section 2.1) at the position of
the degrader foil where the positron plasma was
dumped. These detectors were then calibrated by a
similar procedure to the main CsI diode detector

by placing sources at suitable positions inside the

apparatus and measuring the resulting detector
response, thus again allowing for any material
dependent attenuation. These detectors made it
possible to optimize the positron transfer independent of the main ATHENA detector (see Section
2.1) and provided an independent way of crosschecking the degrader Faraday cup calibration for
positrons.

8. The ATHENA antihydrogen detector

In a homogeneous magnetic eld electrically
neutral H% atoms escape the connement
region and annihilate on the trap electrodes
producing on average about three charged
pions, three high energy gs and two 511 keV gs
(Fig. 21).
The detector [53,54] was designed to allow
extraction of a clean H% signal for background
rates of up to 10 kHz; although the effective
background rate was later found to lie below
100 Hz: Charged particles are detected in two
layers of Si-m-strip detectors covering roughly 80%
of 4p: A three dimensional reconstruction of the p%
annihilation vertex is achieved with s 4 mm
spatial resolution by straight line extrapolations of

511keV

Silicon micro
strips

CsI
crystals

511keV

Fig. 21. Sketch of H% annihilation on the trap wall (not shown).

Solid lines represent charged pions, wavy lines photons from
positron annihilation. The crystals hit by the charged pions or
511 keV gamma photons are indicated. High energy gs from
p0 -decay are not shown.

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

the charged particle tracks. Photons from positron

annihilation 511 keV convert in the CsI
crystals via the photoelectric effect with a
probability of about 25%. The reconstruction
efciency for antihydrogen events can, in principle,
be increased if Compton scattering is also
included. The segmentation into 192 crystals
is required to ensure high enough angular resolution to verify that the two 511 keV photons
are emitted back-to-back. High granularity is
also imposed by background considerations
in p% annihilation. The annihilation releases
2 GeV of center-of-mass energy which is distributed among the pions and later high energy
gs. With 192 crystals the probability of a crystal
being hit by either a pion from the annihilation
vertex or one of the high energy gs is about 5%/
crystal. This background contains (spatially uncorrelated) 511 keV photons produced by g
showers in the surrounding apparatus (e.g. magnet
coils). This background signal will always be
present in our antihydrogen signal but it does
not exhibit the angular correlation of 180 that the
two 511 keV photons from the annihilation of
antihydrogen produce.
PCB (APD's)
Crown
support
Si strip detectors
Crown

12 crystals per row

Hybrids (Si strips)

Main support

APDs
25 cm

Fig. 22. Three-dimensional drawing of the H% detector.

697

8.1. Design and realization of the detector

The most stringent constraint on the detector
design is the low operating temperature of 140 K;
determined by its location a few millimeters from
the cold 15 K trap region in the center of the
experiment. Low temperature is, in principle,
advantageous for semiconductor detectors since
parallel noise from leakage currents practically
vanishes, but constraints increase on mechanics,
cabling and electronics. It also prevents easy access
for debugging in nal working conditions. Fig. 22
shows the cylindrically symmetric design of the
detector operating in a vacuum of 107 mbar: The
overall dimensions are 75 (140) mm inner (outer)
diameter and 250 mm in length. The outer
diameter is limited by the cold bore of the
superconducting solenoid. The inner diameter is
limited by the size of the cryogenic coldnose UHV
vessel containing the electrodes of the nested
Penning trap (see Section 2.1). The inner-most
part of the detector of thickness less than 10 mm
contains two layers each of 16 double sided silicon
m-strip modules, 162 mm long. Separation between
the modules and the 192 crystals is provided by the
500 mm thin aluminum wall of the main support,
which was manufactured from a single piece of
aluminum using electro-erosion. The crystals with
dimensions 17  17:5  13 mm3 are grouped in 16
rows of 12, closely lling the remaining space.
The m-strip modules consist of two double sided
sensors (SINTEF, Norway, 81:6  19 mm2 with a
wafer thickness of 380 mm) glued onto a silicon
mechanical support, and a multilayer ceramic
hybrid, 2 mm thick (Fig. 23).
The sensor p-sides are segmented into 384 AC
coupled strips with an implant width of 32 mm and
a pitch of 46:5 mm: Every third strip is read out
while the two intermediate strips are oating. The
readout strips (139:5 mm pitch) are bonded to the

Fig. 23. Silicon m-strip module.

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

pitch adapter integrated in the hybrid, itself

bonded to a 128 channel VLSI chip (VA2 TA)
glued on the hybrid (see Section 8.2). The 64 DC
coupled pads 1:25  18 mm2 of the sensor n-sides
are oriented perpendicular to the strips. Thin gold
plated aluminum lines on the silicon support
deliver the 128 signals to the bottom of the hybrid
from where they are brought to a second chip
VA2 TA via 128 feedthroughs in the ceramic. For
simplicity, a common ground is used for both
readout chips. This requires the depletion voltage
65 V to drop on the AC couplings of the sensor
strips. A fraction of 3% of the strips had a short
circuit to the implant and had to be disconnected.
However, they could still deliver signals by
capacitive coupling to adjacent strips.
Pure CsI scintillation crystals obtained from
CRISMATEC (France) are used to detect the
511 keV photons. The light yield is excellent at low
temperature (E50; 000 photons=MeV at 80 K
[55]), and the decay time E1 ms ts the shaping
time of the front end electronics. The crystals were
originally coupled to 17  13 mm2 photodiodes
covering the front of the crystals. For cost reasons
they were produced on the same wafers as the
silicon strip sensors, requiring only one additional
mask for the shallow n-implant on the light
sensitive side. However, only a modest signalover-noise ratio (E15 for 511 keV absorption)
could be achieved due to the unavoidable serial
noise, electric contact problems, difcult optical
coupling and low blue light sensitivity. These
diodes were then replaced by avalanche photo
diodes (APD, Hamamatsu, type S8148,
5  5 mm2 ), obtained from the electromagnetic
calorimeter project of the CMS collaboration,
after carefully testing their low temperature
functionality. From 300 diodes not satisfying
quality criteria for CMS, 220 were selected for
operation at a gain of 2073; at 135 K and 230 V
bias voltage. A typical signal-over-noise ratio of
larger than 50 was achieved in spite of the smaller
active area of the APDs, which is the limiting
factor for the energy resolution (C18% FWHM at
511 keV; Fig. 24).
The APDs (capacity CD E80 pF) or the old
photodiodes are also read out with VA2 TA chips.
One chip is located at the end of a 230 mm long

12000

back scatter
22

Na

10000

150 K

8000

Counts

698

511 keV
6000

FWHM = 18%

4000

2000

1275 keV
0
0 60

500

1000

1500

Signal height [keV]

Fig. 24. 22 Na energy spectrum measured by one crystal
equipped with an APD.

Fig. 25. Block diagram of the detector (see text).

and 12 mm wide printed circuit board (PCB)

which connects the 12 APDs of one row to the
12 rst input channels of the chip. A carbon bre
inlay was inserted into the PCB under the area
supporting the chip to match the different thermal
expansion coefcients.
The 16 PCBs of the crystals and the 16 ceramic
hybrids of each silicon m-strip layer are connected
in parallel by 8 cm long kapton cables (33 lines)
each to a cylindrically bent 0:5 mm thick printed
circuit board. This arrangement divides the
detector into three independent subgroups, or
patch panels (Fig. 25). Each patch panel (at
140 K) is connected by a 1:5 m long bundle of 50
shielded cables (LakeShore 1 mm diameter 50 O
cryo-cables with drain wire and aluminium foil

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

independently, but have to be masked during the

chip readout time 150 ms; due to digital cross
talk. The time jitter is typically 120 ns for the mstrips and 300 ns for the crystals.
8.2. The read out chip VA2 TA
The trigger on pure photon events from e
annihilations is an essential component of the
detector. The trigger signal is also needed to
monitor the high instantaneous annihilation rates
of antiprotons and positrons and for general
diagnosis without full read out of the analog
signals.
The VA2 TA chip (Fig. 26) is based on a series
of CMOS VLSI chips (Viking [56], VA), already
well established in high energy physics applications. A 32 channel version (VA32 TA32 set of
separated analog and trigger chips) was rst
successfully tested at low temperatures.
A 128 channel chip with a medium scale input
stage (60 +11/pF e
rms ), a standard shaper E2 ms
and a second faster shaper E75 ns; was developed for our application by Ideas ASA, Norway.
The outputs of the fast shapers are fed to 128
discriminators (see Fig. 26) with common threshold and programmable transition polarity. A
register is loaded to disable noisy channels. Power
consumption, shaping times, threshold voltages or
output signal heights are controlled by 12 DC
levels. These settings are generated as voltages by
the DACs on the digital repeater cards and are
common to all subdetectors. Fig. 27 illustrates for
instance the spread in shaping times for the crystal
readout, as a function of temperature. Optimizing
VA2_TA block diagram

Fast shaper

Threshold

H(s)

Pulse height
discriminators

Preamp
Slow shaper
H(s)

S/H

128 x

Hold in

Serial output

Logic OR

Input #1

Multiplexer

shielding) to a CF150 vacuum ange at room

temperature housing ve 50-pin D-Sub feedthroughs.
Outside the vacuum vessel, ve repeater cards
are plugged directly into the ange, which also
serves as the main grounding point. Three of these
boards, the digital repeater cards, house the main
electronics for controlling the three subdetectors.
They contain a low noise power supply 72 V;
some logic circuitry (GALs) and level translators,
16 voltage-DACs for chip settings, bipolar calibration pulse generators, the receiver for trigger
signals (currents) and a programmable window
discriminator (trigger multiplicity). The cards are
connected by two digital busses to VME modules.
A sequencer (Caen V551B) controls the readout
and ADC sampling by 8 differential TTL lines,
while a 16 bit ECL digital I/O unit (Caen V262) is
used for programming trigger masks, serially
loaded DAC registers and other logic states of
the detector, such as calibration and test mode.
Since no other active electronics could be placed
in the cold section of the detector, the differential
current output buffers of the chips drive the
multiplexed analog signals over 1:5 m long cables
of the same type as the digital lines. These signals
are fed into receivers on two analog repeater cards
on the vacuum ange. One pair of lines is used for
each of the 256 channel m-strip modules, while the
16 crystal PCBs share four lines. Differential line
drivers send the 36 analog signals to 18 double
channel FADCs (Caen V550) located in the same
crate as the other VME modules. The crate is
connected via an MXI-2 bus (National Instruments) to a PC in the ATHENA control room
running LabView as DAQ software. Data (40
kByte/event) can be written to disk with a rate of
typically 100 Hz; mainly limited by the 5 MByte=s
transfer rate of the VME-PC link.
Shaping times for both detector components are
set to 3 ms; adapted to the time constant of light
emission for pure-CsI [55]. The read out controller
freezes all 8384 analog channels simultaneously
and provides a fully synchronized readout, event
by event. The total power consumption amounts
to roughly 5 W: The multiplicity coded trigger
lines from the individual subdetectors (one line
from each subdetector) can, however, be recorded

699

128 x

Trigger output

Fig. 26. Block diagram of the VA2 TA chip.

ARTICLE IN PRESS

T = 298 K

350

300

250

0
0

20

40

60

80

100

120

140

160

180

T = 172 K

20

40

60

80

100

120

140

160

180

200

Ru

= 12 ns

150
100

0
0

50

100 150 200

Trigger delay [ns]

TA shaper output
30

p = 120 ns

25
20
15
10
5
0
100

200

300

400

500

Trigger delay [ns]

T = 140 K

Fig. 28. Timing resolution (left) and shaper output (right) of

the VA2 TA chip.

4
2
0
0

106

50

2
0
0

Counts

Shaping times [s]

Trigger threshold [mV]

M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

700

20

40

60

80

100

120

140

160

180

Crystal channel

Fig. 27. Temperature dependence of the peaking time for the

crystal shapers.

settings lead to an average shaping time of

370:5 ms at 140 K:
Some of the chip settings, in particular the input
transistor current of the preamplier, are quite
sensitive to noise down to mV: Therefore ceramic
decoupling capacitors and RC lters are used on
all DC lines of the hybrids (PCBs). Some
capacitors had to be oversized by up to 500% to
take into account the strong ferroelectric behavior
of dielectrica under DC operation at low temperatures. Power for the digital part of the chip is taken
locally from the analog supply via RC lters (10 O;
20 mF), while further 20 mF tantalum capacitors
buffer the analog power.
Global grounding was optimized empirically by
removing or setting various ground bridges on key
positions of the detector. The two grounds of the
digital and analog parts in the chip were connected
directly to the ground plane on the hybrids and
PCBs. Stable working conditions were observed
with no observable cross talk from ring triggers
nor common mode noise.
Due to ohmic resistances in the service cables
(the 250 shieldings are connected to a single
ground) and the fully oating operation with
respect to the rest of the ATHENA apparatus,
the detector ground may be as high as 50 mV:
Sensitive voltages such as the trigger threshold are
corrected for this offset by voltage dividers
referencing to the hybrid (PCB) grounds. Various
performance measurements were made on the
VA2 TA chips. The trigger timing resolution at
room temperature from a small test detector

CD E20 pF with respect to a plastic scintillator

coupled to a photomultiplier is shown in Fig. 28
(left). Fig. 28 (right) shows the output signal of the
fast shaper (TA), recorded by scanning the trigger
delay over trigger threshold. A peaking time tp of
120 ns was thereby determined for the fast shaper.
8.3. Position resolution of the m-strip modules
The signal-to-noise ratio of the m-strip modules
is about 40 on the strip and 50 on the pad side,
respectively, measured with minimum ionizing
particles at low temperatures. The somewhat
higher noise on the strips, in spite of the p-type
implants, is caused by the large ratio of strip
implant to pitch widths, leading to a rather large
interstrip capacitance of E2 pF=cm: However, a
good efciency for charge collection on the strips
was measured with cosmic rays and a high
precision sE1 mm incident telescope [57]. Charge
loss on the oating strips was negligible. Cosmic
data could also be used to determine the spatial
resolution on the strip side. Fig. 29 shows the Z
distribution, where Z is dened as the fractional
charge collected on one of the two adjacent
readout strips Z QL =QL QR : The two central
peaks are due to capacitive coupling of the
intermediate oating strips to the readout strips.
The differences between predicted hit positions
from the incident telescope and measured position,
using the weighted average of adjacent strips, are
plotted in Fig. 29. We obtained a position
resolution of s 28 mm: However the much worse
overall vertex resolution of approximately
4 mm 1 s in ATHENA is determined by the
unknown particle track curvatures in the 3 T
magnetic eld of the trap region. This is because

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711
35

701

200

30
150

Counts

25
20

100

15

= 28 m

10

50

5
0
0

0.2

0.4

0.6

0.8

-280

-140

140

Position relative to expected [m]

Fig. 29. Z distribution (left) and spatial resolution (right, see text).

we extrapolate the tracks of the charged pions

from only two points in the silicon m-strip detector
and with just these two points and no information
about the charge of the detected pion, we have no
way of determining the track curvature.

9. Control and data acquisition system

The control system is required to control all 27
electrodes of the mixing trap and the 11 electrodes
of the positron transfer section (see Section 7.3),
and to manage communication with the computers
in charge of catching and cooling antiprotons and
accumulating positrons when these particles are
being transferred into the mixing trap.
Voltages on the electrodes of the mixing trap are
supplied by programmable triggerable DACs
(three 100 kHz 16-bit VME-based DACs (Joerger
VDACM) with a buffer depth of 32k steps and one
1 MHz 12 bit PCI-based National Instruments
DAC (6713) with a buffer depth of 16k steps).
Voltages to the transfer electrodes, as well as
triggers to the DACs, are supplied by two PCIbased 10 MHz programmable and triggerable 32channel pulse generators (Becker & Hickl PPG100) with a buffer depth of 64 kB: Additional fast
(rise time o100 ns) pulses of arbitrary shapes are
provided by four 100 MHz 12 bit VXI-based
waveform generators (RACAL Instr. 3151), which
can be connected to any of the 27 mixing trap
electrodes.

The interface to the control system is written in

LabVIEW; in combination, the pulse generator
and the DACs are able to provide the mixing trap
and transfer electrodes with voltage sequences
consisting of several thousand steps and lasting
several hours with microsecond precision. In
addition, these sequences contain provisions for
further activities such as removing electrons from
the transferred antiprotons, controlling the status
of the electron source, and generic triggers and
gates. Once created, these sequences are downloaded to the DACs and pulse generators, leaving
the time-critical execution under hardware control.
The voltage range of the DACs and pulse
generators is increased to 7140 V by amplifying
all voltages with fast B1 ms pre-ampliers. All
voltages with the exception of those provided by
the RACAL DACs are ltered by a 10 MHz lowpass lter to reduce noise; each electrode has an
additional capacitively coupled unltered input for
plasma diagnostics. This input is used on two
electrodes to apply 40 V=100 ns pulses generated
by a 30 MHz GPIB-based pulser (SRS DS345) to
mixed antiproton-electron clouds, in order to
efciently clean the antiprotons of electrons (at
least 95% of the electrons are eliminated with four
pulses). Transfers of antiprotons and positrons are
managed by setting up dedicated sequences of
voltages which are triggered by the remote
computers, once handshaking has been established, at which point sequences steering the
mixing of antiproton and positron clouds take
over. These procedures are fully automated, and

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

702

VME

VME read out

VME busy

VD-71 transient rec.

DAQ
10 MHz

SIS-3806

>

degrader
CsI
SC-12 preamp

Scintillators

HPDs
ADC 2282 / NI-ADC

>

Beam counter
ADC 2282 / NI-ADC

TA
Crystals

Scintillators

>

LeCroy 1176 TDC

TA
Si Outer

V 550

>

V 551 B

TA
Si Inner

Progr. logic module

Trigger Trigger
Level 1 Level 2

Repeater card

Trigger
Level 0

data base

Detector readout
Trap movements

Electrode voltages

TRAP

POS

RF

Sequencer

NI-DAC

Pulser DS345

RACAL 3151

Joerger VDACM

MIX

Sequencer PPG100

PBAR

GPIB
VME
VXI

ETHERNET

Fig. 30. Diagram of the data acquisition system.

can repeat the same experiment an arbitrary

number of times.
The data acquisition system (DAQ, Fig. 30)
produces a record of all activities in the apparatus:
standard detector read out, changes in the voltages
of the electrodes, state of the apparatus, voltage
and timing information from the control system,
and inter-computer communications. In addition
to the central detector readout (see Section 8),
scintillator pulse heights (determined by LeCroy
1182 ADCs), pulse shapes (recorded by INCAA
VD71 transient recorders), and oscilloscope signals are acquired via VME and GPIB modules
readout by a LabView program, and written to a
database. The order in which the data are written
to the database does not necessarily correspond to
the order in which events occur, since several
modules contain multi-event buffers, and the
control sequence for voltages and timings (which
is downloaded to the data base before being
executed) is not measured in real time. To retroactively establish timing, the heart of the DAQ

consists of multi-hit time stamp units (one 8channel and one 32-channel Struck SIS 3806)
which record with up to 1 ms accuracy the time of
each type of activity. Dead time-free acquisition is
guaranteed internally by the use of two counter
banks, one of which is active at any given time,
while the data of the inactive bank is piped into a
64 k FIFO. Switching between the two banks,
which count the number of 100 ns intervals
between switches, is triggered by an OR of all
triggers. Triggers are generated by a 16-input
VME general purpose programmable logic module
(GPPLM, trigger level 2) combining all detector
signals; this module allows the simultaneous
construction of a variety of trigger conditions.
This information is then used by the ofine data
analysis program to establish trigger rates, correlations of one type of event (e.g. antiproton
annihilation rate) with another type (e.g. voltage
changes on the electrodes), time ordering of
activities and detector dead times. Although the
VME read out rate itself is bandwidth limited

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

(corresponding to a detector read out rate of about

100 Hz), trigger rates up to a few MHz are reliably
recorded for 32 types of triggers in parallel, and
with a time resolution of 10 ms; by the 32-channel
time-stamp unit. A higher time resolution of 1 ms is
achieved for a further eight special interest triggers
with the 8-channel time-stamp unit.

10. The online software

The ATHENA online system was developed in
C++ by using the ROOT package [58]. Data from
the apparatus are analyzed in real time and saved
on media by the DAQ program. As described
previously, data acquisition is performed in LabVIEW on a Windows NT machine and the data
are streamed to disk via a Dynamic Link Library
written in C. The data are saved in a raw le by
using a private binary format without any
compression to obtain the maximum possible
speed during acquisition.
The online system consists of two C++
programs: (i) a monitoring program (AtOM,
ATHENA On-Line Monitor), that reads the raw
data from the DAQ and processes them, and (ii)
an inspection utility (AtVIEW) for a quick
analysis of the raw data during the runs. Because
data are saved asynchronously as they are
recorded by the detectors, the online software uses
a hardware time stamp to re-order the events in the
correct time sequence, to correlate data from the
traps with data from the detectors.
The program also lls a set of preselected
histograms used to monitor the experiment and
saves the data as a ROOT tree in a TOF le (TOF,
Time Ordered File). This ROOT le is compressed
by a factor of about three with respect to the raw
le. The results of the online analysis are also
posted on the web for remote monitoring and are
stored on tape at CERN via the CASTOR (Cern
Advanced STORage) server.
The ROOT data tree is then passed to the ofine
software that extracts the detector events from
the tree and performs the vertex reconstruction
and the event analysis, as will be explained in
Section 11.

703

Hence for each run three les are generated: (i)

the raw data produced by the DAQ, (ii) the TOF
le produced by the on-line, and (iii) the le
produced by the ofine programs.

11. The ofine reconstruction software

The ATHENA Off-line software is written in
C++ and uses a set of ROOT macros. It decodes
the detector response, reconstructs the interaction
points of the particles in the inner and external
silicon layers, associates the tracks to these points
(pattern recognition), nds the vertex of the
charged particles and selects the crystals with
511 keV g-signals. A ROOT online monitor provides a display of the reconstructed events. Some
batch macros give a set of histograms to control
both vertex positions of the annihilations and
detector performances.
A Monte Carlo (MC) code has been written to
study the optimum selection criteria for H% events
and to evaluate the probability that a selected H%
event may be a background event. The chain for
processing MC events is identical to that for real
data.
The MC program was conceived as a C++
interface between ROOT and the GEANT code
for the generation of simulated data from nuclear
physics experiments [59]. This interface allows the
call of the FORTRAN routines of GEANT using
ROOT macros. The ATHENA apparatus geometry was implemented in the MC code using the
standard GEANT routines and copied into ROOT
using a set of specialized routines. A phase space
event generator was written, with the possibility to
generate pp% and/or e e annihilation events, in an
uncorrelated or spacetime correlated way. The
ppmesons annihilation reactions considered,
%
together with their branching ratio, are taken from
Ref. [60]. It is possible to call for one channel in
 
particular (for example pp-p
p p p ) or to
%
generate an annihilation event according to its
branching ratio.
It should be noted that in reality p% annihilations
occur on nuclei of the trap electrodes or of the
residual gas lling the trap volume, rather than on
free protons. Although neutral pion multiplicities

ARTICLE IN PRESS
704

M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

on complex nuclei are uncertain, a reduction of

2040% with respect to the pp
% case is expected
[61]. Therefore, the background effects due to the
electromagnetic showers of high energy gs in the
magnet surrounding the traps and the detector are
probably slightly overestimated in our MC.
Finally, the efciency of each one of the 32
silicon modules and the energy response of each
one of the 192 crystals have been implemented into
the MC code, to fully simulate the performance of
the real detector.
11.1. Point reconstruction
The hit coordinates of charged particles from
the silicon m-strip modules are found by standard

Fig. 31. Strippad correlation for silicon modules with one hit
only. The two lines denote the boundaries of the correlation
band used for method (b) described in the text.

clusterization algorithms applied to the pedestal

subtracted ADC values. The x; y; z hit coordinates are calculated from the r  f-strip and zpad information. Ghost points occurring when
two or more particles hit the same microstrip
module are eliminated in two ways: (a) by keeping
the hit conguration that minimizes the vertex
residuals in three dimensions and (b) by associating strips and pads that are correlated in amplitude
when a module is hit by a single particle. Fig. 31
shows the correlation between strip and pad ADC
contents. When a module has multiple hits, we
consider those points as ghosts that have an ADC
cluster contents out of the band indicated in the
gure. It has been our experience that the best
efciency is obtained by using method (a) when
there are two hits on the module and method (b) in
the cases with more than two hits. The fraction of
modules traversed by two or more particles is 13%
and the ambiguity is resolved in 40% of the cases.
Fig. 32 shows the ADC amplitude distribution
normalized to the traversed silicon thickness with
cosmics (a) and for a standard run with antiprotons annihilating on the walls of the recombination trap (b). For cosmics good agreement is
obtained with predictions assuming a Landau
distribution, whereas the Landau t to the
annihilation data deviates at higher dE=dx because of the presence of non-relativistic particles.
Therefore, in Fig. 32(b) the Landau t was
restricted to the peak only (full curve). We recall
that annihilation products are mostly pions of
300 MeV=c average momentum [60], with some

Fig. 32. ADC amplitudes per unit path length in silicon for cosmics (a) and for particles from p% annihilations (b). The measurements
were done using the strip side of the detector and for the cosmic data there was no magnetic eld. The full curves show the tted
Landau distribution. The dashed curve in (b) shows the prediction of the t limited to the peak.

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

705

350

250

Dedicated positron run

Peak Position = 519 keV

Counts / (5 keV)

Counts / (2 keV)

300
200

FWHM = 130 keV

150

100

cold mixing

250

Peak Position = 498 keV

200

FWHM = 78 keV
150

hot mixing
(306 meV)
x 1.6

100

50
50
0

(a)

0
0

100

200

300

400

500

600

700

Energy [keV]

800

(b)

100

200

300

400

500

600

700

800

Energy [keV]

Fig. 33. A typical crystal spectrum in ADC count obtained during a dedicated positron run (a) and during a normal run (b). The lines
indicate the limits of the 511 keV window.

kaons and nuclear fragments. It is these heavier

nuclear fragments that account for the discrepancy
between the Landau curve and the annihilation
data for higher values of dE=dx:
Assuming that the cosmic rays are 4 GeV
muons, the Landau most probable ionization loss
is C3:35 MeV=cm: From the most probable value
from Fig. 32(a) (5641 ADC counts/cm) one
obtains a conversion factor of C0:6 keV=ADC
count.
The reconstruction software was tested by
measuring the pitch between readout strips. The
four cosmic ray hits obtained with no magnetic
eld were tted to straight lines. The minimization
of residuals as a function of a variable pitch in
r  f gave a readout pitch of 133 mm and a mean
residual of 240 mm; in very good agreement with
the true pitch of 139 mm:
To detect antihydrogen a good performance
of the crystal detector is crucial in order to
assure correct reconstruction of the positron
annihilation. During the experiments, the detector
temperature was monitored continuously and the
crystals were calibrated periodically with positron
annihilation data coming from dedicated runs
where the recombination trap was lled with
positrons only.
The analysis of the 192 spectra from the
individual crystals was performed in a semiautomatic way, with computer ts followed by
the visual inspection of the displayed spectra, to
maintain all the windows correctly around the

511 keV signal. The optimal lower and upper

window limits were found to be 345720 and
675720 keV; where the uncertainties refer to the
differences between the crystals.
In Fig. 33 typical crystal spectra obtained during
the dedicated positron calibration runs and during
normal data taking runs [1] are shown. The
511 keV photon signals from the normal runs
appear at the lower end of the energy spectrum,
close to the noise (see also Fig. 24), while
minimum ionizing charged particles which deposit
about 6 MeV=cm in the crystals and pair production from high energy gs generate a at plateau
centered around 10 MeV: The background due to
these p% annihilation products is clearly visible in
Fig. 33(b). Therefore, the identication of the
511 keV signal from antihydrogen required the
dedicated event selection procedure described in
Section 11.3. Concerning the stability of the whole
detector, peak shifts of less than 5% over a 5
months period have been found for each of the 192
crystals.
11.2. Event reconstruction
A very simple pattern recognition algorithm
associates each point in the rf plane on
the external silicon layer to the nearest point
on the inner silicon layer. These two points
dene a charged particle track through
the detector. Particular cases, like three-hit
clusters, are considered as three-point tracks

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

706

which pass through the overlapping inner modules

(see Fig. 35).
A line passing through the two or three points is
then calculated. A straight line t is performed in
the three point case, whereas in the two point case
(about 95% of the tracks) the best t equations are
used only to propagate the errors from the hits on
the track parameters, to estimate the accuracy of
the vertex. As mentioned in Section 8.3 track
curvatures due to the magnetic eld cannot of
course be measured with two layers. Systematic
effects deriving from the straight line approximation are discussed below.
We use two different vertex reconstruction
algorithms, called the 3D (three-dimensional) and
the dual algorithm, respectively. In the 3D
algorithm the track is parameterized by the
direction cosine vectors c and the track length
parameter ti ; the vertex coordinates are found
from the ti values that minimize the sum of
distances
(
D2min

min

)
jri  rj ci ti  cj tj j

ioj

After minimization, the vertex residual in three

dimensions dened as
D2min
R
number of tracks to the vertex

is an estimate of the resulting accuracy. Apart

from this residual estimate, the 3D algorithm does
not permit an easy error propagation, because of
its intrinsic non-linearity. However, knowledge of
the vertex error is necessary for the denition of
the vertex quality, an important parameter in
event selection.
For this reason we also implemented the dual
algorithm. When tracks have a common vertex
xv ; yv in the xy plane, they obey the equation
yv ai xv bi ; which implies that bi ai xv
yv is a straight line in the dual ab plane with
coefcients xv and yv : This is true, with obvious
change of notation, for each projection x1 y1 ; y2
z1 and x2 z2 : For example, for the xy projection
the vertex coordinates are obtained by minimizing

the w2 [62]:
wxv ; yv 2
N
X
i1

s2 b

bi  xv ai  yv
2
2 2
i xv s ai  2xv covai ; bi

where the errors on ai ; bi stem from the propagation of the errors on the coordinates x; y; z of the
silicon strips and pads. In the absence of magnetic
eld theseperrors
approximately by sx
are given
p
sy Df = 12; sz Dz = 12 where Dz C1:2 mm
and Df C260 mm are averages between the pitch
C139 mm and the silicon thickness C380 mm:
No attempt was made to nd the xy error
dependence on the azimuthal angle, since the
magnetic eld effect is overwhelming. Indeed,
realistic input errors are mostly due to the straight
line ts that ignore the track curvatures in the
magnetic eld. By studying this effect with
simulated data, we found that the effective xy
pitch value, with magnetic eld, is C1:8 mm:
Therefore, the input errors have been xed to
Df
sx sy 7  p;
12
Dz
sz p;
12

Df 260 mm

Dz 1:2 mm:

After the ts in the three projections, each vertex

coordinate gets two values with corresponding
errors. The nal coordinates of the vertex are given
by their weighted averages. The comparison
between the vertex coordinates x1 ; x2 ; y1 ; y2 and
z1 ; z2 is a powerful tool to judge the quality of the
vertex reconstruction.
We assign to each vertex a Quality Index
(Q.I.) calculated by assigning a score to each of the
following good vertex properties: (1) a small vertex
residual R (Eq. (2)) in three dimensions:
Rp0:4 cm; (2) a small discrepancy between dual
and 3D algorithms: the vertices should agree
within 4 mm; (3) a good matching of the x1 y1 ;
x2 z1 ; y2 z2 projections in the dual algorithm; (4) a
small vertex error in the dual algorithm:
maxfsx; sy; szgp0:5 cm: (5) more than two
tracks; (6) a small distance o0:1 cm between the
inner silicon layer point and the line joining the
external silicon layer point and the vertex.

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

The efciency and accuracy of the vertex

reconstruction have been studied with real and
simulated data as a function of the Q.I. In Table 2
the reconstruction efciency of real and MC data
is reported, together with the MC vertex resolution
in space (xyz) or in the (xy) projection only.
The results for the (yz) and (xz) projections are
similar. The resolution in the vertex reconstruction
has been veried with real data through a sample
of point-like p% annihilations on the trap electrodes.
The results, shown in Fig. 34, are in very good

707

agreement with the MC results of Table 2. For the

details of p% annihilation imaging, see Ref. [63]. A
reconstructed real four prong annihilation event
on the trap wall is shown in Fig. 35.
11.3. Antihydrogen event selection
The hit selection in the silicon m-strip detector and the vertex reconstruction strongly
enhances p% annihilation events with practically

Table 2
Quality of the reconstruction software with simulated annihilations on the trap walls (MC) and with real data, in space (xy
z) or in the (xy) projection only
Q.I. Projection ZMC Zreal sx (cm) sy (cm) D (cm)
(%)
(%)
0
0

xyz
xy

68

63

0.53
0.68

0.51
0.68

0.55
0.72

6
6

xyz
xy

13

0.30
0.37

0.32
0.39

0.38
0.47

Z is the vertex reconstruction efciency, sx and sy are the r.m.s

deviations of the distribution in x (or y) [reconstructed]-x (or y)
[MC], and D is the distance between generated and reconstructed MC vertices.

Fig. 35. Reconstructed antiproton annihilation event.

Fig. 34. Vertices reconstructed in the xy plane from point-like p% annihilations on trap electrodes. The dimensions are in cm. (a): 2-D
view; the ring indicates the trap section radius 1:25 cm: The measured width of the sharpest peak is s 0:5 cm:; (b): 3-D view of
the histogram shown in (a).

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

708

no background contamination. For the crystals,

however, two main sources of background contribute: (i) short tracks from charged particles in

the crystals and the conversion of high energy gs

lead to low energy background signals in the
511 keV region; (ii) positrons from showers
generated in the surrounding magnet by high
energy gs also lead to 511 keV signals (see Fig.
36). Our MC describes this background well, as
shown in Fig. 37.
These background effects have been greatly
reduced by means of particular H% selection
criteria, studied by extensive Monte Carlo data.
The procedure for antihydrogen event selection,
which fully exploits the high granularity of the
detector, is as follows. Only events with a
reconstructed vertex are considered. Reconstructed tracks are extrapolated toward the crystals and those hit by an extrapolation line are
identied and excluded from the subsequent g
detection together with their eight neighbors. Due
to the limited efciency, some tracks are not
reconstructed and some isolated hits remain.
Hence, also the crystals that lie above a hit on
an external silicon layer module are considered as
traversed by a charged particle and excluded as is
its eight neighbors. After this rst selection,
crystals are scanned for having a signal within
the 350550 keV window and events with exactly
two such crystals are selected. To reduce from
this sample the fake H% events due to the e.m.
cascades from the surrounding magnet, also the

Fig. 36. A Monte Carlo event showing the noise generated on

the detector by the high energy g showers in the surrounding
magnet. Only the crystals with a g signal are shown.

10
10
10

25000

20000

15000
10000

10
5000
1
0

0
0

9 10

(a) crystals with the 0.511 MeV signal

(b)

number of tracks per vertex

Fig. 37. Real p% (bars) and MC (solid line) detector responses: (a) isolated crystals with the energy response in the 511 keV window
(note the logarithmic scale). The events with 2 isolated crystals contains the entries of the opening angle histogram (see Fig. 38 below);
(b) number of reconstructed tracks per event.

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

709

Table 3
Real and Monte Carlo efciencies for events entering into the opening angle histogram of Fig. 38 (left) and into the rst two channels
around cos yC  1 following the selection procedure described in the text

Entries
Signal

MC
data
MC
data

H% events

p%
Background (1)

pe
% e
Background (2)

9:4570:08102

6:3570:08102
6:1270:11102
1:970:1103
1:870:2103

9:7670:10102

5:970:2103

4:670:2103

The MC results are obtained starting from pure H% annihilations on the trap wall, p% annihilations in the open volume of the trap
(background 1) and spatially uncorrelated pe
% e annihilations on the trap wall (background 2). The real data come from pure p%
annihilations on the trap wall.

Fig. 38. Left: Monte Carlo generated cos y distribution of the opening angle y for the H% annihilation (bold line) and for the p%
annihilation background of type (1) (gray) and the uncorrelated pe
% e annihilation background of type (2) (hatched) (see the text).
Right: the same distribution in the case of real data during the cold mixing of p% and positrons (bold line) and the same histogram
(hatched) in the case of hot mixing without recombination. The H% and background distributions are normalized to the same number of
reconstructed vertices.

requirement that the two selected crystals are not

surrounded by other crystals with a signal above
threshold is made. Therefore, the nal sample
consists of events with a reconstructed p% annihilation vertex and two isolated crystals in the right
energy window (called the candidate crystals), with
the possibility to have some other crystals in the
detector with signals outside this window.
The distance between the line joining the center
of the two candidate crystals and the reconstructed
vertex is expected to be zero in H% annihilations.
This corresponds also to the condition cos y 1
where y is the opening angle between the two lines
joining the candidate crystals and the p% vertex (see
also Fig. 21).

Simulations of antihydrogen annihilations are

based on the p% and e annihilations occurring at
the same time and at the same position on the wall
of the recombination trap. We also considered two
types of background, (1) due to a p% that
annihilates on a residual gas atom in the trap
volume and (2) due to p% and e e annihilations at
the same time but in two different locations on the
trap wall, simulating thein terms of detection
worst case of an H% atom that loses its positron
before annihilating.
The MC efciencies for the H% annihilations and
for these two types of background are shown in
Table 3. The simulation shows also that the
relative efciencies between the H% signal and the

ARTICLE IN PRESS
710

M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

backgrounds are independent of the Quality Index

of the reconstructed vertex.
The events and the signal in the table are dened
as the number of entries in the opening angle
histogram of Figs. 38 and that in the rst two
channels, corresponding to 1p cos yp  0:947:
The Monte Carlo histogram for the cosine of the
opening angle is shown in Fig. 38 (left), where a
clear signal around the value cos yC  1 appears
in the case of H% annihilations, in agreement with
the efciencies reported in Table 3. Apart from
this signal, the rest of the histogram shows a
plateau of events due to the 511 keV gs from
positron annihilations that have been lost by
scattering their original direction or to the
accidental ring of crystals due to soft gs from
the e.m. cascades into the magnet or other parts of
the apparatus.
During the runs, we mixed p% and positrons in
the recombination trap at low temperature (cold
mixing) obtaining a clear H% formation signal,
shown with the bold line in Fig. 38 (right). In the
same gure, the hatched histogram shows the
distributions obtained with the recombination
completely stopped by heating the positron plasma
with a radio-frequency signal [1]. The MC and real
data of Fig. 38 show clearly that our apparatus
allows the detailed study of both H% production
and the various sources of background.

Natural Science Research Council, the UK Engineering and Physical Sciences Research Council,
the EU (EUROTRAPS Network) and the Royal
Society.

References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]

Acknowledgements
We would like to thank S. Bricola, P. Chiggiato,
A. Mongelluzzo, G. Chuste, J. Hansen, A. Lanza,
C. Marciano, J. M. Rieubland, G. Sobrero, N.
Vauthier, J. Casas-Cubillos, M. A. Rodriguez
Ruiz, B. Schmid and E. S^ndergaard. Thanks
are also due to the CERN AD team. Professor C.
Surko provided essential advice. This work was
supported by the following funding agencies:
Istituto Nazionale di Fisica Nucleare (Italy),
Conselho Nacional de Pesquisa e Desenvolvimento (CNPq), Fundao de Amparo a Pesquisa do
Estado do Rio de Janeiro (FAPERJ) e Fundao
CCMN/UFRJ (Brazil), Grant-in-Aid for Creative
Basic Research of Monbukagakusho (Japan), the
Swiss National Science Foundation, the Danish

[20]
[21]
[22]
[23]
[24]
[25]
[26]

[27]
[28]

M. Amoretti, et al., Nature 419 (2002) 456.

M. Niering, et al., Phys. Rev. Lett. 84 (2000) 5496.
R. Carosi, et al., Phys. Lett. B 237 (1990) 303.
G. Baur, et al., Phys. Lett. B 368 (1996) 251.
G. Blanford, et al., Phys. Rev. Lett. 80 (1998) 3037.
L.S. Brown, G. Gabrielse, Rev. Mod. Phys. 58 (1986) 233.
M.M. Nieto, M.H. Holzscheiter, Appl. Phys. B 60 (1995)
103.
S. Maury, Hyp. Int. 109 (1997) 43.
P. Riedler, et al., Nucl. Instr. and Meth. A 478 (2002) 316.
C. Amsler, et al., Nucl. Instr. and Meth. A 321 (1992) 69.
R. Medenwaldt, et al., Nucl. Instr. and Meth. B 58 (1991)
1.
M.C. Fujiwara, M. Marchesotti, Nucl. Instr. and Meth. A
484 (2002) 162.
L.K. van Geest, et al., Nucl. Instr. and Meth. A 310 (1991)
261.
R. DeSalvo, et al., Nucl. Instr. and Meth. A 315 (1992)
375.
H. Arnaudon, et al., Nucl. Instr. and Meth. A 342 (1994)
558.
Delft Electronische Producten, Dwazziewegen 2, NL-9300
AB Roden, The Netherlands.
ASACUSA Collaboration, private communication; M.C.
Fujiwara, ATHENA Technical Report 2000, unpublished.
P. Lubinski, et al., Phys. Rev. Lett. 73 (1994) 3199.
C. Gonzalez, F. Pedersen, in: A. Luccio, W. MacKay
(Eds.), Proceedings of the 1999 Part. Acc. Conference
IEEE, Piscataway, NJ, 1999, p. 474.
M.E. Angoletta, et al., CERN Report PS 2001-044 (2001).
M.C. Fujiwara, et al., Hyperne Interact. 138 (2001) 153.
H. Raimbault-Hartmann, et al., Nucl. Instr. and Meth. B
126 (1997) 378.
X. Huang, F. Anderegg, E.M. Hollmann, C.F. Driscoll,
T.M. ONeil, Phys. Rev. Lett. 78 (1997) 875.
C.A. Kapetanakos, A.W. Trivelpiece, J. Appl. Phys. 42
(1971) 4841.
G. Gabrielse, S. Rolston, L. Haarsma, W. Kells, Phys.
Lett. A 129 (1988) 38.
S.L. Rolston, G. Gabrielse, Hyperne Interact. 44 (1988)
233;
M.H. Holzscheiter, et al., Hyperne Interact. 109 (1997)
145.
L. Spitzer, Physics of Fully Ionised Gas, Interscience
Publishers, New York, 1962, p. 585.
J.M. Kriesel, C.F. Driscoll, Phys. Rev. Lett. 85 (2000)
2510;
R. Chu, et al., Phys. Fluids B 5 (1993) 2378;

ARTICLE IN PRESS
M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711

[29]
[30]

[31]

[32]
[33]
[34]
[35]
[36]
[37]
[38]

[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]

C. F. Driscoll, J.H. Malmberg, Phys. Rev. Lett. 50 (1983)

167;
C.F. Driscoll, K.S. Fine, J.H. Malmberg, Phys. Fluids 29
(1986) 2015.
G. Testera, et al., Nucl. Instr. and Meth. A 461 (2001)
253.
M. Amoretti, et al., in: F. Anderegg, L. Schweikhard, C.F.
Driscoll (Eds.), Non Neutral Plasma Physics, Vol. IV, AIP
Conference Proceedings, Melville, NY, Vol. 606, 2002,
p. 45.
J.D. Larson, J.C. Bergquist, W.M. Itano, D.J. Wineland,
Phys. Rev. Lett. 57 (1986) 70;
T.M.O. Neil, Phys. Fluids 24 (1981) 1447.
L. Turner, Phys. Fluids. 30 (1987) 3196.
D.H.E. Dubin, Phys. Rev. Lett. 66 (1991) 2076.
M.D. Tinkle, R.G. Greaves, C.M. Surko, R.L. Spencer,
G.W. Mason, Phys. Rev. Lett. 72 (1994) 352.
D.H.E. Dubin, Phys. Fluids B 5 (1993) 295.
M. Amoretti, et al., Phys. Rev. Lett. 91 (2003) 055001.
M. Amoretti, et al., Phys. Plasmas 10 (2003) 3056.
C.M. Surko, S.J. Gilbert, R.G. Greaves, in: J.J. Bollinger,
R.L. Spencer, R.C. Davidson (Eds.), Non-Neutral Plasma
Physics, Vol. III, AIP Conf. Proc, Vol. 498, 1999, p. 3.
T.J. Murphy, C.M. Surko, Phys. Rev. A 46 (1992)
5696.
R.G. Greaves, M.D. Tinkle, C.M. Surko, Phys. Plasmas 1
(1994) 1439.
C.F. Driscoll, J.H. Malmberg, Phys. Rev. Lett. 50 (1983)
167.
E.M. Hollmann, F. Anderegg, C.F. Driscoll, Phys. Plasma
7 (2000) 2776.
A.P. Mills Jr., E.M. Gullikson, Appl. Phys. Lett. 49 (1986)
1121.
R. Khatri, M. Charlton, P. Sferlazzo, K.G. Lynn, A.P.
Mills Jr., L.O. Roellig, Appl. Phys. Lett. 57 (1990) 2374.
J.P. Merrison, M. Charlton, B.I. Deutch, L.V. J^rgensen,
J. Phys.: Condens. Matter 4 (1992) L207.
L.V. J^rgensen, et al., in: F. Anderegg, L. Schweikhard,
C.F. Driscoll, (Eds.), Nonneutral Plasma Physics, Vol. IV,

[47]

[48]

[49]
[50]
[51]
[52]
[53]
[54]
[55]
[56]
[57]
[58]

[59]

[60]

[61]
[62]
[63]

711

AIP Conference Proceedings, Melville, NY, Vol. 606,

2002, p. 35.
L.V. J^rgensen, M.J.T. Collier, K.S. Fine, T.L. Watson,
D.P. van der Werf, M. Charlton, in: W. Triftsh.auser, G.
.
Kogel,
P. Sperr, Positron Annihilation ICPA-12,
Uetikon-Zurich, Switzerland, Material Science Forum,
Vol. 363365, 2001, p. 634.
M.J.T. Collier, L.V. J^rgensen, O.I. Meshkov, D.P. van
der Werf, M. Charlton, in: J.J. Bollinger, R.L. Spencer,
R.C. Davidson (Eds.), Non-Neutral Plasma Physics, Vol.
III, AIP Conference Proceedings, Melville, NY, Vol. 498,
1999, p. 13.
R.G. Greaves, C.M. Surko, Phys. Rev. Lett. 85 (2000)
1883.
R.G. Greaves, C.M. Surko, Phys. Plasmas 8 (2001) 1879.
G.R. Heyland, M. Charlton, T.C. Grifth, G.L. Wright,
Can. J. Phys. 60 (1982) 503.
M. Charlton, J.W. Humberston, Positron Physics, Cambridge University Press, Cambridge, 2000 (Chapter 6).
C. Regenfus, Hyperne Interact. 119 (1999) 301.
C. Regenfus, Nucl. Instr. and Meth. A 501 (2003) 65.
C. Amsler, et al., Nucl. Instr. and Meth. A 480 (2002) 494.
O. Toker, et al., Nucl. Instr. and Meth. A 340 (1994) 572.
C. Amsler, et al., Nucl. Instr. and Meth. A 480 (2002) 501.
Rene Brun, Fons Rademakers, ROOTAn Object Oriented Data Analysis Framework, Proceedings of the
AIHENP96 Workshop, Lausanne, September 1996; Nucl.
Instr. and Meth. A 389 (1997) 81, http://root.cern.ch/.
CERN Application Software Group, GEANT: Detector
Description and Simulation Tool, CERN Program Library
W5013, Geneva, 1993. http://wwwinfo.cern.ch/asd/geant/.
%
R. Armenteros, B. French, in: E.H.S. Burhop (Ed.), NN
Interactions in High Energy Physics, Vol. 4, Academic
Press, New York, 1996, p. 237.
G. Bendiscioli, D. Kharzeev, Riv. Nuovo Cimento 17 (6)
(1994) 1.
E. Calligarich, et al., Nucl Instr. and Meth. A 311 (1992)
151.
M. Amoretti, et al., Phys. Rev. Lett., in press.