nima paper2004

8/10/2019 NIMA Paper2004

1/33

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

The ATHENA antihydrogen apparatus

M. Amoretti a , C. Amsler b , G. Bonomi c, A. Bouchta c, P.D. Bowe d , C. Carraro a ,M. Charlton e, M.J.T. Collier e, M. Doser c, V. Filippini f , K.S. Fine c, A. Fontana f ,M.C. Fujiwara g, R. Funakoshi g, P. Genova f , A. Glauser b , D. Gr .ogler b , J. Hangst d ,R.S. Hayano g, H. Higaki g, M.H. Holzscheiter c,1 , W. Joffrain a , L.V. J ^ rgensen e, *,

V. Lagomarsinoa

, R. Landuac

, C. Lenz Cesarh

, D. Lindel.of

b

, E. Lodi-Rizzinii

,M. Macri a , N. Madsen b , D. Manuzio a , G. Manuzio a , M. Marchesotti c,P. Montagna f , H. Pruys b , C. Regenfus b , P. Riedler c, J. Rochet c, A. Rotondi f ,G. Rouleau c,2 , G. Testera a , D. P. van der Werf e, A. Variola a , T.L. Watson e,

T. Yamazaki g , Y. Yamazaki ga Genoa University & INFN, Genoa, Italy

b 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, Italyg 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

The ATHENA Collaboration

Accepted 16 September 2003

Abstract

The ATHENA apparatus that recently produced and detected the rst cold antihydrogen atoms is described. Its mainfeatures, which are described herein, are: an external positron accumulator, making it possible to accumulate largenumbers of positrons; a separate antiproton catching trap, optimizing the catching, cooling and handling of antiprotons; aunique high resolution antihydrogen annihilation detector, allowing an clear determination that antihydrogen has beenproduced; an open, modular design making variations in the experimental approach possible and a nested Penning trapsituated 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

ARTICLE IN PRESS

*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

8/10/2019 NIMA Paper2004

2/33

1. Introduction

The ATHENA collaboration recently succeeded

in making the rst cold antihydrogen %H bymixing antiprotons with a dense cold positron

plasma [1]. This was an important milestone on theway to the main goal of the ATHENA experiment,which is a spectroscopic comparison of theproperties of antihydrogen and hydrogen atoms.This will allow direct tests of CPT invariance forleptons and baryons with unprecedented accuracy.For example, the long lifetime of the metastable 2sstate (122 ms) leads to a natural line width of about 1 Hz for the 1s2s transition, offering, inprinciple, the possibility to compare anti-atomswith their matter counterparts with high precision,perhaps reaching 1 part in 10 18 : In fact, anaccuracy close to 10 14 should be achievable forantihydrogen if the 2 photon 1s to 2s transitioncan be measured with a similar precision to thatachieved for hydrogen in 2000 [2]. Such precisionis several orders of magnitudes better than thepresent best direct CPT tests on leptons andbaryons, and comparable to the indirect tests inthe neutral kaon system [3].

The rst observation of atomic antimatter

was made in 1996 at CERN when the forma-tion of antiatoms was reported at LEAR (LowEnergy Antiproton Ring). The internal antiprotonbeam with a kinetic energy in the GeV rangetraversed a Xe gas jet target and produced e epairs. In a very small fraction of these collisions apositron was captured by the antiprotons resultingin nine events attributed to antihydrogen [4].Similar observations were made in a subsequentexperiment at Fermilab [5]. These antihydrogenatoms were all created at kinetic energies in the 1

6 GeV range corresponding to equilibrium tem-peratures in the 10 13 K range. However, antihy-drogen needs to be produced in much largerquantities and at much lower kinetic energies (i.e.the so-called cold antihydrogen) to facilitate theaforementioned precision spectroscopic measure-ments of its properties. Indeed the highest preci-sion is expected for measurements on theantihydrogen atom held at rest in a neutral atommagnetic trap. The recent results from ATHENAhave brought spectroscopy a large step closer by

making copious amounts of cold antihydrogenavailable for experiments [1].

In the context of cold antihydrogen formation

two mechanisms are of importance and theirparameter dependence determines the boundaryconditions for the design of any experiment toproduce and manipulate cold antihydrogen. Thespontaneous radiative recombination of positronswith antiprotons e %p - %H hn depends onthe positron temperature p 1= ffiffiffiffiT p and thepositron density p n and results predominantlyin atomic levels with low principal quantumnumbers. Three-body recombination e e %p - %H e becomes competitive at cryogenictemperatures p 1=T 4:5 and high positron densi-ties p n2 and mainly populates atomic levels withhigh principal quantum numbers. De-excitation tothe ground state is slow and the nascent atom issusceptible to ionization by electric elds.

In the present paper we describe in detail theequipment used by ATHENA to produce anddetect cold antihydrogen atoms. A schematicillustration of the apparatus is shown in Fig. 1 .Low energy antiprotons are extracted from theCERN Antiproton Decelerator (AD). A super-conducting solenoid (3 T) with a cold bore houses

the antiproton capture trap and the antiproton positron mixing trap. The antiprotons are moder-ated through a silicon beam-proling counter, afoil and various windows and are then reected inthe catching trap by a high voltage electrode.About 500700 ns after the arrival of the pulse ahigh voltage potential is raised on the entranceelectrode to capture the antiproton bunch. Theyare then cooled inside the catching trap to meVenergies by Coulomb interactions with a pre-loaded electron cloud. The antiprotons are con-

ned in the radial direction by the magnetic eld of the solenoid and in the axial direction by anelectrostatic eld produced by 10 cylindricalelectrodes having an inner diameter of 2.5 cm.The eld conguration is similar to the one used inPenning traps [6]. Further details can be found inSections 2, 4 and 5.

The positrons from a 1 :4 GBq 40 mCi 22Nasource are moderated in solid neon and transferredinto a longitudinal magnetic eld region wherethey lose energy by collisions with nitrogen gas

ARTICLE IN PRESS

M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711680

8/10/2019 NIMA Paper2004

3/33

and are eventually conned in a system of

cylindrical electrodes, as described in Section 7.They are then transferred to the superconductingsolenoid where they are held in a trap similar tothe one used to store the antiprotons.

The particles stacked in the two traps aretransported to the mixing trap held at a pressurebelow 10 12 mbar : The formation of antihydrogenatoms is studied by observing their annihilationwhen they impinge upon the electrodes of themixing trap. The annihilation detector, a largesolid angle array of silicon microstrip counters and

CsI crystals, surrounds the mixing trap. Itmeasures the charged hadron tracks (mostly pions)emitted by the annihilation of the antiproton, intemporal and spatial coincidence with the twoback-to-back 511 keV photons from e e annihi-lation. The detector is described in more detail inSection 8.

In Section 9 we describe the control and dataacquisition system and in Section 10 the onlinesoftware for time-ordering and rapid real timeanalysis of the data as it arrives is presented.

Finally in Section 11 the ofine analysis software,

Monte Carlo simulation of the apparatus andevent reconstruction and selection schemes aredescribed.

2. General experimental parameters

In the design of the ATHENA apparatusemphasis was placed on an open and exiblesystem. This has allowed for the accumulation of large numbers of positrons in an external appara-

tus with subsequent transfer into the mainapparatus as well as limited freedom in theapproach to antihydrogen formation and the typesof experiments that could be performed. Due tothe temperature dependence of both of theformation mechanisms mentioned in Section 1 itis crucial to maintain the lowest possible tempera-ture in the trapping region. In addition to this thetwo other main experimental conditions necessaryfor making antihydrogen are a very low residualgas pressure and a high magnetic eld.

ARTICLE IN PRESS

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

M. Amoretti et al. / Nuclear Instruments and Methods in Physics Research A 518 (2004) 679711 681

8/10/2019 NIMA Paper2004

4/33

2.1. The superconducting magnet and the cryogeniccold nose

The magnet used for the ATHENA experimentwas taken over from the original PS200T system[7] and consists of a superconducting solenoid witha 150 mm diameter room temperature bore and ahomogeneous eld region of 1 m length. Themagnet is capable of reaching 5 T but for theATHENA experiments was generally operated at3 T : The requirements of the ATHENA experi-ment were signicantly more complex than theantiproton trapping apparatus developed forPS200T such that a number of modications tothe original system were necessary.

To allow easy access to the trap withoutdisturbing the cryogenic system of the magnet, aseparate continuous ow cryostat (cold nose) wasinstalled in the bore of the magnet to cool thevacuum shell housing the trap. The cold nose canbe cooled to a few K, improving the cryogenicpumping speed at the location of the traps. Inaddition, since the electrons and positrons coolvery efciently by synchrotron radiation in the 3 Teld, a low ambient temperature allows produc-tion of very cold positron and electron plasmas

and, via sympathetic cooling, also very coldantiprotons. The cold nose is closed off at theend toward the AD beamline by the degrader foilsand the silicon beamcounter. At the end towardthe positron accumulator it has an open connec-tion to a room temperature vacuum chamber.During normal operation the outside of the coldnose needs to be shielded from room temperatureradiation by lowering the bore temperature of themain magnet. To allow radial space for mountingthe trap, the trap vacuum system, the cold nose,

and nally the detector, a common nitrogen shieldfor both the main magnet system and the cold nosewas designed. Under normal operation a tempera-ture of 130 K was achieved on the bore walls.

2.2. The vacuum system

Electromagnetic traps of the Penning andPenningMalmberg types are used to store, cooland manipulate the charged particles required byATHENA (positrons, antiprotons and the elec-

trons necessary to cool the antiprotons). Thesetraps (with the exception of the positron accumu-lator; see Section 7) are all situated within the

vacuum vessel of the cryogenic cold nose inside thesuperconducting solenoid. The trap environmentaltemperature is representative of the equilibriumtemperature of the antiproton and electron cloudsat the end of the cooling process. In addition,cooling the trap region reduces the surface out-gassing and transforms all the surfaces of theapparatus into a cryogenic pump. The gas owfrom the room temperature region is reduced bythe pumping system installed in that region and bythe trap and cold nose surfaces. Most of the wallsin the room temperature section have been coatedwith a NEG (Non-Evaporable Getter) material.After an initial pumpout this material is activatedby raising the temperature of the walls to 200 Cfor 24 h : This effectively turns the coated surfacesinto vacuum pumps. In addition to this a specialvacuum chamber lled with NEG coated metalstrips was installed. The NEG coated surfacestogether with a 300 l s 1 ionpump constituted theultra high vacuum pumping system in the roomtemperature region. The pressure can only bedirectly measured in the room temperature region,

where 10 11 mbar is routinely achieved. Themaximum temperature of the trap is approxi-mately 15 K when the complete apparatus isinstalled inside the cold nose, though the walls of the cold nose itself reaches temperatures in therange 410 K :

3. Antiproton beam monitoring

Pulses of antiprotons are ejected from the ADwith a momentum of about 100 MeV =c (or akinetic energy of around 5 :3 MeV) [8]. The %p shotslast for 200 ns and are repeated every 100 s :Typically, 2 107 antiprotons are injected intothe ATHENA apparatus. The locations of thebeam detectors are sketched in Fig. 2 . Theantiproton kinetic energy is degraded by 25 mmof stainless steel, located 10 cm upstream of the %pcatching trap, and by 130 mm of aluminum on theentrance electrode.

ARTICLE IN PRESS


8/10/2019 NIMA Paper2004

5/33

The antiproton beam is monitored by threesystems of detectors: (i) a silicon counter providesthe %p trigger for the catching trap high voltageswitch and measures the beam prole; (ii) theexternal beam detector, which comprises the frontand barrel plastic scintillators, detects the annihi-lation in a pulse mode and monitors the beamintensity and stability; (iii) the horizontal and

vertical external annihilation detectors, consistingalso of plastic scintillators, are used to study thetrapped antiprotons by detecting in a singlecounting mode the pions following annihilationeither on the residual gas or on the trap walls.

3.1. The silicon beam counter

The silicon beam counter [9] is a 67 mm thicksilicon diode, 15 mm in diameter, that wasoriginally designed for the Crystal Barrel experi-

ment [10]. The diode ( Fig. 3 ) is segmented into vepads, each connected to an individual signal line.The voltage required to fully deplete the diode is4:5 V: The beamcounter is located in front of theantiproton catching trap (see Section 4.1) and isoperated at temperatures ranging from 10300 K ;in a vacuum of E 10 8 mbar and in a 3 T magneticeld. A very thin counter has to be used to allowmaximum transmission of the low energy anti-protons into the catching trap. The average energyloss of 5 :3 MeV antiprotons in silicon was

estimated from experimental data [11] to be about

11 :4 keV per micrometer of silicon, thus creating3200 electronhole (eh) pairs, compared to 80 ehpairs for minimum ionizing particles. Thus,around 3 :2 1010 charge pairs mm of silicon aregenerated in the ATHENA silicon beamcounterfor an antiproton beam intensity of 10 7 %p per spill.In order to detect this high instantaneous current,a readout system was developed where the signalcurrent is read directly across a 100 O protectionresistor and fed into a digital oscilloscope and anADC. This also allows a direct measurement of the

ARTICLE IN PRESS

_ p

mix. trap

Magnetcoil

Magnetcryostat

_ res.+trans.e+

e+Si beam counter

Degrader

ANNIHILATION DETECTORVERTICAL EXTERNAL

HORIZONTAL EXTERNAL ANNIHILATION DETECTORS

BEAM DETECTORFRONT

BEAM DETECTORBARREL

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

Fig. 3. Photograph of the 67 mm thick silicon counter mountedin a PCB frame. The contacts to the individual pads are madeusing ultrasonic wire-bonds.


8/10/2019 NIMA Paper2004

6/33

spill duration. The signal from the silicon beamcounter is also used to trigger the %p-catching-trap.A typical signal registered on an oscilloscope isshown in Fig. 4 .

Fig. 5 shows the fraction of the total signalmeasured on the central pad for different biasvoltages. For each voltage setting the meanintegrated signal of three antiproton spills isshown. The error bars indicate the standarddeviation. The signal starts to plateau at about30 V but an operating voltage of 100 V was chosen

to achieve better timing performances and toreduce signal losses due to recombination of charge carriers.

3.2. The external beam detectors

The external beam detectors monitor annihila-tion on the degraders ( Fig. 2 ). They consist of twotypes of modules, Front and Barrel, both madefrom 1 cm thick Bicron BC408 plastic scintillators[12]. The two Front detectors are 195 100 mm 2

each and the two Barrel detectors are 800 195 mm 2 each. The modules of the Front andBarrel detectors cover about 0.1% and 3% of 4 p ;respectively, when antiprotons annihilate on thedegraders. Due to the high instantaneous rateC 1014 s 1 during the %p pulse, these detectors donot count single particles, but operate in currentmode, measuring the total charge deposited by theannihilation products. However, the light pro-duced in a 1 cm thick plastic scintillator wouldsaturate the photomultiplier. We therefore usedproximity focused Hybrid Photo Diodes (HPD)[1315] . These comprise a vacuum tube in whichthe photoelectrons are accelerated toward a silicondiode by a high voltage applied to the photo-

cathode. A gain of a few thousand can easily beachieved, low compared to photomultipliersC 107 but sufcient for our application, in whicha large amount of light is generated. We used twodifferent models of HPDs built by DEP [16]namely model PP0350F for the Front and modelPP0350D the for the Barrel detector. A detaileddescription of their characteristics can be foundelsewhere [12].

In order to test the effect of magnetic eld onHPD gain, we measured the pulse height produced

by the HPDs in the ATHENA magnetic eld usinga pulsed light emitting diode (LED), held in thecenter of the photocathode. Measurements withand without magnetic eld were consistent within3%, the difference being mainly due to LEDinstabilities.

3.3. Absolute beam calibration

An absolute calibration was performed byantiproton activation of an aluminum foil [17].

ARTICLE IN PRESS

Fig. 4. Signal on the central pad of the silicon beam counterregistered on an oscilloscope. The spill duration of around 200

ns is clearly visible.

6

4

2

0

I n t e g r a

t e d s

i g n a

l [ x 1 0 - 6 V s

]

100806040200

Bias Voltage [V]

Vfd

Fig. 5. Integrated signal registered on the central pad of thesilicon beamcounter as a function of the bias voltage of thediode. Vfd indicates the voltage that is required to fully depletethe diode.


8/10/2019 NIMA Paper2004

7/33

Antiprotons impinging on aluminium produce24Na (with a yield of 2 :17 0:3% ) which thendecays by g emission with energies of 1369 and

2754 keV [18]. The half life of 24

Na is 15 h : Wedumped the antiproton beam in an aluminumtarget (made of four foils with a thickness of 110 mm ; glued together) located at the entrance of the ATHENA beam pipe. Using the inducedactivity, measured off-line with a calibratedgermanium detector, an average beam intensityper shot of 1:207 0:19 107 %p was obtained [12].As an independent measure of beam intensity, thenumber of antiprotons in the AD ring wasmeasured before extraction by a low noiseSchottky pickup probe [19,20].

During the antiproton irradiation of the alumi-num foil the annihilation signals were detected bythe Front and Barrel external beam detectors andrecorded with ADCs. Using the afore mentionedknowledge of beam intensity the Front and Barrelexternal beam detectors were calibrated as re-ported in Table 1 . Here the measured ADC countsare converted to collected charge. While theSchottky probe gives the number of antiprotonsinside the AD ring with high accuracy C 1%; it ismeasured before electron cooling at 100 MeV =cand before extraction. Hence, possible coolinginefciencies and losses in the ejection line mightnot be taken into account. However, the resultsshown in Table 1 are in excellent agreement withthe activation method.

Fig. 6 shows the correlation between the chargemeasured with the silicon beam counter and thenumber of antiprotons determined with theexternal beam detectors. The former begins tolose linearity, due to saturation, starting at a uxof C 1 107 %p=pulse :

3.4. The external annihilation detectors

The solid angle coverage of the external

annihilation detectors is sufciently small (seebelow) to allow them to be used in singlecount mode, unlike the beam counters. Thismeans they can be used to detect single%p annihilations on the residual gas of the trapsor on the trap walls. The external annihilationdetectors consist of six coincidence pairs of Bicron BC408 plastic scintillators. Each scintil-lator is 10 mm thick, wrapped with aluminizedmylar sheet, glued on one side to a sh tailshaped lucite light guide and read out by a

Philips XP2020 photomultiplier. The horizontalscintillators are 720 300 mm 2 and thevertical scintillators 600 400 mm 2; covering asolid angle of B 30% 4p with slight variationsdepending on the exact position of annihilationinside the traps. The photomultipliers aredoubly screened from the magnetic eld with am-metal shield inside an iron housing andare mounted in a region where the stray magneticeld is small. They are biased to a negative voltageof (22.5) kV, the gain being adjusted by changing

ARTICLE IN PRESS

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

Calibration method Front Barrel

Activation 57 :57 9:5 27:27 4:5Schottky 55 :57 2:1 26:17 1:0

Fig. 6. Correlation between the number of antiprotons per ADpulse measured with the External Beam Detector and thecharge collected by the silicon beam counter.


8/10/2019 NIMA Paper2004

8/33

the bias voltage, while the appropriate thresholdsare determined using minimum ionizing particles,both from cosmic rays and from %p annihilation.

Pairwise coincidences allow operation at a lowerthreshold, reducing the random noise as well assuppressing neutral background from gammas andneutrons.

The discriminated signals are processed viaNIM logic and are recorded with a VME multi-scalar module (Struck Instruments System,SIS3806), which is read out via a MXI bus ontoa PC. The system allows dead-time free monitor-ing of trapped antiprotons with lifetimes longerthan several hours [21]. Typical background ratesare B 20 Hz per coincidence pair, but can bereduced by further requiring charged particlemultiplicities larger than or equal to two. Thesignals can also be used to trigger the antihydrogendetector readout (see Section 8).

The excellent correlation between the beamintensity measured by the external beam detectorand the number of annihilations of trappedantiprotons measured by the external annihilationdetector is illustrated in Fig. 7 . The details of antiproton trapping and cooling are discussed inSection 4.1.

4. The trap system

The trap system is realized by a sequence of

electrodes having 1 :25 cm inner radius and variouslengths. Different sections of this system are usedto perform different functions. Proceeding fromthe antiproton beam entrance and moving towardthe positron accumulator, the rst 12 electrodesare used to catch, cool and accumulate antiprotons(catching trap). The following group of electrodesis referred to as the mixing trap. This is the regionwhere the antiprotons are merged with thepositron plasma. The last group of electrodes isused during the positron transfer and re-captureprocedure.

All the electrodes outside the mixing region arecopper, plated with gold to avoid oxidation.Below, a 0.51 mm thick palladium sub-layer wasdeposited to improve plating performance andavoid the diffusion of gold. Static or varyingvoltages can be applied independently to eachelectrode allowing the electric eld inside the trapto be shaped according to the particular operationrequired.

4.1. The catching trap

The catching trap is composed of 12 cylindricalelectrodes ( Fig. 8 ). The two outermost electrodes(HVL and HVR) supply the high voltage to catchthe incoming AD antiproton bunches. The centralpart of the trap, comprised of 10 electrodes, is usedto conne the cold antiprotons. This sectionincludes seven electrodes designed to produce anharmonic potential (Penning trap) along the axisby an appropriate choice of the electrode lengthsand applied voltages [22]. One of the harmonic

region electrodes was split in four in order to allowthe application of the rotating wall technique forplasma compression [23] or, alternatively, to detectplasma diocotron modes [24].

4.2. The mixing trap

The mixing trap is the region where theinteraction between antiprotons and positronstakes place (see Fig. 9 ). The electrodes havebeen arranged in such a way that the nested

ARTICLE IN PRESS

0

500

1000

1500

2000

2500

3000

3500

0 5 10 15 20 25

Number of antiprotons (x 10 ) +6

A n n

i h i l a

t i o n

C o u n

t s

Fig. 7. Counts measured by the external annihilation detectorsupon the release of trapped antiprotons as a function of antiproton number in the incoming beam as measured by theexternal beam detectors.


8/10/2019 NIMA Paper2004

9/33

conguration [25] can be achieved. In this cong-uration particles of opposite charge can beconned simultaneously and made to interact.During the mixing of the two clouds recombina-tion can take place. In essence the mixing regionconsists of three sections marked RW, e W andLW in Fig. 9; in each of them it is possible to setup a harmonic potential well as described for thecatching trap. The positrons are normally connedin the central of these three traps e W: All theelectrodes of the mixing region are 3 :25 mm thickand are made of aluminum as they are located inthe antihydrogen production zone inside theannihilation detector. The use of aluminuminstead of copper, combined with the thinner wallthickness, makes it possible to signicantly reducemultiple scattering of the annihilation products ontheir passage to the detector, as well as avoidconversion of high energy photons to electron positron pairs, which if they occurred inside thedetector volume would lead to an increasedbackground. Outside the mixing zone, the electro-des are 4 mm thick. As in the catching trap, one of

the electrodes in the positron trap section has beensplit in four azimuthal quadrants to allowfor radial compression using the rotating walltechnique.

4.3. Additional electrodes

To increase the experimental exibility of thetraps a section made up of nine electrodes andapproximately 21 cm in total length was added on

the side of the mixing trap towards the positronaccumulator (R1-9 in Fig. 9 ). These electrodes areused during the transfer of positrons to the mixingtrap and can act as additional particle reservoirsfor positrons or electrons, thus perhaps in thefuture allowing transfer and stacking of positronsduring the recombination phase. At the end of themixing 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 betrapped directly in the mixing section and cooledby either electrons or perhaps directly by apositron cloud produced previously by stackingseveral accumulation cycles of positrons (seeSection 7). In this way high densities and particlenumbers of positrons could be utilized to furtheroptimize the antihydrogen formation process.

5. Antiproton capture, cooling and manipulation

The standard operating procedure for antipro-

tons in ATHENA consists of their capture in thecatching trap, cooling them by collisions with apreloaded cloud of electrons and transferring thecold antiprotons to the mixing trap. The lastaluminum degrading foil is mounted on the trapentrance electrode HVL. The potential of the exitelectrode (HVR) at the other end of the catchingtrap is initially set to V HV : Antiprotons traversingthe foil with axial energy lower than eV HV arereected from HVR and captured by applying avoltage V HV to HVL before they return there. The

ARTICLE IN PRESS

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


8/10/2019 NIMA Paper2004

10/33

antiproton arrival time is determined by the siliconbeam counter (see Section 3.1). The optimum trap

closing time, which depends on bunch durationand energy and on the trap length, was determinedexperimentally.

The number of captured antiprotons is deter-mined by the external annihilation detectors (seeSection 3.4) by lowering the voltage of theentrance electrode (with a 5 ms time constant),thus allowing the antiprotons to annihilate on thedegrader. The time distribution from the detectoris related to the total number of capturedantiprotons and their axial energy distribution.

Fig. 10 shows how the number of capturedantiprotons increases with applied trap potential.Typically around 10,000 antiprotons are capturedat 5 kV for an incident AD ux of 2 :5 107=pulse :Fig. 11 shows how the number of capturedantiprotons decreases when the trap closing timeis increased. The optimum closing time is 500 700 ns : Note the slow decrease in catchingefciency as a function of closing time, indicatingthat a signicant fraction of the antiprotons leavethe degrader with very low energy.

ARTICLE IN PRESS

H V R R

A n t

i p r o

t o n s

f r o m

C a t c h

i n g

T r a p

R W

L W

e W +

P o s i

t r o n s

f r o m

t h e a c c u m u l a t o r

F i g

. 9 .

T h e m i x i n g t r a p . A

n t i p r o t o n s f r o m t h e c a t c h i n g t r a p a r r i v e a t t h e l e f t a n d p o s i t r o n s f r o m t h e p o s i t r o n a c c u m u l a t o r ( S e c t i o n 7 ) a r r i v e a t t h e r i g h t . T h e n e s t e d

t r a p i s m a d e u p o f t h r e e s e c t i o n s e a c h w i t h t h e p o s s i b i l i t y o f m a k i n g a h a r m o n i c w e l l . T h e t w o o u t e r o n e s o f t h e s e ( R W a n d L W ) a r e i n t e n d e d f o r t h e a n t i p r o t o n s a n d t h e

c e n t r a l l a r g e r t r a p

e W i s i n t e n d e d

f o r t h e p o s i t r o n s . T h e p o s i t i o n o f t h e a n t i h y d r o g e n d e t e c t o r ( S e c t i o n 8 ) i s s h o w n . T h e a d d i t i o n a l h i g h v o l t a g e e l e c t r o d e

( H V R R ) i s

m a r k e d .

[kV]HVV0 1 2 3 4 5 6 7 8 9

C a

t c h i n g

E f f i c i e n c y

0

0.1

0.2

0.3

0.4

0.5

0.6

- 3

( x 1 0

)

Fig. 10. Dependence of the catching efciency on the appliedhigh voltage, V HV : The antiprotons were released from the trap1 s after capture. The numbers of captured antiprotons arenormalized to the beam intensity measured with HPD-basedexternal beam detectors.


8/10/2019 NIMA Paper2004

11/33

5.1. Antiproton cooling

Coulomb collisions between antiprotons andelectrons 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 inthe 3 Tesla magnetic eld with a time constant of about 0 :4 s: Ideally, the two species of particles willreach a nal equilibrium temperature equal to thatof the environment. The cooling process is usuallydescribed by the differential equations [27]

dT pdt

T p T e

t c

dT edt

npne

1t c

T p T e T e T t

t e

where T e and T p are the electron and %p tempera-tures, T t is the unperturbated electron tempera-ture, ne and np are the electron and %p densities, t e isthe synchrotron and t c the electron cooling time.The latter is given by

t c 3memp c3

82p1=2nee4 lnL

kT pmp c2

kT emec2

3=2

:

Here mp and me are the %p and electron masses, e istheir electrical charge and L is given by

L 4pe0

nekT e2

3=2

:

The solution of these equations shows that 10 4

antiprotons having energies in the keV range canbe cooled down to less than a few eV within a fewtenths of a second if they overlap completely withan electron cloud of density around 10 7 108 cm 3:

The ATHENA electron source consists of abarium oxide disc cathode (Kimball Physics Inc.)mounted on a movable support in the positrontransfer region (Section 7.3). The electrons areloaded in the %p catching trap before the arrival of the antiproton pulse. This is achieved by shaping anarrow, low-voltage potential well (typically a fewtens of volts) in the central region of the catchingtrap. The electron primary current traverses thecatching trap and reaches the entrance electrodeHVL where it may be dumped on the degrader foilor repelled by it depending on the electric eldconguration. Electrons can be loaded in differentsections of the catching trap by various proce-dures. They differ from each other by the values of the voltages applied to the trap electrodes during

the primary current passage and they lead toclouds having different initial shapes and densitiesand distinct evolution dynamics.

The cooling process is studied by dumping thehot antiprotons from the high voltage well andthen later the cold antiprotons cooled by theelectrons and captured in the narrow internalelectron well. Fig. 12 shows a typical timedistribution taken with the external annihilationdetectors for trapped antiprotons released andannihilating on the degrader at the entrance of the

trap. The electric potential was rst lowered from5 kV to 40 V (with a time constant of B 20 ms)releasing the higher energy antiprotons (HV dumpin the gure), and then from 40 to 0 V in about1 ms ; thus releasing the antiprotons which hadbeen cooled via interaction with preloaded elec-trons (small trap dump). Fig. 13 shows the fractionof cold and hot antiprotons as a function of interaction time. Nearly all antiprotons are cooledin about 60 s : At the end of the cooling process,antiprotons and electrons share the same volume.

ARTICLE IN PRESS

s]Closing Time [0 2 4 6 8 10 12 14 16 18 20 22

C a

t c h i n g

E f f i c i e n c y

10-6

10-5

10-4

10-3

Fig. 11. Dependence of the catching efciency on the trapclosing time delay.


8/10/2019 NIMA Paper2004

12/33

The electrons can then be ejected from the trap byapplying appropriate electric pulses of about100 ns duration which do not affect the heavierantiprotons.

Since the catching process does not inuence thepotentials within the central region where the coldantiprotons and the electrons are collected, severalAD shots can be stacked in the catching trap. Thisis illustrated in Fig. 14 which shows a linearincrease in the number of trapped cold antiprotonswith the number of AD shots.

5.2. Electron plasma characteristics

The high density and low temperature of theelectron cloud makes the Debye length shorterthan the trap cloud extension and means that the

electron cloud is well within the plasma regime.The evolution of the electron plasma can bestudied using destructive or non-destructive diag-nostic systems. The destructive diagnostic simplyconsists of dumping the electron cloud onto thedegrader mounted on HVL (see Fig. 8 ), which alsoacts as a Faraday cup, and reading the collectedcharge by means of a low-noise high-impedanceamplier. A nondestructive diagnostic based onmonitoring of the axial plasma modes is alsoimplemented, and will be discussed in Section 6.

The electron storage time is limited by collisionswith the residual gas, asymmetries and imperfec-tions in the trap geometry and in the elds, leadingto radial transport across the magnetic eld andexpansion of the plasma. This is a well documen-ted phenomenon in the eld of nonneutral plasmas[28]. In ATHENA, electron plasmas of few cm inlength with several 10 8 electrons are routinelyloaded in the catching trap. The electron storagetime depends on the trap length, the depth of thepotential well and the choice of electrodes. Typical

ARTICLE IN PRESS

Number of Shots0 1 2 3 4

A n

t i p r o

t o n s

0

5000

10000

15000

20000

25000

30000

35000

40000

Fig. 14. Dependence of the number of cold antiprotons on thenumber of stacked AD shots. Each AD shots contains about2 107 antiprotons.

20000 20100 20200 20300 204000

20

40

60

80

100

120

140

160

Time [ms]

HV dump

Small trap dump A n n i

h i l a t i o n c o u n t s

Fig. 12. A typical %p annihilation time spectrum measured bythe external annihilation detector (see text). The clock is startedat injection of the %p beam into the trap.

Cooling Time [s]0 10 20 30 40 50 60

N u m

b e r

p

N

o r m a

l i z e

d

0

0.2

0.4

0.6

0.8

1

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.


8/10/2019 NIMA Paper2004

13/33

electron storage times (the time necessary for theparticle number to reduce by a factor two ) rangefrom 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 havingejected them. The background pressure as mea-sured in the room temperature region of theapparatus external to the cryogenic region (seeSection 2.2) was about 10 11 mbar and the trap

temperature 15 K : We observed that the electronsreduce the cold antiproton storage time; withoutelectrons the half-life of the antiproton cloud isusually better than 10 h :

5.4. Antiproton transfer

Cold antiprotons are transferred toward themixing region by adiabatically moving the elec-trode voltages along the traps. The transferefciency in the case when electrons and anti-

protons are moved together are compared withthose when electrons are rst ejected from thecatching trap and only the cold antiprotons are

transferred. Transfer efciencies greater than 90 %are obtained when electrons and antiprotons aretransferred together while a dramatic decrease in

the transfer efciency has been observed if coldantiprotons are moved alone. In addition, whilethe number of cold antiprotons stored in thecatching trap linearly increases with the number of AD shots ( Fig. 14 ), under the experimentalconditions of 2002 the transfer efciency did notfollow the same behavior. Thus when stackingseveral shots in the catching trap prior to transfer,most of the transferred antiprotons (about 75%)originated from the last trapped bunch. Thisbehavior was unexpected and could be related topossible radial (centrifugal) [31] separation of antiprotons and electrons leading to instabilitiesduring the transfer. An improvement in theelectron loading and especially the antiprotontransfer procedure during the ongoing 2003 runhas solved this problem. The normal procedure inthe antihydrogen production runs in 2002 requiredthe stacking and transferring of 3 AD shotsresulting in about 10,000 cold antiprotons avail-able for recombination in the mixing trap.

6. Plasma modes diagnostic

The thermal equilibrium state of a large numberof positrons or electrons conned in a Penningtrap at low temperature is a rigidly rotatingspheroidal plasma [32] with a sharp boundary.Models predict that the density is almost constantwithin the ellipsoid and that it falls off exponen-tially with the Debye length at the plasmaboundary.

Knowledge of the characteristics of the plasma

(i.e. dimensions, density, temperature) can beobtained by means of a nondestructive methodbased on measurement of the rst two axialelectrostatic mode frequencies (dipole, quadru-pole) [33,34].

The modes are excited by applying sinusoidallytime-varying potentials to one trap electrode, whilethe plasma response can be measured by acquiringthe induced current on another electrode (seeFig. 16 ). The ratio of the induced current to theexcitation amplitude is measured as a function of

ARTICLE IN PRESS

Storage Time [s]0 200 400 600 800 1000 1200 1400 1600 1800

N u m

b e r

p

N o r m a

l i z e

d

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

- 4

( x 1 0

)

Fig. 15. Storage time for cold antiprotons with (squares) andwithout (circles) electrons. The antiproton numbers are normal-ized to the beam intensity measured with the HPD-basedexternal beam detectors.The lines are to guide the eye.


8/10/2019 NIMA Paper2004

14/33

the drive frequency. A narrow stepwise frequencysweep (4 ms duration per 5 kHz step) is madeacross the resonant frequency of each mode. For

each frequency step, the amplitude and phase(relative to that of the drive signal) of the voltageinduced by the plasma motion is acquired. Thisexcitedetect process is performed by means of anetwork analyzer (HP4395) integrated with suita-ble attenuation and amplifying circuits. The crosstalk signal between the transmitting and receivingelectrodes is acquired without positrons andsubtracted from the signal measured with theplasma present.

A detailed and simple analytic theory of theelectrostatic modes in nonneutral plasmas exists[33]. In the framework of this theory the frequen-cies of the rst two symmetric axial modes dependon the plasma size, density, and temperature. The

zero temperature model is used to determine theplasma 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 (seeFig. 17 ) can be modeled using a resonant circuitanalogy, where the values of the components arerelated to the plasma properties. The diagnosticalso allows, in addition to the previous para-meters, the plasma length 2z0 to be obtained. Theradius r0 and the particle number N are thendetermined and a complete nondestructive diag-nostic system is obtained. Temperature T shiftsproduced by the application of a radiofrequencysignal resonant on the (1,0) mode are monitored aschanges in the (2,0) mode frequency. The model[35] is used to calculate the induced temperatureincrease. The diagnostic system as well as thetemperature monitoring and control is described inmore detail elsewhere [36,37].

7. Positron accumulator

7.1. Overview and operation

The operation of the ATHENA positronaccumulator is based on the buffer gas captureand cooling of positrons in a PenningMalmbergtrap. The techniques used were pioneered by the

ARTICLE IN PRESS

f [MHz]20 20.05 20.1 20.15 20.2 20.25

A m p

l i t u d e

[ a r b

. u n

i t s

]

0

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0.016

Dipole Mode

f [MHz]31.96 32.2 32.44 32.68 32.92 33.16

A m p

l i t u d e

[ a r b .

u n

i t s

]

0

0.0005

0.001

0.0015

0.002

0.0025

0.003Quadrupole Mode

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

Fig. 16. Schematic diagram of a Penning trap with the modesanalysis and heating circuit.


8/10/2019 NIMA Paper2004

15/33

University of California San Diego positron [38 40] and electron [23,41,42] groups. As describedbelow, up to 2 108 positrons have been trappedin this accumulator prior to transferring themacross a low eld region to the main ATHENArecombination trap.

The accumulator traps and cools a continuousbeam of slow positrons. These are generated bymoderating b particles from a1:4 GBq 40 mCi 22Na radioactive source andguiding them into the trapping region using axialmagnetic eld transport. A cryogenic cold headcapable of reaching 5 :5 K cools down the sourceand makes it possible to grow a solid neonmoderator directly on the source [4345]. Detailsof the design of the source holder and its interfacewith the cryogenic section can be found elsewhere[4648] . Fig. 18 shows the trapping region. ANaI(Tl) detector is located close to a gate valvewhich can be used to isolate the source/moderatorend of the apparatus and the main trappingregion. This facilitates optimisation of moderatorgrowth, with the closed valve used as a simplepositron annihilation target. The absolute beam

intensity was obtained by cross-calibrating theNaI(Tl) detector and performing coincidencemeasurements with a channeltron detector whichwas periodically placed in the beam-line. Modera-tion efciencies (absolute beam intensity dividedby the total positron activity of the source) of around 0.4% are routinely achieved such thatbeam intensities greater than 5 106 positrons/sare available.

The trapping scheme utilizes nitrogen buffer gasto trap and cool the positrons. Initial trapping

occurs during the rst passage of the positronthrough the trap electrodes by electronic excitationof the nitrogen gas. Such a transition is favored innitrogen compared to positronium formation,which is the only other major inelastic channelopen at our kinetic energies. After trapping, axialconnement is provided by applying appropriateelectric potentials to the electrode array, whilst theradial connement is provided by a 0 :14 T axialmagnetic eld. Once trapped the positrons con-tinue to lose energy in collisions with the gas,nally residing in the potential well formed by thevoltages applied to the large diameter trap

electrodes.One of the trapping electrodes (see Fig. 18 ) is

split into six segments to compress the plasma byapplying a rotating electric eld (the so-calledrotating wall technique) [23,42]. In this techni-que the rotating electric eld transfers torque tothe plasma resulting in radial compression. Themethod was recently shown to work well forpositron plasmas [49,50]. In the present apparatusthe electrodes used have a signicantly largerradius C 10 cm than in these earlier experiments,

but results presented below show that the positronplasma can still be inuenced. Rotating wallcompression leads to heating of the plasma, andsince the magnetic eld in the trap is too low forefcient re-cooling by synchrotron radiation,another cooling mechanism has to be used. Wehave successfully used the nitrogen buffer gasalready present in the trap to provide this cooling,despite the fact that this gas has a poor positroncooling rate [39,51,52] . The presence of thesegmented electrodes in the accumulation trap

ARTICLE IN PRESS

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


8/10/2019 NIMA Paper2004

16/33

allow rotating wall compression during positronaccumulation. This reduces positron losses due tocross-eld transport in the presence of the buffer

gas, leading to a larger number of accumulatedpositrons.The detection system used to monitor the

performance of the positron accumulator hastwo main components: (i) a segmented Faradaycup detector consisting of nine plates to extractinformation on the size and position of the plasma.The Faraday cup has an area of 25 cm 2 and it issituated outside the main magnet in a region wherethe magnetic eld is about a quarter of that insidethe trap. This means that the plasma size obtaineddirectly from the Faraday cup is magniedcompared to its actual size inside the magnet andthis has been taken into account to derive the trueplasma dimensions; (ii) a calibrated CsI-photo-diode detector (see Section 7.4) to monitor theannihilation signal generated when the positronsstrike the Faraday cup.

After trapping, the positrons still need to betransferred to the main magnet containing themixing trap. A transfer section was constructed forthis purpose. This section consists of a vacuumseparation valve, a pumping restriction, a number

of transfer electrodes and a transfer magnetcapable of pulsing from 0 to 1 T in 20 ms andstaying at 1 T for 1 s : The performance of thissystem is reported below.

7.2. Optimization and results

An optimization program was undertaken totune the performance of the accumulator. Wevaried the electrode potentials and buffer gaspressures 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 10 8 positrons in a few minutes. Whenusing a suitable frequency and amplitude for therotating wall compression (see below in thissection) and applying this signal for the last 50%of the accumulation time, the lifetime of thepositrons can be doubled in the presence of thebuffer gas whilst maintaining the same accumula-tion rate. The data using the rotating wallcompression ( Fig. 19 ) was tted with a lifetime

of 200 s whilst the data without the rotating wallgave a lifetime of 95 s : It is important to stress thatthese results occur with the buffer gas still presentin the trap. The increase in lifetime with rotatingwall compression shows that annihilation on thegas is not the dominant loss. This points instead toplasma loss due to collision-induced cross-elddrift 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 appliedrotating electric eld. The CsI-photodiode detectorwas used to record the total number of storedpositrons, whilst the segmented Faraday cupmonitored their position. Plasma centering couldthen be observed by the signal increase on thecentral plates of the Faraday cup. The ratiosbetween the signals on the various Faraday cupplates yielded positional information and absoluteplasma sizes.

Fig. 20 (a) shows measured ratios betweensignals from the central region of the Faradaycup with and without rotating wall compression.The central region covers about 20% of the totalarea. The direction of rotation of the appliedelectric eld coincides with that of the natural~E E ~B B rotation of the plasma. Fig. 20 (b) showsthe corresponding total number of positronsobserved by the CsI photodiode detector.The data in Fig. 20 (a) exhibit a broad enhance-ment in the compression in the frequency range

ARTICLE IN PRESS

Accumulation time [sec]

0 200 400 600

A c c u m u

l a t e d p o s i

t r o n s

[ m i l l i o n s ]

0

50

100

150

200

Fig. 19. Accumulation of positrons with (closed circles) andwithout (open circles) rotating wall compression. The lines arets to the data (see text for details).


8/10/2019 NIMA Paper2004

17/33

300600 kHz increasing with amplitude. The factthat the ratio for the central region rises aboveunity means that parts of the positron plasmawhich initially missed the Faraday cup, e.g. due tocross-eld transport, have been compressed intothe central region. Above 600 kHz an abrupt fall-off occurs which coincides with a similar decreasein the total number of stored positrons. Up to600 kHz the total number of positrons is verystable 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 Faradaycups contained ve individual plates, of whichonly three actually recorded a signal when therotating wall was used. By examining e.g. ratiosbetween adjacent plates and comparing to similardata when the Faraday cup was moved slightly off axis, it was possible to derive the position and sizeof the plasma. The plasma had a width of 15 mm(FWHM) when no rotating wall was applied,reduced to 34 mm following compression. Sincethe total number of positrons stayed constant, or

in some cases increased, the central densityincreased by more than a factor of ten. Thiscompression ratio is much larger than thatreported previously for N 2 [50]. This can possiblybe attributed to the higher gas pressures used inthe present study.

7.3. Positron transfer

After accumulation the positrons are transferredto the mixing trap inside the main 3 T magnet. The

nitrogen buffer gas is pumped out and, after thepressure in the positron accumulator has fallenbelow 10 8 mbar ; the valve is opened and a pulsedtransfer magnet is energized for 1 s : This transfermagnet produces a eld of 1 T and thus helpsbridge the low-eld region between the positronaccumulator and the main ATHENA magnet (seeFig. 1 ) while also making it possible to separate thetwo vacuum systems with a pumping restriction.The positrons are released by lowering a gateelectrode with a fall time of 1 ms and trapped byclosing another gate electrode in the main magnet,

3:2 ms later. This traps the positrons initially in theentire length of the mixing trap and the adjacentpositron trapping section. The positron plasma isthen subsequently axially compressed into thecentral harmonic region of the mixing trap. Thiscompression takes a few tens of seconds. Theoverall efciency for transfer, recapture andcompression is about 50%. It should be notedthat this efciency was found to be mainly limitedby the electronics, particularly the closing time of the trapping electrode in the main magnet.

Efciency gains are anticipated when fasterelectronic units are installed. The present settingsallow the delivery of about 75 million positrons forrecombination every 5 min : The lifetime of thecompressed positrons in the mixing trap is quitelong as no signicant loss was observed during ahold period of 4000 s :

We conclude this section by remarking that theoperation of the positron accumulator, includingthe rotating wall compression, was found to bestable and reproducible over a wide range of

ARTICLE IN PRESS

0 200 400 600 800 1000 1200 1400

C e n

t r a l r e g

i o n

d i v i d e d

b y

t o t a l o n

F a r a d a y c u p s w

i t h o u

t R W

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0.1 V co-rotating0.2 V co-rotating0.5 V co-rotating

Frequency [kHz]

0 200 400 600 800 1000 1200 1400

M i l l i o n s o

f p o s i

t r o n s

10

20

30

40

50

60

70

80

(a) (b)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 wallcompression. (b) Corresponding signal from the CsI-photodiode detector.


8/10/2019 NIMA Paper2004

18/33

operating parameters. Thus the size, density andposition of the positron plasma can be predictedprior to transferring it to the main recombination

trap.

7.4. The CsI monitor

A calibrated CsI crystal with photodiode read-out measures the annihilation intensity when thepositron beam is dumped on the segmentedFaraday cup. The calibration was achieved usingthe g-lines at 511 keV and 1 :274 MeV from a weak22Na source of known activity. Since the 511 keVline from positron annihilation is used to deter-mine the number of positrons, contributions atlow energies from the 1 :274 MeV line (e.g.Compton scattering) need to be subtracted. Thiswas achieved using a calibrated 60Co source whichemits gs at 1.17 and 1 :33 MeV ; and henceproduces a spectrum in the 511 keV region similarto the 22Na 1 :274 MeV line.

The sources were placed in turn inside thepositron accumulator at the position of thesegmented Faraday cup so that conditions wereas close as possible to those experienced during theexperiment. In particular the gs produced a

spectrum similar to that of the Compton scatteredphotons from the annihilation of the positronplasma.

Two quantities were extracted from the back-ground-corrected spectra of the 511 keV line: (i)the average energy per photon and (ii) the overallefciency of the detector. The average energy perphoton is used to calculate the number of positrons annihilating on the Faraday cup duringa standard plasma dump. In its operating position,the CsI photodiode detector had an efciency of

0:0497 0:006% including solid angle acceptance.Two further CsI photodiode detectors were usedto monitor the transfer of the positron plasmafrom the positron accumulator to the mixing trap.One was situated at a place from when itparticularly observed any losses in the transfersection and the other was situated directly outsidethe main magnet (see Section 2.1) at the position of the degrader foil where the positron plasma wasdumped. These detectors were then calibrated by asimilar procedure to the main CsI diode detector

by placing sources at suitable positions inside theapparatus and measuring the resulting detectorresponse, thus again allowing for any material

dependent attenuation. These detectors made itpossible to optimize the positron transfer indepen-dent of the main ATHENA detector (see Section2.1) and provided an independent way of cross-checking the degrader Faraday cup calibration forpositrons.

8. The ATHENA antihydrogen detector

In a homogeneous magnetic eld electrically

neutral %

H atoms escape the connementregion and annihilate on the trap electrodesproducing on average about three chargedpions, three high energy gs and two 511 keV gs(Fig. 21 ).

The detector [53,54] was designed to allowextraction of a clean %H signal for backgroundrates of up to 10 kHz ; although the effectivebackground rate was later found to lie below100 Hz : Charged particles are detected in twolayers of Si- m-strip detectors covering roughly 80%of 4p : A three dimensional reconstruction of the %pannihilation vertex is achieved with s 4 mmspatial resolution by straight line extrapolations of

ARTICLE IN PRESS

Silicon microstrips

CsIcrystals

511keV

511keV

Fig. 21. Sketch of %H annihilation on the trap wall (not shown).Solid lines represent charged pions, wavy lines photons frompositron annihilation. The crystals hit by the charged pions or511 keV gamma photons are indicated. High energy gs fromp 0-decay are not shown.


8/10/2019 NIMA Paper2004

19/33

the charged particle tracks. Photons from positronannihilation 511 keV convert in the CsIcrystals via the photoelectric effect with a

probability of about 25%. The reconstructionefciency for antihydrogen events can, in principle,be increased if Compton scattering is alsoincluded. The segmentation into 192 crystalsis required to ensure high enough angular re-solution to verify that the two 511 keV photonsare emitted back-to-back. High granularity isalso imposed by background considerationsin %p annihilation. The annihilation releases2 GeV of center-of-mass energy which is distri-buted among the pions and later high energygs. With 192 crystals the probability of a crystalbeing hit by either a pion from the annihilationvertex or one of the high energy gs is about 5%/crystal. This background contains (spatially un-correlated) 511 keV photons produced by gshowers in the surrounding apparatus (e.g. magnetcoils). This background signal will always bepresent in our antihydrogen signal but it doesnot exhibit the angular correlation of 180 that thetwo 511 keV photons from the annihilation of antihydrogen produce.

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 fromthe cold 15 K trap region in the center of theexperiment. Low temperature is, in principle,advantageous for semiconductor detectors sinceparallel noise from leakage currents practicallyvanishes, but constraints increase on mechanics,cabling and electronics. It also prevents easy accessfor debugging in nal working conditions. Fig. 22shows the cylindrically symmetric design of thedetector operating in a vacuum of 10 7 mbar : Theoverall dimensions are 75 (140) mm inner (outer)diameter and 250 mm in length. The outerdiameter is limited by the cold bore of thesuperconducting solenoid. The inner diameter islimited by the size of the cryogenic coldnose UHVvessel containing the electrodes of the nestedPenning trap (see Section 2.1). The inner-mostpart of the detector of thickness less than 10 mmcontains two layers each of 16 double sided siliconm-strip modules, 162 mm long. Separation betweenthe modules and the 192 crystals is provided by the500 mm thin aluminum wall of the main support,

which was manufactured from a single piece of aluminum using electro-erosion. The crystals withdimensions 17 17:5 13 mm 3 are grouped in 16rows of 12, closely lling the remaining space.

The m-strip modules consist of two double sidedsensors (SINTEF, Norway, 81 :6 19 mm 2 with awafer thickness of 380 mm) glued onto a siliconmechanical support, and a multilayer ceramichybrid, 2 mm thick ( Fig. 23 ).

The sensor p-sides are segmented into 384 ACcoupled strips with an implant width of 32 mm and

a pitch of 46 :5 mm : Every third strip is read outwhile the two intermediate strips are oating. Thereadout strips (139 :5 mm pitch) are bonded to the

ARTICLE IN PRESS

12 crystals per row APDs Main support

Si strip detectors

Hybrids (Si strips)

PCB (APD's)

Crown

Crown support

25 cm

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

Fig. 23. Silicon m-strip module.


8/10/2019 NIMA Paper2004

20/33

8/10/2019 NIMA Paper2004

21/33

shielding) to a CF150 vacuum ange at roomtemperature housing ve 50-pin D-Sub feed-throughs.

Outside the vacuum vessel, ve repeater cardsare plugged directly into the ange, which alsoserves as the main grounding point. Three of theseboards, the digital repeater cards, house the mainelectronics for controlling the three subdetectors.They contain a low noise power supply 7 2 V;some logic circuitry (GALs) and level translators,16 voltage-DACs for chip settings, bipolar cali-bration pulse generators, the receiver for triggersignals (currents) and a programmable windowdiscriminator (trigger multiplicity). The cards areconnected by two digital busses to VME modules.A sequencer (Caen V551B) controls the readoutand ADC sampling by 8 differential TTL lines,while a 16 bit ECL digital I/O unit (Caen V262) isused for programming trigger masks, seriallyloaded DAC registers and other logic states of the detector, such as calibration and test mode.

Since no other active electronics could be placedin the cold section of the detector, the differentialcurrent output buffers of the chips drive themultiplexed analog signals over 1 :5 m long cablesof the same type as the digital lines. These signals

are fed into receivers on two analog repeater cardson the vacuum ange. One pair of lines is used foreach of the 256 channel m-strip modules, while the16 crystal PCBs share four lines. Differential linedrivers send the 36 analog signals to 18 doublechannel FADCs (Caen V550) located in the samecrate as the other VME modules. The crate isconnected via an MXI-2 bus (National Instru-ments) to a PC in the ATHENA control roomrunning LabView as DAQ software. Data (40kByte/event) can be written to disk with a rate of

typically 100 Hz ; mainly limited by the 5 MByte =stransfer rate of the VME-PC link.Shaping times for both detector components are

set to 3 ms; adapted to the time constant of lightemission for pure-CsI [55]. The read out controllerfreezes all 8384 analog channels simultaneouslyand provides a fully synchronized readout, eventby event. The total power consumption amountsto roughly 5 W : The multiplicity coded triggerlines from the individual subdetectors (one linefrom each subdetector) can, however, be recorded

independently, but have to be masked during thechip readout time 150 ms; due to digital crosstalk. The time jitter is typically 120 ns for the m-

strips 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 thedetector. The trigger signal is also needed tomonitor the high instantaneous annihilation ratesof antiprotons and positrons and for generaldiagnosis without full read out of the analogsignals.

The VA2 TA chip ( Fig. 26 ) is based on a seriesof CMOS VLSI chips (Viking [56], VA), alreadywell established in high energy physics applica-tions. A 32 channel version (VA32 TA32 set of separated analog and trigger chips) was rstsuccessfully tested at low temperatures.

A 128 channel chip with a medium scale inputstage (60 +11/pF e rms ), a standard shaper E 2 msand a second faster shaper E 75 ns ; was devel-oped for our application by Ideas ASA, Norway.The outputs of the fast shapers are fed to 128

discriminators (see Fig. 26 ) with common thresh-old and programmable transition polarity. Aregister is loaded to disable noisy channels. Powerconsumption, shaping times, threshold voltages oroutput signal heights are controlled by 12 DClevels. These settings are generated as voltages bythe DACs on the digital repeater cards and arecommon to all subdetectors. Fig. 27 illustrates forinstance the spread in shaping times for the crystalreadout, as a function of temperature. Optimizing

ARTICLE IN PRESS

Preamp

128 x 128 x M u

l t i p l e x e r

Input #1Slow shaper

Fast shaper Pulse height

discriminators

Serial outputHold in

Threshold

Trigger output

H(s)S/H

L o g

i c O R

VA2_TA block diagramH(s)

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


8/10/2019 NIMA Paper2004

22/33

settings lead to an average shaping time of 37 0:5 ms at 140 K :

Some of the chip settings, in particular the inputtransistor current of the preamplier, are quitesensitive to noise down to mV: Therefore ceramicdecoupling capacitors and RC lters are used onall DC lines of the hybrids (PCBs). Somecapacitors had to be oversized by up to 500% totake into account the strong ferroelectric behaviorof dielectrica under DC operation at low tempera-tures. 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 capacitorsbuffer the analog power.

Global grounding was optimized empirically byremoving or setting various ground bridges on keypositions of the detector. The two grounds of thedigital and analog parts in the chip were connecteddirectly to the ground plane on the hybrids andPCBs. Stable working conditions were observedwith no observable cross talk from ring triggersnor common mode noise.

Due to ohmic resistances in the service cables(the 250 shieldings are connected to a singleground) and the fully oating operation withrespect to the rest of the ATHENA apparatus,the detector ground may be as high as 50 mV :Sensitive voltages such as the trigger threshold arecorrected for this offset by voltage dividersreferencing to the hybrid (PCB) grounds. Variousperformance measurements were made on theVA2 TA chips. The trigger timing resolution atroom temperature from a small test detector

C D E 20 pF with respect to a plastic scintillatorcoupled to a photomultiplier is shown in Fig. 28(left). Fig. 28 (right) shows the output signal of thefast shaper (TA), recorded by scanning the trigger

delay over trigger threshold. A peaking time t p 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 modulesis about 40 on the strip and 50 on the pad side,respectively, measured with minimum ionizingparticles at low temperatures. The somewhathigher noise on the strips, in spite of the p-typeimplants, is caused by the large ratio of strip

implant to pitch widths, leading to a rather largeinterstrip capacitance of E 2 pF =cm : However, agood efciency for charge collection on the stripswas measured with cosmic rays and a highprecision s E 1 mm incident telescope [57]. Chargeloss on the oating strips was negligible. Cosmicdata could also be used to determine the spatialresolution on the strip side. Fig. 29 shows the Zdistribution, where Z is dened as the fractionalcharge collected on one of the two adjacentreadout strips Z QL =Q L QR : The two central

peaks are due to capacitive coupling of theintermediate oating strips to the readout strips.The differences between predicted hit positionsfrom the incident telescope and measured position,using the weighted average of adjacent strips, areplotted in Fig. 29 . We obtained a positionresolution of s 28 mm : However the much worseoverall vertex resolution of approximately4 mm 1 s in ATHENA is determined by theunknown particle track curvatures in the 3 Tmagnetic eld of the trap region. This is because

ARTICLE IN PRESS

106 Ru TA shaper output

Trigger delay [ns] Trigger delay [ns]

T r i g g e r t h r e s h o l

d [ m V ]

C o u n t s

= 12 ns

p = 120 ns

0

50

100150

200

250

300

350

0 50 100 150 200 100

30

25

20

1510

5

0200 300 400 500

Fig. 28. Timing resolution (left) and shaper output (right) of the VA2 TA chip.

0 20 40 60 80 100 120 140 160 1800

2

4

6

0 20 40 60 80 100 120 140 160 1800

2

4

6

0 20 40 60 80 100 120 140 160 1800

2

4

6

Crystal channel

S h a p i n g

t i m e s

[ s ]

T = 298 K

T = 172 K

T = 140 K

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


8/10/2019 NIMA Paper2004

23/33

we extrapolate the tracks of the charged pionsfrom only two points in the silicon m-strip detectorand with just these two points and no informationabout the charge of the detected pion, we have noway 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 electrodesof the positron transfer section (see Section 7.3),and to manage communication with the computersin charge of catching and cooling antiprotons andaccumulating positrons when these particles arebeing transferred into the mixing trap.

Voltages on the electrodes of the mixing trap aresupplied by programmable triggerable DACs(three 100 kHz 16-bit VME-based DACs (JoergerVDACM) with a buffer depth of 32k steps and one1 MHz 12 bit PCI-based National Instruments

DAC (6713) with a buffer depth of 16k steps).Voltages to the transfer electrodes, as well astriggers to the DACs, are supplied by two PCI-based 10 MHz programmable and triggerable 32-channel pulse generators (Becker & Hickl PPG-100) with a buffer depth of 64 kB : Additional fast(rise time o 100 ns) pulses of arbitrary shapes areprovided by four 100 MHz 12 bit VXI-basedwaveform generators (RACAL Instr. 3151), whichcan be connected to any of the 27 mixing trapelectrodes.

The interface to the control system is written inLabVIEW; in combination, the pulse generatorand the DACs are able to provide the mixing trapand transfer electrodes with voltage sequencesconsisting of several thousand steps and lastingseveral hours with microsecond precision. Inaddition, these sequences contain provisions forfurther activities such as removing electrons fromthe transferred antiprotons, controlling the statusof the electron source, and generic triggers andgates. Once created, these sequences are down-

loaded to the DACs and pulse generators, leavingthe time-critical execution under hardware control.The voltage range of the DACs and pulsegenerators is increased to 7 140 V by amplifyingall voltages with fast B 1 ms pre-ampliers. Allvoltages with the exception of those provided bythe RACAL DACs are ltered by a 10 MHz low-pass lter to reduce noise; each electrode has anadditional capacitively coupled unltered input forplasma diagnostics. This input is used on twoelectrodes to apply 40 V =100 ns pulses generated

by a 30 MHz GPIB-based pulser (SRS DS345) tomixed antiproton-electron clouds, in order toefciently clean the antiprotons of electrons (atleast 95% of the electrons are eliminated with fourpulses). Transfers of antiprotons and positrons aremanaged by setting up dedicated sequences of voltages which are triggered by the remotecomputers, once handshaking has been estab-lished, at which point sequences steering themixing of antiproton and positron clouds takeover. These procedures are fully automated, and

ARTICLE IN PRESS

= 28 m

-280 -140 1400

50

100

150

200

Position relative to expected [ m]

C o u n t s

0 0.2 0.4 0.6 0.8 10

5

10

15

20

25

30

35

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


8/10/2019 NIMA Paper2004

24/33

can repeat the same experiment an arbitrarynumber 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 voltagesof the electrodes, state of the apparatus, voltageand timing information from the control system,and inter-computer communications. In additionto the central detector readout (see Section 8),scintillator pulse heights (determined by LeCroy1182 ADCs), pulse shapes (recorded by INCAA

VD71 transient recorders), and oscilloscope sig-nals are acquired via VME and GPIB modulesreadout by a LabView program, and written to adatabase. The order in which the data are writtento the database does not necessarily correspond tothe order in which events occur, since severalmodules contain multi-event buffers, and thecontrol sequence for voltages and timings (whichis downloaded to the data base before beingexecuted) is not measured in real time. To retro-actively establish timing, the heart of the DAQ

consists of multi-hit time stamp units (one 8-channel 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 isguaranteed internally by the use of two counterbanks, one of which is active at any given time,while the data of the inactive bank is piped into a64 k FIFO. Switching between the two banks,which count the number of 100 ns intervalsbetween switches, is triggered by an OR of alltriggers. Triggers are generated by a 16-input

VME general purpose programmable logic module(GPPLM, trigger level 2) combining all detectorsignals; this module allows the simultaneousconstruction of a variety of trigger conditions.

This information is then used by the ofine dataanalysis program to establish trigger rates, corre-lations of one type of event (e.g. antiprotonannihilation rate) with another type (e.g. voltagechanges on the electrodes), time ordering of activities and detector dead times. Although theVME read out rate itself is bandwidth limited

ARTICLE IN PRESS

V D - 7

1 t r a n s

i e n t r e c .

S C - 1

2 p r e a m p

A D C 2 2 8 2 / N I - A D C

A D C 2 2 8 2 / N I - A D C

L e C r o y

1 1 7 6 T D C

d e g r a d e r

C s I

H P D s

B e a m c o u n

t e r

S c i n t

i l l a t o r s

V 5 5 0

V 5 5 1 B

J o e r g e r V

D A C M

R A C A L 3 1 5 1

S I S - 3

8 0 6

VME

VMEETHERNET

GPIB

VXI

>

>

>

>

DAQ

MIX TRAP PBARPOS RF

VME read outVME busy

data base

TriggerLevel 0

TriggerLevel 1

TriggerLevel 2

P r o g r . l

o g i c m o d u l e

S e q u e n c e r

P P G 1 0 0

P u l s e r

D S 3 4 5

N I - D A C

R e p e a

t e r c a r d

TASi Inner

TASi Outer

TACrystals

Scintil-lators

10 MHz

Trap movements

Electrode voltages

Detector readout

S e q u e n c e r

Fig. 30. Diagram of the data acquisition system.


8/10/2019 NIMA Paper2004

25/33

(corresponding to a detector read out rate of about100 Hz), trigger rates up to a few MHz are reliablyrecorded for 32 types of triggers in parallel, and

with a time resolution of 10 ms ; by the 32-channeltime-stamp unit. A higher time resolution of 1 ms isachieved for a further eight special interest triggerswith the 8-channel time-stamp unit.

10. The online software

The ATHENA online system was developed inC++ by using the ROOT package [58]. Data fromthe apparatus are analyzed in real time and savedon media by the DAQ program. As describedpreviously, data acquisition is performed in Lab-VIEW on a Windows NT machine and the dataare streamed to disk via a Dynamic Link Librarywritten in C. The data are saved in a raw le byusing a private binary format without anycompression to obtain the maximum possiblespeed 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 quickanalysis of the raw data during the runs. Becausedata are saved asynchronously as they arerecorded by the detectors, the online software usesa hardware time stamp to re-order the events in thecorrect time sequence, to correlate data from thetraps with data from the detectors.

The program also lls a set of preselectedhistograms used to monitor the experiment andsaves the data as a ROOT tree in a TOF le (TOF,

Time Ordered File). This ROOT le is compressedby a factor of about three with respect to the rawle. The results of the online analysis are alsoposted on the web for remote monitoring and arestored on tape at CERN via the CASTOR (CernAdvanced STORage) server.

The ROOT data tree is then passed to the ofinesoftware that extracts the detector events fromthe tree and performs the vertex reconstructionand the event analysis, as will be explained inSection 11.

Hence for each run three les are generated: (i)the raw data produced by the DAQ, (ii) the TOFle 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 inC++ and uses a set of ROOT macros. It decodesthe detector response, reconstructs the interactionpoints of the particles in the inner and externalsilicon layers, associates the tracks to these points(pattern recognition), nds the vertex of thecharged particles and selects the crystals with511 keV g-signals. A ROOT online monitor pro-vides a display of the reconstructed events. Somebatch macros give a set of histograms to controlboth vertex positions of the annihilations anddetector performances.

A Monte Carlo (MC) code has been written tostudy the optimum selection criteria for %H eventsand to evaluate the probability that a selected %H event may be a background event. The chain forprocessing MC events is identical to that for realdata.

The MC program was conceived as a C++interface between ROOT and the GEANT codefor the generation of simulated data from nuclearphysics experiments [59]. This interface allows thecall of the FORTRAN routines of GEANT usingROOT macros. The ATHENA apparatus geome-try was implemented in the MC code using thestandard GEANT routines and copied into ROOTusing a set of specialized routines. A phase spaceevent generator was written, with the possibility togenerate p %p and/or e e annihilation events, in an

uncorrelated or spacetime correlated way. Thep %p - mesons annihilation reactions considered,together with their branching ratio, are taken fromRef. [60] . It is possible to call for one channel inparticular (for example p %p - p p p p ) or togenerate an annihilation event according to itsbranching ratio.

It should be noted that in reality %p annihilationsoccur on nuclei of the trap electrodes or of theresidual gas lling the trap volume, rather than onfree protons. Although neutral pion multiplicities

ARTICLE IN PRESS


8/10/2019 NIMA Paper2004

26/33

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 themagnet surrounding the traps and the detector areprobably slightly overestimated in our MC.

Finally, the efciency of each one of the 32silicon modules and the energy response of eachone of the 192 crystals have been implemented intothe MC code, to fully simulate the performance of the real detector.

11.1. Point reconstruction

The hit coordinates of charged particles fromthe silicon m-strip modules are found by standard

clusterization algorithms applied to the pedestalsubtracted ADC values. The x ; y; z hit coordi-nates are calculated from the r f -strip and z-

pad information. Ghost points occurring whentwo or more particles hit the same microstripmodule are eliminated in two ways: (a) by keepingthe hit conguration that minimizes the vertexresiduals in three dimensions and (b) by associat-ing strips and pads that are correlated in amplitudewhen a module is hit by a single particle. Fig. 31shows the correlation between strip and pad ADCcontents. When a module has multiple hits, weconsider those points as ghosts that have an ADCcluster contents out of the band indicated in thegure. It has been our experience that the bestefciency is obtained by using method (a) whenthere are two hits on the module and method (b) inthe 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 distributionnormalized to the traversed silicon thickness withcosmics (a) and for a standard run with anti-protons annihilating on the walls of the recombi-nation trap (b). For cosmics good agreement isobtained with predictions assuming a Landau

distribution, whereas the Landau t to theannihilation data deviates at higher d E =dx be-cause of the presence of non-relativistic particles.Therefore, in Fig. 32 (b) the Landau t wasrestricted to the peak only (full curve). We recallthat annihilation products are mostly pions of 300 MeV =c average momentum [60], with some

ARTICLE IN PRESS

Fig. 32. ADC amplitudes per unit path length in silicon for cosmics (a) and for particles from %p annihilations (b). The measurementswere done using the strip side of the detector and for the cosmic data there was no magnetic eld. The full curves show the ttedLandau distrib

nima paper2004

Documents