DissertationIEKP-KA/2004-17

Part II of II

Development of a GEM-based TPC Readoutfor Future Collider Experiments

Steffen G. Kappler

Institut fur Experimentelle Kernphysik,Universitat Karlsruhe (TH), Germany

CERN, European Organization for Nuclear Research,Physics Department, Geneva, Switzerland

July 23rd, 2004

for my family

I

II

Contents

1 Introduction 1

2 TESLA - A concept for a Linear Collider project 3

2.1 Physics and requirements at the Linear Collider . . . . . . .. . . . . . . . . . . . . . . 3

2.2 The TESLA Collider . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . 4

2.3 The TESLA Detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . 4

2.4 The TESLA TPC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 6

3 The physics of Time Projection Chambers 9

3.1 Operation Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . 9

3.2 Ionization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . 10

3.3 Electron and ion transport . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . 11

3.4 Gas amplification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . 12

3.5 Counting gases and quenchers . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . 12

3.6 Spatial and momentum resolution . . . . . . . . . . . . . . . . . . . .. . . . . . . . . 13

3.7 Particle identification . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . 14

4 The GEM-technology 17

4.1 The Gas Electron Multiplier . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . 17

4.2 Operation principles of GEM-detectors . . . . . . . . . . . . . .. . . . . . . . . . . . . 18

4.3 The GEM-technology in TPCs . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . 19

4.3.1 Wire-based readout of TPCs . . . . . . . . . . . . . . . . . . . . . . .. . . . . 19

4.3.2 GEM-based readout of TPCs . . . . . . . . . . . . . . . . . . . . . . . .. . . . 20

4.3.3 Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 22

5 Charge carrier transfer in GEM-detectors 25

5.1 Quantification of transport properties . . . . . . . . . . . . . .. . . . . . . . . . . . . . 25

5.1.1 Single GEM-foils . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . 25

5.1.2 Multiple GEM-stages . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . 26

III

5.2 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . 27

5.3 Investigated GEM-geometries . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . 29

5.4 Measurements of electron transmission . . . . . . . . . . . . . .. . . . . . . . . . . . . 29

5.5 Measurements of ion transmission and ion feedback . . . . .. . . . . . . . . . . . . . . 31

5.6 Discussion of results . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . 32

6 Discharges and aging in GEM-detectors 35

6.1 Discharges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . 35

6.1.1 Basic mechanisms and precautions . . . . . . . . . . . . . . . . .. . . . . . . . 35

6.1.2 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . 36

6.1.3 Discharge measurements . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . 37

6.1.4 Discussion of results . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . 40

6.2 Detector aging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . 40

6.2.1 The COMPASS triple-GEM detectors . . . . . . . . . . . . . . . . .. . . . . . 41

6.2.2 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . 42

6.2.3 Aging measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . 44

6.2.4 Discussion of results . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . 47

7 The prototype GEM-TPC 49

7.1 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . 49

7.2 Current setup and detector validation . . . . . . . . . . . . . . .. . . . . . . . . . . . . 52

7.3 Front-end electronics and data acquisition . . . . . . . . . .. . . . . . . . . . . . . . . 53

8 Measurements in high-intensity particle beams 57

8.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . 57

8.2 Trigger and data taking . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . 60

8.3 Reconstruction technique . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . 60

8.3.1 Conventions and coordinate systems . . . . . . . . . . . . . . .. . . . . . . . . 60

8.3.2 Pedestal and noise determination . . . . . . . . . . . . . . . . .. . . . . . . . . 61

8.3.3 Track reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . 62

8.4 Analysis methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . 64

8.5 Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . 67

8.6 Discussion of results . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . 68

9 Summary 71

IV

Chapter 1

Introduction

The present status of the Standard Model could not have been achieved without experimental inputs fromboth hadron and electron collider experiments. If Higgs bosons indeed will be discovered at the LargeHadron Collider (LHC), the Higgs mechanism could be studiedin great detail at a future�� �� LinearCollider in a physics program complementary to that of the LHC. With a high luminosity over the wholerange of center-of-mass energies from 90 GeV to about 1 TeV, precise measurements of the importantquantities are possible, such as masses, couplings, branching ratios, which will be needed to reveal theorigin of electroweak symmetry breaking and to understand new physics. In the most likely scenariowith a light Higgs boson, the Linear Collider has the unique ability to perform a comprehensive set ofclean precision measurements and will allow an unequivocally establishment of theory. Linear Colliderconcepts have been pursued by several groups worldwide overthe past fifteen years, such as the JapanLinear Collider (JLC), the American Next Linear Collider (NLC) and the European project TeV-EnergySuperconducting Linear Accellerator (TESLA).

The physics program envisaged at the Linear Collider requires a high precision detector with a centraltracking system of excellent momentum resolution, good multi-track separation, and precise measure-ment of the specific ionization

�� ���for particle identification. A detector of choice for these purposes

is the Time Projection Chamber (TPC), as for instance foreseen in the Technical Design Report (TDR)of the TESLA project, a device which provides three-dimensional tracking with a minimum amount ofmaterial. In order to cope with the high track density environment present at the Linear Collider, a novelTPC readout based on micropattern gas detectors (MPGDs) is considered to replace the conventionalmultiwire solution. One of the most recent developments in the field of MPGDs is the Gas Electron Mul-tiplier (GEM), which offers narrow and fast signals, improving granularity and thus two-track resolution,an intrinsically suppressed ion feedback and almost no distortions due to

� � �effects.

After an introduction to the TESLA Linear Collider project and the physics of TPCs, this disserta-tion describes the research carried out on functioning and performance of GEM-based detectors. Theachieved insights in the fields of charge carrier transfer [1, 2], discharge behavior [3] and detector ag-ing [4] are summarized and transferred to the application ofthis technology in TPCs. Consecutively aprototype GEM-TPC is described, which was designed and constructed on the base of this knowledge [5].Tracking studies performed with this system in high-intensity hadronic particle beams are presented anddiscussed, and the results finally extrapolated to a large-scale system as required for the Linear Colliderproject [6].

1

2

Chapter 2

TESLA - A concept for a Linear Colliderproject

There is broad agreement in the High Energy Physics community that a Linear Collider for�� �� col-lisions with an initial center-of-mass energy of 350 to 500 GeV, complementary to the Large HadronCollider (LHC), is of fundamental importance for the futuredevelopment of particle physics. The gen-eral feasibility of a linear accelerator has been demonstrated by the successful operation of the SLACLinear Collider (SLC) at Stanford. Over the past fifteen years, several groups worldwide have beenpursuing different design concepts, such as theJapan Linear Collider[7], the AmericanNext Lin-ear Collider (NLC) [8–11] and the European projectTeV-Energy Superconducting Linear Accellerator(TESLA) [12–15], the latter of which will shortly be outlined in this chapter.

2.1 Physics and requirements at the Linear ColliderThe anticipated physics program at the Linear Collider encompasses a kinematic range of center-of-

mass energies from the� peak up to the TeV range as well as a large physics spectra fromdiscovery tohigh precision measurements. To exploit the potential of the Linear Collider, the detector design has tofulfill the following physics-motivated demands [14].

Analysis of di-lepton masses in the process�� �� � � � � � �� �� provides the means to analyzeHiggs production independent of its decay properties via the recoil mass to the di-lepton system. Thecombinatorial background occurring in the analysis of thischannel can be reduced significantly by requir-ing high precision of the measurements of invariant mass andrecoil for the�� �� system. This demandsexcellenttrack momentum resolution, achievable only by a large tracking volume and high magnetic field.The investigation of electroweak symmetry breaking includes a detailed study of the Higgs boson’s de-cay properties: distinguishing between a light Higgs bosondecaying into���, � ��, �� or � � � � representsa challenge for thevertex reconstruction. An extended Higgs sector will most likely be manifested viaproduction and decay of pairs of heavy Higgs particles in processes such as�� �� � � �� � � ���� or�� �� � � � � ������ discriminated from the multi-fermion background due to distinctive signatureswith multiple �-jets. Of course, Standard Model processes such as � � ���� �� � provide equallychallenging requirements for the tagging of� -leptons and� or � flavored jets. Most signatures of newphysics are expected from processes with complex hadronic final states, in which the intermediate de-cays (as � �� , � � � � or � � � ) must be reliably reconstructed in order to efficiently suppressStandard Model backgrounds. The effect of beamstrahlung and initial state radiation (ISR), the com-plexity of the signal final states and the presence of missingenergy in fusion processes and in reactionsinvolving supersymmetry particles reduce the applicability of kinematic constraints. Because of this,good knowledge about the kinematics of the involved partonsis indispensable, resulting in high require-ments oncalorimetryand tracking. It has turned out in past experiments, that multi-parton final states

3

electron sources

(HEP

and x-ray laser)

linearaccelerator

linear accelerator

X-ray laser

ee +

-inte

ractio

n p

oin

tH

EP

experim

ents

e+

sourc

e

aux.

&e

e+

-2

sourc

en

d

damping ringdamping ring

e+pre

accele

rato

r

e- e+

e-

33 km

Figure 2.1:Schematic view of the TESLA Linear Collider concept.

are best analyzed with the Energy Flow Technique, which combines the information from tracking andcalorimetry for an optimal particle flow estimate and thus the original parton fourmomenta. In addition,goodparticle identificationwill add valuable information to the event reconstruction.As missing energyis the main expected signature for the production and decay of supersymmetry particles and other pro-cesses of interest,hermeticityand particle detection capabilities at small angles are required. Excellentmissing energy resolution will increase the sensitivity tosupersymmetry in those cases with small massdifference between the lightest and the next-lightest supersymmetry particle. In addition, it is a signaturefor extra-dimension scenarios. Hermeticity requires goodcoverage of and measurement capability in theforward direction, furthermore being essential for a precise determination of the luminosity spectrum.

2.2 The TESLA ColliderThe TESLA Collider layout (see Fig. 2.1), which is describedin detail in reference [13], is designed

to provide�� �� collisions1 with center-of-mass energies up to 0.5 TeV (extendable to 0.8 TeV) in a 33km long linear accelerator based on superconducting cavities. With this technology, which allows longerpulses of higher intensities, the collider can be operated at a comparably low frequency2 of 1.3 GHz toprovide the design luminosity of� � ���� �cm�s. Its beam structure is characterized by five bunch trainsper second, each consisting of 2820 bunches arriving with bunch spacings of 337 ns; the beam profile atthe interaction point measures 553 nm in� and 5 nm in� direction.

2.3 The TESLA DetectorThe aforementioned physics requirements are addressed in the concept of the TESLA Detector, the

design of which is schematically shown in Fig. 2.2 and described in reference [15]. It provides an overallmomentum resolution of ��� � � � � ���� � ��GeV with less than 10�m systematic error and anenergy flow precision of� �� � ��� ��� �

GeV.

Themuon subsystem(MUON) represents the outermost part of the TESLA Detector and functions notonly for muon detection but also as a tail catcher calorimeter, supposed to provide an energy resolutionof � �� � ��� � ���� ��� �

GeV. It consists of Resistive Plate Chambers (RPCs) connected to 70kdiscrimination channels, 25k ADC channels and 5k TDC channels and covering an active area of about7000 m�.��

Alternative operation modes for�� collisions and�� collisions are also foreseen.��JLC and NLC would operate at 11.4 GHz due to their design basedon room-temperature cavities.

4

SITVTX/

TPC

ECAL

HCAL

COIL

YOKE

20001150

7400

207

4250

2832

2750

0

160320

1700

1908

2977

3850

4450

6450

7450

MUON

Figure 2.2:Schematic cut through one quarter of the TESLA Detector. The abbreviations for the individ-ual detector subsystems are described in the text, dimensions are in mm.

The next layer consists of the coil, which is based on a NbTi technology and provides a solenoidalmagnetic filed of 4T, with an uniformity better than���� . To achieve a maximum performance withthe Energy Flow technique, tracking and both calorimetry subsystems (electromagnetic and hadroniccalorimeter) are located inside the coil, keeping the amount of inactive material in between as low aspossible.

Both calorimetric detectorsfeature high three-dimensional granularity to resolve energy deposits ofclose particles, and to properly combine redundant measurements. The Hadronic Calorimeter (HCAL)has to provide an energy resolution of� �� � �� � ��� ��� �

GeV, a requirement which can pre-sumably be fulfilled by two options: an iron scintillating tile calorimeter with high transverse and lon-gitudinal segmentation or a fully digital calorimeter withimaging capabilities, where the active layersare gas detectors. For the Electromagnetic Calorimeter (ECAL) the competing technologies are a layoutbased on tungsten absorbers with silicon diode pads or a so-called Shashlik-layout. Both layouts can betuned to fulfill the requirement of� �� � �� � ��� ��� �

GeV energy resolution. In addition, twomore calorimetric subsystems are foreseen, a Low Angle Tagger for the 83.1 to 27.5 mrad coverage anda Luminosity Calorimeter providing a fast luminosity feedback veto at 4.6 to 27.5 mrad.

Thecentral tracking systemconsists of a large Time Projection Chamber (TPC) of 546 cm length and170 cm radius with about 200 readout points in the radial direction, a multi-layered pixel micro-vertexdetector (VTX) (between radii of 1.5 and 6 cm), an additionalsilicon tracking detector between VTXand TPC, consisting of cylinders in the barrel (SIT) and discs in the forward region and finally a preciseforward chamber located behind the TPC endplate. The TPC, which is summarized in more detail inthe following section, aims for a momentum resolution of ��� � � � � ���� � ��GeV and a

�� ���

5

beam pipe

25000

362

0

1618

inner field cage

outer field cage

ca

tho

de

me

mb

ran

e

MP

GD

s +

pa

ds

Ar-CH -CO (93:5:2)4 2

E B= 230 V/cm, = 4T

rea

do

ut

ele

ctr

on

ics

320

1700

2730

Figure 2.3:Schematic cut through one quarter of the TESLA TPC. Dimensions are in mm.

resolution better than 5%. The SIT will provide a point resolution of 10�m and thus improve ��� � by 30%. The VTX is designed with the innermost layer only 1.5 cm apart from the interaction point,thus providing impact parameter resolutions of � ���� � ��m � ���m

� � � ����� � GeV��.

2.4 The TESLA TPCThe Linear Collider physics program demands a central tracking system with excellent momentum

resolution, good multi-track separation, and reliable particle identification – all with a minimum amountof material. A detector of choice for these purposes is the TPC, which features not only high granularityfor tracking with a limited number of electronics channels and a minimum radiation length, but alsoallows for precise measurement of the specific ionization

�� ���.

Fig. 2.3 shows the design of the TESLA TPC, which has an activegas volume of 38 m� , i.e. 500 cmlength, 36.2 cm inner and 161.8 cm outer radius. The overall dimensions are 546 cm length, 32 cm innerand 170 cm outer radius, resulting in an estimated radiationlength of 3%

� , mainly resulting from the

material of inner and outer field cage. Assuming a readout paddesign3 with 2mm�

6mm pitch, alignedin 6 mm wide concentric rows, both endcaps will be equipped with in total about 1.2M readout channels.

At the present time, a gas-mixture of Ar-CH�-CO� (93:5:2) is the favored candidate for the TESLATPC. This gas features a drift velocity of 4.6 cm/�s at a drift field of 230 V/cm only, which is a convenientcompromise of 60 kV cathode voltage and 55�s clearing time; longitudinal (transverse) diffusion ina 4T magnetic field is as low as 300 (70)�m for 1cm. Simulations of the TPC with theBRAHMSsoftware [17] have shown, that with this gas a transverse single-point resolution of 70�m (190�m) for10 cm (200 cm) drift and a longitudinal resolution of 0.6 mm (1.0 mm) are feasible. Furthermore, a two-track resolution better than 2.3 mm (10 mm) in transverse (longitudinal) direction, a

�� ���resolution

of 4.3% and a momentum resolution of� �� � ���� � ��GeV (or � �� � ���� � ��GeV) for azimuthal angles� �� � � � ��� (or� �� � � � ��) could be demonstrated. The comparably low content of CH� is

chosen to minimize the cross-section for interactions withneutrons, an important background at theTESLA Collider [16], though the quenching properties with such a gas-mixture are not as good as e.g.��

A number of pad layouts is currently evaluated: rectangularpads, staggered rectangular pads or so-calledChevron pads are only a few examples.

6

with Ar-CH� (90:10). Because of its even lower neutron cross-section, the use of Ar-CF� (90:10) isfrequently proposed. But, due to hard-to-control aging andetching processes, which are to be wellbalanced and depend on gas flow and local irradiation fluxes, the application of CF� in general should bewell-considered.

Due to the bunch structure of the TESLA Collider, events froma number of bunch crossings willbe superimposed while the drift volume is read out, requiring a sufficiently high resolution for both,the transverse direction and for the projected coordinate.In order to cope with the resulting high trackdensity environment, a novel TPC readout based on micro-pattern gas detectors (MPGDs) is considered.Two of the most recent developments in the field of MPGDs are serious candidates for such a realization:Micromegas[18], realizable in TPCs as described in reference [15], andGEMs[19], which are subjectof this dissertation and shall be discussed in detail in the following chapters.

7

8

Chapter 3

The physics of Time Projection Chambers

This chapter provides a brief summary of operation principles and physics processes in Time ProjectionChambers. More detailed discussions on these topics can be found in various articles and text books, forinstance in references [20–23].

3.1 Operation PrinciplesTheTime Projection Chamber(TPC), as introduced in 1974 by D. Nygren [20], is a central tracking

device allowing three-dimensional track reconstruction and particle identification within one detector. Asillustrated in Fig. 3.1, it consists of a typically cylindrical, up to several m� large gas volume interspersedwith a parallel homogenous electric field�

�, which is created by a so-called field cage. The field cage is

composed of equidistant ring-electrodes, placed around the cylinder, and powered with constant voltage

track

track im

age

track im

age

track im

age

track im

age

track im

age

anode

TP

C r

eadout

mem

bra

ne

interaction point

outer field cage

inner field cage

E B||

TP

C r

eadout

anode

E B|| -

outer field cage

inner field cage

cath

ode

gas volume

beam pipe

Figure 3.1: Illustration of the operation principles of a TPC placed around the beam pipe of a colliderexperiment.

9

differences towards the cathode. In most TPCs, the cathode is realized as a metallized membrane, placedin the center of the gas volume and dividing the device into two halves of opposite electric fields. Onthe anode endcaps of the cylinder, still inside the gas volume, position sensitive gas detectors are placed.These TPC readouts typically are multiwire counters in combination with induction pads, placed in aplane underneath and arranged in rows parallel to the wires.

Ionizing particles traversing the gas volume create ”trackimages” of electron-ion-pairs. Exposed tothe uniform electric field, electrons and ions drift along the electric field lines but in opposite directionstowards anode and cathode, respectively. Whereas the slowly drifting ions are simply collected by thecathode, the electrons are multiplied and registered by thecontinuously operated TPC readout. Exploit-ing the two-dimensional information from the position sensitive gas detectors, the knowledge of the con-stant electron drift velocity and the time information of the individual readout cycles, a three-dimensionalimage of the particle trajectory can be reconstructed. For the measurement of particle momenta, a mag-netic field �

�is used, typically oriented parallel or anti-parallel to the electric field. This avoids Lorentz

forces to interfere the drifting electron swarms in direction of the electric field, and reduces transversediffusion – a convenient property largely improving spatial resolution.

Besides tracking capabilities, TPCs also offer the possibility of particle identification. Due to thehigh homogeneity of the gas volume, the energy loss per unit length

�� ���in the gas can be measured

precisely: if operated in proportional gas amplification mode, the signal from the position sensitivegas detectors is proportional to the number of electron-ionpairs from the ionization process and thusproportional to

�� ���. Combination of the energy loss measured along the particletrajectory and the

momentum measured with the magnetic field allows efficient particle identification, as discussed later insection 3.7.

3.2 IonizationWhen fast charged particles pass through gases, or matter ingeneral, both excitation and ionization

of molecules takes place along their paths. In the latter, electron-ion pairs are formed by either directinteraction with the incident particle, or as a result of secondary processes, in which an elevated amountof energy first is transferred to so-called

�-electrons.

The average differential energy loss�� ���

due to electromagnetic interactions in matter of moder-ately relativistic charged particles other than electronsis described by the Bethe-Bloch equation [22]:

����

� ������ ��� ��� � ��� � � � ��� �� �� ���� �� � �

� � � � � � � � � ��� � � (3.1)

where��

is Avogadro’s number,� � � �� are electron mass and radius,� is the speed of light,� is thecharge number of the incident particle,

its velocity (in terms of�), � �� are atomic and mass number

of the material,� its density,�

a density correction,� �� the maximum energy which can be impartedto a free electron in a single collision, and� the ionization constant which can be approximated by� � �� � � ��eV [23].

The minimum energy loss can be found around � � ���, almost independent of the medium. A

number of practical interest is the total number of electron-ion pairs created along a charged particle’strack, which can be calculated with help of the average energy loss of the incident particle per effectivelycreated electron-ion pair, the so called� -value1. For the total number

�of created electron-ion pairs

in a volume of thickness� we find� � ���� � � � �� ��

� � (3.2)

��Due to the contribution of excitation processes to the energy loss, the� -value is substantially greater than theionization energy of the molecules involved.

10

where the�� are weight fractions and� � the individual� -values of the components in a gas-mixture.Typical values range between ten and hundred electron-ion pairs per cm.

The created pairs are distributed over several clusters containing a varying number of pairs2. Thenumber of clusters per tack length� can be used to calculate the probability� for a charged particletraversing a gas volume of thickness� to cause at least one ionization:� � � � �� �� . Consequently, theminimum thickness of a gas volume required to reach 99% detection efficiency is

� ���.

3.3 Electron and ion transportThe electrons liberated in ionization processes quickly loose their energy in multiple collisions with

gas molecules, randomizing their motion and resulting in a thermal equilibrium with the gas. In presenceof an external electric field�

�an oriented motion is superimposed to the random thermal motion. The

average motion of a number of electrons is a drift of constantvelocity ��� parallel to the electric field

��

. Due to the random nature of collisions, the trajectories ofindividual electrons deviate from theaverage. An initially point-like localized cloud of electrons will diffuse during drift longitudinally andtransverse to the direction of the electric field. After drifting a distance

�� (for a time�), the transverse(longitudinal) distribution of the electrons is describedby a Gaussian with a standard deviation of

�� �� � � � � � �� � � � � � � �� � � ��� with� � �� � � � � � �� ���� (3.3)

where � �� � is the field dependent transverse (longitudinal) diffusioncoefficient, and the quantity� � �� �

can be understood as the transverse (longitudinal) diffusion for a unit drift distance, typically given in�m

��cm thus in�m for 1 cm drift distance.

Drift velocity and diffusion are also affected by the presence of a magnetic field��

. In the time be-tween two collisions a circular motion due to the Lorentz force is additionally superimposed. If the fieldsare parallel or anti-parallel oriented, as it is the case in aTPC, macroscopic drift direction and velocity aswell as longitudinal diffusion stay unchanged whereas transverse diffusion is reduced according to [22]:

� � � � � � � �� � (3.4)

where� is the characteristic time between two collisions and� � � � �� ��� � � the cyclotron frequency ofelectrons in the magnetic field. In the case of nonparallel fields, diffusion in all dimensions is affectedand the electron drift deviates from the electric field lineswith a lower effective drift velocity.

The motion of ions in the presence of electric and magnetic fields also is subject to drift and diffusion.Because of their significantly larger masses and higher interaction cross sections compared to electrons,the ion drift velocity is about three orders of magnitude smaller and diffusion is strongly suppressed. Forthis reason, ions tend to follow the electric field lines – even in the presence of a magnetic field. Theirdrift velocity can be described according to [23]:

�� �� � �� �

� � � ��

(3.5)

where� is the gas pressure,� the standard pressure and�� is the ion mobility, typically ranging from1.0 to � �� � ����m� /Vs.

��Depending on the gas-mixtrue, some 65 to 80% of the clusters contain only one electron and the probabilitythat a cluster contains more than five electrons is smaller than 10%.

11

3.4 Gas amplificationIn electric fields exceeding the range of about��� V/cm, electrons can receive enough energy between

two collisions, to excite and ionize further atoms, leadingto so-called gas avalanche multiplication. Onedistinguishes three characteristic regions of gas amplification: proportional region, limited proportionalregion and discharge region.

The proportional region covers a certain range of electric field strength, in which the number of newlyliberated electrons is proportional to the number of initially existing electrons (e.g. in the electric fieldclose to the wire of a proportional counter). Upon motion of� electrons along a range

��,�� new

electron-ion pairs are created [23]:

�� � � � � � �� (3.6)

where�

is the so-called first Townsend coefficient, which represents the inverse of the mean free pathfor ionization and depends on gas-mixture, temperature, pressure and the local strength of the electricfield. After proceeding from

� �to

�� the number electrons increases from� �

to � � according to

� �� � � � � (3.7)

where�

is the multiplication factor or gas gain, which can be calculated by:

� � �

���� � �

�� � ��(3.8)

The integral considers the dependence of the Townsend coefficient on the locally varying strength of theelectric field.

In the region of limited proportionality, which is entered with higher electric fields, the space chargecreated in the gas avalanche can become large enough to significantly deform and reduce the effectiveelectric field strength, and thus to saturate multiplication.

The upper limit for charge multiplication is given by the beginning of ultraviolet photon emissionprocesses, inducing avalanche spread over the whole gas volume and a discharge or spark breakdown isthe consequence. The Raether condition, in detail elaborated in reference [24], gives a phenomenologicallimit for charge multiplication: the maximum avalanche size must not exceed�� to �� electron-ionpairs; the exact number is strongly related to actual chargedensity and electric field strength. Sincethe energy distribution of the electrons is statistically distributed, operations at��� can already becomecritical.

3.5 Counting gases and quenchersGases of different chemical structures are used to perform different tasks in gaseous detectors. While

noble gases can only be excited through radiative absorption or emission of photons, weakly-boundpolyatomic molecules allow radiationless transitions dueto their abundance of vibrational and rotationalstates. The latter type of gases (”quencher”) dissipates a good part of the energy and protects the detectorfrom discharges, whereas the first type (”counting gas”) dominates and sustains the charge multiplicationprocess.

Classical counting gases used in TPCs are Ar or Ne together with quenchers as CH� or CO�. Fig. 3.2shows for instance the drift velocity of three mixtures of Ar, CH� and CO� as a function of the electricfield and illustrates the reduction of transverse diffusionin the presence of a magnetic field.

12

b)

100 1k 10k 100k0

200

400

600

800

1000

electric field [V/cm]

Ar-CH (90:10)4

B || E = 2.5T

Ar-CH (90:10)4

B || E = 5T

Ar-CH (90:10)4

tra

nsve

rse

diffu

sio

n [

m f

or

1cm

]m

a)

drift v

elo

city [

cm

/s]

m

Ar-CH

(90:10)

4

Ar-CO

(90:10)

2

Ar-CH

-CO

(90:5:5)

4

2

10 100 1k 10k 100k1

10

100

electric field [V/cm]

Figure 3.2: a)Drift velocity of three mixtures of Ar, CH� and CO� . b) Transverse diffusion Ar-CH� (90:10)for three values of magnetic field. The shaded areas correspond to typical values for drift fields in TPCs.Gas data computed with the GARFIELD interface to MAGBOLTZ [25].

3.6 Spatial and momentum resolutionThe accuracy of the reconstructed track images is determined by the single point spatial resolution� of a TPC, which in the first order is determined by the contributing number of effectively detectable

electrons��

from the ionization process and their spread due to diffusion when arriving at the anodeafter a drift distance

�� . The number��

is given by the ionization per sample length (normally one padlength

�) minus losses of electrons before gas amplification. Additional contributions originate from

pad geometry, their orientation with respect to the track (�: projected track angle), and other influencesas electronics noise, cross-talk amongst others, all summarized in��� �. If the pad pitch is adapted withrespect to the transverse charge spread, i.e. if more than one pad is hit per row, the single point spatialresolution of a pad row improves over the standard deviationof a rectangular uniform distribution3 andcan be expressed by:

� �� � � �� � ���� � ����� � � � � � �

�� � �� � ��� � �� (3.9)

Typical transverse spatial resolutions achievable with TPCs range from 100 to 200�m.

Transverse momenta� � of particles are determined by measurement of their trajectory curvaturesdue to the magnetic field. The momentum resolution�� �� � achievable with a TPC of outer radius�� ,inner radius�� and magnetic field

�is given by [26]:

�� � � � � � � �

� �� GeV � � � �� � �� � � �� (3.10)

where�� is the sagitta error, described by:

� �� � � �� ��� � �� � � (3.11)

��The standard deviation of a rectangular uniform distribution of width� – corresponding to the pad pitch – is���� �� � � � �� ��.

13

0.1 1 10 100

6

8

10

12

14

16

18a)energ

y loss

[keV

/cm

]dE

/dX

particle momentum [GeV/ ]c sample energy deposit [keV]

b)

0 5 10 15 20

Figure 3.3: a)Example scatter plot of �� ��� vs. particle momentum for reconstructed tracks in theOPAL experiment at CERN [27] and theoretically expected curves for different particle types. b) Examplehistogram of the energy deposit in individual, 1 cm long measurement samples.

Here, the contribution�� ��� due to limited transverse spatial resolution�� is given by the Glucksteinformula

�� ��� � ���

���� �� � (3.12)

with � �� the number of measured space-points used for track reconstruction. The contribution��� � dueto multiple scattering of the incident particle with gas molecules is described by the relation

�� � � � ���� � �

� �� � ����� � � � � �� �

� �� �� � ���

� � �� � �� ��

(3.13)

with the gas-mixture’s radiation length�.

3.7 Particle identificationThe average differential energy loss of charged particles in matter, as introduced in section 3.2, can

be used for identification of different particle types. Because of the relation

�� � � � � �� � � (3.14)

the differential energy loss�� ���

, represented by the Bethe-Bloch equation (3.1), is a universal func-tion of � �� and thus can be used to separate moderately relativistic particles of different masses. Fig.3.3.a illustrates the separation capability of this methodin a scatter plot of

�� ���versus particle mo-

mentum, both measured with reconstructed tracks in the OPALexperiment at CERN [27].

The measurement of�� ���

consists of a large number of charge measurements (typically a hundredto two hundred) along a particle track. Because of the proportionality of the gas detectors used, themeasured charge is proportional to the ionization, and thusto the energy deposit in the sample. Fig.

14

3.3.b shows an example histogram of the energy deposit in theindividual samples, a so-called Landaudistribution. The shape of this curve with a peak at lower anda pronounced tail to higher values (Landau-tail) reflects the already mentioned dual nature of the ionization process4. To determine the averagedifferential energy loss

�� ���of a track, the energy measurements

� � of the samples� are summed upand divided by the sum of the effective sample lengths

� �:����

����� � � ����� � � � (3.15)

A commonly used method to suppress the Landau-tails is the Truncated Mean Method. In this method,the samples are sorted according to their absolute difference to the mean measured energy,

� � � ��� � � � � with �� � ��

���� �� � (3.16)

and only a fixed fraction (typically 60 to 80%) of the samples with the smallest absolute difference� �is finally used for the

�� ���determination.

�� Electron-ion pairs are either formed in soft, direct collisions with low energy transfer or by means of�-electrons.

15

16

Chapter 4

The GEM-technology

This chapter presents the Gas Electron Multiplier, discusses the operation principle of detectors based onthis technology, and outlines its application in the readout of TPCs.

4.1 The Gas Electron MultiplierThe Gas Electron Multiplier(GEM), introduced in 1997 by F. Sauli [19], was initially invented as

a preamplification stage for micropattern gas detectors. Itconsists of a very thin, two-side copper-clad insulating polymer foil which is perforated with a highdensity of holes, as visible in the electronmicroscope photograph in Fig. 4.1. The holes are etched in a photolithographic process, developed by A.Gandi and R. De Oliveira, which can briefly be described as follows. The raw material, an adhesiveless,50 �m thick Kapton foil laminate with two 3 to 5�m thick layers of copper, is coated with a photo-sensitive film. Two identical masks containing the hole pattern are aligned with a microscope. Then,

70 mm

50 mm

50 mm

3 mm

140 mm t

pd

D

Dd

Figure 4.1:Electron microscope photographs and schematic views of a GEM-foil (here: standard geom-etry with �=140 �m, �=50 �m, �=70 �m and �=50 �m).

17

10k

ele

ctric

field

[V/c

m]

E = 2.5kV/cmD

E = 4.0kV/cmI

U =370VGEM

100k

-200 -100 0 100 2001k

z [ m]m

b)a)

ED

EI

IC

ITOP

IBOT

IA

cathode

anode

driftgap

inductiongap

UGEM

Figure 4.2:Operation principles and electric field configuration of the GEM (a)and electric field strengthon a vertical line through the center of a GEM-hole (b). (���� =370 V, ��=2.5 kV/cm and � �=4.0kV/cm.) Plots computed with the software packages MAXWELL [29] and GARFIELD [25].

the raw foil is slid in between and exposed to UV-light. Afterdevelopment of the photo-sensitive layer,a copper-etching process is carried out with conventional solvents. The next step consists of Kapton-etching, in which the two copper layers serve as masks, and during a final copper-etching process, theremaining edges are removed. A more detailed description ofthe production process is provided inreference [28].

As shown in Fig. 4.1, GEM-holes normally are characterized by a double-conical structure which isresult of the volume etching from each side of the foil. In theso-calledstandard geometry, the holesare arranged in a triangular pattern with 140�m pitch and the hole diameter reduces from 70�m at thecopper electrodes to 50�m in the center of the holes.

4.2 Operation principles of GEM-detectorsOn application of a potential difference between upper and lower GEM-electrode, a strong electric

dipole field builds up inside the holes. Fig. 4.2 illustratesthe electric field in a standard geometry GEMat 370 V with external fields of 2.5 kV/cm (drift field above theGEM) and 4.0 kV/cm (induction fieldbelow the GEM). In this configuration, the electric field strength in the hole center reaches 45 kV/cm.

When inserted in the drift field of a gas detector, electrons from the drift volume above are guidedinto the holes of a GEM-foil, where they are multiplied in a gas avalanche amplification process. Mostof the electrons are then released into the volume below the GEM-foil, where they can be collected byreadout electronics or, in order to achieve higher total gains, transferred to another GEM. During driftand gas amplification, the electrons are subject to diffusion, which causes a fractional amount to be lostto the GEM-electrodes or the Kapton walls.

A single GEM-foil can reach gas gains above��� . By stacking several GEMs in a cascade, so-calledmulti-GEM detectors with even higher total gas gains can be built (see Fig. 4.3). The gain obtained witha GEM-detector can be adapted to the needs of the particular application by choosing the number ofGEM-foils and the voltage across each one. The big advantageof this detector type is the separation ofgas amplification and readout stage, resulting in the fact that the readout signal is only due to motion andcollection of the electrons. This separation provides not only a margin of safety in case of dischargesoccurring in the GEM-foils, it also allows high flexibility in the geometry of the readout structure. The

18

2nd

GEM

1 GEMst

ED

EI

drift gap

induction gap

transfer gapET

cathode

anode

IC

ITOP, 1

IBOT, 1

IA

ITOP, 2

IBOT, 2

~ 3 mm

~ 2 mm

~ 2 mm

ioniz

ing p

artic

le

Figure 4.3:Schematic view of a multi-GEM detector with two GEM-foils (”double-GEM detector”).

electric fields in the typically some millimeters wide gaps between the GEM-foils normally amount 2to 5 kV/cm. In order to achieve efficient charge carrier transport, consecutive gaps should have slightlyincreasing values of electric field strength. Detailed studies of gain and charge sharing as a function ofdrift, transfer and induction field have been published earlier [30]. A typical configuration for a double-GEM detector in Ar-CO� (70:30) would be a drift field field of 2.5 kV/cm, a transfer field of 3.0 kV/cmand an induction field of 3.5 kV/cm, where at GEM-voltages of 420 V each, a gas gain of about� � ���would be reached.

An important issue in the discussion of gaseous detectors isthe clearing of the ions created in the gasavalanches (”secondary ions”). Due to their slow drift velocity, ions traversing a detector’s drift volumecan build up macroscopic space-charges. In the GEM-foil, not all secondary ions are emitted into thedrift volume: contrary to electrons, ions tend to follow theelectric field lines, and a good fraction isdirectly cleared by neighboring GEM-electrodes. But in addition, ions are transported to the surface ofthe insulating Kapton walls where they cause so-calledcharging-upeffects. By accumulation of thesesurface-charges, the neighboring electric field is modifieduntil an equilibrium state is reached. Themodification of the electric field also affects the gas gain ofthe GEM-foil. Thus, in the first seconds ofoperation, the gas gain changes slightly towards the final equilibrium-value. The reverse effect, neutral-ization of the deposited charges after switching off the device, happens due to the very high resistivity ofthe insulator on time scales of several hours to days [31].

4.3 The GEM-technology in TPCsDetectors based on the GEM have already found numerous applications in particle physics and in

other fields [31–33]. High rate capability, good localization accuracy and multi-track resolution, togetherwith robustness of operation make GEM-devices a good choicefor harsh radiation environments. Due tothese and the features discussed below, this technology is well-suited to find application in the readout ofTPCs, where it would replace the conventional solution based on wires and induction pads as illustratedin Fig. 4.4.

4.3.1 Wire-based readout of TPCs

In conventional TPCs, the electrons from the drift volume are multiplied by anode wires, whichefficiently exploit ionization statistics and provide excellent energy resolution. The electrons createdin gas amplification avalanches around the wires induce signals on pads which are placed in a planeunderneath. During the last few mm of their drift, when approaching the radial electric field aroundthe anode wires, the electrons encounter a region of nonparallel E and B fields and undergo deflectionsin direction of the wires, commonly called

� � �effect. The ions created in the avalanches drift away

19

GEM

GEM

anode plane with micro pads

2mm

2mm

2-5mm

cathode plane with pads

b)a)

anode wires

Figure 4.4:Schematic comparison of a conventional TPC readout based on wires and induction pads(a) and a GEM-based one with two gas amplification stages and micro pads for charge collection (b).Not shown here: ion-gates, placed in top of the wires and, if needed at all, in top of the first GEM.

from the wire, with an initially high drift velocity. When encountering the region of reduced electric fieldstrength, the ions slow down, and add a slow component to the signal, the so-calledion-tail. In addition,on their way through the sensitive volume, ions can build up macroscopic space-charges altering the driftof electrons from subsequent events. So-called ion-gates,mounted in top of the wires, can be used toefficiently reduce the back-flow with suppression factors down to ���� .

4.3.2 GEM-based readout of TPCs

In a TPC, two or more GEM-foils would be used for gas amplification and micro pads to collectthe electron signal. Simpler mechanics, resulting from thefact that no wires have to be tensed, provideincreased robustness, and due to two-dimensional symmetry, a GEM-readout offers more flexibility interms of pad geometry. Mainly resulting from charge collection instead of induction, GEM-signals arenarrow, fast, and in particular show no ion-tail. With adapted readout-electronics, i.e. shorter shapingtime and higher sampling frequency, granularity of the TPC and thus two-track resolution can be im-proved. In addition, as visible in Fig. 4.5, the electric fields above the GEMs as well as in the regionof gas amplification inside the holes are parallel to the magnetic field. The region of nonparallel fieldsrestricts to thin slices of about hundred�m above and below the GEM-foils, and distortions due to

� ��

effects are strongly reduced. Furthermore, due to the transition from the low drift field (some hundredV/cm) to the intermediate transfer field (some kV/cm) below the first GEM1, a big fraction of the ionscreated in the gas amplification process is guided to the upper GEM-electrode: only ions created in thecenter of the hole, close to the field lines which lead to the cathode (dark plotted lines), will be releasedto the drift volume. As it can be seen in Fig. 4.5.b, the fraction of field lines leading from the amplifica-tion region to the cathode decreases with increasing GEM-voltages, suggesting that at higher gas gains alower fraction of ions is released to the sensitive volume. Asimilar mechanism applies to ions releasedfrom subsequent multipliers approaching the GEM from the lower side: only ions following the fieldlines which lead to the volume above the GEM and do not end on a GEM-electrode will be releasedinto the sensitive volume. In standard geometry GEMs at TPC-like operation conditions, the fraction ofall field lines originating from the transfer region and leading to the sensitive volume equals for simpleelectrostatics reasons to the ratio of external fields

��:�� , as soon as the GEM-voltages exceeds about

ten Volts.

The above considerations on the ion flow are underlined by themeasurements shown in Fig. 4.6.a,

��The GEM facing the drift volume conventionally is referred to as the first GEM.

20

UGEM=200VUGEM=100V UGEM=350V

driftgap

transfergap

B || ED

B || ET

UGEM

b)a)

Figure 4.5: a)Operation principles and electric field configuration of the first GEM in a TPC at ���� =350V, ��=150 V/cm and � �=2.5 kV/cm. b) Electric field configuration of the same GEM at three differentGEM-voltages. One may naively infer that in these configurations all electrons from the drift region aretransferred through the GEM; this is not always the case due to the presence of transverse diffusion thatcan bring them to the electrodes or to the walls even against the field direction. Plots computed with thesoftware packages MAXWELL [29] and GARFIELD [25].

where the effective ion feedback of a double-GEM device operated in Ar-CO� (70:30) with transfer andinduction fields of 3.5 kV/cm is shown: depending on drift field and GEM-voltage, the effective ionfeedback ranges in this example with a two-stage device from��� down to �� . Making use of thecorresponding gas calibration curve from Fig. 4.6.b, the same data is plotted versus effective gain in Fig.4.6.c; one could naively deduce from these curves, that higher gains result in a lower number of ionsreleased to the drift volume. This is not necessarily the case, as evident in Fig. 4.6.d, where the ionback-flow2 is plotted versus gain: in this example, more than a hundred times the number of ions fromionization is additionally injected into the sensitive volume3. However, in case the intrinsic ion feedbacksuppression of a multi-GEM structure would not be sufficientfor a particular application, a gating gridcould be placed in top of the first GEM to block the remaining ion back-flow.

TPCs based on the GEM-technology principally can be operated with the same gases as wire-basedones. The choice of the gas-mixture anyway is driven by the drift properties of electrons in the sensi-tive volume. Ionization, drift velocity and diffusion, which determine efficiency, spatial resolution and�� ���

-measurement precision, have to be adapted to the experimental requirements. In addition, thegas mixture has to provide sufficient quenching, in order to avoid discharges, and should show low ten-dency to detector aging. Admixtures of CH� provide good quenching and a plateau of elevated driftvelocity at low electric fields, for practical reasons desirable in TPCs with long drift volumes because oflowered cathode voltages. Unfortunately, as discussed in chapter 6, CH� fosters detector aging processes.A candidate with excellent aging properties, good diffusion values, but less fast and worse quenching, isCO�. The use of these quenchers together with e.g. Ar as counting-gas are thinkable options. Besidesclassical TPC-gases as Ar-CH� (90:10) or (95:5), a mixture of Ar-CH�-CO� (93:5:2) is a good candidatefor a future use in the Linear Collider TPC.

��The ion back-flow is the number of ions released from the multi-GEM into the drift volume upon multiplicationof one electron.

��Please notice, that compared to a wire-based TPC readout, this number is still low.

21

103

104

effective gas gain

0.1

1

10

100

E = 150 V/cmD

E = 416 V/cmD

eff. io

n feedback [%

]b)

370 380 390 400 410 420 430 440 450 46010

3

104

105

eff.gas

gain

voltage per GEM [V]

E = 150 V/cmD

E = 416 V/cmD

a)

10

100

1000io

nback

-flo

w

effective gas gain

103

104

E = 150 V/cmD

E = 416 V/cmD

370 380 390 400 410 420 430 440 450 460

voltage per GEM [V]

0.1

1

10

100

eff. io

n feedback [%

]

E = 150 V/cmD

E = 416 V/cmD

d)c)

Figure 4.6:Measurement of the effective ion feedback of a double-GEM detector in Ar-CO� (70:30) with� � � � � � � �� kV/cm (a)and the corresponding gas gain (b) as a function of voltage per GEM. Effectiveion feedback (c) and resulting ion back-flow (d) plotted vs. effective gas gain.

4.3.3 Challenges

Precise measurement of charged particle’s momenta requires efficient and accurate reconstructionof their ionization tracks. Spatial resolution in TPCs is mainly determined and unavoidably limited bydiffusion of electrons according to their drift distances.Apart from that, it is a function of effective ion-ization statistics, and thus depending on linearity and strength of charge multiplication and on losses ofelectrons before charge multiplication. The latter also plays a key role in tracks’

�� ���-measurements

for particle identification, and for both reasons has to be kept as low as possible.

At the same time, a uniform response (in space and time) over the whole sensitive area has to beguaranteed during TPC operation. In triple-GEM detectors local gain variations of about 8% (RMS)have been observed [31], which can be imputed to the cumulative effect of variations in the GEM-holediameters. In addition, GEM-detectors are expected to experience a quick gain increase up to 30% dueto charging-up at the onset of irradiation. Although both effects can be corrected for, the latter effect ismore subtle and depends e.g. on the local radiation flux.

22

Another central question is, whether a multi-GEM stage can achieve sufficient intrinsic ion feedbacksuppression or if an ion-gate has to be implemented. The space-charge built up in the sensitive volumedepends on gas gain and ion feedback from the multi-GEM, on ion drift velocity and thus electric fieldstrength in the drift region, as well as on outer conditions as structure and intensity of the beam.

Furthermore, reliable long-term operation in an experimental environment implicates a prior opti-mization in terms of discharges and detector aging, both impacting the TPC design (number of GEM-stages, segmentation of electrodes, choice of materials etc.).

In the following chapters, general measurements on charge carrier transfer, discharges and aging inGEM-detectors are presented. With special respect to the results of these studies, a prototype GEM-TPChas been designed and constructed. Finally, measurements with this detector during tracking studies inhigh intensity particle beams are presented and the performance of TPCs based on GEM-technologydemonstrated.

23

24

Chapter 5

Charge carrier transfer in GEM-detectors

In order to optimize efficiency and spatial resolution and tomake use of the particle identification ca-pability of a TPC via

�� ���-measurement, ionization statistics has to be exploited and electron losses

have to be kept as small as possible.

The studies presented in this chapter focused on the investigation of charge carrier transfer mech-anisms in the GEM, and in particular on the determination of electron and ion transmission in GEMsof various geometries [1, 2]. The method used consists in measuring the transport currents at gas gainsclose to unity, with consecutive extrapolation to higher gains. The fractional ion feedback can be deducedfrom these results in a wide range of external fields and GEM-voltages. Making use of the data, one canestimate the charge carrier transfer properties of a detector with multiple cascaded GEM-foils.

5.1 Quantification of transport properties

5.1.1 Single GEM-foils

Transport and multiplication of electrons in the GEM conventionally are quantified bytransparency� , gas gain� andextraction�, which are defined for a GEM-stage

�as

�� � number of electrons reaching gas amplificationnumber of electrons entering the GEM-stage

(5.1)

�� � number of electrons leaving gas amplificationnumber of electrons reaching gas amplification

(5.2)

��� number of electrons leaving the GEM-stage

number of electrons leaving gas amplification(5.3)

and depend on gas mixture, hole geometry and electric field configuration. In gas amplification mode(� � �), it is a priori – due to cancellations of electron and ion currents on the GEM-electrodes –impossible to disentangle� , � and �. A quantity which can be determined directly is theeffective gasgain

��� �� � �� � �� � �� � �� � ��

�� (5.4)

where�� is the anode current and�� is the current from the ionization process.

�� � �� � �� (5.5)

25

is the so-calledtransmission. This quantity is most meaningful in the case of unit gain, where it simplyreflects the ratio

�� ��� � � number of electrons leaving the GEM-stagenumber of electrons entering the GEM-stage

(5.6)

and directly indicates electron losses. In this regime,electron transmissioncan be determined by mea-surement of the electrode currents:

�� ��� � � ���� �� ���� �� (5.7)

The other way round, theion transmission, which is defined in the case without multiplication as

� �� � number of ions leaving the GEM-stagenumber of ions entering the GEM-stage

(5.8)

can be determined by

� � � ������ �� �� �� (5.9)

with the same setup and reversed electric fields. To characterize the ion back-flow of a GEM-foil inamplification mode, we define the(fractional) ion feedback

��� number of ions leaving the GEM-stage

number of electron-ion pairs created in the GEM-stage(5.10)

and the quantityeffective (fractional) ion feedback

��� number of ions leaving the GEM-stage

number of electrons leaving the GEM-stage

� ��

��(5.11)

which can be determined approximately via

���� � � �� � ��

� � � � � � ��� � ��� � �

� � ����

� �if

�� � � � � (5.12)

with cathode current�� and anode current�� , or exactly via

��� �� � ��

�� (5.13)

in case the ionization current�� is known.

5.1.2 Multiple GEM-stages

Speaking of multi-GEM devices in gas amplification mode, thetotal effective gas gain�

is the prod-uct of the individual effective gains of the� GEM-stages:

� ���

� � ��� (5.14)

It is important to understand, that the last GEM-stage of thecascade normally creates the dominantamount of charge: From

�electrons,

26

� the 1st GEM creates� � � � � � � � � � � � � � � � � � � � � �� � � ��

,

� the 2nd GEM creates��� � � � � � � �� � � �

� � � � � � � � � � � � � � �� � ��� �� � � � � � � �

,

� the 3rd GEM creates��� �

� � �� � �� � � �� � � ��� � � � � � � � ��

,

� and thus the� th GEM creates

��� ���

� ��� � � � electrons and ions.

The number of ions�

released by a multi-GEM structure of� stages into the drift volume is given by

� ���

� � ��� �� � �

� ��

��� � � �� �

���� �

�� � ���

� ���� �

� � � � �� �

���� � � �� � �

���� �

� ��� � ���

�� �� � � � �

� ��

��� � � �� � (5.15)

and the total effective (fractional) ion feedback defined as

� � number of secondary ions reaching the drift volumenumber of electrons leaving the last GEM-stage

(5.16)

can be calculated via

� � �� � �� � �

� � � �� � �� ��� � � � � � �� � ��

�� � � �� �� �

�� � ��

(5.17)

from the curves�� �� and� �� �� for the different GEM-stages. It can also be determined in mea-

surements with

� � �� � ���� � ��

�� if �� �� (5.18)

according to the case of single GEM-foils, see equations (5.12) and (5.13). The resultingion back-flowof a GEM-device, defined as the number of ejected ions per electron, is given by

� �� � � � � (5.19)

i.e. by the product of effective gas gain and effective ion feedback.

5.2 Experimental setupThe measurements described in the following sections have been performed with small size GEM-

foils of �� � �� cm� active area, framed and mounted in versatile detector assemblies. A multi-framegas containment box permits mounting of one or more GEMs on top of a readout circuit board, andpreceded by a drift electrode and a thin window. Distances between electrodes can be varied, makinguse of insulating spacers. For signal readout, a board with parallel micro strips, 150�m wide at 200�mpitch, connected in groups by wire bonding to an external grounding circuit, was used. The detector wasoperated with an open gas flow, and the gas mixture monitored with external flow meters.

To measure electron transmission in GEM-foils, a single-GEM setup as schematically shown in Fig.5.1.a with a 10 mm gap to the cathode and a 2.5 mm gap to the anodewas used. To avoid current can-celations induced by overlapping ionization currents produced in several gaps, the drift region between

27

transm

issio

ngain

x

0 50 100 150 200 250-200

-150

-100

-50

0

50

100

150

200

anode

GEM, bottom

GEM, top

sum

ele

ctro

de

curr

ent

GEM-voltage [V]

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

[-n

A]

b)

cathode

anode

GEM

IC

ITOP

IBOT

IA

X-raysED

EI

driftgap

inductiongap

a)

IC

window

Figure 5.1: a)Schematics of the current measurement with the single-GEM detector. By reversing theelectric fields, one can measure transfer properties both of ions and electrons. b) Example measurementof the electron transmission with a standard GEM in Ar-CO� (70:30) at ��=150 V/cm and � �=2.5 kV/cm.The cathode current �� , which is not shown in this plot, equals due to charge conservation the negativevalue of the sum �� � ��� � � �� �� � ��� .

50 mm

5 mm

120 mm

70 mm

50 mm

5 mm 75 mm

65 mm

b)a)

Figure 5.2:Electron microscope photographs of GEM-foils in conical (a) and cylindrical (b) geometry.

cathode and GEM was irradiated with a constant, high flux of 5.9keV X-rays parallel to the electrodes1.The resulting currents in the range of tens to hundreds nA were measured on the two GEM-electrodes(�� �� and ���� ) and the anode (�� ) in dependence of the GEM-voltage. The current from the driftcathode (�� ) could not be measured directly due to the contribution of currents induced by ionizationproduced in the gap between the drift electrode and the detector window, resulting from the angulardivergence of the collimated X-ray beam.

From the example measurement shown in Fig. 5.1.b it can be seen, that at GEM-voltages below 0 Vall electrons are collected by the top GEM-electrode. With increasing voltage, more and more electronsare collected on the anode – some also on the bottom GEM-electrode. Between 75 V and 100 V theplateau of maximum transmission is reached before gas amplification starts to set in above 100 V, asindicated by the appearance of a positive ion current on the top GEM-electrode.

By reversing the electric fields, one can use the identical setup to measure the ion transmission.Please notice, that in this case the naming convention for electrodes and gaps has to change accordingly��

Please notice, that irradiating the detector perpendicular to the electrodes results in instabilities presumably dueto surface charging of the insulator exposed to the beam [35].

28

a)

ele

ctr

on tra

nsm

issio

ngain

x

0 25 50 75 100 125 150 175 200 225 2500.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

CO (>99%)2

Ar-CO (70:30)2

Ar-CO (90:10)2

Ar-CH (90:10)4

Ar (>99%)

GEM-voltage [V]

b)

100k

30k

10k

3k

1k

300

100

V/cm

Figure 5.3: a)Electron transmission vs. GEM-voltage with a standard GEM at �� � 150 V/cm and � � �2.5 kV/cm in different gas mixtures. b) Illustration of electron losses due to diffusion.

(�� � �� , �� �� � ���� and��

�� � ).

5.3 Investigated GEM-geometriesThe performance of the following three basic GEM-geometries has been investigated. Thestandard

geometry, as introduced in section 4.1 (Fig. 4.1), is characterized by a double-conical hole shape. Somederivatives of this geometry were produced: with similar geometry but on a thinner foil ( � ���m) orwith wider holes ( � ����m,

� � ���m) or with wider holes and pitch ( � ����m,� � ����m,� � ����m).

The conical geometry(Fig. 5.2.a) is produced with a single-side etching process[34]. The holediameters are � � ����m, �

� ���m and� � ���m with � � ����m pitch. In this case, two

orientations with respect to the charge drift direction arepossible, named ”narrow to wide” (N-W) and”wide to narrow” (W-N), respectively.

Thecylindrical geometry(Fig. 5.2.b) is obtained making use of a longer Kapton etching time than forthe standard geometry. Hole diameter and pitch are � ���m,� � ����m. Possible over-etching withpartial removal of the Kapton under the copper layer can hardly be discovered during production time,resulting in lower production yields and qualities of thesefoils. Indeed, it has been observed that GEMsof this geometry intrinsically are weaker in holding the voltage.

5.4 Measurements of electron transmissionFig. 5.3 shows the electron transmission in a GEM of standardgeometry for several gas mixtures

as a function of the applied GEM-voltage. The drift and transfer fields are 150 V/cm and 2.5 kV/cm,respectively; the comparably low value of drift field corresponds to a possible choice for a TPC appli-cation. In mixtures with smaller amounts of molecular quenchers, one observes more electron lossesbefore multiplication, which appears in the graphs as a fastexponential raise. This observation is consis-tent with the assumption that losses are due to transverse diffusion exceeding the hole diameter, in whichcase electrons are lost to the GEM-electrodes and Kapton walls, as illustrated in Fig. 5.3.b, even if noelectric field lines are ending there. In pure CO� diffusion is small and the transmission close to 100%,confirming the correctness of the measurement procedure.

29

a) b)

100 1k 10k 100k0

100

200

300

400

500

600

700

800

900

1000

electric field [V/cm]

CO (>99%)2

Ar-CO (70:30)2

Ar-CO (90:10)2

Ar-CH (90:10)4

Ar (>99%)

drift

gain

1

norm

al m

ode

transvers

ediffu

sio

n[

m for

1cm

]m

100 1k 10k 100k0

200

400

600

800

1000

electric field [V/cm]

Ar-CH (90:10)4

B || E = 2.5T

Ar-CH (90:10)4

B || E = 5T

Ar-CH (90:10)4

drift

gain

1

norm

al m

ode

transvers

ediffu

sio

n[

m for

1cm

]m

Figure 5.4:Transverse diffusion in selected gases without (a) and with (b) magnetic field as a functionof the electric field. The shaded areas correspond to typical field strengths in the drift region of a TPC,inside the GEM-hole at unit gain (���� =100 V) and in normal amplification mode (���� =350 V). Gasdata computed with the GARFIELD interface to MAGBOLTZ [25].

0 25 50 75 100 125 150 175 200 225 2500.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

ED=75V/cm

ED=150V/cm

ED=300V/cm

ED=600V/cm

GEM-voltage [V]

ele

ctr

on tra

nsm

issio

ngain

x

0 25 50 75 100 125 150 175 200 225 2500.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

EI=1.0kV/cm

EI=2.0kV/cm

EI=3.0kV/cm

GEM-voltage [V]

ele

ctr

on tra

nsm

issio

ngain

x

a) b)

Figure 5.5:Electron transmission vs. GEM-voltage with a standard GEM in Ar-CO� (90:10) (a) for differ-ent drift fields �� at � � � 2.5 kV/cm. (b) for different induction fields � � at �� � 150 V/cm.

The subsequent measurements have been performed with a gas mixture of Ar-CO� (90:10). As evi-dent from Fig. 5.4, this gas mixture shows for electric fieldsexceeding 1 kV/cm without magnetic fieldsimilar transverse diffusion as the typical TPC-gas Ar-CH� (90:10) in a magnetic field of 5 T.

From Fig. 5.5 the influence of the electric field configurationon the electron transmission can bededuced: electron transmission is improved at smaller values of the drift field or higher values of theinduction field. The transmission depends on the ratio

��of drift field and GEM-field2, as well as on

��The GEM-field (or better: the electric field in the center of the GEM-hole) is mainly determined by the GEM-voltage; in a standard geometry foil the GEM-field is about 15kV/cm at a GEM-voltage of 100 V (unit gain)and 40 - 70 kV/cm in normal amplification mode.

30

a) b)

0 25 50 75 100 125 150 175 2000.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

t / D / p: 25 / 70 / 140 m[ ]m

t / D / p: 50 / 70 / 140 m[ ]m

t / D / p: 50 / 115 / 140 m[ ]m

t / D / p: 50 / 140 / 280 m[ ]m

GEM-voltage [V]

0 25 50 75 100

ele

ctr

on tra

nsm

issio

ngain

x

0 25 50 75 100 125 150 175 2000.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

standard geom.

cylindrical

conical W-N

conical N-W

GEM-voltage [V]

ele

ctr

on tra

nsm

issio

ngain

x

Figure 5.6:Electron transmission vs. GEM-voltage in Ar-CO� (90:10) at �� � 150 V/cm and � � � 2.5kV/cm for (a) GEMs of different hole shapes and (b) for GEMs of different thicknesses �, hole diameters� and pitches � ; the gray data points refer to the top-, the black ones to the bottom-axis scale.

the ratio� � of GEM-field and induction field. For simple electrostatics reasons,

��is directly correlated

with how well the field lines from the drift region are focusedinto the holes, and�� with how well

the field lines from the hole are extracted into the inductionfield: focusing improves with smaller driftfields or higher GEM-fields, extraction improves with smaller GEM-fields or higher induction fields(compare for instance Fig. 4.2 and Fig. 4.5). Unfortunately, all three parameters are largely limited byouter constraints: the GEM-field (or GEM-voltage) is definedby gas gain requirements, the inductionfield limited by operation stability concerns, as discussedin detail in section 6.1 about discharges, andin TPC applications the value of the drift field is imposed by optimization of diffusion and electron driftvelocity of the gas mixture used.

In Fig. 5.6.a the electron transmission for the different GEM-geometries at fixed values of drift andinduction field is shown: standard, cylindrical and the two orientations of the conical. One can see thatthe conical GEM, when facing the drift volume with the wide side (W-N), permits to reach higher valuesof transmission before the onset of gas amplification. Perhaps more interesting, and confirming thatlosses are diffusion-dominated, are the results shown in Fig. 5.6.b, obtained with double-conical GEMsof different thicknesses, hole diameters and pitches� . All other conditions being identical, the foilwith 140 �m wide holes at 280�m pitch, which offers the most space to the electrons for diffusing, isalmost fully transparent to electrons (but provides much less gas amplification).

5.5 Measurements of ion transmission and ion feedbackReversing the field direction in the detector the positive ions produced in the gas drift towards and

through the GEM. A measurement of the positive ion current onthe various electrodes provides the iontransmission through the foil as a function of fields – of course, there is no charge multiplication.

As visible from the measurements in Fig. 5.7.a, at fixed values of drift and induction fields, withraising values of the GEM-voltage the ion transmission increases before it reaches a plateau. As to beexpected, the value of the plateau seems almost independentfrom the gas mixture used.

Measurements in a wide range of fields and with the various geometries described before have beenperformed in an Ar-CO� (90:10) gas mixture. As shown in Fig. 5.7.b, for ratio valuesin the investigatedrange (� � ���) all data tend to follow the same trend: the ion transmissionplateau value depends almost

31

ion

tra

nsm

issio

n p

late

au

va

lue

0.00 0.05 0.10 0.15 0.20 0.250.00

0.05

0.10

0.15

0.20

0.25

standard geom. E =150 V/cmD

standard geom. E =2.5 kV/cmI

conical W-N E =150 V/cmD

conical W-N E =2.5 kV/cmI

conical N-W E =150 V/cmD

conical N-W E =2.5 kV/cmI

ratio of external fields ED:EI

0 25 50 75 100 125 150 175 200 225 2500.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.10

CO (>99%)2

Ar-CO (70:30)2

Ar-CO (90:10)2

Ar-CH (90:10)4

ion

transm

issi

on

GEM-voltage [V]

a) b)

Figure 5.7: a)Ion transmission vs. GEM-voltage with a standard GEM at ��=150 V/cm and � �=2.5kV/cm in different gas mixtures. b) Ion transmission plateau value vs. ratio of external fields �� :� � inAr-CO� (90:10).

linearly on the ratio�� � � � .

The first GEM in a TPC could for instance be operated with a fieldratio of��

:� �=150:2.5k=0.06. In

this configuration, the ion transmission plateau values vary from � ��� � �� % for the geometry with widerholes to

� �� � � �� % for the standard geometry. Fig. 5.8.a gives the measured ion transmission for thecase of a high drift field

�� � � �� kV/cm and a slightly higher induction field� � � � ��kV/cm, which

applies for instance for intermediate stages in multi-GEM detectors. In pure CO�, the ion transmissionsaturates at a value (0.61) slightly smaller than the field ratio (0.71). The increase at high GEM-voltagesin Ar-CO� (90:10) is an artifact due to stray X-rays or fluorescence photons converting close to theGEM-holes and multiplying.

The measurements discussed above correspond to the transmission of ions uniformly released in thegap below the GEM-foil. For completeness, also the effective ion feedback into the drift gap for the caseof avalanche multiplication in the foil itself was measured. Fig. 5.8.b shows, for a standard GEM, themeasured effective ion feedback, i.e. the ion back-flow normalized to the effective gain. For low driftfields this fraction is below 10%, whereas it reaches almost 100% for high drift fields; the latter case isencountered for instance in intermediate stages of multi-GEM detectors.

5.6 Discussion of resultsAs mentioned above, the reduced electron transmission in some gases can be attributed to losses due

to large values of transverse diffusion causing the electrons to be lost to GEM-electrodes or Kapton walls,despite the fact that all electric field lines from the drift volume pass through the holes. In the region ofunit gain, the field in the GEM-holes varies between 10 and 20 kV/cm (central shaded area in Fig. 5.4.a).The average transverse diffusion after about 100�m drift, corresponding to the field region inside andclose to the holes, is about 16�m (RMS) for pure CO� and about 30�m (RMS) for the Ar-CO� (90:10)mixture; the latter clearly is close to the hole radius (25 to35 �m), explaining the occurrence of losses.

By the same token, one can infer that, in normal gas amplification mode with GEM-fields of 40 to70 kV/cm (rightmost shaded area in Fig. 5.4.a), the diffusion losses in poorly quenched gases are com-parable to those measured in pure CO� at unit gain, and therefore losses of electrons before entering gasamplification should be very small. Of course, this only applies on condition that the drifting electrons

32

0 50 100 150 200 250 300 3500.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

CO (>99%)2

Ar-CO (90:10)2

ion

transm

issi

on

GEM-voltage [V]

eff. gas g

ain

eff. io

n feedback

0 50 100 150 200 250 300 3500.01

0.1

1

10

100

E =150V/cm, E =2.5kV/cmD I

GEM-voltage [V]

E =2.5kV/cm, E =3.5kV/cmD I

eff. gain ion-f.b.

eff. gain ion-f.b.

a) b)

Figure 5.8: a)Ion transmission vs. GEM-voltage with a standard GEM at ��=2.5 kV/cm and � �=3.5kV/cm. b) Effective ion feedback and effective gas gain vs. GEM-voltage in a standard GEM for differentfield configurations.

are sufficiently focused by the electric field into the centerof the GEM-holes – which for drift fieldsbelow 1 kV/cm and GEM-voltages above 300 V should be a given fact. Short of finding a direct way ofdisentangling electron transparency, gas gain and extraction at large GEM-voltages, the above consider-ations anticipate that no charge losses should occur beforemultiplication in a TPC-like device with thefirst GEM used in gas amplification mode, even if using a poorlyquenched gas.

A similar argument holds for the operation in a strong magnetic field, parallel to the drift field. Fig.5.4.b shows the transverse diffusion in an Ar-CH� (90:10) gas mixture, a common choice for TPCs,with three values of magnetic fields. One can see that in the drift region (0.1 to 1 kV/cm) transversediffusion is strongly reduced, a well-known property used for improving spatial resolution. Inside theGEM-holes, although

�and

�are not always completely parallel, the argument applies tothe transverse

components of the field; lateral displacement of the electrons, the� � �

effect, has been demonstratedexperimentally not to affect the electron collection efficiency [36]. Transverse diffusion is somewhatreduced in the region of unit gain, but perhaps not enough to restore full transmission. Thus, in GEM-stages operated at these low gas gains, geometries with wider holes and pitch should be used in order toavoid losses of electrons before multiplication.

While ion transmission turned out to be largely gas-independent the effective ion feedback varies withgas-mixtures and GEM-voltages. Nevertheless, one can estimate the ion back-flow in a GEM-based TPCfrom the data presented.

In a possible scenario with a standard geometry double-GEM operated in Ar-CH� (90:10) with�� �

��� V/cm,�� � � �� kV/cm and

� � � � �� kV/cm at GEM-voltages of���� � ��� V each, one wouldexpect an effective gas gain of about

� �� � ��� , distributed to the two GEMs in a 40:60 ratio in favor ofthe second multiplier (Fig. 5.8.b). The effective ion feedback can be estimated as follows:

� 1st GEM:�� � ��,

� � � � ���, � � � � ����;� 2nd GEM:

��� ��,

��� � ���;

� total effective ion feedback, see equation(5.17):� � � � �� ��� � �� � �� � �� ��

� � �� � � � ����

�� � � �� ,

here� � � ������ � ��� � � ���� � �� , i.e. in the range of some percent;

33

It is worth to emphasize, that the gas amplification in the first GEM-stage contributes to the feedback withan amount which is smaller by one order of magnitude than the amount from the second one. Generallyspoken, the effective ion feedback is dominated by the product of effective ion feedback in the last GEM-stage and the ion transmissions of the other stages; thus, multi-GEM devices with higher induction fieldsor lower drift fields show a lower effective ion feedback.

34

Chapter 6

Discharges and aging in GEM-detectors

Longterm operation of gaseous detectors at high particle fluxes in general is limited by two phenomena,gas dischargesanddetector aging. The results of studies in these fields performed with GEM-detectorswill be discussed in the following sections.

6.1 DischargesPerformance and operation stability of GEM-detectors exposed to high intensity particle beams has

been studied in view of their use in high energy physics experiments [37–41]. For the efficient detectionof minimum ionizing particles with nowadays available high-density readout electronics, effective gasgains in the range��� to ��� (depending on primary ionization etc.) are required. In principle, thiscan be attained with a single GEM-stage, but it has been observed that when operating at these valuesof gain, exposure to high radiation fluxes, or the release of alarge amount of charge in the sensitivevolume may induce breakdowns of the gas rigidity and lead to gas discharges. In this section studiesof the discharge behavior of multi-GEM detectors in variousconfigurations are presented, in which theinfluence of gas gain, number of GEM-stages and gain distribution as well as electric field configurationwere investigated [3].

6.1.1 Basic mechanisms and precautions

Discharges in GEM-detectors typically arise between the two electrodes of a GEM-foil aslocal dis-charges. In rarer cases, the so-calledpropagating discharges, the breakdowns involve neighboring GEM-stages or the anode readout. Apart from resulting detector dead-times1 and the release of huge amountsof positive ions to the drift volume, occurring discharges might inflict irreversible damage on GEM-foilsor readout electronics.

The sequence of events leading to a discharge is initiated when the avalanche size exceeds�� to�� ion-electron pairs, the so-called Raether limit [24], which is strongly related to charge density2 andelectric field strength. The ensuing local field modificationis then large enough to induce a transition ofthe avalanche to a forward-backward propagating streamer,a well-studied process known from parallel-plate counters and wire chambers. In the latter case, the radially decreasing field encountered by thestreamer when receding from the anode, results in stalling of the propagation. In the GEM as well as

��Depending on the powering scheme, GEM-foils require several �s to ms to recover operation voltage after adischarge.

��The actual charge density results from primary ionization (total amount and local density) convoluted withtransport (drift and diffusion) and gas avalanche amplification processes.

35

HVTOP,2

HVBOT,2

HVD

HVTOP,1

HVBOT,1

HV

2nd

GEM

1 GEMst

cathode

anode

a) b)

2nd

GEM

1 GEMst

cathode

anodeprotection resistors

voltage divider

protection resistors

Figure 6.1:Powering schemes for GEM-detectors based on individual supplies (a) and on a resistivevoltage divider network (b).

in most MPGDs, streamers are more likely to be followed by breakdown because of their by orders ofmagnitudes smaller electrode distances.

Although the first rupture of gas rigidity occurs in the high electric field of a GEM, normally in thelast of a cascade, the discharge may propagate forwards or backwards (or both) to other electrodes. A fastpropagation between GEMs has been observed, compatible with a photon-mediated charge breeding byionization of gas or electrode material [42]. The predominance of a fast propagation mechanism betweenGEMs is confirmed by the observation that discharges can propagate between two multipliers, even ifthe electric field is inverted in the transfer region. Propagation towards anode or cathode, which is dueto possible destructive consequences to readout electronics more critical, appears instead to be a slowerprocess involving the transport of charges.

Whilst not affecting the occurrence of the initial discharge, the high voltage powering scheme ofthe GEM-electrodes can drastically alter the sequence of ensuing events. Individual power suppliesfor each electrode turned out to be most convenient for systematic studies during R&D; but even ifpowered through protection resistors, as shown in Fig. 6.1.a, the use of this scheme can result in asudden, uncontrolled increase or decrease of individual GEM-electrode potentials, leading to sustaineddischarges, often with irreversible damages to the foil. Itis therefore recommendable to work with aresistive voltage divider network for powering of the various electrodes, as shown in Fig. 6.1.b. Withsuch a circuit, the course of events in possible discharges becomes more predictable; the propagationprobability, however, appears to be affected by the value ofresistors: the sudden change of the potentialdifference between electrodes initiated by a discharge increases the external fields, and as a consequencethe propagation probability between GEM-stages and to the anode readout. Whilst for obvious reasonsthe best protection is obtained with high value resistors, experimental requirements as particle rate andgas gain determine the maximum values that can be used: the voltage drops across the GEMs due torate-induced currents in the detector, and thus in parallelto the voltage divider, have to be kept withinacceptable limits. Typical resistor values applied in suchcircuits are 0.5 to 1 M� for the voltage dividerand 10 to 20 M� for the protection resistors.

6.1.2 Experimental setup

The measurements have been performed in the non-flammable Ar-CO� (70:30) gas mixture with asetup similar to the one described in the previous chapter. The GEMs used were�� � �� cm� large andof standard geometry. Unless otherwise mentioned, transfer and induction gaps were 2 mm thick, thedrift gap 3 mm thick, and electric fields of 2 kV/cm in the driftand 3.5 kV/cm in the other gaps wereused.

To induce controlled discharges, a collimated radioactive���Am source which emits

�-particles of 5.6

36

cathode

anode

GEM

IC

ITOP

IBOT

IA

ED

EI

driftgap

inductiongap

Am-source

discharge

propagatingdischarge

clock &counter

3912 s

5 e

UC

UTOP

UBOT

UA

oscilloscope

a)

0 2 4 6 8 100

-200

-400

-600

-800

-1000

-1200

-1400

-1600

volta

ge

[V]

time [ s]m

b)cathode

anode

GEM, top

GEM, bottom

241

Figure 6.2: a)Schematics of the discharge measurements, here for instance with a single-GEM detector.b) Time evolution of all electrode voltages in a single-GEM detector upon a localized discharge.

MeV was used. It was mounted either internally, in contact with the drift electrode, or externally in frontof a 3.5�m thin polymer window (Fig. 6.2.a). In both cases, the drift electrode consisted of a fine wiremesh, allowing the

�-particles to penetrate into the sensitive volume. This way, the spectrum of energy

deposited in the drift volume is better defined, with a peak at500 keV; with a coarsely collimated sourcerates of several ten Hz can be achieved, permitting to measure discharge probabilities per

�-particle down

to ���� within one hour.

The appearance of a discharge is recognized electronicallyas a fast and exceedingly large pulse onthe concerned electrodes. Large induced signals are also seen in all other electrodes. An example for thesequence of events in a localized discharge in a single-GEM detector is shown in Fig. 6.2.b. The voltagesof all four electrodes were recorded with a multi-channel digital oscilloscope. As a function of time, thevertical scale shows the time evolution of voltages on cathode, anode and the two GEM-electrodes: thevoltage difference across the GEM drops almost symmetrically to zero and a positive (negative) signal isseen on the cathode (anode). The recovery time of the pulses in the plots (some ten�s) is given by thecircuit used for voltage-readout, a 10 nF capacitor on 50�, and does not correspond to the real responseof the high-voltage circuit (several ms)3.

When using the powering scheme with individual supplies foreach electrode, a discharge can be de-tected as an overload of the power supply, usually set to a limiting current of 1�A. In measurementswhere a voltage divider was used, discharges were monitoredmaking use of an oscilloscope and a count-ing unit either connected to the voltage divider chain or directly to the properly terminated anode readout.

6.1.3 Discharge measurements

It should be emphasized that the probability of the transition from proportional gas amplificationto streamers and discharges at a given electric field configuration depends on numerous internal andexternal factors, such as temperature, humidity and gas flow, as well as on quality and previous historyof the electrodes. Variation of a single parameter during short and medium-term measurements can beconsidered significant. Thus, caution should be used when confronting measurements realized in alteredsituations or with different detectors. The study of discharges induced with

�-particles can only provide a

qualitative indication of a detector’s operation stability in real running conditions. The different spectrumand angular distribution of heavily ionizing tracks will show a particular signature in each experiment,and a definitive, quantitative answer concerning detector performance has necessarily to be obtained

��For this illustration a modified setup with no protection resistors was used.

37

300 350 400 450 500 550102

103

104

105

Single-GEM

Double-GEM

Triple-GEM

effect

ive

gas

gain

voltage per GEM [V]

460 470 480 490 500 510 520 530

-3

-3

-3

0.0

1.0x10

2.0x10

3.0x10

4.0x10-3

disch. probab.dis

charg

epro

babili

typer

a

GEM- [V]voltage

101

102

103

effe

ctivegas

gain

eff. gas gain

a) b)

Figure 6.3: a)Total effective gas gain vs. voltage per GEM for single-, double- and triple-GEM detector.b) Discharge probability measured with the

���Am �-source and effective gas gain vs. GEM-voltage fora single-GEM detector.

there.

Single-, double- and triple-GEM detectors

In Fig. 6.3.a the gain calibration curves for single-, double- and triple-GEM detector are shown. Asin the subsequent measurements, the drift field was chosen to2 kV/cm and transfer and induction fieldsto 3.5 kV/cm. In double- and triple-GEM detector, voltages across GEMs were kept identical. Fig. 6.3.bshows discharge probability per

�-particle and effective gas gain in a single-GEM detector with Ar-CO�

(70:30) as a function of the GEM-voltage. The threshold for discharges is observed around 500 to 510 V,corresponding to an effective gain of� � ��� to

� � ��� . The discharge probability measured as a function ofGEM-voltage4 in the single-, double- and triple-GEM are given in Fig. 6.4.a. The decrease in dischargevoltage is, of course, a reflection of the increasing avalanche size for multiple devices. In Fig. 6.4.b, thesame data is plotted as a function of total effective gain. Inthe condition described, the maximum gainbelow discharge threshold is increased by about one order ofmagnitude at each addition of a multiplier.This can be imputed to charge spread between the individual GEM-foils due to diffusion as well as tothe fact, that for the same total effective gain, the voltageper GEM is lowered and larger charge densitiesare tolerated before breakdown of the gas rigidity. An additional effect of the lowered GEM-voltages isthe reduction of energy liberated in case of a discharge: theenergy stored in a GEM-foil scales linearlywith its capacitance and quadratically with the applied voltage difference.

Gain sharing

As deduced in the previous chapter, in a cascade of equally multiplying GEM-stages, the amount ofcharge created is by far the largest in the last GEM-foil. Forthis reason, lowering the voltage across thelast GEM-foil, reduces not only the charge density created in this stage, it also increases the breakdownlimit. Of course, in order to maintain the total effective gain, the voltage on other stages has to beincreased accordingly. The measurements shown in Fig. 6.5.a underpin this consideration. In this plot,the discharge probability per

�-particle in a double-GEM detector is shown as a function of the voltage

imbalance, defined as the relative voltage offset� � � , with difference� � ���� �� � ���� �� andsum� � ���� �� ���� �� of the two GEM-voltages. For each curve, the total gain was kept constant.

� � For this measurement, the voltage applied to each multiplier was identical.

38

300 350 400 450 500 550

-3

-3

-3

0.0

1.0x10

2.0x10

3.0x10

4.0x10-3

Single-GEM

Double-GEM

Triple-GEMdis

charg

epro

babili

typer

a

voltage per GEM [V]

102

103

104

105

106

10-5

10-4

10-3

10-2

10-1

Single-GEM

Double-GEM

Triple-GEM

dis

charg

epro

babili

typer

a

effective gas gain

a) b)

Figure 6.4:Discharge probability measured with the���Am �-source for single-, double- and triple-GEM

detector vs. GEM-voltage (a) and vs. effective gas gain (b).

25 cm2

50 cm2

100 cm2

0.0 2.5 5.0 7.5 10.0 12.50.0

0.2

0.4

0.6

0.8

1.0

pro

pagatio

npro

babili

ty

induction field [kV/cm]

a) b)a) b)

-0.10 -0.05 0.00 0.05 0.10 0.15 0.2010

-6

10-5

10-4

10-3

10-2

10-1

dis

charg

epro

babili

typer

a

G = 1.0 10x4

G = 1.8 10x4

DU : Uvoltage imbalance

Figure 6.5: a)Discharge probability measured with the�� �Am � ’s for the double-GEM as a function of the

voltage imbalance for two values of effective gas gain. b) Discharge propagation probability of a 10�10cm

�double-GEM with the upper electrode divided into four sectors of 25 cm

�surface vs. induction field.

For the individual measurements one, two and all four sectors were connected in parallel.

One can see that, depending on the total gain, an optimum is reached at voltage imbalances of 5 to 7� infavor of the first GEM. The fact, that for larger imbalances the discharge probability increases again canbe explained with the decrease of the charge density tolerated before breakdown resulting from higherelectric fields at higher voltages in the first GEM-stage.

Discharge propagation

To investigate and quantify the mechanisms of discharge propagation to the anode readout, GEM-voltages and thus effective gas gain were set just above the onset of discharge occurrence; possible effectsof the GEM-voltage on discharge propagation have not been not studied. Propagating discharges canbe well separated from localized discharges by reading the anode-signals with application of adequatetrigger thresholds5.

39

The propagation probability, defined as the probability of aGEM discharge to evolve into a dischargeto the anode readout, depends on the energy of a discharge andthe strength of the induction field. Fig.6.5.b shows the discharge propagation probability of a 10

�10cm� double-GEM, of which the upper

electrode6 is divided into four sectors of 25 cm� surface, as a function of the induction field. For theindividual measurements one, two and all four sectors were connected in parallel, resulting in a modifiedcapacitance and thus a modified average discharge energy. Itcan be seen that with increasing induc-tion field, and thus increasing extraction of the electric field lines from the GEM-hole, the propagationprobability becomes higher; in addition it increases with larger discharge energies.

6.1.4 Discussion of results

Discharges in GEM-detectors begin with a sudden, radiation-induced breakdown of the gas rigidityin the gas amplification region of a GEM-foil – normally the last if in a cascade of multipliers. Thedischarge probability appears to depend on the primary ionization density and on the total gain of thestructure, but not directly on the external fields. Some general properties have been deduced from themeasurements described above.

With multiple structures, the maximum gain below dischargethreshold gain increases by about anorder of magnitude for each stage added; the discharge probability is further reduced by off-settingthe voltages in order to enhance the gain in the first element in the cascade. A second observationhaving important practical consequences is that not only the discharge energy, but also the propagationprobability depends on the capacitance of the GEM. Manufacturing the multipliers with independentlypowered sectors therefore largely reduces the probabilityof energetic discharges propagating into thereadout electronics. Sectorizing the GEMs, of course, comes with the price of small areas at the sectorborders, which for obvious reasons tend to be less efficient.In addition, the propagation probabilityis a fast increasing function of the induction field, and all indications are that it only appears above afield threshold value, around 5 kV/cm in standard operating conditions. For this reason operation atmoderate induction fields is highly recommendable for a sufficient protection of the readout electronicsagainst destructive propagating discharges. Unfortunately, this is opposed to the minimization of the ion-feedback: as demonstrated in the previous chapter, the ion transmission and thus the total ion feedbackof a multi-GEM device decreases with higher values of the induction field. A compromise of reliableoperation stability and a moderate ion back-flow will have tobe evaluated for each TPC applicationindividually, depending on beam properties, track densities and the counting gas used.

6.2 Detector agingDetector aging, i.e. the slow deterioration of performanceduring operation, has been observed since

the early development of proportional counters. The underlying phenomena are highly complex physicaland chemical processes, of which a satisfactory quantitative explanation is not available up to the presentday. In gaseous detectors, one distinguishes between the direct destruction of electrodes and depositsof insulating material on electrodes. The latter are mainlyproducts of polymerization processes, whichcan take place in the plasma-like environment inside the gasamplification avalanches. Involved in thesereactions are the quenching-gas, traces of gas impurities7 and parts of the detector material which are��

Although the electrostatic energy stored between the two electrodes of a GEM (300 to 400 V over 56 pF/cm�)

is larger than the energy stored between GEM and anode readout (typically 700 V over 0.5 pF/cm�), the signal

amplitudes visible on the anode are about a factor of ten larger in the case of propagating discharges. This isdue to the fact, that in propagating discharges almost the full amount of released charge is transported to theanode readout.

��By sectorizing only the upper GEM-electrode, the absolute voltage of the lower stays more stable in the caseof localized discharges, thus less modifying the inductionfield and suppressing discharge propagation.

��Impurities are intrinsically present in any gas, but additional contributions may originate from finite gas tight-

40

exposed to the avalanche (as for instance the anode wire of a proportional counter). A necessary precursorfor polymerization is the formation of molecular fragments, which in avalanches are mainly the productof molecular dissociation due to electron or photon impact.In CH�, a commom quenching-gas forTPCs, free radicals are produced in the reaction CH� + e� � CH�: + H� + e� . The CH�: radicals,typical precursors for the creation of polyethylene, have large dipole moments and preferably stick toelectrode surfaces. On the contrary, quenchers such as CO� will not produce similar reactions unless inpresence of hydrocarbon contamination.

The macroscopic aging processes obviously depend on microscopical quantities such as e.g. cross-sections, energies of photons and electrons, electrostatic forces, chemical reactivity and others, but aclear theoretical formulation is pending. At present, the during operation accumulated electric charge8

per readout area�� ��� (in wire detectors: charge per wire length

�� �� �) is commonly used to quantifya detector’s ”age”. However, different types of gas detectors, detectors operated in different countinggases, at different gas gains or in different radiation regimes can show strongly differing performancesafter the identical amount of accumulated charge. Not only the total dose, but also the locally appliedrate of radiation, gas flow and detector volume influence aging processes. For this reason, the predictivepower of extrapolations of the radiation hardness of gas detectors from accelerated tests or tests on smallsurfaces is limited. Nevertheless, accelerated tests represent the only way to near-term estimate thelong-term performance of newly developed detector technologies.

In the following section an accelerated test for the Ar-CO� based triple-GEM detectors built forthe COMPASS experiment at CERN is presented [4]. The design of these large-size detector moduleshad been optimized to satisfy the low material budget requirement of the experiment while achievinggood spatial and time resolution properties over large surfaces, thus facing similar requirements as TPCreadouts. The goal of this study, in which more than seven years of normal operation in the COMPASSexperiment’s high-rate environment were simulated, was tofinally validate detector design and materialcomposition – which will serve later on in this work as a guideline for the layout of GEM-based TPCs.

6.2.1 The COMPASS triple-GEM detectors

Since spring 2002, twenty triple-GEM detector modules of� � � � �cm� active area, as sketched inFig. 6.6.a, are installed and successfully operated in the Small-Area-Tracker (SAT) of the COMPASSexperiment [43]. In order to minimize probability and energy of spark discharges, the gas amplificationis distributed to three GEM-foils, which are as illustratedin Fig. 6.6.c on one side segmented into 12sectors plus a central disk, and powered by a resistive voltage divider. At nominal operation, the threeGEMs have a potential difference of 425 V on the first, 380 V on the middle, and 340 V on the thirdGEM – following the outcome of the studies presented in section 6.1. The electric fields are chosen to2.5 kV/cm in the drift gap and 3.7 kV/cm in all other gaps. By lowering the voltage on the central disksector of the lowest GEM, this region can be deactivated to allow operation with the high intensity beamtraversing the detector volume.

The detectors are equipped with a two-dimensional projective strip readout, consisting of two layersof 768 strips at 400�m pitch (70�m width in the upper, 350�m in the lower layer, as shown in Fig.6.6.d), engraved on the two sides of a 50�m Kapton foil. With the Kapton removed in the interstices ofthe upper strips, both layers are opened for charge collection.

As shown in the exploded view in Fig. 6.6.b, the strip readoutis glued to a 3 mm thick carrier structuremade of honeycomb, which is covered by 125�m Stesalit skins on each side. The GEMs are glued tothree fibreglass (Vetronite) frames of 2 mm thickness, whichalso contain spacing grids to maintain thegap width constant over the whole area. To avoid possible charging up effects or discharges close to thesegrids, their surface is smoothed with a thin layer of Polyurethane. The last frame is made of 3mm thick

ness or from outgassing processes of materials used in detector or gas system.��

Sum of charge from the ionization process and charge createdduring gas amplification.

41

307 m

m

carrier-structure

drift cathodeframe

GEM foilframe & grid

GEM foilframe & grid

GEM foilframe & grid

readout

carrier-structure

2 mm

3 mm

2 mm

2 mm

a) b)

c)

d)

Figure 6.6: Schematic (a) and exploded (b) view of the COMPASS triple-GEM detector design. c)Photograph of a GEM-foil: the sector borders of the upper electrode are clearly visible. d) Microscopephotograph of the strip readout. (Pitch 400 �m, strip width 70 �m in the upper and 350 �m in the lowerlayer.)

Stesalit, and carries the drift cathode, consisting of a 50�m thick Kapton foil with 5�m copper. Forstability reasons, the whole structure is closed by another3 mm carrier honeycomb, resulting in a totalmodule thickness of 15 mm at a radiation length of� ����

��� only. The counting gas is supplied through

0.5 m long Polypropylene tubes and distributed via grooves machined into the uppermost frame. Thecomponent assembly was done with the non-outgassing glue Araldit AY103 plus hardener HD991 [44].For completion, all components and materials used for detector production are listed in Table 6.1; amore detailed description of the detector construction canbe found in reference [31].

Equipped with an especially developed front-end electronics circuit based on the APV chip [45],the detector reaches full two-dimensional efficiency (� ���) at effective gas gains of� �� � ��� in Ar-CO� (70:30). Spatial resolutions as good as

�� � � �m for minimal ionizing particles (MIPs) and atime resolution of 15 ns (detector including electronics) have been demonstrated [37], underlining theperformance of this detector concept.

6.2.2 Experimental setup

The setup shown in Fig. 6.7.a was used to irradiate a for this purpose dedicated detector modulefrom the production line for the COMPASS experiment with a high flux of X-rays. A small SWPCfor reference measurements was installed in the same gas line in front of the GEM-detector. Charge

42

Component Details Producer / Supplier

Assembly glue ARALDIT AY103 + HD991 (ratio 10:4) CERN store

Grid conditioning NUVOVERN LW Walter Mader AG,

(2 component Polyurethane) 8956 Killwangen (CH)

Carrier structure� � ����

m STESALIT Stesalit AG, 4234 Zullwil (CH)

on 3mm NOMEX honeycomb Socol, 1020 Renens (CH)

Shielding Aluminium (10�

m) CERN store

GEM-foils� � ��

m Cu on 50�

m KAPTON CERN/EST

Drift 5�

m Cu on 50�

m KAPTON CERN/EST

Frame 3mm STESALIT CERN/EP/TA1

Spacers 2mm VETRONITE grids CERN/EST

Gas pipes PP tube (3mm diam.) Angst-Pfister, 8052 Zurich (CH)

Gas outlet F-glass + fitting CERN/EP/TA1

Microstrip readout Cu microstrips on 50�

m KAPTON CERN/EST

HV boards Custom made CERN/EST

Table 6.1:List of components and materials used for the production of the COMPASS triple-GEM de-tectors.

sensitive amplifiers were used to regularly record low rate pulse height spectra of both coordinates of theGEM-detector and the SWPC. Furthermore, the currents in theGEM-detector as well as the rate of theX-ray tube were monitored, and sensors for ambient parameters like temperature, relative humidity, andatmospheric pressure were read out.

The counting gas was supplied with an open flow system. A gas bottle with Ar-CO� premixed in a70:30 ratio was connected via a pressure reducer and a 10m long stainless steel gas line to a mass flowcontrol unit. A constant flow of�� � � ��cc/min was fed to the SWPC and the GEM-detector with shortPolypropylene tubes. The GEM-detector was directly connected to a hygrometer and a 10m long plasticexhaust gas line, avoiding the use of any bubbler. The detector volume of 0.85 liters was exchanged 5.6times per hour, somewhat more frequent than in the COMPASS experiment (3.5 volume exchanges perhour and two detectors in series per gas line).

One quarter of the detector (�� � �� cm�) was exposed to the 8.9keV X-ray beam with an approx-imately Gaussian profile of�� �� � �� � � mm (see Fig. 6.7.b), thus with a surface of 12.6 cm� insidethe one-� -line. The detector was irradiated in two phases: 10 days at� �� � ��� Hz/mm� in the beamcenter (”phase 1”), and 10 days at

� �� � ��� Hz/mm� (”phase 2”). Operated at the nominal effectivegain of � �� � ��� , this resulted in a current density on the readout strips of

� �� nA/mm� and 10 nA/mm�,respectively.

At the beginning of the aging measurement, a gas gain calibration for the GEM-detector was carriedout, which is shown in Fig. 6.7.c. The calibration was performed at

�� ���C temperature,�� �� �� hPaambient pressure and

�� ��� relative humidity.

During the whole irradiation, the following measurement cycle was performed: after each ten minutesof irradiation, a shutter in front of the X-ray tube was closed, the low rate pulse height spectra as wellas all other parameters recorded, and after reopening the shutter the irradiation continued. Controllingand readout was fully automatic and realized with a personalcomputer connected to a CAMAC crate.In order to measure effective gain

�and energy resolution, a Gaussian model was fitted during data

analysis to the pulse height spectra, and position as well asfull-width-half-maximum of the photopeakdetermined.

43

detector withshielding

X-ray tube

SWPC

preamplifiers

3900 3950 4000 4050 4100 4150 420010

2

103

104

105

eff.gas

gain

detector voltage [V]

a) b)

c)

0 5 10 15 20 250

1000

2000

3000

4000

5000

6000

7000

8000

GEM X

GEM Y

rate

[Hz/

mm

]

position [6.4mm]

2

Figure 6.7: a) Photograph of the aging measurement setup: X-ray tube and triple-GEM detector withshielding. b) Beam profile projections in � and � direction and Gaussian fits: �� ��=20�1 mm. c) Gasgain calibration of the for this measurement used GEM-detector (at 22.9�C temperature, 1012.6 hPaambient pressure and 42.3� relative humidity).

6.2.3 Aging measurement

The variations of the ambient parameters during the aging measurement have been recorded andare shown in Fig. 6.8. Whereas the relative humidity of the atmosphere varied between�� ��� and�� ��� , the H�O concentration in the counting gas measured with the hygrometer at the output of thedetector was constant within 2 ppm at 60 ppm. The crucial importance of the H�O concentration forthe discharge probability in GEM-detectors has been demonstrated recently [31]: with a double-GEMdetector, a change from 81 ppm down to 35 ppm lowered the discharge probability by almost three ordersof magnitude (at a gain of� �� � ��� ). Operation stability of the detector thus implies to avoidexcessiveH�O contents in the counting gas, although H�O has been reported to reduce aging processes [46].However, during the aging measurement the currents on the high voltage power supply were monitoredand not a single discharge has been observed.

The influence of the variations of temperature (� � ���C to

�� ���C, thus� ���) and pressure (���� ��hPa to���� �� hPa, thus� ���) on the gas gain are not negligible and had to be considered. Therefore, acorrected and normalized gain� was calculated from the effective gain

�according to

� � �� � �� � � �� � �� � � (6.1)

where� is an index for the two coordinates of the GEM-detector and the SWPC and� �� the ratio of

44

0 48 96 144 192 240 288 336 384 432 480 528

10

20

30

40

50

60

70

temperature [°C]

rel. humidity [%]

elapsed time [h]

1004

1008

1012

1016

1020

1024

1028

am

bie

nt p

ressu

re [h

Pa

]

amb. pressure [hPa]

tem

pe

ratu

re [

°C]

r

el. h

um

idity [

%]

Figure 6.8: Ambient parameters during the aging measurement vs. elapsed time. (The accuracy is�1�C for temperature, �3% for relative humidity, and �0.5 hPa for atmospheric pressure.)

0 24 48 72 96 120 144 168 192 216 240 2640

200

400

600

800

1000

1200

1400

1600

1800

2000

GEM X

GEM Y

SWPC

elapsed time [h]

eff. gain

~ p

uls

eheig

ht [a

.u]

Phase 1

Figure 6.9:Effective gain of GEM-detector and SWPC during phase 1 of the measurement.

temperature and pressure.� � and� � are fit parameters, determined by fitting the exponential model

� � � �� � � � � �� � �� � (6.2)

to the correlation plots of Fig. 6.10 (in order not to bias theaging study, only points from phase 1 of themeasurement were used, Fig. 6.9). The exponential dependence in equation (6.2) is deduced by assuminginverse proportionality of the Townsend coefficient

�to the mass density�, and thus

� � ��� � � �� .

The results of the aging measurement are given in Fig. 6.11. The accumulated charge�� ��� is

given for the one-� -line of the beam profile, thus the charge collected in the center was higher by a factor1.65. One clearly recognizes smaller fluctuations of�� in the normalized corrected gain of the GEM-

45

0.288 0.289 0.290 0.291 0.292 0.293 0.294800

1000

1200

1400

1600

1800

2000

2200

2400

SWPC, A = 20.1, B = 13.5

GEM X, A = 0.040, B = 35.4

GEM Y, A = 0.055, B = 34.3

T/p [K/hPa]

eff.

ga

in ~

pu

lse

he

igh

t [a

.u]

Figure 6.10:Correlation of effective gas gain and temperature to pressure ratio, deduced from phase 1of the measurement; the solid lines are exponential fits of the model

� �� �� � � � � �� �� �� .

0.0 1.0m 2.0m 3.0m 4.0m 5.0m 6.0m 7.0m0.0

0.2

0.4

0.6

0.8

1.0

1.2

SW

PC

GE

MX

GE

MY

Phase 2Phase 1

dQ/dA [C/mm ]2

16

18

20

22

24

26

28

GE

MX

GE

MY

no

rma

lize

d c

orr

ecte

d g

as g

ain

en

erg

y re

so

lutio

n [%

]

* **

Figure 6.11:Normalized corrected gain and energy resolution versus accumulated charge � ��� . Thespikes marked with a star () indicate an exchange of the gas bottle.

detector, completely in phase with the SWPC, and without anyreduction of the gain. Apart from smallfluctuations, also the energy resolution was constant in each of the two irradiation phases, the improvedenergy resolution at the beginning of the second irradiation phase is only due to a higher signal to noiseratio. Indeed, in order to compensate the rate-induced voltage drop in the GEM-foils, the voltage on theresistive voltage divider was slightly increased, resulting in higher GEM-voltages and thus larger signalsduring the low rate recording phase of the spectra.

After accumulating� �� mC/mm� on the one-�-line (�� �� mC/mm� in the center), the measurementwas stopped and a final scan carried out: pulse height spectrawere recorded in nine different positions,one in the center of the beam profile, four at 19.2 mm, and four at 27.2 mm distance from the center. Allspectra were unchanged, without any loss of gain and energy resolution.

46

6.2.4 Discussion of results

The observation, that the investigated triple-GEM detector showed no loss of gain and energy resolu-tion after accumulating more than 7 mC/mm�, which corresponds to� �� � ��

��MIPs/mm� or seven years

of normal operation in the COMPASS experiment, confirms the relative insensitivity of GEM-detectorsto aging when operated with clean Ar-CO� . Two reasons may explain this result: the fact, that the gasamplification is localized inside the GEM-holes (far from any electrodes and walls), and the small effectof possible polymerization deposits on the electric field inthis region.

TPCs are, in contrast to the COMPASS triple-GEM detectors, normally operated with quenchers con-taining admixtures of hydrocarbons, as for instance CH� . For this reason, when speaking of TPCs, wehave to reassess the positive result obtained in the measurement with the Ar-CO� based GEM-detectorsfor COMPASS. As mentioned before, in gases containing CH� , the production of CH�: radicals andthus polymerization are fostered. Therefore, the creationof polymers and deposit of those on electrodesare to be expected. The central issue, namely if the built-upof these deposits is so fast that the detec-tor performance is compromised, will have to be determined in a separate test with the final operationparameters (gas mixture, electric field configuration, gas gain). None the less, another central issue canby all means be deduced from the measurement with the COMPASStriple-GEM detectors: the fact thatno measurable aging occurred with Ar-CO� indicates the absence of significant outgassing of harmfulsubstances and thus the proper choice of materials and assembly procedure.

47

48

Chapter 7

The prototype GEM-TPC

In this chapter, the design of a general purpose R&D prototype chamber is presented, which was built toto study the performance of the GEM-technology in TPCs [5].

7.1 DesignTo allow a high degree of flexibility, the detector layout wasrequired to guarantee easy replacement

and modification of gas amplification stage, anode readout and front-end electronics. A robust design(but with moderate material budget) with a sufficient numberof irradiation windows for X-rays and low-energy

-rays, as well as a pad readout at ground potential was developed to facilitate setup, handling

and manipulation during measurements.

Fig. 7.1 shows a photograph of two prototype TPCs constructed at Karlsruhe University and CERNwith drift cylinders of 20 cm diameter and lengths of 25 cm and12.5 cm, respectively. The endcapsare replaceable, so that the TPC can be equipped with different types of gas detectors and pad geome-tries. Materials have been selected with special attentionto detector aging properties, i.e. following theaforementioned studies for the COMPASS triple-GEM detectors (see Table 7.1).

To guarantee high homogeneity of the electric field, the fieldcage (see Fig. 7.2) was designed in a

a)

c)

b)

Figure 7.1:Photograph of prototype TPCs designed and constructed at Karlsruhe University and CERNof 25 cm (a) and 12.5 cm (b) length as well as the 25 cm long chamber equipped with electronics (c).

49

Component Details Producer / Supplier

Assembly glue ARALDIT AY103 + HD991 (ratio 10:4) CERN store

Detector sealing O-ring� CERN store

Cathode endcap Fibreglass G10 Stesalit AG, 4234 Zullwil (CH)

Cathode electrode 18�

m Cu on 125�

m Kapton CERN store

Gas outlet SWAGELOCK / stainless steel CERN store

Drift cylinder� � ���

m Cu on 125�

m Kapton CERN store

Carrier structure Layers of 150 and 350�

m FERROZELL Ferrozell GmbH, Augsburg (D)

on 3mm NOMEX honeycomb Socol, 1020 Renens (CH)

Anode endcap Fibreglass G10 Stesalit AG, 4234 Zullwil (CH)

Mirco pads Passivated Cu CERN/EST

Gas inlet Custom made / stainless steel CERN/EP/TA1

GEM support PVC�

pillars & screws CERN/EP/TA1

GEM frames Fibreglass STESALIT Stesalit AG, 4234 Zullwil (CH)

GEM-foils� � ��

m Cu on 50�

m KAPTON CERN/EST

Table 7.1:List of components and materials used for the production of the TPC prototype chambers.(�

��PVC and the O-ring are materials which can cause aging problems. These materials have been

used for functionality reasons, i.e. to maintain the modularity of the prototype.)

a)

c)

b)

Figure 7.2:Photograph of the opened drift cylinder (with view on multi-GEM structure and field plate),cathode endcap and detail view of the resistive voltage divider.

double-layer layout with staggered ring-electrodes. The field cage foil in the barrel consists of two layersof 18 �m thick and 3 mm wide rings, etched at 4 mm pitch to a two-side copper-clad Kapton foil (161�m total thickness, as illustrated in Fig. 7.3.a). The rings of the two layers are aligned with an offset of2 mm, corresponding to half a pitch. Fig. 7.3.b shows a cut through the wall of the drift cylinder. Frominside out, the (gas-tight) field cage foil is surrounded by a125�m thick Kapton foil, which guaranteeselectric insulation to the following honeycomb carrier structure (350�m Ferrozell, 3 mm honeycomb,and 150�m Ferrozell).

To confine the electric field, the whole structure is shieldedby 50 �m copper connected to groundpotential. All components are glued with Araldit AY 103 plusHD 991. In the long drift cylinder,five irradiation windows are foreseen by sparing out the outer four layers. To connect the rings to the

50

50 m

150 m3mm

350 m

125 m

161 m

m

m

m

m

m

copper

FerrozellHoneycomb

Ferrozell

Kapton foil

field cage foil

outs

ide

insi

de

18 mm copper

18 m

125 m

161 m

m

m

m

copper

Kapton foil

field cage foil

b)a)

Figure 7.3:Schematic cut through field cage foil (a) and drift cylinder wall (b). The region where theouter four layers are left out defines an irradiation window for X-rays.

gas inlet HV feed-throughPVC pillars

field correction plate

GEM foils

micro pad readout

O-rings

Figure 7.4:Schematic cut through the anode endcap. GEM structure and field plate are mounted withPVC pillars to the fibreglass endcap disk. The disk is screwed to the drift cylinder and gas tightnessguaranteed by O-rings.

resistive voltage divider (which on purpose is kept outsidethe gas volume), the field cage foil was foldedand brought out at one side. This is to avoid field distortionsand heating or pollution of the counting gasdue to the resistors. The two voltage divider chains are equipped with 10 M� resistors of 1� tolerance,and a temperature coefficient of� � ���� �K. The total resistance of the long drift cylinder is 294 M� ,resulting in 340 mW total dissipation at 10 kV potential difference.

The cathode endcap is made of a 125�m Kapton foil with 18�m copper, glued to a 10 mm thickdisk of G10. The anode endcap is currently equipped with a GEMplus micro pad readout. As shown inFig. 7.4, multi-GEM structures of�� � �� cm� active area can be mounted with variable gaps on PVCpillars; a field correction plate on top avoids distortions of the drift field outside the square GEMs. Thewhole structure is carried by a 1.6 mm thick disk (G10) which also contains micro pads, feed-throughfor high-voltage, and grounding pads, as illustrated in Fig. 7.5.a. The 12.4

�1.17 mm� large pads are

arranged at 1.27mm�

12.5mm pitch in 8 rows and 80 columns, covering an area of 101.6�

100 mm�.The contacting scheme shown in Fig. 7.5.b is used to connect the signal pads to soldering pads on thebackside of the disk adapted to the 1/10 inch standard connectors of the front-end electronics.

51

gas inlet

HV feed-through

PVC pillars

micro pad readoutground electrode 2.5

4 m

m

1.2

7 m

m

b)a)

Figure 7.5: a)Tilt view of the fibreglass endcap disk with micro pads, high-voltage feed-through, and gasinlet. b) Illustration of the pitch adapter to standard

� ��� inch connectors: groups of two micro pads at1.27 mm pitch are connected via metallized holes to two solder pads at 2.54 mm pitch on the backsideof the disk.

0 100 200 300 400 500 600 7000

1

2

3

4

5

6

7

measurement

sim. data

ele

ctro

ndrift

velo

city

[cm

/ms]

electric field [V / cm]

630 650 670 690 710 730 75010

2

103

104

effect

ive

gain

GEM voltage UGEM1

+UGEM2

[V]

b)a)

Figure 7.6: a)Gas gain calibration curve for the double-GEM endcap operated in Ar-CH� (90:10) atelectric fields of 150 V/cm in the drift field, 2.5 kV/cm between the GEMs and 3.5 kV/cm in the gap infront of the pad plane. b) Measured and simulated drift velocity in Ar-CH� (90:10) vs. drift field; hereshown for 8.2 cm electron drift distance.

7.2 Current setup and detector validation

For the measurements presented in this dissertation, the 25cm long prototype was equipped with twoGEMs of standard geometry mounted with gaps of 2 mm. The electrode of the field correction plate isplaced with another 2mm distance in top of the upper GEM. As shown in the gas gain calibration curvein Fig. 7.6.a, the double-GEM endcap provides sufficient gasamplification with effective gains up to 10�in Ar-CH� (90:10). In first system checks the energy resolution of the detector in Ar-CH� (90:10) wastested with 5.9 keV X-rays from a��Fe-source. An unchanged resolution of FWHM� �� � �� �� � � ���measured for different drift distances (some cm to almost

��� meter) indicates high purity of the counting

gas and thus the system’s gas tightness. To test the homogeneity of the electric field in the drift cylinder,the electron drift velocity in Ar-CH� (90:10) was measured over several drift distances and for different

52

VME-crate

TPC

ethernet

optical fibre

external trigger signal

Readout-Controller

FEE-cards

PCdata files in

- formatCPT

ROSIE & Gadwall Clock & Trigger Board

Figure 7.7:Schematic view of the front-end electronics (FEE) and data acquisition system (DAQ).

R38

R47b)a)

c)

Figure 7.8: Photograph of the FEE-card (a), close view of the resistors R38 and R47, which wereexchanged i.o. to shift the digitization baseline (b) and photograph of the Readout-Controller (c).

electric fields. The good agreement with simulated data1, as indicated in Fig. 7.6.b, cross-checks theabsence of impurities in the gas due to leaks and verifies the uniformity of the electric field.

7.3 Front-end electronics and data acquisitionFor detector readout, highly integrated front-end electronics (FEE) initially developed for the TPC

of the STAR experiment is used [47]. As illustrated in Fig. 7.7, the readout system is divided into thefollowing components.

Small FEE-cards(Fig. 7.8.a) house two pairs of custom chips for signal shaping, preamplificationand analogue to digital conversion (ADC). The STAR preamplifier/shaper chip (SAS) contains low-noiseintegrating amplifiers and 2-pole shapers with a variable peaking time (60-150 ns) and pseudo-Gaussian��

Gas data computed with the GARFIELD interface to MAGBOLTZ [25].

53

b)a)

Figure 7.9:Photographs of Clock & Trigger Board (a) and Gadwall plus ROSIE-module (b).

pulse shape (180 ns FWHM). The Switched Capacitor Array/ADCchip (SCA/ADC) accommodates 512time-bin switched capacitor arrays, 12-bit Wilkinson ADCs(only 10 bits are effectively used) as well asoutput buffer and multiplexer. Each chip is dimensioned for16 channels, so that in total 32 channels canbe read per FEE-card. The electronics noise of this layout isas low as 1075 electrons (RMS) for 25 pFpad capacitance.

TheReadout-Controller(Fig. 7.8.c) distributes the external acquisition clock signal to the FEE-cards,monitors temperatures and voltages of and regulates the electricity supply for the FEE-cards. Afterreception of a trigger signal, it multiplexes the digital signals from up to 36 FEE-cards and sends themvia 1.2 GBit/sec optical fibre link to the data acquisition system (DAQ).

TheClock & Trigger Board(Fig. 7.9.a), realized as a VME-module, provides a clock of variable fre-quency (10 - 40 MHz) for waveform digitization and forwards the external trigger pulse to the Readout-Controller.

The DAQ systemconsists of a personal computer, the VME-computerGadwall, and theROSIE-module, which is hosted on and shares memory with the Gadwall(Fig. 7.9.b). During data acquisition,the ROSIE-module receives the data flow from the optical fibrelink and stores up to 512 waveformsamples (”time slice”) per channel to the shared memory. An especially developed software [48] runningon the Gadwall reads the data from the shared memory and writes it via 10 MBit/s ethernet into IRQ-files on a NFS-mounted disk of the personal computer. Anothersoftware, which runs continuously onthe personal computer, compresses the newly arrived IRQ-files to CPT-file format, which is used foronline-monitoring and data analysis. In addition, the personal computer is used to control Gadwall andROSIE-module through a serial RS232 connection.

For the studies presented in this dissertation, eight FEE-cards were used to read out the central 32pads of the 8 pad rows (in total 256 pads covering 40.64

�100 mm� active area). In order to minimize

electronics noise, the FEE-cards were directly plugged to the backside of the anode endcap, and theremaining 384 pads connected via 1 M� resistors to ground potential. The peaking time for signalshaping was set to 150 ns and a sampling rate of 19.66 MHz was selected (50.86 ns per time slice).

Different from conventional TPC amplification devices based on multi-wires, where the avalanchesinduce mainly positive2 signals on the readout pads, GEMs provide a purely negative electron collection

54

b)

nrmsn

1.30.3

10k

20k

30k

40k

0 1 2 30

noise level [ADC]

a)

BrmsB

226.023.3

10k

15k

20k

25k

0 256 512 768 10240

5k

baseline [ADC]

Figure 7.10:Pedestal baseline (a) and noise level (b) in 500 sample events for all 256 channels.

0 10 20 30 40 5096

128

160

192

224

256

288

measu

red

charg

e[A

DC

]

time slice [50.86ns]

detector response

baseline noise level+

undershoot

electron signal

10 15 20 25 30 35 40150

175

200

225

250

time

dela

y[n

s]

clock frequency [MHz]

a) b)

Figure 7.11: a)Time evolution of detector plus electronics response to an electron cluster in Ar-CH�(90:10), after several centimeters drift distance. The undershoot after decay of the signal is an artifact ofthe ion tail correction hardware-implemented in the front-end electronics. b) Time delay between triggersignal and first waveform sample vs. clock frequency.

signal. To adapt the circuit accordingly, the pedestal baseline had to be shifted towards positive polarityby exchanging two resistors on the FEE-cards, as illustrated in Fig. 7.8.a. Due to this modification, theaverage baseline moved to

��� � �� ADC-counts, thus making use of about��� of the 10-bit dynamic

range for negative signal detection. With a noise level average of� �� � � �� ADC-counts, the resultingratio of dynamic range to noise accounts with 174 a sufficientvalue (all values are averages for the used256 channels, determined on a run with 500 sample events, seeFig. 7.10).

Due to the absence of the ion tail in the GEM-signal, the hardware-implemented ion tail correctioncreates a signal undershoot, as visible in Fig. 7.11.b – an artifact which clearly affects the multi-track

��The gas amplification avalanches around the wires in fact induce bipolar signals on the pads. But the amplitudeof the positive component dominates over the much lower amplitude of the negative component.

55

separation achievable with this readout system. Fig. 7.11.b shows the time delay between trigger signaland first waveform sample as a function of the sampling frequency, which is needed for electron driftdistance calibration. The curve was determined with periodic, some ns lasting test pulses connected viacables of different lengths (thus a well defined delay) to trigger input on the Clock & Trigger Board andthe preamplifier inputs on the FEE-cards [50].

56

Chapter 8

Measurements in high-intensity particlebeams

The prototype GEM-TPC presented in the previous chapter wassuccessfully operated during one monthof tests without magnetic field in two high-intensity beams.These measurements, in which for the firsttime the tracking-performance of a TPC with GEM-readout in hadronic particle beams was investigated[6], are presented in this chapter.

8.1 Experimental SetupEquipped with the previously described double-GEM structure and readout electronics, the TPC was

tested in two setups in the beam lines T7 and T11 at the Proton-Synchrotron (PS) of CERN. In the firstsetup, a beam of 9 GeV

�� pions and protons with a� � �� mm� profile (FWHM) was available. The

beam line

scintillator

field cagehigh-voltage

cabling

GEM-readouthigh-voltage

cabling

gas supplywith filter

gas exhaustline

TPC

FEE-cards

Readout-Controller

Figure 8.1:Photograph of the prototype GEM-TPC installed in the T7 beam line.

57

b)

d)

a)

c)

10 100 1k 10k 100k0

5

10

15

electric field [V/cm]

drift v

elo

city [cm

/s]

m

transvers

e d

iffu

sio

n [

m for

1cm

]m

longitudin

al diffu

sio

n [

m for

1cm

]m

10 100 1k 10k 100k0

200

400

600

800

electric field [V/cm]

10 100 1k 10k 100k0

200

400

600

800

electric field [V/cm]

transvers

e d

iffu

sio

n [

m for

1cm

]m

Ar-CO (70:30)2

Ar-CH (95:5)4

Ar-CO (70:30)2

Ar-CH (95:5)4

Ar-CO (70:30)2

Ar-CH (95:5)4

Ar (93:5:2)-CH -CO4 2

Ar-CH -CO (93:5:2)4 2

Ar-CH -CO (93:5:2)4 2

10 100 1k 10k 100k0

200

400

600

800

electric field [V/cm]

Ar-CH -CO (93:5:2)4 2

B = 4T

Ar-CH -CO (93:5:2)4 2

B = 0T

Figure 8.2:Drift velocity (a), longitudinal (b) and transverse (c) diffusion of the gas-mixtures used vs.electric field as well as a comparison of transverse diffusion in Ar-CH�-CO� (93:5:2) at 0 and 4T (d).Gas data computed with the GARFIELD interface to MAGBOLTZ [25].

time structure of the beam was characterized by 550 ms lasting spills, arriving with interceptions of twoto fourteen seconds; each of the in average of 429 spills per hour consisted of up to� �� � ��� particles. Inthe latter setup, a beam of 3 GeV

�� pions and electrons with the same time structure but a slightly largerprofile, namely

�� � �� mm� (FWHM), could be used for tests.

As shown in Fig. 8.1 for the T7 setup, the chamber was orientedvertically to create straight tracksparallel to the readout pads with well defined drift distances. Two scintillators were installed in front of(� �� � �� cm�, vertically aligned) and behind (� � � �� cm� , horizontally aligned) the chamber for trigger-

ing the readout electronics. The studies were performed with three different gas-mixtures at atmosphericpressure: Ar-CH� (95:5), Ar-CO� (70:30) and Ar-CH� -CO� (93:5:2), see Fig. 8.2. As indicated in Table8.1, these mixtures show similar ionization but strongly differ in terms of drift velocity and diffusion.The different gases were supplied with an open flow system from premixed gas bottles connected to theTPC via pressure reducer, mechanical mass flow units and several meter long copper tubes. An oxisorberhas been used to keep impurities of O� and H�O in the gas-mixture as low as possible and thus to avoidelectron attachment during drift.

The electric fields were chosen to 2.5 kV/cm in the transfer gap and 3.5 kV/cm in the induction gap,

58

Gas-mixture Drift field Drift velocity Transverse Longitudinal Ionization at 1 atm.�V/cm� �

cm/�s� diffusion for 1 cm diffusion for 1 cm��� ��� �� mm�

Ar-CH� (95:5) 95� ��� � � ��� 720�m 440�m 115

10.9 nsAr-CO� (70:30) 250 0.55 160�m 158�m 116

28.7 ns310 0.70 150�m 155�m

22.1 nsAr-CH�-CO� 240 4.55 453�m 279�m 115

(93:5:2) 6.1 ns

Table 8.1:Properties of the gas-mixtures from Fig. 8.2 at the drift fields used. The drift velocity valuegiven for Ar-CH� (95:5) was measured during the beam test.

0 512 1024 1536 20480

2500

5000

7500

cluster charge [ADC]

m.p

.v.

b)

630 670 710 750 790 83010

2

103

104

105

effect

ive

gas

gain

GEM-voltage sum [V]

Ar-CH4(90:10)

Ar-CH4(95:5)

Ar-CO2(70:30)Ar-CH -CO4 2 (93:5:2)

a)

Figure 8.3:Gas gain calibration curves (a) for Ar-CH� (90:10), Ar-CH� (95:5), Ar-CO� (70:30) and Ar-CH�-CO� (93:5:2). During the beam test, the gas gain was determined in each run by comparison of themost probable value (m.p.v.) in the cluster charge spectra with reference values (b).

both gaps were 2 mm wide. The high voltage on drift cathode andfield cage was adjusted to create driftfields of

� 95 V/cm for Ar-CH� (95:5) in the T7 beam line,

� 250 V/cm for Ar-CO� (70:30) in the T7 beam line,

� 310 V/cm for Ar-CO� (70:30) in the T11 beam line and

� 240 V/cm for Ar-CH�-CO� (93:5:2) in the T11 beam line.

To ensure reliable operation with a minimized risk of discharges, the voltage across the first GEM wasalways chosen 10 V higher than the voltage across the second one (as discussed in chapter 6). This way,studies with effective gas gains in the range��� up to almost��� could safely be performed in presenceof the high-intensity hadronic beams. Fig. 8.3 shows gas gain calibration curves for the various gases.During the beam test, the gas gain was determined in each run by comparison of the the most probablevalue (m.p.v.) of the cluster charge spectra with referencevalues from Ar-CH� (90:10) calibration runsin the laboratory.

59

0 200 400 600 800 1000-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

veto

sci. 1

sci. 2

signal[

a.u

.]

time [ns]

veto

b)a)

0 200 400 600 800 1000 1200 1400 1600-50

0

50

100

GEM signal

trigger signal

time [ms]

veto veto

cu

rre

nt

[a.u

.]

550 ms

220 ms

80 ms

Figure 8.4: a)Spill time structure in T7, determined by reading the currents on the bottom electrode ofthe second GEM; trigger signals were only accepted in the 80 ms wide window defined by the beam linetrigger. b) Schematic view of the coincidence time windows for the two scintillators and the veto signalfrom the beam line.

8.2 Trigger and data takingDuring the measurements presented, the TPC readout was triggered by coincident signals from the

two scintillators. A veto signal, defined with the the beam line trigger, was used to accept trigger signalsonly in an 80 ms long time window, starting 220 ms after beginning of the spill (see Fig. 8.4). With thesesettings, the detector could be read out in conditions with up to � �� � ��� tracks / s mm�, correspondingto in average four tracks per readout cycle.

Within one month of total beam time, numerous runs for efficiency measurement and spatial resolu-tion determination have been taken. Due to the comparably slow readout rate of some Hz achievable withthe present system, only one readout cycle could be performed per spill, resulting in 225 cycles per hour.In order to record as many tracks as possible, 500 time slices, corresponding to 25.43�s, were acquiredper readout cycle. Consequently, the actual drift times, and thus drift distances of the track images couldonly be determined for the one track, which triggered the system. For all others, the drift distance isdefined by the beam position, with an uncertainty corresponding to the beam profile.

8.3 Reconstruction techniqueThe analysis of the events recorded was performed with a software developed especially for this

system [51]. This software package provides online and offline event visualization, cluster and trackreconstruction as well as various histograms for monitoring purposes. In this section, the reconstructionalgorithms and analysis methods used are briefly presented.

8.3.1 Conventions and coordinate systems

The program processes event data in CPT-file format, which contains up to 512 waveform samplesfor each of the 256 channels. The samples are written into a three-dimensional array

� � � � �� , where� is the pad column number (from 0 to 31),� is the pad row number (from 0 to 7), and� the time slicenumber (from 0 to 511). Electron drift time� and drift distance

�� can be calculated according to:

� � � � � � � � � � � � � � ��� � ��� (8.1)

60

a) b)

0 10 20 30 40 50 60-64

-32

0

32

64

96

128

time slice [50.86ns]

responsebaseline -

zero noise level+

electron signal

sig

nal am

plit

ude

[AD

C]

A

0 10 20 30 40 5096

128

160

192

224

256

288

time slice [50.86ns]

electron signal

measure

d c

harg

e[A

DC

]S

Figure 8.5:Electron signal before (a) and after pedestal subtraction and inversion (b).

�� � � � � � �� � � � � � � � � �� (8.2)

where� � � �� ���ns is the sampling pitch,��� the time delay between trigger signal and first timeslice,� ��� � ���ns the delay due to the shaper’s peaking time and�� the electron drift velocity.

When changing from the voxel coordinate system� � � �� to the spatial coordinate system� � � � � ,the following transformation is used:

�� �

���� � � � �� � �� � �� � � ��� � ��

� � � � ���� (8.3)

where� � � � ���mm,� � � �� ��mm,� � � �� ���ns; the offset of 0.5 for� and� takes care, that integervalues for� and� in the voxel system represent the center of the rectangular micro pads in the spatialsystem, the�-coordinate is chosen to be identical to the drift distance

�� .

8.3.2 Pedestal and noise determination

Pedestal baseline�

and noise level� are individually determined for each pad� � � in each eventwith the following iterative process1 (five iterations, counter� from 1 to 5):

� � � � � � �� �

�� � � � �� � �� � � � �� � � �� � � � � � � � � ��� � � � �

�� � � � � �� � � �

��� � � � �� � � � � � � �� � �� � � � �� � � � � � � � � � � ��� � � � �

where� is the number of time slices in which the mentioned conditionis fulfilled; the start values forthe iterative process are chosen to� � �

for the noise level and�� � for the pedestal baseline. In

the final step, the signal amplitudes� are calculated for each voxel (u,v,w) according to

� � � � �� � � � � � � � � � � �� (8.4)

��Please notice, that for obvious reasons the described procedure is only valid for events with low occupancy.

61

so that negative signals�

have positive sign in� (compare Fig. 8.5).

Malfunctioning channels were identified by investigation of the individual noise levels: in the currentsetup 5 channels had to be flagged as noisy and 3 channels as broken. Before launching event recon-struction, the signal amplitudes of these channels are substituted in each time slice by the average valuesof the two neighbor-channels in the adjacent pad rows.

In order to avoid unnecessary repetition of the comparably CPU-intense pedestal and noise deter-mination process, the software allows to store noise levels� � � � and signal amplitudes� � � � �� todisk. Optionally, in order to reduce the total amount of data, the signal amplitudes can be written inzero-suppressionmode: only voxels� � � �� fulfilling the criterion

�� � � � �� � � � �� � � � � � or�� � � � �� � � � �� � �� �� : average noise level (8.5)

are stored, where�� is the noise level average of all 256 pads. With this method, the disk usage of typicalevents can be reduced from 256kB down to about 20kB (corresponding to 7.8%).

8.3.3 Track reconstruction

During the beam tests, the detector was oriented with the beam line pointing into the direction ofthe long pad borders. For this reason, event reconstruction, i.e. cluster finding, space-point reconstruc-tion and tracking, have been optimized for tracks traversing the TPC in direction of the axes� and� ,respectively.

Cluster finding and space-point reconstruction

The cluster finding algorithm searches each pad row in� and� for coherent areas where the signalamplitudes exceed 3 times the noise level:� � � � �� � � � � � � � . Cluster charge

� �, linear noise level

sum� ���� and cluster noise�� consequently are calculated as:

� � � ���� ���� � � � � �� � ���� � ���� ���� � � � � �� � ���� ���� � � � � � (8.6)

Cluster sizes, both absolute and FWHM, are determined by projection of the cluster charge along the�-and� -axis. For further analysis, only clusters are accepted, which fulfill the following criteria:

� � ��

� � � ����� absolute cluster size�� � �

and FWHM cluster size�FWHM� � � for the�-coordinate

� absolute cluster size�� � � and FWHM cluster size�FWHM� � �for the� -coordinate

Space-points are determined from accepted clusters by either center-of-gravity calculation or Gaus-sian fit to the measured signal amplitude� in � and� ; the pad row number provides the correspondingvalue for the�-coordinate. For convenience, space-points are transformed into spatial coordinates byusing equation (8.3).

Tracking

Track reconstruction is done with a combinatorial track finding algorithm based on a linear trackmodel. For this process, all reconstructed space-points with

� �� � � � ���� are registered in a list and

sorted by pad row numbers. In the next step track seeds are created from the list elements located in one

62

Figure 8.6: Example event in Ar-CH�-CO� (93:5:2) after zero-suppression with seven reconstructedtracks. The user’s interface shows all three projections of the event with reconstructed space-points andtracks as well as the time evolution of the signal on one particular pad.

of the central four pad rows. In order to limit the number of seeds – and thus combinatorics – the seedscan be filtered with cuts on length and direction.

Each seed is developed iteratively into a candidate track byadding space-points situated within acone of�� or within a tube of�� �max� ��� �� � �� �� �� around the track (where� � �� �� is the residualwidth in coordinate

�). Each time a space-point is added, the track parameters areupdated. During this

process, the tracks are computed with a fast linear regression algorithm. A� �-based track fit, whichtakes individual errors for the various space-points into account, is applied in the finalizing step only.Once all candidates are developed, they are rated by the empiric track-quality function�, which is basedon reduced�-� track-� � and number of space-points in the track

�:

� � ���� � � � max min� � � �� � � � � � (8.7)

where� � �is the minimum number,

� � � the desired number of space-points per track, and thereduced�-� track-� � given by

� � � ��� � ����� �

�� � ��� �� �� �� � ��� �� �� (8.8)

with �� � the vector pointing from space-point� perpendicular to the track and�� �� the residual width of

coordinate�

in the pad row of space-point�. Amongst all track candidates with four or more space points

63

b)a)

d)c)

250

500

750

a

b

inclination [°]

-5 -4 -3 -2 -1 0 1 2 3 4 50

200

400

600

800

0 4 8 12 16 200

number of acc. tracks per readout cycle

2500

5000

7500

10000

12500

0 25 50 75 100 1250

cluster center [50.86ns]w

2500

5000

7500

10000

12500

0 8 16 240

cluster center [1.27mm]u

Figure 8.7:Cross-check of beam properties, shown for the T7 beam line in a run with Ar-CH� (95:5) and4.03 cm/�s electron drift velocity: number of tracks per 25.43 �s readout cycle (a), track inclination (b)as well as cluster centers in � (c) and in � (d).

and a reduced�-� track-� � � �� the one with the highest track-quality value is selected. All space-pointsbelonging to this track are removed from the list and the procedure resumes with track seed creation onthe remaining space-points. Track finding is finished when either the space-point list is empty or no validtracks could be newly reconstructed within the last cycle.

For eventual analysis, all selected tracks with five or more space-points and a reduced�-� track-� � � � are accepted. Fig. 8.6 shows an example event from the T11 beam line with seven reconstructedtracks.

8.4 Analysis methodsFor the measurements presented below, the TPC itself was used for tracking and as test device. To

avoid biasing, in each measurement only seven out of eight pad rows were used for tracking and theremaining one to independently determine pad row efficiencyand spatial resolution. Therefore, one rowafter the other is designated target row and the track fits arerepeated while excluding space points fromthis particular row. The target row then is searched for the space-point� with the shortest distance�

�to

the track, and the residuals (i.e. the individual components of ��

) are histogrammed. Please notice, that

64

0 2 4 60

500

1000

1500

2000

reduced track-x-z c2

0 2 4 6 8 100

2000

4000

6000

8000

number of space-points per track

b)a)

Figure 8.8:Confirmation of the correctness of the cuts applied on the number of space-points per trackand the reduced � � � track-�

�. Shape and peak position of the latter indicate properly determined

residual widths of the individual pad rows.

in these measurements the refitted tracks serve as referenceobjects and not as subjects to testing, and inorder to provide reliable results, only well reconstructedtracks, selected with additional cuts and criteria,shall be used.

For spatial resolutiondetermination tracks with a reduced�-� track-� � � � �� are considered. Theyhave to consist of six or seven space-points outside the target row, and must not involve flagged pads orborder voxels in� or � . The resolution�� in coordinate

�is given as the residual width� � �� �� without

the contribution from the track uncertainty� ����� �� . It applies

� ���� �� � � �� � ������ �� (8.9)

where the latter term can be approximated by

� ������ �� � � ���� �� : average number of space points per track (8.10)

and the resolution can be calculated by:

� �� � � �� �� �� � � ������ �� � � �� �� �� � � ��� � � ���� ���

��

(8.11)

The widths� � �� �� are determined individually for each pad row� by Gaussian fits to the residuals his-tograms, and spatial resolution finally is determined according to equation (8.11) from the average valueof the rows 1, 2, 4, 5 and 6.2

Efficiency� is understood in the context of the ability to detect a particle track in a pad row withina given spatial resolution, and thus defined as the fraction of tests where in the searched target row aspace-point fulfilling all of the following criteria could be found:

��The border rows 0 and 7 as well as row 3, which contained flaggedchannels in the center of the active area,were excluded from this analysis.

65

a)

0 2000 4000 600085

90

95

100effic

iency

[%]

effective gain

Ar-CO (70:30), 7 cm2

Ar-CH (95:5), 12 cm4

b)

0 2000 4000 600085

90

95

100

effic

iency

[%]

Ar-CO (70:30), 13 cm2

Ar-CH -CO (93:5:2), 6 cm4 2

effective gain

c)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

ave

rag

e r

esid

ua

l w

idth

[m

m]

Ar-CO (70:30), 7 cm2

Ar-CH (95:5), 12 cm4

Ar-CO (70:30), 13 cm2

Ar-CH -CO (93:5:2), 6 cm4 2

x z

d)

0 2000 4000 6000

effective gain

0

20

40

60

80

100

120

140

160

180

200

Ar-CO (70:30), 7 cm2

Ar-CH (95:5), 12 cm4

Ar-CO (70:30), 13 cm2

Ar-CH -CO (93:5:2), 6 cm4 2

0 2000 4000 6000

effective gain

Q:

cn

c

Figure 8.9: a)Pad row efficiency vs. effective gain for Ar-CO� (70:30) and Ar-CH� (95:5) in the T7 beamline. b) Pad row efficiency vs. effective gain for Ar-CO� (70:30) and Ar-CH�-CO� (93:5:2) in the T11beam line. With the improved setup, the plateau of full efficiency is reached at lower effective gas gains.c) Cluster charge to cluster noise ( �

� ��) vs. effective gain. d) Average residual width vs. effective gain.

� � ��

� � � ����� �� � � � � ��� �� �� (

�� ��� ��: required residual width in�)

� �� � � � � ��� �� �� (�� � �� �� : required residual width in�)

In this study, the pad row efficiency is measured for the six inner pad rows cumulatively and the requiredresidual widths are the plateau values achievable at elevated gas gains. Reference tracks are required toshow a reduced�-� track-� � � � �� and must not involve flagged pads or border voxels in� or � . Inaddition, a criterion was applied to avoid the use of tracks with

�-ray emission. Tracks, for which one or

more clusters fulfilling the following two conditions couldbe found, were excluded from the analysis:

� � �� � � � �� � � �� � � � ��� � �

�� �� � �� � � �� � � ��� � ��� � �� � � ��� � ���

66

b)a)

0 5 10 15 20 250.200

0.225

0.250

0.275

0.300

0.325

0.350

0.375

0.400

Ar-CH4 (95:5)

drift distance [cm]

x

zspa

tia

l re

so

lutio

n [

mm

]

-15 -10 -5 0 5 10 150.000

0.100

0.200

0.300

0.400

0.500

x

z

inclination [ ° ]

Ar-CH4 (95:5)

spa

tia

l re

so

lutio

n [

mm

]

Figure 8.10:Pad row spatial resolution in Ar-CH� (95:5) vs. drift distance (a) and track inclination (b)without magnetic field. As expected, with increasing inclination also the residual width in � increases,whereas the one in � stays unchanged.

where �� �is the average cluster charge in the track and��� the average absolute cluster size in coordinate

�in the track. This criterion has been optimized with a hundred hand-selected events from runs with all

three investigated gases containing tracks with�-rays.

8.5 MeasurementsThe analyses of the subsequently presented measurements have been carried out offline and were

based on zero-suppressed data.

Detector alignment and beam properties were cross-checkedin dedicated runs after installation of theTPC in the beam line and upon manipulations. Fig. 8.7 shows anexample of such a run from the T7beam line: the number of tracks per readout cycle indicates the beam intensity, the track inclinations in

�and

as well as cluster centers in� and� proof the proper alignment. From the latter two plots also the

beam profiles in� and� direction can be determined. Fig. 8.8 gives a confirmation ofthe correctness ofthe cuts applied during analysis on the number of space-points per track and the reduced� � � track-� � .Shape and peak position of the latter indicate properly determined residual widths of the individual padrows.

Efficiency was measured in runs where the GEM-voltages were varied while keeping all other param-eters constant. Fig. 8.9.a shows the efficiency measured in the T7 beam line for 12 cm drift in Ar-CH�(95:5) and for 7 cm drift in Ar-CO� (70:30) as a function of effective gas gain. The plateau of full ef-ficiency (�� �� � � ���) is reached with this setup at gains of� � ��� . The same type of measurementwas performed in the T11 beam line after revision of the front-end electronics grounding scheme for 13cm drift in Ar-CO� (70:30) and 6 cm drift in Ar-CH�-CO� (93:5:2). Fig. 8.9.b shows that due to theseimprovements the same plateau value (�� �� � � ���) is reached at gains of

� �� � ��� already. In Fig. 8.9.cand Fig. 8.9.d cluster charge to cluster noise (

� � � ��) and average residual width of these measurementsare shown. The plateau of achievable resolution obviously is reached for values of

� � � �� � ��; theresidual widths obtained at these values consequently had been required during efficiency determination.

By shifting and rotating the TPC in the T7 beam line, the dependence of spatial resolution in� and� on drift distance (and Fig. 8.10.a) and track inclination (Fig. 8.10.b) has been investigated in Ar-CH�(95:5). Tracks were created parallel to the pad plane, the inclination is with respect to the long padborders. Due to the absence of a magnetic field, transverse diffusion is comparably large, as evident

67

0.2

0.4

0.6

0.8

1.0

1.2

Ar-CO (70:30)2

0.0045

2.04 0.14+

4.49 0.07+

0.85 0.07+

Ar-CH (95:5)4

0.0034

3.35 0.16+

7.76 0.08+

1.42 0.08+

0 5 10 15 20 250.0

dE/dX ~ dQ/dX, trunc. mean [a.u.]

c2

A

Q

sQ

c2

A

Q

sQ

b)a)

0 5 10 15 20 250.000

0.050

0.100

0.150

0.200

0.250

Ar-CO2(70:30), 0T

Ar-CH -CO (93:5:2), 0T4 2

drift distance [cm]

spa

tia

l re

so

lutio

n [

mm

]

G=1.810.

3

G=3.010.

3

Figure 8.11: a)�� ��� -measurement with the Truncated Mean Method for 10 cm long tracks; in bothgases, resolutions �� � slightly below 20% are reached. b) Pad row spatial resolution of the � coordi-nate (1.27 mm pad pitch) vs. drift distance in Ar-CO� (70:30) and Ar-CH�-CO� (93:5:2) without magneticfield.

from Fig. 8.2, which is reflected in the moderate spatial resolutions obtained.

Fig. 8.11.a shows�� ���

-measurements from two runs determined by a Truncated Mean Methodusing six out of eight segments of 12.5 mm segment length for 10 cm long tracks. The

�� ���resolution

was found to be slightly below��� in both cases, for Ar-CO� (70:30) at� �� � ��� effective gain and for

Ar-CH� (95:5) at� � ��� effective gain. Fig. 8.11.b shows the spatial resolution in� with Ar-CO� (70:30)and Ar-CH�-CO� (93:5:2) measured in T11 for selected drift distances: in the low-diffusion gas-mixtureAr-CO� (70:30) the resolution ranges down to�� � ��m for a drift distance of 4 cm.

To estimate the ion feedback of the double-GEM structure in these conditions, the currents on cathode�� and bottom electrode of the second multiplier���� �� have been measured for two different drift fieldsas a function of the GEM-voltage sum (due to the mounted front-end electronics, the anode current�� ,which is the actual quantity of interest, could not be measured). Fig. 8.12.a shows the ratio

� � � �� �

���� �� measured in the T11 beam line with Ar-CH�-CO� (93:5:2) for drift fields of 100 V/cm and 230V/cm. Fig. 8.12.b shows an earlier measurement from the laboratory, where the ratio� � ���� ��

���was determined in dependence of the GEM-voltage for the sameexternal fields as for the second GEMin the beam test (2.5 kV/cm above and 3.5 kV/cm below the GEM):between 200 and 400 V, the ratio�is an approximately linear function of the GEM-voltage. Although the curve shown in Fig. 8.12.b wasmeasured with Ar-CO� (90:10), the data can also be applied to Ar-CH�-CO� (93:5:2): as elaborated inchapter 5, transverse diffusion in the GEM-holes and thus charge distribution are similar and extraction(i.e. ��� ) is mainly determined by the electric field configuration, i.e. GEM-voltage and external fields.With these ingredients, the effective ion feedback can be calculated as

� � � � � �, shown in Fig. 8.12.cversus effective gain. Fig. 8.12.d illustrates the linear dependence of the ion feedback on the drift field.By plotting the ratio

� � �� �� � the two curves for different drift fields come together and range –depending on the GEM-voltage – between 1.25 and 1.75.

8.6 Discussion of resultsIn the high-rate environment of 3 and 9 GeV

�� hadronic beams at the CERN Proton-Synchrotron,tracking studies with a TPC and double-GEM readout were performed. With micro pads of� ��� � �� ��mm� pad row efficiencies of�� �� � � ��� could be achieved at effective gas gains of

� �� � ��� only.

68

a)

d)c)

b)

0.75

1.00

1.25

1.50

C =

I:

BO

T,2

I A

160 200 240 280 320 360 400

GEM-voltage [V]

640 660 680 700 720 7400.1

1

10

100

GEM voltage sum [V]

E = 100V/cmD

E = 230V/cmDF

' =

I:

IC

BO

T,2

[%]

1000 10000

effective gain

0.1

1

10

100

F =

CF

'.

[%]

E = 100V/cmD

E = 230V/cmD

effective gain

F :

E/

E(

)D

I

E = 100V/cmD

E = 230V/cmD

1000 100001.00

1.25

1.50

1.75

2.00

Figure 8.12:Ion feedback estimation based on current measurements in the T11 beam with Ar-CH�-CO�(93:5:2), � � =2.5 kV/cm and � �=3.5 kV/cm. a) Ratio � � � �� ���� � �� vs. GEM-voltage sum. b) Ratio� � �� � � �� ��� vs. GEM-voltage, determined with Ar-CO� (90:10). c) Estimated effective ion feedbackcalculated as � � � � � � vs. effective gain. d) Ratio � � ��� �� � � vs. effective gain.

Spatial resolution, investigated in various gases and for different drift distances and inclinations, butwithout magnetic field, ranges down to�� � � �m.

Of course, when discussing TPC studies performed with a prototype providing limited track lengthsand in absence of magnetic fields, the question about transfer of the results to a realistic scenario hasto arise. The TESLA TDR foresees to operate in a 4 T magnetic field, reducing transverse diffusionof electrons in the TDR gas Ar-CH�-CO� (93:5:2) from 450 down to 70�m for 1 cm drift. One canattempt to extrapolate the results of the beam tests in a simplistic model, by comparing spatial resolutionsobtained in different gases and for different drift distances in terms of transverse cluster width due todiffusion. Therefore, we introduce the equivalent drift distance

��� � � as the drift distance in TDR gasat 4 T, after which a cluster shows a transverse width of� �. Fig. 8.13.a shows the spatial resolutions fromFig. 8.11.b versus this equivalent drift distance. The solid line represents the requirements mentioned inthe TESLA TDR, which are met for all measurements, where the detector was operated at effective gainsexceeding� ���� (compare Fig. 8.13.b). To extrapolate the dE/dX resolutionof

��� for 10 cm long tracks

69

0 50 100 150 200 2500.000

0.050

0.100

0.150

0.200

0.250

0.300

TESLA TDR Values

equivalent drift distance [cm]

TESLA TDR Values

Ar-CO2(70:30), 0T

Ar-CH -CO (93:5:2), 0T4 2

0 50 100 150 200 25010

2

103

104

105

eff.gain

equivalent drift distance [cm]

b)a)

Ar-CO2(70:30), 0T

Ar-CH -CO (93:5:2), 0T4 2

spa

tia

l re

so

lutio

n [

mm

]

Figure 8.13: Extrapolation to the Linear Collider TPC: pad row spatial resolution vs. equivalent driftdistance (a) and effective gas gain of the corresponding measurements (b).

to typical track lengths in the TESLA TPC (at least 125.6 cm for non-curling tracks), a rough estimatebased on the improvement of ionization statistics by a factor 12.56 can be applied:

��� �� �� ��� � � ��� .This value is close to the TESLA TDR requirement of 5%.

Eventually, one can deduce from the measurements in the beam, that a GEM-based TPC operated inAr-CH�-CO� (93:5:2) with a 230 V/cm drift, a 2.5 kV/cm transfer and 3.5 kV/cm induction field at fullefficiency, i.e.� � ��� effective gain, will provide high spatial and sufficient

�� ���resolution over the

whole drift volume. Such a device, if operated in this configuration, presumably will show an effectiveion feedback of about 9%, resulting in an ion-backflow of 270 times the ionization. If this intrinsic ionfeedback suppression is sufficient for the application at the Linear Collider is currently being evaluatedin dedicated tests and simulations.

In addition, there are indications that the use of triple-GEM devices with a non-conventional configu-ration of transfer and induction fields can further decreasethe ion feedback. The authors of reference [52]recently reported about measurements with a triple-GEM detector, where the first transfer field as wellas the induction field were set very high (6 kV/cm and 8 kV/cm, respectively) and the second transferfield extremely low (60 V/cm). The voltages were chosen to 310V across first and second and to 350 Vacross the last GEM, resulting in an effective gain of about��� . With this configuration and a drift fieldof 200 V/cm, an effective ion feedback in Ar-CH�-CO� (93:5:2) as low as 0.25% (0.5%) in presence ofa 4T (0T) magnetic field could be demonstrated. However promising, the clarification, if an operation inthis configuration provides sufficient stability against propagating discharges, despite the use of a highinduction field, and an efficient charge extraction from the second GEM and thus the efficient usage ofionization statistics to achieve high spatial and

�� ���resolution, is still pending.

70

Chapter 9

Summary

The physics program of a future Linear Collider project, envisaging center-of-mass energies up to theTeV range, requires a central tracking system with excellent momentum resolution, good multi-trackseparation, and precise measurement of the specific ionization

�� ���for particle identification. A

detector of choice for these purposes is the TPC, as it is for instance foreseen in the TDR of the TESLALinear Collider project.

In order to cope with the high track density environment resulting from the pile-up of events of severalbunch-crossings, a novel TPC readout based on the GEM-technology is considered. Key advantages ofthis technology are narrow and fast signals, improving granularity and thus two-track resolution, an in-trinsically suppressed ion feedback and almost no distortions due to

� ��effects. Besides this, it allows

for simpler mechanics and offers high flexibility in terms ofpad geometry. The major challenges are theefficient utilization ofionization statisticsin order to provide sufficient spatial and

�� ���resolution,

an operation mode offering a maximum of safety in terms of occurrence and propagation ofdischarges,a proper choice of materials to suppress detectoragingas far as possible, and finally a minimum amountof ion back-flow, which is strongly correlated to the gas gain required for efficient and precise trackreconstruction.

In this dissertation, studies focusing on the investigation of charge carrier transfer mechanisms inGEM-detectors have been presented, from which one could infer that a TPC readout with standard ge-ometry GEM-foils can be operated without losses of electrons before multiplication. Based on the resultspresented, one can further predict the charge carrier transfer properties of multi-GEM detectors and esti-mate the ion back-flow in a wide range of external fields and GEM-voltages. In addition, the underlyingmechanisms of discharges in GEM-detectors have been studied in detail and a number of precautionswere proposed, allowing to configure multi-GEM devices according to their gas gain requirements whileguaranteeing a minimized probability for the occurrence ofdestructive discharges (below���� in pres-ence of heavily ionizing

�-particles) and their propagation to readout electronics.In aging tests with

the large-size triple-GEM detectors produced for the COMPASS experiment, the relative insensitivityof GEM-detectors to aging could be demonstrated in an accelerated measurement with an accumulatedcharge of more than 7 mC/mm�, and the absence of significant outgassing of harmful substances andthus the proper choice of the materials used for this detector design could be verified.

Based on these experiences, a highly flexible general purpose R&D prototype TPC was designed andbuilt with components and materials selected with special attention to detector aging properties, i.e. fol-lowing considerations resulting from the detector aging study. One prototype chamber was equipped witha conventional double-GEM structure, micropad readout andan especially for these purposes adaptedreadout system. After first system checks in the laboratory,the prototype detector has been successfullyoperated for a total period of one month in two high-rate particle beams of pions, electrons and pro-tons. In tracking studies performed in these beams without magnetic field, the detector was operated

71

with different gas mixtures at settings optimized for electron collection and discharge stability. Besidesproviding a proof of principle for the operation of GEM-based TPCs in hadronic high-rate environments,an evaluation of the minimum effective gas gain needed for efficient and precise tracking was performedand the

�� ���-measurement precision determined with a newly developed event reconstruction soft-

ware. Pad row efficiencies of�� �� � � ��� could be demonstrated with micro pads of� ��� � �� �� mm�at effective gas gains of

� �� � ��� only. Spatial resolution, investigated without magnetic field in vari-ous gases and for different drift distances and inclinations, ranged down to�� � � �m, and the

�� ���

resolution for 10 cm long tracks was found to be as good as��� .

Finally, the results of these analyses were extrapolated tothe case of the Linear Collider TPC andcompared to the values given in the TESLA TDR; to all appearances, a TPC with double-GEM readoutoperated in Ar-CH�-CO� (93:5:2) at effective gas gains as low as� � ��� is capable to meet the aforemen-tioned requirements, showing an effective ion feedback of 9% resulting from the selected configurationof external fields. The predictions of this extrapolation could meanwhile be confirmed by tests carriedout in magnetic fields up to 4T with cosmic rays [53], underlining the outstanding potential of the GEM-technology for an efficient and stable long-term operation in a large-scale TPC as it will be required in afuture Linear Collider project.

72

Acknowledgement

In the first place, I would like to thankProf. Thomas Muller andProf. Fabio Saulifor offering the subjectspresented in this dissertation to me, as well as for instructing and supporting my research at Karlsruheuniversity and CERN during the past three years.

My work on GEM-detectors has been supervised byLeszek Ropelewski; I would like to sincerely thankhim for teaching me how to work with gaseous detectors. Many thanks also toArchana Sharma, whoaccepted to be my tutor during the CERN Doctoral Student scholarship. I would like to especiallythankJochen KaminskiandBernhard Ledermannfor the exceptionally good collaboration during all ourcommon projects. I also enjoyed much to work withBernhard Ketzer, Klaus Dehmelt, Mustafa CemAltunbasandJan Ehlers. Special thanks toAlfredo Placci, Michael Hauschild, Ron SettlesandMichaelRonanfor advising and supporting the various R&D projects. Of course, the work presented would nothave been possible without the help ofRui De Oliveira, Miranda Van StenisandTobias Barvich, whoskilfully manufactured detector components and parts of the laboratory equipment.

73

74

List of Figures

2.1 Schematic view of the TESLA Linear Collider concept . . . .. . . . . . . . . . . . . . 4

2.2 Schematic cut through one quarter of the TESLA Detector .. . . . . . . . . . . . . . . 5

2.3 Schematic cut through the TESLA TPC . . . . . . . . . . . . . . . . . .. . . . . . . . 6

3.1 Illustration of the operation principles of a TPC . . . . . .. . . . . . . . . . . . . . . . 9

3.2 Properties of common TPC gases . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . 13

3.3 Illustration of the particle identification capabilityvia�� ���

measurement . . . . . . . 14

4.1 Electron microscope photographs of a standard geometryGEM . . . . . . . . . . . . . . 17

4.2 Operation principles and electric field configuration ofthe GEM . . . . . . . . . . . . . 18

4.3 Schematic view of a multi-GEM detector . . . . . . . . . . . . . . .. . . . . . . . . . 19

4.4 Comparison of wire-based and GEM-based TPC readout . . . .. . . . . . . . . . . . . 20

4.5 Operation principles and electric field configuration ofthe first GEM in a TPC . . . . . . 21

4.6 Effective ion feedback of a double-GEM detector in Ar-CO� (70:30) . . . . . . . . . . . 22

5.1 Schematics of the current measurement with the single-GEM detector . . . . . . . . . . 28

5.2 Electron microscope photographs of GEM-foils in conical and cylindrical geometry . . . 28

5.3 Electron transmission in a standard GEM for various gas-mixtures and illustration ofelectron losses due to diffusion . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . 29

5.4 Transverse diffusion in selected gases as a function of the electric field . . . . . . . . . . 30

5.5 Electron transmission in a standard GEM for various external fields . . . . . . . . . . . 30

5.6 Electron transmission in GEMs of different geometries .. . . . . . . . . . . . . . . . . 31

5.7 Ion transmission in a standard GEM . . . . . . . . . . . . . . . . . . .. . . . . . . . . 32

5.8 Ion transmission, effective ion feedback and gas gain ina standard GEM . . . . . . . . . 33

6.1 Powering schemes for GEM-detectors . . . . . . . . . . . . . . . . .. . . . . . . . . . 36

6.2 Schematics of the discharge measurements . . . . . . . . . . . .. . . . . . . . . . . . . 37

6.3 Gas gain calibration for multi-GEM detectors and and discharge probability for a single-GEM detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .38

6.4 Discharge probability for single-, double- and triple-GEM detector. . . . . . . . . . . . . 39

75

6.5 Discharge probability vs. voltage imbalance and discharge propagation probability vs.induction field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . 39

6.6 Schematic view of the COMPASS triple-GEM detector design . . . . . . . . . . . . . . 42

6.7 Photograph, beam profile and gas gain calibration of the aging measurement setup . . . . 44

6.8 Ambient parameters during the aging measurement vs. elapsed time . . . . . . . . . . . 45

6.9 Effective gain of GEM-detector and SWPC . . . . . . . . . . . . . .. . . . . . . . . . 45

6.10 Correlation of effective gas gain and temperature to pressure ratio . . . . . . . . . . . . 46

6.11 Normalized corrected gain and energy resolution versus accumulated charge . . . . . . . 46

7.1 Photograph of the prototype TPCs . . . . . . . . . . . . . . . . . . . .. . . . . . . . . 49

7.2 Photograph of the TPC drift cylinder . . . . . . . . . . . . . . . . .. . . . . . . . . . . 50

7.3 Schematic cut through field cage foil and drift cylinder wall . . . . . . . . . . . . . . . . 51

7.4 Schematic cut through the anode endcap . . . . . . . . . . . . . . .. . . . . . . . . . . 51

7.5 Fibreglass endcap disk with micro pads and illustrationof the pitch adapter . . . . . . . 52

7.6 Gas gain calibration curve of the prototype and drift velocity measurement . . . . . . . . 52

7.7 Schematic view of front-end electronics and data acquisition system . . . . . . . . . . . 53

7.8 Photographs of FEE-card and Readout-Controller . . . . . .. . . . . . . . . . . . . . . 53

7.9 Photographs of Clock & Trigger Board and Gadwall plus ROSIE-module . . . . . . . . 54

7.10 Pedestal baseline and noise level . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . 55

7.11 Time evolution of a signal and trigger signal time delay. . . . . . . . . . . . . . . . . . 55

8.1 Photograph of the prototype GEM-TPC installed in the T7 beam line. . . . . . . . . . . 57

8.2 Properties of the gas-mixtures used during the beam tests . . . . . . . . . . . . . . . . . 58

8.3 Gain calibration curves of the gas-mixtures used duringthe beam tests . . . . . . . . . . 59

8.4 Spill time structure and schematic view of the coincidence time windows . . . . . . . . 60

8.5 Electron signal before and after pedestal subtraction and inversion . . . . . . . . . . . . 61

8.6 Example event in Ar-CH�-CO� (93:5:2) after zero-suppression with seven reconstructedtracks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

8.7 Cross-check of the T7 beam properties . . . . . . . . . . . . . . . .. . . . . . . . . . . 64

8.8 Confirmation of the correctness of the cuts applied during tracking . . . . . . . . . . . . 65

8.9 Pad row efficiency, signal to noise ratio and average residual width vs. effective gain forvarious gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 66

8.10 Pad row spatial resolution in Ar-CH� (95:5) vs. drift distance and track inclination . . . . 67

8.11�� ���

-measurement for 10 cm long tracks . . . . . . . . . . . . . . . . . . . . . .. . 68

8.12 Ion feedback estimation with Ar-CH�-CO� (93:5:2) . . . . . . . . . . . . . . . . . . . . 69

8.13 Extrapolation of the achievable spatial resolution inthe Linear Collider TPC . . . . . . . 70

76

List of Tables

6.1 List of components and materials used for the productionof the COMPASS triple-GEMdetectors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 43

7.1 List of components and materials used for the productionof the TPC prototypes . . . . . 50

8.1 Properties of the gas-mixtures used during the beam tests . . . . . . . . . . . . . . . . . 59

77

78

Bibliography

[1] J. Kaminski, S. Kappler, B. Ketzer, B. Ledermann, T. Muller, L. Ropelewski, F. Sauli,”Chargecarrier transfer in the GEM at low gas gains”, Proceedings of theInternational Workshop onLinear Colliders, Jeju Island (Korea), August 2002.

[2] F. Sauli, S. Kappler, L. Ropelewski,”Electron Collection and Ion Feedback in GEM-Based Detec-tors” , IEEE Trans. Nucl. Sci., vol. 50, pp. 803-808, 2003.

[3] S. Bachmann, A. Bressan, M. Capeans, M. Deutel, S. Kappler, B. Ketzeret al., ”Discharge studiesand prevention in the Gas Electron Multiplier (GEM), Nucl. Instrum. Methods, vol. A479, pp.294-308, 2002.

[4] M.C. Altunbas, K. Dehmelt, S. Kappler, B. Ketzer, L. Ropelewski, F. Sauli, F. Simon,”Agingmeasurements with the Gas Electron Multiplier (GEM)”, Nucl. Instrum. Methods, vol. A515, pp.249-254, 2003.

[5] S. Kappler, J. Kaminski, B. Ledermann, T. Muller, L. Ropelewski, F. Sauli,”Design and Con-struction of a GEM-TPC Prototype for R&D Purposes”, IEEE Trans. Nucl. Sci., vol. 51, no. 4,2004.

[6] S. Kappler, F. Bieser, J. Kaminski, B. Ledermann, T. Muller, M. Ronanet al., ”A GEM-TPCPrototype with Low-Noise Highly Integrated Front-End Electronics for Linear Collider Studies”,IEEE Trans. Nucl. Sci., vol. 51, no. 3, 2004.

[7] N. Akasakaet al., ”JLC Design Study”, KEK-REPORT-97-1, 1997.

[8] T. Abe et al. (American Linear Collider Working Group),”Linear Collider Physics Resource Bookfor Snowmass 2001, Part 1: Introduction”, hepex/0106055, 2001.

[9] T. Abe et al. (American Linear Collider Working Group),”Linear Collider Physics Resource Bookfor Snowmass 2001, Part 2: Higgs and supersymmetry studies”, hepex/0106056, 2001.

[10] T. Abeet al. (American Linear Collider Working Group),”Linear Collider Physics Resource Bookfor Snowmass 2001, Part 3: Studies of exotic and standard model physics”, hepex/0106057, 2001.

[11] T. Abeet al. (American Linear Collider Working Group),”Linear Collider Physics Resource Bookfor Snowmass 2001, Part 4: Theoretical, accelerator, and experimental options”, hepex/0106058,2001.

[12] F. Richardet al., ”TESLA Technical Design Report, Part I: Executive Summary”, online available:http://tesla.desy.de

[13] R. Brinkmannet al., ”TESLA Technical Design Report, Part II: The Accelerator”, online available:http://tesla.desy.de

[14] J.A. Aguilar-Saavedraet al., ”TESLA Technical Design Report, Part III: Physics at an�� �� LinearCollider” , online available:http://tesla.desy.de

79

[15] G. Alexanderet al., ”TESLA Technical Design Report, Part IV: A Detector for TESLA” , onlineavailable:http://tesla.desy.de

[16] G. Wagner,”Neutron Background Studies at the TESLA Collider”, DESY, LC-DET-2001-048,2001.

[17] T. Behnke, G. Blairet al., ”BRAHMS: a Monte Carlo for a Detector at a 500/800 GeV LinearCollider” , DESY, LC-TOOL-2001-005 , 2001.

[18] Y. Giomataris, P. Rebourgeard, J.P. Robert, G. Charpak, ”Micromegas: A High-GranularityPosition-Sensitive Gaseous Detector for High Particle-Flux Environments”, Nucl. Instrum. Meth-ods, vol. A376, pp. 29, 1996.

[19] F. Sauli,”GEM: A new concept for electron amplification in gas detectors” , Nucl. Instrum. Meth-ods, vol. A386, pp. 531, 1997.

[20] D. R. Nygren,”The Time Projection Chamber: A New 4 Pi Detector for ChargedParticles”, PEP-0144, Proceedings of PEP Summer Study, Berkeley, pp. 58-78,1975.

[21] W. Blum, L. Rolandi,”Particle Detection with drift chambers”, Springer Verlag, 1993.

[22] L.S. Brownet al., ”Physical Review D, Particles and Fields, Volume 45”, The American PhysicalSociety, 1992.

[23] F. Sauli,”Principles of Operation of Multiwire Proportional and Drift Chambers”, CERN, CERN77-09, 1977.

[24] H. Raether,”Electron Avalanches and Breakdown in Gases”, Butterworth, London, 1964.

[25] R. Veenhof,”GARFIELD, a drift-chamber simulation program. Version 7.10” , Jan. 2001, onlineavailable:http://consult.cern.ch/ writeup/garfield

[26] M. Gruwe,”Gas studies for the TPC of a detector for the future Linear Collider TESLA”, DESY,LC-DET-1999-003, 1999.

[27] M. Hauschild,”Progress in dE/dX Techniques used for Particle Identification” , Nucl. Instrum.Methods, vol. A379, pp. 436, 1996.

[28] M. Hoch, ”Development of Fast Tracking Detectors: Micro Strip Gas Chamber and GEM GasElectron Multiplier”, dissertation atTechnische Universitat Wien, 1998.

[29] Ansoft, ”Maxwell” , documentation online available:http://www.ansoft.com/products/em/max3d/index.cfm.

[30] S. Bachmann, L. Bressan, L. Ropelewski, F. Sauli, A. Sharma, D. Mormann,”Charge amplificationand transfer processes in the gas electron multiplier”, Nucl. Instrum. Methods, vol. A438 pp. 376,1999.

[31] C. Altunbas, M. Capeans, K. Dehmelt, J. Ehlers, J. Friedrich, I. Konorovet al., ”Construction, testand commissioning of the triple-GEM tracking detector for COMPASS”, Nucl. Instrum. Methods,vol. A490, pp. 177-203, 2002.

[32] F. Sauli,”Gas detectors: achievements and trends”, Nucl. Instrum. Methods, vol. A461, pp. 47,2001.

[33] S. Bachmann, A. Bressan, S. Kappler, B. Ketzer, M. Deutel, L. Ropelewskiet al., ”Developmentand applications of the Gas Electron Multiplier”, Nucl. Instrum. Methods, vol. A471, pp. 115,2001.

80

[34] J. Benlloch, A. Bressan, M. Capeans, M. Gruwe, M. Hoch, J. C. Labbeet al., ”Further develop-ments and beam tests of the Gas Electron Multiplier (GEM)”, Nucl. Instrum. Methods A419, pp.410, 1998.

[35] J. Kaminski, S. Kappler, B. Ketzer, B. Ledermann, T. Muller, L. Ropelewski, F. Sauli,”Status ofGEM-TPC studies”, presented at theInternational Workshop on Physics and Detectors for aLinear Collider, St Malo (France), April 12-15, 2002.

[36] J. Benlloch, A. Bressan, C. Buttner, M. Capeans, M. Gruwe, M. Hochet al., ”Development of theGas Electron Multiplier (GEM)”, IEEE Trans. Nucl. Sci., vol. 45, pp. 234, 1998.

[37] B. Ketzer, M.C. Altunbas, K. Dehmelt, J. Ehlers, J. Friedrich, B. Grube, S. Kappleret al., ”TripleGEM Tracking Detectors for COMPASS”, IEEE Trans. Nucl. Sci., vol. 49, pp. 2403-2410, 2002.

[38] M. Ziegler, P. Sievers, U. Straumann,”A triple GEM detector with two-dimensional readout”,LHCb Int. Note 2000-056, 2000.

[39] A. Bressan, J.C. Labbe, P. Pagano, L. Ropelewski, F. Sauli, ”Two-dimensional readout of GEMdetectors”, Nucl. Instrum. Methods, vol. A425, pp. 262, 1999.

[40] S. Bachmann, A. Bressan, M. Deutel, S. Kappler, B. Ketzer, L. Ropelewskiet al., ”Performanceof GEM detectors in high intensity particle beams”, Nucl. Instrum. Methods, vol. A470, pp. 548,2001.

[41] J.P. Perroud,Test Beam Results Of The Precision Gaseous Detectors For TheLhcb Inner Tracker”,Nucl. Instrum. Methods, vol. A453, pp. 159, 2000.

[42] J. Ostling, M. Wallmark, A. Brahme, M. Danielsson, C. Jacobaeus, P. Fonte, V. Peskov,Prog.Biomed. Opt. Imaging, vol. 3977, pp. 84, 2000.

[43] The COMPASS Collaboration,”COMPASS: A proposal for a COmmon Muon and Proton Appara-tus for Structure and Spectroscopy”, CERN, CERN/SPSLC 96-14, SPSC/P 297, 1996.

[44] R. Bouclier, M. Capeans, C. Garabatos, F. Sauli, K. Silander, ”Effects of outgassing from somematerials on gas chamber aging”, Nucl. Instrum. Methods, vol. A350, pp. 464-469, 1994.

[45] M.J. French, L.L. Jones, Q. Morrissey, A. Neviani, R. Turchetta, J. Fulcheret al., Design AndResults From The APV25, A Deep Sub-Micron CMOS Front-End Chip For The CMS Tracker, Nucl.Instrum. Methods, vol. A466, pp. 359-365, 2001.

[46] J. Va’vra, Physics and chemistry of aging – early developments, Nucl. Instrum. Methods, vol.A515, pp. 1-14, 2003.

[47] M. Anderson, R. Bossingham, F. Bieser, D. Cebra, E.L. Hjort, S.R. Kleinet al., ”A readout systemfor the STAR time projection chamber”, Nucl. Instrum. Methods, vol. A499, pp. 679-691, 2003.

[48] M. Ronan, B. Ledermann,”sq irq” , Gadwall readout software to write data from the Gadwallmodule to IRQ-files,2002-2004.

[49] M. Ronan, S. Kappler,”readIRQFiles”, JAVA software to convert IRQ-files into CPT-files,2002.

[50] J. Kaminski, B. Ledermann,”Readout of MPGD TPCs with the STAR electronics”, IEKP internalnote in preparation.

[51] S. Kappler,”The GTPC Package: Tracking and Analysis Software for GEM TPCs”, presented atthe Linear Collider TPC R&D Meeting, LBNL, Berkeley, Californi a (USA), October 18-19,2003.

81

[52] M. Killenberg, S. Lotze, J. Mnich, A. Munnich, S. Roth,F. Sefkowet al., ”Charge Transfer andCharge Broadening of GEM Structures in High Magnetic Fields” , DESY, LC-DET-2004-003,2004.

[53] J. Kaminski, M. Ball, F. Bieser, M. Janssen, S. Kappler,B. Ledermannet al., ”Development andstudies of a Time Projection Chamber with GEMs”, presented at theVienna Conference on In-strumentation, Vienna (Austria), Feb. 16-21, 2004.

82