Nuclear Instruments and Methods in Physics Research A 454 (2000) 267}271

3D simulation of charge transfer in a Gas Electron Multiplier(GEM) and comparison to experimentq

Archana Sharma!,",*!CERN, PPE Division, 1211 Geneva 23, Switzerland

"GSI, Planckstr.1, D-6429, Darmstadt, Germany

Abstract

A three-dimensional simulation of the electric "eld and avalanche propagation in a Gas Electron Multiplier isperformed. Results on charge transport are compared to experiment and agree within experimental errors; avalanchemechanism and positive ion feedback are studied. The possibilities of single photon detection with full e$ciency frominternal photocathodes are investigated. ( 2000 Elsevier Science B.V. All rights reserved.

Keywords: Gas electron multiplier; Avalanche; Single photon

1. Introduction

In the last few years, a variety of micropatterndetectors [1] has invaded the scene of chargedparticle tracking in a hostile high luminosity envi-ronment replacing the traditional multiwire cham-bers with their higher rate. Made with simpleprinted circuit board technology, with throughholes etched on double-sided metallized kaptonfoils typically 50 lm thick, the Gas Electron Multi-plier (GEM) [2] has been demonstrated to be a ro-bust charged particle detector. Two foils in cascadeform a double GEM [3}5], delimited by a driftelectrode above the "rst foil and a signal collection

qPaper presented at the Symposium on Applications of Par-ticle Detectors in Medicine, Biology and Astrophysics, Siegen,Germany, 6}8 October 1999.

*Corresponding address: CERN, PPE Division, 1211 Geneva23, Switzerland.

E-mail address: archana.sharma@cern.ch (A. Sharma).

electrode below the second. Due to its design, posi-tive ion feedback into the drift is reduced as com-pared to that of a wire chamber [6]. In this worka "rst attempt is made to model the operation ofa GEM using a 3D simulation. The electron driftproperties are investigated and transparency hasbeen computed and compared to experimentalresults; gain and positive ion feedback are alsoestimated.

2. GEM operation: two-dimensional (2D) modeland its shortcomings

Fig. 1 shows the 2D electric "eld in a single GEMcomputed using MAXWELL [7] and GARFIELD[8]. The geometry used throughout this work isas follows unless stated otherwise: 70lm metalhole diameter, 140 lm pitch with staggered rows.With appropriate potentials on the drift electrodesand across the GEM, and a grounded collection

0168-9002/00/$ - see front matter ( 2000 Elsevier Science B.V. All rights reserved.PII: S 0 1 6 8 - 9 0 0 2 ( 0 0 ) 0 0 8 3 9 - 1 NEW FIELDS

Fig. 1. Principle of the GEM drift lines computed from a 2Delectric "eld.

Fig. 2. Comparison of 2D and 3D electric "eld computed alongthe GEM channel.

Fig. 3. Drift of electrons from a track into several GEM holes,the holes, the holes have been sketched by hand to guide the eye.

electrode, electrons enter the drift volume andare multiplied in the high electric "elds in theGEM channels (E

H). The resulting avalanche of elec-

trons provides su$cient gain for charged particledetection.

3. Three-dimensional (3D) model of GEM andelectric 5elds

A 3D model was made using MAXWELL "eldsimulator. The model consists of a basic mirrorsymmetric cell and volume, with a drift and collec-tion electrode on the top and bottom, respec-tively [9]. The "eld computed by using the 3Dmodel di!ers from that of the 2D as exempli"ed inFig. 2. This is due to the metallic surfaces presentboth on top and the bottom of the GEM unac-counted for in the 2D model, as well as the doubleconical kapton well on either side. The electric"elds computed by Maxwell are imported in Gar-"eld for subsequent electron transport study. Fig. 3shows drift of electrons created by a track; one canappreciate the 3D nature of the problem, observingthat some of them go into neighboring staggeredholes.

3.1. Transparency of the GEM

More than on individual "elds, the electricaltransparency of the GEM depends on the ratio ofdrift to the dipole "eld (E

D/E

H), and its optical

transparency [10]; a staggered matrix increasestransparency as compared to straight rows of holes.Relative measurements of electrical transparencyhave been reported [10]. Within the 3D model,

268 A. Sharma / Nuclear Instruments and Methods in Physics Research A 454 (2000) 267}271

Fig. 4. Computed and measured transparency.

1The Monte Carlo version of electron drift was used inGar"eld.

Fig. 5. E!ective gain computation and measurements.

2The signal will also depend on the pre-ampli"er shapingtime, an e!ect which is not discussed here.

transparency has been computed generating uni-formly a matrix of 2500 electrons on the drift elec-trode surface and following their path as they driftand di!use down the channel.1 Fig. 4 shows thecomputed versus measured transparency fora single GEM. The experimental values fall be-tween those computed for holes with 70 and 80 lmdiameter. This could well be the tolerance of themanufacturing process. Calculations for transpar-ency in a non-zero magnetic "eld will be presentedin Section 4.

3.2. Avalanche, ewective gain and positive ions

The previous sections described how electronsmove from the drift region into the GEM channels,and accounted for the loss of the electrons simplydue to drift and di!usion. When electrons encoun-ter the high "eld in the holes, they experience ioni-zing collisions thereby resulting in an avalanche ofelectrons, whose size depends on the dipole "eld.Some of the electrons of the avalanche are lost tothe bottom electrode of the GEM, consequently the&e!ective' or visible gain of the multiplier is lowerthan the total number of ionizing collisions e!ectedby a single electron [5]. The reliability of computa-

tions in this respect is quite low, since the Town-send coe$cient is not very well known at high "elds[11]; gains obtained from calculations di!er up tofactors of 2}3 when increasing the GEM voltage, asshown in Fig. 5. Detailed understanding of thisdiscrepancy is being studied. The majority of theelectrons are produced in the center; a doughnut ofelectron production is seen at the lower metal edge,where the electric "eld is higher. This is a totallylocal e!ect; di!usion completely overtakes the "eldstructure in GEM and &200lm below the GEMsurface there is no trace of this e!ect; thus obliterat-ing the GEM structure for any localizationmeasurements. This results in making the mecha-nical alignment between two foils (double GEM)redundant. Positive ions are produced essentiallyin the whole channel but mostly in the vicinity ofthe lower GEM electrode; and they move to thedrift volume, the fraction depending strongly on thedrift "eld [10]. It should be noted, however, thatthe signal detected on readout (strips/pads) is total-ly due to electron collection, there is no slowcomponent due to the positive ion movement ascompared to a traditional MWPC sense wiresignal.2 The time taken by the positive ions toreach the GEM top typically corresponds to a

A. Sharma / Nuclear Instruments and Methods in Physics Research A 454 (2000) 267}271 269

NEW FIELDS

Fig. 6. Positive ion feedback into the drift volume, induction"eld: "eld between GEM and anode structure.

Fig. 7. Drift lines in a magnetic "eld perpendicular to driftdirection.

Fig. 8. Transparency and gain as a function of magnetic "eld(open circles for transparency, closed circles for gain).

few ls and few tens of ls to reach the drift elec-trode. The fractional number of positive ion feed-back was also veri"ed for a couple of voltagesettings points at low drift "elds, in general theagreement with Ref. [10] is within 20%, as shownin Fig. 6.

4. Operation of GEM in a high magnetic 5eld

The qualitative behavior of performance in thepresence of a strong electric "eld perpendicular tothe drift "eld was investigated with the 2D model[10] (see Fig. 7). Despite several drift lines end-ing on the bottom of the foil, the lateral spreadof the avalanche is enough to compensate forthe loss of electrons due to the Lorentz force.Fig. 8 shows the computed transparency and gainas a function of magnetic "eld (perpendicular tothe electric "eld) with an Ar}CO

2(70}30) gas

mixture, and the electric "eld values as shown inthe inset. One can see that while transparencydrops dramatically (note that for these computa-tions, a low drift "eld was chosen), there is noperceivable e!ect on the gain; measurements re-porting no e$ciency drop in the presence of amagnetic "eld have been published earlier [4].The gas mixture as well as "eld con"guration is byno means optimized; more work is needed in thisdirection.

5. Single electron/photon detection with a GasElectron Multiplier

Large gains have been reported earlier in purenoble gases with single and cascaded GEMs [12].Single photoelectrons from an internal photo-cathode may be detected exploiting the large di!u-sion in pure argon and having electric "elds setupsuch that the ampli"cation is divided into two

270 A. Sharma / Nuclear Instruments and Methods in Physics Research A 454 (2000) 267}271

Fig. 9. Average number of electrons from an avalanche transfer-red through the GEM channel as a function of the distance fromthe centre of a hole and the point equidistant from three adja-cent holes.

Fig. 10. E$ciency as a function of the average number of elec-trons transmitted through the channel.

parts: parallel plate mode in the drift and sub-sequent avalanche development in the channel.E$ciency was investigated, especially for the worstpoint of the photocathode, namely that equidistantfrom the centers of three adjacent holes. Twohundred avalanches were generated from electronsdistributed along a line on the photocathodecorresponding to that drawn from the center ofa hole to the worst point described above. Whendi!usion is switched o! in the program, one gets theexpected low values for transparency, (see curve2 in Fig. 9). With an overall gain of the order of fewtens, in the total system (drift#GEM), there ishowever a non-zero number of electrons transmit-ted through the channel; curve 1. From lines 1 and3, one can see that the number of collected electronsincreases slightly o! the center of the GEM hole,due to the slightly higher "elds and then this num-ber drops since the "eld lines end up on the metalsurface. Moving further along this line, the slightincrease is due to the fact that more electrons "ndneighboring holes as another row of staggeredholes is made available. When gain is high, thesevariations are obliterated by the avalanche statis-tics and di!usion: line 3 of the "gure shows that thevariation is much smaller. Nevertheless, one needsto increase the overall electric "eld in the system,raising the probability of photon feedback. There-fore a compromise has to be found between gainvariation and total e$ciency for single electrontransfer, obviously an experimental issue.

Fig. 10 shows the e$ciency as a function of SNT(N being the number of electrons transmitted elec-trons through the channel for each avalanche) com-puted only for electrons starting from this worstpoint on the photocathode. It is seen that for lowvisible gains, the e$ciency is approximately70}80 %, while once SNT reaches around 50, thee$ciency is 98%, and then does not vary with gain.Therefore, given a minimum e!ective gain, an al-most fully e$cient electron detection is predictedusing a noble gas.

References

[1] F. Sauli, A. Sharma, CERN-EP/99}69, Ann. Rev. Nucl.and Part. Sci. 99 (1999) 341.

[2] F. Sauli, Nucl. Instr. and Meth. A 386 (1997) 531.[3] J. Benlloch et al., IEEE Trans. Nucl. Sci. NS-45 (1998) 531.[4] C. BuK ttner et al, Nucl. Instr. and Meth. A 409 (1998) 79.[5] J. Benlloch et al, Nucl. Instr. and Meth. A 419 (1998) 410.[6] F. Sauli, CERN-EP-TA1, GEM Readout of the Time

Projection Chamber, Internal Report, 1999.[7] MAXWELL Commercial Finite Element Computation

Package, Ansoft Co. Pittsburg, PA.[8] GARFIELD CERN Wire Chamber Field & Transport

Computation Program written by R. Veenhof , Version6.33.

[9] T. Motos-Lopez, A. Sharma, CERN-IT/99-5.[10] S. Bachmann et al., CERN-EP/99-48, Nucl. Instr. and

Meth. A 438 (1999) 376.[11] F. Sauli, A. Sharma, Nucl. Instr. and Meth. A 323 (1992)

280.[12] A. Buzulutskov et al., Nucl. Instr. and Meth. A 443 (2000)

164.

A. Sharma / Nuclear Instruments and Methods in Physics Research A 454 (2000) 267}271 271

NEW FIELDS