cern, geneva. switzerland - kfki

NUCLEAR INSTRUMENTS AND METHODS 162 (1979) 405-428 ; © NORTH-HOLLAND PUBLISHING CO .

MULTIWIRE PROPORTIONAL CHAMBERS AND DRIFT CHAMBERS

G. CHARPAK and F. SAULI

CERN, Geneva. Switzerland

1 . IntroductionProportional and drift chamber operation de-

pends on the different processes associated withthe motion and the interactions of ions and elec-trons in gases under the influence of electricfields, and the basic laws of the underlying phen-omena have been known for decades' -3 ). Theyhave been extensively analysed in the develop-ment of many kinds of particle detectors, whichhave been essential tools used during variousphases of nuclear physics research : ionizationchambers, Geiger counters, proportional counters,spark chambers'-').The work undertaken at CERN in 1967 and

1968 8) brought to light a few essential features ofmultiwire structures and resulted in very fast de-velopment?). These structures had previously beenused on some rare occasions, but without theirmost important properties being perceived. It wasimmediately clear that with the help of the for-midable arsenal of modern electronics a new gen-eration of particle detectors was at hand .Among the early observations that appeared to

be significant we may mention that :- Arrays of anode wires, stretched even at small

distances from each other, sandwiched betweentwo cathode surfaces, constitute independentdetectors.

- The motion of the positive ions permits the lo-calization of avalanches along the wire by usingthe pulses induced on appropriate cathodes.This showed that the chambers were a new,adequate tool for low-energy X-rays or neutronimaging'°) .

- The time lag between the ionizing event andthe appearance of a pulse could be used forhigh-accuracy position measurements'°).

-A spatial resolution of 100jcm was readily ob-tained by measuring the drift-titter in simplestructures").After 10 years of intensive work by many ta-

lented groups, and the use of novel ideas, a greatvariety of gaseous detectors 12 ) are now being used

in nuclear science, in high-energy physics and inmany other fields of applied research 13 ).Various physical phenomena contro! the proper-

ties of these detectors''). Among the most import-ant we may mention:- The energy loss distribution of the radiation be-

ing detected . While in most applications no use-ful information is required from the pulseheights, there are cases of growing importancewhere it is necessary to have a response propor-tional to the energy losses : transition radiationdetectors, identification of particles in the relat-ivistic rise region, etc.

- The drift of electrons and ions in gases underthe influence of moderate electric fields, wherethey do not experience ionizing collisions.

- The multiplication of electrons in short-rangeavalanches produced by the very intense fieldsin the vicinity of the anode wires .

- The propagation of discharges over large dis-tances, mediated mainly by photons emitted bythe atoms excited in the avalanche process.

- The electrostatic properties of multiwire struc-tures which control the charge distribution onthe different electrodes and the relative intensi-ty of the electric fields").

- The charge distributions induced on the differ-ent electrodes by the motion of the liberatedions").Within the scope of this article we will not de-

tail all these factors and refer the reader instead tothe quoted references .We would rather like to clarify, in a qualitative

way, the relations of the most important propertiesof the detectors to the underlying physical phen-omena.

2. The proportional counterWe will shortly summarize here the main pro-

perties of single wire proportional counwrs andrefer the reader for more details to some literatureon the subject' .6 ).

VIII . TRACK VISUALIZATION DETECTORS

406 G . CHARPAK AND F . SAULI

Such a counter consists of a thin wire stretchedat the axis of a conducting cylinder, usually filledwith a gas of small electron affinity . The electronsliberated in the gas by an ionizing radiation driftto the positively charged wire and, when reaching

300

its vicinity, they experience inelastic ionizing col-lisions, giving rise to a multiplicative avalanche . -- 200The electrons liberated in the avalanche are col-lected by the wire, while the positive ions movetowards the cathodes. The pulse detected on thewire is mainly produced by the motion of the pos-itive ions. Indeed the signal induced on a collect-ing electrode is proportional to the potential excur-sion of the moving charges . The electrons, beingproduced very close to the wire, experience a smallpotential drop despite the very high field, whilethe positive ions have to travel all along the pathbetween the immediate proximity of the anodewire, up to the cathode. This is true in the vastmajority of cases, while there are some exceptionsfor high gain operation as we will see later . Theconditions of amplification in multiwire chambersand drift chambers are nearly the same as insingle wire counters, however their recent exten-sive use and the related developments have givenrise to a greater variety of gases being used, witha deeper understanding of their properties . Veryhigh gas amplification has been reached withoutentering into Geiger propagation or sparking, andspecial shapes of the space-time relations for thedrifting electrons have been looked for . More hasbeen learned about the distribution of avalanchesaround anode wires.Most of the new chambers use gases at atmos-

pheric pressure, although high pressures havebeen investigated in relation to an improvement ofthe spatial accuracy 16 ). Low pressures, required forthe use of the chambers in detecting heavily ion-izing particles, have led to spectacular and surpris-ing results, giving rise to detectors with time res-olution well below 1 ns, almost 50 times betterthan if operated at atmospheric pressure'') . Wewill limit our discussion here to chambers workingat atmospheric pressure. In these conditions, andin fields around 105 V/cm, the mean free path ofelectrons for ionizing collisions is close to I urnand, for gains of the order of 10 5 , the maximumradial extension of an avalanche is of the order of20km. At low gains, therefore, a single avalanchedoes not surround the wire, as generally believedso far . Quantitative measurement of the azimuthalspread of avalanches around the wire have recent-

eFig. 1. At moderate multiplication factors, avalanches producedin a cylindrical proportional counter by a localized ionizationhave a very little spread around the anode wire. The curvesshow the pasitive ion current measured for several avalanches asa function of azimuth angle around the anode; an 8 keV X-raybeam (liberating in the gas about 300 electrons) is collimatedaround B=0° . The wire diameter is 20um, the gas fillingargon-CO2-ethyl alcohol (92.5-6-1 .5)18).

ly been made"- 19 ) recording the ion currents atdifferent angles around an anode wire for radia-tions incident at a given azimuth. As an example,fig. 1 shows the angular distribution of the ionsfor different amplification factor in a mixture ofargon (92.5%), C02 (6%) and ethyl alcohol (1 .5%);the lateral extension of the avalanche increaseswith its size, and eventually surrounds the wire ata gain value that depends on the composition .Some powerful quenching mixtures have been

found where the limit of amplification is mainlyset by the space charge and reaches values around101; the density of positive ions in this case growsso high as to cancel the external field stopping anyfurther growth of the avalanche . This is the casefor the so-called magic gas'°) which was very pop-ular at the beginning of the use of multiwire pro-portional chambers when sensitive amplifiers oflow cost were no as cheap and available as now.Recent studies show, however, an interesting fea-ture of this mode of operation'9a1). Two types ofpulses seem to coexist : one corresponding to apropagation of the discharges around the wire, andanother to a radial propagation in an interruptedstreamer mode. The ultra-violet photons give riseto a string of avalanches just like a streamer in aparallel gap spark chamber, but the string is inter-rupted in the low field region far from the anodewires . In this case electrons are collected oversuch a long distance that they experience a large

potential drop and are therefore responsible formost of the fast signal. This is a case where thewell accepted idea that in a proportional chamberthe pulses are produced by the motion of positiveions only appears wrong. The characteristics of thesignal, i.e. large pulse height and fast rise-time,explain why the magic gas is still widely used .With thick wires, having for instance 100 um

diameter, another peculiar behaviour has been

~IIIIIIillllllllllllaaN~taanINNIIII111Nillli11111 ~J) UI11) `DNfI) `111111l 0

1nl~fl IÎl fn1~ÎnnNllinllllllinm11t1111lunlllulunllualnou

lûluululluaulllllluuNluullluuuuluulnulll

IaaIINII1111111NiUIIINIIIINIIIIIIIINIIIIIIIIIIIIIIIn11a11IIIIIIIIIIIIIiIIIIIIUIIIIIIIIIIII(I

fnII1n11u1n11nl1nuNI11111mNihlINNIININNIINIIIIInnIIIlIlllllll=lIIIIIIIIIIII1111111IIIIIIIIiIII11II1NIJ

(N I1111111111nluun

11iiuu=lluuuu

Illllllitl

WIINIII~uI~wINiINUlluumllnlrlairlaa~nautltannmaallumgINIII~iNalaaNitIiNllalllllllH muiit111N1aH11

niNI1N1111

VIIIllllUIIIIflainHillIlllfVIIIHIüour

VIIIIINI

b)

NfiORlulnnunnu[HIM11111111IINIIIIilllllllilllllll

IlllNalllilIIIIIIIIIIIII11NIIt/ 1111

/IIÎIÏII

MULTIWIRE PROPORTIONAL CHAMBERS

il .vuuiii ,Iii 1 .

ÎÎÎI ; '

üiïm+IIIIIIII

VIIIVIIIVIII11111VIIIVIII

Fig. 2. The multiwire proportional chamber consists of a grid ofuniformly spaced thin anode wires, sandwiched between twocathode planes. Cathodes mayeither consist of uniformconduct-ing foils, or of wire grids (a). The coordinate system usedthroughout this paperis also shown, as referred to the chamberstructure. The origin is centred on an anode wire. When asymmetricdifference of potential is applied between anodes andcathodes, an electric field develops as represented in (b) in aplate xz, perpendicular to theanodes . Twomain regions of field,leading to the particular behaviour of the counter, can be iden-tified : a rein of roughly constant feud extending in most ofthegap, and aregion of rapidly increasing fieldaround the wireswhere avalanche multiplication occurs. Moreover, low fieldregionsexist between the wires, with some consequences on thecollection properties of the chamber1 D).

observed, interpreted as a limited Geiger, propaga-tion along the anode wire22). It originates largeand shaped pulses, as in the streamer mode.

3. The multiwire proportional chamber

3.1 . GENERALITIES

A multiwire proportional chamber (MWPC) us-ually consists of a plane of equally spaced anodewires, sandwiched between two cathode planes(fig. 2a). The equipotential lines for such a si .ruc-ture (fig . 2b) show three regions of electric field . Inthe major part of the volume the field is uniform;close to the wire it varies inversely with the dis-tance r to the axis, almost exactly as in a cylin-drical counter (fig. 3) ; and just between the wiresthere is a small region of very low field . If a par-ticle produces ion pairs in the gas of such a struc-ture, the liberated electrons drift towards theanode wire. Reaching its vicinity, they find thesame conditions as in a cylindrical counter : a mul-tiplicative avalanche develops. We find the samevariety of amplification processes : an avalanchesize proportional to the number of initial electronsfor low gains, an amplification saturated by space-charge effects for high gains, a propagation by sec-ondary photons for gases with low quenching pro-perties leading either to a Geiger type propagationalong the wireS23 ) or to a streamer mode, inter-rupted or bridging the electrodes and followed bya spark.

0Y

YW

fo"

fo-1

Distance from center of wire

407

Fig. 3. Theelectric field strength (onan arbitrary scale) alongthez andx directions,as from the definitions of fig. 2.Thedifferentfield regions are identified in the figure by a short description ofthe main physical processes occurring").

Vill . TRACK VISUALIZATION DETECTORS

408

The motion of the ions liberated in the ava-lanches controls all the important information thatcan be extracted from a chamber. This is why aclear qualitative understanding of the mechanismof inducing the positive pulses seems to us a use-ful preliminary .

3.2. PULSES INDUCED BY MOVING

IONS IN A MULTIWIRE STRUCTURE

Consider a positive charge liberated in the vo-lume of a MWPC . Lines of force connect thischarge to the different

andeWhen the

charge is close to the anoi9e wire, most of the linesare ending on the wire (fig . 4) ; as it starts movingaway, however, the density of lines reaching theneighbouring anode wires or the cathodes is in-creasing and pulses of the same sign are inducedon all electrodes except the closest anode wire .This is true as long as the ions are not very farfrom the initial anode wire ; it is an essential phen-omenon in wire chamber structure and is the veryreason for the localizing properties of the wire col-lecting the electrons from an avalanche: it carriesa negative pulse, while all neighbouring wires pro-vide positive signals .

It was at first believed that this was a proof thatthe avalanche surrounds the wire . It is true thatin case of very high amplification, the electronsmay propagate around the wire either by diffusionor as a consequence of secondary electrons beingejected by the photons emitted by the excited at-

To cathode C,

To cathode C2

G . CHARPAKAND F. SAULI

Fig . 4. Schematic illustration of the induced pulses formation ina multiwire chamber. Positive ions, produced around the anodeA,, project lines of force on all surrounding electrodes ; at thebeginning, the wire Ai itself receives most of them. As ionsmove away from the wire, both the adjacent wires and thecathodes receive an increasing number of lines of force, resultingin a positive induced charge on all surrounding electrodes irres-pective of the direction of motion . When a localized positivecharge has drifted far enough from the anode plane, the polarityof the induced signals in some electrodes will eventually invert.Usually, however, only the first fast component of the inductionis exploited .

Fig. 5. Principle of localization of avalanches in a multiwireproportional chamber by the method of centre of gravity of theinduced charges. Ions leaving the vicinity of an anode wireinduce a positive charge distribution on both cathode planes ; ifthese distributions are measured on each event using a conve-nient stripping of the cathode planes along two perpendiculardirections, calculation of theircentre of gravity allows both xandy coordinates of the avalanche to be measured2r).

oms in the avalanche. However, a cylindrical dis-tribution of ions around the wire is not necessaryto justify the near equality of the fast componentof the pulse induced on the neighbouring elec-trodes.The relative distributions of the charges induced

on the various electrodes carry information ; thesedistributions have been studied in chambers hav-ing segmented cathodes, either printed circuitstrips or groups of wires, parallel and orthogonalto the anode wires2r) (fig. 5). The pulses inducedon the cathode strips and on the anode wires havebeen analysed for various conditions of gaseousamplification.

Since the pulses are produced by avalanchesvery close to the wire, the time lag between thecreation of ionization in the gas and the appear-ance of a pulse is a function of the distance of theionizing phenomenas-'e) to the sense wire. It wasrecognized in the early work on MWPCs that thisopens up the possibility of interpolating the posi-tions between wires; however, with a narrow wirespacing, like 2mm, a fundamental ambiguity ap-pears: it was considered impossible to recognizefrom which side of the wire the electrons weredrifting. This has led to the building of drift struc-tures of various types, designed to overcome theright-left ambiguity and use the drift time tomeasure position. We will see, however, that re-cent progress has led to a solution of the ambigui-ty by a proper use of the wealth of informationcarried by the different induced pulses .

3.3. A SAMPLE OF IMPORTANT PROPERTIES

To better illustrate the nature of the informationthat can be extracted from a MWPC, by properhandling of the available signals, we will reporthere some results obtained at CERN in a smallsize chamber of the kind illustrated in fig . 5 2' .2°) .Under either X-ray or charged particle irradiation,the pulse-height distribution on the cathode stripsas well as on the anode plane is recorded and an-alysed.

It is extremely important to know the spatialdistribution of an avalanche and the fluctuationsin its extension for a proper understanding of theaccuracy limitations connected with the avalanche.Sources of soft X-rays are convenient for this stu-dy, since the range of the photoelectrons can besmall enough to consider the avalanches as havinghad a punctual origin . The 1 .4 keV A1 X-rays andthe 5.9 keV X-rays emitted by "Fe sources aremostly used and the study of the centroid of theinduced pulses exhibits very clear properties .For example, fig . 6 shows the distribution of the

centre of gravity for three positions of a collimated1 .4keV X-ray team, 200um from each other,measured along the direction of the anode wires .The width of the distributions, 35um r.m.s ., isconsistently equal to the range of the photoelec-tron in the gas: dispersions intrinsic to the ava-lanche process itself appear therefore to be ex-tremely small.

100

1 .4 KeV

X-rnys

1 = 35 Fcm

y (J-))


Fig. 6. Precision of localization of a soft ([ .4keV) X-ray beamalong the y direction (parallel to the anode wires, see fig . 5) usingthe

method of centre of gravity of the induced pulses. Thecomputed coordinates for three positions of thecollimated beam*umapart are represented, showing a localization accuracy of'sum r.m.s . This value is rather close to the intrinsic physical

it as imposed by the range in the gas of the producedtotoelectronzt).

NU

U

409

Fig. 7. Localization properties in the plane xz, perpendicular tothe anodes, when fully exploiting the information provided bythe induced pulses on the cathodes . In bath pictures, the hori-zontal scale represents the measured centre-of-gravity in the xdirection . while in the vertical direction the ratio of total pulseheights induced on the two cathodes, or z coordinate, ispresented . When uniformly irradiating the chamber with a softX-ray source, the majority of reconstructed points in this repre-sentation cluster in circles centred around the anode wires, witha few events in between corresponding to energy sharingbetween wires (a). Expanding the scales around one wire, andexposing the chamber to three collimated bdams produces theimage shown in (b) : the three beam positions differ by 500pm.the middle one being centred on the wirez 1 ) .

Fig. 7a shows instead the distribution of thecentroids in. a plane orthogonal to the wires, for auniform irradiation to 5.9 keV X-rays .

In the picture, the abscissa represents the cen-troid position along the direction perpendicular tothe anode wires, the x direction as from fig . 5, andthe ordinate the ratio of the pulse heights on thetwo cathodes. Ring-shaped patterns appear,centred on each anode wire, with a diameter closeto 200 um. This proves that the photoelectrons re-leased in the gas by the X-rays, localized in space,give rise to an avalanche within a very narrow az-imuthal spread : this is best illustrated when re-cording the centroid distributions on three colli-mated source positions, hitting the anode wire andat 500 um on each side of it (fig. 7b) . It is clearlypossible to obtain, with X-ray beams orthogonal tothe plane of the chambers, the exact position ofthe initial photoelectron between the anode wires,by extrapolating back the measured azimuthal po-sition along the corresponding line of force.


41 0

With less accurate methods of determining thecentroid of the charge, the response is not a ringbut a simple point, and for a long, time it was con-sidered as a postulate that the ultimate accuracy ofwire chambers was the wire distance .

Fig . 8 shows how use of the azimuthal informa-tion improves the image quality in the radiographyof an object . The position accuracy between wiresis around 150um, which is a serious progress for2 mm wire spacing.When detecting charged particles we usually do

not deal with a single avalanche. Because of thelarge statistical fluctuations in the energy depositprocess, we cannot expect to find in this case ac-curacies as good as for the detection of X-rays . Inparticular, in the direction perpendicular to the

a

b)

Fig. 8 . Absorption radiography of . a simple object (a metal clipholding a ring) rèalized using a bi-dimensional chamber like theone illustrated in fig. 5. In (a) are represented the centres ofgravity, directly as computed from the induced charge distribu-tion ; the anode wire structure clearly appears. In (b), instead, acontinuous response has been obtained taking into account themeasured azimuthal angle of the avalanche and correctingcorrespondingly the value of the coordinate x, perpendicular tothe anode wires 24 ).

G . CHARPAK AND F . SAULI

160

120

80

40

0

Trajectories on eachside of a wire .

hCentroid -~

Fig. 9. Right-left separation in a multiwire chamber. For mini-mum ionizing charged particles orthogonal to the chamber, thecentroid of the avalanches corresponding to trajectories on theright and the left of a wire clearly separate, with an overlap in5% of the cases only corresponding to an error of about 150 armfor the few particles centred on the wires . By measuring the timeof the pulses relative to a fast detector, a high accuracy contin-uous response is also obtained for the coordinate orthogonal tothe wires of charged particles24).

anode wires, electrons produced all along the ion-ization trail are collected at different azimuthal po-sitions around the anode. Still, an accurate mea-surement of the centre of gravity in this caseshows that events are clustered in two distinctpeaks, each corresponding to a definite side of thewire (fig. 9). The peaks overlap in no more than5% of the events, corresponding to an error of

Position ~sFig. 10. Space-time correlation in a multiwire proportional cham-ber. The diagram shows the variation of the time of detection ofcharged particles as a function of the position of the track rela-tive to the wire. The right-left ambiguity can be lifted by thecentroid measurement on cathode strips parallel to the wires, asshown by the previous figure. The horizontal axis stows themeasured centre of gravity along the x direction, the scale beingdefined by the wire spacing (2.5 mm in this caw, correspondingto the lowest vertices of the image) ; the vertical axis representsthe time of collection, spanning about a 30 rs interval24).

u=150,um for those particles close to the wireand giving rise to a right-left ambiguity.

If the time lag of the anode pulse is measured,taking as reference time the passage of the particleas given by a scintillation counter, the display ofthe time versus position along the x directionshows how easy it is to interpolate between wires(fig. 10). This was not so feasible at the early stageof MWPCs since the ambiguity concerning theaide of arrival of the electrons with respect to theanode wires was introducing an error which couldbe removed only by complicated means.

In the y direction, along the anode wires, thepresence of multiple avalanches has a smaller ef-fect on the obtainable accuracy, at least for tracksperpendicular to the chamber; dispersions withabout 50jum r.m .s. have been measured for min-imum ionizing particles").

3.4. OPERATIONAL CHARACTERISTICS OF MWPCS

Since avalanche multiplication occurs in allgases, virtually any gas or gas mixture can be usedin a proportional counter. In most cases, however,the specific experimental requirements restrict thechoice to several families of compounds; lowworking voltage, high gain operation, good propor-tionality, high rate capabilities, long lifetime, fastrecovery, etc., are examples ofsometimes conflict-ing requirements . In what follows, we will brieflyoutline the main properties of different gases inthe behaviour of the proportional counter; a moredetailed discussion can be found in the bibliogra-phy' -7 .I2).Avalanche multiplication occurs in noble gases

at much lower fields than in complex molecules:this is a consequence of the many non-ionizingenergy dissipation modes available in polyatomicmolecules. Therefore, convenience of operationsuggests the use of a noble gas as the main com-ponent ; addition of other components slightly in-creases the threshold voltage. The choice withinthe family of noble gases is then dictated, at least'for the detection of minimum ionizing particles,by a high specific ionization and low cost ; thechoice falls naturally on argon. An argon-operatedcounter, however, does not allow gains in excessOf 10 3-10° without entering into a permanent dis-charge operation; indeed, during the avalancheprocess, excited and ionized atoms are formed .'she excited noble gases can return to the groundestate only through a radiative process, resulting in- .hotoelectron extraction from the cathode, which


may initiate a new avalanche very soon after theprimary.Argon ions, on the other hand, migrate to the

cathodes and are there neutralized extracting anelectron ; the balance of energy is either radiatedas a photon, or by secondary emission, i.e. extrac-tion of another electron from the metal surface.Both processes result in a delayed spurious ava-lanche : even for moderate gains, the probability ofthe processes discussed is high enough to inducea permanent régime of discharge .Polyatomic molecules have a very different be-

haviour, especially when they contain more thanfour atoms. The large amount of non-radiative ex-cited states (rotational and vibrational) allows theabsorption of photons in a wide energy range. Thisis a common property of most organic compoundsin the hydrocarbon and alcohol families, and ofseveral inorganic compounds like freons, C02 , BF3and others . The molecules dissipate the excess en-ergy either by elastic collisions, or by dissociationinto simpler radicals . The same behaviour is ob-served when a polyatomic ionized molecule neu-tralizes at the cathode : secondary emission is veryunlikely . In the neutralization, radicals recombineeither into simpler molecules (dissociation) orforming larger complexes (polymerization). Evensmall amounts of a polyatomic quencher added toa noble gas change completely the-operation of acounter, because of the lower ionization potentialthat results in a very efficient ion exchange. Goodphoton absorption and suppression of the secon-dary emission allows gains in excess of 106 to beobtained before discharge .

Unfortunately, the use of polyatomic organicgases can have a dramatic consequence on thelifetime of counters, when high fluxes of radiationare detected. In fact, some products of molecularrecombination are liquid or solid polymers. Theseproducts deposit on cathodes and anodes, depend-ing on their affinity, and substawially modify theoperation of the counter after integral fluxes ofradiation around 10'-10 8 counts per cmz when us-ing Isobutane as a quencher"). The following pro-cess then takes place (halter effect) 26 ) : when athin layer of insulator develops on the cathode, asa result of the deposit of polymers, positive ionscreated in further avalanches deposit on the outerside of the layer and only slowly leak through theinsulator to be neutralized on the cathode . Whenthe detected radiation flux grows above a thresh-old value (10 2 or 10 3 counts per second per cm2)

VI11 . TRACK VISUALIZATION DETECTORS

41 2

the production rate of the ions exceeds the leakagerate and very quickly a high density of charge de-velops across the thin layer . The dipole electricfield can be so high as to extract electrons fromthe cathode and through the insulator : a régime ofpermanent discharge is therefore induced, even ifthe original source of radiation is removed. Tem-porary suppression of the counter voltage stopsthe discharge ; the counter, however, remains da-maged and an exposure to lower and lower radi-ation fluxes will start the process again. Onlycomplete cleaning can regenerate a damaged coun-ter.

Fluxes of 108 counts per cm 2 are very quicklymet in high-energy beams having typical intensi-ties of 106/s .cm 2 . A solution to the dilemma hasbeen found by taking advantage of the ion ex-change mechanism already mentioned . If a non-polymerizing agent is chosen having its ionizationpotential lower than those of the other constitu-ents of the gas mixture, addition of even a smallquantity of the new quencher will modify the na-ture of the ions neutralized at the cathode into anon-polymerizing species. Propylic alcoholC3H,OH and methylal (OCHACH are often used.

G. CHARPAK AND F. SAULI

a)

b)

Fig. It . Pulse-height distribution in the proportional region for55Fe 5.9 keV X-rays (a) and for minimum ionizing particles (b) ina standard 1=8mm, s=2mm wire chamber. The horizontalscale corresponds to about I mV/div. (on I W) 12),

Integrated rates in excess of 1011 counts per Cm2have been measured, without alteration of thecounter properties 25). Use of organic quencherswith smaller complexity reduces the polymeriza-tion 2'), but appears to be less efficient for break-down protection .A MWPC operated at moderate gains has a ra-

ther good proportional response, as shown infig. 11, where the actual pulse-height distributionsare shown as measured under identical conditionsfor 5.9 keV X-rays and for minimum ionizing par-ticles in a 16 mm gap chamber, operating witha 60-40 argon-isobutane mixture . In both cases,the horizontal scale is about 1 #A/div. Full effi-ciency of detection for minimum ionizing particlescan be obtained with an electronic thresholdaround one tenth of the peak amplitude, i.e. about0.5 mV on 1 W; this is a typically adopted valuefor operation in the proportional region.The timing properties of a proportional chamber

are determined by the collection time of the elec-trons produced by the ionizing tracks ; the peculi-ar structure of the electric field around the wiresallows the separation of three regions, indicated asA, B, and C in fig . 12. Electrons produced in re-gion A are quickly collected (typical drift velocitiesin this region of high fields are above 5 cm/us);tracks crossing the low field region B, however,will produce a characteristic tail in the time dis-tribution . Electrons produced in region C, on theother hand, smoothly drift to the anode wherethey are amplified and collected. The time resolu-tion of a chamber is defined as the minimum gatewidth necessary on the detection electronics forfull efficiency ; it is of around 30 ns for a 2 mmspacing chamber. Fig. 13 shows the typical timedistribution observed in such a chamber, when allwires are connected together (meaning that eachtrack crosses region A or B of at least one wire),while fig. 14 shows the same spectrum for a singlewire and an inclined beam : the long uniform tail

A 8mesh

Î ' mesh

Fig. 12 . Timing properties of a multiwire proportional chamber.Depending on wherethe original charge has been deposited, onecan distinguish a "fast" region A, a "drift" region C, and anintermediate region B contributing to the tail in the fast regiontime distribution (

also fig. 1b)9 ).


Fig. 13. Typical time distribution measurement in a chamberwhen all wires are connected together (total OR) 12).

in this case corresponds to tracks crossing regionC of the considered wire .Addition to a proportional gas mixture of small

quantities of electronegative vapours, like freon(CF313r) za) or ethyl bromide (CZHS) Z8) allows themultiplication factor to be` pushed to values ashigh as 10' before breakdown, at the same timeobtaining a saturated gain condition, i .e. apulse-height distribution which is entirely inde-pendent of the amount of charge lost in the ion-izing event. This particular behaviour was exten-sively studied in the so-called "magic gas" 2°), ar-gon-isobutane-freon, in the volume proportions70-29.6-0 .4 . The appearance of saturation inthese conditions is illustrated by fig . 15, where the

Fig. 14 . Typical time distribution

red on a singlc wire foran inclined boom of minimum ionizing electronsl2).

41 3

Fig. 15. "Magic gas" operation. Pulse-height spectra measuredin a 5 mm gap, 2mm wire spacing multiwire chamber for SSFeX-rays and minimum ionizing electrons at increasing anodicvoltages, in "magic" gas, showing the, pulse-height saturationeffect2O). Notice that the horizontal scales in the three picturesdo not coincide (the average pulse height is increasing by morethan one order of magnitude from (a) to (c).

41 4

pulse-height spectra for minimum ionizing elec-trons and 5.9 keV photoelectrons are compared atincreasingly high operational voltages. In fig . 15a,the amplification is still proportional (the lowerpeak corresponds to fast electrons), in fig. 15b sat-uration appears and it is full in fig. 15c. It hasbeen proved that, under these conditions, onesingle photoelectron provides the full pulseheight").

Fig. 16. Time spectrum in a proportional chamber operated with"magic gas", for an inclined charged particle beam . By compar-ison with fig. 14, one can see the effect of the electronegativevapour addition to the gas in decreasing the effective memory ofthe chamber").

a

Y

cT

Q


RATE (sec'' mm'')3.10

Fig . 17. Space charge effect . When the flux of detected ionizing particles increases, the positive ion charge density in the chamberincreases so much as to reduce the electric field around the anode wires and therefore the gain. The curves show the measured peakcurrent for minimum ionizing tracks, at various anodic potentials (or initial gains) as a function of the particle: flux. Since the gainreduction appears to be connected to the rate per unit length of the anode wires, the horizontal scale is a linear instead of a surfwerate, the real flux is obtained by multiplying the scale by the anode wire spacing. At fixed discrimination threshold on the wireelectronics, the efficiency of detection will drop as a consequence of the gain reduction : the dashed curves in the figure stow thepoints of equal inefficiency for a 5#A threshold3°).

The amount of electronegative gas that can beused in a chamber is obviously limited by the re-quirement of full efficiency . An important conse-quence of the presence of an electronegative gas inthe mixture is a reduction of the tail in the timedistribution, owing to electron capture, as shownin fig . 16'9): the memory of the chamber is there-fore correspondingly reduced and this allows oper-ation of the detector at higher particle rates .

There is, however, an intrinsic limitation in therate capability of MWPCs owing to the formationof a positive space charge by ion produced in theavalanches and migrating to the cathodes . At highenough fluxes, the large positive charge induces amodification of the electric field around theanodes with a consequent reduction of the detect-ed pulse height"), as shown in fig . 17 ; the cham-ber ceases to be a proportional detector, and evenif used only as a digital positioning device efficien-cy will eventually drop for a fixed electronics de-tection threshold. A completeby new mode ofoperation has been recently proposed to overcomethis fundamental limitation; ').

4. Methods of read-out4.1 . ANODE WIRES AS INDIVIDUAL DETECTORS

In most cases, mainly in large detectors, eachanode wire is connected to its own circuit andused as an independent detector ; the read-outelectronics consists essentially of a simple amplifi-er-discriminator followed by some kind of digitiz-ing system . Several high-density specialized cir-cuits have been developed for this purpose and areavailable to the experimenter.

If, for tracks perpendicular to the chamber, oneexpects a single wire to be hit in most of thecases, when detecting inclined tracks, the numberof wires hit on each track (or cluster size) will ob-viously depend on the time gate on the associatedelectronics. If the gate length is the minimum al-lowed by the requirement of full efficiency(around 30 ns), the cluster size will be of one ortwo wires in a ratio depending on the angle; if, onthe other hand, the gate length corresponds to themaximum drift-time (about 200 ns for an 8 mm,ap), the cluster size will correspond to the max-imum allowed by geometry. Fig. 18 shows themeasured average cluster size as a function of the-ingle of incidence of the tracks (a=0° meansracks perpendicular to the wires plane) for a large


Z

Number of wires

Fig. 18 . Distribution of the number of wires hit, or cluster size,in a multiwire proportional chamber operated in standard condi-tions with a large coincidence gate, for several angles of inci-dence of the beamz0 ).

gate width'°). A detailed discussion on cluster sizeand efficiency can be found in Fischer et al .").When using a gas mixture containing an elec-

tronegative vapour, the cluster size is strongly re-duced as an effect of electron capture; fig . 19shows for example how the addition of freon dec-reases the average number of hit wires, still in theregion of full detection efliciency 20) .

4.2 . THE CENTRE OF GRAVITY OFCATHODE INDUCED PULSES

We have seen that the precise measurement ofinduced pulse height distributions in a chamberpermits an unambiguous bidimensional response,

Number of wires touched

41 5

Fig. 19. Distribution of the number of hit wires, for tracksinclined at 30°, as a function of the content of an electronegativevapour (CF3 B,). Reduction of the sensitive zone around thewires results in a reduced cluster sizez°).


41 6 G . CHARPAKAND F . SAULI

continuous in all directions, with an accuracy closeto 50km for the best coordinate and 150 gm forthe worst.

If a chamber is built with cathodes made ofstrips, each equipped with an amplifier, a lineardelay permitting appropriate logic decisions to betaken and a pulse-height digitizer, localization bycomputation of the centroid of the inducedcharges presents the following advantages 2'1 24) :- With quite simple amplifiers, the signal-to-

noise ratio permits, with a narrow gate of30-50 ns, accuracies of the order of 50km alongthe wire .- Such a short gate limits the useful thickness

of the chamber to a few millimetres only . Thus,good accuracies can be maintained even for veryinclined tracks . If the ionization electrons areshared among several wires the charge centroidgives also an accurate interpolation between wires ;with 2 mm wire spacing accuracies of the order of300,um between wires have been reached.- The correlation between the total pulse height

induced on the two cathodes, from a given set ofavalanches, leads to a partial removal of the am-biguities deriving from the detection of multitrackevents. In other words, not only does a single gapprovide the track coordinates, but it also servesthe purpose reached sometimes by a third cham-ber set at some angle to the anode wires of twoorthogonal chambers, when only the anode wiresare being used .The cost of such a method is too high at pres-

ent to be envisaged in most large-size detectors .However these unique features of high accuracytwo-dimensional detection may make it attractiveto solve some particularly difficult experimentalproblems. For instance a centroid read-out methodhas been chosen to implement a large size cylin-drical detectörin a solenoidal field around an in-tersection of a large electron storage ring33 ).

4.3 . ANALOG READ-OUT METHODS

Developed at a time when sophisticated elec-tronic circuits for MWPCs were not yet available,analog read-out methods still find applications ei-ther if low cost and complexity is of primary con-cern, or if the chamber geometry does not allowother solutions (like in cylindrical MWPCs) . Thesimplest ofthem makes use of electromagnetic de-lay lines, capacitively coupled34) to the anode orcathode wire planes (fig . 20) ; the time difference

START

Fig. 20. Coordinate read-out with two electromagnetic delaylines, externally coupled to the cathode wire planes. The timedifference between the anode pulse and the two outputs of thedelay lines is proportional to the xand ycoordinates, respective-ly3a).

between the occurrence of the event (as given byan external device or by the anode pulse itself)and the detection of a signal at the ends of the de-lay line represents the space coordinate.The use of cathodes made of an insulating foil

coated with a high resistance graphite layer per-mits the collection of the induced pulses on delaylines exterior to the chambers35), directly throughthe cathode planes ; this allows selective reading ofsections of the chambers . The wires themselvescan be connected so as to constitute a delayline36.37), or specially built miniature lines can re-place some of the cathode wires for direct pick-upof the induced signals3").Delay line chambers are simple to make . They

provide one of the easiest way to make cylindricaldrift chambers4t). Their use is restricted, however,to applications where the intrinsic good time res-olution of the chamber is not essential, since theyhave a time resolution equal to the total propaga-tion time, usually of several microseconds.

Localization of an avalanche along the anodewires can be obtained also by a charge divisionmethod (fig. 21). The problems encountered in this

' . .

`' 1 .4~ , , . ; .. r1 ." ; , II '.f

1 0,17

TO ADC's ORTO ANALOG DIVIDER


Fig. 21 . Localization of the avalanche coordinates using thecharge division method. In a resistive wire, the ratio of chargesdetected at the two ends is proportional to the coordinate alongthe wire. The time pickoff output can be used for fast coinci-dence selection or, in a drift chamber, to determine the secondcoordinate (see the next sectione3 ).

technique have been analysed by Radeka etal.42 - 43 ). The ratio of charges detected at the twosides of a resistive wire is proportional to thecoordinate ; accuracies of around 0.4% of the wirelength have been obtainedu). Mainly developed toallow the measurement of the coordinate alongthe axis in cylindrical muldwire chambers, themethod is intrinsically bidimensional and allowstherefore ambiguity-free reconstruction of highmultiplicity.The choice of a given analog method is some-

times a matter of taste, but is is quite often im-posed by physical requirements and represents anadaptation to the physical problems. A good illus-tration can be found in a special detector for slowneutrons using a pressurized "He filled cham-ber"). A slow rise-time of the signal, connected tothe high pressure operation of the chamber, per-mits in this case to obtain very good localizationwith a charge division method, while use ofa de-lay line would lead to very inconvenient rise-timeto delay ratios.On the other hand, delay lines are very practical

to read out with a modest accuracy ( - I%) the po-sition of avalanches along the wires of drift cham-bers with widely spaced anode' wires.

It is the art of the experimentalist to choose cor-rectly the best and cheapest read-out adapted to

41 7

his problem without being embarrassed by thegreat variety at hand.

5. Drift chambers5.1 . INTRODUCTION

A drift chamber is a device in which electronsliberated in a medium by ionizing collisions cqn bemoved away from their initial position by approp-riate electric fields and transported to well-deter-mined positions where they are detected.The most commonly used medium so far has

been the gaseous state, although liquid drift cham-bers merit serious consideration for many applica-tions, since the considerable effort invested duringthe few years spent in the investigation of the pro-perties of liquefied noble gases as detection mediafor ionizing particles has shown that they havemany properties similar to those of gases") .The displacement of the electrons can serve

many purposes :- If the precise creation time of the initial ion-ization is known, then the measurement of thatime interval between production and detectionprovides the position of the initial electrons . Theimportance of drift chambers stems from the factthat, as shown in the very early work on this sub-ject"), spatial accuracies of the order of 100,umand timing accuracies of the order of 5 ns can beachieved by rather simple means. The early appli-cation of this technique to various structuresshowed that a detector'"' -49 ) with considerablepotentialities was available .- The transport of electrons along well-definedlines of force may be used to match a given geom-etrical situation where one wants all the electronsproduced along a line of force to give a unique im-age ; for instance in the spherical drift chambers,where all the electrons produced by soft X-raysalong straight lines passing by the centre of asphere produce a single image").- Drift spaces may be filled with divided solidmatter. The solid is used to give rise to appropri-ate reactions with charged or neutral particles. Thesecondaries are detected in the gas in which thematter is embedded : the ionization electrons aremilked out from the gas and transferred to a pro-portional chamber. Efficient detectors for neutralradiations seem feasible" .52).- Large drift volumes offer interesting possibil-ities. All the information contained in a trajectory,

VI11 . TRACK VISUALIZATION DETECTORS

41 8

namely the three-dimensional position of everyelectron cluster and the energy loss along thetrack, can be measured after the transfer of the in-itial electrons to a multiwire chamber. This is thebest approximation to a purely electronic trackchamber. Considerable effort is being invested inthis promising detector under the leadership ofNygren").

Since the laws governing the drift of electron ingases are essential for the understanding and de-velopment of drift chambers we will give here ashort introduction to this field.

5.2. DRIFT AND DIFFUSION OF ELECTRONS IN GASES

The mass of the electrons is so low compared tothat of the molecules among which they movethat in elastic collisions they retain a large fractionof their momentum. Even in low electric fields theelectrons will retain energy gained from the field,and their velocity distribution, which affects theirmobility, will be very different from that of thesurrounding molecules. This subject is treated inall books related to electrical discharges in gas. Ithas been reviewed extensively in a recent articleby Palladino and Sadoulet5°) and in the thesis ofSchultz").

Since the beginning of the century a great var-iety of methods have been applied to measuredrift velocities . When one now reads the descrip-tion of different types of electron shutter methodsused in this field, one wonders why drift cham-bers were not invented much earlier, to localize:fast particles.As an example, fig . 22 shows the drift velocity

of electrons in some gases . Some salient featuresof the laws governing the drifting of electrons arethe following :

//sMMMWIMEMME

"EMNINFEMENME"EMMNMEMEMEN

aa aE aE aE a. m.q ImenaT ea


a. a . lo

di

E

.z

i

EE

d)

E/P IV/cm . mm Ho)

Drift f10d IV-1

Fig. 22. Examples of drift velocity for electrons in gases, inseveral mixtures used in proportional counters . The horizontalscale is given either as reduced fieldE/P, or directly as the fieldat I atm. The drift velocity saturation properties of some gasmixtures are apparentt " 3 .3°).

- The drifting in many pure gases is stronglymodified by impurities, even in small quantities.The effect is striking in noble gases. The additionof 1% N2 to argon changes the velocity by a factorof 5 (fig. 22a). In most detectors, complex mix-tures are used and the data concerning high-pu-rity gases are of little use.

- The dependence of the drift velocity on theelectric field varies widely with the composition ofthe gas mixtures and their nature (fig. 22b). Verytypical in this respect is the case of the Ar-C02and Ar-isobutane mixtures (figs. 22c and d).

In one case the drift velocity is a growing func-tion of the reduced field E/P while in the othercase it nearly becomes a constant. The conse-quence is the different time distribution of thepulses produced in a MWPC filled with these twomixtures. In the first case the distribution ispeaked in the short-time region (corresponding tothe .large fields around the wire) with a long tailin the long-time region (corresponding to the low-field region between the wires). In the secondcase, it is an almost rectangular distribution withsharp edges, corresponding to the nearly constantdrift velocity.

In the drift chambers, where the aim is to mea-sure the position of a particle from the time ofdrift of the electrons to the sense wires, it is clearthat one has to find gases where the velocity is in-dependent of the electric field, unless the mea-surements are limited to a region of constant field.Depending on the structure of the chamber and

the accuracy, this requirement is more or lessstringent.During the drift in electric field, electrons dif-

fuse following a Gaussian distribution ; the changein the energy distribution due to the electric fielddoes, of course, result in a coefficient of diffusiondependent on E.The r.m.s. of the distribution can be written as :

_2E"XP IQs -

~( e

T

r

where ek is the characteristic energy of the elec-tron swarm in the field E, a the gas pressure P,and x is the distance of drift . The expression al-lows inference of the main characteristics of theelectron diffusion, since ek is experimentallyknown for most pure gases, and straighforwardformulations allow computation of its value forgas mixtures 54.55). As an example, fig. 23 showsthe dependence of the characteristic energy on theelectric field for several argon-isobutane mixtures(a) and for several other gases (b). Increase of thegas pressure, at a constant value ofE/P so as topreserve the value of the drift velocity, decreasesthe diffusion, as is apparent from the quoted ex-pression .


r, , IsK in several Ar-Isobutone mixtures

provided by the theory

a)

41 9

500 1000 3000

E IWcm,of 3006K . 1 arm)

Fig. 23 . Characteristic energy of , electrons in several gases andgas mixtures, as a function of the electric field . Since the elec-tron space diffusion depends on the square root of ek, the choiceof a gas with a small characteristic energy is essential in a driftchamber if high localization accuracy is desired-54 ).


420

EV

2xlot

Fig . 24. Computed and experimental dependence of the standarddeviation of electron diffusion from the electric field for 1 cmdrift, in several gases at normal conditions54).

uvN

E

3

lot

IÔ 2

100

500 1000 2000E Iv/am, ot I atml


Fig. 24 gives, as a function of the electric field,the computed value of the standard deviation ofspace diffusion ax, for 1 cm of drift. The thermallimit is also shown, corresponding to a fictitiousgas where the energy of electrons is not increasedby the presence of the field ; carbon dioxide, be-cause of its very low characteristic energy, is veryclose to the thermal limit. For a 75-25 mixture ofargon-isobutane, often used in proportional anddrift chambers, a,,= 2(Aum independently of E.In drift chambers, one obtains the space coordi-nates of ionizing tracks from the measurement ofthe drift-time, in a more or less uniform field, ofthe electron swarm. A small diffusion coefficientleads of course to a better accuracy ; the choice ofcarbon dioxide is, however, often forbidden by thepoor quenching properties of this gas in propor-tional counters and the quoted argon-isobutanemixture is preferred.

Notice that the limiting accuracy with whichone can localize the drifting swarm is not directlygiven by ax, but by its variance, depending on thenumber of electrons necessary to trigger the time-measuring device. For example, if the average

0.5

1 .0

1.5E

(kV /crn)Fig. 25. Measured drift velocity of electrons, along the direction of drift, as a function of the electric field for several values ofmagnetic field H (perpendicular to E). Curves of this kind depend of course on the gas mixture used, in this case argon-isoinr"tone-methylal (67 .2-303-2 .5) 3°).

time of the n drifting electrons is measured, the li-miting accuracy will be tr.,IJn ; on the other hand,if a simple threshold discriminator is used, sensi-tive essentially to a single electron avalanche, oneexpects to find a limiting accuracy given byQ, = (in n)-l or., 54-56).The presence of an external magnetic field

strongly modifies the drift and diffusion propertiesof electrons ; this point has been carefully investi-gated since drift chambers are often operated in-side large spectrometer magnets . In very first ap-proximation, ifw�=wo(E/P) is the drift velocity inthe absence of magnetic field, when applying afield X perpendicular to E, the new value of thedrift velocity is given byz.54)

1

3HWH

=WOJ~11_(fWO

,

f= 2 E'and the electron swarm drifts at an angle B in re-spect to the electric field direction, as given bytg B = wo . Therefore, when increasing the gaspressure at a constant value of E/P, the tangentof the drift angle is correspondingly reduced .

9


H ( kc )

42 1

It appears, however, that such a simple formu-lation only allows a qualitative estimate of the ef-fect of very high magnetic fields in the tesla re-gion. More sophisticated methods of. calculationexist in the framework of the quoted theory ofelectron transport in gases54.55) ; in many cases,however, a direct experimental measurement of.the effect may be necessary . As an example,figs . 25 and 26 show, respectively, the measureddrift velocity and angle as a function of perpend-icular electric and magnetic field, for an ar-gon-isobutane-methylal gas mixture_used in highaccuracy drift chambers'°).A magnetic field does modify also the diffusion

prc-peldes of electrons ; a very strong reduction ofthe, transverse diffusion has been measured in thecase where electric and magnetic field are paral-lel 53), and this property greatly improves the spaceaccuracy in three-dimensional drift chambers .

5.3 . DRIFT-CHAMBER GEOMETRY

.- In its basic form, a single-cell drift chamber con-sists of a region of moderate electric field, fol-

Fig. 26 . Measured drift angle, i .e. the angle between electric fieldand drift direction, as a function of electric and magnetic fields, forseveral gas mixtures as indicated . The magnetic field direction is perpendicular to E. An increase in the gas density reduces the driftangle, at a given value of etectric and magnetic fields 3O) .


422

x =

w dt,ta

Én

3.5 KVAnode050t'm

Cnnrged particle

Fig . 27. Principle of operation of a single-cell drill chamber. Aset of cathode wires, at suitable potentials, generate in the drillspace a region of uniform field . The electrons produced by anionizing event migrate to one end of the celhz).

lowed by a proportional counter (fig. 27). Suitablefield shaping electrodes, wires or strips, as shownin the figure, allow one to obtain the desired elec-trical configuration. Electrons produced at time toby the incoming charged particle migrate againstthe electric field with velocity w, and reach theanode wire where avalanche multiplication occursat a time r, . The coordinate of the track, in respectto the anode wire, is therefore given by

n

Field wires

Anode wire

G . CHARPAK AND F. SAULI

a)

Drift voltage-MVI

scintillation counter

30

Egalpolanlial Ones

equipotentials in a chamber having 30 mm gap and 60 mm between anode wires').

which reduces, for a constant drift velocity, tox=(r,-ra)w . It is obviously very convenient tohave a linear space-time relationship, and this canbe obtained in structures with uniform electricfield. If a large surface of detection is required,however, a simple structure like the one shownleads to unconfortably large working voltages andvery long drift times; nevertheless, chambers ofthis kind having as much as 50cm drift lengthshave been operated, with an over-all drift voltagearound 50 kV and maximum drift-time (o.- mem-ory) of 7rots . The construction of a large 1 GeVspectrometer equipped with such chambers hadbeet undertaken in 1969 and resulted in a remar-kably simple detection system"). For even largersurfaces, or in cases where shorter memory timesare necessary because of the expected particlerates, a multicell structure can be used ; in thiscase, since the region of the anode wire becomesnecessarily part of the active volume, it is not pos-sible to obtain a constant drift field all across thecell .

In principle, a structure identical to that of aMWPC can be used to realize a multiwire driftchamber; however, the low-field region between

-2KVCathode0200/rm

b)

Fig . 28 . Principle of the multiwire drill chambers with uniform cathode planes : (a) the basic geometry and (b) the electric field

Fig. 29. Equipotentials computed in the Harvard-MIT drift chambers. The field wire has been replaced by an 1-shaped profile,insulated from the cathode planes,and serving both the purpose of field reinforcing element and of mechanical stiffener . The gap is26 mm, the anode wire spacing about 10 cm 59).

the anode wires would result in a strong non-lin-earity of the space-time relationship, especially for;large wire spacings . A modification of the originalproportional chamber structure allows the elimin-ation of low field regions in the central plane4a).The anode wires are alternated with thick field-shaping cathode wires that reinforce the electricfield in the critical regions, as shown in fig. 28 5').Chambers of this kind have been built, with widerwire spacings, up to sizes of about 4 x4 m2 58).

Other designs, similar in principle to the one de-scribed, have been developed, which allow a sim-pler construction of large surface, mechanicallyvery stiff, drift chambers ; fig . 29 shows one exam-ple"). Thin aluminium profiles (I-beam), insulated'from the cathode planes and kept at a negativepotential, serve both the purpose of mechanicalspacers and field-reinforcing electrodes. The ca-thodes are grounded, while the anode wires are"maintained at a positive potential to collect andamplify the electrons .

Screening

. .i

. . . .

. . . . . . .

. . . . . .

. .

. .i

. ."~Cotho~ .

elect,...

. . . . . .

drift wires. . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . .


42 3

Insulator

The major limitation of the structures represent-ed lies in the fact that, in order to obtain a rela-tively uniform drift field, the ratio of the gaplength to the wire spacing has to be maintainedclose to unity . For typical convenient wire spac-ings (5-10 cm) this implies rather thick chambers,and therefore a reduced packaging density . More-over, it takes a long time to collect at the anodeall the electrons produced by a track, and thereforemultitrack capability per wire is excluded. Theseconsiderations have led to the development of thestructure shown in fig . 30 °9). Two sets of parallelcathode wires are connected to increasingly highnegative potentials on both sides starting from thecentre of a basic cell ; the anode wire is main-tained at a positive potential, and two field wires,at the potential of the adjacent cathode wires,sharpen the transition from one cell to the next.The equipotential lines are shown on the figure,for a typical choice of operational voltages ; a uni-form field drift region is produced in most of the

Anodic wire

Field wiret HV 2

- HV I

Fig. 30. Principle of construction of the adjustable field muldwire drift chambers. Cathode wires are connected to uniformlydecreasing potentials, starting from ground in front of the anode . Field wires reinforce the field in the transition region to the nextcell30"49 ).


424

cell. Small gap-to-anode wire spacing ratios cantherefore be implemented ; typical values of 6mmand 50 mm have been used for the gap and theanode spacing, respectively. Notice that since thecathode planes are not equipotential, some fieldlines escape from the structure ; subsidiarygrounded screening electrodes, as shown in the fi-gure, guarantee the immunity of the drift fieldfrom external perturbations .

Other structures have been developed, mainly inthe field of very large size detectors.

In a large neutrino detector at CERN, driftchambers of 4.5 mx4.5 m use uniform cathodeswith the addition of several field shaping wiress').In a large magnetic cylindrical detector, extremelysimple layers of drift cells are used, consisting ofthree field wires on each side of an anode wire . Asmall angle between the direction of the wires in


S (mm)

successive layers permits a two-dimensional read-out of a track'°).

5.4. POSITIONING ACCURACY ANDTIME MEASUREMENTS

The ultimate accuracy that can be obtained in adrift chamber depends both on the good know-ledge of the space-time relationship and on thediffusion properties of electrons in gases. From theprevious considerations it appears that, for mostgases commonly used in proportional counters, in-trinsic accuracies (due to diffusion) below 100,umare possible for minimum ionizing particles . Thespace-time correlation, however, may not beknown at this level of accuracy, especially for largechambers where various mechanical tolerances canlocally modify the electric field structure. Verygood results in terms of stability of operation and

Fig. 31 . Measured and computed space-time relationship for the chamber in fig. 30, as a function of the minimum ionizing beamangle of incidence30).

E

b

Fig. 32. Measured intrinsic accuracy in the drift chamber of fig.30,as a function of drift space. The experimental results havebeen decomposed into threecontributions : a constant electronicsdispersion, a physical diffusion term function of the square rootof the drift space, and a contribution of the primary ion pairstatistics61 -56).


reproducibility have been obtained combining thegood electric field characteristics of the structureshown in fig. 30 and the saturated drift velocitypeculiarity obtained in selected gas mixtures.

In these conditions, space-time relationshipslike the ones shown in fig. 31 are obtained ; fortracks perpendicular to the chamber, the correla-tion is strictly linear within the measurement er-rors .

In a structure where the electric field is less uni-form, the space-time relationship is obviouslymore or less deviating from linearity especially ifsome regions of field are low enough not to allowdrift velocity saturation .The intrinsic accuracy of a chamber can be es-

timated by the usual method of measuring thesame track in a set of equal chambers, and com-puting the standard deviation of the difference, ina given chamber, between the measured and thefitted coordinate. A typical experimental resultis presented in fig. 32 61 ), which gives theaccuracy as a function of the drift distance in achamber like the one in fig . 30 . The result can bedecomposed into three main contributions: asquare root dependence on the distance of driftdue to electron diffusion ; a constant electronicsspread estimated to correspond to about 40.um asfrom the figure ; and a contribution of the primaryelectrons' production statistics, particularly import-ant close to the anode wire56 ).As shown explicitly in the expression giving the

r.m.s . of electron diffusion (section 5.2), operationat high pressures improves the space accuracy, asshown in fig . 33 16).

2

425

Fig . 33 . Measured dependence of the accuracy of localization indriftchamber on thegaspressure. Increase. in presults both in adecrease in electron diffusion, and in a reduction in the primaryion pair statistics effect, because of the increase in the specificionization 16 ).

As far as the proportional gain is concerned,drift chambers are, in general, easier to operatethan MWPCs, being essentially isolated propor-tional counters ; all considerations on multiplica-tion factors and gas choice apply as well, with thenecessary modifications due to the geometry.Common practice has shown that commercialgrade purities are sufficiently good for moderatedrift lengths (a few cm), but that the gas tightnessof the chamber and of the tubing has to be care-fully checked . In some cases, a gas monitoring de-vice at the output of a chamber or a chamber sys-tem is advisable .

It should be emphasized that the intrinsic accu-racy given in fig . 32 is the result of a local mea-surement, normally realized in a short run andessentially independent of the detailed space-timecorrelation. To make use of the quoted accuraciesin an actual coordinate measurement, one has, ofcourse, to know precisely the space-time correla-tion or the drift velocity w = w(t). For a givenchamber geometry, the major factors that can in-fluence the drift velocity are : the electric fieldstrength and direction, the atmospheric pressure,the gas .composition and temperature, the presenceof external factors modifying the drift properties(electric or magnetic stray fields), and the me-chanical imperfections . Although it is, in principle,possible to take all these factors into account byproper calibration or monitoring, for a realistic sys-tem it is more reasonable to set definite limits tothe tolerable variations, as a function of the de-sired final accuracy. The choice of a drift-velocitysaturating gas obviously decreases or eliminates

11111 . TRACK VISUALIZATION DETECTORS

42 6

the dependence on the reduced electric field E/P.It appears also") that for several gases and gasmixtures the temperature dependence of w is re-duced at high fields ; for the particular mixtureused by the authors of ref. 30 the calculation givesa value very close to the experimentally measuredone, i.e. dw/w=3x 10 -4 per °C at a field around1.4 kV/cm. At low fields, and for the same mix-ture, the relative variation is about one order ofmagnitude larger. The dependence on gas compo-sition also reduces at saturation, and has beenmeasured in the quoted gas mixture to bedw/w=1 .2 x 10-' for 1% change in the gas.

If the position of a particle is known, then thespace-time correlation yields the time of creationof the ionizing trail. In the drift chambers with ad-justable field, the spatial accuracy o 100/tin corre-sponds to a timing accuracy of 2 ns . Now, the bestway to obtain the position is also to use driftchambers . If the particles come from a given di-rection, one single additional drift chamber is suf-ficient, provided the zero time is given by a scin-tillator. In this case, one does not really need theadditional drift chamber in order to know thetime. It has only the virtue of giving a secondpiece of time information that permits the rejec-tion of particles accidentally crossing the chamberswhile the electrons corresponding to the eventtagged by the scintillator are drifting. From thepoint of view of accidentals, two drift chambers in


Fig . 34 . Time resolution of staggered driftchambers . The pictureshows the time distributions a and 1 , of two chambers in anarrow beam ; the wire spacing is Scm The two chambers arestaggered by half a wire spacing in a magnetic field of 18 kG (thehorizontal scale is 100ns/div.). The peak c is the distribution ofthe sum of the drift times, TI+T2, obtained by simple hard-ware means(fwhm=2 ns). We seethat twosuch chambers definethe time of passage of the particle with a very good accuracy.

series thus correspond to a very sharp detectorwith 5 ns resolution . An example of such use wasgiven by Heintze and Walenta62 ) with proportionalchambers with wires spaced every 5 mm . They re-ported an accuracy of 6 ns (FWHM), which has tobe tempered by the fact that as the response is notGaussian (to be expected with an inhomogeneouschamber) the real resolution time for 100% effi-ciency is close to 25 ns.Fig. 34 shows the resolution obtained with two

staggered chambers, 2mx 2 m, with a wire spac-ing of 5 cm in a magnetic field of 18 kG 63). If thetwo chambers are staggered by half a wire spacing,one has the additional advantage of resolving theright-left ambiguity which plagues multiwire driftchambers.

If, however, two drift chambers are used tomeasure the position by measuring the differencein the detection time between two staggeredchambers, then the third one gives the time. Withthe measured time accuracies for each drift time,one may expect to define the time reference towithin about 4ns. It is also possible with threesuch chambers to select a given sagitta in a mag-netic field, thus permitting a rapid momentum se-lection. Again, with the measured average 100#maccuracy per chamber, with three chambers 10 cmapart, a 1 GeV/c momentum should be measuredto about 2% .

5.5 . THE THREE-DIMENSIONAL DRIFT CHAMBER

In some problems encountered in high-energyphysics a continuous sampling of particle trajecto-ries is necessary . Not only the coordinates but alsothe energy losses are required for the interpreta-tion of very complex trajectory configurations andparticle identification.

Recently, a new approach to the detection prob-lem has appeared, with the introduction of large-volume drift chambers such as the ISIS detec-tor"), the time projection chambers ;) and similardevices'"').These detectors consist essentially of a large gas

container, constituting the sensitive volume, inwhich theelectrons produced by the ionizing tracksdrift along a suitable electric field to an end-capMWPC; through avalanche multiplication thecharges are amplified there to allow electronic de-tection . As can be seen from fig. 35 the cathodeplane CI is built with a wire mesh or a grid hav-ing good electrical transparency for the drifting

C3

Drift volume

... ..... . ... .... . ... . .. . ... . ... . . C I

. .. . . . . . . . . . . . . . . A 1 MWPCc2

Fig. 35 . Volume drift chamberor time projection chamber. Elec-trons produced by ionizing track in a large gas volume drift toand end-cap multiwire proportional chamber where they aredetected . Storage and subsequent analysis of the amplitudedetected on anode and cathode wires as a function of timeallows a full three-dimensional reconstruction of complexevents.

electrons, while cathode C2 can be a conductingfoil or a printed circuit suitably stripped to providea coordinate measurement. The anode plane Acan either be constituted by a single mesh of thinwires, or by an alternation of thin (anode) andthick (field) wires. Suitable guard rings (not repres-ented in the figure) on the periphery of the driftvolume, maintain the uniformity of the applieddrift field ED over all the sensitive volume .

In the time projection chamber detector (fig. 36)the event topology sliced in narrow time intervals(hence the name of the detector) is rapidly storedin a large system of analogue shift registers thatare then multiplexed and read out at a conveni-ently lower speed. The order number of the con-

Endeap Win

TIME PROJECTION CHAMBER

awm pipe


rm

Endcop Wires

oll0Opp\0_3ilPOLAR,

iNapatln NIghVolt.9a EI.-d .

Ist d£/d. wir . . par Saeta12 Speeldl Wlrae per Seetor

Fig. 36. The time projection chamber. It has a cylindrical shape,so as to fit inside a large solenoid magnet .Two independent driftvolumes terminate with an end-cap multiwire proportionalchamber where bah time and energy loss information is

1 extracted. The structure operated at a high gas pressure (10atm)to improve energy resolution53),

42 7

sidered time slice, as referred to the event time,gives the elevation of each projection, while theanodic wire number and the induced pulse-heightdistribution on cathode strips or pads gives thetwo-dimensional coordinates of each segment inthe projection ; the anode pulse height is con-sidered to represent the energy deposit of thecorresponding track segment . The whole detector,cylindrical in shape, is immersed in a solenoidalmagnet such that the electric and magnetic fieldsare parallel : the drift direction of electrons istherefore not modified and, as mentioned in sec-tion 5.4, their transverse diffusion is strongly re-duced.

Although the development of such detectorsseems promising, several less ambitious appro-aches have been undertaken in several laboratoriesdealing with the same problem . In most cases so-lutions have been adopted where independent driftspaces with two-dimensional read-out based onmore conventional approaches like current divisionare used for the sampling of the trajectories44,66 ).We should also mention that serious doubts

have been cast on the possibility of using large vo-lume drift chambers in high flux condition, be-cause of the deleterious effects of positive ionspace charge on the drift properties61 ).

References1 ) S . C. Brown, Basic data of plasma physics (Wiley, New York,

1959).2) L. B. Loeb,Basic processes ofgaseous electronics (Universityof

California Press, Berkeley, 1%1).3) L. G. Christophorou, Atomic and molecular radiation physics

(Wiley, New York, 1971).4) S.A. Korlr, Electron and nuclear counters (Van Nostrand,New York, 1955).

s) W. Franzen and L. W. Cochran, in Nuclear instruments andtheir uses led . A. H. Snail ; Wiley, New York, 1956).

6) S.C. Curran and J . D. Craggs, Counting tubes, theory and

applications (Butterworths, London, 1949).7) P. Rice-Evans, Spark, streamer, proportional and drift cham-

bers (Richelieu, London).s) G. Charpak, R. Bouclier, T. Bressani, J. Favier and C.

Zupancic, Nucl . Instr. and Meth . 62 (1%8) 235.9) G. Charpak, Ann. Rev. Nucl. Sci. 20 (1970) 195.1t,) G. Charpak,D. Rahm andM. Steiner, Nucl . Instr. and Meth .

80 (1970) 13.11) G. Charpak, J. Phys. 30 (1%9) 86 C2 .12) F. Sauli, CERN 77-09 (1977).13) G. Charpak, Nucl. Instr. and Meth . 156 (1978) 1.14) G. A. Erskine, Nucl . Instr. and Meth . 105(1972) 565 .15) E. Mathieson, T. J. Harris, Nucl. Instr. and Meth . 154(1978)

189.16) W. Fart, J. Heintze, K. H. Hellenbrandt and A.H. Walenta,

Nucl . Instr. and Meth . 154 (1978) 175.

Vlll, TRACK VISUALIZATION DETECTORS

42 8

17) A . Breskin, Nucl. Instr. and Meth . 141 (1977) 505.

44 ) C . W . Fabjan, J. Lindsay, F. Piuz, F. Ranjard, E. Rosso, A.16) G. Charpak, G . Petersen, A. Policarpo and F . Sauli, IEEE

Rudge, S. Serednyakov, W . T. Willis, H. B. Jansen and J . D.Trans . Nucl . Sci . NS-25 (1978) 122.

Petersen, Nucl . Instr. and Meth . 156 (1978) 267.19) J. Fischer, M . Okuno and A . H. Walenta, Nucl. Instr . and

45) J . Alberi, J . Fisher, V. Radeka, L. C. Rogers and B. Schoen-Meth . 151 (1978) 451 .

born, BNL 19487 (1975).20) R. Bouclier, G. Charpak, Z . Dimcowski, G. Fischer, F. Sauli,

46) H . H. Chen and J. F. Lathrop, Nucl . Instr. and Meth . 150G. Coignet and G. FIOgge, Nuct. Instr. and Meth. 88 (1970)

(1978) 585.149.

47) J . Saudinos, in Proc . Topical Seminar on Interactions of21) G. Charpak, G. Petersen, A . Policarpo and F. Sauli, Nucl.

elementary particles with nuclei, Trieste, 1970 (INFN, Trieste);J. Saudinos, J . C. Duchazeaubeneix, C . Laspalles and R.Chaminade, Nucl . trient. and Meth . 111 (1973) 77.

48) A. H . Walenta, J. Heintze and B. Schürlein, Nucl . Instr . andMeth. 92 (1971) 373 .

49) G. Charpak, F. Sauli and W. Duinker, Nucl . Insu. and Meth.108 (1973) 413.

50) G. Charpak, C. Demierre, R Kahn, J. C . Santiard and F.Sauli, Nucl . Instr . and Meth. 141 (1977) 449 .

51) A . P. Jeavons, D.W. Townsend, N. L. Ford, K . Kull, A.Manuel, O. Fischer and M. Peter, CERN DD Internal Report77-15 (1977).

52) C . B . Lim, D. Chu, L. Kaufman, V. Perez-Mendez, R Hattn-er and D. C. Price, IEEE Trans. Nucl . Sci . NS-22 (1975)388.

53) Proposal for a PEP Facility based on the Time ProjectionChamber (PEP 4, Dec. 1976), Spokesman : D . Nygren .

54) V . Palladino and B. Sadoulet, Nucl . Instr. and Meth . 128(1975) 323.

55) G . Schultz and J. Gresser, Nucl. Instr. and Meth. 151 (1978)413 .

56) F. Sauli, Nucl. Instr. and Meth. 156 (1978) 147.57) G. Marel, P. Bloch, S. Brehin, B . Devaux, A .M. Diamant-

Berger, C. Leschevin, J. Maillard, Y. Malbequi, H . Martin, A .

Instr. and Meth. 1411 (IY/g) 4/1 .22) S. Brehin, A. Diamant Berger,G. Marel, G. Tarte, R. Turlay,G . Charpak and F . Sauli, Nucl. Instr . and Meth . 123 (1975)225.

23) C . Charpak and F. Sauli, Nucl. Instr . and Meth. 113 (1973)381.

24) G . Charpak, G . Melchart, G . Petersen and F. Sauli, High

G . CHARPAK AND F . SAULI

accuracy localization of minimum ionizing particles, to bepublished in Nucl . Instr. and Meth.

25) G . Charpak, H . G. Fischer, C. R . Grühn, A . Minten, F. Sauli,G. Plch and G. FIOgge, Nuct. Instr. and Meth. 99 (1972)279 .

26) L . Malter, Phys . Rev. 50 (1936) 48.27) D. Friedrich and F. Sauli, CERN EP Internal Report 77-10

(1977).28) G . Grunberg, L. Cohen and L. Mathieu, Nucl. Insu. and

Meth. 78 (1970) 102 .29) M . Breidenbach, F. Sauli and R . Titter, Nucl . Instr. and

Meth .

108 (1973) 23.30) A. Breskin, G . Charpak, F. Sauli, M. Atkinson and G.

Schultz, Nucl. Insu. and Meth . 124 (1975) 189.31) G. Charpak, G . Melchart, G . Petersen, F . Sauli, E. Bourdi-

naud, B . Blumenfeld, C. Duchazeaubeneix, A. Garin, S.Majewski and R . Walczak, CERN 78-05 (1978).

Patoux, J. Pelle, J . Plancoulaine, G . Tarte and R Turlay,32) H . Fischer, F . Piuz, W. G . Schwille, G. Sinapsius and O.

Nucl. lour. and Meth. 141 (1977) 43 .Ullaland, Proc . Inn . Meeting on Proportional and drift cham-

58) D . C. Cheng, W. A. Kozanecki, R. L. Piccioni, C. Rubbia,bers, Dubna (1975) p. 158.

R. L. Sulak, H . J . Weedon and J . Wittaker, Nucl. Instr. and33) J. Davier et al ., Laboratoire de l'Accélérateur linéaire, in

Meth . 117 (1974) 157 .preparation.

59) Experiment R209 at CERN (1i. Newman, private communi-30) R. Grove, K. Lee, V . Perez-Mendez and J . Sperinde, Nucl.

cation).Instr. and Meth . 89 (1970) 257.

60)W. Davies-White et al ., SLAC-PUB-2181 (1978).35) H . Okuno, R. L. Chase, J. Fischer and A. H . Walenta, IEEE

61) N. A. Filatova, ï. S. Nigmanov, V. P. Pugachevich, V. D.Trans . Nucl . Sci. NS-24 (1971) 213.

Riabtsov, M. D. Shafranov, E . N. Tsyganov, D. V . Uralsky,36) G. Charpak, R. Bouclier, T. Bressani, J . Favier and C.

A . S. Vodopiànov, F. Sauli and M. Atac, Nuel. lour. andZupancic, Nucl. Instr. and Meth . 65 (1968) 217.

Meth. 143 (1977) 17.32) D. M . Lee, J . E . Sobottka and H . A . Thiessen, Los Alamos

62) J. Heintze and A . H. Walenta, Nuel . Instr. and Meth . 111Report LA-4968-MS (1972) .

(1973)461 .3s) G. Charpak, Proc . Inn . Conf. on Instrumentation for high-

63) G. Charpak, ESRO SP-109 (1975), p. 109.energy physics, Frascati (1973).

64) W . W. Allison, C. B. Brooks, J. N. Bunch, J. H. Cté, J. L.39) A . Breskin, G. Charpak, B . Gabtoud, F . Sauli, N . Traumner,

Lloyd and R .W. Plenning, Nucl. Instr. and Meth. 119 (1974)W. Duinker and G . Schultz, Nuel. Instr. and Meth. 119

499.(1974) 9.

65) W . Baril et al., PETRA 76/16 (1976).40) M. Atac and W . E . Taylor, Nuel. Instr. and Meth . 120 (1974)

66)W . Farr, B. Granz, J. Heintze, R. D. Hever, P. Lennert, T.147 .

Nozski, A. Rieseberg and A. Wagner, Nucl. Insu. and Meth.41) L . Camilleri et al ., Nuel. Instr, and Meth . 156 (1978) 275.

156 (1978) 283.42) V. Radeka, IEEE Trans. Nucl . Sci. NS-21 (1974) 51 .

67) A. Astbury et al., CERN /SPSC/78-06 (1978).43) V. Radeka and P . Reliait, IEEE Trans. Nucl. Sci. NS-25

6s) D. Freidrich, G. Melchart, B. Sadoulet and F. Sauli, Nuel .(1978).

Instr. and Meth. 158 (1979) 81 .

cern, geneva. switzerland - kfki

Documents