design and characteristics of present-day photomultipliers bound... · design and characteristics...

1968, No. 8/9 267

Design and characteristics of present-day photomultipliersG. Piétri and J. Nussli

The phot omultiplier is a particularly useful instrument for the study of weak. or rapidlyoccurring effects in which light is produced, and if has proved indispensable for manyinvestigations in the field of nuclear physics. The progressive improvement of this deviceduring the last twenty years has been accompanied by the development of an astonishingdiversity of types to suit the various different applications; the Philips catalogue today listsabout a hundred types. The article below gives an account of the general considerationsunderlying the development of photomultipliers and also discusses the design of photo-multipliers which have an extremely rapid response or other unusualfeatures.

------------------ - --._- - -----_

A photomultiplier is a photoernissive cell associatedwith an electron multiplier (fig. J), forming an ampli-fier which, for most of the time, possesses virtuallynegligible inherent noise. Another characteristic fea-ture, at first sight rather paradoxical, is that no noisetempérature can be assigned to it, since its noise isindependent of température. Moreover, the two fun-damental effects on which it depends - photoernissionand secondary emission - are extremely fast. The timeresolution is limited only by the random variations ofthe initial velocities at which the emitted electrons leavethe electrodes.

These features, combined with its very high sensi-tivity, make the photomultiplier a particularly usefulinstrument for studying effects in which light is pro-duced. In certain conditions its sensitivity is in factimmeasurably superior to that of the photographic filmand exceeds that of the retina of the eye, consideringthe amount of information gathered in a time say ofthe order of one second.

The photomultiplier has proved to be indispensablefor the observation and counting of scintillations pro-duced by the passage of nuclear radiation through cer-tain materials (scintillators) [11. The study of thesescintillations makes it possible to determine: a) Theirfrequency, by counting the individual pulses; rates ofup to 107 counts per second are possible. The activityof the source can be deduced from this information.b) The energy of the incident particle, the number ofphotons produced being, within certain limits, propor-tional to this energy. c) The time at which the particleappears, with an accuracy approaching 10-10 second.

G. Piétri, Ing. £.S.£. Radio, L. ès Sciences (deputy director) andJ. Nussli, L. ès Sciences, are with Laboratoires d'Electronique etde Physique Appliquée (L.£.P.),Limeil·Brévannes [Val-de-Marne ) ;France.

Fig. I. Principle of a photomultiplier. The photocathode K,illuminated by the radiation to be detected, emits photoelectronswhich are multiplied by secondary emission from electrodes Slto SlO as they travel towards the anode An. The voltages indicatedat the electrodes are only intended to give an idea of the kind ofmagnitudes to be expected.

cl) The variation of the luminous flux with time duringthe scintillation, with an accuracy approaching 10-9second.

The design of present-day photomultipliers has beenmuch influenced by their widespread application innuclear techniques, but these applications have in turnbeen a great stimulus in the steady improvement in per-formance of these tubes. Photomultipliers are nowavailable for all applications where weak or very fastIuminous effects have to be observed.

The development of Philips photomultipliers wasstarted in 1952 at Laboratoires d'Electronique et dePhysique Appliquée CL.E.P.) near Paris. In 1954 large-scale production was started by Philips in the works at8rive-Ia-Gaillarde with the type ISOAVP, whileL.E. P., after their work on the 53 AYP, started on thedevelopment of fast photomultipliers, in particular the

[1] See for example: G. A. Morton, The scintillation counter,Adv. in Electronics 4,69-107, 1952. J. B. Birks, The theoryand practice of scintillation counting, Pergamon Press, Lon-don 1964. R. Champeix, H. Dormont and E. Morilleau, Aphotomultiplier tube for scintillation counting, Philips tech.Rev. 16, 250-257, 1954/55.

268 Pf-llLlPS TECHNiCAL REVIEW VOLUME 29

56 AVP, and special photocathodes. In about 1960 adevelopment laboratory was set up at Brive. One of thedevelopments from this laboratory has been that ofvery small photo multipliers with cup-shaped dynodes,such as the types 152 AYP and PM 400 [2]. The cur-rent Philips catalogue lists a very wide range of photo-multipliers, consisting of about a hundred differenttypes (fig. 2).

Before describing the work that led up to the pro-duction of these tubes, with special emphasis on thehigh-performance types, we should discuss the physicaleffects which are applied in the photomultiplier, thegeneral principles of the design and construction, andthe essential characteristics.

Fig. 2. Some of the types of photomultiplier nowproduced by Philips in the works at Brive, France.

Principles of the photomultiplier

The operation of a photornultiplier depends on twofundamental types of electronic emission: photoelectricemission and secondary emission.

Photoelectric emission by any material is the resultof three successive processes:1) Absorption of the incident photon by the transfer ofits energy hv to an electron. For any material apotential, known as the photoelectric work func-tion <Ppl), can be defined which is equal to the energydifference between the highest level from whichphotoelectrons can originate and the vacuum level.

If an electron thus excited is to be able to leave thematerial, Einstein's relation must be satisfied:

hv ~ e<Pph.

2) Diffusion of the electron towards the surface, withenergy loss due to collisions. The energy which theelectron retains when it escapes from the surface ishigher the weaker the absorption of the material forlow-energy electrons, and the nearer the electron isto the surface when it is excited.3) Transmission through the surface: wave mechanicsshows that the probability of this is not equal tounity immediately the energy of the electron be-comes greater than the work function, but tends

towards unity at an energy which has at least twicethat value.

Since the materials used nowadays are almost invariably semi-conductors, we should drawattention to an important differencebetween photoelectric emission from semiconductors and thatfrom metals (fig.3). In metals the conduction band is partlyfilled up to the Fermi level. This is therefore an occupied level,that is to say electrons, when excited by photons of correspondingenergy, may be emitted from this level to a vacuum. Since thislevel is also the highest level from which electrons may be emittedby the thermionic effect, the photoelectric work function <PpI> isidentical with the thermionic work function Wll1•

In a semiconductor the Fermi level is in the forbidden zone (in

1968, No. 8/9 PHOTOMULTIPLIERS 269

Is Vac sc Vac Met Vac

Q

Fig. 3. Model of energy bands in an insulator (Is), an intrinsicsemiconductor (SC) and a metal (Met), all with the same ther-mionic work function Wtll but with different photoelectric workfunctions Wph. The occupied energy zones are shown hatched.

the middle of it for an intrinsic semiconductor). This level againdefines the therm ion ie work function, since it is the mean levelof origin for thermionically emitted electrons (equilibrium existsbetween the population at the top of the valence band and at thebottom of the conduction band). In general, however, the occu-pation of the conduction band is much too small to give rise toany appreciable photoelectric effect; photoelectric emission there-fore originates only from the top of the valence band. This showsthat in semiconductors the photoelectric work function Wpll isalways higher than the thermionic work function Wtll [31.

A consequence of the processes (2) and (3) is thatappreciable photoelectric emission can only take placeif the energy of the photons substantially exceeds thethreshold energy. The photoelectrons then emitted willhave a surplus energy which appears in kinetic form.For example, in the case in which we are interestedtheir initial velocity, which will vary statistically inmagnitude and direction, will lie within a range corre-sponding to energies between about 0 and I eV.The conditions for emission mentioned above are

satisfied reasonably well in several semiconductors thatcan be produced in the form of thin films (of thicknessless than 0.1 [Lm) and with a large surface area by thesuccessive evaporation ofvarious layers under vacuum.However, these films - at any rate the ones which aresensitive to visible light - are all alkaline (usually com-pounds of caesium with antimony or with silver andoxygen which are unstable at temperatures higher than150°C and are rapidly destroyed by the least excessof oxygen).Another factor which has to be taken into account in

the preparation of the films is that they are almostinvariably semi-transparent: the light penetrates fromthe side opposite to the one from which the electronsare emitted. The film is deposited directlyon to theglass of the tube, or, to be more exact, on to a windowsituated at the end of a cylinder (see fig. I). With thisarrangement, the light excited in a scintillator coveringthe whole of the outside of the window is received over

the largest possible useful angle. The thickness of thefilm is however, very critical in this case, because thelight has to be absorbed, but not too far from the emis-sive surface. Incidentally, the arrangement describedpermits much easier location of the electrodes used forfocusing the electrons on to the input of the multiplierunit.An electron multiplier consists of a succession of a

variable number of targets called dynodes. When theirsurface is struck by electrons of appropriate energy(primary electrons) these dynodes emit secondary elec-trons. The geometry of the dynodes and any followingelectrodes, and of the applied electric (or perhapsmagnetic) fields, is such that the electrons emitted byeach dynode are accelerated towards the next one andstrike it with an energy between 100 and 500 eV. Thereare several well-known arrangements which satisfythese conditions and for all types now in productionwe have used the system proposed by Rajchman (seefig.1and, for example, fig. 14), which is a linearlyfocused multiplier of a purely. electrostatic type. Thissystemis fairly universally applied since it can generallybe expected to give the most rapid response.Secondary emission obeys laws which are even more

complex and less well-known than the laws of photo-electric emission. Metals are very poor secondary emit-ters. Good secondary emitters are either insulators(MgO, BeO and the alkali halides), or semiconductors(those which are also photoemissive), One advantage.ofinsulators is that they can bejprepared outside the tubebefore assembly, since they are relatively insensitive toair, but they have to be used in the form of extremelythin films if surface-charge effects are to be avoided.At present we are mainly using Ag-MgO-Cs, obtainedby heat treatment of an Ag-Mg alloy in an oxidizingatmosphere, followed by treatment in caesium vapourand oxidation. The resultant material has a stablesecondary-emission coefficient of between 3 and 8 forprimary electrons with an energy of between 100 and300 eV (fig. 4). The value of this coefficient is practi-callyindependent of the primary current up to values ofthe order of 100 [LA/cm2 or more, thus meeting therequirements for linearity.These dynodes have a fairly small useful surface area

for their volume. This makes it necessary to devise amore or less complex focusing electrode system (the"input optical system") between the photocathode andthe first dynode. It will be noted, moreover, that the[21 E. Morilleau, Production industrielIe des photomultiplica-

teurs (technologie et mesures de contröle), Acta electronica 5,39-51, 1961.

[31 An extensive treatment of the photoemissive effect has recent-ly been given in this journal: J. van Laar and J. J. Scheer,Photoemission of semiconductors, Philips tech. Rev. 29,54-66, 1968 (No. 2). The energies eWpll and eWth are referredto in this .article as Ed and cp (= E« - 15)respectively.

270 PHILIPS TECHNICAL REVIEW VOLUME 29

s1

Fig. 4. Variation of the coefficient of secondary emission <5 as afunction of the energy of the primary electrons, for variousmaterials.

form and arrangement of the first two or three dynodesare somewhat different. This region, referred to as thecoupling region between the input optics and the lastpart of the multiplier, which consists of a succession ofidentical stages, is rather criticalon account of themarked curvature of the paths between the first and thesecond dynode.The secondary electrons also leave the material at

random initial velocities, corresponding to energiesabout ten times greater than those of the photoelec-trons. With velocities ofthis order it is difficult to focusthe secondary electrons, and in fact some slight lossalways has to be accepted (of the order of 10%).The electrons emitted by the last dynode are collected

by the anode. This usually consists of a grid situated ata short distance from the last dynode and is traversedby the electrons travelling to that dynode.

Finally, a word or two about certain special problemsencountered in the manufacture of this type of electrontube. It follows from what has been said that the photo-cathode cannot withstand high temperatures or expo-sure to air. It therefore has to be prepared inside thetube itself, on the vacuum pump, before sealing-off. The inside of the tube must therefore contain every-thing that is needed for this preparation - antimonyto be evaporated, as well as a mixture of alkaline saltsand a reducing metal for the preparation of the alkali"in situ". All these materials and the dynodes cannotbe degassed at the high temperatures normally used indegassing the electrodes of ordinary electronic tubes.The only possible course of action is to pump the tubefor a very long time (12 hours or more) at less than400°C. The forming of the cathode then begins by thecold evaporation of the antimony layer (with a thick-ness ofthe order of 10 nm). The caesium for the SbCs3is next prepared in excess by reduction of caesiumchromate; since the reaction with antimony is sufficient-

ly fast only above 100 "C, the tube is raised to a tem-perature of about 120°C. The vapour pressure of thecaesium is then extremely high and a monolayer settleson all the surfaces. This monolayer cannot be removed,since further pumping is prohibited by the presence ofthe cathode. (With no further pumping it is also im-possible to eliminate all the gases liberated during thepreparation of the caesium, so that after the tube issealed off the pressure is not likely to .be much lowerthan 10-6 torr, in spite of the presence of a getter.) Thecaesium deposited on the dynodes considerably reducestheir work function. As we shall see later, this is ofparamount importance for the dark current. Thecaesium which is not absorbed is evacuated in thepump trap.One final remark willillustrate the variety of pro blems

encountered. The glass used for the envelope has tomeet the following specificrequirements: a) Itmust notdevitrify nor go black during the prolonged and delicateoperation of sealing the end window. b) It must beresistant to attack by caesium at the temperature atwhich the cathode is formed. c) It must contain aminimum of potassium, because of the constant pres-ence of the radioactive isotope 4oK. These require-ments are to all intents and purposes only met by theborosilicates (the third condition is never entirely satis-fied), but in this case the metal seals have to be eithertungsten or an alloy of the Fe-Ni-Co type. Thesematerials are difficult to use and necessitate stringentand repeated quality control, in particular to ensure avacuum-tight seal.

Characteristics of a photomultiplier

We shall now recall some of the chief quantities thatcharacterize a photomultiplier.

Photocathode sensitivity

The sensitivity (jk of the photocathode is a functionof the wavelength A of the incident light. It is expressedeither in amperes per watt of incident radiation or bymeans of the quantum efficiency 'YJq, a quantity ex-pressing the relation between the number of electronsemitted and the number of photons incident on thewindow ofthe tube ('YJq < 1). The quantities 'YJq and (jk

are related by the expression:

. . . . . (1)

where h is Planck's constant, c the velocity of light invacuum, and e the electron charge.For greater convenience another sensitivity has been

defined which broadly characterizes the photoelectricproperties of the cathode, namely the sensitivity towhite light emitted bya tungsten filament lamp at a


colour temperature of 2850 "K. It is usually expressedin fJ.A/lumen.The highest quantum efficiencies, obtained for wave-

lengths of the order of 0.4 fJ.m,are about 30%, corre-sponding to about lOOmA/watt. The colour response ofseveral different types of photocathode is shown by thespectral characteristics in jig.5. Each cathode ischaracterized by a particular position of the photo-electric threshold. The Sll cathode is well adapted tothe blue light emitted by scintillators. Since it is theleast troublesome to prepare, this cathode is the onemost widely used. A typical value for its overall sensi-tivity is 60 fJ.A/lumen.

100mA/W i-

f- ~ ~i- ~ "'IC\1-/ 1\\1\1\ \\\ l..-- Te.ci

~

SbCs3

V (S11)

I-

(\Sb.K-Cs......,

1)ySb(No2K)-Cs

I- (S20)

t- Ag·Q·CsI

1\ '<~S1J1

1\ V \ \

f\'-../

\ \1\1 1

50

10

5

a5

O. a3 a4 0.5 0.6 0.7 oe 0.9_À

1.1fLm

Fig. 5. Spectral sensitivity ak of photocathodes in current use.

It is important to note that, like all semiconductorcharacteristics, this sensitivity varies considerably fromone tube to another. The sensitivity of the SU maypermissibly deviate by a factor greater than 2 (i.e. from40 to 90 fJ.A/lm).Again, the sensitivity is never perfectlyconstant at all points of the surface of a photocathode.A useful measure of the homogeneity of the sensitivityis given by the ratio amln/amax [4] (in terms of whiteor monochromatic light; in the latter case amln/amax

[4] G. Piétri and C. Arvin, J. Phys, Radium 19; Suppl., 154A-159A, 1958.

[5] G. Laustriat and A. Coche, J. Phys. Radium 19, 927-929,1958, and 20, 719-720, 1959. A. T. Young, Appl. Optics 2,51-60, 1963.

[6] I. CantarelI, IEEE Trans. on nuclear science NS-ll, No. 3,152-159, 1964.

may be a function of the wavelength). As a rule, how-ever, the characteristic usually quoted is the meanoverall sensitivity.These different quantities are to a slight extent de-

pendent on temperature [5].

Gain; collection efficiency

The gain is defined as the ratio of the anode currentto the photocathode current. It is given by:

. . (2)

where 1}e is the collection efficiency of the input opticalsystem and ~ is the mean secondary-emission coefficientofthe n dynodes. The gain ofthe multiplier proper mayalso be used; this is ~n.

The collection efficiency is in general a function ofthe co-ordinates of the point of the cathode illurni-nated, and also of the wavelength, since, as we shallshow later, this efficiencyis affected by the initial veloc-ities of the photoelectrons.We mayalso define an anode sensitivity aa, equal to

the product of G and the sensitivity alc (in white orin monochromatic light).The coefficient of secondary emission varies almost

linearly with the energy of the primary electrons in therange 100-200 eV, which is the one ordinarily used. Thegain as a function of the tube supply voltage V thusvaries approximately as vn (fig. 6), which means thata very well stabilized power supply must be used. Thehighest gains achieved are of the order of 109.The gain varies slightly with temperature [5] (a few

tenths of a per cent per degree), but it is particularlysubject to variations in the mean output current [6].

109

1/1

///

1/

7 ///I

G

t

10

1.4 1.8 2.2kV

Fig. 6. Variation of the gain G of a multiplier with 14 dynodesof Ag-MgO-Cs as a function of the high voltage.

272 PHILTPS TECHNICAL REVIEW VOLUME 29

Depending on the tube, reversible variations (positiveor negative) in gain of several per cent are observedwhen the anode current varies from 0 to 100 fLA.Cur-rents higher than 1 mA usually cause an irreversibleand progressive decrease of gain.

Dark current

The dark current is the current 10 observed at theanode when no light falls on it. This current is mostfrequently initiated by electrons emitted from the cath-ode, either through thermionic emission or by excitationdue to luminescent effects inside the tube. Instead of 10,it is therefore more convenient to consider the ratio lolG,which varies less than 10 with the high voltage.

Pulse-amp !itude variations

If we consider the photomultiplier in associationwith the electronic circuits designed to process theinformation supplied by the photomultiplier, we mayassign to it a resolving time r. Let N be the number ofphotoelectrons emitted by the cathode during this time.This number will fluctuate from one measurement toanother; the standard deviation of the probabilitydistribution curve will be JIN and the width of thecurve at half-height will be 2.36VN (if the distributionis Gaussian). At the first dynode, with secondary-emis-sion coefficient (h, the number of secondary electronswill also fluctuate, but the relative fluctuation will bereduced by a factor of jI(5î.This effect can appear in various ways:

a) If we investigate the amplitude V of the pulses, duesay to scintillations, and ifwe consider as a functionof V the probability that the amplitude will be be-tween Vand V + d V, we may define a resolution Rwhich is the ratio LIVIVpeak, LIV being the width athalf-height ofthe probability distribution curve, andVpeak the most probable amplitude. To a firstapproximation, R will be of the order of 2.36IVN,if N is the mean number of photoelectrons per sein-tillation ..

b) If we investigate the amplitude of the pulses pro-duced by the photons arriving at a fairly low rate,so that all the pulses are well separated in time, weobtain a distribution curve of the amplitudes, whichis called the single-electron spectrum (SES). Its formis determined by the fluctuations of the secondaryemission of the first dynodes and its width will be ofthe order of JffI. This spectrum is characteristic ofthe photomultiplier itself.

The "mean gain" on of the multiplier system (cf.eq. 2) can be calculated from the mean amplitudeof the single-electron spectrum, i.e. the amplitudeat its "centre of gravity".

c) If we investigate the amplitude of the pulses in the

total absence of light, we obtain the dark-pulsespec/rum (DPS). This spectrum would be identicalwith the SES if all the pulses came from the cathode.This is never the case and the DPS is therefore veryuseful for studying the dark current.

d) If we observe a luminous effect modulated in ampli-tude, we have the well-known Schottky relation:

LlIk = V2e Ik Llf at the cathode. . . (3a)

and

Lila = G V2e Ik Llfp at the anode,. . (3b)

where Af is the bandwidth of the system and p is afactor slightly greater than unity, related to thefluctuations in the secondary emission. This is theSchottky (shot) noise which is associated with thesignal and increases with the square root of thesignal. It is independent of frequency and is there-fore referred to as white noise.

Transit time; speed of response

To establish the speed of response of the photomulti-plier it is useful to define the following quantities:a) Pulse response.

Ifthe cathode is exposed to an infinitelyshort pulseof light, the shape of the resultant electrical pulsecharacterizes the speed of the photomultiplier. Ingeneral this shape varies considerably from one typeof tube to another and is very rarely Gaussian. Itcan include peaks later than the main one ("after-pulses"). The speed of response may be expressedin terms of the width of the pulse at half peak heightand in terms of its rise time.

b) Mean transit time.Again in the case of a very short pulse of light

a certain time elapses between the moment at whichthe cathode is illuminated and the arrival of theresultant avalanche at the anode. The moment ofarrival of this pulse can for example be taken asthe moment of arrival of its "centre of gravity" (orofthe point half-way up the rising edge ofthe pulse).The time elapsed will fluctuate from one measure-ment to another and its mean value is defined as themean transit time. This is in general less than5X 10-8 s, and varies with the voltage V as 1IVv.

c) Transit time fluctuations.Since the instant at which the anode pulse appears

(definedas indicatedabove) fluctuates from one meas-urement to another, it is more aptly described by aprobability distribution. If each pulse is producedby a single photoelectron, the width at half-heightof this distribution gives a much better measure ofthe accuracy of the photomultiplier for measuringvery short times - in particular the time of flight


of particles -- than is given by the "speed of re-sponse".If there are N photoelectrons in each pulse, this

fluctuation will decrease as N increases.d) Frequency response.

Experience shows that when a photomultiplierdetects a light signal modulated in amplitude at afrequency f, the output signal is a decreasing func-tion off A 3 dB (or 6 dB) bandwidth can be defined,corresponding to a decrease of the output signal bya factor of V2(or 2). The lower limit of the passbandis of course f = 0, since the photomultiplier canamplify a direct current.

Linearity

With a constant light flux W the anode current Ja isat first proportional to W, i.e. la = kW, up to thevalue la = 100 (.LA. With short pulses, however, la maysafely exceed lOO(.LA by a considerable margin. Froma certain value onwards, the current la does not increaseas fast as the flux W. The limit of linearity is defined as thevalue of la at which the la = j(W) curve, expressed indB, falls 1 dB below the line kW.

Improvement of the characteristics of the photomultiplier

We shall now review some of the work done in ourlaboratories during recent years with the object ofgaining a better understanding of the physics of theoperation of the photomultiplier, and of improvingtheir characteristics or adapting them to special appli-cations.

Photocathode

We have already noted (and we shall discuss thesubject in more detail later) that the fluctuation effectsare closely related to the number of photoelectronsdelivered by the cathode. Any improvement here there-fore comes from an improvement of the sensitivity ofthe cathode. For example, the sensitivity of the Sllcathode (SbCs3), which has been known for morethan 30 years, has been doubled in the last ten years,and now approaches 100 (.LA/lumen[71. This advancehas been made possible through the constant improve-ment of the vacuum obtained with modern diffusionpumps, the progress made in the methods of cleaningthe windows and materials used in the tubes, and thegreater purity of the products used in the preparationof the cathodes. The deposition of a substrate of Mn02on the window has also been found to have a favourableeffect.The SI cathode (Ag-O-Cs) has also been known for

some considerable time. It is the only one which hasappreciable sensitivity in the infra-red (l.2 (.Lm), but itsquantum efficiency remains poor (0.1 to 1%) in the

whole of the useful wavelength range. The S20cathode(Sb(Na2K)-Cs) is at present the cathode with thegreatest efficiency between 0.5 and 0.8 (.Lm; in whitelight its sensitivity is as high as 250 (.LA/lm. But main-taining its four constituents in the correct ratio over thewhole surface of the cathode presents such problemsthat the tubes which have this cathode are three timesas expensive as equivalent tubes with the Sll cathode.These problems are chiefly related to the presence ofthe focusing electrodes in the input optical system,which obstruct the deposition of the alkalis on thecathode.A new cathode is now being developed which looks

as if it will supersede the SII in photomultipliers usedfor scintillation counters, at least in certain cases. Thisis the Sb-K-Cs cathode, whose quantum efficiencymaybe better than 25% at 0.4 (.Lm.The extension of the sensitivity range of photomulti-

pliers towards the ultra-violet has been and still is thesubject of a great deal of investigation. The sensitivityrange is in fact limited by the transparency of thewindow (fig. 7). Borosilicate glasses are transparent upto about 0.3 (.Lm, provided they are thin enough.

Fig. 7. Optical transmission D ofvarious windows used in photo-multipliers, as a function of wavelength.

Beyond this limit it is necessary to use quartz (downto 0.16 (.Lm), sapphire (0.15 urn), CaF 2 (0.12 (.Lm) orLiF (0.105 (.Lm). Generally speaking it is impossible,both for economic and purely technical reasons, tomake complete tubes with these materials; they areused for the window only. Since these materials havewidely different coefficients of thermal expansion fromthose of the types of glass used, it has been necessary todevelop complex sealing techniques based in each caseon the use of a number of materials with an inter-mediate coefficient of expansion.All cathodes sensitive in, the visible range have good

quantum efficienciesin the whole of the ultra-violet [81

and may therefore be used in photomultipliers. There

[7) V. A. Stanley, IEEE Trans. NS-13, No. 3, 63-71, 1966.[8) L. Dunkelman, Ultraviolet photodetectors, J. quant. Spectr.

rad. Transfer 2, 533-544, 1962.

274 PHILlPS TECHNICAL REVIEW VOLUME 29

is however a new cathode which is of particular interestfor certain applications because of its selectivity. Thisis the Te-Cs cathode, which has a quantum efficiencyof up to 10% at 0.25 (Lmand has a sensitivity which isvirtually zero beyond 0.32 (Lm.This matches exactlythe wavelength range of the solar radiation which istransmitted through the atmosphere. This "solar blind"photocathode is therefore particularly useful for photo-metric measurements of rays which are not transmittedby the Earth's atmosphere.Finally, no window transparent to wavelengths less

than 0.105 (Lmexists (incidentally, air itself is opaqueto wavelengths ofO.18 (Lmand below). For these wave-lengths we have developed windowless photomultipliersin which the cathode (which is no longer semi-trans-parent) consists simply of a metal, e.g. nickel, or aninsulator (BeO, KCl, etc.) [9l. No material obstaclemust be interposed between this cathode and the radia-tion source.

Dark current

The dark current obviously sets a lower limit to thelevel of light signal that can be detected, and it is there-fore a significant factor in the quality of the tube. Togive some idea of magnitudes, the luminous flux equi-valent to the dark current in a photomultiplier with anormal Sll cathode is of the order of 10-11 lumen.Investigation of the amplitude distribution of the

pulses which form the dark current can give consider-able information about the origins of this current,particularly if multichannel pulse-amplitude analysers(kicksorters) are used uoi [11l. The amplitude of apulse is in fact a function of the point of emission ofthe electron, or electrons, which initiated the avalanche.This point may be on the cathode, but it mayalso beon the focusing electrodes or on the dynodes. Theinterpretation is complicated, however, by the fluctua-tions in the pulse amplitude.The most obvious source of dark current is ther-

mionic emission from the cathode, on account of thelow thermionic work function q)th of the layers. (Weshould recall here that q)th is lower than q)Ph in thematerials which are of interest; see page 269.) Ther-mionic emission obeys the law:

The variation of the dark current with temperatureshould therefore make it possible to minimize this cur-rent by simply reducing the temperature. However, ifwe measure the number of stray electrons N emittedper unit time and draw. curves of log N = f(IIT)(jig. 8), we see that this is not true: although a slopecorresponding to an energy nearly equal to q)th is in

70 60 50 40 30 20 100°C

102L_~::------+-,:----+-------:;~a8 a9 1 293- TOK

Fig. 8. Variation of the dark current (number of electrons persecond) as a function of temperature T for various photomulti-pliers. The slope at low temperatures corresponds to a muchlower work function (shown in eV) than that at high tempera-tures.

fact found at a fairly high temperature, a much smallerslope is always found at normal or low temperatures.Under normal conditions, therefore, the thermionicemission calculated by extrapolation of the steeplysloping section of the curve is always much smallerthan the total current: it only represents a small partof it. It is also found that the dark-pulse spectrum(jig.9) changes its shape with temperature: the rela-tive number of pulses with amplitudes correspondingto electrons emitted from the cathode decreases withdecreasing temperature. If the temperature is reducedsufficiently for the thermionic emission to be negligibleand the cathode is then disconnected, thus suppressingits emission, it is sometimes possible to observe a furtherdecrease in the number of pulses in the correspondingregion of amplitudes. Itmay therefore be assumed thatthese pulses were due to photoelectrons initiated byluminescences inside the tube. The number of thesepulses is in fact frequently observed to increase rapidlywith the high voltage.

To limit this contribution to the dark current, wefirst concentrated on obtaining the required gain whileusing the lowest possible high voltage - which amountsto improving the secondary emission - and on deter-mining the optimum number of dynodes. This proved


to be 14 for a gain of lOS; the interdynode voltages arethen of the order of 130 to 140 V.This choice, however, impairs the speed of response,

as will be shown later. It was therefore necessary tofind other preventive measures and to identify the causeof the luminescences. We shall mention just a fewresults of these investigations.It is' difficult to produce electrodes and connecting

leads which have no surface irregularities. A deposit ofcaesium at such irregularities lowers the work functionand thus encourages cold emission. Moreover, thematerial Ag-MgO-Cs, which is a finely grained in-sulator, is in itself an important source of cold emis-sion. The electrons which are emitted are acceleratedby the electric fields throughout the tube and collideeither with insulators (mica or ceramic spacers, theglass wall, and the inside or outside surfaces of thedynodes coated with MgO) where they give rise toscintillations, or with metal surfaces, where they pro-duce soft X-rays, which in turn can excite considerablephotoelectric emission. All of these radiations exciteeither the cathode or the dynodes, particularly the firstdynodes. It should be noted that these effects have afairly high "efficiency". The ratio ofphotoelectric emis-sion to cold emission could be as high as 0.1 if all thephotons were able to reach the cathode directly.We shall now indicate some ofthe preventive meas-

5.-.-----,-------,-------,-------,

21-H,----

Fig. 9. Pulse-amplitude distribution of the dark current in aphotomultiplier, for various temperatures. The probability den-sity dN/d V (in arbitrary units) is shown as a function of pulseamplitude V. (The pulse amplitudes have been normalized to themean amplitude ofthe pulses for single photoelectrons, as deter-mined by the measurement of the single-electron spectrum, seep.272.) .

ures used. To prevent the electrons from reaching thewall ofthe tube, it is sufficient in principle to bring thiswall to the mostnegative potential in the tube, that ofthecathode. Certain tubes are therefore given a conducting'outer coating (colloidal carbon) held at cathode poten-tial. (Incidentally, it has also been found that the lowestand most stable dark currents are often obtained whenthe tube is operated with the cathode earthed and theanode at positive potential.) The luminescences mustbe counteracted in other ways. They often occur in thebase of the tube, since the leads carrying all the poten-tials are close to one another here, and also at thedynodes, between which, at certain points, very highfields exist. To prevent these photons from reaching thecathode, a system of shields is used - in fact the shield-ing system is formed by the electrodes of the inputoptics - and black glass is used to insulate the dynodesand the various electrodes [lOl. All metal parts arecarefully polished and their shapes carefully chosen sothat sharp edges (e.g. at the dynodes) giving rise to highelectric fields do not occur. Various precautions aretaken to reduce the amount of alkali deposited on thesurfaces during the activation of the cathode. In coop-eration with the glass manufacturers we are carrying outinvestigations to see if the percentage of potassiumcontained in the types of glass we use can be reduced.Thermionic emission itself can be reduced by the intro-duetion of the Sb-K-Cs cathode which has a higherwork function. The result of all these measures is thatin some of the tubes the dark current is less than 1000electrons per second (at the cathode).A unique solution consists in the use of the disc-seal

technique. This has been adopted, for example, in theXP 1210, which will be discussed presently (fig. 10).In this arrangement each dynode is connected to ametal ring, sealed in the envelope, by means of a discwhich occupies practically the whole diameter of thetube. The tube is thus subdivided by a large numberof screens, and moreover the potentials are distributeduniformly over the length of the tube. With this struc-ture interdynode potentials of the order of 300 V maybe applied without greatly increasing the dark current.This is a fairly expensive method, however, and is onlyconsidered for special cases (e.g. where extremely fastresponse is required).

[9] M. Brunet, C. Jehanno, C. JulIiot and A. Tarrius, Photo-multiplicateurs sans fenêtre--étude, réalisation, Onde électr.42,746-753, 1962.

[10] G. Piétri, Present state of research and new developments atthe Laboratoires d'Electronique et de Physique Appliquée(L.E.P.) in the field of photomultiplier tubes, IEEE Trans.NS-ll, No. 3, 76-92, 1964.

[11] M. Duquesne, I.Tatischeff and N. Recouvreur, Nucl. Instr.Meth. 41, 13-29, 1966. E. Morilleau and R. Petit (Brive),Photomultiplicateur rapide à faible bruit de fond, Ondeélectr. 46, 111-119, 1966.


To conclude this section we shall consider the partplayed by the residual pressure. By themselves theatoms of the residual gas cannot liberate a single elec-tron. Below 10-4 torr the mean free path of the elec-trons is so long that these atoms have hardly any effecton the multiplication.lfthe gain is fairly high, however,the electrons, which are very numerous in the finalstages, produce ions, and these ions may travel to thecathode and prod uce secondary electrons there, or they

up to several minutes after the extinction of an intenseill umination.

At the values of gain normally used (108), this kindof behaviour is fortunately rare in the photornultipliersnow in production.

Amplitude fluctuations

Even without extensive theoretical analysis [12] itwill be clear that the fluctuations in amplitude originate

Fig. Ia. a) TubeXP 1210(laboratory designationPM 638 B) with a disc-seal construction for thedynodes. Total length16 cm.b) Schematic sectionthrough the tube.

may recornbine, with the errussion of light. There isthus an interaction between the output and the inputwhich could introduce instability. In reality, theseeffects are not instantaneous and thus appear in theform of after-pulses, delayed by a time almost equal tothe transit time in the photomultiplier if the interactionis due to optical coupling, or by a much longer time(several microseconds) if the interaction occurs via ionstravelling to the cathode. In the extreme cases theseafter-pulses may be regarded as dark-current pulsessince they usually appear when there is no longer anyillumination. Such an interpretation is even more tempt-ing in the case of atoms excited at metastable levels:although these atoms are much fewer in number thanthe ions, they are very much more mobile since they arenot attracted by the electrodes of the tube, and are thusable to cause secondary emission from the cathode

a

b

from the stages where the number of electrons con-veying information is lowest; th us, the photocathode Kis the most important source, followed by the first,second and third dynodes, S1,2,3, as progressively lessimportant sources. The contributions from the follow-ing stages are negligible. Every attempt is thereforemade to develop cathodes which have the highest pos-sible sensitivity, and to achieve high coefficients ofsecondary emission, particularly for the first dynode.Normally the second condition is fulfilled by making

[12] E. Breitenberger, Scintillation spectrometer statistics, Progr.nucl. Phys. 4, 56-94, 1955. G. Piétri, Les photomultiplica-teurs, instruments de physique expérirnentale, Acta electroni-ca 5, 7-30, 1961.

[13J M. Brauit and C Gazier, CR. Acad. Sci. Paris 256, 1241-1244,1963. R. F. Tusting, Q. A. Kerns and H. K. Knudsen,IRE Trans. NS-9, No. 3, 118-123, 1962. C. Gazier, Contri-bution à l'étude des propriétés statistiques du gain desphotomultiplicateurs, Thèse d'ingénieur du CNAM, 1963.

1968, No. 8/9 PHOTO MULTIPLIERS 277

Fig. 12. a) Equipotential surfaces of the input optical system in the 56 AVP. Some paths areshown for electrons coming from the cathode, some without initial velocity and others withan initial transverse velocity corresponding to an energy of 0.3 eV. The diagram gives anindication of the area of the first dynode within which photoelectrons can arrive. (The potentialvalues quoted are relative.) .

b) Equipotential surfaces of the region of the first dynodes. The vertical lines are possiblepaths for electrons originating 'from the cathode. These connect with the paths of secondaryelectrons (zero initial velocities) which should all reach the second dynode. All the pathswill almost inevitably have different lengths since the electron beams never arrive normal tothe target surfaces. .

the potential between K and SI two or three timesgreater than that between the other stages, thus assur-ing a value of 0 which may be greater than 8. A highcollection efficiency is also highly desirable [131, butthe uniformity of the collection efficiency between Kand SI and between SI and S2 is equally important.This may be illustrated with an example.Suppose that the cathode is divided into two equal

parts whose sensitivities are in the ratio of 2 : 1 (whichis effectively the same as a non-uniformity of 2: 1 inthe collection efficiency between cathode and S1). Sup-pose further that scintillations appear near to the cath-ode (which therefore each only illuminate a small area)and produce on average n = 30 electrons in the mostsensitive zone and n = 15 in the other, with a proba-bility of -}of appearing in each zone. The probabilitydistribution obtained pen) will be the sum of two distri-butions centred on 30 and on 15with correspondingwidths at half-height of 2.36V30 = 13and 2.36]1T5 = 9.If the cathode were uniform and its sensitivity was equalto the average of that of these two zones, there wouldonly be one distribution with the width 2.3611225 = 11

Pen)

i

o

Fig. 11. Illustrating the effect of non-uniformity in the sensitivityof the photocathode on the pulse-height resolution of a photo-multiplier.

(fig. '11).We have exactly the same situation with thefirst dynode ifthe collection Sr-+S2 is not uniform overthe whole area of SI capable of receiving the electronsemitted from the cathode. This area depends on thefocusing quality of the input optics and on the initialtransverse velocities of the photoelectrons, and it hasa typical diameter Of a few millimetres in tubes ofconventional dimensions (fig. 12a, b). Defects in the

<;-...-... -...__

278 PHILlPS TECHNICAL REVIEW

uniformity of collection will affect more especially thesingle-electron spectrum, whose form is in fact deter-mined by the secondary-emission fluctuations [131.Fig. 13 shows the effect on the SES of deliberate de-focusing by altering the potentials of the input opticsin the 56 AVP tube. The effect of the wavelength ofthe light (this comes into it since it affects the initialvelocities) is represented by an analogous family ofcurves.

S'll.1. .I. ,I, I. I, ,I.

2.12 Va 1.73111 0.83111 111 Va3Va

VK'F,=

\..-----0.38111o

+0.20 Va+0.80 111

dNdV

1

I Va=100 V I

_Ampl.

Fig. 13. Effect on the single-electron spectrum of deliberate de-focusing by altering the potentials of the input optical system inthe 56 AVP. The diagram at the top shows the distribution ofthepotentials on the different electrodes. Beginning with 84, thevoltage between successive dynodes is equal to Voo The curves 1to 4 were obtained for four different values ofthe voltage betweenphotocathode K and focusing electrode FI.

The improvement of the collection efficiency, bothin value and in uniformity, is a problem in electronballistics which can be solved by progressive modifica-tion of the configuration of the electrodes in an electro-lytic tank. It is advisable to avoid the use of accessoryfocusing electrodes where possible, at least if the userhas to adjust their potential himself. Nevertheless, inhigh-performance tubes (e.g. types 56 AVP andXP 1020) voltage adjustments have been accepted to

VOLUME 29

compensate for the effect of the relative inaccuraciesin the geometry and assembly of the electrodes. How-ever, it is preferable to use arrangements in which theauxiliary electrodes can be connected internally to oneof the dynodes. For example, the 56 AyP has a triodeoptical system (cathode, focusing electrode and accel-erator) in which only the focusing electrode has avariable potential, the accelerating electrode beingconnected to SI. The XP 1020 has an extra electrodeconnected to S9. Fig. 14 gives a comparison of thegeometries of these two types, and fig. 15 shows thevariation of their collection efficiencies as a functionof wavelength. The collection efficiency of the inputoptics is defined here as the ratio of the number ofpulses with a height greater than a certain thresholdvalue to the number of photoelectrons emitted duringthe same time. A description of the method of meas-uring this quantity would be beyond the scope of thisarticle [10].

The shape of the single-electron spectrum and thecollection efficiency are of great importance for thedetection of single photoelectrons: since it is alwaysnecessary to count only the pulses above a certainthreshold, it is important that this spectrum shouldcontain only relatively few pulses of small amplitude.

K

56AVP

K

Acc

XP 1020

Fig. 14. Section through the 56 AVP and the XP 1020. K photo-cathode. F(1.2) focusing electrodes. Acc accelerating electrode.Deft deflector. 8(1 ... n) dynodes. G screen grid. All anode. Cocoaxial feed. N connection to standard coaxial cable. P pins.


'Tl

t90r---~---+----r----r---7~--+-/~~/

~ ~ ~ ~ ~ ~ ~ ~~_À

Fig. 15. Variation of collection efficiency 17 of the 56 AVP (twosamples) and the XP 1020 as a function of wavelength.

On the other hand, if the number of photoelectronsin the resolving time is greater than 1, then the fluctua-tion of the number of electrons predominates.Schottky's equation then applies for the detection ofmodulated light, the fluctuations of the secondaryemission being represented by the correcting factor pin equation (3b).

If we neglect non-uniformity in the secondary-emis-sion coefficients (h and ö of the first and the followingdynodes, respectively, the factor p may be written:

öp = 1 + Öl(Ö _ 1) .

For the typical values Öl = 7 and Ö = 3.5, the valueof p is 1.2, but non-uniformity may increase this valueconsiderably.

The validity of Schottky's equation in the detectionof modulated light has been verified experimentally forall cathodes of the usual types between 75 Hz and100 kHz: deviations from the predicted values were 20 %at the most and the noise was in fact found to be"white". In all cases involving the detection of modu-lated light the value of the ratio SIN can therefore becalculated previously, provided that the dark currentis negligible compared with the d.c. component of thesignal. The value IJ, of this d.c. component must in factbe introduced in Schottky's equation to find the noise.

[14] G. Piétri, Glimpse on some problems encountered in thestudy and construction of photomultipliers with fast re-sponse, Le Vide 15, 120-131, 1960.

If m is the depth of modulation, the ratio SIN is givenby:

SN

m'YJc2h2

4epIJf'YJc1k 4epIJ/

It can be seen that the methods of improving thisratio SIN are always the same: improvement of thecollection efficiency of the input optics and the firststages (which increases Ö and reduces p).

When modulated light is detected in the presence ofa non-negligible dark current, Schottky's equation, evenwhen corrected by the substitution of h+ 10 for Ix,is only valid to a first approximation. In fact, if thedark current is not mainly thermionic in origin, itsfluctuations do not obey this law; the noise producedby the dark current in a circuit of bandwidth IJf maydepend on frequency, and it may be larger or smallerthan the value predicted from Schottky's equation,depending on whether the dark current is due to sein-tillations (ambient radioactivity) or to ohmic leakage.

Finally, we should note that in the detection ofmodulated light it is always desirable to derive as muchbenefit as possible from the amplification of the multi-plier by making the gain sufficiently high that theSchottky noise at the anode is much greater than thethermal noise in the load resistance R. This conditionmay be expressed as:

2kT 1G2» -----.

ep RIk

The proper value of the load R is usually determinedby the required bandwidth and the stray capacitances.

. . . . . (5)

(4)

Speed of response

We shall now present a more detailed account of thephysical effects which determine the speed of responseof photomultipliers, and we shall then give a discussionof recent improvements in this field.

1) Effects limiting the speed of response [14l.

The effects to be taken into account are the initialvelocities of the electrons and the induced currentscaused by the movement of the electrons through thetube.

Let Wn be the energy corresponding to the compo-nent of the initial velocity of an electron (of mass m)normal to the emissive surface. This component willhave the effect of reducing the transit time of the elec-tron with respect to the transit time of an electronemitted at zero initial velocity. It can be shown thatthis reduction of time is given by:

At = "1/12m vw: ,LI • • • • • (6)e Eo


where Eo is the electric field at the emissive surface.Since Wn fluctuates from one electron to another witha mean value Wn, this expression with Wn = Wn willgive the order of magnitude of the fluctuation in thetransit time of the electron. For a given geometry thiseffect will thus vary in inverse proportion to the appliedvoltage.

Let us now consider the transverse velocity compo-nent (perpendicular to the mean electron path). Thiscomponent has the effect of spreading out the electronbeam over the surface of the first dynode Sl. In practice,the cross-section ofthe beam at this location is arrangedto be no larger than the useful surface of the dynode,that is to say a surface such that the electrons emittedfrom any point of it wiIl have a reasonable chance ofreaching the next dynode. The effect of the transversevelocities may now be seen from fig. 12a and b: theasymmetry gives a variation in the length of the pathwith the direction and magnitude of the initial trans-verse velocity. On the other hand, the electrons emittedfrom different points of Sl also take different times totravel from Sl to S2, again because of the asymmetry.There is thus a correlation between the spread in thesuccessive stages which makes any accurate calculationof the fluctuation observed at the anode impossible.The difference in transit time for paths starting from

two different points A and B varies with I/I/V. If thesetwo points are the points of arrival of two electronsemitted from the same point on the preceding surfacewith two different values of initial transverse velocity,their distance apart will also vary approximately asI/VV. If the spread in the transit times of electronsemitted from points A and B varies linearly wtth thedistance AB, we may predict that the final spread willvary approximately as I/V.The difference in the average transit times to the first

dynode of electrons starting from the centre and theedge ofthe cathode (times which can easily be measuredand calculated) is referred to more simply as the"centre-to-edge spread".The induced currents arise when the electrons travel-

ling to the anode arrive in its vicinity, and they last foras long as the movement of these electrons [15]. It isdesirable by suitably shielding the anode to try andsuppress the effect of electrons not travelling to theanode and to reduce the effect ofthe "useful" electronswhich are travelling towards it. The .geometry of Sn inthe 56 AVP meets this requirement. However, the elec-trons moving towards Sn pass through the anode,which takes the form of a grid, and these electronsalready induce an unwanted signal. Moreover, the elec-trons emitted by Sn are not all captured immediatelywhen they arrive at the anode. Some pass through itand oscillate for some time around the wires of the

grid, inducing high-frequency currents in the wires.These currents become even larger if their frequencyhappens to coincide with the resonant frequency asso-ciated with the capacitances and inductances of theanode system.The signal is taken off between Sn and the anode: the

connections to these two electrodes form the beginningof the line along which the signal will be transmitted.The impedance of this line is in general very differentfrom that of the standard coaxial cables and there arediscontinuities in it which give rise to various perturba-tions in the shape of the pulse.

Finally, the inductances of the dynode connections,which carry rapidly varying currents, give variations inthe dynode potentials, and hence gain variations, whichalso have the effect of distorting the pulse.

2) Improvement of the pulse response.

All the effects considered above tend in a varyingdegree to lengthen the pulse. It can be said, however,that in all commercial photomultipliers so far in use themain cause of the longer pulse response is to be foundin the initial velocities; the after-pulses observed in the56 AVP (fig. 16) are about the only exception sincethey are due to oscillations of the electrons around theanode. We have seen that it is difficult to calculate theoverall response of a photomultiplier from the responseof the individual stages because of the correlation be- .tween the stages. Useful results can be obtained, how-

56AVPT=2nsL =2.Bns

XP1020T=1.5nsL=2.Bns

XP 1210T=QBnsL=1.2ns

Fig. 16. Pulse response of three fast photomultipliers. In eachcase the rise time 1: is indicated and the width L at half-heightof the pulse.

1968, No. 8/9 'PHOTOMULTIPLIERS 281

ever, if it is assumed that the responses of all the stagesadd as the square. Let aJ{SI, aSIS2 and aS(n-l)Sn bethe standard deviations corresponding to electron tran-sit-time fluctuations in the first, second and nth stages.The responses of the stages beyond the second mayusually be taken as equal, so that we may write:

A.2 = a~Sl + a~lS2 + (n - l)a~(n-l)Sn.

The width at half-height of the pulse response, L, isrelated to the standard deviation A. by: L = 2.36 A.. Inthe 150 AVP, for example, the most important termis a~s" and for practical purposes we have: L =2.36 atcsi R:::I 10 ns. In the development of the 56 AVP,whose speed of response was to be higher, the maineffort was therefore concentrated on the input optics.The pulse responses of all the stages have widths oftheorder of 0.8 ns (except for stage Sl-S2, where the re-sponse is about twice as long). With 14 dynodes, i.e.15 stages including the stage from Sn to the anode,this results in an overall pulse response of about0.8 X VB = 3 ns. (The contribution from the Sn-anodestage is due to the induced currents.)A further substantial reduction in the length of the

pulse response can only be obtained by simultaneouslyreducing the spread in all the stages. This has beendone with the XP 1210, now under development, byincreasing the voltages applied between the dynodesby a factor of 2 or 3, and by modifying the input opticsand the stage Sl-S2. This results in a reduced centre-to-edge spread and a 7 times higher electric field at thecathode. (These modifications had in fact already beenmade for the XP 1020, as we shall see later.)We have already touched on the difficulties caused

by increasing the applied voltages. The anode here is ashielded disc and the signal is collected on a matchedtransmission line. Finally, the inductance ofthe dynodeconneedons has been considerably reduced by using thedisc construction, whose relatively large capacitancealso appreciably improves the dynamic stability of thepotentials. The pulse response obtained has a width of1.1 ns and a rise time of 0.8 ns.If the full benefit is to be derived from a rise time of

the order of a nanosecond, the fast photomultipliersshould have a very high gain (> 108). This can beseen as follows: the electron tube systems which haveto be used for processing the time information obtainedrequire pulse levels of the order of 10 to 20 V (3 to5 V for semiconductor systems) across an impedanceof 50-125 ohms. Amplification is hardly possible, sinceconventional amplifiers of suitable bandwidth haveonly recently become available and they are still expen-sive. A burst of say 10 photoelectrons must thereforeproduce a signalof 10 V across an impedance of50 ohms, which means that the gain has to be 108• To

provide this gain, 14 stages are necessary in the 56AVP,in which the interdynode potentials are of the order of120 V; however, 10 stages are sufficient in the XP 1210.This helps to improve the speed of the device and more-over the total high voltage does not have to be increasedalong with the interdynode potentials.

3) Reduction of the spread in the overall transittime [16].

Fast photomultipliers are very often used not onlyfor determining directly the shape of a light pulse butfor determining as accurately as possible the instant ofappearance of a short flash of light of low intensity,producing very few photoelectrons (sometimes onlyone). An example is the pulse produced by the Cerenkoveffect in a transparent medium by the passage of a high-energy particle. This pulse, which produces 100 elec-trons at most, lasts about 10-11 second, and it is neces-sary to determine the instant at which it appears withan error of less than 10-10 second, since the velocity(and energy) ofthe particle can be determined from thetime the particle takes to cover the distance betweentwo photomultipliers provided with Cerenkov radia-tors.The accuracy of such measurements is limited by the

spread in the total transit time in the photomultiplier.This spread is of course related to the spread in thedifferent stages, but we shall see that the contributionsfrom the different stages do not have the same relativeimportance as in the calculation of the pulse response.What we therefore have to do is to examine the spreadin the time taken by the avalanche initiated by theemission of a single electron by the cathode to traversethe photomultiplier. The transit time of the electronbetween K and SI varies with the point of emission andthe initial velocity, with a standard deviation of aI{SI

(see above). If N electrons are emitted at the sametime, it may be assumed that the position of thecentreof gravity of the resultant pulse fluctuates with a stan-dard deviation of aJ{sl/J!N. Similarly, when (j secondaryelectrons are emitted at the same time by the dynode SI,the fluctuation of the position of their centre of gravityis aSls2/j/(5 (as a first approximation; the fact that (j

fluctuates must also be taken into account). The stan-dard deviation at of the total fluctuation for the caseof a single electron is found by summing the squares,taking care to avoid the difficulties which arise becauseof the correlation between the paths in successivestages. Introducing a as the standard deviation of thetotal fluctuation for the case of N electrons emittedsimultaneously, we arrive at the relation:[15] The theory of the operation of high-frequency diodes and

trio des will be found in the standard works on electronics.[16] E. Gatti and V. Svelto, Nucl. Instr. Meth. 4, 189-201, 1959.

E. Gatti and V. Svelto, Nucl. Instr. Meth. 30, 213-223, 1964.

r---------------------------~--- -~- -----


a2N = a 2 = a2 + a~IS2 (I+~)+~~£__(I+~)t _ KSI ch ch ~l(~-I) ~ ,

. . . (7)

assuming that only the first stage gives values of aSlS2

and ~l different from those of the other stages, aSS

and ~.It can be seen that the contributions from the input

optics and from the stage Sl-S2 are by far the mostimportant, even though all the a values are of the sameorder of magnitude. A substantial reduction of ass:and of aSlS2 would thus greatly assist in reducing thespread in transit time while the effect of this on thepulse response would be insignificant. In view of thisthe input optics and the stage Sl-S2 in the XP 1020were redesigned. In the XP 1210 the increase of theoperating voltages also resulted in a substantial reduc-tion in aSlS2. The table below shows the contributionsfrom the various stages, expressed in nanoseconds, forthree types of fast photomultiplier. It can be seen that,for the last two types, the dominant term is the one dueto the stage Sl-S2; this is accounted for by the specialconfiguration of this stage. To give a numericalexample, the probable error in the measurement of theinstant at which a single photoelectron is emitted fromthecathodeofanXP I2IOis2.36xO.13 = 0.31 ns, whilethe measurement of the instant of arrival at the cathodeof a Cerenkov pulse giving 100 photoelectrons is sub-ject to an error of 0.31/VIOÖ ns, i.e. 3.1 x 10-11 s.

- -- -----~ T 56 AVP! XP 1020 -T XP 1210--+-----~-----I 0.42 I 0.07 I 0.07

aSlS21/ I +~:/61 ---1~0:~1- 0.20 1 0.10

1 0.06 1 0.04 -1-- 0.03

aKS1

aSS -,/1_+ }_/~1_61(6 -I)

aJfN= at 1 0.59 I 0.22 I 0.13

As an illustration, we shall describe a particularlyuseful application of this knowledge of the spread intransit times: the measurement of the time constant ofthe decay of short scintillations, where the duration ofthis time constant is less than or equal to the pulseresponse ofthe photomultiplier (ofthe order of 10-9 s).

Let us assume that the pulse of light can be repeatedas often as required, and that the resolving time issufficiently short (10-10 s for example). We assume that-the scintillation causes the emission from the cathodeof an average of 100 photoelectrons. If we attenuate thisflux by a factor of 1000, each scintillation will then giverise to an average emission of ii = 0.1 photoelectron.This means that, over a large number of measurements,

there will be an average emission of 1 photoelectronfor every 10 scintillations. To be more exact, there isa certain probability that a scintillation will cause theemission of a burst of more than 1 photoelectron, butby taking a fairly small value for ii it can be shown thatthis probability is small compared with the probabilityof the emission of a single electron. In fact, whenPoisson's law is applied, the probability for the emis-sion of n photoelectrons can be written:

iinPn = - exp (-ii).

n!

The probability that a burst of at least 1 electron willbe emitted, that is to say the probability of a signalappearing at the anode, is:

<IJ

~ Pn = I-Po = I-exp(-ii), .. (8)n=1

while the probability of obtaining just 1 electron is:

PI = ii exp (-ii). ..... (9)

It can easily be verified that if i'i is small compared withunity, the ratio Pl/(I - Po) of these probabilities is ofthe order of 1 - np: In the numerical example givenabove (i'i= 0.1) this ratio is equal to 0.95, in otherwords 95 % of the pulses recorded at the anode willoriginate from single electrons.

The shape of the decay curve of the light can nowbe determined by repeating the pulse many times andsampling the intensity at various instants. The accuracyis then equal to the spread of the transit time in thephotomultiplier (i.e. 0.13 ns in the case of the XP 1210).This can be seen as follows. Although every photo-electron received can have been emitted at any arbitraryinstant during the duration of the scintillation, theprobability that a photoelectron will be emitted be-tween tand t + LIt is simply:

f(t)LI t

J f(t)dt '

if f(t) is the curve representing the variation of theluminous intensity with time. In other words, if wedetermine the arrival times of N single photoelectrons,the number of electrons observed between the moments tand t + LIt is equal to:

Nf(t)LltLlN=--.- J f(t)dt

(10)

Thus, except for a constant factor, f(t) is given byLlN/LI t. The speed of the effects that can be investigatedis no longer limited by the pulse response of the photo-multiplier but by the spread of the transit times in thephotomultiplier, since this limits the accuracy withwhich the instant of emission of a photoelectron can

1968, No. 8/9 PHOTOMULTIPLIERS

be measured. It is obvious that a very large number ofmeasurements will be required if an exact curve is tobe obtained, particularly since very few of the measure-ments will con-espond to an observable pulse. Inciden-tally, this procedure can also be usefully applied to thelight pulses in measuring the speed (the pulse response)of the photomultiplier: their duration is measured, andif this is not negl igibly short compared with the re-sponse of the photomultiplier (l.1 ns), the informationobtained enables corrections to be applied.

4) Frequency response (modulated light).

The detection of modulated light is an increasinglyfrequent application of photomultipliers [17]. The func-tion C(w) that gives the gain of a photomultiplier as afunction of modulation frequency can in principle becalculated from the pulse response R(I), using the well-known relation:

C(w) = f R(I)e-jwtdl.

According to this relation the curve representing the

value of co for which C(w)jC(O) = I/V2. The followingtable gives the calculated bandwidth B for the three fastphotornultipliers mentioned earlier:

Tube Passband

56 AYPXP 1020XP 1210

65 MHz100 MHz300 MHz

We should mention that a heterodyne method can beused to detect light modulated at frequencies very muchhigher than the limit of the multiplier bandwidth. Thestage K-Sl can be made smaller in size, thus giving abandwidth considerably higher than 1000 MHz. lf asuitable control electrode is now introduced into thisstage, the electron beam can be modulated (by meansof a local oscillator) at a frequency ft close to the fre-quency fz at which the light is modulated. This gives afrequency conversi.on in which the difference-frequency

(11) signal is brought within the bandwidth of the photo-multiplier. This has been done in our laboratories [18]

with an experimental tu be PM 994 (fig. /7) at a fre-

Fig. 17. Experimental photomultiplier type PM 994, used for the heterodyne detection ofmodulated light. Total length 14 cm.

function C(w) assumes a more extended form the nar-rower the function R(!). Thus, the bandwidth of aphotomultiplier used as a detector of modulated lightbenefits directly from the improvements of the pulseresponse. On page 273, the bandwidth B, i.e. the upperlimit of the passband, has already been defined as the

[17] C Cernigoi, I. Gabrielli and G. Iernetti, Nucl. Jnstr. Meth.6, 193-200, 1960.

[18j J. Nussli, J. Physique, Suppl. Phys. appl., 26, II3A-114A,1965. G. Marie and J. Nussli, CR. Acad. Sci. Paris 258,5179-5182, 1964.

[lUJ G. Piétri, J. Nussli and M. Brauit, Perfectionnements recentsapportés aux photomultiplicateurs rapides destinés à laphysique nucléaire, Electronique Nucléaire, Compte renduColloque int. S.F.E.R., Paris 1963, pp. 793-805. F. Kirsten,UCRL 8706, pp. 1-7, University of California, Febr. 1959.J. H_ Malrnberg, Rev. sci. Instr. 28, 1027-1029, 1957.

quency close to 1000 MHz; the transmitted signalwas that of a television picture.

5) Measurements of speed of response.

Since the phenomena for which we are trying to develop photo-multipliers of suitable time resolution are already at the limits ofthe measurable, it is reasonable to ask how characteristics suchas the extremely fast speed of response can be checked. In suchprocedures it is necessary to use extremely short light pulseswhich are well-defined in time. The shortest pulses that can beproduced are those due to the passage of relativistic chargedparticles in transparent materials, due to the Cerenkov effect.This procedure, however, is seldom feasible, and the light pulsesusually employed are derived from a high-voltage spark dischargebetween two closely adjacent electrodes [19J. This method has theadvantage that absolutely synchronous electric pulses are avail-able which can be used to give a time reference (for triggering an

283


oscilloscope). However, the duration of these sparks is always ofthe order of 10-9 s, which is short enough for tubes like the56 A VP but about equa I to the response ti me of the XP 1210.A proper measurement can therefore only be made if one candeterrnine in some other way the exact shape of the light pulse,say by means of an ultra-fast pbotoemissive cell, like theTVHC 20 or the TVHC 40 (fig. /8; these cells have a rise timeof 0.2 ns). This presupposes that the light from the spark issufficient to produce a signal that does not have to be amplifiedbefore application to the oscilloscope. Knowing the shape of thelight pulse one can then calculate, in certain cases at least, thepulse response of the photornultiplier from the shape of its out-put pulse.

I f, on the other hand, the delay in the triggering of the oscillo-scope is known, one can calculate the mean transit time in thephotornultiplier. If the light produced by the flash is focused ona point of the cathode, and if this point is moved across thecathode surface, the centre-to-edge spread can be estimated fromthe change in the transit time, that is to say from the horizontaldisplacement of the oscilloscope trace. Finally, if the light isattenuated to such an extent that there is only very little probabil-ity of two photoelectrons being emitted sirnultaneously, the jitterobserved on the trace enables an estimate to be made of the transittime fluctuations. (Fortunately there are other methods for thispurpose that are more exact, e.g. photographic recording of alarge number of traces, or time-amplitude couverters [201.)

The frequency response may in principle also be determineddirectly with the aid of Pockels-effect modulators, or by thedetection of beats between the different modes of a laser.

Extension of linearity

The non-linearity which arises because the gain andsensitivity depend on the current introduces so manyproblems that we cannot deal with them all here. How-ever, we should at least look into what happens whenthe anode current becomes a significant fraction o: ofthe d.c. current lp which flows in the resistances con-nected in parallel with each stage and ensures that thedesired distribution of voltage over the stages is main-tained.

Let Vo be the interdynode potential. Between Sn andthe anode there is therefore a voltage drop a Vo, butsince this stage is only a collection stage, a fall in volt-age does not cause a fall in current. However, as thehigh voltage is stabil ized, the excess potential a Vo isapplied to the other stages, which amounts to multi-plying the voltage applied to the photomultiplier by afactor of 1 + «[n. Since the gain varies approximatelyas Von, the gain of the photomultiplier will be multi-plied approximately by

(I + Cf./n)n.~ 1+ a,

provided that a is less than, say, 0.5. The photomultiplieris thus seen to be "superlinear", i.e. its response in-creases faster than the signal.

With light pulses, the situation is different. In thiscase, capacitors can be arranged at the ends of thedynodes to stabilize the potentiais. As a result, during

_J

Fig. 18. Ultra-fast photoernissive cell TVHC. It contains a photo-cathode deposited on a metal disc and a grid-type anode. Thedistance between these two planar electrodes is about 2 mm,which reduces the transit time of the photoelectrons and hencethe duration of the current they induce. This cell is produced bythe transfer technique and may be integral with one end of acoaxial cable.

short intervals very large currents can be obtainedwhich, however, are limited by the space charge in thefinal stages, so that the response will be sublinear. Theonly way to reduce the space-charge effect is to increasethe electric field, particularly on the last dynode. Thatis why the anode is nearly always in the form of a gridsituated at a short distance (I mm) from the last dynode.By applying to the last dynodes progressively highervoltages (doubling at each stage) up to 500 Y betweenSn and the anode, the linearity range of the 56 AYPcan be extended up to 300 mA. In the XP 1210 thegrid-type anode could not be used, as we noted earlier.The plate adopted for the anode in this tube does notpermit such an extended linearity region, the limit beingreached at 80 mA, or 4 Y across 50 ohms. This is nowa-days sufficient, owing to the use of semiconductorelectronic systems.

(12)

[20] A. E. Bjerke, Q. A. Kerns and T. A. Nunamaker, UCRL9976, University of Californ ia, Febr. 1962. G. Present, A.Schwarzschild, I. Spirn and N. Wotherspoon, Fast de-layed coincidence technique: the XP 1020 photornultiplierand limits of resolving times due to scintillator character-istics, Nucl. Instr. Meth. 31, 71-76, 1964. G. Bertolini, Y.Mandl, A. Rota and M. Cocchi, Time resolution measure-ments with XP 1020 photomultipliers, Nucl. Instr. Meth. 42,109-117, 1966.

[2l1 G. Breuze, J.-P. Fertin and R. Petit, Photornultiplicateurrapide à fort courant de sortie et grand dornaine de linéarité,Electronique Nucleaire, Compte reridu Colloqueint. S.F.E.R.,Paris 1963, pp. 113-125.


In certain very special cases even higher limits oflinearity are required. Types XP 1140 and 1141 arefast tubes designed for voltages of 2000 Y between thelast dynodes. The limit of linearity here may be as highas 3 amperes [211.

Study of particular problems

Tubes of unusually large or smal! dimensions

In order to increase the probability of detecting par-ticles whose paths are not known beforehand, it isnecessary to increase the volume and the surface areaof the scintillators and hence the surface area of thecathode used in the photomultiplier. On the other hand,in some equipment the space available is very limited.

The best comprornise between the various character-istics of a photornultiplier has been found for diametersof 40 to 50 mm. Because of the way in which secondaryemission varies with voltage, the interdynode potentialshave to be between 100 and 200 Y (200 to 500 Y be-tween cathode and SI); any significant change of scaleis therefore reflected in a corresponding variation of theelectric fields. Scaling up the dimensions linearly more-over gives a linear increase in the magnitude (andspread) of the transit times. Scaling down the dimen-sions, on the other hand, gives rise to problems ofinsulation and dark current (field emission). A typicalexample is that of the 58 AYP (fig. /9), which had tobe fitted with a cathode of 110 mm - 2.5 times aslarge as the cathode of the 56 AYP - but still gives

Fig. 19. Fast photomultipliers 56 AYP (on the right) and58 AYP. Length of tubes 18 and 27 cm, respectively.

the same speed [14]. Linear scaling-up would havemeant a proportionate increase in response time: itwas therefore necessary to keep the same multipliersystem. Now, in the input optics, if the speed of re-sponse and the dimensions of the arrival area on thefirst dynode were to remain the same, the potentialswould have to have been increased by a factor of (2.5)2and the electric fields by a factor of 2.5. This was un-acceptable for various reasons, in particular becauseof field emission. A compromise solution was found byfitting an extra electrode (like the one in the XP 1020)which is set to the potentialof one of the later dynodes;the negative side of the cernpromise is that some dete-rioration in the centre-to-edge spread and the collectionefficiency has to be accepted.

Another example is the 152 AYP [2], which has acathode of only 19 mm diameter. Linear scaling-downwould have required impossibly high electric fields andimpossibly small dimensions for the dynodes. Reducingthe length of the dynodes would have upset the opera-tion of the multiplier as a result of charge effects on thesurface of the spacers. A multiplier with special cup-shaped dynodes was therefore developed for this tube(fig·20).

Improvement of mechanical strength

Most photomultipliers are not particularly rugged,because of the nature of their construction. Insulationrequirements, for example, may often make it necessaryto use insulators which are more fragile than mica, suchas glass or ceramics. The XP 1210, however, has beenspecially constructed to withstand vibrations with accel-erations greater than 25 g. The 152 AYP can with-stand the same conditions, because of its small size(and some slight modifications). A specially ruggedizedversion of this type (fig. 21) has been developed at theworks at Brive, in which the tube is shaped like a longrectangular block; all the electrodes are connected topins at two opposite faces. With this arrangementthe tube can easily be mounted between two printed-wiring boards and the whole assembly encapsulated ina potting resin.

Future developments

It is unlikely that there will be any major advancesfor some time in the speed characteristics of photornul-tipi iers intended for nuclear research, since the perform-ance of present-day tubes is often right at the limitsrequired for present-day experirnents. Similarly, weshould not expect any further substantial increase inthe quantum efficiency of the photocathode, which isalready very high in modern photocathodes (in bluelight), although we can look forward to a wider applica-tion of these cathodes and, in the longer term, to the


development of cathodes that are more sensitive in the0.7-1 [Lm region. This would meet more effectively therequirements of photometry in the near infra-red

There are, however, two photomultiplier character-istics which are still capable of considerable improve-ment: these are the dark current and the statisticalfluctuations of the gain.

We have seen that the dark current is chiefly relatedto the presence of caesium on surfaces where it is notwanted. At the present time a new manufacturing tech-nique is being studied, the "transfer technique", in

from their amplitude alone the pulses due to bursts ofone, two or three photoelectrons - it would be veryuseful to distinguish between such bursts in investiga-tions of very weak scintillations like those produced bysoft beta rays. The solution is to be found in otherprocesses of electron multiplication. We may mentionthe following:a) Channel multiplication [22]. Already used (in theform of single channels) in "windowless photo multi-pliers" for the detection of low-energy charged particlesor soft X-rays or for far UV detection, this technique

Fig. 20. Small-diameter photomultiplier 152 AVP. Length 10 cm.

Fig. 21. Ruggedized version of the 152 AVP, type PM 400. Length 9 cm.

which the photocathode can be activated in situ with-out contaminating the rest of the tube. The two partsof the tube are mounted separately inside the bell-jarof a vacuum equipment and the cathode is activatedfrom a small evaporator placed right against it. Oncethe cathode has been formed, the two parts are broughttogether - still under vacuum - and a seal 's effectedby an indium pressure-weld. As a result, the spontane-ous emission from the dynodes is eliminated and emis-sion from the cathode reduced to its thermionic com-ponent alone. These advantages are gained at the ex-pense of a somewhat lower gain (so that the voltagehas to be increased).

The statistical fluctuations of the gain are known tobe related to the low val ues of the mean secondary-emission coefficients, which lie between 3 and 8. Thesefluctuations make it impossible, for example, to identify

is now being extended to image intensification, makinguse of multiple channels.b) Multiplication by a field effect in thin insulatinglayers [23].

c) Multiplication by the creation of electron-hole pairsin semiconductor junctions (like those already in usefor detecting nuclear radiations). In this case, photo-electrons are accelerated by a voltage of the order of40 kV and are detected as beta particles in the semi-conductor junction. Encouraging results have alreadybeen obtained in our laboratories [24]. The relativelylow gain (104), however, necessitates the use of speciallow-noise narrow-band amplifiers.

The PM 994 photomultiplier described earlier, whichis used for the heterodyne detection of modulated light,could well be the forerunner of a new type of tube inwhich the information is processed at the photocath-


ode. This technique, applied to pulses, amounts in factto a sampling technique. Another possible techniqueis to use an optical system between cathode and multi-plier to form an image of the cathode, and to scan theimage by means of a magnetic deflection system: veryaccurate information about the position of a point oflight on the surface of the cathode can be obtained inthis way. This method is at present being used for star-tracking equipment in space satellites [251.The material presented above has demonstrated how

the development of photomultipliers is moving towardsa greater specialization of types and towards a betteradaptability to a whole range of different requirements.This evolution will of course be affected by the develop-ment of solid-state nuclear detectors, whose performancein the special field of high-resolution speetrometry isand will remain incomparably superior to that ofphotomuitipliers, at any rate for medium-energy radia-tion [261. For low-energy radiation the penetration ofthe radiation into a junction is too small, and for veryhigh energies too little energy is transferred in a junc-tion.We may conclude that for a long time to come photo-

multipliers will retain their supremacy as the idealinstruments wherever information has to be derivedfrom effects in which light is produced. In the detectionof nuclear radiation, the photomultiplier will continue

in demand as an indispensable instrument (with poten-tial for further perfection) in the investigation ofextremely rapid phenomena or of very high or lowenergy radiations.

[22] J. Adams and B. W. Manley, IEEE Trans. NS-13, No. 3,88-99, 1966. J. Adams and B. W. Manley, The channel elec-tron multiplier, a new radiation detector, Philips tech. Rev.28, 156-161, 1967 (No. 5/6/7).

[23] H. M. Smith, J. E. Ruedy and G. A. Morton, Performanceof a photomultiplier with a porous transmission dynode,IEEE Trans. NS-13, No. 3, 77-80, 1966.

[24] P. Chevalier and J. Nussli, Photomultiplicateur à hauterésolution utilisant un multiplicateur semi-conducteur, C.R.Acad. Sci. Paris 264B, 462-465, 1967 (No. 6).

[25] W. D. Atwill, Electronics 33, No. 40, 88-91, 1960.[26] W. K. Hofker, Semiconductor detectors for ionizing radia-

tion, Philips tech. Rev. 27, 323-336, 1966.

Summary. The manufacture of photomultipliers now encompass-es an astonishing variety of types of tube, in which attention ispaid to the requirements of widely different applications. After adescription of the general principles underlying the design ofphotomultipliers and their characteristics, the article examines thestatistical effects that limit the performance of photomultipliers,in relation to the introduetion of noise and speed of response.It is shown how the study of these effects gives a better under-standing of the relations existing between the various parametersand leads to an overall optimization of performance. Thus, theappropriate use of high electric fields, which give a high speedof response but tend to increase the dark current, has resultedin a pulse response of nearly 1 ns in the fastest photomultipliers,and a transit-time fluctuation of-nearly 0.1 ns. The article alsogives a description of several of the newest developments suchas heterodyne detection of modulated light, and the concludingsection indicates the probable trepds of future development.


Recent scientific publicationsThese publications are contributed by' staff of laboratories and plants which formpart of or co-operate with enterprises of the Philips group of companies, particularlyby staff of the following research laboratories:

Philips Research Laboratories, Eindhoven, Netherlands EMullard Research Laboratories, Redhill (Surrey), England MLaboratoires d'Electronique et de Physique Appliquée, Limeil-Brévannes

(Val-de-Marne), France LPhilips Zentrallaboratorium GmbH, Aachen laboratory, Weisshausstrasse,

51 Aachen, Germany APhilips Zentrallaboratorium GmbH, Hamburg laboratory, Vogt-Kölln-

Strasse 30, 2 Hamburg-Stellingen, Germany HMBLE Laboratoire de Recherches, 2 avenue Van Becelaere, Brussels 17

(Boitsfort), Belgium. B

Reprints of most of these publications will be available in the near future. Requestsfor reprints should be addressed to the respective laboratories (see the code letter) orto Philips Research Laboratories.: Eindhoven, Netherlands.

G. A. Allen: The activation energies of chromium, ironand nickel in gallium arsenide.Brit. J. appl. Phys. (J. Physics D), ser. 2, 1, 593-602,1968 (No. 5). M

G. Baker & D. E. Charlton (Mullard Ltd., Southamp-ton): Low frequency noise in copper doped germaniuminfrared detectors caused by thermal impedance fluctua-tions.Infrared Phys. 8, 15-24, 1968 (No. I).

J. R.A. Beale & J. A. G. Slatter: The equivalent circuitof a transistor with a lightly doped collector operatingin saturation.Solid-State Electronics 11, 241-252, 1968 (No. 2). M

P. Beekenkamp & J. M. Stevels: A suggestion for anew nomenclature to describe the structure of vitreoussystems and their related crystalline compounds.Phys. Chem. Glasses 9, 64-68, 1968 (No. 2). E

F. Berz: Large signal a.c. field effect in depletion layersfor wide gap semiconductors.Surface Sci. 10, 58-75, 1968 (No. I). M

G. Blasse & A. Bril: Investigations on Bi3+-activatedphosphors.J. chem. Phys. 48, 217-222, 1968 (No. I). E

G. Blasse & A. Bril: The influence of crystal structureon the fluorescence of oxidic niobates and related com-pounds.Z. phys. Chemie Neue Folge 57, 187-202, 1968 (No.3-6). E

G. Blasse & A. Bril: Fluorescence of Eu2+-activatedalkaline-earth aluminates.Philips Res. Repts. 23,201-206, 1968 (No. 2). E

G. Blasse, W. L. Wanmaker (Philips Lighting Division,Eindhoven), J. W. ter Vrugt (idem) & A. Bril: Fluores-cence of Eu2+-activated silicates.Philips Res. Repts. 23, 189-200, 1968 (No. 2). E

J. Bonnefous: Propriétés directionnelles de quelquessubstances ferromagnétiques polycristallines. (Thèse,Orsay 1967.)Acta electronica 11,7-112, 123-179, 1968 (Nos. 1,2). L

J. van den Boomgaard: On the system Cr-SiC.Philips Res. Repts. 23, 270-280, 1968 (No. 3). E

G. A. Bootsma: Gas adsorption on cadmium sulphide.Surface Sci. 9, 396-406, 1968 (No. 3). E

G. Bosch (Philips Radio, Gramophone and TelevisionDivision, Eindhoven): Photoconduction of n-type SiC.Philips Res. Repts. 23, 139-141, 1968 (No. 2).

A. J. Bosman & C. Crevecoeur: Dipole relaxation lossesin CoO doped with Li or Na.J. Phys. Chem. Solids 29, 109-113, 1968 (No. I). E

H. Bouma & J. J. Andriessen (Institute for PerceptionResearch, Eindhoven): Perceived orientation of isolat-ed line segments.Vision Res. 8, 493-507, 1968.

G.-A. Boutry &M. Jatteau: Sur la mesure des tempéra-tures de surface par télévision dans l'infrarouge.C.R. Acad. Sci. Paris 266B, 214-217, 1968 (No. 4). L

S. E. Bradshaw (Associated Semiconductor Manufac-turers Ltd., Wembley, England): The effects of gaspressure and velocity on epitaxial silicon deposition bythe hydrogen reduction of chlorosilanes.Int. J. Electronics 23, 381-391, 1967 (No. 4).

1968, No. 8/9 RECENT SCIENTIFIC PUBLICATIONS 289

S. D. Brotherton (Associated Semiconductor Manufac-turers Ltd., Wembley, England): Dependence of MOStransistor threshold voltage on substrate resistivity.Solid-State Electronics 10, 611-616, 1967 (No. 6).

S. D. Brotherton & J. B. Phillips (Associated Semicon-ductor Manufacturers Ltd., Wembley, England): Thecontrol of threshold voltage of complementary pairm.o.s. transistors.lEE Conf. Publn, 30: Integrated circuits, p. 126-131,1967.

K. H. J. Buschow & A. S. van der Goot: Inter.netalliccompounds in the system samarium-cobalt.J. less-common Met. 14, 323-328, 1968 (No. 3). E

H. J. Butterweck: A theorem about lossless reciprocalthree-ports.IEEE Trans. CT-IS, 74-76, 1968 (No. I). E

P. J. Buysman & W. Luiten: The relationship betweenthe structure and the strength of ceramic shell moulds,&: Permeability of ceramic shell moulds.Proc. 12th Conf. European Investment Casters' Feder-ation, Eindhoven 1967, paper No. XI, 12 & 14pp. E

J. R. Chamberlain, A. C. Everitt & J. W. Orton: Opticalabsorption intensities and quantum counter action ofEr3+ in yttrium gallium garnet.J. Physics C (Proc. Phys. Soc.), ser. 2, 1, 157-164, 1968(No. I). M

F. J. du Chatenier: Space-charge-limited photocurrentin vapour-deposited layers of red lead monoxide.Philips Res. Repts. 23, 142-150, 1968 (No. 2). E

R. W. Cooper, W. A. Crossley, J. L. Page & R. F.Pearson: Faraday rotation in YIG and TbIG.J. appl. Phys. 39, 565-567, 1968 (No. 2, Part I). M

H. J. van Daal: Polar optical-mode scattering of elec-trons in Sn02.Solid State Comm. 6, 5-9, 1968 (No. I). E

M. B. Das (Associated Semiconductor ManufacturersLtd., Wembley, England): Switching characteristics ofm.o.s. and junction-gate field-effect transistors.Proc. lEE 114, 1223-1230, 1967 (No. 9).

M. B. Das (Associated Semiconductor ManufacturersLtd., Wembley, England): Dependence of the charac-teristics of MOS-transistors on the substrate resis-tivity.Solid-State Electronics 11, 305-322, 1968 (No. 3).

J. B. Davies & B. J. Goldsmith: An analysis of generalmode propagation and the pulse-shortening phenom-enon in electron linear accelerators.Philips Res. Repts. 23,207-232, 1968 (No. 2). M

A. M. van Diepen (Natuurkundig Laboratorium derUniversiteit van Amsterdam), H. W. de Wijn (idem) &K. H. J. Buschow: Nuclear magnetic resonance inDyAl3, HoAl3, and a-ErAl3.Physics Letters 26A, 340-341, 1968 (No. 8). E

C. Z. van Doorn: Nature of donor-acceptor pairs incadmium sulphide.J. Phys. Chem. Solids 29, 599-608, 1968 (No. 4). E

J. J. Engelsman, J. Knaape & J. Visser: Volumetriedetermination of the UlO ratio in uranium oxides.Talanta 15, 171-176, 1968 (No. 2). E

u. Enz & H. van der Heide: Domain-wall mobility inSi-doped YIG.J. appl. Phys. 39, 435-437, 1968 (No. 2, Part I). E

J. Flinn: Surface properties of n-type gallium arsenide.Surface Sci. 10, 32-57, 1968 (No. I). M

B. Frank, E. Roeder & S. SchoIz: Verdichtung undFormgebung von Glaspulvern.Ber. Dtsch. Keram. Ges. 45, 231-233, 1968.(No. 5). A

J. G. C. de Gast: Berekening en constructie van hydro-statische lagers.Polytechn. T. Werktuigbouw 23, 141-150, 187-196,1968 (Nos. 4, 5). E

J. G. C. de Gast: Een nieuwe variabele voorrestrictievoor hydrostatische dubbelfilm-Iagers en zijn toepassingin precisiegereedschapswerktuigen.Ingenieur 80, 051-62, 1968 (No. 19). E

C. A. A. J. Greebe & W. F. Druyvesteyn: Some con-siderations on Alfvén-wave resonances in non-viscousliquid metals.Philips Res. Repts. 23, 121-130, 1968 (No. 2). E

C. A. A. J. Greebe, W. F. Druyvesteyn & A. J. Smets:Alfvén wave resonances in liquid Na.Physics Letters 26A, 337-338, 1968 (No. 8). E

J. A. Greefkes & F. de Jager: Continuous delta modu-lation.Philips Res. Repts. 23, 233-246, 1968 (No. 2). E

G. J. van Gurp: Flux-transport noise in type-U super-conductors.Phys. Rev. 166, 436-446, 1968 (No. 2). E

E. F. de Haan & K. R. U. Weimer: The beam-indexingcolour television display tube.Roy. Telev. Soc. J. 11, 278-282, 1967/68 (No. 12). E

C. Haas: Spin-disorder scattering and magnetoresist-ance of magnetic semiconductors.Phys. Rev. 168, 531-538, 1968 (No. 2). E

S. H. Hagen: Surface-barrier diodes on silicon carbide.J. appl. Phys. 39, 1458-1461, 1968 (No. 3). E

E. E. Havinga: Band structure and superconductivityof non-transition metals.Physics Letters 26A, 244-246, 1968 (No. 6). E

H. F. van Heek: Current saturation and negative resist-ance in evaporated CdSe layers.Physics Letters 26A, 175-176, 1968 (No. 5). E

290 PHILIPS TECHNICAL REVIEW VOLUME 29

H. F. van Heek: Hall mobility of electrons in the space-charge layer of thin film CdSe transistors.Solid-State Electronics 11, 459-467, 1968 (No. 4). E

J. C. M. Henning: Magnetic, optical and e.p.r. proper-ties of CS3ZnCls with dilute C02+ ions.Z. angew. Physik 24, 281-287, 1968 (No. 5). E

J. C. M. 'Henning: Semi-empirical calculation of 14Nhyperfine coupling factors in heterocyclic anions.Chem. Phys. Letters 1, 678-680, 1968 (No. 13). E

A. R. Hill: Uses of fine focused ion beams with high. current density.Nature 218, 202-203, 1968 (No. 5137). M

A.H. Hoekstra (Philips Lighting Division, Eindhoven):The chemistry and luminescence of antimony-contain-ing calcium chlorapatite.Thesis, Eindhoven 1967.

D•.vanHouwelingen & J. Volger: The superconducting-dynamo properties and applications.Philips Res. Repts. 23, 249-269, 1968 (No. 3). E

G. H. Jonker: The application of combined conduc-tivity and Seebeck-effect plots for the analysis of semi-conductor properties.Philips Res. Repts. 23, 131-138, 1968 (No. 2). E

Jo' A. Kerr (Associated Semiconductor ManufacturersLtd., Wembley, England) & L. N. Large (SERL,Baldock, England): Semiconductor devices made byion implantation.Proc. int. Conf. on applications of ion beams to semi-conductor technology, Grenoble 1967, p. 601-618.

F. M. Klaassen: Thermal noise in space-charge-limitedsolid-state diodes.Solid-State Electronics 11, 377-378, 1968 (No. 3). E

H. Klotz & B. Sebmidl: Verfahrenstechnik zur Rück-gewinnung von Indiumverbindungen aus Abgasen inder Lampenindustrie.Verfahrenstechnik 2, 25-27, 1968 (No. I). A

H. G. Koek, D. de Nobel, M. T. Vlaardingerbroek &P. J. de Waard: Continuous-wave planar avalanchediode with restricted depletion layer. .Proc. IEEE 56, 105, 1968 (No. I). E

J. van Laar, H. J. Endeman (Laboratorium voor Kris-talchemie der Rijks-Universiteit, Utrecht) & J. M.Bijvoet (idem): Remarks on the relation betweenmicroscopie and macroscopie crystal optics.Acta cryst. A 24, 52-56, 1968 (No. I). E

J. G. M. de Lau: High-frequency and microwave prop-erties of hot-pressed fine-grained ferrites.Proc. Brit. Ceramic Soc. 10, 275-284, 1968. E

D. B. Lee (Associated Semiconductor ManufacturersLtd., Wembley, England): Diffusion into silicon froman arsenic-doped oxide.Solid-State Electronics 10, 623-624, 1967 (No. 6).

W. Lems & Tb. Holtwijk: Magnetization reversal ofelectrodeposited FeNi films.J. Physique 29, Colloque C 2, 140-143, 1968. E

P. R. Loeber: Nuclear magnetic resonance of copperand cobalt in metallic conducting spinels.Z. angew. Physik 24, 277-280, 1968 (No. 5). E

J. Loeekx: Syntactische analyse en programmeertalen.Informatie 9, 277-282, 1967 (No. 12). B

F. K. Lotgering: Mixed crystals between binary sul-phides or selenides with spinel structure .J. Phys. Chem. Solids 29, 699-709, 1968 (No. 4). E

F. K. Lotgering & R. P. van Stapele: Magnetic proper-ties and electrical conduction of copper-containingsulfo- and selenospinels.J. appl. Phys. 39, 417-423, 1968 (No. 2, Part I). E

P. Massini & G. Voorn: Optical and photochemicalproperties of chlorophyll a solubilized in aqueous solu-tions of surfactants.Biochim. biophys. Acta 153,589-601, 1968 (No. 3). E

G. Meijer: Rapid growth inhibition of gherkin hypo-cotyls in blue light.Acta bot. neerl. 17, 9-14, 1968 (No. I). E

R. Memming & G. Neumann: On the relationship be-twee surface states and -radicals at the germanium-electrolyte interface.Surface Sci. 10, 1-20, 1968 (No. I). H

R.F. Mitebell: Some newmaterials for ultrasonic trans-ducers.Ultrasonics 6, 112-116, 1968 (No. 2). M

R. F. Mitcbell & M. Redwood (Dept. of Electrical &Electronic Engng., Queen Mary College, University ofLondon): Frequency response of a distributed piezo-electric source of sound.Electronics Letters 4, 107-109, 1968 (No. 6). M

F. D. Morten &R. E. J. King (Mullard Ltd., Southamp-ton): Multi-element infrared detectors for high in-formation rate systems.Infrared Phys. 8, 9-14, 1968 (No. I).

B. J. Mulder: Symmetry of the fluorescence and ab-sorption spectra of anthracene crystals.J. Phys. Chem. Solids 29, 182-184, 1968 (No. I). E

A. G. Nassibian (Associated Semiconductor Manufac-turers Ltd., Wembley, England): Effect of diffusedoxygen and gold on surface properties of oxidizedsilicon.Solid-State Electronics 10, 879-890, 1967 (No. 9).

A. G. Nassibian (Associated Semiconductor Manufac-turers Ltd., Wembley, England): Effect of gold on sur-face properties and leakage current of MOS transistors.Solid-State Electronics 10, 891-896, 1967 (No. 9).

1968, No. 8/9 RECENT SCIENTIFIC PUBLICATIONS 291

M. Noé & G. Vandewiele (Université Libre de Bruxel-les): The calculation of distributions of Kolmogorov-Smirnov type statistics including a table of significaneepoints for a particular case.Ann. math. Statistics 39,233-241, 1968 (No. I). B

J. M. Noothoven van Goor: Densities of charge carriersin bismuth.Physics Letters 26A, 490-491, 1968 (No. 10). E

J. M. Noothoven van Goor & H. M. G. J. Trum:Distribution coefficient and donor valency of telluriumin bismuth.J. Phys. Chem. Solids 29, 341-345, 1968 (No. 2). E

H. Ocken & J. H. N. van Vucht: Phase equilibria andsuperconductivity in the molybdenum-platinum sys-tem.J. less-common Met. 15, 193-199, 1968 (No. 2). E

C. van Opdorp: Evaluation of doping profiles fromcapacitance measurements.Solid-State Electronics 11, 397-406, 1968 (No. 4). E

A. E. Pannenborg: Decision making in research policyat the industrial level.Ciba Foundation and Science of Science FoundationSymp. on decision making in national science policy,1968, Churchill, London, p. 99-103. E

R. F. Pearson & A. D. Annis: Anisotropy of Fe3+ ionsin yttrium iron garnet.J. appl. Phys. 39, 1338-1339, 1968 (No. 2, Part II). M

J. B. H. Peek: The measurement of correlation func-tions in correlators using "shift-invariant independent"functions.Thesis, Eindhoven 1967. E

L. G. Pittaway, J. Smith & B. W. Nicholls: Laser in-duced electron emission using a pulsed argon laser.Physics Letters 26A, 300-301, 1968 (No. 7). M

T. Poorter (Philips Electronic Components and Mate-rials Division, Eindhoven): The relation between re-verse-secondary-breakdown behaviour and the proper-ties of the epitaxiallayer of planar epitaxial power tran-sistors.Philips Res. Repts. 23, 281-308, 1968 (No. 3).

A. Rabenau, H. Rau & G. Rosenstein: Über die Blei-sulfidhalogenide Pb5S2J6 und Pb7S2BrlO.Naturwiss. 55, 82, 1968 (No. 2). A

J. E. Ralph & M. G. Townsend: Near-infrared fluores-cence and absorption spectra ofC02+ and Ni2+ in MgO.J. chem. Phys. 48, 149-154, 1968 (No. I). E, M

P. Reijnen: Phase equilibria in the system MgO-FeO-Fe203.Philips Res. Repts. 23, 151-188, 1968 (No. 2). E

E. Roeder & J. Hornstra: Grain growth during hot-pressing of tantalum carbide.J. Amer. Ceramic Soc. 51, 224-225, 1968 (No. 4). A, E

D. A. Schreuder (Philips Lighting Division, Eindhoven):Road tunnels in the Netherlands.Light and Lighting 60, 370-377, 1967 (No. 12).

L. A. JE. Sluyterman: The rate-limiting reaction inpapain action as derived from the reaction of theenzyme with chloroacetic acid.Biochim. biophys. Acta 151, 178-187, 1968 (No. I). E

M. J. Sparnaay: Physisorption on heterogeneous sur-faces.Surface Sci. 9, 100-118, 1968 (No. I). E

I. Teramoto & A. M. J. G. van Run: The existenceregion and the magnetic and electrical properties ofMnSb.J. Phys. Chem. Solids 29, 347-355, 1968 (No. 2). E

D. R. Tilley: Influence of planar defects on supercon-ducting surface nucleation fields.J. Physics C (Proc. Phys. Soc.), ser. 2, 1, 293-298, 1968(No.2). M

M. G. Townsend: Visible charge transfer band in bluesapphire.Solid State Comm. 6, 81-83, 1968 (No. 2). M

M. G. Townsend & W. A. Crossley: Magnetic suscepti-bility of rare-earth compounds with the pyrochlorestructure.J. Phys. Chem. Solids 29.593-598,1968 (No. 4). E, M

B. Tuck: Conduction band density of states in impureGaAs.J. Phys. Chem. Solids 29, 615-622, 1968 (No. 4). M

A. A. Turnbull & G. B. Evans: Photoemission fromGaAs-Cs-O.Brit. J. appl. Phys. (J. Physics D), ser. 2, 1, 155-160,1968 (No. 2). M

W. A. J. J. Velge & K. H. J. Buschow: Magnetic andcrystallographic properties of some rare earth cobaltcompounds with CaZn5 structure.J. appl. Phys. 39, 1717-1720, 1968 (No. 3). E

M. L. Verheijke: Efficiencies of a coincidence spectrom-eter for positron annihilation radiation. Addendum.Nucl. Instr. Meth. 58, 347-348, 1968 (No. 2). E

Q. H. F. Vrehen: Interband magneto-optical absorp-tion in gallium arsenide.J. Phys. Chem. Solids 29, 129-141, 1968 (No. I). E

H. W. Werner & H. A. M. de Grefte: Ein Massenspek-trometer zur Untersuchung Dünner Schichten.Vakuum-Technik 17, 37-41, 1968 (No. 2). E

J. S. C. Wessels: Isolation and properties of two digito-nin-soluble pigment-protein complexes from spinach.Biochim. biophys. Acta 153,497-500, 1968 (No. 2). E


J. S. van Wieringen: Note of Mössbauer fraction inpowders of small particles.Physics Letters 26A, 370-371, 1968 (No. 8). E

W. J. Witteman & H. W. Wemer: The effect of watervapour and hydrogen on the gas composition of asealed-off C02 laser.Physics Letters 26A, 454-455, 1968 (No. 10). E

P. Wodon: Quelques aspects du calcul non numériquesur ordinateurs.Rev. MBLE 10, 59-67, 1967 (No. 2). B

G. Zanmarchi & C. Haas: Magnon drag at optical fre-quencies and the infrared spectrum of MnTe.J. appl. Phys. 39, 596-597, 1968 (No. 2, Part I). E

Contents ofPhilips Telecommunication Review 27, No. 4, 1968:

J. Larcher & J. Noordanus: Frequency synthesizers for radio equipment (p. 149-166).

Contents of Electronic Applications 28, No. 1, 1968:

B. J. M:Overgoor: Impedance matching network for capacitor microphones (p. 1-5).J. Tuil: Intermodulation in aerial amplifiers (p. 6-21).A. Cense: Recent developments in circuits and transistors for television receivers, VI. Clamping circuit forcolour difference signals (p. 22-28).H. Schmidt: Thermal feedback in integrated circuits (p. 29-39).

Contents of Mullard Technical Communications 10, No. 92, 1968:

J. Lavallee: Voltage regulation of lower-powered alternators (p. 34-39).J. Merrett: Voltage transient and dV/ilt suppression in thyristor bridges (p. 40-51).B. J. M. Overgoor: Junction field effect transistors: their structure and operation (p. 52-56).D. J. G. Janssen: Circuit logic with silicon controlled switches (p. 57-64).D. Humphries & J. Lavallee: Thyristor control of fan cooling and heating systems (p. 65-72).

Volume 29, 1968, No. 8/9 pages 237-292 Published 18th October 1968

Contents ofValvo Berichte 14, No. 1, 1968:

U. J. Pittack: Hochleistungs-Impulsklystrons in Linearbeschleunigern (p. 1-25).E. Ginsberg: Entwurf monolithischer integrierter Schaltungen. Schaltungselemente der integrierten Technik,Grundregeln für die Ausführung integrierter Schaltungen (p. 26-36).

design and characteristics of present-day photomultipliers bound... · design and characteristics...

Documents