*Corresponding author. Tel.: #972-8-934-2044; fax: #972-8-934-2611.

E-mail address: fnsingh@wis.weizmann.ac.il (B.K. Singh).

CsBr and CsI UV photocathodes: new results on quantume$ciency and aging

B.K. Singh*, E. Shefer, A. Breskin, R. Chechik, N. Avraham

Department of Particle Physics, The Weizmann Institute of Science, 76 100 Rehovot, Israel

Abstract

We report on the photoemission properties of 300 As thick transmissive- and 5000 As thick re#ective UV-sensitive CsBrphotocathodes. Following post-evaporation heat treatment at 703C the absolute quantum e$ciency is 35% at 150 nm,with a red boundary cut-o! at about 195 nm. Extensive aging studies of CsBr and CsI photocathodes, under high photon#ux and under ion bombardment in gas avalanche multiplication mode, were carried out for the "rst time withoutexposure to air. The results are compared with the previously published data on CsI aging and the methodology of theaging tests is discussed in details.

1. Introduction

In recent years we have seen considerable activityin the "eld of photon imaging detectors, combiningsolid photocathodes and gaseous electron multi-pliers [1]. Such devices, which are sensitive to singlephotons, can reach dimensions of a square meter [2]and can operate at very high magnetic "elds [3].They may have applications in various "elds, as forexample, the readout of large arrays of scintillatorsand scintillating "bers, as well as of gas scintillatorsin medical imaging and space instrumentation. Inparticular, they are applied in particle and astro-

particle physics, for particle identi"cation by theRing Imaging Cherenkov (RICH) technique [4].

In the UV spectral range, CsI photocathodes (seereview [5]), currently employed in vacuum- andgas-operated imaging detectors, have the best-known quantum e$ciency (40% at 150 nm) andrelatively good stability for short exposure to air.Their high quantum response is due to the goodelectron transport and emission properties, typicalof alkali halides. CVD diamond "lms have alsobeen investigated recently, showing rather goodphotoemission properties (12% at 140 nm), but ata much more restricted spectral range [6]. CsBr,though known for a few decades as a solar-blindphotocathode [7], has not received much attention,due to its much poorer photoemission yield com-pared to that of CsI. Recently, there has been a re-newed interest in CsBr as a protective coating "lm

Fig. 1. A schematic view of the experimental set-up. The photocathodes are deposited (deposition site) and investigated (measurementsite) in the vacuum chamber without exposure to air. The photocathode history is followed by continuously monitoring thephotocurrent and also by measuring the absolute QE at time intervals.

for alkali}antimonide photocathodes [8]. Thecoating of such photocathodes with a few hundredangstrom thick CsI or CsBr "lms was found toprotect them against exposure to oxygen, pavingthe way towards their use in gas avalanche de-tectors for visible light [9,10].

In this work, we have investigated the prepara-tion method and photoemission properties oftransmissive (300 As thick) and re#ective (5000 Asthick) CsBr photocathodes. Post-evaporation heattreatment as well as the e!ect of exposure to ambi-ent air were studied.

Particular emphasis was placed in this work onthe aging of CsBr "lms under the impact of photonsand ions. To date, there has not been a clear under-standing of aging processes of photocathodes.Although a large collection of data exists in theliterature, mainly for CsI "lms, many discrepanciesare seen among the measurements by di!erentinvestigators. We believe that this is due to ourignorance about the exact nature of the agingmechanism and the relevant experimental para-meters that should be followed during the agingtest. Some possible explanations and a compilationof existing data are reviewed in Ref. [5] and dis-cussed in more detail below. We present here the

results of systematic aging studies of CsBr photo-cathodes under photon and avalanche-ion bom-bardment, and compare them with new data of CsIphotocathodes aging carried out exactly under thesame conditions. The current tests were carried outfor the "rst time without exposing the "lms to airand include the recording of relative photocurrentas well as the absolute photoyield. In the following,we discuss the importance of the aging study meth-odology.

2. Experimental technique

The experimental set-up includes a high-vacuumevaporation chamber (pumped with a cryogenicpump to 10~7 Torr), coupled via a CaF

2window

to a vacuum ultra violet (VUV) monochromator(MC), equipped with a 30 W D

2lamp (see Fig. 1).

An important feature of the set-up is that after thesample preparation, its quantum e$ciency (QE)can be measured in situ, by rotating and displacingit from the evaporation site to the measurementlocation. The QE was measured in a re#ectivemode, under vacuum or in CH

4, 99.999% pure.

A positive voltage was applied to a mesh electrode

Fig. 2. Post evaporation heat enhancement of the photoyield ofCsBr photocathodes. (a) Absolute QE of 300 As CsBr, measuredas deposited and after 2 and 4 h at 703C. (b) Absolute QE of5000 As CsBr measured as deposited and after 3 and 6 h at 703C.The data of Taft and Philipp [7] is also shown.

placed at a distance of 3.5 mm from the photo-cathode surface; the photocurrent, induced bymonochromatic UV photons, was recorded fromthe photocathode. In all QE measurements, thephoton #ux at the lamp's peak (160 nm) was of theorder of 8]109 ph/s on a cathode surface of about103mm2. The absolute QE value is derived from theratio of the current measured from the photo-cathode to the current from a calibratedHamamatsu 1460R photomultiplier (Cal. PMT).This was done by alternatively directing the UVbeam to both, by a rotating mirror. This PMT wascalibrated against a NIST vacuum-photodiode[11] and was operated in a photodiode collectionmode (gain 1) with #80 V on its "rst dynode. Thestability of the MC D

2lamp was monitored

throughout the experiments by a second referencephotomultiplier (Ref. PMT), of the same type, andthe measured photocurrent values were correctedcorrespondingly. The photocathode materials, CsIand CsBr (ALFA, ultra-pure quality) were depos-ited from molybdenum and tantalum evaporationboats, respectively, onto a stainless-steel (SS) sub-strate placed 250 mm above the evaporation boats.The alkali}halide powders were melted and out-gassed under a shutter prior to evaporation. Thephotocathode "lms (300 and 5000 As CsBr and CsI)were deposited on the polished stainless-steel sub-strates, pre-coated with a 1000 As thick Al layer.The CsBr evaporation rate was of 0.2}10 As /s and20}70 As /s for the 300 and 5000 As thick "lms, re-spectively. After the measurement of the QE of an`as-evaporateda photocathode, it was heated undervacuum, with the help of a built-in water heatexchanger, to 703C, for a few hours. For photon-aging studies, carried out under vacuum and undergas #ow (CH

4at 1 atm pressure), we used another

30 W D2

lamp, mounted at the top of the mirrorbox (with the mirror removed), directly illumina-ting the photocathode at high photon #ux (seeFig. 1). For aging studies under positive ion bom-bardment (under gas avalanche), we used the MClight at 160 nm wavelength. In these measurements,the UV-induced photoelectrons emitted from thephotocathode were multiplied in a parallel-plate(PP) avalanche mode by applying a high voltage tothe mesh (placed here at a reduced distance of1.5 mm above the photocathode). The absolute QE

spectra were measured occasionally without multi-plication, either under vacuum or in gas. In ion-aging studies, the chamber was operating in #owmode, with 50 Torr of methane. During themeasurements the gas pressure and the temper-ature variations were carefully monitored.

3. Results

3.1. Quantum ezciency

Fig. 2 shows the absolute QE as a function ofwavelength for 300 and 5000 As thick CsBr photo-cathodes. The QE was "rst measured immediatelyafter evaporation (referred to `as-evaporateda in

Fig. 3. The decay of the QE of 5000 As CsBr under exposure toair inside and outside the chamber.

Fig. 2). Thereafter the photocathode temperaturewas gradually raised to 703C, resulting in an en-hancement of the QE values, which saturated after4 and 6 h for thin and thick photocathodes, respec-tively. For both "lm thicknesses we observea strong e!ect of heat treatment, which is larger atlonger wavelength. At 160 nm the QE is almostdoubled for both thin and thick "lms, typicallyreaching values of 15 and 17%, respectively. A shiftin the photoemission threshold (red boundary) byabout 0.7 eV can be observed in Fig. 2. The CsBrQE spectrum measured by Taft and Philipp [7] isalso shown in the "gure and is very similar to our`as-evaporateda data. Our best QE data for heat-enhanced CsBr "lms are shown in Fig. 3, reachingabout 27% at 160 nm.

3.2. Exposure to air

Moisture plays an important role in the decay ofthe photoemission properties of hygroscopicphotocathode materials. This e!ect has been inves-tigated in detail for uncoated [12}16] and coated[17,18] CsI photocathodes. It was lately shownthat the morphology of thin CsI and CsBr "lms isstrongly a!ected by humidity, which could explainpart of their decay [19]. We present here newresults on the decay in photoemission properties ofCsBr photocathodes. Most of the measurementswere performed inside the evaporation chamber, at

room temperature. Air was introduced into thechamber, for di!erent time periods, followed bypumping and QE measurement in vacuum. Abso-lute QE spectra for a 5000 As thick CsBr photo-cathode are shown in Fig. 3. It was observed thata short-term exposure (20 min) to air inside thechamber does not deteriorate the QE value; on thecontrary, the QE slightly improves at longwavelengths. However, a 25% drop in QE wasobserved after 9 h of exposure to humid air insidethe chamber. The exposure of the same photo-cathode for 40 min outside the chamber, resulted ina further 30% decay in QE. The slower decaywithin the evaporation chamber, also observed forCsI [20], is most probably due to a getter e!ect ofCsBr deposited on the chamber walls. The slightincrease in CsBr QE after a short exposure to airmay be due to an adsorption of a monolayer ofwater molecules, which by their electric dipolereduce the electron a$nity, as discussed in detail inRefs. [18,21].

3.3. Aging

Aging studies, which are very time consuming,are always carried out under accelerated condi-tions, namely incident #ux of photons or ionswhich is by several orders of magnitude larger thanthe realistic one. Therefore, some additional pro-cesses may be provoked, such as charging-up ornon-linear e!ects, which could distort the results.The conclusion on the photocathode aging, basedsolely on linear extrapolation of #ux, should there-fore be carefully examined.

Moreover, in the present work we have followedthe decay of the photocathode by measurements ofphotocurrent and of absolute photoyield in vac-uum or in gas. In part of the cases, there is noagreement between the results obtained from thetwo types of measurements. We have full con"-dence in the absolute QE results, always measuredunder the same controlled conditions. The oddbehavior of the photocurrent measurements will bediscussed at length.

In the present aging study, we used di!erentaging conditions, which may be classi"ed into fourdi!erent phases, as de"ned in Table 1. Some moretechnical details are also given in Table 2.

Table 1De"nition of photon- and ion-aging studies conditions

Phase/type Photon #ux(Photon/mm2 s)

Gas/pressure Gain Photocathode

I Photon aging in vacuum &1010 Vac, 10~7 Torr * Thick and thin CsBrII Photon aging in gas &1010 CH

4, 1 atm 1 Thick and thin CsBr

III Ion aging at high gain &106 CH4, 50 Torr 104 Thick and thin CsBr/CsI

IV Ion aging at low gain &106 CH4, 50 Torr 103 Thick and thin CsBr/CsI

Table 2Summary of photon- and ion-aging studies of CsBr and CsI photocathodes. Given are the experimental conditions (see Table 1), the Fig.no. showing the data and the accumulated charge corresponding to twenty-percent loss (TPL) of QE. We also remark on consistencybetween photocurrent and QE measurements. Phases I and II are photon-aging; phases III and IV are ion-aging. The photon #ux foreach data point was extracted from the recorded photocurrent and the measured absolute QE

Photocathode /thickness (As ) Phase Flux TPL of QE(lC/mm2)

Photocurrent vs.QE consistency

Fig. no.

(photon/mm2 s) (pA/mm2)

CsBr/300 As I 6]1010 861 7 Yes 4CsBr/5000 As I 5]1010 685 14 Yes 4CsBr/300 As II 3]1010 334 '230 No 5CsBr/5000 As II 4]1010 795 '130 No 5CsBr/300 As III 6.5]106 1020 171(14% loss) No 7CsBr/300 As IV 1.5]106 21 1.5 Yes 7CsBr/5000 As III 1.7]106 300 Stable No 8CsBr/5000 As IV 1.4]106 30 13 Yes 8

CsI/300 As III 1.3]106 340 34 Yes 9CsI/300 As IV 2.8]106 37 14 Yes 9CsI/5000 As III 1.2]106 330 43 Yes 10CsI/5000 As IV 1.1]106 31 26 (15% loss) Partial 11

3.3.1. Photon agingPhoton-aging studies were "rst carried out in

Phase I, with intense photon #ux under vacuum(10~7 Torr). Fig. 4a shows the relative photocur-rent as a function of total accumulated charge forboth 300 and 5000 As thick CsBr photocathodes;the respective initial photon #ux were 6]1010 and5]1010 photon/mm2 s. In Fig. 4a we observea similar initial decay rate for both "lms and thena faster QE decay for the thinner "lm. The decayrate seems to decrease with increasing accumulatedcharge, for both. A twenty-percent loss (TPL) inrelative photocurrent was observed after 7 and14 lC/mm2 for the 300 and 5000 As thick photo-cathodes, respectively. Following the above aging

process we re-measured the QE spectra of the agedphotocathodes in situ. In Fig. 4b we display theabsolute QE spectra measured for the 300 As thickCsBr photocathode before and at the end of thephoton aging. The drop in QE by 58% at 160 nm isconsistent with the drop of &57% seen for thephotocurrent measurements in Fig. 4a. Similarly, inFig. 4c we show the absolute QE spectra evolutionof a 5000 As thick CsBr photocathode. A QE loss of54% is observed at 160 nm, which is again consis-tent with the relative 55% drop of the photocurrentin Fig. 4a.

A di!erent behavior was observed in Phase IIphoton aging of thick and thin CsBr photocathode,in 1 atm of methane. The aging of 5000 As CsBr

Fig. 4. Photon-induced aging of CsBr photocathodes in vac-uum (Phase I). (a) Relative photocurrent as a function of accu-mulated charge for 300 and 5000 As CsBr photocathodes. (b) Theabsolute QE spectrum of a 300 As CsBr, measured in vacuumbefore and at the end of the photon-induced aging test. (c) Sameas (b) for 5000 As CsBr.

Fig. 5. Photon-induced aging of CsBr photocathode in 1 atmCH

4, gain"1 (Phase II). (a) Relative photocurrent as a function

of accumulated charge for 5000 As CsBr. (b) The absolute QEspectra measured in vacuum before and at the end of the agingtest. (c) Same as (b) for 300 As CsBr.

Fig. 6. The total gain/voltage curve at 50 Torr CH4

of the PPcomprising a polished SS substrate cathode located 1.5 mmbelow the mesh anode.

photocathode was carried out under an incidentphoton #ux of 4]1010 photon/mm2 s (Fig. 5a).For this large #ux, a decrease of 10% in thephotocurrent was initially observed, followed bya plateau up to an accumulated charge of about50 lC/mm2 and a further slow decrease up to127 lC/mm2. After this aging process we againre-measured the absolute QE of the photocathodein vacuum (Fig. 5b), observing no degradation ofthe QE at 160 nm. Similarly, we studied the photonaging of 300 As CsBr photocathode at 1 atm pres-sure of methane with no charge multiplication.The incident photon #ux was about &3]1010 photon/mm2 s. We observed a steep rise in therelative photocurrent, to about 145% of the initialvalue, followed by a slow decrease to 140% ofthe initial value after accumulated charge of87 lC/mm2 (not shown). At this point, the absoluteQE was remeasured in gas as shown in Fig. 5c. Werecord no enhancement in QE, in contradiction withthe observed rise in photocurrent. We continued theaging of this photocathode up to an accumulatedcharge of &230 lC/mm2. The photocurrent "rstdropped sharply and then stayed at about 130% ofthe initial value (not shown). Absolute QE measure-ments were again done at intermediate points ofaccumulated charges of 189 and 230 lC/mm2; onceagain no change in absolute QE was found.

3.3.2. Ion aging of CsBr and CsI photocathodesAs pointed out above, accelerated ion-induced

aging should be performed in order to obtain theresults over a reasonable time-scale. This may beachieved by carrying out the measurements underhigh #ux or high gain, leading to high count rate(although quite limited due to upcharging), or bychoosing operation conditions known to invokeaccelerated aging. The last occurs, for example,under a parallel plate (PP) avalanche mode, inwhich all the avalanche ions are back drifting andsputtering the photocathode at high velocity.Choosing, for example, low gas pressure and highelectric "eld (namely high gain) even further intensi-"es the photocathode damage. In the present work,we used PP avalanche mode in CH

4, at 50 Torr

under various chamber gains. Slower aging is there-fore expected at atmospheric pressure under di!er-ent multiplication geometries.

We "rst measured the absolute gain curve byrecording the UV-induced photocurrent from a SScathode substrate as a function of voltage (Fig. 6).Gains of 104 and 103 were reached at 620 and500 V, respectively, with 50 Torr of methane. Theaging studies of phases III and IV under gas multi-plication were carried out under these conditions,with the SS photocathode replaced by our CsBr orCsI photocathodes. The photocathode was irra-diated by UV light at 160 nm. In each aging test, werecorded continuously the current from the photo-cathode, and in addition we measured the full abso-lute QE spectrum at given time intervals.

3.3.2.1. CsBr. The results obtained with 300 Asthick CsBr photocathode under gas gain of 104(phase III) are shown in Fig. 7a. The incidentphoton #ux was 6.5]106 photon/mm2 s. We ob-serve a two-component decay; "rst the relativephotocurrent drops to 70% of its initial value afteran accumulated charge of 13 lC/mm2. Thereafter,it drops to &20% of its initial value after anaccumulated charge of 158 lC/mm2. At the begin-ning and at the end of this part of the aging processwe measured the absolute QE spectrum in vacuum(lines 1 and 2 in Fig. 7c). We found that the drop ofQE at 160 nm does not follow the 80% drop re-corded in the photocurrent and it is of only 14%.We de"ne the relative quantum e$ciency (RQE) asthe ratio of the absolute QE at 160 nm at any point

Fig. 7. Ion-induced aging of 300 As CsBr photocathode in50 Torr CH

4. (a) Relative photocurrent as a function of accumu-

lated charge at gains 104 and 103. RQE values at 160 nm are alsoshown. (b) An expanded view of the relative photocurrent decayand RQE values at 160 nm at gain"103. (c) The absolute QEspectra measured in vacuum before and at the end of the ion-induced aging tests at gains 104 (curves 1 and 2) and 103 (curves2 and 3).

in time to its value at the beginning of the agingprocess. The RQE values at points 1 and 2 areas well shown in Figs. 7a and b, as open circles1 and 2.

The aging measurements of the same photo-cathode were then continued at a reduced gain of103 (phase IV), in 50 Torr of CH

4. The incident

photon #ux was 1.5]106 photon/mm2 s. The re-sults are shown in Fig. 7a (above 171 lC/mm2) andan expanded view of this part is given in Fig. 7b.A sharp drop of 50% in the relative photocurrent isseen after an additional accumulated charge of only4 lC/mm2. The absolute QE, measured under vac-uum after this aging step, is shown in Fig. 7c (dottedline marked 3) and the RQE as open circle 3 in Fig.7b; this time the results were consistent (&47%decay) with the relative drop of photocurrent ata gain of 103.

The same aging test was done on a 5000 As thickCsBr "lm, in a reverse order. The results are shownin Fig. 8a. At a gain of 103 (phase IV) the incidentphoton #ux was 1.4]106photon/mm2 s. The rela-tive photocurrent shows an initial increase of about12% of its initial value, followed by a drop to 83%after an accumulated charge of 13 lC/mm2. Theabsolute QE values (Fig. 8b) were measured thistime in gas (yielding slightly lower absolute valuesas compared to vacuum, due to the backscatteringe!ect [22]) along the phase IV aging step; the RQE(open circles 1}4 in Fig. 8a) are well consistent withthe photocurrent measurement. Following thesemeasurements, the gas gain was raised to 104 andthe aging process was continued (phase III) on thesame thick photocathode, at an incident photon#ux of 1.7]106 photon/mm2 s. As shown in theFig. 8a we recorded a fast rise in the relative photo-current, followed by a gradual decrease, to about61% of its initial value, after an additional accumu-lated charge of 160 lC/mm2. The absolute QEspectra (Fig. 8c) show no increase or deteriorationwhatsoever, up to the total accumulated charge of174 lC/mm2. The RQE evolution (open circles 4}9in Fig. 8a), based on the absolute QE measure-ments, is very di!erent compared to that of thephotocurrent.

The strong inconsistency between photocurrentand absolute QE measurements under some con-ditions and their good agreement under other

Fig. 8. Ion-induced aging of 5000 As CsBr photocathode in50 Torr CH

4. (a) Relative photocurrent and RQE values at

160 nm as a function of accumulated charge at gains 103 and104. (b) The absolute QE spectra measured in vacuum and in gasbefore and after the aging at gain 103. (c) Same as (b) at gain 104.

Fig. 9. Ion-induced aging of 300 As CsI photocathode in 50 TorrCH

4and gains of 103 and 104 as a function of accumulated

charge. RQE values at 160 nm are also shown.

conditions is puzzling, requiring further studies, asdiscussed in Section 4.

3.3.2.2. CsI. In order to carry out a reliablecomparison between the aging processes of CsBrand CsI photocathodes, the aging measurementshave to be carried out exactly under the sameconditions and protocols. We cannot make use ofany of the published data on CsI, as in each casesome experimental parameters are di!erent than inthe present study. Therefore, we performed two setsof measurements, on 300 and 5000 As thick CsIphotocathodes, under identical conditions to thoseof CsBr. The results for a 300 As CsI photocathodeare shown in Fig. 9, for an initial gain of 104,followed by a gain of 103. At the gain of 104 theincident photon #ux was &1.3]106 photon/mm2 s. A smooth decrease in photocurrent wasobserved, up to an accumulated charge of83 lC/mm2, where it reached about 68% of itsinitial value. Further aging of the same aged photo-cathode, at a gain of 103 and under an incidentphoton #ux of 2.8]106 photon/mm2 s, resulted ina faster decay (Fig. 9). A relative loss of 25% in thephotocurrent was reached for an additional accu-mulated charge of 23 lC/mm2. The RQE evolvedin an excellent accordance with the photocurrentdata (Fig. 9).

Fig. 10a displays the variation of the relativephotocurrent for a 5000 As thick CsI photocathode,

Fig. 10. Ion-induced aging of 5000 As CsI photocathode in50 Torr CH

4. (a) Relative photocurrent and RQE values at

160 nm as a function of accumulated charge with gain"104. (b)The absolute QE spectra measured in gas at di!erent values ofaccumulated charge.

Fig. 11. Ion-induced aging of 5000 As CsI photocathode in50 Torr CH

4. (a) Relative photocurrent and RQE values at

160 nm as a function of accumulated charge at gain"103. (b)The absolute QE spectra measured in gas at di!erent values ofaccumulated charge.

aged at a gain of 104. The incident photon #ux was1.2]106 photon/mm2 s. We observed a fast decayto about 79% of its initial photocurrent value,followed by a gradual decrease to 57% after anaccumulated charge &109 lC/mm2. The drop inthe absolute QE was measured along the agingprocess (Fig. 10b); the RQE evolution at 160 nm,shown in Fig. 10a, is fully consistent with that of thephotocurrent.

Aging results of a second 5000 As thick CsIphotocathode, at a gain of 103, are shown inFig. 11a; the incident photon #ux was1.1]106 photon/mm2 s. We observed an initial fastdrop of the photocurrent, of almost 40%, followed

by a rise and then a slow decrease. Absolute QEmeasurements along the aging study do not repro-duce the variations in the photocurrent; RQE dataat 160 nm is shown in Fig. 11a; it indicates onlya 15% decay in QE, after an accumulated charge of26 lC/mm2.

4. Discussion

4.1. Quantum ezciency and heat enhancement

The QE of `as-evaporateda CsBr "lms is con-sistent with the published data of Taft and

Phillipp [7]. A post-evaporation heating of CsBrphotocathodes at about 703C in vacuum consider-ably enhances their QE, both for 300 and 5000 Asthick "lms. A similar enhancement was previouslyobserved for CsI, CuI and NaI [23]. The heatenhancement is a permanent modi"cation of theQE, measured at room temperature. It should bedistinguished from a di!erent observation, of anincrease in photocurrent when the photocathodetemperature is raised [24,25], which is a temporarye!ect. We are tempted to ascribe the QE enhance-ment to the removal of water from the surface, assome of the previously mentioned authors did, butfurther considerations may contradict this hypoth-esis. Firstly, it is known that e$cient removal ofwater from the surface usually requires temper-atures above 1003C and therefore we are indeed notsure that water is removed in the present process.Secondly, the same heat enhancement e!ect is ob-served on various materials with a very di!erenta$nity to water. Moreover, the opposite phenom-enon was demonstrated on hygroscopic "lms suchas LiF and NaF, whereby a monolayer of waterwas shown to enhance photoemission, due toa dipole moment created on the surface, e!ectivelyreducing the electron a$nity [18,21].

A di!erent explanation for the heat enhancementcould be due to bromine desorption during theheating process. This explanation follows the ap-proach presented in Ref. [23] for interpreting theheat enhancement observed in alkali}iodine photo-cathodes (CsI, CuI and NaI). It is known thatbromides decompose more easily than other hali-des; an example is the decomposition of Ag(Br) inphotoemulsion under exposure to visible light. Wemay assume that part of the CsBr decomposesduring the evaporation and, due to the higher vaporpressure of bromine compared to cesium, this couldresult in a bromine-enriched deposited "lm. The factsupporting this idea is that there exists a stablecompound of CsBr

3[26]. When the deposited "lm

is heated, the excess bromine is presumably released,which enhances the photoelectron emission prob-ability by improving the lattice structure of the "lm.The heat enhancement observed in this work hasa similar spectral dependence to the one observed inRef. [23] for CsI, CuI and NaI photocathodes. Inter-estingly, the enhancement in QE is accompanied by

an extension of the photocathode sensitivity to lon-ger wavelengths, which is consistent with the hy-pothesis of improved electron transport and reducedsurface barrier.

4.2. Exposure to air

The thick CsBr "lms were found to be ratherstable under short exposure to air. 20 min in airinside the chamber did not a!ect the QE; 9 h in airinside the chamber reduced the QE by about 25%;additional 40 min in air outside the chamber furtherreduced the QE by &30%. This behavior of slowerdecay inside the evaporation chamber is very similarto the one observed for CsI photocathodes [20]; it isprobably due to a `gettera e!ect resulting from al-kali}halide deposits on the chamber walls.

It has recently been demonstrated [19], usingscanning electron microscope (SEM) analysis, thatthe morphology of thick and thin CsBr "lms isdi!erent. It was shown that the grain size is smallerin thin "lms, although it is not clear whether this factin#uences the photoemission performance of the"lms. However, the authors of Ref. [19] show thata 750 As thick CsBr "lm has a continuous morpho-logy whereas a thin CsBr "lm of 200 As is discontinu-ous. The authors also demonstrated that, both thinand thick CsBr "lms, when exposed to humidityundergo a drastic morphological transformation,the small grains coalesce into large, separated grains.This is believed to cause the QE decay in air, by theconsiderable reduction in surface coverage.

4.3. Aging

The data from the present study of aging underphoton and avalanche-ion #ux is summarized inTable 2.

4.3.1. Photon agingAging of CsBr photocathodes in vacuum under

intense photon #ux (Phase 1) occurs faster for thin-ner "lms. Currently, we do not have an explanationfor this phenomenon. The drop in photocurrent,measured along the aging experiment, is consistentwith the drop in the QE, measured at intervalsalong the aging process. The aging of CsBr photo-cathodes under intense photon #ux in 1 atm CH

4

(phase 2) showed practically no degradation in QEup to accumulated charges of 230 and 130 lC/mm2

for 300 and 5000 As thick "lms, respectively. Thisobservation is in disagreement with the photocur-rent measurements during the same aging process,reading an increase to 186% and a drop to 60% ofthe initial values, for the thin and thick samples,respectively. These results are still not explained. Itis interesting to note that a similar trend, of pro-longed lifetime under gas as compared to vacuum,was observed by Anderson et al. [24] for CsIphotocathodes (see Table 3 and further discussionbelow).

4.3.2. Ion agingCsBr photocathodes aged under gas multiplica-

tion (50 Torr CH4) in PP geometry (Phases III and

IV) showed a faster decay for 300 As than for 5000 Asthick "lms. They also showed a large di!erence inthe QE decay rate for di!erent gas gains. A twenty-percent loss (TPL) of QE was found at 1.5 and13 lC/mm2 for thin and thick photocathodes, re-spectively, at a gain of 103, in agreement with thephotocurrent measurements. A TPL of QE wasfound at considerably larger accumulated chargevalues, 171 and ''174 lC/mm2, respectively forthin and thick photocathodes, at a gain of 104. Thisis in disagreement with the photocurrent measure-ments that showed a faster drop for the thinner "lmand an increase followed by a drop for the thickerone. This behavior is not quite understood. In prin-ciple, the disagreement between absolute QEmeasurements and photocurrent measurementscould be related to a possible change of the detectorgain (the photocurrent is a product of the QE andthe gain). However, unlike thin multiplying anodewires known to increase in diameter due to polym-erization under intense irradiation that results indecrease of gain, a PP geometry is known to bestable. In addition, the above disagreement doesnot persist, but appears only in some cases and notin others. We may argue that at a gain of 104 thechamber had some transient instabilities due todefects in the cathode or to local upcharging, in-ducing current #uctuations (Such #uctuations werealso reported by Va'vra et al. [27] and by Krizan(private communication)). This could presumablyexplain the discrepancy between true QE variations

and the measured photocurrent variations. Ata gain of 103 such instabilities probably do not existand hence the QE and current measurementsagreement. It is interesting that a thin "lm agesfaster than a thick one also under ion bombard-ment. The nature of the ion aging is however notclear. Obviously, the avalanche ions sputter thecathode surface and cause modi"cation of its sur-face. However, the faster decay under gain of 103compared to that at 104 seems to contradict ourintuition, because we would have expected largersputtering damage to the photocathode underhigher electric "eld (more energetic ions). The ex-perimental data may indicate that ion sputteringalso rejuvenates the aged photocathode surface,and under some conditions the rejuvenation ratecould exceed that of the aging. But unless thishypothesis is further investigated, for exampleby systematically varying the gas type andpressure and the gas}ions energy, we couldnot draw any real conclusion on this matter.Rejuvenation of CsI photocathodes was discussedby Anderson et al. [24].

The same ion-induced aging process was carriedout under identical conditions on CsI photo-cathodes of 300 and 5000 As thickness. For bothphotocathodes there seems to be a faster decay atthe lower gain although the di!erence is smallercompared to that discussed for CsBr. A good agree-ment exists between photocurrent and QEmeasurements at the higher gain but only a limitedagreement is seen at the lower gain.

We may compare our results with previouslypublished data on aging of CsI photocathodes,summarized in Table 3 (which is an extension ofTable 2 from Ref. [5]). We should note that unlikethe present measurements, which were made in situ,all studies quoted in this table were carried out onsamples that were exposed to air for a short timeprior to the aging process.

In Table 3 values of TPL of photocurrent forphoton aging of CsI in vacuum vary from 0.1 to8 lC/mm2 and in gas they are consistently larger.Anderson et al. [24] indeed measured both in vac-uum and in gas, "nding the same trend, similar toour observation on CsBr. Moreover, they founda dependency on the gas type and pressure. Theyinterpreted the photon-induced QE decay by the

Tab

le3

Asu

mm

ary

ofpho

ton

and

ion-indu

ced

agin

gst

udie

sof

CsI

photo

cath

odes

inva

cuum

and

gas

Aut

hor

Thic

knes

s(As)/

subst

rate

Exp

ose

dto

air

Flu

xD

etec

tor

Gas

,pre

ssure

Gai

nTP

Lof

photo

curr

ent

(lC

/mm

2)

QE

consist

ent/

rem

arks

(photo

n/

mm

2s)

(pA

/mm

2)

Dan

gend

orf

[14]

'20

00As

/Al

1m

in10

13(1

85nm

)2]

105

PP

CH

420

Torr

12

Notpro

vide

d10

112]

105

20Torr

100

4Sam

e10

112]

105

100

Torr

100

8Sam

e10

870

010

Torr

350

0.4

Sam

eA

nder

son

etal

.[2

4]50

00/A

l10

min

Unkn

ow

n(1

80nm

)&

70PP

vacu

um

1&

0.5

Notpro

vide

dSa

me

&70

C2H

620

Torr

14

Sam

eSa

me

&70

CH

420

Torr

1A

4Sam

eSa

me

&70

i-C

4H

1020

Torr

1A

4Sam

eSa

me

600

i-C

4H

1020

Torr

127

Sam

eSa

me

Unkno

wn

i-C

4H

1020

Torr

3.5]

104

A4

Sam

eLu

etal

.[2

5]50

00/A

lSh

ort

lyU

nkn

ow

n(19

5nm

)U

nkno

wn

MW

Vac

uum

10.

1(tw

oco

mpo

nen

ts)

RQ

E,

consist

ent

Sam

eC

2H

620

Tor

r30

013

(190

nm

)R

QE

,co

nsist

ent

Unkn

ow

n(18

0nm

)C

2H

620

Torr

104

30(1

90nm

)tw

oco

mpo

nen

tsO

nly

RQ

E

Sam

eC

2H

620

Torr

105

15(1

90nm

)tw

oco

mpo

nen

tsSa

me

Kriza

net

al.[

13]

5000

/Cu

Yes

105(

180

nm)

200

MW

CH

41

atm

105

100

RQ

E,

consist

ent

5000

/Cu#

Sn/P

bSa

me

200

CH

41

atm

105

20Sa

me

9000

/Cu#

Sn/P

bSa

me

200

CH

41

atm

105

100

Sam

eV

a'vr

aet

al.[2

7]50

00/S

S#

Al

2}5

min

1.2]

1010

(185

nm)

300

MW

CH

41

atm

120

Notpro

vide

d50

00/S

S#

Al

104(

185n

m)

16C

H4

1at

m10

51

Abs

.QE,

consist

ent

5000

/Cu#

Sn/P

bSa

me

CH

41

atm

105

7Sam

e50

00/C

u#N

i/A

uSa

me

CH

41

atm

105

90Sa

me

Rab

uset

al.[2

8]50

00/S

S#

RSG

!Y

es5]

1012

(150

nm

)30

0PP

Vac

uum

18

Only

abs.

QE

5000

/Cu#

Ni/A

u#

RSG

!

Sam

eV

acuum

18

Sam

e

This

wor

k50

00/S

S#

Al

No

1.2]

106(

160

nm)

330

PP

CH

4,50

Torr

104

43A

bs.Q

E,

consist

ent

Sam

e31

CH

4,50

Torr

103

26(1

5%lo

ss)

Abs

.QE,

par

tial

cons.

!RSG

"re

sin-

stab

ilize

dgr

aphi

te.

photolysis process, whereby neutral Iodine atomsare created and evaporated from the surface, leav-ing it rich with Cs. This process, proposed byDangendorf et al. [14] and studied with X-ray-induced spectroscopy, was not entirely con"rmed.Anderson et al. [24] provide indirect evidence forthis process by the recovery of the QE after heating,which presumably evaporates the excess of Cs andreturns the surface to its original composition.Under gaseous environment, this process is con-siderably slowed down, but Anderson et al. donot provide an explanation. We may assume thatthe gas molecules hitting the surface supply themissing negative charge and thus prevent theIodine evaporation process. Alternatively, wemay speculate that the hydrocarbonic gas is poly-merized on the surface and prevents the evapor-ation of the volatile species. Such a hypothesiscould be checked by investigating the aging underother, `non-aginga gases, such as Ar/CO

2. We be-

lieve that the same mechanism, of surface modi"ca-tion by evaporation of one atomic species, could beassumed in the CsBr case. Similar to our data onCsBr, Lu et al. [25] observed a two-componentdecay, which they attribute to surface (fast) andbulk (slow) aging phases, however, they do not pro-vide direct evidence for this hypothesis. Theseauthors have also observed, in similarity to our dataon CsBr, a faster decay at longer wavelengths.Finally, we note that there also seems to be a de-pendence of the decay rate on the substrate material.

Previous data on Ion-induced aging of CsI undergas multiplication of 105 also shows a large spreadof TPL photocurrent values, from 1 to 100 lC/mm2,which depend on gas type and on sample prepara-tion details. Some authors (Va'vra [27] and Krizan(private communication)) report on a #uctuating(by up to a factor 1.4) behavior of the photocurrentduring measurements, and Anderson et al. [24]consistently observe an initial increase of photocur-rent by 5}10% before observing a decrease. Thesample preparation (substrate) seems to be a veryimportant factor in this aging process. The com-parison between photocurrent measurements andthat of absolute (or relative) QE is not complete,but seems to be quite consistent whenever exists.Varying decay slopes were observed by Krizan [13]and Lu [25], and a more complicated behavior

(initial increase) is indicated in the data of Ander-son et al. [24].

We may conclude that although none of the pre-viously published data on CsI can be exactly com-pared to our results on CsI or CsBr, as they di!er byvarious experimental parameters (vacuum/gas-typepressure, spectral range of the aging light, light #uxand current density, exposure to air prior to theaging study, photocathode deposition technique, thesubstrate material and its preparation, etc.), somegross similarities do exist. The photon aging and theion-induced aging results show a large spread ofdecay rates, which can be expected in view of thedi!erent experimental conditions.

In addition, it is not even clear which are therelevant parameters that should be controlled inorder to have a reproducible set of results. Anexample is the sample temperature or the photon#ux, which seem to be very relevant to the agingprocess [24] but were not controlled in the sameway and even sometimes disregarded by the di!erentauthors. Another example is gas impurities (e.g.water and oxygen), that might cause a decay of thephotocathode by chemically reacting with its surfaceand thus modifying the electron escape probability[13]. The large spread in the results may also berelated to the fact that most of the studies, which arevery time consuming, including the present one, (andexcept those of Anderson et al. [24]) were carriedout on a single sample per experimental condition.The sample-to-sample variations are unknown,which makes it impossible to accurately assess anylevel of agreement between di!erent sets of data.

Finally, we would like to point out that themechanism, or rather mechanisms, responsible forthe photocathode aging are not yet fully under-stood. The aging probably occurs due to an accu-mulation of several processes, occurring on thesurface and in the subsurface layers, which changethe electronic structure of the material. There arespeculations about the role of charge transportthrough the layer: according to one hypothesis, inmaterials of ionic lattice such as CsI and CsBr itmay lead to displacement of ions, i.e. build-up oflattice defects. According to another assumptionmentioned above, neutralized species (I, Br) arecreated and are evaporated from the surface,thus modifying the surface stoichiometry and the

electron a$nity. Some discussion of the chargetransport and its consequences was given by Va'vraet al. [27]. We observe a large di!erence in the TPLvalues under gains of 103 and 104 in the case ofCsBr, but a much smaller di!erence for CsI underthe same conditions. This could indeed be relatedto their di!erent electrical properties } higher resis-tivity of CsBr compared to CsI. The substrate ma-terial and its preparation certainly play a role in theaging process, but this parameter too is not quiteclear. We refer the reader to Refs. [12,19] for fur-ther information. The role of the gas in slowingdown the photocathode decay was mentionedabove and the reader may refer to Anderson et al.[24] for more information.

5. Summary

We have presented new data on the absolute QEof semitransparent and re#ective CsBr photo-cathodes. Following a post-evaporation heat treat-ment, these photocathodes reach 35% QE at150 nm. They may be attractive for some applica-tions as solar-blind photocathodes.

We have presented a new systematic study of theevolution of the QE of CsBr and CsI photo-cathodes, under photon and avalanche-ion bom-bardment, carried out for the "rst time in situ. Inthese aging tests we have measured the photo-current in a continuous way, and the absoluteQE at several time intervals. It is quite clearfrom our CsBr studies that the photocathode decaydeduced from photocurrent measurements is notnecessarily equivalent to that obtained by measur-ing the absolute QE. The "rst is easier to perform,but, in our opinion, it has to be backed by at leasta relative QE measurement, in vacuum or in gas,under well-controlled conditions. We have dis-cussed in detail the various phenomena observed inaging studies in this work as well as in previousworks on CsI, and indicated some possible physicalprocesses responsible for the experimental observa-tions. However, the whole aging process is still notunderstood and requires further studies, preferablydone in combination with surface analysis that willpoint more clearly at the nature of the materialstructure and surface modi"cation.

Acknowledgements

The work was partly supported by the IsraelScience Foundation. A. Breskin is the W.P. ReutherProfessor of Research in the peaceful uses of atomicenergy.

References

[1] A. Breskin et al., Nucl. Instr. and Meth. A 442 (2000) 58,and references therein.

[2] A. Di Mauro et al., Nucl. Instr. and Meth. A 433 (1999)190.

[3] F. Piuz et al., Nucl. Instr. and Meth. A 433 (1999) 178.[4] J. Seguinot, T. Ypsilantis, Nucl. Instr. and Meth. A 343

(1994) 1, and references therein.[5] A. Breskin, Nucl. Instr. and Meth. A 371 (1996) 116, and

references therein.[6] A. Laikhtman et al., Appl. Phys. Lett. 84 (1998) 1443.[7] E.A. Taft, H.R. Philipp, J. Phys. Chem. Solids 3 (1957) 1.[8] A. Buzulutskov et al., Nucl. Instr. and Meth. A 400 (1997)

173.[9] E. Shefer et al., Nucl. Instr. and Meth. A 419 (1998) 612,

and references therein.[10] E. Shefer et al., Nucl. Instr. and Meth. A 433 (1999) 502.[11] A. Breskin et al., Nucl. Instr. and Meth. A 343 (1994) 159.[12] J. Almeida et al., Nucl. Instr. and Meth. A 367 (1995) 337.[13] P. Krizan et al., IJS Report, IJS-DP-7087, October 3 1994.[14] V. Dangendorf et al., Nucl. Instr. and Meth. A 308 (1991)

519.[15] G.W. Fraser et al., ESA SP-356 (1992) 97.[16] C. Lu, K.T. McDonald, Nucl. Instr. and Meth. A 343

(1994) 135.[17] A. Buzulutskov et al., J. App. phys. 77 (1995) 2138.[18] A. Buzulutskov, A. Breskin, R. Chechik, Nucl. Instr. and

Meth. A 372 (1996) 572.[19] T. Boutboul et al., Nucl. Instr. and Meth. A 438 (1999) 409.[20] A. Buzulutskov et al., Nucl. Instr. and Meth. A 371 (1996)

147.[21] A. Buzulutskov, A. Breskin, R. Chechik, J. Appl. Phys. 81

(1997) 466.[22] A. Di Mauro et al., Nucl. Instr. and Meth. A 371 (1995) 137.[23] A. Buzulutskov et al., Nucl. Instr. and Meth. A 366 (1995)

410.[24] D.F. Anderson et al., Nucl. Instr. and Meth. A 323 (1992)

626.[25] C. Lu et al., Nucl. Instr. and Meth. A 366 (1995) 60.[26] Handbook of Chemistry and Physics, R.C. Weast (Ed.),

Chemical Rubber Co., Cleveland, 1972, p. 13}81.[27] J. Va'vra, A. Breskin, A. Buzulutskov, R. Chechik,

E. Shefer, Nucl. Instr. and Meth. A 387 (1997) 154.[28] H. Rabus, U. Kroth, M. Richter, G. Ulm, J. Friese,

R. Gernhauser, A. Kastenmuller, P. Maier-Komor, K.Zeitelhack, Nucl. Instr. and Meth. A 438 (1999) 94.