IEEE Transactions on Nuclear Science, Vol. NS-29, No. 3, June 1982

PRESENT PERFORMANCES AND POTENTIAL APPLICATIONSOF MIERCUIzRC IODIDE DETECTORS

Asher HolzerKevex Corporation1101 Chess Drive

Foster City, CA 94404

Summary

The need for room temperature detectors capableof high detec-iiorn efficiency for x-ray and y-ray spec-troscopy has long been recognized. This has resultedin substantial work on potentially suitable high-Zsemiconductors . Detectors imade from mercuric io-dide have proved to be the most attractive candidates,particularly for low-energy x-ray applications. Suchdetectors are now commercially available.

Mercuric iodide exhibits the wide-band gap re-quired for room temperature operation and the highatomic number (Z) required for efficient photons de-

tection . However, in common with other II-VI com-pound semiconductors, poor hole collection resultsfrom deep hole traps. Fortunately, for low-energyx-rays incident on the negatively-biased contact, thesignal is dominated by electron collection. Thereforein this case, very good energy resolution can some-times be obtained, opening up the possibility of wide-ranging applications for x-ray fluorescenice spectro-scopy. Good resolution, long-term stability, and alack of polarization effects all represent the objec-tives of work on these detectors; these factors andtheir control will be discussed here. The performance

9of recent detectors has been discussed elsewhereThe paper will focus on three aspects of develop-

isent of HgI2 detectors:i) Methods of purification and crystal growing.

Several methods are used and these will be discussed.ii) Fabrication of detectors.

iii) Applications and performance of detectors.

Mercuric iodide detectors are described here bothfrom the point of view of fabrication methods and fromthe user's point of view. The paper highlights theperformance achieved at different photon energies.Through a short background about the various ways thatHgI2 detectors are being fabricated, one can get anidea about the basic advantages and limitations ofHgI2 detectors.

There are several methods to purify the raw mate-rial,to grow crystals and to fabricate detectors.Each of them has its own advantages and limitations inregard to the performances at low energies versus highy energies, the geometry of the detectors, dead layer,stability, etc. This paper presents these methodsand emphasizes what is the most attractive method fordifferent applications.

Crystal Growth

Raw material of 99 percent purity is readilyavailable from commercial sources. Repeated sublimi-

10 11nation is used to purify this raw material ' . Thisis done by heating the HgM2 to the sublimination pointin a continuously pumped quartz ampoule so that impu-rities wnich have higher vapor pressure than HgI2 arepumped away while the HgI2 is left behind. Repeatingthe same process without pumping allows the HgI2 tosublimate and recrystallize at a cool part of the sys-tem while leaving behind impurities with lower vaporpressure. This is done at several tempera-tures toproduce good selection of the HgI2 from most othercommon impurities such as HgI, Hg2I2, HgCl, HgCH2HgI,

and others1l. Polymers recently have been placed inthe purification sy-stem to get even better selection.If this process is tuned correctly, very pure rawmaterial results while preserving good stoichiometry(i.e., the ratio of Hg and I a-toms). Other purifica-tion techniques cmay produce partial dissociation ofHgI2 causing the escape of iodine. Consequently, ad-ditional iodine must be added; since it is almost im-possible to determine the amount of iodine lost andto control the amount of iodine which is added, it isvery difficult to achieve good stoichiometry by thismethod. Such material produces a low yield of gooddetectors.

HgI2 crystals are grown from the vapor phase us-

ing three different techniques. The first method13which was developed by Scholz and improved by E. G.

G. Santa Barbara, produces single crystals as big as15cm3 (1 cu. in.). In this method, the crystal isgrown in a pumped Pyrex ampoule maintained with a

very accurate and well defined temperature profile. Adynamic process which oscillates the temperature ofof the crystal and the source of raw material is

12employed . The crystal is cooler than the rest ofthe ampoule for most of the time, which allows it togrow. Periodically, the temperature gradient in theampoule is changed slightly so the crystal is hotterthan the rest of the ampoule and evaporated for awhile. This preferentially promotes the growth of amore perfect crystal. This effect results from thefact that imperfection sites have higher surface en-ergy; therefore, they are the first to be evaporated.The method is very complicated and processing requiresseveral weeks and very careful control. Even then,at the present time, the yield of good crystals islow, so such crystals are quite expensive. At thepresent time, this is the only way to produce materialthat can yield large area detectors ( > 2 x tcm), andthick detectors suitable for efficient high-energyphoton detection.

The second growth method, that of platelets de-

veloped by Faile , uses 1 percent of a carrier mate-rial (usually polyethylene) together with the rawmaterial which is located at one end of a sealedpumped ampoule in a furnace at about 230C while theother end is held at room temperature. In approxi-mately a day, the material sublimates to form a massof many plates at the end of the ampoule. Since theprocess is not well controlled, most of the platesthat grow are very thin and small, but some are ofadequate size and properties to give good performances

as nuclear detectors . This method is very attrac-tive because of its simplicity, but its uncontrollednature makes it impossible to guarantee a constantsupply of good quality detectors. The very low yieldof good detectors achieved by this method has resultedin the development at Kevex Corporation of a newprocess which can give higher yield than any otherprocess but applies much better control of the

11growth and transport parameters . While the methodis still limited to small crystals (relative to theScholz Method) it can be tuned to give either a highyield of small plates (several mm2 in area and 500 mmthick) or larger plates (approximately 6 x 6 x 2 mm)which are capable of detecting the higher energy ofy -rays (up to 662 keV). A controlled doping of

00 1 8-9499/82/0600-1119$00.75 c 1982 IEEE 1119

impurities which probably compensate trapping centersallows better collection efficiency of electrons with-out sacrificing the collection efficiency of theholes.

The role of the carrier polyethylene in these114

processes is not fully understood . Several phenom-ena relating to this have been observed. The poly-ethylene appears to bind loosely to the HgI2, andbehaves as a transport agent in the process, while itdissociates from the HgI2 at the low temperature endof the ampoule near the crystal growth sites. It isalso likely that the polyethylene is a very efficienttransport agent for HgI2 but that it is very ineffi-cient for the transport of Hg2I2 and other impurities.This has the effect of purifying the crystals and im-proving their stoichiometric qualities.

The Temperature Oscillation Method (TOM) ofScholz is the only known way of growing large crystals(.15cm3) at present. Therefore, for large volume de-tectors, this is the only feasible crystal productionmethod. Furthermore, difficult processes such ascutting and cleaving are required to produce detectorsfrom such crystals, so low yields of good detectorsresult and their cost is high. The performance oflarge detectors is usually quite poor due to severehole trapping and non-uniformity of the material.These problems appear to be serious limitations indetectors made from TOM material.

It seems that improving the collection efficiencyof holes that might be caused by non-stoichiometry,impurities, internal delaminations in the crystals,dislocations, etc., is the main present limitation togetting high yield of good quality large HgI2 detec-tors for high y -rays using the TOM method. Non-uniformity also limits the performance of large-areaHgI2 detectors.

An essential requirement for using HgI2 for lowenergy x-ray spectroscopy is to have a low-noise elec-tronic system. In most cases, the major contributionto the energy resolution is electronic noise. There-fore, a low capacitance detector is required. Forthese kinds of applications, using the TOM method hasa disadvantage since the crystals are too big.

The new growth method, which was developed

at Kevex Corporation gives a high yield of moderatesized plates that are ideal for x-ray applicationswhen low capacitance is essential to give good energyresolution. Some plates are thick enough to be usedfor higher-energy y-ray spectroscopy, but the mainadvantages of this method are for the range of 1-60keV.

Detector Performances

In 197610 the 662 keV y -ray line of Cs 37 was

measured with an energy resolution of 4.5 keV FWHM(Figure la) using a HgI2 detector. This was achievedby developing an iodine treatment to improve thestoichiometry of the crystal grown by the TOM method.These detectors also had the capability to producephotopeaks for the 1332 keV and the 1173 keV y -rays

of a Co source (Figure lb). This capability is un-

usual because it is very difficult to compensate forthe iodine lost during the purification of the raw

material for the TOM method. Figure lc shows a spec-

trum obtained on the 60 keV line of AmWhen HgI2 detectors are used to detect high-ener-

gy y-rays, a polarization phenomenom is often ob-served. When high-energy photons are absorbed deep ina thick HgI2 detector, a high probability exists fortrapping of carriers (especially holes) which results

in changes in the internal field as illustrated inFigure 2. Depending upon the quantity and trapping!

detrapping properties of the deep traps and on thefree charge-carrier density (due to radiation andleakage currents) a positive space charge builds upat the negative electrode side of the detector and anegative space charge occurs near the anode. Sincethis is a dynamic process governed by the trapping/detrapping process and the amount of free charge-carriers (i.e., radiation history and leakage cur-rents), the internal field changes with time causingan instability in the detector performance.

This polarization effect exists in most of thethick detectors ( >1 mm thickness) and some thinnerdevices if used for high-energy photons. The polari-zation effect causes instability in the resolution,intotal counting efficiency, full-energy peak efficien-cy, photopeak symmetry, and in the peak position whilethe detector is under bias and radiation. It is pos-sible to improve the detector behavior by carryingout a periodic treatment to reduce the polarizationeffect. This is illustrated in F'igure 3a. There areseveral options which tend to establish a stable in-ternal electric field configuration either by clean-ing the traps or by filling them in a homogenous way.These options are as follows:

a) Remove bias from the detector for a period oftime. Illuminating the crystal at zero bias speedsup the detrapping process.

b) Reverse the polarity of the bias for a shorttime therebybuilding a new profile of the internalfield.

c) Illuminating the detector under bias, whichcauses high leakage currents which fills the holeand electron traps in a uniform way. This usuallyresults in relatively lower performance but stablebehavior. The effect of this treatment is prolongedbecause detrapping times are very long since themiddle of the band gap at 1.05 ev in. corresponds toa detrapping time in the range of many hours.

A polarized detector might have a large regionin which the electric field is very low. Therefore,the detection performances such as full peak count-ing efficiency, FWHM, etc., will deteriorate. Atlower bias, the polarization effects are more severe;the detection performances are less stable both quan-tity- and percentage-wise as described in Figure 3b.

Another comnmon effect at high energies ( > 60 keV)is asymmetry of the photopeak. This is due to incom-plete hole collection resulting in a low-energy tailon peaks. This problem is almost absent when low-energy x-rays are incident on the negative electrode.Here, signals are primarily due to electron collectionand the hole contribution is negligible. Also, therelative contribution of noise to the peak width ishigher, resulting in almost Gaussian peaks. Some im-provement in the tailing can be achieved by rejectingslow pulses. A serious penalty in efficiency resultsfrom this procedure as shown in Figures L4 and 5. InFigure 4, the improvement in tailing due to acceptanceof only those events of <20 ns rise time is accom-

panied by a loss of 90 percent of the events as com-

pared with accepting rise times up to 400 ns. Figure5 shows the effect of accepting rise times up to 60 ns

compared with 20 ns. The efficiency changes by a

factor of 5 between the 20 and 60 ns rejections times.Discriminating these pulses practically colli-

mates the detector to a very thin layer wnich contri-butes to the throughput. Reducing the influence of

some of the trapping celnters (which have detrappingtime similar to the collection time) is possible byselecting only those pulses which have rise time fast-er than 0.8 jsec. while the peaking time in theamplifier is 23 psec. Thus the 'trapping-detrappingnoise' can be reduced. Using this technrique allows usto discriminate pulses which some of their chargecarriers had trapped. There is no certainty thatthose pulses will be high enough (within the collec-

1120

tion time, which is similar to the detrapping time)to enter the photopeak. These pulses might contributeto the left tail of the photopeak or to the back-ground.

Figure 6 compares a Terbium fluorescence spec-

trwn, excited by Am source, with and withoutsuch rise time discrimiination. Not only has the energyresolution been improved but the photopeak has becomemore symmetric and the background has been dramati-cally reduced. These factors are very important forseveral applications when trying to detect low countrate peaks buried in the background from other peaks.

Application to Low-Energy X-Ray Spectrometry

For detection of low-energy x-rays in the rangeof 1 keV to 30 keV, the use of platelets is veryattractive because small area and relatively thin de-tectors can be employed here. The geometry of HgI2detectors for low-energy applications is a compromisebetween electronic noise parameters (capacitance,leakage current.) and counting efficiency.

Fabrication procedures and handling techniquesplay a major role in obtaining low leakage current,high breakdown voltages and elimination of dead layersin HgI2 detectors. Since the basic physics of HgI2detectors and the contacts interface is not well un-derstood, and the technology for handling HgI2 has notyet been developed, the fabrication process involvesmuch personal skill and experience. An advantage ofthe platelet growth method is the fact that it is pos-sible to fabricate detectors without cleaving crystalssince the growth system can be tuned correctly toproduce mostly plates in the right thickness range

without cleaving. It has been shown 5 that the cleav-age causes trapping centers and internal delaminationsparallel to the C-planes of the tetragonal lattice.Following strict fabrication procedures, it is possi-ble to fabricate devices with a high yield capable ofholding lOOOV bias with leakage currents in the pico-amp range. The detectors can also exhibit practicallyno dead layers under the contacts, show no polariza-tion effects, and produce symmetric peaks in the low-

energy x-ray range. Results have been reported '2showing the capability to efficiently measure x-raysfrom 1.5 keV up to 10 keV with an energy resolution(FWHM) in the range 300 to 400 kV. The relativelylow cost, good resolution and room temperature opera-tion opens up the possibility of many applications.One of the most attractive applications is their usein XRF (x-ray fluorescence) analysis for non-destruc-tive sorting of metal alloys as illustrated in Figure7. One of the most difficult pairs of metals to sortis nimonic alloy #75 and nimonic alloy #90. The maindifference between these alloys is the existence ofabout 4 percent of Fe which can be detected as shownin Figure 7a.

Conclusion

Mercuric iodide detectors have great promise foruse as room temperature detectors in x-ray fluores-cence analysis applications and in other applicationsof low-energy x-ray spectroscopy. Their applicationto high-energy photon spectroscopy is limited by holetrapping and polarization effects even though tech-niques exist to reduce the effects of polarization. Aconsiderable amount of work needs to be done to gain abetter understanding of the physics of the material,of detectors and the contact processes. Until thisunderstanding develops, ad hoc procedures executedwith great care are required to produce satisfactorydetectors.

Acknowledgements

The author wishes to thank Dr. F.S. Goulding forhis assistance and contributions and for his criticalreading of this marnuscript.

References

1. M. Singh et al, 29th Annual Conf. on Applicationsof X-ray Analysis, Denver, August (1980).

2. G. C. Huth, A. J. Dabrowski, M. Singh, T. E. Econ-omu and A. L. Turkevich, Proc. of the 27th AnnualConf. on Applications of X-ray Analysis, Denver,August 1978.

3. S. Caine, A. Holzer, I. Beinglass, M. Schieber,and E. Lowental, "Fabrication and Evaluation ofHgI2 M:iniature Probes for In-Vivo Medical Applica-tion", IEEE NS-25 (1978).

4. S. Caine, A. Holzer, I. Beinglass, G. Dishon, E.Lowental, and M. Schieber, "Direct Current andPulse Counting Measurements on In-Vivo MercuricIodide Medical Probes", Nuclear Instruments andMetal, 110 (1978).

5. J. Nissenbaum, A. Holzer, M. Roth, M. Schieber,29th Annual Conf. on Applications of X-ray Analy-sis, Denver, Colorado, August (1980).

6. L. Scharager, P. Siffert, A. Holzer, and M. Schie-ber, IEEE Nucl. Sci. Trans., NS-27, 276 (1980).

7. H. L. Malm, T. W. Randorf, M. Martini, and K. R.Zanio, IEEE Trans. Nucl. Sci., NS-20, No. 1, 500(1973).

8. S. P. Swierkowski and G. A. Armantrout, IEEE Nucl.Sci. Trans., NS-22 (1975).

9. R. C. Whited and M. M. Schieber Nucl. Inot. Meth.,162, 113 (1979).

10. I. Beinglass, G. Dishon, A. Holzer, S. Ofer, andM. Schieber, "High Energy Gamma Spectra Detectedwith Improved HgI2 Spectrometers at Room Tempera-ture", Applied Physics Letter 30, No. 11, 611(1977).

11. A. Holzer, "New Method of Growing HgI2 Crystalsfor Low X-ray Spectroscopy", to be published.

12. M. Schieber, I. Beinglass, G. Dishon, A. Holzer,"GGrowth of Large Crystals of HgI2 from the VapourPhase by Temperature Oscillation Methods", CrystalGrowth and Materials Proc. First Eurojean Conf. onCrystal Growth, Eds. E. Kaldis and H. J. Scheel(North-Holland. Amsterdam 1977), p. 279.

13. H. Scholz, Acta Electronics, 17, 69 (1974).

14. S. P. Faile et al., J. of Crys. Grow., 50 (1980).

15. U. Glebert, Y. Yacoby, I. Beinglass, and A. Holzer,IEEE Trans. Nucl. Sci., NS-24, No. 1, 135 (1977).

16. A. Holzer and M. Schieber, IEEE Nucl. Sci. Trans.,NS-27, No. 1 (1980).

1121

negative positiveelectrode electrode

a

-J

LLJ

F-1-LJLUJ

Fig. la: A Cs 37 spectrum taken with 500 Mrm thickdetectors. The energy resolution on the662 keV peak is 4.5 keV.

Fig. 2: The internal field in "polarized" HgI2detectors due to (a) non-complete holecollection, (b) with a contribution of non-complete electrons of collections and (c)smoothedl6.

a

8'I

n) c0

U:9_,0

0 0 *

Fig. lb: A Co6 spectran; tGhe full energy peaks aresmaller than the Compton edges due to asmall probability of complete absorptionthrough the photoelectric effect. Two HgKa escape peaks are observed at 1260 keVand 1100 keV.

Fig. lc: An Am spectrum; the energy resolution of

the 60 keV peak is 1.92 keV. (Ref: Nucl.Inst. and Method, 150 (1978), M. Schieber,I. Beinglass, G. Dishon, A. Holzer, andG. Yaron, "State of the Art of CrystalGrowth and Nuclear Spectroscopic Evaluationof Mercuric-Iodide Radiation Detectors").

9I .t'v\

7

1004kFv

6() 4al ''- *At\ e -W-*A46f)- Tane without bias a min

c rodiotioan10 min

I,0 10 20 0 l0 20 0 10 20 0 10 20

TIME 103 sec

Fig. 3a: Behavior of a "polarized" HgI2 detectorshowing (a) the photopeak channel position,(b) the energy resolution, (c) the countingrate in the photopeak of Am241, as a func-tion of time. After 20 x 103 sec. the de-tector performance reached a steady state;bias and radiation were then removed for10 seconds, 1 minute and 10 minutes. Thecounting rate just after application biasis taken as 100 percentl6.

1122

cb

POSITION

Am

FWHM= 1.2keV Pulser

I I I

r

I I I *I

* a

4)NC

2

4LIL

V.-i" N\

41(

I

lo(

8(

4C

0 S * v * fielda

o~~~~~8 3X 104 cE

0 _ \ 6 *bC .6X105.

O , 63X 103 -10d

0 3x lo, Z

10 20 30 40 50 E-0 x 10 3 sec

TIME

bias

150( V

S,OO V

:o00 v

150 V

tIlc

0

Fig. 3b: Count rate versus time for different biasvoltages on a 0.5 mm thick detector. Thedeterioration is larger when the field onthe detector is smaller. The faster attain-ment of steady state condition at the higherbias is due to higher leakage currents. Thedetection efficiency is represented as apercentage of that at the highest bias(1500 V) just after application of the bias.After each measurement, the bias was reducedto zero.

wLri(1d)w U 2 1)11..- -

wi__dow UbU"C -Ow

96Energy KeV

Fig. 5: A Co57 spectra with a 0-20 ns window com-

pared to a 0-60 ns acceptance window forrise time.

(a)

381 500 642 716

Channel

(b)

122

122 KeV

136 KeV

Fig. 4: A Co57 spectra taken (a) with discrimina-

tion against all pulses with rise times morethan 400 ns, (b) while accumulating onlythose pulses with rise times below 20 ns.

The system rise time is 14 ns. The energyresolution on the 122 keV peak improved from

5 keV to 1.8 keV and the peak positionshifted by 1%. The reduction in the count-

ing efficiency is about one order of magni-tude.

308 500 649575 725

Channel

1123

cnI

ow<zta:

cn

0

C)

0(9

6(

.0.C-)

b ~~~~~~~~~~~~~~~~~FeKa

5 6 7 8

0 10 20 30 40 50 60 70

Energy, keV NiKaNimonic 90

Fig. 6: Terbium fluorescence spectrum excited by CrKaAm21 source, (a) without and (b) with NiK,/rise time discrimination of all pulsesslower than 1 psec. The energy resolutionof the 414 keV peak improved from 4.2 keV to2.4 keV and the peak position shifts to theright by about 700 eV when using the risetime discrimination. The peaking time in CrK/lthe main amplifier was 23 usec.

5 6 7 8

0 10 20 30 40

Energy, keV

Fig. 7a: An XRF spectrum from nimonic 75 using 109Cdexcitation. This high nickel alloy contains19.5% of Cr, 4.2% of Fe, and traces ofcobalt, copper and titanium. The scatteredpeak from the silver line of the 109Cdsource can also be observed.

Fig. Tb: A nimonic 90 XRF spectrum. The main differ-ence existing between nimonic 75 and nimonic90 is the existence of 4% Fe seen in thespectrum of Fig. 7a.

1124