A novel cavitation probe design a d some preliminary measurements of its application to megasonic cleading

Law- A. O u d ~ f o r ~ r m d M ~ ~ ~ d P h ~ ~ ~ l O I 3 N E 4 ( W I S i m & ~ e . Rhrh i~mn 99105

( ~ a c i v g d 5 S- 2001; nviscd 28 May 2002, .ocepfed 12 Jmw 2002)

Cavi~tionuandIcctivcphyaicslma:~frrroafamulng~w.~,~ har wida appli~tio~w in w h d i v e fields as sonocbemisny, in which &emid radians 8x1 initiated or adersad , or in I~IC el& componmt ckaning indusoy in which particles (and other matmi&) arc r e d fmm gllf~~es. Howew, devices derigncd to acP as cavitation monitors have had lime s u m , @y because lhcir inhusiycms often atfow thc cavitalion field itself. Rcscntal hen is a brief desniprion of a dquc cavitation monitor that niilizcs Uae phawmmon of sonolwnincsaKe as an indirect guanti&r of cevitution. It ap- to work efficiently over a b d mnge of acoustic &Id iacnrities and its applicadon w megasonic cleaning has provided intorenting ! and valuabk mdchu into this lcdmolonv. 8 2W2 Acwwticol Socufy of Amrrira i

A s ~ ~ a f ~ d ~ i a m h m shriak, tim requirencn~~ m mnrmt @ c u b mmminatim in the submicron size mgc on poducc wafa mfam be- comts cxrrcwly critical. Miuintiriag "killer pniiclcq" thar is, @elm on &c orda of oretenth of tho Line wid& of silicca devkviees, -8 that submimdizc pmiclts shauld be ranovcd during wet pmxssing. For example, the W particle size is 0.04 jun for a Wmegabit silicon device. Al-. hugh high-velocity sprv ckaning tshniques haw bsan shown to remove larger size particles fmm the surf.%& they do mt nrmve particlea in the anbmicmn rim range baause of thc formation of bovndnry layers on the wafer s n r f e Boundary laya &ems have bkn shown to d i q p m , or be overcome, when a megasonic wafer cleaning technique is used with both SGl (ammonium hydroxiddhydrogen peroxideldeionized w m ) and SC-2 (bydmgen chloride1 hydrogen p ~ d e l d ~ i O l l i 7 . d W) S~lutiOtIS (SpcISOtl er d., 1991). Thus, megamic cleaning is baoming a widely nMd technique for rcmoving panicla fmm wafer surfaces: howtva. this technique is still poorly undnatood md is dif- ficult to monitor.

' I b m a p p m t o b e r w o p c i a i p l l c l u r a e s o f ~ ~ ~ rhrough which megasonic cleatlcrs could ~ L i s h cl- ing: (a) through a dlnet action of m d field with the particle, and (b) thmugh an indirect acdos viz., acoustic cavimtion. Consik 6rsl a dim3 dimaction of tbe sound field with the attached partick, i.c.. the oscillating acoustic Reld e x m per io$c. fm directly m particuke m e r at- tached lo a bunday or surface. Thue oscillating forces might evcntudly overcome the amsnive auacbmcnt fares and fiee th e c l e . In an -tic field. a small particle of

%=a=& 4. km.pl.-ddg

u l e l l . o i d d i l l o P o l l U c b a c k d f a m . ~ . d m g f a r c c on obstaclu in the flow. Ho-. awustic pmIiclc wloei- tic3 (at (armgarOllie dcaldng inrmoitics) are of^^ dis- plaowncnt rn mult in aignifi~~nt palticle motion. Far ex- ample, at a hequency of 1 W, and ao arrmstic intensity of 1 w/cm2. the acou~tic particle displacemnt in watu is on the order of a fcw .4ngsm,ms (Kinokr and Eky, 1962). Ob- viously. this displacement is iasufiiciem to mmve pmiicles with hires on the orda of a few hundredths of a mimn. Some authors contend that &tion prssun fom=z, some- times d c d primary Bj- form (CNQ 1975), d d also m l t in particle removal. Amrding to Meoon (1990). the* foxes are principally responsible for mga.waic clean- ing. However, r e n t papers by 0l.m (1988) and by Gem and Hasheminejad (1991) dminc the conditions nmcssnry for optimal particle m v a l abd their t b r i e s " . . . suggest very high trrqumcies (-1 GHz) will be q u k d m m o w submicron coataminants Fmm wafers" (Olson. 1988).

A more plausible mcCbanism for a submicmn pniick lmKlval is ruwt ic cavitation. l l i s modality ba tbne dif- f m l spacifs ways in which puticles can be removed ?be first mechmnism c d d be described as amustic micros- ing. In this case, a minoscopic air bubble pRexisting in the liquid (or nuckated on a solid smfaca) uod,=rgocs stable, I larpamplimde pulsalions whish in turn cansea rapid move ment of the liquid as it follows the oscillating bubble M- i ary. If any asyniinetry exists in the flow paitern (brought ; about by, say, a d y boundary), intense microsaaming i panems develop which can lead to significant shear ~aesm 1 imposed along the boundar).. Ibis mirroabeaming phenom- i emn is normally associated with tbe prcscoce of sf&& cavi- , tation activity, which gcnsrally occurs when the cavita- I liquid IS suumted with gas. However, -aming must! aka occur for more vigororrs cavitation as it is relatsd to the VO~UIUC oscillations of cavitation bubbles. Acoustic miam- I

!

;enaval@&, 195% N& 196% khhkowh and I k ? naiaa. 19m). Otha mEhaoinm such is SchlicUng and Eck- an JcMning may be iuvolvcd, but lhir role is consided to bc milla.

A saond cavi la t i~~la tcd , pkaM ~ w o a of pmick rcmovalistlnwghapharomnonthatswbecaUedmi~ jec impaa..lf lo o s d k i n g Wle is mdqoing relatively laqc disphmnmt ucuniions, t b n tk bubbk-Wan colhpac velocities can also be quite Lagc and the motion is inherently unstable. In thir cuse. the bubble is said to be inutially con- mlled and my aryrnmc~y in the flow field around the bubble normally results in an asymmcuic (nonsphtrkd) bubble coll-. Given this sccmrio, oa wall tends to col- lap= fmtcr thsn the other, &ring in the development of a microscopic liquid ju that prqxls itself through the bubble. p e r m s apposite w d , and violently impacts the very bovndary that cmutd thc asymmetry to develop in the first place. This liquid jet, which can attain supa~onic vclocitics. is often thonghl to be the principal mehanism for cavitaclon m i o n damage. Jet formulation is normally associated with tbc pwei!ce of inedhl cavitatlw, which p n d y ocem when the cavifating liquid is pda l ly dcgascd and the acoostic p m s m amplimdcs are relatively large. For mon infnmatioo on this phwmewa. see. for example, papers by Naudc and Ellis (1961). Lautuborn and Bolle (1975) and Crum (lW9.1988). Incidentally. ulvasonic c l e m opmt- ing at Mohcrtz fnquencies are w t suitable for wafer surface clcaning bccausc thae cavitation jets can damagc the sub strate; bowever. when megahutz fnqnencic~ are used (in gassy liquids such as SC-I and SC-2), the jet impacts me appamtly not sufficiently violat lo induce this mderirablc damage.

Athirdcnviatim-mlatedmslumismifrrcwltofthe violent implosion of tba collapsing gm bobbk: Whm thc bubble implodcq it can creak a shock wave in the liquid sonoonding the bubbk. The local p~csru~*i in this s h d wavc aro known to be M high aa a few kilobars--pnssuns sufficient to damage & hod- of numb. and thus, a p lmtial W i d e m v d mcchaoism. If che implagiog shock wavc can damage tk d a c e , it can clearly l m n ligbtly anPehed mtamimtion Allbough m cavitation march- m b c b Ja impacIs arc b e pr iwi i &anism for cavi- tation damage. it is not possible to eliminate shock wave effects as a pmible damage mum, and thus also captabla of particle temoval. However, shock waves an typically p duced by M a l nuhcr than stable cavitation, and am them fm more likely to be present When the cavitating liquid is modciaoely degassod or when the acoustic psstm ampli- Nde is very large, or when the acoustic fqucncy is in thc kilobere mgbmnditiws rhat iuc not normally met in w d a l megamic cleaners. Accoldingly. we m l u d e that cavitation-generated shock wvui do not play a major role in mcgasonic cleaner particle r ~ l ~ ~ d . Studies of shock- wave-associared acoustic cavitation dnm.1ge have bera p- formed by EUis (1966) and Tomits et at (1986).

A.pheUomCUOn tbaf %associatad with eavitatim ir h t of sonoluminesancc (SL), in which thc collapsing ovitation

J.kourBoc.AmW.~I?. ,Ha3,PL~.~-ZOCQ

ba~k heats the gm comsiood M bubb~t to i m - dePoent tempnhues. Although SL is gmra l ly ass- with inatial, dm than slaMe cavitation, it is commonly lmom thaf dtrrsonie clemcm, ovsn megasonic cleanag g e n e high levels of SL. In- using luminol, a wovc- shifting chemicali o m can wi ly m'rvich the d c d eye thc SL g m c d in an ulbasoaic clamer and detect arcas of nonuniformity of the amastic &Id.

To briefly summark panicle m v a l horn s i l i m wa- fer surfaces is most pubably associated d t h cavitation. At the a m ~ c intensities used in ulhasonic chum% parricu- M y megasonic Cleanas. Lhis cavita6on prodwxY SL. which an h d e M d with optical dctcctors. 'Ibus, an inshumeat tbat cwld detect thc SL pmducd by Lhm cl- can sene u a real-time, alhcit indirect, monitor of thc effectiveness of them cleaners.

II. APPARATUS AND SOW RESULTS

In the semiconduatn iodusty, silicon wafers lmdogo mauy cleaning stsgq a number of which utilize megasonic ckaning devices. Cavitation is a difficult phenomenon to moaitor; in a rtlll sms. it is a s~cbutic phcnomcnon, rs- salting fmm thc sumdipitmu spmial encounter of a pmx- W g nucleation site with d y the right temporal acoustic conditions. Nevcnheless. its presence is esmtial to cleaning.

When a numkr of ZOO-- wafen an cleaned in a megasonic cleaning taok, they arc typically pland in the tank for approximately 20 mio, along with a tm wafer. i b e wafm am either mschanically 'moved back and fonh to en- sure complete ruxrustic/cavitatiw field coverage, or. more R-

abtly. tbe acoustic field is itself spatially "moved" by an .ppmpristc amporal v m o n of thc voltage to thc various transduas~ lining tk cleaning lank. W~thin the Unk dso is a tmt wafer, which is ~ m o v c d and scanned with a micro- p h l e &kctor to murc that Ule ncoessary level of e l m - in$ is obtained. However, them is little if any effective real- time monitoring of the cleaning efliciency or effectiveness of the megawnic sysfca T h w arc a number of evenu mat can ormr wilhin a megasonic cleaner that could result in unac- eeplable c l d g :

(i) Oncofthc~tnadusns11191linethekmkcan bewmc lea d l i c k a ~ &dug in a "cold" spot or a ngion of i d k l i v e claning; altlmmgh some device8

.have lmnsdum maniwring electmnfcs, a minor change in trandwx efficiency can mult in such a wld spot--kc= cavitation is a threshold pbeonn- cnoo and is nonlincsrly dated to the aansducer's e l a t r i d eharadrristicr.

@) clcaniag solurion can bcfaac M a t i w for sup pcning cavilarion. If is well l t n m lhat at velocity of sound in a liquid is strongly dependent on the mm- ~ i b i l i t y of the liquid, which is in mm atrsmely xnsitive to rbc amount of gas bubbles pmmt In- daeq a "volume fraction" of gas on & ordm of lo-' can mult in significant changes in the soual velocity. Thus. heating of the clcming liquid a inadequate da-

g a s i n g c a u l c r u l t i a a ~ l o s s o f ~ c f - liciency. It is vay dimeult to mwita this cffcd m d t l o g megmnic cleanus.

(61) lbm ?bcrr MI real-time m i t o r of ihc emnt of clean- ing. Thc mmin clearung rims ww obraincd thiungb empirid sfudiw If ~ I C conditiom are famablc. complete cleating could ocnv WIU half Umt Bme: alrtmativtly, d ihy m unfavorable. mid ?As Wt wa- fcr ts not dquntely clam, a full 20-rn rcclc~ung is ~ f teaper fomrdThus , th i~~y~maLespooruscof cleaning cfficimcy.

~kscandnnmberof~tdl~lhodcrofwg+soplic - deaans m l d bs eliminated if n real& qvlmlion uumi-

taiog &vice wac available to mcaarn the efkctivenc~ of the nqpwnic cleaning pmass We have daigoed a u d r p davicc that is capable of dacctiqg cbe purosc of sonolumi- m.mmm with d d e r a b k sparial n~dlution, and iadiratly ihm& his ddcction of sono ldcxence , to detect the spatial and tQnporal enistmce of cavitation &If. Our Ourti& dcsigo for a cavitation ikkwtor is shown in Fig. I, and a photognylh of one of om promtype models is shown in Fig.

. 2. This design if very Ilexiblc and pamill the cavitation field to be measured and lldrmor4td in real time by a data acquisi- tion system. ' h e design is also quite simple and straightfor. wad.

~ b c cmbal the ,probe is a d l dosed ell *ch is made of m3al or swu & opaqvc mamid W p m .&mion by phomvls that origime from any other &on than the test voloms. An "acoustic window" that thc acoustic &?Id to encr the cell is ycom

pllshea by tdilkhg a 2S~puhick aoinlcrr ateel film I% film is rmnpIntly a p p to photonr bra is wsrly t n u q u a t to the acadc fleld. It is ilw thick cwogb to survive any cavitation damage that might be p m d d by rhe meglnonic clcaner.

A second imponant fcmvc of OPI dsoign is Ih.t it ma- surcs the cavitation p a d d in the fluid of in- mt some unrelated anificial fluid. lbis e&ct is rmmplishcd by conmcting a Iorlum path for the fluid to enter tk test region, but one such lbt extanal ph&m Can Mt We detamincd that this palh can w( bc a simpln one. but of several turns; otherwise, even if the ~ n ~ t i ~ codficicnt is quite low. ~ h c backgromd has such a high density of visible photoes &at "leak thmugh" cw be a major pmblem. We have coashuctcd this path by uid and crmr, with the rcsult b t nu bacLgmuml ~auot is on the order of M ~ ~ l u t d W. while typical cwnt kwls in the region of intens fav i t s t ioneanbcomtheaderafafewhualnd~d

Ihc t& region is viewed by a photomultiplier tube (F'M'I9 that is sensitive to a range bclwccn Z70 and 650 nm. although we an aware of thc water cutoff Ulat occurs below about 300 om. ~n& feature of our design is to use a p l a a m v e x hs wirh a focal length of 30 mm ta dircd n m y ph&om ar possible onto the collecting m a of the dcrsna, although wirh am sensitive PMT. wc still have a h g e dynamic' range in vrmp of the number of SL a m i s above background Incidentally. we find that a photpdiodk although more robost acd must less exparive. does mt pro- vide sufficient scnridvitv for this d o n . We have lcamcd also mar we can ure optical ha1 a, physically s ~ p m the PMT from the detector itself. Using an optical fiber per- mits us to conshuct he &U small enough to fit in the space behuem two wafcrs in a megasonic clcancr, tor example, thus permitting us to examine quite dkmly the actual cavi- tation field experienced by the dm wafers themsclvcli. Without the op&ai fiba, the diameter of thc sensitive arca of the mbe is on the mda of I sm. which is danmimd bv the sire'of the PMT. However, as w h be s h - n lam in &dam p t ~ m t e d (see Fig. 5, for example), spatial discrimiclatioo of the SL Bux a n be dersrmimd with much higher precision.

a * . I ~ ~ O f 6 0 W i O . a q w n t i a l r m n - m . m E v ~ r i a u - w 1 1 ~ i " 4 i . curd m dbe Rgurr. The nmte ol pbomnr-wbyhpohd* lhir snlirs rsqunee is plotled m the vmiul&. Nmc thacimwmirq ihe drsaiE.lpmumthemwdm-mb y m d . ~ ~ l * l h l r . d E l ~ . ~ ( ~ c i f D n m L k ~ o f c s " i l a t a o g==ud md, Urns, pmunahly. cm thc EMS ef ikbcy.

Iu Pi& 3 and 4 we pwidc rcnnc initial dm on mC nreurfvl quation of o~u prok, wblch demon- mates ckarly how it am k of collaiderably value in dew- mining rhe tunpal and sgatial channerisrics of mt cavira- tion field of a mcgmdc cleaner. In Fig. 3, we show the rssulrr of a aardy of dK effect of i d m m d u c ~ voltage on !he cavitation field of a co- mcgasonic clcaac~ 'Ihc data in Fig. 3 have imponant implications in IJE megas- mic cleaning community. Until thwe data were obfaineb and as far as we how. rbey lue unique-it war our n w d pfmmption that megamic cleanas wwLd work mors ef- fectively at electrical powers in excess of XY) W, indcui. some manufaciurcn advertise that their system have the ca- pability of operating ar Mowau powcr levels. We believe &at rhis cffca of reduced cffcctiverea at high powers is

NOT doc 0 limitaim in mC & t m d u c h aye rahs but is a c a m Uuid efieu. Specifically, if there is TOO much cavitation, Ilu in& volume fraaiw of gas in the liquid preventa h e roustic field fmm propagating. Such an e&cc has bem obsmved by me s o w c ~ a y am- munity. It is not umamnable to expecf, and gll&oM evi- dcna snppab our view, that t h e a& tinas when magar- d c clcn~~cr. apsators find that their sy- is not claaoing effectively, and subsequently increPse the power beyood the

I ramtion level, which mntinues to reduce their cleaning ef- fecrivenus. 'The uss of a cavitation pmbc would prevent Ulir ! cffat from occurring.

~ ~ 4 d e m 0 ~ ~ ~ 8 d i P c o v a y t b l . m d c w h i k t e s t i n g o a r p m b e . I t i s w e l l i w m ~ p b existing gasfillcd nucki 8c l ~a~sary form- iqcFp

m.4 l a m i . ~ . & & d p c n ~ m x b d m d I a m h Irveldavinaoll.airxlra-- ~ W N i m m t ~ m P I l Q c slun.-inwMchdLcliqoi6wa prrnmsa m & 4 S m M fa U ~ ~ . m d m s n t h e s l t M V Y ~ N m ( h * L k ~ O T uvil&a&tyb&Ydyimd m n ~ * a g n r m d ~ u d l m i ~ r - fivilyc"morhmccourrrdurex- pcriman. W,lh tho solib ve E b o w b c r r c l n w b i d I l i r w s b ~ through ihc cluDt liquid duhg op d o n . N w 6x1 k kvcl Of activity .b m m than lhm dmcr h i g h We f h r m l i d ~ . ~ * e s b o v . - i " Wbichtix Liqmdrrrpcrmintdt~rn mined undioavbsd fu r y d rmp mm, but vhm ihs c l r n a wsr ac ~ m i k o v r a t h c W c r . N o O Ihc pdwl iaacasc b Vlc lncl of mrira6a sctivity. We.= uavaa if Lk b.arimr sm in ~~iti"ily is.&3xiz ell ar cavimon m origio.

P I F . I ~ d a M n m . c x - p s a e d m p I & m ~ ~ m i ~ &!dm&.t*minin- lbm+ly md map*oly . we &ow ibo mmlu of ransing h Odd of -w w.-ic e*.ncr rrm . l i l i ~ f ~ ~ L " d - i s i r . orammicll m o ~ U pneDaDd "s fmnred".gtbe"drr ,+! . ip &sd by dl f m umhcem: h- cw, by %amlmg ova a r c p n of l h r r e a f m c w a i r - i h * "m!d~"u-ofrrdmdmi- ~ . ~ ~ a i m i a h ~ illp bllb.

tim; k, e .tabpleiely bmqpmw liquid hts m nor- mmdy b i i cwiwion Ihwhold. In maaC cxprimantS na examid the he flux dcpmding oo (a) the initial conditioas and R) tho ps8nce of a source of gas nuclei. W k n m a l l 4 tka Liquid to xwnain undisturbed within the mepa- onicclclna for 10 min, and fben masmd the SL activity as cbe clcaaet wm engaasd, we. ohsaved that pftcr an initial apih, the kvcl of cantntim activity was quite low, shoam ar the solid UiangIes io tbc figun, and slowly, but W l y . showed dsrca~ing activily over a paid of a h 1 2 min. We p u m c t h a t m w u e g r a d u n U y ~ c h c ~ g nucleation siw. If, on the other band, we eontinuad to sup ply nucleation sites. by bubbling air tbmugh the Liquid. born before and dming cleana opwuJion, we o b s d a similar bcbavior as Wow h a m by the d i d circlw in the figure. except that iu *' casc tix lmcl of cavitation sctivity was omn ltvm t h m tims higher. Fmdly, we allowed the system to remain nedisiwbcd for several minutes, tben we engaged both tbc cleaner and fbe bubbler. As shown in Fig. 4 as the solid squnrcg thc level of cavitatiar acthrity gradually in- c m s d from rbe l o w value, obmved without the buubblcr. to b e higher value, o w e d when tbe bubbler was active. One could muke fbe simple obsavation that it would bc bem to bubble gas UIIMlgh the mgawnic cleana wbilo it is in a c l d g modc.

1n~i~.4,wel lsbwticadthuwiminaf&sk~ds~fba megasonic cleaner was engaged, a .large apl% of cavita-

lion activity cccumd Wax wc wcn unable to msaant the electrical drive to tbe c l ~ ~ i l ~ r u a d m . we are u s m e if this uansient is an elWical one. OI a cavitarion-rrlatrd OM,

~oany,~~h0~i.~41.~thc~of&1YI~&p0be to dctamiac the h- of cavif+ion field pm- ducdbyarypicaZannmucialm~eckansr.~ibis care. wc aramioed a cleaner that laLd fovr madocas to

~ ~ ~ ~ e a v i t r r t i o n . ~ n v n i o m ~ o f ~ j t i o a , c i M a 1 1 1 u f t h c ~ c e s s a r c c n ~ a o y oneofthe irarisducas am sngagal, or I~IC tnnrduara am mgaged sequenhnl1y-a normal mode of opaatiw. It is pesumod that. in s q w d i . 1 opedon, there is no red to move the silimn wafen in the tank bacaubc ths canmion activity IS

i F i w 5 dummates that. on Ux enntnry, when ihc clean-

I homogmmusly dishibuted over mC entire cl~auing region. '

ing region was cxamirrd d u h g sequenrid opcntion. tbe vatiow bnnsdwm g e n e a k i d of cavitation activity with regions bmcen thc rrPnrduccts whcrc very little cavi- Wion is occurriag. During ddrmal cleaning opaations, it is aot unusual to obrme u c l d bads on dm: these data , provide a plausible caviution-depndcnt explanahon fa the cxistena of mese bands.

IIL SUMMARY AND CMICU)SIONS

We repar? hem 6n a Dovcl design for a easitatirm pobc a d pmml mmc plimbary data on its qqk!~~tirn to me gasolue cleaning. It qpcacs to bc quite mbun and mark- ably Jcositivc. pcnnittiog us to obtain real-dme cavitation occlneDee data. ExpcrimeW perfcmned in n megasonic clcamr provided urea& i n f o d o n on cavitation be- hna, suggxtiq this prob bas M& applications in indus- trial cavilation research.

Thiswol lrwaasoplnmdinpmtbytha~Scimcc Fo&m and the Defense Advanced Resesrch Rojsca Agency. 2

a S L A ( l W q . - 8 i a t n c . f a ~ a i n a - d ~ " ~ . m sar ~ m . s.1363-137a .,

-LA ( m , ~ . - u d M . ~ b ~ i ~ J . k m n . ~ A m . 4 0 . 1 3 1 - 1 3 5 .

~ L h ( ~ . ~ ~ ~ ~ ~ m s c b t s i m k -1 csknll cwnmpdrn in ESW I. VrrnoM 140.15111-1596.

L A. o. "Snd-" my. lww +I. 72-29, EMS. S. A. 0959). -0"illtio. m i c m h t m i ~ ' . 1. *mraL *Am 31.

54-6s. LUlu.AT.U%@. ~ J m ~ r b o s L u ~ ~ ~ ~ " ~ SirthNrnlH@@pmiaSplpahuaaclobQ1966

O c a . T . L ~ H a b a n i K j a d . M . ( l ~ l ) . ' r i n a r ~ ~ o l U ul- ck~ ia f i pmMaq" I. I\mua Sac. Am. W. 3238-3U9.

W - h I.. nd B u e A (1993). "Nomaid h W i m of& m p m i e =laming technip," Roc. InsL Bavir. Sei. pp. 386-593.

ne~k. L. E. a. adey. A R (1%21. ~-lr .f-, M d Wiy. New Yat). p. 121.

W.. .nd Balk. H. 0. " G x p l r i d imd+dfl*i- t Y l w b r W l s d . ~ s i . 1 b o n i g b b a b m d d a r d i d ~ , ' 1 . F l d d

H s r m V P . ~ " R n i d o & c d m m ~ ' B e q d ~ . " i o ~ = m = k a m d f ~ s - b d ~ ~ m ~ m , r d ~ b ~ ~ . & .M UdaceI l k k b , M Y a t ) . ~ 359-382. , .

~, Nax& C. F., llld Ellis. A 7. (WI). "On Ux mchwm d CdvitWm m m&&m 1 r d

. . N-. W. M. (W. ' ' Aw@ic rtmamiw." la M A t z m i t x Pr*.

+bm*lU&&,&cdtry Ww.P.MloaiAcb9mdaNaYmk).gp %<**I

. . - - . -. . -, . - . . . - . % v e r s o n W . L ~ . M . K d ~ . P . I . ( n n ) . r n m a

of X'-IISC.2 m g w r uda demmg." io Pknnlurp #lk Sormd b ,e -dml .S,-dun m Clram. T-n a D..

12 pp. 10-17. Wt, Y.. Shimq A, ud +. I (1)IL). of

p r - ~ i m m d ~ p i l f ~ b y ~ v l ~ & -hm" Roc. b. S p p . m lhiwi4 57-82