ontheeffectof supplyfrequencyonthepositive Calculations Hg-Ar Ac discharge columnof a low-pressure P C DROPTandJ POLMAN Eindhovcn, Philips Rcscarchlaboratories, NV Philips' Glocilampenfabriekcn, Netherlands MS reccivcd28 October 1971.in rcvisedform 29 Novcmber1971 Abstract. Resultsof a thcorctical study of the positivecolumn of a Hg-Ar dischargc sustainedby an Ac cuffent with a sinusoidalwaveform(per= 3 Torr,psrs= 6 x 10-3Torr, inefficiencyforwemision withinc.reasing R= 1.8cm,Ier=0.4{.55 A) showa decrease supply frequency/ for f<500H2, and an increascin efrciency for />500H2. The efhciencycao be slightly higher than the Dc efficiencyfor a properly chosenfrequency (/)l0kHz). A comparisonwith experimentalresults, availablein the literature for comparabledischarges,is made, indicating a qualitativeagresment.In addition, the temporal behaviour of a number of plasma quantities is pres€otedfor various supply frequencies. 1. Inhoduction * ; $ It is known from experiments that the efficiency for production of Hg resonance radiation by the Ac mercury-argondischargeof a ffuorescentlamp increaseswith increasing supplyfrequency(Campbell1960). of the electrode Meyersand Strojny(1959)ascribethis increasemainlyto a decrease by Koedamand Verwey(1965)have losseswith increasing frequencyl Measurements decrease as/increases from of3 Torr the anodelosses shownthat for an argonpressure 50 to 400Hz, and that for/> 400Hz theselossesare practicallyindependentofl, that radial expansionof the arc with increasingfrequency Campbell(1960)suggests is responsible for an increase in efficiencyof the positivecolumn.Suchan expausionhas not beenobserved, of Readand Kerry (1965)and Hasebe however,in the experiments and Ka"kizima(1967). A third opinion has beenput forward by the last authors,who useda rathersimplesingle-stage ionizationmodel of the positivecolumnto showthat with increasingfrequency. the oc componentof the electrontemperature?" increases temperature(Ie) causesan increasein the efficiency. This increaseof the time-averaged resultsof numericalcalculationson theAc Hg-Ar positive Thepresentpaperdiscusses column,basedon a modelof the dischargeplasmawith a five-levelschemeof the Hg atom.It will be shownthat a minimumin efficiencyof the positivecolumnis found for f=500 Hz and a maximumfor f=60 kHz. Furthermore,a comparisonof our results with comparableparametersis made. with experimentaldata availablefor discharges for several Finally,thetemporalbehaviourof a numberof plasmaquantitiesis presented supplyfrequencies. t Permanent address: C-entral l,aboratory Ligbt Division, NV Philips' Gloeilampenfabrieken, Findlqvgn, Netherlands. 562 calculations on the efecr of supplyfrequency on thepo,titiuecolumn 563 In a previous paper (Polman et al 1972) the properties of the positive column in modulated and pulsedmercury-rare gasdischargeswere theoretically and experimentally investigated.The theoreticalmodel was basedon the time-dependentcontinuity equations for the densitiesof Hg(63P0),Hg(63Pr), Hg(6sPg)atoms and the electrons, and on the energyconservationequation for the electrons.The assumptionwas made that the radial profiles (describedby Bessel functions) are constant as a function of frequency. The electronvelocity distribution was assumedto be maxwellian.It wasfound that the average electron temperature(re) has a minimum value at a frequency for which a maximum Ie modulation occurs. Reasonableagreementwas obtaiued between the results of the calculations and the experiments.The reader is referred to the paper mentioned, where a complete descriptiouof the nonfinear effectsgiving rise to the (re) minimum is given. Here it is only mentionedthat the sameeffects,which will be explained in 92, also play a role in the properties of the ec discharge. 2. Results of the calculations The calculationshave been carried out, as in the above paper by polman et al (1972), for a Hg-Ar positivecolumn with pressures/a1 : 3 Torr, psr: g x lO-s Torr (determined by the tube wall temperatureof 42 oC),and tube radiusR: t .8 cm. Table l. Parametersof the Hg-Ar oc positivccolumn, with p1"= 3 16rr, pse=6x lQ-sTorr and R:1.8 cm Parameter Calculation Experiment I (A) E(V m-t) P (W rn-t) t (%) kT"(eY 0'4 73 2e.2 73.8 1.0 0'4 8l 32.4) 69 1.0 neo (m-3) r(6sPe) (m-s) zo(68Pr)(m-3) (m-3) no(63Pg) 5 .4 x1 0 1 7 5 .7 x1 0 r ? g .4 x 1 0 1 ? 1 4 .2 x l0 r z 5x101? 4xl 0rz 4x l 0r7 7x 101? For the oc dischargeunder theseconditions,the parametersof the positive column have been calculatedfrom our model. Similar calculationshave previously been made by Cayless(1963).Our results are given in table l, together with experimentaldata. The data for the electricfield strengthE, power input P per metrecolumn length,electron temperaturere and electrondensity reo at the tube axis are taken from verwey (1961); the excitedarom densitiesns(63P)at the tube axis from Koedam and Kruithot (1962); and the efficiency from Koedam et al (1963). The efficiency is defined as the ratio '7 betweenthe radiationoutput and the power input per metrecolumn length.In our model only the Hg(63P)levelsareconsidered.Further calculations,taking Hg(6tPr) into account, have shown that in Hg-Ar oc dischargesunder various conditionsapproximatelygo% of the radiation is emittedin the 253.7nm line (63P1 61Se),and that the efficiencyas referredto in this paper is representativefor that of -the total radiation output. For the ec dischargethe current waveform is given by I:Issin(2.J}). Figure I showsfor this casethe calculatedefticiencyagainstpolverinout for variousfreouencies t,,t ' 564 P C D ropandJP ol man = the curve for /err: 0'4 A /and for a number of effectivecurrentvalues/"11 *f ol/2. Front 500 Hz' At this for/near smallest it is seenthat the efficiencyantl the power input are the average (j"e)' consequently and particularfrequencythe averageetectrontenlperature (l0kHz5/54MHz) the frequencies hi-eher elecrricfield are found to be lorvest.For discharge oc of the that than higher be slightly to found of thepositivecolumnis efficiency rvith the same power input. For/> 4MHz the efficiencyapproachesthe DC efficiency again. P (W m-r) production of uv radiationagainstpower columnefficiency Figure1. Positive '1for supplyfrequencies / (kHz)anda numberof inputp pernetrecolumnlengthfor various currentvalueslett:pe,t:3Torr; pfls:Sx l0-3Torr; R: l'8 cm; A, f"n: effective 0 4 A; B, /err:0'45A; C' /en:0'55A' !:: -{r i. rt Three time constantsare important for the understandingof the behaviour found, namely the ambipolar diffusion time rp of the plasma, the energy relaxation time ta of the electronsand the averagelifetime z* of the excitedHg atoms.It is noted that the above quantities are badly defined, since the decay of the densitiesand the electron temperatureis not exponentialin the afterglow of this type of discharge(Polman and Drop 1972).For the presentargument,however,we shall considerthescquantitiesto be ps, t*:25 ps 7D-5 ms)' In consiants.In the presentdischargerB(t*(te('e-2 will be denoted by a (:)"71 current lc of the the following, the angular frequency n*. by atoms Hg(63P) of and the concentration If <.,(rp-r, the dischargecan be regardedas a slowly varying oc dischargeand its propertiescan be derived from those of a oc discharge.There will be only a small n"rltion of Ig during the period of the current,apart from a short time interval near the zerocrossingof the current.lf a)rz-r, t1e,Teand n* remainconstantin time and the propertiesare similar to thoseof a oC discharge.If rp-t4 u4tE-l, there is no ne moduiation, but a considerable?" modulation is found with a frequency that is twice the supply frequency.Becauseof nonlinearitiesin the system,a decreaseof the average temperatu.e(Ie) below its oc value occurs' due to the 7e modulation (Polman et al there is also an n* modulation, The uv emissionvaries with Te and lgTr). If -(r*-t, will be lower than in the nc case.Forrf:500H2, the effect of the intensity the average calculations on the efect of supplyfrequency on the positiue column 565 ]"" modulation is found to be largest,and consequently(re) and will be smallest.It 1 is noted, however,that for large Ts modulation deviations in the tail of the electron velocity distribution may occur (Polman et al 1972).In that case,the efficiencyfor f:500H2 may be somewhathigher than has been calculatedfrom our model. In the remaining frequencyronge r*-l 4 no n* modulation occurs and, due to the -4rr-r, nonlinear dependenceof the excitation rate on Ze, the n* density adjustsitself to that value correspondingto the highest14 during the period, which is higherthan that of the oc discharge,as will be shownlater. Moreover,because(f.) is still low, the elasticlosses will be lower tban in the nc discharge. Consequently r7is higherthan its oc valuein this frequencyrange, as has actually beencalculatedfor l0 kHzsf SaMHz (seefigure l). Finally,in orderto illustratethe temporalbehaviourof the variousplasmaquantities, we have plotted in figures2-4 someresultsof the calculationsfor/:50 Hz, 500Hz and 32kHz with,I"u:g'4A. (The valuesof the densitiesgiven are the densitiesat the tube 0 r l4 n l? 3rl 4 n Figure 2. One half'period of the current | field strength d electron temperature lile, electron density zeo and excited atom densitieslo (6sp) at the tube axis: Iorr:0.4A; f=50Hz axisdenotedas axo).In the 50 Hz case(figure2), where<o-ro-r, it is seenthat the electron temperature is nearlyconstantandcloseto thatof the0'4 A oc discharge (l.e: 1.0ev) apart from a short time intervalnear the zero crossingof the current.This has been confirmedby a microwavenoise measurement of ru as a function of time, giving an agreementwithin l0f with the resultsof the calculations.The electronconcentration xes aod the Hg(63Pr)atom concentration ,,0(63P1) have approximatelythe sameu,aveform as the current, but .E',n0(63pg)and to some extent Io and ro(6spo)show a rapid increaseat the beginningof a half-periodof the current.This is due to the fact that the electrondensityis very low in this time interval.Becausea certainamount of current mustffow, the drift velocityr)dand consequently theelectricfield f must increase steeply, causingariseof rg and, in this way,a risein thc productionof Hg(63p)atoms.Sincethe i'.. ': : il: ...:' 566 P C D roP andJP ol ntan Hg(63P0)and Hg(63P.r)atoms can only decayby collisionswith electlons,and sinceneo is stitl low, the concentrationsof the Hg(63P0)and Hg(63P2)atoms increasemore than that of the Hg(63P1)atoms,which can also decayby a radiativetransition. Figure 3 givesthe dischargequantitiesfor/:5gg Hz' wherethe maximum r" modulaoccurs' tion is found-.Figure4 showi that forf:32 kHz only a small changein n0(63P1) €n o \ a. u. I F Q t- rlz 3r 14 Figure 3 One half-periodof the current I, field strenglhE, electrootemperature*G, elitron density n"o and excited atom density no(63Pr)at the tube axis: Ietr=o'4A; .f:500 Hz. TI o E ff; ,..-t',*',,*?j.j ',,1::*:i.l :.; ' ..i.:l ,, '.1 * ^n cJ \ rl4 ql 2 3 r l4 Figure 4. One half-periodof the current I, field strength4 electrontemperatureftIe' elitroo density z.o and excited atom deosity zo(63Pr)at the tube axis: /err=0'4A; f:32kHz Calculations on the efect of supplyfrequency on the positiae column 567 and that the maximum value of 7e is higher than its oc value (seealso table l). Further the time interval betweenthe maximum values of calculationsshow that at/:l0kHz E and I is zero, while zs(63P1)has adelay of rl4 comparedto L Ie and.I are approximately in phase. With increasingf it has been found that, compared to the maximum value of 1, for/> 100kHz the maximum value of E showsa delay which is largest (a/10) at/:500 kHz and which decreasesagain to zero atf:4 MHz. We thus concludethat the dischargeis approximately 'inductive' for 50 Hz<.f<l kHz, 'resistive'from I kHz to At this l00kHz, then slightly'capacitive'andfinally'resistive'againnearf:4MHz, Iast frequency the modulation depth of Ie is decreasedso much that the discharge has the sameproperties as a Dc discharge(with the samepower input). 3. Discussion In our model of the lc positive column the efficiencyr1is found to have a minimum at ,f-5mHz; it increaseswith increasing/to slightly above its oc value for/il0kHz, and finally approachesthe Dc value for/:4 MHz. It is intcresting to compare our results with experimentaldata. Unfortunately, most experimentsas describedin the literature have beeucarried out under conditions different from those used in our calculations.However, a qualitativecomparisoncan be made. It should be noted that in the computations an ideal sinusoidal waveform has been postulated,while in practice slight deviations from this waveform will occur, especially tielow about I kHz. Readand Kerry (1965)measuredthe intensity of 253'7nm emissionas a function of the frequencyin the positive column of a Hg-Ar dischargewith constantpositive column power P:50 W m-1. Their resultsshow a slight minimum near/=200-500 Hz, and an increasewith increasingf to a value slightly above the oc value forfT2A kHz. However, this last effect is not very clearly beyond the experimental error. With regard to the in conditions,the agreementwith the resultsof our calculationslooks qualidifferences tatively rather good. Damelincourt and Scoarnec(1970) have qualitativelymeasuredn(63P2)againstfrequency and have found an increaseof this density with increasingf from 5@Hz to 20kHz,just as found in our calculations. On the dischargeas a whole-that is, including the electroderegions-more data are available(eg Campbell 1960).However, comparisonwith our resultsis difficult, since the anode lossesare known to depend on the supply frequency(see$1 of this paper). Above 500Hz the electrodelossesremain nearly constant(Koedam and Verwey 1965), and any changcin efficiencymust be due to changesin positive column efficiency.In with increasing fact, it hasbeenfound that the efficiencyof the fluorescentlamp increases (Campbell calculations. However, with our 1960), in accordance frequencyup to 20 kHz (Hasebe Kakizima 1967)is and cannot be exceeded the oc efficiency that the statement not supportedby our results,which show that the efficiencycan slightly exceedthe pc value, especiallyif the frequencyis taken higher than the valuesused so far. The increasein q with increasingfrequencywill be the more pronouncedthe lower the electrontemperatureof the discharge,for example,when using a higher current. Indeed, our theoreticalresults indicate that the efficiencyof the positive column, as compared with the 50 Hz value, is more favoured by frequencyincreasewhen using higher input powers(seealso figure l). t',a 568 P C Drop and J Polman References CampbellJ H 1960IIIum. EngngSS247-56 CaylcssM A L96JBr. J. appl. Phys'14 863-9 (London: IEE) pp 207-1I DamelincourtJ J and ScoamecL 1970Proc.Int, Con!. on GasDischarge,r 31-44 HasebeK and Kakizima T 1967Electl EngngJapanE7 Koedam M and Kruithol A A 1962.Pirur'ca2E E0-100 Koedam M, Kruithof A A and fuemensI 1963Physica29 565-84 Koedam M and Verwey W 1965 Proc. Tth Int. Conf. on Phenomenain Ionized Gases@elgrade: GradevinskaKnjiga) pP 392-5 MeyersG A and Strojny F M W 1959lllunr.Engng3465-70 Polman J and Drop P C 1972J. appl. Phys'submitted for publication PolmanJ, van dcr Werf J E and Drop P C 1972J'fiyr. D : Appl' Phvs' 5 266-79 Read T B and Kerry D C 1965 Br,J. appl. Phvs.16 453-5 VerweyW 196l Philips Res.Rep-Suppl'2 l-ll2 ;-,i) *i f: ii, li 1',: