resonant rfnetwork antennas for large ... - plasma technology€¦ · in both antenna cases,...

lOP PuBLISHING PLASMA SOURCES SCIENCE AND TECHNOLOGY

Plasma Sources Sci. Technol. 22 (2013) 055021 (lOpp) doi: 10.1088/0963-0252/22/5/055021

Resonant RF network antennas forlarge-area and large-volume inductivelycoupled plasma sourcesCh Hollenstein 1, Ph Guittìenne" and A A Howling!

l Ecole Polytechnique Fédérale de Lausanne (EPFL), Centre de Recherches en Physique des Plasmas,CH-IOI5 Lausanne, Switzerland2 HELYSSEN Sàrl, Route de la Louche 31, CH-I092 Belmont-sur-Lausanne, Switzerland

Received 25 February 2013, in final form 4 June 2013Published 27 September 2013Online at stacks.iop.org/PSST /22/055021

AbstractLarge-area and large-volume radio frequency (RF) plasmas are produced by differentarrangements of an elementary electrical mesh consisting of two conductors interconnected bya capacitor at each end. The obtained cylindrical and planar RF networks are resonant andgenerate very high RF currents. The input impedance of such RF networks shows thebehaviour of an RLC parallel resonance equivalent circuit. The real impedance at theresonance frequency is of great advantage for power matching compared with conventionalinductive devices. Changes in the RLC equivalent circuit during the observed E-H transitionwill allow future interpretation of the plasma-antenna coupling. Furthermore, high powertransfer efficiencies are found during inductively coupled plasma (lCP) operation. For theplanar RF antenna network it is shown that the E-H transition occurs simultaneously over theentire antenna. The underlying physics of these discharges induced by the resonant RFnetwork antenna is found to be identical to that of the conventional ICP devices described inthe literature. The resonant RF network antenna is a new versatile plasma source, which canbe adapted to applications in industry and research.

(Some figures may appear in colour only in the online journal)

1. Introduction

Resonant radio frequency (RF) network antennas areinteresting new plasma sources with numerous potentialapplications in plasma processing as has recently beendemonstrated [1,2]. They show many advantages comparedwith the conventional capacitively coupled plasma orinductively coupled plasma (ICP) devices and couldtherefore replace these plasma reactors for certain industrialapplications.

Resonant network antennas can be up-scaled withoutgreat difficulty to run as large-area or large-volume plasmasources with significantly less effort than the commonly usedcapacitive and inductive plasma sources. They can be adaptedto the size and geometry of the optimal plasma volume forprocessing. RF power matching of the reactor is considerablyeasier due to the real impedance of the resonant networkantenna. Consequently, the high RF currents or high RFvoltages as observed in capacitive or inductive plasma reactors

0963-0252/13/055021 +10$33.00

can be avoided. Furthermore, the impedance is only veryweakly dependent on the antenna size and therefore does notlead to new complications when up-scaling.

In this paper, we demonstrate that resonant RF networkscan be arranged to form large-area or large-volume plasmasources with properties similar to conventional ICP devices.As theoretical considerations [3] from the medical applicationsof birdcage coils and recent antenna design studies show,closed and open configurations of the antenna for plasmaproduction are possible and can be described using the samemathematical formulation.

In this paper, we give examples of an open networkantenna as a large-area planar plasma source and of a closednetwork antenna as a cylindrical plasma source. Both arecomposed of similar electrical meshes.

Low RF power operation of the planar RF network antennadriven at different normal modes is investigated in view ofapplications for surface treatment of large flat areas, suchas film coating [4]. The cylindrical RF network antenna or

© 2013 lOP Publishing Ltd Printed in the UK & the USA

Plasma Sources Sci. Technol. 22 (2013) 055021 Ch Hollenstein et al

Ca)

Figure 1. Schematic of the 16-leg cylindrical (birdcage) RFantenna, showing the electrical circuit and opposite points of RFfeeding and grounding. The vacuum vessel is a glass cylinder closedat the top and bottom by grounded metal plates.

birdcage antenna has not yet been investigated for applicationas an ICP source. Operation at different normal modesshows the capability of this antenna type for large-volumeplasma applications. The cylindrical plasma source can beused as a volume plasma source for immersion of manywork pieces or for treatment of a large substrate. Otherpossible applications are plasma sources for neutral beaminjectors in fusion devices, or for high-density ion thrusters.In both antenna cases, impedance measurements are useful ininvestigating the antenna-plasma interaction such as the E-Htransition. The similarity of certain results for the two differentplasma sources shows that it is possible to construct complexRF antennas out of the same basic network element, which canbe adapted depending on the application.

Furthermore, the same resonant RF network antenna,which generates plasma by inductive coupling in the presentcase, can be further developed for wave-excited plasma sourcessince this type of antenna has been shown to efficiently excitehelicon waves in the presence of a magnetic field [5]. ResonantRF network antennas are a new versatile type of antenna forlarge-area and large-volume plasma production as demandedby the growing plasma processing community.

2. Background of resonant network RF antennas

The closed birdcage coil shown in figure 1 is well knownfrom nuclear magnetic resonance (NMR) applications wherethe uniform transverse RF magnetic field produced by them = 1 mode is used to excite the nuclear spins within asample. The structure and electrical properties of such a closedcylindrical birdcage antenna have been extensively describedfor the non-dissipative case in [3]. A planar antenna is made

(b) metal screen

Idielectric

grounded metal baseplate

Figure 2. Schematic and electrical circuit of the 23-leg planar RFnetwork antenna. (a) Top view showing the RF feeding point andthe four ground connections. (b) Side view showing the metalscreen used to confine the plasma above the antenna.

by unwrapping a birdcage coil to form a flat, open structure.The planar antenna can also be described as a segment of aladder network, as shown in figure 2.

The theoretical background of the cylindrical and planarRF network antennas has already been described in theliterature [3]. The calculation of the electrical circuit for theclosed and open antennas is distinguished only by the boundaryconditions inherent to their structure. By solving Kirchhoff'sequations over the antenna with these corresponding boundaryconditions, it is shown that resonant modes arise for bothclosed and open antennas. Each mode m is characterized by aresonance frequency determined by the distributed inductanceand capacitance components and the mutual inductances, andby specific current and voltage distributions in the N antennalegs. The phase change per section is Zmtt / N for the closedantenna, and mit /N for the open antenna. The current andvoltage amplitudes are sinusoidally distributed in space andare temporally in phase. The theoretical and measured currentdistributions in the legs of the cylindrical and planar antennashave been given in [1, 2].

In the neighbourhood of a mode resonance the impedanceof the complex network antenna can be described by a parallelresonance r, L, C equivalent circuit. This general resultapplies to the cylindrical (closed) and planar (open) antennas,as demonstrated by the experiments described in this paper.The simultaneous measurement of RF current and voltage andtheir phase relation gives the antenna impedance and withthis the equivalent circuit components r, L and C can bedetermined. The interpretation of the measured changes ofthese components caused by the plasma can therefore giveinsights into the plasma-antenna coupling, as will be discussedin future work.

An important aspect of proper RF antenna operation is thelocation of the RF feeding and grounding connections on the

2

Plasma Sources Sci. Techno!. 22 (2013) 055021 Ch Hollenstein et al

....__300E (a)

m=6.c~ 200CDto)

m=4cca 1001:1CD m=2C-.E O

8 10 12 14 16 18frequency (MHz)

....__600E (b) m=1.c~ 400CDCJe~ 200CDC-

.57 8 9 10 11 12 13

frequency (MHz)

antenna. There are a large number of different RF antennaarrangements possible in view of geometry and also in viewof the RF operation and of the plasma obtained. In this paper,only one set of electrical connections is investigated for bothantenna types. The influence of other connection schemes onthe antenna behaviour and plasma operation is not discussedin this paper.

3. Experimental setup

Two different resonant RF network assemblies are investigatedin this paper: a cylindrical and a planar RF antenna.

The cylindrical RF antenna was built following theconcept of a high-pass birdcage coil [3]. The antenna wasmounted outside a Pyrex glass tube of diameter 32 cm andlength 50 CID. In figure 1, the RF antenna consists of 16 copperlegs equally spaced by 6.7 cm interconnected with capacitorsof2.47 nF. The uncooled antenna was fed at the mid-point andoperated usually with the opposite point grounded, as shown inthe figure. These electrical connections favour them = 1modestructure of the birdcage antenna. Two half-cylinder screenscould be mounted at a variable distance outside the antennaassembl y to provide fine-tuning of the resonance frequencyif a fixed excitation RF frequency was to be used [4]. Theoperating pressure range of the cylindrical RF network antennawas 10-3-10-1 mbarin argon; the maximum RFpowerforthisantenna assembly was 300-500 W.

The planar RF network antenna is built from elementarymeshes composed of 20cm-Iong copper rods and 2.6nFcapacitors similar to those used in previous works [1, 2]. Thewhole planar antenna, comprising 23 copper rods with 2.5 cmspacing as shown in figure 2, is placed inside a vacuum vessel ofdiameter 0.5 m and length 0.75 m. In order to avoid parasiticdischarges, the antenna is embedded in a metallic box filledwith silicone and covered on the plasma-facing side by a thinglass plate. Since the antenna is placed within the vacuum,the antenna-plasma distance (silicone plus glass plate) is only5mm, shorter than that in conventional ICP devices becausea dielectric window thick enough to withstand atmosphericpressure is not necessary. A grounded screen was placed7 cm in front of the antenna to define a discharge gap astypically encountered in plasma processing. The RF powerwas fed into the middle of the antenna whereas four pointswere grounded, as shown in figure 2, thus favouring the m = 6mode. This planar RF network antenna corresponds to theelectrical configuration used in previous investigations [1, 2].The RF input power was limited to 200 W due to the RF currentrating of the capacitors. However, by using a water-cooledantenna and parallel assemblies of capacitors to allow highertotal currents, RF power up to 3 kW has already been usedfor coating applications [4]. The vacuum base pressure was10- 7 mbar and these experiments were also performed in argonat pressures of 10-3-10-1 mbar.

A wideband power amplifier was used for tuningseveral antenna normal mode frequencies to investigateplasma operation at different normal modes. Thefrequency-dependent behaviour of the RF network antennasand of the resulting plasmas could thus be studied. The

Figure 3. Magnitude of the impedance of the RF network antennaswithout a plasma, measured using a network analyser: (a) planar RFnetwork antenna and (b) cylindrical (birdcage) antenna.

amplifier frequency range was 10-100 MHz, with a maximumpower of 100W, and a pi-matching circuit to connect the RFgenerator and the antenna.

For electrical characterization of the RF network antennaswithout a plasma, a commercial RF network analyser was used,as shown in figures 3(a) and (b). During plasma operation,electrical characterization was performed using simultaneousmeasurements of RF voltage, RF current and their phaserelation. Commercial voltage and current probes were used.The probes were mounted as close as possible to the RF antennato minimize perturbations to the measured impedance. In thisway, the antenna impedance for different normal modes wasmeasured in the presence of a plasma. The frequencies aroundthe normal mode resonances were scanned using the RF poweramplifier. At each frequency step, the antenna was carefullymatched to the RF generator. The high-frequency probeswere calibrated by comparing impedance measurements withthe RF network analyser measurements in the absence of aplasma. Figure 4 shows a typical electrical probe and networkmeasurement of the impedance for the m = 3 normal modeof the cylindrical RF network. Note that the impedance andpower input of the resonant antennas can easily be measuredwith sufficient accuracy using voltage and current probes, incontrast to highly reactive capacitive or inductive reactors forthe following reason: for a conventional inductive or capacitiveplasma source, it is difficult to measure the input impedanceand power directly into the reactor using voltage, current andphase measurements at the reactor input, because the phase</Jis very close to +900 or -900 for these reactive-impedancereactors. This means that the phase must be measured veryaccurately to avoid large errors in the measurement of power =V I cos(</J)because cos(</J)is close to zero and is very sensitiveto small errors in the phase measurement. The matching

3

Plasma Sources Sci. Technol. 22 (2013) 055021

(a) 200 (b)100

180 80

160 60

........ 140 40E

Gi" 20i5 120 !........ DaCD ! O~ 100ca CD"C 120CD 80a. .&::..§ a.

60 -40

40 -60

20 -80

Ch Hollenstein et al

-100 L---.l'------1. _ __J

7.9 8.15 8.4 8.6frequency (MHz)

o _.....c:;;_-L-_--3f9-==

7.9 8.15 8.4 8.6frequency (MHz)

Figure 4. Comparisonof networkanalyserimpedancemeasurements(lines)with resultsobtainedfromRF voltageandRFcurrentprobemeasurements(points)withouta plasma (50 W) forthem == 3 modeof the cylindricalRF antenna:(a) magnitudeand(b) phase.

box for these reactors transforms their reactive impedanceto a resistive impedance (usually, close to 50 Q) and so thepower input to the matching box and the reactor system can beeasily measured, but now the power measurement includes anunknown power loss in the matching box. The advantage ofa resonant network plasma source is that the impedance andpower measurements can easily be made directly at the reactorinput (i.e. after the matching box), because the phase is closeto zero and, therefore, the measurements are not very sensitiveto small errors in the phase measurement.

As shown in previous works, the positions of the RFfeed point and the grounded points of the cylindrical andplanar antennas playa crucial role for the antenna and plasmaoperation [l, 2]. Several possible connection configurationscan be chosen but a detailed study of their influence on theantenna operation and the produced plasma is beyond the scopeof this paper. For this work, the above-mentioned connectionsof the RF feeding and grounding points of the cylindrical(figure l) and planar (figure 2) antennas were kept the same forall results shown. For the cylindrical antenna, the connectionsfor m = l operation were chosen, whereas for the planarantenna the same electrical connections (m = 6) as in theprevious work [1,2] were adopted. As a consequence, someof the modes exhibited very low input impedance.

For plasma diagnostics, Langmuir probes, light intensityand optical emission spectroscopy were used. Single anddouble Langmuir probes were applied to measure longitudinaland vertical density profiles and to detect the transition froma capacitively coupled discharge to an inductively drivendischarge (E-H transition). The profile measurements were

used to determine the uniformity of the plasma density and totest the proper operation of the antenna by comparison withthe calculated electric and magnetic fields.

4. RF antenna characterization

Figure 3 shows the impedance of the planar and cylindricalantennas without a plasma as measured by a commercial RFnetwork analyser. Both antennas show the presence of normalmodes whose frequencies are determined by the antennaelectrical components [1-3]. The reason for the fact thathigher modes are found at low frequencies is that the chosennetwork is a high-pass circuit [3]. A simple parallel resonanceRLC equivalent circuit can describe the impedance close toeach normal mode frequency to a good approximation. Theimpedances, real at the resonance frequency, cover a range ofup to several hundred ohms, a range favourable for efficient RFpower matching, provided that the connection configuration iscorrectly chosen for the required mode. For a given resonancefrequency, the RF current and RF voltage variations along theantennas were determined by RF probe measurements withouta plasma. These distributions were used to determine the exactmode structure on the antenna. The mode numbers identifiedand investigated in the frame of this paper are indicated infigures 3(a) and (b) for the planar and cylindrical RF antennas,respectively.

The magnitude of impedance for the planar antennaoperating at normal modes m = 2 and m = 4 in the presence ofa plasma is shown for different pressures in figures 5 and 6. Theantenna impedances were measured for different RF powers ateach pressure. The magnitudes for the m = 6 mode looksimilar to those given in figures 5 and 6. The magnitude andphase of the impedances show the typical dependence of theimpedance around the resonance frequency of an elementaryRLC parallel resonance circuit. The magnitude of impedanceat the resonance frequency is one of the main advantages of thepresent RF scheme compared with the usual inductive devicesusing solenoids or spiral coils. The normal mode frequency isslightly shifted towards higher frequency with increasing RFpower and the real impedance decreases, indicating RF powerdissipation in the plasma. The antenna resistance is reducedby a factor up to about 5. Figure 7 shows the impedancemeasurements for the cylindrical antenna, which basicallyshows the same behaviour of impedance as a function of theexcitation frequency as the planar antenna. However, in thiscase the observed frequency shift in the presence of the plasmais larger than for the planar antenna. In addition, the frequencyshift for the m = 1 mode is more pronounced than for them = 3 mode.

Frequency shifts could be a disadvantage when using afixed RF frequency generator in plasma processing, since theantenna impedance could shift out of the resonance conditionas the power increases, leading to impedance mismatch andhigh reflected power if the shift is large. The frequency shiftcan be compensated using metallic screens mounted around thecylindrical antenna: by adjusting the screen-antenna distance,typically by about a centimetre, the normal mode frequencycan be tuned to a fixed RF frequency power generator [4].

4

Plasma Sources Sci. Technol, 22 (2013) 055021 Ch Hollenstein et al

-e--ow 60 (a)-t--25W~50W .....--.

E-B--SOW .c

Q. 40CI)ucca'"C 20CDCl.

.5

60 (a)

Ê 50~Q_ 40CDus 30"C& 20.Ë

10

O~----~----~----~------~--~~14.15 14.2 14.25 14.3 14.35

frequency (MHz)14.4

_...._.15E.cO.,._..CD 10ucCI"CCD

.§ 5

-e--ow--\-25W-1~50W-B--SOW

(b)

17.55 17.625 17.7 17.75frequency (MHz)

Figure 5. Impedance measurements: the magnitude of impedancefor a planar RF network antenna at gas pressure 2 x 10-3 mbar as afunction of the applied RF power: (a) m = 4 and (b) m = 2.

The influence of a metallic screen has also been observed inNMR applications and is based on mirror currents inducedby the antenna in the metallic screen [3]. The frequencyshifts for the planar antenna are much smaller and can becompensated in advance during the antenna construction byadjustment of the distance between the antenna and its metallicbaseplate. Careful design of the planar antenna and its casingis an important factor for successful RF operation.

The real impedance of the antenna at a normal moderesonance frequency can be used to determine the powertransfer efficiency to the plasma as follows: the impedancemeasurements presented above have shown that the inputimpedance of the antenna network in figure 1 or 2 behavessimilarly to a parallel resonance circuit in the neighbourhoodof a mode resonance. This means that the impedance neareach resonance can be represented by an equivalent circuitof capacitance C in parallel with inductance L and resistancer. The values of C, L and r can be found by curve fittingto the experimental impedance measurements, and they havedifferent fitted values for each resonance. The impedanceof an r, L, C parallel resonance circuit is purely resistive atresonance and is given by

s.; = (w~acCvacrvac)-1,

where Rvac is the input impedance measured in vacuum withouta plasma, and Wvac = J Lvac C yac is the angular frequency

---+-O-+--- OW--+---- 25W-)~(-50W-'-4D-80W

14.2 14.25 14.3 14.35frequency (MHz)

14.4

20~--~--~--~--~-~..-,.. (b) O ow~ 15

25WO

>( 50W~ D SOWCDCl 10cca"aCI)Cl. 5.6

OL.--~----'-----~--~--_--L-___J17.475 17.55 17.625 17.7

frequency (MHz)17.775

Figure 6. Impedance measurements: the magnitude of impedancefor a planar RF network antenna at gas pressure 10-1 mbar as afunction of the applied RF power: (a) m = 4 and (b) m = 2.

at the corresponding mode resonance. On plasma ignition,the power dissipation in the plasma due to coupling with theantenna manifests itself as an increase in ryac, which is thedissipative element in the equivalent parallel resonance circuit.This is responsible for the fall in the antenna resonance inputresistance, as shown by the inverse dependence between Rand r in (1). More specifically, the ohmic resistance to currentin the plasma is transformed by the antenna-plasma inductivecoupling into an effective resistance in the antenna elementsas described in [6], whose combined impedance results inthe equivalent circuit resistance. Explicit expressions for therelations between the plasma resistance, the antenna elementresistances and the combined equivalent circuit resistancewould require further analysis, but the equivalent circuitdescription is sufficient for our purposes in this paper.

The resonance frequency is observed to increase by 8win the presence of a plasma. Since the equivalent circuitcapacitance Cvac is dominated by the antenna capacitors whosecapacitance is substantially larger than any antenna-plasmasheath capacitive coupling, we will assume here that the modefrequency increase, 8w, is due to a decrease in Lvac. In thepresence of a plasma, equation (1) therefore becomes

(1) R load = (w¡2oad C load rload) -1 , (2)

where Rload is the reduced input impedance when the antennais loaded by coupling to the plasma, {ù¡oad = Wvac + 8w, rload

5


200~~--~----~--~--~~ 1...-------,------.-----.--------r---,

Ê 150.t:O~

----t-O-+-- ow---I1---30W---'!lof---- SOW-4--4--- 70W

(a)

100CDCJCCI

"CDc..5 50

o8.1

600.........E.t:O 400~CDuc:ftS"C 200CDe,

.5O12

8.2 8.3 8.4 8.5frequency (MHz)

o ow+ 25WX SOWD 70W

12.25 12.5 12.75 13frequency (MHz)

Figure 7. Impedance measurements: the magnitude of impedancefor a cylindrical RF network antenna at gas pressure 10-1 mbar as afunction of the applied RF power: (a) m = 3 and (b) m = 1.

is the increased equivalent resistance due to plasma coupling,and the capacitance is unchanged, Cload = Cvac. Finally, thepower transfer efficiency 1] from the antenna to the plasma isgiven by the fraction of the equivalent resistance due to theplasma coupling, i.e.

Yi = rload ~ r vac = l ~Wroad R10ad

rload w~ac R yac

R10ad ( 8w )~1- -- 1+2- ,s.; Wvac

using (1) and (2), where Sco is the frequency shift due to theplasma, and Rvac and R10ad are the measured antenna realimpedances at resonance in vacuum, and with the plasma,respectively. Figure 8 shows the obtained power transferefficiencies of the planar and cylindrical RF antennas operatedat different normal modes. Both RF network antennas show apower transfer efficiency of 70-90% for the inductively drivendischarge obtained at higher RF powers. These values arecomparable to power transfer efficiencies of conventional ICPdevices [6]. Resonant RF network antennas are therefore asefficient as the conventional ICP sources investigated in theliterature.

Furthermore, from the frequency dependence of theimpedance changes due to the plasma, the equivalent

>.g 0.8CDë;IECD 0.6..~(I)CI! 0.4...

-~-m=1-+-+-m=3---m=2- .......-m=4

20 40 60 80input power (W)

Figure 8. Measured power transfer efficiency for the cylindrical RFantenna (m = 1, m = 3; p = 10-2 mbar) and for the planar RFnetwork antenna (m = 2, m = 4; p = 2 X 10-3 mbar).

400

~ 300.,C.siC 200....r:..~

100

o~~~~~~~~_----~----~O 25 50 75

power(W)100

(3)

Figure 9. Visible light intensity versus RF power. Planar RFnetwork antenna for mode m = 2 «+) 10-2 mbar, (o)2 x 10-3 mbar).

capacitance, inductance and resistance can be determinedusing the r , L, C equivalent circuit model. However, before themechanisms of antenna-plasma interactions can be elucidated,these impedance changes must first be related to changes inthe antenna component values in figures 1 and 2. Antennamodelling to understand the relation between the equivalentcircuit impedances and the antenna component impedances isunderway.

5. Plasma characterization

Single Langmuir probes and visible light emission were used tomonitor the transition from a capacitively coupled discharge toan inductively coupled discharge as the RF power is increased,as also observed in conventional ICP devices [7,8]. Figure 9

6


400


250

........:::ICI..._.......CCD....::l

100 CJCO:¡:¡t!:::s..asI)

C.2

100O

200

.......::sca..........Ce 1000..::su

S 500:¡:;I!::s~ o~~~--~------~--------~------~fi) o 25 50 75.21000~----~------~----~------~

OL-~~~~~--~------~----~o 25 50 75-e-- 0.001mbar

-t-- 0.01 mbar

---*- 0.007 mbar

500

o~~~~~~~--~--------~------~o 25 50

power(W)75 100

Figure lO. Ion saturation current versus RF power for variouspressures in the case of the planar RF network antenna: (a) m = 2,(b)m =4and(c)m =6.

shows the visible light intensity as a function of the RF powerfor two gas pressures for the m == 2mode of the planar antenna.A jump in the light intensity is an indication that the dischargeunderwent a transition from a capacitively coupled dischargeto an inductively coupled discharge. At low gas pressures,the E-H transition occurs typically around 60-70 W whereasat higher gas pressures, the RF power for transition is around20W. Similar results were obtained for all investigated modesand gas pressures, for the planar and also for the cylindricalRF antenna.

The measurements of the ion saturation current also showa typical change around the same RF power as for the visiblelight intensity. In figure lO the ion saturation current is shownas a function of the RF power for different gas pressures andnormal modes (m == 2, 4 and 6). In all cases, a typical jumpin the ion saturation current is observed. For all three of thenormal modes investigated for the planar RF antenna, the jumpappears at a lower RF power for higher gas pressures. Similarbehaviour of the ion saturation current can also be observed forthe normal modes of the cylindrical RF antenna, as shown infigure 11. The clear jump in the ion density and also in the lightemission (not shown) indicates a transition of the discharge.All these observations are in close agreement with reportedobservations of the so-called E-H transition in conventionalICP devices [7,8].

The impedance at the resonances presented in figures 5and 6 shows only a small reduction in the antenna equivalentcircuit resistance for low RF powers, whereas at higher RFpowers, substantial resistance reductions are found for all the

25 50 75power (W)

100

Figure 11. Ion saturation current versus RF power for the cylindricalRF network antenna at 10-1 mbar: (o) m = 1 and (+) m = 3.

normal modes investigated. Using equation (1), this leads tothe conclusion that the RF power transfer efficiency is lowduring the initial, capacitively dominated phase, as shownin figure 8. Only a small amount of RF power is absorbedin the discharge during the capacitive phase. High powertransfer efficiencies are obtained only after the transition tothe inductively dominated plasma phase.

All these experimental observations clearly indicate that,at low RF powers, both types of RF network antennas create acapacitively coupled plasma, whereas at higher RF powers, atransition occurs towards a fully inductively coupled discharge.This behaviour is thus identical to that found in conventionalICP devices [6-11].

Other electrical quantities are also influenced by thetransition from a capacitively to an inductively coupleddischarge. As can be seen in figures 5-7, the frequency shift ofthe resonance is small for low RF power capacitively coupleddischarges. At a higher RF power, larger frequency shifts areobserved. The frequency shift can be due to changes in theequivalent circuit capacitance and/or inductance. However,the antenna capacitance is substantially bigger than the plasmasheath capacitance, so that the plasma does not directlyinfluence the total antenna capacitance. Therefore, theobserved increase in frequency is due to a reduction in theantenna equivalent circuit inductance, which can be attributedto inductive coupling. A dissipative network model is currentlybeing developed to interpret these measurements in terms ofthe effect of plasma coupling on the antenna component values.

Spatial profiles of the ion saturation current were measuredin both RF antenna systems. Density profiles are useful inexamining the correct operation of the antenna in terms ofpower distribution corresponding to the driven mode, andin measuring the uniformity of the ion density, which is animportant parameter for industrial applications. Longitudinal(axial) and transverse (radial) ion saturation profiles weremeasured for the planar (cylindrical) antennas, respectively.In figures I2(a) and (c), longitudinal profiles for the planar

7


~ 0.8 - - - 0.002 mbar::l --- 0.004 mbar (a)"~.. 0.6ce...::l 0.4CJco 0.2;:ces...::l O.....ces 1(I)

c.2 0.5

O 10 20 30 40 50 60distance (cm)


...-...::lces 0.4~..c 0.3!..::lCJ 0.2c:O 0.1:¡:;L'!::l O.." 1(I)

C.2 0.5

O 10 20 30 40 50 60distance (cm)

Figure 12. Longitudinal ion saturation current profiles, measured2 cm above the planar antenna, for different pressures, modes(a) m == 4 and (c) m == 2 at 50 W RF power. The circles in (b)and (d) mark the positions of the antenna legs, and the trace givesthe normalized magnitude of the calculated magnetic field 2 cmabove the antenna.

antenna are shown for modes m = 4 and m = 2 for differentgas pressures. Figures I2(b) and (d) show the calculatedmagnetic field produced by the corresponding mode structure.The three local density maxima in figure 12(c) correspond tothe regions of maximum RF current flowing in the antenna,corresponding to the mode m = 2. This is also the case forthe m = 4 mode operation in figure I2(a), but due to theparticular distribution of the RF current maxima within theantenna, a narrower, more peaked density profile is observed.The similarity of the profiles shows that the plasma densitydevelops simultaneously over the whole antenna, as is alsothe case when the RF power increases in figure I3(a): At alow RF power, a capacitively coupled discharge is producedeverywhere, and also, at a higher RF power, the transition to aninductive discharge occurs over the entire antenna, as shown byfigure I3(b). A complete study of the density profiles obtainedfor the planar RF antenna shows that the choice of the modenumber, and also of the antenna internal electrical connections,all strongly influence the uniformity of the ion density, at least

:3 O.5~--~----~--~----~--~----~--~~CU~1: 0.4tU"-"-B 0.3cO~ 0.2..=1ä 0.1tilCO O~~~-~---~-----~--~---~--~~

O 10 20 30 40 50distance (cm)

60 70

:; 0.5~----~----~----~----~----~----~.~ (b)...¡ 0.4.....~ 0.3co.. 0.2'm..:::J16 0.1fi>s::~2 O

20 30 40 50 60power(W)

70 80

Figure 13. (a) Longitudinal ion saturation current profile for theplanar RF network antenna in mode m == 2 as a function of the RFpower (25,30,40, 50,60, 70, 80 W), and (b) ion saturation currentversus RF power taken at positions 15, 24, 39, 54 and 63 cm in (a).

,.... 600::lces,__,...... 500ee..:::J 400CJcO:¡:; 300l!:::J.. 200ces(I)

c::.2 100

OO

• m=2, SOW

• m=2, BOWO m=4,50WD m=4,80W

246 8distance (cm)

Figure 14. Transverse profile of the ion saturation current versusdistance above the planar antenna for modes m == 2 and m == 4.

for these low RF powers. At higher RF powers, more uniformdensity profiles are always observed [1].

Transverse profiles of the ion saturation current measuredin the 7 cm gap above the planar antenna are presented infigure 14 for m = 2 and m = 4 at different RF powers. At low

8


........._. 160:::sas.__.. 140...C!... 120:lUC 100O..t! 80:l.. OasI) 60 •C m=1,70W.2 40 E3 m=3,50W

• m=3,70W20-20 -10 O 10 20

distance (cm)

Figure 15. Axial ion saturation current profiles for the cylindricalantenna (p == 10-1 mbar).

pressures and RF powers, rather symmetric ion saturationprofiles are observed, a sign of the capacitive nature of thedischarge. At higher pressures and RF powers, the maximumof the profile is slightly shifted towards the antenna, indicatingan inductively driven discharge. The position of the ion densitymaximum roughly coincides with the calculated skin depth.Profile measurements show similar behaviour for different gapsizes. If there is no metal screen to confine the plasma abovethe planar antenna, the plasma fills the whole vacuum chamber.An arrangement of several planar plasma sources inside thecylindrical vacuum vessel wall could therefore also be used toobtain a very large volume plasma.

Axial ion density profiles for the cylindrical RF antennaare shown in figure 15 for m = 1 and m = 3 for differentRF powers. The behaviour of the other modes and moredetailed investigations into the influence of the mode structureon the plasma are currently underway. Within the RF powerrange limited by the present design of the RF antennas,double Langmuir probe measurements of the plasma showedthat electron densities of a few 1016 m-3 at 140W with atypical electron temperature of approximately 2.5 eV could beproduced by the cylindrical RF antenna.

The azimuthal ion saturation profile for the cylindricalRF antenna operated at m = 1 is shown in figure 16(a).The profile shows the same behaviour as the calculatednormalized magnitude of magnetic field produced by the coil(figure 16(b)). The magnitude of the magnetic field wascalculated for the case of vacuum with the COMSOL software.The fine structure of the magnetic field is due to the differentlegs of the birdcage coil. Finally, figure 17 shows the radialion saturation profile of the plasma produced by the cylindricalantenna. The antenna was used in the m = 1 mode at an RFpower of 65W at 4 x 10-2 mbar. The radial profile revealsa rather constant ion saturation current in the middle of thedischarge, whereas most inductively coupled discharges havemore or less pronounced hollow density profiles.

:ì500~------~------~------~----~~C 400e...s 300

~¡;; 1cGa"tJ 0.8"tJGi¡¡:: 0.6CJ:¡:¡

! 0.4c:Dca

~ 0.2 (b)o o~------~------~------~------~c o

co;¡ 200l!::s.....~ 100 (a).2 o~------~------~--------~------~

o 90 180 270angle (degrees)

360

90 180 270angle (degrees)

360

Figure 16. (a) Azimuthal profile of the ion saturation current in theplasma of the cylindrical antenna at R == 16cm (m == 1, Prf == 65 W,P == 4 X 10-2 mbar). (b) Calculated normalized magnetic fluxdensity at the same radial position as the ion saturation profile in (a).

~::sca.....".. 1250cl! 1000..::sU 750CO:¡:¡ 500t!::s 250.....II(II Oc.2

x (em)

Figure 17. Radial ion saturation current profile at the mid-plane ofthe cylindrical antenna (m == 1, Prf == 65W, P == 4 x 10-2 mbar).

6. Conclusion

Large-area and large-volume RF plasma sources can be built upusing different arrangements of the same elementary electricalmeshes to form planar, cylindrical or even other shapes ofresonant RF network antennas. A typical example of anelementary mesh consists of two conductors interconnected

9


by a capacitor at each end. When driven at one of itsresonant frequencies, very high currents are generated withinthe antenna structure. A resonant RF network antenna cantherefore be used as a source for an inductively coupledplasma, which can be scaled up to large areas or large volumesmore easily than for conventional sources [12]. The inputimpedance of all the various antennas, close to a normalmode resonance, shows the characteristics of an RLC parallelresonance equivalent circuit. At a resonance frequency, theantenna presents a real input impedance, thus facilitatingmatching to RF power generators. Changes in the RLCequivalent circuit values due to coupling with the plasma canbe used to determine plasma parameters, for example, a simpleestimation of the power transfer efficiency.

The plasmas produced by these antennas show thetransition from the capacitively coupled to the inductivelycoupled regime as found in conventional ICP devices usingsolenoids or spiral coils. The underlying physics of dischargesdriven by the resonant RF network antenna is essentiallythe same. The E-H transition occurs simultaneously overthe entire antenna. Therefore, it is expected that for bothantenna types similar performances and plasma parameters asfor conventional ICP sources can be reached.

The resonant RF network can be used to construct awide variety of RF antennas. A large-area planar antenna hasalready been successfully used in the packaging industry [4]and other interesting applications to other plasma processesare in preparation. With addition of an external magneticfield, the network antenna also opens the way to new wave-excited plasma sources. The resonant RF network antennais an important new plasma source, which can be adapted toapplications in research and industry.

Acknowledgments

The Swiss Commission for Technology and Innovation (GrantNo 9896.2 PFIW-IW) supported part of this work. In addition,

the authors would like to thank the CRPP and CRPP staff fortheir excellent support.

References

[1] Guittienne Ph, Lecoultre S, Fayet P, Larrieu J, Howling A Aand Hollenstein Ch 2012 Resonant planar antenna as aninductive plasma source J. Appl. Phys. 111083305

[2] Lecoultre S, Guittienne Ph, Howling A A, Fayet P andHollenstein Ch 2012 Plasma generation by inductivecoupling with a planar resonant RF network antennaJ. Phys. D: Appl. Phys. 45082001

[3] Jin J 1998 Electromagnetic Analysis and Designs in MagneticResonance Imaging (Boca Raton, FL: CRC Press)

[4] Guittienne Ph, Fayet P, Larrieu J, Howling A A andHollenstein Ch 2012 Plasma generation with a resonantplanar antenna 55th Annual SVC (Surface Vacuum Coaters)Technical Conf (Santa Clara, CA, 28 April-3 May 2012)

[5] Guittienne Ph, Chevalier E and Hollenstein Ch 2005 Towardsan optimal antenna for helicon waves excitation J. Appl.Phys. 98 083304

[6] Piejak R B, Godyak V A and Alexandrovich B M 1992 Asimple analysis of an inductive RF discharge PlasmaSources Sci. Technol. 1 179

[7] Cunge G, Crowley B, Vender D and Turner M M 1999Characterization of the E-to-H transition in a pulsedinductively coupled plasma discharge with internal coilgeometry: bi-stability and hysteresis Plasma Sources Sci.Technol. 8 576

[8] Kortshagen U, Gibson N D and Lawler JE 1996 On the E-Hmode transition in RF inductive discharges J. Phys. D:Appl. Phys. 29 1224

[9] Godyak V 2003 Plasma phenomena in inductive dischargesPlasma Phys. Control. Fusion 45 A399

[lO] Lieberman M A and Boswell R W 1998 Modeling thetransition from capacitive to inductive to wave-sustained rfdischarges J. Phys. IV 8 Pr7-145

[11] Zhao S-X, Xu X, Li X-C and Wang Y-N 2009 Fluid simulationof the E-H transition in inductively coupled plasma J. Appl.Phys. 105 083306

[12] Wendt A E and Mahoney L J 1996 Radio frequency inductivedischarge source design for large area processing PureAppl. Chem. 68 1055

lO

resonant rfnetwork antennas for large ... - plasma technology€¦ · in both antenna cases,...

Documents