Polymerization of fluorocarbons in reactive ion etchingplasmasStoffels, W.W.; Stoffels - Adamowicz, E.; Tachibana, K.

Published in:Journal of Vacuum Science and Technology. A: Vacuum, Surfaces, and Films

DOI:10.1116/1.581016

Published: 01/01/1998

Document VersionPublisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)

Please check the document version of this publication:

• A submitted manuscript is the author's version of the article upon submission and before peer-review. There can be important differencesbetween the submitted version and the official published version of record. People interested in the research are advised to contact theauthor for the final version of the publication, or visit the DOI to the publisher's website.• The final author version and the galley proof are versions of the publication after peer review.• The final published version features the final layout of the paper including the volume, issue and page numbers.

Link to publication

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal ?

Take down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

Download date: 11. Jul. 2018

Polymerization of fluorocarbons in reactive ion etching plasmasW. W. Stoffels,a) E. Stoffels, and K. TachibanaDepartment of Electronic Science and Engineering, Kyoto University, Yoshida-Honmachi, Sakyo-ku,Kyoto 606-01, Japan

~Received 14 May 1997; accepted 3 October 1997!

Polymerization reactions in radio frequency fluorocarbon plasmas of CF4, C2F6, and C4F8 have beenstudied by electron attachment mass spectrometry~EAMS!. In these plasmas polymerization occursreadily and molecules containing up to ten carbon atoms~the mass limit of the mass spectrometer!have been found. The densities of large polymers increase with increasing size of the parent gas. Ina fluorine-rich environment like a CF4 plasma the detected polymers are mainly fully saturated withF ~CnF2n12). As the amount of fluorine in the parent gas decreases, also the degree of saturation ofthe polymers decreases, which is clearly seen in C2F6 and C4F8 plasmas. The unsaturated polymersare more reactive, so they can stick more easily to surfaces and possibly create thick polymer films,which are often observed after discharge operation. The polymerization rate depends on thechemical activity of the plasma, which can be easily enhanced by increasing the radio frequencypower. The positive ions, extracted from the plasma, are generally somewhat smaller than theneutral polymers and their fluorine content is lower. This is probably due to dissociation of neutralsduring their ionization by plasma electrons and to ion collisions in the sheath region. Finally, wehave shown that EAMS has considerable advantages in the study of electronegative plasmas andpolymerization processes in comparison with traditional mass spectrometry. Unlike the traditionalmass spectrometry, employing ionization by high energy electrons, EAMS much better preservesthe structure of high polymers, allowing us to detect them as large negative ions. ©1998American Vacuum Society.@S0734-2101~98!03601-X#

inpr

tcully

thcess

laomto

io

rghid

leththe

entega-spe-forres.hingfacen beure

eras.ep

t onre-

antivethe-

ofetchbri-on-ringer

thetheof

s.at

chni

I. INTRODUCTION

Reactive ion etching of semiconductor materials usfluorocarbon gases is one of the most important surfacecessing methods in the semiconductor industry. In spitethe widespread usage of this technique and extensivesearch effort invested in understanding plasma–surface inactions, the exact etching mechanisms are still under dission. In order to improve performance in technologicaimportant areas like selective etching of SiO2 over Si orhigh-aspect ratio etching, one needs still more insight incomplex plasma chemistry and its impact on the surfa1

Numerous experimental studies of these chemical procehave been performed, involvingin situ plasma diagnosticssuch as infrared spectroscopy, mass spectrometry andinduced fluorescence, and surface diagnostics like ellipsetry, x-ray photoelectron spectroscopy, and attenuatedreflection Fourier transform infrared spectroscopy~see Refs.2 and 3 and the references therein!. Moreover, elaboratemodels for plasma chemistry and plasma surface interacthave been developed.4,5 Especially in reactive media likefluorocarbon plasmas, the combined action of high eneions and active species is essential for a successful etcprocess. Therefore most emphasis has been put on unstanding the ion and radical kinetics, while relatively littattention has been paid to plasma polymerization. Neverless it is clear that the latter process can drastically alterplasma chemistry. The plasma-produced polymers hav

a!On leave from the Department of Physics, Eindhoven University of Tenology, PO Box 513, 5600 MB Eindhoven, The Netherlands. Electromail: stoffels@discharge.phys.tue.nl

87 J. Vac. Sci. Technol. A 16 „1…, Jan/Feb 1998 0734-2101/9

go-ofre-er-s-

e.es

ser-

tal

ns

ynger-

e-ea

lower ionization threshold and a higher electron attachmcross section than the parent gas, so both positive and ntive ion formation are enhanced in the presence of thesecies. Furthermore, polymerization reactions form a sinkradicals and can lead to formation of macroscopic structuThese dust grains have been found in many reactive etcand deposition plasmas and they are responsible for surdamage. Many studies have revealed that large grains caproduced by gas-phase polymerization in low pressplasmas.6

Next to gas phase polymerization, formation of polymfilms on the surface is an important issue in etching plasmA clear example of this effect is side wall passivation in detrench etching.7,8 Elaborate studies by Oehrleinet al.7 re-vealed that the presence of such films has a major impacthe etching performance of the plasma. It influences thesistivity of the coated surface, which will be reflected byaltered plasma impedance and consequently the effecpower input. Charging of a deposited layer influencespositive ion flux towards the surface.9 This leads to undesired effects in high aspect ratio etching, where chargingthe layer deposited on a side wall of a trench reduces therate at the bottom and deteriorates the quality of the facated structure. Moreover, large dust grains which can ctaminate the substrate surface, are often formed duplasma processing by flaking or sputtering of the polymlayer.6 Finally, the presence of a polymer layer changessurface conditions, which in many cases seriously affectsplasma chemistry by changing the sticking probabilitiesplasma species and by introducing new surface reaction

The deposition mechanisms of these polymer films are

-c

878/16 „1…/87/9/$10.00 ©1998 American Vacuum Society

aen

tea

atcenereth

thc

tteveoea

e-

thuxtivth

thione

thic-ffi

thonthtivoh

aslere

on-

ha

rice

ih

er

ti

em-ov-sidengthe

of

arelec-

n torgyns.ol-agesrly

tre-

ichentthe

tivee-e,

l ashas

anr is

mp.

er-the

BPis a

88 Stoffels, Stoffels, and Tachibana: Polymerization of fluorocarbons 88

present poorly understood. Polymer layers on the surfaceeasily formed in fluorocarbon discharges with high conctrations of radical species, like CF, CF2, and CF3.

10–12How-ever, generally these radicals themselves do not contributhe polymer film formation. Contrariwise, in many casesnet radical fluxfrom the surfaceinto the plasma glow hasbeen observed.10,11,13This together with the observation, thsmall radicals stick poorly on a polymer-coated surfagives a strong indication that they are not the building stoof the polyfluorocarbon films, the deposition mechanismmaining thus an open question. It has been establishedsmall radicals are predominantly formed close to or atsurface and several production mechanisms have beensidered: surface conversion of radicals, ion-induced spuing of the polymer layer, reactions with high energy positiions in the sheath, etc. Chemical conversion of radicals frthe plasma has been excluded, as all these species appbe produced at the surface.10 Purely physical sputtering byenergetic ions has also been ruled out, based on the absof radical fluxes from a polymer film, exposed to ion bombardment in an argon plasma. However, the importancepositive ions in the radical production process is clear:ouflux of radicals from the surface is correlated to the fland energy of the ions, reaching the electrode. A quantitaanalysis reveals that the net outflux of radicals away fromsurface exceeds the influx of positive ions.10 Therefore, asingle ion must create several~up to ten! radicals. It has beensuggested that a positive ion gains enough energy insheath region to produce many radical species by colliswith neutrals. Indeed, energy-resolved mass spectromshows that positive ions undergo many collisions insheath under the same conditions in which a high radoutflux is observed.14 This mechanism certainly can contribute to radical production, yet it requires a considerable eciency of the ion-induced process in order to explaineffect quantitatively. At present, knowledge of such collisiprocesses and their cross sections is very limited. Anoway to produce several small radicals from a single posiion involves fragmentation of ions containing several carbatoms. Although mass spectrometric measurements sthat in most fluorocarbon plasmas CF3

1 is the dominant posi-tive ion,14 also high polymeric ions are present in these plmas, as discussed further in this work. Finally, compchemical sputtering of the polymer film has been considein which both positive ions and radicals~e.g., fluorine! areinvolved. In this kind of process the yield of radicals per ican be much larger than unity.15 Thus, this scheme can provide an explanation for ion-induced radical production.

The remaining problem is the actual deposition mecnism of the polymer layer. Recently, Boothet al.16 have sug-gested that this process is related to gas phase polymetion. The growth of a fluorocarbon film on the surfaproceeds by sticking of large polyfluorocarbons, formedthe discharge and diffusing towards the surface. This mecnism can be verified by checking the presence of polymspecies in the discharge.

As discussed above, a study of gas phase polymeriza

J. Vac. Sci. Technol. A, Vol. 16, No. 1, Jan/Feb 1998

re-

to

,s-at

eon-r-

mr to

nce

ofe

ee

es

tryeal

-e

erenow

-xd,

-

za-

na-ic

on

processes can provide valuable insights in the plasma chistry. Many important aspects of etching plasmas are cered: understanding of deposition processes leading towall passivation and surface radical production; clarifyiformation mechanisms of dust particles; and testingplasma chemical models. In this work we present a studypolymer formation in CF4, C2F6, and C4F8 low-pressure ra-dio frequency discharges. The polymeric neutral speciesdetected by means of a newly implemented technique: etron attachment mass spectrometry~EAMS!. While the tra-ditional mass spectrometry uses electron impact ionizatiotransform neutrals to positive ions, EAMS uses low-eneelectrons, which attach to neutrals and yield negative ioThe EAMS technique, when applied to electronegative mecules, has been recently proven to have many advantover ionization mass spectrometry, one of them is its clealess destructive character.17 The traditional electron impacionization mass spectrometry applied to heavy moleculessults in a very complicated fragmentation pattern, in whheavy ions are rather scarce. In contrast, only a few differnegative ions are produced from a given neutral parent,fragmentation is limited and consequently heavy negaions are readily visualized in EAMS. This structurpreserving feature of EAMS facilitates detection of largfragile molecules, like polymers.

II. EXPERIMENTAL SETUP

A detailed description of the experimental setup as welof the electron attachment mass spectrometry techniquebeen published previously.17,18 In summary, a capacitivelycoupled low-pressure 13.56 MHz plasma is generated inopen configuration, as shown in Fig. 1. The plasma reactoevacuated by a roots blower and a turbo molecular puThe powered electrode is a square (737 cm) aluminumplate, separated from a grounded aluminum shield undneath by 5 mm ceramic spacers in the upper part of

FIG. 1. Schematic view of the experimental setup; TMP, RP, and Mindicate turbo molecular, roots, and mechanical booster pumps, MFCmass flow controller.

leu

nene-fole

rogast-thm

aon

edeem

-oa.teiethsiedepeit

pe

attiincins

onnth

ea

caieoe,

ee

e areost

e-tronec-hisigh.

d atol-

s byingoss

ofuntedeV,red

hallrom-lec-ion,ell-in

theved argy,ndthech-

er-

yent.

of, sochative-

verent

ble

is aix-

canther

89 Stoffels, Stoffels, and Tachibana: Polymerization of fluorocarbons 89

reactor, where the diagnostic system is mounted. The etrode is connected to a vacuum manipulator, which allowsto adjust the position of the electrode. The plasma is geated using a RF Power Products RF10S-type generator ahomebuilt p-type matching network. The power levels rported in this work are the nominal values, not correctedthe losses in the matching network and cables. The stainsteel walls of the reactor and the quadrupole mass specteter ~QMS! head act as the grounded electrode. Processis supplied by a Tylan mass flow controller and a typicflow rate is 2–4 sccm. In this open configuration the redence times can be relatively long~about 10 s at the lowesapplied flow rate!, which is optimal for plasma polymerization, as it allows the produced species to accumulate indischarge. The gas pressure, measured by a Baratronnometer, is typically 15–30 mTorr. The measurementsperformed using a Hiden EQP500-type QMS with an ienergy selector. In our setup the grounded QMS orifice~150mm diameter! is situated about 5 cm away from the powerelectrode. The electron filament in the QMS, normally usas an ionization source, is now operated at low electronergies to allow for electron attachment to sampled plasspecies. The pressure in the attachment~or ionization! cham-ber is kept below 1026 Torr. The design of the Hiden ionization cage allows for an electron energy resolution of ab0.1 eV. Negative ions, formed from plasma species bytachment of filament electrons, are detected by the QMScase of F2 detection, a background signal must be subtracfrom the data. This signal, emerging at electron energhigher than 8 eV, is ascribed to sputtered material ofheated electron filaments. For other ions no backgroundnal has been observed. The maximum mass, detectablthe QMS is 510 amu and the sensitivity of the systemcreases substantially with increasing ion mass. Mass dedent transmission can be accounted for by calibration wsome known source gases and according to the Hiden sfications it decreases approximately as 102M /100, whereM isthe mass in amu. This correction has not been incorporin the data presented below, which means that the reladensities of large polymers are substantially higher thandicated by their count rates. The differences in the attament efficiencies of the various ions have not been takenaccount, mainly due to the absence of quantitative crosection data for the polymeric species. In this work we csider large polymers, typically attaching at low electron eergies~0–2 eV!. For these species it can be assumed thatabsolute values of the attachment cross sections incrslightly with increasing number of carbon atoms.19 TheEAMS data are collected by making an electron energy sfrom 0 to 10 eV. When the resonances from different specare well pronounced, the count rates of ions produced frthe corresponding molecules are separated. For examplcase of a CF4 plasma, the contributions of CF4, C2F6, andC3F8 can be distinguished in the electron energy dependsignals of F2 or CF3

2 ions. However, when no fine structur

JVST A - Vacuum, Surfaces, and Films

c-sr-d a

rssm-asli-

ea-

re

dn-a

utt-Indseg-by-n-hci-

edve-

h-tos---ese

ns

min

nt

is visible in the resonances, the count rates presented herintegrated over the electron energy. This is the case for mheavy ions detected.

Alternatively, neutral polymeric species have been dtected by the traditional mass spectrometry, using elecimpact ionization. The system is tuned for maximum dettion and an electron energy of 50–70 eV is used. In trange the corresponding ionization cross sections are hOn the other hand, the polymers are easily dissociatethese high energies and consequently only few heavy mecules can be detected. The detection of large polymerionization mass spectrometry cannot be improved by uslower energy electrons, as in this case the ionization crsections are also low.

Finally, the study is completed with measurementspositive ion fluxes from the plasma. To determine the corate of positive ions, the Hiden QMS settings are optimizfor each mass. The ions are collected at an energy of 20which is close to the plasma potential under the consideconditions.

III. RESULTS AND DISCUSSION

A. Dissociative attachment processes

Before discussing polymerization processes, we sshow some features of electron attachment mass spectetry. The method is based on electron attachment to etronegative species. In contrast to electron impact ionizatattachment is a resonant process, which requires a wdefined electron energy. The negative parent ion, formedthis process often carries a substantial internal energy. Inabsence of collisions with a third particle, small negatiions can stabilize by dissociation, yielding a lighter ion anneutral. The fragments carry away the excess kinetic enewhich is the sum of the kinetic energies of the neutral aelectron before attachment and the electron affinity ofneutral minus the dissociation energy. For example, attament to CF4 with maximum cross sections at electron engies between 6–8 eV, results in the formation of F21CF3,F21CF21F, or F1CF3

2. In the first case (F21CF3) the F2

ion has a significant kinetic energy~1–2 eV!, which influ-ences its detection.17 For larger molecules, attaching onllow-energy electrons such energy effects are not presMoreover, large molecules have many internal degreesfreedom, into which the excess energy can be distributedthe parent ion may stabilize without dissociation. In sucases the parent neutral can be also observed as a negion, like, e.g., SF6

2, resulting from nondissociative attachment to SF6. For fluorocarbon polymers, however, we hano evidence of nondissociative attachment. At least, panegative ions of fully saturated fluorocarbons (CnF2n12

2 )have not been observed, most likely due to their unfavoraelectronic configuration.

The resonant character of the attachment processgreat advantage of EAMS. Even if several species in a mture produce the same negative ion, their contributionsoften be separated, as each component yields a ra

uasv

ativ

nonhemrint

,a-ele.n

harge

-rgerheaatuthth

eventt

ned

ea-lly,

asy thellect

-.-

ersh

Fthe

pre-rs

dot

a-at-rbondete

s

As

thaturs

entt

f the

90 Stoffels, Stoffels, and Tachibana: Polymerization of fluorocarbons 90

peaked, distinct resonance at his own electron energy. Sseparation is much more troublesome in ionization mspectrometry, where the cross sections are broad and olapping is inevitable. In Fig. 2, the EAMS count rate of F2 ina CF4 plasma is shown. The resonance at 8 eV is duedissociative electron attachment to CF4. The signal at lowerenergies is observed only in the presence of the plasma,based on the electron energy it is attributed to dissociaattachment to C2F6 at 4 eV and to C3F8 at 3 eV.19 A similarresonance due to plasma-produced C2F6 is identified in theCF3

2 signal, but the C3F8 contribution is absent. Separatioand detection of species present in a plasma is onlyexample, demonstrating the advantage of EAMS. Anotimportant application is the study of the negative ion cheistry in a plasma. The same negative ions, detected duthe EAMS measurements are also abundantly present inplasma. In the considered CF4 discharge F2 and CF3

2 areformed by dissociative attachment from the same neutralsin the QMS system; F2 is the dominant plasma anion. Negtive ions are the major negative charge carriers in low prsure fluorocarbon plasmas, with densities exceeding the etron density by more than an order of magnitude20

Naturally, such high negative ion densities influence maplasma properties and clarifying the ion formation mecnisms is essential for a good understanding of the dischaThe negative ion signal, shown in Fig. 2 represents thefective cross section for F2 formation by dissociative attachment in a discharge. In combination with the electron enedistribution function in the plasma, this allows us to detmine the negative ion formation rate, without knowing tdensity of the various contributing species and/or theirtachment cross sections. It can be easily seen in Fig. 2 ththis CF4 plasma, with a typical electron temperature of abo3 eV, dissociative attachment to the parent gas is notdominant process for negative ion formation. Instead,conversion products C2F6 and C3F8, with their higher crosssections for dissociative attachment and lower electronergy thresholds provide the bulk of the negative ions, ethough the C2F6 and C3F8 densities are at most a few perceof the CF4 density.17 This directly confirms previous indirecmeasurements, relating infrared absorption data of C2F6 with

FIG. 2. EAMS of F2 observed in a 20 W CF4 plasma at 15 mTorr pressurand 2 sccm flow rate. The peak at 8 eV is due to dissociative attachmeCF4, while the peaks at 3 and 4 eV result from dissociative attachmenC3F8 and C2F6 formed in the plasma.

J. Vac. Sci. Technol. A, Vol. 16, No. 1, Jan/Feb 1998

chser-

to

nde

er-g

he

as

s-c-

y-e.f-

y-

t-intee

n-n

negative ion densities and their formation rates, determiby laser-induced photodetachment.20 In future studies theEAMS data will be compared with mass spectrometric msurements of negative ion fluxes from the plasma. Generaextraction of negative ions from a discharge is difficult,these species cannot pass the potential barrier, created bplasma. Many successful attempts have been made to cothe negative ions in the afterglow.21 However, the majordrawback is that this is not anin situ method and the acquired data may be influenced by some afterglow effects

After this introduction to the EAMS technique we proceed with the discussion on neutral and ionic polymformed in CF4 discharge, followed by a comparison witC2F6 and C4F8 plasmas.

B. The CF4 plasma

Figure 3 shows the polymers, extracted from a C4

plasma at 30 mTorr, 4 sccm, and 9 W, and detected usingEAMS technique. The counts of CnF2n2k are integrated overan electron energy scan of the QMS and the results aresented as a function ofn andk. Thusn denotes the numbeC atoms and a higherk corresponds to species with lesfluorine than the fully saturated fluorocarbon~CnF2n12 , k522!. In our case CnF2n11 (k521) is the largest detecteanion for a givenn; parent ions of fluoroalkanes have nbeen observed. Figure 3 shows clearly that polymer formtion readily occurs, as molecules with up to nine carbonoms are observed. Nevertheless, the dominant fluorocaanion is CF3

2, which has more than an order of magnitumore counts than C3F7

2, the next most abundant anion. Nothat mainly anions with an odd number of fluorine atomhave been detected. Species like CnF2m11

2 have no unpairedelectrons, so their formation is energetically preferred.can be seen, saturated anions (k521) are relatively abun-dant in a CF4 plasma. The resonances of CnF2n11

2 for n.2appear at low electron energies~0–2 eV!. The fact that littleenergy is needed to form a heavy polymeric ion suggestsno extensive fragmentation of the parent molecule occ

toto

FIG. 3. EAMS count rates of polymers, formed in a 9 W, 30mTorr, 4 sccmCF4 plasma. The data are arranged in series CnF2n2k . Thus a largen im-plies a larger polymer, while a largek denotes relatively little fluorine. TheEAMS counts for each species are integrated over an energy scan oattaching electrons.

thteat

le

-t

onine

roeinthansen

oC

oriti

t

d

at

edtMic

endo

e

inofmio

ea

eis

isndi-

ertoh

areicgasde-theor aesortrialare

her.to

ontheing

MShe

F0 W

91 Stoffels, Stoffels, and Tachibana: Polymerization of fluorocarbons 91

during dissociative attachment. This gives some clues inidentification of the parent neutral species in the complicamixture, which is formed in the plasma. Thus, it is likely ththe observed ions originate from fluoroalkanes (CnF2n12).The ions with k50, although energetically less favorabthan series with an oddk, are still visible, especially forlower n values. At highern values also some fluorinestripped negative ions emerge (k51,3), although their counrates are considerably lower. In contrast to the CnF2n11

2 ions,more electron energy is needed to form the unsaturated iThis indicates that these unsaturated anions are maformed from saturated parent molecules in an attachmprocess, followed by multiple bond breaking, e.g.,

CnF2n121e→CxFy21Cn2xFz1~2n122y2z!F.

Of course, some unsaturated ions may also be formed funsaturated parent neutrals. However, in a fluorine-richvironment like a CF4 plasma unsaturated species, containrelatively little F, are not expected to be abundant. Forsame reason polymerization is not very efficient, as all dgling bonds are readily saturated by the excessively prefluorine and consequently the growth of the carbon chaiterminated. Thus, in comparison with less fluorine-rich C2F6

and C4F8 chemistries, discussed below, the count rateslarge polymers and especially unsaturated species in a4plasma are relatively low. In a CF4 chemistry only limiteddeposition of polymer layers has been observed. Therefthe above measurements support the idea of film deposby sticking of polymers to the surface.

It is remarkable that C2F52 has a significantly lower coun

rate than the next higher anion C3F72. This feature, which is

observed also in C2F6 and C4F8 discharges, must be relateto the specific formation process of C2F5

2. Its major produc-tion channel is dissociative attachment to C2F6. This reactionhas indeed a low cross section and the preferred dissociattachment channels of C2F6 yield F2 ~see Fig. 2! and CF3

2

with cross sections about 1000 and 300 times highrespectively.19 Of course, most fluorocarbon species, incluing large polymers, can form F2 by dissociative attachmenand consequently many resonances contribute to the EAsignal of F2. These numerous resonances overlap, whmakes the analysis of the F2 signal less convenient for thstudy of polymer formation in plasmas. On the other hathe F2 signal is the most interesting one in the studynegative ion formation in the discharge, where F2 is thedominant negative ion.

Figure 4 shows the radio frequency power dependencthe count rates of saturated polymeric anions (CnF2n11

2 ) in aCF4 plasma. The electron density increases with increaspower level,22 resulting in an increased chemical activitythe plasma. This leads to enhanced dissociation and polyization, which is reflected by an increase of polymeric ansignals with power. In Fig. 4~a! the total electron energyintegrated count rates are displayed, so count rates of hions increase faster with power than the CF3

2 and F2 signals.The increase of the latter signals is due to resonances, eming at low electron energies. The signals resulting from d

JVST A - Vacuum, Surfaces, and Films

ed

s.lynt

mn-ge-nt

is

fF

e,on

ive

r,-

Sh

,f

of

g

er-n

vy

rg--

sociative attachment to CF4 stay roughly constant, as thereno significant depletion of the parent gas under these cotions. This can be seen in Fig. 4~b!, where the major low-energy resonances in the CF3

2 and F2 signals, attributed toC2F6, and the CF4 resonance are shown separately. Anothcontribution to the F2 count rate is dissociative attachmentF2 at 0–1 eV. The latter is readily formed in the fluorine-ricCF4 discharge and can be easily monitored by EAMS.

The polymerization reactions in low pressure plasmasrelatively slow. In a CF4 plasma, the saturated polymerproducts are removed from the discharge region by theflow rather than transformed into even larger species orstroyed at the surface. This is visualized in Fig. 5, wheregas flow dependence of some polymer signals is shown f20 W, 30 mTorr CF4 plasma. The polymer density decreaswith increasing gas flow, indicating that flow is the majdestruction process for these species. Note that in indusreactors the residence times at the applied high flow ratesshorter, but also the power densities are substantially higThus efficient gas phase polymer formation is expectedoccur during industrial surface processing.

In order to compare the neutral chemistry with the ichemistry, we have collected the positive ions, arriving atQMS head. This is representative for the ion flux, reachthe grounded electrode. The measured CnF2n2k

1 ion countrates in a 30 mTorr, 30 W, and 4 sccm CF4 plasma are shownin Fig. 6. These results should be compared with the EAdata shown in Fig. 3. The most striking difference is t

FIG. 4. ~a! EAMS count rates of some selected polymers, formed in a C4

plasma at 30 mTorr pressure and 4 sccm flow rate for 5, 10, 20, and 3radio frequency power inputs.~b! The count rate of CF3

2 formed by disso-ciative attachment to CF4 in gas~0 W! and plasma, and CF3

2 and F2 formedby dissociative attachment to plasma produced C2F6 as a function of power.

alis

-

atonthwiti

Se

ralliagevoletra

heeen

d iniveavytronric

aes

n,uo-

a

mas,r-. Inap-

ag-

d

CFor

-a

re.

92 Stoffels, Stoffels, and Tachibana: Polymerization of fluorocarbons 92

large amount of unsaturated positive ions. The dominnegative ions, observed by EAMS are saturated speciesCF3

2, C2F52, and C3F7

2, whereas in the positive ion masspectrum small ions like CF2

1 and CF1 are abundant. Moreover, larger ions with a highk value, like C3F3

1 are readilyobserved. Possibly, these ions are created in a dissociionization process or fluorine is stripped from saturated ias they collide in the sheath region. Another difference ismuch faster, monotonous decrease of the ion count rateincreasing number of carbon atoms. Even though the posion data is collected at a higher power level than the EAMdata ~Fig. 3!, conditions in which polymerization is morefficient ~see Fig. 4!, large positive ions (n.5) are notpresent in the plasma. Presumably, this is also due to fmentation during the ionization step and dissociative cosions with neutrals in the sheath region. Note that ion frmentation due to collisions will be more efficient in thsheath at the powered electrode, due to a higher sheathage. Thus, the positive ions, arriving at the powered etrode are expected to be smaller. Polymeric ions and neuhave been investigated by Howlinget al.23 in a SiH4 deposi-tion plasma and similar observations have been made. Tauthors have performed an extensive study of plasma chistry and dust particle formation, by means of mass sptrometry of positive ions and neutrals. Large negative io

FIG. 5. EAMS count rates of some selected polymers, formed in a4plasma at 30 mTorr pressure and 20 W radio frequency power input fgas flow rate from 1 to 6 sccm.

FIG. 6. Polymeric positive ions formed in a CF4 plasma at 30 mTorr pressure, 30 W radio frequency power, and 4 sccm gas flow rate. The ionsarranged in CnF2n2k

1 series. The fluorine ion F1 is cross hatched atn50,k51.

J. Vac. Sci. Technol. A, Vol. 16, No. 1, Jan/Feb 1998

ntke

iveseithve

g---

lt-c-ls

sem-c-s

and negatively charged dust particles have been detectethis radio frequency silane plasma, while no large positions nor neutrals have been visualized. The lack of heneutrals may be due to the detection method by elecimpact ionization mass spectrometry, in which polymemolecules easily fragmentize.

C. The C2F6 and C4F8 plasmas

Figure 7 shows the EAMS data in a C2F6 plasma at radiofrequency power levels of 7 and 20 W, respectively. As inCF4 plasma ~Fig. 3!, the species belonging to the seriCnF2n11

2 are most abundant for lown values. In the lowpower case@Fig. 7~a!# C3F7

2 and C4F92 have higher count

rates than CF32. This indicates that the first polymerizatio

step C2F51C2F5→C4Fx1••• is quite efficient. Furthermoreit can be easily seen that the degree of saturation with flrine is lower than in the CF4 plasma, especially in the highpower case@Fig. 7~b!#. Larger saturated species~k521, n.7! are not observed, while in comparison to CF4 there aremany more unsaturated polymers, also with highn values.This result is not an artifact of the EAMS method, butfundamental difference between the CF4 and C2F6 plasmachemistries. Saturated species are present in both plasbut in a C2F6 plasma, with a lower fluorine content, unsatuated neutrals are also expected to be well representedaddition, the resonances of the unsaturated negative ionspear at low electron energies, which suggests limited frmentation of parent species~see discussion above!. It is thusplausible to assume, that the negative ions with highk val-ues, detected in the C2F6 case, originate from unsaturate

a

re

FIG. 7. Polymer formation, studied by EAMS in a C2F6 plasma at 20 mTorrpressure, 2 sccm flow rate, and~a! 7 W and ~b! 20 W radio frequencypower. The polymers are arranged in CnF2n2k series. The series with anevenk number have very low count rates, so they are not displayed he

inm

ga

hal

is

aonn

rorv

io

ant

er

ally

ndsaedisighturesof the

s ofs isnle

temtualtedisgainly-

ciesture

tral

they

re-

Sn-ingedeex-de-nen-

ect-r,

-e

s

e

93 Stoffels, Stoffels, and Tachibana: Polymerization of fluorocarbons 93

molecules/radicals. Note that in a C2F6 plasma large poly-mers~regardless ofk! are more abundant than in CF4. This isagain the consequence of a lower fluorine content of C2F6,which enhances polymerization, as the number of termtion reactions with fluorine atoms is lower. Abundant fildeposition on the surface has been observed in a C2F6

plasma, which is an evidence of correlation betweenphase polymer formation and film deposition.

The count rates of positive ions arriving from the C2F6

plasma, collected by the QMS, are shown in Fig. 8. Textracted positive ions are slightly smaller than the neutrdetected by EAMS in C2F6 ~Fig. 7!. This is similar to the CF4case~see Figs. 3 and 6!. As in CF4, positive ions contain lessfluorine than neutrals. Again the dominant positive ionCF3

1, but also C2F51, C2F4

1, and C2F31, which are directly

formed from the C2F6 parent gas, are abundant. Note ththere is a clear preference to form ions with paired electr(CnF2m11

1 ), as these are more stable. For the negative ioformed in the EAMS method, this effect is even more pnounced, as ions with an unpaired electron are not obse~see Figs. 3 and 7!, with an exception ofk50 series in a CF4plasma. This agrees well with the negative ion formatreactions, reported by Spyrouet al.24

Figures 9 and 10 show the neutral and positive ion dobtained in a C4F8 plasma. The ratio of F to C for the paremolecule is lower than in the case of C2F6 and CF4. This isreflected by an even lower fluorine content of the polym

FIG. 8. Polymeric positive ions formed in a C2F6 plasma at 20 mTorr pressure, 4 sccm gas flow rate, and 7 W radio frequency power. The ions ararranged in CnF2n2k

1 series.

FIG. 9. Polymer formation, studied by EAMS in a C4F8 plasma at 15 mTorrpressure, 4 sccm flow rate, and 8 W radio frequency power. The polymerare arranged in CnF2n2k series.

JVST A - Vacuum, Surfaces, and Films

a-

s

es,

tss,-ed

n

ta

s

than in the previously discussed plasmas. It is especiclear for the positive ions, reaching the surface~Fig. 10!.Positive ion chemistry in this plasma is very complicated athe presence of cations of fully saturated fluorocarbon~CnF2n12

1 , k522! is striking. As the parent gas is alreadylarge molecule, significantly larger polymers can be formin this system. Unfortunately the mass limit of our system510 amu, so we cannot detect species with more than eor nine carbon atoms. The presence of even larger structcan nevertheless be expected, based on the comparisoncount rates in a C4F8 plasma with those observed in C2F6 andCF4. In the latter cases a fast decrease of the count rateneutrals and ions with increasing number of carbon atomobserved, while in a C4F8 plasma the count rate distributiois flat and high polymeric species still yield considerabsignals. Note that as the transmission of the QMS syssignificantly decreases with increasing ion mass, the acdensities of the high polymers are even higher than indicahere. In C4F8 formation of polymer layers on the surfacemore efficient than in the chemistries discussed above, asupporting the proposed film deposition mechanism by pomer sticking. Unlike CF4 and C2F6, in C4F8 also some poly-mers with an unpaired electron~evenk! are detected, thoughtheir count rates are much lower than those of the spewith paired electrons. Nevertheless, this interesting feacan be a consequence of a high radical density in a C4F8

plasma. As the fluorine content in this plasma is low, neuspecies with dangling bonds (CnF2m11) are expected to berelatively abundant. These species are very reactive andreadily form negative ions. Even though CnF2m

2 are not themost favorable negative ions, attachment to radicals cansult in formation of detectable amounts of these species.

Finally, we want to demonstrate the advantage of EAMover the traditional detection method by electron impact ioization of neutrals. Figure 11 shows the data collected uselectrons of 50 eV to transform the neutral polymers formin a C4F8 plasma into positive ions. This figure should bcompared with Fig. 9, which shows the neutral speciestracted from the discharge under identical conditions andtected by means of EAMS. It is obvious that in ionizatiomass spectrometry the high energy electrons cause fragmtation of the polymers and consequently the maximal detable size is only about half of that for EAMS. Moreove

FIG. 10. Polymeric positive ions formed in a C4F8 plasma at 15 mTorrpressure, 4 sccm gas flow rate, and 8 W radio frequency power. The ions ararranged in CnF2n2k

1 series.

naralnt

siyinn,F

raasrmnma

thbn

onrgthethow

rizn

toa

eto

ore

allyontgre-ntThistive,onent,ls of

ev-icals

ngthein-aymaro-mis-

beer-sur-

ofch

M.

ol. B

nd

de

de

ys.a

gy,

ondon

d by

mec-a

d bn

94 Stoffels, Stoffels, and Tachibana: Polymerization of fluorocarbons 94

large amounts of fluorine are stripped during ionization aconsequently all possible combinations of C and Fpresent in the mass spectrum. As a result, there is virtuno structure left in these data and the only information ocan obtain is that there are some polymers present inplasma, but no conclusions can be drawn about theirdistribution and degree of saturation with fluorine. Anywathe high polymers, which are most interesting from the poof view of polymer deposition or dust particle formatioremain hidden. Similar data sets to the one presented in11 have been also collected in CF4 and C2F6 plasmas. How-ever, in contrast to the EAMS method, no significant diffeences in the polymer chemistry of the considered fluorocbon gases have been found using ionization mspectrometry. Applying lower energy electrons to perfothe ionization might be a way to limit the undesired fragmetation. This is clear from the comparison of the data froFig. 11 with the positive ion count rates from the plasm~Fig. 10!. Positive ions are produced in the plasma, whereelectron temperature is a few eV. Ionization is performedthe electrons from the tail of the energy distribution functiowith energies just above the threshold of the ionization. Csequently, there is significantly less fragmentation and lapositive ions are present in the plasma. However, fordetection of neutrals by ionization mass spectrometry lowing the electron energy is not a practical solution, asionization cross sections at the threshold are extremely l

IV. CONCLUDING REMARKS

We have shown that fluorocarbon gases easily polymein low pressure discharges. Both polymeric neutrals apositive ions have been detected; heavy neutrals appearrelatively abundant. The study has been performed by meof a new technique: electron attachment mass spectromwhich is proven to be extremely efficient for the detectionelectronegative polymers.

We have found that the polymerization proceeds mrapidly in fluorine-poor conditions. It has been establish

FIG. 11. Polymer formation in a C4F8 plasma at 15 mTorr pressure, 4 sccflow rate, and 8 W radio frequency power, studied by ionization mass sptrometry. The polymers are arranged in CnF2n2k series. These data are obtained using 50 eV electrons to ionize the neutral polymeric speciesdetect the resulting positive ions, which in contrast to the data obtaineEAMS under the same conditions~Fig. 9! results in a much poorer detectioof polymeric species.

J. Vac. Sci. Technol. A, Vol. 16, No. 1, Jan/Feb 1998

delyeheze,t

ig.

-r-s

-

ey,-eer-e.

edbensry,f

ed

before that the presence of silicon in the plasma drasticlowers the fluorine content, mainly due to efficient formatiof the etching product SiF4.

22 Therefore, it is expected thapolymer formation will be especially efficient in an etchinenvironment, in which a silicon substrate is inserted. Moover, the polymers formed in a fluorine-poor environmeare less saturated and they may contain dangling bonds.makes the gas phase formed polymers especially reacwhich will enhance the deposition of fluorocarbon filmsareas, which are not exposed to a heavy ion bombardmsuch as reactor walls, grounded electrodes, and side waltrenches on a processed wafer.

For the formation channels of the observed polymers seral processes can be considered. In all cases small radplay a crucial role, either as building blocks or in terminatireactions. This means that apart from the question ofradical formation process at the surface, discussed in thetroduction, the density of radicals and polymers are anywstrongly linked. As gas phase polymerization under plasconditions has many important implications for surface pcessing, it cannot be neglected in a study of plasma chetry of processing discharges. More research effort shouldput in clarifying the polymerization mechanisms and undstanding the influence of polymers on the plasma andface.

ACKNOWLEDGMENTS

The authors~E. and W. W. Stoffels! wish to acknowledgethe support of the Japanese Society for the PromotionScience~JSPS!, the European Commission and the DutOrganization for Fundamental Research~NWO!.

1J. W. Coburn, J. Vac. Sci. Technol. A12, 617 ~1994!.2G. M. W. Kroesen and F. J. de Hoog, Appl. Phys. A: Solids Surf.56, 479~1993!.

3Plasma Diagnostics, edited by D. Auciello and D. L. Flamm~Academic,New York, 1989!.

4M. J. Kushner, W. Z. Collison, M. J. Grapperhaus, J. P. Holland, andS. Barney, J. Appl. Phys.80, 1337~1996!.

5D. Edelson and D. L. Flamm, J. Appl. Phys.56, 1522~1984!.6Various contributions in Plasma Sources Sci. Technol.3, ~1994!, and J.Vac. Sci. Technol. A14, ~1996!.

7 G. S. Oehrlein, J. F. Rembetski, and E. H. Payne, J. Vac. Sci. Techn8, 1199~1990!.

8V. A. Yunkin, I. W. Rangelow, J. A. Schaefer, D. Fischer, E. Voges, aS. Sloboshanin, Microelectron. Eng.23, 361 ~1994!.

9K. Kurihara and M. Sekine, Plasma Sources Sci. Technol.5, 121 ~1996!.10M. Haverlag, W. W. Stoffels, E. Stoffels, G. M. W. Kroesen, and F. J.

Hoog, J. Vac. Sci. Technol. A14, 384 ~1996!.11M. Haverlag, E. Stoffels, W. W. Stoffels, G. M. W. Kroesen, and F. J.

Hoog, J. Vac. Sci. Technol. A12, 3102~1994!.12M. Kitamura, H. Akiya, and T. Urisu, J. Vac. Sci. Technol. B7, 14

~1989!.13J. P. Booth, G. Hancock, N. D. Perry, and M. J. Toogood, J. Appl. Ph

66, 5251 ~1989!; N. St. Braithwaite, J. P. Booth, and G. Cunge, PlasmSources Sci. Technol.5, 677 ~1996!.

14R. J. M. M. Snijkers, Ph.D. thesis, Eindhoven University of TechnoloEindhoven, The Netherlands, 1993.

15D. C. Gray, I. Tepermeister, and H. H. Sawin, Proceedings of the SecInternational Symposium on Reactive Plasmas and 11th SymposiumPlasma Processing, edited by T. Goto, Yokohama, Japan, publisheOrganizing Committee of ICRP-2/SPP-11, 1994, p. 751.

-

ndy

adla

ibae

rt

n,

of

pl.

95 Stoffels, Stoffels, and Tachibana: Polymerization of fluorocarbons 95

16J. P. Booth, G. Cunge, P. Chabert, F. Neuilly, J. Derouard, and N. Sghi, Proceedings of the Third International Conference on Reactive Pmas and 14th Symposium on Plasma Processing, edited by K. Tachand Y. Watanabe, Nara, Japan published by Organizing CommitteIRCP-3/SPP-14, 1997, p. 75.

17E. Stoffels, W. W. Stoffels, K. Tachibana, Rev. Sci. Instrum.~in press!.18W. W. Stoffels, E. Stoffels, and K. Tachibana, Jpn. J. Appl. Phys., Pa

36, 4638~1997!.19S. R. Hunter and L. G. Christophorou, J. Chem. Phys.80, 6150~1984!.

JVST A - Vacuum, Surfaces, and Films

e-s-naof

1

20E. Stoffels, W. W. Stoffels, D. Vender, M. Haverlag, G. M. W. Kroeseand F. J. de Hoog, Contrib. Plasma Phys.35, 331 ~1995!.

21L. J. Overzet, Y. Lin, and L. Lou, J. Appl. Phys.72, 5579~1992!.22E. Stoffels and W. W. Stoffels, Ph.D. thesis, Eindhoven University

Technology, Eindhoven, The Netherlands, 1994.23A. A. Howling, L. Sansonnens, J.-L. Dorier, and Ch. Hollenstein, J. Ap

Phys.75, 1340~1994!.24S. M. Spyrou, I. Sauers, and L. G. Christophorou, J. Chem. Phys.78,

7200 ~1983!.