plasma diagnostic study of silicon nitride film growth in

Plasma diagnostic study of silicon nitride film growth in aremote Ar-H2-N2-SiH4 plasma : role of N and SiHn radicalsCitation for published version (APA):Kessels, W. M. M., Assche, van, F. J. H., Hong, J., Schram, D. C., & Sanden, van de, M. C. M. (2004). Plasmadiagnostic study of silicon nitride film growth in a remote Ar-H2-N2-SiH4 plasma : role of N and SiHn radicals.Journal of Vacuum Science and Technology A: Vacuum, Surfaces, and Films, 22(1), 96-106.https://doi.org/10.1116/1.1631294

DOI:10.1116/1.1631294

Document status and date:Published: 01/01/2004

Document Version:Publisher’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 version of the article upon submission and before peer-review. There can beimportant differences between the submitted version and the official published version of record. Peopleinterested in the research are advised to contact the author for the final version of the publication, or visit theDOI 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 pagenumbers.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.

If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, pleasefollow below link for the End User Agreement:www.tue.nl/taverne

Take down policyIf you believe that this document breaches copyright please contact us at:[email protected] details and we will investigate your claim.

Download date: 03. Jan. 2022

https://doi.org/10.1116/1.1631294

https://doi.org/10.1116/1.1631294

https://research.tue.nl/en/publications/9ecf2370-0e44-4fb7-b91b-fa61972c43b8

Plasma diagnostic study of silicon nitride film growth in a remoteAr–H2 – N2 – SiH4 plasma: Role of N and SiH n radicals

W. M. M. Kessels,a) F. J. H. van Assche, J. Hong, D. C. Schram,and M. C. M. van de Sandenb)

Department of Applied Physics, Center for Plasma Physics and Radiation Technology, Eindhoven Universityof Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlands

~Received 11 February 2003; accepted 6 October 2003; published 25 November 2003!

A remote expanding thermal plasma operated on an Ar–H2– N2– SiH4 mixture has been studied byseveral plasma diagnostics to obtain insight into the plasma processes and the hydrogenatedamorphous silicon nitride (a-SiNx :H) growth mechanism from the N2– SiH4 reactant mixture.From Langmuir probe measurements, ion mass spectrometry, and threshold ionization massspectrometry, it is revealed that the Ar–H2– N2 operated plasma source leads mainly to N and Hradicals in the downstream region. The H radicals react with the SiH4 admixed downstream creatinga high SiH3 density as revealed by cavity ringdown spectroscopy. By cavity ringdownmeasurements, it is also shown that Si and SiH have a much lower density in the downstreamplasma and that these radicals are of minor importance for thea-SiNx :H growth process. Theground-state N radicals from the plasma source do not react with the SiH4 injected downstreamleading to a high N density under thea-SiNx :H deposition conditions as revealed by thresholdionization mass spectrometry. From these results, it is concluded that N and SiH3 radicals dominatethe a-SiNx :H growth process and the earlier proposed growth mechanism ofa-SiNx :H from theN2– SiH4 mixture @D. L. Smithet al., J. Vac. Sci. Technol. B8, 551~1990!# can be refined: Duringdeposition, ana-Si:H-like surface layer is created by the SiH3 radicals and at the same time thisa-Si:H-like surface layer is nitridated by the N radicals leading toa-SiNx :H formation. This growthmechanism is further supported by the correlation between the SiH3 and N plasma density and theincorporation flux of Si and N atoms into thea-SiNx :H films as deposited under various conditions.© 2004 American Vacuum Society.@DOI: 10.1116/1.1631294#

bidca

gsndnieoonthtoonicusoip

is,x

asreicig

resd

s.the

heyen-

cessss

de-tin-ces-d

abletes

mor-

ulkonth

asm

I. INTRODUCTION

The deposition process of silicon nitride depositedplasma-enhanced chemical vapor deposition has been winvestigated as silicon nitride films have numerous applitions in the microelectronics industry.1 However, applica-tions of silicon nitride films keep emerging and this brinforth different demands on the properties and process cotions of the film. For example, plasma deposited silicontride films are currently emerging as multifunctional antirflection coatings on the next-generation crystalline silicsolar cells.2 Besides antireflection performance, the silicnitride films induce bulk passivation of defect states inlower-quality silicon stock material that is currently usedreduce the price of photovoltaic electricity. Moreover, silicnitride can also induce surface passivation of the cells whincreases the efficiency of a cell. In the photovoltaics indtry, it is, however, essential to obtain a sufficiently high prduction throughput while keeping the investments in equment relatively low. The deposition rate of silicon nitridetherefore, a critical parameter for this industry. Another eample of an emerging application of silicon nitride films isencapsulation layers for protection of devices from moistuAlthough this application itself is not new, a class of organmaterial-based devices is at stake and this requires h

a!Electronic mail: [email protected]!Electronic mail: [email protected]

96 J. Vac. Sci. Technol. A 22 „1…, Jan ÕFeb 2004 0734-2101Õ200

yely-

i---n

e

h-

--

-

.

h-

quality encapsulation layers deposited at low temperatu~below ;100 °C!. Important examples are polymer- anorganic-based light emitting devices for flexible display3

Another issue for these nonsemiconductor industries, isuse of precursor gases other than NH3 and SiH4 . Althoughso far usually the best film properties~for the microelectron-ics industry! are obtained from these precursor gases, trequire extensive investments with respect to safety andvironmental aspects that would rather be avoided.

These demands on the properties and production proof the film, together with the fact that the deposition proceof silicon nitride has not yet been completely unraveledspite the significant progress in the last decade, make conued research of the deposition process of the material nesary. In this article, silicon nitride film growth will be studiefor the expanding thermal plasma~ETP! using N2 and SiH4

as precursor gases. This remote plasma has shown to beto deposit different materials at very high deposition rawhile maintaining good film properties.4,5 Recently, it hasbeen shown that the ETP can produce hydrogenated aphous silicon nitride (a-SiNx :H) films at deposition rates upto 20 nm/s and that these films are very well suited as bpassivating antireflection coatings for multicrystalline silicsolar cells.6 This result has been obtained for boNH3– SiH4 and N2– SiH4 mixtures.7

Although the deposition process of silicon nitride hmostly been studied for silicon nitride deposited fro

964Õ22„1…Õ96Õ11Õ$19.00 ©2004 American Vacuum Society

oithe-

t

etio

hee

rflm

thiodio

N

ndosd

he

iaunRw–3th

o

cN

te

wn

ltsrtle

isten-

a isn

intos in

g.al to

heallyd and

o-theeterofa-y

ron

ingleandhe

at a

97 Kessels et al. : Plasma diagnostic study of silicon nitride film 97

NH3– SiH4 plasmas, some studies have carried outN2– SiH4 plasmas. Specifically, the seminal work of Smand co-workers8–10 has led to considerable insight into thdeposition mechanism ofa-SiNx :H films. From mass spectrometry and film studies, Smith and co-workers10 pointedout that the formation ofa-SiNx :H is not governed by Si–Nprecursor species from the plasma when using N2. Instead,they proposed that the deposition process proceeds byreaction of silane radicals (SiHn) and nitrogen radicals~N! atthe surface. This mechanism has been concluded from msurements of N radicals in the plasma by threshold ionizamass spectrometry~TIMS!. Silane radicals SiHn were notexperimentally observed in their study but their role in tdeposition process was inferred from the deposition procof hydrogenated amorphous silicon (a-Si:H) from SiH4

plasmas. Other research groups have focused on the sureactions and the evolution of the film properties during figrowth, e.g., byin situ infrared spectroscopy on the film.11,12

Although these studies reveal important information on~sub!surface reactions, there is usually no direct informaton the plasma species present. Therefore, in these stuone can only speculate on the nature of the radicals andarriving at the substrate during film deposition.

In this article, we will focus on the determination of theradical and SiHn radical densities in the Ar–H2– N2– SiH4

operated ETP. The N radicals have been investigatedTIMS while the densities of the silane radicals Si, SiH, aSiH3 have been determined from cavity ringdown spectrcopy ~CRDS!. The SiH2 radicals have not been investigateby cavity ringdown in the ETP setup so far, however tinformation on the Si, SiH, and SiH3 densities provide asufficient basis to draw conclusions about the SiHn radicalchemistry in the plasma and the correspondinga-SiNx :Hgrowth mechanism. The results obtained by the plasma dnostics will be compared to the film properties obtainedder various plasma conditions as have been reported in6. As mentioned earlier, the deposition rates associatedthese ETP conditions are very high, i.e., in the range of 4nm/s. Finally, we will show that our results are in line withe results obtained by Smith and co-workers8–10 in a con-ventional rf plasma and from the SiHn radical densitiesfound, we will refine the proposed growth mechanisma-SiNx :H.

In Sec. II, the ETP setup and the plasma diagnostic teniques used will briefly be described. The detection ofradicals by TIMS will be addressed in more detail~Sec. II C!.In Sec. III, the results of the investigations will be presenand discussed. First, the reactive species emanating fromplasma source of the ETP will briefly be addressed~Sec.III A ! as these species induce the dissociation of the dostream injected SiH4 . Section III B contains the results othe N radical and the SiHn radical density in the downstreamplasma. Some aspects of the growth mechanisma-SiNx :H will be discussed in Sec. IV on the basis of resureported in this study as well as on the basis of data repoin literature. In Sec. V, the main conclusions of this articwill be summarized.

JVST A - Vacuum, Surfaces, and Films

n

he

a-n

ss

ace

enies,ns

by

-

g--ef.ith5

f

h-

dthe

n-

of

ed

II. EXPERIMENT

A. Expanding thermal plasma deposition setup

A diagram of the ETP with the plasma diagnosticsgiven in Fig. 1. The ETP setup has been described exsively in previous publications4,6 and, therefore, we will onlybriefly address the main characteristics fora-SiNx :H depo-sition.

In a dc operated cascaded arc plasma source, a plasmcreated in an Ar–H2– N2 gas mixture. This plasma creatiotakes place at subatmospheric pressure~typically 300 Torr!and at the exit of the plasma source, the plasma expandsa low-pressure deposition chamber. The reactive speciethe plasma can subsequently react with the SiH4 gas that isinjected into the downstream plasma by an injection rinThe plasma operating conditions have been chosen equthose used in Ref. 6, which reports about thea-SiNx :H filmproperties obtained by the Ar–H2– N2– SiH4 operated ETP.Briefly, the gas flows used are 55 standard cubic cm3/s ~sccs!Ar for all conditions, 0–5 sccs H2, 0–10 sccs N2, and 0–15sccs SiH4 . The pressure in the downstream region is in trange of 130–150 mTorr and the cascaded arc is typicoperated at a current-controlled arc current of 45 A anvoltage of 230 V. For convenience, the deposition rate aSi/N atomic ratio obtained for various SiH4 flows are repro-duced in Fig. 2 from Ref. 6.

The a-SiNx :H films have been deposited at samples psitioned at a substrate holder but, for the present study,substrate holder has been replaced by a mass spectromwith a similar geometry. The reactor is pumped by a stacktwo roots blowers and a forepump with a high pumping cpacity ~2600 m3/h!. Overnight, the reactor is pumped bmeans of a turbopump reaching a base pressure of 1026 Torr.

B. Probe measurements and ion mass spectrometry

The density of ions and electrons, as well as the electtemperature for the nondepositing Ar–H2– N2 plasma ema-nating from the plasma source, has been measured by sLangmuir probe measurements. The probe configurationanalysis procedure are described in detail in Ref. 13. TLangmuir probe measurements have been carried out

FIG. 1. ETP setup with plasma diagnostics.

oxr

ee1

305lyo

ytsanispoee

onal

thsh

e

ofic

ereIMthe

ofhe

toillas

arehe

-

aree 3argenergy’’the

at-the

s nol-

rvedn-

f

The

er

mi

r therlyhighon-leand


distance of 6 cm from the cascaded arc nozzle, i.e., apprmately at the position where the SiH4 gas is injected undea-SiNx :H deposition conditions.

Information on the ions present in the plasma has bobtained by ion mass spectrometry as described in Ref.The mass spectrometer used is a Hiden Analytical EPIC~Hiden Analytical Ltd., 420 Europa Blvd., Warrington WA7UN, UK!. The ions in the plasma are extracted effusiveinto the mass spectrometer by a knife-edged extractionfice with a diameter and thickness of 50mm. By means of aturbopump with a pumping capacitySof 56 l/s, the pressurein the mass spectrometer is kept below 1026 Torr duringplasma operation~at ;150 mTorr!. The ions are separated btheir mass-over-charge ratio@expressed in atomic mass uni~amu!# in a triple-stage mass filter and detected by a chneltron ~DeTech 415!. The mass range of the mass filter300 amu. Ion mass spectrometry is applied for the nondeiting Ar–H2– N2 plasma as well as for one condition of thdepositing Ar–H2– N2– SiH4 plasma. In the latter case, thmeasurement time~integration time per step in amu! taken isvery short to avoid the influence of clogging of the extractiorifice by film deposition. However, this reduces the signto-noise ratio of the measurement.

C. Threshold ionization mass spectrometry

The presence and density of ground-state N radicals inplasma have been investigated by mass spectrometry, uthe so-called TIMS or appearance potential technique. Ttechnique has also been applied by Smith and co-workers8–10

in their studies of thea-SiNx :H deposition mechanism. Inthe present study, a Hiden Analytical EPIC 300 mass sptrometer@with ‘‘plasma sampling mass spectrometer~PSM!upgrade’’ for TIMS#14 has been installed at the positionthe substrate holder. This enables the detection of the radspecies contributing to film growth.

The gas extraction takes place through the same orificdescribed in Sec. II B, as are the pumping capacity and psure in the mass spectrometer. Under operation in the Tmode, gas species are ionized by electrons emitted by ariated iridium filament located in the ionizer region of th

FIG. 2. Deposition rate and Si/N atomic ratio of thea-SiNx :H films as afunction of the SiH4 flow. The data~taken from Ref. 6! are obtained forthese plasma conditions: 55 sccs Ar, 10 sccs N2 , 5 sccs H2 , 45 A arccurrent, and 150 mTorr. The error bars for the N/Si ratio are an upper li

J. Vac. Sci. Technol. A, Vol. 22, No. 1, Jan ÕFeb 2004

i-

n3.0

ri-

-

s-

-

eingis

c-

al

ass-So-

mass spectrometer. This ionizer region is situated ‘‘linesight’’ at a distance of 24 mm from the extraction orifice. Telectron emission currentI e is controlled at a relatively lowcurrent ~2–20 mA! to avoid space-charge buildup andkeep the filament temperature low. The low temperature wminimize thermal decomposition of plasma species, suchN2 at the filament. From the ionizer, the ionized speciesled through a Bessel box type of energy filter followed by ttriple-stage mass filter and the channeltron detector.

The ionization of the ground-state N radicals (N1e→N112e, Ei515.2 eV) is distinguished from the dissociative ionization of N2 molecules (N21e→N11N12e, Ea

525.3 eV) in the ionizer by scanning the electron energyEof the electrons emitted by the filament. The electronsmonoenergetic with a spread of less than 0.5 eV. Figurshows a typical electron energy scan at a mass-over-chratio of 14 amu for N radicals. The ‘‘plasma on’’ conditioshows a clear hump at energies above the threshold enEi for the ionization of N radicals. For the ‘‘plasma offcondition, a large signal appears for energies abovethreshold energyEa for dissociative ionization of N2 intoN1. The small signal at lower electron energies can betributed to the ionization of N radicals that are created byaforementioned thermal decomposition of N2 at the hot fila-ment in the mass spectrometer. Furthermore, Fig. 3 giveindication for dissociative ionization of excited nitrogen moecules in the mass spectrometer, which would be obseby an additional hump in the signal for the plasma on codition in the electron energy region of 15–25 eV.

The densityn of N radicals is derived from the signal othe mass spectrometer~corrected for the small contributionto the signal by thermal decomposition! by

I ~E!5aI es~E!n with Ei,E,Ea , ~1!

in which s(E) is the cross section for ionization anda is thegeometry and mass dependent proportionality constant.determination of the unknown constanta is avoided by link-ing the signal due to N radicals to a signal due to oth

t.

FIG. 3. Electron energy scan at a mass-over-charge ratio of 14 amu fodetermination of the N radical density. The plasma on condition cleashows the signal created by N radicals at low energies. The signal atenergies is lower for the plasma on condition than for the plasma off cdition due to the depletion of N2 by the plasma. The electron energy scahas been recalibrated using the threshold energy for ionization of Arother noble gases~see Ref. 14!.

hion

reurioiain

ityof

-a

ne

ntascieua

athx

np-

n

r

e

hehare

tht

cte

N

-

rgehet inhistateionr

S,rp-

andis

In-sideeteeirtheare

adi-o

dueon

trateeen

edthe

im--de-

asto

a


species with a known number density in the plasma. For twe have taken the Ar signal during the plasma off conditifollowing the procedure proposed by Singhet al.15 Singhet al.15 pointed out that it is necessary to use the signal cated by ionization of a monoatomic species in this procedThe application of a signal created by dissociative ionizat~e.g., N2) will introduce considerable error because dissoctively ionized species have a different extraction efficiencythe mass filter.

In the procedure of quantifying the N atom radical densby using the known density of Ar atoms in the plasmacondition, differences in the ionization cross sections~takenfrom Refs. 16, 17, and 18!, as well in mass filter transmittance and channeltron detector efficiency, are taken intocount. The product of mass filter transmittance and chantron detector efficiency has am21 dependence on the massmof the analyzed species.13,15 Furthermore, the difference insurface reactivity for Ar and N atoms needs to be taken iaccount. Both Ar and N atoms enter the ionizer of the mspectrometer by the effusively extracted beam. Ar specan however have wall collisions inside the ionizer withobeing lost at the walls and, therefore, Ar can build upbackground density in the ionizer. N radicals are more retive than Ar, and the N radicals can possibly react ationizer walls. Depending on the surface reactivity of N, epressed by the surface loss probabilitygN , the N radicalscan only build up a smaller background density or evenbackground density in the mass spectrometer. For examin the case thatgN51, the N radicals will have no background density componentnion,back in the ionizer but only adensity component in the effusive beamnion,beam. As de-scribed in Ref. 14, in this specific case ofgN51, the Nradical density derived from the comparison with the Ar desity needs to be corrected by the correction factorC:

C5nion,beam1nion,back

nion,beam. ~2!

For the particular mass spectrometer used, this factoC5;35.14 When the surface loss probabilitygN of N radicalsin the ionizer is smaller than unity, Eq.~2! no longer holdsand the factorC is smaller. The lower limit isC51, in thiscase, the N radicals are as unreactive with the wall as thatoms. To determine the correction factorC for our measure-ments, information aboutgN on stainless steel is needed. Tsurface loss probability of N radicals on stainless steelbeen reported in Refs. 19 and 20 for different pressuThese values are shown in Fig. 4. AlthoughgN is only a fewpercent at typical plasma operating pressures, it is cleargN increases for decreasing pressures. On the basis oftrend, we assume, therefore, thatgN51 in the ionizer con-sidering the fact that the pressure in the ionizer is 1026 Torr.This assumption is also in line with the fact that N reaeasily at stainless steel unless the stainless steel is covwith a ‘‘protective’’ N2 layer formed over the saturatedlayer adsorbed at the stainless steel. This N2 layer has agreater fractional surface coverage at higher pressures.19 Inconclusion, we can assume thatC5;35. We need to men


s,,

-e.n-

f

c-l-

oss

t

c-e-

ole,

-

Ar

ss.

athis

sred

tion that this assumption is rather crude and leads to a lasystematical uncertainty in the N radical density in tplasma as determined by TIMS. However, as pointed ouRef. 21, the N radical density determined by TIMS using tassumption is in good agreement with the ground-sN(4S) density determined by the two-photon absorptlaser-induced fluorescence~TALIF ! method applied undesimilar conditions.22

D. Cavity ringdown spectroscopy

Silane radicals SiHn have been measured by CRDwhich is a very sensitive and relatively easy-to-apply absotion spectroscopy technique.23 The high sensitivity is ob-tained because the technique is based on multipassingdue to the fact that the change in absolute light intensitynot measured like in traditional absorption spectroscopy.stead, the decay rate of a short laser light pulse trapped ina high-quality optical cavity, created by two highly reflectivmirrors, is measured. For an ‘‘empty’’ cavity, the decay radepends on the reflectivity of the mirrors as well as on thdistance. When an absorbing medium is created insidecavity, e.g., by plasma ignition such that plasma radicalsgenerated, the decay rate of the laser light~with a wave-length corresponding to the absorption spectrum of the rcal! in the cavity will be enhanced. Comparison of the twdecay rates directly yields the magnitude of absorptionto the radicals. From this absorption value, informationthe density of the radicals can be extracted.

For the present work, the radicals Si, SiH, and SiH3 havebeen measured at a distance of 3.5 cm from the subsholder. The experimental details and procedures have bdescribed in detail in Refs. 24 and 25 fora-Si:H depositionconditions and, therefore, only the basics will be describhere. The Si radicals have been detected by scanninglaser over the Si(4s 3P1←3p2 3P1) transition at 251.9 nm.The Si absorption peak in this wavelength region is superposed on the broadband SiH3 absorption spectrum as discussed below. The total ground-state Si density has beentermined by assuming thermal equilibrium for a gtemperature of 1500 K.25 Special attention has been given

FIG. 4. Surface loss probabilitygN of N radicals on stainless steel asfunction of the partial pressure of N2 as reported by Adams and Miller~seeRef. 19! and Singhet al. ~see Ref. 20.!

uro

e

dnatio

th

fre

-luth

ot

esntha

ul. IrN

w-spion

andor

ciesure-tion

ns

5ob-elyam

ny-nceo-le inthe

byc-

tirio

d ionor the


the laser pulse energy used in order to avoid optical sattion of the transition. SiH radicals have been measuredthe A 2D←X 2P electronic transition around 414 nm. Thlaser has been scanned over theQ1(11.5) rotational line at413.46 nm and the total SiH density has been determineusing a rotational SiH temperature of 1500 K. This rotatiotemperature has been estimated from the relative distribuof the Q1(14.5) andR2(1.5) rotational lines of SiH whichwas determined for some characteristic conditions ofAr–H2– N2– SiH4 plasma.25 The SiH3 radicals have beenmeasured by theA 2A1←X 2A1 electronic transition. Thecorresponding absorption band is featureless and ranges;200 to;260 nm~due to the predissociative nature of thexcited state! with a maximum around 215 nm.26 The SiH3

radicals in the Ar–H2– N2– SiH4 plasma have been measured around 217 nm and 250 nm and the absorption vaat these wavelengths measured are shown in Fig. 5, togewith the SiH3 absorption spectrum reported by Lightfoet al.26 The absorption values for thea-SiNx :H depositingplasma fit well to the SiH3 spectrum, just as is the case of thAr–H2– SiH4 plasma fora-Si:H growth. This strengthenthe conclusion that, indeed, SiH3 is measured in absorptioand that the measurement is not significantly disturbed bypresence of clusters or particles. As discussed in Refs. 2425, absorption and light scattering by clusters/particles coin principle, lead to an apparent ‘‘broadband absorption’’would, however, be expected that the scattering behaviosignificantly different for Si clusters/particles than for Si–clusters/particles.

III. RESULTS AND DISCUSSION

A. Plasma source: Downstream Ar–H 2 – N2 plasma

Understanding of the downstream plasma processesSiH4 requires information on the Ar–N2– H2 plasma emanating from the cascaded arc plasma source. The reactivecies emanating from this source will initiate the dissociatand ionization reactions of the injected SiH4 . These reactive

FIG. 5. Parts of the normalized SiH3 absorption spectrum as measuredCRDS in ana-SiNx :H anda-Si:H depositing plasma. The absorption spetrum reported by Lightfootet al. ~see Ref. 26! is given for comparison andis in arbitrary units. The inset shows the correlation between the absorpvalues measured at 217 nm and 251.8 nm as obtained under vaAr–H2– N2– SiH4 plasma conditions.


a-n

byln

e

om

eser

endd,tis

ith

e-

species have already been extensively addressed for ArAr–H2 plasmas13,27and here we will consider the species fAr–N2 and Ar–N2– H2 plasmas.

First, we concentrate on the electrons and ionic spewhich have been investigated by Langmuir probe measments and ion mass spectrometry. Figure 6 gives informaon the electron temperatureTe , electron densityne and iondensityni , and the relative abundances of the different ioin an Ar–N2 and Ar–N2– H2 plasma. The Ar–N2 plasma isgenerated with an Ar flow of 55 sccs and a N2 flow of 10sccs while the Ar–N2– H2 plasma is obtained by addingsccs H2 to these conditions. The electron temperaturetained from the Langmuir probe measurements is relativlow for both plasmas, as is generally found for downstreplasmas in the ETP technique.13 This low Te is caused by thefact that no power is coupled to the plasma electrons amore in the downstream region. An important consequeof the low Te is that electron-induced ionization and dissciation processes of precursor gases are very improbabthe downstream region. The electron temperature forAr–N2– H2 plasma is somewhat lower than for the Ar–N2

plasma~for which Te is slightly higher than for a pure Ar

onus

FIG. 6. Electron temperatureTe , electron densityne and ion densityni , andthe relative abundancy of the different ions in~a! the Ar–N2 and ~b! theAr–N2– H2 plasma as determined by Langmuir probe measurements anmass spectrometry. The data in the ion spectra have been corrected fm21 dependence of the mass spectrometer for the ion massm. The N2 flowis 10 sccs and the H2 flow for the Ar–N2– H2 plasma is 5 sccs.

reis-

e

eo

du

on

altsona

seh

sathamn

e

omeedeeom

en

ile

ugig

aha

or

even

wellre 7N

erce,

-.

che

theiblyr anthe

h-di-dta

in

bi-n-

ard-


plasma!.13 Such a decrease inTe when H2 is admixed hasalso been observed in previous work.13

The electron and ion densitiesne and ni in the Ar–N2

plasma are very high and only slightly lower than for a puAr plasma.13 It is, therefore, expected that ion-induced dsociation of SiH4 will be very important in an Ar–N2– SiH4

plasma. This expectation is based on the equivalent casan Ar–SiH4 plasma for deposition ofa-Si:H.13,14 In theAr–SiH4 plasma, ion-induced SiH4 dissociation reactions arpredominant and this leads generally to relatively poa-Si:H material properties. High-qualitya-Si:H is only ob-tained whenne andni are significantly reduced by adding H2

to the Ar in the cascaded arc plasma source. A drastic retion in ne and ni is also observed in Fig. 6: When H2 isadded to the Ar–N2 plasma in the cascaded arc, the idensity is reduced roughly by a factor 10. The value forne

andni in the Ar–N2– H2 plasma is, furthermore, about equto the value forne andni in the Ar–H2 plasma as obtained aa distance of;6 cm from the plasma source. Moreover, alfor a-SiNx :H films, an improvement in film quality has beeobserved when admixing H2 in the cascaded arc plasmsource~i.e., an Ar–N2– H2 plasma! compared to workingwith only an Ar–N2 plasma.6

Figure 6 shows the relative abundance of the ions prein the plasmas as measured by ion mass spectrometry. Tmeasurements are performed at a distance of;37 cm fromthe plasma source and, therefore, the data do not necesrepresent the relative abundancy of ions at the position ofinjection ring. The ‘‘composition’’ of the ions in the plasmis expected to change when the ions flow from the plassource to the mass spectrometer due to gas phase reactiothe plasma. In the Ar–N2 plasma and at the position of thmass spectrometer/substrate holder, argon ions, Ar1, havethe highest density followed by atomic nitrogen ions, N1,and molecular nitrogen ions, N2

1 @Fig. 6~a!#. Furthermore,hydrogenated nitrogen ions (NHx

1 and N2Hx1) are observed

in which the hydrogen atoms are expected to originate frresidual H2 in the reactor. Residual H2 is present becausmeasurements with H2 have been carried out throughout thion mass spectrometer experiments and no special procefor pumping down the reactor has been applied in betwthe measurements. Furthermore, some signal due to athydrogen ions H1 is observed for the Ar–N2 plasma.

When a H2 flow of 5 sccs is admixed to the Ar–N2plasma@Fig. 6~b!#, the ion composition in the plasma~againin front of the mass spectrometer! changes drastically. Thesignal due to Ar1 disappears similar to what happens whgoing from a pure Ar plasma to a Ar–H2 plasma.13 One typeof ion, NH4

1 , is predominantly present in the plasma whthere is also a small signal due to N2H2

1 ions. These ionsmay be created by ion–molecule reactions of N1 and N2

1

ions emanating from the plasma source and moving throthe H2 background density in the reactor. Consequently, F6~b! does not exclude the possibility that N1 is mainlypresent in the Ar–N2– H2 plasma at the position of the SiH4

injection ring. Very remarkable, however, is that the signdue to H1 has completely disappeared despite the fact t


of

r

c-

ntese

rilye

as in

urenic

h.

lt,

in this case, H2 is actively admixed to the plasma source. Fcomparison, H1 ions are the dominant ions in an Ar–H2

plasma used for the deposition ofa-Si:H.As explained in Sec. III C, the N radical density in th

plasma has been measured by TIMS. The N density is giin Fig. 7 as a function of N2 flow for the Ar–N2– H2 plasmawhile one data point is given for the Ar–N2 plasma~for a N2

flow of 10 sccs!. We concentrate on the Ar–N2– H2 plasmabecause this leads to lower electron and ion density asas to better material properties as addressed above. Figushows that the N radical density increases with increasing2

flow. The highest N density is reached at 10 sccs N2 and isequal to;1531012cm23. This high density shows that thcascaded arc plasma source is an efficient N radical soueven when it is operated with an Ar–N2– H2 mixture. Whenoperated with Ar–N2, the N density is higher, but the difference with the Ar–N2– H2 plasma is less than a factor of 2Considering the much lower ion density for the Ar–N2– H2

plasma, the ratio of N radical density to ion density is muhigher for the Ar–N2– H2 plasma and, therefore, thAr–N2– H2 plasma is preferred fora-SiNx :H deposition.6

Atomic H radicals are other species that emanate fromAr–N2– H2 operated plasma source and these are possvery important for the downstream plasma processes. FoAr–H2 operated plasma source, it is even found thatplasma source operates mainly as an atomic H source.14,28Asa matter of fact, this fact is exploited for depositing higquality a-Si:H films because the reactions between H racals and SiH4 radicals lead to film growth that is dominateby SiH3 radicals.29 Although we have no experimental daon H radical density for the Ar–N2– H2 plasma, we alsoexpect a considerable flux of H radicals from the sourcethis case, and reactions between H and SiH4 are believed tobe relatively important in the downstream Ar–N2– H2

plasma. We expect that the high flux of atomic H, in comnation with the high ratio of the N density and the ion desity, is the reason for the higher-qualitya-SiNx :H that is

FIG. 7. Density of N radicals in the Ar–H2– N2 plasma in front of thesubstrate holder~mass spectrometer! as determined by TIMS. The H2 flow is5 sccs. For comparison, also the density of N radicals for an Ar–N2 plasmais given for a N2 flow of 10 sccs. The indicated error bars are standdeviations~1s intervals! as determined from multiple data points. Systematic errors, mainly arising from the used correction factorC @Eq. ~2!#, arenot taken into account.

thes

lend

aeyseiHtenaenay

te

om

sin

e

Nlsralo

der.H

ngSi

yet

r

artd

eled

in

nd

ndone

by

iH

.


obtained for Ar–N2– H2– SiH4 plasmas than forAr–N2– SiH4 .

B. Silicon nitride deposition: Ar–H 2 – N2 – SiH4 plasma

From Sec. III A, it can be concluded that the Ar–H2– N2

operated cascaded arc operates mainly as a source of NH radicals and that the influence of ions and electrons indownstream dissociation reactions is limited under thconditions. In line with literature,8,9,31 it is, therefore, ex-pected thata-SiNx :H film growth is dominated by radicaspecies and not by ionic species. Accordingly, first the dsity of N radicals and SiHn radicals will be addressed ansubsequently the ionic species in the Ar–H2– N2– SiH4

plasma will briefly be considered.The N radical density has been investigated by TIMS

the position of the substrate holder and in Fig. 8, it is givas a function of the SiH4 flow. The N density shows a verparticular behavior because the N density first increawhen SiH4 is admixed and decreases again for higher S4

flows. This behavior will be addressed in more detail la~Sec. IV! but first, the relatively high density of N radicals ithe SiH4 plasma itself will be discussed. This high density,well as the fact that the N density increases initially whSiH4 is admixed, suggests that ground-state N radicalsunreactive with SiH4 . This has also been pointed out bPiperet al.30 who found an upper limit for the reaction rabetween ground-state N and SiH4 of ,8310214cm3 s21.This extremely low reactivity of the N radicals with SiH4 isrelated to the fact that the abstraction reaction of H frSiH4 by N radicals is endothermic by;0.45 eV ~comparethis with the abstraction reaction of H from SiH4 by H whichis exothermic by;0.54 eV!.30,31The fact that the N radicalsare unreactive with SiH4 furthermore implies that N radicalin the expanding plasma survive at least up to the powhere they reach the substrate holder. Therefore, surfacactions between N radicals and the depositinga-SiNx :H filmneed to be considered. The extremely low reactivity of theradicals with SiH4 furthermore indicates that the N radicado not significantly contribute to the generation of neutSi–N species in the plasma, which will be addressed be

FIG. 8. Density of N radicals in the Ar–H2– N2– SiH4 plasma as a functionof the SiH4 flow. The N2 flow is 10 sccs and the H2 flow is 5 sccs. Theindicated error bars are standard deviations~1s intervals! as determinedfrom multiple data points.


andee

-

tn

s

r

s

re

tre-

lw.

The SiHn radicals, Si, SiH, and SiH3 , have been detecteby CRDS at a position of 3.5 cm from the substrate holdTheir densities are given in Fig. 9 as a function of the Si4

flow. Figure 9 reveals that the density of SiH3 ~ranging from1012– 1013cm23) is much higher than the density of SiH~ranging from 1010– 1011cm23) and Si ~about 1010cm23).Both SiH3 and SiH show a clear increase with increasiSiH4 flow but the increase levels off at higher flows. Thedensity shows no clear trend with SiH4 flow. Although not allpossible silicon-containing radicals have been consideredfor this plasma, these observations indicate that SiH3 radicalsare very important fora-SiNx :H film growth. Assuming areasonablely high sticking probability of SiH3 radicals on thea-SiNx :H surface~of roughly the same magnitude as foSiH3 on a-Si:H),32 the high SiH3 density indicates that theSi atoms incorporated into the film are for a large pbrought to the surface by SiH3 radicals. The radicals, Si anSiH, have a much smaller contribution toa-SiNx :H filmgrowth: Even if their sticking probability is equal to on~which is about the case for SiH as has been reveaexperimentally33 and which is generally assumed for Si!,32

their density is, by far, too low to explain thea-SiNx :Hgrowth rate. Furthermore, both the SiH3 density and thea-SiNx :H deposition rate increase with increasing SiH4

flow. This fact, which will be considered in more detailSec. IV, also indicates that SiH3 dominatesa-SiNx :H forma-tion. These observations are similar to those fora-Si:H. Forhigh-quality a-Si:H films, there is clear evidence that SiH3

dominates film growth to a very large extent and that Si aSiH have only a minor contribution.14,24,25,29For the presentplasma conditions, it is expected that the SiH3 radicals aremainly created by H abstraction reactions from SiH4 by theH radicals emanating from the plasma source:

H1SiH4→SiH31H2. ~3!

It is, however, not excluded that vibrationally excited ametastable N2 and N atoms also contribute to the productiof SiH3 to some extent.30 The Si and SiH radicals can bcreated by several pathways, e.g., by reactions initiatedions and by sequential H abstraction reactions.25 Si and SiHcan also possibly be created by H abstraction from Sn

FIG. 9. Density of Si, SiH, and SiH3 radicals in the Ar–H2– N2– SiH4

plasma as a function of the SiH4 flow for the same conditions as in Fig. 8The error bars represent the 1s interval.

rm

iH

a

e

Nth.at

ietsi-n

d.

strs

igpectld

c-

de

nono

ec-

ityhef a

.tail

we

in

thisinlyns

ilarl,cu-ni-dN

ac-

wn

byt

hen

andre-

N-nin

masessrre-


radicals by N radicals because these reactions are exothe~abstraction of H by N from SiH3 and SiH2 are exothermicby ;0.28 eV and;0.19 eV, respectively!.30 The latter reac-tions can contribute to the fact that both the Si and Sdensities are higher for the Ar–H2– N2– SiH4 plasma thanfor the Ar–H2– SiH4 plasma under similar plasmconditions.25

On the basis of these results, we postulate at the momthat Si atoms are mainly brought to thea-SiNx :H film bySiH3 radicals while N atoms are brought to the film byradicals. The ions in the plasma have been studied forcondition with 8 sccs SiH4 by ion mass spectrometry. In Fig10, the ion mass spectrum is given. The signal-to-noise ris relatively poor because a very short integration timeused in order to keep the orifice of the mass spectromfrom clogging under this condition with a very high depotion rate~;20 nm/s!. Figure 10 shows that although the ioflux from the cascaded arc is relatively low, these ions leathe production of SinNmHx

1 ions by ion–molecule reactionsThis is similar to the observation of SinHm

1 ions inAr–H2– SiH4 plasmas in the ETP setup used to depoa-Si:H.13 By comparing the ion mass spectrum to specobtained for Ar–H2– SiH4 plasmas, the presence of N atomin the ions is clearly revealed from the ion spectrum in F10. However, the exact assignment of the peaks in the strum is complicated by the fact that Si atoms have exatwice the mass of N atoms. A possible assignment, basethe observations in the Ar–H2– SiH4 plasma, is given in Fig.10. Furthermore, H1 and NH4

1 ions are observed in the spetrum. In summary, Fig. 10 indicates that the SinNmHx

1 ionshave a relatively low contribution toa-SiNx :H film growth,similar to the case ofa-Si:H film growth by the ETP.

Apart from these SinNmHx1 ions, it needs to be mentione

that no neutral Si–N precursor species with a significant dsity have been observed in the Ar–N2– H2– SiH4 plasma.Similar to the results of Smithet al.,8,9 mass spectrometry othe stable gas molecules in the plasma has revealedN-related peaks~mainly N2 and a very small signal due tNH3) and Si-related peaks~mainly SiH4 and a very smallsignal due to Si2H6) as well as peaks due to Ar and H2. This

FIG. 10. Ion mass spectrum~i.e., with ionizer off! for the Ar–H2– N2– SiH4

plasma with a SiH4 flow of 8 sccs SiH4 . The N2 flow is 10 sccs and the H2flow is 5 sccs. The data in the spectrum have not been corrected for them21

dependence of the mass spectrometer for the ion massm.


ic

nt

e

ioser

to

ita

.c-yon

n-

ly

is different from the observations for NH3– SiH4 plasmaswhere neutral Si–N species are found to dominatea-SiNx :Hfilm growth.8,10 It has to be mentioned that the mass sptrometry measurements on the stable gas molecules~i.e., notin the TIMS mode! in the Ar–N2– H2– SiH4 plasma are notvery sensitive and will not reveal the presence of low-densSi–N precursor radicals. A low density of SiN radicals in tplasma might, for example, be indicated by observation osmall ‘‘ghost’’ peak in the SiH spectrum around 414 nm34

This has, however, first to be investigated in more debefore conclusions can be drawn.

IV. DISCUSSION ON THE a-SiNx :HGROWTH MECHANISM

On the basis of the results presented in Sec. III B,postulate that the Si atoms in thea-SiNx :H film are mainlybrought to the substrate by SiH3 radicals, while the N atomsin the film originate from N radicals in the plasma. This isline with the growth mechanism fora-SiNx :H as proposedby Smith et al.8,9 who proposed that SiHn radicals and Nradicals react at the surface to producea-SiNx :H in a rfplasma. This present study extends the insight intogrowth mechanism because we have evidence that maSiH3 is responsible for the Si atoms in the film and that ioand other radicals~such as Si and SiH! play only a veryminor role. Our results are also in agreement with a simreaction mechanism proposed by Hanyaloglu and Aydi12

apart from the fact that they concluded that excited molelar nitrogen (N2* ) species, and not N radicals, cause thetridation of the surface Si—Si bonds. Because they observeno optical emission from N radicals, they inferred that theradical concentration is too low to cause the nitridation retion. Therefore, they suggested that it is more likely that N2*species cause the nitridation because it is very well knothat N2-based plasmas contain various~relatively long-lived!electronically excited N2 species.35 Furthermore, Piperet al.30 have shown that vibrationally excited N2 can causedecomposition of silane. Plasma and surface reactionsexcited N2 ~either electronically or vibrationally excited burepresented in a general way as N2* ) should, therefore, not beunderestimated. On the other hand, N2* species might notsurvive up to the point where they reach the substrate wthey are very reactive with species, such as SiH4 .

Although we cannot exclude a possible influence of N2*species in thea-SiNx :H growth mechanism, we will providemore evidence for the growth mechanism based on NSiH3 radicals as schematically illustrated in Fig. 11: The cation of ana-Si:H-like surface layer on thea-SiNx :H bySiH3 radicals which, at the same moment, is nitridated byradicals. To do so, the SiH3 and N densities will be correlated with thea-SiNx :H film properties, such as depositiorate and N/Si ratio in the film. These film properties, givenFig. 2, as a function of the SiH4 flow for a constant N2 flow,are taken from a previous study using the same plasconditions.6 Figure 2 shows that the deposition rate increawith increasing SiH4 flow while the N/Si ratio decreases. Amentioned earlier, the increase in the deposition rate co

nge

to

le

-

po-theor

for

at

st-he.igh

t for

Nux

e NFig.

s.en-

-werin

s inte N

en

heof

a.

als

y

of


sponds with the increase in SiH3 density for increasing SiH4flow ~Fig. 9!. This can be considered in more detail by usithe information on the Si atomic density in the film for thdifferent conditions.6 From the Si atomic density of the film~in cm23! and the deposition rate~in nm/s!, the Si incorpo-ration flux ~in cm22 s21! can be calculated and correlatedthe SiH3 density in the plasma. This is done in Fig. 12~a!. Itis clear from this Fig. 12~a! that the Si incorporation fluxincreases linearly with increasing SiH3 density. This is also

FIG. 11. Schematic ball-and-stick model of the proposeda-SiNx :H growthmechanism from a N2 /SiH4 plasma. This model is based on the work bSmith et al. ~see Refs. 8, 9, and 12! and in this article.

FIG. 12. ~a! Incorporation flux of Si atoms into thea-SiNx :H film vs theSiH3 density in the plasma and~b! incorporation flux of N atoms into thea-SiNx :H film vs the N density in the plasma. The incorporation fluxesSi and N have been calculated from the data in Fig. 2~and as presented inRef. 6!.


expected when SiH3 radicals are almost solely responsibfor the Si atoms incorporated into thea-SiNx :H film. Thiscan be inferred from:25,29,36

Si incorporation flux5NSiRdep51

4nn

s

12b/2USiH3

, ~4!

with NSi as the Si atomic density in the film,Rdep as thedeposition rate,n as the SiH3 density,n as its thermal veloc-ity, and s andb as the SiH3 sticking and surface loss probability, respectively~with s<b). Moreover, using Eq.~4!, itis possible to check whether the numbers for the Si incorration flux are compatible with the absolute numbers ofSiH3 density. Assuming a thermal velocity of 1000 m/s fSiH3 ~corresponding with a gas temperature of 1500 K!,25

the slope of the fit in Fig. 12~a! is perfectly compatible withvalues ofs50.15 andb50.30. In principle, no informationabout s and b is available fora-SiNx :H, but the valuesmentioned are very close to those usually reporteda-Si:H. For a-Si:H, the reported values ofb usually rangebetween 0.25 and 0.30, while it is generally assumed ths51/2b.32,37 Although the s and b values might differ tosome extent fora-SiNx :H, the agreement strongly indicatethat the SiH3 density and Si incorporation flux are compaible. It provides further evidence that the Si atoms in ta-SiNx :H originate from SiH3 radicals from the plasmaMoreover, this assessment clearly shows that a very hdensity of radicals in the plasma is necessary to accounthe high deposition rate of thea-SiNx :H in the ETP.

For the N atoms, a similar comparison between theradical density in the plasma and the N incorporation flcan be made. However, unlike the SiH3 radicals whose pro-duction increases for higher SiH4 flows, the production rateof N radicals is unaffected by the SiH4 flow. In the presentcase, the N density can only be affected by a change in thloss rate, either by gas phase or surface reactions. In12~b!, the N incorporation rate~calculated from the N atomicdensity in the film and the deposition rate!6 is given as afunction of the N density in the plasma. Figure 12~b! showsthat the N incorporation flux is higher for low N densitieThis behavior can be understood fairly easily: As the N dsity is given by the balance between production~which isconstant in the present case! and loss, a higher N incorporation flux leads to more surface loss and, therefore, to a loN density. In other words, this means that the N densitythe plasma is governed by the surface loss of N radicalthe plasma and, consequently, by reactions that incorporaatoms into thea-SiNx :H film. It should be noted in thisrespect that the N incorporation flux has an upper limit givby the stochiometric limit of thea-SiNx :H (x5N/Si ratio5;4/3!. This consideration sheds also more light on tparticular behavior of the N radical density as a functionSiH4 flow in Fig. 8. When more SiH4 is admixed, more Natoms are incorporated into thea-SiNx :H film and thisleads, therefore, to a lower N radical density in the plasmThis implies that the surface loss probabilitygN ~in principle,we can distinguish for these depositing conditions alsosticking and surface loss probability for the N radica!

tc

pl

ticao

b,sitee

miHxis.ssth

ndieie

he-

edu.ee

s t, bh

or

a

he

ud

an

alswetingro-

matiesma

pe-di-di-anter.theedur-

s-

.

n.alon

e Home

h-

e

re,to

ema-hea

y a

to. It


changes with SiH4 flow. When assuming for simplicity thaall N radicals that are lost at the surface stick at the surfawe can calculate the corresponding value ofgN for the dif-ferent N radical densities/plasma conditions. For examfor a N density of;6031012cm23 ~corresponding with Nincorporation flux of ;631016cm22 s21), we find gN

5;0.04 while for a N density of;12031012cm23 ~corre-sponding with N incorporation flux of;231016cm22 s21),gN5;0.007. AlthoughgN is not known for N radicals dur-ing a-SiNx :H growth, these calculated values are realisMoreover, Fig. 4 indicates that these calculated valuesapproximately equal or smaller than the surface loss prability gN of N radicals at stainless steel for a partial N2

pressure of;20 mTorr. A decrease of the surface loss proability of the N radicals when SiH4 is admixed to the plasmatherefore, also explains the initial increase of the N denwhen SiH4 is admixed in Fig. 8 and when the stainless streactor walls become coated witha-SiNx :H ~see Sec. III B!.

In conclusion, the results on the densities of the plasspecies and, in particular, the correlation between the S3

and N radical density and the Si and N incorporation fluprovide strong evidence for the proposed growth mechanof a-SiNx :H from a N2 /SiH4 plasma as illustrated in Fig11. Furthermore, this correlation yields reasonable valuethe surface loss probability of SiH3 and N radicals and leadto a consistent explanation of the remarkable behavior ofN density as a function of the SiH4 flow ~Fig. 8!. However,we cannot totally exclude a significant role of other reactioand plasma species. Not all silane radicals have been stuwhile also possible Si–N precursor species have not binvestigated with highly sensitive diagnostics. Future studof radicals, such as SiH2 , NH, and NH2 ~all detectable withCRDS! can provide more information on the reactions in tplasma and thea-SiNx :H growth mechanism. The hypothesized role of nitridation reactions by N2* can be clarified bystudying the density of excited N2 molecules and relatingtheir density to the N incorporation flux. It should be nothere that importance of N2* was hypothesized by Hanyalogland Aydil on the basis of optical emission experiments12

Their conclusion is, therefore, only indirect and much morsubject of debate than the more direct evidence on thradical density presented in this article.

Another way to test the proposed growth mechanism iinvestigate the surface reactions of radicals, for exampledetermining the surface loss probability of the radicals. Tpresent study reveals that SiH3 on a-SiNx :H has a surfaceloss probability which is of a similar magnitude as fa-Si:H. This is not unlikely in the case that the SiH3 leads toan a-Si:H-like surface layer. For the N radicals, we findsurface loss probability ona-SiNx :H which is slightly lowerthan for stainless steel. This is fully compatible with tobservations in Fig. 8 but experimental data ongN fora-SiNx :H would resolve the matter. Furthermore, these sties can be coupled toin situ film diagnostic studies ofa-SiNx :H growth. One particular interesting issue that cbe addressed by these studies is the fact that thea-SiNx :Hfilms deposited from N2 and SiH4 are generally relatively


e,

e,

.reb-

-

yl

a

,m

of

e

sed,ens

aN

oy

e

-

rough and have a poor step coverage.9 Smith et al.9 attrib-uted this to high sticking probabilities of both the N and SiHn

radicals, however, this work has indicated that the N radichave a relatively low surface reactivity. In future studies,will address this apparent discrepancy and other interesissues byin situ infrared reflection spectroscopy and spectscopic ellipsometry~combined withex situatomic force mi-croscopy! studies.

V. CONCLUSIONS

Silicon nitride film growth by the remoteAr–H2– N2– SiH4 ETP has been studied by several plasdiagnostics revealing direct information about the densiof species in the plasma, as well as insight into the plasprocesses and the growth mechanism of thea-SiNx :H films.The conclusions can be summarized as follows:

~1! The different plasma diagnostics have been applied. Scial attention has been given to the detection of N racals by TIMS. The determination of the absolute N racal density is complicated by the fact that N radicals cbe lost at the inner walls of the mass spectromeTherefore, only an estimate of the absolute value ofN radical density is possible, but the value obtainshows good agreement with TALIF measurements. Fthermore, the detection of SiH3 radicals in theAr–H2– N2– SiH4 plasma by CRDS has clearly been etablished by probing the broadband SiH3 A 2A1

←X 2A1 transition at two well-separated wavelengths~2! The cascaded arc plasma source operated on Ar–H2– N2

that creates reactive species for SiH4 dissociation, leadsmainly to N and H radicals in the downstream regioFor an Ar–H2– N2 plasma, the ratio between N radicdensity and ion density is very high, while the electrtemperature is too low for SiH4 dissociation by elec-trons. The fact that under these conditions also a largflux emanates from the plasma source is inferred frexperiments on an Ar–H2 plasma under comparablconditions. Furthermore, comparison of the Ar–N2

plasma to the Ar–N2– H2 plasma has revealed a muchigher ion density in the Ar–N2 case which is in agreement with the fact that the properties of thea-SiNx :Hfilms deposited from Ar–N2– SiH4 plasmas are inferior.

~3! The N and SiH3 radicals have a very high density in thdownstream Ar–H2– N2– SiH4 plasma, while the Si andSiH density is relatively low. The latter two species atherefore, not expected to contribute significantlya-SiNx :H growth. The SiH3 radicals are believed to bgenerated by abstraction reactions by the H atoms enating from the plasma source. The N radicals in tSiH4 plasma, originating directly from the plasmsource, do not react with SiH4 . No Si–N species havebeen observed apart from SinNmHx

1 ions which are cre-ated by ion–molecule reactions and which have onlsmall contribution toa-SiNx :H growth.

~4! The N and SiH3 radical densities have been correlatedthe film properties obtained under various conditions

eat-

ofo

eddththhee

arteheth

o

an

hehethe

s

’’er-s

ands,g

s

jk,D

. Jhy

D

de

T., J.

J.aics

. B

C.

ac.

gs,

H.

sma

tion

ls,

C.

m.

sma

, J.l.,

. G.

C.

M.

.

yn,

nol-

ech-

. C.


has been observed that the Si incorporation flux is linin the SiH3 density providing evidence that most Si aoms are brought to the film by the SiH3 radicals. Thecorresponding value of the surface loss probabilitySiH3 shows good agreement with the reported valuesa-Si:H growth and lies, therefore, well in the expectrange fora-SiNx :H. The N density on the other handecreases with increasing N incorporation flux whenSiH4 flow is increased. This can be understood fromfact that the production of N radicals is unaffected by tSiH4 flow, whereas the loss of the N radicals increaswhen more N atoms are incorporated into the film~i.e.,when the N surface loss probability increases!. The cor-responding values of the surface loss probabilitycompatible with the proposed values for stainless sand they provide a straightforward explanation for tremarkable behavior of the N density as a function ofSiH4 flow.

~5! The presented experimental observations provide msupport for the a-SiNx :H growth mechanism fromN2 /SiH4 plasmas as presented in literature. The formtion of thea-SiNx :H takes place by the formation of aa-Si:H-like surface layer by SiHn radicals and, at thesame time, this surface layer is converted intoa-SiNx :Hby nitridation reactions by N radicals. Furthermore, tinsight into this growth mechanism is extended by tpresent study. First of all, there is more support thatnitridation takes place by the N radicals. Second, therclear evidence that the formation of thea-Si:H-like toplayer is dominated by SiH3 radicals and not by radicalsuch as Si and SiH.

ACKNOWLEDGMENTS

This work has been carried out within the ‘‘Sunovationproject which is part of the E.E.T. program of the Nethlands Ministry of Economic Affairs, the Ministry of Education, Culture, and Science, and the Ministry of Public Houing, Physical Planning, and Environment. The researchone of the authors~W.M.M.K.! has been made possible byfellowship of the Royal Netherlands Academy of Arts aSciences~KNAW !. J. P. M. Hoefnagels, A. A. E. Stevenand Y. Barrell are gratefully acknowledged for performinthe CRDS measurements and Dr. W. M. Arnoldbik~UtrechtUniversity! for providing the ERD analysis. The technicianM. J. F. van de Sande, J. F. H. Jansen, A. B. M. Hu¨sken, andH. M. M. de Jong are thanked for their assistance.

1A. Grill, Cold Plasma in Materials Fabrication~IEEE, New York, 1994!.2A. G. Aberle, Sol. Energy Mater. Sol. Cells65, 239 ~2001!.3H. E. A. Huitema, G. H. Gelinck, J. P. B. H. van der Putten, K. E. KuiC. M. Hart, E. Cantatore, P. T. Herwig, A. J. J. M. van Breemen, andM. de Leeuw, Nature~London! 414, 599 ~2001!.

4W. M. M. Kessels, R. J. Severens, A. H. M. Smets, B. A. Korevaar, GAdriaenssens, D. C. Schram, and M. C. M. van de Sanden, J. Appl. P89, 2404~2001!.

5J. W. A. M. Gielen, W. M. M. Kessels, M. C. M. van de Sanden, andC. Schram, J. Appl. Phys.82, 2643 ~1997!; R. Groenen, J. L. Linden,


r

fr

ee

s

eel

e

re

-

eis

-

-of

.

.s.

.

H. R. M. van Lierop, D. C. Schram, A. D. Kuypers, and M. C. M. vanSanden, Appl. Surf. Sci.173, 40 ~2001!.

6W. M. M. Kessels, J. Hong, F. J. H. van Assche, J. D. Moschner,Lauinger, W. J. Soppe, A. W. Weeber, and M. C. M. van de SandenVac. Sci. Technol. A20, 1704~2002!.

7J. Hong, W. M. M. Kessels, F. J. H. van Assche, H. C. Rieffe, W.Soppe, A. W. Weeber, and M. C. M. van de Sanden, Prog. Photovolt11, 125 ~2003!.

8D. L. Smith, Mater. Res. Soc. Symp. Proc.165, 69 ~1990!.9D. L. Smith, A. S. Alimoda, and F. J. von Preisig, J. Vac. Sci. Technol8, 551 ~1990!.

10D. L. Smith, J. Vac. Sci. Technol. A11, 1843~1993!.11A. D. Bailey III and R. A. Gottscho, Jpn. J. Appl. Phys., Part 134, 2172

~1995!.12B. F. Hanyaloglu and E. S. Aydil, J. Vac. Sci. Technol. A16, 2794~1998!.13W. M. M. Kessels, C. M. Leewis, M. C. M. van de Sanden, and D.

Schram, J. Appl. Phys.86, 4029~1999!.14W. M. M. Kessels, M. C. M. van de Sanden, and D. C. Schram, J. V

Sci. Technol. A18, 2153~2000!.15H. Singh, J. W. Coburn, and D. Graves, J. Vac. Sci. Technol. A18, 299

~2000!.16H. C. Straub, P. Renault, B. G. Lindsay, K. A. Smith, and R. F. Stebbin

Phys. Rev. A54, 2146~1995!.17D. Rapp, P. Englander-Golden, and D. D. Briglia, J. Chem. Phys.42,

4081 ~1965!.18K. Stephan, H. Helm, and T. D. Ma¨rk, J. Chem. Phys.73, 3763~1980!.19S. F. Adams and T. A. Miller, Plasma Sources Sci. Technol.9, 248~2000!.20H. Singh, J. W. Coburn, and D. B. Graves, J. Appl. Phys.88, 3748~2000!.21W. M. M. Kessels, J. Hong, E. Langereis, P. J. van de Oever, R. A.

Engeln, and M. C. M. van de Sanden~unpublished!.22S. Mazouffre, I. Bakker, P. Vankan, R. Engeln, and D. C. Schram, Pla

Sources Sci. Technol.11, 439 ~2002!.23Cavity Ring Down Spectroscopy—A New Technique for Trace Absorp

Measurementsedited by K. W. Busch and M. A. Busch~AmericanChemical Society, Washington, 1998!.

24W. M. M. Kessels, A. Leroux, M. G. H. Boogaarts, J. P. M. HoefnageM. C. M. van de Sanden, and D. C. Schram, J. Vac. Sci. Technol. A19,467 ~2001!.

25W. M. M. Kessels, J. P. M. Hoefnagels, M. G. H. Boogaarts, D.Schram, and M. C. M. van de Sanden, J. Appl. Phys.89, 2065~2001!.

26P. D. Lightfoot, R. Becerra, A. A. Jemi-Alade, and R. Lesclaux, ChePhys. Lett.180, 441 ~1991!.

27M. C. M. van de Sanden, J. M. de Regt, and D. C. Schram, PlaSources Sci. Technol.3, 501 ~1994!; R. F. G. Meulenbroeks, R. A. H.Engeln, M. N. A. Beurskens, R. M. J. Paffen, M. C. M. van de SandenA. M. van der Mullen, and D. C. Schram, Plasma Sources Sci. Techno4,74 ~1995!; S. Mazouffre, M. G. H. Boogaarts, J. A. M. van der Mullenand D. C. Schram, Phys. Rev. Lett.84, 2622~2000!.

28M. C. M. van de Sanden, R. J. Severens, W. M. M. Kessels, R. FMeulenbroeks, and D. C. Schram, J. Appl. Phys.84, 2426~1998!.

29W. M. M. Kessels, M. G. H. Boogaarts, J. P. M. Hoefnagels, D.Schram, and M. C. M. van de Sanden, J. Vac. Sci. Technol. A19, 1027~2001!.

30L. G. Piper and G. E. Caledonia, J. Phys. Chem.95, 698 ~1991!.31M. J. Kushner, J. Appl. Phys.71, 4173~1992!.32J. P. M. Hoefnagels, A. A. E. Stevens, M. G. H. Boogaarts, W. M.

Kessels, and M. C. M. van de Sanden, Chem. Phys. Lett.360, 189~2002!.33W. M. M. Kessels, P. R. McCurdy, K. W. Williams, G. R. Barker, V. A

Ventura, and E. R. Fisher, J. Phys. Chem. B106, 2680~2002!.34R. Walkup, Ph. Avouris, R. W. Dreyfus, J. M. Jasinski, and G. S. Selw

Appl. Phys. Lett.45, 372 ~1984!.35G. J. H. Brussaard, Ph.D. dissertation, Eindhoven University of Tech

ogy, The Netherlands.36P. Kae-Nune, J. Perrin, J. Guillon, and J. Jolly, Plasma Sources Sci. T

nol. 4, 250 ~1995!.37W. M. M. Kessels, M. C. M. van de Sanden, R. J. Severens, and D

Schram, J. Appl. Phys.87, 3313~2000!.

plasma diagnostic study of silicon nitride film growth in

Documents