HAL Id: jpa-00246051
https://hal.archives-ouvertes.fr/jpa-00246051
Submitted on 1 Jan 1989
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Reactive ion beam etching of silicon with a new plasma ion source operated with CF4 : SiO2 over Si selectivity
and Si surface modification
C. Lejeune, J.P. Grandchamp, J.P. Gilles, E. Collard, P. Scheiblin
To cite this version:
C. Lejeune, J.P. Grandchamp, J.P. Gilles, E. Collard, P. Scheiblin. Reactive ion beam etching of
silicon with a new plasma ion source operated with CF4 : SiO2 over Si selectivity and Si surface
modification. Revue de Physique Appliquée, Société française de physique / EDP, 1989, 24 (3),
pp.295-308. �10.1051/rphysap:01989002403029500�. �jpa-00246051�
Reactive ion beam etching of silicon with a new plasma ion source operated with CF4 : SiO2 over Si selectivity and Si surface modification
C. Lejeune, J. P. Grandchamp, J. P. Gilles, E. Collard and P. Scheiblin
Institut d’Electronique Fondamentale, Université Paris XI et C.N.R.S. (Unité associée N° 22), Bâtiment 220, 91405 Orsay, Cedex, France
(Reçu le 23 juin 1988, révisé le 21 juillet 1988, accepté le 26 septembre 1988)
Résumé.
2014Nous présentons des résultats de gravure
sous unfaisceau d’ions réactifs délivré par
un nouveautype de canon à ions
2014la Source d’Ions Reflex Electrostatique Maxi-SIRE - alimenté en gaz
CF4 pur. Ils concernent la silice et le silicium monocristallin et démontrent que les conditions d’irradiation peuvent être optimisées de façon à définir un procédé de gravure à la fois très sélectif et anisotrope de la silice vis-à-vis du silicium et qui n’entraîne pas d’altérations irrémédiables du silicium sous-jacent ; les facteurs de
qualité en font une alternative très valable aux procédés actuels sous plasmas CHF3 ou CF4/H2, pour lesquels
les dommages induits par l’hydrogène sont bien connus. Pour le procédé proposé, avec des ions de 500 eV
sousincidence normale, les faits essentiels sont : i)
unesélectivité SiO2/Si de 19 est obtenue pour l’opération de la décharge de source
auvoisinage de
sapression minimale de fonctionnement, ce qui entraîne une très forte fragmentation des neutres injectés ; ii) les vitesses de gravure associées à cette sélectivité sont respectivement
de 130 nm/min et 7 nm/min pour SiO2 et Si, résultats normalisés à une densité de courant d’ions de 1 mA cm-2 ; iii) la couche de blocage fluorocarbonée qui se forme sur le silicium et assure l’atténuation de son
attaque, peut être enlevée par
unsimple bain de 60 s dans l’acide fluorhydrique concentré (50 %) ; iv) ce
traitement laisse un silicium propre dont les qualités électriques
nesont que faiblement altérées vis-à-vis de celles d’un échantillon témoin ; la procédure standard de guérison des dommages, c’est-à-dire
untraitement
enplasma oxygène suivi d’un recuit lent sous azote, semble donc pouvoir dans
cesconditions conduire à de très bons résultats. Des informations concernant les cinétiques et les mécanismes de croissance de la couche de résidu et d’évolution des dégâts superficiels du silicium ont été obtenues grace à des mesures ellipsométriques,
des mesures de caractéristiques électriques de contacts métal-silicium et des spectres d’analyse Auger (en
surface et en profondeur). Les résultats sont rapportés et discutés en mettant en avant les effets associés à la dose d’irradiation par les ions et à la pression de fonctionnement du canon à ions.
Abstract.
2014Reactive Ion Beam Etching is obtained from
a newspecific ion gun, the Electrostatic Reflex Ion Source (Maxi-ERIS), which is operated with pure CF4 gas. The reported results concern both silicon dioxide and single-crystal silicon. They show that the operation of the source discharge down to its minimum pressure which implies
anextensive fragmentation of the injected neutrals, provides
avery convenient process for selective etching of SiO2
overSi,
abasic problem in semiconductor technology. From the characteristic
performances which are achieved, this process appears as
afair alternative solution to the standard reactive ion etching process with CF4/H2 or CHF3 (in
aplasma environment). It is known that these latter ones lead to
deep lying modifications of the Si single-crystal, which are attributed to hydrogen-induced extended defects.
For the proposed RIBE process with
a500 eV beam at normal incidence the main features are : i) selectivity SiO2/Si : 19/1 ; ii) etch rates : 130 nm/min and 7 nm/min, respectively for SiO2 and Si, data normalized to a
1 mA cm-2 current density ; iii) the blocking carbonaceous film which is formed over the silicon and insures the slow-down of the etch rate may be removed by
asimple dip for 60
sin concentrated hydrofluoric acid (50 %) ; iv) such a post-etching treatment
2014without further plasma oxidation or thermal annealing - leaves
a
clean Si substrate, the electrical properties of which
areonly slightly altered
ascompared to a control sample.
Informations about the kinetics and mechanisms of the formation of both the overlayer and the near-surface
damage are obtained from ellipsometry, Auger electron spectroscopy, Auger sputter profiling and metal-
silicon contact electrical measurements. They
arereported and discussed with
aspecial emphasis
onthe effect
of both the ion exposure dose and the operation pressure of the ion gun.
Classification
Physics Abstracts
61.80J
-81.60
-81.60C
Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/rphysap:01989002403029500
1. Introduction.
Selective etching of Si02 over Si is a basic problem in
semiconductor technology. In a pure CF4 discharge
excited with RF power, Si and Si02 are etched at
very similar rates. Therefore anisotropic and selec-
tive etching of silicon dioxide over silicon is generally
achieved by the combination of an ion activated process (Reactive Ion Etching - RIE) in a plasma
environment and the presence of hydrogen either in
the etchant molecule or in the gas mixture [1, 2].
Unfortunately, hydrogen atoms and ions have been
shown to be responsible of damage and contami- nation of the silicon crystal down to depths as large
as 30-50 nm [3-5]. They are very difficult to cure and alternative processes are required. Reactive Ion Beam Etching (RIBE) using pure halocarbon gases may provide a solution [6-10]. However, the selec- tivity has to be further increased and, furthermore the compatibility of the process with the VLSI circuit
requirements has to be investigated. Therefore the contamination and damage of the silicon near-sur-
face have to be analyzed first to determine their effects and their origin and second in order to find a
solution which may restore the surface to a device-
quality state.
In section 2, the experimental apparatus and pro- cedures are described. A specific ion gun has been
developed for RIBE which can be operated with
fluorocarbon gases without lifetime problems and
with reduced contamination. It has a new ionization chamber : the Electrostatic Reflex Ion Source
(ERIS) and a plasma bridge neutralizer. For the RIBE procedure, the neutrals are injected into the
source discharge chamber. For the present ion gun the pressure within this latter chamber may be as low
as 1 x 10-4 mbar whereas the corresponding pres-
sure within the interaction chamber is 1 x
10- 5 mbar. The composition of the ion beam is on-
line recorded ; it will be hereafter referred to as a
« CF+x ion beam
».The gun also delivers a flow of reactive neutrals and radicals which may affect the ion beam-sample interactions.
In section 3, we first report the variations of the etch rates and the resulting Si02 over Si selectivity as
functions of the gun operation pressure ; they are
discussed briefly from a comparison to the associated
variation of the ion beam composition. Then we report results of measurements concerning the ir-
radiation of unmasked Si single-crystal ; they were
devoted to improve the understanding of the
mechanisms which insure : 1) the growth and thick-
ness limitation of the CF-carbonaceous overlayer ; 2) the Si etch rate slow-down, and 3) the Si near-
surface damage and contamination. Both the tran- sient and steady states have been analyzed using the following complementary diagnostics : 1) on-line
variation of the SiF4 partial pressure ; 2) ellipsomet-
ric parameter (0394 2013 03C8) variations ; 3) Auger sputter
profiles (ASP), (100 eV Ar+ sputter beam) ; 4) elec-
trical evaluation of Metal-Si contacts (Mercury-Sili-
con probe diode). Both as-etched and wet-cleaned
samples as obtained after a 60 s dip in HF 50 %,
have been compared. Such a simple post-etching
treatment has been chosen
-instead of the more
standard oxygen plasma treatment followed by a
concentrated HF dip
-in order to minimize the
consumption of the underlying silicon substrate.
The influence of the CFx ion beam energy and incidence has been studied, but in this paper we report and discuss mainly the influence of the gun
operation pressure and the exposure dose effects.
2. Expérimental set-up and procedures.
A schematic diagram of the RIBE apparatus is shown in figure 1.
2.1 THE ION GUN : ELECTROSTATIC REFLEX ION SOURCE (MAXI-ERIS).
-A specific ion gun is
used, the Maxi-ERIS [11]. The source discharge is a
three-electrode structure, with a hot tantalum cathode and a small graphite anode which are both located within the cylindrical source chamber (250 mm diameter). This latter is negatively biased
with respect to the hot cathode and thereby insures
the electrostatic containment (reflex effect) of the primary ionizing electrons which have been initially
accelerated from the cathode. This gun has been
especially developed in order to satisfy the various
requirements of ion-beam assisted processes
-either deposition or etching. In particular a steady-
state operation may be reached for the operation
with fluorocarbon gases which generally leads to the deposition of insulating films on the chamber walls
or electrodes [12]. Because of the small anode the temperature of which may be far above room
temperature no deposition occurs on the anode
which conversely is slowly etched. Its material must be chosen according to the chemically reactive plasma environment. It has to be pointed out that
standard ion sources using a magnetic confinement and a DC potential excitation have a very short lifetime in relation with the extinction of the dis-
charge [8, 13]. With fluorocarbon gases such as
CF4, the hot cathode tantalum wire has been shown
to be passivated after about one hour of operation
due to the formation of tantalum carbide. The beam contamination is then strongly reduced and for typical operation conditions as such reported in this
paper the cathode lifetime is about 100 h [14]. The
extraction optics is a three grid system constructed of stainless steel as the ionization chamber. It may deliver a 7 cm diameter beam in the energy range 0.2-2 keV at current densities of up to 1 mA cm- 2,
as measured on the sample holder located 15 cm
Fig. 1. - Schematic diagram of the RIBE apparatus. For
aRIBE process the neutral gas is injected into the
sourcedischarge chamber. For
aCAIBE process, gases
arealso injected directly into the interaction chamber.
downstream the optics. Graphite may also be used
as a base material for the grids, but not molybdenum
which is etched and thus leads to beam and substrate contamination. The positive ion beam current is
electron compensated on the target in order to allow
the irradiation of insulating material (Si02 ). The
electrons are delivered by an auxiliary discharge
which has the same structure but a much smaller chamber : the Medium-ERIS [15]. It is fed with
argon ; a plasma bridge insures the coupling between
the neutralizer discharge plasma and the ion-beam plasma.
The interaction chamber is pumped by a liquid- nitrogen-trapped diffusion pump (7001/s for Argon).
The base pressure (pi) in this chamber is 5 x 10-7 mbar and the operation pressure range is 1 x 10- 5 mbar up to 10 x 10- 5 mbar during typical
RIBE experiments, the neutrals being injected into
the source chamber. This pressure, recorded both by
a capacitance manometer and a Penning ionization
gauge, will be used as a characteristic parameter of
the RIBE process. The pressure (ps ) within the
source chamber - also recorded by a capacitance
manometer - is an order of magnitude higher. The
REVUE DE PHYSIQUE
APPLIQUÉE. -
T. 24, N’ 3, MARS 1989ion beam composition is on-line recorded with a
magnetic-mass-spectrometer (MMS). The compo- sition of the stable neutrals present within the interaction chamber is recorded with a quadrupole
mass spectrometer (QMS) ; this latter is located within an independent chamber differentially pumped and connected to the main chamber
through a 2 mm diameter hole. Because of the DC excitation (80-130 V) of the hot cathode discharge
linked to the efficient electrostatic containment, the primary electrons have very large dissociation and ionization yields. Therefore, the injected neutrals
may be extensively fragmented both as CFy radicals
and CF+x ions through stepwise processes. The
fragmentation is more extensive as the discharge voltage and current increase, and as the injected
neutral flow decreases. The variation of this latter
implies the variation of the pressures in both the ionization and interaction chambers. In figure 2 are
shown typical ion beam and neutral phase compo- sition spectra for two values of the CF4 flow rate, which correspond to the extreme values of the pressure range of interest for RIBE with the present ion gun. The spectra of the neutral phase (Figs. 2a,
21
Fig. 2.
-Ion beam and neutral composition for two values of the CF4 gas flow rate, as injected into the
sourcedischarge chamber. a) and b) : neutrals and ions for the resulting highest pressure HP
=9
x10- 5 mbar within the interaction chamber. c) and d) : the corresponding spectra for the lowest pressure LP
=1
x10- 5 mbar. The beam current density
wasrespectively 0.5 and 0.25 mA cm- 2 for HP and LP.
2c), clearly shows that heavy fluorocarbon molecules such as C2F6, C3F6, C4F6, C3F8 and C4F8 are also synthetized. The existence of these species may be
imputed to the presence of the large amount of
unsaturated radicals such as CF and CF2 both within
the source chamber and the ion beam chamber. A discussion about the plasma chemistry which governs the overall behaviour of the ion gun would be
interesting but is nevertheless out of the scope of the
present paper. It must be pointed out that these spectra are associated to the steady-state operation
of the gun, that means conditions for which both the thermal equilibrium and the chamber conditioning steady state are reached. Starting from a clean vessel
the latter requires about 2 h [14]. From figure 2 it is clearly seen how the pressure is a sensitive parameter in order to modify and/or control the ion beam and neutral flow composition, although this latter is rather difficult to be determined with the present
experimental set-up.
2.2 SAMPLE EXPOSURE AND POST-ETCHING TREAT- MENT.
-The samples to be exposed to the ion beam
were introduced via a load-lock system. They were
stuck with a carbon paste on the water-cooled substrate holder ; this latter is driven by a motor system which allows the choice of both the sample
position and orientation within the beam cross-sec-
tion. A small Faraday cup is included in the sample
holder to measure the irradiation current density.
The emission or the subsequent formation of stable
neutrals associated to the beam sample exposure are on-line recorded by the QMS - in particular the SiF4 partial pressure.
Masked samples with HPR 204 resist, post-backed
at 110 °C, were used for etch rate measurements.
They were exposed for a given period of time so that
a step of about 150-200 nm is obtained, in order to
reduce the uncertainties associated to the profilome-
ter measurement and to the transient variation of the silicon etch rate (Sect. 3.2a). Thermal silicon dioxide
(600 nm) grown on silicon substrate and Si single- crystal were used for the present results. The Si substrates were n-type, (100)-oriented crystals with resistivity ranging between 4.0 and 6.0 ilcm. For the
analysis of the CF+x/Si interaction kinetics and
mechanisms, unmasked samples were considered. A standard organic cleaning was the only pre-exposure treatment.
After the RIBE exposure, the wafers were dipped
into absolute ethanol as soon as they were taken out
of the load-lock chamber. This procedure has been
shown to reduce the sample contamination and
oxidation in particular for those having the thinnest
carbonaceous overlayer. They will be subsequently
called as-etched samples. It has already been shown
that 02 plasma post-etching treatments are efficient for the removal of C, F-films grown on silicon [4, 16]. However a silicon dioxide layer is also produced, thereby consuming the underneath silicon substrate up to 2-3 nm depth. Thus after a HF dip, the near-
surface of the silicon substrate is also removed and it is no longer possible to evaluate the damage and
contamination of the very interesting interfacial
layer. In order to preserve the integrity of this layer,
different post-RIBE wet procedures were evaluated concerning their ability to remove the carbonaceous
overlayer. A 60 s dip in concentrated hydrofluoric
acid (50 %) was shown to be efficient from ellip- sometric, Auger and electrical characterizations and for standard exposure conditions, as discussed in
section 3. One may expect that such a wet procedure
did not remove the underneath silicon as far as this latter had not suffered important oxidation. They
will be called wet-cleaned samples. They were also
maintained in absolute ethanol until they were
characterized. For samples which were left at room
air after etching, the time for the HF dip which was required to remove the overlayer decreases ; after a week, 30 s were effective.
2.3 SAMPLE CHARACTERIZATIONS.
2.3.1 Ellipsometric measurements.
-A manual
Rudolph Research T436 was used to determine the
ellipsometric angles. The polarizer-compensator- sample-analyzer configuration (PCSA) was chosen.
d and w were measured in all four zones at a 70°
angle of incidence and at a 546.1 nm wavelength
[17]..
2.3.2 Auger Electron Spectroscopy (AES) ; Auger Sputter Profiling (ASP).
-Auger Electron Spec- trometry was performed from nonderivative spectra collected in the EN(E) mode. They were quantitat- ively exploited by the use of the peak to background
ratio (Px/B), hereafter referred to as the Auger
ratio. The advantages of the method have already
been discussed [18]. The electron gun delivered a
3 keV, 0.5 J..LA beam, and the CMA aperture angle
was 42.18°. For ASP, a low energy ion gun has been
developed in order to reduce the knock-on spreading effect ; it delivered a 100 eV-50 03BCA Ar+ ion beam in
a 7 mm diameter spot.
2.3.3 Electrical evaluation : Mercury Probe-Silicon Diode.
-A mercury-silicon contacting device was
used in order to analyze the Schottky-barrier diode
as established between mercury and silicon : it is the
Mercury Probe-Silicon Diode [19]. The Hg-Si con-
tact to be tested had a 1 mm2 area ; the samples were
stuck backside on a metal electrode with a silver paste which insured a large area contact (100 mm2).
I-V ; C-V ; 1/C2-V and G-V characteristics were
recorded with this device which can be used locally
and instantaneously to form a diode.
3. Expérimental results.
3.1 ETCH RATES AND Si02/Si SELECTIVITY. - The
dependence of the etch rates and Si02/Si selectivity
versus the operation pressure
-pi
-is shown in
figure 3. The etch rate values have been normalized to a 1 mA cm- 2 current density, as it is usual in
RIBE data in order to make easier the comparison
between various experiments. Of course this pres- entation of the results has only a practical meaning if
the ion gun is able to deliver a current density up to this value and if the beam composition is not
affected by the necessarily linked modification of the
source parameters. Generally, for a low operation
pressure, the beam composition is strongly modified
as the discharge current is increased and a nonlinear behaviour of the Si etch rate is observed, with a deficit as compared to a linear extrapolation [10].
Our own experiments confirm this feature. Con-
versely the Si02 etch rate is not as much affected, so
that higher selectivity might be achieved with a
further increase of the beam current density.
For the present work the Si etch rate of
7 nmlmin/mAlcm 2 corresponds more physically to
an effective sputtering yield of 0.1 Si atoms/imping- ing CFx ion (500 eV ions at normal incidence). This
value is smaller than the 0.2-0.3 values reported in
the literature for beams extracted at about the same
energy from more standard source discharges oper- ated with CF4 [6-8]. However it must be pointed out that, for these latter experiments, the operation
pressures within the ion sources and process cham- bers (respectively about 10-3 and 10-4 mbar) are higher than those of the present device. The 0.1
Fig. 3.
-Si02 and Si etch rates
asfunctions of the neutral pressure in the interaction chamber and resulting the Si02/Si etch selectivity (steady-state values). Data
arenormalized to
anion-current density of 1 mA cm - 2. Beam
energy W+
=500 eV, and incidence 0
=0°.
sputtering yield value compares to the values re-
ported for the operation of such standard ion sources
but operated with CHF3 [16]. This different be- haviour may be attributed to the extensive fragmen-
tation of the injected neutrals and/or to the low operation pressure which both may be achieved with the Maxi-ERIS ion gun. The pressure variation
implies the variation of the composition of the beam,
but also the variation of both the composition and
flux of the neutrals and radicals which are emitted
out of the ionization chamber, as already mentioned.
As far as the ion beam is concerned, its composition dependence versus pressure is shown in figure 4,
where are plotted the variations of the peak height
ratio of the main ions. The reference is CF3’ , the
most abundant ion in the usual RIE plasma environ-
ment. The high selectivity values, in the range 10-20,
are achieved for the lowest part of the pressure variation range. The comparison between figure 2
and figure 4 shows that such high values are associat-
ed to the more extensive fragmentation of the
neutrals and to a large increase of both the C+ and F+ monoatomic ions, the dominant ion
being then CF+ . Values reported for CFx bombard-
ment of silicon show that the etch yield decreases as
far as the F/C ratio decreases [20]. A value of 0.1 at
500 eV was reported for the sputtering yield of
silicon by CF+ ions ; it is in agreement with the value derived from this work. However as shown in
figure 2, the ion spectrum is so rich that this compari-
son has only a qualitative meaning.
Fig. 4.
-Dependence of the CFx ion beam composition
versus
pressure pi to be associated to the data in figure 3.
The source discharge voltage is 120 V. The discharge
current is adjusted (about 2 A) in order to provide
aconstant beam current density (j+
=0.5 mA cm-2). Yet
for the lowest pressure value j + decreases to
0.25 mA cm- 2 because of the pressure limitation of the
discharge current.
What about the contribution of the neutrals and radicals in the present RIBE process ? As already known, the slow-down of the Si etch rate can be attributed to the formation of a C, F-blocking overlayer [2, 5]. This latter can develop in RIBE
processes with CF4, as will be discussed later on, but not in the more classical Reactive Ion Etching (RIE)
with pure CF4 in a plasma environment because of the role played by the dense population of reactive
neutrals. In typical RIE conditions the neutral to ion flow density ratio range is 10- 2-10- 3, the ion flow density having about the same value as in the present RIBE experiment (0.25-0.5 mA cm- 2). On the other
hand, for this latter this ratio has a much smaller value. If we consider
-as a first estimation
-only
the stable neutrals which are thermalized within the interaction chamber (as they are recorded by the QMS), the following values are obtained :
Low pressure operation :
High pressure operation :
Two comments must be added. Firstly, as shown
in figure 2a-c, a large variety of fluorocarbon mol- ecules are present in the interaction chamber and their respective sticking coefficient and further on
influence may be quite different. Secondly, radicals
and heavy molecules are ejected from the ion source
and impinge straight on the sample. They contribute
to an increase of the above values of the ratio. A
more accurate estimation would require the analysis
of the nature and flow of the neutral species which
are emitted outward the ion source. As far as concerns the RIBE process by itself, experiments
have been performed in order to estimate the effects
of these particles : 1) Similar measurements as those
reported above but done in a farther cross section of
the beam, located 37 cm downstream the grids [21] ; 2) Influence of the electrons delivered by the neut-
ralizer discharge and 3) Exposure of the samples to
the ion source plasma when the optics bias potentials
were turned off. Data will be reported in a further
paper.
The present results show that a Si02/Si etch rate selectivity up to 20 may be achieved for the RIBE mode and the operation of the gun at the lowest pressure (1 x 10- 5 mbar). In a practical process
higher values might also be expected with the
present apparatus from the use of two modifications
to the present primitive RIBE experimental process.
First, if the operation pressure is varied du ring the etching of the silicon dioxide film of given thickness ;
it is possible to begin at the highest value and then
lower the pressure as far as the etching proceeds
towards the silicon substrate. The mean value of the silicon dioxide etch rate will then be higher, because
for the pressure range which is considered in figure 3,
it varies by a factor of 2. The etch time of the oxide film will then reduced, and if the mean value of the silicon dioxide etch rate is used to define a concept of
«effective selectivity », a higher value of this
latter might be expected from a convenient monitor-
ing of the RIBE pressure. It must be noticed that whatever is the pressure in this range the etching
remains anisotropic and that the operation pressure of an ion source is easily monitored. Second, from the direct injection of fluorocarbon gases within the interaction chamber known as the CAIBE mode for
Chemically Aided Ion Beam Etching. Our measure-
ments show that starting from the lowest pressure value as reported in figure 3 (1 x 10-5 mbar) without
additional gases (RIBE procedure), the injection of CF4 up to 10 x 10- 5 mbar, does not modify the Si
etch rate but conversely increases that of Si02 by a
factor of 1.3, and thereby involves a selectivity value
of about 25.
3.2 CARBONACEOUS OVERLAYER GROWTH AND CLEANING.
3.2.1 SiF4 partial pressure.
-The on-line variation of the SiF4 partial pressure (SiF’ peak height) as a
function of the Si irradiation ion dose is shown in
figure 5. Three values of the pressure were con-
Fig. 5.
-On-line variation of the SiF4 partial pressure (SiF’ peak height in the QMS)
as afunction of the Si irradiation dose. Parameter : the gas flow rate and the
resulting pressure p; within the interaction chamber.
W+
=500 eV ; 0
=0° ; Si(100) ; electron-compensated
ion beam. The given time scale is related to the high and
medium pressure (0.5 mA cm-2 ion current density).
sidered, in order to demonstrate the influence of both the ion dose and neutral pressure. For these measurements the characteristic time of the vacuum
equipment was estimated to be about
T =2 s ; it
affects mainly the signal increase which would be very sharp in the limit of a vanishing characteristic time. The sample introduction speed and the ion beam density profile have also to be considered for a more accurate quantitative analysis of these graphs.
Experiment with RCA clean silicon demonstrates that the effect of the native oxide layer does not
affect significantly the characteristic features of the above graphs, as they are discussed now. The
maximum Ro of each curve corresponds to the silicon etching. Then the silicon removal rate is decreasing
as a result of the carbonaceous overlayer growth on
the Si surface. A critical dose Dl does appear which may be attributed to the establishment of the steady
state of the overlayer thickness and structure, the etch rate (Rs) being then constant. The resulting
attenuation factor Rs/ Ro of the etch rate is clearly
seen ; it varies from 0.45 to 0.15, for the considered pressure variation range. The lower the attenuation
factor, the higher the dose DI, a feature that will be
further imputed to the increase of the overlayer
thickness (see Fig. 9). Here again Dl presents a sharper increase in the lowest part of the pressure range ; typical values are :
The steady state relative values of the etch rate
(R,) as derived from these graphs have been com- pared to the etch rate values as measured from profilometer data. They are proportional. From this
result it may be assumed that when the overlayer steady state is reached, whatever is the pressure, similar mechanisms are implied in the removal of the underneath silicon. In the lack of a better infor- mation, if we assume that for the transient period of
the overlayer formation the etch rate variation is
also given by the graphs in figure 5, it is possible to
obtain a crude estimation of the silicon thickness el which is removed for an exposure dose Di: A
value of about 4 nm is obtained, almost independent
of the operation pressure. Similar graphs to those in figure 5 are also obtained when other RIBE process parameters are modified, such as the incidence, the
energy or the nature of the injected neutrals. An
important feature which must be emphasized is that
for the present RIBE experiments the overlayer
thickness increases with increasing the energy, in the studied energy range (250-1 500 eV) ; see reference [22] for the preliminary results related to the energy influence and for the operation with CHF3.
The data shown in figure 5 imply a time depen-
dence of silicon etch rate. Conversely, for the
present experimental conditions there is no blocking overlayer grown on the silicon dioxide, the etch rate of which does not depend on the exposure time, excepted for a slight initial increase imputed to the
variation of the surface temperature. In such cases for which a blocking overlayer formation is required
to insure the etch rate slow-down of the underneath substrate and thereby to provide for the film-to-
substrate selectivity, this latter selectivity concept has to be more precisely defined and used. In
particular, the values estimated from long time
exposure, i, e. the
«steady-state selectivity
»values
as those reported in figure 3, are not sufficient to
determine the optimum overetch time for a given
process. The time dependence of the etch rate and the resulting removed thickness during the initial
transient period of the silicon exposure to reactive
species have to be considered [2].
3.2.2 Ellipsometric data and post-RIBE cleaning.
-In figure 6a is shown the variation of d as a function of the ion dose for the same three pressures as in
figure 5. The initial decrease of à may be attributed
to the thickness increase of a thin overlayer grown
on the substrate. However the critical dose Dl, as
defined previously in figure 5, now corresponds approximately to a minimum value of d. Beyond Dl a slight increase of à is observed up to a higher
dose D2 beyond which a plateau is reached. The value of D2 is about twice that of Di. Such a
variation of d versus the ion dose has been attributed
to the final reorganization of the near-surface of the silicon substrate, once the overlayer thickness is
Fig. 6.
-a) Variation of the ellipsometer angle à as
afunction of the Si irradiation dose. The parameters are the
same
asin figure 5 ; b) Variation of .L1 for both as-etched and wet-cleaned samples processed at the lowest pressure . value LP.
constant and as long as the Si interface regresses,
because two phenomena interfere for the determi- nation of the ellipsometric parameters : the substrate
damage (crystal defects and atom incorporation) and
the overlayer growth. The first amorphization step which requires a dose of about 2 x 1015 cm- 2 [20]
cannot be seen in the present exposure scale. The saturation value of à depends on the properties of
both layers. Yet for low layer thicknesses, a quanti-
tative analysis is not accurate [17]. Nevertheless the above variations are significative of the sequential
effects of the reactive ion bombardment as will be discussed later on. In order to investigate more precisely for the contribution of the overlayer in the
d values, the measurements have been done on wet- cleaned samples. The data are plotted in figure 6b
for the samples processed at the lowest pressure.
They show that :
i) for the lowest dose in the scale, say D DI, the
HF dip has no effect on 0394. Times higher than 60 s
have been tested without effect. The substrate is not able to be cleaned by this wet procedure. The AES spectrum of the wet-cleaned sample surface is shown in figure 7a ; it demonstrates the presence of a large
amount of both carbon and fluorine, but the SILVV peak is nevertheless seen. The AES spectrum of the as-etched sample is about the same as that for
the wet-cleaned sample. It may be assumed that carbon and fluorine are incorporated within the silicon lattice, as discussed by Chuang et al. [23] ;
Fig. 7.
-AES spectra
-peak
overbackground EN(E)
mode - for as-etched and wet-cleaned samples processed
at the lowest pressure : a) Wet-cleaned after
asmall exposure dose (1.2
x1016 cm-2) ; b) and c) Wet-cleaned and as-etched after exposure for the critical dose D2
=1.5 1017 cm-2, as defined in figure 6.
ii) for D > Dl, the wet cleaning treatment leads
to an increase of d which is approximately constant
and equal to 17°. In figure 7b and 7c are shown the
AES spectra of wet-cleaned and as-etched samples
which both have received the same CFx ion dose D2 (1.5 x 1017 cm- 2). The as-etched spectrum is
typical of Si with its C, F-blocking overlayer. The
oxygen peak is not intense (Po/B
=0.02) and the SiLVV, (89 eV) peak is not seen. The wet-cleaned
sample spectrum corresponds typically to the case of
silicon contamined in room air, during the transfer time to the Auger diagnostic chamber ; Po/B
=0.08 corresponds approximately to one Si02 mono- layer. The carbon peak is very small (PCIB
=0.03), and no residual fluorine signal is seen : the
substrate seems really
«clean », from the AES point
of view. The d value of the etched + wet-cleaned silicon sample for D
>D2, is approximately equal to
that of a silicon with its native oxide. Although the ellipsometric measurements were done under a dry nitrogen flow, a native oxide overlayer does exist.
Furthermore, as shown in reference [24], a silicon damaged under Ar+ ion bombardment may have, when the saturation is reached and according to the light wavelength, the same d value as that of a
substrate cleaned in an ultra-high vacuum. Therefore
a more sensitive characterization must be done in order to further investigate the residual near-surface
damage (Sect. 3.4).
Fig. 8.
-Auger Sputter Profiles (peak
overbackground EN(E) mode), through the C, F-overlayer and the Si
near-surface. The parameter is the pressure the values of which
arethose given in figure 3 and referred to
asHP, MP and LP ; Ar+ sputter-beam : 100 eV ; 0.13 mA cm- 2.
The given overlayer thickness values
areestimated from the C- and Si-signal variations.
3.3 OVERLAYER AND NEAR-SURFACE IN-DEPTH ANALYSIS (ASP).
-ASP was used to estimate the thickness and the composition profile of the car-
bonaceous overlayer. In-depth profile scans are
shown in figure 8, for as-etched samples having
received an exposure dose higher than D2, and for
the three typical pressures already considered. Of
course the method may involve uncertainties, in particular concerning the fluorine which desorbs under electron impact [23, 25] and may chemically
react under the ion bombardment. The F-profiles in figure 8 correspond to the steady state of the electron induced desorption. A further paper will report
more details on AES, ASP measurements and the
comparison of the results with XPS data. We com- ment the main features of the in-depth analysis (ASP) in relation with the present address.
As the sputtering of the residue proceeds the Si- signal grows and the CKLL signature changes beyond
a critical dose from a graphitic shape to a shape
which is suggestive of the presence of Si-C. In order to clarify the discussion, this dose was chosen to define a conventional C, F-film/Si interface (1). A question now arise : do the Si-C bonds exist within the silicon near-surface or are they induced by the
Ar+ sputter ion bombardment (100 eV ions) ? The
answer to this question is important ; first in order to
determine the appropriate post-etching treatment to restore the silicon crystal to a device quality state [4]
and, second because Si-C bonds have been identified from XPS data on silicon exposed to CF4/H2 RIE,
for conditions of selective etching of Si02 over Si [2, 3]. As shown in figure 7b the AES spectrum of a wet-cleaned Si surface does not show incorporated
carbon. Therefore it may be assumed that the C, F- overlayer/silicon interface is relatively abrupt, at
least for the irradiation conditions involving the
formation of a rather thick carbonaceous overlayer (low pressure operation). The exponential decrease
of the carbon signal in the graphs of figure 8, as observed for the low pressure values, may be considered as a corroboration of this feature. Such a
variation is in fact predicted when an atomic layer is sputtered, if the removal rate of the monolayer is
assumed to be proportional to the surface coverage
as a consequence of the statistical nature of the
sputtering process [26]. Recent XPS data on Si
samples exposed to RIBE with CHF3 in the present apparatus have been reported by Cardinaud et al.
[22]. No Si-C bonds were detected even for 1500 eV ions in the Si2p detailed peak as obtained from a
monochromatized radiation. It may be expected that
the same behaviour is also valid for RIBE with
CF4.
The lower the RIBE processing pressure, the
higher the sputter dose required to remove the major fraction of the C, F-overlayer. For the lowest
pressure a plateau is clearly seen in the carbon
profile with about the same Auger ratio PC/B ~
0.75, whatever are the energy, incidence and gas
(CF4 or CHF3). It corresponds to the bulk of the carbonaceous layer grown under the ion bombard-
ment.
Whatever is the pressure, the fluorine profile
shows two regions which correspond either to a
difference of concentration or to a difference in the bond strength of the fluorine with the surrounding
atoms. The inner region, near the film-Si interface may be imputed to the direct incorporation into the
carbonaceous overlayer of the fluorine atoms which
are produced by the dissociation of the impinging
fluorocarbon molecular ions such as CF+x. Con- versely the steep fluorine decrease which is seen at the topmost part of the carbonaceous film may be
imputed to the formation of a mixed layer. This
latter would be formed under the bombardment of reactive neutrals, neutralizing electrons and ener-
getic ions. The synergetic effects of the irradiation of this mixed layer with the three types of particles
determine the balance between the deposition and
the removal of the carbon at the interface between the C, F-film and the gas phase. On the other side of
the C, F-film, and beyond the interface I, the fluorine steeply decreases within the silicon. This feature may be partly attributed to the Ar+ activated
etching of the Si in the presence of fluorine. It has been shown that the chemical enhancement of
sputtering is very high at low ion energy [27].
Anyhow, no more fluorine is seen in the AES spectrum of a wet-cleaned sample which has not been bombarded by Ar+ (Fig. 7b). We may assume
that, as well as the carbon atoms, the fluorine atoms
are not incorporated deeply into the silicon lattice.
For the rather low pressure, the Si signal shows an exponential increase here again, as that which might
be expected from the existence of a sharp interface I.
Assuming this fact, this portion of the graphs has
been used to determine the required dose to sputter
a thickness equal to the electron escape depth (SiLVV electrons mean free path through graphite
was taken as : A e
~3.8 Â). The overlayer thickness
values given in figure 8 were then derived, assuming
a constant sputter erosion through the entire over-
layer. The comparison of the slopes of the SiLVV signal increase, in figure 8, demonstrates that the
sputtering yield of the blocking overlayer depends
on the process pressure. The associated variation of the composition of the incoming ion and neutral flows may be supposed to affect the physico-chemi-
cal nature of the carbonaceous residue layer. XPS
measurement would give further information about the chemical bonds and thickness of this film [22].
Whereas at the interface 1, the (PSi/B)I Auger
ratio is independent of the pressure and is equal to 0.4, the (PC/B)I Auger ratio increases as the
pressure increases and is respectively 0.2, 0.3 and
0.4, values to be compared to 0.75 corresponding to
the bulk carbonaceous film. This increase of the carbon percentage at the interface may be explained
from the decrease of the overlayer thickness which
as a consequence leads to a deeper incorporation of
the carbon atoms into the silicon lattice, associated
to a less sharp interface between the overlayer and
the silicon. These features are corroborated by the following results dealing with the electrical evalu- ation of the residual contamination and damage of
the silicon.
3.4 NEAR-SURFACE SILICON DAMAGE : CONTACT
ELECTRICAL EVALUATION. - The current-voltage
characteristics of Hg/Si contacts of RIBE-exposed
Si are shown in figure 9 for wet-cleaned samples.
Both forward and reverse 1 V characteristics are
given for three exposure doses and compared to
those of a control sample which also has received the wet cleaning treatment, but without beam exposure.
The dose effect on the near-surface damage is clearly
demonstrated.
*