Ion and electronassisted gassurface chemistry—An important effect in

plasma etching
J. W. Coburn and Harold F. Winters

Citation: J. Appl. Phys. 50, 3189 (1979); doi: 10.1063/1.326355

View online: http://dx.doi.org/10.1063/1.326355
View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v50/i5
Published by the American Institute of Physics.

Additional information on J. Appl. Phys.

Journal Homepage: http://jap.aip.org/
Journal Information: http://jap.aip.org/about/about_the_journal
Top downloads: http://jap.aip.org/features/most_downloaded
Information for Authors: http://jap.aip.org/authors

Downloaded 16 May 2013 to 141.161.91.14. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jap.aip.org/about/rights_and_permissions
 lon- and electron-assisted gas-surface chemistry-An important
effect in plasma etching
J. W. Coburn and Harold F. Winters
IBM Research Laboratory. San Jose. California 95193
(Received 23 October 1978; accepted for publication 13 November 1978)
The extent to which gas-surface chemical reactions can be enhanced by energetic
radiation (primarily ions and electrons) incident on the surface is described. Emphasis is
placed on chemical systems which lead to volatile reaction products. In particular, the
reactions of Si, Si02, and Si3N4 with XeF 2, F 2, and Cl 2 are examined experimentally.
Possible mechanisms for the radiation-induced enhancement are discussed and some
technological implications of this process in plasma etching technology and lithography
are considered.
PACS numbers: 52.40.Hf, 82.65.Nz, 79.20. - m
I. INTRODUCTION
There are many instances where it has been observed
that the rate at which a gaseous species reacts with a solid
surface in a vacuum can be enhanced by simultaneous irra-
diation with energetic ions or electrons. The most common
example of this effect is probably the carbon buildup ob-
served on an electron-iradiated surface in a vacuum contain-
ing hydrocarbon vapors. I This effect has caused problems in
electron microscopy I and spectroscopy2 and it is also known
in these fields that these carbon layers can be removed by
electron irradiation in the presence of oxygen. J Moreover,
electron-stimulated oxidation has also been noted on various
solid surfaces where the oxide is not volatile.'- Io
Several new reactions involving radiation-enhanced
processes are reported in this paper. It is shown that reac-
tions between silicon and XeF 2,F2, and Cl 2 to form volatile
products (e.g., SiF. and SiCl.) are greatly enhanced by argon
ion bombardment. It is also shown that Si JN.,Si0 2, and SiC
react with XeF 2to form volatile products in the presence of
either electron or ion bombardment, whereas no reaction is
observed without radiation. Interpretation and results for
ion-bombardment-enhanced chemistry are presented in Sec.
III A, while similar information for electrons is presented in
Sec. III B.
The reactions mentioned above and the conditions un-
der which they occur explain many experimental observa-
tions made by investigators studying etching phenomena in
a glow discharge environment. For example, the etching rate
of a surface subjected to energetic ion bombardment in a
plasma etching system (i.e., the target or cathode surface) is
usually substantially larger than the etching rate of a surface
held at ground potential or at the floating potential. Exam-
ples of this type have been reported by Hosokawa et al. 11 for
the halocarbon plasma etching of silicon and by Holland and
Ojha 12-14 for the oxygen plasma etching of carbon. The mag-
nitude of the etch rates which are observed are such that the
enhancement caused by ion bombardment cannot be easily
explained by simply superimposing a physical sputtering
process onto the chemical etching process.
3189 J. Appl. Phys. 50(5), May 1979 0021·8979179/053189·08$01.10
Downloaded 16 May 2013 to 141.161.91.14. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jap.aip.org/about/rights_and_permissions
As was suggested in a previous paper,15 the etching of
solid material by exposure to gas-phase particles with or
without a plasma is usually described by the following se-
quence of steps: (1) nondissociative adsorption of gas-phase
species at the surface of the solid being etched; (2) dissocia-
tion of this absorbed gas (i.e., dissociative chemisorption);
(3) reaction between adsorbed atoms and the solid surface to
form an adsorbed product molecule, e.g., SiF. (ads); (4) de-
sorption of the product molecule into the gas phase; and (5)
the removal of nonreactive residue (e.g., carbon) from the
surface. Illustrative examples of each of these five steps are
now presented schematically for silicon interacting with CF J
radials.
Step 1. Nondissociative adsorption
CFJ(gas) + Si-CFJ(ads) + Si,
Step 2. Dissociative adsorption
CF3 (ads) + Si-C(ads) + 3F(ads) + Si,
Step 3. Formation of products molecule
4F(ads) + Si-SiF.(ads),
Step 4. Desorption of product molecule
SiF.(ads)-SiF.(gas),
Step 5. Residue removal
C(ads) + 4F(ads)---+CF.(gas).
The first step almost always occurs since attractive
forces between the undissociated molecule and the surface
usually exist. This step may involve adsorption into a so-
called "precursor state" where the molecule is mobile and
diffuses across the surface until it dissociates; possibly at a
step, kink, vacancy, or other defect. The resulting atoms may
continue to diffuse or may become attached to a given sur-
face site. On the other hand, it is possible that the first four
steps occur almost simultaneously, i.e., there is very little
surface diffusion. An example of a possible reaction of this
type is
CFlgas) + F(ads) + Si(solid)-SiF.(gas) + C(ads).
Nevertheless, analysis of the etching reaction in terms of
© 1979 American Institute of Physics 3189
these five steps is useful independent of the detailed mecha-
nism which results in formation of the product molecule.
Finally, the first two steps would be combined and labeled
"adsorption" for atomic fluorine but the conceptual frame-
work would be similar.
We now discuss mechanisms which could cause radi-
ation-enhanced etching and relate them to steps 2-S. There
are at least three processes which may increase the rate of
dissociative chemisorption (step 2): (1) Dissociative chemi-
sorption may occur exclusively or preferentially at defect
sites produced by ion or electron bombardment. (2) Electron
and/or ion bombardment may cause an adsorbed molecule
to dissociate in situations where this would not normally
happen. In this regard, it should be noted that many of the
molecules used in plasma etching do not spontaneously dis-
sociate on the surface of interest but do dissociate under ion
or electron bombardment. 15 Electron-induced dissociative
processes on the surface are likely to be similar to those
which take place in the gas phase, i.e., the adsorbed molecule
is excited to an electronic state which subsequently dissoci-
ates. Ion bombardment, on the other hand, probably causes
fragmentation as a consequence of momentum transfer to
the adsorbed molecule. (3) Dissociative chemisorption often
occurs on an atomically clean surface but may not occur on
the same surface with a monolayer of adsorbed gas. Removal
of the adsorbed layer by electron bombardment (electron-
stimulated desorption) or ion bombardment (physical or
chemical sputtering) may allow the dissociative chemisorp-
tion reaction to proceed, which in turn promotes etching.
For example, it is expected that a monolayer of oxygen, car-
bon, or chlorine would slow the etching of silicon exposed to
a flux of fluorine atoms. When the adsorbed layer is removed
by radiation, the etch rate should be enhanced.
If a surface reaction is to involve more than monolayer
chemisorption, then the species adsorbed on the surface
must be able to migrate into the second and deeper layers,
forming new chemical bonds and often new molecular spe-
cies. This is step 3, product formation, and it often requires
an activation mechanism to proceed, i.e., a monolayer is
formed and the reaction stops unless the substrate is held at
elevated temperature or there is ion or electron bombard-
ment. Damage-enhanced diffusion, knock-on collisions, and
bond breaking may promote the reaction in the presence of
ion bombardment. Electron collisions may also break bonds
which allow adsorbed atoms to penetrate and react with the
substrate. Although the precise mechanisms are unclear, it is
certain that electron and ion bombardment cause step 3 to
occur in some instances where the chemical reaction does
not proceed in the absence of radiation.
The consequences of the above processes are largely de-
termined by the volatility of the species which are formed. If
the species are volatile, they will desorb, causing etching of
the surface (step 4). If the species are involatile, a layer of
reaction product will form with its thickness being deter-
mined by the mobility of the chemically active species
through the layer (e.g., the oxidation of aluminum).
J. Appl. Phys., Vol. 50, No.5, May 1979
3190
Downloaded 16 May 2013 to 141.161.91.14. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jap.aip.org/about/rights_and_permissions
There are several mechanisms which could cause radi-
ation-enhanced etching by influencing step 4. For example,
an involatile product molecule may be sensitive to electron-
stimulated desorption and therefore a reaction which would
not occur without radiation may occur under the influence
of electron bombardment. Moreover, an involatile product
molecule may be weakly bound to the surface and as a conse-
quence of this weak binding the sputtering yield of the prod-
uct molecule could be substantially larger than the sputter-
ing yield of the original substrate. Therefore, a combination
of physical sputtering and chemical reaction would produce
a radiation-enhanced etch rate which is substantially larger
than would be expected from consideration of the separate
individual etch rates. A model based on this mechanism has
recently been proposed by Mauer et al. 16 to explain the en-
hanced etching of silicon caused by ion bombardment in a
glow discharge. It should be noted, however, that evidence
from the literature"·l8 and data presented in Sec. III indicate
that adsorbed layers usually decrease the rate that the under-
lying material is sputtered. We believe that the etch rate en-
hancement caused by ion bombardment is, in most cases,
due to ion-enhanced gas-surface chemistry and not to a
chemical enhancement of momentum transfer sputtering.
The discussion of the experimental results to be presented
below is based on the assumption that the latter process is
not important. We are not aware of any experimental evi-
dence which demonstrates conclusively that this assumption
is true, but some recent data relevant to the question is pre-
sented in the Appendix. Finally, one must consider the possi-
bility of an involatile residue being formed by one or more of
the atomic species present in the molecular gas, e.g., ad-
sorbed carbon from CF4 , Unless this residue is removed by
some mechanism (step S), it will terminate the etching reac-
tion. Electron and ion bombardment could enhance residue
removal by many of the same mechanisms mentioned above
in connection with step 2-4.
II. APPARATUS
Two vacuum systems invoking two somewhat different
approaches have been used for this study. Most of the elec-
tron-beam-enhanced gas-surface chemistry has been carried
out in a bakeable ultrahigh vacuum system which contains a
CMA for Auger analysis, a quadrupole mass filter, and an
ion gun. In this system, which has been described previous-
ly,19 an electron beam can be directed onto a clean surface in
the presence of a background pressure (- 10- 1 Pa) of active
gas. Most of the ion-enhanced gas-surface work has been
carries out in a diffusion-pumped high-vacuum system
where the active gas is directed onto the surface by means of
a small tube (1.6 mm in diameter) maintained at room tem-
perature and placed close to ( - 3 mm) the sample surface.
The effective pressure of the gas at the sample surface is in
the range 10- 3_10- 1 Pa. An ion beam can be directed onto the
sample coincident with the flux of active gas. The ion beam,
which is not mass analyzed, is obtained by electron impact
ionization of a gas (usually argon) admitted to the system at
a pressure of about 2 X 10-2 Pa. The experimental arrange-
ment is shown schematically in Fig. 1. In order to ensure that
the active gas does not significantly affect the composition of
J.W. Coburn and H.F. Winters 3190
 Ion
~--Source
Reservoi'r
Manometer
Quartz
Crystal
Microbalance
FIG. 1. Schematic diagram of the apparatus used to study ion-assisted gas-
surface chemistry. The gas injection tube is 1.6 mm inside diameter and is
about 3 mm from the quartz crystal microbalance. The gas flow is deter-
mined from the rate of pressure increase in the reservoir when the shut-off
valve is closed.
the ion beam, the flow rate of active gas injected through the
small tube onto the surface was kept to less than 1% of the
argon gas flow injected through a conventional leak valve.
In both these experimental systems, the etch rates were
measured using quartz crystal microbalances. Thin films
( - 2 Jlm thick) were deposited by conventional vapor depo-
sition methods onto the vender-supplied (Inficon No. 321-
25) quartz crystals prior to their installation in the vacuum
systems. All experiments were conducted at room
temperature.
The experiments reported in this paper involve the use
ofXeF 2, F 2, and C1 2. These gases react slowly with reservoirs
and vacuum system walls to produce involatile halides and
gaseous impurities such as CO 2, CO, and COF2 • The reser-
voirs containing these gases were exposed for a significant
period of time prior to experiments in order to minimize
artifacts. Nevertheless, gaseous impurities were always
present.
III. EXPERIMENTAL AND DISCUSSION
A. Ion-enhanced gas~surface chemistry
1. Experimental
Ar+ ions were used exclusively in this study for two rea-
sons: (1) to eliminate any chemical contribution from the
incident ions, and (2) the sputtering yield of Ar+ ions is suffi-
ciently large that the slight contamination of the ion beam
due to the reactive background gas will not appreciably in-
fluence the sputtering rate of the surface. This would not be
true, for example, if He+ ions were used and a small fraction
of CI+ ions was introduced into the He+ ion beam.
A chemical system which effectively demonstrates ion-
enhanced gas-surface chemistry is XeF2 on Si where the re-
action product (SiF4 ) is volatile at room temperature and
3191 J. Appl. Phys., Vol. 50, No.5, May 1979
Downloaded 16 May 2013 to 141.161.91.14. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jap.aip.org/about/rights_and_permissions
where the residue (Xe) is also volatile and does not impede
the reaction by remaining on the surface. The etch rate ofSi
as determined from the quartz crystal microbalance is plot-
ted versus time in Fig. 2. The Si surface is first exposed to the
XeF2gas (t < 200 sec), then to an Ar+ beam and the XeF2 gas
together (200 < t < 640 sec), and finally to the Ar+ beam only
(t> 640 sec). The unique etching action of XeF2 on Si and
other materials has been discussed previously.zo In Fig. 2 the
cooperative interaction between the Ar+ beam and the XeF 2
gas is shown clearly in that etching ofSi obtained with both
the Ar+ ions and the neutral XeF2 molecules simultaneously
incident on the surface is about eight times the sum of the
etching rates of each species measured separately. The tran-
sient response of the etch rate immediately following the
initiation of Ar+ bombardment (200 < t < 300 sec) is believed
to be partially due to a direct interaction between the ion
beam and the quartz crystal microbalance caused by either
slight heating or bombardment-induced strain in the crystal.
This transient may also be due to a decrease in the steady-
state surface coverage of fluorine caused by the Ar+ bom-
bardment. The transient decrease in the etch rate following
the shutoff of the XeF2gas (640 < t < 750 sec) is a result of the
fact that it was not possible in our apparatus to instanta-
neously decrease the XeF2 gas flow to zero and this transient
behavior is believed to reflect the XeF2 flux decay.
Similar ion-enhanced chemistry has been seen in our
laboratory with F2 on Si, F2 on C, Cl 2on Si, and O 2on C. The
behavior of all these systems is very similar and the results
obtained for Cl z on Si are shown in Fig. 3. In this figure the Si
surface was subjected to Ar+ bombardment only for
0< t < 220 sec. The flux ofCl 2gas was initiated at t = 220 sec
and the large transient immediately following the initiation
of the Cl 2flow is due to adsorption of chlorine on Si. The rate
of chlorine adsorption overwhelms the etching process for a
short period of time causing the Si sample to gain mass (neg-
ative etch rate for 220 < t < 260 sec). Note that the steady-
state etch rate has been increased by a factor of almost 4 by
the introduction of C1 2 • When the ion beam is turned off, the
r-XeF2Gas+A,+ Jon Beam + XeF 2 Gas ---I- A'+ Jon Beam -1
Only Only
70
'. .......
:sE 60
............................
;;;:
50
'"
',i;
0: 40
.c:
w
B
30
c:
0
.\! 20
Vi
10
................... .....
OL-~10~0~~20~0~~3~OO~~4~0~0-.5~0~0-~60~O~~700~-··~·~~~~~..~"',~~~'
Time (sec)
FIG. 2. Ion-assisted gas-surface chemistry using Ar' and XeF, on silicon
(volatile reaction product). Ar' energy = 450 eV, Ar' current = 0 (t < 200
sec), Ar' current = 2.51lA (t> 200 sec), XeF, flow = 2X 101! mol/sec
(t < 660 sec), and XeF, flow = 0 (t> 660 sec). (The Ar' current density and
the XeF, flux are not uniform over the Si surface. The effective area for the
'Ar' current and the XeF, flux are estimated at 0.1 and 0.3 em', respectively.)
J.W. Coburn and H.F. Winters 3191
 etch rate drops quickly to zero since Cl 2alone does not etch
Si at a detectable rate ( < 0.1 A/min) under the conditions of
this experiment. The situation for F2 on Si, F2 on C, and O 2
on C is very similar in all respects to the Cl 2on Si system.
It is well known that the sputtering rate of materials,
which chemisorb an active gas to form an involatile com-
pound, is decreased significantly when inert gas sputtering is
carried out in the presence of the active gas. This effect has
been demonstrated using oxygen gas on several occa-
sions. n . 1s We have chosen the system fluorine gas on alumi-
num to illustrate the effect in our apparatus. The reaction
product AIF3 is in volatile at room temperature. The results
are shown in Fig. 4 where F2 is injected onto the Al surface
during Ar+ bombardment and the sputter-etch rate decreases
by a factor of 2. A similar result was obtained using O 2on Si.
It should be mentioned that in certain situations (low ion
current density or poor vacuum conditions) an increase in
the etch rate of Al was observed when F2 gas was injected
during Ar+ sputtering. This observation is believed to be an
artifact caused by background gases which impede the sput-
tering rate of Al in the absence ofF2. It is suggested that the
F2 gas assists in the removal of the adsorbed contaminants
resulting in an increase in the sputter-etch rate of the Al
substrate.
2. Interpretation
As was mentioned earlier, there are several ways in
which the incident radiation can enhance the gas-surface
reaction rate: (1) by increasing the efficiency of the initial
adsorption-dissociation step, (2) by expediting the formation
of reaction products by supplying energy to the surface, 12. 14
or (3) by increasing the rate of desorption of the product
molecule. 12 .14
~Ar+ Ion Beam-f--Ar+ Ion Beam + CJ 2 Gas---1
Only
c
~ 10 .. .
...... ., .... ... .....
c<r:
III
+-'
....
ro
c:
-5
+-'
5
W
C
o ... ... ...........
..
u
100 200 300 400 500
Time (sec)
FIG. 3. Ion-assisted gas-surface chemistry using Ar' and el, on silicon
(volatile reaction product). Ar' energy = 450 eV, Ar' current = l.0ttA, el,
flow = 0 (t < 220 sec), and el, flow = 7 X 10" mol/sec (t> 220 sec).
3192 J. Appl. Phys., Vol. 50, No.5, May 1979
Downloaded 16 May 2013 to 141.161.91.14. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jap.aip.org/about/rights_and_permissions
Ar + Ion Beam-+-- Ar + Ion Beam + F 2 Gas ---l
c:
E
......
o<T::
2::
r- Only
15 ..........•.............•.
'" 10
a:
.!: ..........................................
B
w 5
E
:>
c: j I
0
'E 100 200 400 500 600 700
:>
« Time (sec)
FIG. 4. Ion-assisted gas-surface chemistry using Ar' and F, on aluminum
(in volatile reaction product). Ar' energy = 450 eV, Ar' current = 3.0 ttA,
F, flow = 0 (t < 270 sec), and F, flow = 2 X 10" mol/sec (t> 270 sec).
In the case of the fluorine-silicon system (i.e., both XeF 2
and F2 on Si), the fact that XeF 2will etch Si at very large rates
in the absence of radiation 20 and the fact that the Si etch rate
is proportional to the flux of XeF2 on the surface20 provide
good evidence that the adsorption-dissociation process is the
rate-limiting step. That is to say, once F atoms are present on
the Si surface, the formation and desorption of SiF4proceeds
spontaneously at a rapid rate at room temperature. There-
fore, one can conclude that the mechanism responsible for
the Ar+ -enhanced XeF 2-Si reaction rate shown in Fig. 2 is an
ion-bombardment-induced increase in the rate at which
XeF 2dissociatively chemisorbs on Si. There are three possi-
ble ways in which the Ar+ bombardment could accomplish
this result: (1) creation of damage or defects on the Si surface
which initiate the spontaneous dissociative chemisorption of
XeF 2, (2) dissociation of molecular XeF 2 which is weakly
bonded to the Si surface by direct impact, or (3) removal of
nonvolatile residue. Attempts have been made to detect Xe
on a Si surface by Auger electron spectroscopy20 and no Xe
has been seen even when the 2OOO-eV electron beam is trans-
lated rapidly over the surface so as to eliminate beam-in-
duced artifacts. Moreover, only small amounts of involatile
residue are observed during XeF 2etching of silicon in the
absence of radiation. This strongly suggests that the process
responsible for the effect shown in Fig. 2 is ion-bombard-
ment-induced surface damage increasing the rate at which
XeF 2 dissociatively chemisorbs on Si.
We cannot extend this latter conclusion to the F2 on Si
system and in fact the very low reaction coefficient for F2 on
Si at room temperature (_10- 5-10- 6)21 suggests the possibil-
600 ity of the presence of weakly bonded molecular fluorine
which could be dissociated by direct impact.
Unlike the fluorine-silicon system, there is no evidence
that CI atoms on Si spontaneously form the volatile SiCl4
molecule at room temperature. The fact that chlorine is ob-
served to be chemisorbed on Si 22.23 suggests that SiCI. is not
easily formed. Consequently, we do not have enough infor-
mation to determine the role of the Ar+ ions in the enhanced
etching shown in Fig. 3 for the Cl 2-Si system. It is suspected,
however, that the role of the ion bombardment is to assist in
the formation of SiCl. by providing energy to the surface.
J.W. Coburn and H.F. Winters 3192
 9000r----,----,----.-----,----,----,,---o can be etched at appreciable rates and the etching occurs

8000
45/1(1, 50111([/(111 2, 1500 eV
/
°
only on the region of the sample which is SUbjected to the
electron bombardment. Although the behavior of all these
materials is similar, for illustrative purposes two different
7000
Rate '0 600 J\ Imin
/ experimental procedures are used in Figs. 5 and 6.
In Fig. 5, initially only the electron beam is incident on
~
Cl
w
6000 1° the Si3N4 with the background presure being 5,3 X 10-4 Pa
(4 X 10- 6 Torr) and no etching is observable. After about 800
sec, XeF 2 is introduced into the system to about 8 X 10-2 Pa
>
a
~ 5000 1° (6x 10-4 Torr) and significant etching is immediately pro-

/0
0:
duced. The opposite procedure is shown in Fig. 6 for Si02
z"
,;: 4000 where the pressure ofXeF 2is 8 X 10-2Pa (6 X 10-4Torr) at the
beginning of the run and no etching is observed. After ap-
u.
a
~ 3000
::J
a
/0 proximately 800 sec, the electron beam is activated and etch-
ing immediately occurs.
~
<t
2000
i 2. Interpretation
l
.i Auger spectra of Si02, Si 3N 4, and SiC surfaces taken
1000
electron beam on
I Xe F 2 mtroduced
Vo during exposure to XeF2 indicate that fluorine is present on
/Po4XlO·G Torr ~ 4
p o 5.8xl0· Torr the surface20 but no xenon can be seen. Therefore, the rate-
o e-.-e-.-_-.-e-.,- limiting step in these systems must be the formation of the
SiF4molecule and one would expect that the electron-stimu-
l~ted etching shown in Figs. 5 and 6 is a result of the electron
o 250 500 750 1000 1500 1750 beam enabling the F atoms to react with these materials to
TIME I SECONDS I form SiF4, It is known that electron bombardment ofSi0 2
FIG, 5, Electron-assisted gas-surface chemistry using 1500-eV electrons
and XeF, on Si,N., P(total) = 6X 10-' Torr with most of the ambient gas
9000r----.1----~----,-1---,1-----r-,--~1----~
being xenon, Neither exposure to XeF, nor an electron beam produces etch-
ing by itself. Simultaneous exposure produces an etch rate of - 600 A/min,

8000 - -

B. Electron-enhanced gas-surface chemistry SI02

1. Experimental 7000 -
P= 6x 10 4 Torr.

As we mentioned earlier, the electron-enhanced studies 45/1«, 50mu/cm7, 1500eV

were carried out in a UHV system which was backfilled with Rate" 200 '\ Imln
6000-
the active gas of interest. The gas-jet approach was not used
in these investigations. The electron beam was provided by aw
the coaxial gun ofa cylindrical mirror analyzer used for Au- 6 5000-
~
ger spectroscopy. The electron current to the sample was w
cc
typically 45 /-lA, and the current density was estimated to be
4000-
50 mA/cm2 by using a Faraday cup with a 5 X 1O-4_cm 2aper-
ON

1°
ture. The electron beam could be positioned precisely at the u.
a / •
center of the sample by observing the beam-induced frequen- I-- 3000 I- 0 1•
-
Z /
cy shift of the quartz oscillator and adjusting the sample ::> I·
a
manipulator in an appropriate manner. ~ I"
<t 2000 I- /. -
Etch rates for Si02and Si3N4 were investigated using I •
the quartz crystal microbalance and also by etching com- I"
pletely through a film of known thickness. Penetration 1000 I- • • -
Xe F 2 in SY5te'i electron beam on
through the film was assumed to have occurred when the
underlying metal film was detected using Auger
spectroscopy.
o I.. _._o_._._.~.r -

The most interesting system, both technologically and I I

mechanistically, is the electron-stimulated etching ofSi0 2, o 500 1000 1500 2000 25000 3000
Si3N., and SiC in the presence ofXeF2gas. It has been shown TIME I SECONDS)
previously20 that XeF 2does not chemically attack Si02,
FIG, 6. Electron-assisted gas-surface chemistry using 1500-eV electrons
Si 3N., or SiC in the absence of radiation nor does electron
and XeF, on SiO" P(total) = 6X 10-' Torr with most of the ambient gas
radiation by itself produce etching. However, with electron being xenon. Neither exposure to XeF, nor an electron beam produces etch-
bombardment in the presence of XeF2, all these materials ing by itself. Simultaneous exposure produces an etch rate of - 200 A/min_

3193 J. Appl. Phys., Vol. 50, No.5, May 1979 J.W, Coburn and H.F. Winters 3193

Downloaded 16 May 2013 to 141.161.91.14. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jap.aip.org/about/rights_and_permissions
 will produce elemental Si on the surface2',25 which can be
readily removed by XeF 2• The chemistry involved is likely to
be more complex than this simple reduction and subsequent
etching of the elemental Si, but this example illustrates a
possible mechansim.
IV. TECHNOLOGICAL IMPLICATIONS
Although the role of the ions has not been fully recog-
nized in plasma etching technology, it has been established
that the anisotropic or directional nature of the etching pro-
cess is associated with ion bombardment of the etched sur-
face. 26 It is our opinion that most if not all of the directional
etching observed in plasma etching systems is caused by ra-
diation-enhanced gas-surface chemistry where the radiation
is usually energetic positive ions. However, as shown in this
paper, energetic electrons may also playa significant role in
certain situations. It should be mentioned that a wafer held
at ground potential in a glow discharge environment can still
be subjected to energetic positive ion bombardment in that
there are several conditions under which the plasma poten-
tial can be high enough to cause energetic ion bombardment
of grounded surfaces. 27 ,28 Although it is difficult to eliminate
the possibility of radiation-enhanced gas-surface chemistry
in a glow discharge system, the "tunnel" configur~tion, '9
which comes close to achieving this situation, exhibits most-
ly isotropic etching behavior. The way in which directional
etching is derived from radiation-enhanced gas-surface
chemistry can be recognized directly when one realizes that
the energetic ions are incident normally onto the etched sur-
face having been accelerated from near thermal energies
across the sheath between the plasma and the wafer. The
ions strike only the bottom surface of an etched groove or
hole, not the sidewalls. Therefore, whereas the etching of the
bottom surface proceeds at the enhanced rate, the etching of
the sidewalls proceeds at the slower unenhanced rate. That
is, the anisotropy in the etching is determined by the ratio of
the radiation-enhanced etch rate to the normal etch rate
without radiation. This ratio can be very large for certain
materials systems (e.g., fluorocarbon etching ofSiO,) or can
be made large by intentionally suppressing the normal etch
rate (e.g., addition ofH 2 to CF. while etching Si).30 In certain
situations, essentially vertical sidewalls can be obtained un-
der conditions of energetic ion bombardment. 31 ,3'
Radiation-enhanced gas-surface chemistry has some
implications in the area of fine line lithography. Broers et
al. B have generated 8o-A-wide lines where the first step in-
volves irradiating a surface (in this instance a selfsupported
thin film to reduce electron backscattering) with a finely fo-
cused electron beam in the presence of hydrocarbon vapors
to produce a thin narrow carbon line. The carbon line then
serves as a mask for subsequent etching procedures. This
electron-induced carbon deposition can be thought of as a
type of radiation-enhanced gas-surface chemistry leading to
an involatile product. The electron-enhanced oxidation of
carbon could presumably be used to produce an image of the
opposite polarity to that of Broers et al. J3 by writing on a thin
carbon film with an electron beam in the presence of a pres-
sure ( ~ 10-' Pa) of oxygen.
3194 J. Appl. Phys., Vol. 50, No.5, May 1979
Downloaded 16 May 2013 to 141.161.91.14. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jap.aip.org/about/rights_and_permissions
The electron-stimulated etching ofSiO" SiJN., and SiC
in principle allows the resistless etching of Si to be accom-
plished, The pattern of interest could be written directly into
a thin Si02 film on Si by electron beam writing in the pres-
ence ofXeF,. The patterned SiO, film could then serve as a
mask for subsequent etching of the Si if it is required. How-
ever, problems of electron dosage would appear to preclude
this approach as an on-line pattern generation procedure.
Finally, a role of oxygen in the SiF 4 -O, gas etching ofSi
can be suggested. If pure SIFt etch gas is used, the Si liberated
by the dissociation of SIFt and deposited on surfaces will
compete effectively with the Si wafers for available fluorine,
resulting in low etch rates. However, when oxygen is added,
SiO, will be formed on the surfaces in the system and on
those surfaces not SUbjected to energetic radiation the etch
rate ofSiO, will be small. Consequently, more fluorine is
available to etch the Si wafers which are SUbjected to
radiation,
V. CONCLUSION
Conclusions based on the experiments and analysis re-
ported in this paper can be summarized as follows:
(1) The etching reaction produced by fluorine atoms
interacting with Si02, Si)N., and SiC can be simulated using
a flux ofXeF, as a source of fluorine.
(2) XeF2 does not etch SiO" Si)N., or SiC in the absence
of radiation nor does electron radiation by itself produce
etching, However, these materials are chemically etched
when both XeF, and radiation (ions and electrons) are
present.
(3) Both ions and electrons can be used to produce radi-
ation-enhanced chemistry and, therefore, anisotropic etch-
ing (i.e., straight sidewalls).
(4) The rate-limiting step in the etching ofSi0 2, Si)N.,
and SiC appears to be formation of the product molecule
SiF 4 ,
(5) The etching of silicon by active gases (e.g., XeF 2, F,)
is enhanced by argon ion bombardment. The mechanism
believed to be operative in producing this phenomena is that
ion bombardment enhances the rate that the active gas is
dissociatively chemisorbed at the silicon surface.
(6) Many of the differences observed between the plas-
ma etching of silicon and its compounds can be attributed to
the fact that fluorine atoms spontaneously attack silicon,
while both fluorine atoms and radiation (or elevated tem-
perature) are required to achieve sustained etching ofSiO"
Si)N., and SiC.
(7) The fact that fluorine etches SiOh Si)N., and SiC
only where electrons hit the surface creates the possibility for
maskless lithography. This process would, however, require
large electron dosages.
(8) It has been suggested that CF3-type radicals would
preferentially attack SiO" whereas F atoms would preferen-
tially attack silicon. The present work shows that this is not
completely true since fluorine rapidly attacks SiO, in the
presence of radiation.
J.W. Coburn and H.F. Winters 3194
 ACKNOWLEDGMENTS
The authors are very appreciative of the technical skills Si F4 Pressure Determined from Si F3+ Mass Peak
of R. W. Sadowski and J. E. Schlaegel and have benefited P (XeF2 )= 2xl0- 7 Torr
from numerous technical discussions with B.N. Chapman, P (Ar) = lxl0- 4 Torr
M. Chen, E. Kay, K. Lee, and V. Minkiewicz. 8
p (Xe) ~ 3x 10- 6 Torr
APPENDIX
i-_ •
Mauer et al. 16 have suggested that the increase in the
etch rate of silicon caused by ion bombardment in a fluoro- I
I' • • ......
>- 6 I
carbon glow discharge could be due to a chemical enhance- I
ment of the physical sputtering process. The implication is ~ I
~ I
that partially fluorinated silicon (i.e., SiF, SiF 2, SiF3) may '" 5 I
,'-- Silicon rotated into XeF2
have a much higher sputtering yield than elemental silicon. ~
:::J , and Ar+ (500 eV) Beams
The large difference in the sputtering yield of fluorinated ~'" 4
,
,
carbon (Teflon) relative to carbon is an example of such a Cl.

li.""" I
situation. The fact that the etch product observed in plasma (/) 3 I
etching systems during the etching of silicon is almost exclu- I
I
sively SiF4 cannot be used to eliminate this process for the I
2 I
following reason. If the primary etch product were SiF, SiF2, I • • •
or SiF3, as would be predicted from this chemically en- : j*" ••
hanced sputtering model, these species would probably react 1 ..!~: : :: :', ,l'--Silicon rotated into
with fluorine atoms in the gas phase or on nearby surfaces to Background Gas XeF2 Beam
form SiF4 which then appears in the effluent gas as the etch OL-__ ~ __ ~ ____ ~ __ ~ __ ~ ____ ~ __ ~

product. o 200 400 600 800 1000 1200 1401

Recently an experiment has been performed in which
the density of active fluorine is so low that the evolution of
SiF4cannot be accounted for by such a two-step process. The
data is shown in Fig. 7 in which the SiF3+ peak is monitored
as a Si sample is rotated into ajet ofXeF 2 gas with and
without argon ion bombardment. It should be mentioned
that the parent SiF / peak exhibits a similar behavior to that
shown in Fig. 7 for SiF3+ and therefore the SiF / peak repre-
sents the SiF4 partial pressure. The ion-enhanced etching of
silicon is indicated by the much larger evolution of SiF4when
the sample is exposed to XeF 2 with Ar+ bombardment as
compared to XeF 2 exposure alone. . uct formed when a silicon surface is subjected to ion bom-
In this experiment, the background pressure ofXeF2 is
about 3 X 10- 5 Pa (2 X 10-7 Torr) correpsonding to a flux of
XeF2 of about 3 X 1013 molecule/cm 2 sec. This partial pres-
sure was measured mass spectrometrically using published
fragmentation patterns for XeF2 .34 At these pressures it is
clear that gas-phase processes can be neglected and there-
fore, ifan SiFx radical (x = 1,2,3) reacts with XeF 2 to form
SiF4, the reaction must occur after the SiFx radical has ad-
sorbed on a wall of the vacuum system. However, if we as-
sume the SiFx species have a cross section of about 10- 15 cm 2 ,
the mean time for a XeF rSiF x collision is about 30 sec at this
low XeF2 pressure. Furthermore, the reaction probability in
the absence of ion bombardment (as is the case for all wall
surfaces in this system) cannot exceed about 0.1 since the
reaction rate can be enhanced by a factor of 10 by ion bom-
bardment (Fig. 2). Therefore, the mean reaction time for
converting SiFx species to SiF4 would appear to be of the
order of 300 sec in this environment. However, the evolution
of SiF4 with ion bombardment responds essentially instanta-
neously as is shown in Fig. 7. This, in our view, indicates that
the SiF4 is formed on the silicon surface and not in a two-step
process involving physical sputtering of SiFx and subse-
Time (seconds)
FIG. 7. The SiF,pressure as a silicon surface is rotated into a beam ofXeF,
molecules with and without Ar' ion bombardment. The SiF, pressure was
determined mass spectrometrically. The ratios of the pressure increases (top
curve to bottom curve), indicate the magnitude of ion-assisted etching. The
experiments were performed in a standard UHV system of the type often
used for surface experiments. The experimental arrangement was similar to
that described in Fig. I.
quent fluorination. Furthermore, the relative magnitudes of
the SiF4 increases depicted in Fig. 7 corresponds very rough-
ly to the relative etch rates measured on the quartz crystal
microbalances, suggesting that SiF4 is the major etch prod-
bardment in the presence of active fluorine.
This argument has implicitly assumed that the two-step
process involves adsorption of the SiFx radical on a wall and
subsequent fluorination by XeF2 • It is possible that the SiFx
radical could combine with fluorine which is already present
on the wall presumably in the form of a fluoride of the wall
material. If this process were to occur, the SiF4 evolution
would be essentially instantaneous as shown in Fig. 7 and
although we consider this process to be unlikely, we have no
data which can be used to eliminate this possibility.
'W.A. Knox, Ultramicroscopy 1,175 (1976).
'H.H. Madden, J. Vac. Sci. Techno!. 13, 228 (1976).
'H.G. Heide, Lab. Invest. 14, 1134 (1965).
'Y. Margoninski, O. Segal, and R.E. Kirby, Surf. Sci. 53, 488 (1975).
'W. Ranke and K. Jacobi, Surf. Sci. 47 525 (1975).
'J. Verhoeven and J. Los, Surf. Sci. 58,566 (1976).
'Y. Margoninski, J. App!. Phys. 47,3868 (1976).
'J. Verhoeven, Proceedings of the Seventh International Vacuum Congress
and the Third International Conference on Solid Surfaces, edited by R.
Oobrozemsky, F. Riidenauer, FP. Viehbi:ick, and A. Breth (R. Oobro-

3195 J. Appl. Phys .• Vol. 50, No.5, May 1979 J.W. Coburn and H.F. Winters 3195

Downloaded 16 May 2013 to 141.161.91.14. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jap.aip.org/about/rights_and_permissions
 zemsky et al., Vienna, 1977), p. 915. UK.C. Pandey, T. Sakurai, and H.D. Hagstrum, Phys. Rev. B 16, 3648
'J.H. Neave and B.A. Joyce, J. Phys. D 10, 243 (1977). (\977).
IOH.G. Tompkins, Surf. Sci. 62, 293 (1977). "S. Thomas, J. Appl. Phys. 45, 161 (\974).
"N. Hosokawa, R. Matsuzaki, and T. Asamaki, Jpn. J. Appl. Phys. Suppl. "8. Carriere and B. Lang, Surf. Sci. 64, 209 (1977).
2, Pt. 1,435 (1974). "G.e. Schwartz, L.B. Zielinski, and T. Schopen in Etching, edited by M.1.
"L. Holland and S.M. Ojha, Vacuum 26, 53 (1976). Rand and H.G. Hughes (Electrochemical Society Princeton, 1976) p. 122.
ilL. Holland and S.M. Ojha, Vacuum 26,233 (1976). 'TW. Coburn and Eric Kay, J. Appl. Phys. 43, 4965 (1972).
"L. Holland, 1. Vac. Sci. Technol. 14, 5 (1977). "J.L. Vossen, J. Electrochem. Soc. 126, 319 (1979).
"H.F. Winters, J. Appl. Phys. 49,5165 (1978). "R. Bersin and M. Singleton, U.S. Patent No. 3879597.
i61.L. Mauer, 1.S. Logan, L.B. Zielinski, and G.e. Schwartz, 1. Vac. Sci. 'oR.A.H. Heinecke, Solid-State Electron. 18, 1146 (1975).
Technol. IS, 1734 (1978). "H.W. Lehmann and R. Widmer, J. Vac. Sci. Techno!. IS, 319 (1978).
"O.E. Yonts and D.E. Harrison, J. Appl. Phys. 31, 1583 (1960). "P.M. Schaible, W.e. Metzger, and J.P. Anderson, J. Vac. Sci. Techno!. 15,
liM. Cantrel and M. Marchal, 1. Mater. Sci. 8, 1711 (1973). 334 (1978).
"J.W. Coburn, H.F. Winters, and T.1. Chuang, J. Appl. Phys. 48,3532 "A.N. Broers, W.W. Molzen, J.1. Cuomo, and N.D. Wittels, App!. Phys.
(1977). Lett. 29,596 (1976).
2"H.F Winters and J.W. Coburn, Appl. Phys. Lett. 34, 70 (1979). )'W.E. Falconer, M.J. Vasile, and FA. Stevie, J. Chern. Phys. 66, 5335
"A.K. Kuriakose and J.L. Margrave, 1. Phys. Chern. 68, 2671 (1964). (1977)
"J.E. Rowe, G. Margaritondo, and S.B. Christmas, Phys. Rev. B 16, 1581
(1977).

3196 J. Appl. Phys., Vol. 50, No.5, May 1979 J.W. Coburn and H.F. Winters 3196

Downloaded 16 May 2013 to 141.161.91.14. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jap.aip.org/about/rights_and_permissions