Highly selective etching of silicon nitride over silicon and silicon dioxide
B. E. E. Kastenmeier, P. J. Matsuo, and G. S. Oehrleina)
Department of Physics, The University of Albany, State University of New York, Albany, New York 12222
共Received 29 January 1999; accepted 27 August 1999兲
A highly selective dry etching process for the removal of silicon nitride 共Si3N4) layers from silicon
and silicon dioxide (SiO2) is described and its mechanism examined. This new process employs a
remote O2 /N2 discharge with much smaller flows of CF4 or NF3 as a fluorine source as compared
to conventional Si3N4 removal processes. Etch rates of Si3N4 of more than 30 nm/min were achieved
for CF4 as a source of fluorine, while simultaneously the etch rate ratio of Si3N4 to polycrystalline
silicon was as high as 40, and SiO2 was not etched at all. For NF3 as a fluorine source, Si3N4 etch
rates of 50 nm/min were achieved, while the etch rate ratios to polycrystalline silicon and SiO2 were
approximately 100 and 70, respectively. In situ ellipsometry shows the formation of an
approximately 10-nm-thick reactive layer on top of the polycrystalline silicon. This oxidized
reactive layer suppresses etching reactions of the reactive gas phase species with the silicon.
© 1999 American Vacuum Society. 关S0734-2101共99兲05706-1兴
I. INTRODUCTION
The stripping of silicon nitride mask material after the
local oxidation of silicon 共LOCOS兲 is a possible source of
device damage during integrated circuit 共IC兲 fabrication. The
pad oxide can suffer degradation during the overetch. Furthermore, etchants can reach the underlying Si substrate
through imperfections in the pad oxide, and etch the substrate at a significant rate. This effect leaves behind craters in
the substrate, and is referred to as ‘‘pitting.’’ Current dry
processes used for the Si3N4 stripping step favor this undesired effect, since they typically etch Si much faster than
Si3N4.1–4 Therefore, a process which etches Si3N4 selectively
over both SiO2 and Si is desirable.
In previous work by this group2,5 and by Blain et al.,3 the
etching of Si3N4 in the afterglow of CF4 /O2 /N2 and NF3 /O2
discharges has been investigated. The flow of the fluorine
source, CF4 or NF3, was kept at constant values for most
experiments, and O2 and N2 were added in varying amounts.
The etch rate of Si3N4 was found to be correlated to the
density of nitric oxide 共NO兲 in a linear relationship for both
gas mixtures. No correlation was observed between the
Si3N4 etch rates and the density of atomic fluorine. The F
atom concentration, determined from the polycrystalline silicon etch rate2,6 and mass spectrometry, was found to be
higher by at least a factor of 20 than necessary to sustain the
measured Si3N4 etch rates. We concluded that the arrival of
NO is the etch rate limiting step in Si3N4 etching, and F
atoms are available in abundance for Si3N4 etching.
The etch rates of SiO2 共Refs. 2, 5兲 for the same parameters were found to be independent of the NO concentration,
but followed the F radical density very well. The CF4 based
chemistry produces the desired ratio between the etch rates
of SiO2 and Si3N4. With O2 and N2 added to CF4, ratios of 10
or slightly higher could be obtained easily. The etch rate of
a兲
Electronic mail: [email protected]
3179
J. Vac. Sci. Technol. A 17„6…, Nov/Dec 1999
Si3N4 typically was about 30 nm/min, that of SiO2 3 nm/min.
Typical etch rate values for SiO2 and Si3N4 in the afterglow
of NF3 /O2 discharges were 60 and 80 nm/min, respectively,
too high to achieve selective etching of Si3N4 relative to
SiO2.
Etching of Si4,7 always occurred at rates much faster than
those of Si3N4. Etch rates as high as 300 nm/min for
CF4 /O2 /N2 and 700 nm/min for NF3 /O2 gas mixtures were
measured. These high etch rates can be explained with the
high F density and the spontaneous reaction of the Si surface
with F atoms. The etch rate of Si varies linearly with the
fluorine density.6 A decrease of the Si etch rate is consistently observed for additions of 20% or more of O2 to the
CF4 or NF3 plasma. This effect is due to the oxidation of the
Si surface in the presence of O and/or O2. The oxidation
makes the Si surface very similar to that of SiO2 during
etching, and in the limit of very high flows of O2, the etch
rates of both materials become equal.
Based on the above observations, a new process for the
etching of Si3N4 was developed, that provides high selectivities to both SiO2 and Si. A mixture of O2 and N2 was used as
the primary discharge gas, to which small amounts of a fluorine source 共CF4 or NF3) were admitted. Three mechanisms
contributing to a high Si3N4 /Si selectivity are exploited by
choosing an O2 /N2 based chemistry, rather than a fluorine
based one.
共1兲 Decreased F atom density. The density of atomic fluorine in the reaction chamber is very low as compared to
CF4 /O2 /N2 and NF3 /O2 gas mixtures. The Si etch rate decreases in the same way as the F density, since they are
linearly correlated. The etch rate of Si3N4, on the other hand,
should not be affected as strongly, because F is available in
abundance in the conventional gas mixtures and the etching
of Si3N4 in those cases is not limited by the arrival rate of F.
0734-2101/99/17„6…/3179/6/$15.00
©1999 American Vacuum Society
3179
3180
Kastenmeier, Matsuo, and Oehrlein: Highly selective etching of silicon nitride
FIG. 1. Schematic of the chemical downstream etcher used in this investigation. The gases are fed into the sapphire applicator, where a microwave
discharge is ignited. The species effluent from the plasma travel through
tubing of variable length and lining material to the reaction chamber. The
sample is placed on the center of an electrostatic chuck. A quadrupole mass
spectrometer is mounted on the chamber on top of the sample, and monochromatic ellipsometry is used to determine etch rates.
共2兲 Oxidation of the Si surface. The second mechanism
that suppresses the Si etch rate is the oxidation of the Si
surface in the presence of O and O2 in the gas phase. The
oxidized Si surface is very inert to the attack of F atoms,
therefore, the Si etch rate is significantly decreased. In addition to oxygen species, NO, which is produced in the O2 /N2
discharge, was found to contribute to the oxidation of Si
surfaces.4 The surface of Si3N4 is only slightly oxidized during etching in the presence of O and O2, not enough to influence the etch rates.5,8
共3兲 Si3N4 etch rate enhancement by NO. The etch rate of
Si3N4 is proportional to the density of NO. The O2 /N2 chemistry produces a significant amount of NO in its afterglow,
which allows for high Si3N4 etch rates. The Si etch rate is
less influenced by NO addition, and in the surface oxidation
limited regime no enhancement of the etch rate is observed.4
This process also provides selectivity of Si3N4 over SiO2.
The etch rate of SiO2 in the absence of ion bombardment
depends only on the F density, and is expected to be smaller
than the Si etch rate.
In Sec. II the setup of the remote plasma etching tool and
the experimental procedure are described. Subsequently, etch
rates of Si3N4, poly-Si, and SiO2 are reported for both CF4
and NF3 as sources of F. Gas phase experiments 共Ar actinometry and mass spectrometry兲 are conducted to gain information about the concentration of reactive species. The surface of poly-Si suring etching is examined by ellipsometry,
and the close correlation between the decrease of the etch
rate and the formation of a reactive layer on the Si is demonstrated.
II. EXPERIMENT
The remote plasma etching tool has been described before
in detail.2,4,5 Figure 1 shows a schematic of the apparatus
used for the experiments. Mixtures of O2, N2, and the fluorine source 共 CF4 or NF3) are excited using an ASTEX 2.45
GHz microwave applicator with a sapphire coupling tube.
The species produced in the plasma travel through a transJ. Vac. Sci. Technol. A, Vol. 17, No. 6, Nov/Dec 1999
3180
port tube to the cylindrical reaction chamber. The length of
the tube was fixed at 75 cm, and the inside was lined with a
Teflon liner. Samples of size 1 in. by 1 in. are glued on a 5
in. carrier wafer, which is placed on an electrostatic chuck in
the reaction chamber. The materials used for this investigation are low-pressure chemical vapor deposition 共LPCVD兲
Si3N4 and thermally grown SiO2. Surface modifications of
Si were studied using crystalline Si and polycrystalline Si,
etch rates of silicon were determined using polycrystalline
silicon. The temperature of the sample is monitored with a
fluoroptic probe which contacts the backside of the sample.
It was kept constant at 10 °C for all experiments. Helium at a
pressure of 5 Torr was fed between the surface of the electrostatic chuck and the carrier wafer in order to obtain a good
heat conduction. Etch rates are measured in situ by monochromatic ellipsometry 共wavelength 632.8 nm兲. Some analytical experiments were performed in order to gain a qualitative understanding of the mechanisms responsible for
achieving the high selectivity. A fiber optic cable for optical
emission experiments of the discharge was mounted on the
housing of the applicator. The spectrograph used in this investigation is a 30 cm optical multichannel analyzer. A quadrupole mass spectrometer is mounted on top of the reaction
chamber such that the distances from the discharge to the
sample and to the orifice are the same. Some samples were
moved to the surface analysis chamber through the ultrahigh
vacuum 共UHV兲 wafer handling system without exposure to
air. Results of similar analytical experiments on a CDE system are reported in greater detail in previous
publications.2–5,7
The experiments were conducted with a sapphire applicator at 1000 W of microwave power and a chamber pressure
of 600 mTorr. Flows of O2 and N2 were kept constant at 800
and 110 sccm, respectively, for most experiments. These parameters are referred to as the standard conditions.
III. RESULTS AND DISCUSSION
The Si3N4 etch rates were measured as a function of CF4
addition to a O2 /N2 plasma 共see top panel of Fig. 2兲. The
etch rate increases linearly with the flow of CF4. The parameter for the two curves in the top panel of Fig. 2 is the
amount of oxygen fed into the discharge. A lower flow of O2
共300 sccm兲 yields an etch rate about 15 nm/min higher than
the flow of O2 at standard conditions 共800 sccm兲. The highest
etch rates, obtained with 46 sccm of CF4, are 24 nm/min for
800 sccm of O2, and 39 nm/min for 300 sccm of O2.
Nitrogen trifluoride was used as an alternative fluorine
source to CF4. The plasma chemistry of NF3 is significantly
different than that of CF4.5,9,10 The threshold for electron
impact dissociation of NF3 is 0 eV,11,12 that of CF4 is 12.6
eV.13 This results in a higher degree of dissociation of NF3 in
a discharge. In fact, in the high density microwave discharges employed for the work reported here, 100% dissociation is typically achieved.5,14 The dissociation of CF4 for
3181
Kastenmeier, Matsuo, and Oehrlein: Highly selective etching of silicon nitride
3181
FIG. 3. Etch rates of poly-Si for CF4 and NF3 as source of fluorine at
standard conditions. The samples were HF dipped before the experiments to
remove the native oxide layer. The initial high etch rate is suppressed as an
oxygen rich reaction layer forms during etching.
FIG. 2. Etch rates of Si3N4 as a function of the flow of CF4 共top panel兲 and
NF3 共bottom panel兲.
similar discharge parameters varies between 40% and 60%.
Therefore, the production of free fluorine radicals from NF3
is higher than from CF4. Moreover, the concentration of NO
in the afterglow of the discharge might be affected by the
additional N atom feed.
The etch rates of Si3N4 as a function of the flow of NF3
are shown in the bottom panel of Fig. 2. The measurements
were performed under standard conditions, i.e., the flow of
O2 was kept constant at 800 sccm. The etch rates of Si3N4
are proportional to the flow of NF3, and significantly higher
if NF3 is used instead of CF4. An addition of 46 sccm of CF4,
e.g., yields an etch rate of 24 nm/min, whereas the etch rate
for the same flow of NF3 is twice as high 共49 nm/min兲.
The etch rates of polycrystalline silicon were found to
depend strongly on the initial surface conditions of the
sample, and on the etch time. These etch rate variations all
can be explained with the presence of a native oxide layer on
the Si surface, and the formation of a reactive layer during
etching. A layer of native oxide on a polycrystalline silicon
film suppresses the etching almost completely. As an example, the etch rate of poly-Si at standard conditions and 30
sccm of CF4 is as low as 0.07 nm/min with the native oxide
layer present. If the native oxide is removed immediately
before processing by dipping the sample in HF for 3 s, the
etch rate is 0.49 nm/min, a sevenfold increase. In Fig. 3 the
etch rates of poly-Si are plotted as a function of time. The
native oxide layer was removed before the experiments by
HF dipping. Both samples were treated at standard conditions, with CF4 and NF3 as a source of fluorine. The etch rate
of a ‘‘clean’’ Si surface 共no native oxide and no reactive
layer兲 can be as high as 20 nm/min. As etching proceeds, the
reactive layer forms on the Si and impedes etching reactions.
The final etch rates are 0.50 nm/min for both CF4 and NF3.
The etch rate decreases faster if CF4 is used as fluorine
source. Extrapolation of the curves to the x axis, using the
initial slopes, yields 12 s for CF4, and 40 s for NF3 as estiJVST A - Vacuum, Surfaces, and Films
mates for the decay time. All etch rates reported in the following are ‘‘final’’ rates, measured after 300 s of etching.
Figure 4 shows the etch rates of silicon as a function of
CF4 共top panel兲 and NF3 addition 共bottom panel兲. Again, the
parameter of the two curves in the top panel is the amount of
O2 fed into the plasma. Strong variations of the etch rate with
respect to the O2 flow are observed. The etch rate is significantly suppressed if 800 sccm of O2 are used. No etch rate
suppression is found at a low flow 共300 sccm兲 of O2. At 30
sccm of CF4, for example, the etch rate for 300 sccm of O2 is
19 times larger than that for 800 sccm. The Si etch rate is
proportional to the flow of NF3 at small flows up to 20 sccm.
At higher NF3 flows, the etch rate assumes a plateau value at
0.5 nm/min.
The etch rates of silicon dioxide are plotted in Fig. 5. The
top panel shows etch rates when CF4 is used as a fluorine
source. Etching occurs only at the low flow of O2 共300
FIG. 4. Etch rate of poly-Si vs the flow of CF4 共top panel兲 and NF3 共bottom
panel兲. The ellipsometric determination of an etch rate for 300 sccm of O2
and flows of CF4 higher than 30 sccm was not possible, probably due to the
formation of roughness on top of the poly-Si.
3182
Kastenmeier, Matsuo, and Oehrlein: Highly selective etching of silicon nitride
FIG. 5. Etch rates of SiO2. At the high O2 flow with CF4 added, the etch
rates were too small to be detected by ellipsometry 共⬍0.05 nm/min兲.
sccm兲, and etch rates never exceed 1 nm/min. The etch rates
at 800 sccm of O2 were too small to be detected by ellipsometry 共⬍0.5 nm/min兲. In the bottom panel, the etch rates for
800 sccm of O2 and NF3 as a fluorine source are shown. The
etch rates increase linearly for NF3 flows up to 30 sccm, and
then level out at 0.8 nm/min.
The Si3N4 /Si etch rate ratio obtained from Fig. 2 and Fig.
4 is shown in Fig. 6. If CF4 is used as a fluorine source 共top
panel兲, the highest selectivity is achieved at the high O2 flow
and intermediate flows of CF4. At these conditions, the selectivity assumes a peak value of around 40. The selectivity
for the low O2 flow is much less due to the higher Si etch
rates. If NF3 is used 共bottom panel兲, the selectivity increases
3182
FIG. 7. Ratio of the Si3N4 and SiO2 etch rates, determined from the data
shown in Fig. 2 and Fig. 5. The curve for the high O2 flow in the top panel
is missing, since the selectivity at these conditions is extremely high.
with increasing flow of NF3. A selectivity value of 100 is
reached at standard conditions.
In Fig. 7, the ratio of the Si3N4 and SiO2 etch rates is
shown. These data are obtained from Fig. 2 and Fig. 5. In the
case of CF4 as fluorine source 共top panel兲, the curve for the
high O2 flow is missing. These values are extremely high
共⬎500 or infinity兲 due to the unmeasurably small SiO2 etch
rates at these conditions. The selectivities for NF3 as a fluorine source 共bottom panel兲 are at around 60 for intermediate
flows of NF3, and have an increasing tendency at higher
flows.
The selectivities reported here are improved significantly
as compared to those obtained in earlier work. In CF4 based
etching of Si3N4, a typical selectivity to the underlying pad
oxide of 10 is achieved. For a NF3 based chemistry, which is
predominantly used for chamber cleaning after deposition
processes, this value is less than half as big. In both chemistries, Si is etched at a significantly faster rate than Si3N4. In
Table I, the Si3N4 etch rates and selectivities to SiO2 and Si
are summarized for these processes, together with the data
presented above.
Several experiments were performed to investigate quantitatively the importance of the three mechanisms mentioned
above for achieving high Si3N4 /SiO2 and Si3N4 /Si selectivities while maintaining a high Si3N4 etch rate. As described in
previous work,2,5 argon actinometry and quadrupole mass
TABLE I. Typical values of the Si3N4 etch rate and the selectivities to SiO2
and Si. See Refs. 2 and 4 for the work on CF4 /O2 /N2, and Refs. 5 and 7 for
NF3 /O2.
CF4 /O2 /N2 NF3 /O2 O2 /N2 /CF4 O2 /N2 /NF3
FIG. 6. Ratio of the Si3N4 and Si etch rates, determined from the data shown
in Fig. 2 and Fig. 4.
J. Vac. Sci. Technol. A, Vol. 17, No. 6, Nov/Dec 1999
Si3N4 etch rate 共nm/min兲
Sel Si3N4 /SiO2
Sel. Si3N4 /Si
30
10
0.1–0.6
80..350
0.7–4
0.1–0.8
10..20
⬎500
30
30..50
60
80–100
3183
Kastenmeier, Matsuo, and Oehrlein: Highly selective etching of silicon nitride
3183
FIG. 9. Signal intensity of the NO⫹ peak at amu 30 in the reaction chamber.
FIG. 8. Variation of the F atom concentration in the discharge region, as
determined by Ar actinometry. See the text for the parameters used for this
measurement.
spectrometry were used to measure the change of the concentration of species relevant for etching. For the actinometry measurements,15,16 argon at a constant flow rate of 60
sccm was injected into the discharge. Figure 8 shows the
intensity ratio of the F emission at 703.7 nm and the Ar line
at 811 nm. This ratio is proportional to the F atom density in
the discharge. The top panel shows the ratio for CF4 , the
bottom panel for NF3 as a source of fluorine. The ratio is
approximately 4 times higher if NF3 is used as compared to
CF4 , indicating a higher F atom concentration in the case of
NF3 . This is consistent with the observation that NF3 is dissociated to a higher degree than CF4 in microwave
discharges.5 For the present experiments, mass spectrometry
measurements of the NF⫹
3 peak 共amu 71兲 for plasma-on and
plasma-off states show the complete destruction of NF3 . The
corresponding measurement of the CF⫹
3 intensity 共amu 69兲
suggests a dissociation of the CF4 of about 20%–30%, if it is
assumed that the contribution of CF3 radicals produced in the
discharge of the signal intensity of the peak at amu 69 is
negligible.
A significant amount of NO is present in the reaction
chamber during the experiments reported here. The intensity
of the NO⫹ peak at amu 30 is shown in Fig. 9. In the case of
CF4 共top panel兲, the signal intensity decreases by almost
50% as a small amount of CF4 共15 sccm兲 is added to the
O2 /N2 discharge, and then decreases further as more CF4 is
injected. The same initial decrease is observed for NF3 added
to O2 /N2 共bottom panel兲, but then, in contrast to the CF4
case, the NO density increases as more NF3 is injected.
The proportionality of both the actinometrically determined F density in the discharge and the Si3N4 etch rate with
the amount of F-containing gas added suggests that the Si3N4
etch rate is limited by the amount of F atoms available for
etching reactions. No correlation is observed between the
NO density in the reaction chamber and the Si3N4 etch rate.
JVST A - Vacuum, Surfaces, and Films
This is in contrast to CF4 and NF3 based processes, in which
small amounts of O2 and/or N2 are added to the F-containing
gas. In these cases, F atoms are available in abundance, and
the Si3N4 etch rates were found to be limited by the NO
density.2,5
As mentioned above, a thick reactive layer on top of the
Si surface inhibits etching reactions. The formation of this
layer was monitored in situ during the etching by monochromatic ellipsometry. Figure 10 shows the time evolution of
FIG. 10. Evolution of the ellipsometric variables ⌿ and ⌬ during the etching
of a film of polycrystalline silicon 共total thickness 250 nm兲 on a SiO2 layer
共100 nm兲 on a Si共100兲 substrate. The experiments were performed under
standard conditions, with addition of 30 sccm of CF4 共top panel兲 or 30 sccm
of NF3 共bottom panel兲. The time interval between two data points is 1.1 s.
Simulations of ⌿/⌬ for different reactive layer thickness 共0, 10, and 12.5
nm兲 on top of the poly-Si are included. An increase in ⌬ is equivalent to the
removal of the poly-Si and increase in ⌿ means the growth of the oxidized
reactive layer.
3184
Kastenmeier, Matsuo, and Oehrlein: Highly selective etching of silicon nitride
the ellipsometric variables ⌿ and ⌬ as a function of time
共dotted curves, top panel for CF4 , bottom panel for NF3 ). A
multilayer stack consisting of poly-Si 共total thickness 250
nm兲 on top of SiO2 共100 nm兲 on a Si共100兲 substrate was
etched. The lines in Fig. 10 show calculations of the ellipsometric variables ⌿ and ⌬. These can be calculated by assuming values for the thickness and the optical parameters of
each layer of the stack. Also the reactive layer can be included in those calculations as a film of a certain thickness
on top of the poly-Si. In Fig. 10, the results of these calculations for three different thicknesses of the reactive layer 共0,
10, and 12.5 nm兲 are included. Etching of the poly-Si corresponds to an increase of the value of ⌬. At the same time,
information about the reactive layer thickness is obtained
from ⌿. An increase of ⌿ means an increase of the reactive
layer thickness.
The top panel of Fig. 10 shows the etching of polycrystalline Si and formation of the reactive layer under standard
conditions with 30 sccm of CF4 added. Immediately after the
plasma is ignited and tuned 共‘‘start’’兲 the reactive layer
grows at a considerable rate, and Si is removed at the same
time. The time interval between two measurement points is
1.1 s. The reactive layer assumes a steady state thickness of
approximately 10 nm. At this thickness value the etch rate of
Si is slowed down to 0.4 nm/min. At the moment the discharge is terminated 共‘‘end’’兲, the overlayer thickness has
increased to about 12.5 nm. The etch time from start to end
was 475 s, and 16.7 nm of Si was removed in that time. After
termination of the discharge 共‘‘postplasma’’兲, more surface
modifications occurred.
NF3 was used instead of CF4 for the experiment shown in
the bottom panel. As in the case of CF4 , growth of the reactive layer and poly-Si etching occur simultaneously. However, some significant differences are observed: The formation of the reactive layer happens at a slower rate than in the
case of CF4 etching, and the reactive layer does not achieve
a steady state thickness, but continues to grow. Also the
postplasma modifications show a different trend.
IV. CONCLUSIONS
A novel remote plasma chemical dry etching 共CDE兲 process which enables selective etching of Si3N4 over silicon
and SiO2 with an etch rate ratio greater than 30:1 has been
demonstrated. It uses high flows of O2 and N2, and relatively
small additions of CF4 and NF3 as a source of fluorine.
For the experiments reported here, the silicon nitride etch
rate is limited by the flux of F atoms to the surface. Nitric
oxide is available in abundance and does not influence the
Si3N4 etch rate. This is the inverse of processes in which
small amounts of O2 and/or N2 are added to a CF4 or NF3
discharge.
The significantly decreased F atom density, especially for
the case of CF4 as a fluorine source, is not sufficient to etch
Si3N4 with a high selectivity over silicon. Also the Si3N4 etch
rate boosted by the high NO density in the chamber is not
sufficient for a high etch rate ratio. From the data presented
J. Vac. Sci. Technol. A, Vol. 17, No. 6, Nov/Dec 1999
3184
above, it can be concluded that the formation of an etchinhibiting reactive layer on top of the Si is the dominant
mechanism for achieving a high Si3N4 etch rate ratio. The
etching of the virgin silicon surface proceeds at a rate of
approximately 20 nm/min. In situ ellipsometry shows the
formation of a reactive layer on top of the polycrystalline
silicon during the etch process within a matter of seconds.
The etch rate of silicon is decreased to a level comparable to
that of SiO2 after the reactive layer has formed.
Finally, the authors would like to suggest that imperfect
gate oxides can potentially be improved by the process investigated here. The reactive layer that forms on top of the Si
during etching is highly oxidized with some F content, and
thus similar to the gate oxide. It is possible that the Si underneath voids or imperfections in the gate oxide are oxidized during the overetch period after the Si3N4 removal,
thus improving the electrical properties of the gate.
ACKNOWLEDGMENTS
The authors would like to thank Matt Blain, Gary Powell,
John Langan, Rob Ellefson, and Lou Fries for their support
and fruitful discussions. Marc Schaepkens, Theo Standaert,
and Christian Völker are thanked for their helpful ideas and
technical assistance. We would like to acknowledge Matrix
Integrated Systems, Sandia National Laboratories, Air Products and Chemicals, and Leybold Inficon for financial support of this study.
1
N. Hayasaka, H. Okano, and Y. Horiike, Solid State Technol. 31, 127
共1988兲.
2
B. E. E. Kastenmeier, P. J. Matsuo, J. J. Beulens, and G. S. Oehrlein, J.
Vac. Sci. Technol. A 14, 2802 共1996兲.
3
M. G. Blain, T. L. Meisenheimer, and J. E. Stevens, J. Vac. Sci. Technol.
A 14, 2151 共1996兲.
4
P. J. Matsuo, B. E. E. Kastenmeier, J. J. Beulens, and G. S. Oehrlein, J.
Vac. Sci. Technol. A 15, 1801 共1997兲.
5
B. E. E. Kastenmeier, P. J. Matsuo, G. S. Oehrlein, and J. G. Langan, J.
Vac. Sci. Technol. A 16, 2047 共1998兲.
6
C. J. Mogab, A. C. Adams, and D. L. Flamm, J. Appl. Phys. 49, 3796
共1978兲.
7
P. J. Matsuo, B. E. E. Kastenmeier, G. S. Oehrlein, and J. G. Langan, J.
Vac. Sci. Technol. A 17, 2431 共1999兲.
8
R. A. Gottscho, C. W. Jurgensen, and D. J. Vitkavage, J. Vac. Sci. Technol. B 10, 2133 共1992兲.
9
J. A. Barkanic, D. M. Reynolds, R. J. Jaccodine, H. G. Stenger, J. Parks,
and H. Vedage, Solid State Technol. 32, 109 共1989兲.
10
N. J. Ianno, K. E. Greenberg, and J. T. Verdeyen, J. Electrochem. Soc.
128, 2174 共1981兲.
11
R. M. Reese and V. H. Dibeler, J. Chem. Phys. 24, 1175 共1956兲.
12
J. C. J. Thynne, J. Phys. Chem. 63, 1586 共1969兲.
13
L. G. Christophorou, A. J. K. Olthoff, and M. V. V. S. Rao, J. Phys.
Chem. Ref. Data 25, 1341 共1996兲.
14
B. E. E. Kastenmeier, P. J. Matsuo, G. S. Oehrlein, and J. G. Langan
共unpublished兲.
15
V. M. Donnelly, D. L. Flamm, W. C. Dautremont-Smith, and D. J. Werder, J. Appl. Phys. 55, 242 共1984兲.
16
R. E. Walkup, K. L. Saenger, and G. S. Selwyn, J. Chem. Phys. 84, 2668
共1986兲.