Nickel electroless deposition process on chemically pretreated Si(100) wafers in aqueous alkaline solution

Article abstract of DOI:10.1016/S0013-4686(02)00838-1

Pretreating Si wafer surfaces with hydrochloric acid and hydrogen peroxide mixture (HPM) or ethanol was found to enhance the reactivity of chemical Ni deposition on Si(100) wafers in a simple bath of NiSO₄-(NH₄)₂SO₄ at pH 9.0. This phenomenon was identified as the acceleration of anodic reaction involved in chemical Ni deposition reaction on Si wafer surfaces, namely oxidation process of surface Si. Just after immersing into the alkaline bath, a reactive surface where oxidation reaction of Si was accelerated was formed on Si wafer surface with HPM or ethanol pretreatment. On the reactive surface, uniform and glossy Ni deposition film was obtained. In order to clarify the effects of HPM- or ethanol-pretreatment on anodic reaction, the pretreated Si(100) wafers were immersed into aqueous alkaline solution excluding NiSO₄, resulting acceleration of Si oxidation compared to the unpretreated Si(100) surface. The progress of surface reactivity was also clarified by open circuit potentials (OCP), XPS, and ex-situ ATR FTIR. Moreover, by using this pretreatment, selective deposition onto nano-patterned Si substrate was performed. A two-dimensional array of fine Ni dots (diameter ca. 80 nm) was successfully fabricated.

Full text of DOI:10.1016/S0013-4686(02)00838-1

Electrochimica Acta 48 (2003) 1295ꢀ1300
/
www.elsevier.com/locate/electacta
Nickel electroless deposition process on chemically pretreated Si(100)
wafers in aqueous alkaline solution
Daisuke Niwa, Nao Takano, Taro Yamada, Tetsuya Osaka *
Department of Applied Chemistry, School of Science and Engineering; Kagami Memorial Laboratory for Materials Science and Technology,
Waseda University, Shinjuku-ku, Tokyo 169-8555, Japan
Abstract
Pretreating Si wafer surfaces with hydrochloric acid and hydrogen peroxide mixture (HPM) or ethanol was found to enhance the
reactivity of chemical Ni deposition on Si(100) wafers in a simple bath of NiSO4ꢀ(NH4)2SO4 at pH 9.0. This phenomenon was
/
identified as the acceleration of anodic reaction involved in chemical Ni deposition reaction on Si wafer surfaces, namely oxidation
process of surface Si. Just after immersing into the alkaline bath, a reactive surface where oxidation reaction of Si was accelerated
was formed on Si wafer surface with HPM or ethanol pretreatment. On the reactive surface, uniform and glossy Ni deposition film
was obtained. In order to clarify the effects of HPM- or ethanol-pretreatment on anodic reaction, the pretreated Si(100) wafers were
immersed into aqueous alkaline solution excluding NiSO4, resulting acceleration of Si oxidation compared to the unpretreated
Si(100) surface. The progress of surface reactivity was also clarified by open circuit potentials (OCP), XPS, and ex-situ ATR FTIR.
Moreover, by using this pretreatment, selective deposition onto nano-patterned Si substrate was performed. A two-dimensional
array of fine Ni dots (diameter ca. 80 nm) was successfully fabricated.
# 2003 Elsevier Science Ltd. All rights reserved.
Keywords: Electroless deposition; Silicon; Chemical pretreatment; Metallic nano-dot arrays
1. Introduction
We have surveyed silicon wafer-scale fabrication
process by means of electroless deposition of metals to
make nano-structures on Si wafer because of its
simplicity and ability to fill in fine patterns. We also
investigated the mechanism of electroless Ni deposition
on Si(100) wafers aiming for fabricating fine metal dot
Fabrication of nanometer-scale structures over silicon
surfaces on a mass-production basis is a prospective
process in relation to the development of nano-devices
such as single electron transistors [1ꢀ
recording media [4ꢀ6], highly integrated sensors [7,8],
/3], patterned
arrays [16ꢀ19]. For this purpose, we proposed a process
/
/
composed of two steps, that is, the treatment for
nucleation and then the treatment for particle growth,
by using two different deposition bathes. In the first
nucleation step, the bath containing Ni2ꢁ ion without
reducing agent was used and then in the second particle-
growth step, the bath with reducing agent was used. This
process was designed because the particle growth
proceeds more easily than the nucleation on Si wafer
surface when the bath contains the reducing agent [18].
Obviously, the nucleation step is the key process in
forming metal dots with satisfactory properties.
and so on, which are all based on silicon wafer surfaces.
Among modern methods of silicon manufacturing, wet
processes such as electrodeposition and electroless
deposition are attractive in mass-production of fine
structures [9ꢀ12]. Practically, the electrodeposition
/
technique is attracting a grate attention in recent Cu
interconnection technology, and various types of appli-
cation have been studied [9,13]. The electroless deposi-
tion technique is also notable in many aspects and is
extensively applied in electronic circuitry fabrication
[14,15].
Electroless deposition of Ni from a bath of simple
NiSO4ꢀ(NH4)2SO4 solution is accompanied by the
/
oxidation of the Si surface [17]. When the Si wafer was
immersed in this simple bath right after the HF
treatment, however, partial and non-uniform nickel
* Corresponding author. Fax: ꢁ
E-mail address: osakatet@waseda.jp (T. Osaka).
/
81-3-3205-2074.
0013-4686/03/$ - see front matter # 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0013-4686(02)00838-1

1296
D. Niwa et al. / Electrochimica Acta 48 (2003) 1295ꢀ1300
/
deposition occurred on the Si surface. On the other
hand, the uniformity of Ni nuclei formation was
excellently improved, and the rate of deposition reaction
was drastically accelerated, when the nucleation step
was performed after a wet chemical pretreatment of Si
wafer. [18,19] The wet pretreatment was immersion in
either HPM (a mixture of HCl and H2O2 and H2O
heated at 80 8C) or ethanol.
To investigate the behavior of anodic reaction of
electroless Ni deposition, the aqueous alkaline solution
excluding NiSO4 from Bath I was used (listed in Table 1
as solution I?).
The bath temperature of electroless deposition was
adjusted at 75 8C for the Si wafers pretreated in
ethanol, and at 80 8C in the other cases. All bathes
were adjusted at pH 9.0 with NH4OH.
In the present study, we report our investigation of
the changeover of Si surface conditions affecting on Ni
deposition behavior by monitoring the surface by X-ray
photoelectron spectroscopy (XPS), the electrochemical
open circuit potential (OCP) measurement, and ex-situ
attenuated total reflection (ATR) FTIR. In addition, to
conform the effects of chemically pretreatment, selective
deposition was performed onto nano-patterned Si sub-
strate.
For better quality of deposited Ni films, two types of
pretreatment solutions, HPM (a mixture of one part of
35% HCl, one part of 30% H2O2 and five parts of H2O
heated at 80 8C) and ethanol (99.5% ethanol), were
used. The pretreatment was performed just after the
formation of H-terminated surface. All the experiments
were performed on a class 100 clean bench.
The surface of the specimens was examined by using
an X-ray photoelectron spectroscope (XPS; JEOL Ltd.,
JPS-9000MX) furnished with monochromatized Al Ka
radiation, pass energy of 10 eV, slit width of 5 mm,
photoelectron take-off angle of 15 degree. The overall
energy resolution was 0.1 eV. To measure the OCP, a
one-compartment cell was assembled with a Si wafer as
the working electrode and a salt-bridge-connected Ag/
AgCl as the reference electrodes. The OCP measurement
was performed by a potentiostat/galvanostat (Hokuto
Denko Corporation, HABF-501) equipped with an
arbitrary function generator. Infrared absorption spec-
2. Experimental
The substrates used were n-type Si(100) wafers
(Phosphorus-doped with a resistivity of 8ꢀ12 V cm,
/
Shin-Etsu Handotai Co. Ltd.), which had naked edges.
For fabrication of fine dot arrays the n-Si(100) [resis-
tivity of 8ꢀ12 V cm] wafers obtained as covered with a
/
tra were measured using FTIR (Parkinꢀ
trum One) with liquid N2 cooled HgCdTe (MCT)
detector. NꢀSi(100) wafer (0.5ꢃ50ꢃ
20 mm3) with 45
/Elmer, Spec-
thermally SiO2 resist layer (ca.10 nm in thickness) with
arrays of bores (ca. 80 nm in diameter) was used. These
patterned substrates were prepared by conventional
electron beam lithography process in our laboratory.
The wafers were treated with SPM (a mixture of four
parts of concentrated H2SO4 and one part of 30% H2O2
heated at 80 8C, 10 min) followed by rinsing with 18
MV deionized water. The wafers were immersed in 1%-
HF aqueous solution for 30 s to prepared clean,
hydrogen terminated Si(100) surfaces (H-terminated
Si(100)).
/
/
/
bevels on each of the long side was used this measure-
ment.
3. Results and discussion
Fig. 1 shows photographs of three types of specimens
after immersion in the simple bath I. These specimens
were H-terminated, HPM-pretreated and ethanol-pre-
treated Si(100) wafers on which Ni deposited, respec-
tively. Apparently, Ni films on the HPM- and the
ethanol-pretreated Si wafers are considerably different
from the one without pretreatment. Uniform and glossy
Ni films were formed on Si wafers pretreated with HPM
or ethanol, whereas on the surface of H-terminated Si,
nickel was just partially deposited.
As was already reported, the amount of silicon oxide,
which was formed during Ni deposition reaction,
increases as the acceleration of Ni deposition reaction
[19]. These phenomena seemed to be caused by the
difference of reactivity of the surfaces. To simplify the
phenomena by separating the oxidation of Si and the
reduction of Ni2ꢁ, we first examined the specimens that
were immersed in solution I? by XPS in order to
investigate the surface conditions which affect the
For the chemical deposition of nickel, the simple bath,
consisted of only NiSO4 and (NH4)2SO4 listed as ‘bath I’
in Table 1, was employed. In the case of specimen
pretreated with ethanol for the electroless deposition,
0.01 mol cmꢂ3 of ethanol was added into the bath. The
ethanol-pretreated wafer was immersed into the bath
before the ethanol droplets dried out.
Table 1
Chemical compositions and operating conditions of baths
Chemicals (mol dmꢂ3
)
Bath I
Solution I?
NiSO4 × 6H2O
(NH4)2SO4
Ethanola
/
0.10
0.50
0.01
ꢀ
0.50
0.01
/
pH: 9.0 (adjusted with NH4OH). Bath temperature: 80 or 75 8C.
Ethanol was added only in the case of using ethanol-pretreated
substrate.
a

D. Niwa et al. / Electrochimica Acta 48 (2003) 1295ꢀ
/1300
1297
Fig. 1. Photographs of three types of specimens after immersion in bath I for 5 min. Each substrate of these specimens was: (a) H-terminated; (b)
HPM-pretreated; and (c) ethanol-pretreated Si(100) wafer.
behavior of anodic reaction involved in Ni deposition
reaction.
were shown in Fig. 3 as dotted lines. The H-terminated
Si wafer surface was scarcely oxidized for 300 s, whereas
the other two types of pretreated surfaces were readily
oxidized within 60 s. In the case of HPM-pretreated
specimen, there was a thin oxide layer (ca. 0.6 nm in
thickness relative to thermally SiO2) on the surface,
which was formed during the pretreatment [18,20]. The
variation of Si3ꢁ and Si4ꢁ peak intensity shown in Fig.
2b indicates that the thin oxide layer was once dissolved
Fig. 2 shows the XPS of Si 2p taken from three types
of the specimens. In Fig. 2, the two peaks were observed
at 99.3 eV and 103.2 eV. The main peak of 99.3 eV is
corresponding to bulk Si3/2. The chemical shifts and
FWHM of Si4ꢁ, Si3ꢁ(Hꢀ
SiO3) are 4.0 (0.80), 3.0 (0.78),
/
respectively [20]. Therefore, it seems that the other peak
indicates the existence of Si3ꢁ and Si4ꢁ. Growing up of
Si3ꢁ and Si4ꢁ henceforward means the increase of
silicon oxide layer. The oxidation behavior of each
specimen is clearly distinguishable from the others,
especially between pretreated specimens from the un-
pretreated ones. The summarized time courses of Fig. 2
in the first 10ꢀ30 s, but the oxide was not etched out
/
completely. At 10 s, the thickness of the oxide is 0.3 nm
calculated by using the value of the electron escape
depth at this setup (bulk Si; 2.7 nm, Thermally SiO2; 3.4
nm). After that, the oxidation reaction readily pro-
ceeded. The sequence implies that the active surface is
emerged after the etching of thin oxide layer formed by
HPM. On the other hand, ethanol-pretreated surfaces
were oxidized slowly in the first 10 s and then the
oxidation accelerated. This result indicates that the
active surface where oxidation of silicon surface occurs
easily is formed in the very initial stage.
In order to clarify the time variation of surface
conditions during immersion in the aqueous alkaline
Fig. 2. X-ray photoelectron spectra of Si 2p recorded on: (a) H-
terminated; (b) HPM-pretreated; and (c) ethanol-pretreated surfaces of
Si(100) wafers after immersion in solution I? for various times. The
binding energy of Si 2p 3/2 peak of H-terminated specimen is set as a
standard for each spectrum.
Fig. 3. XPS peak area intensity ratio of Si3ꢁꢁSi4ꢁ and Si0 recorded
for three types of specimens. The dotted lines represent specimens
which were immersed in solution I?, whereas the solid lines represent
ones immersed in bath I.
/

1298
D. Niwa et al. / Electrochimica Acta 48 (2003) 1295ꢀ1300
/
Fig. 4. Time variation of open circuit potential (OCP). The specimens
were immersed in solution I?. The dotted lines represent the behavior of
H-terminated specimen, and the black-solid and the gray-lines
represent that of ethanol- and HPM-pretreated ones, respectively.
Fig. 5. ATR spectra of the ethanol- and HPM-pretreated specimen
immersed in solution I?. The gray lines; initial, the solid lines; After
immersion at 20 s.
solution without Ni2ꢁ (solution I?), the OCP was
measured and shown in Fig. 4. The OCP behaviors of
the two pretreated specimens are again quite different
from that of H-terminated one. The OCP of HPM-
pretreated Si shows peak behavior of anodic side in the
initial stage, which means that there was the thin oxide
layer formed by HPM pretreatment, and the oxide layer
was etched by the alkaline solution. After the peak, the
OCP shifts to cathodic side, which implies active surface
emersion. The OCP of ethanol-pretreated Si shows a
little shift to anodic side in the very initial stage, which
implies a quite thin oxide layer formation. The existence
of this quite thin oxide layer corresponds to electro-
chemically active surface formation, because the anodic
reaction proceeded faster after formation of the quite
thin oxide layer as apparent from the results of XPS.
After the emerging of active surface, both of HPM- and
ethanol-pretreated specimens show much lower poten-
tial than H-terminated one. These results suggest that
extremely active surfaces are formed in aqueous alkaline
solution by the two types of pretreatment.
more or less one monolayer of hydrophilic oxide layer
including suboxide bonds. Besides, based on these
results, the formation process of the active surface is
discussed. In the case of HPM-pretreated surface, a very
thin layer of silicon oxide (ca. 0.6 nm) on the wafer
surface by the HPM is dissolved by the aqueous alkaline
solution, and subsequently formed a suboxide rich
surface. At ethanol-pretreated surface, it seems that
the ethoxylation proceeds at surface at the initial stages.
A hydrogen-terminated Si surface is often alcoxylated in
alcohol (RꢀOH) such as methanol, ethanol and iso-
/
propanol [21,22]. It is considered that the alcoxylated
surface easily reacts with water shown as Eq. (1).
SiꢀORꢁH2OꢄSiꢀOHꢁROH
(1)
By this reaction, the suboxide surface that is active
surface is expected to form. Therefore, we can conclude
that the same active surface is constructed at both
chemically pretreated Si surface immersed in aqueous
alkaline solution.
By reviewing the results of XPS and OCP at the same
time, it is understood that the active surface was formed
We have dealt with the silicon wafer surfaces treated
in the aqueous alkaline solution excluding metal salt
(solution I?) in above. Now, we move on to the surface
change in an aqueous alkaline solution including NiSO4,
which is the electroless deposition bath (bath I).
The results of XPS and OCP measurement for the
three types of specimens treated with bath I were shown
in Figs. 3 and 6, respectively. Similar to the results of
specimens immersed in solution I?, it is seen that the
active surface was formed in the very initial stage, and
that there was a quite thin, hydrophilic oxide layer on
the surface. The apparent difference of the behavior in
bath I from that in solution I? is that it took shorter
in the aqueous alkaline solution within the first 20ꢀ30 s
/
and that the active surface was about one monolayer of
oxidized silicon. To clarify the chemical structure of the
active surface, ex-situ ATR FTIR was measured (Fig.
5). The specimens were prepared by immersion on
pretreated surface in solution I? at the time of 20 s. It
is confirmed that both active surfaces are constituted
with Siꢀ
and SiꢀOH (3500ꢀ
CꢀH bonds. Especially, suboxide bonds are major one.
/
H (2050ꢀ
/
2150 cmꢂ1), suboxide (2250 cmꢂ1),
3750 cmꢂ1) bonds, and are not exist
/
/
/
This suboxide rich surface is very active for oxidation
reaction. It is indicated that the reactivity of silicon
wafer surfaces comes from the existence of very thin,
time, around 5ꢀ10 s, for active surface formation.
/

D. Niwa et al. / Electrochimica Acta 48 (2003) 1295ꢀ
/1300
1299
Fig. 6. Time variation of open circuit potential (OCP). The specimens
were immersed in bath I. The dotted lines represent the behavior of H-
terminated specimen, and the black-solid and the gray-lines represent
that of ethanol- and HPM-pretreated ones, respectively.
The difference between anodic reaction rates in bath I
and solution I? is apparently shown in Fig. 3. It is
reasonable that the existence of Ni2ꢁ has a great
influence on improving the reactivity of anodic reaction
of Ni deposition on Si wafers.
By comparing the results of specimens in solution I?
and bath I, we conclude that the anodic behavior in
solution I? is reproduced in the behavior in bath I.
In addition to ethanol and HPM pretreatments, we
examined several sorts of reagents for pretreatment. The
pretreatment with organic solvents such as acetone and
isopropyl alcohol (IPA), or with oxidative solutions
such as SPM and APM (mixture of NH4OH, H2O2 and
H2O) were utilized for the comparison. All these
pretreatments encouraged the formation of active sur-
face in bath I, which resulted in the enhancement of the
Ni deposition rate (Fig. 7). Moreover, a surface covered
with a thin thermal oxide layer (less than 1 nm in
thickness) was also utilized as substrate for Ni deposi-
tion in bath I. Although there was a minor difference in
deposition reaction rate, glossy Ni films were formed on
all of the specimens.
It is shown that the improvement of surface reactivity
was caused by the formation of quite thin, ca. one
monolayer of, hydrophilic suboxide surface, and that
the pretreatements encourages the formation of active
surface. The pretreatments, promoting formation of a
quite thin oxide layer, is effective to deposit the very
homogeneous and uniform Ni films from the Ni2ꢁ
solution without reducing agent. It may come from
hydrophilicity of surface produced from ca. one mono-
layer of oxide layer, when considering the hydrophobic
surface of H-terminated Si.
Fig. 7. Time variation of open circuit potential (OCP). The specimens
were immersed in bath I. (a) Mildly oxidized series; (b) alcohol dipped
series.
deposition process for fabricating metallic fine struc-
tures. Previously, we performed a selective Ni deposition
onto patterned Si substrates, and the Ni dot arrays of
270 nm in diameter were successfully formed by using
two-steps electroless process [16,17]. Now, we try to
apply to fabricate a more fine structure that is the
metallic nano-dot arrays. Firstly, the patterned substrate
was pretreated by HPM. The substrate was immersed in
Bath I for initiating nucleation. Then, the deposited
wafer was immersed in conventional NiP bath including
reducing agent; sodium hydrophosphate for promoting
grain growth over the nuclei. Fig. 8 shows SEM images
of the patterned specimen before or after this two-step
electroless deposition treatment. It is seen the bores were
filled with Ni and a two-dimensional array of dots with
a diameter of ca. 80 nm was successfully formed. From
the result, it was revealed that the wet chemically
pretreatment was much effective for fabricating fine
metal structure onto Si wafer.
Based on knowledge acquired as above, the wet
chemical pretreatment was applied to selective Ni

1300
D. Niwa et al. / Electrochimica Acta 48 (2003) 1295ꢀ1300
/
Fig. 8. SEM images of (a) nano-patterned substrate, (b) Ni dots fabricated by selective electroless deposition on a patterned silicon substrate. The
dot diameter and period are 80 nm apart and 320 nm, respectively.
[6] M. Farhoud, M. Hwang, H.I. Smith, M.L. Schattenburg, J.M.
4. Conclusion
Bae, K. Yousef-Toumi, C.A. Ross, IEEE Trans. Magn. 34 (1998)
1087.
[7] M.J. Schoning, A. Kurowski, M. Thust, P. Kordos, J.W.
A sequence of changing silicon surface condition in
¨
aqueous alkaline solution with and without Ni2ꢁ was
investigated by XPS and OCP measurement in order to
clarify the effects of the pretreatment of Si wafers. From
the results of XPS, it was found that the two pretreat-
ments, such as HPM and ethanol, accelerated the anodic
reaction of Ni deposition, i. e. oxidation reaction of
silicon. The improvement of the reactivity of silicon
wafer surfaces coincided with cathodic shift of open
circuit potential. The active surface was formed within
Schultze, H. Luth, Sensors Actuators B 64 (2000) 59.
¨
[8] M. Lehmann, W. Baumann, M. Brischwein, R. Ehret, M. Kraus,
A. Schwinde, M. Bitzenhofer, I. Freund, B. Wlof, Biosens.
Bioelectron. 15 (2000) 117.
[9] P.C. Andricacos, C. Uzoh, J.O. Dukovic, J. Horkans, H.
Deligianni, IBMJ. Res. Dev. 42 (1998) 567.
[10] M.-S. Chen, S.L. Brandow, C.S. Dulcey, W.J. Dressick, G.N.
Taylor, J.F. Bohland, J.H. Georger, Jr, E.K. Pavelchek, J.M.
Calvert, J. Electrochem. Soc. 146 (1999) 1421.
[11] V.M. Dubin, Y. Shacham-Diamand, B. Zhao, P.K. Vasudev,
C.H. Ting, J. Electrochem. Soc. 144 (1998) 898.
[12] Y. Shacham-Diamand, Y. Sverdlov, Microelectron. Eng. 50
(2000) 525.
20ꢀ30 s in the simple alkaline solution (solution I?), and
/
within 10 s in alkaline Ni2ꢁ solution (bath I). This
improvement of surface reactivity was caused by the
formation of quite thin hydrophilic suboxide rich sur-
face promoted by the foregoing chemical pretreatments.
This pretreatment was also essential process for
fabricating metallic nano-structure on Si.
[13] J.J. Kelly, C. Tian, A.C. West, J. Electrochcem. Soc. 146 (1999)
2540.
[14] E.J. O’Sullivan, A.G. Schrott, M. Paunovic, C.J. Sambucetti, J.R.
Marino, P.J. Bailey, S. Kaja, K.W. Semkow, IBMJ. Res. Devel.
42 (1998) 607.
[15] H. Honma, T. Kobayashi, J. Electrochem. Soc. 141 (1994) 730.
[16] T. Osaka, N. Takano, S. Komaba, Chem. Lett. 7 (1998) 657.
[17] N. Takano, N. Hosoda, T. Yamada, T. Osaka, J. Electrochem.
Soc. 146 (1999) 1407.
References
[18] N. Takano, N. Hosoda, T. Yamada, T. Osaka, Electrochim. Acta
44 (1999) 3743.
[1] L. Guo, E. Leobandung, L. Zhuang, S.Y. Chou, J. Vac. Sci.
Technol. B15 (1997) 2840.
[2] L. Guo, E. Leobandung, S.Y. Chou, Science 275 (1997) 649.
[19] N. Takano, D. Niwa, T. Yamada, T. Osaka, Electrochim. Acta 45
(2000) 3263.
[20] T. Hattori, K. Takase, H. Yamazaki, R. Sugino, Y. Nara, T. Ito,
Jpn. J. Appl. Phys. 28 (1989) L296.
[3] R. Groger, M. Barczewski, P. von Blanckenhangen, Surf. Sci.
¨
454ꢀ455 (2000) 76.
/
[21] J.N. Chazalviel, J. Electroanal. Chem. 233 (1987) 35.
[22] T. Bitzer, N.V. Richardson, D.J. Schiffrin, Surf. Sci. 382 (1997)
L686.
[4] S.Y. Chou, J. Magn. Soc. Jpn. 21 (1997) 1023.
[5] S.Y. Chou, P.R. Krauss, W. Zhang, L. Guo, L. Zhuang, J. Vac.
Sci. Technol. B15 (1997) 2897.