Surface termination and hydrogen bubble adhesion on Si(100) surfaces during anisotropic dissolution in aqueous KOH

Abstract

The formation and growth of hydrogen bubbles on a Si(1 0 0) surface during its anisotropic etching in aqueous KOH has been investigated. Quantitative data on bubble size, lifetime and density on the etching surface was obtained and their dependence on KOH concentration, applied potential and temperature were measured. In situ FTIR measurements demonstrated a strong dependence of bubble attachment on surface termination and hence on the hydrophilicity of the Si(1 0 0) surface during etching. The formation of surface defects and the geometry of bubble imprints have been directly characterised with scanning probe microscopy. The analysis of hillock formation and statistical considerations show that the adhesion of hydrogen bubbles during anisotropic etching of silicon is a source of surface roughness and pyramid formation.

Full text

Page 1

Surface termination and hydrogen bubble adhesion on Si(100)
surfaces during anisotropic dissolution in aqueous KOH
Wolfgang Haissa,*, Philipp Raischa, Lennart Bitscha, Richard J. Nicholsa,
Xinghua Xiab, John J. Kellyb, David J. Schiffrina
aUniversity of Liverpool, Chemistry Department, Crown Street, Liverpool L69 7ZD, UK
bDebye Institute, Utrecht University, P.O. Box 80000, 3508 TA Utrecht, Netherlands
Received 17 October 2005; received in revised form 16 May 2006; accepted 24 July 2006
Available online 18 September 2006
Abstract
The formation and growth of hydrogen bubbles on a Si(100) surface during its anisotropic etching in aqueous KOH has been inves-
tigated. Quantitative data on bubble size, lifetime and density on the etching surface was obtained and their dependence on KOH con-
centration, applied potential and temperature were measured. In situ FTIR measurements demonstrated a strong dependence of bubble
attachment on surface termination and hence on the hydrophilicity of the Si(100) surface during etching. The formation of surface
defects and the geometry of bubble imprints have been directly characterised with scanning probe microscopy. The analysis of hillock
formation and statistical considerations show that the adhesion of hydrogen bubbles during anisotropic etching of silicon is a source of
surface roughness and pyramid formation.
? 2006 Elsevier B.V. All rights reserved.
Keywords: Silicon etching; Bubble adhesion; Micromachining; Anisotropic etching; FTIR spectroscopy; Si(100); Hillock formation
1. Introduction
Etchingofsiliconinalkalinesolutionsiswidelyemployed
in micromachining [1,2]. Beams and membranes for
mechanical sensors and actuators [3,4], V-groove structures
used for the passive alignment of glass fibres in optoelec-
tronic devices and complex micro-electromechanical sys-
tems (MEMS) [5,6] can be easily fabricated using the
anisotropicdissolutionofsinglecrystalsilicon[7,8].Modern
exacting demands [9] require the production of defect-free
structures with a well-defined anisotropic ratio and high
quality surface finish with surface features of nanometer
dimension. For this reason it is important to elucidate the
mechanism of formation of micro-pyramids and of large-
scale roughness during the wet chemical etching of silicon.
Two hydrogen molecules are released for each silicon
atom dissolved [10] and Si(OH)4 is regarded as the
primary reaction product [11,12]. The overall reaction is
[13,14]
Si þ 2H2O ! 2H2þ SiðOHÞ4!
OH?
soluble silicates
ð1Þ
Etching defects and surface roughness can result if the
hydrogen bubbles produced remain long enough on the
surface to mask it from the etching solution [15–19].
The purpose of this work was to investigate the physico-
chemical aspects of bubble adhesion from measurements
of bubble size, lifetime and IR spectroscopy of Si(100) sur-
faces in contact with aqueous KOH. In addition, scanning
probe techniques have been used to elucidate the role of
bubbles in the formation of surface defects (i.e. micro-
pyramids) and/or large-scale surface roughness during
etching.
0022-0728/$ - see front matter ? 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.jelechem.2006.07.027
*Corresponding author. Tel.: +44 1517943556; fax: +44 1517943588.
E-mail address: w.h.haiss@liv.ac.uk (W. Haiss).
www.elsevier.com/locate/jelechem
Journal of Electroanalytical Chemistry 597 (2006) 1–12
Journal of
Electroanalytical
Chemistry

Page 2

2. Experimental methods
2.1. Chemicals and samples
Unless otherwise stated, samples were Czochralski
grown, phosphorous doped n-Si(100) wafers (1–20 X cm;
orientation tolerance ±2?; Compart Technology, UK) or
boron-doped p-Si(100) wafers (1–30 X cm; orientation tol-
erance ±0.5?; Okmetic, Finland). The silicon working-elec-
trode was mounted in a Kel-F holder and the exposed
surface limited with an ‘‘o’’-ring. Electrical contact was
made to the backside of the sample with Ga/In eutectic.
Milli-Q water and analytical grade KOH (Fluka puriss
p.a.) were used throughout. As described below, different
procedures were used to remove the native oxide, depend-
ing on the type of measurement performed.
2.2. Voltammetry and etch rate measurements
For these measurements, the native oxide was removed
before each measurement by dipping the wafers in 1 M
HF + 2 M NH4F (Merck, p.a.) for 1 min followed by rins-
ing with water. The electrochemical experiments were car-
ried out with a PAR 283 potentiostat (EG&G Princeton
Applied Research) controlled by a personal computer via
a LabVIEW interface (National Instruments). The cur-
rent–potential scans (voltammograms) were measured in
the dark and the potential was scanned from negative to
positive values at a scan rate of 1 mV/s. The etch rate
was determined by measuring the etched depth as a func-
tion of etching time using a surface profiler (Dektak II,
Veeco Instruments).
2.3. Video microscopy
The samples were rinsed first in ethanol and then in
water and finally blown dry in argon (Pure shield, BOC)
before mounting on the sample holder. A circular area of
0.95 cm2was exposed to the solution. The etching solutions
were not deaereated. The native oxide was stripped directly
in the etching solution at 60 ?C. This procedure avoided
excessive roughening of the surface at the beginning of
etching, which was observed for HF-dipped samples.
Etching was carried out in a 200 cm3Pyrex glass vessel
equipped with an optical flat. The sample was illuminated
from the side with a slide projector. The potential of the sil-
icon surface was controlled with a home-built potentiostat
and the temperature was kept constant with a PYE UNI-
CAM temperature controller. A television camera (MK
11 Series Tube Camera; DAGE-MTI, Inc.) in conjunction
with a Nikon objective (Micro-Nicor, 105 mm; f/4) was
used to image the surface. The images were displayed on
a television monitor with a ·15 magnification factor, stored
in a Toshiba V-858B video recorder and subsequently ana-
lysed on a PC using a commercial frame grabber (All-In-
Wonder 128 AGP 16 MB; ATI Technologies, Inc.). The
lifetime (s) of the hydrogen bubbles on the surface (i.e.
the time between initial formation and detachment) was
determined from the video images using a stopwatch. For
lifetimes shorter than 1 s the number of video frames for
which the bubble was visible were counted, resulting in this
case in a time resolution of 40 ms (±1 frame). Bubble diam-
eters were measured directly from the frozen video frame
on the computer screen before detachment of the bubbles
took place. With this simple method the bubble diameter
(d) was determined with an accuracy of ±0.02 mm. The
screen was calibrated using the sample holder as an internal
standard.
2.4. Scanning probe techniques
In contrast with the video microscopy experiments, the
native silicon oxide layer was removed prior to etching by
immersionin2 M HF (Flukapuriss p.a.) for2 minto ensure
that etching commenced immediately after immersion in the
KOH solution. After the HF dip the samples were rinsed
with water and blown dry in a stream of argon (Pure Shield,
BOC). Samples were etched whilst lying horizontally at the
bottom of a glass vessel filled with 2 M KOH. The etching
was performed in the dark at room temperature and no
sample holder was used. At the end of the etch period, the
samples were rinsed in water for 30 s and subsequently
immersed in 0.8 M sulphuric acid for 5 min to remove any
residual KOH. This was followed by another 2 min water
rinse. The samples were finally blown dry in a stream of
argon. A surface profiler (Dektak II, Veeco Instruments)
was used to characterise the long-range defects (imprints)
caused by hydrogen bubbles during the early stages of etch-
ing. The radius of the diamond tip used was 1 lm. A Molec-
ular Imaging PicoScan atomic force microscope was
employed for ex situ AFM imaging as previously described
[20]. The surface of the etched samples was examined with a
Philips XL30FEG scanning electron microscope.
2.5. FTIR spectroscopy
FTIR spectroscopy measurements were carried out in
the attenuated total reflectance (ATR) configuration [21].
Prisms were made from float zone grown Si(100) wafers
(1 kX cm, Compart Technology, UK). All FTIR measure-
ments were carried out at room temperature. Both a
Bruker IFS 66v/S and a Bio-Rad FTS-40 infrared spec-
trometer were employed. The potential was controlled with
a potentiostat (DT2101, Hi-Tek Instruments, UK). The
counter electrode was a platinum mesh and an oxidised
Ag wire was used as a pseudo reference electrode. This elec-
trode was calibrated with respect to a saturated calomel
electrode (SCE) and all potentials are referred to the SCE
scale. Infrared absorption spectra were acquired using the
SNIFTIRS (Subtractively Normalised Interfacial FTIR
Spectroscopy) technique [22]. Briefly, successive series of
interferograms were collected at the reference potential
E1and at the sample potential E2. The relative change in
transmitted IR intensity is given by
2
W. Haiss et al. / Journal of Electroanalytical Chemistry 597 (2006) 1–12

Page 3

DI
I
¼IðE1Þ ? IðE2Þ
IðE1Þ
ð2Þ
where I1 and I2 are the infrared intensities transmitted
through the silicon crystal at the potentials E1and E2. With
this convention an upward (positive) going band corre-
sponds to greater spectral absorbance of the sample at
E2. For the measurements at steady-state conditions a time
of 60 s was allowed to elapse between acquisition of spectra
at E1and E2. All spectra were corrected to remove sloping
baselines [21].
The ATR prisms were polished with diamond paste
(0.05 lm) and then rinsed with ethanol and water. For
the in situ experiments, the ATR prism was not etched with
HF to eliminate the native oxide since this treatment, fol-
lowed by contact with aqueous KOH, led to large surface
roughness and substantial loss of the IR signal. Instead,
the polished and rinsed prism was immersed directly in
the KOH solution under study and smooth surfaces were
obtained after etching approximately 1 lm of the surface
at a potential of ?1 V. This preparation method ensured
that the crystal did not roughen during the initial etching,
an important consideration for the KOH solutions of lower
concentration.
3. Results and discussion
3.1. Voltammetry and etch rate measurements
The dissolution of Si(100) in aqueous KOH is a com-
plex process in which two mechanisms are involved [23–
25]. In the chemical mechanism silicon is dissolved in a
hydroxide-catalysed, thermally activated chemical reac-
tion, in which water is the reactive species. Simultaneously
with this, an electrochemical dissolution pathway may take
place. Most of the silicon dissolves by the chemical dissolu-
tion pathway.
Fig. 1 compares voltammograms of p- and n-Si(100) in
2 M KOH at 70 ?C with the etch rates measured at different
potentials. The general features of the current–potential
curves measured at other temperatures (20–70 ?C) and
KOH concentrations (2–10 M) were similar to those shown
in Fig. 1 and to results described in the literature [12,26–
29]. At potentials positive of the open circuit potential
(OCP), a pronounced increase of current is observed that
reaches a maximum at Ep and finally drops sharply to
low values. The strong drop of the current is associated
with the formation of a surface oxide [10,14,29–31] and
consequently, the rapid drop in the etch rate to zero is
due to anodic passivation. At potentials negative of the
OCP a negative current is observed for n-Si(100) due to
the hydrogen evolution reaction proceeding under accumu-
lation conditions. No current flow is observed for p-
Si(100) in this potential range in the dark since the surface
is depleted. The very large decrease in the etch rate
observed for n-Si in the potential range from ?1.3 to
?2 V is due to a significant change in the Helmholtz poten-
tial [32]. This change does not take place for p-Si since it is
under depletion conditions in this potential range and
therefore a constant etch rate is observed for this material.
3.2. Size and lifetime of attached bubbles
Fig. 2 shows the influence of potential on bubble size
and number density for n-Si(100) etched in 2 M KOH at
71 ?C. Bubble nucleation occurs randomly over the surface
and the average size and density of attached hydrogen bub-
bles increases as the potential is changed from ?1.07 V
(Fig. 2a) to ?1.17 V (Fig. 2b). A drastic increase of bubble
size is observed at more negative potentials (?1.63 V,
Fig. 2c). At even more negative potentials (E = ?2 V,
Fig. 2d), the surface ceases to etch (see Fig. 1b) and electro-
chemical hydrogen evolution takes place at specific sites.
Fig. 3 shows images of p-Si(100) during etching in KOH
under various conditions to illustrate the influence of KOH
concentration (cKOH), temperature and potential on bubble
adhesion. At high KOH concentration and at anodic
potentials (Fig. 3a) only a few bubbles are visible on the
surface due to their rapid detachment and their small size
compared with those observed at more negative potentials
(compare Fig. 3a at ?0.90 V with Fig. 3b at ?1.27 V) indi-
cating a stronger adhesion of bubbles to the surface as the
potential is made more negative. An increase in tempera-
ture does not change significantly the average size of the
-2.0 -1.8-1.6-1.4
U / V vs SCE
-1.2-1.0 -0.8-0.6
0
20
40
60
80
0.0
0.4
0.8
1.2
1.6
current / mA
etch rate (μm/hr)
-2.2 -2.0 -1.8 -1.6 -1.4 -1.2 -1.0 -0.8 -0.6
U / V vs SCE
0
20
40
60
80
0.0
0.4
0.8
1.2
1.6
current / mA
etch rate (μm/hr)
Fig. 1. Dependence of the etch rate of p-type (a) and n-type (b) silicon on potential in 2 M KOH solution at 70 ?C (circles). The current–potential curves
for each type of silicon under the same conditions are shown for comparison (solid lines).
W. Haiss et al. / Journal of Electroanalytical Chemistry 597 (2006) 1–12
3

Page 4

bubbles (compare Fig. 3b and c), but a decrease of cKOH
from 6 to 2 M increases the average size by an order of
magnitude (compare Fig. 3c and d). Figs. 2 and 3 clearly
demonstrate that the average bubble size is strongly depen-
dant on E and cKOH. This dependence was found to be
qualitatively similar in the potential region positive to the
Fig. 2. Potential dependence of bubble formation on n-Si(100) in 2 M KOH sample holder diameter = 1.13 cm; (a) E = ?1.07 V (current maximum); only
a few small bubbles are adhering to the surface for a very short time and most of the hydrogen is nucleating at the sample holder edge. (b) E = ?1.17 V;
bubble size and density is increasing compared to (a). (c) E = ?1.63 V; the size of the bubbles increases dramatically. (d) E = ?2.0 V; the nucleation of
bubbles takes place mainly at specific surface sites. At this potential the surface is no longer etching and direct electrochemical evolution of hydrogen is
observed (compare with Fig. 1(b)). T = 71 ?C.
Fig. 3. Bubble formation on p-Si(100) during etching in KOH under various conditions. Sample holder diameter = 1.13 cm. (a) 6 M KOH; E = ?0.9 V;
T = 23 ?C; (b) 6 M KOH; E = ?1.27 V; T = 23 ?C; (c) 6 M KOH; E = ?1.27 V; T = 51 ?C. The increase of temperature decreases the average bubble size
marginally but more bubbles are formed due to an increase of etch rate (compare with (b)). (d) 2 M KOH; E = ?1.23 V; T = 51 ?C. The bubble lifetime
and the average bubble size increases dramatically at lower KOH concentrations (compare with (c)).
4
W. Haiss et al. / Journal of Electroanalytical Chemistry 597 (2006) 1–12

Page 5

anodic peak for n- and p-Si, whereas the differences men-
tioned above are observed at potentials negative of the
OCP.
The dependence of bubble diameter (d) on bubble life-
time (s) is shown in Fig. 4. The average diameter for a
sequence of 35 bubbles and their lifetime is shown as an
open circle in this figure. It was observed that the forma-
tion and detachment of bubbles on the surface is a random
process. Bubbles do not emerge at special sites but their
formation is distributed evenly over the whole area and
bubble lifetime shows a broad distribution. The experimen-
tally observed relation between bubble diameter and the
bubble lifetime in Fig. 4 can be described by
dðsÞ ¼ aðTÞs1=2
where the coefficient a depends on temperature. This repre-
sentation is consistent with a bubble growth model pro-
posed by Scriven, which relates radial mass transfer to
spherical phase growth controlled by diffusion [33]. This
model leads to [34]
ð3Þ
d ¼ 4bðktÞ1=2
where b is a dimensionless growth parameter and k is the
temperature-dependent diffusivity.
Fig. 5a shows the potential dependence of the average
lifetime of hydrogen bubbles on p-Si(100) in 2 M KOH.
Remarkably, a change of potential from ?1.2 to ?0.85 V
reduces s by two orders of magnitude. The average dia-
meter d scales approximately with s1/2, independently of
potential (Fig. 5b). Fig. 6 shows similar measurements on
n-Si(100). A potential change from ?1.63 to ?1.07 V
decreases the bubble size by approximately an order of
magnitude and the bubble lifetime by two orders of magni-
tude. The pronounced influence of KOH concentration on
bubble lifetime, measured at the OCP, is also shown in this
ð4Þ
figure; an increase in KOH concentration from 1.7 to 6 M
decreases s by nearly two orders of magnitude.
3.3. Surface morphology
Fig. 7 shows typical profile scans through the centre of
circular bubble imprints observed on n-Si(100) etched only
for a few minutes in 2 M KOH. In these measurements,
the exposure time to the solution was small to avoid inter-
ference from excessive surface roughness. The profiles
observed are strongly dependent on dc, the diameter of
the circular region where the surface is not in direct contact
with the solution. Small bubbles (dc? 20 lm) result in
Fig. 4. Dependence of the lifetime (s) of 35 different hydrogen bubbles
formed during etching of n-Si(100) in 1.7 M KOH at 24 ?C on their
diameter d (solid squares). The solid line shows a non-linear regression fit
(d = 0.15t1/2). The open circle represents the average of the data and the
error bars represent the standard deviation.
Fig. 5. (a) Potential dependence of the average lifetime (s) of hydrogen
bubbles on p-Si(100) formed during etching in 2 M KOH at 74 ?C (open
circles) and 52 ?C (filled circles). (b) Average lifetime (s) of bubbles from
(a), plotted against their average diameter (d). Error bars represent the
standard deviation.
Fig. 6. Dependence of the average lifetime (s) of hydrogen bubbles on n-
Si(100) formed during etching in 2 M KOH at 71 ?C on their average
diameter (d) for different potentials (solid circles). The room temperature
measurements show the influence of KOH concentration on bubble size
and lifetime (open squares). The number next to the symbols represents
the potential at which each data point was measured. Error bars represent
the standard deviation.
W. Haiss et al. / Journal of Electroanalytical Chemistry 597 (2006) 1–12
5