Reactive species generated during wet chemical etching of silicon in HF/HNO3 mixtures

Article abstract of DOI:10.1021/jp0608168

The role of intermediate species generated during wet chemical etching of silicon in a HF-rich HF/HNO₃ mixture was studied by spectroscopic and analytical methods at 1°C. The intermediate N₂O₃ was identified by its cobalt blue color and the characteristic features in its UV-vis and Raman spectra. Furthermore, a complex N(III) species (3NO ⁺·NO₃^-) denoted as [N₄O ₆²⁺] is observed in these solutions. The time-dependent decay of the N(III) intermediates, mainly by their oxidation at the liquid-air interface, serves as a precondition for the study of the etch rate as function of the intermediate concentration measured by Raman spectroscopy. From a linear relationship between etch rate and [N₄O₆²⁺] concentration, NO⁺ is considered to be a reactive species in the rate-limiting step. This step is attributed to the oxidation of permanent existing Si-H bonds at the silicon surface by the reactive NO⁺ species. N₂O₃ serves as a reservoir for the generation of NO⁺ leading to a complete coverage of the silicon surface with reactive species at high intermediate concentrations. As long as this condition is valid (plateau region), the etch rate is constant and yields a smooth silicon surface upon completion of the etching. If the N₂O₃ concentration is insufficient to ensure a coverage of the Si surface by NO ⁺, the etch rate decreases linearly with the N₂O ₃ concentration and results in a roughening of the etched silicon surface (slope region).

Full text of DOI:10.1021/jp0608168

J. Phys. Chem. B 2006, 110, 11377-11382
11377
Reactive Species Generated during Wet Chemical Etching of Silicon in HF/HNO3 Mixtures
†
Marco Steinert, J o1 rg Acker,* Matthias Krause, Steffen Oswald, and Klaus Wetzig
Leibniz Institute for Solid State and Material Research Dresden (IFW Dresden), P.O. Box 270116,
D-01171 Dresden, Germany
ReceiVed: February 8, 2006; In Final Form: April 12, 2006
The role of intermediate species generated during wet chemical etching of silicon in a HF-rich HF/HNO
3
mixture was studied by spectroscopic and analytical methods at 1 °C. The intermediate N
2 3
O was identified
by its cobalt blue color and the characteristic features in its UV-vis and Raman spectra. Furthermore, a
+
-
2+
complex N(III) species (3NO ‚NO
3
) denoted as [N
4
O
6
] is observed in these solutions. The time-dependent
decay of the N(III) intermediates, mainly by their oxidation at the liquid-air interface, serves as a precondition
for the study of the etch rate as function of the intermediate concentration measured by Raman spectroscopy.
From a linear relationship between etch rate and [N
2
+
+
4
O
6
] concentration, NO is considered to be a reactive
species in the rate-limiting step. This step is attributed to the oxidation of permanent existing Si-H bonds at
+
+
2 3
the silicon surface by the reactive NO species. N O serves as a reservoir for the generation of NO leading
to a complete coverage of the silicon surface with reactive species at high intermediate concentrations. As
long as this condition is valid (plateau region), the etch rate is constant and yields a smooth silicon surface
2 3
upon completion of the etching. If the N O concentration is insufficient to ensure a coverage of the Si surface
+
by NO , the etch rate decreases linearly with the N
etched silicon surface (slope region).
2 3
O concentration and results in a roughening of the
1
. Introduction
of nitrogen oxides and nitrous acid were discussed by Abel et
8
-10
al.
Robbins et al. concluded that nitrous acid is acting as an
The HF/HNO3 etching system is perhaps the mostly widely
active oxidizing agent because of the following observations.
Traces of added NaNO2 increase the etch rate due to the
1
used isotropic etchant for silicon. Applications, e.g., the removal
of work damage or roughness (after sawing of ingots) or the
texturing of surface, can be found in the semiconductor industry
4
,5
formation of nitrous acid and prevent a commonly observed
induction period. This period is characterized by an initially
lower etch rate because the oxidizing agent nitrous acid has to
2
,3
as well as in a production step for solar cell fabrication.
A systematic study of the etching mechanism was done by
Robbins and Schwartz.4-6 According to these authors, the
etching of silicon in HF/HNO3 mixtures follows a chemical
process with two basic reaction steps. In the first step, silicon
is formally oxidized by HNO3 (eq 1) followed by the dissolution
of formed SiO2 by HF (eq 2). The overall reaction is written as
eq 3.
6
be generated by the etch process itself.
However, neither an identification of the assumed intermedi-
ates nor their relevance for the kinetics of etching has been
experimentally proven so far. This paper is devoted to the
intermediates in the etching of silicon in HF-rich, concentrated
HF/HNO3 mixtures. The spectroscopic identification of the
intermediate nitrogen species, trapped when the etching is
performed at 1 °C, as well as their impact on the etch rate of
silicon is presented. Additionally, XPS measurements were
performed for characterization of the wafer surface after etching.
The presented results give new insight into the reaction
mechanism of isotropic acidic etching of silicon, particularly,
the nature of the rate-limiting step.
3Si + 4HNO f 3SiO + 4NO + 2H O
(1)
(2)
3
2
2
SiO + 6HF f H SiF + 2H O
2
2
6
2
3Si + 4HNO + 18HF f 3H SiF + 4NO + 8H O (3)
3 2 6 2
The crucial step in this reaction is the oxidation of silicon by
nitric acid. The electrochemical nature of the process was first
mentioned by Turner, who described the simultaneous dissolu-
2
. Experimental Section
7
Analytical-grade nitric acid (65 wt %, 14.45 M) and hydro-
tion of silicon at local anodic sites and reduction of the oxidizing
fluoric acid (40 wt %, 22.77 M) were used for all etch mixtures
reported herein as volume percentage [% (v/v)]. An overall
1
agent at local cathodic sites. Kooij et al. also assumed the
electrochemical behavior of silicon etching because of the
creation of holes which are injected into the valence band due
to the reduction of HNO3.
Yet unresolved is the question about the nature and the role
of intermediates formed by the reduction of HNO3. Equilibria
volume of 50 mL of etch solution after HF and HNO were
mixed was filled in wide-mouthed bottles of HDPE (high-
density polyethylene) and thermostated during the series of
experiments to the reaction temperature by using a cryostat
(Polystat K12-2, Huber K a¨ ltemaschinen GmbH).
3
Each etch solution was then pre-aged by dissolving a defined
amount of silicon (e.g., 1.4 g of Si). To prevent the loss of SiF4
and an uncontrollable warming of the solution, silicon slices of
approximately 70 mg (boron-doped, thickness of 675 µm,
*
To whom correspondence should be addressed. Phone: +49 351 4659-
6
94. Fax: +49 351 4659-452. E-mail: j.acker@ifw-dresden.de.
†
Present address: Institute for Analytical Chemistry, Dresden University
of Technology, D-01062 Dresden, Germany.
1
0.1021/jp0608168 CCC: $33.50 © 2006 American Chemical Society
Published on Web 05/20/2006

11378 J. Phys. Chem. B, Vol. 110, No. 23, 2006
Steinert et al.
resistivity of 24-36 W cm) were dissolved successively. Not
until one piece of Si had been dissolved completely was the
next one added to the etch solution. This pre-aging process
dissolved the entire amount of silicon in ∼2 h.
For the etch rate determination, a 10 mm × 10 mm silicon
specimen [(111) orientation, boron-doped, thickness of 325 µm,
resistivity of 10 W cm, arithmetic average roughness of 1.1 nm]
was held between the ends of a tweezers and immersed in the
etch solution for 5-180 s. The reaction was quenched by
immersing the sample in a beaker filled with a large volume of
deionized water and rinsing thereafter. The etch rate (r) was
obtained by differential weighing.
To perform a series of experiments at a quasi-constant silicon
content, the time of etching had to be fitted to the etch rate in
such a way that for any etch rate measurement only 1-7 mg of
silicon was dissolved additionally. For the interpretation of
surface morphology photographs, it is important to notice that
with a decreasing etch rate a smaller silicon content was
dissolved that can be attributed to the constant etch time during
a series of experiments.
Figure 1. (a) Correlation among etch rate r, nitrite concentration, and
arithmetic average roughness (R
°C, m(Si)diss ) 1.40 g, 50 mL of 70% (v/v) HF and 30% (v/v) HNO
b) XPS profiles of the Si2p region of the silicon surface after etching.
a
) of the etched silicon surfaces [ϑ )
1
3
].
(
The use of a monochromator results in Si2p3/2 and Si2p1/2 splitting.
Selected samples are numerated; the numeration is valid throughout
this paper.
The concentrations of nitrite, nitrate, and fluoride ions in the
etch solution were determined by ion chromatography (Deutsche
METROHM GmbH & Co. KG) by dilution of an aliquot of
the kinetic measurements is the nitrite concentration measured
immediately after the last piece of silicon had been dissolved.
If one begins from that maximum value within a given
experimental series, the nitrite concentration decreases with time
0
1
.5 mL of the etch mixture with deionized water to a factor of
:5000. The UV-vis measurements were carried out with a
1
1
Specord 210 spectrometer (Analytik Jena AG) using UV
following first-order kinetics.
semimicro cuvettes (d ) 10 mm, 220-900 nm; Brand GmbH
Figure 1 shows the etch rate as a function of the nitrite
concentration. To follow the series of experiments on a time
scale, the x-axis of Figure 1 has to be read from right (high
nitrite concentration) to left (low nitrite concentration). The
region of high nitrite concentrations, where the etch rate is
independent of the nitrite content, is denoted as a plateau region.
Below a certain value, a linear relationship to the etch rate was
observed and denoted as the slope region. These two remarkable
dependencies between etch rate and nitrite concentration were
reproducibly found for (i) different temperatures (T e 15 °C),
&
Co. KG). The same cuvettes were used for Raman measure-
ments.
Time-dependent Raman scattering of etch solutions was
+
excited by 514.5 nm laser radiation of an Innova 305 Ar ion
gas laser (Coherent). The laser power was 25 mW. The scattered
light was collected in a 180° backscattering geometry, analyzed
by a T 64000 triple monochromator spectrometer (Jobin Yvon)
-1
set to a spectral band-pass of 2 cm , and detected using a liquid
nitrogen-cooled CCD camera.
(ii) various amounts of predissolved silicon, and (iii) different
XPS measurements were carried out at a PHI 5600 CI
etch bath compositions with a hydrofluoric acid content of
(Physical Electronics) system using a hemispheric energy
>
70% (v/v).
analyzer. Typical measurement parameters were as follows:
excitation with monochromatic Al KR X-rays at 350 W,
measuring area ∼800 µm in diameter, pass energy of 12 eV,
Activation energies of 41 kJ/mol for the plateau region and
4
4 kJ/mol for the slope region, each at a constant nitrite
-
10
concentration for different temperatures, are an indication that
the reaction mechanism is the same in both regions and
and base pressure of 2 × 10
mbar. The samples were
measured immediately after drying, however, after a short
1
1
independent of nitrite concentration.
(minutes) transport on air.
XPS studies of the etched silicon surfaces show a complete
hydrogen termination also, which is also independent of the
nitrite concentration of the etch mixture (Figure 1). Neither
Si-O bonds nor Si-F or Si-O-F bonds (103.3-105 eV) were
detected. If Si-F bonds were formed during etching of silicon,
a replacement with Si-OH bonds would occur during water
rinsing and a change in the Si2p peak should occur.1 Only a
complete hydrogen termination of the silicon surface and its
resistivity against oxidation explains that no native oxide was
observed in the Si2p spectra. This shows that any changes in
the etch rate must originate by species in the etch solution.
Surface morphology studies of etched silicon wafers were
carried out with an optical depth profilometer [MicroProf, Fries
Research & Technology GmbH (FRT)] receiving roughness
parameters, photographs, three-dimensional pictures, and surface
profiles. The sample area of 3 mm × 3 mm was profiled with
2
00 × 200 points. The roughness of the polished wafers was
2,13
measured using a Dektak 3030 surface profilometer (Veeco
Instruments Inc.).
All species that appear in brackets in the text or figures denote
concentrations in units of grams per liter.
1
1
We supposed earlier that the nitrite ion is not the reactive
species in the etching process but rather a sum parameter for
all intermediate nitrogen species in the +3 oxidation state such
3
. Results and Discussion
3.1. Etch Rate as a Function of the Nitrite Concentration.
+
2+
A high concentration of reactive intermediates relieves their
detection and enables the investigation of their influence on the
reaction mechanism of silicon dissolution over a long period.
This condition is achieved by predissolving a certain amount
of silicon in a freshly prepared mixture of concentrated HF and
HNO3 at 1 °C to stabilize the intermediate nitrogen species
as NO , N2O3, and [N4O6 ] [in the following denoted as N(III)
species] in the concentrated etch solution. Only the dilution of
the concentrated etch solutions by a factor of 1:5000 for the
ion chromatographic analysis causes the full conversion of all
N(III) species to nitrite ions.
3.2. UV-Vis and Raman Spectroscopy of Intermediates
in Colored Etch Solutions. UV-vis spectra of blue or green-
blue etch solutions exhibit, independent of the mixing ratio of
(
visible as blue- or green-colored etch solutions) and to slow
their decomposition into NO and/or NO2. The starting point for

Wet Chemical Etching of Silicon in HF/HNO3 Mixtures
J. Phys. Chem. B, Vol. 110, No. 23, 2006 11379
Figure 2. UV-vis spectra of blue-colored etch solutions that differ
in composition.
HF and HNO3, a broad peak at 638 nm (Figure 2). With an
increasing nitric acid content, the peak maximum at 638 nm
increases due to an enhancement (darkening) of the blue color.
Simultaneously, the nitrite concentration increases from 21.5
to 28.7 g/L and then to 32.0 g/L for 50% (v/v) HNO3.
The blue and green colors are originated by N2O3, according
to a UV-vis study of N2O3 in aqueous solutions by Doherty et
1
4
al. The formation of N2O3 is confirmed by Raman spectra of
various colored etch solutions showing a broad peak at 258
-
1
cm .
According to a Raman study on different isomers of N2O3 in
NO/O2 matrices by Nour et al., this line is assigned to the ν6
N-N stretch mode of the asymmetric nitrogen(III) oxide (as-
N2O3).15 This line is the strongest Raman line of as-N2O3 at
all. Further characteristic N2O3 vibrations were detected at 195,
Figure 3. (a) Raman spectrum of a blue-colored etch solution after
silicon dissolution [ϑ ) 1 °C, m(Si) ) 1.4 g, 50 mL of 70% (v/v)
diss
3
HF and 30% (v/v) HNO ]. (b) Raman spectrum a after exposure of the
-
1
sample to air for several hours. (c) Difference Raman spectrum (a -
b). (d) Raman band profile analysis of spectrum c for the low-energy
wavenumber range. (e) Raman band profile analysis of spectrum c in
the NO stretching range; the intensity axis was extended by a factor of
6
23, 784, and 1288 cm . The generation of green and blue-
green etch solutions might be ascribed to a superposing of blue
(N2O3) and yellow brown (nitrous gases) (Figure 3).
A second N(III) species was identified by Raman spectros-
2
0 compared to that of panel d.
-
1
copy by a peak at 2240 cm . According to the work by Harrar
1
6
+
-
et al., this peak can be associated with the species (3NO ‚NO3 )
The correlation of Raman peak area for the second N(III)
-
1
2+
(
2246 cm ), further denoted as [N4O6 ]. Moreover, a very
weak band corresponding to the NO ion at 2382 cm was
detected. These Raman results confirm the work of Kelly et
2+
species, [N4O6 ], versus nitrite concentration shows a com-
pletely distinct behavior as seen in Figure 5. At high nitrite
+
-1
2+
concentrations, the [N4O6 ] Raman peak area is more or less
+
al., who suggested the formation of NO in a HF/HNO3 etch
constant; however, with a certain nitrite concentration, a linear
1
7
mixture. Further relevant Raman bands of colored etch
solutions are attributed to nitric acid, nitrate, dissolved N2O4,
and the hexafluorosilicic anion (Figure 3), as listed in Table 1.
Other nitrogen species in the +5 oxidation state, for instance,
2+
dependence between the amount of [N4O6 ] in concentrated
etch solution and the nitrite concentration measured in a dilute
solution exists. The right-hand side of Figure 5 shows a linear
dependence between the kinetic parameter etch rate and the
+
2+
NO2 , whose strongest Raman line would have been observed
[N O ] Raman peak area and identifies the reactive species
4
6
-
1
16
+
at 1400 cm were not found. Moreover, the Raman spectra
did not reveal any hint on the presence of nitrite ions, since the
strongest Raman line (ν1) at 1329 cm in aqueous solution
was not observed.
in the rate-limiting step. It is supposed that the NO ion is the
essential constituent in the [N O ] complex ion rather than
2
+
4
6
-
1
the nitrate ions, which are present in a manifold excess and
considered to stabilize the nitrosonium cation in form of the
2
7
2
+
3
.3. Etch Rate versus N(III) Intermediates. The etch rate
4 6
[N O ] complex species.
measurements using the concentrated, blue-colored etch solution
from Figure 1 were followed by Raman spectroscopy. The inset
On the basis of the results that were obtained, the following
interpretation can be given.
-
1
in Figure 4 shows the evolution of the 258 cm Raman band
attributed to N2O3 as function of time, i.e., the nitrite ion
concentration. If the etch solution is exposed to air, the N(III)
intermediates undergo an oxidative decay.11 The signal intensity
decreases as the initially dark blue solution fades out and
vanishes finally if the solution becomes transparent. The plot
of the N2O3 Raman peak area versus the nitrite concentration
(i) By the dilution of concentrated etch solutions, the
2
+
intermediate N(III) species N O and [N O ] are fully
2
3
4
6
converted into nitrite ions. A quantification is given by the
14
relationship in eq 4, introduced by Doherty et al., and expanded
+
by the species nitrosonium ion, NO . Even if the other species
mentioned in eq 4 were not spectroscopically identified, their
presence at low concentrations should not be ruled out.
(Figure 4) shows a linear relationship between the amount of
-
+
+
N2O3 dissolved in the concentrated etch solution and the nitrite
concentration measured in a 1:5000 dilution by ion chroma-
tography.
(
[NO ])
) ([N O ] + [NO ] + [HN O ] +
diluted solution 2 3 2 3
2
+
[
H NO ] + [HNO ])
(4)
2
2
2
concentrated solution

11380 J. Phys. Chem. B, Vol. 110, No. 23, 2006
Steinert et al.
TABLE 1: Wavenumbers of the Raman Lines Which Are
Characteristic for the Species Found in HF/HNO
3
Etch
Solutions
observed
literature
wavenumbers
wavenumbers
-
1
-1
species
HNO
(cm
)
(cm )
929a
928,16 925,18,19 92620
3
1
6
18,19
20
1
6
305
90
1294, 1300,
1303
b
19
680
720
-
18
NO
3
722
038, 1047
812
1
6,18
1
1050
810, 80921
1
6
N O
2 4
1
6,21
1
384
1380
2246
2387
2+]
2240
2382
195
16
22
[
N
4
O
6
+
NO
1
5
Figure 5. Correlation between the Raman peak area of [N O62+] at
N O
2 3
205
4
258, 260,23 253,21 26615
1
6
-1
2
6
7
1
57
23
84
288
2246 cm and the nitrite concentration obtained by ion chromatography
1
1
5
5
2+
627
784
1288
1866
(a) and etch rate dependence on [N
O
4 6
] concentration (b).
1
2
5
4
be given by the following series of equilibria between the
different N(III) species in concentrated etch solutions (eq 5).
cis-N
2
-
O
2
1869
404
2
25
SiF
6
395-400
25,26
6
54
656
+
+H O
2
+2H+
+2H+
+
-
H
+
2
H O
1640
HN O
y
z N O y z 2HNO y z 2H NO2
y
z
2
3
+
2
3
2
-2H+
2
-2H+
+
H
-H2O
9
1
56
447
cuvette
cuvette
+
+
2H O + 2NO (5)
3
a
This Raman band results from the symmetric stretching vibration
of HNO and coincides with a signal from the employed plastic cuvette.
The Raman band at 690 cm-1 is also observed for HF/HNO
mixtures
without dissolved silicon and is attributed to the NO bending vibration
(680 cm ). The shift is considered to be a consequence of
In the plateau region, the concentration of N2O3 is sufficiently
3
b
+
high to establish a high concentration of NO (the concentration
3
+
2
of N2O3 is found to be proportional to the square of the NO
-
1
of HNO
3
concentration, in agreement with eq 5) that completely covers
interaction between both acids, e.g., hydrogen bonds.
+
the silicon surface, yielding a region with a constant NO
concentration and a uniform etch rate. From a certain concentra-
+
tion of N2O3, the generated NO concentration is insufficient
for a complete surface coverage and the etch rate decreases
+
linearly to the surface concentration of NO . The right plot of
Figure 5 showing a linear correlation between etch rate and
2
+
[
N4O6 ] Raman peak area is considered as evidence for the
+
assumed coverage phenomena and shows that NO is involved
as the reactive species in the rate-determining step of the
dissolution of silicon. The similar activation energies for the
plateau and slope region, the always present hydrogen termina-
tion of the silicon surface, and the linear relationship between
the [N4O6 ] concentration and the etch rate are seen as evidence
of a uniform reaction mechanism that is independent of the
concentration of the N(III) intermediates.
.4. Morphology of Etched Silicon Surfaces as a Function
2+
1
1
3
of Nitrite Concentration. A characteristic feature of this etching
system is the evolution of the surface morphology and roughness
parameters after etching of initially polished silicon samples
(with an initial arithmetic average roughness of 0.7 nm) in the
dependence on the N(III) concentration. The arithmetic average
roughness (Ra) values for a 3 mm × 3 mm area are shown as
function of the nitrite concentration in the left plot of Figure 1.
The corresponding micrographs of the surface as well as the
three-dimensional views of samples 1, 3, and 5 are depicted in
Figure 6. The plot of the Ra values versus the nitrite concentra-
tion gives two different regions in analogy to the already found
etch rate dependence as a function of the nitrite concentration.
This finding points to the previously proposed coverage
phenomena as an explanation of the etch rate and surface
morphology evolution shown in Figure 1. The reactive N(III)
species are generated at the silicon surface due to the reduction
of nitric acid. This can be visibly observed in an unstirred
solution having a high content of already dissolved silicon by
the evolution of a blue-green-colored cloud covering a piece of
silicon after its immersion in the etch mixture. This leads to a
high and localized concentration of reactive N(III) species near
2 3
Figure 4. Relationship between the Raman peak area of N O at 258
-
1
cm and the nitrite concentration obtained by ion chromatography.
The inset shows the time-dependent N Raman signal (top curve,
maximum nitrite concentration; bottom curve, nitrite concentration near
2
O
3
0
g/L).
The complete conversion of the N(III) species, mainly N2O3
and [N4O62 ], to nitrite ions by dilution of concentrated etch
solutions underlines the importance of the nitrite concentration
as a useful sum parameter (eq 4) in characterizing the reactivity
of etch mixtures in industrial applications.
+
(
ii) The correlation between the peak area of [N4O62+] and
the nitrite ion concentration in Figure 5 does not yield a linear
dependence like that for N2O3 in Figure 4. The region with a
+
constant NO concentration with a large amount of nitrite, i.e.,
high N2O3 concentration (plotted as a function of nitrite
concentration and time in Figures 5 and 7, respectively), leads
to the assumption that N2O3 serves as a reservoir for the
+
+
generation of NO . A plausible method of NO formation can

Wet Chemical Etching of Silicon in HF/HNO3 Mixtures
J. Phys. Chem. B, Vol. 110, No. 23, 2006 11381
Figure 6. Micrographs of etched silicon pieces (top row) and corresponding three-dimensional views (bottom row).
to decide whether the observed pitting is caused mainly by a
simple absence of N(III) species or by a facilitated attack of
nitric or hydrofluoric acid.
3
.5. Fate of the Intermediates. Exposing the etch solution
to air causes a time-dependent decay of all N(III) species as
illustrated in Figure 7. The decay of the nitrite concentration
was shown to follow first-order kinetics, whereas the nitrate
concentration increases identically with the same reaction order
1
1
and rate. Furthermore, evidence for a direct oxidation of the
N(III) species by oxygen from the air was found, since the molar
ratios between the amounts of decomposed nitrite and of formed
nitrate for given time intervals amount to values ranging from
0.9 to close to unity. The deviation from unity can be explained
by outgassing caused by the disproportionation of N2O3 into
NO and NO2. The oxidation of the N(III) intermediates on air
is easily underlined by the following observation. (i) A
concentrated colored etch solution stored without oxygen, e.g.,
in a tightly sealed reaction vessel, can hold its color at 1 °C for
several days. (ii) If the bottle is opened, a decoloration of the
solution starts from the liquid-gas interface and proceeds slowly
into the bulk solution. (iii) Stirring accelerates the nitrite decay
via an increase in the area of the liquid-gas interface and faster
oxygen uptake. The rate constants of nitrite decay and nitrate
increase were found to be almost identical with different stirring
Figure 7. Time-dependent decomposition of N(III) species, nitrite
2+
4 6 2 3
concentration obtained by ion chromatography, and [N O ] and N O
concentration obtained by Raman spectroscopy.
the silicon surface and therefore to a full coverage of all possible
surface reaction sites. This situation remains unchanged as long
as the local N(III) species concentration is sufficiently high and
results in the observed plateau region in Figure 1. The
concentration of the reactive species decreases with time mainly
via an oxidative degradation (see the next section). At a certain
concentration, indicated by the transition from the plateau to
slope region, the full coverage is no longer given, leading to a
linear relationship between the N(III) concentration and Ra value.
Stirring experiments underline this hypothesis.11 It was found
that the etch rates in both the slope and plateau regions are
dramatically lowered by stirring. The linear relationship between
the etch rate and nitrite concentration in the former slope region
1
1
rates.
In light of the Raman studies, these findings could be
explained by a direct oxidation of N2O3 (as the major contribu-
-
tion to the nitrite concentration) to NO3 by air oxygen formally
described by eq 6.
N O + O + H O f 2HNO
3
(6)
2
3
2
2
-
remains; however, the slope ∆r/∆[NO2 ] decreases with an
Furthermore, the absence of nitrous acid and nitrite ions in
the Raman spectra is seen as an indication that the commonly
assumed disproportionation reaction of nitrous acid into nitric
acid and NO in eq 7 does not occur or occurs to an only
negligible extent.
increase in the stirring speed. The former plateau region is turned
into a slope region with a linear relationship between the etch
rate and nitrite concentration that decreases with an increase in
the stirring rate.
Such unusual behavior can be explained only by stirring the
species involved in the rate-limiting step away from the surface.
As a consequence, species from the bulk solution, such as nitric
or hydrofluoric acid, are transported to the surface. Then, the
surface reactive sites are faced with different solution species
with distinct reactivity. However, at present, it is not possible
3
HNO f HNO + 2NO + H O
(7)
2
3
2
4
. Conclusions
Performing the etch experiments at 1 °C instead of at room
temperature leads to a stabilization of nitrogen intermediates

11382 J. Phys. Chem. B, Vol. 110, No. 23, 2006
Steinert et al.
because of reduction of nitric acid. Another important advantage
for kinetic investigations can be seen in the decrease of the etch
rate. As the etch rate is lowered, the reaction heat can be
removed from the silicon surface more efficiently, leading to a
better controllability of etching.
the etching reaction because the mentioned observations make
clear the fact that the reduction of nitric acid does not end with
NO as the final product.
To clarify the mechanism of wet chemical etching of silicon
in HF/HNO3 mixtures, further detailed investigations of the
liquid phase and the gas phase are needed to characterize the
reduction of nitric acid to the final reduction products.
On the basis of the spectroscopic identification of N2O3 and
+
-
2+
(
3NO ‚NO3 ) (denoted as [N4O6 ]) as intermediary species
2+
and the linearity between the [N4O6 ] concentration and the
etch rate, the following mechanistic model is established.
Assuming the simplified formalism of eqs 1 and 2 as a valid
description of the isotropic etching of silicon in HF-rich
solutions, the oxidation of the silicon surface by nitric acid (eq
Acknowledgment. We gratefully acknowledge the European
Regional Development Fund 2000-2006 and the Free State of
Saxony for funding within the project “SILCYCLE” under
Contract 8323/1293 at the S a¨ chsische Aufbaubank (SAB).
1
) is the rate-limiting step while the subsequent dissolution of
References and Notes
28-31
SiO2 (eq 2) proceeds much faster.
According to Robbins
et al.,4 the diffusion of the minority component (HF or HNO3)
,5
(1) Kooij, E. S.; Butter, K.; Kelly, J. J. Electrochem. Solid State 1999,
, 178.
2
to the silicon surface limits the overall reaction.
(2) Tool, C. J. J.; Coletti, G.; Granek, F. J.; Hoornstra, J.; Koppes, M.;
The reduction of nitric acid as an oxidant from the +5 to +2
oxidation state with NO as the final gaseous product proceeds
in two consecutive steps: (i) the fast reduction of N(V) to N(III)
and (ii) the rate-limiting step of the reduction of N(III) to N(II).
Both partial reactions are illustrated by the very simplified
reaction scheme (eqs 8 and 9):
Kossen, E. J.; Riffe, H. C.; Romijn, I. G.; Weeber, A. W. Proceedings of
the. 20 EPVSEC, Barcelona, 2005, p 578.
th
(
3) Einhaus, R.; Vazsonyi, E.; Duerinckx, F.; Horzel, J.; Van Ker-
th
schaver, E.; Szlufcik, J.; Nijs, J.; Mertens, R. Proceedings of the 26 IEEE
PVSPC, Anaheim, 1997, p 167.
(4) Robbins, H.; Schwartz, B. J. Electrochem. Soc. 1959, 106, 505.
(5) Robbins, H.; Schwartz, B. J. Electrochem. Soc. 1960, 107, 108.
(6) Schwartz, B.; Robbins, H. J. Electrochem. Soc. 1961, 108, 365.
(7) Turner, D. R. J. Electrochem. Soc. 1960, 107, 810.
+
Si
+
(8) Abel, E.; Schmid, H. Z. Phys. Chem., Stoechiom. Verwandtschaftsl.
1928, 132, 55.
9) Abel, E.; Schmid, H. Z. Phys. Chem., Stoechiom. Verwandtschaftsl.
initiation:
HNO3 -
8 N O + NO
(8)
2
3
Si(IV)
(
+
NO+
+4F-
+2HF
1928, 134, 279.
propagation: Si(0)
8 Si(IV)
8 SiF4
8 H SiF
(10) Abel, E.; Schmid, H.; Badad, S. Z. Phys. Chem., Stoechiom.
Verwandtschaftsl. 1928, 136, 135.
-
2
6
-4e
(9)
(
11) Steinert, M.; Acker, J.; Henssge, A.; Wetzig, K. J. Electrochem.
Soc. 2005, 152, C843.
12) Gr a¨ f, D.; Grundner, M.; Schulz, R. J. Vac. Sci. Technol., A 1989,
, 808.
13) Takahagi, T.; Ishitani, A.; Kuroda, H.; Nagasawa, Y. J. Appl. Phys.
(
XPS measurements of the silicon surface after etching reveal
a hydrogen-terminated silicon surface as result of a preceding
7
(
3
2-34
dissolution of oxidic silicon surface species by HF.
The
1991, 69, 803.
hydrogen-terminated silicon surface possesses a hydridic char-
acter and is therefore stable in acidic solution in chemical
analogy to SiH4. The rate-limiting step is now assumed as the
(14) Doherty, A. M. M.; Haine, N.; Jones, E.; Stedman, G. J. Chem.
Soc., Perkin Trans. 2 1996, 2055.
(
(
15) Nour, E. M.; Chen, L. H.; Laane, J. J. Phys. Chem. 1983, 87, 1113.
16) Harrar, J. E.; Rigdon, L. P.; Rice, S. F. J. Raman Spectrosc. 1997,
+
attack of NO , a small and positively charged ion capable of
28, 891.
(
17) Kelly, M. T.; Chun, J. K. M.; Bocarsly, A. B. Appl. Phys. Lett.
approaching the hydridic surface, either under formation of
intermediate Si-OH surface moieties or by an oxidative
cleavage of the Si-Si back-bonds of the Si-H surface species.
This gives rise to two main questions. (i) How does the fast
reduction of HNO3 to the N(III) intermediates proceed, and with
which steps in the oxidation of silicon can be associated? A
possible answer can be given with respect to the model by
1
994, 64, 1693.
(
18) Ingold, C. K.; Millen, D. J. J. Chem. Soc. 1950, 2612.
(19) Cohn, H.; Ingold, C. K.; Poole, H. G. J. Chem. Soc. 1952, 4272.
(20) McGraw, G. E.; Bernitt, D. L.; Hisatsune, I. C. J. Chem. Phys.
965, 42, 237.
1
(
21) Hisatsune, I. C.; Devlin, J. P.; Wada, Y. J. Chem. Phys. 1960, 33,
14.
(22) Sharp, D. W. A.; Thorley, T. J. Chem. Soc. 1963, 3557.
(23) Bradley, G. M.; Siddall, W.; Strauss, H. L.; Varetti, E. L. J. Phys.
Chem. 1975, 79, 1949.
24) Nour, E. M.; Chen, L. H.; Strube, M. M.; Laane, J. J. Phys. Chem.
7
7
Turner, where the reduction of nitric acid is an electrochemical
process. According to Lehmann,35 the etching of silicon using
concentrated HF/HNO3 mixtures is analogous to the anodic
oxidation of silicon in HF solutions at high current densities.
The injection of holes into the valence band of silicon can cause
a cleavage of Si-Si bonds which may further react with water
molecules, yielding a Si-O-Si bond. Such a quite unspecific
reaction step might account for the isotropic character of the
(
1
984, 88, 756.
(25) Badachhape, R. B.; Hunter, G.; McCory, L. D.; Margrave, J. L.
Inorg. Chem. 1966, 5, 926.
26) Woodward, L. A.; Anderson, L. E. J. Inorg. Nucl. Chem. 1956, 3,
62.
(
3
(27) Laane, J.; Ohlsen, J. R. Prog. Inorg. Chem. 1980, 27, 465.
(28) Judge, J. S. J. Electrochem. Soc. 1971, 118, 1772.
(29) Verhaverbeke, S.; Teerlinck, I.; Vinckier, C.; Stevens, G.; Car-
36
acidic etching.
tuyvels, R.; Heyns, M. M. J. Electrochem. Soc. 1994, 141, 2852.
30) Osseo-Asare, K. J. Electrochem. Soc. 1996, 143, 1339.
(
ii) Yet unresolved is the question of the final state of NO+
(
after its reaction with silicon. The evolution of brown gas as a
consequence of the oxidation of NO by air is broadly attenuated
at 1 °C compared with that at g25 °C. A quantitative analysis
of the gas atmosphere during the acidic etching of a multi-
crystalline silicon with a HF/HNO3/H2O mixture gave a
(31) Knotter, M. D. J. Am. Chem. Soc. 2000, 122, 4345.
(32) Jakob, P.; Chabal, Y. J. J. Chem. Phys. 1991, 95, 2897.
(
33) Ubara, H.; Imura, T.; Hiraki, A. Solid State Commun. 1984, 50,
73.
34) Yablonovitch, E.; Allara, D. L.; Chang, C. C.; Gmitter, T.; Bright,
6
(
T. B. Phys. ReV. Lett. 1986, 57, 249.
(35) Lehmann, V. Electrochemistry of Silicon; Wiley-VCH Verlag
GmbH, Weinheim, Germany, 2002.
considerable amount of N2O as the reaction product.37 Further-
more, N2O was identified as the major reaction product in the
(
36) Hines, M. A. Int. ReV. Phys. Chem. 2001, 20, 645.
1
electrochemical oxidation of silicon in HF/HNO3 mixtures. It
(37) Weinreich, W. Diploma Thesis, Freiberg University of Mining and
has to be concluded that eq 3 is an insufficient description of
Technology, Freiberg, Germany, 2005.