Influence of the electrolyte viscosity on the structural features of porous silicon

Article abstract of DOI:10.1016/S0038-1098(01)00036-9

The influence of an increase in the electrolyte viscosity, obtained by reducing the chemical bath temperature, replacing a volume fraction of ethanol with the heavier glycerol and increasing the anodization current density, was studied in porous silicon layers formed on p-type silicon. Synchrotron X-ray reflectivity and cross-sectional transmission electron microscopy techniques were used for 300-nm-thick porous silicon layers. For greater thicknesses (10 μm), the temperature effect was investigated by multicrystal X-ray diffraction with a laboratory source. It was found that higher average porosities and lower roughnesses at the interface between porous silicon and substrate are obtained when the electrolyte viscosity increases. Moreover, the formation of flatter porosity depth-profiles is favored by lowering the electrolyte temperature. The importance of these effects on the performances of optical devices based on porous silicon is underlined.

Full text of DOI:10.1016/S0038-1098(01)00036-9

PERGAMON
Solid State Communications 118 (2001) 85±90
www.elsevier.com/locate/ssc
In¯uence of the electrolyte viscosity on the structural features
of porous silicon
a,
M. Servidori , C. Ferrero , S. Lequien , S. Milita , A. Parisini , R. Romestain ,
b
c
a
a
d
*
a d,e
S. Sama , S. Setzu , D. Thiaudi eÁ re
f,g
a
CNR-Istituto LAMEL, Via P. Gobetti 101, 40129 Bologna, Italy
b
European Synchrotron Radiation Facility, BP 220, 38043 Grenoble Cedex, France
c
Laboratoire Pierre Sue, CEA-CNRS, CEN Saclay, 91191 Gif sur Yvette, France
Laboratoire de Spectrom e trie Physique, BP 87, 38402 St. Martin d'H eÁ res Cedex, France
d
e
Physics Department, Cagliari University, 09042 Monserrato, Italy
Centre de Recherche sur les Mat e riaux aÁ Haute Temp e rature, 1d Av. de la Recherche Scienti®que, 45071 Orl e ans Cedex, France
f
g
Laboratoire pour l'Utilisation du Rayonnement Electromagn e tique (LURE), 91898 Orsay Cedex, France
Received 28 September 2000; received in revised form 9 January 2001; accepted 24 January 2001 by J.F. Sadoc
Abstract
The in¯uence of an increase in the electrolyte viscosity, obtained by reducing the chemical bath temperature, replacing a
volume fraction of ethanol with the heavier glycerol and increasing the anodization current density, was studied in porous
silicon layers formed on p-type silicon. Synchrotron X-ray re¯ectivity and cross-sectional transmission electron microscopy
techniques were used for 300-nm-thick porous silicon layers. For greater thicknesses (10 mm), the temperature effect was
investigated by multicrystal X-ray diffraction with a laboratory source. It was found that higher average porosities and lower
roughnesses at the interface between porous silicon and substrate are obtained when the electrolyte viscosity increases.
Moreover, the formation of ¯atter porosity depth-pro®les is favored by lowering the electrolyte temperature. The importance
of these effects on the performances of optical devices based on porous silicon is underlined. q 2001 Elsevier Science Ltd. All
rights reserved.
Keywords: A. Porous materials; C. X-ray scattering; E. Synchrotron radiation
PACS: 61.43.Gt; 61.10.Kw; 61.16.Bg
1
. Introduction
In recent papers [5,6] the in¯uence of J and T on the
porosity and PS±Si interface roughness was studied by
weighing, optical, and mechanical techniques in p-type
samples having PS thicknesses $5 mm. It was found that
an increase in J produces higher porosities and ¯atter inter-
faces between PS and Si, and that, for given J and electrolyte
composition, a decrease in T from room temperature to
2318C reduces remarkably the PS±Si interface roughness.
The temperature effect was ascribed to an increase in the
electrolyte viscosity which favors asperity smoothening
under anodization regime tending to electropolishing [7].
The same behavior was observed when a volume fraction
of the ethanol is replaced by a heavier (more viscous) alco-
hol like glycerol [6]. However, in those investigations the
authors determined average porosity values by gravimetric
measurements, thereby losing information on the porosity
Porous silicon (PS) is nowadays a very attractive material
for manufacturing silicon on insulator structures [1], optical
devices [2,3] and solid-state chemical microsensors [4]. The
wide ®eld of applications is due to the fact that the properties
of the material are very dependent on the type and resistivity
of the starting silicon (Si) substrate and on the electro-
chemical parameters used for the anodization process.
These latter are essentially the composition of the electro-
lyte (HF, H O and alcohol), the current density J and the
2
chemical bath temperature T.
*
Corresponding author. Tel.: 139-051-639-9184; fax: 139-051-
39-9216.
E-mail address: servidori@lamel.bo.cnr.it (M. Servidori).
6
0038-1098/01/$ - see front matter q 2001 Elsevier Science Ltd. All rights reserved.
PII: S0038-1098(01)00036-9

8
6
M. Servidori et al. / Solid State Communications 118 (2001) 85±90
depth-gradient in thick layers. All these aspects are very
important in the ®eld of PS-based optical devices.
2
. Experimental
PS layers with nominal thicknesses of 300 nm were
prepared by anodizing in the dark p-type (boron doping,
±9 V cm resistivity) Si substrates in a solution of HF/
8
H
2
O/ethanol ˆ 35:35:30. The J values were 16.6, 166 and
2
3
33 mA/cm , and T ˆ 24 or 2318C was used for anodi-
2
zation with J ˆ 16:6 mA=cm : For each of the three J
values, 10% of ethanol was replaced by glycerol. Moreover,
1
0-mm-thick PS layers were formed on doped Si at T ˆ 24
2
or 2318C using J ˆ 16:6 mA=cm without glycerol.
XRR measurements were made at the ID1 beamline of
the European Synchrotron Radiation Facility (ESRF) in
Grenoble (France). The energy of 7.1 keV was selected by
a pair of Si crystals oriented in ꢀn; 2n† 311 symmetric
Bragg-case diffraction, and
a single 111 symmetric
re¯ection Si analyzer was placed downstream the sample
to prevent the detector from collecting the strong diffuse
X-ray scattering from the porous structure. The beam
diameter at the sample was reduced to 50 mm by means of
slits. Longitudinal specular XRR scans were obtained from
Fig. 1. (a) Experimental (full circles) and calculated (line) XRR
spectra taken with 7.1 keV synchrotron X-rays from sample 2 in
Table 1. (b) Porosity pro®le resulting from the best ®t in (a).
3
00-nm-thick PS by rotating sample and analyzer in u=2u
coupling, and the measurements were simulated by a
computer code based on optical multislice theory [8]
which takes multiple scattering into account. From the
best ®t, obtained by routines of minimization between
experiment and calculation, layer thickness, porosity pro®le,
roughnesses at the PS±air and PS±Si interfaces were
determined.
depth-pro®le, as these measurements require chemical
removal of the PS layer in sodium hydroxide solution.
In this work the in¯uence of J, T and 10% substitution of
ethanol with glycerol on porosity, interface roughness and
porosity depth-pro®le is studied. Synchrotron X-ray re¯ec-
tivity (XRR) spectra and cross-sectional transmission
electron microscopy (XTEM) images were taken from
XTEM observations were performed at LAMEL Institute
with a Philips CM30-T microscope operating at 300 kV.
The instrument was equipped with a Gatan 794 multiple
scan camera allowing digital image recording on a
1024 £ 1024 pixel charge coupled device array.
3
00-nm-thick PS layers formed on p-type Si substrate.
Moreover, rocking curves (RCs) were obtained by multi-
crystal X-ray diffraction (MCXRD) from 10 mm PS thick-
ness on p-type Si to verify the in¯uence of T on the porosity
MCXRD measurements were made at the Laboratoire
de Spectrom e trie Physique (LSP) with a Philips MRD
Table 1
PS thickness (t) and interface roughness (s) obtained from XRR measurements and XTEM observations as a function of the current density (J),
the electrolyte temperature (T) and the glycerol percent. Discrepancies among some of the t and s values obtained from XRR and XTEM are
evident (see text for details)
2
J (mA/cm )
Sample
T (8C)
Glycerol (% vol)
t
XRR (nm)
tXTEM (nm)
s
XRR (nm)
s
XTEM (nm)
pÅ (%)
1
2
3
4
5
6
7
8
16.6
16.6
166
24
24
0
10
0
303 ^ 6
28 ^ 2
343 ^ 4
324 ^ 4
369 ^ 4
386 ^ 4
402 ^ 4
269 ^ 3
267 ^ 3
7.9 ^ 0.5
13.6 ^ 0.5
6.7 ^ 0.5
4.8 ^ 0.5
5.8 ^ 0.5
5.7 ^ 0.5
4.9 ^ 0.5
36.2 ^ 0.4
37.5 ^ 0.4
39.8 ^ 0.5
40.0 ^ 0.4
45.0 ^ 0.6
37.9 ^ 0.4
38.2 ^ 0.5
24
331 ^ 6
354 ^ 7
10 ^ 2
9 ^ 2
166
24
10
0
333
333
24
24
10
0
373 ^ 8
256 ^ 5
5 ^ 2
5 ^ 2
16.6
16.6
2 31
2 31
10

M. Servidori et al. / Solid State Communications 118 (2001) 85±90
87
at 7.1 keV X-ray energy. The p pro®le resulting from the
best ®t of Fig. 1a is reported in Fig. 1b, where a depth
gradient is evident. The low fringe contrast observed in
Fig. 1a is mainly due to the roughness at the PS±Si interface
(s) and in part also to the porosity gradient. The in¯uence of
the two parameters on the shape of XRR spectra can be
veri®ed by theoretical calculations, which show that a ¯at
p distribution with s ˆ 0 gives rise to a high fringe visibility
all over the range of incidence angles. By increasing s, this
visibility is smoothed out only in the high angle region of the
XRR spectrum, while a porosity gradient lowers the fringe
contrast also at low angles. Thanks to the minimization
routines, the correct p pro®le and s value are then obtained.
The PS thickness (t) and the PS±Si interface roughness
resulting from the simulation of the experimental XRR
spectra and from XTEM intensity pro®le measurements
are summarized in Table 1 as a function of J, T and glycerol
percent. In the last column, the values of the average por-
P
P
osity pꢀ ˆ
p t = ti are reported, where i is the index of
i i
i
i
each layer in which the whole PS thickness was divided
(
Fig. 1b) and p
i
is the porosity within each layer of thickness
i
t . The roughness values determined by XRR at the PS±air
interface are not shown in Table 1, as an almost constant
value of about 2 nm was obtained irrespective of changes in
the anodization parameters. The XRR data for all the
samples but sample 1 are reported in Table 1. This was
due to an unexpected and fast aging phenomenon observed
for PS under the synchrotron beam. The need to obtain
reliable spectra from stable samples required so many
sample alignment procedures and XRR measurement
collections as to exceed the time allocated at ESRF for all
the planned experiments. As will be described elsewhere
Fig. 2. Interface roughness (a) and average porosity (b) vs ano-
dization current density for the two different electrolyte tempera-
tures and alcohol compositions. The sample numbers are reported
near the respective symbols. The lines are guides to the eye.
1
diffractometer installed at a conventional Cu Ka source.
The instrument was equipped with a two-crystal, four-
re¯ection (220) germanium monochromator. RCs were
obtained from 10-mm-thick PS and were reproduced by a
dynamical diffraction code developed for implanted Si
layers [9], suitably modi®ed to take into account the reduc-
tion by porosity of the Fourier coef®cients of 4p times the Si
polarizability [10]. The depth pro®les of porosity (p) and
[
12], the PS aging was the reason why the pÅ values in
Table 1 are remarkably lower than the average porosities
of 0.55±0.75 reported in Refs. [5,6] for samples anodized
under the same experimental conditions.
Discrepancies among some of the t and the s values
obtained from XRR measurements and XTEM images are
evident in Table 1. These might arise from the much larger
sample surface probed by XRR in comparison with that
imaged by XTEM. However, the overall agreement is
satisfactory at exception of the t values for sample 6. In
the case of sample 6, both techniques give a low s value,
i.e. a nearly ¯at PS±Si interface, indicating that the dis-
crepancy cannot be attributed to the effects of local rough-
ness variations on the t values determined by XTEM
imaging a small portion of the sample. More probably, at
the origin of this disagreement are two main reasons. On the
one hand, in XTEM intensity pro®le measurements, small
systematic errors in the determination of the PS layer
thickness could be possible owing to the presence of a Fres-
nel fringe at the interface between PS and the denser Si
region. On the other hand, due to the grazing geometry of
XRR measurements, surface regions close to the rim of PS
layer, where inhomogeneities are expected, can contribute
to the measured spectra.
perpendicular lattice mismatch (m ) between PS and
'
substrate were determined by the same minimization
routines used for the simulations of the XRR spectra, and
were found to be linearly related to each other, according to
what was already observed [11].
All the samples were anodized at LSP and measured by
XRR and MCXRD just after the anodization. Some weeks
later, a few of them were thinned for XTEM observations.
3
. Results
Fig. 1a shows the experimental (dots) and calculated
line) XRR pro®les taken from a sample anodized
(
with electrolyte composition HF/H O/ethanol/glycerol ˆ
2
2
3
5:35:20:10, J ˆ 16:6 mA=cm and T ˆ 248C: Two inten-
sity decays are observed at incidence angles of about 0.20
and 0.258, corresponding to the critical angles for total
external re¯ection of PS layer and Si substrate, respectively,

8
8
M. Servidori et al. / Solid State Communications 118 (2001) 85±90
Fig. 3. XTEM images and SAED patterns (insets) taken from sample 1 (a) and sample 8 (b).
To comment upon the XRR and XTEM results in easier
It is worth underlining that, though the values of pÅ are
considerably lowered as a consequence of the aging
phenomenon mentioned above, the trends of pÅ and s with
J, T and glycerol percent follow those already observed in
Refs. [5,6] for a much greater PS thickness t $ 5 mm:
The strong difference in s is clearly evidenced in the
XTEM phase contrast micrographs of Fig. 3a and b, relative
to sample 1 and sample 8, respectively. We can also observe
the typical feature of particle-like Si regions (black dots in
Fig. 3a and b) in PS formed on p-type Si substrate. However,
in the case of sample 1 (Fig. 3a), a tendency to an alignment
of the Si particles along ®laments is observed in the region
close to the Si substrate, clearly indicating that a porosity
gradient is present in this sample. In the case of sample 8
(Fig. 3b), this tendency appears to be less pronounced
indicating a more homogeneous pore morphology in the
PS layer. The crystalline nature of the Si particles is
evidenced in the selected area electron diffraction patterns
(SAED) taken in PS regions close to the sample surface and
reported in the insets to these ®gures.
way, the s and pÅ data of Table 1 are plotted in Fig. 2 as a
function of J for the two temperatures and alcohol compo-
sitions. The general picture can be described as follows:
1
. PS±Si interface roughness (Fig. 2a). Anodizations at
room T without glycerol lead to well observable s
decreasing with increasing J (samples 1, 3, 5). For
given J, the s reduction produced by glycerol is more
marked the lower J is (samples 1, 2 and 3, 4) and is
comparable with the ¯attening effect due to the increas-
ing in J (samples 5, 6). Anodizations at low T and the
lowest J involve a further, though modest, s reduction to
the same value measured at the highest J, independently
of the alcohol composition in the electrolyte (samples 7,
8
).
2
. Average porosity pÅ (Fig. 2b). For anodizations performed
at room T, pÅ increases with J and the increase is faster
when the electrolyte contains glycerol (samples 2, 4, 6).
For low anodization T at the lowest J, an enhancement in
pÅ is observed (sample 7), whereas the 10% replacement
of ethanol with glycerol does not lead to an appreciable
further increase in pÅ (sample 8).
As to the porosity depth-pro®les obtained in the different
experimental conditions, we report the data relative to the
effects of J for the samples 2 and 6 (Fig. 4) and T for the
Fig. 4. Porosity pro®les obtained from samples 2 and 6 in Table 1.
Fig. 5. Porosity pro®les obtained from samples 2 and 8 in Table 1.

M. Servidori et al. / Solid State Communications 118 (2001) 85±90
89
2
Fig. 6. Experimental (open circles) and calculated (line) RCs for a 10-mm-thick PS formed with J ˆ 16:6 mA=cm at T ˆ 24 (a) and 2318C
(
(
b). Depth pro®les of the perpendicular lattice mismatch between PS and Si substrate (c) and porosity (d) resulting from the best ®ts in (a) and
b).
samples 2 and 8 (Fig. 5). In both ®gures, the gradient in the p
depth-distribution decreases with increasing J and reducing
T.
plane and the wavelength dispersion of the beam exiting the
monochromator. An additional blurring effect is given by
possible depth gradients of m
d, larger gradients in the m
'
. As can be seen in Fig. 6c and
and porosity pro®les are
The phenomenon of gradient variation as a consequence
of a change in T was investigated also for 10 mm PS thick-
nesses by MCXRD technique. Fig. 6a and b show the
'
produced by the higher T, underlining again the bene®cial
effect produced by lowering T. The comparison between
Figs. 5 and 6d indicates that this effect is not only preserved
but also enhanced in thicker PS. The surface peaks observed
in the m' and p pro®les of Fig. 6c and d, respectively, are
very probably due to pore size widening induced by the
prolonged anodization time necessary to form thick PS
layers and/or by Si skeleton consumption associated with
the beginning of post-anodization PS oxidation to air. As a
®nal remark, an average porosity pꢀ ˆ 0:59 ^ 0:05 results
from the p pro®le in the sample anodized at T ˆ 2318C
(Fig. 6d). This value is much higher than that of sample 8
in Table 1 and agrees well with the one expected from the
used anodization conditions [5,6] in the absence of aging.
This suggests that the aging phenomenon under laboratory
X-rays is not so fast as that observed at the ID1 beamline of
ESRF.
experimental RCs (circles) obtained with J ˆ
2
1
6:6 mA=cm for T ˆ 24 and 2318C, respectively. In
both spectra, the substrate peak at the reference angle Du ˆ
0
and an intensity contribution given by PS at lower angles
are present. The PS intensity pro®les are quite different for
the different T, consisting of a double peak for T ˆ 248C and
a single peak for T ˆ 2318C: To extract quantitative
information, the RCs were simulated by constraining the
whole m' and p pro®les to be con®ned within the nominal
10 mm thickness. This constraint is imposed by the fact that
for a very thick epilayer of even constant m , the angular
'
frequency of the intrinsic intensity oscillations in the RC,
driving the diffraction code towards the determination of the
surface layer thickness, is so narrow as to be completely
smoothed out by the angular divergence in the diffraction

9
0
M. Servidori et al. / Solid State Communications 118 (2001) 85±90
be veri®ed in a subsequent paper [15], where results of
4
. Discussion and conclusions
grazing incidence small angle X-ray scattering (GISAXS)
measurements made at the ESRF-ID1 beamline will be
reported. The GISAXS data were taken from the same
samples investigated here, immediately after the XRR
measurements.
This paper deals with the in¯uence of some anodization
parameters (current density, electrolyte temperature and
percent of ethanol-replacing glycerol in the chemical bath)
on the macroscopic structural characteristics of single PS
layers formed on p-type Si substrates. It was already
observed [5,6] that increasing J and glycerol percent, and
decreasing T produce PS layers with higher porosity and
lower roughness at the interface with the substrate. In
addition to this, the present paper shows that a lowering in
T favors the formation of ¯atter porosity depth-pro®les.
These effects are linked to an increase in the electrolyte
viscosity, which results from lowering T, as well as
replacing part of ethanol with the heavier glycerol. As to
the effect of J, this can be interpreted as a manifestation of a
limited diffusion of the chemical species which becomes
more apparent when the reaction is driven on a shorter
timescale by increasing the current density. Indeed, it is
generally admitted that above a critical current density
electropolishing prevails above pore formation [7]. There-
fore, in an electrolytic regime of viscosity-induced limited
diffusion, higher porosities, more homogeneous porosity
depth-pro®les and ¯atter interfaces are obtained.
References
[
[
1] G. Bomchil, A. Halimaoui, R. Herino, Microelectron. Engng 8
1988) 293.
(
2] M.G. Berger, S. Frohnhoff, W. Theiss, U. Rossow,
H. Muender, Porous Silicon Science and Technology,
Springer, New York, 1994, p. 345.
[
[
3] A. Loni, L.T. Canham, M.G. Berger, R. Arens-Fisher,
H. Muender, H. Luth, H. Arrand, T.M. Benson, Thin Solid
Films 276 (1996) 143.
4] P. Maccagnani, L. Dori, P. Negrini, Proc. Eurosensor XIII
Conference, The Hague, 1999, p. 817.
[5] S. Setzu, G. L e rondel, R. Romestain, J. Appl. Phys. 84 (1998)
3129.
[6] S. Setzu, PhD thesis, Universit e J. Fourier, Grenoble, France,
1999.
[
[
[
7] D.R. Turner, J. Electrochem. Soc. 105 (1957) 402.
8] L.G. Parrat, Phys. Rev. 95 (1954) 359.
Good structural characteristics of PS are very important
from the point of view of optical devices. For instance, light
transmission in PS-based waveguides can undergo severe
losses through diffuse scattering at interface roughnesses
9] M. Servidori, F. Cembali, S. Milita, X-Ray and Neutron
Dynamical Diffraction: Theory and Applications, Series B:
Physics, vol. 357, Plenum Press, New York, 1996 (p. 301).
[
13,14] and have perturbed propagation modes owing to
[
[
10] A.A. Lomov, D. Bellet, G. Dolino, Phys. Status Solidi (B) 190
1995) 219.
11] D. Bellet, G. Dolino, Thin Solid Films 276 (1996) 1.
different optical refractive indices associated with porosity
depth-gradients. As described in this work, these detri-
mental behaviors can be removed by choosing appropriate
anodization parameters.
(
[12] S. Sama, C. Ferrero, S. Lequien, S. Milita, R. Romestain, M.
Servidori, S. Setzu, D. Thiaudiere, J. Phys. D: Appl. Phys.
(2001) in press.
In previous papers [5,6] quite different optical refractive
indices were measured in PS layers having the same average
porosity, though obtained with different J and T values. To
account for this result, the hypothesis was made that pores of
different sizes are formed in the layers. This hypothesis will
[13] G. Le
Ârondel, R. Romestain, F. Mad e ore, F. Muller, Thin Solid
Films 276 (1996) 80.
14] G. L e rondel, R. Romestain, Thin Solid Films 297 (1996) 114.
15] M. Servidori et al., in preparation.
[
[