A miniature fuel cell with porous Pt layer formed on a Si substrate

Article abstract of DOI:10.1149/1.3545070

A novel miniature fuel cell is proposed with monolithically fabricated Si electrodes. We recently discovered that a porous Pt layer can be formed uniformly on Si substrate by immersing porous Si into a Pt plating bath containing HF. To exploit the porous Pt as a catalyst layer, we explored the conditions for porous Pt formation. After formation of the porous Pt catalyst layer on a Si wafer, fuel channels were opened by applying dry etching from the opposite side of the Si wafer until the etching had reached to the porous layer. The porous Pt layer could be used as a stopping layer for the dry etching, and a through-chip porous layer was obtained. Two such Si electrodes were hot-pressed onto either side of a polymer electrolyte membrane, and 230 m thick prototype cells were constructed. By feeding H₂ and O₂ to the anode and cathode side, respectively, we obtained power generation at a peak power density of 145 mW/cm².

Full text of DOI:10.1149/1.3545070

Journal of The Electrochemical Society, 158 (4) B355-B359 (2011)
0
B355
013-4651/2011/158(4)/B355/5/$28.00 VC The Electrochemical Society
A Miniature Fuel Cell with Porous Pt Layer Formed on
a Si Substrate
a,b, ,z
a
a
*
Masanori Hayase,
Tomoya Fujii, and Jose Geraldo Alves Brito-Neto
a
Faculty of Science and Technology, Tokyo University of Science, 2641 Yamazaki, Noda-shi, Chiba-ken 278-8510, Japan
Research Center for Green and Safety Sciences, Tokyo University of Science, 1-3 Kagurazaka, Shinjuku, Tokyo
b
1
62-8601, Japan
A novel miniature fuel cell is proposed with monolithically fabricated Si electrodes. We recently discovered that a porous Pt layer
can be formed uniformly on Si substrate by immersing porous Si into a Pt plating bath containing HF. To exploit the porous Pt as
a catalyst layer, we explored the conditions for porous Pt formation. After formation of the porous Pt catalyst layer on a Si wafer,
fuel channels were opened by applying dry etching from the opposite side of the Si wafer until the etching had reached to the po-
rous layer. The porous Pt layer could be used as a stopping layer for the dry etching, and a through-chip porous layer was obtained.
Two such Si electrodes were hot-pressed onto either side of a polymer electrolyte membrane, and 230 lm thick prototype cells
were constructed. By feeding H
2
power density of 145 mW/cm .
2
and O
2
to the anode and cathode side, respectively, we obtained power generation at a peak
VC 2011 The Electrochemical Society. [DOI: 10.1149/1.3545070] All rights reserved.
Manuscript submitted October 5, 2010; revised manuscript received December 28, 2010. Published February 16, 2011.
Microelectromechanical systems (MEMS) fabrication technol-
ogy is an important tool for miniaturizing the fuel cell structure to
micrometer scales, and it is well suited for mass production. Many
studies now exist for fuel cell miniaturization using MEMS techni-
Novel Electrode Design and Fabrication Process
6
In our previous work, the fuel channels were formed on the Si
substrate by alkaline etching, and the catalyst metal was deposited
onto a through-chip porous Si layer formed at the bottom of the
channels, as shown in Fig. 1a. The catalyst metal deposition was
problematic, however. First, it was not easy to ensure that the pores
uniformly reach to the opposite side of the membrane by anodiza-
tion alone. The resulting porous layer was very brittle and would of-
ten break during the production steps. Moreover, the fuel channels
produced by anisotropic wet etching were narrower at the base than
at the top, reducing the area for catalyst usage.
1–16
ques.
Various MEMS fabrication techniques have been used. In
early studies, for example, Kelley et al. created a catalyst layer sup-
porter on a Si chip and demonstrated that the resulting miniaturized
direct methanol fuel cell (DMFC) has almost the same performance
1
as state-of-the-art fuel cells having conventional structure. In most
of the studies, conventional catalyst layers are used, in which carbon
black with catalyst metal particles is sprayed. However, MEMS
fabrication techniques employ basically monolithic structures, and
treatment with powders such as carbon black is not appropriate. To
adapt the construction process to MEMS fabrication, various
approaches exist, but the performance of the prototypes was gener-
To overcome these problems, we opted for dry etching to open
the fuel channels into the Si substrate, as shown in Fig. 1b. Dry etch-
ing ideally produces channels of rectangular cross section with the
deep reactive ion etching (DRIE) process, which maximizes wafer
utilization. The process order was also modified. We opened the
fuel channels, after synthesizing the catalyst layer on the opposite
side of the wafer. This modified process should allow us to circum-
vent difficulties related to the production and manipulation of a
through-chip porous Si layer. The porous Pt layer should enable this
5
–14
ally poor.
Much attention has been paid to forming catalyst
layers, which involve a large surface area. Recently, Kuriyama et al.
proposed a novel thin Si based fuel cell using carbon nanotubes as a
catalyst support; a relatively high power output was reported.
8
1
1–15
Novel approach of electrolyte have also been explored,
and
1
5
chemically modified porous Si showed promising results. In gen-
eral, polymer electrolyte fuel cells (PEFCs) have been employed in
realizing miniature fuel cells, whereas miniaturization of solid oxide
6
strategy, because the etching speed of the SF plasma is much lower
at the porous platinum catalyst layer than for crystalline Si. In other
words, the porous Pt layer should act as a stopping layer for the dry
etching.
1
6
fuel cells (SOFCs) have been recently actively studied.
We also have proposed a Si based miniaturized fuel cell, in
which a through-chip porous Si layer was used to make the cata-
lyst. Porous Si is easily produced by anodization; it is an attractive
Figure 2 shows the fabrication process of the fuel cell. First, a
Cu layer is patterned on a Si wafer, which acts as an electrical con-
tact and a dry etching mask. The Si wafer is then anodized in HF
solution, and a porous Si layer is formed. Immersion plating is per-
formed immediately after the porous Si is formed, to give a porous
Pt catalyst layer. Fuel channels are etched by plasma until their
bottoms reach to the porous layer. Two Si electrodes are hot-pressed
with Nafion solution onto either side of a polymer electrolyte mem-
brane (PEM). The cells are stacked; the fuel channels are open at
the top at this point.
6
material due to its large surface area and compatibility with MEMS
fabrication techniques. Studies by other groups, exist, but the high
5
resistivity of porous Si leads to poor performance of the fuel cells.
Recently, we found that a porous Pt layer can be obtained by
immersing high-porosity porous Si into a Pt plating bath containing
1
7
HF.
X-ray energy-dispersive spectroscopy (EDS) analysis
revealed that the porous layer consists of mostly Pt; little signal
of Si was observed. In the plating bath, it is assumed that Si
reduces Pt ions, and a metallic Pt deposit forms on the porous Si
surface, while Si itself forms Si oxide. The Si oxide is removed
by HF, which is added to the plating bath. This reaction continues
until almost all Si has been replaced by Pt. The resistivity of the po-
rous Pt should be low, and the porous Pt should act as a suitable cat-
alyst for miniaturized fuel cells. In this paper, we propose a novel
design of the miniaturized fuel cell electrodes. Prototype fuel cells
were built, and preliminary results of power generation tests are
shown.
Experimental
To verify the new electrode fabrication strategy, we prepared a
prototype fuel cell with the new Si electrodes.
Si wafer preparation.— The Si wafers used in this work were
n-type,110 lm thick, (100)-oriented, and mirror-polished on both
sides. Two different kinds of wafers were used: (1) P-doped, 0.01–
0
.02 X cm resistivity, (“medium-doping wafers,” below); and (2)
as-doped, 0.001–0.003 X cm resistivity (high-doping wafers). The
Si wafers were cut into pieces, 15 mm square, which were soaked in
concentrated HF to remove the thermal oxide layer and thoroughly
washed with water, then ethanol, and dried under air.
*
z
Electrochemical Society Active Member.
E-mail: mhayase@rs.noda.tus.ac.jp.
Downloaded on 2015-06-01 to IP 138.251.14.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

B356
Journal of The Electrochemical Society, 158 (4) B355-B359 (2011)
strength is important when the porous Pt layer is used as an etching
stop layer and catalyst layer. Our preliminary experiments with the
lm thick porous Pt layers found that it could stand dry etching,
7
but cracks were often observed in it. To improve the mechanical
strength of the porous Pt layer, therefore thicker layers were sought
by optimizing the anodization current, anodization time, and plating
time.
Fuel channels.— After the porous Pt catalyst layer was prepared,
the fuel channels were opened on the Si substrate by dry etching
from the backside. No DRIE machine was available in our facility,
so that a conventional parallel plate type reactive ion etching (RIE),
6 2
Samco RIE 10N (Japan), was used. The flow rates of SF and O
gases were 18 and 4 sccm, respectively. The pressure in the reaction
chamber was about 10 Pa. At 30 W power, 70–80 min of etching
was needed to reach the porous Pt layer. Dry etching was interrupted
a few times so that the specimen could be observed by optical
microscopy, until the porous layer became visible.
Figure 1. Schematic view of monolithic Si electrodes for miniature fuel
cells. (a) Previous design, as proposed in Ref. 8. The through-chip porous
region is made before Pt catalyst deposition. (b) Design proposed in this
study. The through-chip porous region is made after Pt catalyst deposition.
Prototype assembly.— For electrical contact, a thin gold layer
was formed at the border of the catalyst layer, as described in our
17
previous work, because direct electrical contact between the
porous Pt catalyst layer and Si does not appear to be adequate
enough, Two Si electrodes were then hot-pressed onto either side of
a PEM. This was a TSF-1301 model PEM obtained from Toagosei,
Co., Ltd. (Japan). TSF-1301 is a pore-filling electrolyte membrane,
which, advantageously, does not swell when exposed to water or
methanol. Nafion solution was used as an adhesive in the assembly
process as follows: 0.8% Nafion solution was spread on both elec-
trodes, and the PEM was immersed in a 1.7% Nafion solution. [5%
Nafion solution from Wako Chemicals (Japan) was diluted to the
specified concentration with isopropyl alcohol.] The membrane and
electrodes were hot-pressed at 373 K and 0.09 MPa for 30 min and
cooled to the room temperature.
Copper mask.— A conventional parallel electrodes type dry etch-
ing machine was used for opening of the fuel channels in the later
process. A metal mask was required to etch Si deeply using this
machine; a 200 nm thick Cu film was sputtered onto the back side
of the wafers, which also acts as an electrical contact for the
anodization process. Using a photolithographic patterning and sub-
8
sequent wet etching with 40 wt % FeCl solution, the Cu film on the
Si wafer was successfully patterned with 100 lm width stripes of
pitch 200 lm within a 5 Â 5 mm area.
Porous Pt layer.— The apparatus and procedure for the porous
17
Pt formation were the same as described in our previous paper. In
summary, a porous Si layer was produced on the single-crystal Si
substrate by anodization in an HF-containing electrolyte. After
anodization, the porous Si layer was immersed in a solution contain-
Cell testing.— Cell polarization experiments were performed
inside a thermostatic chamber. The top of the fuel channels on the
prototype fuel cell is open, because a cell stack is planned. At this
point, the prototype cell was set into an aluminum casing, to which
fuel and air supply lines were connected. As shown in Fig. 3, fuel
was fed from the inlet hollow into the edge of the fuel channels and
passed through the channels on the Si electrode to the outlet hollow.
The clamping pressure was adjusted by the spring. It was varied
from 0.02 to 0.2 MPa in preliminary testings, and no obvious differ-
ence in cell performance was observed. Then, we chose 0.05 MPa
2–
ing [PtCl6] and HF. Hydrofluoric acid, sulfuric acid, and pure
(
(
99.5%) ethanol and acetone were obtained from Wako Chemicals
Japan) and were used as received. [PtCl ] solutions were prepared
6
2
–
by dissolving solid H PtCl 6H O, which was obtained from Wako
2
6
2
Chemicals (Japan). All solutions were prepared using ultrapure
water (Milli-Q, Millipore). Anodization and plating took place at
2
83 K inside a refrigerated chamber under ambient illumination.
The anodization electrolyte was prepared by mixing concentrated
hydrofluoric acid (46 wt %), ultrapure water, and ethanol in 3:5:2
mass ratio. The anodization area was defined by a Kalrez O-ring
2 2
as the clamping pressure. Humidified H and dry O were fed to the
anode and cathode sides of the cell, respectively. The cell potential
was scanned at 2 mV/s using a potentiostat (Hokuto-Denko
HABF5001, Japan) controlled by a computer running our own data
acquisition software.
(
diately after acetone rinsing, following the anodization. The plating
inner diameter, 9.25 mm). Immersion plating was performed imme-
2
–
bath consists of a 20 mmol/L [PtCl
00 mmol/L HF in 1 mol/LH SO .
In our previous paper, only possible porous noble metal layers
6
]
solution containing
3
2
4
1
7
Results and Discussion
of thickness ꢀ 7 lm were studied. We had little knowledge of the
Porous Pt layer.— Porous Pt formation has not yet been investi-
gated in detail, and there are several relevant parameters. Our
maximum possible thickness of the porous Pt layer. Mechanical
Figure 2. (Color online) Fabrication pro-
cess of the miniature fuel cell. A porous
Pt catalyst layer is synthesized with po-
rous Si, and the porous Pt layer works as
an etching stopping layer. Two electrode
chips are combined with a PEM sheet and
a prototype cell results.
Downloaded on 2015-06-01 to IP 138.251.14.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 158 (4) B355-B359 (2011)
B357
Figure 3. Al alloy casing for power generation tests. Cell stacking is
designed; the proposed cell has open fuel channels. For testing of single cell,
casing is needed for the fuel supply.
present goal is demonstration of the new Si electrode fabrication
process and prototype fuel cell construction. Optimization of the
parameters for porous Pt layer formation is beyond our present
scope, and the experiments proceeded by trial and error.
First, the effect of porous Si morphology on the metallization was
investigated. The pore diameter and porosity of porous Si tend to
increase with the anodization current density, whereas the pore wall
roughness tends to be smooth with large anodization current
Figure 5. The thickest porous Pt layer obtained in this study.
18
(STEM) image reveals roughness on the scale of a few nanometers,
on 100 nm scale agglomerate. Medium-doping Si was used, and
the porous Si was formed with anodization current density of
density.
Figure 4 shows the typical external appearance and cross-
sectional electron microscope images of the porous layer after the
immersion plating. The scanning transmission electron microscope
2
7
0 mA/cm . Porous Pt layer formation appears to be difficult,
because a dense Pt deposit forms on the porous area close to the top
surface and impairs the flux of reagents to deeper regions. Longer
immersion times scarcely enhanced the porous Pt deposit zone.
To improve this part of the process, we used anodization current
2
densities greater than 100 mA/cm for medium-doping Si and greater
than 50 mA/cm for high-doping Si in an attempt to create a thick po-
2
rous Pt layer. The anodization and plating times were extended, and
with each new set of conditions, porous Pt synthesis was attempted.
The thickest layer that we were able to obtain is shown in Fig. 5.
Medium-doping Si was used, and an anodization current of 100 mA/
2
cm was applied for 1100 s. Immersion plating was then performed
with 10 mL solution for 3.5 h. Beyond this point, the mechanical sta-
bility of the porous Si layers themselves appeared to place a limit on
the thickness of the porous Pt layers. Porous Si layers thicker than
these were very fragile and peeled away spontaneously from the sub-
strate after anodization. Within the rather limited experimental condi-
tions, we did not observe significant morphological differences in the
resulting porous Pt. Agglomerates was observed of spherical nanopar-
ticles with diameters between 40 and 100 nm in all cases. There were
some differences in the agglomerate density, but we did not study
this phenomenon in any detail. Further study is needed for optimiza-
tion of the porous Pt layer.
Figure 4. (Color online) Typical porous Pt layer appearance. (a) External
view. Anodization and plating took place in a fluorocarbon resin vessel
sealed by an O-ring; the dark circle is the porous region. (b) Cross-sectional
SEM image. The specimen chip was cleaved, and the cross-section was pre-
pared. (c) Magnified cross-sectional SEM image around the middle depth of
the porous Pt layer. (f) Further magnified STEM high-angle annular dark
field (HAADF) image.
Figure 6. Cross-sectional observation of a porous catalyst layer formed
under the same conditions used for prototype fabrication. SEM image and
EDS elemental mappings of Si and Pt are shown. The anodization current
was reduced linearly with time from 60 mA/cm to zero for mechanical
stability.
2
Downloaded on 2015-06-01 to IP 138.251.14.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

B358
Journal of The Electrochemical Society, 158 (4) B355-B359 (2011)
Figure 7. Cross-sectional SEM image of a Si electrode. The specimen chip
was mounted in an epoxy resin, and the cross-section was prepared by pol-
ishing. A through-chip porous Pt region was produced successfully by dry
etching.
Prototype testings.— For fuel-cell electrode production, we used
the following synthesis conditions: high-doping wafers; anodization
performed for 1380 s, with current density decreasing linearly from
6
2
0mA/cm down to 0; and immersion plating for 1800 s. Figure 6
shows a cross section. EDS Si and Pt analyses were also performed,
and the EDS quantitative analysis showed Si/Pt atomic ratio to be 3/
Figure 9. Polarization curves and power density of prototype cells at 313 K.
Humidified H and dry O were supplied. Peak power densities were
observed at 54 and 145 mW/cm for the stripe channel and the square chan-
nel designs, respectively.
2
2
9
7, 66/34, and 96/4 at points a, b, and c in Fig. 6, respectively. Our
2
aim in decreasing the current linearly to zero was to reduce the
porosity of the porous Si layer near its bottom and improve the
stability of the specimen during the dry etching step.
torsional deformation caused cracks in the 5 mm length through-
chip porous region. Many factors could be optimized to improve
performance, but the poor yield impaired prototype constructions.
We then reduced the reaction area from 5 Â 5 to 3 Â 3 mm, and the
shape of the fuel channel openings was changed from 5 mm length
stripes to 100 Â 100 lm square openings, as shown in Fig. 10. Apart
from the channel design, the fabrication process and the operation
conditions were identical to those with the stripe channels. This
design change greatly improved the yield, and several prototypes
were constructed. The best result is shown in Fig. 9. Peak output
Figure 7 shows a typical scanning electron microscope (SEM)
cross-sectional view of one such sample after dry etching of the fuel
channels. Above the Pt-rich region of the porous layer, the Si-rich
porous region is still visible. As expected, the etching rate was sig-
nificantly reduced around this region relative to that of bulk Si. It is
possible to etch the channel all the way down to the Pt-rich porous
Pt layer, but we found it better to stop short of that for the sake of
the mechanical stability of the specimen chip. Figure 8 shows a
photograph of the membrane electrode assembly (MEA) after hot-
pressing and the assembled test device. The thickness of the
prototype cell was 230 lm, far thinner than conventional fuel cells.
Figure 9 shows a polarization curve (stripe channel) obtained by
feeding 8 sccm humidified H and dry O at 313 K to the anode and
2
2
was increased from 54 mW/cm (stripe channel) to 145 mW/cm
square channel). The increase was due to the higher production
yield, because more prototypes could be tested.
(
2
2
2
At current densities greater than 250 mA/cm , there is an
increase in the slope of the polarization curve. This suggests that the
transport of reactive species begins to limit the device performance
at high currents. The polarization curve is still relatively linear
throughout the whole range, however, indicating that we are still far
from the transport limiting regime. The principle mechanism re-
sponsible for voltage loss in the device appears to be the ohmic drop
through the polymer electrolyte membrane and its interface layers
with the electrodes. We are currently working to characterize and
cathode sides, respectively. The overall power density was signifi-
cantly improved over our previous prototype whose peak power
6
2
density was 1.5 mW/cm . The output power density was still far
inferior to that of conventional larger type PEFCs, however.
The production yield of the electrode chips was quite poor,
because the through-chip porous region was fragile. In particular,
Figure 10. (Color online) External appearance of a prototype cell with
square channels. The production yield of prototypes with stripe channels
shown in Fig. 8 was prohibitively low, so the size of the through-chip porous
region was reduced. Following this modification, several prototypes were
Figure 8. (Color online) External appearance of a prototype cell with stripe
channels. Two Si electrodes were combined with a PEM sheet.
2
constructed. A maximum power density of 145 mW/cm was obtained.
Downloaded on 2015-06-01 to IP 138.251.14.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 158 (4) B355-B359 (2011)
B359
improve this part of the device. Although the production procedure
was identical, the peak output varied from 20 to 145 mW/cm , and
Tokyo University of Science assisted in meeting the publication costs of
this article.
2
reproducibility was poor. The hot-pressing of the electrodes and the
PEM appears to be the most delicate step in the assembly of the cell
and is the source of most of the problems affecting the performance
of the device. For instance, the contact resistance between the PEM
and the electrodes seems to be the major factor responsible for the
ohmic drop in the polarization curve. Further improvements to the
hot-pressing step are clearly necessary.
References
1
2
3
4
5
6
7
.
.
.
.
.
.
.
S. C. Kelley, G. A. Deluga, and W. H. Smyrl, Electrochem. Solid-State Lett., 3,
407 (2000).
S. J. Lee, A. Chang-Chien, S. W. Cha, R. O’Hayre, Y. I. Park, Y. Saito, and F. B.
Prinz, J. Power Sources, 112, 410 (2002).
T. Ito, S. Kaneko, and M. Kunimatsu, Electrochem. Solid-State Lett., 12, B154
2009).
A. Kamitani, S. Morishita, H. Kotaki, and S. Arscott, J. Micromech. Microeng.,
8, 125019 (2008).
G. D’Arrigo, C. Spinellaa, G. Arenab, and S. Lorenti, Mater. Sci. Eng., C, 23,
318 (2003).
M. Hayase, T. Kawase, and T. Hatsuzawa, Electrochem. Solid-State Lett., 7, A231
2004).
K. Min, S. Tanaka, and M. Esashi, J. Micromech. Microeng., 16, 505 (2006).
(
1
Conclusions
1
We have proposed a novel microfuel cell design, employing
monolithic Si based electrodes with fuel channels opened by dry
etching and using porous Pt layers as catalyst and etch-stop layer.
By modulation of the porous Pt synthesis conditions so as to yield
thick layers, we obtained stability of these prototype electrodes with
respect to dry etching and built fuel channels with bottoms open to
through-chip pores. This led to an overall increase in the output
power density of our devices, and a significant increase in peak
power density over our previous prototype was demonstrated. There
remains scope for improvement in the interface between the poly-
mer electrolyte and the electrodes, and this is currently under
investigation.
(
8. N. Kuriyama, T. Kubota, D. Okamura, T. Suzukib, and J. Sasahara, Sens. Actua-
tors, A, 145–146, 354 (2008).
Y. Zhang, J. Lu, H. Zhou, T. Itoh, and R. Maeda, J. Microelectromech. Syst., 17,
020 (2008).
9.
1
1
0. Y. Zhang, J. Lu, H. Zhou, T. Itoh, and R. Maeda, J. Micromech. Microeng., 19,
15003 (2009).
11. S. Tominaka, S. Ohta, H. Obata, T. Momma, and T. Osaka, J. Am. Chem. Soc.,
30, 10456 (2008).
1
1
2. T. Pichonat and B. Gauthier-Manuel, J. Micromech. Microeng., 15, 179 (2005).
3. T. Pichonat and B. Gauthier-Manuel, J. Power Sources, 154, 198 (2006).
1
14. J. P. Esquivel, T. Senn, P. Hernandez-Fernandez, J. Santander, M. Lorgen, S.
Rojas, B. Lochel, C. Cane, and N. Sabate, J. Power Sources, 195, 8110 (2010).
1
1
1
5. S. Moghaddam, E. Pengwang, Y. B. Jiang, A. R. Garcia, D. J. Burnett, C. J.
Brinker, R. I. Masel, and M. A. Shannon, Nat. Nanotechnol., 5, 210 (2010).
6. H. Huang, M. Nakamura, P. Su, F. Fasching, Y. Saito, and F. Prinz, J. Electro-
chem. Soc., 154, B20 (2007).
Acknowledgments
This study was partly supported by JSPS KAKENHI (22510126)
and NEDO of Japan. J.G.A.B.-N. acknowledges JSPS for a postdoc-
toral fellowship.
7. J. G. A. Brito-Neto, S. Araki, and M. Hayase, J. Electrochem. Soc., 153, C741
(2006).
18. V. Lehmann, R. Stengl, and A. Luigart, Mater. Sci. Eng., B, 69–70, 11 (2000).
Downloaded on 2015-06-01 to IP 138.251.14.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).