,.,
..
,_
..I
..:<.>7!.:,,;&..-&::
J.
Am.
Ceram.
Soc.,
78
[21
479-86
(1995)
Processing
of
Silicon Carbide-Mullite-Alumina Nanocomposites
Yoshio
Sakka*
National Research Institute
for
Metals,
2-1,
Sengen-1, Tsukuba, Ibaraki
305,
Japan
Donald D. Bidinger* and Ilhan
A.
&say*
Department
of
Chemical Engineering and Princeton Materials Institute, Princeton University,
Princeton, New Jersey
08544-5263
Nanocomposite materials in the form
of
nanometer-sized
second-phase particles dispersed in a ceramic matrix have
been shown to display enhanced mechanical properties. In
spite
of
this potential, processing methodologies to produce
these nanocomposites are not well established. In this paper,
we describe a new method for processing SiC-mullite-
Al,O, nanocomposites by the reaction sintering
of
green
compacts prepared by colloidal consolidation
of
a mixture
of
SIC and A1,0, powders. In this method, the surface of the
Sic particles was first oxidized to produce silicon oxide and
to reduce the core
of
the SIC particles to nanometer size.
Next, the surface silicon oxide was reacted with alumina to
produce mullite. This process results in particles with two
kinds
of
morphologies: nanometer-sized Sic particles that
are distributed in the mullite phase and mullite whiskers in
the
SIC
phase. Both particle types are immersed in an
Al,O, matrix.
I.
Introduction
ERAMIC-MATRIX
nanocomposites have been receiving
C
increasing attention largely due to their significantly
enhanced mechanical properties, low-temperature densifica-
tion, machinability, and superplastic behavior.’-3 For instance,
in the pioneering studies of Niihara and his colleagues, fracture
strengths as high as 1.5 GPa and toughnesses as high as
7.5
MPa.m”* have been reported in systems where nanometer-sized
Sic was dispersed in A1,O1, MgO, Si,N,, and mullite matri-
ces.’., Niihara classified nanocomposites into two general
~ategories.~ One category consists of composites of only nano-
meter-sized grains. The other consists
of
composites where
nanosized particles are distributed within the intra- andor
intergrain regions of micrometer-sized grains. The main advan-
tage in using nanocomposites of the first category is that they
can be shaped by superplastic deformation, whereas the main
advantage
of
the nanocomposites of the second category is their
enhanced high-temperature stability against grain coarsening.
In this paper, we deal with the processing of the second type
of
nanocomposite with a novel reaction sintering method.
Nanocomposites produced by Niihara’s group have been
formed mainly by using composite powders mixed on the nano-
meter scale by conventional milling of micrometer-sized pow-
ders. Although the mechanisms
of
size reduction to nanometer
scale and the entrapment of particles within each other are not
clearly understood, this conventional milling approach has cer-
tainly been effective in processing nanocomposites with unique
properties despite the fact that milling of powders generally
results in contamination or reproducibility problems. Alterna-
tively, chemical vapor deposition
(CVD)
has also been success-
fully used to process similar nanocomp~sites.~ However, as
stated by Niihara? the use of CVD
to
fabricate large and com-
plex-shape components is not suitable
for
mass production.
We propose the use of reaction sintering as a more economi-
cal and reliable method for producing nanocomposites. In this
study, we chose the Sic-mullite-Al,03 system because it has
been shown that Sic-mullite, SiC-AI,O,, and mullite-Sic
systems have excellent mechanical
proper tie^.^^
Our
procedure
consists of the three steps shown in Fig. 1.’ First, we consoli-
dated micrometer-sized Sic and A1,0, powders homoge-
neously through colloidal consolidation (slip casting), which is
known to be an excellent processing route for improved
mechanical behavior.’.’ Second, partial oxidation treatment was
conducted to oxidize the surface of the Sic particles to
SiO,
and thus to reduce the size of the Sic particles to nanometer
size. Finally, we reacted the surface oxide and A1,0, to produce
mullite. As we illustrate in the following sections, the advan-
tages of this method are that
(1)
the reduction of the inclu-
sion phase to nanoscale can be achieved without milling and
(2)
because of a volume increase during reaction sintering, the
sintering shrinkage is low.
11.
Experimental Procedure
The a-Al,O, powder used in this study was Sumitomo
Chemical’s (AKP-50) high-purity alumina (299.995%) with a
mean particle diameter of
0.21
pm
and a specific surface area
of
9.5 m’lg. Two types
of
0-Sic powders were used: Superior
Graphite’s HSC059 containing C
(0.85
wt%), Si
(0.03),
N
(0.21),
and
0
(0.80) as major impurities, and Mitsui Toatsu’s
MSC-20 containing
SiO,
(0.11 wt%) as a major impurity.
Figure
2
shows TEM photographs of the two powders. Superior
Graphite’s powder is relatively coarse (indicated as SiC(C))
with a wide particle size distribution, a mean particle diameter
of 0.56 pm, and a specific surface area of 15.0 m2/g. Mitsui
Toatsu’s powder is finer (indicated as SiC(F)) with a narrow
particle size distribution, a mean particle diameter
of
0.15 p,m,
and a specific surface area of 21.3 mZ/g.
M. Sacks--contributing editor
Reaction
Sin terina
Consolidation Oxidation
-
Manuscript
No.
194576. Received May 12,1993; approved
March
23,1994.
Supported
by
the
U.S.
Air Force Office
of
Scientific Research under Grant
No.
AFOSR-F49620-93- 1-0259.
Fig.
1.
Schematic illustration
of
the
Process
steps
used to
oroduce
‘Member, American Ceramic Society.
niocomposites
by
reaction sintering.
’
479
480
10
0-
k
a
-10-
!i
-20
4
-30:
-40
Journal
of
the American Ceramic Society-Sakka et al.
-
1
-
Vol. 78, No.
2
Fig.
2.
TEM
photographs
of
(a)
SiC(C)
and
(b)
SiC(F)
powders.
The zeta potential of the two Sic powders in suspension was
obtained by the Smoluchowski equation'' from the electropho-
retic mobility (Otsuka Electronics ELS-800 and Be1 Japan Zeta-
sizer 4) determined on diluted suspensions containing
0.01M
NaCl
(or
0.01M
KCl) to control ionic strength. The pH was
adjusted using NaOH (or KOH) and HC1.
Stable colloidal suspensions with a solid content of 45 vol%
were prepared electrosterically in distilled water with an NH,
salt
of
poly(methacry1ic acid) (PMAA, Darvan C) at pH 10 as
described in the next section. NH,OH was used to adjust the
pH. After ultrasonic vibration (Sonic Materials Vibracell6OOW
Unit, Danbury, CT) was applied for
10
min to facilitate the dis-
persion of the powder agglomerates, the suspension was stirred
using a magnetic stirrer for over 12
h
at room temperature.
Degassing of the suspension was achieved in a bell jar con-
nected to a vacuum pump. A colloidal consolidation technique
employing a gypsum mold was used to consolidate the colloidal
particles. The compacts were then dried overnight at
100°C.
Thermogravimetric analysis (TGA; TGA7 thermogravi-
metric analyzer, Perkin-Elmer, Norwalk, CT) was conducted to
determine the oxidation level of Sic by weight increase. The
dried compacts were put into a platinum pan and heated to pre-
determined
soak
temperatures at a heating rate of 10"C/min in a
stream of air.
After partial oxidation treatment in air, reaction sintering was
conducted in an alumina crucible in a stream
of
Ar using a
graphite furnace at a heating rate of 25"C/min and a cooling rate
of
5"C/min. The densities of the green compacts and the sin-
tered bodies were measured by the Archimedes method using
kerosene or distilled water, respectively. The pore channel size
distribution of the compacts was determined by mercury poro-
simetry.''," The pore channel size distribution was obtained
using standard values for the mercury surface energy (0.48
N/m) and the contact angle (140"). Phase analysis was con-
ducted by X-ray diffraction
(XRD;
X-ray diffractometer, Phil-
ips Electronic Instruments, Inc, Mahwah, NJ) using Ni-filtered
CuKol radiation. Sintered samples were polished down to a
1-km surface finish with diamond paste and then thermally
etched at 1450°C for
20
min in an
Ar
atmosphere. The resulting
microstructures were evaluated by scanning electron micros-
copy (SEM; Philips 515 scanning electron microscope) and
transmission electron microscopy (TEM; Philips
300
transmis-
sion electron microscope).
111.
Results and
Discussion
(I)
Consolidation
Process
In preparing colloidal suspensions, controlling the interac-
tions between particles has a significant influence on the stabil-
ity of a suspension. In our system, an electrosteric stabilization
approach was preferred over an electrostatic one since it was
not possible to disperse both Sic and A1,0, equally well at the
same pH level. When only electrostatic dispersion was used,
SIC dispersed best under basic conditions since the zeta poten-
tial is higher in basic solutions, as seen in Fig.
3,
whereas A1,0,
dispersed best under acidic
condition^.'^-'^
Consequently, an
NH, salt of PMAA was used as an electrosteric stabilizer (0.4
g/m2)14.15 to improve the stability of A1,0, under basic condi-
tions
so
that a low-viscosity composite suspension could be pre-
pared at pH 10.
Figure 4 shows pore channel size distributions
of
the A1,0,-
15SiC(C) compacts consolidated by the colloidal consolidation
of the suspensions with different values of pH. A green compact
with a narrow pore channel size distribution with small pores
could be obtained by adjusting the pH.
During colloidal consolidation of binary suspensions, a key
problem is the segregation
of
particles due to either gravita-
tional"
"
l6
or
thermodynamic phase separationI7 effects. The
best solution
for
minimizing particle segregation is to prepare
the suspensions as highly concentrated as possible."
''
In
our system, to check if segregation and/or phase separation
occurred, the green densities and X-ray intensity ratios along
the perpendicular axis
of
both Al20,-15SiC(C) and A1,0,-
15SiC(F) compacts (approximately
3.5
cm height) were
measured. Because these displayed similar values within
experimental error, it was concluded that significant segrega-
tion and/or phase separation did not occur while using a 45
vol% solid suspension. As shown in Figs.
5
and
6,
relatively
Fig.
3.
Plots
of
zeta potential
vs
pH
for
SiC(C)
and
SiC(F).
February
1995
Processing
oj
Silicon Carbide-Mullite-Alumina Nanocomposites 48
1
I
0.2
k
2
'3s
2.
f
0.1
a
g
'i3
5
0.0
.I
1
.....
.1
Pore
Channel
Size
urn)
Fig. 4.
Pore channel size distributions
of
the Al,O,-lSSiC(C) com-
pacts consolidated by
the
colloidal consolidation of
the
suspensions
with different values of pH.
I
A
I2
0
3
-S
iC(
F)
A1203-SiC(C)
50
60
70
Green Density
(%)
Fig.
5.
with various amounts
of
SiC(C) and SiC(F) powders.
Green densities
of
the colloidally consolidated composites
high green densities (Fig.
5)
and narrow pore channel size dis-
tributions (Fig. 6) obtained with all of these compacts were also
an indication of the consolidation of the particles without sig-
nificant segregation.
-c-
AI20315SiC(F)
-
A1203-15SiC(C)
----t
A1203-30SiC(C)
-
Al20*5SiC(C)
0
2.
0.0
.Ol
.1
1
Pore
Channel Size
Wrn)
Fig.
6.
composites with various amounts
of
SiC(C) and SiC(F) powders.
Pore channel size distributions of the colloidally consolidated
(2)
Oxidation
Process
The typical weight losses andor gains when heating A1,0,-
15SiC(F) and AI,O,-lSSiC(C) at a heating rate of 10"C/min are
shown in Fig. 7. Upon heating to 700°C weight
loss
was
observed in three regions for the Al,O,-lSSiC(C) sample: in
the first region (up to 300"C), the second (around 350"C), and
the third (around 650°C).
To
determine the origin of the weight
loss,
a TGA experiment on the AI,O, compact after colloidal
consolidation was conducted. By comparison with the weight
loss
of AI,O3-15SiC(C) and A1,0,, it is concluded that the first
weight loss was due to desorption
of
water and the second is
due to decomposition of the surfactant. Although slight weight
loss of A1,0, was observed due to the chemically bonded
water (i.e., hydroxy groups) in the temperature range of
400-
lOOO"C, the weight
loss
was different from that of the third
region. Temperature-programmed desorption measurementZ",2'
of SiC(C) powder in oxygen atmosphere was conducted, where
the evolved gases were monitored with a quadrupole spectrom-
eter. The evolution of CO and CO, was observed in the third
region. Therefore, in the third region the weight
loss
was
mainly due to the evolution of carbon oxide from carbon impu-
rity. In the case
of
A1,O3-15SiC(F), the third region was not
observed because carbon was not a major impurity in the
SiC(F). Above 700°C for A1,0,-15SiC(C) and above 500°C for
A1,O3-15SiC(F), the weight
loss
reached a maximum value
l.e.2
{
Y.
I
A120
3-15SiC(c)
99.2
I
I I
I
I I
Temperaturn
('C)
I
1
I
Jo
so.
7.e
900
in
Fig.
7.
TGA curves
of
colloidal consolidated compacts of Al,O,-lSSiC(C)
and
Al,O,-lSSiC(F) at a heating rate of 10"C/min
in
a stream of air.
482
Journal
of
the American Ceramic Society-Sakka et al.
Vol.
78,
No.
2
and, because
of
the onset of Sic oxidation,
a
weight gain was
observed at higher temperatures. The plateau value at the maxi-
mum weight loss was used as the reference point to approxi-
mate the fraction
of
silica that formed as an oxidation product
on the surface of the Sic particles.
The size
of
the Sic core particles was controlled by
determining the fraction of Sic oxidized during heat treatment.
Figure
8
shows the weight gain and oxidation fraction. Because
of its finer particle size, the oxidation fraction of the SiC(F) sys-
tem is higher than that of the SiC(C) system. The oxidation
behavior of Sic powder in this temperature range is believed to
be the following passive oxidation reaction:”
(1)
The oxidation reactions at 1400°C are depressed by comparison
Sic
+
20,
+
SiO,
+
CO,
1.04
1.03
I
c)
P
B
.g
1.02
1.01
g
e
a
-0.6
v
th
-102
1.00r
’
10.0
0
200 400
600
800
Time
(min)
(a)
with those at 1300”C, especially for the SiC(C) system, as seen
in Fig.
8.
This phenomenon may be due to the significant sinter-
ing occurring simultaneously with the oxidation reaction at
1400°C. Therefore, partial oxidation treatments were conducted
below 1400°C. Many authors have reported that oxidation of
Sic powders in various oxidizing atmospheres follows para-
bolic reaction rate
kinetic^.^^^^'
The oxidation fraction can be
represented by the following Jander’s equatiod8
where
f
is the oxidation fraction,
t
is the reaction time, and
k
is
the rate constant.
As
shown in Fig.
9,
when
1
-
(1
-
f)”’
is
plotted versus the square root of time, a linear variation is
observed at low temperatures (1000-1200°C) and also within
the initial stages of higher temperature oxidation treatments,
1
1
3
1
1
c)
$1
*
B
1
1
.05
.04
-03
.02
.o
1
.oo
0
200
400
600
800
Time
(min)
(b)
Fig.
8.
during isothermal holding in a stream of air.
Weight increase ratio (left-hand side) and oxidation fraction (right-hand side)
of
(a) A1,0,-15SiC(C) and (b) AI,O,-lSSiC(F) composites
0.6
0.6
8
m88.8
0000
0.5
0.5
0.4
0.4
m
\
-
%
r
I
F
r
m
n
Y-
1
I
0.3
I
W
I
0.3
c-
W
7
0.2
0.1
0.2
0.1
0.0
0.0
20 30
0
10
20
30
0
10
dt
(rninllz)
dt
(minl/z)
(a)
(b)
Fig.
9.
Relationship between square root
of
time and
1
-
(1
-
f)”’
for
(a) A1,O3-I5SiC(C) and (b) AI,O,-ISSiC(F) composites during isothermal
holding in a stream
of
air.
suggesting that oxidation is rate limited by a diffusion process.
lo-'
f
Using
Eq.
(2), we can calculate the rate constants at every tem-
perature. Arrhenius plots of the rate constants in the tempera-
ture range of 1000-1300°C are shown in Fig.
10.
The activation
energies of AI,O,-SiC(F) and AI20,-l5SiC(C) are 256 and 216
1
o-'
kJ/mol,
respectively. Activation energies reported in the litera-
ture vary from 134 to 498 kJ/m01.*~-~~ The large amount of scat-
(powder, polycrystal, and single crystal) with varying concen-
trations of impurities, which can alter the oxidation kinetics sig-
nificantly. In our case, the activation energies are close
to
the
-
rl
*c
ter has been attributed to the different types of materials
Y
Y
1
o-~
activation energy of the ionic oxygen diffusion in vitreous silica
(298 kJ/m01)'~ but not to that of the molecular oxygen diffusion
(1
13
kJ/mol).,' Therefore, the present oxidation reaction seems
to proceed via ionic oxygen diffusion through a silica film.
AI20~15SiC(F)
Al2Os-lSSiC(C)
10-8
Q.05
{
0
.05
m
5;90
(2)
Oxidation
at
13OOOC
for
5
h
in air
0,
-
1
d
4.65
3.28
{
II
;
1
I..,,.,,,,
5.80.
5.88
m
.
N
(1)
As
colloidal
4.05
3.20'
m
2.45
1.80
1.25'
2
a.e0'
e.a5.
0.20.
8.85.
'
.
P
I
0.45.
0.05.
15
20
25
30
35
40
45
50 15
20
25
30
35
40 45
50
20
(degree) 28
(degree)
Fig.
11.
5
h
in air, and (3) after reaction sintering at
1600°C
for
2
h
in Ar.
X-ray diffraction patterns of (a) A1,03-15SiC(C) and
(b)
AI2O,-15SiC(F):
(1)
as colloidal consolidation,
(2)
after oxidation at 1300°C for