Full text of DOI:10.1111/j.1151-2916.2001.tb00601.x

J. Am. Ceram. Soc., 84 [1] 13–22 (2001)
journal
Microstructural Control of a 70% Silicon Nitride–
30% Barium Aluminum Silicate Self-Reinforced Composite
Feng Yu,* Nandakumar Nagarajan,* Yi Fang,* and Kenneth W. White*
Department of Mechanical Engineering, University of Houston, Houston, Texas 77204–4792
The processing response of a 70% silicon nitride–30% barium
aluminum silicate (70%-Si3N4–30%-BAS) ceramic-matrix
composite was studied using pressureless sintering, at temper-
atures ranging from 1740°C, which is below the melting point
of BAS, to 1950°C. The relationship between the processing
parameters and the microstructural constituents, such as
morphology of the ␤-Si3N4 whisker and crystallization of the
BAS matrix, was evaluated. The mechanical response of this
array of microstructures was characterized for flexural
strength, as well as fracture behavior, at test temperatures up
to 1300°C. The indentation method was used to estimate the
fracture resistance, and R-curves were obtained from modified
compact-tension samples of selected microstructures at room
temperature.
factors usually are considered when selecting a sintering aid:
sinterability, refractoriness of the secondary phase, and ease of
crystallization of the additives. Unfortunately, the use of more-
refractory sintering additives normally precludes pressureless-
sintering methods, because the associated high viscosity of the
liquid phase restricts both densification and the ␣-Si3N4–␤-Si3N4
phase transformation. Excessive sintering temperatures also must
be avoided, to minimize the decomposition of Si3N4.
Barium aluminum silicate (BAS), which is based on the
composition BaO⅐Al2O3⅐2SiO2, has been studied for decades, in
both synthetic and naturally occurring forms.12–17 The polymor-
phism of stoichiometric BAS at atmospheric pressure is repre-
sented graphically in Fig. 1. At temperatures less than the melting
temperature of BAS (1760°C), BAS first undergoes a solid-state
phase transformation at 1590°C, where the stable hexagonal BAS
(hexacelsian) transforms to the monoclinic phase (celsian), which
remains thermodynamically stable through room temperature.
However, this hexagonal-to-monoclinic reconstructive transforma-
tion is extremely sluggish, which causes hexagonal BAS to persist
in a metastable state at temperatures of Ͻ1590°C. When under-
cooled, the metastable hexacelsian will transform rapidly to an
orthorhombic phase at ϳ300°C. This reversible transformation is
accompanied by a volume change of ϳ3%; this change in volume
usually is destructive.13,14 Therefore, the persistence of the hexa-
celsian phase at temperatures of Ͻ1590°C generally is regarded as
an undesirable feature of the BAS ceramic systems.
I. Introduction
HE strong interest in silicon nitride (Si3N4) ceramics for service
T
temperatures through 1500°C results largely from good high-
temperature strength, good thermal shock resistance, and relatively
good resistance to oxidation, in comparison with other high-
temperature structural materials.1–4 However, Si3N4 components
have high manufacturing costs and poor reliability, which are
major barriers to their extensive use in engineering applications.
High processing pressures often are required for densification, as
well as for promotion of the ␣-Si3N4–␤-Si3N4 phase transforma-
tion. This requirement incurs a manufacturing-cost penalty, be-
cause of the need for specialized equipment, and restricts the
geometric shape of the components. To fabricate complex-shaped
components from dense, high-strength Si3N4 with a minimum of
machining, pressureless sintering has been an interesting area of
study in the Si3N4 research community.1,2,5–7 To apply this
technique in silicon nitride ceramics, the selection of a sintering
aid becomes critical.
Recent work18–22 has shown that ␤-Si3N4 whiskers can be
grown in situ from ␣-Si3N4 in the presence of liquid BAS, which
implies that the liquid phase with
a
composition of
BaO⅐Al2O3⅐2SiO2 has adequate wettability and solubility with the
Si3N4 starting particles. The liquid BAS phase is essential to
ensure both the progress of the ␣-Si3N4–␤-Si3N4 transformation
and the attainment of full densification. The unique combination of
superior refractory properties (the service temperature for BAS can
be as high as 1590°C and the maximum melting point of celsian
BAS is 1760°C), low density, high modulus, low thermal expan-
sion, good oxidation resistance, and low dielectric constant makes
BAS–Si3N4-whisker-reinforced ceramic composites particularly
interesting for high-temperature applications.
Si3N4 is difficult to densify without the use of sintering
additives, primarily because of the highly covalent bond character,
which results in a low self-diffusion coefficient of nitrogen.8
Pressureless densification typically requires a sintering-additive
content of Ͼ15 vol%, compared to the ϳ5–10 vol% of sintering
additives that is required for pressure-assisted densification.1–3 In
addition, eutectic liquids, which form from reactions between
Si3N4 particles and the additives during densification, typically
result in residual amorphous phases after sintering. These residual
amorphous phases often are the phases that adversely influence the
high-temperature performance (such as creep and high-
temperature strength) of the sintered body.1,2,9–11 Therefore, three
Hwang and Newman22 hot-pressed Si3N4 at 1750°–1925°C at a
pressure of 35 MPa, using ϳ10 wt% of BAS as a sintering
additive. The as-sintered material had an average flexural strength
of 817 MPa at room temperature. Its fracture toughness (KIC), as
measured from the chevron-notched bend-bar specimen, was
I-W. Chen—contributing editor
Manuscript No. 190128. Received June 3, 1998; approved December 20, 1999.
This work was generously supported by the Texas Higher Education Board
Advanced Technology Program, under Grant No. 03652-919.
Presented, in part, at the 99th Annual Meeting of the American Ceramic Society,
Cincinnati, OH, May 6, 1997 (Paper No. C-038-97).
Fig. 1. Schematic representation of the possible phase transitions in the
BAS system under atmospheric conditions.
*Member, American Ceramic Society.
13

14
Journal of the American Ceramic Society—Yu et al.
Vol. 84, No. 1
reported to be 5–6.9 MPa⅐m1/2 at room temperature. BAS showed
good crystallization capability, by crystallizing to hexacelsian
during cooling, which contributed to the high-temperature strength
of the as-sintered material. Ceramic-matrix composites with com-
positions of Ͼ20 vol% of BAS as matrix material can be processed
to full density without high pressure.18,20,21 Richardson et al.20
densified Si3N4 with 30 vol% BAS via pressureless sintering and
observed an average flexural strength of 565 MPa and an average
KIC of 5.74 MPa⅐m1/2 at room temperature.
II. Experimental Procedure
(1) Processing
Seventy volume percent of Si3N4 powder (E-10, UBE Indus-
tries, Yamaguchi, Japan) was mixed with 30 vol% of BAS
constituent powders (analytical reagent grade BaCO3, from
Mallinckrodt, Inc., Paris, KY; SM8 Al2O3, from Baikowski Interna-
tional Corp., Charlotte, NC; and 2034DI SiO2, from Nyacol Products,
Inc., Ashland, WA) in isopropyl alcohol for 48 h in a ball mill, using
Si3N4 grinding media. The 30-vol%-BAS–70-vol%-Si3N4 composi-
tion, which has been selected for the present study, corresponds to that
which was developed previously,20 for ease of sinterability and
acceptable mechanical and electromagnetic performance. Green pel-
lets were compacted at a pressure of 50 MPa and then packed in
graphite crucibles with a Si3N4-based powder bed. Samples were
sintered at temperatures of 1740°, 1770°, 1820°, 1870°, 1920°, and
1950°C, each for 30 min in a nitrogen atmosphere. Higher sintering
temperatures were not attempted, because of unacceptable surface
degradation. Sintering times of 5, 30, 60, 120, and 240 min also were
applied to the samples that were processed at 1920°C.
Because of the destructive character of the phase transformation
at 300°C, care should be exercised to avoid hexacelsian in these
composites. However, all the above-mentioned studies of Si3N4–
BAS systems reported hexacelsian as the major phase. Various
methods that have been used previously in monolithic BAS, such
as hot pressing and seeding with celsian particles, were ineffective
in promoting the formation of celsian.17,19,21,22 In addition, neither
electron diffraction nor X-ray diffraction (XRD) analysis has been
effective for detecting the hexacelsian-to-orthorhombic BAS trans-
formation, because only small changes in the O-atom positions
within a common framework characterize this transformation.23
Using high-temperature XRD techniques, Quander et al.21 moni-
tored the phase transformation in a Si3N4–BAS composite in situ
up to 550°C and observed no significant change in the XRD
spectrum with the test temperature. However, their dilatometry
studies identified the potential existence of this transformation in
Si3N4 with a high BAS content.21,24 The dilatometric indication of
transformation for the 30% Si3N4–70% BAS composite was
significant, whereas that for the 70% Si3N4–30% BAS composite
was almost negligible. The magnitude of the volume change also
decreased significantly as the degree of ␣-Si3N4–␤-Si3N4 phase
transformation increased. All these features indicated that the
presence of more ␤-Si3N4 whiskers had a tendency to preclude the
hexacelsian-to-orthorhombic transformation. Richardson et al.20
conducted thermal fatigue tests on both 30-vol%-Si3N4–70-vol%-
BAS and 70-vol%-Si3N4–30-vol%-BAS composites, between
room temperature and 600°C. They reported no decrease in
strength after thermal cycling with the 70-vol%-Si3N4 composite,
which indicated that either this transformation was suppressed or it
was not destructive anymore.
(2) Microstructural Characterization
The density of the as-sintered samples was measured using the
water-immersion method. Phase characterization via XRD was
performed on the bulk material. Sintered materials were sectioned,
ground, and plasma-etched, using a mixture of carbon tetrafluoride
(CF4) and oxygen, to characterize the whisker morphology. Quan-
titative microstructural analysis was performed on digitized scan-
ning electron microscopy (SEM) images that were obtained at
three magnifications (3400ϫ, 6200ϫ, and 12000ϫ), where the
diameter and length of each ␤-Si3N4 whisker were determined
using commercially available image analysis software (NIH Im-
age, National Institute of Health, Washington, DC). This method
defines the grain length and diameter as the maximum and
minimum grain projection, respectively. Some samples in our
study contained residual ␣-Si3N4, which was present in a primarily
spherical shape; therefore, we have defined whiskers as being
particles with an aspect ratio of Ͼ4. The diameter/length and the
apparent aspect ratio of the whiskers were determined statistically,
using the method proposed by Braue et al.25 and Mitomo and
co-workers.26,27 At least 1200 whiskers were measured for the
samples that were processed under each condition. A transmission
The economic incentives that are associated with the near-net-
shape production of pressureless-sintered components provide the
basis for this study. This study intends to present our examination
of the relationship between the pressureless-sintering conditions,
the microstructure, and the resulting mechanical properties.
Fig. 2. Weight percentage of ␣-Si3N4–␤-Si3N4 phase transformation (␤/(␣ϩ␤)) and relative density of as-sintered samples, each as a function of sintering
temperature (for a constant sintering time of 30 min).

January 2001
Microstructural Control of a 70%-Si3N4–30%-BAS Self-Reinforced Composite
15
electron microscopy (TEM) system (JEM-2000FX, JEOL, Tokyo,
Japan) that was fitted for energy-dispersive X-ray spectroscopy (EDS)
was used to conduct microstructural observations. High-resolution
transmission electron microscopy (HRTEM) images also were ob-
tained, to explore the potential presence of amorphous phases.
temperatures, then later crystallizing to a more-refractory com-
pound.
With the extension of sintering time at 1920°C, the density and
the phase transformation exhibit behavior that is similar to that
previously described. A relative density of 97% of TD, with ϳ68%
of the ␣-Si3N4 being transformed to ␤-Si3N4, was achieved after
only 5 min at 1920°C. When the sintering time was extended to
120 min, no residual ␣-Si3N4 phase could be detected via XRD in
the as-sintered material.
During the liquid-phase sintering of Si3N4, the material under-
goes particle rearrangement, solution–diffusion–precipitation, and
grain growth, assuming that the liquid phase provides good
wettability and solubility for Si3N4. The solution–reprecipitation
transformation of Si3N4 may either help or hinder the densifica-
tion, depending on whether a skeleton of whisker forms.30 In the
present case, a relative density of Ͼ90% of TD was attained, with
only ϳ35% ␣-Si3N4 transformed (Fig. 2), which illustrates its
(3) Mechanical Behavior
The flexural strength of the material that was processed under
each condition was obtained in three-point bending, at test tem-
peratures up to 1300°C, in laboratory air. These specimens were
cut from the as-sintered pellets in dimensions of 3 mm ϫ 1.5
mm ϫ 30 mm. The surfaces of the specimens were ground using
a 400-grit diamond wheel. At least seven specimens were tested
under each test condition.
Indentations were made on the broken flexural-strength-testing
bars, in ambient air, using a universal testing machine (Instron,
Danvers, MA) that was fitted with a diamond pyramid indenter.
The contact loads varied over a range of 35–45 kg, and the
peak-load contact time was 90 s. At least twelve valid impressions
on each selected sample were used to estimate the fracture
toughness, according to the approach devised by Anstis et al.28. A
Young’s modulus of 242 GPa was measured using ultrasonic
methods.
The R-curve behavior of the samples that were processed at
1920°C for 30, 120, and 240 min was determined from modified
compact-tension samples with dimensions of 17 mm ϫ 30 mm ϫ
3.5 mm at room temperature. Actual crack extension during
loading was calculated using both boundary-element and finite-
element analyses. The final R-curve results were the average of at
least two compact-tension specimens that had been tested under
the same conditions.
III. Results and Discussion
(1) Density Measurement and Phase Characterization
The weight loss of the as-sintered samples remained at Ͻ2% for
sintering temperatures up to 1920°C. The 1950°C/30-min samples
had glassy surface layers with an average thickness of ϳ0.3 mm
and experienced a weight loss of at least 4%. This surface
condition seemed to be more sensitive to sintering temperature
than to sintering time. In comparison, the 1920°C/240-min sample
experienced a weight loss of ϳ2%, with a smooth, glassy surface
layer that was ϳ0.2 mm thick.
The 1740°C/30-min sample had a measured density of 3.14
g/cm3, which was the lowest density measured in our study. The
relative density of each sample was obtained by comparison to a
calculated theoretical density (TD) of 3.25 g/cm3, as shown in Fig.
2.
The progress of the ␣-Si3N4–␤-Si3N4 phase transformation,
relative to the processing temperature, which was calculated in a
manner that was described by Gazzara and Messier,29 also is
presented in Fig. 2. The phase transformation became detectable at
temperatures less than the melting temperature of BAS. Complete
transformation was detected via XRD in the samples that were
processed at 1950°C for 30 min. The liquid eutectic temperature of
the BaO–Al2O3–SiO2 system is in the range of 1175°–1760°C,
depending on the composition.13,14 In Si3N4–BAS composites,
extensive BAS phase formation via solid reaction among three
constituent powders is observed at temperatures as low as
1300°C.21 However, the distribution of BAS constituent powders
may not be stoichiometrically perfect, which can result in early
local melting (Ͻ1760°C). For example, reaction between the
residual silica (SiO2) layer on the surface of the ␣-Si3N4 starting
powders, which has been identified using both HRTEM observa-
tion and EDS methods, and the constituent powders is difficult in
the low-processing-temperature range. However, at higher pro-
cessing temperatures, this surface silica could result in a SiO2-rich
area and form a liquid layer that surrounds the ␣-Si3N4 particles at
temperatures much less than that of eutectic BAS. This phenom-
enon encourages early ␣-Si3N4–␤-Si3N4 phase transformation.
Therefore, in this material system, the use of BAS constituent
powders instead of BAS exhibits an advantage of initially assisting
pressureless sintering by forming a liquid phase at relatively low
Fig. 3. SEM micrographs showing the whisker morphology of the (a)
1920°C/30-min, (b) 1920°C/60-min, and (c) 1920°C/240-min samples.

16
Journal of the American Ceramic Society—Yu et al.
Vol. 84, No. 1
minimal contribution to densification. To obtain good strength and
high toughness from silicon nitride ceramics, complete densifica-
tion and phase transformation should be achieved. Therefore, the
sintering progress of this composite is controlled by the phase-
transformation rate.
Only the hexagonal allotrope of the BAS matrix was identified
via XRD. However, neither electron diffraction nor XRD can
conclusively exclude the existence of orthorhombic BAS in this
composite. We assume that the absence of accommodation rea-
sonably precludes the presence of this low-temperature phase,
which is consistent with the reports in the literature.18–22 No
microcracks, which normally accompany the hexagonal-to-
orthorhombic transformation, were observed using either TEM or
SEM methods.
that the material contains no pores, with ␤-Si3N4 whiskers oriented
randomly in a fine, continuous matrix of BAS. Isotropic mechan-
ical performance is expected from this type of microstructure.
Figures 4(a) and (b) respectively illustrate the influence of
sintering time and temperature on the average whisker length and
average apparent aspect ratio. At 1920°C, the average whisker
length increased from ϳ2 ␮m after 5 min to ϳ5 ␮m after 240 min.
However, changes in the aspect ratio are much more subtle,
relative to sintering time; the average apparent aspect ratio of most
samples is in the range of 17–18.5. The rapid increase of both the
length and aspect ratio at 1740°–1770°C most likely results from
the large-scale melting of the BAS matrix and phase transforma-
tion in this temperature range.
The morphology of the ␤-Si3N4 whiskers in the present study
closely resembles that of conventional hot-pressed silicon nitride
ceramics, except for the absence of very large whiskers (Ͼ15 ␮m).
Statistical analyses suggest a normal distribution of whisker
width/length for all samples, with no indication of significant
bimodal whisker growth, as shown in Fig. 5. Grain coarsening is
evident with the extension of sintering time (Fig. 5); however, the
(2) Whisker Morphology
Figure 3 demonstrates the typical microstructures and, in
particular, the development of whisker morphology, relative to
sintering time. SEM observation of the as-sintered samples showed
Fig. 4. Average whisker length and apparent aspect ratio, each as a function of (a) sintering time and (b) sintering temperature.

January 2001
Microstructural Control of a 70%-Si3N4–30%-BAS Self-Reinforced Composite
17
Fig. 5. Distribution of whisker width for the 1920°C/30-min, 1920°C/60-min, and 1920°C/210-min samples.
rate of diameter increase is small. Longer sintering times promote
growth of all whiskers, which shifts and broadens the entire
distribution curve.
may surround several whiskers. Neither cavities nor interface debond-
ing that was associated with the crystallization, as reported for a
silicon nitride ceramic with a crystallized YSiAlON matrix,7 was
observed via TEM.
Some authors31–33 reported that bimodal microstructures, which
have abnormally grown grains in a fine and uniform matrix, can be
obtained in silicon nitride ceramics using a seeding method with
the assistance of a high processing pressure. The presence of a
certain amount of abnormally grown grains will promote com-
bined improvements in both toughness and strength.33 Emoto and
Mitomo31 proposed that, after completion of the phase transfor-
mation, the driving force for ␤-Si3N4 grain growth scaled with the
difference in particle size, and abnormal grain growth occurred
only when the difference of particle size exceeded a critical value.
In addition, the abnormal grain growth is assumed to be acceler-
ated by the ␣-Si3N4–␤-Si3N4 transformation.34–36 Normally, this
transformation is enhanced with increasing processing pressure, as
noted by Vincenzini and Babini37 in an earlier paper. In our study,
the rate of ␣-Si3N4–␤-Si3N4 phase transformation is comparably
slower, because of the high viscosity of the BAS liquid, as well as
our choice of the pressureless-sintering method. Therefore, the
grain-size distribution remains relatively narrow, in comparison
with that of hot-pressed silicon nitride materials, which predictably
suppresses abnormal grain growth in a steady state.
(3) Crystallization of the Barium Aluminosilicate Matrix
As noted previously, the BAS glass-ceramic serves as a liquid-
phase-sintering aid for the ␣-Si3N4–␤-Si3N4 phase transformation
and remains as a structural matrix. Volume fractions as high as
30% are used, in contrast to sintering-aid contents of typically
Ͻ10% in conventional hot-pressed silicon nitride ceramics; there-
fore, a pronounced influence on the composite properties can be
expected.
Compared with other silicate glasses, BAS glass is unusually
refractory, with a dilatometric softening point of 925°C.15 How-
ever, for high-temperature applications, crystallization to the
crystalline BAS phase offers additional improvement. TEM methods
identified almost-complete crystallization of the BAS matrix in all
samples. Figure 6(a) is a bright-field TEM image that shows the
microstructure of ␤-Si3N4 whiskers and a BAS matrix of a 1920°C/
120-min sample. Figure 6(b) is the corresponding dark-field image,
៮
៮ ៮
obtained using a [1101] diffracted beam from the BAS [1213]
diffraction pattern, as indicated in the inset in the upper left-hand
corner. The BAS regions that have the same orientation are shown in
bright contrast in Fig. 6(b). Further illustrating the completeness of
crystallization, BAS pockets, almost without exception, fully crystal-
lized in the “corners” that are formed by whiskers (see Figs. 6 and 7).
The TEM dark-field image in Fig. 6(b) also shows that one BAS grain
Fig. 6. TEM images illustrating the crystallization of BAS in the
1920°C/120-min sample ((a) bright field and (b) dark field).

18
Journal of the American Ceramic Society—Yu et al.
Vol. 84, No. 1
Fig. 7. HRTEM image showing the triple point of two whiskers (marked as “W”) and one BAS grain (marked as “BAS”) in the 1920°C/240-min sample.
Fig. 8. Vickers’s hardness, as a function of sintering time.
The use of materials at higher temperatures requires a refractory
matrix composition, as well as a microstructure that does not
soften because of the presence of residual glassy phases. The
almost-complete crystallization of the BAS matrix in this composite
illustrates an important advantage for its high-temperature applica-
tions. Because the crystallization is spontaneous, no additional heat
treatment (such as the extensive post-sintering annealing time re-
quired to obtain crystallized grain-boundary phases in other silicon
nitride systems) is needed.1,3,5–7 The excellent crystallization capa-
bility of BAS liquid also has been reported elsewhere.18–22
Using HRTEM methods, some residual glassy phase was
observed, with no exception in any samples, predominantly at
triple-grain junctions of whiskers and BAS grains (see Fig. 7).
Also, an interlayer of glassy phase (ϳ1 nm thick) persisted
between the whiskers. Within the resolution capability of HRTEM,
the thickness was almost constant. Also, different sintering proce-
dures did not seem to change the characteristics of this thin film.
microstructure. This value is somewhat less than that of conven-
tional silicon nitride materials, which usually is 17–20 GPa,
probably because of the 30 vol% BAS content. We have attributed
the decreasing hardness with sintering time (see Fig. 8) to the
progressive replacement of ␣-Si3N4 with the softer and larger
␤-Si3N4 grains.38,39 After completion of the ␣-Si3N4–␤-Si3N4
phase transformation, the decreased softening rate is governed
only by changes in grain size.
At a fixed sintering temperature of 1920°C, the room-
temperature flexural strength peaked after ϳ2 h of sintering (see
Fig. 9). However, the fracture toughness (KIC) continued to
increase as the sintering time increased. The improvements in the
fraction of ␣-Si3N4–␤-Si3N4 phase transformation and the whisker
morphology that result from increasing the sintering time from 5
min to 120 min were believed to contribute to the increases in both
strength and toughness. The highest flexural strength of 962 Ϯ 70
MPa, which was obtained from the 1920°C/120-min samples,
corresponded to an average KIC value of 5.4 MPa⅐m1/2. This
combination of properties agrees with that of conventional hot-
pressed silicon nitride ceramics. The combined features of a dense,
fine, and homogeneous structure, along with the in situ formation
(4) Mechanical Characterization
The Vickers’s hardness, which was in the range of 14.9–16.2
GPa, changed with differences in the phase composition and

January 2001
Microstructural Control of a 70%-Si3N4–30%-BAS Self-Reinforced Composite
19
Fig. 9. Flexural strength and fracture toughness (KIC), each as a function of sintering time (for a constant sintering temperature of 1920°C).
Fig. 10. Flexural strength of the 1770°C/30-min, 1870°C/30-min, 1920°C/30-min, and 1950°C/30-min samples, as a function of test temperature.
of a hexacelsian matrix, make Si3N4–BAS materials rather unique.
When the coarser microstructure was obtained after sintering at
1920°C in 240 min, the KIC value increased to 6.2 Ϯ 0.1 MPa⅐m1/2
but the flexural strength decreased considerably, to 852 Ϯ 5 MPa.
An important feature that occurs in all silicon nitride ceramics
at elevated temperatures is a degradation in strength, which
generally has been recognized as a result of subcritical crack
growth or creep. The residual glassy phase at the grain boundary
and the triple junctions of grains can soften when the environmen-
tal temperature is greater than its softening temperature, which
results in grain-boundary sliding, separation, void formation, and
cracking. Both the amount and viscosity of the glassy phase are
important to the high-temperature performance of silicon nitride
ceramics.
retain room-temperature strength to 600°C; the strength begins to
decrease at temperatures from 600°C to 920°C. The strength
decrease in this temperature range is believed to share a common
origin with conventional silicon nitride materials. The inconsis-
tency between this temperature range and the softening point of
BAS glass implies that the composition of the residual glassy
phase is different from that of stoichiometric BAS.
The flexural-strength–test-temperature curves of the 1870°C/
30-min, 1920°C/30-min, and 1950°C/30-min samples all show
characteristic plateaus in the temperature range of 920°–1120°C.
For the 1770°C/30-min sample, this plateau occurs at a higher
temperature range. Similar behavior has been reported in several
silicon nitride ceramics with a crystallized matrix phase, which
generally is considered to be the main contributor to this phenom-
enon.3,5–7,19–22,40 In this case, the crystallized BAS matrix is
assumed to restrict the further softening of the residual glassy
phase with increases in temperature. The flexural strength of the
Figure 10 shows the trend of flexural strength with test
temperature for samples that have been sintered at different
temperatures with a constant sintering time of 30 min. All samples

20
Journal of the American Ceramic Society—Yu et al.
Vol. 84, No. 1
1870°C/30-min, 1920°C/30-min, and 1950°C/30-min samples de-
creases considerably at Ͼ1120°C. At 1300°C, no significant
difference in strength is observed among these three samples.
When the test temperature is Ͻ900°C, the load–displacement
curve of each sample that is obtained from the bending test is
linear, which characterizes brittle behavior. When the test temper-
ature is Ͼ900°C, however, some samples exhibit plastic deforma-
tion before failure. At 1120°C, the 1770°C/30-min sample still
maintains brittle behavior. However, the load–displacement
curves of the 1920°C/30-min and 1950°C/30-min samples exhibit
both linear and nonlinear regions.
indication of the softening conditions, as illustrated in Fig. 11.
Generally, a high sintering temperature results in an early decrease
in the proportional limit from unity. At the same test temperature,
the proportional limit normally decreases as the sintering temper-
ature increases. This trend agrees well with the strength–test-
temperature behavior that has been discussed previously. At
1300°C, the 1950°C/30-min samples have the lowest proportional
limit, with an average value of ϳ0.38. This low value indicates
that the sample underwent a significant amount of plastic defor-
mation and also implies that more residual glassy phase that
accommodates the plastic deformation may be present.
The proportional limit—i.e., the ratio between the maximum
load within linear portion and the failure load—offers another
The flexural strength of samples that have been sintered at
1920°C but for various times exhibits behavior that is similar to the
Fig. 11. Proportional limit calculated from the load–displacement curve of samples sintered at different temperatures for constant time, as a function of
test temperature.
Fig. 12. Flexural strength of the 1920°C/5-min, 1920°C/60-min, 1920°C/120-min, and 1920°C/240-min samples, as a function of test temperature.

January 2001
Microstructural Control of a 70%-Si3N4–30%-BAS Self-Reinforced Composite
21
increase of test temperature (see Fig. 12). However, the starting
temperature for the first significant decrease in strength varies. For
the 1920°C/60-min sample, a significant decrease in strength
occurs at 840°–920°C, which is close to the softening temperature
of BAS glass. When sintering time was extended to 120 min, the
strength decrease shifted to the temperature range of 740°–840°C.
For the 1920°C/240-min sample, a large decrease in strength
occurs even earlier, at 600°–740°C. In the temperature range of
920°–1120°C, all samples that have been sintered at 1920°C
maintain a flexural strength of ϳ77%–80% of that at room
temperature, except the 1920°C/5-min sample, which retains
ϳ88% of the room-temperature flexural strength. The flexural
strength of 1920°C/120-min sample, which is the strongest among
all samples at room temperature, was 782 Ϯ 64 MPa in this
temperature range. This value is higher than that of most silicon
nitride ceramics with a crystallized matrix that have been pro-
cessed either with or without pressure.
Several authors19,22 indicated that the ␤-Si3N4 component in the
Si3N4–BAS system was, in fact, a ␤-SiAlON. The amount of
aluminum substitution in the ␤-SiAlON in this system could be in
the range of 0.76–4.44 equiv.%, depending on the amount of
alumina (Al2O3) that was used in the material preparation and
processing conditions. Higher sintering temperatures and longer
processing times can encourage the dissolution of aluminum into
the ␤-Si3N4 lattice, which drives the chemistry of the remaining
liquid further away from the BAS eutectic composition. Hwang
and Newman22 used quantitative phase analyses to compare the
amount of crystallized BAS matrix after different processing
conditions; they reported less hexacelsian in samples that were
processed at higher temperatures or longer times. Therefore,
higher sintering temperatures and longer processing times seemed
to hinder complete crystallization of the BAS liquid.
Because of the limitation of the instrument, the chemistry of the
glassy phase could not be obtained in our study. However, TEM
observation has revealed the characteristic core structure of a
␤-Si3N4 grain, as shown in Fig. 13(a). Although the contrast of this
dark core typically is weak, it could be identified in most ␤-Si3N4
grains by adjusting the orientation. EDS analysis (Figs. 13(b) and
(c)) illustrates the difference in aluminum content between the core
and the shell. The core structure, which contains virtually no
aluminum, is believed to be pre-existing ␤-Si3N4 nuclei. The
aluminum content within the shell structure, which may result
from the formation of a SiAlON solid solution, varies with the
location and size of the whiskers. Therefore, the composition of
residual glass may be away from the BAS stoichiometric point and
close to the BaO–SiO2 side of the BaO–Al2O3–SiO2 ternary phase
diagram. The average aluminum content of whiskers of the
1920°C/5-min sample is ϳ8% less than that of similarly sized
whiskers of the 1920°C/240-min sample. The early significant
decrease in strength of the 1920°C/240-min sample implies a
larger departure in the chemistry of the residual glassy phase from
that of BAS, when the samples are sintered under this condition. In
contrast, the 1920°C/5-min samples, which have the shortest
sintering time, suffered only a minimal loss of strength.
R-curve behavior has been characterized only on the 1920°C/
30-min, 1920°C/120-min, and 1920°C/240-min samples at room
temperature, as presented in Fig. 14. The 1920°C/120-min and
1920°C/240-min samples show distinct rising R-curves, relative to
crack extension. The 1920°C/240-min sample has the largest
increase in toughness (KIC ϭ 2.2 MPa⅐m1/2). The KIC value of the
1920°C/120-min sample increases from 4.9 MPa⅐m1/2 to 5.98
MPa⅐m1/2 within a crack extension of 300 ␮m. The R-curve data
agree well with the KIC values that have been determined using the
indentation method. For the 1920°C/30-min sample, however, a
large discrepancy exists between the values that have been
measured using each method.
IV. Conclusions
Composites with a composition of 70 vol% silicon nitride–30
vol% barium aluminum silicate (70 vol% Si3N4–30 vol% BAS)
can be fabricated via pressureless sintering. In this composite, the
BAS glass-ceramic serves as an effective liquid-phase-sintering
aid, to attain full densification and complete the ␣-Si3N4–␤-Si3N4
Fig. 13. (a) TEM micrograph of a ␤-Si3N4 grain with a core structure.
Figures 13(b) and (c) respectively show EDS spectra obtained from the
core and the shell.

22
Journal of the American Ceramic Society—Yu et al.
Vol. 84, No. 1
8K. Kijama and S. Shirasaki, “Nitrogen Self Diffusion in Silicon Nitride,” J. Chem.
Phys., 65, 2668–71 (1976).
9S. H. Knickerbocker, A. Zangvil, and S. D. Brown, “High-Temperature Mechan-
ical Properties and Microstructures for Hot-Pressed Silicon Nitrides with Amorphous
and Crystalline Intergranular Phases,” J. Am. Ceram. Soc., 68 [4] C-99–C-101 (1985).
10S. Dutta and B. Buzek, “Microstructure, Strength, and Oxidation of a 10 wt%
Zyttrite-Si3N4 Ceramic,” J. Am. Ceram. Soc., 67 [2] 89–92 (1984).
11D.-S. Cheong and W. A. Sanders, “High-Temperature Deformation and Micro-
structural Analysis for Silicon Nitride–Scandium(III) Oxide,” J. Am. Ceram. Soc., 75
[12] 3331–36 (1992).
12T. Ito, X-ray Studies on Polymorphism; p. 19. Maruzen Co. Ltd., Tokyo, Japan,
1950.
13H. C. Lin and W. R. Foster, “Studies in the System BaO–Al2O3–SiO2, I. The
Polymorphism of Celsian,” Am. Mineral., 53, 134–44 (1968).
14D. Bahat, “Homogeneous and Heterogeneous Polymorphic Transformations in
Alkaline Earth Feldspars,” J. Mater. Sci., 5, 805 (1970).
15N. P. Bansal and M. J. Hyatt, “Crystallization Kinetics of BaO–Al2O3–SiO2
Glasses,” J. Mater. Res., 4 [5] 1257–65 (1989).
16C. H. Drummond III and N. P. Bansal, “Crystallization Behavior and Properties
of BaO⅐Al2O3⅐2SiO2 Glass Matrices,” Ceram. Eng. Sci. Proc., 11 [7–8] 1072–86
(1990).
17C. H. Drummond III, “Glass Formation and Crystallization in High-Temperature
Glass-Ceramics and Si3N4,” J. Non-Cryst. Solids, 123, 114 (1990).
18F. Yu, C. R. Ortiz-Longo, K. W. White, and D. Hunn, “The Microstructural
Characterization of In Situ Grown Si3N4 Whisker-Reinforced Barium Aluminum
Silicate Ceramic Matrix Composite,” J. Mater. Sci., 34, 2821–35 (1999).
19H. Pickup and R. J. Brook, “Barium Oxide as a Sintering Aid for Silicon Nitride,”
Br. Ceram. Soc. Proc., 39, 69–76 (1987).
Fig. 14. R-curves measured from the 1920°C/30-min, 1920°C/120-min,
and 1920°C/240-min samples.
20K. K. Richardson, D. W. Freitag, and D. Hunn, “Barium Aluminosilicate
Reinforced In Situ with Silicon Nitride,” J. Am. Ceram. Soc., 78 [10] 2662–68 (1995).
21S. W. Quander, A. Bandyopadhyay, and P. B. Aswarth, “Synthesis and Properties
of In Situ Si3N4-Reinforced BaO⅐Al2O3⅐2SiO2 Ceramic Matrix Composites,” J.
Mater. Sci., 32, 2021–29 (1997).
phase transformation, and remains as a structural matrix that is
reinforced by the whiskers. Si3N4 whiskers nucleate and grow in
random directions in an almost completely crystallized matrix of
hexacelsian BAS. Although small amounts of amorphous phase
remain in some grains junctions, the configuration of the interface
between whiskers seems to approach thermodynamic stability,
which suggests little opportunity for improved crystallization of
the grain-boundary glass that is present between whiskers.
High flexural strength (962 Ϯ 70 MPa) can be obtained from
samples that have been sintered at 1920°C for 120 min with a
fine-grained microstructure. The fracture toughness of this mate-
rial is ϳ5.4 MPa⅐m1/2. Rising R-curves are obtained from samples
that have been processed at 1920°C for 120 and 240 min. The
crystallized BAS matrix significantly benefits the high-
temperature strength of this composite. The composite can main-
tain this high strength up to a temperature of 1120°C. At 1300°C,
the composite exhibits a flexural strength of ϳ500 MPa.
22C. J. Hwang and R. A. Newman, “Silicon Nitride Ceramics with Celsian as an
Additive,” J. Mater. Sci., 31, 150–56 (1996).
23Y. Takeuchi, “A Detailed Investigation of the Structure of Hexagonal
BaAl2Si2O8 with Reference to Its ␣–␤ Inversion,” Mineral. J., 2 [5] 311–32 (1958).
24A. Bandyopadhyay, P. B. Aswarth, W. D. Porter, and O. B. Cavin, “The Low
Temperature Hexagonal to Orthorhombic Transformation in Si3N4 Reinforced BAS
Matrix Composite,” J. Mater. Res., 10, 1256–63 (1995).
25W. Braue, G. Wotting, and G. Ziegler, Ceramic Microstructure ’86; p. 250.
Plenum Press, New York, 1987.
26M. Mitomo, M. Tsutsumi, H. Tanaka, S. Uenosono, and F. Saiton, “Grain Growth
During Gas-Pressure Sintering of ␤-Silicon Nitride,” J. Am. Ceram. Soc., 73 [8]
2441–45 (1990).
27M. Mitomo and S. Uenosono, “Microstructural Development during Gas-
Pressure Sintering of ␣-Silicon Nitride,” J. Am. Ceram. Soc., 75 [1] 103–108 (1992).
28G. R. Anstis, P. Chantikul, B. R. Lawn, and D. B. Marshall, “A Critical
Evaluation of Indentation Techniques for Measuring Fracture Toughness: I, Direct
Crack Measurements,” J. Am. Ceram. Soc., 64 [9] 533–38 (1981).
29C. P. Gazzara and D. R. Messier, “Determination of Phase Content of Si3N4 by
X-ray Diffraction Analysis,” Am. Ceram. Soc. Bull., 56 [9] 777–80 (1977).
30N. Nandakumar, “The Liquid Phase Influence on Microstructure Evolution in the
Silicon Nitride–BAS/SAS System”; Master’s Thesis. University of Houston, Hous-
ton, TX, 1999.
31H. Emoto and M. Mitomo, “Control and Characterization of Abnormally Grown
Grains in Silicon Nitride Ceramics,” J. Eur. Ceram. Soc., 17, 797–804 (1997).
32W. Dre␤ler, H.-J. Kleebe, M. J. Hoffmann, M. Ru¨hle, and G. Petzow, “Model
Experiments Concerning Abnormal Grain Growth in Silicon Nitride,” J. Eur. Ceram.
Soc., 16, 3–14 (1996).
Acknowledgment
The authors wish to thank Mr. Nunez for his assistance in the Young’s modulus
measurement.
33E. Sun, P. F. Becher, C.-H. Hsueh, K. P. Plucknett, and S. B. Waters,
“Microstructural Design of Silicon Nitrides with Improved Fracture Toughness”;
presented at the 99th Annual Meeting of American Ceramic Society, Cincinnati, OH,
May 4–7, 1997 (Paper No. SXVI-011-97).
References
34D.-D. Lee, S.-J. L. Kang, G. Petzow, and D. N. Yoon, “Effect of ␣ to ␤(␤Ј) Phase
Transition on the Sintering of Silicon Nitride Ceramics,” J. Am. Ceram. Soc., 73 [3]
767–69 (1990).
1G. Ziegler, J. Heinrich, and G. Wotting, “Review Relationships between Process-
ing, Microstructure and Properties of Dense and Reaction-Bonded Silicon Nitride,” J.
Mater. Sci., 22, 3041–86 (1987).
35D. Suttor and G. S. Fischman, “Densification and Sintering Kinetics in Sintered
Silicon Nitride,” J. Am. Ceram. Soc., 75 [5] 1063–67 (1992).
2F. F. Lange, “Silicon Nitride Polyphase Systems: Fabrication, Microstructure, and
Properties,” Int. Metall. Rev., 1, 1–20 (1980).
36F. F. Lange, “Fracture Toughness of Si3N4 as a Function of the Initial ␣-Phase
Content,” J. Am. Ceram. Soc., 62 [7–8] 428–30 (1979).
3M. J. Hoffmann, “High-Temperature Properties of Si3N4 Ceramics,” MRS Bull.,
10 [2] 28–32 (1995).
37P. Vincenzini and G. N. Babini, “The Influence of Secondary Phase on
Densification, Microstructure and Properties of Hot-Pressed Silicon Nitride”; pp.
425–51 in Sintered Metal-Ceramic Composites. Elsevier Science Publishers B.V.,
Amsterdam, The Netherlands, 1984.
4M. J. Hoffmann, “Analysis of Microstructural Development and Mechanical
Properties of Si3N4 Ceramics”; p. 59 in Tailoring of Mechanical Properties of Si3N4
Ceramics. Edited by M. J. Hoffmann and G. Petzow. Kluwer Academic Publishers,
Dordrecht, The Netherlands, 1994.
38C. Greskovich and G. E. Gazza, “Hardness of Dense ␣ and ␤-Si3N4 Ceramics,”
J. Mater. Sci. Lett., 4, 195–96 (1985).
5T. Hayashi, H. Munakata, H. Suzuki, and H. Saito, “Pressureless Sintering of
Si3N4 with Y2O3 and Al2O3,” J. Mater. Sci., 21, 3501–508 (1986).
6N. Hirosaki, A. Okada, and K. Matoba, “Sintering of Si3N4 with the Addition of
Rare-Earth Oxides,” J. Am. Ceram. Soc., 71 [3] C-144–C-147 (1988).
7M. Cinibulk, G. Thomas, and S. M. Johnson, “Grain-Boundary-Phase Crystalli-
zation and Strength of Silicon Nitride Sintered with a YSiAlON Glass,” J. Am.
Ceram. Soc., 73 [6] 1606–12 (1990).
39R. F. Coe, R. J. Lumby, and M. F. Pawson, Special Ceramics 5; p. 361. Edited
by P. Popper. British Ceramic Research Association, Stoke-on-Trent, U.K., 1972.
40K. Komeya, M. Komatsu, T. Kameda, Y. Goto, and T. Tsuge, “High-Strength
Silicon Nitride Ceramics Obtained by Grain-Boundary Crystallization,” J. Mater.
Sci., 26, 5513–16 (1991).
Ⅺ