J Am. Cwam Sac.

, 78 [91 2345-52 (1995)

Effect of the Grain Boundary Thermal Expansion Coefficient on the

Fracture Toughness in Silicon Nitride
Irene M. Peterson' and Tseng-Ying Tien'
Department of Materials Science and Engineering, The IJniversity of Michigan, Ann Arbor, Michigan 48109

The effect of thermal expansion mismatch stress between and Tien,5Bonnell,6 and Hwang.' The solidpiquid equilibria in
silicon nitride and different grain boundary phases on the the Y,Si,Al/O,N system were determined by Wisnudel* and
fracture toughness of silicon nitride was investigated. Huang and tier^.^ The compositions of the sintering liquid in the
Different sintering aids in the Y-Mg-Si-Al-0-Nsystem pro- Y,Mg,Si,Al/O,N system were determined as part of this study.
duced silicon nitride specimens with very similar micro- The compositions of some of the crystalline grain boundary
structures but different grain boundary phase compositions phase in equilibrium with silicon nitride have been reported.
and different values of fracture toughness. The fracture Subsolidus phase relationships in the Si,Al,Mg/O,N system
toughness of the silicon nitride increased as the thermal were reported by Nunn et a1.I' Both cordierite and N phase are
expansion coefficient of the grain boundary phase compatible with silicon nitride." Subsolidus phase relation-
increased. The presence of tensile residual stress at the grain ships in the Si,AI,Y/O,N system were reported by Naik and
boundary caused by thermal expansion mismatch between Tien" and by Sun et a1.l' Both YAG and Y,Si,O, are compatible
the silicon nitride and the grain boundary phase enhanced with silicon nitride. The thermal expansion coefficients of crys-
crack deflection and grain bridging. tallized Y,Mg,Si,Al/O,N compositions were measured for this
study.
The effect of the grain boundary phase on toughness and
I. Introduction crack propagation was demonstrated by Tanaka et al.," who
densified silicon nitride without additional sintering aids. They
B ECAUSE of its strong covalent bonding and slow diffusion
rates, silicon nitride is very difficult to densify without a
sintering aid. Sintering aids affect the properties of the silicon
reported a fracture toughness of 3.0 MPa.mo5 for their fully
dense material. The addition of 20 vol% Si,N, whiskers raised
nitride ceramic in two ways: by their effect on microstructure the fracture toughness only slightly, to 3.5 MPa.mn5. Their
development and by the properties of the grain boundary phase photo of the crack path revealed nearly transgranular fracture,
formed by the sintering aids. There is a large amount of litera- with little sign of crack bridging, deflection, or pullout by the
ture on the effect of grain diameter and aspect ratio on fracture whiskers. Therefore the whiskers did not contribute much to the
toughness.'-' However, several authors have observed that sili- fracture toughness. Clearly, elongated grains are useful only if
con nitride samples which were sintered with different sintering the fracture path is primarily intergranular.
aids had the same microstructure but drastically different values Bridging grains are formed when the crack is deflected
of fracture toughness." The differences in fracture toughness around the silicon nitride grain, leaving the grain intact. The
were attributed to differences in the grain boundary phase. The formation of bridging grains is affected by the presence of
thermal expansion coefficient, elastic properties and toughness residual stresses in the material. Residual stresses caused by
of the grain boundary phase, and the bond strength between the thermal expansion mismatch between the grain boundary phase
matrix and the grain boundary would be expected to affect the and the silicon nitride interact with the crack tip. For smooth
crack path. Despite the importance of the behavior of the grain grains, tensile stress in the grain boundary phase would be
boundary phase, the effect of the grain boundary phase on crack expected to draw the crack along the weaker grain boundary
morphology and fracture toughness is poorly understood. The phase, enhance debonding, and lead to intergranular f r a c t ~ r e . ' ~
purpose of this study was to determine the effect of residual Compressive stress in the grain boundary phase would be
stress on crack path and fracture toughness in silicon nitride. To expected to limit debonding and lead to transgranular fracture.
accomplish this, sintering aids were used which produced simi- The use of thermal expansion mismatch stresses to improve
lar microstructures, but which produced grain boundary glasses fracture toughness was demonstrated by Taya et al. in a Sic/
TiB, composite.I5Local residual stresses were also observed to
with different thermal expansion coefficients.
affect the crack path in a SiC/glass composite."
In order to determine the thermal expansion coefficients of
The conditions for crack deflection along a fiber were deter-
the grain boundary phases, model glasses with the same compo- mined by Evans and He." The crack will deflect if
sition as the grain boundary phase were prepared. The composi-
tions of the amorphous grain boundary phase can be found GJG, TJT,
from the composition of the sintering liquid. The solid/liquid
equilibria relevant to many of the systems in this study have where G, = the crack energy release rate along the interface,
been determined by other authors. The solidfliquid phase equi- G, = the crack energy release rate through the fiber, T, = the
libria in the Mg,Si,Al/O,N system were determined by Hayashi fracture energy of the interface, and T, = the fracture energy of
the fiber.
The ratio of the critical energy release rates depends on the
alignment between the loading stress and the fibers, and the
R. RajGontributing editor
elastic properties of the fiber and interface. The GJG, ratios
were calculated as a function of the angle of crack deflection by
Evans and Marshall'8 for the case where the elastic properties
Manuscript No. 193343. Received August 5 , 1994; approved April 2, 1995. of the fiber and matrix were the same. The analysis was
Presented at the 96th Annual Meeting of the American Ceramic Society, Indianapo-
lis, IN, April 25-27, 1994 (Paper No. SIII-3-94). extended by He and Hutchinson'' to account for an interface
Supported by the U S . Department of Energy Office of Transportation Technologies between materials with different elastic constants. The elastic
Advanced Materials Development Program, through Oak Ridge National Laboratories
Contract No. DE-AC05-840R2140. constants of the material were taken into account through
'Member. American Ceramic Society. Dundar's parameter, aDundar:
2345
2346 Journal of the American Ceramic Sociew- Peterson and Tien Vol. 78, No. 9

determined by the size and shape of the silicon nitride grains,

the residual stresses present in the material, and the details of
the stress vs separation curve of the bridging grains.*' The
where G , = shear modulus of the silicon nitride, G, = shear models attribute the residual stresses to thermal expansion mis-
modulus of the glass, v , = Poisson's ratio of the silicon nitride, match between grains of different orientation~.~~ Additional
and v 2 = Poisson's ratio of the glass. residual stresses are introduced by thermal expansion mismatch
The ratio of critical energy release rates, G,/G, was calculated between the grain boundary phase and the silicon nitride.
as a function of the angle of deflection and aDunda by He and According to Becher's model, the increase in fracture tough-
Hutchinson. l 9 ness caused by bridging grains can be expressed as
Local residual stresses change the fracture resistance of the
fiber and interface according to models proposed by Evans AKgb = [APb~,,E'd/2(1- v ~ ) ] ' ' ~ (6)
et a1.20and Cutler and Virkar." When the silicon nitride is under where AKfb= the increase in fracture toughness from bridging
compressive stress, the fracture resistance of the silicon nitride grains, APb = the area fraction of bridging grains, T~~ = the
particles increases: frictional sliding stess, E" = Young's modulus of the compos-
ite, d = grain length, v = Poisson's ration of the composite.

K , = K,, + 2u,, $ According to both models, the fracture toughness of the

silicon nitride is expected to increase linearly as the area frac-
tion of bridging grains increases.
AK,,, = 2u,, $ 11. Experimental Procedure
where uhn= local average compressive stress in the silicon
nitride, A = distance over which stress is exerted (grain diame- (1) Silicon Nitride
ter), K R = fracture resistance for silicon nitride/particle com- Silicon nitride samples with 5, 10, and 15 vol% of four
posite, K,, = fracture toughness of silicon nitride without different sintering aids were prepared. The sintering aids were
particles, and AK,,, = change in critical stress intensity factor combinations of the commonly used sintering aids cordierite
in the silicon nitride caused by residual stress. (Mg,AI,Si,O,,) and YAG (Y3Al5OJ. The compositions of the
The fracture resistance of the glass decreases because the sintering aids are shown in Table I.
glass is under tensile residual stress. If the silicon nitride is The powders were attrition-milled with silicon nitride for 4 h.
under compression and the glass is under tension, the relative A three-step firing schedule was used for the samples which
fracture energy ratio would decrease, which would promote contained 15 vol% sintering aid, as shown in Table 11. The
crack deflection: samples were densified at 1640C for 3 h, then held at higher
temperature to grow the grains, and then held at 1640C for 2 h
T = -K2 to reequilibrate the grain boundary phase.
E To determine the effect of the volume fraction of grain
boundary phase on the fracture toughness of silicon nitride, a
set of samples with 5 , 10, and 15 vol% of the four different
additives was sintered at 1640C for 3 h.
The samples were cooled rapidly by turning off the power to
where E,, and E, are Young's moduli of the silicon nitride the furnace. X-ray diffraction was used to determine the phases
and glass. The effects of residual stresses on apparent fracture present, which were P-Si,N, and glass. The compositions of the
resistance had been observed in tests of crack propagation along grain boundary glasses were determined by the lever rule from
bimaterial interfaces by Charalambides et a1.** A residual stress the volume fraction of glass present in SEM photos of the
of 10 MPa at the interface of a sandwich specimen changed the microstructure, and compared to the compositions determined
measured fracture resistance from the intrinsic material value from the bulk glasses.
of 2 J.rn-* to 7 J.m-*. Some samples underwent further heat treatment for 15 h at
There are many equations in the literature to calculate resid- 1250"-1350"C to crystallize the grain boundary phase. X-ray
ual stresses. One widely used one is based on Eshelby's inclu- diffraction was used to determine the phases present. The densi-
sion theoryz3as arranged by HsuehZ4and Taya et ~ 1 . : ' ~ ties of the samples were measured using Archimedes' method.
The fracture toughness was measured by the indentation
method with loads of 10 kg. The fracture toughness was calcu-
lated using the equation of Anstis et ~ 1 . : "

(7)
and
where K , is the fracture toughness, E is Young's modulus, H i s
f," the hardness, P is the peak load on the Vickers indenter, and c
u, = --I &"
- is the crack length from the center of the indent to one tip of
where ub,,and us are the stresses in the silicon nitride and glass the crack.
matrix, a,, and ag are the thermal expansion coefficients of The microstructures and crack paths were examined using
the silicon nitride and glass matrix, AT is the difference in the SEM. When the elemental contrast in the SEM was insuffi-
cient to reveal the microstructure, the samples were etched in a
temperature over which the stresses are locked in, f , , is the
volume fraction of the silicon nitride, vsn and v g are Poisson's
ratios of the silicon nitride and glass, and E,,, and E, are Young's
moduli of the silicon nitride and glass. Table I. Compositionsof
The glass was assumed to be the continuous phase, and the Sintering Aids
particles are Si,N,. AT was taken as the difference between the Cordierite YAG
Composition (wt%) (wt%)
glass transition temperature of the grain boundary glass and
room temperature, which was approximately 1000C. C 100 0
Models developed by Becherz5 and Swansson et a1." for 8OCl2OY 80 20
alumina relate fracture toughness to the presence of bridging 60C/40Y 60 40
grains behind the crack tip. The crack tip bridging stresses are Y 0 100
September 1995 Effect of the Grain Boundary Thermal Expansion Coefficient on the Fructure Toughness in Silicon Nitride 2341

Table 11. Sintering and Crystallization Schedules

Name Steo 1 Steo 2 Steo 3
1640 (3) 1640C 3 h
1800 (3) 1640C 3 h 1800C 3 h 1640C 2 h
1840 (1) 1640C 3 h 1840C 1 h 1640C 2 h
1840 (3) 1640C 3 h 1840C 3 h 1640C 2 h
1840 (6) 1640C 3 h 1840C6 h 1640C 2 h
1880 ( 3 ) 1640C 3 h 1880C 3 h 1640C 2 h

molten salt solution for 15 to 30 s. The distance between bridg- that the glasses were completely amorphous; no Si,N, precipi-
ing grains was measured from the SEM photographs from five tates were detected. The thermal expansion coefficients of the
indentation cracks for each sample. The portions of the crack glass compositions closest to the solubility limit of Si,N were
which showed space between the faces were examined for measured using an alumina rod dilatometer calibrated with a
evidence of bridging grains. The cracks were photographed in fused silica standard.
segments of approximately 10 pm. Grains which remained The glasses underwent a crystallization treatment at 1250"-
intact across the crack face were immediately identified as 1350C for 15 h. The thermal expansion coefficients of the
bridging grains. Grains which had fractured at a distance from devitrified samples were measured using an alumina rod dila-
the crack plane and which showed evidence of being pulled out tometer calibrated with a fused silica standard.
from their sockets were also considered to have been grain
bridges. 111. Results
The angle between the direction of crack propagation and a
silicon nitride grain in its path determines whether the crack
will propagate through the grain or be deflected. For a sample (1) Compositions and Thermal Expansion Coefficients of
with a given composition, the crack would be deflected around
Glasses
the grain if the angle between the advancing crack and the grain The compositions and thermal expansion coefficients (as deter-
was smaller than a critical angle. When the crack was deflected mined from the model glasses) of the grain boundary glasses
along a grain boundary, the angle between the macroscopic are shown in Table 111. The compositions determined from the
direction of the crack and the direction of the deflected segment lever rule from the silicon nitride samples were the same as the
was measured from SEM micrographs. compositions determined from the model glasses, and with
The mean angle of crack deflection and the maximum angle those reported in earlier The crystalline phases and
of crack deflection were measured. The entire lengths of five their thermal expansion coefficients are listed in Table IV. The
indentation cracks per sample were used to determine the thermal expansion coefficient of the cordierite-rich glasses
angles of crack deflection. The cracks were divided into seg- decreases on crystallization. The thermal expansion coefficient
ments roughly 10 pm in length. of the YAG-rich glasses increases on crystallization.
The grain sizes of the samples were measured by the disinte- (2) Effect of Thermal Expansion Coefficient of the Grain
gration method. The samples were disintegrated in a molten salt Boundary Phase on Crack Propagation in Silicon Nitride
bath of K0H:NaOH in 1:1 molar ratio. To remove the surface All the sintered samples were at least 99% dense. The micro-
grains, the samples were etched for 3 min, rinsed in water, and structures produced using the different sintering aids in the
placed in water in an ultrasonic cleaner for 5 min. The sample Y,Mg,Si,Al/O,N were very similar, as shown in Fig. 1 for
was then etched again for an additional 3 min, rinsed in water, samples sintered at 1640C for 3 h, followed by 1840C for 1 h
and placed in water in the ultrasonic cleaner again for 5 min. and 1640C for an additional 2 h. The average length and width
The loose grains from the second etching step were collected of the silicon nitride grains are shown as a function of the
from the water by vacuum filtration. SEM photos of the col- amount of YAG in the sintering aids in Figs. 2(a) and (b). Note
lected grains were used to determine the average length and that the effect of sintering aid chemistry on grain morphology
width of the grains. At least 100 grains were measured to was negligible in this particular case, primarily because the
determine the length and width values. grain growth temperature was very low and the volume frac-
(2) Composition and Thermal Expansion Coefficients of tions of liquid were the same. Therefore, any changes in frac-
Grain Boundary Phases ture toughness for samples treated at the same temperature were
Glass compositions with the same ratios of oxide powders as not due to changes in grain shape or size. However, the crack
in the sintering aids were prepared. Silicon nitride additions of paths are very different when different sintering aids are used,
5, 10, and 15 wt% were added to the oxide powders. The MgO,
Y,O,, A1,0,, SiO,, and Si,N, powders were mixed by hand.
The powders were melted under 10 atm nitrogen pressure at Table IV. Thermal Expansion Coefficients of
1650"-1700"C for 3 h. The weight losses were <1 wt%. The Crystalline Grain Boundary Phases
maximum solubility of silicon nitride was determined using Phase 01 ( X 10h/"C)
X-ray diffraction and SEM. Electron probe microanalysis indi- YAG (Y,A1,0,,) 8.0
cated that the cation ratios had not changed during melting. N-phase (MgAI,Si,O,N,) 2.3*
Glass compositions as close to the solubility limit of Si,N, as *The thermal expansion coefficient of N phase wa$ measured
possible were fabricated. X-ray diffraction and SEM showed by Nunn."

Table 111. Grain Boundary Glass Compositions and Thermal Expansion Coefficients
Composition (mol fraction)
01
MgO Y*O, A120, SiO, Si,N, (X lOhYC)*

0.20 0 0.21 0.51 0.08 3.4

0.18 0.03 0.24 0.46 0.09 3.7
0.16 0.08 0.28 0.39 0.10 4.1
0 0.25 0.42 0.08 0.25 4.6
*Thermal expansion of coefficient of p-Si,N,: 3.0 X lO-"/"C.
2348 Journal of the American Ceramic Socieg- Peterson and Tien Vol. 78, No. 9

fM 1640(3)
3 + lSOO(3)
c
1840(3)
* 1880(3)

0 20 40 60 80 100
Wt % YAG in 15 vol % sintering aid

B
1640(3)
lSOO(3)
+ + 1840(3)
1880(3)
.-e 0
E
c7
0 I . 1 . I

0 20 40 60 80 100
Wt % YAG in 15 vol % sintering aid

Fig. 2. (a) Effect of sintering aid composition on length of grains.

(b) Effect of sintering aid composition on width of grains.

thermal expansion coefficient of the grain boundary phase

increased, as shown in Figs. S(a) and (b). The range of mean
crack deflection angles is the same as the range reported for
Si,N, samples of similar toughness by Faber and Evans.
The effect of crystallization on fracture toughness is shown
Fig. 1. (a) Microstructure and crack path in silicon nitride sintered in Fig. 6. When the thermal expansion coefficient of the grain
with 15 vol% of 100 wt% cordierite. This sample was plasma etched. boundary phase increased on crystallization, the fracture tough-
(b) Microstructure and crack path in silicon nitride sintered 15 vol% of ness increased upon crystallization. When the thermal expan-
60 wt% cordierite/40 wt% YAG. This sample was plasma etched. sion coefficient of the grain boundary phase decreased on
crystallization, so did the fracture toughness.

as shown in Fig. 1. The sample sintered with cordierite alone (2) Effect of Volume Fraction of Sintering Aid
shows a nearly straight crack path with almost no bridging The grain length increased slightly as the amount of sintering
grains. In contrast, the sample sintered with YAG shows crack aid increased, as shown in Fig. 7 for silicon nitride sintered with
deflection and bridging grains. cordierite. The fracture toughness of the silicon nitride did not
The fracture toughness would be expected to increase as the change when different volume fractions of magnesium alumi-
distance between bridging grains decreased and as the amount nosilicate sintering aid were used. The fracture toughness
of crack deflection increased. As expected from the photos of increased slightly with increasing volume fraction of YAG, as
the crack paths, the fracture toughness increased as the thermal shown in Fig. 8.
expansion coefficient of the grain boundary phase increased, as
shown in Fig. 3. IV. Discussion
The fracture path was characterized by the distance between
bridging grains and the angles of crack deflection. The distance The material properties used in the calculations in this sec-
between bridging grains was measured from the SEM photo- tion are listed in Table V. Using a modified version of Eshelbys
graphs. The distance between grains decreased as the thermal inclusion methodI5 discussed earlier, the residual stress in the
expansion coefficient of the grain boundary phase increased, as silicon nitride and in the grain boundary phase can be calculated
shown in Fig. 4. from Eqs. (5) and (6), as shown in Fig. 9 for 1.5 vol% of
The mean and maximum angles of crack deflection were sintering aid. When the thermal expansion coefficient of the
determined from the SEM micrographs. The grain size of the grain boundary phase is greater than that of the silicon nitride,
sample had only a small effect on the crack deflection angles, both the tensile stress in the grain boundary and the compres-
as shown in Fig. 5. The angles increased slightly as the grain sive stress in the silicon nitride grains increase as the thermal
size of the sample increased. The angles increased as the expansion coefficient of the grain boundary phase increases.
September 1995 Effect of the Grain Boundary Thermal Expansion Coefficient on the Fracture Toughness in Silicon Nitride 2349

0 5 0 0

Y
0

4E 1800(3)
1840(3)
'
1640C
1840C
mi
1840(6) ~ 1880C
1880(3)
6
i:
L
*
c
k
3.0 3.4 3.8 4.2 4.6 5.0
7

3.4 3.6 3.8 4.0 4.2 4.4 4.6

Thermal expansion coefficient
Thermal Expansion Coefficient
(x 106PC) (x 10 6/ "C)

Fig. 3. Effect of thermal expansion coefficient of grain boundary (a>

phase on fracture toughness of silicon nitride sintered with Y20,-
Al20,-Mg0-Si0, additives.

The residual stress in both phases can also be calculated as

a function of the volume fraction of sintering aid, as shown 9
c
0

in Fig. 10. 70 0 0 Maximum angle

Tensile stress at the grain boundary and compressive stress
in the silicon nitride grains promote crack deflection and the
formation of bridging grains. The expected T,/T, ratio was esti-
mated (roughly) as a function of residual stress using Eq. (4)
and the material properties listed in Table V. The distance over
which stress was exerted was interpreted as the average grain
diameter for silicon nitride. The initial toughness value of the 30 I'''I ' ' I
I

3.2 3.4 3.6 3.8

I ' I ''I
4 4.2 4.4 4.6 4.8
* ' ' 'j
aluminosilicate glass was taken as 0.9 MPa.mo5.28 The initial
toughness of silicon nitride was taken as the single-crystal value Thermal expansion coefficient
of 2.0 MPa.mn5.13The distance over which stress was exerted ( x 106/0c )
was interpreted as the average diameter of the silicon nitride
grain. Since the maximum angle of crack deflection occurs
when TJT, = G,/G,, the predicted maximum angles of deflec-
tion can be found from the calculated TJT, ratios and the calcu- Fig. 5. (a) Effect of sintering temperature and thermal expansion
lated Dundar's parameter, aDundar. The value of aDundar was coefficient on mean crack deflection angle. (b) Effect of thermal expan-
calculated as 0.5. The critical angle of deflection was found sion coefficient on maximum crack deflection angle (sintering step 2 at
from the graphs in He and Hutchinson's paper.I9The calculated 1840C for 3 h).
critical angle of deflection is compared to the measured critical
angle of deflection in Fig. 11. These calculations were per-
formed to demonstrate that the change in crack deflection angle for crack deflection. The linear fraction of the crack which was
follows the trends expected from theory; no further significance bridged is l/d, where 1 is the grain diameter and d is the distance
is implied. between bridging grains. The linear fraction of the crack which
The fracture of grains which become bridges depends on the was bridged increased linearly as the fraction of grains expected
fracture of grains which intersect the crack at an angle favorable to intersect the crack at an angle less than the critical angle
increased, as shown in Fig. 12 for silicon nitride sintered with
combinations of cordierite and YAG.
The area fraction of bridging grains, A,,, can be estimated
from the grain diameter and the distance between bridging
grains:27

A =-
l2
gb 2d2 (9)
1800C
6' 1840C If a sample was broken in half and the fracture surface was
0
examined, the fraction of the fracture surface which consisted
8 1880C of broken grain bridges would be the area fraction of bridging
4- grains. In the samples in this study, the area fraction is just a
0 mathematical convenience, since only the linear portion of the
crack which is bridged can be measured. Figure 13 shows the
2 " " . " ' . ' . linear increase in fracture toughness with the area fraction of
3.4 3.6 3.8 4.0 4.2 4.4 4.6 bridging grains, in agreement with theory.25" For modeling
Thermal expansion coeffcient purposes, it should be possible to predict the area fraction of
bridging grains from the predicted maximum deflection angle
(x 10 /"C) and the grain diameter.
Changes in the chemistry of the glass have also been reported
Fig. 4. Effect of thermal expansion coefficient on distance between to change crack deflection angles by altering the intrinsic bond-
bridging grains. ing strength of silicon nitride/glass interface, even in the
23.50 Journal of the American Ceramic Society-Peterson and Tien Vol. 78, No. 9
6.0 Cordierite
YAG
5.6

5.2
Quenched
After 4.8
crystallization
treatment
4.4
3
"U
; 4 *
V
C 4.0
0 20 40 60 80 100
k 4 6 8 10 12 14 16
Wt % YAG
Volume % Additive
in CordieritelYAG sintering aid
Fig. 8. Effect of volume fraction additive on fracture toughness of
Fig. 6. Effect of crystallization on fracture toughness. silicon nitride.

absence of residual stress effects." However, the differences in grain boundary triple points following crystallization treat-
chemistry between the first three grain boundary phase compo- ments in samples of P-Si,N, sintered with yttria and strontia."
sitions shown in Table 111are mostly caused by substitution of SEM examination of the samples in this study did not reveal
cations which are not directly involved in bonding across the any pores in the grain boundary pockets. If there were pores,
interface. Therefore, the effect of the change in chemistry on they may have been too small to detect by SEM.
the interfacial bond strength is expected to be relatively small Although the density of the grain boundary phases increased
for the three samples which contain cordierite in the sintering during crystallization in all of the samples, the fracture tough-
aid. ness increased only in the samples which had the highest vol-
The changes which occur in fracture toughness after crystal- ume fraction of YAG and decreased in others. Since the thermal
lization of the grain boundary phase can't be attributed entirely
to changes in the thermal expansion coefficient of the grain
boundary phase. Some of the change in fracture toughness after
crystallization may be due to the stresses or porosity arising
from volume changes which occur on crystallization.
According to earlier studies, both YAG and N-phase have
higher densities than their respective oxynitride glasses. The
density of the grain boundary phase in the Mg,Si,Al/O,N sam-
ples increases from the density of the glass at 2.6 g/cm' to the
density of N-phase, which is 2.9 g/cm3, an increase of ll%.31
The density of the grain boundary phase in the Y,Si,Al/O,N
samples increased from the density of the glass, at 3.9 g/cm3 to
the density of Y,AI,O,,, at 4.55 g/cm', an increase in density of
almost 17%. The volume changes can be accommodated by
viscous flow of the grain boundary phase during crystallization, I Thermal Expansion Coefficient
by solution/precipitation of the silicon nitride grains during
crystallization, or by the development of pores in the grain (x 106/"C)
boundary phase.3z Extensive crystallization would not have
been possible unless most of the stresses produced during crys- Fig. 9. Residual stress as a function of thermal expansion coefficient
tallization were re1ieved.l2 Nunn reported that small shrinkage of the grain boundary phase for Si,N, containing 15 vol% grain bound-
pores appeared in the grain boundary phase after crystallization ary phase.
in samples of P'-SiAlON sintered with cordierite additives."
Shrinkage pores have also been reported in the corners of the

2.01 : :
300

0
-100 , , , ,r , , &Silicon
,7, Nitride
/,w]

-200
0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
4 6 8 10 12 14 16 Volume fraction Si3N4
Volume % Additive
Fig. 10. Residual stress as a function of the volume fraction of grain
Fig. 7. Effect of small volume fractions of cordierite additive on boundary phase. Thermal expansion coefficient of the grain boundary
length of Si,N, grains. phase = 6 X lOP/"C.
September 1995 Effect of the Grain Boundary Thermal Expansion Coeflcient on the Fracture Toughness in Silicon Nitride 235 I

Table V. Elastic Properties Used in Calculations expansion coefficient of YAG is higher than that of the Y-Si-
Al-0-N glasses, the thermal expansion coefficient of the grain
~ ~~~~~~~

Silicon nitride Glass

boundary phase which contains a large fraction of YAG would
E (GPa) 300 I00 increase after crystallization. On the contrary, the thermal
V 0.3 0.3
K O(MPa.m*) 2.0 0.9 expansion coefficient of grain boundary phases which contain a
large amount of N-phase would decrease after crystallization.
The data in Fig. 6 suggest that the change in fracture toughness
on crystallization is influenced by the change in thermal expan-
sion coefficient of the grain boundary phase.

V. Conclusions
Thermal expansion mismatch between the grain boundary
glass and the silicon nitride has been shown to influence the
fracture toughness of silicon nitride. The thermal expansion
mismatch causes residual stresses in the material, which inter-
act with the advancing crack. Residual tensile stress at the grain
boundary has been shown to increase the number of bridging
grains and increase the fracture toughness. Although the ther-
mal expansion coefficient of the grain boundary phase is not the
only factor which affects the fracture behavior of silicon nitride
50 100 150 200 250 ceramics, it has a large and predictable effect. Further studies
on the effect of the chemistry of the sintering aid on microstruc-
Grain Boundary Stress, MPa tural development of the silicon nitride/glass interfaces and on
the bond strengths between the silicon nitride and the glass
Fig. 11. Measured maximum deflection angles and calculated would further improve understanding of the effect of the grain
deflection angles. boundary phase on the fracture toughness of silicon nitride.

References
P. F. Becher, Microstructural Design of Toughened Ceramics, J. Am.
Ceram. Soc., 74 [2j 255-69 (1991).
,M. Mitomo and S. Uenosono, Microstructural Development during Gas
Pressure Sintering of &Silicon Nitride, .I. Am. Cerum. Soc., 75 [ l j 105-
108 (1992).
P. F. Becher, H. T. Lin, S. L. Hwang, M. J. Hoffman, and I-W. Chen, The
Influence of Microstructure on the Mechanical Behavior of Silicon Nitride
Ceramics; pp. 147-58 in Materials Research Society Symposium Proceedings,
Vol. 287. Edited by I-W. Chen, P. F. Becher, M. Mitomo, G. Petzow, and T-S.
Yen. Materials Research Society, Pittsburgh, PA, 1993.
A. Pyzik and D. R. Beaman, Microstructure and Properties of Self-
Reinforced Silicon Nitride,J. Am. Ceram. Soc., 76 [I 11 2737-44 (1993).
T. Payashi and T. Y. Tien, Formation and Crystallization of Oxynitride
Glasses in the System Si,Al,Mg/O,N, Yogyo Kyokuishi, 94 [l] 44-52 (1983).
6D. A. Bonnell, Determination of Crystallization Mechanisms and Correla-
tion of Grain Boundiuy Morphology to Mechanical Properties in Silicon
0.0 0.2 0.4 0.6 0.8 1.0 Nitride; Ph.D. Thesis. University of Michigan, Ann Arbor, MI, 1986.
7C. M. Hwang, The System of SiA10N-Y3AIs0,, and SiAION-Cordierite:
Fraction of grains intercepting crack Sintering and Grain Growth; Ph.D. Thesis. The University of Michigan, Ann
Arbor, MI, 1988.
at angle < max deflection angle M. Wisnudel, Solid-Liquid Reaction in the System SiAION-Y,Al,O,,;
M.S. Thesis. University of Michigan, Ann Arbor, MI, 1991,
Fig. 12. Linear fraction of crack which is bridged increases as the 9Z. K. Huang and T. Y. Tien, Solid-Liquid Reaction in the System Si,N,-
Y,Al,O,,-Y,Si,O, under 1 MPa of Nitrogen, J. Am. Cerum. Soc., 77 [lo] 2763-
fraction of grains oriented favorably for deflection increases (sintering 66 (1994).
step 2 at 1840C for 3 h). S. D. Nunn, H. Honke, L. J. Gaukler, and T. Y. Tien, Suhsolidus Phase
Relationships in Part of the System Si,AIMg/N,O, Am. Ceram. Soc. Bull., 57,
321 (1978).
I. K. Naik and T. Y. Tien, Suhsolidus Phase Relations in Part of the System
Si,Al,Y/N,O,J. Am. Cerum. Soc., 62,642 (1979).
*W.-Y. Sun, T. Y. Tien, and T.3. Yen, Sub-solidus Phase Relationships in
Part of the System Si,Al,Y,/N,O: The System Si,N,-AIN-YN-Al,O,-Y,O,,
J.Am. Ceram. Soc., 74 [ I l l 2753-58 (1991).
I. Tanaka, G. Pezzotti, Y. Miyamoto, T. Okamoto, Fracture Toughness of
Si,N, and Its Si,N, Whisker Composite without Sintering Aids, J. Muter. Sci.,
26,208-10 (1991).
I4G. H. Campbell, M. Ruhle, B. J. Dalgleish, and A. G. Evans, Whisker
Toughening: A Comparison between Aluminum Oxide and Silicon Nitride
Toughened with Silicon Carbide, J . Am. Ceram. Soc., 73 [3] 521-30 (1990).
1800(3) M. Taya, S. Hayashi, A. S. Kobayashi, and H. S. Yoon, Toughening of a
1840(3) Particulate Reinforced Ceramic Matrix Composite by Thermal Residual Stress,
.I.Am. Cerum. Soc., 73 [5j 1382-91 (1990).
A 1880(3) H. Cai, N. P. Padture, B. M. Hooks, and B. R. Lawn, Flaw Tolerance and
Toughness Curves in Two-Phase Particulate Composites: SiC/Glass System,
J. Eur. Cerum. Soc., 13, 149-57 (1994).
A. G. Evans and M. Y. He, Interface Debonding and Fiber Cracking in
0.0 0.2 0.4 Brittle Matrix Composites, J . Am. Ceram. Soc., 72 [12] 2300-303 (1989).
A. G. Evans and D. B. Marshall, Overview No. 85: The Mechanical Behav-
Area Fraction Bridging Grains ior of Ceramics Matrix Composites,Acru Metall., 37 [lo] 2567-83 (1989).
M. Y. He and J. W. Hutchinson, Kinking of a Crack out of an Interface,
.I.Appl. Mech., 56 [6] 270-78 (1989).
Fig. 13. Fracture toughness increases linearly as the area fraction of A. G. Evans, A. H. Heuer, and D. L. Porter, The Fracture Toughness of
bridging grains increases. Ceramics, Proc. Iut. Con$ Fruct., 4th, I, 529-56 (1977).
2352 Journal of the American Ceramic Society- Peterson and Tien Vol. 78, No. 9
2R. A. Cutler and A. V. Virkar, The Effect of Binder Thickness and Residual *G. R. Anstis, P. Chantikul, B. R. Lawn, and D. B. Marshall, A Critical
Stresses on the Fracture Toughness of Cemented Carbides, J. Mater. Sci., 20, Evaluation of Indentation Techniques for Measuring Fracture Toughness: I,
3557-73 (1985). Direct Crack Measurements,J. Am. Ceram. Soc., 4 [ 9 ] 533-38 (1981).
*P. G. Charalambides, H. C. Cao, J. Lund, and A. G. Evans, Development of *K. T. Farber and A. G. Evans, Crack Deflection Processes-11. Experi-
a Test Method for Measuring the Mixed Mode Fracture Toughness of Bimaterial ment, Acta Metull., 3 1 [4] 577-84 (1983).
Interfaces, Mech. Muter., 8,269-83 (1990).
23T. Mura; pp. 184 in Micromechanics of D e f e m in Solids, 2nd rev. ed. lop. F. Becher, S.-L. Hwang, and C.-H. Hsueh, Using Microstructure to
Martinus Nijhoff, Dordrecht, The Netherlands, 1987. Attack the Brittle Nature of Silicon Nitride Ceramics, MRS Bull., 10 I21 23-
24C.H. Hsueh, Sintering Behavior of Powder Compacts with Multiheteroge- 27 (1995).
neities,J. Muter. Sci., 21,2067-72 (1986). S. D. Nunn, Processing and Properties of S i c Whisker Reinforced Si,N,
25P. F. Becher, Reinforced Ceramics Employing Discontinuous Phases; Ceramic Matrix Composites; Pb.D. Dissertation. The University of Michigan,
pp. 475-83 in Ceramic Transactions, Vol. 19, Advanced Composite Materials. Ann Arbor, MI, 1991.
Edited by M. Sacks. American Ceramic Society, Westerville, OH, 1990. W. Pompe and H. Kessler, Internal Stresses in Silicon Nitride and Their
z6P.L. Swansson, C. J. Fairbanks, B. R. Lawn, and Y. Mai, Crack-Interface Influence on Mechanical Behavior; pp. 353-64 in Tailoring of Mechanical
Grain Bridging as a Fracture Resistance Mechanism in Ceramics: 1, Experimen-
tal Study on Alumina,J. Am. Cerum. Soc., 70 [4] 279-89 (1987). Properties ofSi,N, Ceramics. Edited by M. J. Hoffman and G. Petzow. Kluwer
S . J. Bennison and B. R. Lawn, Role of Interfacial Grain-Bridging Sliding Academic Publishers, Dordrecht, Netherlands. 1994.
Friction in the Crack Resistance and Strength Properties of Non-Transforming 3. A. Bradley and K. R. Karasek, Effect of Thermal and Stress Exposure
Ceramics, Acta Metall., 37 [ 101 2659-71 (1989). on Grain Boundaries of Silicon Nitride, Ultramicroscopy, 29,9-17 (19x9). 0