nfiSlS

 

' Ilillllllllllllllllllllllllllll

3 1293 01413 7867

This is to certify that the

thesis entitled

COLLOIDAL OPTIMIZATION OF FREEZE?
DRIED Sij REINFORCED Si3N4 CERAMIC
COMPOSITES

presented by l
Patrick R. Sneary

has been accepted towards fulfillment
of the requirements for

Masters Materials Science

degree in

 

 

0—7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

l
T
1

II

 

 

LIBRARY

Michigan State
University

 

 

 

PLACE N RETURN
TO AVOID FINES

BOX to remove this checkout from your record.
return on or before dnto duo.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Cimna-pd

fi-

COLLOIDAL OPTIMIZATION OF FREEZE-DRIED SICw
REINFORCED SIsN4 CERAMIC COMPOSITES

By

Patrick R. Sneary

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF SCIENCE
Department of Materials Science and Mechanics

1996

i. ll;

.III (

VIII

ABSTRACT
COLLOIDAL OPTIMIZATION OF FREEZE DRIED SiC“, REINFORCED SI3N4
CERAMIC COMPOSITES
by

Patrick R. Sneary

Si3N4 is a candidate for many advanced structural engineering applications because
of its excellent thermal and wear properties, however, application of Si3N4 is limited due to
its low fracture toughness. Because the fracture toughness of ceramic materials is highly
dependent upon initial flaw sizes, improvement of the fracture toughness of ceramics must
begin with green processing. Inefficient packing Of powders within a green ceramic
compact can lead to an inhomogeneous pore structure making it very difficult to Sinter to
full density without hot pressing or hot isostatic pressing (HIPing). Optimizing processing
of Si3N4 components, as well as the addition of a toughening phase such as SiC whiskers, '
has been shown to improve the fracture toughness of Si3N4. This study examines the effect
of colloidal optimization on freeze-dried Si3N4 monolithic and composite powders
consolidated by pressureless sintering. Results are given as a function of Sij loading and

processing pH.

fl

'5];

' .

ACKNOWLEDGMENTS

I would like to thank my advisor Dr. Melissa Crimp, colleague Zhengmao Yeh and
the countless number of great fiiends I have made while at Michigan State. Your patience
with me and confidence in me made this thesis possible, and your friendship allowed me to
keep my sanity. A special thanks to my parents whose love and support gives me the

strength to accomplish anything.

iii

TABLE OF CONTENTS

LIST OF TABLES .................................................................................................. vi
LIST OF FIGURES ............................................................................................... vii
INTRODUCTION ................................................................................................... 1
REVIEW OF LITERATURE .................................................................................. 3
Colloidal Forces ............................................................................... 4

Attractive Forces .................................................................. 7

Repulsive Forces ................................................. i ................. 8

Diffuse Double Layer ............................................... 9

Stem Theory ........................................................... 14

Zeta Potential ......................................................... 15

HHF Theory ........................................................... 18

Stability Prediction ............................................................ 22

Freeze-Drying ............................................................................... 23

Liquid Phase Sintering .................................................................. 26

Initial Stages ...................................................................... 28

Intermediate Stage: Solution-Reprecipitation ................... ‘31

Final Stage: Grain Growth ................................................ 35

iv

Processing Considerations ................................................. 36

Liquid Phase Sintering of Si3N4 .................................................... 39

Powder Morphology ......................................................... 41

Impurities .......................................................................... 43

YZO3+A1203 as Sintering Aid ........................................... 45

Sij Additions to Si3N4 ................................................... 49

Mass Loss ......................................................................... 50

Sintering Time and Temperature ...................................... 51

Fracture Toughness ....................................................................... 52

Determination of KIC ........................................................ 52

Whisker Toughening Mechanism ................................................. 57
EXPERIMENTAL PROCEDURE ....................................................................... 63
RESULTS ............................................................................................................. 74
Colloidal Optimization ................................................................. 74

Freeze-Dried Powders .................................................................. 81

Green Density ........................................................................ . ...... 90

X-ray Analysis of Green Powders ................................................ 92

Sintered Density ........................................................................... 99

X-ray Analysis of Sintered Compacts ......................................... 112

Sij Toughening of Si3N4 ........................................................... 131
DISCUSSION ..................................................................................................... 155
Green Processing of Sij- Si3N4 Composites ............................ 155

V

Densification ............................................................................... 1 58

CONCLUSIONS ................................................................................................. 1 62
RECOMMENDATIONS .................................................................................... 1 64
REFERENCES .................................................................................................... 165

vi

LIST OF TABLES

Table 1 Double layer thickness with respect to electrolyte
concentration. Adapted from [1] ..........................................................................
Table 2 Effect of Sintering mechanism on a powder compact. Adapted
fi'om [21] ................................................................................................................
Table 3 Purity, particle size, and surface area of starting ceramic
powders ..................................................................................................................
Table 4 Range of homostability and heterostability for Sij/Si3N4
composite system as predicted by Suspension Stability© ....................................
Table 5 The effect of processing pH and ball-milling on the
tapdensities of freeze-dried powders. Powders include 8% wt. Y203 - 2% wt.
A1203 as Sintering aids. Values listed are in g/cm3 ...............................................
Table 6 The effect of whisker loading and selected component ball-

milling on tap densities of powders freeze-dried at pH=11with Sintering aids.
Values listed are in g/cm3 ......................................................................................

Table 7 Green density as a function of pH and whisker loading
(g/cm3) ...................................................................................................................
Table 8 Green densities of specimens CIPed at 250MPa with 8 wt.%
Y203/ 2 wt.% A1203 added as Sintering aids ........................................................
Table 9 Effect of ball-milling on green densities of specimens CIPed at
250MPa with 8 wt.% Y203 / 2 wt.% A1203 added as Sintering aids ....................

vii

12

32

63

81

81

89

90

91

92

LIST OF FIGURES

Figure 1 Normalized double layer potential versus interparticle
separation according to Debye-Huckel. (a) shows variation with electrolyte
concentration (b) shows variation with electrolyte valence for 10'3 M
symmetrical electrolyte. Adapted fiom [1] ..........................................................

Figure 2 Stem potential ‘I’d is evaluated at the shear plane which
represents the charged surface, adsorbed ions, and suspension medium which
move as one unit. Charged surface is represented by shaded spheres along the
y-axis. Adapted from [5] ......................................................................................

Figure 3 Simplifying parameter h as defined by Derjaguin from
equation (14). Adapted from [10] ........................................................................

Figure 4 An energy barrier to coagulation is necessary for suspension
stability. Plot is the result of adding Vr + Va for a stable suspension. Adapted
from [5] ..................................................................................................................

Figure 5 Evolution of a green ceramic compact Sintered with the
assistance of a liquid phase. I = initial stage, II = intermediate, and III = the
final stage of sintering. Adapted from [21] ..........................................................

Figure 6 Repeating atomic unit layers for Si3N4. Adapted from [63] .......
Figure 7 Plot showing the sensitivity of a and [3 powders to Sintering
temperature. [3 powders are more difficult to Sinter at low temperatures.

Adapted fi'om [64] .................................................................................................

Figure 8 The effect of 02 content on the densification of Si3N4 powders
with respect to surface area. Adapted from [64] ..................................................

viii

13

16

20

21

27

42

44

46

Figure 9 Schematic of Vickers indent technique. Radial cracks result

upon unloading and are subsequently measured and related to ceramic
hardness and fracture toughness. Adapted from [77] ..........................................

Figure 10 Average agglomerate size vs. pH for as-received ultrasonicated
Si3N4 powders as determined by Andreason Pipette Method ..............................

Figure 11 Sedimentation density vs. pH for as-received ultrasonicated
SI3N4 and Sle .......................................................................................................

Figure 12 Zeta-potential vs. pH for as-received ultrasonicated Si3N4 and
Sij as determined by ESA ..................................................................................

Figure 13 Stability Ratio vs. pH for as-received ultrasonicated Si3N4 and
Sij ........................................................................................................................

Figure 14 Freeze-dried, ultrasonicated and ball-milled Si3N4 powders
processed at pH 3 with Sintering aids ....................................................................

Figure 15 Freeze-dried, ultrasonicated and ball-milled Si3N4 powders
processed at "natural" pH (7) with Sintering aids ................................................

Figure 16 Freeze-dried, ultrasonicated and ball-milled Si3N4 powders
processed at pH 11 with Sintering aids ..................................................................

Figure 17 Freeze-dried and ultrasonicated Si3N4 powders processed at pH
11 with Sintering aids. ...........................................................................................

Figure 18 Freeze-dried, ultrasonicated and ball-milled 2OV/o Sij/Si3N4
powders processed at pH 11 with Sintering aids. .................................................

Figure 19 Freeze-dried ultrasonicated and ball-milled 20V/o SiC“/Si3N4
powders processed at pH 11 with Sintering aids. Sij added after milling .........

Figure 20 X-ray diffraction results for Ube Sn-e-lO as-received Si3N4. A
= a- SI3N4 and B = B'SI3N4 ...................................................................................

Figure 21 X-ray diffraction results for Tokai Whisker TWS-l 00 as-
received B-SiC whiskers ........................................................................................

Figure 22 X-ray diffraction results for as-received Y203 powder. Circled
region indicates undefined peak. ..........................................................................

54

75

77

78

80

82

83

84

85

86

87

93

94

95

Figure 23 X-ray diffraction results for as-received a-A1203 powder ..........

Figure 24 X-ray diffraction results for ball-milling residue from powders
ball-milled at pH 7.5 and 11. A = a-Si3N4, Y = Y203. High background may
indicate the presence of an amorphous Y-containing phase .................................

Figure 25 X-ray diffraction results for monolithic Si3N4 ball-milled at pH
7. A = Ot-Si3N4, Y = Y203 ....................................................................................

Figure 26 X-ray diffraction results for monolithic Si3N4 ball-milled at pH
1 1. A = a'Sl3N4, Y = Y203 ..................................................................................

Figure 27 X-ray diffraction results for 10 v/o Sij-Si3N4 ball-milled at pH
7. A = a-SI3N4, Y = Y203, S = SIC“, ....................................................................

Figure 28 X-ray diffraction results for 10 V/o Sij-Si3N4 ball-milled at pH
11. A = Ot-Si3N4, Y = Y203, S = SIC“, ..................................................................

Figure 29 X-ray diffraction results for 10 v/o Sij-Si3N4 ball-milled at pH
3. A = a‘SI3N4, Y = Y203, S = Sij ....................................................................

Figure 30 X-ray diffraction results for monolithic Si3N4 ball-milled at pH
3. A = a-Si3N4, Y = Y203. Note the Y203 peaks are missing. The new
uncharacterized crystalline phase is indicated by * ..............................................

Figure 31 Density as a function of whisker loading for ball-milled Si3N4
specimens Sintered 4 hours. Results show higher densities for composites
with ball-milled (reduced aspect ratio) Sij in addition to milling of Si3N4...

Figure 32 Density as a function of time for selected component ball-
milled powders. High density is achieved more quickly when all
components are milled. Milling Sij is more important for high density
than milling Si3N4 alone.After 4 hours, compacts containing unmilled
components showed only limited densification. This point is circled for
clarity ................................................................................................................

Figure 33 Density as a function of time and whisker loading for
powders where all components were ball-milled. Final density decreases as
whisker loading increases. Compared to unmilled monolithic powder, ball-
milled monolithic powder shows a lower final density and lower
densification rate ...................................................................................................

96

98
100
101
102
103

104

105

106

108

109

Figure 34 Density as a function of pH for ball-milled and unmilled
monolithic and ball-milled 10 VI", Sij-Si3N4 .....................................................

Figure 35 X-ray diffraction results for ball-milled monolithic Si3N4.
Sintered 4 hours. B = B-Si3N4 ..............................................................................

Figure 36 X-ray diffraction results for unmilled monolithic Si3N4.
Sintered 4 hours. B = B'SI3N4 ..............................................................................

Figure 37 X-ray diffraction results for ball-milled Si3N4 with 10 "/o
Sij. Sintered 4 hours. B = B'SI3N4, S = B' Sij, * = Y203'SI3N4 ................

Figure 38 X-ray diffraction results for ball-milled Si3N4 with 20 v/o
Sij. Sintered 4 hours. B = B-Si3N4, S = [3- Sij * = Y203OSi3N4 ................

Figure 39 X-ray diffraction results for ball-milled Si3N4 with 10 v/o
Sij added after milling. Sintered 4 hours. B = B-Si3N4, S = [3- SiCW * =
Y203OSi3N4 .......................................................................................................
Figure 40 X-ray diffraction results for ball-milled Si3N4 with 20 v/o
SiCw added after milling. Sintered 4 hours. B = B-Si3N4, S = [3- Sij, * =
Y203-Si3N4 ............................................................................................................

Figure 41 X-ray diffraction results for monolithic Si3N4 ball-milled at
pH 11. Sintered 1 hour. B=B-Si3N4. ................................................................

Figure 42 Plot showing stability of Sij with respect to time and
temperature. Adapted from Whitehead et al.[73] .............................................

Figure 43 Y203-SiOz-Si3N4 phase diagram. Adapted from Jack et al.
[110] .......................................................................................................................

Figure 44 Oz equivalents phase diagram. Adapted from [55] ..................

Figure 45 X-ray diffraction results for ball-milled Si3N4 processed at
pH 3 with 10 "/o Sij. Sintered 4 hours. B = B-Si3N4 S = [3- Sij .................

Figure 46 X-ray diffraction results for ball-milled Si3N4 processed at
pH 7 with 10 v/0 Sij. Sintered 4 hours. B = B-Si3N4, S = [3-8ij ................

Figure 47 X-ray diffraction results for unmilled monolithic Si3N4
processed at pH 3. Sintered 4 hours. B = B-Si3N4 .............................................

xi

111

113

114

115

116

117

118

119

121

123

124

125

126

127

Figure 48 X-ray diffraction results for unmilled Si3N4 processed at pH
7. Sintered 4 hours. B = B-Si3N4 ........................................................................

Figure 49 X-ray diffraction results for monolithic Si3N4 ball-milled at
pH 3. Sintered 4 hours. B = B-Si3N4 ..................................................................

Figure 50 X-ray diffraction results for monolithic Si3N4 ball-milled at
pH 7. Sintered 4 hours. B = B-Si3N4 ....................................................................

Figure 51a Micrograph taken from secondary crack found along the
fracture surface. a) shows evidence of both crack deflection and crack
bridging away from the crack tip in a monolithic Si3N4 .......................................

Figure 51b b) Micrograph shows a crack bridge in a secondary crack on a
10 V/o Sij specimen .............................................................................................

Figure 52 Fracture Toughness as a function of whisker loading and
ball-milling for specimens processed at pH 11 .................................................

Figure 53 Fracture surface of monolithic ball-milled and ultrasonicated
Si3N4 processed at pH 11 .......................................................................................

Figure 54 Fracture surface of ball-milled and ultrasonicated Si3N4 with
10 V/oSij added after milling. Processed at pH 11 ............................................

Figure 55 Fracture surface of ball-milled and ultrasonicated Si3N4 with
20 V/oSij added after milling. Processed at pH 11 ............................................

Figure 56 Fracture surface of 10 v/o Sij reinforced Si3N4, all
components ball-milled and ultrasonicated at pH 11 ...........................................

Figure 57 Fracture surface of 20 v/o Sij reinforced Si3N4, all
components ball-milled and ultrasonicated at pH 11 ...........................................

Figure 58a Image of typical porosity in Sintered specimens. Shows
elongated B-Si3N4 grains growing into and filling porosity .................................

Figure 58b Insert showing irregular B-Si3N4 grain grth within the
porosity ..................................................................................................................

Figure 59a Micrograph showing a region of two distinct grain size
regimes in monolithic Si3N4 processed and ball-milled at pH 3 ..........................

xii

128

129

130

134

135

136

139

140

141

142

143

144

145

146

Figure 59b Higher magnification micrograph of the large grain size
region. The morphology of B-Si3N4 appears much like that found in the
porous regions of other Sintered specimens ..........................................................

Figure 60a Changes in pore morphology with processing pH. Acicular
pore morphology found in monolithic specimens ball-milled at pH 3 ................

Figure 60b Typical pore morphology for all other processing conditions .....

Figure 61a Figures a-c show regions of inhomogeneity found in
specimens regardless of processing pH. a) shows either glass rich regions, or
large B-Si3N4 grain cross-sections in a 10 V/o Sij specimen ..............................

Figure 61b Shows glass rich-region in a monolithic specimen processed at
pH 7.5 .....................................................................................................................

Figure 61c Shows a large glass region associated with a large pore in a 10
"/o Sij specimen ..................................................................................................

Figure 62 Micrograph showing large B-Si3N4 grains growing from a
glass-rich region. The hexagonal morphology becomes more obvious as
grains grow larger than 1 um. The micrograph also evidences the growth of
B-Si3N4 through the liquid phase as the B-Si3N4 grain appears to be
precipitating out from a glass-rich area .................................................................

Figure 63 Schematic showing metal hydroxide interaction with the
Si3N4-solution surface ...........................................................................................

Figure 64 Comparison of density and fracture toughness for Specimens
freeze-dried with and without ball-milling skin ....................................................

Figure 65 Comparison of densification with time for specimens fieeze—
dried with and without ball-milling skin. Standard deviation for each data

point is too small to be seen clearly on the plot ....................................................

xiii

147

149

150

151

152

153

154

157

160

161

INTRODUCTION

Dry pressing of ceramic powders remains a popular method for consolidating green
bodies for sintering. However, dry processing methods are often ineffective for producing
ceramic composites. Dry milling of ceramic components provides limited homogeneity by
physically breaking apart agglomerates and mixing of components. Silicon carbide
whiskers (Sij) are a popular addition to toughen monolithic ceramics, however, dry-
milling to achieve adequate homogeneity often results in excessive whisker damage
eliminating their toughening effect. Homogeneity can be achieved by mixing composite
components in colloidal suspension and adjusting suspension conditions to provide
optimum dispersion. Whiskers can then be added to the ceramic matrix in as-received form
without sustaining damage. Ball-milling can still be employed to reduce Sij aspect ratios
to examine the effect of decreasing whisker aspect ratio on density and fracture toughness,
but ball-milling is not relied upon for composite homogeneity.

Dispersed composite suspensions are subsequently freeze-dried to remove to
remove the supernatant leaving behind a dry composite powder suitable for pressing. The
improved homogeneity provided by colloidal processing is especially impOrtant for
pressureless Sintering of Sij-Si3N4 composites. Pressureless Sintering of monolithic Si3N4

l

to assist in porosity removal. In addition, the added pressure supplied by these techniques
suppresses the decomposition reaction and allows higher temperatures to be used to
increase densification. However, these techniques are expensive and do not lend
themselves well to large scale production like pressureless sintering. The purpose of this
study is to determine whether colloidal Optimization of freeze-dried suspensions provide
adequate homogeneity of components to allow full densification by pressureless sintering.
A survey of literature indicates nearly all Sij-Si3N4 composites are produced by hot
pressing, and few attempts have been made to produce these composites by pressureless

sintering.

REVIEW OF LITERATURE

In general, a colloid is defined by particles in a suspension with a linear dimension
between 10'3 and 1 pm [1]. Smaller particles would be indistinguishable from solution, and
larger particles, common to ceramic processing, would be more highly influenced by
gravitational forces. Most of the research in colloidal theory involves examining how small
particles of one material are distributed throughout a medium, and whether these particles
remain discrete, or agglomerate. Ideas involving surface chemistry and interfacial energies
are significant to colloids because of the increased surface area of dispersed particles within
the suspension. This dependence on surface area is shown by the classification of colloids
as either lyophobic (solvent hating), or lyophilic (solvent loving), or hydrophobic and
hydrophilic when referring to aqueous suspensions. Unfortunately, these distinctions define
only extremes, and there are many intermediate states. Hydrophilic colloids are
thermodynamically stable, meaning there is a strong interaction between the medium and
the particle surface. The interactions are strong enough to break-up the dispersed phase
causing an entropy increase. Hydrophobic colloids are never thermodynamically stable,
therefore, the kinetics of coagulation is extremely important. Some may have i a highly

reactive surface, but because only a small number of surface atoms can interact with the

medium (low wetabililty), there is an increase in Gibb’s fi'ee energy. Therefore, lyophobic
colloids can remain stable only if there are strong repulsive forces to keep individual
particles fi'om agglomerating. Consequently, they are always thermodynamically unstable
and their stability, relative to ceramic processing times, is purely kinetic. Study of
hydrophobic colloids is most pertinent to ceramic processing. Lyophillic colloids are
affected by the same forces, but also by strong effects specific to the solvent (medium)
which can be difficult to predict. Therefore, we will discuss the behavior of lyophobic
colloids and how their behavior can effect ceramic processing. Issues significant to
processing include, particle size, flocculation (agglomeration), interparticle forces and
stability, as well as estimation of parameters needed to classify and predict stability of
composite ceramic colloids. Because of the “cloudy” nature of colloids, the effects of
aqueous dispersant on the surface of Si3N4 (hydrophilic interactions) should be examined
for greater understanding of a system which is not entirely hydrophobic

COLLOIDAL FORCES

There are attractive, repulsive, and steric forces which can act on particles in a
colloidal suspension. Steric forces are generated by the physical interaction of adsorbed co-
polymers on the surface of the suspended particles.

Because lyophobic particles are thermodynamically unstable, their stability relies on
mechanisms that prevent flocculation during collisions with other particles induced by
brownian motion. This can be accomplished if suspension particles have same Sign electric
charges which create a repulsive force, or by adsorbed polymer chains that physically

prevent close approach. Steric interactions, however, will not be evaluated because we will

be considering systems without additions of surfactants and dispersants. Charges
accumulate at the solid liquid interface of colloidal particles upon dispersion with each
phase taking on an opposite sign. The sources of these charges are referred to as potential
determining ions. For ionic materials, the source of ions is obvious (generally unequal ionic
dissolution in suspension medium), but the sources within other materials can be more
Obscure, usually involving ionization of particle surface groups or adsorption of ions from
electrolyte suspension. For aqueous suspensions, HI and OH' are often potential
determining ions and the interparticle potential can be controlled by changing the
suspension pH [1]. These charges impart an electrical potential on the particles which
attracts counter ions from the medium. Some counter ions become bound to the particle
surface while others remain only loosely associated with the particle due to thermal motion.
This association creates a diffuse electrical double layer around the particle which
represents the gradual reduction of potential around the particle. As electrolyte
concentration increases (source of counter ions, which ideally do not react with the particles
surface and, therefore, have no effect on determining surface potential), the number of
available counter ions increases and the double layer becomes more compact, i.e.: repulsive
interactions between particles become smaller (potential drops more quickly). Therefore,
particles must approach closer for interactions to occur and attractive van der Waals forces
become greater at smaller separation distances.

Stability can be viewed in a macroscopic or microscopic manner. The latter
involves individual collisions between particles and examining the forces involved to

determine whether a collision will result in flocculation. One in every 1010 collisions may

be considered kinetically stable. Microscopic stability will be examined more closely later,
however, it is known macroscopically, that flocculation can result from changes in pH or
electrolyte concentration [1].

Colloidal stability is dependent upon the strength of the surface charge (determined
by potential determining ions) and the size of the double layer (determined by counter ions).
Therefore, even if the surface potential is high, flocculation can still occur if the extent of
the double layer is reduced by electrolyte addition. A point of rapid coagulation can be
isolated where the addition of electrolyte causes nearly every collision to result in
flocculation. This is referred to as the critical coagulation concentration (c.c.c.). The c.c.c.
is easy to determine experimentally and is more dependent upon counter ion valency than
electrolyte concentration itself [1]. This is referred to as the Schultz-Hardy rule which
shows the sensitivity of lyophobic colloids to electrolyte concentration, the rationale for
which is called DLVO theory (named after the pioneers in colloidal study Deryaguin,
Landau, Verwey, and Overbeek). DLVO theory explains why particles tend to flocculate
unless kinetic effects are retarded. The kinetics of flocculation depend on the frequency and
energy of particle encounters and whether the thermal energy is great enough to overcome
the potential barrier between particles. This balance of attractive and repulsive forces is
measured in terms of a potential energy barrier as developed by DLVO theory for the
prediction of colloidal stability. To compute the magnitude of the potential energy barrier,

formulations for attractive and repulsive forces between particles must first be examined.

Attractive Forces

van der Waals forces are the dominant attractive forces between particles in
suspension and can be divided into 3 distinct types of attraction London, Debye, or
Keesom. These forces involve the attraction of single molecules, however, it is assumed
that colloidal particles can be treated in the same yet additive manner [1]. The potential
energy of this attraction can be written as CD=sx'lz-Bx‘6, where x is the separation distance, 8
represents the repulsion due to the close approach of atoms, and B represents the type of
attraction. The potential energy of attraction is always negative. van der Waals attraction
arises from the interactions of dipoles within an electric field produced by either a
permanent or induced dipole from another particle. London forces are the result of the
interaction of mutually induced dipoles and are the dominant attractive force between
particles in suspension unless the particles are highly polar. While van der Waals forces are
usually considered between single molecules, colloidal particles are treated in an additive
manner (cluster of molecules) resulting in long range (tens of nanometers) attraction.
Hamaker [2] developed a relation for the attractive forces without retardation between
dissimilar spherical colloidal particles with radii a, and aj, where H is the separation

distance.

 

2
Va=——’i y 2 +y 2 +210g x +xy+x (1)
12 x +xy+x x +xy+x+y x2+xy+x+y

A= Hamaker constant
X: H/(31+32)

y= al/az

The Hamaker constant can be difficult to calculate because it contains many parameters that
are hard to find or measure for materials of interest, however, substituting general values of
components, the Hamaker constant can be shown to have values around 10'13-10'l2ergs [1].
Lifshitz [3] developed a method for determining the Hamaker constant of materials in a
vacuum using macroscopic properties such as the dielectric constant or the refractive index.

This relation was further simplified by Bleier [4] to

 

2
A,(kT) z 113.7 (8",- l) (2)

(a, +1)5(e, +2);
where a is the material dielectric constant. The effective Hamaker constant between
particles in suspension is lessened by the fact that particles will attract molecules Of the
medium located between them. The effective Hamaker constant can be computed knowing
the Hamaker constant of the medium by the following formulation [1].
Aa=(Aim-AmmxAjm-Amm) (3)

Therefore, the dispersion medium has a positive effect on colloidal stability by reducing van
der Waals forces responsible for flocculation. In addition, electrostatic forces between
particles can create a repulsive force strong enough to prevent flocculation by van der
Waals forces.
Repulsive Forces

Repulsive forces between particles are the result of an electrochemical double layer
which defines the interface between isolated colloidal particles and the suspension medium.
The distribution of charge surrounding a colloidal particle is the result of two different types

of ions, potential determining ions (p.d.i.) and indifferent ions. P.d.i.s can adsorb to the

surface of particles and affect surface charge while indifferent ions have no effect on
particle potential. However, indifferent ions (supplied by the electrolyte) of opposite charge
can be found in higher concentration around the charged particle, thus balancing the charge
by surrounding the particle. This ‘sea’ of charges surrounding the particle, as well as the
adsorbed p.d.i.s form the double layer. Particle charge can be manipulated by the addition
of p.d.is from a common source, such as the addition of HNO3 and KOH resulting in H+
and OH ions and KNO3 electrolyte. The charge density and potential drop at the interface
of a colloidal particle and the medium can be described by their dependence upon the
concentration of ions, their valence, and their specific tendencies to adsorb. Therefore, it
must be understood how the potential or density of charge carriers changes as a function of
distance from the particle. This was first modeled by Guoy and Chapman as a diffuse
double layer [1].
Difi'use Double Layer

Simplifying assumptions must be made in order to mathematically model the
diffuse double layer. Ions are treated as point charges in a Boltzmann distribution,
therefore, charged particle surfaces are large relative to the ions and can be assumed to be
flat. Further simplifications include a uniform dielectric constant throughout a medium
containing a single symmetrical electrolyte of charge 2. Allowing We to equal the potential
at the particle surface, and 141 being the potential as a function of distance (x), the number of

+ and - ions per unit volume can be determined at points where the potential is w.

n+ = no exp( -7511) and n_ = no “42—273) (4)

 

no= bulk ionic concentration

10

z = ion charge number
e = electron charge

k = Boltzmann constant
T = temperature

The charge density (p) is then defined as p = ze(n+-n,). Charge density and potential are

related by the Poisson equation [5] yielding the following differential equation.

2
2
d ‘2}, = zen” sinh—Ze‘P (5)
dx 8 kT

 

This can be solved for the following boundary conditions to yield an expression for w.

Boundary Conditions include: w=wo

 

 

x=0
w=0 and dw/dx=0 at x=oo
l
\11 = 2k_T ln[ +7 exp(—1<x):l (6)
ze l—y exp(—1<x)
where
l l
__ e"I’(Ze‘I’/2kT)—1 _ [Zeznoz2 ) 5 _ [ZeZNAcz2 ) 5 (7)
y - exp(ze‘1’/ 2kT )+ 1 — skT — ekT

N A=Avagadro’s Number
c = electrolyte concentration

While this equation provides valuable insight into the interpretation of K, as
described below, the Debye-Huckel approximation is generally applied for stability
prediction because it yields relatively simple equations which allow qualitative examination
of the double layer [5]. The above combines an equation derived from Poisson (which
implies potentials associated with various charges are additive) and Boltzmann which

describes the relationship between charge and potential exponentially. This creates an

ll

inherent inconsistency in the equation, therefore, it can only be solved for limiting cases and
approximations must be made to smooth out these inconsistencies [1]. This is referred to as
the Debye-Huckel approximation. Applying the Debye-Huckel approximation allows for a

simplified function of ‘1’.

ze‘l’o
2kT

 

 

<<1 therefore: exp 99:11 z 1+ 28‘?"
2kT 2kT

resulting in...
‘1’ = ‘1’” exp(—1<x) (8)

The approximation assumes low potentials (z25.7mV) and is solved explicitly using
zie‘P=kT for monovalent indifferent ions. While the Debye-Huckel approximation is
weakened in its application by the limitation of low potentials, it is Simpler to understand
than more complex models and when these models are limited to low potentials, they
produce the Debye-Huckel result [1]. The approximation also qualitatively agrees with the
effects of electrolyte concentration and valence produced by more complex models. The
Debye-Huckel approximation also allows for interpretation of the parameter 1c.

Because 1: appears in the dimensionless parameter exp(—1<x), it must have the unit of
reciprocal length. Therefore, 161 has been called the double layer “thickness” and all
distances pertaining to the double layer are deemed small or large relative to 16'. 1c is also
concentration dependent. Equation (7) can be used to calculate the double layer “thickness”

with respect to electrolyte strength and concentration. (Table 1).

12

Table 1 Double layer thickness with respect to electrolyte concentration. Adapted from

 

 

 

 

 

 

 

 

 

 

[1]-

Molarity 2.1 1: (ml) x“(m)
0.001 1:1 1.04::108 9.61x10'9
0.001 2:2 2.08x108 4.81x10'9
0.001 3:3 3.12::108 3.20::10'9
0.01 1:1 3.29::108 3.04::10‘9
0.01 2:2 6.58x108 1.521109
0.01 3:3 9.87x108 1.01::10‘9

0.1 1:1 1.04::109 9.61x10"°
0.1 2:2 2.08x109 4.81::10'10
0.1 3:3 3.12::109 3.20::10'lo

 

 

 

 

 

 

The results show a similar thickness to that of the capacitor model [1], however, the difiuse
model is superior in its physical description and also shows how 16' varies inversely with
electrolyte strength (2) and with c”2 for a symmetrical electrolyte. These results are shown
graphically in Figure 1. Note that the curves are marked where x corresponds to K ’1.

Clearly this is not the thickness Of the double layer, however, if we examine the idea of

electroneutrality at a charged interface for the equation of potential as a function of distance

1:40

 

and assume total charge from the particle is balanced by the medium which stretches to 00,

then integration of the charge density yields

Figure 1

13

tel

 

 

 

1.0
0.8
1-1 elecrrotyte
0.0 ..
0.001 M
0.4 ,_
0.01 M
0.2 1.
0.10 M
2 o 1 1 1 i 1 1
0 1 1 1 T 1 1
1.0 (bl
0,3 .. 0.001 M electrolyte
0.6 1-
3.3 2-2 1 1
0.4 ..
0.2 -
1 I L 4 I l
2.0 4.0 6.0 8.0 10.0 12.0
x (nml

Normalized double layer potential versus interparticle separation according
to Debye-Huckel. (a) shows variation with electrolyte concentration (b)
shows variation with electrolyte valence for 10'3 M symmetrical
electrolyte. Adapted from [1].

14
o = {31111) (10)
which is the same as treating the charged surface and the diffuse double layer as the plates
of a parallel plate capacitor [1] by replacing ‘1’0 with A‘I’and d with 161 where d is the
separation distance, hence, 1c" again becomes the double layer thickness.

While inconsistencies exist in the Poisson-Boltzmann equation, they can be shown
to be minor under the conditions found in colloidal suspensions, even if the potential is not
small [1]. Debye-Huckel Shows ‘I’ varies exponentially with x, Guoy-Chapman reveals that
the complex ratio 7 varies with x. This expression can be reduced to Debye-Huckel by
assuming low potentials and expanding the exponentials. This can also be reduced to the

limiting case for high potentials yielding
‘1' = flk—T—exp(—Kx) (11)
ze

which along with Debye-Huckel bridge the true potential. Debye-Huckel, however, is the
most useful approximation when modeling colloidal ceramic systems due to the
assumptions made for greater mathematical simplicity.
Stern Theory

Limitations to the Gouy-Chapman model are realized by examining the expression

for charge density.

 

 

o ___ SkTK Sin}(ze‘l’0) (12)
21tze 2kT
This relation yields 23 ions A'2 for the case of 0.1M solution of 1:1 electrolyte [1]. This is

physically impossible and is due to the fact that Gouy-Chapman treats ions as point charges.

15

Stem addresses this problem by dividing the diffuse double layer into 2 regions, the Stern
layer and the diffuse layer [5]. (Figure 2) The Stern layer is composed of specifically
adsorbed ions and is separated from the diffuse layer by the Stern plane located
approximately one hydrated ion radius from the particle surface. Stern theory allows for
finite ion size by applying a Langmuir-type adsorption isotherm to the particle surface and
as the surface approaches saturation, the potential drop across the Stern layer becomes
constant. This sets up a situation similar to a parallel plate capacitor where the capacities of
the Stern layer (C1) and diffuse layer (C2) are given by

C1: 0" ; C2 :2"— Therefore, ‘Pd = C‘W" (13)
‘1’ — ‘1", ‘1", Cl + C2

 

 

Outside the Stern layer the double layer is still accurately described by Gouy-Chapman or
Debye-Huckel, except ‘Pd rather than Ll’o is used. While Stern theory removes the limitation
of Gouy-Chapman, it introduces parameters which are difficult to estimate, such as 1c, 4),
and 8 [5], but as far as colloidal stability is concerned, we are mainly concerned with the
outer layer. Stern actually extends the validity of Debye-Huckel by using the lower ‘Pd
potential (therefore, Debye-Huckel remains valid for higher true surface potentials because
these cases are represented by the appropriate Stern potential). This becomes a greater
advantage in stability prediction, because the potential at the Stern layer can be
approximated by the zeta-potential, which can easily be determined experimentally.
Zeta-Potential

The experimentally measured C-potential is the potential at the surface ‘of shear

(Figure 2). This imaginary plane separates adsorbed ions on the charged surface from the

l6

Compact Layer Diffuse Layer

.0

 

‘— Shear Plane

    

 

 

Distance ——+

Figure 2 Stem potential ‘Pd is evaluated at the shear plane which represents the
charged surface, adsorbed ions, and suspension medium which move as one
unit. Charged surface 1s represented by shaded spheres along the y-axis.
Adapted fi'om [5].

l7

electrolyte solution. C-potential is assumed to be slightly smaller than wd due to the effect
Of solvent which may also move with the particle within the surface of shear. This error is
ignored for lyophobic colloids and becomes pronounced only at high electrolyte
concentrations [5].

Q-potentials can be measured experimentally by electrokinetic phenomena. This is
due to the fact that when the double-layer is subject to an electric field, the charged surface
is excited to move in the appropriate direction and mobile counter ions move in the other
carrying solvent with them. The opposite is also observed. An electric field is set-up if the
charge surface is excited to move relative to the diffuse double layer. C—potential can be
measured by the relative movement of charged particles in a liquid by many different
electrokinetic phenomena including electrophoresis, electro-osmosis, and streaming
potential [5]. Because electrokinetic phenomena are directly related to the mobile part of
the double layer, it is ideally suited for calculation of the C—potential. While this does not
directly measure surface potential 100, C has an assumed identity with W which can be used
with Stern theory to predict colloidal forces present. Q—potential is most commonly
calculated from the electrophoretic mobility which is determined by elctrokinetic
experiment. Mobility is sensitive to double layer characteristics and is calculated by two
different relations, the Huckel equation for small xa and the Smoluchowski equation for
large xa [5]. Knowing the relationship between Q-potential and mobility, C-potential can be
calculated by Electrokinetic analysis (ESA) [6,7]. The ESA effect uses an alternating
electric field which generates sound waves due to the induced oscillatory movement of

suspended particles in the field. O’Brien [8] first showed how the sound wave signal could

18

be used to interpret dynamic mobility. For Ka>>1 (particle radius >> double layer
thickness), dynamic mobility is related to the static mobility and therefore, the Q-potential.
While this measurement of Q-potential is involved mathematically, it has been shown to
compare well with electrophoretic results using ceramic powders such as Si3N4 and A1203
[6,9]. Thus it is a suitable method for use in stability prediction [10].
HHF Theory

The ideas of spontaneous charge generation by particles in suspension have been
outlined. Repulsive forces generated by spontaneous charge formation act to reduce
coagulation, while van der Waals forces can result in coagulation of identically charged
particles. However, when two different species are in suspension, mutual coagulation can
occur by oppositely charged particles. A study by Hogg et a1. [12] (HHF) quantitatively
explains mutual coagulation. HHF theory predicts surface charges by using DLVO theory
coupled with the Debye-Huckel approximation, and ignores chemical interaction between
suspended species. The two components must be of similar chemical type and have the
same potential determining ions. Debye-Huckel is used to solve for the distribution of
potential as a function of distance between particles by defining them as two parallel plates
of differing potential. This result is then used to compute V], the interaction energy
between two flat double layers as a firnction of surface potential and separation distance
between plates. V, is subsequently used in the formulation presented by Derjaguin [11] for

the interaction between two dissimilar spherical particles.

V, = ['2::hV,dh (14)

19

Where h is defined in Figure 3. h is then mathematically related to the separation distance d

and the integral is evaluated to yield equation (15).

.. 8a,az(‘112m‘{1202 ) 2\I’mwoz 1"" exp(—KH0)

VR - 4(0l +02) [(Wzor +‘I’202)ln[1-exp(—KHo)] +ln(1-CXP(-2KH0))](15)

This relation holds for interacting double layers of dissimilar particles and reduces to

 

 

Derjaguin’s result for similar particles [11]. The assumptions made mathematically require
‘1"), and ‘Poz 5 25mV and the double layer thickness be small with respect to particle size.
However, experimentation revealed (above equation) to be a good approximation for cases
up to 50mV [12] and Ka>lO [13].

HHF theory uses the repulsive force (equation 15) and adds this to the attractive van
der Waals force (1) to compute a total potential energy of interaction as demonstrated in
Figure 4. This curve represents the case where interacting particles have significant (yet
still falling within the assumption of “low”) same sign potentials. Vt displays a maximum
at small separation distances which physically represents an energy barrier that must be
overcome for coagulation to occur. When potentials are very small or of Opposite sign,
there is no energy barrier and coagulation can readily occur. Therefore, Vt plays a
significant role in the kinetics of mutual coagulation (stability).

When a barrier to coagulation exists, the rate of coagulation is defined by a factor
W. Defined by Fuchs [l4] and modified by Overbeek [15] for dissimilar particles, W is the

number of collisions resulting in coagulation over the total number of collisions.

r
V, dr
Kit—)r—z (16)

 

W=(a, +aj) a]. exp

0‘“!

20

Component 2

Component 1

 

 

 

 

 

Figure 3 Simplifying parameter h as defined by Derj aguin from equation ( l4).
Adapted from [10].

21

 

 

 
   

  

 

20

15—

1o: . “c
I [9
:.»'-A :1.
'3 o

 

 

Total Potential Energy of Interaction, V,

 

 

 

 

 

 

O
-5 -4>.—
-1 0 4e
-15 {e l l
O 1 2 3
K Ho
Figure 4 An energy barrier to coagulation is necessary for suspension stability. Plot

is the result of adding Vr + Va for a stable suspension. Adapted from [5].

22

Provided the number of collisions is dependent only upon the relative number of particles
[12]. Three stability ratios are needed to characterize a system of 2 dissimilar particles.
W, 1, W22, for homocoagulation and W 1 2 for heterocoagulation, or agglomeration Of
dissimilar particles. Together, these define an overall stability ratio W,.

i.
W1:

11

__P_
W11

% + 5:— + (l 7)
Where P is the probability a collision of type 11, 22, or 12 will occur. Wilson and Crimp
[10] used a modified version of HHF theory in the development of a stability prediction
model.
Prediction of Stability

Wilson and Crimp [10] have shown that the theories of stability presented above,
can be mathematically modeled to predict conditions under which composite stability
occurs using a modified version of HHF theory. The HHF model is well suited for this
application because it includes the kinetic aspects of stability and provides quantitative
expressions for the stability of non-identical particles. These modifications include using @-
potential to estimate the surface potential of suspended particles (by applying Stern theory
to characterize double layers) and using Bleier’s method of determining the Hamaker
constant for ease of application to different composite ceramic systems. Wilson and Crimp

showed success applying their model, Suspension Stability©, to predict the stability of Si3N4

and SiC powders in aqueous suspensions.

23

Once a suspension of appropriate ceramic components has been stabilized, ceramic
components must be removed from suspension and formed to near net shape. One method
for accomplishing this is freeze—drying.

FREEZE-DRYING

Freeze-drying is used to remove the water from disperse composite suspensions and
leaves behind a dry, intimately mixed powder suitable for pressing [l6]. Drying of ceramic
powders at elevated temperatures can cause changes in powder characteristics (i.e.: surface
groups present, oxidation) and can bring about the formation of hard agglomerates which
require milling and sieving to remove. Powders containing hard agglomerates rarely press
to homogeneous green compacts [l6]. Freeze-drying is a simple alternative to this
procedure. The freeze-drying process can be divided into 3-steps; (1) pre-freezing (2)
primary drying and (3) secondary drying [17,18].

Pre-freezing

Because freeze—drying involves the sublimation of water from the suspension, the
suspension must first be adequately frozen to ensure success of the freeze-drying procedure.
The major volume of a suspension to be freeze—dried is the supernatant (water and an added
amount of acid or base). The acid or base will act to change the freezing point of the water
and forms a eutectic mixture. As the water freezes, solute is rejected from the freezing
water and forms pockets of liquid with a fi'eezing point lower than 0°C. Pockets of eutectic
liquid must all be frozen, or they will instigate melting under the low pressure of freeze-
drying conditions. Pre-freezing is usually carried out in a shell freezer [18]. Shell freezing

involves fieezing a suspension to the walls of a cylindrical vessel which increases the

24

surface area to volume ratio of frozen suspension. This acts to improve the efficiency of
primary drying-
Primary drying

After pre-freezing is complete, conditions must be quickly established to promote
sublimation of aqueous suspensions before melting begins. Temperature and pressure must
be carefully adjusted to keep the suspension within the sublimation range, but not so cold
that sublimation occurs too slowly. Efficient freeze-drying takes place just below the triple
point, or lowest freezing eutectic, low enough to avoid incipient melting, and high enough
to establish a significant vapor pressure above the surface of the frozen suspension.
Increasing vapor pressure encourages sublimation from the frozen suspension to the
condenser. By keeping the condenser pressure as low as possible, a vapor pressure
differential is created and vapor moves from the suspension to the condenser. The flow of
water vapor from the suspension to the condenser is greatly improved by placing the entire
system under a vacuum. A vacuum pump also acts to remove non-condensable gases from
the system.

Because it takes ten times the energy to sublime water as it does to fi'eeze it, energy
needs to be added to the suspension to keep the temperature gradient with the condenser as
strong as possible. Leaving the cylindrical vessel at room temperature provides enough
heating to counteract the evaporative cooling that takes place during sublimation. This
becomes most important as the drying front moves through the thickness of the frozen shell.
Water molecules must travel through the dried layer above, therefore, driving force must

remain high in order for the water to escape to the condenser.

25

Secondary drying

Secondary drying occurs after the suspension appears dry. Upon completion of
primary drying, the material may still contain up to 8% residual moisture [18]. Secondary
drying occurs by isothermal desorption of bound water from the ceramic powder. The
outside temperature of the vessel is raised while keeping pressure and condenser at the same
levels. Secondary drying is carried out for 1/3 to 1/2 the time of primary drying. While
freeze—drying offers an effective way of preparing mixed composite powders, little research
has focused on the topic.

M.W. Real [16] demonstrated freeze-dried ceramic powders could be pressed to
form homogeneous compacts. At water contents of 4:1 by volume and greater, freeze-dried
powders form a fragile plate-like structure which breaks down easily by hand into a fine
powder suitable for processing of engineered ceramics.

The properties of engineered materials, especially engineering ceramics, is
dependent upon its internal flaw structure. This internal structure is indicative of the
processing of the material which controls the internal pore structure. The distribution of
pores can, in turn, have an effect on grain structure and second phase distribution within a
composite. Therefore, control of final properties involves control over grain size along with
pore and second phase distribution which all requires knowledge of how pores are removed
from a densifying compact. How green structure effects sinterability is addressed in the

next section.

26

LIQUID PHASE SINTERING

Sintering in the presence of a liquid provides enhanced densification at lower
temperatures than free Sintering [19]. Disadvantages arise from the fact that there are many
interfaces and energies of individual components to consider in liquid phase Sintering, as
well as solubilities of phases involved and viscosity of the accompanying liquid. Therefore,
analytical treatments of liquid phase Sintering tend to be more qualitative in nature than
those associated with solid state Sintering, and applications of liquid phase Sintering also
tend to be less predictable in nature. Liquid is provided by the addition of a second phase
material with a lower melting (or glass forming) temperature than the matrix. Liquid phase
Sintering is characterized by three distinct over-lapping steps [20,21] as shown in Figure 5.
When liquid forms, interparticle fiiction drops and capillary forces accelerate particle
rearrangement. Densification by rearrangement is rapid and depends upon particle size,
green density and the amount of liquid present. Solution-reprecipitation occurs
concurrently, but is overshadowed by rearrangement during the initial stages of sintering.
In general, the solubility of a grain is inversely proportional to its size. Therefore, small
grains preferentially dissolve in the liquid setting up a concentration gradient and material is
precipitated on to larger, less soluble, grains. This is termed "Ostwald ripening" and
contributes to densification as well as grain growth. The last stage of persistent liquid phase
Sintering involves solid state Sintering mechanisms. While solid state Sintering is occurring
throughout the entire process, its effects dominate only upon completion of the first two

steps. This stage is characterized by grain coarsening and either pore elimination or growth

27

 

Amount of Sintering

 

 

 
  
      
 

/, e ’ //
4517 1 '1 1
’41? ,7," ’/-. 5

//.<”/ , ’ 1’1
4‘“ ’14

4444

          
       

 

 

 

 

 

Time

 

 

Figure 5 Evolution of a green ceramic compact Sintered with the assistance of a liquid
phase. I = initial stage, II = intermediate, and 111 = the final stage of
sintering. Adapted from [21].

28

depending upon entrapped gases within the pore structure. Prolonged final stage Sintering
may degrade the final properties of liquid phase Sintered components [21].
Initial Stages

While this is the shortest of the 3 stages, rearrangement often results in significant
densification. Microstructural changes occurring during the first stage include melting and
solid-liquid solubility, penetration and fiagmentation, and rearrangement. These steps,
however, may be more difficult to characterize in a more complex ceramic system where
viscosity, interfacial energy, and the amount of liquid are continually changing.
Intermediate compounds may also form and change the quality of the liquid present.
Despite these shortcomings, the classic models of liquid phase Sintering can be used to
characterize successful Sintering systems.

Melting and Solubility

The solubility ratio between the additive and matrix is an important aspect of initial
heating during liquid phase sintering. High solid solubility in the liquid improves
rearrangement by surface smoothing thus reducing interparticle fiiction. A low solubility
ratio is characterized by pore formation at the prior additive particle sites. This greatly
reduces rearrangement and can cause compact swelling rather than densification. In
general, the greater the solid solubility, the greater the final density and the less additive
needed to achieve final density [21]. As temperature increases, the liquidus moves closer to
the solid composition and the liquid can hold more solid in solution thus increasing the
amount of liquid present. Rearrangement can also be enhanced by using smaller additive

particle sizes which can improve distribution of the additive within the compact thus

29

forming a more homogeneous liquid phase. Large additive particle sizes are detrimental to
final pore removal, and it is common to ball-mill additive and matrix particles together for
improved homogeneity, reduced particle sizes, and to break apart hard agglomerates for
improved packing [20]. After the initial heating stage where diffusivity and solubility
dominate Sintering behavior melt formation begins.

During melt formation, the most important characteristic of the liquid is how well it
wets the solid. A wetting liquid will spread through the compact by capillary forces and the
major portion of densification that can occur by rearrangement will occur within minutes of
liquid formation. Of course, this rate is limited by the thickness of the liquid layer and the
viscosity of the liquid. Poor wetting liquids are evidenced by liquid exuding from the
compact surface [21]. The compact will remain intact while the liquid forms a bead on the
exterior. Further densification is dependent upon the solubility between solid and liquid
causing particle smoothing and reduced interparticle fiiction allowing for easier, faster

particle rearrangement [21]. Both wetting and high solid solubility are important liquid
characteristics and necessary for complete densification.

Penetration and Fragmentation

Penetration describes the movement of liquid through pore and grain structures

Within a compact driven by either capillary or reactionary forces. This driving force is
Strong initially because the liquid has not yet achieved an equilibrium composition and,
therefore, has a high curvature which aids solubility and allows the liquid to break solid-
solid interfaces [21]. Penetration is dependent upon liquid viscosity and reactivity [as well

as 0n dihedral angle. This is the angle made when the liquid intersects a solid-solid

30

interface. This angle is dependent upon the ratio of solid-solid to solid-liquid interfacial
energy. While dihedral angle is important, this work centers around a current additive
system, so good wetting conditions are assumed. F ragrnentation can result in a grain
structure smaller than the starting particle size when liquid wets grain boundaries and
breaks apart agglomerates. Compact swelling can occur during fiagmentation, but later aids
densification by secondary rearrangement and reducing the overall particle size which,
again, improves rearrangement and will aid the solution-reprecipitation process.

Rearrangement

The formation of a wetting liquid creates attractive forces between particles and
causes a hydrostatic pressure on existing pores. Influenced by reduced interparticle friction
due to particle smoothing, particles repack to a higher coordination. Rearrangement can be
broken down into two stages, primary and secondary. Primary rearrangement involves
individual particles while secondary rearrangement involves clusters or fragments of
individual particles. Secondary rearrangement becomes important when liquid is unevenly
distributed throughout the compact. This uneven distribution causes clusters of primary
particles which may or may not fragment. As these clusters grow and repack they release
liquid to the intercluster region. Kingery [20] showed that time is not a significant factor in
rearrangement and later demonstrated this experimentally with oxide and carbide-metal
systems [22]. More important factors include volume fraction liquid, particle size, dihedral
angle, and green density.

Liquid present in 35V/o can facilitate full densification by rearrangement [20], but

most systems are kept below 20 ”ID to avoid shape distortion, therefore, other processes such

31

as solution-reprecipitation must be active in order to achieve full density. Rearrangement
processes are less efficient with non-spherically shaped particles. Kingery showed [20] that
capillary forces vary inversely with particle size. Experimental results confirm that small
particle sizes are beneficial for rearrangement [23,24]. High green densities create more
mechanical interlocking between particles and have less vapor phase present.
Consequently, the capillary force for rearrangement is reduced. In general, high green
densities reduce the densification rate, but increase the final density of the compact due to
the lower initial porosity. Rearrangement consequently must change this initial pore
structure for densification to occur [25].

The pore structure changes in the initial stages of Sintering due to the formation of
new pores at former additive particles sites, and pore coarsening due to rearrangement.
Capillary action draws liquid from the site of the additive particle leaving a pore behind.
(Another reason why small additive particle sizes are important.) Pore refilling can occur
only if glass composition remains consistent during prolonged Sintering [21]. As the liquid
spreads and capillary forces activate, new contacts are made which increase the rigidity of
the compact. It's at this point the initial stage processes slow and the intermediate stage of
liquid phase Sintering dominates.

Intermediate Stage: Solution-Reprecipitation

While solutionizing and reprecipitation occur simultaneously with rearrangement,
their effects on the compact do not dominate until rearrangement is complete. The
intermediate stage of liquid phase Sintering is characterized by grain shape accommOdation,

densification, coalescence and neck growth, and pore filling. There must be an appreciable

32

solubility of the solid in the liquid additive for Sintering in the intermediate stage to be
successful. Grain growth also occurs concurrently with the intermediate stage, but it is
generally treated as the final stage of liquid phase sintering.

Grain shape accommodation

Because most useful liquid phase Sintering systems contain small amounts of liquid
sZOV/o, there is still porosity within the compact after rearrangement. Grain shape
accommodation acts to reduce this porosity. Grain shape accommodation occurs by
dissolution of small particles and dissolution at the contact points of larger particles.
Dissolved material reprecipitates away from contact points [21,26], thus altering particle
shapes and acting to fill pore space. Grain shape accommodation is a mechanism for both
densification and grain growth [27].

Densification

The rearrangement process leaves behind densely packed particles surrounded by a
thin layer of liquid. The liquid at contact points causes capillary pressure between particles
creating a compressive load within the compact which Kingery [20] has shown increases
the solid solubility at those points. This results in material transfer away from the contact
points which causes particle centers to approach one another resulting in densification. This
is termed contact flattening and is one of three mechanisms used to describe densification of
liquid Sintered compacts. A second mechanism involves dissolution of small grains and
reprecipitation onto larger grains, Often resulting in grain shape accommodation [27]. A
third mechanism involves solid state diffirsion. A compact with small amounts of liquid or

a partially wetting liquid [28], may rely on solid state diffusion for additional densification,

33

however, solid state diffusion is much slower than diffusion through a liquid and its effects
are apparent only in the later stages of sintering. The effects these mechanisms have on the
compact are summarized in Table 2.

Table 2 Effect of Sintering mechanism on a powder compact. Adapted from [21].

 

 

 

 

 

 

 

Factor Contact Dissolution of Neck Growth
Flattening Fines
Material Source contact zone small grains grain boundary
Transport Path liquid liquid solid
Transport Rate rapid rapid slow
Shape Accommodation yes yes yes
Grain Coarsening no yes no
Solubility in liquid is is not required
required required

 

 

 

 

 

 

There are differing opinions on which of the densification mechanisms is dominant.
Observations by Kaysser et a1. [29] found densification to correlate with the onset of rapid
grain growth. Examination revealed a direct correlation between densification and
dissolution of fine grains [27]. According to research on glass phases in ceramics
[20,30,31] the liquid width between ceramic grains is only a few atomic diameters.
Therefore, contact flattening would not be a dominant densification mechanism for these
systems. Regardless of the dominant mechanism, shrinkage is proportional to time“3
[2 3,32]. Shrinkage is enhanced in the intermediate stage by high solid solubility, small

particle sizes, and longer times in the intermediate range. Temperature effects are most

34

strongly observed through enhanced diffusivity, although it may have an effect on solubility
and surface energy, especially in mixed component ceramic systems [21].

Neck growth and coalescence

At the end of the intermediate phase of Sintering contacting particles show a stable
neck with size dependence on dihedral angle, grain size, and volume fraction solid present.
Neck growth occurs by material transfer from either the grain contact point, which results in
densification, or the grain surface which results in no densification. Rate ‘of neck growth
depends upon the mechanism involved, either coalescence, solid state diffusion, or solution-
reprecipitation. Neck growth can be a detriment to densification because if a continuous
solid particle structure is formed, densification will slow, ending the intermediate stage.

Coalescence describes the fusing of two dissirnilarly sized grains into a single grain,
most commonly by grain boundary movement [21]. Coalescence is used as a mechanism to
describe grain growth and the reduction in grain number during sintering. Coalescence is a
function of the number of grain contacts and is therefore more prevalent in compacts with
high green densities. Coalescence of particles can also contribute to densification by filling
the neck region between two spherical particles in contact [26].

Pore filling

While grain shape accommodation assists pore filling at the beginning of the
intermediate stage of Sintering, pores formed at former additive sites may exist throughout
most of the intermediate stage. The coarser the additive particle size, the longer the
incubation time for filling (i.e.: small pores fill first). Pore filling requires grain growth and

accommodation. In order for pore filling to occur, the liquid meniscus radius must exceed

35

the pore radius. Therefore, there is a critical grain size for pore filling [33] and prolonged
Sintering may be required for grains to grow large enough for the liquid to reach conditions
favorable for pore filling. However, as pores fill, the effective liquid content of the system
is reduced and it becomes harder to fill succeeding pores.
Final Stage Sintering: Grain Growth

Because densification and microstructural coarsening proceed by similar
mechanisms, it is important to understand coarsening in order to achieve full density but
maintain rnicrostructural control. Further densification in the final stages of Sintering is
highly dependent upon pore characteristics and entrapped gasses within the pores. Because
pore structure typically scales the particle size, smaller particle sizes should leave behind
smaller, easier-tO-remove pores. Sintering in a non-soluble gas atmosphere increases the
probability of trapping gasses within the pore structure. Pore elimination is further
inhibited by the resulting continuous solid structure due to neck formation between solid
grains within the compact, typical of final stage sintering.

Continued densification

Densification can continue despite the skeletal nature of the powder compact via
continued solution-reprecipitation driving grain shape accommodation for final pore
removal. This is possible only if inhibiting factors such as entrapped gasses, decomposition
products [34], gross imperfections in packing [35,36], and reaction products with the
atmosphere [23] can be avoided. Another source of entrapped gasses is from
decomposition of the compact material. Release of vapor fiom reaction produCts can

produce increasing vapor pressures within the pore structure and cause swelling.

36

Grain growth

While grain growth is occurring along with densification, the final stage of Sintering
is characterized by the dominance of grain growth. The driving force for grain growth in
liquid phase Sintering is the decrease of interfacial energy due to the decrease in particle
curvature and subsequent reduction of the solid-liquid interface. Grain growth, as with
densification, can be diffusion controlled or reaction controlled and again with multi-
component systems tends to be reaction controlled. With either controlling mechanism, the
rate of grain growth is increased by increasing the volume fraction solid and raising the
temperature [21,37]. Grain growth can also be inhibited by the presence of a second
immiscible solid phase such as SiC whiskers in Si3N4 [38,39].
Processing Considerations

Together with the Sintering mechanisms discussed previously, processing concerns
determine the Sintering behavior of a powder compact. A major concern of industry
involves dimensional control [21]. Shrinkage is high when a liquid phase is present during
Sintering, but dimensional control is improved for compacts with a high green density.
Homogeneity is also important to maintain equivalence of dimensional changes from
compact to compact.

Particle characteristics

In general, small particle sizes have an advantageous effect on sintering. A narrow
particle size distribution is also beneficial [19]. A large particle size distribution or
aggregates within the starting powders can induce exaggerated grain growth [19,21] or

create gross packing deformities [35,40] resulting in abnormally large pores. Grain shape

37

can also have an effect on compaction. Spherical particles Show better rearrangement
because more capillary pressure can be developed between particles and spheres leading to
greater compact uniformity. Irregular powders can increase the green strength, but show
reduced rearrangement and densification due to greater interparticle friction as well as more
non-uniformity in green density [41].

Green properties

High green densities reduce rearrangement due to increased interparticle friction.
Green density plays a stronger role in Sintered density as the amount of grain boundary
liquid decreases. Increased green densities also improve dimensional control, but because
shrinkage varies directly with green density, non-uniformities in density can lead to uneven
shrinkage. Because initial stage Sintering effects are reduced by increasing green densities,
effective intermediate stage Sintering becomes more important.

Heating and cooling rates

Diffusional homogenization can have a detrimental effect on liquid phase sintering.
Homogenization of the liquid composition due to slow heating rates can affect wetting
characteristics and in extreme cases, suppress liquid formation altogether [21]. In general,
rapid heating rates improve densification, while slower heating rates and isothermal holds,
below liquid formation temperatures prove, detrimental [42,43]. However, if heating rates
are too rapid, they can lead to entrapped porosity. While cooling rates are rarely optimized,
post-Sintering treatments can be employed to improve mechanical properties (i.e.: anneals to

induce grain boundary crystallization) [44]. The greater the cooling rate, the greater the

38

saturation of solid in the liquid phase and the greater the Sintered strength. However, rapid
cooling at the solidification temperature can cause solidification porosity [21].

Temperature and Time

Liquid phase Sintered materials Show a strong temperature sensitivity. As
temperature increases, diffusion rates increase, wetting and solubility increase, viscosity
decreases, and the amount of liquid increases. Time and temperature must be optimized in
order to avoid excessive grain coarsening, harmful decomposition reactions or intermediate
compound formation which can reduce the amount and viscosity of available liquid. This is
especially important for pressureless Sintering of Si3N4. Increasing temperatures above
185 0°C enhances decomposition and inhibits densification mechanisms [45]. Most
densification models for liquid phase Sintering Show density to be a weak function of time
and increased Sintering times often lead to degradation of mechanical properties [21].

Atmosphere

The proper sinterirrg atmosphere acts to reduce formation of harmful films, such as
oxides, or as is the case with Si3N4, suppress decomposition. Insoluble atmospheres can
reduce densification rates when gasses become entrapped in the porosity. Densification
rates are then controlled by the diffusivity of the gas within the matrix [45]. Sintering in a

vacuum can eliminate this problem, but will not suppress decomposition reactions.

39

LIQUID PHASE SINTERING OF Si3N4

Sintering of Si3N4 to near theoretical density often requires the assistance of liquid
phase Sintering along with hot pressing or HIPing [46-49]. Hot pressing and I-IIPing can
produce dense high quality Si3N4 components, but lack the versatility pressureless Sintering
lends to automation and the production of complex shapes [50]. Si3N4 is difficult to densify
because of its low thermal conductivity and strong covalent bond structure giving it a low
self-diffusivity [44]. Self-diffusivity cannot be improved by increasing the Sintering
temperature because decomposition of Si3N4 becomes a problem above 1825°C [34,51].
Therefore, higher temperatures cannot be used to boost the diffusivity without the addition
of a high pressure N2 gas environment [52]. Diffusivity, and thus densification, can be
enhanced at lower temperatures by addition of a second phase which forms a liquid at
Sintering temperatures. Typical additives, or Sintering aids, to Si3N4 include MgO, Y203,
and Y203+A1203. While other Sintering aids have been examined (ie:Ce02-, La203, Sm203
[44,51]), the former remain the most popular currently in use. MgO was one of the first
Sintering aids used in pressureless Sintering of Si3N4 [45]. Work by Mitomo et a1. [53]
indicated Si3N4 could be pressureless Sintered to >90% theoretical density with additions of
5% wt. MgO with strengths that rivaled hot pressed Si3N4 with the same additions. Studies
by Terwilliger [54] yielded similar densities but indicate substantial strength degradation at
temperatures above lOOO°C. Lange [54] associates this degradation with amorphous grain
boundary phases. Reduced crystallinity is due to the lack of Mg solubility within B-Si3N4,
and Ca impurities which lowers the softening temperature of magnesium silicate along

grain boundaries [56]. Additions of Y203 alone show improvements over MgO, with

40

greater high temperature strengths. Densification of Si3N4 with MgO additions is highly
dependent upon impurities present within the Si3N4, therefore, densification of high purity
Si3N4 requires hot-pressing to achieve high densities [57,5 8].

Work by Gazza indicated Si3N4 doped with Y203 could be hot-pressed to high
densities (3.26g/cm3) comparable to those with MgO additions, but with high temperature
strengths due to the formation of refractory phases, such as yttrium silicates, which yields
less residual glassy phase at grain boundaries [56,59]. Si3N4 materials hot pressed with
Y203 can exhibit poor oxidation resistance at intermediate temperatures. Weaver et. a1. [60]
showed these problems to be associated with high impurity levels in the Si3N4 starting
powder. Weaver demonstrated high purity Si3N4 with <8% wt. YZO3 to have excellent
oxidation resistance, mechanical properties, and high temperature strengths, when
compared to Si3N4 with MgO additives. Both Y203 and MgO, however, have shown
limited success for pressureless Sintering of high purity, low 02 content Si3N4.

Much of the present research involves additions of both Y203 and A1203 to Si3N4.
The addition of A1203 lowers the glass forming temperature making it possible for
pressureless Sintering of Si3N4 to high densities below the decomposition temperature.
Work by Loehman and Rowcliffe [59] with Si3N4-Y203+A1203 results in shrinkage curves
that display the three stages of liquid phase Sintering defined by Kingery [20] with one
exception. Their results show enhanced shrinkage in the intermediate stage of Sintering,
which they attribute to improved solid-liquid reactivity due to the phase transition from
either or or amorphous Si3N4 to B-Si3N4 during Sintering [59]. By drawing upon the

mechanisms of liquid phase Sintering, and current research on the Si3N4/Y203+A1203

41

system, the specific characteristics Of Si3N4 Sintering can be examined. These include
properties of the main constituent, such as, impurities, and morphology, along with
compatibility with Sintering aids and effect on decomposition which can be inferred from
mass loss during sintering. Addition of Sij will also have an effect on Sintering behavior.
Generalities which effect all Sintering systems, such as particle size and degree of
agglomeration have been discussed earlier.
Powder Morphology

Si3N4 exists in two crystalline forms, 01 and B. It has been argued that a-Si3N4 is not
a true nitride, but an oxynitride with composition Si23N3OO [61]. However, evidence also
exists showing a-Si3N4 to be purely Si3N4 [62], with z 0.05% 02 as an impurity. The
relationship between 01 and B is not well understood, partly due to the fact that the B—m
transformation has never been observed. 01 appears to be stable at low temperatures and in
high P02 regions, while B appears stable at any temperature. The structure of the two phases
is represented by Figure 6, where B consists of repeating AB layers and 01 consists of
ABCD repeating layers. Because AB and CD are mirror images, 01—)B transformation must
occur by breaking a Si-N bond and occurs only through a gas, or liquid phase [61]. a-Si3N4
is especially suited for Sintering because of its higher 02 content and the fact that it is
thermodynamically unstable with respect to B-Si3N4 at Sintering temperatures [64], which
improves dissolution of Si3N4 in the liquid phase. Ichicawa [63] reports no improvement in
densification as a function of a content, but regardless of whether 01 content improves

densification, it has been shown to improve the final strengths of Sintered pieces [64]. If B-

42

 

Figure 6 Repeating atomic unit layers for Si3N4. Adapted from [63].

43

grains are present in the starting powder, there will be a continuous growth of globular
shaped B at the expense of 01, if few or no B grains are present, the liquid phase must first
reach a supersaturation to nucleate rod-like B grains.

The mean aspect ratio, 5 , of B grains after sintering can be estimated by
characteristics of the starting powder (21' =1+01 /B) [55]. High aspect ratio B-grains act as an
in-situ form of strengthening the material by crack deflection and additional fracture energy
consumption [65-67]. While powders containing pre-existing B, despite their lower green
densities, can also be densified, they demonstrate reduced mechanical properties due to their
equiaxed microstructure. They are also more sensitive to Sintering temperatures as shown
in Figure 7.

Impurities in Si3N4

Commercial Si3N4 is produced by four major processes, nitridation, vapor phase
deposition, carbo-thermal reduction, and diimide precipitation [64], each producing a
powder with unique sizes, morphology, and impurities. The focus of this study concerns
Si3N4 powders produced by diirrride precipitation. These powders (along with those
produced by CVD) have fine particle sizes with very low impurity levels compared to the
other powders. Typical grades, such as Ube’s E- 1 O, are characterized by subrrricron
equiaxed particles of extremely uniform particle size and morphology [61]. Si3N4 produced
by diimide precipitation contains trace levels of Cl remaining fiom the $0., which is not
present in the other powders mentioned. Impurities commonly found in Si3N4, such as Ca,
Na, and K [56], stimulate glass formation at elevated temperatures. Residual levels of

carbon act to reduce liquid formation by reacting with oxygen and SiOz to form CO or SiO

44

 

 

 

 

 

 

Density (g/cma)
3.2
2.85 —l
2.5—;
g
l /
2.15-— /
’ /
- e
F g / /
1.8 ’ 1 1 1 1 1 - 1
1400 1500 1600 1700 1800
Sintering Temperature °C

 

 

 

Figure 7 Plot showing the sensitivity of or and B powders to Sintering temperature. B
powders are more difficult to Sinter at low temperatures. Adapted from [64].

45

which volatilize at high temperatures to reduce sinterability. Other impurities include, Mg,
Al, and Fe, which are dissolved by the liquid phase and act to increase the amount and
reduce the viscosity of the liquid phase [64]. While these impurities have a positive effect
on densification, they have a negative affect on mechanical properties, especially high
temperature properties. Therefore, while diimide particles have a lower affinity for liquid
production, their mechanical properties should be superior if densification is complete. The
most important impurity present in Si3N4 is oxygen.

As the amount of Sintering aid is reduced, equiaxed powders with high surface areas
are needed to produce high densities. Reactions with oxygen present in the form of Si02 at
the surface and as a solid solution within the (It-phase, assist formation of the liquid silicate
phase. This is shown in Figure 8. High oxygen contents assist densification of coarser
powders, but low oxygen content powders can be Sintered to high density as particle surface
area is increased [64]. Just as with other impurities, however, the best mechanical
properties are found with low 02 containing powders.

YZO3+Ale3 as Sintering Aid for Si3N4

Y203+Ale3 are the most common Sintering aids used with Si3N4 because Si3N4 is
soluble in many of their products and their glassy phases can be crystallized for added high
temperature strength [68,69]. A1203 lowers the softening temperature of oxynitride glasses
[59], therefore allowing more effective liquid phase Sintering by lowering liquid viscosities
and allowing pressureless Sintering at temperatures below decomposition. Al is present in
the residual glassy phase of Sintered specimens, or in solid solution with B- Si3N4 as B’-

Si3N4 which contains Al substitutions for Si and O substitutions for N [44]. Y cannot

46

 

Theoretical Density
100%

 

High 0,

  

95% -

 

 

 

90%n
. / Low 02
/
/
/
/
r / 20
85% r— / / /
1r /
80% r 1 1 r { 1 1 1 1 :
10 15 20 25

Specific Surface Area (ml/g)

 

 

 

Figure 8 The effect of 02 content on the densification of Si3N4 powders with respect
to surface area. Adapted from [64].

47

replace Si within the B phase, and most of the Y203 present is found in crystalline grain
boundary phases containing no Al. Therefore, Al can be used to lower glass forming
temperatures and viscosities, then form solid solutions with B- Si3N4 and YSiOzN thereby
reducing the residual glassy phase present after liquid phase sintering.

Oxynitride glasses begin to form within the compact at temperatures as low as
1350°C [59] by reactions onZO3-A1203-Si02 (Si02 found on the surface of Si3N4 [70]), as
well as N2 from the Sintering atmosphere. At temperatures above 1550°C, Si3N4 can
dissolve in the molten glasses present and several different second phase glasses and
compounds can be formed depending upon the relative amounts of each of the constituents
[71].

The dissolution rate of Si3N4 is dependent upon particle size and crystallinity.
Small (<1 micron) on and B particles dissolve rapidly and reprecipitate in the B form. Due to
their relative instability, 01 particles dissolve more readily forming a supersaturation within
the liquid and form idiomorphic rod-like B grains [64]. B present in the starting powder,
dissolves under nearly equilibrium conditions and forms spherical B grains [64].
Amorphous Si3N4 particles dissolve even more rapidly due to the free energy difference in
the amorphous-crystalline transition and they also form rod like B-Si3N4 [59]. The driving
force for these reactions emanates from free energy differences between reactants and
products due to the formation of more stable compounds and reduction of solid-liquid
interface.

Because the content of the liquid is continually changing, it is almost impossible to

know the composition of the liquid at any given time. Grain boundary phases are difficult

48

to examine after Sintering, because they are often too thin to be examined by x-ray or
electron probe rrricroanalysis. Loehman [59,71] and Rowcliffe [59] compare relative
amounts of Sintering additives to a study [71] of oxynitride glasses. By assuming that the
oxygen content within the starting Si3N4 reflects the SiOz content and a moderate residual
nitrogen content of 6 at.% [59], predictions can be made about composition and volume
percent of second phase present by examining the glass forming region of Y-Si-Al-O-N
glasses. Therefore, relative levels of additives and 02 present in the starting powder affects
the amount and quality of the liquid phases formed. Increasing amounts of equivalent
liquid will increase the densification rate, but generally have little effect on the final density
[59]. Cirribulk et al. showed direct addition of YSiAlON glasses allows for more control
over residual glassy phases and yielding high Sintered densities [44]. Examination of
successful Si3N4-YZO3-AIZO3 Sintering systems should provide insight to applying Y203
and A1203 to diimide derived Si3N4 powders.

Crosbie et a1. [42] studied diimide derived Ube E-lO Si3N4 with additions of 10
wt.% 31,03 and 2.26 wt.% .41203 yielding final densities from 2.35-3.19 g/cm’ (swam
theoretical). They discovered density to be highly dependent upon heating rate from an
intermediate hold, as well as on the temperature of intermediate hold. Sintering time
proved to be only a minor factor. The highest densities resulted fiom an intermediate hold
at 1700°C with a 600°C/hr. ramp to Sinter at 1800°C for 2 hours with a sample mass loss of
5 0.5% total mass. Ichikawa [63] examined densification of Si3N4 as a firnction of Y203
and A1203 content. Sintering at 1750°C for 4 hours and keeping A1203 content at 2 wt.%,

he discovered density to decrease with decreasing Y203 content for several different

49

powders. The decrease was attributed to a decrease in the amount of liquid causing grain
morphology, particle size distribution, and agglomeration to play a more important role in
densification. When A1203 was varied from l-3wt.%, no changes in density were found.
Cirribulk et. al. [44] found optimum densification to occur at 1825-18500C, but used 1.5
MPa N2 to minimize decomposition at these temperatures using 5 wt.% Y203 and 5.5wt.%
A1203 in an effort to form YAG at grain boundaries.
SiCM Additions to Si3N4

By the same mechanisms acicular B-Si3N4 grains in-situ strengthen the matrix,
additions of Sij can act to strengthen Si3N4, but their existence in the starting powder
affects sinterability of the composite. Sinterability is reduced by the presence of Sij (even
if green density problems are avoided) by increasing diffusion distances between
solutionizing and reprecipitating Si3N4 [72]. Whisker additions have also been shown to
inhibit the 01—)B transition in Si3N4 [58]. Ueno et al. showed this delay to be a function of
Sintering temperature with the retarding effect becoming smaller with increasing
temperature. Therefore, longer Sintering times or higher Sintering temperatures are required
to provide adequate amounts of B to achieve high densities by pressureless sintering.
Unfortunately, longer Sintering times result in stronger interfacial bonding between matrix
and whiskers and toughness is not improved over monolithic values [58].

SiCw maintain structural integrity during Sintering [73] and, therefore, must
rearrange to accommodate the shrinking compact. This rearrangement interferes with
shrinkage at 20-30 vol.% Sij because SiC is insoluble within the liquid phase [[74].

Nearly all examples of consolidating Si3N4/Sij composites involve 111ng or hot pressing

50

along with the addition of liquid Sintering additives. Hot pressing of Sij composites leads
to alignment of whiskers creating an anisotropic material. Densities near theoretical have
been reported by hot pressing and 111ng whisker composites [3 8,39,48]. Grain boundary
phases present can have a strong effect on interfacial energies and, therefore, on fiacture
toughness. Weak interfacial bonding can result in more whisker pullout and crack
deflection around whiskers, both of which act to improve fiacture toughness. Studies of
Sij/Si3N4 interfaces [48] indicates residual glassy phase can be 5 times thicker around
SiCW than between Si3N4 grains. It’s suggested that this is due to increased amounts of
Si02 on the whisker surface which leads to reduced fracture toughnesses due to improved
interfacial bonding. Smith et al. [7 5] showed fracture toughness of whisker toughened
alumina was reduced by vacuum hot pressing rather than pressing in Ar due to the presence
of 02 under vacuum conditions. High surface 02 contents have been shown to degrade
high temperature properties of composites during processing or in service unless whiskers
were etched or coated to reduce the influence of SiOz [48].

Mass Loss (Suppression of Decomposition)

Mass loss is a function of specific surface area and impurities present in the Si3N4
powder. Increased surface area provides a greater area over which vaporization reactions
can occur. At temperatures 5 1850°C reactions with Si02 and metal oxides account for
most of the mass loss in Si3N4 [55].

Si,N4 + 3 3102 —> 6SiOT + 21121

$1,144 + 3/a M20, —+ 35101 + N21 + 2/a M3N,

51

Mass loss alters the relative amounts of constituents in the liquid phase [55] which can
shorten intermediate phase sintering. Mass loss can be reduced by Sintering in a powder
bed which is more surface active than the Si3N4 specimens [49]. Therefore, the
preponderance of mass loss will occur at the surface of the powder bed [55]. In general,
irnide derived powders provide small mass losses due to their high purity and moderate to
low oxygen levels. Samples with high mass loss are characterized by open porosity
throughout the Sintering cycle due to vaporization of constituents from the liquid phase
[64].
Sintering Time and Temperature

While the time and temperature trends apparent in liquid phase Sintering systems
have already been outlined, it is worthwhile to examine specific examples from the Si3N4/
Y203+A1203 system. Pressureless Sintering of Si3N4 must occur below 1825°C [34,64] in
order to avoid decomposition of Si3N4. Some research has Shown the benefit of using non-
isotlrermal, [42,43,45] or two stage Sintering processes. The two stage process involves a
hold above glass forming temperatures before Sintering at 1800°C [42,64]. This has been
shown to improve densities and reduce mass loss. Few explanations for the success of these
criteria have been offered other than extending intermediate stages of liquid phase Sintering
at the lower hold temperature [42]. This is essential for high purity Si3N4 powders because
the liquid phase formed will be of higher density (viscosity) and, therefore, Sintering will

have a greater time dependence.

52

FRACTURE TOUGI-INESS
Determination of ch

Because ceramics are inherently brittle, their failure is governed by the growth Of
cracks. A process which is often approximated using linear elastic fracture mechanics [76].
Fracture mechanics is used to determine a materials resistance to fracture relative to the
flaw size distribution. Ceramics fracture by the combined effects of acting stresses and flaw
severity, which acts to magnify stresses at the crack tip to a point where atomic bonds at the
crack tip break [76]. This magnification is measured by the stress intensity factor K, where
the subscript defines the failure mode. Failure occurs at the point where KI=KIC which is
the critical stress intensity factor or fracture toughness. There are many well accepted
testing procedures for determination of KlC all with their respective advantages and
disadvantages [76]. Measurement of Krc by crack length measurements is a simple and
effective method of determining Krc with greater specimen economy. Indentations
produced by diamond cone indenters most closely resemble real flaws [76] and yield very
reproducible results. A very successful technique for measuring Krc by indentation has
been presented by Anstis et a1. [77] using Vickers-produced radial cracks.

S. Palmqvist [78] was one of the first to realize indentation cracks could be used to
determine hardness and ch through his work with metal carbides. The idea of measuring
Km for ceramics began with using spherical indenters to induce a hertzian cone crack [76].
The crack front, however, lies entirely underneath the surface of the material and, therefore,
is useless for characterization of opaque materials. A Vickers diamond pyramid indenter

provides a symmetrical impression for hardness determination and characteristic radial

53

cracks (Figure 9) on ceramic materials which lends itself to reliability and reproducibility
[77]. In addition to the radial crack system, a lateral crack is also generated. If indentations
are formed with a high load, this crack may begin to interact with the radial system and
cause chipping of the indented surface obstructing crack measurement. Therefore, Optimal
loading should be large enough to produce measurable cracks, but low enough so that
chipping is avoided.

Models for the behavior of cracks induced by diamond cone indenters within
isotropic homogeneous materials, reveals two components to the forces acting on the crack
system, an elastic (reversible) and a residual (irreversible) force. The elastic portion is
compressive and the residual portion is tensile. Therefore, cracks grow due to the residual
force once the restraining elastic force is removed upon unloading of the indenter. When
these cracks are well developed (c>>a) they can be modeled as center loaded half-penny

cracks which results in a residual stress intensity factor of

&=x, am

3
C 2

where x, is a constant which can be approximated by [77]
E %
xx=4td (m)

where (j is a material independent constant for Vickers induced radial cracks and H is the
R,V

hardness. H is determined by the geometry of the indent and the load applied.

54

 

 

 

 

 

 

 

 

P
Medial Crack L
Cross-Section / Side View
of Radial Crack
4 b
2c
<—+
2a

Top View

 

Figure 9 Schematic of Vickers indent technique. Radial cracks result upon unloading
and are subsequently measured and related to ceramic hardness and fracture
toughness. Adapted from [77].

55

 

(20)

Where 010 =2 for Vickers indenters, and a is taken as the half-diagonal length of the indent.

(I was "calibrated" [77] using existing fracture toughness data for a wide range of

R,V

materials resulting in (j =.Ol6+/-.OO4. By combining these equations Krc can be calculated
R V

assuming the radial cracks remain stable in their post-indentation equilibrium configuration;
c=co.

antlfijflfi] (21)

In the case of many ceramics, however, equilibrium conditions do not prevail. Slow crack
effects [77] extend radial cracks beyond c0 reducing the residual stress intensity factor and
diminishing the result below the true Krc value. Therefore, the model above under predicts
fracture toughness for most ceramics. Using a similar methodology, Singh et al. [81]

developed a formula for calculating Krc for whisker toughened Si3N4.

1% (22)

Km = comm/4%) %[5

a

Equation (22) provided results similar to identical specimens tested by double torsion,
therefore, the formulation is considered accurate for whisker toughened ceramics.

Other factors can also affect crack stability, such as low H/E values, and materials

with open networked structures such as silicate glasses. Material under the indenter may

densify, thus reducing the elastic response. Microstructure will also effect indentation

56

results. Fluctuations in c0 can result from inhomogeneity on a scale >lmicron, and as the
grain size increases to sizes comparable to the indent size, the fiacture pattern can be
disrupted by local grain-failures [79].

There has been debate about whether the assumption of half-penny cracks can be
made when evaluating fracture from a Vickers induced flaw. In a review of indentation
induced cracking in ceramics and glasses by Cook and Pharr [80], they demonstrate a
substantial diversity in crack configurations between materials. While the final crack
morphology of most "normal" ceramics is that of a half-penny, Cook and Pharr argue that
half-penny cracks could not only be formed by radial cracks propagating downward (as
Anstis et al. would suggest) but also could coalescence with the median crack formed from
beneath the plastic zone under the indenter [80]. In a study by Kaliszewski et al., indents
were made through a lead acetate solution to strain the radial and median cracks for SEM
study. Examination of Y203 stabilized Zr02 with various toughnesses revealed radial
cracks turning toward one another and eventually forming half-penny cracks when a critical
load was exceeded [81]. This lends additional validity to the models presented by Anstis et
al. and Singh et al.

Experimental application of indentation methods reveals silicate glasses are subject
to substantial errors if crack lengths are not measured immediately afier indentation. Unlike
the Silicate glasses, ceramics showed no slow crack growth. This could be due to the fact
that ceramics Show higher crack velocities relative to ch than glasses, therefore, crack
growth due to factors other than the residual driving force may have already occurred before

cracks could be examined.

57

Overall, indentation methods require knowledge of E, H, and crack-length-to-load
data. H and crack-length—to—load data is collected by microscopic evaluation and E can be
computed for Sij reinforced Si3N4 by

E = Eoe'3‘06” (23)
where p is the porosity and E0 is 297, 310, 323, and 335 GPa for 0-30 v/o SiCW respectively
[82,83]. Testing requires fewer specimens and offers more variability in specimen size,
however, the surface must be polished to an optical finish. Materials tested should have a
well behaved indentation response, most glasses, softer ceramics, and coarse grained
materials should be avoided. Cracks should be well developed (22a) (therefore,
indentation methods don't lend themselves to testing high toughness ceramics, because
cracking is by nature less well developed) with no evidence of chipping (lateral crack
interaction). Lastly, the test surface should contain no pre-existing stresses prior to
indentation.

WHISKER TOUGHENING MECHANISM

SiC whiskers have been added to several CMC systems in an effort to improve
fracture toughness [48,66,82,84]. Whiskers are single crystals with a high aspect ratio.
When incorporated within a ceramic matrix, whiskers can interact with propagating cracks
to slow or alter their progress. The crack requires more energy to propagate, therefore,
more energy is required for failure and toughness is subsequently improved. SiC whiskers
have seen wide spread application in CMC's and seem to perform better than other materials
in service, despite their oxidation at high service temperatures [48]. Experimental studies of

whisker toughened ceramics indicate five toughening mechanisms that may operate to

58

increase fracture toughness [48]. These include; crack deflection [66,76,85,86], crack
bowing [87], microcracking [67,76,84,88,89], whisker pullout [48,66,90,91] and crack
bridging [67,76,92]. While advanced fracture mechanics has been applied to account for
toughening due to these mechanisms, there are still arguments over which mechanisms are
responsible for the toughening. It is often reasonable to assume that several mechanisms
operate in conjunction with one another and the prevalent mechanism is material dependent.
Microcracking

Microcracking can act to improve Kit: by relaxing stresses at the crack tip.
Relaxation occurs by a reduction in the elastic modulus and by creating a dilatational strain
in the microcracked zone surrounding the crack tip [84]. Microcracking is driven by
microstructural residual stresses caused by thermal contraction mismatch between randomly
oriented grains and between matrix grains and second phase additions. In a study by Evans
and F u [88] they showed toughening due to modulus reduction was maximized in small
grain sizes while toughening due to dilatational strains should be maximized at a larger
critical grain size. Beyond the critical grain size, some microcracks form thermally, and do
not contribute to toughening because they are not a part of the stress-strain hysteresis, in
other words, they act to reduce the initial modulus, not the modulus at a propagating crack
tip. Experimentation showed toughening at larger grains sizes, but toughening at small
grain sizes was not realized. Evans concluded that a discrete microcrack zone could not be
sustained at small grain sizes. Another possible explanation involves the grin size
dependence on fracture toughness for non-cubic materials such as Si3N4 [76]. Non-Cubic

materials exhibit thermal expansion anisotropy (TEA) which can give rise to local stress

59

that may, in turn, lead to microcracking at grain facets when subject to applied tension.
However, these non-cubic systems exhibit toughening over only a narrow distribution of
grain size (51/2 the critical size needed for spontaneous microcracking) for optimum
toughening due to microcracking. Microcracking may work in conjunction with other
mechanisms and can result in R-curve behavior as the microcrack zone develops [67,76,89].
Microcracking that occurs upon external loading has been shown to improve fracture
toughness by 10% [89]. However, pre-existing microcracks, such as those produced during
processing by high cooling rates or grain sizes above those necessary for spontaneous
rrricrocracking, show no significant toughening [89].
Crack Bowing

Crack bowing originates from the interaction of a propagating crack with a strong
dispersed second phase. The crack front is lengthened by the interaction of two or more
second phase particles with the crack there by increasing fracture energy and the strength of
the material. However, crack bowing has not been widely reported as a strengthening
mechanism in whisker toughened ceramics which may be due to the fact that whiskers are
weak when subjected to shear stresses [87].
Crack Bridging

Bridging is distinct fiom bowing, deflection, and microcracking in that toughening
doesn’t occur by interaction with the crack tip, but at a large distance from the tip along
opposing crack faces. Toughening due to crack bridging occurs by friction and mechanical
interlocking of rough opposing fracture surfaces and ligamentary bridges created by

whiskers and elongated grains connecting the fracture surfaces. Bridging is believed to be

60

the major mechanism responsible for Krc > 12MPam”2 in Sij reinforced A1203
composites [93]. This has been confirmed by SEM studies of whisker toughened alumina
[87,94]. These bridges of unbroken material act to shield the crack tip from applied
stresses. Increased grain size and aspect ratio should increase the degree of toughening due
to bridging. The number of bridging sites will also influence toughening and the degree of
R—curve behavior expected [76]. Because of this R-curve behavior, whisker toughened
ceramics cannot be classified with a single value of fracture toughness. Resistance to crack
growth increases as crack length increases due to an increase in the number of traction sites.
Therefore, strength is no longer uniquely related to initial flaw size, but is subject to stable
crack growth over a range of flaw sizes (below a critical size) as determined by the R—curve
and load configuration [67].
Whisker Pullout

While whisker pullout and debonding lengths are smaller than those associated with
fibers, toughening from whisker pullout can be estimated from Thouless and Evans'
relations derived from fiber reinforced CMCs [90]. How pullout affects toughness is highly
dependent upon Weibull Modulus, but is always proportional to pullout length. Typical
whisker pullout lengths are from 1-2 micron [94,95]. Pullout lengths, and thus toughening,
can be improved by preventing strong bonding between whiskers and the surrounding
matrix. Using smooth whiskers to prevent mechanical interlocking along with control over
elements present on whisker surfaces, can have a positive effect on pullout. Carbon
coatings were shown to improve Krc of 20% Sij/A1203 by 50% [48], but SiOz on the

surface of SiC whiskers can strengthen interfaces and reduce pullout [96]. Most whisker

6]

composites do not appear to show significant whisker pull-out, however, most whisker
toughened ceramics do show obvious evidence of crack deflection [66].
Crack Deflection

Crack deflection mechanisms in whisker toughened ceramics arise from residual
strain due to elastic modulus or thermal mis-match between phases, or from weak
interfacial bonding between phases [86]. Enhancing crack deflection can result in up to a
40% increase in Krc [76]. Crack deflection is evidenced by tilting and twisting of the
propagating crack fiont which effects the strain energy release rate and subsequently
toughens the material [66,84]. This effect has been modeled for both 3-D random whisker
configurations [86] and 2-D [66] situations where whiskers show preferred alignment, as in
hot pressing or slip casting. The 3-D model examined the effect of rod-shaped particles
with large aspect ratios (3-12) on the tip of a tilted or twisted crack. It was determined that
the crack front twist angle has a much greater effect on deflection toughening than does the
tilt angle of the crack tip. The 2-D model predicts greater toughening due to improved
crack deflection when the direction of crack propagation is normal to the whisker plane,
thus insinuating that whisker alignment imparts greater toughening than a random
distribution. Both models also indicate that toughening becomes volume fraction invariant
above 20 ‘70 whisker reinforcement. Therefore, no further toughening can be imparted by
crack deflection with greater whisker loadings. Studies by both Faber et al. [86] and Carter
et al. [85] show increased fracture toughness as median crack deflection angles increase.
The increased degree of crack deflection is also evident by a roughening of the fiaCture

surface [76]. Crack deflection is not due solely to the presence of whiskers. Monolithic

62

ceramics exhibit crack deflection which is dependent on grain shape. This makes the Sij
contribution to crack deflection difficult to determine for Si3N4 because Si3N4 already
exhibits a great deal of crack deflection due to elongated B-grains [86]. Therefore, if the [3-
grain aspect ratio were reduced by decreased sinterability from Sij additions, toughness

could theoretically decrease.

63

EXPERIMENTAL PROCEDURE

Starting Materials

Particle sizes and impurities as well as other useful information is listed in Table 3.

Table 3 Purity, particle size, and surface area of starting ceramic powders.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Components Si3N4 Sij Y203 A1203
(99.0%) (99.99%)
N >38.0%wt N/A N/A Si _<_ 40 ppm
0 1.35%wt N/A N/A ‘Na 5 10 ppm
Cl <100(ppm) N/A N/A Mg 5 10 ppm
Fe <100(ppm) 0.05% <0.005 Cu 5 10 ppm
Ca <50(ppm) 0.05% <0.005 Fe _<_ 20 ppm
Al <50(ppm) 0.08% <0.005% N/A
% Crystal >99.5 N/A N/A N/A
% B/(a+B) <5.0 100 N/A 100% or
SSA (mz/g) 11.6 24 N/A 6-10
Aspect Ratio 1 10-40 1 1
Diameter (mm) 0.5 0.3-0.6 1.0-3.0 0.3
Length (mm) N/A 5-15 N/A N/A '
Density (g/cm3) 3.22 3.20 5.01 3.98

 

 

 

 

 

 

 

64

Starting ceramic powders were selected from various manufacturers. Si3N4 from Ube
Industries, grade SN-EIO, SiC whiskers from Tokai Whisker, grade TWS-lOO, were the
major constituents that made up the composite, and Y203 from Rhone-Poulenc, and A1203
from Sumitomo Chemical, grade AKP-SO, were used as sintering aids.
Particle Sizing

Testing began with a particle size analysis of the as-received Si3N4 and SiC whisker
to confirm manufacturer's reported particle sizes and to determine relative particle sizes as a
function of pH. The Andreason Pipette method was chosen to determine particle sizes [97].
This method is based upon Stoke's Law, the principal law of sedimentation of particles in a
fluid, which states that larger (heavier) particles will sediment (fall through a fluid medium)
faster than smaller particles under the force of gravity. Slurries of Si3N4 and SiC whisker
were prepared in 104M KNO3 electrolyte (prepared with deionized water) to simulate
processing conditions. Each suspension was ultrasonicated for three, 5 minute cycles to
disperse any pre-existing agglomeration. This was repeated for each slurry over a pH range
of 2-11. The pH values were adjusted by adding appropriate amounts of HNO3 (acid) and
KOH (base). All acids, bases and salts used were of reagent grade or better. The slurry was
subsequently placed in a 500ml graduated glass cylinder and shaken by hand for two
minutes. The suspension was then allowed to sediment. Samples of 15ml were extracted
from the Slurries at 3, 10, 60, 120, 240 minutes and for longer periods when needed, into
weighed and marked 50ml glass beakers. The beakers were placed into a furnace at 120°C
for 24 hours to evaporate the water, and the beakers weighed a second time. Maximum

particle diameters were determined as a function of suspension medium viscosity and

65

density, settling distance of particles over a given time, density of particles, and acceleration

 

of gravity using Stoke's Law.
d = J_l§"fl_ (24)
g(Dr + D2 )t
where:
d= diameter of coarsest particles which remain in suspension at a depth below
h (cm) and at a time t (sec)
= liquid viscosity (timer: 0.00818 poise)
= height (cm)

Dl= density of the particles (g/cm3)

D2= density of liquid (Dwm,=0.996 g/cm3)

g= acceleration of gravity = 980 cm/s2

t= time (sec)
Representative particle sizes from each pH tested were determined from the time when 50%
of the particles in suspension were under sized. This was calculated by the following

relationship.

%Undersized = E *100% (25)

2 l

where:

W,= weight of solids in sample withdrawn

W2= weight of total solids in cylinder at time of sample drawing

V1= volume of the sample drawn

V2: volume in the cylinder at the time of drawing
A correction factor for the asymmetric effects of the SiCW falling through an aqueous
medium under the effects of gravity [98] was included in the calculation of an equivalent
spherical diameter to characterize the whisker agglomerate size. Once particle size values

were determined as a function of pH, further sedimentation experiments were conducted to

characterize the behavior of Si3N4 and Sij suspensions as a function of pH.

66

Sedimentation

A 400ml suspension of either Si3N4 or SiCw was prepared at 104M KNO3, brought
to pH 2 and ultrasonicated for 3 five minute cycles to simulate processing conditions. Upon
which, 20ml was drawn off and placed into a test tube. The tube was shaken by hand for 2
minutes and then allowed to sediment for several days. Suspension densities were
subsequently measured to estimate the degree of agglomeration as a function of pH. This
experimental stability was then compared to stability estimates from C—potential data. C-
potential curves were generated from data determined by Electrokinetic Sonic Amplitude
(BSA) testing.
ESA Testing

ESA curves were determined using 220ml Slurries of 0.5 "/o solids in 104M KNO3
electrolyte. The pH of each suspension was lowered to pH 2 with HNO3 and ultrasonically
dispersed for five 3 minute cycles to remove any pre—existing agglomeration. ESA curves
were generated using a computer controlled Matec 8000 ESA system equipped with an
autotitrator, overhead stirrer and probes for pH, temperature, and conductivity. This system
uses O'Brien’s [8] theory of electro-acoustic effects in dilute suspensions of spherical
particles to calculate the suspension Q—potential fi'om the measured ESA signal. These
theories are valid if; (1) the suspension is <10V/0, (2) the electrical double layer is thin
relative to the particle radius (1ca>50), and (3) the standard deviation of the particle size is <
20% the mean particle size. SiCw do not meet the final criteria, therefore, electrochemical

data collected from the whiskers will be more subjective.

67

A 1 MHz electrical field is applied to the suspension which generates a sound wave
at the electrode surface. The pressure amplitude of the wave (electrokinetic sonic
amplitude) and the phase angle relative to a standard of known polarity (10 v/o Ludox) were
measured. For suspensions meeting the three criteria stated above, the ESA signal is
proportional to the density difference between solid particles and the liquid phase, solid
volume fraction, and the dynamic (frequency dependent) electrophoretic mobility [8].
Under these conditions, dynamic mobility is approximately equivalent to the electrophoretic
mobility and is therefore related to the potential at the shear plane, or Q—potential. Using
this technique to characterize the behavior of ceramic powders has been validated in a
number of studies [6,8]. Slunies were subjected to base titration over a range from pH 2-
11. The ESA data for Si3N4 powder and SiC whisker was collected by two different
methods. The first method using the manufacturer's reported particle size over the entire pH
range, and the second method was a point-to-point determination of the ESA using the
varying particle sizes determined by the Andreason Pipette method. Both techniques
yielded similar curves with identical i.e.p.s (iso-electric points). The point-to-point method
resulted in nearly the same curve with increased intensity of the zeta potential values due to
a higher size correction factor caused by the increased agglomerate size. It was decided to
use the conventional ESA curve with a singular particle size because more is known about
how to interpret these curves (i.e.: homostability at approximately 25mV) [99] and the
additional agglomerate size information didn't change the inherent shape of the curve. ESA
testing was conducted on A1203 and Y203 Sintering aids to estimate their homostability'as a

function of pH. Q-potential results could not be collected for Y203 however, due to

68

solubility in acidic solutions. This has a substantial effect on processing conditions which
will be discussed in more detail later. The calculated ESA and particle size values along
with Hamaker constants for Si3N4 and SiC were input into Suspension Stability©, a
computer program developed by Wilson and Crimp [10] to predict the stability of
composite colloidal suspensions by applying the modified version of DLVO/HHF theory
discussed earlier.
Stability Calculations

Hamaker constants for each component were determined by a simplified Lifshitz
method developed by Bleier [4] for particle interactions in a vacuum. Because these
ceramic components are interacting in an aqueous medium, an effective Hamaker constant
is determined which subtracts out the effects of molecules from the medium altering the
interactions of the composite particles. Results from Suspension Stability0 revealed three
different regions of relative stability between Si3N4 and Sij at pH 3, 6, and 11. Colloidal
CMC processing began by comparing how these changes in pH affected suspension
parameters and green densities.
Suspension Preparation

Sample processing was carried out by freeze-drying composite colloidal
suspensions followed by consolidation and Sintering of the green bars. Ceramic
suspensions were produced by first preparing a 104M solution of KN O3 and adjusting that
solution to the desired pH with either HNO3 or KOH. The appropriate amount of Si3N4 was
added to the electrolyte and the pH adjusted to return the suspension to the processing pH.

SiC“, were added in varying percentages (10 and 20V/0) and the pH again corrected. Some

69

Si3N4 slurries were ball-milled for 24 hours before the addition of Sij to reduce the
number of hard agglomerates present in the as-received powder [100]. Ball-milling was
carried out in 250ml Nalgene bottles containing alumina grinding media. Total ceramic
powder additions to the process Slurries form a 5:1-7:1 ratio of electrolyte to solids. This
has been shown to be an optimum range for freeze-drying ceramic powders [16]. Y203 and
A1203 are added to the composite suspensions to promote liquid phase Sintering [42,64] at
weight percentages of (4%, 2%), (6%, 2%), and (8%, 2%) respectively. Sintering aids were
added to Si3N4 suspensions prior to the ball-milling step to improve mixing with the Si3N4
and to reduce particle size of the Y203. Addition of Sintering aids offered no change to
suspension pH at pH 11, however, Y203 is soluble in aqueous acid solutions [101] and will
consume the acid additions to the suspension until it is completely dissolved in the
suspension. Composite suspensions of Si3N4/Sij were ultrasonically dispersed for 3
cycles at five minutes each and pH adjusted at the end of each cycle. Ultrasonication was
performed in an ice bath to ensure that changes in pH were due to forming of new surface
area from splitting soft agglomerates and not due to temperature changes. After the Slurries
were ultrasonically dispersed, they were poured into 300ml glass vessels to be freeze—dried.
Freeze-drying

The vessels were stoppered and placed on rollers in a methanol bath at -40°C for 30
minutes. Shell freezing the suspension to the walls of the vessel for more efficient freeze-
drying [l 8]. The fiozen suspension is then introduced to a vacuum of < 2kPa and the
condenser is dropped to -5 0°C to promote sublimation of the frozen water from the

suspension. The freeze—drying process is complete when the exterior of the vessel reaches

70

room temperature (approximately 12 hours). Secondary drying was not carried out.
Freeze-drying has a distinct advantage over mechanical mixing of powder and whiskers
because of the added control of agglomeration invoked by colloidal manipulation. Soft
agglomerates formed during fi'eeze-drying were removed by agitation (30 nrinutes either by
hand, or by very low speed ball mill) in a Nalgene bottle containing 3 alumina cylinders
used as grinding media in the previous ball-milling step. Once sofi agglomerates were
removed (visual inspection), the powder was uni-axially pressed to form specimen bars.
Uni-axial Pressing

Near-net shaping of freeze-dried Si3N4/Sij composite powder was accomplished
by first cold uni-axially pressing the dry powder in a steel double action die (65mm x 6mm
x 15mm) at 15MPa. The die cavity was filled with powder and vibrated for 15 seconds to
promote mriform settling of the powder within the die. The level of powder in the die was
then topped off and subjected to vibration a final time. Force was applied to the die by a
Model C Carver Laboratory Press. Further consolidation of the green specimens was
accomplished by cold isostatic pressing (CIPing).
CIPing

Specimens were relatively fragile afier low pressure uni-axial pressing and require
further consolidation for greater green strength. This was accomplished by CIPing at
250MPa (35,000psi). The CIP used was a model produced by [so-Spectrum. Specimens
were vacuum sealed in latex bags to protect them from the pressurizing fluid. Bagged
specimens were placed in a basket and immersed in a mixture of water and anti-fi'eeze

within the pressure vessel. Pressure is applied by an air driven pump which delivers more

71

fluid into the vessel thus raising the pressure within. Top pressure was held for 2 minutes
and released over the course of 30 seconds by hand. Upon completion of CIPing, specimen
bars were subsequently wei red and dimensioned using calipers to calculate green
densities.
Sintering

Composite green specimens were Sintered in an Astro 1000-8 4560 graphite
resistance firmace by Thermal Technologies in an atmosphere of flowing nitrogen. Flow
was monitored by the consistent flow of emission gasses through silicon oil. Sintering
began from room temperature to 1100°C under vacuum conditions. A N2 pressure of 70
kPa was maintained in the furnace chamber above 1100°C. Furnace temperature was
controlled by two methods. The fumace is equipped with a computerized controller
attached to a BOT-2 thermocouple (graphite vs. boron-doped graphite) supplied by Thermal
Technologies. Temperature was also monitored with an optical pyrometer sighted through
a quartz window port to ensure accuracy of temperature measurement. Specimen bars were
packed in a 70 vol.%BN/30 vol.%Si3N4 powder bed to minimize weight loss during
Sintering [49]. Specimens were then Sintered under varying schedules to evaluate how
changes in Sintering time and temperature affect density, microstructure, and relative 01/[3
content.
Density Measurements

Densities of Sintered ceramic specimens were determined by ASTM C 373-88
[102]. This method is similar to Archimedes principle in that it involves the comparison of

wet and dry weights, but employs techniques that allow for more successful detection of

72

surface porosity. A digital balance was used to weigh specimens, and evaluate specimens
suspended in water.
Microstructural Evaluation

Specimens were prepared for microstructmal evaluation by first removing 0.5mm
from the surface with a high-speed diamond saw. The exposed surface was then polished
with either a series of graded SiC polishing papers and varying sizes of alumina suspensions
down to 0.05 micron or diamond slunies of 10, 6, and 1 um respectively. Polished surfaces
were examined by optical microscope, and subsequently gold coated for examination by
SEM. Gold coatings were applied by DC sputtering to all specimens under a partial
pressure of argon. Each specimen was coated for 3 minutes yielding an approximately
30nm thick coating of gold on the polished surface. The specimens were examined in an
Hitachi 4500-8 SEM.
Characterization of Crystallinity

X-ray diffraction with a Scintag model 2000 XDS diffractometer was used to
determine the crystalline phases present in densified samples. Cu ku radiation was used in
0.030 steps through 20-50° 20 to create X-ray diffraction peaks for characterization of or/B
ratio in Si3N4 and crystallinity of sintering aids. Samples were either sectioned or surface
layers were ground to prevent surface layers from effecting results. Relative amounts of

or and B can be calculated from the following relationship [52],

Ia(102) + 101(210)
l-898(Im(102) + Ia(2l0))+(15(102) + 113(210)

 

Fa = (1.898) _ (26)

73

however, peak heights can be used instead of computing integrated peak intensities for use
in equation (26) [103]. This simplification is valid for Si3N4 ceramics because peak
broadening is small for brittle materials and also appears to be constant over the appropriate
range of 29 for Si3N4 [103].

Determination of K“;

Fracture toughness values were evaluated by the previously described indentation
technique [82]. Sintered samples were sectioned and polished to 111m diamond paste and
indented using a Buehler Vickers indenter. Indents were made using varying loads from
100-200N. Indents were positioned to create cracks along both the longitudinal and
transverse axis of the specimens. The Buehler indenter is equipped with an optical
microscope for immediate evaluation of crack lengths and the measuring apparatus is
sensitive to 0.1um. Indented surfaces and fi'acture surfaces (created in a three-point-bend
fixture) were subsequently prepared for SEM examination to characterize toughening

mechanisms.

74

RESULTS

Colloidal Optimization

As the pH of a particle suspension changes, surface characteristics of the powder in
suspension may also change giving rise to either an increase or a decrease in surface
potential. This change in potential will alter the attractive and repulsive forces between the
particles in suspension and determine the stability of the suspension. When attractive forces
dominate, individual particles agglomerate to form larger particle bundles which settle out
of suspension. Therefore, when Si3N4, or any ceramic powder, is in suspension at any pH
other than its stable pH, it will behave as a system of agglomerates much larger than the
reported particle size.

Particle Sizing

The average particle size, or observed agglomerate size, was determined for Si3N4
as a function of pH using the Andreason pipette method (Figure 10). Results indicate
observed particle sizes vary with changes in pH. The largest agglomerates were formed at
pHs from 2-8 with sizes well above the reported manufacturers particle size of 0.3um.
Therefore, the flocculation of Si3N4 in aqueous suspension is clearly effected by the
suspension pH. Only at pH 2 and 9-11 is it possible for Si3N4 to be present in suspension

with little or no agglomeration.

75

 

Particle Diameter (in Microns)
25 .

 

207
15]

10"

 

 

 

 

 

 

 

Figure 10 Average agglomerate size vs. pH for as-received ultrasonicated Si3N4
powders as detennined by Andreason Pipette Method.

76

Sedimentation

Sedimentation of both Si3N4 and Sij powders revealed improved powder packing
with increased suspension stability (Figure 11). With an increased degree of stability, the
effective particle size in suspension is reduced (as shown previously). Therefore, when
suspensions of varying pH are allowed to sediment, those suspensions that are the most
stable will sediment slower and to a higher density than unstable suspensions. The pH at
which the highest sedimentation densities occur is pH 2-3 and 10-11 with Si3N4 showing a
4-fold increase in density and Sij a 2-fold increase at pH 11 and a small increase in
density at pH 2.

Sedimentation and agglomerate size results provide physical evidence against which
to compare theoretical results obtained through Q—potential testing and stability prediction
calculations performed using DLVO/HHF theory.

ESA Testing

C—potential results obtained by ESA as a function of pH, for Si3N4 and Sij are
shown in Figure 12. As defined by Stern theory, Q-potential data can be used to
approximate the surface potential of particles in suspension and thus the repulsive forces
responsible for suspension stability. Because van der Waals forces (attractive forces) will
not change as a function of pH, Q-potential can be used to estimate stability. Hunter [99]
estimates the minimum potential required for stability in most systems to be between i 25-
50mV. Using the lower bound for the results in Figure 12, A1203 should be stable in an
acidic environment, while Si3N4 and Sij are very near the minimum potential for stability.

It is important to mention that this estimate does not account for any variation in particle

77

 

Density (glcma)

 

 

 

 

 

8 91011

 

 

 

 

+Si|icon Nitride "er-Silicon Carbide

 

 

 

Figure 11 Sedimentation density vs. pH for as-received ultrasonicated Si3N4 and Sij.

78

 

Zeta Potential (mV)

 

30

20-

10-

 

-10_

-20 _

-30 _

 

 

 

 

 

 

 

 

+ Silicon Nitride + SiC(w)

 

 

 

Figure 12 Zeta-potential vs. pH for as-received ultrasonicated Si3N4 and Sij as
determined by ESA.

79

size. If a particle is large (>1p.m), the charge is spread over greater area, the force of gravity
on the particle is greater, and because van der Waals forces are treated in an additive
manner, the attractive forces will be stronger. Therefore, it is more likely that Si3N4 will be
stable at pH 2 compared to Sij due to the larger size of the Sij and increased surface
area. The basic side of Figure 12 indicates Si3N4 and possibly Sij could be homostable
(i.e.: stable with respect to itself) while the homostability of A1203 is unlikely. At a neutral
pH, applying the approximation indicates that neither of the components would be
homostable and all three components would agglomerate. While C—potential data is a
convenient way to estimate homostability, it alone is not enough to estimate interparticle
interactions between different components. DLVO/HHF theory can be applied using the @-
potential results to estimate stability between different components in suspension.

Stability Calculations

Stability calculations were modeled using Suspension Stability©, as developed by
Wilson and Crimp [10]. Stability calculations were completed for interactions between
Si3N4/Sij, the results of which are shown in Figure 13. Hunter [104] defines the behavior
of colloids with W values from 1-20 as marginally stable. Therefore, a conservative
measure of stability could include values of W > 20 as stable. Table 4 indicates relative
stability between components as predicted from the stability plot of Figure 13 by assruning

W> 20 is indicative of relative stability.

80

 

 

Stability Ratio (W)

 

 

 

 

 

 

 

 

2e- Silicon Carbide (W11) + Silicon Nitride (W22)

 

* Composite (W12)

 

 

Figure 13 Stability Ratio vs. pH for as-received ultrasonicated Si3N4 and Sij.

81

 

 

 

 

 

 

 

 

Table 4 Range of homostability and heterostability for Sij/Si3N4 composite
system as predicted by Suspension Stability©.

Stability Si3N4/Si3N4 SiCW/Sij Si3N4/Sij
Ratio W Interactions Interactions Interactions
Range of pH 2-4
Predicted and pH 10-11 pH 10-11
Stability pH 7-11

Freeze-dried powders

While colloidal forces determine the relative particle sizes in suspension, the final
agglomerate sizes of freeze-dried powders is quite different. Large soft agglomerates form
upon freeze drying yielding particle sizes much larger than the colloidal agglomerate sizes,
usually 2 50 um (Figures 14-19). SEM micrographs and tap densities reveal the effect of
pH and ball-milling on freeze dried powders. Processing without the addition of acid or
base (pH 7.5) yields finer freeze-dried particle size as revealed by SEM analysis (Figures
14-16) and by tap densities Table 5.

Table 5 The effect of processing pH and ball-milling on the tap densities of

freeze-dried powders. Powders include 8% wt. Y203 - 2% wt. A1203 as
sintering aids. Values listed are in g/cm3.

 

 

 

 

 

 

 

 

 

Si3N4 Suspension pH=3 pH=7.5 pH=11
Ultrasonicated 0.70 0.92 0.85
Ultrasonicated 0.88 1 .03 l .00
& Ballrnilled
Ultrasonicated 0.63 0.85 0.945
& Balhnilled
with 10 v/() SiC

 

82

 

Figure 14 Freeze-dried, ultrasonicated and ball-milled Si3N4 powders processed at pH
3 with sintering aids.

room

<r——1>

 

Figure 15 Freeze-dried, ultrasonicated and ball-milled Si3N4 powders processed at
"natura " pH (7) with sintering aids.

84

too pm

<l—1>

 

Figure 16 Freeze-dried, ultrasonicated and ball-milled Si3N4 powders processed at pH
11 with sintering aids.

85

mar u
<r—r>

 

Figure 17 Freeze-dried and ultrasonicated Si3N4 powders processed at pH 11 with
sintering aids.

86

 

Figure 18 Freeze-dried, ultrasonicated and ball-milled 20V/o Sij/Si3N4 powders
processed at pH 1 1 with sintering aids.

87

DEE

 

Figure 19 Freeze-dried ultrasonicated and ball-milled 20"/o Sij/Si3N4 powders
processed at pH 1 1 with sintering aids. Sij added after milling.

88

Although colloidal studies show pH 7.5 to be a region of agglomeration, the effect of the
dissolution of Y203 must also be considered. The dissolution of any constituent phase into
the suspension could effect the surface charge on particles in suspension and hence the
stability.

SEM rnicrographs show ball-milling to have a small effect on particle size and
morphology. Particles appear smaller on average with a more rounded shape and a tighter
size distribution as shown in Figures 16 and 17 for monolithic powder dispersed at pH 11.
This may be due to a smaller size distribution of primary particles created by ball-milling.
As a function of pH, particle sizes appear smallest at pH 7.5 with ball-milling. This is
confirmed by tap density measurements where pH 7.5 yields the highest tap density. While
pH 7 .5 yields the highest tap density for monolithic Si3N4, pH 11 shows the highest tap
densities for the composite powder (Table 5). This is due to the instability of SiCW at pHs
10.5 as demonstrated by the stability prediction in Figure 13. Processing at pH 3, on the
other hand, shows poor tap densities with the addition of Sij. While Sij should be
relatively stable at pH 3, YZO3 is chemically unstable and will dissolve in suspension
[101,105] during ultrasonication and ball-milling. Subsequent precipitation of metallic Y3+
complexes on the Si3N4 surface [101] yields a harder powder upon freeze-drying.

Pressing characteristics of all powders were relatively consistent with the exception
of powders containing 2. 20 "/o Sij loading. These powders tended to delarninate after
uni-axial pressing and pH 3 powders produced extremely warped specimens upon CIPing.

This is due to the hard agglomerates formed during processing at pH 3.

89

Selected component milling of pH 11 powders revealed reductions in tap density
with whisker loading (Table 6). Composite powders were processed at pH 11 due to their
higher tap densities and consistent powder quality. Milling of all components yielded the
best tap densities, while milling of Si3N4 alone improved densities more than milling of
whiskers alone. Therefore, deagglomeration of Si3N4 is important for improving the
packing of freeze-dried powders. This is finther demonstrated by the fact that the 10 v/()
Sij ball—milled powder shows a higher tap density than does unmilled monolithic Si3N4.
The addition of 20 "/O Sij had a major effect on tap densities (Table 6). The increased
interparticle friction will act to lessen green densities and reduce rearrangement during
liquid phase sintering thus reducing the Sintered density. Figures 17 and 18 show increased
surface roughness of fieeze-dried powders with 20 v/0 Sij. The advantage of ball-milling
Sij can be seen by a reduced surface roughness of the powder.

Table 6 The effect of whisker loading and selected component ball-milling on

tap densities of powders freeze-dried at pH=11 with sintering aids. Values
listed are in g/cm3.

 

 

 

 

Si3N4 Suspension Monolithic 10 "/0 SiC- 20 V/0 SiC-
Si3N4 Si3N4 Si3N4
All Components 1.00 0.94 0.76
Ballmilled
Only Si3N4 1.00 0.89 0.70
Ballmilled
Only Sicw 0.85 0.75 --
Ballmilled

 

 

 

 

 

 

90

Green Density

Green densities were measured as a function of both whisker loading and pH.
Green specimens were first prepared without sintering aids from ultrasonicated suspensions.
The resulting green densities of uni-axially pressed specimens were determined as a
function of pH. Results show the highest green densities for composite specimens at pH 11
where there is heterostability and homostability of both constituents in the composite
suspension (Table 7). The trend in green density results is due to whisker stability, good at

pH 11, fair at pH 3, and poor at pH 6,

 

 

 

 

 

Table 7 Green density as a function of pH and whisker loading (g/cm3).
Whisker Processed at Processed at Processed at
Loading pH 3 pH 6 pH 11

0% 1.33 +/- .005 1.27 +/- .005 1.32 +/- .005
10% 1.21 +/- .015 1.22 +/- .015 1.28 +/- .015
20% 1.24 +/- .016 1.15 +/- .01 1.27 +/- .008
30% 1.11 +/- .013 1.14 +/- .013 1.18 +/- .01

 

 

 

 

 

 

as shown by sedimentation as well as Q-potential and W results. Monolithic specimens
show the highest density at pH 3 and 11 where Si3N4 is stable. CIPing at 350MPa resulted
in an increase in green densities, but the relative difference in green density as a function of
pH did not change. Green density as a function of V/o whisker content was also analyzed.
In general, green densities decrease with whisker loading as whiskers increase interparticle
fiiction and reduce rearrangement during cold pressing [106]. Scatter in [the data is at a

maximum for specimens with 10 v/() whiskers. This is due to the fact that delamination of

91

the green specimens [107] begins to occur at 20 v/o Sij, therefore, only the best specimens
are suitable for evaluation, thus seeming to reduce inconsistencies in the data. Specimens at
30 v/0 Sij were difficult to produce at all pHs. CIP densities are not presented for
specimens without sintering aids due to inconsistencies in pressing pressure during testing.
These results demonstrate the dependence of green density on colloidal stability. Freeze-
dried powders pack to higher densities when they are produced from more stable
suspensions for both the monolithic and composite systems. This is due to improved
homogeneity and smaller mean agglomerate sizes under suspension conditions.

Tables 8 and 9 show how green densities were affected by the addition of sintering
aids. Composites were not processed at pH 3 or 7 without ball-milling due to incomplete

sintering of specimens reinforced with as-received whiskers. Green densities without ball-

 

 

 

 

Table 8 Green densities of specimens CIPed at 250MPa with 8 wt.% YZO3/ 2 wt.%
A1203 added as sintering aids.
Whisker Processed at Processed at Processed at
Loading pH 3 pH 7 pH 11
0% 1.74 i .01 1.69 i .00 1.83 i .01
10% -- -- 1.73 i- .01
20% -- -- 1.68 i .02

 

 

 

 

 

 

milling seem to follow the same trends seen before the addition of sintering-aids with green
density being the highest at pH 3 and 11 and decreasing with whisker loading. Table 9

shows the effects of ball-milling on green density. Ball-milling of suspensions increases

92

 

 

 

 

Table 9 Effect of ball-milling on green densities of specimens CIPed at 250MPa
Whisker Processed and Processed and Processed and
Loading milled at pH 3 milled at pH 7 milled at pH 11

0% Unmeasurable 1.76 i .01 1.83 _+_ .01
10% 1.73i.01 1.72:.02 1.77:.015
20% -- -- 1.72 i .01

 

 

 

 

 

 

green densities for all processing pH and whisker loadings. Ball-milling in combination
with ultrasonication removes hard agglomerates from suspension and improves suspension
homogeneity. Suspensions ball-milled at pH 3 produced green specimens that were too
warped to measure the volume by calipers. Freeze-dried powders produced from ball-
milled suspensions at pH 11 produced the highest green densities. Green densities of ball-
milled pH 11 powders with 20 v/o Sij were comparable to those of monolithic specimens
without ball-milling. In addition to increasing the green densities, ball-milling of pH 7 and
pH 11 suspensions produced a dark-gray hydrophobic residue on the surface of the
suspensions. This residue was difficult to remove and was therefore still present in powders
freeze-dried at these pHs. This residue was collected and characterized along with as-
received and fieeze-dried powders by X-ray diffraction.

X-Ray Analysis of Green Powders

X-ray analysis began with as-received powders. These results were then used to
characterize the green specimens. Figures 20 - 23 show diffraction peaks for as-received ct-

Si3N4, B-Sij, YZO3, and ct-Ale3 powders respectively. All data match the expected

93

 

Intensity

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

20 25 30

2 Theta

 

 

 

Figure 20 X-ray diffraction results for Ube Sn-e-10 as-received Si3N4. A = a- Si3N4
and B = B-Si3N4.

94

 

Intensity

 

 

1.4.4.11. m “4.1.. I A. J‘AA‘
v

30 35 40 45 50
2 Theta

 

 

 

 

 

 

 

Figure 21 X-ray diffraction results for Tokai Whisker TWS-100 as-received B-SiC
whiskers.

95

 

Intensity

 

 

 

 

 

 

2 Theta

 

 

 

Figure 22 X-ray diffiraction results for as-received Y203 powder. Circled region
indicates undefined peak.

96

 

 

Intensity

 

 

 

 

 

 

 

 

20 25 30 35 40 45 50
2 Th eta
Figure 23 X-ray diffraction results for as-received 01-A1203 powder.

 

97

spectrum (from JCPDS--the Joint Committee on Powder Diffraction Standards) with the
exception of Y203 which showed an additional peak at z 47.50 29 along with the shifting of
smaller peaks from their standard positions. This may indicate the as-received YZO3
powder is a mixture of various crystalline forms of YZO3. Analysis of the ball-milling
residue is shown in Figure 24. The residue could not be completely separated fi'om
suspension components, therefore, peaks from pH 11 suspension should be present along
with the diffraction data from the residue. The major peaks can be characterized indicating
the presence of a-Si3N4 and Y203 with the relative Y203 peak at 3.060 being much higher
than in the green powder. This indicates a relative increase in the amount of Y203 with
respect to Si3N4, however, it does not explain the hydrophobic behavior of the residue.
Another possibility arises from the high background seen in Figure 24. Background
correction software indicates this could be due to an amorphous phase that is forming
during ball-milling where the diffraction peaks are the result of the inability to separate the
suspension from the residue. Regardless, this skin forms on all suspensions ball-milled at
neutral or basic pHs and is a part of all freeze-dried powders prepared by ball-milling in this
pH range.

X-ray analysis was carried out on green specimens to evaluate the chemical state of
the composite prior to sintering. Ball-milled green powders were chosen because if there
are any chemical changes in the suspension components, they should be more noticeable in
the ball-milled powders due to longer processing times and increased particle surface areas
and interactions. Suspensions containing 10 V/o SiCW were also examined to see if the

whiskers played a role in green chemistry. X-ray diffraction results for composite green

98

 

Intensity

 

Am

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2 Theta

 

 

 

Figure 24 X-ray diffiaction results for ball-milling residue from powders ball-milled at
pH 7.5 and 11. A = a-Si3N4, Y = Y203. High background may indicate the
presence of an amorphous Y-containing phase.

99

powders are shown in Figures 25 - 30. Results for pH 7 and 11 are nearly identical (Figures
25 - 26). Both show the presence of or-Si3N4, YZO3, and B-Sij with no other crystalline
phases present. Results for 10 "/o B-Sij are also identical for pH 7 and 11 (Figures 27 -
28). Both show increased relative YZO3 peaks over their monolithic counterparts. This is
probably due to the relative increase in YZO3 content to Si3N4 content with the addition of
Sij. Composite suspensions prepared at pH 3 resulted in similar results, but with much
smaller Y203 peaks (Figure 29 - 30). This is due to the fact that YZO3 is soluble in nitric
acid solutions [105]. Diffraction results of monolithic Si3N4 ball-milled at pH 3 showed no
evidence of Y203. It is believed all the YZO3 was dissolved in suspension during ball-
milling and reprecipitated during freeze-drying into another crystalline compound. This is
evidenced in Figure 22. These unknown peaks could not be characterized as an Yttria
containing nitrate, nitrite, nitride, hydroxide, or ammonia containing compound. Therefore,
it is unclear what chemical state the YZO3 is in before sintering of pH 3 processed freeze-
dried powders.

Sintered Density

Small amounts of some chemical species was evolved during sintering and
deposited on crucible and furnace graphite walls. These were not evaluated as mass losses
were low and decomposition did not appear to be prevalent. Sintered densities were first
evaluated as a function of whisker loading for specimens processed at pH 11. All Sintered
results are from specimens containing 8% Y203 / 2% A1203 by weight. Results are
shown in Figure 31 for ball-milled Si3N4 reinforced with both as-received and ball-milled

Sij and Sintered for 4 hours at 1800°C. Ball-milled Sij provide enhanced

 

100

 

Intensity

 

 

 

 

 

 

 

 

 

 

2 Theta

 

 

 

Figure 25 X-ray diffraction results for monolithic Si3N4 ball-milled at pH 7. A = or-
Si3N4, Y = Y203.

101

 

Intensity

 

 

 

 

 

 

 

 

 

2 Theta

 

 

 

Figure 26 X-ray diffiaction results for monolithic Si3N4 ball-milled at pH 11. A = 01-
Si3N4, Y = Y203.

102

 

Intensity

 

 

 

 

 

 

 

 

 

 

   

20
2 Theta

 

 

 

Figure 27 X-ray diffraction results for 10 V/o Sij-Si3N4 ball-milled at pH 7. A = or-
Si3N4, Y = Y203, s = Sij.

103

 

Intensity

 

A A

 

 

 

 

 

 

 

 

   

2 Theta

 

 

 

Figure 28 X-ray diffraction results for 10 v/o Sij-Si3N4 ball-milled at pH 11. A = or-
Si3N4, Y = Y203, S = Sij.

104

 

Intensity

 

 

 

 

 

 

 

 

   

2 Theta

 

 

 

Figure 29 X-ray diffraction results for 10 v/o Sij-Si3N4 ball-milled at pH 3. A = 0L-
Si3N4, Y = Y203, S = SICW.

105

 

Intensity

 

 

 

 

 

 

 

 

 

 

 

 

Figure 30 X-ray diffraction results for monolithic Si3N4 ball-milled at pH 3. A = or-
Si3N4, Y = Y203. Note the Y203 peaks are missing. The new
uncharacterized crystalline phase is indicated by *.

106

 

 

 

 

 

 

g/cm3 % Theoretical Density
98.8%
-91.2%
\
l \ .
2.75 —— \ \ 783.6%
‘1
2.5 i 76%
0% 10% 20%

VI, Whisker Loading

 

 

 

“ 'As-received Whiskers +BaII-milled Whiskers

 

 

 

Figure 31 Density as a function of whisker loading for ball-milled Si3N4 specimens
Sintered 4 hours. Results show higher densities for composites with ball-
milled (reduced aspect ratio) Sij in addition to milling of Si3N4.

107

densification of the composite by improving homogeneity and reducing the aspect ratio
of the Sij. Sij can act to reduce rearrangement during liquid phase sintering and
decrease the final Sintered density and/or increase the time to full density. This is shown
more clearly in Figure 32 for powders containing 10 "/o Sij. The highest densities are
achieved when either all components are milled, or only when Sij are milled. However,
when all components are milled, the densification rate increases. When only the Si3N4
matrix is milled, rearrangement is hampered and intermediate phase sintering must occur
over a longer time to achieve full density. This is shown in Figure 32. The slope of a
plot of density versus sintering time for powders with only matrix milling show the
highest sintering rate from 1-2 hours. This is due to the fact that more liquid is necessary
to allow for initial rearrangement to occur. Sintering then continues at nearly the same
rate as materials with reduced aspect ratio whiskers. Therefore, it is expected that
powders with milled Si3N4 and as-received whiskers could achieve the same densities as
the other milled powders with increased sintering time. Powders that received no ball-
milling showed minimal densification after 4 hours.

Monolithic powders achieve higher densities more quickly than those reinforced
with ball-milled whiskers, as shown in Figure 33. As whisker loading increases, even
with reduced aspect ratio whiskers, final density after 4 hours sintering decreases.
However, because the 20 "/o composite still shows an appreciable sintering rate after 4
hours, the final density difference between 10 v/o and 20 V/o should lessen with increased

sintering time.

108

 

g/cm3 % Theoretical Density
3.29 ., :r 100%

 

g 96.96%
‘N 93.92%
. ‘ 90.88%

_ 87.84%

 

 

- 'V. 84.8%

 

Time (hours)

 

+AI| Components milled °+‘On|y matrix milled

3" No milling *‘Only SiC milled

 

 

 

Figure 32 Density as a function of time for selected component ball-milled powders.
High density is achieved more quickly when all components are milled.
Milling Sij is more important for high density than milling Si3N4 alone.
After 4 hours, compacts containing unmilled components showed only .
limited densification. This point is circled for clarity.

109

 

g/cm3 % Theoretical Density
, a 100%

 

3.29
3.19 I 96.96%
3.09 ,«' W 93.92%
2.99 90.88%

2.89 ~ . 87.84%

 

 

Time (hours)

 

 

 

—°— Monolithic '+' Milled Monolithic
+ Modified Milling +20% SiC

 

 

 

Figure 33 Density as a function of time and whisker loading for powders where all
components were ball-milled. Final density decreases as whisker loading
increases. Compared to unmilled monolithic powder, ball-milled
monolithic powder shows a lower final density and lower densification
rate.

110

Final density as a function of processing pH is shown in Figure 34 for monolithic
compacts with and without ball-milling, and 10 v/0 SiCw compacts sintered for 4 hours at
1800°C. Exanrining unnrilled monolithic powders reveals the highest sintered densities at
pH 3. This follows logically from green density results, where pH 3 yielded the highest
green densities for monolithic unmilled powders. Examining ball-milled monolithic
powders, an entirely different trend is revealed. Powders processed at pH 7 give the
highest sintered density, while the final density of powders processed at pH 3 and 11 go
down with respect to their unmilled densities. This may be due to the effects of ball-
rrrilling on the green chemistry at these pH. At pH 3 there is a complete solutionizing of
YZO3, and reprecipitation into another crystalline form, while at pH 11 there is a
substantial amount of skin formed on the ball-milled suspensions which is incorporated
into the powder during freeze-drying. These changes in the chemical state of the
sintering aids seems to have a major effect on the sintered density of ball-milled
monolithic specimens. These chemical changes, however, do not have the same effect on
sintered density for the composite specimens.

Composites of 10 v/() Sij show the highest sintered densities with powders
processed and ball-milled at pH 3 and 11. This is due to the high degree of
agglomeration of Sij and Si3N4 that occur at pH 7. Whisker agglomerates have been
shown to substantially decrease sinterability of composite green compacts [58].
Although processing the composites at pH 3 and 11 should substantially reduce whisker
agglomeration, the highest densities of 10 v/o Sij were still considerably lower than the

lowest densities for monolithic specimens with identical sintering additives and

lll

 

g/cm3 % Theoretical Density

 

 

 

 

 

3.29 , 99.98%
t
3.24 i , ,—~98.46%
\
\ -l
. \ .
3.19-- \I --96.94%
r i
i 1. .
. a x} ,
3.14»- , ,c -—95.42%
fr i 1
3.09 l l l i 93.9%
2 4 6 8 1o 12

 

 

 

*‘Ball-milled +No milling +10% SiC Whisker

 

 

 

Figure 34 Density as a function of pH for ball-milled and unmilled monolithic and
ball-milled 10 v/o Sij-Si3N4.

112

processing conditions. This is due to the fact that whiskers impede rearrangement during
the initial stages of liquid phase sintering and, in addition, increase difiusion distances
during intermediate stages of liquid phase sintering. This slows the or—)B transformation
and the driving force for liquid phase sintering of Si3N4. Sij have also been shown to
have a rough surface finish due to a high number of stacking faults in B-Sij. This
roughness increases surface area of the Sij and thereby increases the thickness of glassy
phase surrounding the whisker after sintering [108]. Therefore, more liquid phase is
required to densify the composite as compared to the monolithic compact and because
sintering aid additions were equal for all specimens, whisker reinforced compacts could
not achieve the same densities as the monolithic compacts, unless there was a surplus of
liquid phase for the monolithic specimens during sintering. Another possible reason for
the lower densities could be the chemical state of the YZO3 at pH 3 and 11. If the effects
described above have a negative effect on sintering in monolithic specimens, they may
also have a negative effect on the sinterability of the composite. Examining the
crystalline phases present in the sintered compacts may help to answer this question.

X-Ray Analysis of Sintered Components

X-ray diffraction patterns for selected specimens processed at pH 11 are shown in
Figures 35 - 41 . Results indicate Si3N4 is present only in the [3 form for all [specimens after
sintering for as little as 1 hour (Figure 41). Diffraction peaks were evaluated using d—
spacings fi'om a crystallography catalog and compared with a study by Chen et al. [109]
who studied x-ray diffraction (XRD) data for YZO3/A1203 assisted Si3N4/Sij sintered

composites. Patterns were also compared to XRD data from the initial starting powder

 

113

 

Intensity

 

 

 

 

 

 

 

 

20 25 30 35 4O 45 50
2 Theta

 

 

 

Figure 35 X-ray diffraction results for ball-milled monolithic Si3N4. Sintered 4
hours. B = B'SI3N4.

 

 

114

 

Intensity

 

 

 

 

 

 

 

 

 

2 Theta

 

 

 

Figure 36 X-ray diffraction results for unmilled monolithic Si3N4. Sintered 4 hours.
B = B-Si3N4.

115

 

Intensity

 

 

 

 

 

 

 

 

2 Theta

 

 

 

Figure 37 X-ray diffraction results for ball-milled Si3N4 with 10 V/o Sij. Sintered 4
hours. B = B'Si3N4, S = B' Sij * = Y203OSI3N4.

116

 

Intensity

 

 

 

 

 

 

 

 

 

 

 

 

2 Theta

 

 

Figure 38 X-ray diffraction results for ball-milled Si3N4 with 20 vlo Sij. Sintered 4
hours. B = B-Si3N4, S = B- Sij * = Y2030Si3N4.

117

 

Intensity

 

 

 

 

 

 

 

 

 

20 25 30
2 Theta

 

 

Figure 39 X-ray diffraction results for ball-milled Si3N4 with 10 "/o Sij added after
milling. Sintered 4 hours. B = B-Si3N4, s = 13- Sij, r = Y2030Si3N4.

118

 

Intensity

 

 

BABl

 

 

7 a: y , * B 7‘ S *
.11...-1151. .. I 51.111111... ....1'1.. .15 1‘

20 25 30 35 40 45 50
2 Theta

 

 

 

 

 

 

 

 

 

 

Figure 40 X-ray diffraction results for ball-milled Si3N4 with 20 v/o Sij added after
milling. Sintered 4 hours. B = B-Si3N4, S = B- Sij, * = Y203OSi3N4.

119

 

Intensity

 

 

 

 

 

 

 

 

20 25 30 35 40 45 50
2 Theta

 

 

Figure 41 X-ray diffraction results for monolithic Si3N4 ball-milled at pH 11.
Sintered 1 hour. B=B-Si3N4.

120

which was predominately a. B-SiC was retained within whisker reinforced sintered
specimens, but the degree of degradation of B-SiC was not evaluated. However, literature
indicates SiC whiskers are inert at the sintering temperatures employed in this study,
therefore, we would expect degradation to be minimal as demonstrated by Chen et al. [109].
Whitehead et al. [73] demonstrated B-SiC whiskers resist recrystallization up to 2050°C and
effects on the morphology at 1850°C in Ar and N2 atmospheres were insignificant (Figure
42) showing only the formation of a Sinter-boundary after 2 hours. B-Sij polytype is stable
even with large morphology changes, such as whisker spherodization at high temperatures.
However, temperatures above 2050 °C would be required for this to occur. B-Sij will
degrade in oxidizing atmospheres, or may become more unstable in the presence of an
oxynitride silicate glass [49]. Possible reactions include;
zsic + 3102 + N2 :1 31,114 + 2coT
or
3SiC + 33102 + 2N2 <:> 3131\14 + 3Si0 + 3coT

with the later reaction occurring over the former when whisker surface contact with the
liquid is high. However, if equilibrium lies to the right of the equation, the Si3N4 matrix
could become unstable. Zheng et al. [49] studied the suppression of these reactions by the
use of various powder beds during sintering. It was determined that a powder bed of 50%
BN and 50% Si3N4 minimized weight losses. Whitehead et al. [73] indicates
recrystallization of Sij at 1600°C in the presence of liquid Silicon. Therefore, the glass
composition of sintered composites could effect whisker stability. Boron carbide, a known

sintering aid for SiC, could also effect whisker stability in the sintered compact. However,

 

121

 

Temperature °C
2,150 ,. _. .1, , ..

 

2,050 ~ '

   

? Regionofseta ? x
1.9501 "Stability" _ _ ‘

 

 

 

1.850 ‘ 1 1 ”1 1

Time (Minutes)

 

 

 

Figure 42 Plot showing stability of Sij with respect to time and temperature.
Adapted from Whitehead et al.[73].

122

x-ray results indicate polytype stability and show no evidence of BC. Morphological
changes would be evidenced by spherodization of Sij which would be obvious during
SEM analysis of fracture surfaces. No evidence was found to suggest such degradation.
Therefore, it is believed that the effect of the liquid phase on the Sij during sintering is
minor. Whitehead also notes that color change of whiskers can indicate degradation. It is
believed that Sij get their green color from N2 impurities. Loss of N2 during sintering is
indicated by a color change from green to gray. All sintered composite specimens
maintained a bright green color, indicative of nitrogen retention and whisker stability [73].
Grain boundary phases remained glassy in monolithic Si3N4 as evidenced by
diffraction patterns shown in Figures 35 - 36. However, the presence of a third crystalline
phase in the whisker reinforced specimens is clear fi'om XRD data (Figures 37 - 40). These
five additional diffraction peaks coincide with Y203OSi3N4 as characterized by JCPDS
standard powder difi‘raction file. Y203OSi3N4 is shown in the Y203 -Si3N4-Si02 phase
diagram [110] in Figure 43. Shih et al. [111] reports a very similar diffraction pattern for
Si3N4-Sij composites with Y203-A1203 additives reporting a YSiOzN or K—phase as the
additive crystalline phase. While there are many crystalline Y203 -Si3N4-Si02 compounds
with similar diffraction patterns, measurement of relative intensities fi‘om the composite x-
ray peaks indicates the Y203 OSi3N4 compound. Because Y203 OSi3N4 is a relatively low 02
compound (Figure 44), this coincides with the fact that the starting Si3N4 is low in 02.
X-ray results for specimens processed at pH 3 and 7 are shown in Figures 45 - 50.
Composite specimens processed at these pHs also show evidence of a third crystalline

phase, however, it is of different composition than that characterized in pH 11 specimens

123

  

9' 320711

reg-on ot mellng 1650'C . .
Y203 72530311,. 51:, N4

 

Figure 43 Y203-Si02-Si3N4 phase diagram. Adapted from Jack et al. [1 10].

 

EQUIVALENT 36 OXYGEN —-D--

124

Y4£fl$l°4l30

vsao vzsao
227 5
2vzo3

 

 

35")? I I I .1 T V
__ 1640°c Y95'5021N3 .1
1650°C /

0.8 r- I . "‘

_ . asoo°c , .1

0.6 r- '1
vsaoz~

fi

0.4 _1

. Y286303N4 _

0.2 "‘

 

Si3N4

Figure 44

 

 

1 L 1 1 1 1 1 1 1 4,
0.2 0.4 0.6 0.8 4YN

EQUIVALENT % YTTRIUM-——.-

02 equivalents phase diagram. Adapted from [55].

1:

125

 

Intensity

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

‘

Figure 45 X-ray diffraction results for ball-milled Si3N4 processed at pH 3 with 10
"/o Sij. Sintered 4 hours. B = B-Si3N4 S = B- Sij.

126

 

Intensity

 

 

 

 

,M

| I T

20 25 30 35 40 45 50
2 Theta

 

 

 

 

 

 

 

Figure 46 X-ray diffraction results for ball-milled Si3N4 processed at pH 7 with 10
‘1/0 Sij. Sintered 4 hours. B = B-Si3N4, S = B-Sij.

127

 

 

 

 

 

 

 

 

Intensity
B
B
B
B B
B
B B
20 25 30 35 40 45 50
2 Theta

 

 

 

Figure 47 X-ray diffraction results for unmilled monolithic Si3N4 processed at pH 3.
Sintered 4 hours. B = B-Si3N4.

128

 

Intensity

 

 

 

 

 

 

 

 

 

 

 

 

 

2 Theta

 

 

 

Figure 48 X-ray diffraction results for unmilled Si3N4 processed at pH 7. Sintered 4
hours. B = B'Si3N4.

129

 

Intensity

 

B B

 

 

 

 

 

 

   

 

I l
20 25 3O 35 4O 45 50
2 Theta

 

 

 

Figure 49 X-ray diffraction results for monolithic Si3N4 ball-milled at pH 3.
Sintered 4 hours. B = B-Si3N4.

130

 

Intensity

 

 

 

 

 

 

I l I
20 25 30 35
2 Theta

 

 

 

 

Figure 50 X-ray diffraction results for monolithic Si3N4 ball—milled at pH 7. Sintered
4 hours. B = B-Si3N4.

131

and remains uncharacterized (Figures 45 - 46). Like the composite specimens, the unmilled
sintered compacts at pH 3 and 7 also show very similar results, but with yet another
different uncharacterized crystalline phase present (Figures 47 - 48). Ball-milling at these
pHs reveals yet another set of unique x-ray results (Figures 49-50), with both pH 7 and pH
3 yielding different uncharacterized results. The diffraction peaks at pH 3 are especially
unique in that the B-Si3N4 peaks have split into doublets. This seems to indicate the
substitution of a larger atom into the B-Si3N4 lattice. Al is known to be soluble in B-Si3N4
and may be responsible for the peak splitting. However, these peaks correspond only to a
shift in B-Si3N4 and not to SiAlON, therefore, a solid solution of A1203 in B-Si3N4 is more
likely.
Sij Toughening of Si3N4

Because the toughening mechanisms discussed earlier behave in a material
dependent manner, Sij dispersed in a Si3N4 matrix may toughen in a different manner
than when its dispersed in other matrix materials. In fact some studies have shown no
appreciable toughening in Si3N4 systems due to the addition of Sij [82]. Summarizing
possible toughening mechanisms, little toughening would be expected due to crack bowing
because of the low shear strength of SiCw [87]. Only minimal toughening could occur by
crack deflection, because deflection is already prevalent in monolithic Si3N4 due to TEA
and high aspect ratio beta-grains [76,86]. Therefore, Sij could only strengthen in an
additive manner, or by some additional deflection due to E differences (511 and 297 GPa

respectively for Sij and Si3N4) [103].

132

The addition of SiC“, could toughen by microcracking, due to the difference in thermal
expansion coefficients (3.3 - 4.5x10*S K“ for SiC and 3.0x10’6 K“ for Si3N4) [82], but again
because Si3N4 is non-cubic, microcracking can be present without the addition of a second
phase due to TEA. Microcracking is also highly dependent on grain size, therefore, this
toughening mechanism will only occur for a narrow range of grain size materials. In
addition, microcracking mechanisms of toughening have shown only 10% increase in
toughness over the monolithic material [88]. SEM video films of in situ stressing of a crack
indicates microcracking around the crack tip of both monolithic and whisker composite
material, however, microcrackin g seems more pronounced when whiskers are present [82].
The relative thermal expansion coefficients suggest whiskers are in tension and Si3N4 is in
compression giving rise to a tensile stress at the interface which is favorable for crack
propagation through the interface and improved pull-out [82].

Finally, whisker pullout and bridging mechanisms could act to toughen Si3N4.
While pullout has not been mentioned prevalently in many studies, bridging has been
shown to be prevalent for both Sij reinforced, and monolithic B-Si3N4. Again, Sij
would be strengthening in an additive manner, but if Sij aspect ratios are higher than beta-
grains, bridging distances would be increased and the number of traction sites would be
increased by the presence of Sij. However, when fracture toughness is evaluated by
indentation techniques, a single value fracture toughness is generated. Because bridging
and pullout mechanisms act at large distances from the crack tip, their toughening may be
more easily assessed from examining R-curve behavior as opposed to single values of

fracture toughness [76]. Experimental results seem to follow the above theories for Sij

133

reinforcement of Si3N4. Olagnon et al. [82] reports crack bridging to lead to the highest
degree of toughening among all observable active toughening mechanisms.

Evidence of active fracture toughening mechanisms could be observed by SEM
from secondary cracking on fracture surfaces and from polished and indented surfaces used
to determine fracture toughness. Figure 51a and b indicate crack bridging occurring in both
monolithic and whisker reinforced B-Si3N4. Optical micrographs of indentations also reveal
crack deflection in both monolithic and Sij reinforced B-Si3N4. Therefore, there is
evidence of toughening due to high aspect ratio constituents in both the monolithic and
composite specimens. Other investigators have also reported little improvement of
properties by the addition of Sij .

Olagnon et al. [82] sintered specimens at 1850°C under 1 MPa N2 for 6 hours using
varying amounts of Y203 and A1203 as sintering aids with 20 "In Sij. They reported Krc
of z 7 MPam”2 for 20 v/o Sij, however, were able to produce KIC = 7.6 MPam”2 for
monolithic Si3N4. The monolithic material exhibited more signs of debonding and a greater
average grain size and aspect ratio. Therefore, it was concluded whiskers suppressed grain
growth and rough surfaces of the whiskers allow less debonding to occur. Therefore, while
Sij encourage some strengthening mechanisms, they suppress others active in the
monolithic Si3N4.

Result of K“; Determination for Si3N4 and Sij-Si3N4

Fracture toughness results from indentation testing are shown in Figure 52 as a

function of whisker loading and whisker aspect ratio for monolithic and composite

Specimens produced at pH 11. The bars under 0 vol.% whisker loading compare ball-milled

134

 

Figure 51a Micrograph taken from secondary crack found along the fracture surface. a)
shows evidence of both crack deflection and crack bridging away from the
crack tip in a monolithic Si3N4.

135

 

Figure 51b b) Micrograph shows a crack bridge in a secondary crack on a 10 v/o Sij
specimen.

136

 

K", (MPa m1”,

 

   

 

 

3635.

0% 10% 20%
% Whisker Loading

 

 

 

 

 

All Components Milled .As-received Components

 

 

 

Figure 52 Fracture Toughness as a function of whisker loading and ball-milling for
specimens processed at pH 11.

137

matrix to unmilled, with the ball-milled Si3N4 showing a higher fiacture toughness. For
10% whisker loading, specimens containing as-received whiskers in a ball-milled Si3N4
show a slightly higher fracture toughness than specimens where all components were milled
together. This higher toughness is due to the maintenance of high whisker aspect ratios
which, as discussed earlier, more readily improve fracture toughness. However, high aspect
ratio components also reduce densification, therefore, components with 20 v/0 unmilled
whiskers showed little densification and degraded mechanical properties.

In a study by Zheng et al. [49] they showed no increase in fiacture toughness with
the addition of 10-30 "/0 SiCw to Si3N4 for pressureless sintered compacts (K,C=4 MPamm),
but did show improvements in KIC for hot pressed composites of the same composition. In
fact, most studies that show improved toughness with the addition of SiCW consolidate
specimens by hot pressing [50,112]. The improved toughness during hot pressing could be
due either to improved density (as compared to pressureless sintering), or alignment of
Sij. Liu et al. [66] showed whisker alignment improved crack deflection toughening.
Aligned whisker planes can be modeled in 2d rather than the 3d analysis by Faber and
Evans [89]. In the 2d analysis there is a contribution to toughening from crack front tilting
and twisting, as opposed to the 3d model which shows most of the toughening should occur
from twisting alone. Therefore, the toughness increase seen in hot pressing studies may not
be the sole result of higher densities, but a result of whisker geometry.

Results of Sij reinforcement should also be compared to in situ reinforced Si3N4.
As discussed above, Si3N4 shows evidence of all the major toughening mechanisms without

the addition of a second phase. By optimizing beta-grain size and aspect ratio, Lange et al.

All.

138

[55] showed fracture toughnesses of monolithic Si3N4 could be substantially improved.
Current studies report toughnesses in the range of 8-11 MPam”2 [48, 91]. Which are of the
same order as Si3N4 reinforced with Sij.
SEM Analysis of Fracture

SEM analysis can be used to characterize grain size as well as the degree of
toughening from the microstructure and microstructural homogeneity. Ball-milled Si3N4
processed at pH 11 shows increased porosity with whisker loading (Figures 53 - 55). High
levels of porosity are seen on 20 v/o as-received Sij reinforced specimens. This surface
shows little toughening with fracture occurring either transgranularly, or by a thick
continuous glassy phase. Porosity is reduced by adding Sij before milling and
toughening fiom the microstructure can be seen to substantially increase (Figures 56 - 57).
When 10 VIC SiCW is added, the porosity and toughening differences due to ball-milling are
not as obvious on the fracture surface. Both appear tough with little porosity, however,
monolithic ball-milled Si3N4 also reveals a tough microstructure with virtually no porosity.
Average grain size appears to decrease with added Sij, however, this difference is very
slight. Green porosity had the most significant effect on grain size. Preferred grain growth
is obvious in B-Si3N4 and all specimens exhibited this characteristic. B-Si3N4 grows to fill
porosity and grain diameters of B-Si3N4 are substantially increased for grains growing into
porosity as shown in Figure 58 a) and b). This abnormal grain growth is especially
prevalent in ball-milled specimens processed at pH 3 (Figure 59 a) and b)). These pockets
of larger grain size material may be due to large acicular pre-existing pores that were nearly

filled by large precipitating B-Si3N4 grains. The pore structure for specimens produced

139

 

Figure 53 Fracture surface of monolithic ball-milled and ultrasonicated Si3N4
processed at pH 11.

140

 

Figure 54 Fracture surface of ball-milled and ultrasonicated Si3N4 with 10 V/oSij
added after milling. Processed at pH 1 1.

141

 

Fig“ re 55 Fracture surface of ball-milled and ultrasonicated Si3N4 with 20 v/oSij
added afier milling. Processed at pH 11.

142

 

Figure 56 Fracture surface of 10 v/() Sij reinforced Si3N4, all components ball-milled
and ultrasonicated at pH 11.

143

 

Figure 57 Fracture surface of 20 v/o Sij reinforced Si3N4, all components ball-milled
and ultrasonicated at pH 11.

144

 

- 1
<—-‘——— 11>

Figure 58a Image of typical porosity in sintered specimens. Shows elongated B-Si3N4
grains growing into and filling porosity.

145

H

31‘ u

 

Figure 58b Insert showing irregular B-Si3N4 grain growth within the porosity.

146

   

Figure 59a Micrograph showing a region of two distinct grain size regimes in
monolithic Si3N4 processed and ball-milled at pH 3.

147

 

Figure 59b Higher magnification micrograph of the large grain size region. The
morphology of B-Si3N4 appears much like that found in the porous
regions of other sintered specimens.

 

148

from pH 3 ball-milled powders is unique to all the other specimens, as shown in Figure 60
a) and b). This is the only major difference between fiacture surfaces as a function of pH.
All specimens show evidence of glass rich regions and inhomogeneities regardless of
processing history (Figure 61 a), b) and c).

It is extremely difficult to discern between B-Si3N4 and Sij on SEM fracture
surfaces. Their diameters are similar, however, B-Si3N4 exhibits a hexagonal cross-section
which is easy to identify when grains grow > lum (Figure 62). These larger grains are
often associated with porous or glass rich areas in the microstructure as was demonstrated in
Figure 58. Many toughening mechanisms are also evidenced on the fiacture surfaces.

Monolithic and 10 v/o reinforced materials show evidence of grain pull-out on their
fracture surfaces. Pull-out seems to be more evident on monolithic specimens as opposed
to whisker reinforced specimens. Areas of both crack bridging and crack deflection can be
isolated on secondary cracks on the surface of both monolithic and composite fracture
surfaces as was shown in Figure 51. These quantities can be more equivalently compared

from indentation induced cracking however.

149

new

 

Figure 60a Changes in pore morphology with processing pH. Acicular pore
morphology found in monolithic specimens ball-milled at pH 3.

150

 

110mm
/=/

 

Figure 60b Typical pore morphology for all other processing conditions.

15]

 

Figure 61a Figures a-c show regions of inhomogeneity found in specimens regardless of
processing pH. a) shows either glass rich regions, or large B-Si3N4 grain
cross-sections in a 10 "/o Sij specimen.

 

152

 

Figure 61b Shows glass rich-region in a monolithic specimen processed at pH 7.5.
area.

153

   

Figure 61c Shows a large glass region associated with a large pore in a 10 v/0 Sij
specimen.

Figure 62

[54

 

Micrograph showing large B-Si3N4 grains growing from a glass-rich region.
The hexagonal morphology becomes more obvious as grains grow larger
than 1 pm. The micrograph also evidences the growth of B-Si3N4 through
the liquid phase as the B-Si3N4 grain appears to be precipitating out from a
glass-rich

 

155

DISCUSSION

Green Processing of Si3N4/Sij Composites

The results of colloidal optimization are clearly demonstrated by sedimentation and
Andreason pipette sizing of agglomerates. Agglomerate sizes are largest and sedimentation
densities are lowest for pH 4-7 for Si3N4 and sedimentation densities are low for Sij for
pH < 11 (Figures 10 - 11). It is interesting to note that these numbers match exactly the
regions of homostability for Si3N4 and SiCW as predicted by Suspension Stability© [10] and
outlined in Figure 13. Stability predictions also correlate well with green density results for
specimens pressed without sintering aids. Those pressed at pH 6, where Si3N4 and Sij are
agglomerated, show the lowest green densities for monolithic and composite uni-axially
pressed specimens. Monolithic specimens pressed from powders processed at pH 3 and 11,
which showed the highest levels of stability, showed the highest green densities as well.
This is due to removal of agglomeration and improved homogeneity of the suspension,
which results in a freeze-dried powder with better pressing characteristics. Green densities
are highest for composite powders produced at pH 11 because the enhanced stability of SiC
whiskers provides fewer whisker agglomerates which impede green consolidation. From
the above results, it’s easy to see how applying colloidal stability theories (HI-IF/DLVO)

can provide consistent, higher green densities. However, these theories apply to

 

156

suspensions whose components are chemically inert with exception of the potential
determining ions [1]. If components are soluble in the processing environment, such as
Y203 in HNO3 [105], the ability of these theories to predict suspension behavior begins to
break down [101].

Reduction of agglomeration of colloidally produced ceramics can improve green
homogeneity by reducing the number of flaws caused by deficient particle packing.
Agglomeration can be reduced by manipulating particle charge to increase the double layer
interaction between particles [1,99,104]. In order to do this, it’s important to understand
how changes in the suspension medium affect the surface of Si3N4.

The charges present on the Si3N4 are determined by the active chemical groups on
the powder surface and how they change as aqueous suspension conditions change i.e.:
changes in pH (H+ and OH' concentrations). Crimp [112] used infrared spectroscopy to
characterize surface groups present on Ube SNE-lO Si3N4 powder. Results showed the
presence of both silanol (acidic behavior) and amine groups (basic behavior). The i.e.p. of
Si3N4 has been shown to decrease (or more closely resemble the behavior of SiOz) as the
surface 02 content increases [7]. A silica rich surface layer is the result of
thermodynamically unstable Si3N4 reacting with 02 to form SiOz.

Si3N4(s) + 302(g) <:> 3SiOz(s) + 2N2(g)
Hackley et al. [101], however, reports a surface of hydrolyzed Si3N4 (Figure 63) comprised
of acidic silanol and basic silazane (Si=NH). In aqueous solution, the silanol and silazane
generate surface charge by adsorption of H+ and OH which act as potential determining

ions. In addition to the p.d.i.s, the surface of Si3N4 can react with hydrolysable metal

157

 

 

Hydroxide ppt

Initial Si,N. surface New surface

Figure 63 Schematic showing metal hydroxide interaction with the Si3N4-solution
surface.

158

cations forming mono- and polynuclear complexes. Hackley suggests the following
reactions my occur with metal ions such as Y3+ at pH < 9, the pKa for Si2=NH2+. At more
alkaline pH, Si=NH may also
: Si-OH + MZ+ <—) : Si-OM‘Z'”+ + H“
or
2: Si-OH + MZ+ <—> (: Si-O)2 M‘Z‘z’+ + 211*

become involved in complexation. Hackley simulates composite suspension conditions by
electrokinetic titration of Si3N4 in the presence of Y(NO3)3 electrolyte. The surface of Si3N4
reacts with as little as 0.001M Y(NO3)3 at pH > 6 resulting in the precipitation of Y(OH)3
on the surface of Si3N4. As precipitation continues, the electrokinetic behavior (Q—potential)
of Si3N4 changes to that of Y(OH)3. While Y(OH)3 appears a logical candidate for the
precipitation that occurs during ball-milling at pH 7 and 11, it was not detected in the
residue or in green ball-milled specimens. Green specimens were also checked for the
presence of Y(NO3)3 revealing nothing. Therefore, while it is clear that YZO3 is soluble in
aqueous suspensions and especially in nitric acid solutions [101,105], it is still unclear what
chemical complexes are being formed. Understanding the chemical state of the YzO3,
however, is important because it clearly has an effect on green density and may also have an
effect on grain boundary phases in the sintered state.
Densification of Sij-Si3N4

Densities of tired specimens correlate well with composite green densities, showing
decreased densification as whisker loading increases. The lowest composite sintered

densities occurred at pH 7 despite the fact that the highest monolithic densities also

 

159

occurred at pH 7 (Figure 34). This can be explained by examining what is happening to
individual composite components at pH 7. First, Sij is agglomerated at pH 7 which will
hamper liquid phase sintering and reduce final densities of the composite. It would also
seem that because the Si3N4 matrix is also agglomerated at pH 7 that this should reduce
sintered densities of the monolithic material. However, chemical effects during ball-milling
may be hampering the densification of specimens produced at pH 3 and 11 more so than the
agglomeration at pH 7. In order to see if the ball-milling residue had an adverse effect on
densification, suspensions were milled and most of the skin that formed on top of the
suspension was removed to lessen the relative amount in the freeze-dried powder. The
effect this had on density, densification time, and fracture toughness is outlined in Figures
64 - 65. Specimens sintered more quickly and to higher densities when the ball-milling
residue was reduced. However, fracture toughness decreased. While fracture touglmess can
be improved by reduction in flaw or porosity size and fiequency, the Km of liquid phase
sintered ceramics is highly dependent upon the interface between grains, both matrix
/matrix, and matrix/whisker. Therefore, the fracture energy of the grain boundary phases

present play a key role in the toughness properties of both monolithic and composite Si3N4.

 

160

 

 

 

Density (g/cma)
Fracture Toughness (MPa m1 l2)

 

 

 

 

 

 

 

 

Fracture Toughness

 

Figure 64 Comparison of density and fiacture toughness for specimens freeze-dried
with and without ball-milling skin.

 

161

 

glcm3 % Theoretical Density

 

  
 

.- 97.28%
i

3.2 1'

- 94.24%

 

 

 

3 .L i 91.2%
1 2 3 4
Time (hours)

 

 

 

 

+Monolithic 'A‘Milled Monolithic *Modified Milling

 

 

 

Figure 65 Comparison of densification with time for specimens fi'eeze-dried with and
without ball-milling skin. Standard deviation for each data point is too
small to be seen clearly on the plot.

162

CONCLUSIONS

The suspension behavior of as-received Si3N4 and Sij powders can be accurately
modeled using a modified version of DLVO/HHF theory [10] to predict the stability of both
monolithic and composite suspensions. However, when hydrolysable metal ions are
introduced into the suspension, such as the case when Y203 is added to suspensions,
changes to the surface chemistry [101], and hence, surface charge of constituents in
suspension become difficult to predict. As a consequence, green density no longer has a
direct correlation with suspension stability. YZO3 may react in suspension to produce
currently uncharacterized compounds in the green state. Thus, hampering the ability to
isolate the variables that most directly effect the sinterability and toughness of Si3N4 and
Sij-Si3N4. HNO3 acid solutions are capable of completely solutionizing YZO3 during
green processing, and basic solutions produce a residue during milling which hampers
densification, but may produce a grain boundary phase that enhances fracture toughness.
The addition of Sij to Si3N4 also has an effect on grain boundary phases. Compacts
produced from powders at pH 11 with Sij produce crystalline YZO3-Si3N4 which is not
present in their monolithic counterparts.

Ball-milled Si3N4 specimens reinforced with 10% as-received (high aspect ratio)

Sij produced the highest fracture toughness value of 9.1 MParrr”2 slightly higher than

163

monolithic Si3N4. However, both toughness values could be improved with larger grain
size [65]. SEM analysis of fracture surfaces not only evidenced crack deflection, pull-out,
and bridging for both monolithic and composite specimens, but shows wide spread
inhomogeneities in specimens produced at all pH values (Figures #61 a-c). Finally,
chemical reactions occuning during colloidal processing can impact the final properties of
sintered compacts by affecting liquid formation during sintering, and adding to the already
large number of parameters that must be taken into account when sintering any ceramic

composite.

164

RECOMMENDATIONS

In order to gain more control over the final sintered microstructure, sintering aids
could be added as Y-Al-Silicate glasses. Pre-prepared glasses (or crystalline phases such as
YAG) could be made with varying composition and added to Si3N4 in a powder form.
Materials prepared as sintering aids could also be mechanically tested to determine which
compositions may have the greatest toughening effect, or perhaps the greatest high
temperature strength. Colloidal processing and optimization could still be carried out with
these sintering aids though Q—potential testing and stability prediction. Solubility testing
could also be carried out to determine the compositions most suitable for colloidal
processing. In addition, changes in the sintering schedule could be made to vary the final

grain size of both monolithic and composite specimens to evaluate further toughening.

m

 

i"

BIBLIOGRAPHY

[N

 

10.

11.

165

BIBLIOGRAPHY

Paul C. Hiemenz, BrineinlemfeolloidandSurfaeLChemim, Marcel Dekker,
New York, New York, 1986

EM. Elirnelech, “Kinetics of Capture of Colloidal Particles in Packed Beds Under
Attractive Double Layer Interactions”, J. Col. Int. Sci, [146], p.337, 1991.

EM. Lifshitz, Soviet Phys. JETP, [2], p. 73, 1956.

Alan Bleier, “Fundamentals of Preparing Suspensions of Silicon and Related
Ceramic Powders”, J. Amer. Ceram. Soc, [66], pp. C79-C81.

Duncan J . Shaw, WWW Butterworth-
Heinemann, Oxford, 1992.

Michael James, Robert J. Hunter, and Richard W. O’Brien, “Effect of Particle Size
Distribution and Aggregation on Electroacoustic Measurements of 6; Potential”,
Langmuir, [8], pp. 420-23, 1992.

Egon Matijevic, “Colloid Science of Composite systems”, Serengegffieramrg
CW, Larry L. Hench and Donald R. Ulrich eds. ,.pp 463- 81, John
Wiley and Sons, New York 1986.

R.W. O’Brien, “Electro-Acoustic Effects in a Dilute suspension of Spherical
Particles”, J. Fluid Mech, [190], pp. 71-86, 1988.

M.J. Crimp, MW, Case Western University, 1985.

BA. Wilson and M.J. Crimp, “Prediction of Composite Colloidal Suspension
Stability Based upon the Hogg, Healy, and F uerstenau Interpretation”, Langmuir,
[9], pp. 2836-43, 1993. ~

Derjaguin, Kolloid—Z, [69], p. 155, 1934; Acta Physiochim., [10], p. 333, 1939.

r“:

 

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

166

R. Hogg, T.W. Healy, and D.W. Fuerstenau, “Mutual Coagulation of Colloidal
Dispersions”,Trans. Faraday Soc, [62], pp. 1638-50, 1960. *

Verwey and Overbeek, W, Elsevier, New York, 1948.

Fuchs, Z. Physik, [89], p. 736, 1934.

J. T. G. Overbeek, “Kinetics of Flocculation” in W
Inmmhlefiystems, HR. KruytEd., pp. 278- 301, Elsevier, New York, 1952.

M.W. Real, “Freeze Drying Alumina Powders”, Proc. Brit. Ceram. Soc. , [38], p. {
59-66, 1986.

JD. Mellor, W, Academic Press, London, 1978.

Labconco Industry Service Publication, WW
Lalmatom, 1993-

RE. Zovas, R.M. German, K.S. Hwang, and Cl. Li, "Activated and Liquid-Phase
Sintering--Progress and Problems", J. Metals, [35], p. 28-33, 1983.

 

W.D. Kingery, "Densification During Sintering in the Presence of a Liquid Phase. 1.
Theory", J. Appl. Phys. [30], p. 301-306, 1959.

RM. German, Liqmdfihasefiimefing, Plenum press, New York and London, 1985.

W.D. Kingery, E. Niki, and MD. Narasirnhan, "Sintering of Oxide and Carbide-
Metal Compositions in Presence of a Liquid Phase", J. Amer. Ceram. Soc., [44], p.
29-35, 1961.

W.D. Kingery and MD. Narasirnhan, "Densification During Sintering in the
Presence of a Liquid Phase. 2. Experimen ", J. Appl. Phys., [30], p. 307-310,
1959.

Yu. V. Naidich, I.A. Lavrinenko, and VA. Evdokimov, "Densification During
Liquid-Phase Sintering in Diamond-Metal Systems", Soviet Powder Met. Metal
Ceram., [11], p.715-718, 1972.

V. Smolej, S. Pefovnik, and WA. Kaysser, "Rearrangement During Liquid Phase
Sintering of Large Particles", Powder Met. Inter., [14], p. 34-36, 1982.

W.J. Huppmann, W.A. Kaysser, D.N. Yoon, and G. Petzow,"Progress in Liquid
Phase Sintering", Powder Met. Inter. ,[ll], p. 50-51, 1979.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

167

Duk N. Yoon and Winfiied J. Huppmann, "Grain Growth and Densification During
Liquid Phase Sintering of W-Ni", Acta Met. , [27], p. 693-98, 1979.

G.H. Gessinger, H.F. F ischmeister, and H.L. Lukas, "A Model for Second-Stage
Liquid-Phase Sintering with A Partially Wetting Liquid", Acta Met, [21], p. 715-24,
1973.

WA. Kaysser, M. Zivkovic, and G. Petzow, "Shape accomodation during grain
growth in the presence of a liquid phase", J. of Mat. Sci, [20], p. 578-84, 1985.

Rishi Raj, "Morphology and Stability of Glass Phase in Glass-Ceramic Systems", J.
Amer. Ceram. Soc, [64], p. 245-48, 1981. I

L.K.V. Lou, T.E. Mitchell, and AH. Heuer, "Impurity Phases in Hot-Pressed
Si3N4", J. Amer. Ceram. Soc, [61], p. 392-96, 1978.

TH. Courtney, "Densification and Structural Development in Liquid Phase
Sintering", Met. Trans. A, [15A], p.1065-74, 1984.

Hyo-Hoon Park, Seong-Jai Cho, and Duk N. Yoon, "Pore Filling Process in Liquid
Phase Sintering", Met. Trans. A, [15A], p.1075-80, 1984.

H. Dean Batha and E. Dow Whitney, “Kinetics and Mechanism of the Thermal
Decomposition of Si3N4”, J Amer. Ceram. Soc, [57], p. 365-69, 1973.

F .W. Dynys and J .W. Halloran, "Influence of Aggregates on Sintering", J. Amer.
Ceram. Soc, [67], p. 596-601, 1984.

F .F . Lange, “Sinterability of Agglomerated Powders”, J. Amer. Ceram. Soc, [67],
p. 83-89, 1983.

SS. Kang and D.N. Yoon, “Kinetics of Grain Coarsening During Sintering of Co-
Cu and F e-Cu Alloys wiht Low Liquid Contents”, Metall. Trans. A, [13A], p. 1405-
11, 1982.

J. XinXiang and R. Taylor, “Dispersion of SiC Whiskers in SiC Whisker
Reinforced Si3N4 Composite and Related Microstructural Characteristics”, Brit.
Ceram. Trans, [93], p. 11-15, 1994.

01. Shih, J .M. Yang, and A. Ezis, “Microstructure and Properties of Reaction-
Bonded/Hot-Pressed Sij/Si3N4 Composites”, Composites Sci. and T ech., [43], pp.
13-23, 1992.

W.H. Rhodes, “Agglomerate and Particle Size Effects on Sintering Yttria-Stabilized
Zirconia”, J. Amer. Ceram. Soc, [64], p. 19-22, 1981.

 

.r'i

41.

42.

43.

45.

46.

47.

48.

49.

50.

51.

52.

168

J .N. Brecker, “Analysis of Bond Formation in Vitrified Abrasive Wheels”, J. Amer.
Ceram. Soc., [57], p. 486-89, 1974.

GM. Crosbie, J .M. Nicholson, and ED. Stiles, “Sintering Factors for a Dry-Milled
Silicon Nitride-Yttria—Altunina Compostion”, Ceram. Soc. Bull, [68], p. 1202-06,
1989.

M. Kramer, “Rate-Controlled Sintering of Si3N4: Densification Kinetics and
Microstructural Development”, J. Mat. Sci. Letters, [14], p. 778-80, 1995.

M.K. Cinibulk, G. Thomas, and SM. Johnson, “Grain-Boundary-Phase
Crystallization and Strength of Silicon Nitride with a YSiAlON Glass”, J. Amer.
Ceram. Soc, [73], p.1606-12, 1990.

G.R. Terwilliger and RF. Lange, "Pressureless Sintering of Si3N4", J. of Mat. Sci,
[10], p. 1169-74, 1975.

RM. German and KS. Churn, “Sintering Atmosphere Effects on the Ductility of
W-Ni-Fe Heavy Metals”, Metall. Trans. A, [15A], p. 747-54, 1984.

OJ. Shih, J .M. Yang, and A. Ezis, “Microstructure and Properties of Reaction-
Bonded/Hot-Pressed Sij/Si3N4 Composites”, Composites Sci. and Tech, [43], pp.
13-23, 1992.

Murat Bengisu and Osman T. Inal, "Whisker Toughening of Ceramics. Toughening
Mechanisms, Fabrication and Composite Properties", AnnuaLRexiemeetefiali
Seienee, M. C. Flemmings, J. B. Watchman, Jr. E. N. Kaufmann, and J. A.
Gioordmaine, Eds., [24], p. 83-124, Annual Reviews Inc., Palo Alto, Ca., USA.

X. Y. Zheng, F. P. Zeng, M. J. Pomeroy, and S. Hampshire, "Reingorcement of
Silicon Nitride Ceramics with Whiskers and Platelets", Bfifishfleramieheeeedings

Eahrieatienleehnelegy, R. W. Davidge and D. P. Thompson Eds., [45] p. 187- 98,
The Institute of Ceramics Shelton, Stoke-on-Trent, Staffs. U. K. 1990.

S.T. Buljan, J .G. Baldoni, and ML. Huckabee, “Si3N4-SiC Composites”, Am.
Ceram. Soc. Bull, [66], pp. 347-52, 1987.

H. F. Priest, G. L. Priest, and G. E. Gazza, "Sintering of Si3N4 Under High Nitrogen
Pressure", J. Amer. Ceram. Soc, [60], p. 81,1977.

0. Abe, “Sintering Process of Y203 and A1203-doped Si3N4”, J. Mat. Sci, [25], pp.
4018-26, 1990.

 

53.

54.

55.

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

169

M. Mitomo, M. Tsutsumi, E. Bannai, and T. Tanaka, "Sintering of Si3N4", Ceram.
Soc. Bull, [55], p. 313, 1976.

G.R. Terwilliger, “Properties of Sintered Si3N4”, J. Amer. Ceram. Soc, [57], p. 48-
49, 1973.

F .F . Lange, "Fabrication and Properties of Dense Polyphase Silicon Nitride",
Ceram. Soc. Bull, [63], p.1369-74, 1983.

GE. Gazza, "Hot-Pressed Si3N4", J. Amer. Ceram. Soc, [56], p. 662, 1973.

C. Greskovich and C. O’Clair, “Effect of Impurities on Sintering Si3N4 Containing
MgO or YZO3 Additives”, Ceram. Soc. Bull, [57], p. 1055-56, 1978.

K. Ueno, S. Sodeoka, and J. Hirooka, “Influence of Sintering Conditions on the
Mechanical Properties of SiC Whisker/Si3N4 Ceramic Composite”,lournal of
Ceramic Society of Japan, Int. Ed, [100], p. 525-29, 1992.

RE. Loehman and DJ. Rowcliffe, "Sintering of Si3N4-YZO3-A1203", J. Amer.
Ceram. Soc, [63], p. 144-48, 1980.

G.Q. Weaver and J .W. Lucek, “Optimization of Hot-Pressed Si3N4-YZO3
Materials”, Ceram. Soc. Bull, [57], p. 1131-36, 1978.

M. Mitomo, “Thermodynamics, Phase Relations, and Sintering Aids of Si3N4”,
W, S. Somiya, M. Mitomo, and M. Yoshirnura, Eds, p. 1-11, Elsevier
Applied Science, New York, 1990.

K. Kato, Z. Inoue, K. Kijirna, I. Kawada, H. Tanaka, and T. Yamane, “Structural
Approach to the Problem of Oxygen Content in Alpha Silicon Nitride”, J. Amer.
Ceram. Soc, [58], p. 90-91, 1975.

K. Ichikawa, “The Influence of Si3N4 Powder Characteristics on Sintering
Behavior”, W, S. Somiya, M. Mitomo, and M. Yoshirnura, Eds., pp.
107-16, Elsevier Applied Science, New York, 1990.

G. Wotting and G. Ziegler, "Powder Characteristics and Sintering Behavior of
Si3N4-Powders", Powder Met. Int, [18], no.1, p. 25-32, 1986.

A. J. Pyzik and D. F. Carroll, "Technology of Self-Reinforced Silicon Nitride",
WW, MC. Flemmings, J. B. Watchman, Jr., E..N
Kaufrnann, and J. A. Gioordmaine, Eds, [24], p. 189- 214, Annual Reviews Inc.,
Palo Alto, Ca, USA, 1995.

 

66.

67.

68.

69.

71.

72.

73.

74.

75.

76.

77.

78.

170

H. Liu, Karl-L. Weisskopf, and G. Petzow, "Crack Deflection Process for Hot-
Pressed Whisker-Reinforced Ceramic Composites", J. Amer. Ceram. Soc., [72],
p. 559-63, 1989.

DB. Marshall and J .E. Ritter, "Reliability of Advanced Structural Ceramics and
Ceramic Matrix Composites-A Review", Am. Ceram. Soc. Bull., [66], p. 309-17,
1987.

J .E. Weston, P.L. Pratt, and B.C.H. Steele, “Crystallization of Grain boundary
Phases in Hot-Pressed Silicon Nitride Materials”, J. Mat. Sci. , [13], p. 2137-46,
1978.

A. Tsuge, K. Nishida and M. Komatsu, “Effect of Crystallizing the Grain-
Boundary Glass Phase on the High-Temperature Strength of Hot-Pressed Si3N4
Containing Y203”, J. Amer. Ceram. Soc., [58], p. 323-26, 1975.

RE. Loehman, "Preparation and Properties of Yttrium-Silicon-Aluminum
Oxynitride Glasses", J Amer. Ceram. Soc., [62], p. 491-94, 1979.

R.K. Bordia, R. Raj, “Analysis of Sintering of a Composite with a Glass or Ceramic

Matrix”, J Amer. Ceram. Soc., [69], p.C55-C58, 1986.

Al Whitehead, T.F. Page, and I. Higgins, “Degradation of SiC Whiskers at
Elevated Temperatures”, Ceram. Eng. Sci. Proc., [10], pp. 986-97, 1989.

PD. Shalek, J .J . Petrovic, G.F. Hurley, and F .D. Gac, “Hot-Pressed SiC Whisker/
Si3N4 Matrix Composites”, Ceram. Soc. Bull, [65], p. 351-56, 1986.

SM. Smith, R.O. Scattergood, J .P. Singh, and K. Karasek, “Effect of Silica and
Processing Environment on the Toughness of A1203 Composites”, J. Amer.
Ceram. Soc., [72], pp. 1252-55.

Stephen W. Freirnan, "Brittle Fracture Behavior of Ceramics", Ceram. Soc. Bull.,
[67], p.392-402, 1988.

G.R. Anstis, P. Chantikul, B.R. Lawn, and DB. Marshall, "A Critical Evaluation
of Indentation Techniques for Measuring Fracture Toughness: 1, Direct Crack
Measurements", J. Amer. Ceram. Soc., [64], p. 533-538, 1981.

S. Palmqvist, "Occurrence of Crack Formation During Vickers Indentation as a
Measure of the Toughness of Hard Metals", Arch. Eisenhuttenwes., [33], p. 629-
633, 1962.

 

79.

80.

81.

82.

83.

84.

85.

86.

87.

88.

89.

90.

91.

171

AG. Evans, WWW; Ed. S.W. Freiman,
ASTM Spec. Tech. Publ., No. 678, Philadelphia, pp. 112-35, 1979.

RF. Cook and GM. Pharr, "Direct Observation and Analysis of Indentation
Cracking in Glasses and Ceramics", J Amer. Ceram. Soc., [73],pp. 787-817, 1990.

MS. Kaliszewski, G. Behrens, A.H. Heuer, M.C. Shaw, D.B. Marshall, G.W.
Dransmann, R.W. Steinbrech, A.Pafares, F. Guiberteau, F .L. Curnbrera, and Arturo
Dominguez-Rodriguez, "Indentation Studies on Y203 -Stabilized ZrOZ: 1,
Development of Indentation- Induced Cracks", J Amer. Ceram. Soc., [77], p.
1185-93, 1994.

C. Olagnon, E. Bullock, and G. Fantozzi, “Properties of Sintered SiC Whisker-
Reinforced Si3N4 Composites”, J Euro. Ceram. Soc., [7] p. 265-73, 1991.

mismananaLEncxelopediaoLCmnpositesXQLfl, Stuart M. Lee, Ed., VHC, New
York, New York, 1990.

A.G. Evans, "Perspective on the Development of High-Toughness Ceramics", J
Amer. Ceram. Soc., [73], p. 187-206, 1990.

D.H. Carter and GP. Hurley, "Crack Deflection as a Toughening Mechanism in
SiC-Whisker-Reinforced MoSizF", J. Amer. Ceram. Soc., [70], p. C79-C81, 1987.

K.T. Faber and A.G. Evans, "Crack Deflection Processes-I. Theory & II.
Experimen ", Acta Metall., [31], p. 565-84, 1983.

G.H. Campbell, M. Ruhle, B.J. Dalgleish, and A.G. Evans, “Whisker Toughening:
A Camparison Between Aluminum Oxide and Silicon Nitride Toughened with
Silicon Carbide”, J Amer. Ceram. Soc., [73], p. 521-30, 1990.

A.G. Evans and Y. Fu, "Some Effects of Microcracks on the Mechanical
Properties of Brittle Solids-II. Microcrack Toughening", Acta Met, [33], p. 1525-
31, 1985.

A.G. Evans and KT. Faber, “Crack-Growth Resistance of Microcracking Brittle
Materials”, J Amer. Ceram. Soc., [67] p.255-60, 1984.

MD. Thouless and A.G. Evans, "Effects of Pull-Out on the Mechanical
Properties of Ceramic-Matrix Composites", Acta Met., [36], p. 517-22, 1988.

P. Sajgalik, J. Dusza, and M.J. Hoffrnann, "Relationshop between Microstructure,
Toughening Mechanisms, and Gracture Toughness of Reinforced Silicon Nitride
Ceramics", J Amer. Ceram. Soc., [78] p. 2619-24, 1995.

92.

93.

94.

95.

96.

97.

98.

99.

100.

101.

102.

103.

I O4.

172

RP. Becher, E.R. Fuller Jr., and Peter Angelini, "Matrix-Grain-Bridging
Contributions to the Toughness of Whisker-Reinforced Ceramics", J Amer. Ceram.
Soc., [74], p. 2131-35, 1991.

Paul F. Becher and George C. Wei, “Toughening Behavior in SiC-Whisker-
Reinforced Alumina”, J Amer. Ceram. Soc., [67], p. C267-C269, 1984.

RF. Becher, Chun-Hway Hsueh, P. Angelini, and TN. Tiegs, "Toughening
Behavior in Whisker-Reinforced Ceramic Matrix Composites", J Amer. Ceram.
Soc., [71], p. 1050-61, 1988.

AA. Morrone, S.R. Nutt, and S. Suresh, “Fracture Toughness and Fatigue Crack
Growth behavior of an Ale3-SiC Composite”, J Mat. Sci, [23], pp. 3206-13,
1988.

J. Homeny, W.L. Vaughn, and MK. F erber, “Silicon Carbide Whisker/Alumina
Matrix Composites: Effect of Whisker Surface Treatment on Fracture Toughness”,
J Amer. Ceram. Soc., [73], pp. 394-402, 1990.

Personal Communication with Brett Wilson

M. J. Crimp and R. F. Preston, “Evaluation of Sidimentation as a Technique for the
Separation of Sub-Micron Diameter Silicon Carbide Whiskers”, Britishfleremre

PreeeedmgeEabneanenleehnelegy, R. W. Davidge and D. P. Thompson Eds, [45]
p. 187- 98, The Institute of Ceramics Shelton, Stoke-on-Trent, Staffs. U. K, 1990.

R.J. Hunter, Emmdaflenejnflelleiifieieneem, Oxford University Press,
Oxford, 1986.

Z. Yeh, P.R. Sneary, and M.J. Crimp, “Deagglomeration of Si3N4-Sij Colloidal
Systems”, submitted to J Amer. Ceram. Soc, May, 1996.

VA. Hackley and 8.6. Malghan, “The Surface Chemistry of Silicon Nitride Powder
in the Presence of Dissolved Ions”, J Mat. Sci, [29], pp. 4420-30, 1994.

ASTM C 373-88, “Strandard Test Method for Water Adsorption, Bulk Density,
Apparent Porosity, and Apparent Specific Gravity of Fired Whiteware Products”,
1988.

CF. Gazzara and DR. Messier, “Determination of Phase content of Si3N4 by X-
Ray Diffraction Analysis”, Amer. Ceram. Soc. Bull, [56], pp. 777-80, 1977.

Robert J. Hunter, ZejePetenfiaLinflellgjiSeienee, Academic Press, London, 1981.

105.

106.

107.

108.

109.

110.

111.

112.

173

William R. Linke, SolubrlrtresoflnorgameandMetalflrgamecompoundsfl,
American Chemical Society, Washington, DC, 1965.

M. Yang and R. Stevens, “Fabrication of SiC Whisker Reinforced A1203
Composites”, J Mat. Sci, [25], pp. 4658-66, 1990.

Robert R. Thompson, “Mechanics of Powder Pressing: II, F inite-Element Analysis
of End-Capping in Pressed Green Powders”, Amer. Ceram. Soc. Bull. , [60], pp. 244-
51, 1981.

PD. Shalek, J .J . Petrovic, G.F. Hurley, and FD. Gac, “Hot-Pressed SiC
Whisker/Si3N4 Matrix Composites”, Amer. Ceram. Soc. Bull, [65], pp. 351-56,
1986.

S.T.J. Chen and J .M. Yang, “Quantitative Determination of the Stability of SiC
Whisker in Reaction-Bonded/Hot Pressed Si3N4 Composites by X-Ray Diffraction”,
J Mat. Sci, [27], pp. 5123-32, 1992.

A. W. J. M. Rae, D. P. Thompson, and K. H. Jack, “The Role of Additives 1n the
Densification of Nitrogen Ceramics”, in Ceramreifiorflghfiequnnanee
Applieatienle, John J. Burke, Edward N. Lenoe, and R. Nathan Katz, Eds. ,.pp
1039- 67, Brook Hill Publishing Co., Chestnut Hill, Mass, 1978.

C.J. Shih, J .M. Yang, and A. Ezis, “Processing of SiC Whisker-Reinforced Si3N4
Composites”, Ceram. Eng. Sci. Proc., [10], pp. 1064-71, 1989.

Byong-Taek Lee and K. Hiraga, “Microstructure and Micro-Indentation Fracture
of SiC-Whisker-Reinforced Si3N4 Composite Studied by High-Resolution Electron
Microscopy”, Materials Transactions, JIM, [34], pp. 930-36, 1993.