Ceramic Powder Processing Insights
Ceramic Powder Processing Insights
, 72 [ I ] 3-15 (1989)
journal
Powder Processing Science and
Technology for Increased Reliability
Fred F. Lunge*
Materials Department, College of Engineering,
University of California, Santa Barbara, California 93106
- Flocced (pH 7)
storage even when acted upon by cen-
trifugal forces.
has the potentid to access mixing unifor-
mity during processing as a nondestruc-
+ ,030
5
N
Figure 2 also shows that two or more
0
-
0.6- *-*-By
-8
.0.25 R powder phases, separately treated as
tive evaluation tool.
When a multiphase body is observed
-t-J
2 0.4- 0 summarized in the preceding paragraph,
N -0.20 ‘Z by SEM, the X-rays collected produce an
2 can be mixed to form multiphase slurries.
-0.15 EDX spectrum that quantitatively defines
0 2- -4” If the different slurries are colloidally com-
Suspended (pH 2.5) E the atomic fraction of each element. If
.a10 I patible, i.e., do not floc one another, they
8 different elements are associated with
can be redispersed (by again adding the
O i I -0 different phases, the content of each
0 0.2 0.4 0.6 0.8 10 2 proper surface agent) and mixed. More
phase within the area scanned can be de-
t Normalized distance t commonly, the phases are not colloidally
termined. At low magnifications,the EDX
Bottom TOP compatible. These systems can still be
spectrum defines the fraction of each
Fig. 3. ZrlAl count raho (obtained from an mixed because flocced mixtures can be
phase within the large body. With reasona-
EDX spectra) versus the normalized distance mechanically redispersed by a device that
ble counting periods, the standard devia-
detailing the ZrOP and Al2O3 distibution for produces a high shear-ratefield (an ultra-
tion for different areas examined at low
mixed powder slurries centrifuged in the dis- sonic horn, high-speedrotor, etc.). As the
magnifications is low (usually <30) and
persed and flocced states mechanically dispersed mixture leaves the
associated with counting statistics. At very
high shear-rate field, it flocs to form a new
high magnifications, the area examined
may not be representative of the large
body; e.g., the scanned area may only
cover one of the many phases, and the
+Both A1203 and Zr02 can be dispersed at pH 2
and are colloidally compatible when mixed, I e , they deviation of the spectrum, relative to the
remain dispersed Both can be flocced at pH 8 large body, can be very large. At some
January 1989 Powder Processing Science and Technology for Increased Reliability I
intermediate magnification, the standard also shown that the volume fraction of par-
deviation will begin to depart from that ticulates can be further increased with the
produced by counting statistics. At this proper particulate-size distribution.18 As
magnification,the area scanned is statisti- shown in Fig. 5, the addition of a given
cally identical (within an acceptable stan- fraction of finer particulatesdecreasesthe
dard deviation somewhat larger than that slurry viscosity.
due to counting statistics) with the large Once molded, the rheological proper-
body. The size of this area thus defines ties of the slurry must be dramatically al-
the smallest area that contains the same tered to allow shape retention during
phase distribution as the whole body. This unmolding. This change can be induced
area (A,) can be defined quantitatively by a variety of different phenomena that
and used to represent phase uniformity. include freezing (used in injection mold-
The better the mixing, the smaller A,. A, ing), gelation,lg and in situ flocculation.
is an extrinsic property of the multiphase Each can change a flowable slurry into a
material that depends on processing.15 firm body without fluid-phase removal.
Values of A, can be related to different Rheological changes are not without prob-
mixing methods and mixing periods, e.g., lems; they can produce strain gradients.
different resident periods for mixed slur- For example, volume changes can ac-
ries within a high shear rate field, and to company freezing, and freezing initiates
the properties of the mixed slurry itself, at the surface. In addition, if capillary pres-
Lange and Miller16 have shown that A, sure arises during fluid removal, bodies
can be related to the viscosity of flocced, molded with dispersed slurries will shrink
two-phase slurries, and they have sug- unless an attractive, touching-particle net-
gested that an in-line viscosity measure- work is formed first (e.g., by in situ floccu-
ment could be used to determine phase lation). Shrinkage and expansion gradients
uniformity (defined by a number, i.e.,A,) can lead to significant stresses and/or dis-
during processing, ruptive phenomena. The potential prob-
lem of mass and phase segregation
IV. Consolidation from the produced by sedimentation within a highly
Slurry State filled, moldable dispersed slurry needs fur-
Once colloidally treated, powders ther study.
should not be dried. Slurries contain soh- Moldable slurries can also be achieved
ble salts, produced, e.g., by a reaction when flocced bodies are rapidly sheared.
with the powder itself, which can bond Forces produced by differential acceler-
touching particles when the last bit of li- ations break apart the attractive particle
quid evaporates from their pendular ring. network to produce Newtonian rheology;
That is, agglomerates, previously eliminat- the network reforms shortly after the shear
ed by colloidal treatment, will reform dur- field is removed. Firm, fully saturated bod- Fig. 4. Micrographs of dense, AI2O3-ZrO2
ing drying. Drying also reexposes the ies containing between 50 and 60 vol% composite ceramic formed by centrifuging a
particles to uncontrolled environments particulates can be produced by either dispersed slurry, illustrating particle-sizeand
which can reintroduce inclusions. Colloi- centrifuging14 or pressure filtering20 low- phase distributions at the (A) bottom and (B)
dally treated slurries could be piped direct- top of the specimen
er viscosity, flocced slurries. When these
ly to a consolidation machine. firm bodies are coupled to high-intensity
A major issue for exploiting the colloi- ultrasound, they fluidize to fill a cavity.
dal approach is to directly form powder Once the ultrasound is removed, the rheol-
compacts from the slurry state. Two con- ogy of the slurry quickly reverts to a firm,
ventional slurry consolidation methods, slip molded body. The author's experience
casting and tape casting, can be directly with this innovative molding method sug-
used with colloidally prepared powders, gests that attenuation of the ultrasound is
and, with some innovative changes, injec- one problem that must be addressed.
tion molding could be adapted to do the Pressure filtration produces a fully satu-
same. Each of these conventional forming rated powder compact as particles within
methods is limited, e.g., slip casting best a slurry are partitioned as liquid flows
produces thin-walled bodies, and a large through a filter leaving particles behind to
amount of polymer must be removed af- form a consolidated layer. Hard ferrite par-
ter injection molding. For these reasons, ticles within a slurry can be aligned with
investigators are exploring alternative a magnetic field prior to consolidation.
slurry-shaping methods where either the Pressure filtration is used to form ceramic
solidlliquid ratio remains constant during magnets with a variety of applications be-
shaping, i.e., slurry molding, or the parti- cause the desiredlpermanant magnetic
cles are partially partitioned from the liquid field is produced during consolidation.21 Fines fraction
during shaping, e.g., pressure filtration. With the development of moldable cast-
To avoid excessive shrinkage during flu- ing dies made of a porous plastic materi- Fig. 5. Viscosity (extrapolatedto zero shear
id removal andlor densification, molding al, an innovative pressure filtration rate) versus fine fraction for A1203 (0.55 volume
methods require slurries containing the fraction) slurries, dispersed with a polyelectro-
machine has recently been introduced22 lyte, containing coarse (0.8 pm) and fine (0.2
highest possible fraction of particles. Mold- that enables the rapid production of large, pm) powders.18
ing also requires flow. As detailed by Ak- relatively complex shaped bodies. Al-
say and co-workers,17 pourable ceramic though these new filter presses are cur-
slurries containing in excess of 60 vol% rently marketed to produce functional
of particulates require highly repulsive in- articles (from plates to sinks) with traditional
terparticle forces. In addition, they have clay-based slurries, they represent the first
8 Journal of the American Ceramic Society-Lange Vol. 12, No. 1
/-
).65-
c p consolidated layer (h)formed under con- particles have not formed a continuous
a,
stant pressure ( f )is parabolically related network (the case of a dispersed slurry),
Floeced slurry
$
5 so-
e 55- lnterpanlcie
~IepuISIve
to time (f) by the slurry pressure is identical with the fluid
a IOICeS)
pressure and equal to the pressure (Pa)
50-
y exerted by the plunger. Ambient, at-
45f 9 I mospheric pressure Po exists on the ex-
ternal side of the filter. The differential
where u is the liquid viscosity and vo and pressure across the filter and consolidat-
v/ are the volume fractions of particulates ed layer (fa- Po) is the driving force for
within the slurry and consolidated layer, fluid flow. A gradient in fluid pressure ex-
respectively The permeability (k) is in- ists across the consolidated layer; i.e., at
versely related to the resistance of fluid the slurry-layer interface, f f = f a ,and,
flow through the consolidated layer As- neglecting the pressure gradient across
suming that the particles are identical the filter, f r = Po at the filter-layer inter-
spheres with diameter d , one can estimate face. Because the total pressure within the
the permeability of a consolidated layer consolidated layer must be equal to the
with the Kozeny-Carman relation, which applied pressure, the particle network
models the layer as a bundle of tortuous must support a pressure gradient equal,
capillary tubes with hydraulic diameters but opposite, to the fluid pressure. That is,
resembling slits 2 at the filter-layer interface, the network
pressure, P, = f a ,and, at the slurry-
d*(1 - v,)3 layer interface, f , = 0. Because network
k = pressure will produce particle rearrange-
36c$ ment, during filtration, the packing densi-
ty will be greatest at the filter-layer
where the Kozeny constant (c) defines the
interface and decrease to the layer-slurry
shape and tortuosity of the flow chanels
interface. Once the plunger meets the
(c = 5 for many systems).
consolidation layer, filtration will continue
until the differential fluid pressure dissi-
Applied pressure = Pa n pates; i.e., the fluid pressure decreases to
ambient pressure. During this period, the
gradient in the network pressure dissipates
to the applied pressure to produce a uni-
form particle-packingdensity across the
consolidation layer.
Liquid pressure = f a With the information presented in the
preceding paragraph and Fig. 6, note
Network pressure = Po that, during pressure filtration, flocced slur-
Liquid pressure = Po ries produce large gradients in packing
Network pressure = Pa density relative to dispersed slurries. For
this case, there is no clear demarcation
between the slurry and the consolidated
Ambient pressure = Po layer as there is for dispersed slurries.
Schematic of pressure distribution The mechanics of particle packing dur- The mechanics of particle packing have
during pressure filtrationshowing that a pres ing filtration and the mechanics of only been addressed with static models,26
sure gradient exists wlthln the network Of Con pressure-filteredbodies have been exam. which does not address particle re-
solidated particles ined for aqueous slurries containing A1203 arrangement.Why dispersed slurries pro-
powder with a mean diameter of approx- duce high packing densities at very low
imately 0 6 pm 20 The electrostatic meth- pressures has not yet been detailed, but
od (pH control) was used to produce it must be related to the effect of repulsive
either dispersed or flocced slurries Figure interparticle forces on rearrangement.
6 shows the relative densitv of consolidat It is rarely recognized in ceramic tech-
January 1989 Powder Processing Science and Technology f o r Increased Reliability 9
nology that powders exhibit nonlinear elas- growth of this region during strain release
tic stress-strain behavior similar to that will be constrained by the rest of the
described by Hertzz7 when spheres are body, and it will produce tensile stress
pressed together. Walton28 has reported similar to an inclusion. These tensile
that the compressive stress-strain re- stresses can produce radial cracks.
sponse of a powder can be expressed as Cracks are more frequently observed at
a = 5 ~ 3 ' 2 , where B depends on the rela- higher filtration pressures where more
tive density of the powder compact (aver- stored strain is released. As detailed else-
age number of contacts per particle) and where,20 cracking after pressure filtration
the elastic properties of the particles and can be avoided by increasing the
is independent of particle size. Figure 8(A) resistance of the compact to crack
describes this response for A1203 powder growth, e.g., by adding small amounts
compacts as determined with strain recov- (<2 wt%) of certain polymers to the slur-
ery measurements after pressure filtration ry, which apparently bonds the particles
of both flocced and dispersed sIurries.20 as they form the consolidated layer dur-
As illustrated, relatively small stresses pro- ing pressure filtration.
duce large strains and the compact be- Figure 8(B) also illustrates that bodies
comes stiffer as the stress is increased. lt formed with dispersed slurries relieve
is not the porosity that produces this be- their stored strain within a much shorter
havior, but the large displacements be- period relative to bodies formed with
tween particle centers when a "point" flocced slurries. The reason is that the
contact is elastically compressed into an body formed with the dispersed slurry is
area contact. Thus, after a powder has still a fluid after pressure filtration, albeit,
been consolidated and the pressure has with a higher viscosity relative to the ini- Recoverv strain
been released, large elastic strains are tial slurry. That is, the consolidated body
recovered and the compact grows. itself can flow to release stored strain af-
The greater the consolidation pressure, ter filtration. Bodies formed from dis-
the greater the recoverable strain. Inclu- persed slurries will continue to flow after
sions within the powder which are either removal from their die cavities, much like
stiffer (e.g., dense agglomerates, whiskers, "silly putty," which has similar dilatant
or fibers) or more compliant (organic in- rheology.
clusions) will store less or more strain rela- Pressure filtration is not the only
tive to the powder compact, respectively, method to fractionate particles from slur-
during consolidation. Figure 8(A) also il- ries to form a shape. Centrifugation of Pressure = 80 MPa
lustrates that the elastic response of a flocced slurries can accomplish the same
dense A1203 inclusion (f = 400 GPa) and task and avoids the mass segregation
a very compliant polymer inclusion (€ = 1 produced by centrifuging dispersed slur-
GPa). The differntial strain relieved by the ries.14 Although it is not as well developed
inclusion relative to the powder will pro- as pressure filtration, with innovative de-
0 200 400 600 800 1l 0
duce detrimental stresses during strain sign, centrifugation offers great potential Time (s)
recovery. Likewise, the powder compact in shaping colloidally prepared powders.
can be damaged by metal die cavities
which constrain the strain recovery of pow- V. Densification Fig, 8. (A) Elastic strain recovered after
Powder compacts are densified at different saturated A1203powder bodies were
ders pressed within them.
temperatures where mass transport consolidated from either dispersed or flocced
For consolidated dry powders, strain slurries by presure filtration at different applied
recovery is nearly instantaneous with eliminates the void phase. Mass transport
pressures. Note that the nonlinear stress-strain
pressure release. As shown in Fig. 8(B), is driven by the excess free energy as- relation is similar to that expected from Hertzi-
the strain recovery for compacts pro- sociated with the surface area of the pow- an elastic behavior, viz., o = Bc3'2.20 (8)Time-
duced by pressure filtration is time depen- der. The difference in surface curvature dependent strain recovered after pressure
dent20 e.g., a compact produced from where the particles touch (net negative filtration (80 MPa) of dispersed and flocced
a flocced slurry will continue to release curvature), relative to the rest of the sur- A1203 slurries.20 Note that the recovery peri-
face of the particle (positive curvature), od for the body formed with the flocced slurry
strain and grow many hours after pres- is much longer.
sure release. This time-dependent strain- causes mass to "fill" the contact region
release phenomenon arises because fluid between touching particles. This process
(liquid or air) must flow into the compact is called sintering. If mass is removed
to allow the compressed particle network from between particle centers, mass
to grow and relieve its stored strain. transport will lead to shrinkage and den-
Release of the applied pressure will cause sification. If the particles are crystals, a
the fluid within the compact to share the grain boundary forms and grows as sin-
stored network strain and thus place the tering proceeds. In considering the free
fluid in tension. Fluid flow after pressure energy change of a powder compact, the
release is driven by the negative fluid increased energy associated with grow-
pressure within the compact relative to ing grain boundaries must be taken into
the ambient pressure outside the com- account as well as the decrease in parti-
pact. Because the fluid must flow from the cle surface energy. With this ther-
surface to the interior, the strain within the modynamic consideration, one can
compact is not uniformly released; i.e., show29 that sintering will stop, i.e., mass
strain is first released from the surface of transport to form a neck between touch-
the body. ing particles terminates, when
In practice, strain is likely to be relieved
-dAsidAb = yb/ys (3)
first at one region on the surface. The
10 Journal of the American Ceramic Society-Lange Vol. 72, No. 1
where dA, is the change in surface area, which continuously alters the network
dAb is the change in grain-boundary configuration.
area, and y b and ys are the energy per Although mass transport to the contact
area associated with grain boundaries region may stop when Eq. (3) is satisfied,
and particle surfaces, respectively. adjacent sintered grains will have differ-
When a linear array of touching parti- ent radii of curvature that will drive inter-
cles sinter, they form an equilibrium con- particle mass transport. Interparticlemass
figuration which satisfies Eq. (3).If mass transport will cause coarsening; i.e.,
is removed from between particle smaller grains disappear as larger grains
centers, the shrinkage of the linear array grow. It can be shown that the decrease
can be predicted with knowledge of in free energy for interparticle mass trans-
yb/ys. Although the linear array can be port is much smaller than the decrease
used to estimate shrinkage during sinter- associated with transport to the contact
ing, it does not enclose and define a void region. That is, the differential curvature
space and thus cannot be used to fully between two particles is much smaller
understand and predict the disappear- than the differential between particles and
ance of pores within powder compacts. their contact region. In addition, during
The void space within a powder com- coarsening, the grains become larger.
pact can be structuraily defined as pores, Larger grains result in larger radii and a
which are defined by irregular polyhedra lower driving force for interparticle mass
of touching particles connected to one transport. Thus, one might expect the ki-
another to form the particle network.30 netics for interparticle mass transport
Each polyhedron contains one pore. The leading to coarsening (grain growth) to
number of touching particles surround- be slower than the transport to the con-
ing and defining each pore is called the tact regions (sintering).
coordination number of the pore (n) If we argue that rapid transport to the
(equivalent to the number of vertices that contact region leads to shrinkage by sin-
define each polyhedron).The question is tering and the development of a metasta-
whether or not sintering can eliminate ble network similar to that shown in Fig.
each and every pore defined by the con- 9(A) and that further shrinkage is con-
nective network of different polyhedra. trolled by slower interparticle transport,
To answer this question Kellet and then the kinetics of densification should
Lange'g used Eq. (3) to analytically de- be separated by two regimes: an initial
termine the sintered, equilibrium config- regime controlled by sintering kinetics
uration of different regular polyhedra and a subsequent regime controlled by
Fig. 9. Micrographs of spherical ZrOn( + 8 formed with identical, touching spherical coarsening kinetics. Because the sinter-
mol% Y203) particles produced by electrostatic particles. It was shown that pores within ing regime results in shrinkage to a
atomization of zirconia acetate (A) after sinter- all polyhedra shrink (decrease their metastable network with a given relative
ing by heating to 1300°C for 10 h, ( 6 )same volume) during sintering, but only pores density, further densification must be con-
area after further heat treatement at 1300°C for with a coordination number less than a trolled by coarsening kinetics. Thus, the
18 h, and (C) 1400°C for 4 h, where similar net- critical value (n d n,) disappear. Pores two regimes will be separated by the rela-
work has undergone coarsening and further with n > n, shrink to an equilibrium size. tive density of the metastable network
densification 3' The critical coordination number is relat- produced by sintering.
ed to the ratio, yblys;29 the greater yblys, Figure 10 shows the shrinkage strain
the greater n,. It was therefore conclud- rate determined when different speci-
ed that all pores within a powder compact mens cut from a single powder compact
will shrink during sintering, but not all of A1203 were heated to 1550°C at differ-
would disappear; i.e.,sintering alone may ent heating rates.32 In each case, the
not result in full densification. shrinkage strain rate increases to a max-
The sintered network of touching Zr02 imum and then decreases.The maximum
spherical particles developed by heating shrinkage strain rate corresponds to the
at 1300°C for 10 h is shown in Fig. 9(A). inflection in relative density versus tem-
I
Note in Figs. 9(B) and (C) that when the perature curves commonly obtained in
1 20"Clmm
same area is viewed after a subsequent densification experiments. For each
number will not only require extensive ing two-phase powders, e.g., via the
grain growth to satisfy its thermodynam- colloidal approach. Namely, the inclu-
ics for instability, but its disappearance sions must be effective in retarding grain
will also be kinetically limited once it is un- growth before the powder compact
stable. achieves a relative density >0.9, where
rapid grain growth is commonly observed
VI. Control of Grain Growth in single-phase materials.35
The preceding section concluded that Grains are classified by their size
some grain growth via coarsening is re- (volume) and polyhedron type, the latter
quired to fully densify a powder compact. defined by the number of As
Although much of this coarsening takes grains decrease their volume, they
place by the disappearance and growth progressively and sequentially decrease
of adjacent grains through intergranular their number of faces. During grain
mass transport, grain boundaries can growth, the number of grains per unit
also move as they would in a fully dense volume decreases; i.e., some grains
body before full density is achieved. Be- decrease their volume and disappear.
cause matter needs only to diffuse over Only tetrahedral-shaped grains disap-
atomic distances, grain-boundarymotion pear. Thus, in modeling the effect of per-
will lead to more rapid grain growth than turbations on grain growth, one need only
the coarsening phenomenon. Grain- to model the shrinkage and disappear-
boundary motion will also trap unstable ance of simply shaped, tetrahedral
pores within grains before they have the grains. The tetrahedral grain decreases
opportunity to disappear via grain- its free energy as it decreases its volume.
boundary diffusion. In addition, if a near- (Grains with many faces (>14) will
ly dense body remains at the densifica- decrease their free energy by growing.)
tion temperature for a short time, grains The derivative of the free energy with
in many ceramics can grow more than respect to the volume of the grain (the
an order of magnitude larger than the Laplace equation) is proportional to the
average particle size used to form the grain-boundary energy per unit area y(),
powder compact. Some grains in ceram- and inversely proportional to the radius
ics with noncubic crystalline structures of curvature (r) of its grain boundaries,
can grow to an enormous size relative to viz., dE1dV = 2yglr. dEldV has the
their surrounding grains. Although the dimension of stress and is referred to as
reasons for abnormal grain growth are the driving stress (od) to decrease the
still unclear, abnormally large grains can volume of a grain. For the tetrahedral
be a dominant, strength-degrading flaw grain, the radius of curvature is positive
population in many important ceramics. (grain boundaries are concave when
For these and other reasons, grain looking from within), and r = 0.6Dt,
growth via boundary motion must be con- where Dt is the diameter (grain size) of
trolled to optimize both density and the equivalent spherical volume of the
properties. tetrahedron. Thus, for the tetrahedron,
I b Second-phaseinclusions have become ~d = 3.3yglDt.
Grain volume increasingly important in controlling grain Zener43 was the first to explain how in-
size in ceramics. Inclusions generally give clusions retard grain growth. Zener's con-
Fig. 13. (A) Two-dimensional schematic of rise to residual stresses because of cept is visualized by Fig. 13(A), which
grain boundaries of a tetrahedral grain inter- differential thermal contraction and, there- shows, in two dimensions, how a shrink-
acting with spherical inclusions. Different po- fore, are usually thought of as strength- ing tetrahedral grain interacts with spher-
sitions illustrate the shape of the boundary as degrading flaw populations themselves. ical inclusions of radius R.44 As the
it encounters and breaks away from the inclu- But, it has been shown, both through shrinking grain encounters the inclusion,
sions. (B)Energy of the tetrahedral grain ver- experimentation37 and theory,38 that, if an increasing proportion of the boundary
sus its volume as it encounters and breaks the inclusions are less than a critical size,
away from inclusions. Three different curves
area is removed from the grain. When N
their residual stress will not induce inclusions are simultaneously intersected,
are for three different retarding stresses
produced by inclusions microcracking either during cooling from the maximum amount of grain-boundary
the densification temperature or during area removed is N n W , which cor-
subsequent stressing. Inclusions with a responds to a decrease in free energy of
size that will not induce microcracking NnRnyb. If the grain shrinks further, the
can be used to both control grain growth grain-boundary area adjacent to the in-
and engineer new composites for desired clusions must bow out as it attempts to
properties without the fear of degrading break away. This bowing and the fact that
strength. For example, large andlor ab- the grain boundary regains its area
normal grain growth can be prevented by occupied by the inclusion during bowing
the addition of an appropriate inclusion causes the free energy of the system
phase as demonstrated by the dramatic to increase. Using line tension argu-
strengthening achieved with additions of ments, Zener showed that, as the grain
SIC to AI2O3,39Zr02 to p"-AI2O3,40 and boundary breaks away, the inclusions
AI2O3 to cubic ZrO2.4' In each of these exert a maximum restraining "stress,"
cases, the inclusion phase eliminates the or = 0.75fyb/R, where f is the volume
most detrimental flaw population, viz., fraction of inclusions. Thus, the net driv-
large andlor abnormal grains. ing stress for the grain boundaries of a
For ceramics, inclusons used to con- tetrahedral grain to break away from the
trol grain size can be introduced by mix- inclusions is
January 1989 Powder Processing Science and Technology f o r Increased Reliability 13
undergo grain growth themselves. As
grain growth in the major phase pro-
ceeds, coalescence of inclusions at four-
If it assumed that the kinetics of grain dis- grain junctions will continue and the ra-
appearance (growth) are portional to on, tio of inclusion size to grain size will be
then Eg. (4) shows that inclusions will maintained constant during further grain
reduce the kinetics for grain growth. growth. Immobile inclusionsthat effective-
The free energy versus volume func- ly retard grain growth at lower tempera-
tion for inclusion encounter and break tures can become mobile at higher
away is schematically shown in Fig. 13(B) temperatures, where they become much
for the cases where on > 0, on = 0, and less effective because of coalescence.45
on < 0. The case of on < 0 represents Because inclusions are introduced into
the condition (f > 4.4R/Dt) where the ceramics by mixing powders, their distri-
tetrahedral grain is trapped within a bution is a critical processing issue. When
potential well and will not decrease its the inclusions are not uniformly distribut-
volume unless thermally activated to ed during powder mixing, grain growth
"jump out."44 in some regions will be less hindered than
Zener's concept and Eq. (4) provide others. In the extreme case, e.g.,where
rules for the use of inclusions to hinder major phase agglomerates are not re-
grain growth. First, the inclusion must be moved by colloidal fractionation and in-
smaller than the grains. Recent experi- clusions cannot be introduced into these
ments indicate that even when f is large, agglomerates, the region containing few
inclusions will have no effect on grain inclusionswill undergo rapid grain growth
growth until their mean grain size is >I .5 as shown in Fig. 14. In addition, when
times the inclusion size.45 Second, the very small volume fractions of inclusions
smaller the inclusion size, the greater the are used, their intrinsically poor distribu-
hindrance to grain growth at a given tion can exaggerate abnormal grain
volume fraction. Third, larger volume frac- growth relative to single-phasematerial.49
tions will produce greater hindrance. Thus, controlling grain growth with inclu-
These rules appear to be obeyed for cer- sions starts with uniformly prepared
amics when the inclusions are immobile, powders.
as assumed by the analysis.45.46 It is still Control of grain growth can also be
questionable if inclusions will completely achieved with minor additions of select-
prevent grain growth, as suggested for ed cations which exhibit slight solid SOIL-
the case where on << 0. bilitv.50 It is usuallv Dresumed that these
Inclusions can also retard the motion "im'purities" will p'rifer to concentrate at
Fig, ,4. Micrograph showing that ZrOp In-
of three- and four-grain junctions. Be- boundaries where they better accommo- clusions (white phase) could not restrain grain
cause it requires much less energy for the date the Special "Crystal chemistries" Of growth within &03agglomerate during den.
inclusionto either break away from a four- grain-boundary structures, i.e., ion-size sification as in surroundina matrix where
I
grain junction and slip onto a three-grain ratio, charge balance, and coordination phases are well mixed.
junction or break away from a three-grain number. If the grain boundary were to
junction and slip onto a two-grain junc- break away from this planer "cloud of im-
tion (grain boundary) than to break away purities," i.e,, if boundary motion were
from a grain boundary and slip into the more rapid than the diffusion required to
crystal, inclusions are much more effec- keep the impurities moving with the
tive in retarding grain growth when they boundary, then the planer cloud of impu-
interact with grain boundaries.44 rities with area Ab would now reside wi-
The analysis of the preceding para- thin the grain. Grain-boundary break
graph assumes that the inclusions are away will increase the free energy by
spatially fixed, i.e., immobile. If the inclu- dG = AGAb dDt, where AG is the free
sions exhibit sufficient self-diffusion.they energy per unit volume for the impurity
can become mobile and can be dragged cloud residing within the crystal relative
along by the moving boundaries.47 to the boundary and dDt is the change
Although mobile inclusions can still be ef- in grain size. The restraining stress for the
fective in retarding grain growth, as shrinkage of a tetrahedral grain with
Evans48 has shown for analogous pore Ab = aDt2 and volume, V = pDt3, is
drag, they soon coalesce to become larg- or = dG/dV = (a/3p)AG,where a and p
er and less effective. The kinematics of are numerical factors.51 The net driving
coalescence can be viewed as follows.44 stress for boundary break away during
Mobile inclusions are first collected by grain shrinkage is given by
moving grain boundaries. Three-grain
junctions move together to eliminate grain
boundaries and push together inclusions
already collected by the disappearing Similar to the case for inclusions, when
boundary. Finally, four-grain junctions on > 0, the cloud of impurities will
move together to collect all inclusions on momentarily hinder grain growth, but
three-grain junctions to form a new four- boundaries can easily break away With
grain junction which contains all collect- further grain growth, a grain size will be
ed inclusions. With grain growth, mobile obtained where D, > 1O(a//3)(yb/AG)and
inclusions are relocatedto four-grainjunc- on < 0 Here, the cloud of impurities will
tions, their lowest free energy location. severly restrain grain growth and grain
During coalescence, inclusions sinter and growth might be expected to be limited
14 Journal of the American Ceramic Society-Lange Vol. 12, No. 1
by the diffusion kinetics of the impurity research directions. The concepts lead-
cloud.sz Thus, the impurities will have ing to the colloidal methodology to frac-
very little effect when the grains are small, tionate and consolidate powders were
but a major effect as the grains grow larg- developed out of a need to eliminate het-
er. It is obvious that the greatest effect on erogeneities common to powders that
grain growth is obtained by choosing im- produce unwanted flaw populations in
purities that maximize AG. structural ceramics.' Although the fun-
The same reasoning has been applied damental understanding of interparticle
to the case where phase partitioning is forces is relatively mature, relations be-
concurrent with grain growth. It was tween surfactant chemistries and ceramic
observed53 that grains of the same struc- surface chemistries leading to a predic-
ture develop different compositions dur- tive choice of surfactants and conditions
ing partitioning. Grain growth in the that optimize and alter interparticleforces
two-phasecompositional region is severly consistent with ceramic powder process-
limited relative to single-phase regions but ing is an open research field. Likewise,
second-phase grains which might hinder research leading to a predictive under-
grain growth, as discussed earlier in this standing of the slurry rheology as affect-
paper for inclusions, are not observed53 ed by interparticle forces, particulate
as initially postulated.54 If compositional volume fraction, size, size distribution,
differences between grains did not alter and mixed-phasecontent will significantly
during grain growth (i.e., if boundary mo- aid new ceramic processing technology.
tion is faster than the period required to Our understanding of powder consolida-
equilibrate compositional gradients), it tion mecahnics, viz., how particles rear-
was reasoned that the boundary would range during powder flow under
leave behind a "ghost" boundary within conditions of constrained dilatancy, is
the growing grain where the lattice nonexistent but critical to optimize pack-
parameters change abruptly because of ing density to optimize densification kinet-
the compositional gradients. If the com- ics. The mechanics of powder compacts
positional gradient between the two themselves, ranging from strain release
grains is a step function, then the ghost after consolidation to fracture phenome-
boundary would appear as a coherent in- na, is a research void that must be filled
terface. The different lattice parameters to learn how to consolidate powders from
on either side of the ghost boundary slurries without producing disruptive
would produce a strain energy density, phenomena. In the past, densification
Use= KEZIE,where E is the strain due to research has emphasized kinetic models
the change in lattice parameters, E is the of sintering that neglect the effects of par-
elastic modulus of the material and K is ticle arrangement. The link between coar-
a dimensionless constant.53 In the same sening kinetics and densification kinetics
manner used to develop Eq. (5),the net would be one fruitful research topic. Our
driving stress for shrinkage of the tetra- ability to control grain growth would be
hedral grain is given by greatly enhanced with a fundamental un-
3.3 €2 derstanding of the effect of impurities on
on = ad - 0, = -yb - K- the crystal chemistry of grain-boundary
Dt E
structures and how compositional gra-
(6) dients produced by partitioning effect
As descirbed for impurity clouds, grain-boundary motion and partitioning
the grain boundary could easily move kinetics itself. This research would be
to form a ghost boundary when D, directed to increase property reliability via
is small (on > 0), whereas as grain increased processing reliability, i.e.,
growth proceeds, a condition will arise processing reliability equals material relia-
(Dt > 3 . 3 y b H K ~ 2 where
) grain growth will bility.
be limited (un < 0) by the diffusion kinet-
ics required to equilibrate the composi- References
tion; i.e., grain-growth kinetics will be 'M. V. Swain, "R-Curve Behavior of Magnesia-
related to partitioning kinetics. In systems Partially-StabilizedZirconia and its Significance for
Thermal Shock", pp 355-70 in Fracture Mechanics
that exhibit extended solid solution, grain- of Ceramics, Vol. 5. Edited by R. C. Bradt. A. G.
growth control will be more pronounced Evans, D. P. H. Hasselman, and F. F. Lange. Ple-
when small compositional differences num Press, New York. 1983.
produce large differences in lattice ZD. B. Marshall and A. G. Evans, "Failure Mechan-
isms in Ceramic-FiberlCeramic-MatrixComposites."
parameters. J Am Ceram. SOC.,68 [5] 225-31 (1985).
The use of impurity clouds and phase 3F. F. Lange and M Metcalf, "Processing-Related
partitioning to control grain growth is one Fracture Origins 11, Agglomerate Motion and
of accidental discovery due to a lack of Cracklike Internal Surfaces Caused by Differential
Sintering," J. Am. Ceram. SOC., 66 [6] 398-406
understanding of how to systematically (1 983).
chose and engineer compositions that uti- 4R. G. Frev and J. W. Halloran. "ComDaction Be-
lize and optimize these important havior of Spray-DnedAlumina," J Am &ram Soc ,
phenomena. 67 [3] 199-203 (1 984)
5W H Rhodes. "Aqqlomerate and Particle Size
VII. Concluding Remarks Effects on Sintering Yttia-StabilizedZirconia," J. Am.
Ceram. SOC.,64 [l] 19-22 (1981).
This review of processing science and 6F. F. Lange, "Sinterability of Agglomerated Pow-
technology leading to ceramics with in- ders," J. Am Ceram. Soc..67 [2] 83-89 (1984).
creased reliability suggests certain 7F. F. Lange, B. I Davis, and E. Wright,
January 1989 Powder Processinx Science and Technology .for Increased Reliability 15
Am. Ceram. Soc , 69 (11 66-69 (1986). ics of Densification. 111, Experimental Relation be-
8V Engle and H. Hubner, "Strength Improvement tween Grain Growth and Pore Closure"; unpublished
of Cemented Carbides by Hot Isostatic Pressing." work.
J. Mater. So.. 13 [9] 2003-13 (1978). 34F. F. Lange and B. J. Kellett, "Thermodynam-
9B. J. Kellett and F. F Lange, "Experiments on ics of Densification: 11, Grain Growth in Porous
Pore Closure During Hot Isostatic Pressing and Forg- Compact and Relation to Densification," to be pub-
ing," J. Am. Ceram. SOC,71 [l] 7-12 (1988). lished in J. Am Ceram. SOC.
'OF F. Lange, "Processing-Related Fracture Ori- 35T. K Gupta, "Possible Correlation Between Den-
gins: I, Observations in Sintered and Isostatically sity and Grain Size During Sintering." J Am. Cer-
Hot-Pressed AI2O3/ZrO2Composites," J. Am. Cer- am SOC.,55 [5] 276-77 (1972).
am. SOC, 66 [6] 396-98 (1983). 36E. A Barringer and H. K. Bowen. "Synthesis and
"J. W. Goodwin. "The Rheology of Dispersions"; Processing of Submicrometer Ceramic Powders";
pp. 246-93 in Colloid Science, Vol. 2 D. H. Everett pp 482-96 in Science of Ceramic Chemical Process-
(Senior Reporter). The Chemical Society, London, ing. Edited by L L. Hench and D. R. Ulrich. Wiley,
UK, 1975. New York. 1986
12M. D Sacks, "Rheological Science in Ceramic 37D B. Binns, "Some Physical Properties of Two-
Processing"; pp. 522-38 in Science of Ceramic Phase Crsytal-GlassSolids"; pp. 31 5-35 in Science
Chemical Processing Edited by L. L. Hench and D of Ceramics Edited by G H. Steward Academic
R. Ulrich. Wiley, New York. 1986 Press, New York, 1962
13J. N Israelachvili. Intermolecular and Surface 3 8 0 J Green, "Microcracking Mechanisms in Cer-
Forces. Academic Press, London, UK, 1985. amics"; p 457 in Fracture Mechanics of Ceramics,
14F. F. Lange, "Forming a Ceramic by Floccula- Vol. 5. Edited by R. C. Bradt, A. G Evans, D. P. H.
tion and Centrifugal Casting." U.S. Pat. No. Hasselman. and F F. Lange. Plenum Press, New
4 624 808, Nov. 25, 1986. York, 1983.
15F. F Lange and M M. Hirlinger. "Phase Distri- 39F. F Lange and M Claussen, "Some Process-
bution Studies Using Energy Dispersive X-ray Spec- ing Requirements for Transformation-Toughened
tral Analysis," J. Mater. So. Len., 4, 1437-41 (1985). Ceramics"; p 493 in Ultrastructure Processing of
16F. F Lange and K. T. Miller, "A Colloidal Method Ceramics, Glasses, and Composites. Edited by L L
to Ensure Phase Homogeneity in fi"-AI2O3/ZrO7 Hench and D. R Ulrich. Wiley. New York, 1984.
Composite Systems," J. Am. Ceram SOC.,70 [12] 40D. J. Green, "Transformation Toughening and
896-900 (1987). Grain-SizeControl in P-AI2O3-ZrO2Composites." J.
17J. Cesarano 111, I. A. Aksay. and A. Bleier. "Sta- Mafer So., 20 (71 2639 (1985).
bility of Aqueous u-A1203 Suspensions with 41F. J Esper, K H. Friese, and H Geier, "Mechan-
Poly(methacry1ic acid) Polyelectrolyte." J. Am. Cer- ical. Thermal, and Electrical Properties in the System
am. SOC, 71 [4] 250-55 (1988). of Stabilized ZrO2(Y2O3)lu-AI7O3"; pp. 528-36 in Ad-
18K.P. Darcovich and I. Aksay. "Particle-Sized Dis- vances in Ceramics, Vol. 12. Science and Technol-
tribution of Dense Ceramic Suspensions"; unpub- ogy of Zirconia 11. Edited by N Claussen, M Ruhle.
lished work and A. H. Heuer. American Ceramic Society, Colum-
19R. D. Rivers, "Method of Injection Molding Pow- bus, OH, 1985
derMetal Parts," U S. Pat. No. 4 113 480, Sept. 12, 475 K Kurtz and F M. A. Carpay, "Microstruc-
1978 ture and Normal Grain Growth in Metals and Cer-
ZOF F. Lange and K T. Miller. "Pressure Filtration: amics: Part I, Theory." J Appl. Phys , 5 1 [ l l ]5725
Kinetics and Mechanics," Am Ceram SOC.Bull , 66 (1980).
[lo] 1498-504 (1987). 43C Zener, kindly quoted by C S. Smith, Trans.
"(a) S. Strijbos, "Pressure Filtration of Permanent Mefall. SOCAIME. 175, 15 (1949).
Magnetic Powders"; in Proceedings of the Confer- "F. F Lange, "Controlling Grain Growth"; pp
ence on Hard Magnetic Materials. Edited by H. Zi~l- 497-508 in Ceramic Microstructures '86: Role of In-
stra Bond voor Materialenkennis, The Hague, terfaces. Edited by J. Pask and A. G. Evans, Plenum
Netherlands, 1974. (b) C A. M. Van den Broek and Press, New York. 1988.
A. L Stui]ts, "Ferroxdure," Phrlps Tech. Rev., 37 [7] 45F. F. Lange and M. M Hirlinger, "Grain Growth
157-75 (1977). in Two-Phase Ceramics. A1203 Inclusions in Zr02."
ZZGebruder Netzsch, Maschinenfabric GmbH and J. Am. Ceram. SOC.,70 [ l l ] 827-30 (1987).
Co , Technical Information Bulletin,GK 012, 0-8672. 4 6 0 . L Olgaad and B. Evans, "Effect of Second-