JOURNAL OF MATERIALS SCIENCE 25 (1990) 3449 3455
Powder preparation and compaction behaviour
of fine-grained Y-TZP
W. F. M. GROOT ZEVERT, A. J. A. WINNUBST*, G. S. A. M. THEUNISSEN,
A. J. BURGGRAAF
University of Twente, Faculty of Chemical Technology, Laboratory for Inorganic Chemistry,
Materials Science and Catalysis, P.O. Box 217, 7500 AE Enschede, The Netherlands
Two wet chemical preparation methods are described for yttria-doped tetragonal zirconia
powders. Both methods yield powders with an extremely small crystallite size (8 nm) and a
narrow size distribution. The agglomerate and aggregate structure of these powders have been
investigated by several techniques. Gel precipitation from an alkoxide solution in water
("alkoxide" synthesis) results in a ceramic powder with irregular-shaped weak and porous
agglomerates, which are built up from dense aggregates with a size of 18 nm. Gel precipitates
formed from a metal-chloride solution in ammonia ("chloride" synthesis) do not contain
aggregates. Both types of agglomerate are fractured during isostatic compaction. Hydrolysis
and washing under (strong) basic conditions probably decrease the degree of aggregation.
The aggregate morphology and structure are key parameters in the microstructure development
during sintering of a ceramic. Several characteristics of these powders are compared with
those of a commerical one (Toyo Soda TZ3Y).
1. Introduction
Ceramic materials which consist of 100% yttria-doped
tetragonal zirconia polycrystals (Y-TZP) exhibit a
high strength and toughness compared to other
ceramics. For zirconia doped with 3 mol % Y203 the
ceramic grain size must be less than 0.8 #m to remain
fully tetragonal [1]. However this critical grain size is
much smaller (about 0.3 #m) when the ceramic has to
remain tetragonal after ageing in water at elevated
temperatures [2, 3]. In order to obtain dense and pure
tetragonal zirconia ceramics with these small grain
sizes, special requirements must be fulfilled by the
powders used for the sintering process. These ceramic
powders must be homogeneous in composition and
highly sinteractive (low sintering temperature) in
order to decrease grain growth during sintering. They
can be prepared by means of wet chemical methods
such as sol-gel or gel precipitation. These fine grained
(submicrometer) powders are not monodispersed but
consist of microstructural elements due to clustering
[4-8]. These elements are: (1) primary crystallites; (2)
aggregates in which, the primary crystallites are held
together by neck areas which are formed by a reaction,
e.g. sintering of the powder by surface diffusion or
even diffusion and precipitation at necks in the sus-
pension [4]; (3) agglomerates. The aggregates or
the individual primary crystallites are held together
in an agglomerate structure by relatively weak (e.g.
Van der Waals, capillary) attractive forces.
During wet-chemical preparation these crystallite
"clusters" (2 and 3) are essentially formed in the early
stages of the preparation cycle [9]. The state of these
clusters determines, for example, surface area, flow
* Author to whom all correspondence should be addressed.
0022-2461/90 $03.00 + .12 9 1990 Chapman and Hall Ltd.
behaviour, densification processes during compaction
and the final grain size in the ceramic. The clusters in
the powder can be detrimental to sintering when they
are still present in the green compact prior to sinter-
ing. In those types of compacts no homogeneous pore
size distribution is present and the pores within the
compact, which can be much larger than the crystal-
lites, only disappear during sintering at extremely
high temperatures or do not disappear but even
grow [5, 10]. These porosity inhomogeneities are a
source for inhomogeneous sintering, which can give
coarse-grained regions in the sintered compact. Those
inhomogeneities induce nonuniform sintering rates
and back stress effects, which in turn create transient
(or residual) stresses [11].
In studies of the microstructural development of
cubic zirconia it has been found that the aggregates
with a size of some tens of nanometres already sinter
to full density at very low temperature (about 1170 K)
forming larger single crystallites [8]. These crystallites
(size about 15 nm) probably form the initial grain size
with which the normal grai n growth process starts and
consequently the aggregate size is important for the
final grain size of the dense ceramic. This relation is
studied in the current programme to obtain dense
ceramics with a grain size of 0.1 #m and smaller.
In this paper a detailed study is made of the change
in agglomerate morphology during compaction
of fine-grained and weakly agglomerated powders.
Three different ceramic powders were used. Two pow-
ders prepared by hydrous-gel-precipitation methods
and a commercially available powder (Toyo Soda
TZ3Y).
3449
2. Experimental procedure
2.1. Powder
preparation
The zirconia-yttria powders were prepared by two
different gel-pre~pitation techniques. Both implied
the hydrolysis of a diluted Zr-Y precursor solution in
an excess of hydrolysing agent.
For one method (the "alkoxide" method), freshly
prepared Zr-t-amyloxide and Y-isopropoxide were
used. A 0.2 M solution of these alkoxides in benzene
was slowly added to a large excess of water (molar
ratio H20:metal = 250). The hydrolysis was per-
formed in a dispersion turbine reactor as described
previously [8]. After hydrolysis, the gel was washed
three times with water in order to remove organic
solvents. This hydrous gel was filtered and wet-milled
with ethanol in plastic containers using teflon balls.
Subsequently the gel was washed three times with
an excess of ethanol in order to remove the free
water within the gel. All the washing steps were
performed with large amounts of washing agent (gel
concentration about 0.1 M) using a high-energy disc
turbine [8].
The precursor for the second method consists of
commercial ZrC14 (Merck) and YC13 (Cerac) which
were dissolved in HC1 (0.4 M). The precursor solution
was added slowly to an excess of a 25% ammonia
solution. This method is called the "chloride" method.
During hydrolysis the pH remained at 11 or more. The
hydrolysis and washing procedure were performed in
the same reactor as described for the alkoxide method.
Washing of the chloride-derived gel with water must
be performed very thoroughly in order to remove
chlorides from the solution. For that reason the
gel was washed about eight times in a solution with
a gradually decreasing amount of ammonia. Sub-
sequently the gel was washed three times with an
alcohol to remove free water in the same way as
described for the "alkoxide" method.
Both gels ("alkoxide" and "chloride" derived) were
filtered, dried in air for 16h (390K), dry-milled (in
a plastic container equipped with teflon balls) and
calcined at 820 K for 2 h after which they were dry-
milled again.
The microstructure of the alkoxide and chloride
powders were compared with a commercially available
powder with about the same composition (Toyo Soda/
TOSOH; TZ3Y).
2.2. Characterization
Several techniques were used to characterize the pow-
ders. The amount of zirconia and yttria was analysed
by X-ray fluorescence spectrometry using a Philips
PW 1410 spectrometer. The amount of chloride was
determined by dissolving the powder in sulphuric acid
during long-term refluxing. In order to counteract the
formation of gaseous hydrochloric acid, a cold finger
was used during refluxing. The solution was after-
wards titrated with an AgNO3 solution.
A Philips X-ray diffractometer PW 1710 using
CuKe radiation was used to determine the phase-
composition and the primary crystallite size of the
powder. The Scherrer equation was used to deter-
mine the crystallite size from X-ray line broadening
measurements. In this case correction for instrumental
broadening and K~-splitting has been applied. The
crystallite-size of the as-calcined powders were also
verified by means of TEM (Jeol 200 CT).
The specific surface area was determined according
to the BET method using the equipment as reported
by Bosch and Peppelenbos [12].
The tap density was measured by vibrating a test-
tube containing the powder until the powder volume
remained constant. The powders were isostatically
pressed up to 500 MPa. Compaction behaviour was
tested by measuring the density as a function of com-
paction pressure. The overall density was determined
using the Archimedes technique (in mercury).
The pore-size distribution and the pore volume in
the compacts were calculated from a nitrogen
adsorption/desorption isotherm at 77 K using a Carlo
Erba Sorptomatic Series 1800. The desorption branch
was used according to the method described by
Dollimore and Heal [13]. Pores with radii larger than
about 10nm were determined by mercury-intrusion
porosimetry. These measurements were performed
with a Carlo Erba Porosimeter Series 200 at pressures
up to 200 MPa.
The as-calcined powders were isostatically pressed
to 400MPa before sintering. For these sintering
experiments the temperature was raised at a rate of
1 K min- 1 to the desired temperature. At this tempera-
ture the compact was sintered for 10h in air.
3. Results and discussion
The X-ray fluorescence measurements showed that the
composition of the powders was near to the nominal
composition: ZrO2 + 3tool % Y203 (Table I). The
phase analysis by X-ray diffraction revealed that only
traces of monoclinic zirconia were observed in the
as-calcined "alkoxide" and "chloride" powders. In
contrast the Toyo Soda powder clearly contained
monoclinic and cubic crystallites.
The amount of chloride in the as-calcined powders
is important because there is an indication that sinter-
ing starts at higher temperatures when the ceramic
powder contains more chloride. For undoped zirconia
made by the same "chloride" method it was shown
that the temperature at which transformation from
tetragonal to monoclinic zirconia takes place decreases
with increasing amount of chloride in the as-calcined
powder [141.
The specially developed gel-washing procedure
for the "chloride" method showed good results in
decreasing the amount of chloride in the resulting
calcined powder. The chloride concentration in the
chloride powder was four times larger when, after
hydrolysing in ammonia, the gel was only washed with
water instead of an ammonia/water solution.
The primary crystallite sizes of the as-calcined
chloride and alkoxide powders were comparable
(8nm). A TEM picture of the chloride powder is
shown in Fig. 1. The commercial Toyo Soda powder
contained crystallites which were about four times as
large (34nm). The surface area of the Toyo Soda
powder was much lower than the SBET of the chloride
and alkoxide powders. The reason for the observed
3450
TA B LE I Powder properties
Synthesis Composition C1 content Crystal structure Crystallite size, d SBE x
(tool % YOts) (wt %) (%) (nm) (m2g l)
alkoxide 5.9 < 0.001 t > 99 8 101
chloride 5.7 0.015 t > 99 8 124
Toyo Soda 5.5 0.010 m= 16 34 21
c=10
t= 74
t = tetragonal, m = monoclinic, c = cubic.
differences in the SBET values between the three pow-
ders will be discussed later.
The powders were isostatically pressed up to
500 MPa. During compaction overall density measure-
ments were made on several compacts. These relative
densities were plotted as a function of the logarithm of
the applied pressure (Fig. 2). The curves show two
linear parts (parts 1 and 2) with a point of intersection
at pressure, Py. The linear part at high pressures (part
2, Fig. 2) can be described by the empirical equation
P - Pv = mln
(P/Py)
(1)
where m is a constant, # the density and P the pressure.
The subscript y stands for the yield point. The values
of Py and m are given in Table III. M is probably a
constant which is related to the microstructure of the
powder. The calculated values for rn (Table III)
show large differences. Halloran [5, 6] used the same
equation but did not prove the applicability nor report
any m values.
According to Van de Graaf
et al.
[8] the intersection
Py is the strength of the largest microstructural ele-
ment. After compaction at a pressure around Py these
elements are all gradually fragmentated, while they
are only rearranged at pressures below Py. These ele-
ments are called the agglomerates. However, if a pow-
der is granulated the granules are the first elements to
be broken. The Toyo Soda powder is a granulated
powder as is confirmed by TEM (Fig. 3). Clearly this
powder consists of spherical-shaped granules. The Py
values differ significantly for the powders studied
(Table III). For the chloride powder, the Py value
(intersection) is much larger than for the other two.
In this paper the density, packing capability (mor-
phology) and strength of the agglomerates will be
discussed.
In order to obtain information about the density
of the agglomerates the pore-size distributions after
isostatic compaction of the powders has to be con-
sidered (see Fig. 4). The range of the pore radii
within the powder compacts can be divided into three
sections as indicated in Fig. 4: (1) R ~< 3nm, the
intra-aggregate pores (pores between the crystallites);
(2) 3 ~< R < 100nm, the inter-aggregate pores; (3)
R > 100 nm, the inter-agglomerate pores or the inter-
granular pores in case of the Toyo Soda powder. The
TAB LE 11 Packing properties of the as-calcined powders
Synthesis TD (%) SD (%) A (TD - SD) (%)
alkoxide 12 5 7
chloride 19 17 2
Toyo Soda 22 22 0
agglomerate or granular density was calculated from
the sum of the pore volumes of the inter-and intra-
aggregate pores.
After compaction at P = 4 MPa the agglomerates
themselves remain intact, because for each powder the
applied pressure is less than Py (Table III). The
agglomerate density in these powder compacts is
therefore the same as in the unpressed powder. The
pores with a radius smaller than R = 100 nm (sections
2 and 3 in Fig. 4a) have to be taken into account for
calculating the agglomerate density by means of the
equation
Oagg~. =
[x/(x +
y)] x 100% (2)
where x
= l/ko t (~o t
is the theoretical density) and y is
the volume (in units of gram powder) of the nitrogen
which is adsorbed in the intra-agglomerate pores
as determined by nitrogen adsorption/desorption
measurements. The agglomerate densities obtained in
this way are given in Table III. From these results
it can be seen that the agglomerates of the alkoxide
powder are more porous than the chloride agglomer-
ates. The Toyo Soda granules are rather dense.
The difference in tap density values (TD) as given in
Table lI can be explained by the difference in agglom-
erate densities. A ceramic powder with a high agglom-
erate density shows a high tap density.
In order to obtain large compacts it is not only
important to have dense agglomerates but also to have
agglomerates with a good packing capability. The
density of a powder after it is poured into a tube or
mould without applying any external force is a good
characteristic for the resulting packing capability of
the powder. This density is determined by an extra-
polation of the linear part of the density-compaction
Figure 1
Transmission electron micrograph of a calcined "chloride"
powder.
3451
J
40-
?
*~ 30-
[-
13
>e 20-
10-
50-
I I I III I I J I-~ lit I I [ [ [
5 10 50 100 500
applied pressure (MPa)
Figure 2 Densification behaviour of the ZrO2-Y203
powders during isostatic compaction. ~ : Pore size dis-
tribution measured at these pressures. (zx) Alkoxide,
(O) Toyo Soda, (0) Chloride.
pressure curve at low pressures (part 1 in Fig. 2) down
to P = 0.1 MPa. The density of this point (further
referred to as SD) can deviate from the density
obtained after putting some energy to the powder
system by vibrating the tube. This latter density is
called the tap density (TD). Values for both densities
are given in Table II. From these results it is clear that
SD and TD are dependent on the preparation method
of the powder. Also a significant difference in SD and
TD for the alkoxide powder is observed. These
phenomena will be discussed now. A measure for
good packing capability is the packing factor of the
powder (p) which is defined by Equation 3 according
to Halloran [5, 6]
p
= SD/kOaggl.
(3)
Here SD is used instead of TD because SD is a value
of powder packing without any external force added
during stacking. The fc c packing is the ideal situation,
which leads to a p value of 0.74 when spherical
particles with uniform diameter are assumed. Accord-
ing to Fedors [15], however, a value of 0.63 is the
most probable p value when mechanical vibration
is employed to a powder which consists of non-
deformable particles with a uniform size. This p value
was determined by assuming that about one-half of
the particles was packed in a (hexagonal) close pack-
ing (= fc c) and the other half in a simple cubic array
(= s c with p = 0.52). In Table III the p values of the
investigated powders are given. The p values of the
chloride and Toyo Soda powders are relatively high.
This indicates that the agglomerates of the chloride
powder and the granules of the Toyo Soda powder
stack rather well. Probably the particles can move
easily with respect to each other and are spherically
shaped. This also explains why, for these two powders,
the SD and TD values were about equal. The p value
for the alkoxide powder is rather low. The alkoxide
agglomerates pack poorly prior to compaction. The
difference in SD and TD for this powder was also
rather large. Both effects indicate that the agglom-
erates of the alkoxide powder are irregularly shaped.
Another important characteristic of a powder
agglomerate is its strength. Indicative for the agglom-
erate strength is the Py value as given in Table III. The
larger Py value in the chloride powder indicates that
the agglomerates in this powder are stronger than the
agglomerates in the alkoxide powder or the Toyo
Soda granules. For the chloride powder, however, the
density-compaction curve is not linear between
P = 50 and 100MPa (see Fig. 2). These two values
probably represent the lowest and highest strength of
the agglomerates. The overall density at the point of
maximal agglomerate strength (P = 100MPa) is in
good agreement with the agglomerate density as given
in Table III. For the alkoxide and Toyo Soda powder
the density-compaction curve is clearly divided into
two sharply separated linear parts. This indicates that
the maximum strength for the agglomerates/granules
of the powder is very close to its average strength. So
the agglomerate strength in these two powders is rather
uniform. The relative densities of the powder com-
pacts at the pressure at which all agglomerates are
fragmented (yielding point in Fig. 2) is in good agree-
ment with the agglomerate densities (Table III). This
is an indication, that the internal agglomerate (or
granule) structure remains the same at pressures
below the yielding point.
In order to investigate the change in microstructure
during compaction, the pore-size distributions are
determined from powder compacts pressed at 100
and 400 MPa (Figs 4b and c). After compaction at
T A B L E I I I Agglomerate properties
Synthesis Agglomerate strength, Agglomerate density m-value* (%) SD-value (%) p factort
Py (MPa) p aggl (%)
aIkoxide 45 22 11 5 0.22
chloride 80 30 14 17 0.56
Toyo Soda 40 36:~ 7 22 0.61
:~Granule density for the Toyo Soda powder.
*According to Equation I.
t According to Equation 3.
3452
Figure 3
Scanning electron micrograph of a Toyo Soda powder.
100MPa the large pores which are ascribed to the
inter-agglomerate pores (section 3, Fig. 4) disappear.
This again is evidence that the strength of all agglom-
erates or granules is less than (or equal to) 100 MPa.
In Table IV the mean pore sizes of the inter- and
intra-aggregate pores are given as functions of the
applied pressure. For the Toyo Soda Powders, Lecloux
et al.
[7] reported an inter-aggregate pore size of
28 nm, which is in agreement with the inter-aggregate
pore size after compaction at 4 MPa (see Fig. 4a and
Table IV). After isostatic compaction at 100 MPa the
alkoxide and Toyo Soda powder compacts show
smaller pores in section 2, Fig. 4 (the inter-aggregate
pores) if compared with compaction at 4MPa. This
decrease in inter-aggregate pore size takes place
between
Py
and 100 MPa. This effect is not observed in
the chloride powder because the difference between Py
and 100MPa is small and because the maximum
strength of the chloride agglomerates is 100 MPa. So
after compaction at 100 MPa all the agglomerates of
the chloride powder are rearranged but the internal
agglomerate structure remains the same.
A decrease in inter-aggregate pore size is observed
in all types of powder by further compaction up to
P = 400 MPa (see Table IV and Fig. 4c). The pores,
in the alkoxide and chloride powders are as small as R
= 3 to 4rim and in the Toyo Soda compact R =
17 nm. The Toyo Soda compacts did not clearly show
any pores in region 1. The pores in this region can
probably be ascribed to pores within the aggregates,
the intra-aggregate pores. This will now be discussed.
The presence of aggregates can be demonstrated in
the following way. As shown previously, the primary
crystallites of the alkoxide and chloride powder were
comparative in size (8 nm). The Toyo Soda powder
had a crystallite size of 34nm (see Table I). These
crystallite sizes were determined by X-ray line
broadening measurements. Crystallite sizes can also
TABLE IV Pore-size distribution data for various pressures
and sections 1 and 2
Synthesis 4 MPa 100 MPa 400 MPa
1 (nm) 2(nm) l(nm) 2(nm) 1 (nm) 2(nm)
Alkoxide 0-3 15 0-3 10 0 3 4
Chloride * 7 0-3 7 0-3 3
Toyo Soda * 28 * 24 * 17
9 No pores detected in this section.
3.0- (~-
I
1 ,. 2 I 3
o5
,
o
.....
1 O[ (b)
0.5 , ~J ii
"0 /I
/
iI
I i
10 100 1000
pore radius (nm)
Figure 4
Pore-size distribution within the powder compacts after
isostatic pressing at: (a) 4MPa, (b) 100MPa, (c) 400MPa. ( )
Alkoxide, (----) Chloride, ( ....... ) Toyo Soda.
be calculated from the specific surface area (SBET). In
that case an assumption is made that the total surface
of a spherically shaped crystallite is accessible for the
absorbed gas. The crystallite diameter (d=) calculated
from
SBE T
is
ds = 6/(ot SBEs)
(4)
where Ot is the theoretical density and SBET the specific
surface area. The crystallite sizes as determined by
X-ray line broadening (d X) and the crystallite sizes as
calculated from Equation 4 (d=) are given in Table V.
The ds value of the chloride powders fits well with the
d,- value, while for the alkoxide and Toyo Soda pow-
ders these values do not match. The difference in
crystallite size value found with X-ray line broadening
and with SBET can be explained by the fact that in these
types of powder the primary crystallites are clustered
and form an aggregate. In this way Argon, used for
the BET measurement, cannot completely cover the
crystallite surface. So these differences between d= and
dx give some indication of the degree of aggregation.
The size and structure of aggregates is also deter-
mined by the number of necks, which are formed
by reaction (sintering) during calcination. According
to Ramsay [16] the number of necks (and therefore an
indication of the number of crystallites per aggregate)
TABLE V Aggregate properties
Synthesis d X (XRLB) d~ (BET) Necks*, Particles
(nm) (nm) n
cp sc
Alkoxide 8 l0 15 8 l0
Chloride 8 8 0 - -
Toyo Soda 34 47 96 42 50
*Necks = number of necks within an aggregate.
+ particles = number of particles per aggregate, cp = close packing
assumed, sc = simple cubic packing assumed.
3453