J. Am. Ceram. Soc.
, 80 [6] 146976 (1997)
Experimental Design Applied to the Chemical Durability of
SolGel-Derived Zirconias
Robert S. Pavlik Jr.,*, Lisa C. Klein,* and Ronald A. McCauley*
Ceramics Department, RutgersThe State University of New Jersey, Piscataway, New Jersey 088550909
A statistical design was used to investigate the effects of
various processing conditions on the chemical durability of
solgel-derived zirconias. Eight processing variables were
investigated: water:alkoxide molar ratio, alcohol:alkoxide
molar ratio, nitric acid:alkoxide molar ratio, mixing tem-
perature during hydrolysis, firing temperature, heating
rate, soak at firing temperature, and firing atmosphere.
Processing variables were set at high and low limits in a
main-effects statistical design. The design identified four
critical processing variables (listed in descending order
of significance): firing temperature, firing atmosphere,
water:alkoxide ratio, and nitric acid:alkoxide ratio. Sol
gel-derived zirconias were characterized by differential
scanning calorimetry, X-ray diffractometry, and mercury
intrusion. Suggestions are offered that may help to explain
why these particular processing conditions influenced the
chemical durability of solgel-derived zirconias.
I. Introduction
URING recent years, ceramic membranes have been pre-
D pared by the solgel process. Membranes composed of
alumina, silica, titania, and zirconia are discussed in the
literature.14 It is possible to synthesize ceramic membranes
with pore diameters of a few nanometers.5 Ceramic membranes
are expected to compete with more-conventional polymeric
membranes in application areas that require high temperatures
and extreme pH conditions. The chemical durability of a
membrane material is important in applications such as caustic
material recovery and in membrane-cleaning procedures. For
example, it is common practice to use sodium hydroxide to
remove trapped organic matter from inorganic membranes.6
There are several advantages of using zirconia instead of
silica, titania, or alumina, including alkaline resistance and
biocompatiblity.79
Zirconia exhibits three polymorphs at 1 atm pressure. The
monoclinic phase (m-ZrO2 , 5.83 g/cm3) is stable from room
temperature to 1170C. Between 1170 and 2370C, the tetrago-
nal form (t-ZrO2 , 6.10 g/cm3) is thermodynamically stable.
Above 2370C, but below the melting point (2680C), the cubic
form (c-ZrO2 , 6.09 g/cm3) exists.10 Density differences between
phases can cause microcracking when thermal cycling occurs
between the monoclinic and tetragonal phases. All three
phases have been reported to exist below 500C in zirconia
powders synthesized by chemical routes.11 Amorphous zirconia
(a-ZrO2 ) phases have been detected in zirconia gels up to 450C
by perturbed-angular-correlation (PAC) measurements.12,13 It is
generally accepted that a metastable tetragonal (t-ZrO2 ) form
crystallizes first from an amorphous phase at 250C in
B. Dunncontributing editor
Manuscript No. 192128. Received December 22, 1995; approved June 3, 1996.
Financial support was given by Texaco Foundation and the Center for Ceramic
Research, a New Jersey Commission on Science and Technology Center.
*Member, American Ceramic Society.
Current address: Corning, Incorporated, HP-88-01-2, Corning, NY, 14831.
1469
alkoxide-derived zirconias.14 Reasons for the existence of meta-
stable zirconia phases at room temperature have been discussed
in the literature and remain controversial. Particle-size consid-
erations,15 structural similarities between amorphous gels and
crystalline phases,16 and impurities17,18 have been reported. The
monoclinic form crystallizes at 430C.12,13 The phase devel-
opment sequence of alkoxide-derived zirconias up to 1080C
has been reported12,13 and is given in Eq. (1).
a-ZrO2 t-ZrO2 ( small a-ZrO2 )
m-ZrO2 t-ZrO2 m-ZrO2 (1)
The chemical durability of gel-derived zirconia is reported to
be influenced by both solgel-processing variables and firing
conditions.7,8 Different crystalline phases result in different
chemical durabilities.68 Based upon conventional corrosion
fundamentals, the tetragonal phase should be more durable than
either the monoclinic phase or an amorphous phase because of
the basic differences in structural density among these phases.19
Processing conditions can be controlled to tailor the phase
assemblage and, therefore, the durability of zirconia gels.
This study examines the caustic durability of zirconia gels
processed under different solgel-processing and firing condi-
tions. Because of the many variables and their possible interac-
tions, a statistical design has been utilized.
A PlackettBurman statistical design is a reliable and effi-
cient method for identifying critical variables.20 One type of
PlackettBurman design is a main-effects test that identifies
critical variables and eliminates weak factors and interactions.
Moreover, this type of test is based upon a minimum number of
trials. After the critical variables are identified, a separate and
more-manageable interaction test, involving only significant
(critical) variables, is performed.
This paper presents the results of a main-effects statistical
test and how it has been utilized to optimize the durability of
zirconia gels in 1N sodium hydroxide. Suggestions are pre-
sented that may help to explain the effect of processing condi-
tions on the chemical durability of gel-derived zirconias.
II. Experimental Procedure
Solgel-derived zirconia powders were prepared by combin-
ing zirconium n-propoxide (Lot E08B15, Johnson Matthey
Electronics, Wayne, PA), 2-propanol (Lot 910683, Fisher Sci-
entific, Pittsburgh, PA), and distilled/deionized water with or
without nitric acid in a narrow-necked flask. Addition of water
to zirconium n-propoxide instantaneously caused precipitation
of zirconia particles. Hydrolysis and condensation can be
described by the following reaction sequence:
Zr(O(CH2 )2CH3 )4 H2O
ZrO(O(CH2 )2CH3 )2 2((CH2 )2CH3OH) (2)
2ZrO(O(CH2 )2CH3 )2 ZrO2 Zr(O(CH2 )2CH3 )4 (3)
(1) Powder Preparation
Eight different compositions were prepared according to the
statistical design (see Tables I and II). Order of addition
remained the same for all compositions.
1470 Journal of the American Ceramic Society Pavlik et al. Vol. 80, No. 6

Table I. Main Effects Test in 12 Trials

Effect
Run
No. x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11

1
2
3
4
5
6
7
8
9
10
11
12

Table II. Process Variable Limits

Variable xi limit limit
Water:alkoxide ratio (mol%) x1 10 1
Alcohol:alkoxide ratio (mol%) x2 10 2.5
Firing atmosphere x3 No getter With O2 getter
Acid:alkoxide ratio (mol%) x4 0.125 0
Mixing temperature during hydrolysis (C) x6 50 25
Firing temperature (C) x7 675 400
Soak at firing temperature (min) x8 300 30
Heating rate (C/min) x9 20 5

Zirconium n-propoxide was diluted in 2-propanol using an
alcohol:alkoxide molar ratio of either 2.5 or 10. Acidified
((0.125 mol of nitric acid)/(mol of zirconium n-propoxide)) or
unacidified water was added to a diluted alkoxide mixture.
Distilled/deionized water was added at a water:alkoxide molar
ratio of either 1 or 10. Mixing took place for 0.5 h at either 25
or 50C. After mixing, excess liquid was decanted, and the wet
precipitate was dried at 120C for 24 h in a drying oven.
Powders were washed with distilled/deionized water, dried, and
stored in a desiccator. The powder preparation procedure is
represented in Fig. 1. Powders were compacted and subjected
to different firing conditions according to the statistical design
(see Tables I and II).
(2) Compact Preparation
Powders were partially calcined at 200C for 0.5 h to remove
surface moisture. Powders were ground with a mortar and pes-
tle and sieved between mesh sizes 250 and 325. Quantities of
powders were weighed and placed in a hardened steel die with
a punch diameter of 10 mm. The moisture that remained in the
powders acted with sufficient binding action to form compacts.
Powders were uniaxially pressed at 280 MPa pressure.
Pressed compacts were fired under different conditions spec-
ified by the statistical design (see Tables I and II). Compacts
were fired in air. Commercial-grade steel wool was placed
around samples during firing, where necessary, to provide
more-reducing conditions. For discussion purposes, the steel
wool often is referred to as an oxygen getter. Compacts were
heated at either 5 or 20C/min ramps to a firing temperature
(400 or 675C). The temperatures were chosen to promote the
development of different crystalline phases without an exces-
sive decrease in surface area. Porous bodies are required for
membrane applications. The compacts remained at the firing
temperature for either 30 or 300 min. After the compacts were
fired, their weight and dimensions were measured. Compacts
were used instead of powders for durability tests, because
compacts better represent commercial, bulk (unsupported)
membranes.
(3) Durability Test
Fired compacts were placed into an alkaline-resistant basket.
The compact within the basket was suspended in a beaker
containing 250 mL of 1N sodium hydroxide (pH 13) using a
platinum wire. The top of the beaker was covered with plastic
film to reduce evaporation. The solution in the beaker was
agitated with a magnetic stirrer, and the temperature of the
solution was maintained at 25C.
After the compacts were immersed for 9 d, they were
removed from the caustic solutions and washed with distilled/
deionized water until the pH of the wash water became neutral.
Next, a second rinse was made with 2-propanol. The samples
were calcined to 400C to drive off excess solvent. The com-
pacts were immediately taken from the calcining furnace and
weighed. Weight loss was measured, and an apparent durability
coefficient (K d with units of g/cm2) was calculated according to
the expression
W
Kd (4)
S
where W is the net weight loss (initial minus final weight after
9 d immersion, in grams) and S is the surface area of the com-
pact calculated from measured dimensions. The smaller the K d
value the better the durability.
(4) Statistical Design
Statistical designs are often used to study several variables
economically. This main-effects test includes eight processing
variables that were suspected to affect the chemical and physi-
cal properties of solgel-derived zirconias. The eight variables
(see Table II) are water:alkoxide molar ratio (x1 ), alcohol:
alkoxide molar ratio (x 2 ), firing atmosphere (x 3 ), nitric acid:
alkoxide molar ratio (x4 ), mixing temperature during hydrolysis
(x 6 ), firing temperature (x 7 ), soak at the firing temperature (x 8 ),
and heating rate (x 9 ).
A symbol (x i ) is designated to each variable, where i
1, 2, 3, , 11. Each numeral corresponds to a specific variable
and column in the design; e.g., x1 is the variable water:alkoxide
molar ratio and is contained in the first column. The main-
effects test is given in Table I, where a plus sign denotes an
upper limit, and a minus sign denotes a lower limit. Columns
x 5 , x10 , and x11 have been reserved for degrees of freedom (error
terms) in the test. The upper and lower limits for each of the
variables listed above are given in Table II.
If each of the eight variables is assigned upper and lower
limits, and a factorial experiment is used, the number of permu-
tations is 28, requiring 256 experiments. One approach is to use
June 1997 Experimental Design Applied to the Chemical Durability of SolGel-Derived Zirconias 1471

Fig. 1. Solgel-derived zirconia powder preparation scheme.

a full fractional factorial design, but this sometimes contains alumina powders (A-16 Alumina, Alcoa, Bauxite, AR). Mer-
insignificant interaction terms. An alternative approach is to use cury porosimetry measurements were performed (Model 9305
a main-effects test.20 To minimize the number of runs, main- pore sizer, Micromeritics, Norcross, GA) to gain information
effects tests are used to identify significant (critical) variables related to the microstructure of compacts. Surface areas of
and eliminate weak factors. Interactions are considered in a compacts were estimated using mercury intrusion data
separate manner only after the critical variables have been according to a method outlined in Ref. 23.
identified. The test given in Table I allows eight variables to be
tested in only twelve trials. III. Results
(5) Characterization (1) Main-Effects Test
Differential scanning calorimetry (DSC, Model 911, TA Apparent durability coefficients (K d ) calculated for all com-
Instruments, Wilmington, DE) was performed to gain structural pacts after 9 d in 1N sodium hydroxide are given in Table III.
information. Samples weighing 10 mg were heated at 10C/min All coefficients ranged between 1.73 104 and 24.68
in dry air from room temperature to 600C. Phase determina- 104 g/cm2, and sampling error was calculated to be 2%. Low
tion was performed using XRD methods (Kristalloflex Diffrak- coefficients were expected because zirconia is quite stable over
tometer, Siemens, Madison, WI). Samples were exposed to the entire pH scale, especially in excess base.8
nickel-filtered CuK radiation over a 2090 2 range using a
step width and count time of 0.02 and 1 s, respectively. The
tetragonal and monoclinic phases were identified using JCPDS Table III. Response Data from the
card file numbers 17-0923 and 36-0420, respectively. Higher- Main Effects Test
order geminate peaks were used to distinguish between the Run Kd
cubic and tetragonal phases. At temperatures below 700C, the No. (104 g/cm2)
tetragonal phase and not the cubic phase was expected in gel- 1 6.72
derived zirconias.1214 The weight fractions of the monoclinic 2 15.92
and tetragonal phases were calculated according to a method 3 20.18
delineated by Toraya et al.21 that predicted a sampling error 4 24.68
of 1%3%. This method did not consider the presence of 5 3.44
amorphous phases. Average crystallite size of the tetragonal 6 13.32
(111) peak situated at 30.2 2 was calculated using the Scher- 7 11.97
rer formula.22 Instrumental broadening was corrected for using 8 2.30
9 8.06
10 23.43
11 23.40

Joint Committee on Powder Diffraction Standards, International Centre for Diffrac- 12 1.73
tion Data, Newtown Square, PA.
1472 Journal of the American Ceramic Society Pavlik et al. Vol. 80, No. 6

Using the data given in Table III, numerical intercept with an asterisk are only best guesses. First-approximation
(b values) values were calculated according to standard statis- determinations should be treated with caution. For example,
tical methods.24 The b values are presented in the following a weak factor prediction given in Table IV calls for a low
expression: alcohol:alkoxide molar ratio to minimize weight loss. The
information given in Ref. 8 demonstrates otherwise; i.e., zirco-
y 12.93 104 2.5 104 x1 1.0 104 x 2 nia gels derived from higher alcohol:alkoxide molar ratios (10)
4.1 104 x 3 2.4 104 x4 0.0x 6 with high water:alkoxide ratios and high acid:alkoxide ratios
show better durability in sodium hydroxide than do gels with
5.3 104 x 7 0.6 104 x 8 1.1 104 x 9 (5) lower alcohol contents.
The grand mean is given by the first term (12.93 10 ) in 4 The apparent durability coefficient data are normally dis-
Eq. (5) and indicates how meaningful the level of the trials is tributed. To determine this, residuals have been calculated
on average over a long period of time. Equation (5) contains according to
critical variables and weak factors. The magnitude of the b Residual K dj K davg (9)
values determines each variables significance. For example,
the magnitude of the coefficient in front of variable x 7 (firing where K davg is the average of the K d data and is equal to 12.93
temperature) is the largest; therefore, variable x 7 is the most 104. K dj corresponds to each K d value listed in Table III.
significant main effect in the study. By contrast, x 6 (mixing Each value is determined from its run number ( j), where j
temperature during hydrolysis) is the least significant main 1, 2, 3, , 12.
effect. Calculated residual values are listed in Table V. These residu-
Confidence limits were calculated from the no-effect b als are plotted as a function of run number in Fig. 2. The
values (NEBVs) according to standard statistical methods. The random nature of the data implies homogeneity, i.e., no extrane-
NEBVs (b 5 , b10 , and b11 ) indicate the amount of variation ous factors. In Table V, the residuals are ranked from lowest
within the responses. A standard deviation term () can be (most negative) to highest (most positive). Probability values
obtained from the NEBVs according to also are given in Table V and can be calculated according to
1/2
rank
(b 0) (b 0)2 (b11 0)2

2
T 1/2
5 10
(6) Probability (%) 100 (10)
3 T1
where T is the number of trials in the test. The standard devia- where T is the number of trials, i.e., 12. The residuals are
tion term for this test has three degrees of freedom and has been plotted against probability on a probability grid in Fig. 3. These
calculated to be 5.20. Using this value, the confidence limits data are linear and imply the existence of a normal distribution.
can be calculated according to (2) Interaction Test
Confidence limits (t 1/2) (S)/(T) 1/2
(7) The interaction test involved only critical variables identified
from the main-effects test. Because fewer variables were con-
where t is the student t value at a particular level. Significant sidered in the interaction test, more degrees of freedom for error
factors should lie outside the confidence limits. For example, at were able to be pooled. As a consequence, the test was expected
90% confidence, x 7 and x 3 are the only significant variables. to result in better prediction capability, and intercept coeffi-
However, in the analysis of this test, confidence limits have cients (b values) were expected to change slightly.
been relaxed below the 90% level to compensate for small
distortions resulting from interactions.
Variables having b values outside of the confidence limits are Table IV. Optimal Settings for Each Variable to
considered significant, and the variables that have b values that Minimize Weight Loss
fall near zero are considered insignificant. The main-effects test
Variable xi Setting
does not consider interactions, but interactions tend to distort b
values somewhat. Therefore, variables having b values that fall Water:alkoxide molar ratio x1 10 ()
within the calculated confidence limits but are far from zero can Alcohol:alkoxide molar ratio* x2 2.5 ()
be considered significant in this test for the purpose of finding Firing atmosphere x3 With O2 getter ()
interaction terms. The expression given in Eq. (8) is similar to Acid:alkoxide molar ratio x4 0.125 ()
Eq. (5) but contains only significant (or critical) variables. Mixing temperature during
hydrolysis (C)* x6 50 ()
y 12.93 104 2.5 104 x1 4.1 104 x 3 Firing temperature (C) x7 675 ()
Soak at firing temperature (min)* x8 300 ()
2.4 104 x4 5.3 104 x 7 (8) Heating rate (C/min)* x9 20 ()
The regression fit of Eq. (8) is 84.5%. This regression value *Best guesses.
is expected to increase as interaction terms are included in the
expression. In Eq. (8), x1 (water:alkoxide molar ratio), x 3 (firing
atmosphere), x4 (nitric acid:alkoxide molar ratio), and x 7 (firing Table V. Residual Values by Rank and Run Number and
temperature) are considered critical variables. The remaining Their Corresponding Probabilities
processing variables (given in Table II) are considered weak
Residual value Probability
factors. Rank Run No. (104) (%)
The sign of each b value determines how each variable
should be set to minimize weight loss or K d . For example, the b 1 12 11.20 7.69
value corresponding to firing temperature has a negative sign. 2 8 10.63 15.38
3 5 9.49 23.08
Therefore, the higher firing temperature limit must be used to 4 1 6.21 30.77
keep weight loss to a minimum (more negative). This is a 5 9 4.87 38.46
straightforward procedure for critical variables. However, it is 6 7 0.96 46.15
more arbitrary for weak factors (coefficients near zero) because 7 6 0.39 53.85
of the difficulty in drawing conclusions from them; weak factors 8 2 3.02 61.54
show little or no effect. Nevertheless, a determination can be 9 3 7.25 69.23
made for these variables to a first approximation. Table IV pre- 10 11 10.47 76.92
dicts how each variable should be set to minimize weight loss 11 10 10.50 84.62
12 4 11.75 92.31
in this test field using this procedure. The factors in Table IV
June 1997 Experimental Design Applied to the Chemical Durability of SolGel-Derived Zirconias
Fig. 2. Residual plot.
Fig. 3. Probability plot.
After critical variable data from the main-effects test were
analyzed using a computer algorithm and standard statistical
methods,9 two interaction (second order) terms were identified.
A stronger interaction term, x4 x 7 (nitric acid:alkoxide molar
ratio and firing temperature), and a weaker interaction term,
x1 x 7 (water:alkoxide molar ratio and firing temperature), were
identified. Similar to the main effects, the relative strength of
each interaction term was indicated by the magnitude of the
corresponding coefficient preceding the term.
The interaction terms involve setting each critical variable at
its upper limit ( ), such that the product of the two
critical variables results in a new product holding the sign of an
upper limit. However, to minimize weight loss, the upper limit
product must hold a negative sign. This is demonstrated in
the expression
y 12.93 104 3.4 104 x1 3.8 104 x 3 Endotherm
3.0 10 x4 5.3 10 x 7
4 4
2.6 104 x4 x 7 1.7 104 x1 x 7 (11)
1473
Negative signs precede the interaction terms in Eq. (11). A
regression fit for Eq. (11) is 96.5%. Equation (11) is expected
to describe chemical durability of the gel-derived zirconias in
1N sodium hydroxide under the experimental conditions.
(3) Characterization
DSC was performed to obtain structural information.
Table VI summarizes the DSC results for the zirconia gel sys-
tems derived from different water (x1 ) and acid (x4 ) contents.
All the samples in Table VI demonstrate initial endothermic
reactions that are attributed to loss of solvent and water in
gels.12,13 Low-temperature exotherms, located in the 250
290C range, are present for all systems. These exotherms are
typical of oxidation in alkoxide systems; byproduct evolution
has been reported accompanying crystallization of a gel into the
tetragonal phase from an amorphous state.8 Higher-water gels
have slightly sharper exotherms at this location. Also, gels
with higher acid contents demonstrate sharper exotherms at this
location. Sharper exothermic peaks may suggest the conversion
of more amorphous phase into a crystalline state.12,13 Acidified
gels appear to crystallize sooner than unacidified gels. Exother-
mic reactions detected in the 440450C range are crystalliza-
tion peaks associated with the monoclinic phase and coincide
with dehydroxylation and partial organic burnout.8,13 In both
cases, gels made under acidified conditions show sharper exo-
thermic peaks than unacidified gels.
XRD was used to determine phases present and average
crystallite sizes. The weight fractions of the monoclinic and
tetragonal phases were calculated using the relationships given
in Ref. 21. These relationships do not consider the existence of
phases other than the monoclinic and tetragonal phases in a
zirconia system. All compacts showed the same general behav-
ior. Up to 300C, the samples were predominantly amorphous
to XRD. At 350C, the metastable tetragonal form was the
major phase. As firing temperature increased, the tetragonal
phase diminished. This was typical behavior for alkoxide-
derived zirconias.8 Average crystallite size of the tetragonal
phase was determined using the Scherrer formula.22 Table VII
lists the weight fractions of the tetragonal and monoclinic
phases and average crystallite sizes obtained under different
processing conditions. The weight fraction of the tetragonal
phase decreased with increased firing temperature (x 7 ) while x1 ,
x 3 , and x4 remained constant. At 400C, 65% of the compact
was in the tetragonal form, whereas only 27% remained tetrago-
nal at 675C. The average crystallite size increased by a factor
of 1.7 from 107 nm at 400C to 179 nm at 675C (compare
rows A and B in Table VII). Under more-reducing firing condi-
tions (x 3 is ), and, while x1 , x 3 , and x 7 remained constant, a
larger weight fraction of the tetragonal phase and a higher
average crystallite size were determined (compare rows A
and C in Table VII) compared with gels processed under more-
oxidizing conditions. Gels fabricated with lower weater:alkox-
ide molar ratios and with constant x 3 , x4 , and x 7 parameters
had slightly lower tetragonal content and a smaller average
crystallite size (compare rows B and D in Table VII) compared
to gels with higher water content. Acid addition (x4 0.125)
resulted in a higher tetragonal weight fraction and a comparable
average crystallite size compared to unacidified gels (compare
rows A and E in Table VII) under otherwise similar conditions.
Surface areas of compacts were calculated from mercury
intrusion data according to a method given in Ref. 23. Surface
areas were calculated for compacts fabricated under acidified
Table VI. Summary of DSC Results
Temperature (C)/strength of peak
x1 10, x1 10, x1 1, x1 1,
x4 0 x4 0.125 x4 0 x4 0.125
70/w 80/w 110/w 90/w
Exotherm 260/w 250/s 290/vw 255/m
Exotherm 450/s 445/m 440/m 440/vs
vw is very weak, w is weak, m is medium, s is strong, and vs is very strong.
1474 Journal of the American Ceramic Society Pavlik et al. Vol. 80, No. 6

Table VII. XRD Results: Phase Determination and and the monoclinic phase becomes apparent at 430C. These
Crystallite Size for Gels under Different Processing Conditions results do not indicate if, or quantify how much, residual amor-
Average phous phase exists in these systems above 300C. Borrowing
Composition (wt%)
crystallite size from the results of recent TDPAC studies on these zirconia
Condition Tetragonal Monoclinic (nm)
gels,12,13 it is evident that amorphous (to XRD) zirconia phases
A: x1 10 are present up to firing temperatures of 450C. Amorphous
x3 phases are less durable than their crystalline counterparts.8,19
x4 0.125 No residual amorphous phases are expected in these gels at
x 7 400C 65 35 107 675C.12,13 This means better corrosion resistance and a lower
B: x1 10 K d value for gels fired to higher temperatures. These sug-
x3 gestions may explain the better durability demonstrated by
x4 0.125 compacts heated at higher temperatures in this test.
x 7 675C 27 73 179 (B) Firing Atmosphere (x 3 ): Compacts fired in the pres-
C: x1 10 ence of an oxygen-getter material (more-reducing firing condi-
x3 tions, x 3 is ) showed better durabilities than compacts fired
x4 0.125 without an oxygen-getter material present. Firing atmosphere
x 7 400C 75 25 138 has been reported to influence the surface area and phases
present in zirconias.8,2729 Compacts fired under more-reducing
D: x1 1 conditions showed a higher average crystallite size than com-
x3
x4 0.125 pacts fired in air. Comparison of rows A and C in Table VII
x 7 675C 25 75 149 demonstrates this fact. The compact fired under reducing condi-
tions demonstrated an average crystallite size of 138 nm,
E: x1 10 whereas a compact fired in air evinced an average crystallite
x3 size of 107 nm. Larger particle size indicates less surface area
x4 0 for attack and better corrosion resistance.19 This is demon-
x 7 400C 54 46 101
strated by lower K d values for compacts fired under more-
F: x1 1 reducing conditions. Compacts fired under reducing conditions
x3 also showed a higher weight fraction of the tetragonal phase
x4 0.125 than compacts fired under more-oxidizing conditions (compare
x 7 400C 71 29 114 rows A and C in Table VII). Under more-reducing conditions
(row C), 75 wt% of the compact remained in the tetragonal
form, but only 65 wt% of the compact remained in the tetrago-
conditions with high water:alkoxide molar ratios and reducing nal form under oxidizing conditions (row A of Table VII).
firing conditions. At 400 and 675C, respective surface area Based upon conventional corrosion fundamentals, the tetrago-
values of 4.16 and 2.58 m2 /g were determined. The surface nal phase should be more durable than either the monoclinic or
area of the compact decreased by a factor of 1.6 as the firing an amorphous phase because of the basic differences in struc-
temperature increased. The crystallite size for this compact tural density between these phases.19 The higher tetragonal
system decreased by a factor of 1.7 when the firing temperature content detected in compacts fired under more-reducing condi-
increased from 400 to 675C. This may suggest that crystallite tions should explain, at least in part, the higher durability
size is a satisfactory indicator of surface area for these observed for these compacts. Better durability was illustrated
compacts. by a lower K d value for compacts fired under more-reducing
firing conditions.
(C) Water:Alkoxide Ratio (x1 ): Compacts prepared using
IV. Discussion of Results higher water:alkoxide molar ratios (e.g., 10) showed better
durabilities than compacts derived from lower ratios (e.g., 1).
(1) Main Effects Zirconia gels fabricated from high water:alkoxide ratios (10)
(A) Firing Temperature (x 7 ): Firing temperature has the have been reported to show more uniform shrinkage, higher
strongest influence on chemical durability in this test. Compacts densities, and higher strengths after firing because of stoichiom-
fired at higher temperatures (675) have better durabilities than etry and particle morphology considerations.30,31 This suggests
those fired at lower temperatures (400C). Firing temperature that gels made with higher water contents should be more
can influence durability in many ways and has been reported to durable. Gels synthesized with more water showed comparable
affect such properties as surface area, crystallite size, and tetragonal contents to gels prepared with lower water contents.
phases present in ziroconia gels.1113,25,26 These properties can This is demonstrated in Table VII by comparing rows B and D.
influence the rate of chemical attack and the apparent durability Because the tetragonal content is not drastically different, the
coefficient. When the firing temperature was increased from influence of water content on durability may be better explained
400 to 675C, the surface area decreased from 4.16 to through crystallite size. Comparison of rows B and D in
2.58 m2 /g (a factor of 1.6). Surface area of zirconia gels has Table VII demonstrates an average crystallite size of 179 nm
been reported to decrease rapidly at temperatures 400C.25 for high-water gels and a crystallite size of 149 nm for low-
Less weight loss due to corrosion is expected for a compact water gels. Larger particle size indicates less surface area for
with less surface area,19 and lower K d values were observed for attack and better corrosion resistance.19 This may explain why
compacts with less surface area and fired at higher temperatures lower K d values were observed for zirconia compacts prepared
in this test. Average crystallite size, which may be a good with higher water:alkoxide molar ratios.
indicator of how surface area changes in a compact under these (D) Acid:Alkoxide Ratio (x4 ): Compacts prepared under
test conditions, increased with increased firing temperature. acidified conditions (x4 0.125) showed better durabilities
Comparison of rows A and B in Table VII demonstrates this than compacts prepared without an acid catalyst. Using acid
fact. Between 400 and 675C, the average crystallite size during zirconia gel preparation has been reported to enhance
increased from 107 to 179 nm (a factor of 1.7). Larger particle durability in both acidic and basic corrosion environments.8,32
size indicates less surface area for attack and better corrosion DSC results (Table VI) showed that acid addition promotes
resistance.19 This means a lower K d value. Phase development stronger crystallization of a gel from its amorphous form into a
in a compact also can affect durability. Taken together, the DSC metastable tetragonal crystalline state. This was demonstrated
and XRD results demonstrate that an amorphous zirconia phase by sharp exotherms for acidified systems. Sharp exotherms
crystallizes into the metastable zirconia form just below 300C, (near 250C) may suggest more conversion from an amorphous
June 1997 Experimental Design Applied to the Chemical Durability of SolGel-Derived Zirconias
state into a crystalline state.27 Acidified gels showed lower-
temperature exotherms as compared to unacidified gels. This
showed that acidified gels crystallized sooner and had more
time to grow from an amorphous structure into a crystalline
form. This may suggest that acidified gels are more crystalline
than unacidified gels and, consequently, more durable. Results
for an acidified gel system and its counterpart are given in
rows A and E in Table VII. Comparison of these systems
shows that, although crystallite size remains about the same,
the tetragonal content in the acidified system is higher by 11%
under otherwise similar conditions. Acid-derived zirconia gels
are known to contain more tetragonal content than gels precipi-
tated under more neutral conditions.33 Based upon conventional
corrosion fundamentals, the tetragonal phase should be more
durable than either the monoclinic or an amorphous phase
because of the basic differences in structural density between
these phases.19 This may explain why lower K d values were
observed for acidified gels.
(2) Interactions
Two interactions were identified from this main-effects test.
A strong interaction, nitric acid:alkoxide molar ratio (x4 ) with
firing temperature (x 7 ), and a weaker interaction, water:alkoxide
molar ratio (x1 ) with firing temperature, are discussed.
(A) Acid:Alkoxide Ratio (x4 ) and Firing Temperature (x 7 ):
A strong interaction was found between terms x4 and x 7 . When
set at their upper limits (0.125 and 675C, respectively), weight
loss was minimized, which meant a lower K d value and better
durability. DSC results (Table VI) showed that acid addition
promoted strong crystallization of a gel from its amorphous
form into a tetragonal crystalline state. This was demonstrated
by sharper exotherms for acidified systems. Sharp exotherms
(near 250C) may suggest more conversion from an amorphous
gel into a crystalline state,27 specifically into the metastable
tetragonal phase. Results for this gel system (x4 0.125, x 7
675C) and another (x4 0, x 7 400C) are given in rows B
and E, respectively, in Table VII. Comparison of these systems
shows that, although the amount of the tetragonal phase is
higher in the system represented in row E, it very well may
contain more amorphous (to XRD) phase. The presence of
an amorphous phase decreases durability relative to a more-
crystalline system.19 In Table VII (rows B and E), average
crystallite size for the acidified gel fired at 675C is 1.8 times
larger than an unacidified gel fired to 400C. Average crystallite
size may be a good indicator of how surface area changes in a
compact. In any event, larger particle size indicates less surface
area for attack and better corrosion resistance.19 This is demon-
strated by a lower K d value for this system.
(B) Water:Alkoxide Ratio (x1 ) and Firing Temperature (x 7 ):
A weaker interaction was present between terms x1 and x 7 .
When set at their upper limits (10 and 675C, respectively),
weight loss was minimized, which meant better durability. Gels
synthesized with higher water contents and higher firing tem-
peratures showed lower tetragonal contents compared with gels
prepared with lower water contents and higher firing tempera-
ture. This is demonstrated in Table VII by comparing rows B
and F. Zirconia gels fired at lower temperatures are expected
to contain more amorphous phase than gels fired at higher
temperatures.12,13 A more-crystalline compact is expected to
demonstrate better durability.19 Better durability also can be
explained by crystallite size. The average crystallite size of the
gels given in row B (x1 10, x 7 675C) is 1.6 times as
large as the gel system (x1 1, x 7 400C) given in row F of
Table VII. Larger particle size indicates less surface area for
attack and better corrosion resistance.19 This may explain why
lower K d values were observed for compacts prepared with
higher water:alkoxide molar ratios and fired to 675C.
V. Conclusion
Several processing variables were investigated using a statis-
tical design to determine their influence on chemical durability
1475
of zirconia gel compacts in sodium hydroxide. Apparent dura-
bility coefficients ranged between 1.73 104 and 24.68
104 g/cm2. Four critical processing variables were identified:
water:alkoxide molar ratio (x1 ), firing atmosphere (x 3 ), nitric
acid:alkoxide molar ratio (x4 ), and firing temperature (x 7 ). The
optimal conditions for each variable within the test limits were
presented. Two interaction terms also were identified: x1 x 7 and
x4 x 7 . Suggestions were offered that may help to explain why
particular processing conditions and interactions influenced the
chemical durability of these solgel-derived zirconias. Sugges-
tions included surface area, crystallite size, degree of crystallin-
ity, and tetragonal phase content. Considerations for future
work include an investigation of properties outside the statisti-
cal design test limits. Because the internal and external surface
areas of compacts influenced the durability, a better response
parameter may be the weight loss divided by the total surface
area for a given immersion time.
Acknowledgment: Corning, Incorporated, is acknowledged for techni-
cal and statistical support.
References
1
C. C. Yu and L. C. Klein, Characterization of Supported Alumina Mem-
branes Formed by an in situ SolGel Method, J. Am. Ceram. Soc., 78 [11]
314952 (1995).
2
C. J. Brinker, N. K. Raman, M. N. Logan, R. Sehgal, R. A. Assink, D. W.
Hua, and T. L. Ward, StructureProperty Relationships in Thin Films and
Membranes, J. Sol-Gel Sci. Technol., 4, 11733 (1995).
3
S. H. Hyun and B. S. Kang, Synthesis of Titania Composite Membranes by
the Pressurized SolGel Technique, J. Am. Ceram. Soc., 79 [1] 27982 (1996).
4
L. Klein, C. Yu, R. Woodman, and R. Pavlik, Microporous Oxides by the
SolGel Process: Synthesis and Applications, Catal. Today, 14, 16573 (1992).
5
A. F. M. Leenars and A. J. Burggraaf, The Preparation and Characterization
of Alumina Membranes with Ultrafine Pores, Part 2. The Formation of Sup-
ported Membranes, J. Colloid Interface Sci., 105, 2740 (1985).
6
R. R. Bhave, Inorganic Membranes: Synthesis, Characteristics, and Applica-
tions; p. 140. Van Nostrand Reinhold, New York, 1991.
7
M. Nevant and L. C. Klein, Durability of Porous Zirconia Powders in
Caustic, Mech. Corros. Prop., A, 61/62, 13742 (1991).
8
R. S. Pavlik Jr., L. C. Klein, and R. A. McCauley, Zirconia Gels in Concen-
trated NaOH, J. Am. Ceram. Soc., 78 [1] 22124 (1995).
9
R. S. Pavlik Jr., Zirconia Membranes by the SolGel Method; Ph.D. Thesis.
RutgersThe State University of New Jersey, Piscataway, NJ, 1994.
10
R. Stevens, An Introduction to Zirconia: Zirconia and Zirconia Ceramics,
Publication No. 113, 2nd ed.; p. 12. Magnesium Elektron, Ltd., Flemington, NJ,
July 1986.
11
K. S. Mazdiyasni, Powder Synthesis from MetalOrganic Precursors,
Ceram. Int., 8 [2] 4256 (1982).
12
P. C. Rivas, J. A. Martinez, M. C. Caracoche, A. R. L. Garcia, L. C. Klein,
and R. S. Pavlik Jr., Perturbed Angular Correlation Study of Zirconias Produced
by the SolGel Method, J. Am. Ceram. Soc., 78 [5] 132934 (1995).
13
J. A. Martinez, P. C. Rivas, M. C. Caracoche, A. M. Rodriguez, A. R. Lopez-
Garcia, R. S. Pavlik, and L. C. Klein, PAC Study of Zirconia Powders Obtained
from Acid Solutions; pp. 42427 in Hyperfine Interactions, C1, 1996.
14
R. Srinivasan, R. J. DeAngelis, G. Ice, and B. H. Davis, Identification of
Tetragonal and Cubic Structures of Zirconia Using Synchrotron X-radiation
Source, J. Mater. Res., 6 [6] 128792 (1991).
15
R. C. Garvie, The Occurrence of Metastable Tetragonal Zirconia as a
Crystallite Size Effect, J. Phys. Chem., 69 [4] 123843 (1965).
16
J. Livage, K. Doi, and C. Mazieres, Nature and Thermal Evolution of
Amorphous Hydrated Zirconium Oxide, J. Am. Ceram. Soc., 51 [6] 349
53 (1968).
17
R. Srinivasan, M. B. Harris, S. F. Simpson, R. J. DeAngelis, and B. H. Davis,
Zirconium Oxide Crystal Phase: The Role of the pH and Time to Attain the
Final pH for Precipitation of the Hydrous Oxide, J. Mater. Res., 3 [4] 787
97 (1988).
18
A. Ayral, T. Assih, M. Abenoza, J. Philippou, A. Lecomte, and A. Dauger,
Zirconia by the Gel Route, J. Mater. Sci., 25, 126874 (1990).
19
R. A. McCauley, Corrosion of Ceramics; Ch. 2. Marcel Dekker, New York,
1995.
20
R. L. Plackett and J. P. Burman, The Design of Optimum Multifactorial
Experiments, Biometrika, 33, 30525 (1946).
21
H. Toraya, M. Yoshimura, and S. Somiya, Calibration Curve for Quantita-
tive Analysis of the MonoclinicTetragonal ZrO2 System by X-ray Diffraction,
J. Am. Ceram. Soc., 67, C-119C-121 (1984).
22
H. P. Klug and L. E. Alexander, X-ray Diffraction Procedures for Polycrys-
talline and Amorphous Materials, 2nd ed.; Ch. 9. Wiley, New York, 1974.
23
H. M. Rootare and C. F. Prenzlow, Surface Areas from Mercury Porosi-
meter Measurements, J. Phys. Chem., 71 [8] 273336 (1967).
24
W. Knotts, Problem Solving Using Experimental Design Techniques,
Ceram. Eng. Sci. Proc., 16 [3] 11922 (1995).
1476 Journal of the American Ceramic Society Pavlik et al.
25
Ph. Colomban and E. Bruneton, Influence of Hydrolysis Conditions on
Crystallization Phase Transition and Sintering of Zirconia Gels Prepared by
Alkoxide Hydrolysis, J. Non-Cryst. Solids, 147/148, 201205 (1992).
26
X. Turrillas, P. Barnes, D. Hausermann, S. L. Jones, and C. J. Norman,
Effect of Chemical and Heat Treatment on the Tetragonal-to-Monoclinic Trans-
formation of Zirconia, J. Mater. Res., 8 [1] 16368 (1993).
27
G. Gimblett, A. A. Rahman, and K. S. W. Sing, Thermal and Related
Studies of Some Zirconia Gels, J. Chem. Technol. Biotechnol., 30, 5164
(1980).
28
N. A. Andreeva and L. V. Kozlovskii, Effect of Disturbing Compositional
Stoichiometry on the Sintering of Zirconia in a Vacuum, Refractories (Engl.
Transl.), 2, 12426 (1970).
Vol. 80, No. 6
29
D. E. Collins, K. A. Rogers, and K. J. Bowman, Crystallization of Metasta-
ble Tetragonal Zirconia from the Decomposition of a Zirconium Alkoxide Deriv-
ative, J. Eur. Ceram. Soc., 15, 111924 (1995).
30
B. E. Yoldas, Effect of Variations in Polymerized Oxides on Sintering and
Crystallize Transformations, J. Am. Ceram. Soc., 65, 387393 (1982).
31
B. E. Yoldas, Zirconium Oxides Formed by Hydrolytic Condensation of
Alkoxides and Parameters that Affect Their Morphology, J. Mater. Sci., 21,
108086 (1986).
32
S. J. Wang, S. H. Chan, and W. M. Hung, Structure and Thermal Behavior
Studies of Zirconia Treated by Different Concentrations of Sulfuric Acid,
Mater. Res. Bull., 26, 137983 (1991).
33
B. H. Davis, Effect of pH on Crystal Phase of ZrO2 Precipitated from
Solution and Calcined at 600C, J. Am. Ceram. Soc., 67, C-168 (1984).