Materials Science and Engineering A 498 (2008) 208–215

Contents lists available at ScienceDirect

Materials Science and Engineering A

journal homepage: www.elsevier.com/locate/msea

Effect of the addition of mullite–zirconia to the thermal shock

behavior of zircon materials
N.M. Rendtorff ∗ , L.B. Garrido, E.F. Aglietti
Centro de Tecnologia de Recursos Minerales y Cerámica (CETMIC: CIC-CONICET-UNLP), Camino Centenario y 506. C.C.49 (B1897ZCA) M.B. Gonnet. Buenos Aires, Argentina

a r t i c l e i n f o a b s t r a c t

Article history: Ceramics and refractories are subjected to local temperature and atmosphere gradients due to the suc-
Received 18 February 2008 cessive application of heating cooling cycles, during service. These situations originate thermal stresses
Received in revised form 21 July 2008 that cause certain degree of damage to the material defining the potential usage of the refractory. The
Accepted 25 August 2008
favorable effect of the addition of a second phase for improving the thermal shock resistance (TSR) of
dense ceramics have been proved previously. However the addition must be controlled because it can
Keywords:
excessively affect the densification and microstructure. The objective of this work is to study the TSR of
Zircon
pure zircon ceramic materials and to quantify the influence of the addition of mullite–zirconia (MZ) grains
Mullite–zirconia
Composites in a range between 15 and 45 wt.%. Cool water quenching test was carried out on slip cast prismatic bars
Mechanical properties in order to study the TSR. For the thermal shock tests temperatures differences (T) between 200 and
Thermal shock resistance 1200 ◦ C were applied. The elastic dynamic modulus (E) was evaluated by the impulse excitation technique
as a function of the temperature difference of the quenching test. The decrease in the E modulus was also
correlated with the number of thermal cycles. The materials were characterized by density measurements
and mechanical properties. The XRD and Rietveld method were employed for determining the crystalline
phase composition. Finally, the microstructure of the materials was examined by SEM. Significant changes
in the E modulus were observed for T over 400 ◦ C for all composites studied.
© 2008 Elsevier B.V. All rights reserved.

1. Introduction erties in structural applications and as an additive to other phases

Zircon (ZrSiO4 ) is good refractory material because it does
not undergo any structural transformation until its dissociation
at about 1700 ◦ C. It exhibits many attractive properties such as
excellent chemical stability, a very low thermal expansion coef-
ficient (4.1 × 10−6 ◦ C−1 ) from room temperature to 1400 ◦ C and low
heat conductivity coefficient 5. 1 W/m ◦ C at room temperature and
3.5 W/m ◦ C at 1000 ◦ C.
Sintered zircon with high purity can retain its bending strength
up to temperatures of 1200 and l400 ◦ C [1]. These properties make
zircon a potential candidate as a useful structural ceramic, espe-
cially in such fields where a sudden change of temperature may
occur. These materials are widely used in the steel and glass indus-
tries [2–4]. The presence of impurities decreases the dissociation
and usage temperatures. Moreover, pure zircon dense materials
degradation by a thermal shock is sometimes the limiting property
for its applications.
On the other hand mullite–zirconia (MZ) electrofused powders,
commercially available, have demonstrated good refractory prop-
∗ Corresponding author. Tel.: +54 2214840247; fax: +54 2214710075.
E-mail address: rendtorff@hotmail.com (N.M. Rendtorff).
forming a composite material [5–10].
The thermal shock resistance is improved by the deliber-
ate introduction of thermal stress concentrators in the form
of microstructural inhomogeneities in the material. Thus, the
dispersion of zirconia grains in a mullite matrix, an improved
thermomechanical behavior is obtained as a result of microcracks
formation and by the dissipation of elastic energy related to zirconia
martensitic transformation [11].
The first approach to determine the thermal stresses of brittle
material is the thermo-elastic theory [12,13], which is focused in the
initiation of the fracture. A second approach focuses on crack prop-
agation for conditions of thermal shock more severe than those for
crack initiation [13]. A unified theory of the thermal shock resis-
tance considering the initiation and crack propagation in brittle
ceramics was developed by Hasselman [14] who presented analyt-
ical solutions for the extent of the crack propagation as a function
of the severity of the thermal shock. Therefore, several theoretical
thermal shock parameters were proposed previously [15] to predict
the thermal shock resistance (TSR) of a certain material. These are
functions of the global properties of the material.
The principal objective of the present work is to study the
influence of the added amount of mullite–zirconia on the mechan-
ical properties and the thermal shock resistance of dense zircon

0921-5093/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.msea.2008.08.036
 N.M. Rendtorff et al. / Materials Science and Engineering A 498 (2008) 208–215 209

materials. The elastic modulus (E) was the property chosen to ent T. Dilatometry of samples sintered at 1600 ◦ C was performed
follow the material damage. This can be done because E is related using a Netszch dilatometer up to 1400 ◦ C at a heating rate of
to the microstructure integrity, meaning the number, size and 10 ◦ C/min. The thermal expansion coefficient ˛ of these materi-
shape of cracks developed. E was measured by a dynamic method, als up to 1000 ◦ C was determined. Microstructural examination
using the impulse excitation technique (sonic velocity), that is was conducted with a scanning electron microscope SEM (Jeol
a simple and non destructive standard test [16]. In this way the JSM 6360 LV) after polishing the probes surface until to 1 ␮m dia-
severity of the treatment was studied by determining the relative mond paste. Flexural strength ( f ) was measured on the bars with
drop in elastic modulus after exposure at a given thermal cycle. rectangular section using the three-point bending test. (Universal
Also the retention of E was evaluated after application of repeated testing machine INSTRON 4483) 40 mm of span and a displacement
thermal cycles. Finally, the TSR was compared with the behavior rate of 2.5 mm/min were employed. The fracture toughness (KIC )
predicted from parameters R, R and Rst which were obtained and the fracture initiation energy ( NBT ) were evaluated by the
from the thermoelastic and energy balance theories respectively. single edge notched beam method (SENB) [10,20] using a three-
point bending universal testing machine. Samples of dimensions
(7.5 × 7.5 × 50 mm3 ) were notched with diamond saw of 0.3 mm
2. Experimental
thickness, with depth between 0.3 and 2.5 mm. The three-point test
was carried out at room temperature with a rate of 0.1 mm/min. In
2.1. Slip casting of dense zircon based materials
this method KIC is given by:
 C   C 2  C 3  C 4 
A commercial zircon powder was used as principal raw material. 3QLC 1/2
With: ZrO2 = 64–65.5 wt.%, SiO2 = 33–34 wt.%, Fe2 O3 ≤ 0.10 wt.% KIC = A0 + A1 + A2 + A3 + A4
2WD2 D D D D
and TiO2 ≤ 0.15 wt.%, specific gravity of 4.6 g/cm3 , melting point
of 2200 ◦ C and mean diameter (D50 ) of 1.5 ␮m (Kreutzonit Super, (1)
Mahlwerke Kreutz, Germany).
where Q is the load applied to the notched bar in kg, L is the span in
Electrofused mullite–zirconia was used as second raw
meters, C is the depth of the notch in meters, D is the thickness of
material. With Al2 O3 = 44.0 wt.%, SiO2 = 18.0 wt.%, ZrO2 = 39.1%
the specimen in meters, W is the width of the specimen in meters,
TiO2 ≤ 0.3 wt.%, Fe2 O3 ≤ 0.1 wt.% and Na2 O ≤ 0.2 wt.%, specific
and A0 , A1 , A2 , A3 y A4 are functions of the ratio (L/D) described in
gravity of 3.74 g/cm3 melting point of 1850 ◦ C and mean diameter
[10,21]. Eq. (1) can be approximated by the following equation:
(D50 ) of 5 ␮m (MUZR, ElfusaLT, Brazil).
√
Zircon and MZ mixtures were prepared with 0, 15, 25, 35 and KIC ∼
= f C (2)
45 wt.% (in weight basis) of MZ and were called Z0, Z1, Z2, Z3, and
Z4, respectively. Aqueous 80 wt.% suspensions of the mixtures at pH where  f is the flexural strength in MPa. The calculated values of
9.1–9.2 were prepared by adding the powder to aqueous solutions KIC , together with E, were used to estimate the surface energy for
with suitable content of dispersant (Dolapix CE64, Zschimmers and the area created by the crack propagation ( NBT ) by the subsequent
Schwartz) and NH4 OH. After mixing, the suspensions were ultra- equation [10,20]:
sonicated for 20 min. The prismatic bars 7.5 × 7.5 × 50 mm3 were 
KIC = 2NBT E (3)
produced from well dispersed suspensions by slip casting in a plas-
ter molds. where  NBT can be expressed:
Dried probes were fired at heating rate of 10 ◦ C/min up to 1600 ◦ C
for 2 h. KIC 2  2 C
NBT = = f (4)
2E 2E
2.2. Characterization techniques Finally, the critical crack length (Lc ) was estimated from Eq. (2):
 2
After drying at 110 ◦ C green density of bars was determined by KIC
Lc = √ (5)
mercury immersion. Density and open porosity of sintered sam- f 
ples were determined by the water absorption method. Theoretical
density was calculated taking into account the density of each The critical length was also proposed by Hasselman [14] in terms
component: 3.16 g/cm3 for mullite, 5.89 g/cm3 for zirconia and of the surface energy less than a constant; the values obtained by
4.56 g/cm3 for zircon. Crystalline phases formed were analyzed by the two methods differ in a constant:
X-ray difraction (XRD) (Philips 3020 equipment with Cu–K␣ radi-  EUef
ation in Ni filter at 40 kV – 20 mA). The Rietveld method [17,18], Lc = R = (6)
t 2 (1 − )
a quantitative analysis, was carried out to characterize the present
crystalline phases in the materials. The XRD patterns were analyzed This method (SENB) can only be used if the sizes of the defects
with the program FullProf, which is a multipurpose profile-fitting (pores, cracks etc.) in the microstructure of the materials are smaller
program, including Rietveld refinement [19]. The starting crystal- than Lc .
lographic data for each phase was extracted from the literature.
The dynamic elastic modulus E of the composites was measured 3. Results and discussion
by the excitation technique with a GrindoSonic, MK5 “Industrial”
model. For the TSR experiments, the water quenching method was 3.1. Characterization of the sintered Z0–Z4
used. Thermal cycles with quenching temperature differentials, T
of 400, 600, 800, 1000 and 1200 ◦ C were applied. Sintered sample In Table 1, the physical characteristics and mechanical prop-
was heated at a selected temperature in an electrical furnace in erties of the materials sintered at 1600 ◦ C–2 h are shown. All the
air atmosphere for a period of 90 min and then cooled in a water materials studied had relative densities over the 90% of the theoret-
bath at 25 ◦ C. After quenching, samples were dried at 100 ◦ C and ical and the open porosity was less than 5%. The values of the elastic
then the TS severity effect on E as well as its variation with the modulus and the flexural strength are comparable with literature
number of applied thermal cycles was determined for the differ- data [22–24].
210 N.M. Rendtorff et al. / Materials Science and Engineering A 498 (2008) 208–215

Table 1 Table 2
Physical and mechanical properties of the materials Initial and final crystalline phase compositions of the studied materials in wt.%

Apparent density (g/cm3 ) Apparent porosity (%) ˛ (10−6 ◦ C−1 ) Lc (␮m) Starting Final

Z0 4.1 2 5.0 103 Zircon Mullite ZrO2 Zircon Mullite m-ZrO2 t- ZrO2
Z1 4.1 2 5.0 58
Z0 100 0 0 96.5 0 3.5 0
Z2 4.0 3 5.0 76
Z1 85 9 6 83 10 7 0
Z3 3.9 4 5.0 72
Z2 75 15 10 72 15 12 ≤1
Z4 3.9 5 5.0 54
Z3 65 21 14 62 21 16 ≤1
Z4 55 27 18 45 27 27 ≤1

10 wt.%. Zircon decomposition was not completely reversible due

to the presence of impurities and the relatively rapid cooling rate
(5 ◦ C/min) and therefore ZrO2 remained after cooling.
The degree of dissociation was more important in the material
with a higher proportion of MZ. Mullite proportion did not change
during the sintering. The partial dissociation of zircon with an
increase in the amount of ZrO2 implied that a silica rich glassy phase
was present. Therefore 5–10 wt.% of this non crystalline phase was
present in the composites. The final amount of t-ZrO2 in all the
materials can be neglected.
Fig. 3 shows the evolution of the crystalline phases (estimated
using the Rietveld method) as a function of MZ addition. In this
graph the initial composition is also represented. The zircon content
decreased near 5 wt.% whereas the amount of mullite remained
unchanged with the thermal treatment. Also, a small quantity of
zircon (≤5%) decomposed to ZrO2 and SiO2 for materials with less
than 35 wt.% of MZ whereas for Z4 the dissociated amount was
Fig. 1. Complete XRD pattern of the studied materials.
almost 20 wt.% of the initial zircon.

3.2. Crystalline phases present 3.3. Dilatometry

Diffractograms of the materials are shown in Fig. 1. For Z0 the
main crystalline phases are zircon accompanied by a little bad-
deleyite (m-ZrO2 ) originated by the zircon thermal dissociation.
On the other hand for the composite materials the peaks of the
initial components are present: zircon, baddeleyite and mullite
(3Al2 O3 ·2SiO2 ). In Fig. 2, a detail of the previous DRX pattern, the
gradual increase of the amount of the second phases (MZ) can be
seen.
Table 2 shows the initial composition of the materials before
processing and the final crystalline composition of the materials
evaluated by the Rietvield method. Less than 5 wt.% of the zir-
con thermally dissociated during sintering at 1600 ◦ C for Z0–Z3
compositions, except for Z4 in which dissociation was nearly
The dilatometric curves of the Z0 and Z4 materials are shown
in Fig. 4. The Z0–Z3 materials showed a linear expansion curve
with similar linear expansion coefficient, (˛ ≈ 5. 10−6 ◦ C−1 ) with
the exception of the Z4 composition which presented a hysteresis
loop area, due to the volume change associated to the m↔t trans-
formation. This change is not seen in the other materials perhaps
because of the small amount of m-ZrO2 and the presence of glassy
phase covers the volume change of the m-ZrO2 grains.
The m↔t transformation began at 1100 ◦ C on heating and indi-
cated that an important content of m-ZrO2 was present. This
transition occurred when the probes were sintered and when a
thermal shock test for values of T = 1000 and 1200 ◦ C was applied.

Fig. 3. Final crystalline composition of the studied materials evaluated by the

Fig. 2. A detail of the XRD pattern of the studied materials. Rietveld method in terms of the amount of MZ powder in the initial composition.
 N.M. Rendtorff et al. / Materials Science and Engineering A 498 (2008) 208–215
Fig. 4. Dilatometric curve of the materials sintered at 1600 ◦ C for 2 h; expansion%
as a function of temperature.
At 1600 ◦ C the stable phase is the tetragonal form, but on cool-
ing the monoclinic phase transformation took place in those grains
having a size higher than the critical one. For T = 1200 ◦ C a com-
plete transformation is expected, while a partial one will occur for
T = 1000 ◦ C.
3.4. Elastic modulus E0 and flexural strength  f
The elastic modules of the composites (Fig. 5) are similar to
those previously reported [1,24]. Results by the impulse excita-
tion technique are reliable and comparable with others obtained
by other methods such as three- and four-point bending methods
or indentation techniques [25].
The influence of the addition of MZ can be described as fol-
lowing: when it acts as an additive (very low concentrations) the
value of the elastic modulus is slightly increased. For proportions
of the second and third crystalline phases (i.e. mullite and zirco-
nia) higher than 15 wt.% the elastic modulus of these materials
decreased almost linearly. This behavior is clearly shown in Fig. 5.
The flexural strength  f (Fig. 5) showed a gradual increase with
the amount of MZ in the original composition. These values are
slightly lower than those given in the literature for pure mullite
(254 MPa) [12], mullite–zirconia (about 215 MPa) [13] and pure zir-
con (150–320 MPa for various products processed at 1600 ◦ C [14]).
The difference may be explained by the presence of pores, microc-
racks resulting from the zirconia transformation during cooling and
the presence of a non crystalline phase.
Fig. 5. Dynamic elastic modulus (E) and flexural strength ( f ) as functions of the
amount of MZ powder in the initial composition.
211
Fig. 6. Fracture toughness KIC and initiation surface energy  NBT as functions of the
amount of MZ powder in the initial composition.
3.5. Toughness and initiation fracture energy
Many methods are currently used to measure the fracture tough-
ness of ceramics materials. The SENB method is a simple and
inexpensive that has been positively used in several occasions for
this kind of materials [10,21].
At first glance the values of fracture toughness (KIC ) and the frac-
ture initiation energy ( NBT ) (Fig. 6) are similar to the ones found in
the literature for these materials [21]. The effect of the addition of
MZ was different for low and high proportions. An increase of the
 NBT and KIC values is observed with the composition as showed in
Fig. 6. With 15 wt.% of addition the fracture toughness and the sur-
face energy slightly decreased in comparison with the pure zircon
material.
3.6. Microstructure
SEM micrographs of the five materials are shown in Figs. 7–11.
All the samples had a dense microstructure with low residual poros-
ity which is similar to that reported in previous works [24]. There is
no a preferential direction in the microstructures of these materials.
The images demonstrate that the sintering was achieved; moreover,
some grains are joining.
In the pure zircon material (Z0, Fig. 7) no zirconia grains can be
observed neither detected with EDAX. In the Z1–Z4 composite a
zircon matrix (middle gray) can be seen with zirconia grains (light
Fig. 7. SEM micrographs of the Z0 (z: zircon, p: pore).
212 N.M. Rendtorff et al. / Materials Science and Engineering A 498 (2008) 208–215
Fig. 8. SEM micrographs of the Z1 (m: mullite, z: zircon, b: zirconia, p: pore).
Fig. 9. SEM micrographs of the Z2 (m: mullite, z: zircon, b: zirconia, p: pore).
gray– d ≤ 7 ␮m) and mullite grains (dark gray). Phase identification
was carried out by EDAX.
Mullite grains in the composites with more than 25 wt.% of MZ,
formed a nearly continuous phase. Hence it cannot be considered
as an additive.
Fig. 10. SEM micrographs of the Z3 (m: mullite, z: zircon, b: zirconia, p: pore).
Fig. 11. SEM micrographs of the Z4 (m: mullite, z: zircon, b: zirconia, p: pore).
3.7. Thermal shock behavior
3.7.1. Effect of quenching temperature difference (T) on the
E1/E0 ratio.
The TS behavior of the materials was evaluated by measuring the
decrease in E1/E0 ratio where E0 and E1 are the original dynamic
elastic modulus and those values after one quenching process,
respectively. The elastic modulus decreases if any damage exists,
like new cracks nucleation or propagation of existing ones [26].
Fig. 12 shows the evolution of E1/E0 ratio with T indicating that
the degree of damage was not significant up to T = 200 ◦ C for pure
zircon and up to T = 400 ◦ C for all the composites. Thus, the Tc
of the materials was near to 200 and 400 ◦ C, respectively.
In the range of T = 800–1000 ◦ C, the degree of damage was
clearly different between materials, the performance of the com-
posite enhanced with the added amount of MZ. This result agreed
well with that reported in literature [10,27,28] where the incorpo-
ration of a second phases gave better TSR and materials with higher
toughness.
Most of probes dramatically failed under a thermal shock
with T = 1200 ◦ C, which did not allow extrapolating the obtained
results to the actual operating conditions.
The general tendency of TSR may be well explained and pre-
dicted by the classical thermal shock resistance thermoelastic
models and Hasselman unified theory (Table 3). The observed
Fig. 12. Effect of quenching temperature difference (T) on the E1/E0 ratio.
 N.M. Rendtorff et al. / Materials Science and Engineering A 498 (2008) 208–215
Table 3
Theoretical thermal shock resistance parameters
R = (Tc ) R
Z0 65 0.030
Z1 81 0.019
Z2 88 0.017
Z3 101 0.014
Z4 138 0.008
behavior corresponded well with a non-static, kinetic, crack prop-
agation [13,14,26,28].
3.7.2. Effect of the number of quenching cycles on the E/E0 ratio
for different T
The effect of the repeated cycles at T = 800 ◦ C for the different
compositions is shown in Fig. 13 as the change of the E/E0 ratio
being E0 and E the elastic modulus for original and quenched sam-
ples, respectively, with the numbers of quenching cycles. There was
a rapid E/E0 drop after few cycles, and then a gradual reduction up to
achieve a practically constant value with increasing the number of
cycles. This is a typical behavior previously reported in the literature
[9]. All the composites (Z1–Z4) retained about 80% of the E0 at n = 10
cycles showing the main microstructure degradation occurred by
applying three thermal cycles. The effect on the elastic modulus of
cracks and microcracks was reviewed by Bradt and latter by Stiffler
and Hasselman [30,31]. Generally the resulting elastic modulus is a
function of the number and lengths of the cracks, so the decrease in
the elastic modulus after a thermal shock is directly related with the
nucleation and propagation of the cracks in the material. The crack
growths if elastic energy originated by the thermal shock is bigger
than the energy required to create the new surface of the cracks.
Thus, Figs. 13 and 14 show that after two or three cycles, the elastic
modulus did not decrease with the next thermal shock test which
means that the energy provided by the thermal shock was not high
enough to propagate the existing cracks. A higher addition of MZ
caused a better TSR. For the pure zircon material (Z0) the reduc-
tion in E/E0 was almost twice as that of the composites, indicating
a greater degradation. However, E/E0 remained unchanged for n
cycles higher than four cycles proving that no greater crack prop-
agations occurred. The thermal shock behavior for T = 1000 ◦ C of
the different materials is shown in Fig. 14. The E/E0 ratio followed
a similar variation with the number of cycles applied to the one
observed in tests with T = 800 ◦ C (Fig. 13). But in this case the Z0,
Z1 and Z2 materials retained a similar E, close to 50% of the original
Fig. 13. Evolution of E/E0 ratio with the number of thermal cycles at T = 800 ◦ C
applied for the different materials.
213
Rst
3.8
3.5
4.4
4.9
5.7
Fig. 14. Evolution of E/E0 ratio with the number of thermal cycles at T = 1000 ◦ C
applied for the different materials.
modulus E0 which is lower relative to T = 800 ◦ C. The remain-
ing E/E0 for Z3 and Z4 were 60% and 70% respectively. The lower
degree of damage with increasing MZ was probably due to the par-
tial occurrence of martensitic transformation during the thermal
treatment using a T = 1000 ◦ C.
The mismatch between the thermal expansion coefficients of
the different phases introduces micro-cracks in the grain bound-
aries. Moreover, microcracks developed during the processing by
the martensitic transformation. Contrarily to an intuition thought,
an increase in the number of the micro cracks improves the TSR of a
certain material, because the elastic energy provided by the thermal
shock is divided into a greater amount of cracks. Thus the energy on
each crack becomes not large enough to create a new surface, and
the cracks can not propagate. This is might be the reason why the
addition of micro crack “generators” improves the TSR. And in fact
the addition of MZ accomplishes this requirement in the two ways,
due to the mismatch of the thermal expansion coefficients of the
new phases and the volume change associated with the zirconia
transformation.
3.7.3. Theoretical thermal shock resistance parameters
The thermoelastic approximation proposed by Kingery estab-
lishes that the fracture will occur when the thermal stress equal
or overcomes the fracture strength of the material [28,15]. There R
and R parameters are defined, for h = ∞ and h = constant (h = heat
transference coefficient), respectively:
f (1 − )
R= (7)
E˛
kf (1 − )
R = Rk = (8)
E˛
where  f is the flexural strength,  is the Poisson ratio, E is the elastic
modulus, ˛ is the thermal expansion coefficient and k is the thermal
conductivity. The R parameter is the critical thermal difference for
the crack initiation Tc .
The energy balance approximation was developed by Has-
selman [14] and establishes that a sample will fracture if the
thermoelastic energy is superior to the energy required for the
creation of the crack surfaces. Assuming that the only energy trans-
ferred is the elastic energy from the thermal stresses. The following
expressions are deduced for the R and R parameters:
E
R = (9)
f 2 (1 − )
214 N.M. Rendtorff et al. / Materials Science and Engineering A 498 (2008) 208–215
 Uef E
R = Uef R =
f 2 (1 − )
where  f is the flexural strength,  is the Poisson ratio, E is the
elastic modulus, and Uef is the surface effective energy.
For the conditions used in this work it is reasonable to assume
that the influence of the Biot (h = ∞) can be neglected to calculate
R in the first model. Since materials studied can be taken as family
of materials, the thermal conductivity (k) and surface energy (Uef )
are expected to be similar. Then the analysis can be reduced using
Eqs. (7) and (9).
The unified theory of thermal shock fracture initiation and crack
propagation in brittle materials [14] includes the parameters men-
tioned before, and defines a new parameter, the thermal stress crack
stability parameter (Rst ). The Rst parameter depends on the fracture
surface energy (), the elastic modulus (E) and the linear thermal
expansion coefficient (˛), or in terms of the fracture toughness (KIC )
once a crack has been initiated, thermal shock failure is controlled
by the nature of the crack propagation through the material. This
parameter can be used to predict the behavior of materials with
sufficiently long cracks under severe thermal stresses.

1/2 KIC
Rst = = √
˛2 E 2E˛
Table 3 shows the values of the approximations calculated for
the materials. According to Hasselman [14] the first model (R) is
applicable to brittle materials where the crack initiation is deter-
mining in the TSR behavior. The R predicted, that TSR improved
with a major addition of MZ, showing the increase of Tc Thus, the
thermoelastic model correlated satisfactorily well with the exper-
imental data of E1/E0 ratio vs. quenching temperature difference
(T) and the number of quenching cycles for T = 800 and 1000 ◦ C.
The second approximation (R ) is valid where the initiation
of the cracks is unavoidable. In our materials, R does not cor-
relate the TSR behavior with the increase of MZ. This fact was
expected because of the inverse proportionality of this parame-
ter with square of the mechanical strength property that increased
with the amount of MZ in the material as seen in Fig. 5.
The microstructure of any ceramic material exhibits grains,
pores, grains boundaries, microcracks and other defects especially
materials obtained by slip casting. But the extent and length of
the cracks must be considered. The SEM micrographs clearly show
that these materials exhibited defects or microcracks, inherent to
the processing, and others due to the presence of m-ZrO2 and its
martensitic transformation. The microcracks and residual stress are
mainly generated by the thermal expansion mismatch between zir-
conia particles and zircon–mullite matrix. Zircon and mullite have
similar thermal expansion coefficient. These microcracks increase
with MZ addition. However, these defects are in all the cases smaller
than 10 ␮m, and are in “nucleation” stage.
The defects size resulted smaller than the critical crack length
(Lc ) (Table 1), estimated by Eq. (5), for all the materials studied [29].
The Rst parameter is mainly use to evaluate thermal shock dam-
age resistance behavior of refractory materials, where large cracks
are present. A numerical correlation between Rst results and exper-
imental data was observed, but as long cracks were not detected
we cannot conclude that the increase in Rst obeys to the presence
of long cracks and their stability to propagation.
A more simple analysis can be done; experimentally TSR
increased with the increase of the MZ proportion, the same hap-
pened with some properties like,  f ,  NBT , KIC , and the inverse of
the elastic modulus (1/E). Hence these facts could be extended to
other similar material as preliminary approach to predict the TSR.
Finally, the improved resistance to thermal shock can be
attributed to the microcracks distribution developed by zirco-
nia particles because it also increased with the amount of
(10)
MZ.
4. Conclusions
The influence of mullite–zirconia grains (MZ) on thermal shock
resistance of Zircon based composites prepared by slip casting has
been studied. When MZ proportion is over 25 wt.% it cannot be
considered as a simple additive.
The partial zircon decomposition occurred during the sinter-
ing being the degree of dissociation more important at high MZ
addition. Therefore, the composites and also the samples prepared
from the pure zircon had a zirconia content slightly higher that the
expected from the started compositions. Consequently the zirconia
content reached up to 22 vol% and the samples contained a minor
amount of non crystalline phase.
Addition of MZ grains increased zirconia content of sintered
composites, which probably reduced zircon densification result-
ing in an increase in porosity. Moreover, the volume change of
the zirconia phase transformation which is the cause of microcrak-
ing formation increased the porosity with increasing MZ addition.
The local differences in thermal expansion coefficients probably
(11)
contributed to microcraking. Therefore, initial elastic modulus E0
linearly reduced with MZ due to the changes in composition and
also to a slightly higher porosity. Whereas the strength of the com-
posites increased with the incorporation of zirconia, connecting
mullite grains (SEM).
The KIC and  NBT of 15 wt.% MZ composites were a little low
relative to that of pure zircon whereas further MZ addition resulted
in a considerable increase in KIC , reaching KIC for 45 wt.% of MZ twice
the initial value of that of pure zircon.
The thermal shock behavior of composites was typical of the
brittle ceramic materials. The Tc of the composites is between
200 and 400 ◦ C.
The damage of the quenched samples was satisfactorily evalu-
ated by a dynamic elastic modulus measurement.
The addition of MZ provided an increase in the TSR As expected,
the thermal shocks made with temperatures near the martensitic
transformation (T = 1000 or 1200 ◦ C) caused a more important
damage.
Such improvement corresponded well with an increase in KIC
and high surface energy  NBT for fracture propagation. In addition
the effect of the dispersion of an appropriate content m-ZrO2 grains
(microcraking, branching, etc.) was favorable.
The thermoelastic approximation (R) and the unified theory (Rst )
predict the experimentally observed thermal shock behavior of
these dense zircon–mullite–zirconia composites.
References
[1] Y. Shi, X. Huang, D. Yan, Ceram. Int. 23 (1997) 457–462.
[2] E.A. Thomas, T. Weichert, Proceedings UNITECR, 1989, pp. 730–760.
[3] J. Mori, N. Watanbe, M. Yoshimura, Y. Oguchi, T. Kawakami, A. Matsuo, Proceed-
ings UNITECR, 1989, pp. 541–553.
[4] C. McGarry, T. Wehrenberg, Proceedings UNITECR, 1989, pp. 826–834.
[5] H. Zender, H. Leistner, H. Searle, Interceram 39 (6) (1990) 33–36.
[6] P. Descamps, S. Sakaguchi, M. Poorteman, F. Cambier, J. Am. Ceram. Soc. 10
(1991) 2476–2481.
[7] N. Claussen, J. Jahn, J. Am. Ceram. Soc. (Note) 3–4 (1980) 228–229.
[8] L.B. Garrido, E.F. Aglietti, Ceram. Int. 5 (2001) 491–499.
[9] N.M. Rendtorff, L.B. Garrido, E.F. Aglietti, Ceram. Int. 34 (2008) 2017–
2024.
[10] V. Pandolfelli, V. Salvini, Proceedings UNITECR, 1993, pp. 282–291.
[11] M. Hamidouche, N. Bouaouadja, H. Osmani, R. Torrecillas, G. Fantochi, J. Eur.
Ceram. Soc. 16 (1996) 441–445.
[12] M. Hamidouche, N. Bouaouadja, C. Olagnon, G. Fantozzi, Ceram. Int. 29 (6)
(2003) 599–609.
[13] D.P.H. Hasselman, J. Am. Cer. Soc. 46 (1963) 535–540.
[14] D.P.H. Hasselman, J. Am. Cer. Soc. 52 (1969) 600–604.
 N.M. Rendtorff et al. / Materials Science and Engineering A 498 (2008) 208–215
[15] D.P.H. Hasselman, Am. Ceram. Soc. Bull. 49 (12) (1970) 1033.
[16] ASTM C1171-05 Standard Test Method for Quantitatively Measuring the Effect
of Thermal Shock and Thermal Cycling on Refractories.
[17] D.L. Bish, J.E. Post, Am. Mineral. 78 (1993) 932–940.
[18] Rodriguez, Caravajal, Abstracts of the Satellite Meeting on Powder Diffraction
of the IUCr, Toulouse, France, 1990, p. 12.
[19] H.M. Rietveld, J. Appl. Crystallogr. 2 (1969) 65–71.
[20] H. Harmuth, E. Tsechegg, Veitsch-Radex-Rundschau 1–2 (1994) 465–
542.
[21] M. Hamidouche, N. Bouaouadja, R. Torrecillas, G. Fantozzi, Ceram. Int. 33 (2007)
655–662.
215
[22] T. Mori, H. Yamamura, H. Kobayashi, T. Mitamura, J. Am. Ceram. Soc. 75 (9)
(1990) 2420–2426.
[23] C. Veytizou, J. Guinson, Y. Jorand, J. Eur. Ceram. Soc. 22 (2002) 2901–2909.
[24] Y. Shi, X. Huang, D. Yan, Mater. Lett. 23 (1995) 247–252.
[25] M. Radovic, E. Lara-Curzio, L. Riester, Mater. Sci. Eng., A 368 (1) (2004) 56–70.
[26] C. Baudin, boletín de la Sociedad Española de Cerámica y Vidrio, 1993, p. 4.
[27] P.M. Kelly, L.R. Francis Rose, Prog. Mater. Sci. 47 (5) (2002) 463–557.
[28] W. Kingery, J. Am. Ceram. Soc. 38 (1955) 3–15.
[29] C. Aksel, Ceram. Int. 29 (2003) 311–316.
[30] R.C. Bradt, Science of Ceramics, vol. 11, Gotemborg, Sweden, 1981.
[31] R.C. Stiffler, D.P.H. Hasselman, J. Am. Ceram. Soc. 66 (3) (1983) C52–C53.