786672
786672
786672
genehmigte Dissertation
vorgelegt von
aus Châlons-Sur-Marne
Diese Dissertation ist auf den Internetseiten der Universitätsbibliothek online verfügbar
Acknowledgments
Acknowledgment
The present work has been performed during my time as research assistant at the Institute
of Mineral Engineering (GHI) of the RWTH Aachen University.
This work is dedicated to the memory of my mother, Mrs Mauricette Traon, died during
my time as research assistant at GHI, and who was for me the principal source of
motivation to obtain a doctorate in engineering.
I would like to thank Univ.-Prof. Dr. rer. nat. Rainer Telle above all for the help and
support in attaining both academic and professional goals. I feel particularly grateful to
Prof. Dr. Telle for having given me the opportunity to carry out my thesis at GHI, where
I could deepen my knowledge in terms of scientific and linguistic skills that I can
henceforth put into practice in my current professional life.
During this time at GHI, I could take part to plenty of international conferences as well as
Federation for International Refractory Research and Education (FIRE) meetings that
turned out to be enriching professional opportunities by making acquaintance of
important persons of the academic or industrial world. In this sense, I would like to
dedicate special words to Univ.-Prof. Dr.-Ing. Jacques Poirier for the considerable
support in terms of knowledge transfer, for agreeing to act as a second examiner and for
his interest in the present study.
I would like to thank Dr.-Ing. Thorsten Tonnesen for the intensive support of the present
work and for his endless encouragements that greatly helped me to fulfil my academic
and professional expectations.
I would like furthermore to thank all staff members of the Institute of Mineral
Engineering for their professionalism over the years and for the pleasant working
atmosphere. I have in this sense a particular thought for my former colleagues Dr.-Ing.
Benjamin Schickle, M. Sc. Simon Etzold, Dipl.-Ing. Lise Loison and M. Sc. Jonas
Schnieder for their collaboration and the meaningful discussions that I could share with
them over the years.
I would like to thank finally Ms. Petra Schott and Mr. Volker Schmitz for their crucial
assistance in executing experimental procedures.
Pereira, A.H.A., Otani, L.B., De Anchieta Rodrigues, J., Traon, N., Tonnesen, Th., Telle,
R., Refractories Manual (Interceram), 02/2011, “The influence of nonlinear elasticity on
the accuracy of thermal shock damage evaluation by the impulse excitation technique”,
(ISSN 0020-5214), pp. 98-102
Pereira, A.H.A., Otani, L.B., De Anchieta Rodrigues, J., Traon, N., Tonnesen, Th., Telle,
R., Refractories Manual (Interceram), 06/2011, “The influence of nonlinear elasticity on
the accuracy of thermal shock damage evaluation by the impulse excitation technique”,
(ISSN 0020-5214), pp. 388-392
Traon, N., Tonnesen, Th., Telle, R., Refractories Worldforum 4, 01/2012, “The
understanding of the microstructural changes of refractory castables after thermal shocks
through damping measurements”, (ISSN 1868-2405), pp. 119-124
Traon, N., Tonnesen, Th., Franca, G. D., Piccolo, P., Telle, R., Refractories Worldforum
7, 08/2015, “Investigation of the Elastic Properties of Dense Ceramics of the Binary
System Al2O3 – ZrO2 after Thermal Shocks”, (ISSN 1868-2405), pp. 119-126
Etzold, S., Traon, N., Tonnesen, T., Telle, R., Interceram – International Ceramic
Review, Vol. 65, Issue 7, 04/2016, Elastic and Mechanical Fatigue at High Temperatures
of High-Alumina Castables with Addition of Partially Stabilized Zirconia, (ISSN 0020-
5214), pp. 19-23
i
List of abbreviations and symbols
A: Al2O3
AH3: Al2O3·3H2O
C: CaO
C3A: 3CaO·Al2O3
C 5 A3 : 5CaO·3Al2O3
CA: CaO·Al2O3
CAH10: CaO·Al2O3·10H2O
CA2: CaO·2Al2O3
CA6: CaO·6Al2O3
C12A7: 12CaO·7Al2O3
C2AH8: 2CaO·Al2O3·8H2O
C3AH6: 3CaO·Al2O3·6H2O
H: H2 O
Z: ZrO2
Ca-ZrO2: CaO-Partially Stabilized Zirconia
Mg-ZrO2: MgO-Partially Stabilized Zirconia
Y-ZrO2: Y2O3-Fully Stabilized Zirconia
α: α-corundum
β: β-alumina
mon or M: Monoclinic
tet or T: Tetragonal
cub or C: Cubic
P: Periclase
PSZ: Partially Stabilized Zirconia
FSZ: Fully Stabilized Zirconia
TZP: Tetragonal Zirconia Polycrystal
ZTA: Zirconia Toughened Alumina
Y-TZP: Yttria-Stabilised Tetragonal Zirconia Polycrystals
LTD Low Temperature Degradation
ii
List of abbreviations and symbols
iii
List of abbreviations and symbols
Tan (θ): Tangent of the point of the displacement curve with a maximum slope [-]
αmatrix: Coefficient of thermal expansion of the matrix [K-1]
αparticle: Coefficient of thermal expansion of the particle [K-1]
α: Coefficient mismatch [K-1]
iv
List of abbreviations and symbols
v
List of abbreviations and symbols
vi
List of abbreviations and symbols
vii
List of abbreviations and symbols
applied [mm²]
σMOR: Modulus of rupture (MOR) [MPa]
Mmax: Bending moment at the point of failure [N.m]
Wres: Moment of resistance [m³]
FBend,max: Maximum force applied on the test piece [N]
Ls : Distance between the points of support of the test piece [mm]
Tεmax: Temperature of maximum expansion [°C]
T0,5: Temperature at which the sample exhibits a percentage change in the
height of 0,5 % compared to Tεmax
v5h – 25h: Rate of the percentage change in the height of the test piece during the last
20h of the CiC experiment when the temperature is constant [%.h-1]
SVresidual after i TS: Residual sound velocity in the length of the specimen after i thermal shock
cycles [%]
SV after i TS: Sound velocity in the length of the specimen after i thermal shock cycles
[µs]
SV after sintering: Sound velocity in the length of the specimen after sintering [µs]
MORresidual after i TS: Residual Modulus of Rupture of the specimen after i thermal shock cycles
[%]
MOR after i TS: Modulus of Rupture of the specimen after i thermal shock cycles [MPa]
MOR after sintering: Modulus of Rupture of the specimen formulation after sintering [MPa]
αf : Non-linear coefficient related to the flexural resonant frequency [-]
frAmp i: Flexural resonant frequency of the peak point i in the time sinusoidal
signal decay function [Hz]
fr Amp max: Flexural resonant frequency of the peak point in the time sinusoidal signal
decay function with the maximum amplitude [Hz]
βf : Non-linear coefficient related to the damping of the flexural resonant
frequency [-]
ζAmp i: Damping of the flexural resonant frequency of the peak point i in the time
sinusoidal signal decay function [-]
ζ Amp max: Damping of the flexural resonant frequency of the peak point in the time
sinusoidal signal decay function with the maximum amplitude [-]
ts,L: Travel time of the acoustic signal through the prismatic sample length [s]
vL: Ultrasonic velocity of the acoustic signal through the length of the sample
[m.s-1]
flong: Longitudinal resonant frequency of the prismatic specimen [Hz]
ρ: Density [g.cm-3]
K: Correction factor [-]
vb: Ultrasonic velocity of the acoustic signal through the width of the sample
[m.s-1]
tb,L: Travel time of the acoustic signal through the prismatic sample width [s]
viii
List of abbreviations and symbols
vt: Ultrasonic velocity of the acoustic signal through the thickness of the
sample [m.s-1]
tt,L: Travel time of the acoustic signal through the prismatic sample thickness
[s]
αT: Coefficient of thermal expansion at temperature T [K-1]
L0 : Initial length of the sample at room temperature [mm]
T0 : Room temperature [K]
ΔL0: Length change at room temperature extrapolated from the first data point
[mm]
ΔLT: Length change at temperature T [mm]
χE: Non-linear coefficient related to the stiffness with increasing voltage for
excitation generation [GPa/V]
ΔE: Stiffness difference [GPa]
ΔV: Difference of voltage delivered to the electromagnet for excitation
generation [V]
ΔV: Difference of voltage delivered to the electromagnet for excitation
generation [V]
ς: Non-linear coefficient related to the flexural damping with increasing
amplitude of impulse excitation [-]
MAX 𝜁𝑓𝑓 : Maximum damping value of the flexural resonant frequency for a defined
thermal shock cycle in a range of amplitude of impulse excitation from
30 % to 100 % [-]
MIN 𝜁𝑓𝑓 : Minimum damping value of the flexural resonant frequency for a defined
thermal shock cycle in a range of amplitude of impulse excitation from
30 %
to 100 % [-]
T1,0: Temperature at which the sample exhibits a percentage change in the
height of 1,0 % compared to Tεmax [°C]
T2,0: Temperature at which the sample exhibits a percentage change in the
height of 2,0 % compared to Tεmax [°C]
εmax: Maximum expansion during RuL or CiC [%]
ε0: Expansion at time t = 0 h during CiC [%]
ε1: Expansion after 1 h dwell time at the tested temperature during CiC [%]
ε2: Expansion after 2 h dwell time at the tested temperature during CiC [%]
ε5: Expansion after 5 h dwell time at the tested temperature during CiC [%]
v1,0: Creep velocity during the first hour of dwell time at the tested temperature
during CiC [%.h-1]
ix
List of abbreviations and symbols
v2,0: Creep velocity during the first two hours of dwell time at the tested
temperature during CiC [%.h-1]
v5,0: Creep velocity during the first five hours of dwell time at the tested
temperature during CiC [%.h-1]
Tmon-tet: Transformation temperature of zirconia from monoclinic to tetragonal
[°C]
Ttet-mon: Transformation temperature of zirconia from tetragonal to monoclinic
[°C]
TTS: Temperature of thermal shock at air [°C]
χE,T: Non-linear coefficient related to the relative stiffness with the relative
voltage for impulse generation at temperature T [% of Young’s modulus /
% of voltage for impulse generation] i.e. [-]
𝜒𝜁𝑓,𝑇 : Non-linear coefficient related to the damping of the flexural resonant
frequency with the relative voltage for impulse excitation at temperature T
[% of flexural damping / % amplitude of excitation] i.e. [-]
Δζf: Difference of damping of the flexural resonant frequency [% of flexural
damping]
x
Summary
Summary
In the present study, the elastic properties of high alumina castables are assessed at room
and elevated temperature through Resonant Frequency Damping Analysis (RFDA). In
addition, several factors prone to influence the RFDA results accuracy are studied. As
Impulse Excitation Technique (IET), RFDA measurements require an impulse excitation
to generate an acoustic signal. The intensity of such an impulse influences the
measurement of the elastic properties depending on the crack density within the
microstructure of the material. The evolution of the non-linearity of the Young’s modulus
and of the damping can be correlated with damage mechanisms, namely in regard to
crack initiation and crack propagation on one side as well as crack unification and crack
widening on the other side. Furthermore, castables exhibit a heterogeneous
microstructure in terms of pore distribution due to manufacturing method. RFDA
measurements allow the determination of a volumetric measurement of the sample elastic
properties without stressing however such a material heterogeneity. Moreover, the pore
shape also influences the elastic properties of the material as well as the crack path
throughout the porous material matrix. Beyond highlighting those influencing
parameters, the present study also aims to examine the impact of zirconia addition with
regard to thermal shock resistance improvement. A first approach consists in
investigating the dense Al2O3/ZrO2 binary system to reveal the more considerable stored
elastic energy of alumina rich compositions and the enhancement of friction phenomena
with increasing zirconia content. The second approach consists in studying the influence
of ZrO2 addition into high alumina castable. Three different stabilized zirconia are tested:
CaO-ZrO2 with a main tetragonal phase and a moderate retained monoclinic phase
(~ vol-%), MgO-ZrO2 with a major non-stabilized monoclinic phase (around 80 vol-%)
and Y2O3-ZrO2 with a major cubic phase and a negligible retained monoclinic phase
(< 5 vol-%). The higher the retained monoclinic content, the more damaged the
microstructure after sintering. The nature of zirconia phase also influences the thermal
shock resistance of the castable in heating and/or cooling conditions. According whether
radial tensile stresses (during cooling phase) or radial compressive stresses (during
heating phase or during martensitic transformation during the cooling phase) are
generated at the interfaces between zirconia particles and the matrix, crack propagate
tangentially or orthogonally to the zirconia particle surface respectively. Toughening
mechanism in terms of phase transformation induced crack closure can take place in the
CaO-ZrO2 based castable formulation, wherein tetragonal phase can transform into
monoclinic. Finally, high temperature RFDA (HT-RFDA) measurements are correlated
with high temperature RuL measurements to reveal the influence of the viscous phases
and their viscosity on crack healing mechanisms and creep behaviour. Both tests reveal
1
Summary
2
Zusammenfassung
Zusammenfassung
3
Zusammenfassung
4
Introduction and motivation
The present survey focusses first on the development of stable interfaces in refractory
castables subjected to thermal, mechanical and thermo-mechanical stresses through the
addition of zirconia into high alumina refractory castable formulation. Toughening
mechanisms are expected to take place through the incorporation of such a raw material
[SCH04] [HAN00] [SAL17]. More precisely, phase transformation induced crack closure
constitutes the microstructural mechanism that is aimed to be achieved in order to
enhance the fracture toughness of the material [CAR07] [RIT88]. Such mechanisms are
based on the volume expansion of zirconia during martensitic transformation from the
tetragonal stabilized phase to the monoclinic phase. Thus, when a crack is propagating
through the material microstructure, the propagation of this crack is impeded by such
local particles volume expansion.
However, those toughening mechanisms depend on the nature of the zirconia phase and
subsequently on the choice of the raw material. Therefore three different zirconia are
tested in the present work as substitute aggregates in a tabular alumina based castable
formulation: a partially stabilized zirconia doped with CaO exhibiting a moderate
retained monoclinic content and tetragonal phase as main stabilized phase, a zirconia
doped with MgO exhibiting monoclinic phase as main phase with some cubic phase, and
a fully stabilized zirconia doped with Y2O3 exhibiting a negligible retained monoclinic
content and cubic phase as main stabilized phase.
5
Introduction and motivation
Damage mechanisms are first identified in case of thermal shock in cooling conditions in
air according to DIN EN 993-11 and in heating conditions by performing quenching tests
in molten aluminum. Those tests are carried out to distinguish the microstructural damage
induced when radial tensile stresses are generated at the grain boundary of aggregates
from those induced when radial compressive stresses are generated. Those stresses are
either generated by the transformation of the aggregate with ensuing volume expansion
like in case of non-stabilized zirconia or by the difference of the thermal expansion
behaviour between the aggregate and the matrix, also called coefficient mismatch
[TES06]. As refractory materials are subjected to cycling thermal shocks during
industrial application, high temperature thermal shocks (HTTS) are furthermore
performed between two elevated temperatures to examine the kinetics of damage
mechanisms in real cycling conditions as well as to study the impact of the repeated
transformation of zirconia on the resulting elastic properties of the castables.
The assessment and the characterization of the thermal shock resistance of castables
constitute the second main research focus of the present study. If the examination of the
microstructure subsequent to progressive thermal shock experiments is an essential
experimental procedure to identify damage mechanisms in terms of crack initiation, crack
propagation, crack unification and crack widening, non-destructive technique for material
characterization is required to follow the evolution of the material properties after each
single quenching test.
Among the commonly used technique of measurements of material elastic properties,
namely Sound Transmission (ST) [NON98], Ultrasonic Pulse Echography (UPE)[BRI08]
[HUG02] or Resonant Beam technique (RBT) [PRA04], the Impulse Excitation
Technique (IET) [MOR06] [RAT96] [TOG10] [ROE97] has the further advantage, in
addition to determine the Young’s modulus of the material, to provide with damping
values that are related with the enhancement of friction phenomena within the material
microstructure as soon as new crack flanks are generated. In the present study, Resonant
Frequency Damping Analysis (RFDA) based on IET measuring principle [IMC12] is
performed on the tested castables in order to correlate the evolution of the Young’s
modulus with the evolution of the damping and with the evolution of the microstructure.
If in theory the formation of cracks within the castable microstructure generated by
progressive thermal shocks causes the degradation of the material stiffness as well as the
enhancement of friction between crack flanks and subsequently the enhancement of
damping, such a correlation appears to be challenging. If RFDA measurements first
reveal probative results in case of micro-damage assessment in dense ceramic materials
[COP13] [ABE02], the implementation of such a measurement technique to porous and
heterogeneous materials like refractory castables faces however challenges in terms of
interpretation of the monitored results.
6
Introduction and motivation
Indeed, the present study shows the non-linear behaviour of the acoustic signal in
function of the heterogeneity of the material in terms of pore distribution and pore shape.
Furthermore, as RFDA requires a mechanical impulse in order to generate the acoustic
signal, discrepancies of the monitored resonant frequency are also observed for different
impulse excitation intensities. If the determination of the elastic properties of ceramic
materials through RFDA measurements is defined according to ASTM C 1548-02 or DIN
EN ISO 12680-1, the intensity of the impulse excitation as non-explicitly defined
experimental factor in the standard is a source of non-linearity depending on the damage
of the material. This shows in particular the precautions that need to be taken when
inter-laboratory elastic properties results are discussed. Those parameters are examined in
a first part of the present survey in order to get a critical and constructive look on the
assessment of the thermal shock resistance of porous and heterogeneous castables
through RFDA measurements. Furthermore, the interpretation of the damping behaviour
of castables subjected to severe thermal gradients is rendered challenging when the crack
density within the material microstructure is critical. In such a case, the energy delivered
during thermal shock causes in addition to the formation of new crack flanks, the
widening of pre-existent cracks. This survey shows that the evolution of the damping
behaviour of a refractory castable should be interpreted in the sense that further crack
propagation phenomena involving an increase of friction compete with crack widening
resulting in a decrease of friction.
RFDA measurements are also performed in the present survey at elevated temperatures in
order to better understand the elastic behavior of the tested castables at application
temperatures. Such measurements also aim to reveal the consequences in terms of
stiffness evolution when the martensitic transformation of zirconia takes place. Cycling
RFDA measurements at elevated temperature are also carried out as fatigue tests to reveal
the impact of the repeated zirconia transformation on a long-term stiffness evolution.
Such measurements enable to isolate the influence of the thermal stresses on the material
properties. Mechanical stresses are then added to these thermal stresses by performing
cycling Refractoriness under Load (RuL) measurement between same temperatures. The
comparison of HT-RFDA and HT-RuL experimental results reveals the contribution of
the mechanical stresses to the hysteretic behaviour of the zirconia transformation
temperature that keeps on being shifted toward higher temperatures during fatigue tests.
Furthermore, both experiments also show the influence of the formation of viscous
phases at elevated temperatures on the creep behaviour as well as on crack healing
mechanisms.
7
Introduction and motivation
zirconia addition to castable formulation on the stiffness and damping behaviors requires
the examination of the elastic behavior of dense ceramic materials from the Al2O3/ZrO2
binary system. Such a study aims to understand the interaction between zirconia and
alumina particles in terms of friction by preventing further effect induced by the porosity.
This also aims to examine the crack path after thermal shocks with increasing zirconia
content in order to correlate the resulting damping behavior with the configuration of the
crack.
The objective of the present study is to examine beneficial and detrimental effects of
stabilized zirconia addition on the thermo-mechanical and thermo-elastic behaviour of
high alumina castables. The study focusses more particularly on the thermal shock
resistance of those castables. As refractory materials are subjected to severe repeated
thermal gradients during application, this survey proposes to study the evolution of the
interfaces between zirconia particles and the matrix after progressive heating thermal
shocks, cooling thermal shocks or a combination thereof. Therefore, three zirconia
exhibiting monoclinic, tetragonal or cubic as main phase and different retained
monoclinic content are tested.
The present study has also the objective to reveal the challenging implementation of
RFDA measuring technique for assessing the thermal shock resistance of such
heterogeneous castables by showing the importance of influencing factors such as
structural material properties like porosity, and experimental parameters such as intensity
of impulse excitation on the resulting elastic properties values. RFDA measurements also
aim to reveal the effects of the martensitic transformation of zirconia on the resulting
material stiffness at elevated temperature in cycling conditions and to reveal the influence
of the zirconia raw material and more particularly of the doping agent on the formation of
viscous phases considered as the driving force of creep and microstructure healing
mechanisms.
Therefore, the elastic behaviour of dense ceramic materials of the Al2O3/ZrO2 binary
system after sintering and after thermal shocks is preliminary examined in order to better
understand the friction phenomena occurring at the interfaces between alumina and
zirconia particles. Then a further part of the present work is dedicated to the measurement
technique based on RFDA and more particularly to the parameters influencing the
monitoring of the elastic properties, namely the heterogeneous pore distribution within
the material, the pore shape and the intensity of the impulse excitation. This part of the
study is crucial for the knowledge transfer from the dense ceramic materials of the
Al2O3/ZrO2 binary system to porous high alumina castables. Finally, the thermal shock
resistance of tabular alumina based castables with addition of partially or fully stabilized
zirconia is examined in heating conditions through quenching test in molten aluminum, in
8
Introduction and motivation
cooling conditions through quenching tests in air or in both conditions through High
Temperature Thermal Shock (HTTS) experiments. The thermal properties of the tested
castables are completed by performing fatigue tests through cycling HT-RFDA and
HT-RuL experiments, in order to examine the influence of the doping agent contained in
the zirconia raw material on the creep behaviour and on the damage healing mechanisms
occurring at elevated temperature.
9
State of the scientific and technical knowledge
10
State of the scientific and technical knowledge
in spite of their low quantity, considerably influence the mixture performances during
processing in terms of rheological and flowability properties of the slurry as well as the
setting time and setting duration or the enhancement of the mechanical strength of green
bodies [OLI09] [AUT13] [OLI03] [LTI11].
Chemical bonded
Refractory castables
(Dense or heat insulating)
Regular Cement
Hydraulic bonded
Castable
Medium Cement
Castable
Low Cement
Castable
Deflocculated
castables
Ultra Low Cement
Castable
No Cement
Castable
Deflocculated castables are subdivided in four different classes according to their CaO
content as documented on table 2.1. Even if Medium Cement Castable continue to retain
a significant market share, the trend is to decrease the cement content and / or the lime
content in the refractory castable formulation, responsible for the water demand and the
ensuing apparent porosity generation and the degradation of the mechanical properties
after sintering process. For high temperature applications, the use of Low Cement
Castable (LCC), Ultra Low Cement Castable (ULCC) or No Cement Castable (NCC)
aims to shift the temperature of phase transformation with low melting point to a higher
temperature range.
11
State of the scientific and technical knowledge
Table 2.1: Distinction of the deflocculated castables according to their CaO content
[ROU01]
CaO content [wt-%]
Type
Min. Max.
Medium-Cement-Castable
> 2,5 -
(Type MCC)
Low-Cement-Castable
> 1,0 ≤ 2,5
(Type LCC)
Ultra-Low-Cement Castable
> 0,2 ≤ 1,0
(Type ULCC)
No-Cement-Castable
0 ≤ 0,2
(Type NCC)
Besides the admixtures, the particle arrangement influences the flow and rheological
properties of deflocculated castables. According to the requested rheological properties
for application, processing technology of castable formulation is different as illustrated
on figure 2.2.
Vibra flow mixtures are obtained with a sufficient water content to fill the voids between
the aggregates, which still remain in contact. In that case, an external energy is needed
for the flowability of the mixture.
Free flow mixtures are obtained with an addition of more water and/or more fine
particles, so that aggregates are not in contact anymore in the slurry. No external energy
is needed for the flowability of the mixture.
Control Particle Size Distribution mixtures are designed with a high content of fillers that
occupy the voids between the aggregates. The rheological properties of such mixtures are
controlled by the addition of dispersive agents at reduced water addition.
Fig. 2.2: Illustration of the different processing technology of castable formulations and
their ensuing rheological and flow properties [ELK]
12
State of the scientific and technical knowledge
In addition to the decrease of the cement content in refractory castable formulations and
the effectiveness of dispersive agents, the use of fine and ultra-fine with an ensuing
higher specific surface area results in optimizing the water demand of the slurry. This
aims to the prevention of the degradation of the mechanical strength of the refractory
material in the temperature range containing the transition between hydraulic and ceramic
bonding system shown in table 2.2. Thus, improved bulk density as well as a more
suitable thermal shock resistance are achieved [SCH04].
Table 2.2: Type of bonding system and their temperature range of transition in unshaped
materials according to Gelsdorf and Schellberg [SAL07]
Type of bonding Temperature range State after
system Start approx. [°C] End approx. [°C] consolidation
Ceramic 1000 1200-1500 -
Hydraulic 20 500-600 over 1000°C ceramic
Organic-chemical 50 250 over 1000°C ceramic
Inorganic- over 1000°C also
200-250 1000-1450
chemical ceramic
Inorganic-organic- over 1000°C also
50 1000-1450
chemical ceramic
TA CAM
M
P
Fig. 2.3: Microstructure overview of a Low Cement refractory Castable based on tabular
alumina with addition of partially stabilized zirconia (doped with 14,28 mol-% CaO)
after sintering at 1500°C/6h (TA: Tabular Alumina aggregate, Z: Partially Stabilized
Zirconia aggregate, CAM: Calcium Aluminate containing bonding phase and P:Pore)
13
State of the scientific and technical knowledge
Figure 2.3 reveals the heterogeneous microstructure of a dense high alumina refractory
castable type LCC with addition of partially stabilized zirconia after sintering at 1500°C
for 6 hours. The large particle size distribution from aggregates to matrix and the
presence of pores due to the processing are noticeable on this picture.
2.2.1.0 Introduction
Aluminum oxide (Al2O3) can appear in different crystallographic forms. The technical
relevant crystallographic forms are α and γ [TIE94]. The α-phase commonly called
corundum is the only thermodynamically stable phase of alumina. The lattice structure of
this phase is trigonal-rhomboedral with a theoretical density of 3,99 g.cm-3 and a
theoretical Young’s modulus of 400 ± 20 GPa [RIC05] [SAL07]. Corundum exhibits a
high melting point at 2050°C and a coefficient of thermal expansion of 8,310-6 K-1.
Transition aluminas, such as γ-Al2O3 or ε-Al2O3, are generated during the Bayer process
after dehydration of aluminum hydroxide. Those crystal structures are reactive due to
their thermodynamical instability, except for β-Al2O3 stabilized by the combination with
Na2O from the Bayer process and crystallized under the following form: Na2O·11Al2O3
and hence does not correspond to a pure alumina phase. Thus 99,5 wt-% of the alumina
14
State of the scientific and technical knowledge
that constitutes the high alumina castable formulations is pure alumina mainly in the
structure of α-corundum and the subsidiary 0,5 wt-% are composed of β-Al2O3 [CAR07].
The temperature ranges of the thermal transformation sequence of the aluminum
hydroxide are illustrated on figure 2.4.
The term reactive alumina relates to pulverulent and partly metastable Al2O3 such as
γ-Al2O3 with a primary crystal size between 0,3-1,5 µm and a subsequent high specific
15
State of the scientific and technical knowledge
surface area between 2 and 10 m².g-1 inducing a considerable surface energy and a
considerable driving force during the sintering process [POH98]. Reactive alumina
exhibits a density of about 98 % of the theoretical density of alumina and the presence of
the previous mentioned impurities causes the lowering of this density [KOL04].
In the study hereby, reactive alumina provided by Alteo with the following specification
PFR is incorporated in all studied formulations. This reactive alumina is characterized by
a specific surface area of 6,4 m².g-1, a d50 of 0,5 µm and a d90 of 1,8 µm [ALT].
Cements are hydraulic binders, i.e the addition of water is required for the setting, while
calcium aluminate hydrate phases are forming [WÖH09]. For applications at elevated
temperatures, cements with a low SiO2 content are necessary to prevent from the
formation of viscous phases. For this reason, high alumina containing cements have been
produced since the 1920’s for possible refractory applications. Contrary to Portland
cement containing 19 to 23 wt-% SiO2 for civil engineering application, the cement that
composes the deflocculated castable formulations is constituted of a low silica content.
In the study hereby, cement with the following specification Secar 71 provided by
Kerneos is incorporated in all formulations. This raw material contains up to 0,8 wt-%
SiO2 and exhibits a density between 1,04 and 1,23 g.cm-3 [KER06]. Besides this main
impurity, Secar 71 also contains up to 0,4 wt-% Fe2O3 according to supplier information.
16
State of the scientific and technical knowledge
Fig. 2.5: CaO-SiO2-Al2O3 phase diagram with typical compositions of Portland cement
(PZ for Portland Zement [German translation]) and Calcium aluminate cement (TZ for
Tonerde Zement [German translation]) [MTDATA10]
Fig. 2.6: CaO-SiO2-Al2O3 phase diagram at 1800K (1527°C). The presence of silica
increases the viscous phase content [MTDATA10]
17
State of the scientific and technical knowledge
Nevertheless, as shown on figures 2.5 and 2.6, formation of viscous phase rich in CaO
can occur at 1500°C corresponding to the sintering temperature of the castable in the
study hereby and to a possible application temperature. This viscous phase formation
temperature can be moreover lowered by the presence of further impurities in the
castable. Thus, this phase formation can appear at relative low temperature (at
approximatively 1100°C) according to the nature and the content of the impurities.
For high alumina castable formulations, calcium aluminate cement with a high alumina
content (approximatively 70 wt-%) is commonly used in order to guarantee the
refractoriness of the final product. In contact with water during processing, cement first
reacts to form calcium aluminate hydrate phases as illustrated on figure 2.7. This step is
called hydration and is characterized by a viscosity increase of the slurry because of the
dissolution of the calcium aluminate phases and the nucleation of calcium aluminate
hydrate phases. The water demanded for castable elaboration is partly used for the
hydration and also to ensure the workability of the slurry, i.e during the casting under
vibration.
Unhydrated Tabular
Cement Alumina
Temperature [°C]
Fig. 2.7: Phase transformation in a high alumina castable according to the temperature
[GIV75]
For a suitable clarity of the illustration of cement phase evolution with temperature, the
following nomenclature has been taken into account: C for CaO, H for H2O and A for
Al2O3. Following to the hydration at room temperature, the dehydration of the calcium
aluminate hydrate phases and aluminum hydrate phases takes place in a temperature
18
State of the scientific and technical knowledge
range from 100°C to around 400°C. From 1000°C, the sintering process begins with the
formation of calcium di-aluminate (grossite) and calcium hexa-aluminate (hibonite).
Calcium aluminate phases, which constitute the refractory cement, in the form of CA,
CA2 (grossite) and C12A7 (mayenite) react with mixing water to first form metastable
CAH10 and alumina gel AH3 at temperature under 20°C [CAR04].
Monocalcium aluminate (CA) with a relative low melting point (around 1600°C) is the
component of Calcium Aluminate Cements that speeds up the hardening after the initial
setting due to the high CaO content. This component represents from 40 to 70 wt-% of
the cement composition.
Calcium dialuminate (CA2) exhibits a highest refractoriness than the previous phase but
requires a longer setting time. Therefore, such a calcium aluminate phase does not exceed
a quarter of the cement composition. Nevertheless, calcium dialuminate plays the role of
catalyst for monocalcium aluminate hydration. Setting time of calcium aluminate cement
can therefore be designed according to the CA and CA2 contents.
As shown on figure 2.7, besides the calcium aluminate phases, alumina phase hydration
also occurs after mixing with water in case of calcium aluminate cement with addition of
corundum in order to enhance the refractoriness and to improve the mechanical strength
of the castable formulation.
Figure 2.8 shows the influence of the Al2O3 / CaO ratio on the refractoriness and the
melting point of calcium aluminate phases.
19
State of the scientific and technical knowledge
Fig. 2.8: Al2O3-CaO binary diagram with intermediary calcium aluminate phases C3A,
C12A7, CA, CA2 and CA6 [LEA56]
20
State of the scientific and technical knowledge
Such calcium aluminate hydrate phases contribute as hydraulic binders to the mechanical
performance of the material up to the dehydration and the ceramic bonding formation at
elevated temperature.
During the dehydration, crystallized water releases the crystal lattice of calcium
aluminate hydrates in form of water vapor that releases the structure of the castable
throughout the apparent porosity. The dehydration kinetics must be controlled to avoid
any microstructural or structural damage in case of high heating rate [SCH04].
Each hydrate breaks up at different temperature and such a material characteristics
constitutes the basis to determine qualitatively and quantitatively the nature of the
calcium aluminate hydrate phases according to curing conditions through DTA-TG
experiments [CAR04]. Table 2.3 shows the temperature ranges corresponding to the
dehydration of the main calcium aluminate hydrate phases and alumina hydrate phases.
Table 2.3: Temperature range of calcium aluminate hydrate phases and alumina hydrate
phases dehydration [CAR04]
Nature of the hydrate phase commonly
Temperature range of dehydration
present in a calcium aluminate cement
[°C]
mixture
AH3 (amorphous gibbsite) 100
AH3 (crystallized gibbsite) 210 - 300
CAH10 120
C2AH8 170 - 195
C3AH6 240 - 370
The dehydration taking place in refractory castables during the heating leads to a
decrease of the mechanical strength in a temperature range from 350°C to 1000°C, while
CA phase reacts with alumina to form CA2 [MYH94]. The sintering process continues
with the reaction between alumina and CA2 to form CA6 from 1200°C up to the sintering
temperature equal to 1500°C within the framework of this work [LEE98].
Calcium hexa-aluminate (CA6) corresponds to the only non-hydrating phase with the
highest refractoriness in the pure calcium aluminate system as this hydrate exhibits an
incongruent melting point from 1833°C. This phase presents a theoretical density of
3,79 g.cm-3 [UTS88] and a theoretical Young’s modulus of 260 ± 10 GPa [CRI88].
21
State of the scientific and technical knowledge
Zirconium dioxide also called zirconia is usually added to refractory formulations based
on alumina with purpose of thermal shock improvement because of the high strength, the
high fracture toughness, the thermal shock resistance, the corrosion resistance [HUE14]
as well as the more considerable thermal conductivity compared to alumina [CAR08].
Zirconium dioxide can crystallize in three different lattice structures: monoclinic,
tetragonal or cubic as illustrated on figure 2.9.
At room temperature, the stable lattice structure of pure zirconia is monoclinic (mon)
with the following mineralogical specification baddeleyite. At 1170°C, zirconium dioxide
transforms into the tetragonal crystalline structure (tet). This transformation is
martensitic: i.e the transformation is polymorphic, diffusionless, and spontaneous,
characterized by a rearrangement of the crystal lattice resulting in a deformation of the
unit cell [GOT14]. At 2370°C, tetragonal zirconium dioxide transforms into cubic (cub)
[KOL04]. A high melting point of 2710°C characterizes zirconium dioxide [CAR08].
According to the doping agent, zirconium dioxide exhibits a Young’s modulus between
240 and 320 GPa and a high density that is notably dependent on the crystalline structure:
5,56 g.cm-3 (mon), 6,10 g.cm-3 (tet) and 5,80 g.cm-3 (cub) [CAR08]. Like the modulus of
elasticity, the coefficient of thermal expansion of zirconium dioxide also depends on the
lattice structure. The monoclinic crystalline structure reveals an anisotropic behavior with
regard to the expansion with a coefficient of thermal expansion between 6,810-6 K-1 and
1410-6 K-1 according to the three dimensional crystalline direction. The tetragonal and
the cubic forms reveal a coefficient of expansion depending on the nature of the doping
agent: between 8,010-6 K-1 and 1010-6 K-1 for the tetragonal structure and between
10,010-6 K-1 and 10,810-6 K- 1 for the cubic form [SAL07].
A further characteristic of zirconium dioxide is the discrepancy with regard to the
martensitic transformation temperature during the heating and the cooling phases. In spite
of the reversible character, the transformation of the tetragonal crystalline structure to the
22
State of the scientific and technical knowledge
monoclinic structure during the cooling down takes place at 950°C for a pure zirconium
dioxide. This observation of the thermal behavior of zirconium dioxide explains the
hysteresis shown on figure 2.10. It is worth mentioning that the presence of doping agent
results in a shift of the temperature range of transformation to lower temperature and a
widening of this temperature range. Above a concentration threshold in doping agent,
depending on the nature of this doping agent, zirconium dioxide is fully stabilized in the
cubic crystalline structure and no transformation is to be noticed. In addition to the nature
of the doping agent, the particle size as well as the environment of the zirconia particle
can influence the widening of the hysteresis. As far as pure zirconium dioxide is
concerned, nucleation of monoclinic particles may be delayed during the cooling down
and induce this hysteretic behavior [SAL07].
The shift of the transformation temperature also occurs during the heating phase. Thus
the monoclinic to tetragonal transformation takes place between 1060°C and 1070°C for
a zirconium dioxide doped with 2 mol-% up to 4 mol-% CaO [DUR87].
As illustrated on the dilatometric curve, the linear shrinkage during the heating phase and
the linear expansion occurring during the cooling phase induce a volume change of
zirconium dioxide: a first volume shrinkage during the heating phase and the
23
State of the scientific and technical knowledge
The single cubic phase of FSZ exhibiting no phase transformation from room
temperature to the melting point is obtained by addition of 16 mol-% CaO, 16 mol-%
MgO or 8 mol-% Y2O3 [CAR08].
Within the framework of this study, different stabilized zirconia aggregates are
incorporated into a tabular alumina based castable formulation in order to examine the
influence of the stabilization on the mechanical and elastic properties of the castables.
The first aggregate is composed of 14,28 mol-% CaO, the second one of 10,15 mol-%
MgO and the last one is constituted of 5,73 mol-% Y2O3. The chemical analysis of each
aggregate is examined in chapter 3.1.2.3 (see table 3.9). The binary phase diagrams of the
combination of each doping agent and zirconium dioxide are illustrated in the following
figures 2.11, 2.12 and 2.13.
24
State of the scientific and technical knowledge
The ZrO2-CaO binary phase diagram shown on figure 2.11 reveals the stability of the
tetragonal crystalline structure under temperature of 2400°C for concentration in CaO
below 6 mol-%. After a further cooling down of such compositions, monoclinic and
tetragonal crystalline structures can co-exist in a relative reduced temperature range. A
further cooling down will induce the transformation from tetragonal to monoclinic and
the formation of Ca6Zr19O44 (see F2 on fig. 2.11). For higher calcium oxide
concentrations, from 6 mol-% to 17 mol-%, cubic and tetragonal structures are stable
above 1140°C. The quenching of such compositions leads to the transformation from
tetragonal to monoclinic particles embedded in a cubic matrix [HEI10] [SER04]. At room
temperature, segregated monoclinic crystalline phases and Ca6Zr19O44 are formed. In the
study hereby, the chemical analysis shown on table 3.9 detailed in chapter 3.1.2.3 reveals
a CaO quantity of 14,28 mol-% in the studied Ca-ZrO2 (zirconium dioxide doped with
CaO). According to the manufacturing parameters (cooling rate, temperature of post-cure
heat treatment, dwell time) each crystalline structure of zirconia can be achieved.
Therefore, X-Ray Diffraction analysis is essential for such a product to estimate both
stable zirconium dioxide phase and retained monoclinic phase content to study their
influence on the mechanical and elastic properties of high performance castable after
thermal shock tests.
25
State of the scientific and technical knowledge
The binary ZrO2-MgO phase diagram shown on figure 2.12 reveals the first stable solid
solution at temperatures above 1400°C. A solid solution with a cubic crystalline structure
can be formed at this temperature with the following eutectic composition: 87 mol-%
ZrO2 and 13 mol-% MgO [SER04]. In order to stabilize a Partially Stabilized Zirconia
with MgO as doping agent (Mg-PSZ), the optimal MgO content should be between 8 and
10 mol-%. In this MgO content range and through a quenching process up to 1400°C
with a cooling rate between 500 and 1000 K.h-1 followed by a post-cure heat treatment at
this temperature of 1400°C, cubic phase of zirconium dioxide with clusters of tetragonal
phase can be achieved [SAL07]. In the present study, the chemical analysis shown on
table 3.9 detailed in chapter 3.1.2.3 reveals a quantity of MgO slightly higher than
10 mol-% (10,15 mol-%) in the studied Mg-ZrO2 (zirconium dioxide doped with MgO).
According to this chemical composition and by assuming similar processing parameters,
the cubic crystalline structure might be obtained with a non-negligible concentration of
tetragonal particles.
26
State of the scientific and technical knowledge
In the following study, two different stabilized zirconia doped with yttria are studied.
The chemical analysis shown on table 3.10 detailed in chapter 3.1.2.3 referring to the
zirconia added to the high alumina refractory castable composition reveals a quantity of
Y2O3 of 5,73 mol-% in the studied Y-ZrO2. According to the binary phase diagram
shown on figure 2.13, such a quantity of yttrium dioxide, slightly superior to the
peritectoid composition (around 4,75 mol-% of Y2O3) induces a stabilization at room
temperature of both cubic and tetragonal crystalline structures with a residual and minor
monoclinic phase content. This last mentioned crystalline structure constitutes the critical
point with regard to zirconium dioxide application in refractory castables, as this typical
crystalline structure transforms into the tetragonal structure during the heating phase with
a reversible transformation during the cooling down with an ensuing and considerable
volume expansion that can generate cracks in the microstructure of the refractory
castable. According to the nature and the toughness of the matrix surrounding such
zirconium dioxide particles, the generated crack density may be so considerable that the
addition of zirconium dioxide has more drawbacks than benefits. Therefore, in the study
hereby, a phase analysis of each tested zirconium dioxide through X-Ray Diffraction is
performed to estimate this monoclinic crystalline structure content.
27
State of the scientific and technical knowledge
The X-Ray Fluorescence analysis performed on the yttria stabilized zirconia used for the
elaboration of dense ceramic through slip casting reveals the presence of 97,14 mol-% of
ZrO2 for a quantity of 2,86 mol-% of Y2O3 as shown on table 3.2 in chapter 3.1.1.2.
According to the binary phase diagram shown on figure 2.13, such a quantity of yttrium
dioxide induces a stabilization at room temperature of the tetragonal crystalline structure
with a residual and non-negligible monoclinic phase content. A more considerable
retained monoclinic content is expected in this raw material in comparison with Y-ZrO2
incorporated into the castable formulation.
The fracture mechanics describes the nucleation and the propagation of cracks within a
material inducing a material failure due to the ensuing rupture [GRO11]. The crack
propagation is generated by a driving force, which permanently faces the material
microstructure resistance [RIT88].
As far as polycrystalline materials are concerned such as all tabular alumina grain
fractions studied in this work, cracks are able to propagate either at the grain boundaries,
such cracks are intragranular, or throughout the grains, such cracks are transgranular as
illustrated on figure 2.14.
In case of materials with weak bonds between primary crystals that form the
polycrystalline grain, cracks are also able to propagate along phase boundaries
throughout the polycrystalline grain to form transgranular cracks [GRO11].
Fig. 2.14: Illustration of transgranular crack propagation (left) and intragranular crack
propagation (right) [GRO11]
By assuming a specimen as linear elastic, the crack opening can be described according
to the nature of the stress applied on the specimen as shown on figure 2.15. Concerning
the first mode of solicitation, the stress is perpendicularly applied to the crack
28
State of the scientific and technical knowledge
propagation direction. In case of mode II, the stress is parallely applied to the crack
propagation direction and for mode III, the stress is tangentially applied to the crack
propagation direction [GRO11].
Taking into consideration the non-linear effects taking place at the crack tip,
corresponding to a macroscopic negligible region of the damaging area, the hypothesis of
a linear elastic specimen is not appropriate anymore. This assumption is however
sufficient as far as brittle materials are concerned.
Fig. 2.15: Nature of crack opening according to the stress applied on the specimen
[GRO11]
The approach of a linear elastic behavior of a specimen can be considered with the help
of the energy concept defined by Griffith [CAR07] or with the help of the stress intensity
concept defined by Irwin [GRO11].
Fig. 2.16: Energy concept according to Griffith – Griffith crack with a length of 2c
[CAR07]
29
State of the scientific and technical knowledge
According to Griffith, formation of new crack surfaces at wake region of the damaging
area results in a decrease of the stored elastic energy [CAR07].
Starting from a crack with a characteristic crack length of 2c, as illustrated on figure 2.16,
the surface energy of both crack edges also called crack flanks is defined as follows:
Considering that during crack flanks formation, the elastic energy Uel is equal to or
higher than the surface energy UOF, this energy concept can be defined as a function of
external stresses and material stiffness.
2 𝑐2
𝜋𝜎𝑒𝑥𝑡 𝑐
𝑈𝑂𝐹 = 𝑈𝑒𝑙 = Equation 2.6
𝐸
From this theoretical standpoint, Griffith equation defines the required stress leading to
crack progress in a material.
2𝐸𝛾
𝜎𝑓 = √ Equation 2.7
𝜋𝑐 𝑐
The required stress σf for crack propagation in a material is dependent on the material
stiffness (E), the material specific surface energy (γ) and the critical crack length (c).
Irwin extended this approach of Griffith by defining stress intensity factors also called
K factors. Those factors describe the surrounding areas of the crack tip as shown on
figure 2.17 that may be subjected to plastic deformation under stresses (ex. metals and
alloys). According to the nature of crack opening parameters, KI for mode I,
30
State of the scientific and technical knowledge
KII for mode II or KIII for mode III are defined. As the stress intensity factor must satisfy
the boundary conditions of a linear elastic behavior, this approach is valid in the
delimited KI field contained in a modelled circle with radius R and with center the crack
tip and beyond the plastic area characterized by a plastic deformation. This plastic area is
represented on figure 2.17 by a circle with radius rp and center the crack tip [GRO11].
Fig. 2.17: Irwin concept for crack opening of the first mode [GRO11]
If by hypothesis KI field is clearly bigger than the region defined by rp radius, the
surrounding area of the crack tip can be described by the means of KI. If this stress
intensity factor achieves a critical value, then the crack can progressively propagate
[GRO11].
The critical stress intensity factor also called fracture toughness is defined as follows:
𝐾𝐼𝑐
𝜎𝑓 = Equation 2.8
√𝜋𝑐𝑐
Concerning refractory materials with brittle elastic failure behavior and characterized by
no plastic deformation at the crack tip during crack propagation, the last mentioned
equation defines with a suitable accuracy the fracture toughness of the material. If the
surrounding area of the crack tip exhibits a non-linear behavior, this concept turns out to
be erroneous [SCH04].
Griffith approach of fracture mechanics is a part of a strategy to optimize the material
fracture energy. The fundamental idea to enhance the material fracture energy is to
prevent the material failure through the nucleation of cracks that further propagate
straight on through the matrix. The bypass of the crack from its initial and quasi straight
on path is energy consuming and results in a delay in the material fracture. Three
different concepts can be distinguished.
31
State of the scientific and technical knowledge
The first concept also called Irwin concept (K concept) focusses on the minimization of
the local stresses situated at the level of the crack tip. Therefore, several options are
available:
KIc: the maximization of the critical stress intensity factor by increasing the
curvature radii at the level of the crack tip or by taking into account the thermal
behavior (coefficient of expansion) of the present phase at the level of the crack
tip. In both cases, the maximization of the critical stress intensity factor is
achieved through a material design and the possible incorporation of a second
phase revealing a thermal behavior that can defy the crack propagation. In such a
context, the addition of an allotropic phase with a characteristic volume expansion
during the transformation initiated by the stresses induced by the crack turns out
to be a reliable solution.
c: the minimization of the defect size or crack length. The optimization of the
material processing results in limiting the nucleation of pre-existent cracks.
32
State of the scientific and technical knowledge
3 MPa.m1/2 and slowly increases with the crack length to reach a constant values close to
6 MPa.m1/2 when the crack length is bigger than 2 mm, the initial fracture toughness of
Mg-PSZ material is around 6 MPa.m1/2 and increases up to constant values around
15 MPa.m1/2 after the crack reaches a length of 1 mm. In addition to the microstructural
mechanisms occurring at the crack tip leading to the fracture toughness enhancement,
such as the volume expansion during the martensitic transformation and the possible
coefficient of expansion mismatch with a second phase, some microstructural phenomena
independently on the crack length can occur in the wake process region that can explain
those considerable fracture toughness values of zirconium dioxide.
Table 2.4: Examples of strengthening mechanisms taking place at the level of crack tip
[RIT88]
Class Example of strengthening mechanism
Crack deflection
and crack
meandering
Deflection Meandering
Zone shielding
Transformation toughening Microcrack toughening
33
State of the scientific and technical knowledge
Fig. 2.18: Evolution of the fracture toughness of Al2O3 and a Partially Stabilized Zirconia
Mg-PSZ with the crack length [SAL17]
Figure 2.19 gives an overview of the possible microstructural mechanisms that can occur
in the wake region of a crack contributing to the improvement of the material fracture
toughness.
Crack branching can take place when the crack tip is in contact with a region of the
material applying compressive stresses facing the propagation of the crack. In a normal
case, the path of the crack is deviated but according to the angle of attack of the crack,
this crack can be split in two cracks progressing in two different regions. The stress field
initially localized at the primary crack tip is equally divided to form a new stress field
localized at the crack tip of both new cracks. The requested energy as driving force for
the propagation of these new cracks is higher than in the initial case where only one crack
is propagating throughout the matrix.
Pull-out effect is typical for fiber reinforced materials as explained by Stang et al.
[STA86]. The pull-out failure can be described as a partial or total debonding of a second
phase with the matrix. This debonding is in many cases facilitate by the thermal behavior
discrepancy between the second phase and the matrix. Such a microstructural observation
is possible for fiber like or needle like particle with a length much higher than the
diameter or the thickness. Pull out effect can also induce crack flank friction phenomena.
34
State of the scientific and technical knowledge
Crack bridging can be observed in presence of a viscous phase containing in most of the
case some silicates with a characteristic low melting point and a viscosity depending on
the temperature and the presence or not of alkalis or earth alkalis or other impurities like
iron oxide. Crack bridging can be achieved through a second phase in the form of
aggregate, fiber, whisker or ductile particle as explained by Alves and illustrated on
figure 2.20. In such cases the crack propagates throughout the matrix, meets the second
phase and continues the progression on the other side of the second phase without
inducing microcracks in the second phase or a debonding of this second phase.
Fig. 2.19: Microstructural phenomena occurring in the wake region of a crack leading to
the enhancement of the material fracture toughness [SAL18]
35
State of the scientific and technical knowledge
Fig. 2.20: Effective crack bridging mechanisms occurring in the wake region of the crack
[ALV14]
Crack flank friction occurs between the crack flanks in case of typical interlocking
structures. To better understand this notion of crack flank friction, damaged
microstructure after progressive thermal shocks can be considered as a tribological
system. Indeed, the induced crack separates two microstructural parts of the ceramic
specimen. Applying a load or mechanical solicitation on one of this part of the specimen
results in relative motion of this part on the surface of the opposite part. This relative
motion is responsible for friction between the areas of both parts in contact. The
importance of crack flank friction phenomena depends, as illustrated on figure 2.21, on
the nature of the two bodies in contact and more specifically on the hardness of both
parts, the surface roughness of both contact bodies, the amplitude of the mechanical
solicitation, the presence or not of material between the two contact bodies and the nature
of the environment at the interface between both bodies [SAL18].
36
State of the scientific and technical knowledge
Influence factor
Friction
Wear
In the study hereby, the tribological system can be considered as friction and wear
phenomena between aggregates of tabular alumina or zirconia and the calcium aluminate
based matrix through the induced crack network after progressive thermal shocks. In a
first part of the study, friction phenomenon of the Al2O3-ZrO2 system in dense ceramics
are studied to better understand the friction phenomenon occurring in refractory castable
formulations characterized by an apparent porosity of around 18 vol-%. The
characterization of this tribological system requires to understand the influence of
zirconia particles on the wear and friction of high alumina castable.
In his doctoral thesis, Pasaribu studied the wear and friction of zirconia and alumina
ceramics doped with CuO [PAS05]. Under similar experimental conditions, the
coefficient of friction of both experimental systems alumina / alumina, and
alumina / zirconia are determined (see figures 2.22 and 2.23).
37
State of the scientific and technical knowledge
Under similar experimental conditions, namely a sliding velocity of 0,1 m.s-1 and a
normal load of 10 N, the coefficient of friction of alumina sliding against zirconia is
much higher than the coefficient of friction of alumina doped with 1 wt-% CuO sliding
against alumina (0,7 [-] vs 0,45 [-]). However a lubricating effect is observed by adding
copper oxide in crystalline structure of alumina or zirconia. Thus, the coefficient of
friction of alumina sliding against alumina is estimated to 0,65 [-]. Therefore, an addition
of zirconia to a pure alumina system results in a light increase of the microstructural
friction phenomena. This friction enhancement can also be triggered by the grain
morphology and the roughness of the grain surface as illustrated on figure 2.24.
Fig. 2.24: Crack flank friction between partially stabilized zirconia particles
(with 14,28 wt-% CaO) with different grain shapes and the calcium aluminate cement
based matrix after three thermal shock tests at high temperature between 900°C and
1500°C
38
State of the scientific and technical knowledge
The third approach that aims to enhance the fracture energy of a material is the plasticity
model. Even if ceramic materials exhibit a linear elastic behavior, a catastrophic failure
following this linear elastic behavior and a low strain when a stress is applied on the
material, in some cases ceramic materials can reveal a non-negligible plastic deformation.
This non-linear strain to stress evolution can be caused by three different crystallographic
modifications illustrated on figure 2.26.
Initial structure
39
State of the scientific and technical knowledge
2.3.4.1 Dislocation
Dislocation can take place in ceramic materials with a typical high symmetry lattice
structure. A complete crystalline plane of the initial structure slides up to a metastable
state, where the crystallographic structure and the charge neutrality is maintained.
Crystalline structures with a high Burger’s vector required a considerable external energy
to induce such a lattice modification, and therefore such dislocations are less likely in the
present work.
Crystal twinning is a symmetrical intergrowth of crystals. In that case the lattice points of
a crystal are shared as lattice points in another crystal structure. Both crystals have a
defined lattice orientation. That constitutes a diagnostic feature for the mineral
identification. Crystal twinning can be distinguished under three different forms
according to the nature of the symmetric intergrowth: twin plane, twin axis and twin
center.
The R-curve or resistance curve of a quasi-brittle material is the curve representing the
apparent variation of fracture energy with crack extension as shown on figure 2.27. This
curve can be considered as the envelope of the curves of the energy release rate in
function of the crack extension in a specimen. The crack driving energy G must reach a
threshold G0 before the crack starts to grow. By assuming a fully developed fracture
process zone travelling within the material microstructure in a similar way, the resistance
curve of the material can increase with crack extension. As refractory ceramic materials
are heterogeneous, based on coarse grains and multiple phases, such materials do not
automatically exhibit a linear elastic behavior during the fracture. The continuous
40
State of the scientific and technical knowledge
propagation of a local plastic zone at the level of the crack tip during crack extension
explains the increase of the R-curve up to reach a defined crack driving energy Gf value
corresponding to a crack length cf. In addition to the enhancement of the elastic behavior
of refractory materials with regard to crack propagation phenomena, the design of high
performance materials also aims to achieve a beneficial non-linear behavior of the crack
resistance in function of the crack length. This design takes into consideration the
incorporation of phases leading to local plastic deformation of the microstructure. Within
the framework of this study, stabilized zirconium dioxide is added to traditional high
alumina castable formulations to study the ensuing effects of the martensitic
transformation and its resulting plastic deformation zone development at the crack tip on
the crack propagation resistance. The optimization of the microstructural behavior with
regard to fracture mechanism aims therefore to achieve rising R-curve profile up to high
crack length values through microstructural toughening.
Gf
G0
cf
Rising R-curve profiles for refractory castables are expected to be achieved through the
martensitic transformation of zirconium dioxide, as this phase transformation induces
crack closure due to the considerable volume expansion of zirconium dioxide particles
during the transformation of tetragonal crystalline phase to the monoclinic structure. The
tested zirconium dioxides of this study are doped with MgO, CaO or Y2O3. Hannink et al.
showed the importance of the nature of the doping agent on the R-curve behavior of
zirconia [HAN00]. As illustrated on figure 2.28, a rising R-curve behavior is observed for
Partially Stabilized Zirconia with MgO as doping agent. The toughness of such a material
reaches around 14 MPa.m1/2 for crack above 1 mm, while a flat R-curve behavior is
exhibited by the zirconia doped with yttria. The toughness of this material is around
6,5 MPa.m1/2 independently of the crack length.
41
State of the scientific and technical knowledge
Fig. 2.28: Typical R-curves for ceramic materials obtained from measurements of long
cracks in compact tension samples [HAN00]
The thermal shock resistance of refractory materials is a technical and high relevant
material characteristic describing the ability of a material to withstand progressive and
abrupt heating and cooling down phases. As ceramic materials are brittle, such materials
are vulnerable to failure during considerable heat transfer or abrupt temperature change
[HAS69]. During application, ceramic materials and specifically refractory materials are
subjected to heating and cooling phases. At macroscopic scale, each part of the ceramic
piece, and at microscopic scale each chemical compound of the ceramic piece has a
specific thermal behavior. From this standing point, differently induced thermal stresses
can be defined.
First order thermal stresses are defined at macroscopic scale of a ceramic piece,
according to its volume, is subjected to a thermal gradient during the quenching test. In
such conditions, even if the material can be considered as homogeneous, thermal stresses
are induced because of the non-homogeneous temperature distribution within the
specimen. Sample part characterized by a considerable discrepancy of thermal behavior
with the neighboring sample part is supposed to be firstly damaged. Following this crack
nucleation, cracks propagate as the thermal gradient is progressively shifting throughout
the sample.
Second order thermal stresses are defined at microscopic scale as ceramic pieces and
more particularly refractory pieces are heterogeneous with regard to the structure with the
presence of solid phases, amorphous phases and pores and with regard to the chemistry
42
State of the scientific and technical knowledge
with the presence of different phases. Each chemical compound reveals a specific thermal
behavior, namely a specific coefficient of thermal expansion. Even if the temperature is
assumed to be equal in the whole ceramic piece, each compound has a different behavior
inducing some stresses at the interfaces between the different phases. Therefore, cracks
tend to be nucleated at the level of the grain boundaries of heterogeneous materials and at
the phase boundaries of single phase materials [KIN55].
Fig. 2.29: Evolution of the mechanical strength of a ceramic material in function of the
temperature difference [HAS70]
43
State of the scientific and technical knowledge
1/2
𝜋𝛾(1 − 2𝜇)2 16(1 − 𝜇2 )𝑁𝑉 𝑐 3 1
∆𝑇𝑐 = [ ] × [1 + ] Equation 2.9
2𝐸𝑁𝑉 =0 𝛼 2 (1 − 𝜇2 ) 9(1 − 2𝜇) √𝑐
The unified theory of thermal shock fracture initiation and crack propagation in brittle
ceramics consists in defining the total energy (UT) per unit area as the sum of the strain
(USE) and surface (USA) energies of the cracks.
With:
UT: Total energy per unit area [J.m-2]
USE: Strain energy of the crack [J.m-2]
USA: Surface energy of the crack [J.m-2]
44
State of the scientific and technical knowledge
𝜎2 Equation 2.11
𝑈𝑆𝐸 =
2𝐸𝑁𝑉=0
By assuming:
As the strain energy of the crack is supposed to be thermo-activated and dependent on the
material coefficient of expansion, USA can be defined as follows:
𝐸 Equation 2.15
𝐸𝑁𝑉=0 =
1 + 2𝜋𝑁𝑆 𝑐²
45
State of the scientific and technical knowledge
Therefore the critical temperature difference (Tc [K]) can be defined by the following
equation:
2𝛾0
∆𝑇𝑐 = √ + (1 + 2𝜋𝑁𝑆 𝑐²)
𝛼 2 𝐸𝜋𝑐 Equation 2.18
It is worth mentioning that the equation 2.18 does not thoroughly correspond to the first
definition of the critical temperature difference given by the equation 2.9. Indeed, another
system analysis with the assumption of the nucleation of cylindrical pores with radii and
length equal to c within the microstructure of an isotropic material in terms of thermal
behavior leads to the following definition of the strain (USE) and surface (USA) energies:
With:
With:
𝐸
𝐸𝑁𝑉=0 =
16(1 − 𝜇²)𝑁𝑉 𝑐 3
1+ Equation 2.21
9(1 − 2𝜇)
Salvini et al. extended the unified theory of thermal shock fracture initiation and crack
propagation in brittle ceramics of Hasselman by considering the energy dissipation
mechanisms during crack propagation [SAL12]. The critical temperature difference is
defined according to the final dimension of the crack cf and the initial crack dimension c0.
This new approach considers the energy required for crack initiation c0 and for crack
propagation γNBT and the work of fracture γWOF, while Hasselman considered a global
46
State of the scientific and technical knowledge
effective surface energy. The implementation of the fracture surface energy ratio in the
Hasselman theory gives a more accurate definition of the critical temperature difference
for refractory materials and a more realistic quantitative evaluation of the volumetric
crack density, as mechanisms of crack interaction with material microstructure are taken
in account.
−1 3
3(𝛼∆𝑇𝑐 )²𝐸 16(1 − 𝜇²)𝑁𝑖 𝑐03 16(1 − 𝜇²)𝑁𝑓 𝑐𝑓′ 2 Equation 2.22
{[1 + ] − [1 + ]−1 } = 2𝜋𝑁𝑓 𝛾𝑊𝑂𝐹 𝑐𝑓′ − 2𝜋𝑁𝑖 𝛾𝑁𝐵𝑇
2(1 − 𝜇) 9(1 − 2𝜇) 9(1 − 2𝜇)
Fig. 2.30: Dependency of the critical temperature difference with the crack half-length
according to the number of cracks [HAS69]
47
State of the scientific and technical knowledge
Hasselman defined the minimum thermal strain required for initiation of crack
propagation and for propagation of initially short cracks as a function of crack length and
crack density as shown on figure 2.30. Regions of stability and instability can therefore
be defined according to the damaging grade of the ceramic material microstructure and
the temperature difference. It is worth mentioning that the energy required for crack
initiation for a defined crack length is higher than the energy required for the propagation
of pre-existent cracks of the same length. Furthermore, the energy required for the
propagation of cracks has to be higher when the crack density is considerable. Thus,
some research works treat of the formation of microcracks during sintering process to
increase the required energy for the propagation of those cracks. Such a microstructural
damaging should be controlled in order not to achieve critical mechanical strength
properties in accordance with the application.
By assuming the nucleation and the propagation of short cracks, the definition of the
critical temperature can be simplified as follows:
2𝛾 2𝐸𝛾 𝐾𝐼𝑐
∆𝑇𝑐 = √ = √ = = 𝑅 [K]
𝛼2 𝐸𝜋𝑐 𝛼2 𝐸²𝜋𝑐 𝛼𝐸 √𝜋𝑐
Equation 2.24
The first thermal shock parameter is indeed defined as the critical temperature difference
within the framework of thermal shock experiments respecting the main boundary
condition of a homogeneous distribution of the temperature within the ceramic specimen.
The thermal stresses induced by such a thermal shock test are not provided by a thermal
gradient but by the non-homogeneous behavior with regard to thermal expansion of each
compound of the material.
The definition of this first thermal shock parameter can be completed by implementing
the thermal conductivity λ or a stress reduction term Ø of the ceramic material in case of
constant heat transfer or constant heat transfer rate respectively during the thermal shock
procedure:
𝐾𝐼𝑐 𝐽
𝑅′ = ∙ 𝜆 [ ] Equation 2.25
𝛼𝐸 √𝜋𝑐 𝑐𝑚 𝑠
48
State of the scientific and technical knowledge
The thermal shock parameters R, R’ and R’’ reveal the crack initiation resistance of
ceramic materials under specific experimental conditions without thermal gradient during
the thermal shock tests. The role of thermal conductivity on crack stability in parameter
R’ to satisfy the boundary conditions with regard to the nature of the heat transfer is
introduced by Kingery [KIN55]. Therefore, a material exhibiting a high thermal
conductivity is required to limit the stress fields within the microstructure of ceramic
specimen due to a non-homogeneous temperature distribution. To sum up, the first
thermal shock parameter, R, gives a critical temperature difference magnitude that can be
withstood by a ceramic material. The second thermal shock parameter R’ reveals the
influence of the thermal conductivity of the ceramic material on the heat transfer within
the material during the thermal shock test and gives a first approach with regard to the
kinetics. And the third thermal shock parameter R’’ shows the maximal cooling rate
applied on the ceramic material without crack initiation [SAL07].
Even if the most significant conclusion of Hasselman theory is to reduce the probability
of fracture initiation by designing a material exhibiting suitable mechanical, thermal and
elastic properties to enhance those R, R’ and R’’ thermal shock parameters, experimental
conditions of thermal shock tests can not satisfy the boundary conditions. Indeed, in
addition to pre-existent cracks due to processing manufacture, crack nucleation is
induced by thermal stresses existing in the part of ceramic materials subjected to severe
heat transfer and thermal gradient.
By assuming the nucleation of cracks and microcracks in ceramic material microstructure
during application as an inevitable damaging process, research works with purpose of
material thermal shock improvement focus on the crack propagation resistance. In this
context, Hasselman defines the following parameters:
𝐸 1
𝑅′′′ = [ ] Equation 2.27
𝜎²(1−𝜇) 𝑃𝑎
𝐸𝛾
𝑅′′′′ = [𝑐𝑚] Equation 2.28
𝜎²(1−𝜇)
𝛾 𝐾 Equation 2.29
𝑅𝑆𝑡 = √ [ ]
𝛼²𝐸 𝑚−0,5
49
State of the scientific and technical knowledge
The definition of the thermal shock parameters relative to the crack propagation
resistance illustrates the considerable complexity of ceramic materials designing to
satisfy both crack initiation and crack propagation resistance. While a high mechanical
strength and a low stiffness are required to limit the formation of cracks, such a
combination favors the propagation of cracks [AKS03] [HAS69].
Since the introduction of the “Unified Theory of Thermal Shock Fracture Initiation and
Crack Propagation in Brittle Ceramics” of Hasselman, research is focused on the
extension of Hasselman theory [SAL12] or on the definition of new thermal shock
parameters.
Boccaccini et.al. elaborated a method to predict the service life for refractories by
correlating the residual fracture strength and the variation of the ultrasonic velocity after
progressive thermal shock cycles [BOC08]. This method is based on the definition of the
resistance to fracture initiation on refractory materials quantified by the parameter R. By
expecting no significant variation of thermal expansion and Poisson’s ratio of the
material after quenching test, the focus can be laid on the rate of σf/E ratio as the
stiffness (E) of material is supposed to be more sensitive to crack initiation in comparison
with the fracture strength (σf).
Rodrigues et al. correlated the residual fracture strength of refractory materials after
progressive thermal shocks with a new established parameter defined by interpreting the
wedge splitting curve of material [MIY14]. Not only the energy of fracture and the work
of fracture can be determined through this experiment, an innovative parameter (R x) can
be defined by implementing the maximal horizontal load, the value of the tangent on the
inflexion point in the load displacement curve during the crack propagation stage and the
maximal load displacement value as shown on figure 2.31. Such an interpretation of the
wedge splitting curve has been made possible mostly with fundamental research
performed on this measuring technique by Harmuth and Tschegg, who managed to
correlate the energy of fracture and the work of fracture with stable crack propagation
[HAR95] [HAR96].
50
State of the scientific and technical knowledge
Fig. 2.31: Typical load-displacement curve obtained from steady crack propagation in a
specimen for a fracture energy test and definition of the curve parameters implemented in
the definition of the thermal shock parameter Rx
𝐹max ∙ 𝑐𝑚𝑎𝑥
𝑅𝑥 = [𝑁 . 𝑚 ] Equation 2.30
|𝑇𝑎𝑛 (𝜃)|
Besides the thermal shock parameters, Hasselman defined the stored elastic energy as the
thermal stresses stored in the material microstructure acting as driving force for crack
propagation when energy release can take place. This energy is at disposal of the crack
and depends on the stress to strain behavior of the material [SAL18] as shown on
figure 2.32.
This stored elastic energy (Uel) is defined as follows:
Fig. 2.32: Strain to stress behavior of a ceramic material according to Hooke’s law and
illustration of the stored elastic energy [SAL18]
51
State of the scientific and technical knowledge
𝜎2 1 Equation 2.31
𝑈𝑒𝑙 = [ ]
2𝐸 𝑃𝑎
The strong correlation between the elastic response of a ceramic material with regard to
mechanical strength is also illustrated in the definition of the stored elastic energy.
According to this principle, a minimization of the stress and subsequently a minimization
of the mechanical strength in addition to a maximization of the stiffness may lead to a
reduction of the stored elastic energy and furthermore to a theoretical improvement of the
crack propagation resistance. This statement is in correlation with the definition of the
thermal shock parameters R’’’ and R’’’’.
The definition of the thermal shock parameters and the critical temperature difference of
a ceramic material takes into account the thermal behavior of the ceramic material by
systematically implementing the coefficient of thermal expansion α and sometimes the
thermal conductivity λ. The contribution of the thermal conductivity consists in providing
further information with regard to the heat transfer and the possible thermal gradient
according to the experimental conditions. In the ideal case, the coefficient of thermal
conductivity should be high with a view to ensure a steady state of the system with a
homogeneous distributed temperature within the material microstructure. The
implementation of the thermal conductivity contributes furthermore to define in a kinetics
aspect of the cooling rate threshold leading to crack nucleation when exceeded.
The coefficient of thermal expansion provides further information concerning the strain
response of the ceramic material during the thermal shock. However, and as mentioned in
chapter 2.3.6, most of the refractory material compositions are heterogeneous. This
heterogeneity is the second source of crack development if the material is subjected to
abrupt temperature difference. Indeed, each phase that composes the material is
characterized by a specific thermal behavior and a specific coefficient of expansion. Two
coefficients of expansion have to be taken into consideration: the coefficient of expansion
of the matrix αmatrix and the coefficient of expansion of the particle αparticle. The different
thermal behavior between the matrix and the particle is at the origin of a stress field at the
grain boundary of the particle causing the crack nucleation. Therefore, the design of
refractory formulations has to consider this thermal discrepancy known as coefficient
mismatch and defined as follows:
If the coefficient α has a negative value during the cooling phase, a debonding of the
particle from the matrix takes place. The radial tensile stresses at the interface
52
State of the scientific and technical knowledge
Fig. 2.33: Microstructural effects occurring at the interface between the matrix and the
particle during the cooling phase according to the discrepancy of thermal behaviour
[TES06]
53
State of the scientific and technical knowledge
can be regarded as a martensitic transformed process zone facing the crack propagation
through applied compressive stresses, as illustrated on figure 2.34.
54
State of the scientific and technical knowledge
Fig. 2.35: Stress-strain curve of zirconia doped with 10 mol-% yttria single crystals along
the <112> compression axis in the temperature range 500°C - 1400°C [BAU96]
55
State of the scientific and technical knowledge
Plasticity model
The ensuing volume expansion of the martensitic transformation of zirconia during the
cooling phase from tetragonal to monoclinic crystalline structure may induce in some
materials depending on the matrix the formation of a plastic region at the crack tip. This
plastic region migrates during crack propagation, which explains a quasi-ideal plastic
behavior of the ceramic material containing metastable tetragonal zirconia particles. This
non-linear elastic behavior can be as a matter of example noticeable on stress-strain
curves as shown on figure 2.35 obtained through compressive tests realized on zirconia
monocrystals doped with yttria [BAU96]. This plastic behavior can be designed by the
choice of the doping agent, the content of the doping agent, the zirconia particle shape
and the compactness of the shaped material.
In rarer cases, crystal twinning can also take place. That also contributes to plastic
deformation development. Nevertheless, such a microstructural phenomenon occurs more
frequently in dense nano-materials or dense materials containing over-doped zirconia
particles as shown on figure 2.36 [TAK85].
56
State of the scientific and technical knowledge
conditions. Nevertheless models are defined to determine the influence of crystalline and
amorphous phases as well as the impact of the porosity and the pore structure on the
ensuing elastic behavior of those heterogeneous materials.
Based on the same principle, the shear modulus is defined in case of proportional twist
deformation when the material is submitted to shear stresses [SAL07]:
Fig. 2.37: Tensor matrix illustrating the stress to strain behavior of a linear elastic
material according to the three space directions [HUG13]
57
State of the scientific and technical knowledge
The elastic tensor of a material can however be simplified for monocrystalline material
by implementing the crystal symmetries of the material in the trigonal structure of
alumina [AND95], the orthorhombic structure of mullite [HIL01], the cubic structure of
andalusite [RAL84] or the hexagonal structure of silicon carbide α [KAM97].
The equation 2.35 reveals the contribution of each phase that composes the refractory
material on the stiffness of the material. This equation results from Voigt’s model
[VOI10] and Reuss’ model [REU29] with a parameter J equal to 1 and -1 respectively.
1
N J
Ec vi EiJ Equation 2.35
i 1
Within the framework of Voigt’s model, the supposed boundary condition consists in a
uniform distribution of the strain (J = 1), while the Reuss’ model is based on a uniform
distribution of the stresses (J = -1).
Both Voigt’s and Reuss’ models are commonly considered as upper and lower limits of a
multi-phase materials stiffness when the effect of Poisson’s ratio can be neglected
[LIU09].
As a matter of example, for composite materials containing continuous and unidirectional
fibers, the Young’s modulus Ec of the material according to Voigt corresponds to the
material stiffness when the load is applied (when the Young’s modulus is measured
58
State of the scientific and technical knowledge
during static procedure) in a direction parallel to the fibers, while the Young’s modulus
Ec of the material according to Reuss corresponds to the material stiffness when the load
is applied in a direction perpendicular to the fibers. Therefore, both Voigt and Reuss
models depend on the material microstructure in terms of nature of phases, shape of
grains within the microstructure and preferential direction of said grains.
Hashin and Shtrikman defined another model, also called HS-Model, that gives another
estimation of the upper and lower limits of a multi-phase material when the moduli of
elasticity as well as the Poisson’s ratio of each phase is known and when elastic
properties of each phase are quite similar [HAS63]:
vinc
K Km
1 3vm
Equation 2.36
K i K m 3K m 4Gm
vm
K Ki
1 3vinc
Equation 2.37
K m K i 3K i 4Gi
vinc
G Gm
1 6K m 2Gm v m
5Gm 3K m 4Gm
Equation 2.38
Gi Gm
vm
G Gi
1 6K i 2Gi vinc
5Gi 3K i 4Gi
Equation 2.39
G m Gi
9 KG
E
3K G Equation 2.40
E
G
2(1 ) Equation 2.41
3K 2G
Equation 2.42
6K 3G
59
State of the scientific and technical knowledge
The lower and upper limits of stiffness according to Hashin and Shtrikman, 𝐸𝐻𝑆 − and
𝐸𝐻𝑆 + respectively are subsequently defined as follows [HAS63]:
9𝐾 + ∙ 𝐺 + Equation 2.43
𝐸𝐻𝑆 + =
3𝐾 + + 𝐺 +
9𝐾 − ∙ 𝐺 − Equation 2.44
𝐸𝐻𝑆 − =
3𝐾 − + 𝐺 −
Hashin and Shtrikman’s analytical model HS- is more suitable to estimate the Young’s
modulus of a heterogeneous material with perfect bonding between the matrix and the
inclusions. In that case, both matrix and inclusions exhibit a similar thermal behavior
with regard to the thermal expansion as described in equation 2.32 (α close to 0). Model
materials constituted of a matrix with a lower coefficient of expansion than that of the
inclusion (α < 0) reveal a stiffness fitting Reuss analytical model, while on the contrary
a material composed of a matrix with a higher coefficient of expansion than that of the
inclusion (α > 0) shows stiffness values much lower than the predicted values of each
analytical model [TES06].
60
State of the scientific and technical knowledge
2.4.3 Influence of the porous areas and the pore configuration on the stiffness
The porosity has an influence on the main material properties, namely thermal and
electrical conductivities or elastic properties. First analytical approaches to define the
porosity dependency of the material stiffness used to be developed from empirical
models treating of the influence of the porosity on the thermal and electrical
conductivities, namely the Maxwell model developed in 1873 [MAX73]. Since then,
numerous analytical approaches are proposed taking into account the nature of the
porosity (apparent or closed), the pore size distribution and the pore shape. The
stiffness-porosity relationship in refractory castable turns out to be challenging to
understand and to model because of the complexity of this material property. Indeed,
independently of the nature of the manufacturing process, refractory materials exhibit a
pore network composed of a random distribution of pores of different size. Those pores
can be located within the matrix to form the apparent pores network or within the
aggregates without contact with the matrix to form a closed porosity. It could be of
interest to analyse the influence of each parameter (pore distribution, pore shape and
nature of porosity) on the resulting elastic properties of such heterogeneous materials.
Zivcova et al. examined the relevance of the analytical approaches defined previously in
chapter 2.4.2 within the framework of multi-phase materials by considering inclusions as
pores [ZIV09].
The upper limit of the effective Young’s modulus of porous ceramics of arbitrary
microstructure corresponds to the Voigt’s model [VOI89] and is mathematically defined
as follows:
𝐸𝑉 = 𝐸0 ∙ (1 − 𝜈𝑃 ) Equation 2.45
EV: Young’s modulus of the porous material according to Voigt’s model [GPa]
E0 : Young’s modulus of the pore free material / solid phase [GPa]
νP : Volume fraction of pores [-]
In case of materials with an isotropic microstructure, the upper limit can be defined
according to Hashin and Shtrikman approach [HAS63] [PAB07]:
+
1 − 𝜈𝑃 Equation 2.46
𝐸𝐻𝑆 = 𝐸0 ∙ ( )
1 + 𝜈𝑃
61
State of the scientific and technical knowledge
It is worth mentioning that the lower limit of the effective Young’s modulus of a porous
material cannot be defined according to both approaches of Voigt and Hashin and
Shtrikman as those approaches degenerate to zero for all finite porosity values [ZIV09].
Those approaches are valid if the Young’s modulus of each phase that constitutes the
heterogeneous material is similar. By considering the apparent porosity as the second
phase, this boundary condition is not respected anymore [RAV94].
By assuming a realistic solid phase with very small porosities, and isometric pore shape
and with a Poisson’s ratio in the range 0,1 - 0,4 [-], a dilute limit approximation for the
effective Young’s modulus can be defined [ZIV09] [TOR02]:
62
State of the scientific and technical knowledge
The parameter bʹ is an empirical constant characterizing the material. For a same phase,
this parameter can fluctuate in a wide range of values and depend on the manufacturing
process. Spriggs defined this model for alumina, and the typical values of bʹ are between
1 and 5 [-] [SPR61]. Although this analytical approach fits the evolution of the material
stiffness, this does not take into consideration other parameters that could also influence
this porosity dependency like the overlapping of pores, the pore shape or pore
distribution.
Another meaningful non-linear equation that results from the linear approximation in the
dilute limit is also defined as follows [ZIV09]:
Besides the apparent porosity, the pore shape also influences the porosity dependency of
the material stiffness as proved by Roberts et al. in case of spherical pores overlapping,
oblate ellipsoidal pores overlapping or solid spheres overlapping [ROB00]. More
sophisticated analytical approaches are required to predict the stiffness of material
according to the porosity and the pore shape.
Wachtman proposed a polynomial model including the volume fraction of pores and
empirical parameters depending on the pore geometry [WAC69] [SAL05]:
63
State of the scientific and technical knowledge
Since the study of Eshelby on the determination of the elastic field in materials
containing ellipsoidal inclusions [ESH57], more sophisticated predictions implementing a
parameter related to the pore configuration are available.
Boccaccini proposed a power-law relation by integrating an adjustable parameter S
offering the opportunity to examine the influence of complex pore geometries on the
material stiffness [BOC93] [BOC97]:
2 𝑆 Equation 2.52
𝐸 = 𝐸0 (1 − 𝑣𝑃3 )
This form parameter S is defined in function of the ratio z/x corresponding to the length
over width ratio and the angle φ between the x-axis and the stress direction:
1 1
𝑧 3 𝑧 −2 2
𝑆 = 1,21 ( ) [1 + (( ) − 1) 𝑐𝑜𝑠²(𝜑)]
𝑥 𝑥 Equation 2.53
64
State of the scientific and technical knowledge
Figure 2.38 represents the geometrical shape of the pore and the signification of the axes
x and z.
Considering high alumina refractory castables, the Young’s modulus of the solid phase
E0 that should be considered in the equation corresponds to the Young’s modulus of a
multi-phase material constituted of pure alumina and CA6.
65
State of the scientific and technical knowledge
direct relation between the compactness or the formulation design and the crack flank
friction can be deduced.
Andreassen model (Equation 2.54) and Funk & Dinger (Equation 2.55) model are
currently widespread approaches to define the compactness of ceramics formulations by
assuming the particles as spheres with a diameter d and as non-reactive during the
heating treatment [AND30] [DIN92].
q
d Equation 2.54
CPFT 100
Dmax
d q q
d min
CPFT 100
D
Equation 2.55
q
q
max d min
The main discrepancy between both analytical approaches of particle packing is the
implementation of the minimal particle size in the Funk & Dinger model, while in the
Andreassen model, an infinite distribution of fine particles is assumed. Such an
assumption is naturally not verified in real case. Therefore, Funk & Dinger approach
leads to an accurate design of the refractory material formulation by taking into
consideration the whole spectrum of particle size distribution as shown on figure 2.39
[MYH96].
66
State of the scientific and technical knowledge
by Myhre et al. prove that a distribution modulus of 0,30 [-] should not be exceeded in
order to provide a high compactness as well as suitable flow properties of the slurry
[MYH96]. Another option to optimize the particular packing consists in extending the
particle size spectrum of the formulation. This approach is nevertheless partially limited
as the addition of big size aggregates induces a degradation of some material properties
such as a characteristic brittleness number over 2 [-] synonymous with generation of
crack network in the material microstructure resulting in unsuitable ensuing elastic and
mechanical properties. Since the 1960’s, research works are focused on the incorporation
of fine particles like reactive alumina or microsilica to enhance the material compactness
[SAK09] [KHO11] [LIU07] [MYH94].
Fig. 2.39: Illustration of Andreassen model and Funk & Dinger model applied on a
castable mixture with dmin = 0,5 µm, Dmax = 4000 µm and q = 0,26 [-]
67
State of the scientific and technical knowledge
Higher order terms can be added to the exponent for the region where the density is very
close to the packing density. This supposes an optimal packing of spherical particles.
Based on the same principle as pore shape dependency models related to material
stiffness, sophisticated approaches aim to define the Young’s modulus of damaged
materials according to crack shape. Katchanov considers penny shaped microcracks with
a typical crack size a, and a crack density NV defined as follows [KAT92]:
Equation 2.57
1
𝑁𝑉 = ∑ 𝑎𝑖3
𝑉
The definition of the Young’s modulus of such a damaged material presenting penny
shaped microcracks depends on the orientation of the cracks.
Considering a random distribution of the cracks, the moduli of elasticity are defined as
follows:
𝐸0
𝐸=
16 (1 − 𝜇0 2 )( (1
− 3𝜇0 )/10)
1+( ( ) ∙ 𝑁𝑉 ) Equation 2.58
9 (1 − 𝜇0 /2)
68
State of the scientific and technical knowledge
𝐺0
𝐺=
16 (1 − 𝜇0 )((1 − 𝜇0 )/5)
1+( ( ) ∙ 𝑁𝑉 ) Equation 2.59
9 (1 − 𝜇0 /2)
𝐸0 Equation 2.60
𝐸=
16 (1 − 𝜇0 2 )
1 + 𝑘0 ( ( ) ∙ 𝑁𝑉 )
3 (1 − 𝜇0 /2)
69
State of the scientific and technical knowledge
𝑇𝑚 Equation 2.61
𝐸𝑇 = 𝐸𝑇=0 − 𝐵𝑆 𝑇𝑒𝑥𝑝 (− )
𝑇𝐴𝑏𝑠
Above 1000°C, the presence of a low content of impurities such as SiO 2 or Na2O in
alumina induces the formation of viscous phases with high viscosity. This viscous phase
formation is assumed to be responsible for the further stiffness decrease at elevated
temperatures [WEB99].
70
State of the scientific and technical knowledge
The formation of viscous phase is attested by an increase of the crack flank friction
[MYH94].This damping increase can be interpreted as:
a so-called brittle / ductile transition because of the viscosity decrease of the
viscous phase with increasing temperature [PEZ05]. That describes the transition
from brittle fracture to creep under dislocation mechanisms.
a release of stored elastic energy through crack flank friction phenomenon
occurring at the level of the grain boundary, where the viscous phase is
concentrated [SCH00a]. In this temperature range, where the damping values of
the material sudden increase, the fracture toughness can be enhanced by the
reduction of the stresses localized at the level of the crack tip through friction
phenomena and viscous phase bridging. This last mentioned toughening
mechanism may occur in case of refractory materials, as cracks mainly propagate
along the grains, where both high-intensity stress field and viscous phases are
localized.
71
State of the scientific and technical knowledge
Fig. 2.40: Comparison of computed (analytical and numerical) and experimental Young’s
modulus values in function of the temperature of a model material constituted of a
vitreous glass matrix and 30 vol-% spherical alumina inclusions [JOL08]
The influence of the grain size on the generation of a crack network during the cooling
phase can be defined through a fracture mechanics approach and the principle of
brittleness number BN. The brittleness number corresponds to the ratio between the
elastic energy at fracture time and the work of fracture. The brittleness number is
assessed as follows [HAR97]:
2
𝜎𝑐,𝑡 ∙𝑑 Equation 2.63
𝐵𝑁 =
𝐺𝑠,𝑓 ∙ 𝐸
72
State of the scientific and technical knowledge
If the brittleness number is smaller than 2 [-], the work of fracture required for crack
initiation is considerable. In that case, big cracks cannot be formed in the material
microstructure.
By increasing the particle size, the brittleness number increases, which causes the
generation of microcracks. However microcrack toughening mechanisms can take place,
which explains the decrease of the material brittleness as the specific energy of fracture
increases and in the same time the mechanical strength decreases. Furthermore, crack
nucleation induces an increase of the crack flank friction and a decrease of the material
stiffness [BRI08]. As both stored elastic energy and specific fracture energy depend on
the crack density within the material microstructure, the evolution of the material
brittleness in function of the particle size depend on the eventual toughening mechanisms
generated by the microcrack network [HAR97].
The calculation algorithm for the Young’s modulus assessment through Resonant
Frequency Damping Analysis defined in equation 3.16 is a function of the resonant
frequency and the specimen geometry. However, the specimen geometry, initially
integrated in the calculation and corresponding to the sample geometry at room
temperature, is supposed to evolve due to the thermal expansion behavior of the material.
Therefore, several studies correlate the temperature profile of the material stiffness with
the material thermal expansion behavior to better understand untypical Young’s modulus
evolution with temperature [TES06] [BAB11].
Considering the data acquisition through Resonant Frequency Damping Analysis, the
Young’s modulus of the material at elevated temperature can be corrected by
implementing the coefficient of thermal expansion [IMC12]:
2
𝑓𝑓,𝑇 1 Equation 2.64
𝐸𝑇 = 𝐸𝑅𝑇 ∙ ( ) ∙( )
𝑓𝑓,𝑅𝑇 1 + 𝛼∆𝑇𝑇−𝑅𝑇
73
State of the scientific and technical knowledge
To provide a suitable comparison of the obtained Young’s modulus values through those
different measurement techniques, it is worth mentioning that each measurement is
performed under different physical condition, namely different loading distribution, or
stress rate, stress amplitude, adiabatic or isothermal deformation. This discrepancy is the
main reason explaining the different Young’s modulus values according to the
measurement technique [WEG13].
74
State of the scientific and technical knowledge
three-point bending tests strongly depend on the applied load, the loading rate and the
eventual irreversible material deformation. Nevertheless, tested materials, also named
cermet, mainly exhibit at room temperature a linear elastic behavior up to fracture as
explained by Cutard et all but show a strong plastic deformation at elevated temperatures
[CUT96].
To conclude, a comparison between destructive and non-destructive measurement
techniques makes sense if the stress/strain behavior of the material reveals a perfect
linear-elastic and reversible behavior while a comparison between non-destructive
measurement techniques provide a similar trend if different materials are tested. It is also
important to notice that each measurement technique strongly depends on the accuracy of
the sample dimensions measurement also considered as an error-source for Young’s
modulus determination.
2.4.10 Influence of the crack flank friction on the material damping behavior
All those microstructural phenomena are responsible for the dissipation of the stored
elastic energy. The crack flank friction can be mathematically defined as follows
[ROE97]:
75
State of the scientific and technical knowledge
∆𝑊
𝑄−1 =
2𝜋𝑊𝑒𝑙,𝑚𝑎𝑥 Equation 2.65
Damping regroups each source, which is likely to attenuate the acoustic signal of a
specimen. Two difference sources can be distinguished: internal sources corresponding to
the crack flank friction phenomena occurring at microstructural scale and external
sources. Indeed, the environment of the specimen can also induce an attenuation of the
acoustic signal, such as the material support on which the specimen is lying or the wave
guide for high temperature measurements. Therefore, the crack flank friction cannot be
more considerable than the damping.
Equation 2.66
−1
𝐷𝑎𝑚𝑝𝑖𝑛𝑔 ≥ 𝑄
1 1 U
Qbkg QRT C exp act
Equation 2.67
RTAbs
76
State of the scientific and technical knowledge
Research works of Pezzoti and Schaller revealed the strong correlation between the
damping increase induced by numerous dislocations at grain boundaries of
polycrystalline materials and the macroscopic creep behavior of technical ceramics
[SCH00] [PEZ96]. Such an analogical approach leads to an estimation of the requested
activation energy for creep through the determination of the crack flank friction
background [NOW72] [SCH00].
1 1
Qbkg Equation 2.68
1 n
The fractional exponent parameter n ranges between 0 [-] and 1 [-] and is equal to 1 [-]
for ideal viscoelastic behavior.
As the viscoelasticity is also determined by thermally activated processes, the relaxation
time τ can be defined according to the Arrhenius approach [SIM14]:
77
State of the scientific and technical knowledge
By combining both equations 2.68 and 2.69 and by applying a logarithmic conversion,
the following equation is obtained leading to a possible estimation of the energy of
activation Uact by calculating the slope of the curve ln (Q-1) as a function of 1/TAbs
[SIM14]:
−1
𝑈𝑎𝑐𝑡 Equation 2.70
ln(𝑄𝑏𝑘𝑔 ) = −𝑛 [𝑙𝑛𝜔𝜏0 − ]
𝑘𝐵 𝑇𝐴𝑏𝑠
Such a correlation between the high temperature crack flank friction background and
creep shows that both properties are controlled by volume diffusion assisted climb of
dislocations [WEL04].
In order to better understand the analogy between the creep behavior of ceramics and the
crack flank friction at high temperature, the fundamental knowledge relative to the
material deformation under mechanical condition at elevated temperature should be first
explained. Refractoriness under Load (RuL) corresponding to the material deformation
with increasing temperature under constant load, and the Creep in Compression (CiC)
corresponding to the material deformation under constant load at a defined elevated
temperature, both constitute important material properties to predict the service life of
those materials. The creep is more generally defined as the time dependent deformation
at high temperature under constant load [GOT14] [CAR07]. The creep and creep rate
evolution with time are illustrated on figure 2.42.
78
State of the scientific and technical knowledge
increasing σ, T
increasing σ, T
Fig. 2.42: a) creep evolution with time - b) creep rate evolution with time [GOT14]
The curve describing the creep evolution at constant load and temperature with time as
well the curve describing the creep rate evolution at constant load and temperature with
time can be divided into three characteristic parts [GOT14] [JIN14]:
I. Primary creep
II. Stationary creep
III. Tertiary creep
79
State of the scientific and technical knowledge
decrease of the glassy phase viscosity results in the enhancement of the grain
reorientation within the matrix leading to deformation and creep [BAK94]. This
thermo-activated particular packing rearrangement favored by lubricating effect of
low-viscosity viscous phase is completed as soon as the local structure reached a
maximum compactness [KOL04].
U act / m
, TAbs F exp
Equation 2.71
RTAbs
Daniel Glymond reported in his PhD dissertation the effect of zirconia addition on the
toughening mechanisms and the creep behavior of mullite based products [GLY13].
Glymond examined the beneficial effects of monoclinic zirconia, yttria stabilized zirconia
and ceria stabilized zirconia to mullite on the activation energy required for creep. The
reference mullite product exhibits a low strain rate and a quite low activation energy at
elevated temperature. The addition of each zirconia results in an increase of the strain rate
at elevated temperature. The addition of monoclinic zirconia causes the decrease of the
activation energy, while the addition of ceria stabilized zirconia does not influence the
value of the activation energy and the addition of yttria stabilized zirconia induces an
increase of the activation energy. As shown on figure 2.43, even if the activation energy
could be improved through stabilized zirconia addition, more considerable strain rate at
lower temperature can be achieved as far as such composites are concerned. Therefore,
zirconia has a detrimental effect on the creep resistance relative to the same mullite
80
State of the scientific and technical knowledge
Fig. 2.43: Arrhenius plot measured data at 26,4 MPa compressive stress for mullite and
mullite zirconia composites and comparison with the literature [GLY13]
According to Glymond, the driving force of the creep enhancement is the difference in
oxygen diffusion coefficient responsible for material transport. Thus, a higher
concentration of oxygen vacancies favours material transport and subsequently creep.
Zirconia addition results in an increase of the creep rate, independently on the degree of
stabilization and the doping agent, as this oxide exhibits a considerable ionic conductivity
at elevated temperature. Furthermore this creep rate can be more considerable if the
doping agent favours the presence of oxygen vacancies in the crystal lattice. The strain
rate of a composite material can be defined in function of the oxygen diffusion
coefficient as follows according to a rearrangement of the Nabarro-Herring diffusion law
[HER50] [NAB67]:
14Deff
, TAbs Equation 2.72
d ² k BTAbs
81
State of the scientific and technical knowledge
Dryden et al. proposed another approach to explain the discrepancy of creep rate induced
by the addition of stabilized zirconia by assuming the formation of thin grain boundary
film at the grain boundaries. According to the quantity or thickness and the viscosity of
this viscous phase, the rearrangement of the particular packing under compression can be
favored inducing higher creep rate [DRY89]:
82
State of the scientific and technical knowledge
3
GB
, TAbs
1 Equation 2.73
3 GB d
Three different cases can be considered according to the nature of the stabilization:
Ca-PSZ (high alumina Low Cement Castable with addition of Partially Stabilized
Zirconia with CaO as doping agent):
In that case, the viscous phases correspond to calcium aluminate phases with high
local concentrations of CaO. Taking into consideration the Al2O3 – CaO binary
diagram presented on figure 2.8, melting point temperature at around 1400°C can
be achieved for eutectic compositions. Such chemical conditions can induce the
formation of viscous phases at relative low temperature with decreasing viscosity
with increasing temperature and favor creep mechanisms.
Mg-PSZ (high alumina Low Cement Castable with addition of Partially Stabilized
Zirconia with MgO as doping agent):
In that case, the Al2O3–CaO–MgO ternary diagram has to be taken into
consideration (see figure 2.45). On this diagram, it is noticeable that low
concentration of MgO in a calcium aluminate phase decreases the melting
temperature up to temperatures around 1300°C for the eutectic. Like CaO, MgO
addition enhances creep rate through formation of viscous phases at the grain
boundaries.
83
State of the scientific and technical knowledge
84
State of the scientific and technical knowledge
Y-FSZ (high alumina Low Cement Castable with addition of Fully Stabilized
Zirconia with Y2O3 as doping agent):
Figure 2.46 shows the phase equilibria ternary diagram of CaO-Al2O3-Y2O3. On
this diagram, eutectic of CaO and Al2O3 with low concentration in Y2O3 present
melting temperatures around 1500°C that are almost 200°C above the melting
temperatures of the eutectics Al2O3-CaO or Al2O3-CaO-MgO. This formation of
viscous intergranular phases at higher temperatures can explain a theoretical
suitable creep resistance of calcium aluminate bonded refractory materials with
addition of Fully Stabilized Zirconia with yttria as doping agent.
85
State of the scientific and technical knowledge
Fig. 2.47: Stiffness and damping of an yttria stabilized tetragonal zirconia (Y-TZP) with
2 mol-% Y2O3 up to 1000 K with the help of Impulse Excitation Technique [ROE03]
Fig. 2.48: Evolution of the resonant frequency and crack flank friction of an
alumina - alumina/zirconia composite at high temperature [LAM07]
86
State of the scientific and technical knowledge
As fundamental research work dealing with the ensuing damping behavior of dense
ceramic materials after thermal shock experiments, Coppola et al. studied the correlation
between the fracture energy of two different SiC materials and the resulting crack flank
friction development after thermal shock experiments with different temperature
difference [COP73]. The resulting fracture strength and damping after thermal shock are
illustrated on figures 2.49 and 2.50 according to the temperature difference.
Fig. 2.49: Fracture strength change and damping change with increasing thermal shock
temperature difference of a SiC material with a low fracture strength [COP73]
87
State of the scientific and technical knowledge
Fig. 2.50: Fracture strength change and damping change with increasing thermal shock
temperature difference of a SiC material with a high fracture strength [COP73]
According to Coppola et al., both materials exhibit an increase of the crack flank friction
with increasing thermal shock temperature difference up to 700°C for the SiC material
with a low fracture strength and 600°C for the SiC material with a high fracture strength,
while crack are nucleating and propagating through the material. Above those threshold
temperatures, damping values are constant, while cracks are further propagating and
unifying. Beyond the evolution trend of the damping with the thermal shock temperature
difference, it is worth mentioning that materials exhibiting a low fracture strength show
high values of damping and inversely, materials exhibiting a high fracture strength show
low values of damping.
However, the interpretation according to Coppola et al. of the damping behavior is
questionable. If the SiC material with a high fracture strength exhibits indeed a clear
increase up to 600°C followed by almost constant values characterized by a high
scattering, such a two-step damping evolution is not so obvious concerning the SiC
material with a low fracture strength. As shown on figure 2.49, damping changes increase
quite linearly with increasing furnace temperature. That shows the challenging
interpretation of damping evolution for materials having a low strength.
Based on this fundamental survey, current studies are focusing on the damping evolution
of refractory castables to characterize their thermal shock resistance. Pereira et al.
performed post-mortem Young’s modulus measurements after progressive thermal shock
experiments at different temperatures on high alumina castables containing electrofused
alumina aggregates through Impulse Excitation Technique (see figures 2.51 and 2.52)
[PER10].
88
State of the scientific and technical knowledge
While the Young’s modulus follows the similar trend independently on the temperature
difference (a two-step decay behavior), damping measurements provide further
information with regard to damaging process. At temperature difference below 750°C, a
local maximum crack flank friction peak is observed after three, four and five thermal
shock cycles at temperature difference of 650°C, 700°C and 750°C respectively. Those
local considerable damping values can be associated with different damage stages of
crack or microcrack extension and stress relaxation affecting the damping behavior and
not the stiffness of the material [PER10]. To conclude, the sensitivity of damping
measurements is expected to provide further information of multiscale damaging process
that cannot be revealed by the only Young’s modulus acquisition.
Fig. 2.51: Retained Young’s modulus of Fig. 2.52: Resulting damping of the high
the high alumina castable after progressive alumina castable after progressive thermal
thermal shocks according to the shocks according to the temperature
temperature difference [PER10] difference [PER10]
89
State of the scientific and technical knowledge
material are proven, the toughening mechanisms taking place in such porous and
heterogeneous materials are not scientifically detailed [MIY15]. Toughening mechanisms
in castables are however documented in terms of thermal expansion of monoclinic
zirconia after decomposition of zirconium silicate at high temperature [GUA15]. Tiny
microcracks are nucleated in the surrounding regions of monoclinic zirconia particles
inducing a microcrack toughening [LIU12]. The influence of monoclinic zirconia
addition introduced either under the form of electrofused zirconia or Al2O3-ZrO2-SiO2
aggregates is examined in a ULCC castable [REN12]. Such an addition of non-stabilized
zirconia induces the deterioration of the mechanical properties, namely MOR, elastic
properties, namely Young’s modulus as well as the degradation of the fracture toughness
KIC and the fracture surface energy to initiate crack propagation γNBT. Structural
reinforcement of high alumina castables can also be performed through the addition of
zirconia as fiber material [DON15]. In this case, the reinforcement is mainly due to the
typical shape of the zirconia raw material, while no information related to the nature of
the phase is given.
Structural materials exhibiting a high stiffness and a low damping (case of ceramic
materials):
│E*│. tan ψ = 0,1 GPa for structural metals
│E*│. tan ψ = 0,01 GPa for steel
│E*│. tan ψ < 0,001 GPa for aluminum alloy and main ceramic materials
│E*│represents the absolute value of the complex Young’s modulus and the product
│E*│. tan ψ = 0,6 GPa represents the blue line on fig. 2.53 distinguishing conventional
materials that occupy the region to the left of that line and the materials suitable for
90
State of the scientific and technical knowledge
vibration damping applications that occupy the region to the upper right of the diagonal
line. ψ represents the angle of the representative vector of complex Young’s modulus E*
in a (E’; E’’) map.
Under conditions of static loading, the Young’s modulus is defined according to Hook’s
law as shown on fig. 2.32 and equation 2.33. When dynamic loading in the form of
vibration is applied, the crack flank friction of the specimen resist the exciting force. In
case of viscous crack flank friction phenomena, a shift between the stress and strain
under steady state vibration conditions occurs. This can explain the following
mathematical definition of the stiffness [MON13]:
𝐸 ∗ = 𝐸 ′ + 𝑗𝐸 ′′ Equation 2.74
91
State of the scientific and technical knowledge
Where:
𝐸 ′′ Equation 2.76
𝜂 = ′ = tan 𝜓
𝐸
The term tan ψ is commonly known as loss factor and is also the reciprocal of the quality
factor Q of the resonances produced in the materials.
1 Equation 2.77
𝜂= = 𝑄−1
𝑄
Thus the damping of a defined resonant frequency like the flexural frequency f f measured
through Resonant Frequency Damping Analysis can be defined as follows like in
equation 3.23:
𝐿𝑜𝑠𝑠 𝑓𝑎𝑐𝑡𝑜𝑟 𝜂
𝐷𝑎𝑚𝑝𝑖𝑛𝑔 = 𝜁𝑓𝑓 = = Equation 2.78
(𝜋𝑓𝑓 ) (𝜋𝑓𝑓 )
92
State of the scientific and technical knowledge
a much smaller volume than the hard phase composed of the aggregates. This bond
system is nevertheless subjected to strong deformation while the hard phase is mostly
insensitive to the deformation. This bond system also includes the contact between the
grains and the microcracks. This discrepancy of behavior with regard to deformation is at
the origin of the non-linear behavior of ceramic materials.
After defining the contribution of the heterogeneous structure of geomaterials on the
non-linear behavior from the Hertzian concept based on the theory of Landau [LAN86],
Ostrovski highlighted the influence of the present microcracks in the bond system on the
non-linear behavior. An extension of the Hertzian concept is proposed by implementing
the effects of the microcracks and a quadratic non-linear coefficient is defined. Ostrovsky
concluded that a heterogeneous material exhibit a higher non-linear behavior if this
material is constituted of coarse grains. Furthermore the crack shape can strongly
influence the elastic properties of a material. If cracks with conical surfaces are present in
the bond system, the Young’s modulus of the material is not affected while the non-linear
parameter of the material drastically increases [OST01]. An illustration of the Hertzian
concept is shown on figure 2.54.
Fig. 2.54: Illustration of the Hertzian model: heterogeneous materials are constituted of
hard phases (aggregates) separated by a soft bond system [OST01]
In dense crystalline materials, other phenomena can generate some non-linearity in the
elastic behavior of the material. With increasing stresses and more particularly shear
stresses, the dislocation density increases in the material microstructure. The material
becomes softer, which results in non-linearity [OST01].
The non-linear response of a damaged ceramic material can be in short a helpful indicator
of microstructural damage assessment and a tool to monitor progressive damage. This
93
State of the scientific and technical knowledge
indicator has already been performant in the characterization of materials for medicine
application [DUC02] or for civil engineering application [ABE00]. In figure 2.55, an
illustration reveals the discrepancy between the propagation of a linear wave and a
non-linear wave throughout the material [DUC02]. In such experimental conditions,
which suppose a low density of cracks propagating in a dense matrix, the treatment of the
non-linearity of the acoustic response may provide further information concerning the
nature of the damaging in agreement with the microstructure examination and in an
optimal case may justify the damping behaviour of the ceramic pieces.
Fig. 2.55: Linear propagation and non-linear propagation of the acoustic signal
throughout a material [DUC02]
Figure 2.56 shows the evolution of the resonant frequency, the modal damping ratio and
the non-linear coefficient α1 of a Ti-alloy specimen during the fatiguing test. The results
reveal the discrepancy in terms of sensitiveness of each specimen characteristic
mentioned previously during a progressive damage. While the examination of the
evolution of the resonance frequency does not show any change synonymous with
microstructural damaging after four million cycles of excitation, the modal damping ratio
94
State of the scientific and technical knowledge
as well as the non-linear coefficient α1 exhibit an increase after 2,2 million and below one
million cycles respectively. The microcrack generation from the sample excitation
inducing the increase of damping and non-linearity coefficient α1 is first detected by the
non-linearity coefficient α1 and more than one million cycles later by the damping. Thus,
the examination of the non-linear acoustic response of material may provide with further
information with regard to microstructural damage at early stage.
Fig. 2.56: Detail of the variation of the linear and non-linear parameters in titanium
prismatic bars at early stages of the fatiguing process [ABE02]
If the contribution of the material heterogeneity on the non-linear acoustic response turns
out to be permanent, the operator of impulse excitation technique measurements can limit
to a lesser extent the contribution of the signal acquisition and treatment on the non-linear
response. Those key experimental indicators of non-linearity are also listed by Ostrovski
[OST01]. The most common physical phenomena resulting in the non-linearity of the
signal treatment are the following:
95
State of the scientific and technical knowledge
Among the refractory materials, some of them can rapidly react with the air moisture
such as doloma based materials [GHO12] [YEP07] or MgO based materials [NAN06]
[SUT10] and to a lesser extent alumina based materials [AHA02]. The presence of water
in the pores network of the material, even in small amount, can strongly affect the
acoustic response of the material. For low saturation level of moisture, the formation of
molecular layers of adsorbed water results in the activation of internal molecular forces in
a range of saturation level corresponding to the material hydrophility. Figure 2.57 shows
96
State of the scientific and technical knowledge
Fig. 2.57: Modification of the Hertzian contact at the grain boundary of a polycrystalline
material in presence of molecular layers of adsorbed water [OST01]
Therefore, some parameters have to be taken into consideration for room temperature
impulse excitation technique measurements in order to limit the effect of the moisture on
the non-linear behavior of the material, such as:
Degree of humidity
Pressure of the atmosphere
Apparent porosity of the material
Distribution, size and shape of the pores
Adsorption ability of the material
Miyaji et al. proved the impact of the air moisture on the sample relaxation after high
temperature Young’s modulus measurements through Resonant Frequency Damping
Analysis. Thus, a high alumina refractory castable, which does not exhibit a reversible
stiffness evolution during cooling, can recover to its initial stiffness before the high
temperature Young’s modulus experiment through room temperature reactions with the
air moisture [MIY12].
97
Materials and experimental procedure
3.1 Materials
In a first part of this research work, damping behavior of dense ceramics based on
Partially Stabilized Zirconia and alumina is studied. Samples are prepared by slip casting
after mixing powder and additives in an attrition mill for 60 min with 1200 rotations per
minute. Therefore, a reactive alumina CT 3000 LS SG (D50: 0,780 µm [ALM]) provided
by Almatis and a partially stabilized zirconia, doped with 3 mol-% of yttria, TZ-3YS-E
(D50: 0,644 µm [TOS04]) produced by Krahn Chemie GmbH are used. Those values of
mean diameter correspond to those of the material as received.
The Particle Size Distribution (PSD) of both products illustrated on figure 3.1 concern
the laser diffraction results obtained with the help of a Malvern Mastersizer 2000 of the
tested powders after mixing process in the attrition mill. The measurement is performed
through the dispersion of the slurry in water with ultrasound and a stirring speed of
2500 rotations per minute.
Table 3.1: Mean value of D10, D50 and D90 of both studied materials CT 3000 LS SG and
TZ-3YS-E after attrition according to the Rayleigh-Mie Method (with an
absorption index of 0,1 [-] for both materials and a refraction index of
1,780 [-] for the alumina based material and 1,445 [-] for the zirconia based
material)
98
Materials and experimental procedure
a mean diameter, which is significantly higher than the wave length of the laser beam.
For both materials, the implementation of the parameters such as the refraction index, the
diffraction, the reflection as well as the light absorption by the studied particles are taken
into account to give a more realistic distribution of the grains in the ultra-fine particle size
range.
Fig. 3.1: Particle size distribution of both studied materials CT 3000 LS SG and
TZ-3YS-E after milling process in attrition mill for 60 min with 1200 rotations per
minute
Figure 3.1 and table 3.1 reveal for both powders a similar bimodal distribution with a
slightly lower mean diameter for the zirconia based material with a value of D50 of
0,150 µm against 0,285 µm for the reactive alumina. The large second peak of the
particle size distribution of CT 3000 LS SG may however reveal some agglomeration of
the reactive alumina particles.
X-Ray Fluorescence (XRF) analyses are performed on the different raw materials. Those
measurements are performed with the help of a Philips PW 2404 XRF Spectrometer on
annealed samples. The determination of the Loss of Ignition (LOI) is performed
according to DIN EN 1744-1 [DIN EN 1744-1]. The samples are first ground in an agate
mortar, then dried at 110°C up to reach a constant weight. The assessment of the LOI is
carried out at 1050°C by determining the mass difference before and after the heat
treatment.
99
Materials and experimental procedure
The results concerning the raw materials CT 3000 LS SG and TZ-3YS-E are shown on
table 3.2. The aim of this chemical analysis is to measure the quantity of the doping agent
in TZ-3YS-E in order to place the composition ZrO2 – Y2O3 in the corresponding binary
diagram and to give further information with regard to the stabilized phase of zirconia
after its manufacturing.
Table 3.2: Chemical composition of the raw materials CT 3000 LS SG and TZ-3YS-E
Al2O3 Na2O SiO2 K2O Fe2O3 CaO ZrO2 Y2O3 HfO2 LOI
Materials [wt-%] [wt-%] [wt-%] [wt-%] [wt-%] [wt-%] [wt-%] [wt-%] [wt-%] [wt-%]
CT 3000
99,85 0,03 0,04 0,08 0,00 0,00 0,00 0,00 0,00 0,41
LS SG
TZ-3YS-E 0,20 0.06 0,02 0,00 0,01 0,00 92,06 4,99 3.30 0,01
In addition to the grain size distribution, the nature of the phases of each product is
identified through X-Ray Diffraction with the help of a diffractometer Bruker D8
Advance to qualitatively assess the crystallographic phase composition of ceramic
samples by using a Ni-filtered CuKα radiation ( = 1,54060 Å). Such a phase
identification is performed in Bragg-Brentano configuration on sample holder containing
the powder, whose upper surface must be cautiously flat to enhance the measurement
quality.
100
Materials and experimental procedure
α
α
α
α
α α α
Figure 3.2 reveals the only presence of α-corundum in CT 3000 LS SG coinciding with
the PDF Nr. 01-070-5679. That supposes that this reactive alumina was preliminary
calcined at high temperature (above 1000°C) so that the conversion of the metastable
alumina such as γ-Al2O3 or ε-Al2O3 in α-corundum is completed and this raw material
does not exhibit any β-Al2O3 as illustrated in figure 2.4. According to supplier data,
alumina CT 3000 LS SG has a Na2O content of around 300 ppm and is therefore a special
purpose low soda high reactive alumina [ALM]. This Na2O content is furthermore
verified by the X-Ray Fluorescence analysis performed on this raw material as previously
detailed in table 3.2. This can explain the absence of β-Al2O3 on the XRD pattern of this
raw material. Reference to chemical analysis
101
Materials and experimental procedure
zirconia. Such a data is required to understand the effect of the monoclinic to tetragonal
transformation on the resulting elastic and mechanical properties of the material after
thermal shock.
TC
CT
T T
M
M TTC TC
MC M CT
M M T CT T
M MM TM MM M T MC
Porter and Heuer defined a polymorph method for the calculation of the volume fraction
of the monoclinic phase by implementing the integrated intensities of the (111) m, (111̅)m
and (111)c peaks [POR79]. This polymorph method was based on a previous study
carried out by Garvie et al. [GAR72] and later extended or refined by Miller et al.
[MIL81], Toraya et al. [TOR84], Schmid [SCH87] and Howard et al. [HOW90]. The
calculation of the volume fraction of monoclinic phase is performed according to
equation 3.1 [POR79] [MIL81].
With:
Im: Integrated intensity of the monoclinic phase in (hkl) lattice plane [counts]
Ic : Integrated intensity of the cubic phase in (hkl) lattice plane [counts]
(111): (hkl) lattice plane with Miller indices h=1, k=1 and l=1
(111̅): (hlk) lattice plane with Miller indices h=1, k=1 and l=1 or -1
102
Materials and experimental procedure
Alumina (CT 3000 LS SG) and zirconia (TZ-3YS-E) are mixed together according to the
formulations listed in table 3.3. Some adjuvants are added in quantities mentioned in
table 3.4 to form the slurry:
1. distilled water
2. polyethylene glycol PEG 400 provided by Merck Millipore and used as chemical
binder
3. a defloculant called Dolapix CE64 supplied by Zschimmer & Schwarz
4. an antifoaming agent called Contraspum KWE also supplied by
Zschimmer & Schwarz
The slurry is next mixed in an attrition mill for 60 min with 1200 rotations per minute.
The resulting homogeneous suspension is cast into a water-permeable gypsum form with
the following geometry: 70 mm x 10 mm x 10 mm. After a 24h-curing at room
temperature, the dried samples are removed from the form and each surface of the sample
is lightly polished with sandpaper to calibrate the final geometry of the sample, so that
the minimum geometry is reasonable for accurate elastic properties measurements
through Resonant Frequency Damping Analysis, i.e. 50 mm x 5 mm x 5 mm.
103
Materials and experimental procedure
Table 3.3: Formulation of the dense samples based on Al2O3 and ZrO2 elaborated by slip
casting
Material Specification Content [wt-%]
Solid
Al2O3 CT 3000 LS SG 100-x
ZrO2 TZ-3YS-E x
Total 100
Additives
Distilled water H2Odis 41
Binder Polyethylene glycol PEG 1
400
Defloculant Dolapix CE64 1
Antifoaming agent Contraspum KWE 0,043
104
Materials and experimental procedure
The specimen are finally sintered at 1550°C for 2 h according to the sintering program
illustrated on figure 3.4.
1500°C/2h
5K/min
5K/min
500°C/2h
2K/min
Fig. 3.4: Sintering program of tested formulations of the Al2O3-ZrO2 binary system
In a second part of this research work constituting the main focus of the dissertation,
damping behavior of high alumina refractory castables based on tabular alumina is
examined. Low Cement Castables are tailored with the use of Secar 71 cement from
Kerneos with a calcium oxide content of around 30 wt-% (29,3 wt-% according to a
furher chemical analysis detailed in table 3.8). Tabular alumina used in the refractory
castable formulations as well as the reactive alumina are provided by Alteo, while
CaO-Partially Stabilized Zirconia (14,28 mol-%) [Ca-ZrO2], MgO-Partially Stabilized
Zirconia (10,15 mol-%) [Mg- ZrO2] and Y2O3-Fully Stabilized Zirconia (5,73 mol-%)
[Y-ZrO2] are supplied by Industriekeramik HochrheinGmbH. Some adjuvants are added
in quantities mentioned in table 3.4 to form the mixture:
1. room temperature water to control the nature of the calcium aluminate hydrate as
detailed on figure 2.7 and the subsequently rheological properties and setting time
of the slurry
2. a deflocculant called Castament FS40 from BASF based on polyethylene glycol
3. citric acid used as retarder
105
Materials and experimental procedure
Alumina and zirconia are mixed together with cement for 2 min without additives in
order to ensure a suitable homogeneity of the dry mixtures with the help of a Hobart
mixer from Toni Technik Baustoffprüfsysteme GmbH. The admixtures are afterwards
progressively added to the dry mixtures while mixing process is in progress. Deflocculant
and retarder are preliminary diluted into the room temperature water to avoid the
presence of any agglomerate. The slurry is mixed for 5 min more to favour a
homogeneous distribution of the adjuvants in the mixture. After those 5 min, a manual
mixing process is carried out to remove the fine powder that can remain at the surface of
the Hobart mixer. Then, a final mixing process is applied for two minutes. The obtained
slurry is then ready for casting into appropriate forms. This manufacturing process step
will be treated in a further chapter.
The quantity of the different raw materials is defined to fulfil the following conditions:
1. A CaO content of 1,5 wt-% to study the behavior of Low Cement Castables (Low
Cement Castables are defined in table 2.1)
2. The tailored tested compositions must exhibit the same compactness and therefore
be characterized by the same Andreassen packing coefficient (see equations 2.73
and 2.74). This criterion will be treated in the further part 3.1.2.2.
106
Materials and experimental procedure
Table 3.5: Formulation of the low cement high alumina refractory castables
Ref Ca-PSZ Mg-PSZ Y-FSZ
Materials Specification
[wt-%] [wt-%] [wt-%] [wt-%]
Solid
CA Cement Secar 71 5 5 5 5
Reactive alumina PFR 12,5 12,5 12,5 12,5
0 ‒ 0,045
Tabular Alumina 10 10 10 10
mm
Tabular Alumina 0 ‒ 0,3 mm 10 10 10 10
Tabular Alumina 0,2 ‒ 0,6 mm 10 5 5 5
Tabular Alumina 0,5 ‒ 1 mm 17,5 8,75 8,75 8,75
Tabular Alumina 1 ‒ 3 mm 35 35 35 35
Ca-ZrO2 0,2 ‒ 1 mm 13,75
(14,28 mol-% CaO)
Mg-ZrO2 0,5 ‒ 1 mm 13,75
(10,15 mol-% MgO)
Y-ZrO2 0,2 ‒ 1 mm 13,75
(5,73 mol-% Y2O3)
Total 100 100 100 100
Additives
Water H2O at RT 5 5 5 5
Defloculant FS 40 0,1 0,1 0,1 0,1
Retarder Citric acid 0,03 0,03 0,03 0,03
The influence of the porosity and more particularly of the pore geometry on the elastic
properties of castables is also a focus of this study. Starting from the reference
formulation Ref, some pore forming agents are added. Those pore forming agents are
characterized by different geometry:
1. Ref-sph-n refers to the formulation Ref plus addition of a spherical pore forming
agent
2. Ref-cyl-n refers to the formulation Ref plus addition of a cylindrical pore forming
agent
3. Ref-fib-n refers to the formulation Ref plus addition of a fibre-like pore forming
agent
For each formulation, n refers to the volume fraction of the pore forming agent and is
contained between 0 and 10 vol-%.
107
Materials and experimental procedure
Additives
Water H2O at RT 5 6 6 6
Defloculant FS 40 0,1 0,1 0,1 0,1
Retarder Citric acid 0,03 0,03 0,03 0,03
Spherical n [vol-%]
Pore forming Cylindrical n [vol-%]
agent
Fibre-like n [vol-%]
It is worth mentioning that the water content in the designed porous refractory castables
has to be raised to 6 wt-% to guarantee suitable rheological properties during the
manufacturing process because of the water absorption by the organic pore forming
agent.
The quantity of the different additives, namely water and defloculant is optimized to
enhance the workability and the rheological properties of the mixtures in a reasonable
time frame for casting. The flowability of the mixture is determined according to
DIN EN ISO 1927-4 [DIN EN ISO 1927-4] and defined as optimum when the slump after
30 s vibration (with a vibration amplitude of 0,75 cm) is equal to 20 ± 2cm.
The refractory castable is then ready for casting into preliminary lubricated metallic mold
fixed on a vibrating table. In a first step, the half of the mold is filled with the mixture
and a first vibration cycle of 1 min is applied. A second vibration cycle is applied after
completely filling the mold. A third vibration cycle is required for releasing the entrapped
air bubble, which can be responsible for apparent porosity increase after sintering.
108
Materials and experimental procedure
As refractory castables are composed of calcium aluminate cement, the specimen require
minimum green mechanical properties for the removal of the specimen from the metallic
mold. Therefore, the samples are placed for 24 h in a humid chamber for curing and
setting. After those 24 h, the removal can take place and the specimen are placed for
further 24 h in a humid chamber for curing completion [GAZ93] [ANT13]. After 48 h
curing, the samples are placed in a dryer at 110 ± 5°C for 24 h for the release of the
retained environmental water and the dehydration of some hydrate phases such as
amorphous AH3 and CAH10 [SCH00b] [GUI98].
Fig. 3.5: Geometry of tested samples after sintering at 1500°C for 6 h before grinding
With:
109
Materials and experimental procedure
The Particle Size Distribution illustrated on figure 3.6 concerns the laser diffraction
results obtained with the help of a Malvern Mastersizer 2000 of the tested zirconia
mentioned in table 3.4 and the grain fraction of tabular alumina 0,2 ‒ 0,6 mm and
0,5 ‒ 1 mm that are replaced. The measurement is performed through the dispersion of
each powder in water with ultrasound and a stirring speed of 2500 rotations per minute.
As the boundary conditions defined by Fraunhofer are thoroughly fulfilled, namely the
opacity of the particles and a mean particle diameter, which is significantly higher than
the wave length of the laser beam, the Rayleigh-Mie Method is not applied on those
powders.
The black dashed line of figure 3.6 constitutes the reference particle size distribution to
be replaced by the different zirconia and corresponding to the particle size distribution of
the middle-grain fraction of tabular alumina (TA). It is worth mentioning that both
Ca-ZrO2 and Y-ZrO2 exhibit a similar grain size distribution however shifted to finer
grain size. This statement is further proved on table 3.7. In case of MgO-Partially
Stabilized Zirconia, a monomodal particle size distribution is observed contrary to the
other studied powders and more particularly the tabular alumina.
110
Materials and experimental procedure
Fig. 3.6: Particle size distribution of tested zirconia Ca-ZrO2, Mg-ZrO2, Y-ZrO2 that
substitute the middle grain fraction of tabular alumina (TA) corresponding to 50 wt-% of
0,2 – 0,6 mm and 50 wt-% of 0,5 – 1 mm
Table 3.7: Mean value of D10, D50 and D90 of zirconia based raw materials Ca-ZrO2,
Mg-ZrO2 and Y-ZrO2 according to the Fraunhofer Method
Material D10 [µm] D50 [µm] D90 [µm]
Ca-ZrO2 52,896 155,315 687,174
Mg-ZrO2 525,390 785,215 1053,218
Y-ZrO2 78,117 281,436 610,222
Table 3.8: Compactness of the tested castable formulations Ref, Ca-PSZ, Mg-PSZ and
Y-FSZ according to Andreassen
Castable Formulation Andreassen coefficient [-]
Ref 0,30
Ca-PSZ 0,28
Mg-PSZ 0,30
Y-FSZ 0,29
111
Materials and experimental procedure
X-Ray Fluorescence (XRF) analyses are performed on the raw materials Ca-ZrO2,
Y-ZrO2 and Mg-ZrO2 with the same equipment as described in chapter 3.1.1.2. The
results concerning the high alumina materials, namely the reactive alumina, the tabular
alumina and the calcium aluminate cement, are shown on table 3.9.
Table 3.9: Chemical composition of the high alumina raw materials of refractory
castables
Al2O3 Na2O SiO2 TiO2 Fe2O3 CaO LOI
Materials
[wt-%] [wt-%] [wt-%] [wt-%] [wt-%] [wt-%] [wt-%]
Reactive
99,79 0,01 0,10 0,02 0,01 0,07 0,23
alumina
Tabular
99,5 0,05 0,06 0,26 0,09 0,04 0,18
alumina
Cement 69,8 0,22 0,30 0,02 0,18 29,3 0,13
112
Materials and experimental procedure
The chemical analysis of the zirconia based raw materials, i.e. Ca-ZrO2, Mg-ZrO2 and
Y-ZrO2 is shown on table 3.10. The aim of this chemical analysis is to measure the
quantity of the doping agent in order to place the composition ZrO2 – doping agent in the
corresponding binary diagram and to give further information with regard to the
stabilized phase of zirconia after its manufacturing.
113
Materials and experimental procedure
As far as Y-ZrO2 is concerned, the quantity of ZrO2 is equal to 86,75 wt-% while the
quantity of the doping agent Y2O3 is 9,67 wt-%. A conversion in mol-% reveals the
presence of 90,59 mol-% of ZrO2 for 5,51 mol-% of Y2O3 by taking into account the
quantity of impurities. The main impurities of Y-ZrO2 are HfO2 with a quantity of
1,59 wt- % and Al2O3 with a quantity of 1,29 wt-%. The rest of impurities, namely CaO,
Fe2O3, TiO2 and SiO2, are comprised between 0,12 wt-% and 0,31 wt-% and represent
0,70 wt-% of the powder chemistry. By neglecting the impurities, the zirconia Mg-ZrO2
is composed of 94,27 mol-% of ZrO2 and 5,73 mol-% of Y2O3. This composition is
documented in the ZrO2- Y2O3 binary diagram in chapter 2.2.2.3 on figure 2.13.
The nature of the phases of each raw material required for castable formulation is
identified through X-Ray Diffraction with the help of the same measurement technique as
the one described previously in chapter 3.1.1.2. This phase identification is performed in
Bragg-Brentano configuration on sample holder containing the powder, whose upper
surface must be cautiously flat to enhance the measurement quality. Concerning the
middle grain size fractions and the coarse grains of tabular alumina, the powder is firstly
ground in an abate mortar and sieved under 63 µm. As for the calcium aluminate cement
Secar 71, the powder is preliminary dried at 110°C to dehydrate the CAH10 phases and
the amorphous AH3 eventually generated by cement aging (the measurement is for safety
reason performed on a just received product). As far as zirconia is concerned, to avoid the
martensitic transformation of the metastable tetragonal particles due to the sample
preparation, X-Ray Diffraction analyses are performed on samples sieved under 200 µm
without mechanical solicitation.
XRD patterns of high alumina products, i.e. reactive alumina PFR, tabular alumina and
the calcium aluminate cement Secar 71 are respectively illustrated on figures 3.7, 3.8 and
3.9 in an angle range from 5° to 90° 2θ.
Figure 3.7 reveals the only presence of α-corundum in PFR. That supposes that this
reactive alumina was preliminary calcined at high temperature (above 1000°C) so that the
conversion of the metastable alumina such as γ-Al2O3 or ε-Al2O3 in α-corundum is
completed and this raw material does not exhibit any β-Al2O3 as illustrated in figure 2.4.
According to table 3.8, reactive alumina PFR has a Na2O content of 0,01 wt-% and is
therefore a special purpose low soda high reactive alumina. This can explain the absence
of β-Al2O3 on the XRD pattern of this raw material. PDF data bank JCPDS PDF
Nr. 01-070-5679 is used for corundum identification phase.
114
Materials and experimental procedure
α
α
α α α
Fig. 3.7: XRD pattern of reactive alumina PFR as received (α: α-corundum)
Figure 3.8 reveals the presence of α-corundum as major phase and also β-Al2O3 in tabular
alumina. The presence of β-Al2O3 can be explained by a higher soda content in this
alumina based product. According to table 3.8, tabular alumina has a Na2O content of
0,05 wt-%. PDF data bank JCPDS PDF Nr. 01- 070-5679 and Nr. 00-021-1096 are used
for corundum and β-Al2O3 phase identification respectively.
α
α α
α
α α
β α
Fig. 3.8: XRD pattern of tabular alumina in proportion of Ref formulation as shown in
table 3.4 after grinding and sieving under 63 µm (α: α-corundum; β: β-alumina)
115
Materials and experimental procedure
Figure 3.9 reveals the presence of CA and CA2 phases as main calcium aluminate phases
in cement Secar 71. Mayenite is also present as minor calcium aluminate phase. The
retained alumina that did not react with calcium oxide during the cement manufacturing
process is present as β-Al2O3. According to table 3.8, cement Secar 71 exhibits a high
soda content of 0,22 wt-%, which can explain the nucleation of β-Al2O3. PDF data bank
JCPDS PDF Nr. 01- 079- 1560 is used for β-Al2O3 identification, while the calcium
aluminate phase identification of CA, CA2 and C12A7 is performed by the use of PDF
Nr. 01-070-0134, 00- 023-1037 and 00-009-0413 respectively.
₂
₂
₁
₂ ₂₁
₁
₁₂ ₂ ₂ ₁₁
₂ ₁ ₁₁ ₁ ₂ ₂ ₁ ₂₁ ₁ ₂
₁ ₁₂ ₁ ₁ ₁ ₁₁ ₁ ₁ ₂₂ ₁ ₁
β ₁ ₂ ₁ ₃ ₂₁ ₁₁ ₁ ₁ ₁ ₂ ₂ ₂₂₁ ₁₁₂₁ ₁ ₂ ₁₁ ₂ ₂₂ ₁₂ ₂ ₂₂
Fig. 3.9: XRD pattern of calcium aluminate cement Secar 71 after drying at 110°C
(1: CA – 2: CA2 – 3: C12A7 – β: β-Al2O3)
XRD patterns of zirconia based powders, i.e. Ca-ZrO2, Mg-ZrO2 and Y-ZrO2 are
illustrated on figures 3.10, 3.11 and 3.12 respectively in an angle range from 5° to 90° 2θ
with the exception of Mg-ZrO2, whose measurement is performed in an angle range from
5° to 70° 2θ. XRD analyses are performed on powder sieved under 200 µm. The zirconia
is not previously ground to avoid the possible phase transformation due to the sample
preparation.
PDF data bank JCPDS PDF Nr. 00- 037- 1484 is used for zirconia monoclinic phase
identification, while the phase identification of tetragonal and cubic phases are performed
by the use of PDF Nr. 01-081-1544 and 00-026-0341 respectively.
As shown on figure 3.10, the zirconia exhibits the three crystalline phases. However the
peak related to the monoclinic phase reveals an extreme low intensity. The tetragonal and
116
Materials and experimental procedure
cubic phases can be considered as the main stable phases in Ca-ZrO2, while the
monoclinic phase is regarded as the minor phase. As a comparison, the peaks related to
the monoclinic phase in Ca-ZrO2 reveal an integrated intensity of around 900 counts at
2θ = 28,15°, while the same peaks in TS-3YS-E show an integrated intensity around
2900 counts (figure 3.3). The integrated intensity is measured by measuring the area
enclosed by the peak and the estimated background. Those integrated intensities can be
compared as the main peak reveals a similar integrated intensity in both experiments
(around 18500 counts for the peak at 2θ = 30,03°). As a monoclinic content is calculated
between 15 and 22 vol-% 5 vol-% in TS-3YS-E, a lower monoclinic content is
expected to be shown by Ca-ZrO2.
TC
TC
TC
TC
TC CT
M M M TC TC
Fig. 3.10: XRD pattern of Ca-ZrO2 as received after sieving under 200 µm
(M: monoclinic – T: tetragonal – C: cubic)
Figure 3.11 reveals the presence of periclase MgO in the raw powder Mg-ZrO2. A part of
the doping agent did not react with zirconia during the manufacturing process and is
consequently not present in the form of a solid solution or the kinetics of the cooling
phase during the manufacturing process is not enough sufficient to lead to the
stabilization of the solid solution. PDF data bank JCPDS PDF Nr. 00- 037- 1484 is used
for zirconia monoclinic phase identification, while the phase identification of tetragonal
and cubic phases are performed by the use of PDF Nr. 00-014-0534 and 00-054-1266
respectively. The identification of periclase is performed through the use of PDF
Nr. 00-045-0946.
117
Materials and experimental procedure
M
TC
MC
M
T TC TC
M M
M M M
M M T
M
M P M MM M M
M MM CCC
Fig. 3.11: XRD pattern of Mg-ZrO2 as received after sieving under 200 µm
(M: monoclinic – T: tetragonal – C: cubic – P: periclase)
Das et al. stressed the challenging experimental conditions to stabilize zirconia with
MgO. Indeed, MgO tends to crystallize at the grain boundaries of cubic phase of zirconia
during the cooling process and occupy more precisely the inter-granular positions
between polygonal and sub-rounded cubic zirconia grains. This phenomenon is enhanced
with increasing content of MgO doping agent. This typical structure of MgO-zirconia
results in a degradation of the mechanical properties of the zirconia product [DAS12].
Contrary to Ca-ZrO2, the monoclinic crystalline phase can be considered as the main
crystalline phase in Mg-ZrO2, as the peaks related to the tetragonal and cubic phases
reveals a lower integrated intensity as the main peak related to the monoclinic phase at
2θ = 28,15° showing an integrated intensity of around 6900 counts, while the main peak
that can be related to the tetragonal or cubic phase at 2θ = 30,03° reveals an integrated
intensity of around 3700 counts.
XRD pattern of Y-ZrO2 is illustrated on figure 3.12. PDF data bank JCPDS PDF
Nr. 00-037-1484 is used for zirconia monoclinic phase identification, while the phase
identification of tetragonal and cubic phases are performed by the use of PDF
Nr. 01-070-4435 and 01-082-1246 respectively.
By analogy with Ca-ZrO2, the zirconia exhibits the three crystalline phases. However the
peaks related to the monoclinic phase reveal an extreme low intensity. The tetragonal and
cubic phases can be considered as the main stable phases in Ca-ZrO2, while the
118
Materials and experimental procedure
monoclinic phase is regarded as the minor phase. The peaks related to the monoclinic
phase in Ca-ZrO2 reveal an integrated intensity of around 900 counts while the same
peaks in Y-ZrO2 show an integrated intensity around 800 counts for the peak at
2θ = 28,15°. However a careful comparison between Ca-ZrO2 and Y-ZrO2 should be
done as the main peak reveals a different integrated intensity in both experiments (around
18500 counts for Ca-ZrO2 vs 11800 counts for Y-ZrO2 for the peak at 2θ = 30,03°). Like
Ca-ZrO2, a low monoclinic content is expected to be shown by Y-ZrO2.
TC
CT
T
T C
C CT CT
M M M CT CT
Fig. 3.12: XRD pattern of Y-ZrO2 as received after sieving under 200 µm
(M: monoclinic – T: tetragonal – C: cubic)
The volume fraction of the monoclinic content of each zirconia raw material is measured
according to equation 3.1 and the Miller method. The results are summarized in
table 3.11.
Table 3.11: Content of the monoclinic phase in the tested raw zirconia materials
calculated according to Miller method (see chapter 3.1.1.2 and equation 3.1)
ZrO2-Material Volume fraction of monoclinic
Type and quantity of doping agent content [vol-%]
Ca-ZrO2 (14,28 mol-%) 1,5 ± 5
119
Materials and experimental procedure
Taking into account the uncertainty of the Miller calculation method, two groups of
materials can be distinguished:
1. Mg-ZrO2 that exhibits a considerable volume fraction of monoclinic content
(above 70 vol-%). As illustrated on the MgO – ZrO2 phase diagram (figure 2.12),
an insufficient cooling rate of this raw material leads to the formation of
monoclinic phase. Therefore, MgO exhibits a low effectiveness with regard to
zirconia stabilization explaining such a high volume fraction of monoclinic
content in this raw material.
2. Ca-ZrO2 and Y-ZrO2 that show a low volume fraction of monoclinic content
(under 5 vol-%). In spite of the low amount of monoclinic phase, a clear
distinction of the tetragonal and cubic phases is extremely challenging, and
therefore an accurate quantitative calculation of metastable tetragonal content in
those raw materials would have been audacious. It is furthermore worth
mentioning that the low residual monoclinic content in Y-ZrO2 is in agreement
with the interpretation of the ZrO2-Y2O3 binary phase diagram illustrated in
figure 2.13. Moreover, a lower monoclinic content is calculated in Y-ZrO2 in
comparison with TZ-3YS-E as it could be expected by the interpretation of the
binary diagram from the respective yttria content (2,40 mol-% and 4,75 mol-% for
TZ-3YS-E and Y-ZrO2 respectively)
120
Materials and experimental procedure
As shown on figure 2.10, the doping agent influences the transformation temperature of
zirconia and more particularly decreases the transformation temperature. While pure
monoclinic phase of zirconia (baddeleyite) transforms into tetragonal at 1170°C during
the heating phase, this transformation takes place at 1151°C for Ca-ZrO2, 1106°C for
Y-ZrO2 and 1034°C for Mg-ZrO2. In a further part of the study, thermal shock
experiments in air are performed on refractory samples based on those zirconia raw
materials. In function of the experimental procedure and more particularly the quenching
temperature, the understanding of the thermal behaviour of zirconia is essential to
understand the thermal shock damage behaviour of the sample.
A hysteretic behaviour is observed for each zirconia raw material during the cooling
phase. The measured transformation temperature through DTA – TG analysis can be
compared to those shown on the phase stability diagram of each system ZrO2 – doping
agent (CaO, MgO or Y2O3).
1. Ca-ZrO2: the tetragonal to monoclinic transformation occurs at 1075°C, while
figure 2.11 shows a transformation taking place at 1000°C for 6,76 wt-% as
measured by XRF analysis. A shift of 75°C is thus observed between the phase
diagram and the DTA-TG analysis
2. Mg-ZrO2: the tetragonal to monoclinic transformation occurs at 964°C, while
figure 2.12 shows that this transformation takes place at around 880°C for
10,15 mol- %. Indeed, the MgO-quantity of 10,15 mol-% in Mg-ZrO2 is superior
to the quantity of MgO characterizing the peritectoid of composition 7,5 mol-%
MgO for 92,5 mol-% ZrO2. Therefore, the martensitic transformation in Mg-ZrO2
should take place at the peritectoid reaction temperature, which is slightly above
880°C. A shift of around 80°C is thus observed between the phase diagram and
the DTA-TG analysis
3. Y-ZrO2: the tetragonal to monoclinic transformation occurs at 988°C, while
figure 2.13 shows that this transformation takes place at around 550°C for
121
Materials and experimental procedure
5,73 mol- %. This questionable and considerable shift of more than 400°C
between the phase diagram and the DTA-TG analysis may show that the
endothermic peak observed at 988°C may be caused by another external effect,
while no endothermic peak is exhibited in the DTA-TG diagram between 500°C
and 600°C. It may be assumed that DTA-TG analysis is not an appropriate
technique of measurement for the determination of the martensitic transformation
temperature in case of zirconia doped with yttria as yttria reveals a significant
stabilizing effect, proved by the low amount of retained monoclinic content. For
such a material, dilatometric measurements turn out to be more relevant.
The true density corresponds to the ratio of the mass of a quantity of dried material to its
true volume. This material property is measured according to DIN ISO 8130-2 with the
help of a gas comparison pycnometer AccuPyc 1330 from Micromeritics. The different
aggregates that constitute the castable formulation, i.e. tabular alumina, Ca-ZrO2,
Mg-ZrO2 and Y-ZrO2 are separately ground in an agate mortar, dried at 110°C and sieved
under 63 µm. Around 1 g of each ground material is placed in a pycnometer, which is
sealed in a defined compartment of the instrument with a defined volume. This technique
of measurement is based on the gas displacement method. As inert gas, helium
constitutes the displacement medium. This gas is admitted in the compartment, where the
sample is localized and then expanded into another precision internal volume. A
computation of the sample solid phase volume is performed through the observation of
the pressures filling the sample chamber and then discharging it into a second empty
chamber. Pores of one angstrom diameter are rapidly filled by helium molecules, so that
only the solid phase of the sample is able to displace the gas. The gas displacement
density is given by dividing this volume into the sample weight.
For each sample, five measurements are performed. Table 3.13 gives the mean value of
the true density of tabular alumina, Ca-ZrO2, Mg-ZrO2 and Y-ZrO2.
122
Materials and experimental procedure
Table 3.13: True density of the raw materials of tested refractory castable formulations
measured with the help of an air comparison pycnometer according to
DIN ISO 8130-2
Material True density [g.cm-3]
Tabular alumina 3,9627 0,0014
Ca-ZrO2 5,7181 0,0005
Mg-ZrO2 5,6360 0,0004
Y-ZrO2 5,9849 0,0007
In chapter 2.2.1.1, the density of tabular alumina is calculated by 3,55 g.cm-3 [BUE07].
However this density corresponds to the density of the raw product taking the closed
porosity into consideration. The tabular alumina measured according to DIN ISO 8130-2
is preliminary ground and sieved under 63 µm. The obtained true density is close to the
theoretical density of corundum (3,99 g.cm-3) [KOL04]. The low amount of closed
porosity in this small grain fraction can explain this deviation of the true density with a
value of 3,9627 g.cm-3.
Concerning the zirconia base raw materials, the discrepancy of the true densities can be
attributed by the stabilized phases. Indeed, Mg-ZrO2 that exhibits the monoclinic phase as
major phase according to both X-Ray Diffraction and DTA-TG analyses shows a true
density of 5,6360 g.cm-3, which is close to the theoretical true density of monoclinic
zirconia phase (ρmon = 5,56 g.cm-3) [KOL04]. The true density of Ca-ZrO2 and Y-ZrO2 is
higher than the one of Mg-ZrO2. That proves a higher stabilization level of those raw
powders, which is in agreement with both X-Ray Diffraction and DTA-TG analyses. It
seems to be challenging to identify the major stabilized zirconia phase in those products
through the measurement of the true density. Nevertheless, the considerable true density
of Y-ZrO2 (5,9849 g.cm-3) may be caused by a high cubic phase content as this stabilized
phase reveals a theoretical true density ρcub of 6,10 g.cm-3 [KOL04]. Ca-ZrO2 exhibits a
true density of 5,7181 g.cm-3, which is close to the theoretical true density of the
tetragonal phase of zirconia (ρtet = 5,80 g.cm-3) without necessarily revealing the
tetragonal phase as major phase [KOL04].
Two different sintering programs are taken into consideration according to the nature of
the castable as illustrated on figure 3.13.
Profile 1 is applied for the pure tabular alumina based castables, i.e. Ref, Ref-cyl-n,
Ref-sph-n and Ref-fib-n. The specimen are fired at 1500°C for 6 h with a heating and
cooling rate of 2 K.min-1.
123
Materials and experimental procedure
Profile 2 is applied for the tested castables containing zirconia, i.e. Ca-PSZ, Mg-PSZ and
Y-PSZ. The specimen are also fired at 1500°C for 6 h with a heating rate after debinding
at 450°C of 1 K.min-1 and a cooling rate of 1 K.min-1. The temperature is maintained at
1050°C for 2 h during the cooling phase to limit the increase of the thermal stresses due
to coefficient mismatch as explained on figure 2.33 or due to the martensitic
transformation of zirconia as explained on figure 2.9.
Profile 3 is applied for the tested castable Ca-PSZ in order to examine the influence of
the sintering temperature on the thermal shock resistance. The specimen are fired at
1300°C for 6 h with a heating rate after debinding at 450°C of 1 K.min-1 and a cooling
rate of 1 K.min-1. The temperature is maintained at 1050°C for 2 h during the cooling
phase to limit the increase of the thermal stresses due to coefficient mismatch as
explained on figure 2.33 or due to the martensitic transformation of zirconia as explained
on figure 2.9.
1500°C/6h 1500°C/6h
1300°C/6h
2K.min-1
1050°C/2h 1050°C/2h
2K.min-1
1K.min-1 1K.min-1
450°C/6h
Fig. 3.13: Sintering program of the tested high alumina refractory castables
124
Materials and experimental procedure
In order to guarantee the accuracy of the measurement of the elastic, mechanical and
thermal properties of the castables, grinding operation is applied on sample surfaces.
According to the sample geometry, two procedures can be distinguished:
1. For prismatic bars: the upper surface of the sample (casting surface) is ground
with the help of a cup wheel appliance MPS2 from G&N GmbH in order to obtain
sample with the required thickness and plan-parallel faces. After such a grinding
operation, the deviation of sample thickness must be ± 0,1 mm.
2. For cylindrical samples: by means of the same operating device, the required
cylindrical geometry, i.e. (50 ± 0,5) mm height und (50 ± 0,5) mm diameter is
achieved by grinding the lower and upper surfaces of the samples. This geometry
is requested for Cold Crushing Strength measurements. For Refractoriness under
Load and Creep in Compression measurements, a co-axial inner hole of 12,5 mm
throughout the sample height is performed with the help of a diamond hollow drill
of a standing drill type AB4 S8 from Alzmetall. Such a geometry is requested
according to DIN EN ISO 1893:2008 [DIN EN ISO 1893:2008].
The different sample geometries after grinding are illustrated on figure 3.14
(160 mm x 40 mm x 40 mm on the right side, 150 mm x 25 mm x 25 mm in the middle
and two 50 mm x 50 mm cylinders on the left side, one of them has an inner hole of
12,5 mm of diameter for RuL and CiC measurements)
Fig. 3.14: Geometry of tested samples after sintering at 1500°C for 6 h before grinding
125
Materials and experimental procedure
In the following part, the term dense shaped product refers to product having a true
porosity lesser than 45 vol-% when measured in accordance with DIN EN 993-1. The
true porosity is the sum of the apparent porosity determined according to DIN EN 993-1
and the closed porosity determined according to DIN EN 993-2.
The bulk density (ρb) of a dense shaped specimen corresponds to the ratio of the mass of
the dry material of a porous body to its bulk volume. This property is expressed in g.cm-3
or kg.m-3. The apparent porosity of a dense shaped specimen (πa) corresponds to the ratio
of the total volume of the open pores in a porous body to its bulk volume. This property
is expressed as a percentage of the bulk volume. Both properties can be assessed through
the Archimedes’ principle and the implementation of three different masses according to
DIN EN 993-1 [DIN EN 993-1].
Those three different masses refer to:
1. the weight of the specimen after drying at 110 ± 5°C and after achieving a
constant mass (m1)
2. the apparent mass when the sample is immersed in water with which it has been
impregnated under vacuum for 30 min (m2)
3. the mass in air of the specimen while still soaked with the liquid (m3). In the study
hereby, porosity measurements are performed with water
126
Materials and experimental procedure
𝑚1
𝜌𝑏 = .𝜌
𝑚3 − 𝑚2 𝑙𝑖𝑞 Equation 3.2
With:
𝑚3 − 𝑚1
𝜋𝑎 = . 100
𝑚3 − 𝑚2 Equation 3.3
With:
One objective of this study hereby is to examine the influence of the porosity and the
pore geometry on the resulting elastic properties. Pore forming agents are added to the
reference castable formulation up to achieve an apparent porosity after sintering close to
40 vol-%. As those values of porosity are close to the threshold value of 45 vol-%
separating the dense refractory materials from the insulating materials, porosity
127
Materials and experimental procedure
𝑉𝑏 = 𝐿 . 𝑏 . 𝑡 Equation 3.4
With:
𝑚1
𝜌𝑏 =
𝑉𝑏 Equation 3.5
With:
The apparent porosity (πa) of the porous specimen is defined according to equation 3.6:
𝜋𝑓 𝜌𝑏
𝜋𝑎 = [1 − − ] . 100
100 𝜌𝑡 Equation 3.6
With:
128
Materials and experimental procedure
3.2.2 Closed porosity and total porosity of Ref, Ca-PSZ, Mg-PSZ and Y-FSZ
The closed porosity (πf) is the ratio of the total volume of the closed pores in a porous
body to its bulk volume, expressed as a percentage of the bulk volume. The true
porosity (πt) is the ratio of the total volume of the open and closed pores to the bulk
volume of the material, expressed as a percentage of the bulk volume. The true porosity
is consequently the sum of the apparent porosity and the closed porosity. Such properties
are determined according to DIN ISO 8130-2 with help of a gas comparison pycnometer.
The instrument used is the AccuPyc 1330 from Micromeritics. The measurement
principle is previously detailed in chapter 3.1.2.6.
The total porosity is defined as follows according to equation 3.7
𝜌𝑡 − 𝜌𝑏
𝜋𝑡 = . 100
𝜌𝑡 Equation 3.7
With:
As mentioned before, the closed porosity (πf) is deduced from the total porosity (πt) and
the apparent porosity (πa) according to equation 3.8:
𝜋𝑓 = 𝜋𝑡 − 𝜋𝑎 Equation 3.8
With:
129
Materials and experimental procedure
To determine the closed porosity and the total porosity of the complete tested
formulations, a representative volume of a sample is ground in an abate mortar after
sintering. The ground sample after drying is ready for measurement. The calculation of
the closed porosity (πf) requires the assessment of the apparent porosity according to
DIN EN 993-1.
The Cold Crushing Strength (CCS) corresponds to the maximum load per unit area,
applied under specified conditions (compressive stresses) at room temperature, that a
refractory sample will withstand before failure occurs. This property is determined in
accordance with DIN EN 993-5 [DIN EN 993-5] with the help of a hydraulic
compression testing device from Losenhausenwerk Düsseldorfer Maschinenbau AG
(year of construction 1953 – maximum vertical load: 100 000 kg). Contrary to the
evaluation of the apparent porosity or bulk density presented before that are independent
of the sample geometry, the Cold Crushing Strength is determined on cylindrical or cubic
samples when steadily increasing compressive load of 1,0 ± 0,1 MPa.s-1 is applied on the
specimen in the cast direction (compacting direction). The Cold Crushing Strength is then
calculated from the maximum load indicated at failure and the mean cross-sectional area
over which the load is applied as shown on equation 3.9.
With:
The bending strength or Modulus of Rupture (MOR) corresponds to the maximum stress
that a prismatic dense shaped specimen can withstand when it is bent in a three-point
bending device at room temperature. This property is determined according to
DIN EN 993-6 [DIN EN 993-6] with the help of a three-point bending device from Toni
Technik Baustoffprüfsysteme GmbH (maximum load 20 kN). The prismatic sample lies on
130
Materials and experimental procedure
a two-point support with a distance between the points of support of 125 ±1 mm. The
load is applied vertically in the middle of the sample and the rate of increase of stress
shall be 0,15 ± 0,015 MPa.s-1 for dense shaped refractory product (case of refractory
castable) or 0,05 ± 0,005 MPa.s-1 for shaped insulating refractory product.
The Modulus of Rupture is the ratio between the bending moment at the point of failure
(Mmax) to the moment of resistance (W) and is calculated according to the following
equation 3.10.
𝑀𝑚𝑎𝑥 3 . 𝐹𝐵𝑒𝑛𝑑,𝑚𝑎𝑥 . 𝐿𝑠
𝜎𝑀𝑂𝑅 = =
𝑊𝑟𝑒𝑠 2 . 𝑏 . 𝑡2 Equation 3.10
With:
Even if a distance between the points of support of 125 ± 1mm is only appropriate
according to standard for the prismatic geometry 150 mm x 25 mm x 25 mm, this
distance is also applied for sample with geometry 160 mm x 40 mm x 40 mm as no
distance is advised for this geometry.
The hot bending strength or Hot Modulus of Rupture (HMOR) corresponds to the
maximum stress that a prismatic dense shaped specimen can withstand when it is bent in
a three-point bending device at elevated temperature. This property is determined
according to DIN EN 993-7 [DIN EN 993-7] with the help of a static electromechanical
test facility type 5500 Series from Instron Deutschland GmbH equipped with a Bluehill
Software and an electrically test frame type 1186. The measurement principle is similar
as the one for determination of the Modulus of Rupture. The three-point test set up is
coupled with an electric furnace with MoSi2 heating elements. The maximum load is
131
Materials and experimental procedure
200 kN. Like the Modulus of Rupture, the Hot Modulus of Rupture is calculated
according to equation 3.10.
Nevertheless, some geometrical parameters have to be taken into consideration as shown
on equation 3.11 and 3.12.
𝑡 1
<
𝐿𝑠 4 Equation 3.11
𝑡 1
>
𝑏 3 Equation 3.12
With:
In spite of damage generated by thermal shock experiments and the high test temperature
(up to 1500°C), the designed castables still show a sufficient bending strength at elevated
temperature, so that the rate of increase strength is 0,15 ± 0,015 MPa.s-1.
The proof temperature shall be a multiple of 100°C and the heating rate shall be between
2 K.min-1 and 10 K.min-1 up to the test temperature ± 10 K and a heating rate between
4 K.min-1 and 6 K.min-1 is considered as optimum. In the study hereby, HMOR tests are
performed at 1300°C.
132
Materials and experimental procedure
The Refractoriness under Load (RuL) corresponds to a particular measure of the behavior
of a refractory material subjected to the combined effects of load, rising temperature and
time. This material property is defined according to DIN EN ISO 1893
[DIN EN ISO 1893] with the help of a testing facility constituted of a lowerable and
raisable displaceable furnace unit from Netzsch with a data acquisition unit developed by
Hesse Instrument. The measurement of the length change of the sample is performed by a
sensor from Hottinger Baldwin Messtechnik GmbH. The measurement principle is the
following: a cylindrical test piece of geometry 50 ± 0,5 mm height and 50 ± 0,5 mm
diameter with a co-axial inner hole of 12,5 ± 0,5 mm is placed in a furnace. This
specimen is subjected to a specified constant compressive load and is heated at a
specified rate of temperature increase until a prescribed deformation or subsistence
occurs. The temperature acquisition is performed by means of a thermo-couple placed in
the inner hole of the specimen. The deformation of the test piece is recorded as the
temperature increases, and temperatures corresponding to specified proportional degrees
of deformation are determined.
The applied vertical compressive load depends on the nature of the specimen and shall
be:
1. 0,20 MPa for dense shaped refractory products (case of refractory castables)
2. 0,05 MPa for insulating materials
The heating rate shall be between 4,5 K.min-1 and 5,5 K.min-1 and can be 10 K.min-1 for
dense shaped refractory products up to 500°C.
From this thermo-mechanical behavior and more particularly from the geometry change
of the sample according to the temperature, particular points can be noticed such as the
temperature of maximum expansion Tεmax and the percentage change in the height of the
test piece as a function of the temperature. For instance, T0,5 corresponds to the
temperature at which the sample exhibits a percentage change in the height of 0,5 %
compared to Tεmax.
A Refractoriness under Load measurement is considered as completed if the temperature
of the furnace reaches 1700°C or if the percentage change in the height of the test piece
reaches 5 %.
133
Materials and experimental procedure
The applied vertical compressive load depends on the nature of the specimen and shall
be:
3. 0,20 MPa for dense shaped refractory products (case of refractory castables)
4. 0,05 MPa for insulating materials
The heating rate shall be between 4,5 K.min-1 and 5,5 K.min-1 until the test temperature
and can be 10 K.min-1 for dense shaped refractory product up to 500°C.
From this thermo-mechanical behavior and more particularly from the geometry change
of the sample according to the temperature, particular points can be noticed such as the
rate of the percentage change in the height of the test piece during the last 20 h of the
experiment when the temperature is constant. This term is written v5h – 25h and is
expressed in %.h-1.
In the study hereby, Refractoriness under Load measurements are performed on the tested
refractory castables between two high temperatures, namely 900°C – 1500°C, with the
help of the test facility described above with a mechanical load of 0,20 MPa. According
to DTA-TG results illustrated on table 3.11, the martensitic transformation of the tested
zirconia is comprised in this temperature range. The aim is to observe and study the
134
Materials and experimental procedure
Fig. 3.15: Cycling Refractoriness under Load temperature profile tested on refractory
castables Ref, Ca-PSZ, Mg-PSZ and Y-FSZ between 900°C and 1500°C
The thermal shock resistance of a dense shaped refractory material is defined as the
damage resistance generated by an abrupt quenching test from 950°C to room
temperature when compressed air is applied on a surface of the refractory specimen. In a
135
Materials and experimental procedure
first part, the standardized procedure to assess the thermal shock resistance of refractory
materials is described. In a second part, the same experimental procedure with the
exception of the quenching temperature is detailed. Indeed, a part of this study hereby
treats of the thermal shock behavior of the tested refractory castables at different
quenching temperatures from 750°C to 1050°C.
As stated above, the thermal shock resistance of dense shaped refractory castables is
assessed through quenching test in air at 950°C according to DIN EN 993-11
[DIN EN 993-11].
Fig. 3.16: Experimental set up of quenching test in air according to DIN EN 993-11 on
high alumina castables with geometry 160 mm x 40 mm x 40 mm (L x b x t)
136
Materials and experimental procedure
𝑆𝑉𝑎𝑓𝑡𝑒𝑟 𝑖 𝑇𝑆
𝑆𝑉𝑟𝑒𝑠𝑖𝑑𝑢𝑎𝑙 𝑎𝑓𝑡𝑒𝑟 𝑖 𝑇𝑆 = . 100
𝑆𝑉𝑎𝑓𝑡𝑒𝑟 𝑠𝑖𝑛𝑡𝑒𝑟𝑖𝑛𝑔 Equation 3.13
With:
SVresidual after i TS: Residual sound velocity in the length of the specimen after i thermal
shock cycles [%]
SV after i TS: Sound velocity in the length of the specimen after i thermal shock
cycles [µs]
SV after sintering: Sound velocity in the length of the specimen after sintering [µs]
The Modulus of Rupture of the specimen obtained through three-point bending test after
sintering and after thermal shock is recorded and a residual bending strength can be
calculated by dividing the recorded modulus of rupture of the specimen after quenching
test by the mean value of the modulus of rupture of the specimen formulation after the
sintering as shown on equation 3.14.
𝑀𝑂𝑅𝑎𝑓𝑡𝑒𝑟 𝑖 𝑇𝑆
𝑀𝑂𝑅𝑟𝑒𝑠𝑖𝑑𝑢𝑎𝑙 𝑎𝑓𝑡𝑒𝑟 𝑖 𝑇𝑆 = . 100
𝑀𝑂𝑅𝑎𝑓𝑡𝑒𝑟 𝑠𝑖𝑛𝑡𝑒𝑟𝑖𝑛𝑔 Equation 3.14
With:
137
Materials and experimental procedure
MORresidual after i TS: Residual Modulus of Rupture of the specimen after i thermal shock
cycles [%]
MOR after i TS: Modulus of Rupture of the specimen after i thermal shock cycles
[MPa]
MOR after sintering: Modulus of Rupture of the specimen formulation after sintering
[MPa]
In addition to the measurement of the retained Modulus of Rupture and retained sound
velocity in the length of the specimen, the elastic properties of the specimen, namely the
Young’s modulus, the shear modulus, the Poisson’s ratio as well as the damping of the
resonant flexural and torsion frequencies are assessed after sintering and after each
thermal shock cycle through Resonant Frequency Damping Analysis. This experimental
procedure is detailed in chapter 3.2.8.3.1.
A total of 10 thermal shock cycles are performed. After the 1st, 3rd, 5th, 7th and 10th
thermal shock cycle, 4 samples are put aside in order to measure the retained modulus of
rupture (MOR) and to examine the evolution of the formulation microstructure with
progressive thermal shock cycles. Starting from 20 samples, 8 samples remain after the
whole set of thermal shock cycles to guarantee representative values of elastic and
mechanical properties.
One focus of the study hereby is to examine the influence of the temperature difference
on the resulting elastic properties of the tested castable Ca-PSZ. Therefore, the same
experimental procedure described before in chapter 3.2.7.1.1 according to
DIN EN 993-11 is taken into account with the major exception that the quenching
temperature is either below or above the standardized temperature. Three quenching
temperatures are chosen: 750°C, 850°C and 1050°C.
The elastic properties of the specimen, namely the Young’s modulus, the shear modulus,
the Poisson’s ratio as well as the damping of the resonant flexural and torsion frequencies
are assessed after sintering and after each thermal shock cycle through Resonant
Frequency Damping Analysis on samples with geometry 160 mm x 40 mm x 40 mm
(L x b x t). This experimental procedure is detailed in chapter 3.2.8.3.1.
A total of 10 thermal shock cycles are performed. After the 1st, 3rd, 5th, 7th and 10th
thermal shock cycle, 4 samples are put aside in order to measure the retained modulus of
rupture (MOR) and to examine the evolution of the formulation microstructure with
progressive thermal shock cycles. Starting from 20 samples, 8 samples remain after the
138
Materials and experimental procedure
whole set of thermal shock cycles to guarantee representative values of elastic and
mechanical properties.
Furthermore, another focus of the study treats of the generation of cracks in dense
samples elaborated by slip casting through quenching test at 400°C in water. The
composition of these dense samples is detailed in chapter 3.1.1.3. The samples are
sintered at 1550°C for 2 h according to the sintering profile illustrated on figure 3.4. The
thermal shock resistance of dense shaped refractory materials is determined according to
DIN 51068 [DIN 51068].
According to this standard, cylindrical samples of geometry 50 ± 0,5 mm diameter and
50 ± 0,5 mm height are placed after sintering in a dryer at 110 ± 5°C for 45 min up to
reach a constant mass. This step is regarded as pre-heating phase. Then, the samples are
placed in a further furnace at test temperature of 950 ± 50°C for 15 min. The samples are
then quenched in running water at temperature between 10°C and 20°C for 3 min. After
each quenching test, the samples are placed in the dryer up to reach a constant mass and
the cycle can be repeated. In case of quenching test in water, no further experiment is
carried out contrary to thermal shock test in air with measurement of the resulting sound
velocity and modulus of rupture. This procedure must be repeated as long as the
specimen failure occurs. If the sample can withstand 30 cycles, the procedure is stopped.
The number of performing cycles per sample is recorded.
In the study hereby, the quenching test in running water are performed on prismatic
samples with the following geometry 50 mm x 5 mm x 5 mm (L x b x t). To limit the
microstructural damage in those dense samples characterized by a low crack propagation
resistance, and because of the considerable heat capacity of water (4,187 kJ.kg-1.K-1)
[INC90] the quenching temperature is fixed at 400°C to favor crack initiation without
catastrophic crack propagation and sample failure.
After quenching test, the samples are dried. The elastic properties of the specimen,
namely the Young’s modulus, the damping of the resonant flexural frequency and the
non-linearity of the resonant flexural frequency are assessed after sintering and after each
thermal shock cycle through Resonant Frequency Damping Analysis. This experimental
procedure is detailed in chapter 3.2.8.3.1.
A total of 3 thermal shock cycles are performed. After each thermal shock cycle,
5 samples are put aside in order to examine the evolution of the formulation
microstructure with progressive thermal shock cycles. Starting from 20 samples,
5 samples remain after the whole set of thermal shock cycles to guarantee representative
values of elastic properties.
139
Materials and experimental procedure
A further focus of this study is the examination of the influence of the annealing medium
on the resulting elastic properties of tested refractory castables Ref, Ca-PSZ and Y-FSZ.
Buyco et al. proved that the heat capacity of molten aluminum is constant from the
melting point up to 1700 K [BUY70]. The value of the heat capacity of the aluminum at
quenching test temperature is around 2,813 cal.g-1.K-1, i.e. 11,777 kJ.kg-1.K-1. This value
of heat capacity is considerably above the heat capacity of the air at room temperature,
which is 1,0054 kJ.kg-1.K-1. Therefore, an experimental procedure is developed at the
Institute of Mineral Engineering/Aachen for the purpose of dipping refractory samples in
molten aluminum at 800°C in order to generate the thermal shocks occurring in
metallurgical applications. No standard consequently defines this kind of quenching
method. While quenching tests in air or in water provide abrupt cooling temperature
difference, the thermal shock resistance under heating conditions is examined through
quenching tests in molten aluminum at 800°C.
The chemical analysis of the used aluminum performed through atomic absorption
spectrometry at the Institute for Machine Elements and Machine Design from RWTH
Aachen University is shown on table 3.14.
Table 3.14: Chemical analysis of the used aluminum through atomic absorption
spectrometry (Results from the Institute for Machine Elements and
Machine Design / RWTH Aachen University)
Element concentration
Element
[vol-%]
Al 88,80
Si 9,64
Fe 0,694
Mn 0,304
Mg 0,269
Ni 0,089
Ti 0,045
Cu 0,040
Zn 0,040
Sr 0,034
others 0,045
This chemical analysis also shows that the previously suggested heat capacity of molten
aluminum may differ from that of pure aluminum. However, it can be assumed that the
140
Materials and experimental procedure
heat capacity of the tested alloy is considerably above the heat capacity of the air at room
temperature and also above the heat capacity of water at room temperature.
Fig. 3.17: Experimental testing device for quenching test of prismatic refractory
castables in molten aluminum at 800°C
After quenching test, the samples are cooled down up to room temperature. The
aluminum present at the surface of the refractory castable specimen is detached. The
elastic properties of the specimen, namely the Young’s modulus, the shear modulus, the
Poisson’s ratio, as well as the damping of the resonant flexural and torsion frequencies
are assessed after sintering and after each thermal shock cycle through Resonant
Frequency Damping Analysis. This experimental procedure is detailed in
chapter 3.2.8.3.1.
A total of 10 thermal shock cycles are performed. After the 1st, 3rd, 5th, 7th and 10th
thermal shock cycle, 4 samples are put aside in order to measure the retained modulus of
141
Materials and experimental procedure
rupture (MOR) and to examine the evolution of the formulation microstructure with
progressive thermal shock cycles. Starting from 20 samples, 8 samples remain after the
whole set of thermal shock cycles to guarantee representative values of elastic and
mechanical properties.
Thermal shock experiments are also carried out between two high temperatures with the
help of a specific designed furnace composed of two heat chambers (heated by MoSi2
heating elements) separated by a movable gate. This test facility is designed by
Therm-Aix and can work up to 1750°C. The gate can be pneumatically opened and the
samples, which are lying on a movable sample holder, can be transferred from one heat
chamber to the other. After transfer of the samples from one chamber to the other, the
removable gate returns to its initial closed position. An illustration of this high
temperature thermal shock testing device is shown on figure 3.18.
Movable gate
Thermocouple
Sample
Movable
Sample holder
Fig. 3.18.: Illustration of the high temperature thermal shock testing device
In the study hereby, a temperature difference of 600°C is chosen between 1500°C and
900°C. According to such experimental conditions, both effects of the coefficient
mismatch described in chapter 2.3.8 expected to take place during the heating phase and
those resulting from the repeated transformations of zirconia and more particularly after
volume expansion of the zirconia particles during the cooling phase described in
chapter 2.2.2 cause the generation of a crack network in the sample microstructure.
Therefore, HTTS experiments constitute a skilled combination of both heating and
cooling thermal shock conditions. The temperature range (900°C ‒ 1500°C) is the same
as the one illustrated before within the framework of cycling Refractoriness under Load
(chapter 3.2.6.3) and will also be used within the framework of cycling Resonant
142
Materials and experimental procedure
Fig. 3.19: Temperature profile of the high temperature thermal shock experiments
between 1500°C and 900°C
143
Materials and experimental procedure
After the high temperature thermal shock treatment, post mortem analysis is performed
on each sample through Room Temperature Resonant Frequency Damping Analysis
(RT-RFDA) according to the measurement principle described in chapter 3.2.8.3.1 to
assess the resulting elastic properties such as the Young’s modulus and the damping of
the resonant flexural frequency. High Temperature Resonant Frequency Damping
Analysis is also performed on each sample up to 1500°C to record the stiffness evolution
with increasing temperature of the damaged sample according to the temperature profile
described in chapter 3.2.8.3.2. The stiffness evolution with increasing temperature of the
damage sample is compared to the stiffness evolution with increasing temperature of the
sample after sintering.
In the present part, different techniques known in the art for measuring material elastic
properties under static or dynamic procedure are illustrated. Then, Resonant Frequency
Damping Analysis (RFDA) is particularly stressed as this measurement technique is the
one used within the framework of the present study.
The Young’s modulus is typically defined as the slope of the stress to strain curve
obtained during a tensile test in cases on the condition that the material exhibits an elastic
linear behavior. This definition of the Young’s modulus obeys to Hooke’s law as shown
on figure 2.32 [SAL18]. Another static procedure alternative consists in examining the
linear behavior of a specimen during a three-point-bending-strength test [LAT05]
[PET13]. Nevertheless, the measurement of the modulus of elasticity through static
procedure can be called into question as the obtained results depend on the load, the
loading rate and eventual irreversible deformation of the specimen [AND03] [BEL12].
As the strain acquisition during three-point bending tests turns out to be challenging in
spite of a meticulous sample preparation and a sophisticated testing preparation, the focus
is shifted to dynamic procedures [LAT05].
Since the 1950’s, dynamic procedures for Young’s modulus assessment have been
appreciated as non-destructive approaches [LEE98]. The measurement techniques are
based on the treatment of the acoustic signal propagating throughout the specimen. This
sound propagation strongly depends on the stiffness and the density of the material.
144
Materials and experimental procedure
Sound transmission through a specimen can be treated to measure the elastic moduli of
the specimen according to the geometry, the isotropic behavior and the density [LEE03]
[NON98]. Indeed, when a sound wave impacts on the surface of a solid specimen, a part
of the sound wave energy is reflected, another part is absorbed and the last part is
transmitted through the specimen. The experimental setup of this measurement technique
is illustrated on figure 3.20.
Fig. 3.20: Apparatus for in situ dynamic determination of Young’s modulus of refractory
materials up to 1650°C through sound transmission measurements [NON98]
145
Materials and experimental procedure
Fig. 3.21: Experimental setup used for Young’s modulus assessment at high temperature
by Ultrasonic Pulse Echography [BRI08]
Resonant Beam Technique (RBT) consists in exciting a specimen suspended from loops
attached to transducers to bending vibrations. This experimental procedure is based on
measuring velocities of propagating elastic pulses or wave packets or on the
determination of frequencies of resonant vibrations. A sinusoidal signal excitation is
caused by the transmitter unit and the resonant peaks are then detected by numerical
differentiation [PRA04]. The experimental setup of Resonant Beam Technique is
illustrated on figure 3.22.
Fig. 3.22: Experimental setup used for Young’s modulus assessment at high temperature
by Resonant Beam technique with the help of Elastotron 2000 [PRA04]
146
Materials and experimental procedure
Fig. 3.23: Schematic illustration of the Impulse Excitation Technique measurement with
the help of a punctual impact [TOG10]
Fig. 3.24: Schematic illustration of the Impulse Excitation Technique measurement with
the help of an automated tapping device [IMC12]
147
Materials and experimental procedure
Time
Fig. 3.25: Damped sinus function of the resulting acoustic signal of a sample after
mechanical excitation delimited by an exponential decay envelope [HAR14]
148
Materials and experimental procedure
A Fast Fourier Transformation (FFT) applied on the acoustic signal sinusoidal function is
required for the assessment of the specimen fundamental resonant frequencies, namely
the flexural, torsion and longitudinal. The elastic properties of the specimen depend on
those resonant frequencies in addition to the specimen geometry and density
[DIN 12680-1].
First the Young’s modulus is calculated as follows as far as prismatic bar specimen are
concerned [IMC12]:
The term T1 of the equation 3.16 corresponds to a correction factor defined according to
[DIN12680-1] as follows:
𝑡 2 𝑡 4
𝑇1 = 1 + 6,585 ∙ (1 + 0,0752𝜇 + 0,8109𝜇2 ) ∙ ( ) − 0,868 ∙ ( ) − 𝐽
𝐿 𝐿 Equation 3.17
With:
𝑡 4
8,34 ∙ (1 + 0,2023𝜇 + 2,173𝜇2 ) ∙ ( )
𝐽= [ 𝐿 ] Equation 3.18
𝑡 2
1 + 6,338 ∙ (1 + 0,1408𝜇 + 1,536𝜇 ) ∙ ( )
2
𝐿
The Shear Modulus is calculated as follows as far as prismatic bar specimen are
concerned [IMC12]:
149
Materials and experimental procedure
4 ∙ 𝐿 ∙ 𝑚 ∙ 𝑓𝑡2 𝐵
𝐺= ∙[ ]
𝑏∙𝑡 (1 + 𝐴 ) Equation 3.19
𝑏 𝑏 2 𝑏 3
[0,5062 − 0,8776 ∙ ( ) + 0,3504 ∙ ( ) − 0,0078 ∙ ( ) ]
𝑡 𝑡 𝑡
𝐴= Equation 3.20
𝑏 𝑏 2
12,03 ∙ ( ) + 9,892 ∙ ( )
𝑡 𝑡
𝑏 𝑡
+
𝐵= 𝑡 𝑏 Equation 3.21
𝑡 𝑡 2 𝑡 6
4 ∙ ( ) − 2,52 ∙ ( ) + 0,21 ∙ ( )
𝑏 𝑏 𝑏
The calculation of both Young’s modulus and shear modulus leads to an calculation of
the Poisson’s ratio of the material that is iteratively implemented in the equation of the
Young’s modulus and the shear modulus to refine the values of each elastic property
[IMC12]:
𝐸 = 2𝐺 (1 + 𝜇) Equation 3.22
Besides the assessment of the elastic moduli through the acquisition of the fundamental
resonant frequencies of the specimen, Resonant Frequency Damping Analysis enables to
determine the damping behavior of each resonant frequency. The physical notion of
damping describes the exponential decay of the signal amplitude x(t) and the decrease of
the vibration energy Ws(t) with time. Because of damping, the vibration energy gradually
decreases and is transformed into other forms of energy like heat [HAR14].
150
Materials and experimental procedure
Damping ζ is therefore derived from the decay constant of the signal envelope δ of the
corresponding resonant frequency f [ROE98]:
𝛿 Equation 3.23
𝐷𝑎𝑚𝑝𝑖𝑛𝑔 = 𝜁𝑓𝑓 = [-]
(𝜋𝑓𝑓 )
With:
Damping defined in equation 3.23 is related to the decay constant of the flexural resonant
frequency ff.
The logarithmic decrement method is applied on the acoustic signal x(t) to determine the
decay constant δ of the signal envelope. This mathematic approach consists in a
logarithmic conversion of the ratio of two oscillating signal amplitudes xi separated by a
time period Td as illustrated on figure 3.25.
𝑥(𝑡)
ln ( )
𝑥(𝑡 + 𝑇𝑑 )
𝛿= Equation 3.24
𝑇𝑑
In the present section, the experimental procedure with regard to RFDA within the
framework of the present study is more accurately described.
As Impulse Excitation Technique (IET), RFDA is a non-destructive testing method based
on the recording of the resulting acoustic signal of a specimen after mechanical excitation
151
Materials and experimental procedure
According to the propagation of the resonant frequency throughout the prismatic sample,
the position of the automated tapping device required for the acoustic excitation of the
sample and the position of the microphone required for the recording of the resulting
152
Materials and experimental procedure
acoustic signal must be optimized. Figure 3.27 shows the optimized position of the
automated tapping device and that of the microphone to record the flexural frequency, the
torsion frequency or both of them. RT-RFDA measurements are performed on prismatic
samples with geometry 150 mm x 25 mm x 25 mm, 160 mm x 40 mm x 40 mm or
50 mm x 5 mm x 5 mm according to the thermal shock experiment beforehand carried out
(in air, in water, in molten aluminum or at elevated temperatures).
The blue lines of figure 3.27 represent the amplitude of the acoustic signal within the
sample. In order to assess the Young’s modulus, the shear modulus and the Poisson’s
ratio, the room temperature RFDA measurements are performed in the flexural/torsion
mode. In that case, the specimen lays on a sample holder constituted of two thin steel
wires separated by a distance of 0,552 x L from each other (with L: the length of the
prismatic sample) corresponding to the node distance of the flexural wave propagation
throughout the prismatic sample. The wires are placed in agreement with the flexural
mode while the microphone and the automated tapping device are placed in accordance
with the torsion mode.
Wires
Wires
Wires Hammer
Hammer
To guarantee the reproducibility of the RFDA measurements, the distance between the
automated tapping device and the lower surface of the sample shall be 5 mm in order that
the intensity of the mechanical impulsion is the same. The distance between the
microphone and the upper surface of the specimen shall be 5 mm. The sample is carefully
centered place on the sample holder so that the influence of the suspension wire on the
acoustic signal propagation throughout the specimen is limited. The distance between the
end of the specimen and the suspension wire shall be 0,224 x L (with L: the length of the
prismatic sample). By respecting those experimental criteria, the influence of the
suspension wires on the damping behavior of the specimen flexural resonant frequency is
153
Materials and experimental procedure
minimized while the influence of the suspension wires on the damping behavior of the
specimen torsion resonant frequency shall be taken into consideration.
After entering the geometrical parameters of the specimen, the Sonelastic Software type
RFDA MF V 7.0.2 - NLB from IMCE (Integrated Material Control Engineering,
Diepenbeek, Belgium) gives the node distance that shall be taken into account. After
placing the specimen in a centered position of the suspension wires, a mechanical
impulse is generated by the automated tapping device placed 5 mm below one corner of
the prismatic sample. The resulting acoustic signal is recorded by the microphone placed
5 mm over the corner of the specimen, which is diagonally opposite. The frequency
spectrum of the specimen is converted from the acoustic signal by a Fast Fourier
Transformation (FFT). The software implements the sensed flexural and torsion
frequencies in equations 3.16 and 3.19 to calculate the Young’s modulus and the shear
modulus respectively. From those both elastic properties, an iterative algorithm is applied
to give an accurate measurement of the Poisson’s ratio according to equation 3.22.
Finally, the logarithmic decrement method is used for each resonant frequency to assess
its damping value according to equation 3.24.
Figure 3.28 a) shows the sample holder used for the determination of the Young’s
modulus, the shear modulus and the Poisson’s ratio of refractory castable specimen in
flexural/torsion mode, while figure 3.28 b) shows the sample holder used for the
determination of the Young’s modulus of dense ceramic material elaborated by slip
casting in flexural mode.
Fig. 3.28: Experimental configuration for the assessment of the elastic properties of
a) refractory castables (flexural/torsion mode)
b) dense ceramic material (flexural mode)
154
Materials and experimental procedure
Indeed, as far as the dense ceramic specimen elaborated by slip casting are concerned, the
geometry of the sample does not allow the acquisition of the torsion frequency.
Therefore, the flexural mode is applied and an calculation of the Poisson’s ratio shall be
done for a possible calculation of the Young’s modulus. In spite of the difference
chemical compostion of the tested materials of the Al2O3-ZrO2 binary system, the same
Poisson’s ratio of 0,2 [-] is implemented in the calculation algorithm of the Young’s
modulus.
Generally, the room temperature measurements on dense or porous ceramic samples are
conducted in order to restrict the non-linear effects to the unique contribution of the
ceramic material detailed in chapter 2.10.1. The preliminary drying of the sample before
measurement guarantees the absence of thin layer of adsorbed water in the sample
porosity as described in chapter 2.10.3. Moreover, after each measurement, the frequency
spectrum of the sample is analyzed in order to discern any wave modulation or
non-collinear interaction of acoustic beams. Such phenomena can disturb the
measurement of the shear modulus, as the torsion frequency is situated in the high
frequency field, where the harmonics and other wave modulations are supposed to be
located. To obtain representative damping values of the sample, a time of 1 min is
respected between each measurement to limit physical phenomena such as slow
dynamics or non-linear resonance frequency shift detailed in chapter 2.10.2.
Concerning the dense shaped specimen elaborated by slip casting, five measurements in
flexural mode for each specimen are performed to obtain a representative Young’s
modulus of the specimen after each thermal treatment, i.e. sintering process and
155
Materials and experimental procedure
quenching test in water. As 5 samples are measured after each thermal shock cycle, and
as 5 measurements are performed per sample, the mean value of the elastic properties is
assumed to be representative of the whole set of specimen for a given thermal shock
cycle. To avoid the undesirable effects of moisture on the resulting elastic properties of
the material, the samples are placed before each RT-RFDA measurement in a dryer at
110 5°C up to reach a constant mass and the measurement takes place 15 min after
removing the samples from the dryer. Because of the high thermal conductivity of the
dense material and the small prismatic geometry, an idle time of 15 min between the
removal of the samples from the dryer and the room temperature RFDA measurement is
assumed to be sufficient.
The calculated damping values through Resonant Frequency Damping Analysis must not
be confused with the crack flank friction of the specimen. The damping behavior, in
addition to be dependent on the crack flank friction of the material, strongly depends on
the intensity of the impulse delivered by the automated tapping device [MAN09] as well
as on the nature of the sample holder and the specimen size [ROE97]. Therefore, the
interpretation of the evolution of the relative damping behavior is reliable if the initial
damping value is discussed. A material can present a low damping increase after
progressive thermal shock cycles and meanwhile exhibits extremely high absolute
damping values after sintering. And conversely, a material can present a considerable
damping increase after progressive thermal shock cycles and meanwhile exhibits low
absolute damping values after sintering.
156
Materials and experimental procedure
1750°C). HT1750 RFDA test facility is designed to perform high temperature elastic
properties measurements in normal atmosphere or in semi-automatic controlled inert gas
(4 - 6 bar) with argon or nitrogen on two different samples placed on separated sample
holders. Figure 3.29 shows an overview of the experimental set up HT1750 with the
appropriate sample holders.
Sample holders
Electromagnetic automated
tapping device
Fig. 3.29: Overview of the HT1750 experimental set up with the appropriate sample
holders for HT-RFDA measurements
157
Materials and experimental procedure
is generated in the middle of the lower surface of the sample by the automated tapping
device. Contrary to the RT-RFDA test facility where the intensity of the impulsion can be
adjusted according to the distance between the electromagnetic hammer and the sample,
the intensity of the impulsion in the HT1750 test facility is computer controlled, while the
distance between the electromagnetic hammer and the sample is always 5 mm. The
lowerable displaceable furnace unit enclosed the test facility and can be programmed by a
computer controlled temperature profile. Three openings are located in the upper part of
the furnace. Two microphones are plugged into two of them while the thermocouple is
plugged in the third one. The resulting acoustic signal of the sample is guided by an
aluminum oxide tube with a diameter of 5 mm up to the microphone. Like in RT-RFDA
measurements, the distance between the sample and the signal acquisition system is
always 5 mm. Similarly, the distance between the thermocouple and the upper surface of
the sample is also 5 mm in order to measure with more accuracy the temperature
surrounding the sample, which is not as a consequence the temperature inside the sample
like for Refractoriness under Load or Creep in Compression measurements. To limit the
influence of the ambient noise on the acoustic signal quality, a high-pass filter with
frequency threshold of 1000 Hz is used.
One focus of the study hereby is to investigate the influence of the impulse intensity on
the resulting elastic properties of the tested refractory castables Ref, Ca-PSZ, Mg-PSZ
and Y-FSZ. HT-RFDA measurements are performed on prismatic samples with geometry
150 mm x 25 mm x 16 mm (L x b x t). A preliminary study has shown that a sample
height of 16 mm is suitable for the acoustic signal acquisition at elevated temperature.
Therefore, the HT-RFDA tests samples are first ground up to 16 mm thickness for
measurement. As the maximum grain size of the formulation is 3 mm, a square section of
the sample 25 mm x 16 mm (b x t) is still sufficient to guarantee a representative volume
of the castable formulation. HT-RFDA measurements are performed up to 1500°C for 3 h
with a heating rate of 4 K.min-1 and a cooling rate of 2 K.min-1. The temperature profile
of HT-RFDA measurements is shown on figure 3.30. The measurement is repeated on
each representative sample of the castable with different impulse intensities from 20 % to
80 % with a gradual step of 10 %. For the zirconia based castables, i.e. Ca-PSZ, Mg-PSZ
and Y- FSZ, the measurements are performed on different samples per tested impulse
intensity as the transformation of zirconia does not allow any reversible stiffness
evolution during the cooling phase unlike castable Ref. The automated tapping device
excites the sample each 120 s. In combination with the reasonable high heating and
cooling rates, this frequency of excitation of the sample leads to the acquisition of
sufficient measuring points and also to a sufficient idle time between two measurements
for a suitable sample relaxation limiting the slow dynamics effects [OST01]. As high
temperature RFDA measurements focus on the flexural frequency evolution with
158
Materials and experimental procedure
1500°C/3h
4K/min
2K/min
Fig. 3.30: Temperature profile for HT-RFDA measurements to examine the influence of
the impulse intensity on tested castable stiffness with increasing temperature
1500°C/3h
2K/min
4K/min
Fig. 3.31: Temperature profile for HT-RFDA measurements to examine the stiffness
evolution of the tested castables with increasing temperature
159
Materials and experimental procedure
Another aspect of the HT-RFDA measurements treats of the stiffness evolution of the
tested castables Ref, Ca-PSZ, Mg-PSZ and Y-FSZ after sintering. Therefore, the
measurement is performed on prismatic samples with geometry
150 mm x 25 mm x 16 mm (L x b x t). The samples are heated up to 1500°C for 3 h with
a heating and cooling rates of 2 K.min-1 and 4 K.min-1 respectively as shown on
figure 3.31. The intensity of the impulse delivered by the automated tapping system each
60 s is 40 %.
The stiffness evolution of tested refractory castables Ref, Ca-PSZ, Mg-PSZ and Y-FSZ
after thermal shock at elevated temperature between 1500°C and 900°C described in
chapter 3.2.7.4 is examined through HT-RFDA measurements. Therefore, the
measurement is performed on prismatic samples with geometry
150 mm x 25 mm x 16 mm (L x b x t). The samples are heated up to 1500°C for 3 h with
a heating and cooling rate of 4 K.min-1 and 2 K.min-1 respectively as shown on
figure 3.30. The intensity of the impulse delivered by the automated tapping system each
60 s is 30 %. A lower impulse intensity is chosen in comparison with the HT-RFDA
measurements performed on sintered samples to limit the non-linear effects related to the
sample excitation as those samples exhibit a non-negligible crack density.
The impact of the stiffness correction through the implementation of the coefficient of
thermal expansion is examined on a Ca-PSZ sample to stress the importance of the
martensitic transformation on the thermal behavior of the sample and subsequently on its
stiffness. Nevertheless, it is obviously noticeable that both curves follow almost the same
evolution with temperature. If no particular discrepancy is visible during the heating
phase, the corrected Young’s modulus values slightly differ from the original values
during the cooling phase and more particularly when the martensitic transformation of
zirconia takes place. It is worth mentioning that the correction curve shows a sudden
160
Materials and experimental procedure
crack closure during the martensitic transformation resulting in the sudden increase of the
stiffness followed by a drastic decrease of the stiffness due to material damaging
generated by the transformation. The original stiffness evolution only shows the
post-transformation microstructural damage. In case of the study hereby, the influence of
the thermal behavior of the refractory castable on the stiffness correction at elevated
temperature is assumed to be marginal. Therefore, the High Temperature Resonant
Frequency Damping Analysis measurements presented in the study show the stiffness
evolution of the tested castables without correction with regard to the thermal behavior.
The fatigue of the refractory castables Ref, Ca-PSZ, Mg-PSZ and Y-FSZ is studied
through cycling HT-RFDA measurements. The samples are submitted to temperature
change between 1500°C and 900°C with a heating and cooling rate of 2 K.min-1. The
temperature profile is similar as the temperature profile for cycling RuL measurements
defined in figure 3.15 for cycling Refractoriness under Load measurements. In spite of a
different nature of temperature measurement, the objective is to compare the resulting
thermo- mechanical and thermos-elastic behaviors of the tested castables in the same
temperature range [900°C – 1500°C] without additional mechanical stresses in case of
161
Materials and experimental procedure
Within the framework of the examination of the resulting elastic properties of dense
ceramic materials of the Al2O3-ZrO2 binary system, the nature of the quenching test in
water at 400°C generates a low crack density in the samples microstructure. The
combination of dense materials with an apparent porosity close to zero exhibiting a low
crack density constitutes interesting boundary conditions for a trustful treatment of the
non-linear behavior of the tested materials because of the extreme high quality of the
resulting acoustic signal. The non-linearity of the flexural resonant frequency, which is
treated in this part, corresponds to the theoretical decrease of the resonant frequency with
the amplitude increase of the signal and conversely to the increase of the resonant
frequency in the time sinusoidal decay function. As the amplitude of the specimen
acoustic signal is proportional to the amplitude of the mechanical excitation impulse
generated by the automated tapping device, the following studied phenomenon refers to
the non-linear resonance frequency shift presented in chapter 2.10.2. A crack-free
material should exhibit the same resonant frequency in the whole acoustic signal
function. This decrease is observed for damaged samples and is expected to be strongly
dependent on the crack density within the material.
After processing the acoustic response of a specimen, the signal can be split into several
time frames. Each time frame can be considered as individual acoustic response of the
specimen. This tuning of the acoustic signal is provided by an extra non-linear behavior
analysis module also supplied by IMCE (Integrated Material Control Engineering,
Diepenbeek, Belgium) coupled with the Sonelastic Software type
RFDA MF V 7.0.2 - NLB. Each elastic property, namely flexural resonant frequency and
162
Materials and experimental procedure
its respective damping behaviour, can be determined in each time frame composing the
whole acoustic signal response.
The operator controls the size of the measurement window (time frame) and the size of
the measurement step performed in the time frame. For each formulation, the acoustic
signal is divided into 45000 arbitrary units, which are dimensionless. For a possible
results comparison, the size of the time frame and the measurement step are kept constant
for each tested formulation of the Al2O3-ZrO2 binary system and are both of them 400 of
this arbitrary unit. A parameter αf is defined to illustrate the non-linear behaviour of the
flexural resonant frequency and is defined as the ratio between the flexural frequency in
the time frame i divided by the flexural frequency at the beginning of the time sinusoidal
decay function, where the amplitude of the signal is the highest. The definition of
parameter αf is shown on equation 3.25:
𝑓𝑟𝐴𝑚𝑝𝑖
𝛼𝑓 = | | Equation 3.25
𝑓𝑟𝐴𝑚𝑝𝑚𝑎𝑥
With:
A similar coefficient related to the non-linear behavior of the damping of the resonance
flexural frequency in the acoustic signal decay function could also be defined with
exactly the same approach as previously as far as parameter αf is concerned. The
expression of the non-linearity coefficient βf related to the damping of the flexural
resonant frequency is defined in equation 3.26:
𝜁𝐴𝑚𝑝𝑖
𝛽𝑓 = | | Equation 3.26
𝜁𝐴𝑚𝑝𝑚𝑎𝑥
With:
163
Materials and experimental procedure
ζAmp i: Damping of the flexural resonant frequency of the peak point i in the time
sinusoidal signal decay function [-]
ζ Amp max: Damping of the flexural resonant frequency of the peak point in the time
sinusoidal signal decay function with the maximum amplitude [-]
The parameters αf and βf are calculated for each sample after each thermal treatment, i.e.
sintering and quenching test in water at 400°C described in chapter 3.2.7.2. The results
that will be presented refer to the mean value of the parameters αf and βf of the
5 specimen.
The influence of the porosity and more particularly of the pore geometry on the resulting
elastic properties of castables is examined in the study hereby. To observe the eventual
anisotropy of the castables due to a heterogeneous distribution of the pores after
sintering, ultrasound velocity measurements are performed with help of an ultrasonic
measuring instrument type Pundit G.N.S. Instruments London. A piezoelectric signal is
emitted by a transmitter on one square face 40 mm x 40 mm (b x t) of the prismatic
sample with geometry 160 mm x 40 mm x 40 mm (L x b x t). This signal is captured by a
transducer placed on the opposite face of the prismatic sample. The travel time of the
signal through the length of the prismatic sample is measured and noted ts,L.
The ultrasonic velocity of the signal through the prismatic sample is defined as the ratio
of the sample length to the travel time as shown on equation 3.27:
𝐿
𝑣𝐿 =
𝑡𝑠,𝐿 Equation 3.27
With:
From this travel time in the length of the sample, a calculation of the longitudinal
frequency can be realized according to the following equation 3.28.
164
Materials and experimental procedure
1
𝑓𝑙𝑜𝑛𝑔 =
2𝑡𝑠,𝐿 Equation 3.28
With:
2
4. 𝐿2 . 𝑓𝑙𝑜𝑛𝑔 .𝜌
𝐸𝑙𝑜𝑛𝑔 = .𝐾 Equation 3.29
109
With:
𝜋 2 . 𝜇2 . 𝑡 2
𝐾 =1+ Equation 3.30
12. 𝐿2
With:
165
Materials and experimental procedure
Besides the further measurement of the Young’s modulus of the prismatic specimen in
longitudinal mode, ultrasonic measurements are also performed in the width and the
thickness of the sample in function of the nature of the pore forming agent and the
quantity of the pore forming agent. Such measurements can prove the anisotropic elastic
properties of the porous castables, when the evolution of the ultrasound velocity in each
dimensional direction of the sample does not follow the same tendency in function of the
pore forming agent quantity.
The ultrasonic velocity of the acoustic signal through the prismatic sample width and
thickness is respectively defined as the ratio of the sample width or thickness to the travel
time through the width or the thickness of the sample as shown on
equations 3.31 and 3.32:
𝑏
𝑣𝑏 =
𝑡𝑏,𝐿 Equation 3.31
𝑡
𝑣𝑡 =
𝑡𝑡,𝐿 Equation 3.32
With:
vb: Ultrasonic velocity of the acoustic signal through the width of the sample [m.s-1]
b: Width of the sample [m]
tb,L: Travel time of the acoustic signal through the prismatic sample width [s]
vt: Ultrasonic velocity of the acoustic signal through the thickness of the sample
[m.s-1]
t: Thickness of the sample [m]
tt,L: Travel time of the acoustic signal through the prismatic sample thickness [s]
166
Materials and experimental procedure
Figure 3.33 shows the measurement facility for dilatometric measurements prepared for
thermal expansion analysis of two prismatic Ca-PSZ samples.
1 ∆𝐿 𝑇 − ∆𝐿0
𝛼𝑇 = ∙
𝐿0 𝑇 − 𝑇0 Equation 3.33
With:
167
Materials and experimental procedure
Fig. 3.33: Test facility for a double dilatometric measurements prepared for thermal
expansion analysis of two prismatic Ca-PSZ samples
The key focus of the study hereby is to correlate the resulting elastic properties of
ceramic materials after progressive thermal shock cycles and the microstructural
evolution of those ceramic materials. Therefore, an equipment consisting of a microscope
LEO 440i from Oxford Instruments providing a magnification from 30 up to 100 000 and
an integrated energy dispersive X-Ray analysing system (EDX) is used to investigate the
samples microstructure. The Scanning Electron Microscopy (SEM) is performed on
polished samples previously embedded in a resin to prevent the further microstructural
damaging during polishing. The inspection of the microstructure is performed through
secondary electrons (SE) that generate the topographical effects and the back scattering
electrons (BSD) that create the contrast due to different atomic masses of the different
studied elements such as zirconia and alumina. After an appropriate sample preparation,
namely a meticulous fine polishing of the sample surface, the sample is dried at
110 ± 5°C and then vaporized either with carbon or gold in order to form a nanometer
thick conductive layer for a possible energy dispersive X-Ray analysis. The experience is
conducted in high vacuum.
168
Results
4. Results
The results of the study hereby focus on the interpretation of damping behaviour of
ceramic materials through Resonant Frequency Damping Analysis after progressive
thermal shocks. To better understand the friction phenomena induced by cracks in
alumina – zirconia based refractory castables, a first part of the study is dedicated to the
examination of those crack flank friction phenomena in dense Al2O3-ZrO2 binary system
materials. Afterwards, the examination of the damping behaviour after thermal shocks of
alumina – zirconia based refractory castables that constitutes a second part of the study
can be discussed with more accuracy and more relevance. If the centre of gravity of the
study consists in correlating microstructural damage induced by thermal stresses with the
crack flank friction of the studied materials, the relevance of the data acquisition through
room temperature and high temperature Resonant Frequency Damping Analysis is also
investigated through the non-linear effects that are expected to take place in function of
the excitation amplitude required for the signal acquisition and treatment.
In a first part, the emphasis is laid on the examination of the damping behavior in dense
ceramic materials of the Al2O3-ZrO2 binary system. By assuming a negligible apparent
porosity in the following samples elaborated through slip casting, the examination of the
elastic properties in correlation with the microstructure characterization may constitute a
solid data base to better understand the friction phenomena between zirconia and alumina
particles in presence of a low crack network. Prismatic samples with composition
explained in table 3.3 are elaborated by slip casting, sintered at 1550°C for 2 hours and
the resulting elastic properties are determined after sintering and after quenching test in
water as described in chapter 3.2.7.2 through Resonant Frequency Damping Analysis.
Three quenching tests are performed. By studying the whole spectrum of the Al2O3-ZrO2
binary system, this part of the study aims to examine the influence of the nature of the
interface between the particles on the damping behaviour. Indeed, by adding some
zirconia to a pure alumina based formulation, the contact surface between alumina
particles and zirconia particles increases until zirconia becomes the major phase of the
system. Thus, each configuration with regard to the nature of the interface can be
examined: zirconia/zirconia, zirconia/alumina and alumina/alumina.
169
Results
Fig. 4.1: Evolution of the sample bulk density in the Al2O3-ZrO2 binary system with
increasing zirconia content according to DIN EN 993-1
The bulk density of the samples exponentially varies from 3,95 g.cm-3 for the formulation
based on pure alumina to 6,06 g.cm-3 as far as the formulation based on pure zirconia is
concerned. It is worth mentioning that the bulk density of the manufactured samples
based on pure alumina and pure zirconia measured by Archimedes’ principle according
to DIN EN 993-1 is close to the true density of the raw materials determined with the
help of a gas comparison pycnometer according to DIN ISO 8130-2. Those last
mentioned measurements have shown a true density of CT 3000 LS SG and TZ-3YS-E of
3,9542 g.cm-3 ( 0,0002 g.cm-3) and 5,9835 g.cm-3 ( 0,0035 g.cm-3) respectively. The
170
Results
low discrepancy can be explained by the low apparent porosity exhibited by the different
samples because of the manufacturing process. Figure 4.2 shows the evolution of the
apparent porosity in function of the zirconia content in the samples.
Fig. 4.2: Evolution of the sample apparent porosity in the Al2O3-ZrO2 binary system with
increasing zirconia content according to DIN EN 993-1
The aim of this diagram is to show that no tendency can be distinguished from the
zirconia content dependency of the apparent porosity. Then, this random apparent
porosity evolution varies between extreme low values, i.e. 0,2 vol-% and 0,5 vol-%.
On figure 4.1, a linear trend line is added to the experimental bulk density values in order
to show that the measured density values for each binary formulation is lower than their
theoretical density according to the rule of mixtures. As shown on figure 4.2, the apparent
porosity of the tested materials in the whole binary system is almost constant and has a
negligible effect on the bulk density evolution.
The Young’s modulus of each sample of the formulations of the Al2O3-ZrO2 binary
system listed in table 3.3 is determined through Resonant Frequency Damping Analysis
in flexural mode with an estimation of the Poisson’s ratio of 0,20 [-] after sintering and
after quenching test in water according to DIN 51068 at 400°C as explained in
chapter 3.2.7.2. The mean values of the stiffness as well as the standard deviation of the
stiffness of the materials after progressive thermal treatment comprising sintering and
171
Results
quenching test are shown on figure 4.3 in function of the zirconia content. In the
following part, the abbreviation TS refers to Thermal Shock.
Fig. 4.3: Evolution of the Young’s modulus in the Al2O3 – ZrO2 binary system after
progressive quenching tests in water at 400°C according to DIN 51068
After sintering, the Young’s modulus of the tested formulations evolves from 403 GPa
for the pure alumina based composition to 208 GPa for the pure zirconia based
composition. Those measured values are in agreement with the theoretical stiffness value
as far as alumina is concerned (see chapter 2.2.1.0: 400 ± 20 GPa [RIC05] [SAL07]). As
for the composition based on zirconia, the measured Young’s modulus is below the
theoretical Young’s modulus value of partially stabilized zirconia (see chapter 2.2.2:
between 240 and 320 GPa [CAR08]). The decrease of the stiffness is quite linear with
increasing zirconia content and almost coincides with the theoretical evolution of the
material stiffness by taking into consideration the contribution of each phase according to
Voigt’s model [VOI10] except for the composition 50A-50Z that exhibits a higher
Young’s modulus as the expected stiffness according to the slope of the curve as well as
for the formulations rich in zirconia with a zirconia content above 62,5 wt-%.
After the first quenching test in water, an expected decrease of the stiffness is observed in
the whole Al2O3 – ZrO2 binary system. The trend of the curve is quite similar as the one
after sintering with an almost linear decrease of the Young’s modulus in two steps: a first
decrease concerns the formulations characterized by a zirconia content between 0 wt-%
and 87,5 wt-% with respective Young’s modulus values of 341 GPa and 209 GPa,
followed by a more considerable decrease for the tested formulations with a zirconia
content above 87,5 wt-%. The composition based on pure zirconia exhibits a Young’s
172
Results
modulus of 169 GPa. The decrease of the absolute Young’s modulus values appears to be
smaller with increasing zirconia content starting from a stiffness difference of 62 GPa for
100A-0Z and 31 GPa as far as 0A-100Z is concerned. A high Young’s modulus is also
observed for the formulation 50A-50Z after quenching test with a value of 282 GPa,
which is above the linear decay of the stiffness with increasing zirconia content.
After the second quenching test in water, a further stiffness decrease of the tested
formulations is observed for a zirconia content between 0 wt-% and 25 wt-%. The
Young’s modulus of the pure alumina based formulation is 317 GPa: a decrease of
24 GPa is therefore observed. This decrease becomes smaller with increasing zirconia
content: 25 GPa for 87,5A-12,5Z, 7 GPa for 75A-25Z and 2 GPa for the formulation
62,50A-37,50Z. For zirconia content above 25 wt-%, the retained Young’s modulus is
exactly the same as the Young’s modulus after the first thermal shock cycle.
After the third quenching test in water, a further and low decrease of the Young’s
modulus is observed for the formulations 87,5A-12,5Z and 75A-25Z with a stiffness
decrease of 7 GPa and 2 GPa respectively. Concerning the other tested formulations
including 100A-0Z, the Young’s modulus values are exactly the same as the Young’s
modulus values after the second and consequently the first thermal shock. It is worth
mentioning that the Young’s modulus of the formulation 50A-50Z remains considerable
after each thermal treatment. After three thermal shock cycles, the Young’s modulus of
100A-0Z is 319 GPa and 165 GPa for 0A-100Z.
Fig. 4.4: Retained Young’s modulus in the Al2O3 – ZrO2 binary system after progressive
quenching tests in water at 400°C according to DIN 51068
173
Results
Figure 4.4 aims to show the evolution of the retained Young’s modulus after progressive
quenching tests in water with increasing zirconia content. This diagram shows that the
retained Young’s modulus after three quenching tests in water is almost the same for
100A-0Z and 0A-100Z with 79,1 % and 79,3 % respectively. After each thermal shock
cycle, the retained Young’s modulus reveals a peak for the formulation 50A-50Z with a
retained value of around 87,5 %. The further decrease of the stiffness after the second and
the third thermal shock cycle is also visible for the formulations 100A-0Z, 87,5A-12,5Z
and 75A-25Z. The formulations can be divided into two different groups: the
formulations with a zirconia content between 37,5 wt-% and 87,5 wt-% that exhibit a
retained Young’s modulus of around 85 % after the third quenching test and the
formulations based on a low zirconia content from 0 wt-% to 25 wt-% as well as the
formulations with a considerable zirconia content from 93,75 wt-% to 100 wt-% that
exhibit a retained Young’s modulus of around 80 %.
To examine the crack flank friction phenomena at the interfaces of alumina and zirconia
particles generated by the microcracks induced by the quenching test in water, the
damping behaviour of the flexural resonant frequency for each tested formulation is
investigated after sintering and after thermal shock. The mean value of the damping of
the flexural resonant frequency (ζf) in function of the zirconia content is illustrated on
figure 4.5.
Fig. 4.5: Damping evolution of the flexural frequency (ζf) in the Al2O3 – ZrO2 binary
system after progressive quenching tests in water at 400°C according to DIN 51068
174
Results
After sintering, the tested formulations exhibit low damping values in agreement with the
considerable bulk density, the low apparent porosity and the high stiffness. Those
damping values linearly increase with the zirconia content. Thus, the formulation based
on pure zirconia show damping values around four times higher than those of the
formulation based on pure alumina with 1,86.10-4 [-] and 7,51.10-4 [-] respectively.
After the first quenching test in water, the damping values of each formulation increase
as expected. The initiated cracks that propagate throughout the matrix generate crack
flank friction between the crack walls. This increase of the damping behaviour strongly
depends on the zirconia content. Formulations with a zirconia content comprised between
0 wt-% and 75 wt-% exhibit similar damping values from 1,61.10-3 [-] for 100A-0Z to
2,29.10-3[-] for 62,5A-37,5Z. For a higher zirconia content (from 75 wt-%), the damping
values of the formulations drastically increase to reach 4,87.10-3 [-] as far as 0A-100Z is
concerned.
After the second thermal shock cycle, a further increase of the damping is not observed.
Furthermore, the damping values after the second thermal shock cycle are in general
lower than those measured after the first thermal shock. A chaotic and oscillating
behaviour is noticeable. For a zirconia content between 0 wt-% and 62,5 wt-%, the
damping values are fluctuating between 1,24.10-3 [-] for 37,5A-62,5Z and 2,83.10-3 [-] for
87,5A-12,5Z. For a zirconia content of 62,5 wt-% the damping continuously and
drastically increases to reach 3,28.10-3 [-] for 0A-100Z.
After the third thermal shock cycle, a similar oscillating behaviour of the damping is
noticeable, even if the damping values tend to be lower than after the second thermal
shock cycle. A peak of damping is observed for a zirconia content of 12,5 wt-% with a
value of 2,77.10-3 [-]. A part from this peak, the damping is almost constant with values
fluctuating between 1,16.10-3 [-] for 100A-0Z and 1,82.10-3 [-] for 75A-25Z for a zirconia
content comprised between 0 wt-% and 87,5 wt-%. For high zirconia content, a
considerable increase of the damping is noticeable like after the first and the second
thermal shock cycles to reach a damping value of 3,06.10-3 [-] for 0A-100Z.
The expected low crack density, induced by the relative low quenching temperature in
water, that are propagating throughout a dense matrix is favourable for the examination
of the non-linear behaviour of the samples acoustic signal. The processing of both
non-linear behaviour of the flexural resonance frequency and of the damping of the
flexural frequency is explained in chapter 3.2.9. Parameters αf and βf are defined
according to equations 3.25 and 3.26 respectively to investigate the non-linear behaviour
of each formulation of the Al2O3 – ZrO2 binary system with regard to the resonant
175
Results
flexural frequency and its damping respectively. The examination of the non-linear
behaviour of the acoustic signal aims to confirm some observations with regard to the
evolution of the Young’s modulus and the damping of the resonant flexural frequency
after progressive thermal shocks. Moreover, the extreme high sensitivity of the
non-linearity of the acoustic signal may also show other microstructural evolution of
some compositions at early stage of damage.
Figure 4.6 shows the evolution of the non-linear coefficient αf with increasing zirconia
content after sintering and after the first quenching test. The examination of the
non-linear behaviour of the acoustic signal, related to both flexural frequency and
damping of this frequency, results in the calculation of parameters αf and βf with an
extreme high standard deviation. It can be consequently assumed that after the second
thermal shock cycle in water, the crack density within the samples is so high that the
study of the non-linearity of the acoustic signal becomes questionable. As the
discrepancy between the order of magnitude of coefficient αf before and after thermal
shock is meaningful, a logarithmic scale is plotted.
As expected, the formulations of the Al2O3 – ZrO2 binary system exhibit low values of
parameter αf after sintering. This statement is in correlation with the high Young’s
modulus values close to the theoretical stiffness values and the low damping values.
Without microstructure examination, the study of the elastic properties of the
formulations, namely the Young’s modulus, the damping of the resonant flexural
frequency and the non-linearity of this resonant frequency concludes to a crack free
microstructure. Thus, the formulation 75A-25Z reveals a parameter αf of around
0,006 [-], while the formulation 50A-50Z shows a parameter αf of around 0,002 [-].
After the first thermal shock cycle, the parameter αf increases for each formulation as the
microcrack network induced by the thermal stresses disrupts the acoustic signal.
Furthermore, a two-step behaviour is noticeable: for a zirconia content comprised
between 0 wt-% and 87,5 wt-%, the parameter αf oscillates between 0,043 [-] as far as
100A-0Z is concerned and 0,112 [-] for 75A-25Z. Concerning the formulations with a
zirconia content above 87,5 wt-%, the parameter αf drastically increases to reach 0,226 [-]
for the formulation 0A-100Z. By analogy with the damping evolution, the evolution of
the parameter αf after the first thermal shock in water in function of the zirconia content
is similar to the evolution of the damping of the flexural resonant frequency after the
second thermal shock.
176
Results
Fig. 4.6: Evolution of the non-linear coefficient αf related to the flexural resonant
frequency after the first quenching test in water with increasing zirconia content
By analogy with the non-linear behaviour of the resonant frequency, figure 4.7 shows the
evolution of the non-linear coefficient βf related to the damping of the flexural resonant
frequency with increasing zirconia content after sintering and after the first quenching
test.
Because of the higher sensitivity of the damping in comparison with that of the Young’s
modulus assessment, the examination of the non-linear behaviour of the damping appears
to be more challenging than the non-linear behaviour of the flexural resonant frequency
detailed before.
177
Results
After the first quenching test in water, parameter βf reveals an oscillating behaviour for
zirconia content from 0 wt-% to 87,5 wt-% with values between 8,4 [-] for the
formulation 37,5A-62,5Z and 4,9 [-] for the formulation 50A-50Z. It should be noticed
that the parameter βf for some formulations, particularly with a low zirconia content up to
25 wt-%, is lower after the quenching test than after sintering. For a zirconia content
above 87,5 wt-%, parameter βf drastically increases to reach a value of 10,1 [-] for the
formulation 0A-100Z. In general, a chaotic and oscillating behaviour is noticeable for
both parameters αf and βf for zirconia content comprised between 0 wt-% and 87,5 wt-%
after thermal shock. For higher zirconia content up to 100 wt-%, both parameters
drastically and linearly increase with the zirconia content. Such a behaviour is also
observed by the damping evolution after the second thermal shock.
Fig. 4.7: Evolution of the non-linear coefficient βf related to the damping of the flexural
resonant frequency after the first quenching test in water with increasing zirconia content
178
Results
Because of the considerable difference of atomic mass between alumina and zirconia, a
high contrast is noticeable between both phases. Zirconia particles appear light-coloured
and in bright, while alumina particles are dark and black. It is worth mentioning that in
formulations based on both phases, it appears to be challenging in some cases to
distinguish alumina particles with open pores.
In general, a homogeneous distribution of the alumina particles in zirconia based matrix
as far as formulations exhibiting zirconia as the major phase are concerned is noted.
Conversely, a homogeneous distribution of the zirconia particles in alumina based matrix
as far as formulations exhibiting alumina as the major phase are concerned is noticed.
This statement is important for the interpretation of the crack flank friction phenomena as
it can be assumed that the interfaces between alumina and zirconia particles are
homogeneously distributed in the sample microstructure. On figures 4.16, 4.17 and 4.18,
some clusters of dark areas are visible within the bright zirconia matrix. However, those
dark surfaces are obviously constituted of few micrometres size pores locally surrounded
by a high alumina particles concentration.
179
Results
Fig. 4.8: Microstructure overview of the Fig. 4.9: Microstructure overview of the
formulation 100A-0Z after sintering formulation 50A-50Z after sintering
Fig. 4.10: Microstructure overview of the Fig. 4.11: Microstructure overview of the
formulation 25A-75Z after sintering formulation 12,5A-87,5Z after sintering
Fig. 4.12: Microstructure overview of the Fig. 4.13: Microstructure overview of the
formulation 6,25A-93,75Z after sintering formulation 0A-100Z after sintering
180
Results
In spite of the local thermal stresses at the interfaces between the alumina and zirconia
particles during the sintering process, no microcrack is visible in each sample
microstructure. The representative samples of the tested formulations can consequently
be considered as crack free materials exhibiting an extreme low apparent porosity.
Regarding the interfaces between the zirconia and alumina particles, spherical particles of
alumina are embedded in a zirconia rich matrix as far as the formulations with a zirconia
content above 50 wt-% are concerned. On the other part, spherical particles of zirconia
are embedded in an alumina rich matrix as far as the formulations with a zirconia content
below 50 wt-% are concerned. The dark needle-like structures present in the matrix of the
formulations, particularly in the formulations rich in zirconia, are presumably pores due
to the manufacturing process. As for the formulation 50A-50Z, the microstructure is
constituted of homogeneous distributed zirconia and alumina clusters without particular
shape. Thus, 50 wt-% corresponds to a threshold zirconia content value delimiting a
system composed of spherical zirconia particles in an alumina rich matrix and a system
composed of spherical alumina particles in a zirconia rich matrix.
181
Results
Fig. 4.14: Magnification (x 5000) of the Fig. 4.15: Magnification (x 5000) of the
formulation 100A-0Z microstructure after formulation 50A-50Z microstructure after
sintering sintering
Fig. 4.16: Magnification (x 5000) of the Fig. 4.17: Magnification (x 5000) of the
formulation 25A-75Z microstructure after formulation 12,5A-87,5Z microstructure
sintering after sintering
Fig. 4.18: Magnification (x 5000) of the Fig. 4.19: Magnification (x 5000) of the
formulation 6,25A-93,75Z microstructure formulation 0A-100Z microstructure after
after sintering sintering
182
Results
Scanning Electron Microscopy analyses are performed on polished samples of the same
formulations 100A-0Z, 50A-50Z, 25A-75Z, 12,5A-87,5Z, 6,25A-93,75Z and 0A-100Z
after one quenching test in water. Overview of the microstructure of those formulations
are shown on figures 4.20 – 4.25. The aim of the microstructure examination of the tested
samples after quenching is to reveal the density of the cracks induced by the thermal
shock and to show the crack path throughout the sample matrix.
For each formulation a main network composed of two or three cracks is observable. If
the crack density is quite similar in each studied formulation, the main discrepancy lies in
the localization of the cracks. Two different groups of formulations can be distinguished
after the quenching test. Formulations 50A-50Z, 12,5A-87,5Z and 6,25A-93,75Z show
cracks that propagate throughout the sample. Formulations 100A-0Z, 25A-75Z as well as
0A-100Z show cracks that propagate in corners of the sample. Those last mentioned
cracks can also be induced by the propagation of pre-existent cracks during the polishing.
In both cases, the cracks are not isolated within the sample microstructure, as those
cracks are unified.
Concerning the observation of the crack path, in each microstructure the cracks are
mostly propagating in a straight line throughout the sample. A low curvature radius is in
general noticed for each formulation except for the formulation 6,25A-93,75Z that
exhibits a sharp-edged crack unification on the bottom right corner of the sample.
Concerning the other formulations, the cracks do not unify in such a very sharp point
forming several angular fragments, which are more or less associated with each other.
No overall conclusion can be drawn with regard to the influence of the zirconia content
on the crack density and the nature of the crack path after quenching test. The local
damage of the samples containing a high alumina content and the absence of cracks
crossing over the samples from one side to another, as it can be observed for the
formulations containing a high zirconia content, maybe constitutes an indicator that
shows that the damage process of those samples is not completed.
183
Results
Fig. 4.20: Microstructure overview of the Fig. 4.21: Microstructure overview of the
formulation 100A-0Z after one quenching formulation 50A-50Z after one quenching
test in water test in water
Fig. 4.22: Microstructure overview of the Fig. 4.23: Microstructure overview of the
formulation 25A-75Z after one quenching formulation 12,5A-87,5Z after one
test in water quenching test in water
Fig. 4.24: Microstructure overview of the Fig. 4.25: Microstructure overview of the
formulation 6,25A-93,75Z after one formulation 0A-100Z after one quenching
quenching test in water test in water
184
Results
The influence of the zirconia content on the nature of the crack configuration clearly
appears on the SEM pictures of the different formulations after quenching test. The
formulations mainly constituted of alumina exhibit cracks with a saw-tooth profile as
shown on figure 4.26. Such a crack configuration results in an interlocking structure with
a random granular stacking at the level of the crack walls. Such a crack geometry is
conductive to crack flank friction phenomena between the interlocking interface
structures with a mean distance under 1 µm. On the other side, as far as the formulations
containing zirconia as main phase are concerned, the cracks are propagating straight on
throughout the microstructure and the distance between the crack flanks remain almost
the same in the whole crack length. If the content of zirconia phase seems to influence the
crack geometry, it is however challenging to draw a direct conclusion of a zirconia
content dependency of the mean distance between the crack walls. Indeed, formulation
25A-75Z shows a primary crack network with a mean crack width of around 2-3 µm as
shown on figure 4.28, while the formulation 12,5A-87,5Z reveals a primary crack
network with a mean crack width of around 1 µm as shown on figure 4.29. The crack
width can reach a maximum value of around 6-7 µm as far as the formulation 0A-100Z is
concerned as shown on figure 4.31.
As the crack width is in general higher than the mean diameter of the zirconia and
alumina particles, it is assumed that the connection of the cracks corresponds to a crack
unification more than to a crack branching. Nevertheless, on figure 4.29, the main crack
is split into two cracks with a mean crack width almost equal to the half of the crack
width of the main crack. In that case, a crack branching could also be interpreted as
microstructural toughening mechanism. Except for this formulation 12,5A-87,5Z, the
connection of the cracks in the other studied formulations takes place under a
right-angled corner configuration. This microscopic observation also concerns the
formulation 100A-0Z as shown on figure 4.26. Therefore, the nature of the crack
connection does not depend on the zirconia content.
185
Results
Fig. 4.26: Magnification (x 5000) of the Fig. 4.27: Magnification (x 5000) of the
formulation 100A-0Z microstructure after formulation 50A-50Z microstructure after
one quenching test in water one quenching test in water
Fig. 4.28: Magnification (x 5000) of the Fig. 4.29: Magnification (x 5000) of the
formulation 25A-75Z microstructure after formulation 12,5A-87,5Z microstructure
one quenching test in water after one quenching test in water
Fig. 4.30: Magnification (x 5000) of the Fig. 4.31: Magnification (x 5000) of the
formulation 6,25A-93,75Z microstructure formulation 0A-100Z microstructure after
after one quenching test in water one quenching test in water
186
Results
After examining the elastic properties of the Al2O3 – ZrO2 binary system in dense
materials, the following part of this study aims to stress some parameters that have to be
taken into consideration in order to discuss constructively the relevance of elastic
properties in heterogeneous and porous materials such as refractory castables. Those
influencing parameters can be distinguished in two different groups: first, the intrinsic
and anisotropic properties of castables caused by the manufacturing process and the
heterogeneity of the formulation and on the other side the experimental parameters
related to the measurement technique for the generation of the acoustic signal.
Concerning the investigation of the anisotropic properties of castables, in a first part, the
influence of the heterogeneous structure of the vibrating castables is discussed by
examining the local Young’s modulus in function of the sample thickness. In a second
part, the influence of the apparent porosity as well as the pore geometry on the resulting
elastic properties of the porous refractory castable is detailed according to the volume
content and the nature of pore forming agent added to a reference composition.
In order to study the influence of the manufacturing process on the ensuing elastic
properties, the following procedure is carried out. First, prismatic samples of formulation
Ref with geometry 160 mm x 40 mm x 40 mm (L x b x t) are elaborated according to the
manufacturing process detailed in chapter 3.1.2.1. After a sintering process at 1500°C for
6 hours according to the sintering profile 1 illustrated on figure 3.13, five samples are cut
in small pieces by respecting the geometries illustrated on figure 4.32. A thermal shock
test in air according to DIN EN 993-11 at 600°C is performed on five other samples of
the same formulation Ref. A temperature of 600°C is chosen to limit the microstructural
damage of the castables for a relevant calculation of the Young’s modulus, the damping
of the resonant frequency, the non-linearity of the flexural frequency as well as the
non-linearity of the damping of the flexural resonant frequency. After the thermal shock
test, those samples are embedded in a resin and cut into small pieces respecting the
geometries illustrated on figure 3.14. The sample preparation in resin is primordial to
avoid or limit the further microstructural damage related to the cutting operation.
187
Results
Fig. 4.32: Illustration of the measurement grid used in the non-uniform damage test
The elastic properties, namely the Young’s modulus and the flexural damping are
determined through Resonant Frequency Damping Analyses (RFDA) according to
ASTM C 1259- 08 detailed in chapter 3.2.8.1 after sintering and after the thermal shock
cycle on the test sample. Those measurements are performed first on the whole sample
and on each piece, which constitutes the sample after cutting operation. The mean value
of the elastic properties is calculated for each layer A, B, C and D and the following
results essentially focus on the anisotropy of the elastic properties according to the
stacking layer in the densification direction.
Figure 4.33 shows the mean value of the Young’s modulus of each layer A, B, C and D
of formulation Ref after sintering at 1500°C for 6 hours. The stiffness mean value of the
samples after sintering and before the cutting operation is 120 GPa and this mean value is
illustrated by the red line on the figure 4.33.
On figure 4.33, the anisotropy of the elastic behaviour of prismatic samples is illustrated
by the discrepancy of the Young’s modulus between the different layers that compose the
whole sample. Thus, two different groups of layers can be distinguished. First, the upper
layer A of the sample exhibits a lower Young’s modulus than that of the other layers B, C
and D with a value of 106 GPa. This stiffness value is moreover below the Young’s
modulus mean value of the samples (120 GPa). The other layers B, C and D show a
similar Young’s modulus value comprised between 124 GPa for layer C and 128 GPa for
layer B. Those stiffness values are consequently slightly above the mean value of the
Young’s modulus of the samples. It is worth mentioning that the average of the Young’s
modulus of layers A, B, C and D is equal to 121 GPa corresponding to the mean value of
the Young’s modulus of the whole samples.
188
Results
Fig. 4.33: Mean value of Young’s modulus of Ref samples after sintering and after
cutting operation according to measurement grid of figure 4.32
Figure 4.34 shows the mean value of the damping of the flexural resonant frequency for
each layer A, B, C and D of formulation Ref after sintering at 1500°C for 6 hours. The
damping mean value of the samples after sintering and before the cutting operation is
3,06.10-4 [-] and this mean value is illustrated by the red line on figure 4.34.
On figure 4.34, the anisotropy of the elastic behaviour of prismatic samples previously
illustrated by the discrepancy of the Young’s modulus in chapter 4.2.1.1.1 is also
revealed by the discrepancy of the damping behaviour of the flexural resonant frequency
between the different layers that compose the whole sample. The upper layer A of the
sample exhibits a higher damping than that of the other layers B, C and D with a value of
27,23.10-4 [-]. The other layers B, C and D show a similar damping value comprised
betwen 12,39.10-4 [-] for layer D and 13,98.10-4 [-] for layer B. Thus, a damping gradient
is noticeable as the damping values decrease with the distance to the upper surface of the
sample. A direct correlation between the Young’s modulus and the damping of the
flexural resonant frequency is observed: the lower the Young’s modulus, the higher the
damping. The examination of the damping behaviour after sintering leads to the
conclusion of a homogeneous microstructure concerning the lower layers B, C and D of
the refractory castable samples and a microstructure of the upper layer that shows a lower
stiffness and a higher defect concentration. However, the damping values of each layer
189
Results
are much higher than the damping value of the whole sample before the cutting
operation. This can be caused by a further microstructural damage during the cutting
operation in spite of the experimental precautions.
Fig. 4.34: Mean value of the flexural resonant frequency damping of Ref samples after
sintering and after cutting operation according to measurement grid of figure 4.32
The expected low crack density, induced by the relative low thermal shock temperature
in air at 600°C, may be favourable for the examination of the non-linear behaviour of the
samples acoustic signal. The processing of both non-linear behaviours of the flexural
resonance frequency and of the damping of the flexural frequency is explained in
chapter 3.2.9. Parameters αf and βf are defined according to equation 3.25 and 3.26 to
investigate the non-linear behaviour of each layer of Ref samples as well as of the whole
samples with regard to the resonant flexural frequency and its damping respectively. The
examination of the non-linear behaviour of the acoustic signal aims to confirm some
observations with regard to the evolution of the Young’s modulus and the damping of the
resonant flexural frequency after thermal shock in air. Moreover, the extreme high
sensitivity of the non- linearity of the acoustic signal may also show other microstructural
evolution according to the distance to the annealing medium at early stage of damage.
190
Results
The mean values of the coefficients αf and βf related to the non-linear behaviour of the
flexural resonance frequency and the damping of the flexural resonance frequency
respectively, for each layer A, B, C and D as well as for the whole sample after sintering
and before the cutting operation are illustrated on figure 4.35. As the discrepancy
between the order of magnitude of coefficients αf and βf is meaningful, a logarithmic
scale is plotted.
The non-linear behaviour of the flexural resonant frequency is extreme low after
sintering. The values of the coefficient αf are comprised between 0,001 [-] for the layer B
and 0,008 [-] for the layer A. Contrary to the Young’s modulus and the flexural damping,
no tendency is noticeable with regard to the evolution of the non-linear coefficient αf with
the distance to the upper surface of the sample. If the high value of parameter αf for the
layer A is in good correlation with the low value of the Young’s modulus and the high
value of damping, it should be noted that the parameter αf is minimum for the layers B
and C with values of 0,001 [-] and 0,002 [-] respectively but the layer D that exhibits
comparable Young’s modulus and damping values to those of the layers B and C shows a
high value of αf (0,005 [-]). The parameter αf of the whole samples is equal to 0,005 [-]
which is higher than the average value of the parameter αf of each layer (0,004 [-]).
Fig. 4.35: Mean value of non-linear coefficients αf and βf of Ref samples after sintering
and after cutting operation according to measurement grid of figure 4.32
The non-linear behaviour of the flexural resonant frequency damping is low after
sintering but nevertheless more considerable than the non-linear behaviour of the flexural
resonance frequency. The values of βf can be extremely low such as the βf-value of the
191
Results
layer B (0,620 [-]) or very high like the βf-value of the layer A (7,835 [-]). By analogy
with the parameter αf, no tendency is noticeable concerning the parameter βf according to
the sample thickness. If the high value of parameter βf for the layer A is in good
correlation with the low value of the Young’s modulus and the high value of damping,
the layers C and D in spite of high Young’s modulus values and low damping values also
show high βf-values (4,152 [-] and 2,914 [-] respectively. The parameter βf of the whole
samples is equal to 5,544 [-] which is higher than the average value of the parameter βf of
each layer (3,880 [-]).
Figure 4.36 shows the mean value of the Young’s modulus of each layer A, B, C and D
of formulation Ref after a thermal shock test in air at 600°C. The stiffness mean value of
the whole samples after thermal shock and before the cutting operation is 112 GPa and
this mean value is illustrated by the red line on the figure 4.36.
The thermal shock test at 600°C induces as expected little microstructural damage as the
Young’s modulus of the samples is 120 GPa before the thermal shock and 112 GPa after
thermal shock, i.e. a decrease of 7 %. A decrease of the Young’s modulus is observed for
each layer. However the amplitude of the decrease is not similar for each layer. The layer
A reveals a high decrease of the stiffness from 106 GPa before thermal shock to 86 GPa
after thermal shock, i.e. a relative decrease of 19 %. This degradation of the elastic
properties of layer A can be explained by the local high thermal gradient as this upper
layer of the samples is first in contact with the annealing medium (compressed air). The
layers B and C reveal a low decrease of the stiffness from 128 GPa before thermal shock
to 115 GPa after thermal shock for layer B, i.e. a relative decrease of 10 % and from
124 GPa before thermal shock to 112 GPa after thermal shock for layer C, i.e. a relative
decrease of 9 %. This shows that during such a thermal shock test in air, only the upper
layer of the sample is damaged and therefore the crack network may be localised in the
upper layer of the samples. The layer D also reveals a high decrease of the stiffness, like
layer A, from 126 GPa before thermal shock to 82 GPa after thermal shock, i.e. a relative
decrease of 35 %. This degradation of the elastic properties of layer D can be explained
by the local high thermal gradient as this lower layer of the samples is in contact with the
steel plate that is also able to quickly conduct the thermal flux coming from the hot
samples. Nevertheless, a so high decrease of the stiffness of the lower layer of the
samples is not expected in this order of magnitude. It is worth mentioning that the
average of the Young’s modulus of layers A, B, C and D is equal to 99 GPa. This mean
value of the material stiffness does not correspond anymore to the mean value of the
Young’s modulus of the whole samples equal to 112 GPa as it could be the case after the
192
Results
sintering. It could be concluded that the cutting operation may induce some
microcracking in the material microstructure that already exhibits some cracks induced
by the thermal shock experiment.
Fig. 4.36: Mean value of Young’s modulus of Ref samples after thermal shock and after
cutting operation according to measurement grid of figure 4.32
Figure 4.37 shows the mean value of the damping of the flexural resonant frequency for
each layer A, B, C and D of formulation Ref after thermal shock test. The damping mean
value of the whole samples after thermal shock and before the cutting operation is
4,82.10-4 [-] and this mean value is illustrated by the red line on the figure 4.37.
As expected, the damping value of the samples flexural resonant frequency increases
after thermal shock. This value was 3,06.10-4 [-] after sintering and the damping value
after thermal shock reaches 4,82.10-4, i.e. an increase of 125 % is observed as far as the
damping behaviour of the whole samples before cutting operation is concerned.
However, the increase of the damping is only observed for the layer D with an initial
value before thermal shock of 12,39.10-4 [-] and a value of 16,31.10-4 [-] after thermal
shock, i.e. an increase of 32 % is observed for the layer D. The increase of the damping
of layer D is in good correlation with the considerable decrease of the Young’s modulus
of this same layer. Concerning the layer A that also exhibits a considerable decrease of
the stiffness, no damping increase is observed. Indeed the damping value of the layer A
was initially 27,23.10-4 [-] and is 20,59.10-4 [-] after thermal shock. As far as the layers B
193
Results
and C are concerned, the damping values are almost constant after thermal shock with an
initial damping value of 13,98.10-4 [-] before thermal shock and a value of 13,34.10-4 [-]
after thermal shock for layer B and an initial value of 12,87.10-4 [-] before thermal shock
and a value of 12,77.10-4 [-] after thermal shock for layer C.
Fig. 4.37: Mean value of the flexural resonant frequency damping of Ref samples after a
thermal shock in air at 600°C and after cutting operation according to measurement grid
of figure 4.32
A direct correlation between the Young’s modulus and the damping of the flexural
resonant frequency is nevertheless observed: the lower the Young’s modulus, the higher
the damping. However, layer D should exhibit the highest damping values as the relative
Young’s modulus decrease of this layer is the most considerable after thermal shock. The
examination of the damping behaviour after thermal shock in air leads to the conclusion
of a microstructure microcracking in the lower surface of the samples that generate crack
flank friction phenomena. Microstructure micracracking that takes place in the upper
layer of the sample does not generate crack flank friction phenomena. This can be
explained by the nucleation of large cracks. The layers B and C revealing a low stiffness
decrease and constant damping evolution show a lesser microstructural damage.
However, the damping values of each layer are much higher than the damping value of
the whole sample before the cutting operation. This can be caused by a further
microstructural damage during the cutting operation in spite of the experimental
precautions. Therefore, the conclusions from damping measurements should be called
into question.
194
Results
The mean values of the coefficients αf and βf related to the non-linear behaviour of the
flexural resonance frequency and the damping of the flexural resonance frequency
respectively for each layer A, B, C and D as well as for the whole sample after thermal
shock and before the cutting operation are illustrated on figure 4.38. As the discrepancy
between the order of magnitude of coefficients αf and βf is meaningful, a logarithmic
scale is plotted.
The non-linear behaviour of the flexural resonant frequency is still extremely low after
thermal shock. The values of αf are between 0,017 [-] for the layer C and 0,182 [-] for the
layer A. No evolution of the parameter αf is observed for the whole sample after thermal
shock with an initial value of 0,005 [-] and a resulting value of 0,005 [-], while an
increase of the non-linear behaviour is expected because of the microstructure
microcracking. However, an increase of parameter αf is observed for each layer. Such an
increase can also be caused by the further damage resulting from the cutting operation.
Concerning the layers A and B, the relative increase of the parameter αf is quite similar:
the increase of parameter αf for the layer A is 2062 % from 0,008 [-] before thermal
shock to 0,182 [-] after thermal shock and the increase of parameter αf for the layer B is
2367 % from 0,001 [-] before thermal shock to 0,022 [-] after thermal shock. Concerning
the layers C and D, the relative increase of the parameter αf is much lower: the increase
of parameter αf for the layer C is 900 % from 0,002 [-] before thermal shock to 0,017 [-]
after thermal shock and the increase of parameter αf for the layer D is 465 % from
0,005 [-] before thermal shock to 0,029 [-] after thermal shock.
By analogy with the damping, the non-linear behaviour of the flexural resonant frequency
follows a first decrease from the layer A to layer C and increases to the lower part of the
samples. A direct correlation between the Young’s modulus, the damping of the flexural
resonant frequency and the parameter αf is observed: the lower the Young’s modulus, the
higher the damping and the higher the parameter αf. However, layer D should exhibit the
highest parameter αf value as the relative Young’s modulus decrease of this layer is the
most considerable after thermal shock. The examination of the non-linear behaviour of
the flexural resonant frequency after thermal shock in air leads to the same conclusion as
the previous one drawn from the examination of the flexural damping, i.e. microstructure
damage take place preferentially in the lower and upper surfaces of the samples.
195
Results
of 5,5 [-] and a resulting value of 8,1 [-], i.e. an increase of 45 %. Moreover an increase
of parameter βf is observed for each layer. Such an increase can also be caused by the
further damage resulted from the cutting operation. No tendency is to be observed related
to the relative increase of parameter βf with the different layers. Indeed this relative
increase can be considerable, like for layer B with an increase of 54 628 % from 0,1 [-]
before thermal shock to 28,0 [-] after thermal shock. In second position, the layer D
shows an important increase of 861 % from 2,9 [-] before thermal shock to 28,0 [-] after
thermal shock. Layer A shows a less considerable increase of 414 % from 7,8 [-] before
thermal shock to 40,2 [-] after thermal shock and layer C shows the lowest increase
(193 %) of parameter βf from 4,2 [-] before thermal shock to 12,2 [-].
Fig. 4.38: Mean value of non-linear coefficients αf and βf of Ref samples after thermal
shock in air at 600°C and after cutting operation according to measurement grid
of figure 4.32
A direct correlation between the Young’s modulus, the damping of the flexural resonant
frequency and the parameter αf and βf is observed: the lower the Young’s modulus, the
higher the damping and the higher the parameter αf and βf. However, layer D should
exhibit the highest parameter βf value as the relative Young’s modulus decrease of this
layer is the most considerable after thermal shock. The examination of the non-linear
behaviour of the flexural resonant frequency damping after thermal shock in air leads to
the same conclusion as the previous one drawn from the examination of the flexural
damping and the non-linear behaviour of the flexural resonance frequency, i.e. a
preferential microstructure microcracking takes place in the lower and upper surfaces of
the samples. A gradual microstructural damage is supposed to take place form the layer A
196
Results
to the layer C while the microstructural damage of the lower layer D is of the same order
of magnitude as the one occurring in the upper surfaces of the samples in direct contact
with the annealing medium. This can be caused by the severe thermal transfer between
the samples and the steel plate.
Scanning Electron Microscopy analyses are performed on polished samples of each layer
A, B, C and D after thermal shock test in air at 600°C. Overviews of the microstructure
of the sample layers are shown on figures 4.39 – 4.42. The aim of the microstructure
examination of the tested samples after thermal shock is to reveal the density of the
cracks induced by the thermal shock, to show the crack path throughout the sample
matrix and to show the material microstructure in undamaged area in order to study the
evolution of the material microstructure in the casting direction.
The examination of the microstructure of the different sample layers if the crack
networks is not taken into consideration shows a noticeable porosity gradient. Indeed, the
pore concentration increases in the microstructure form the layer D to the layer A. This
reveals the inherent problems related to the manufacturing of vibrating castables.
According to the rheological properties of the mixture, the air containing in the mixtures
can more or less easily move towards the upper surfaces of the samples to be withdrawn
from the sample. Moreover, the capillary forces can cause the transport of fine materials
in direction of the upper surfaces of the samples. Both phenomena can explain the more
considerable heterogeneous structure of the upper layer A of Ref samples and more
generally the formation of a more and more anisotropic structure of the samples when the
manufacturing process does not fit the rheological properties of the mixture.
Concerning the evolution of the crack density in function of the sample layer, a dense
crack network is visible in the microstructure of the layer A shown on figure 4.39. Those
cracks can appear very large on the figure and reveal the severe thermal transfer that the
upper layer A is submitted during the thermal shock. Those cracks are mainly
propagating straight on throughout the matrix and radially to the big pores until
continuing its crack path beyond the pores.
197
Results
Fig. 4.39: Microstructure overview of the layer A of a Ref sample after a thermal shock
in air at 600°C – Crack meandering through pore network
Fig. 4.40: Microstructure overview of the layer B of a Ref sample after a thermal shock
in air at 600°C – dense microcrack network
Fig. 4.41: Microstructure overview of the layer C of a Ref sample after a thermal shock
in air at 600°C – Network of cracks with small crack width
198
Results
Fig. 4.42: Microstructure overview of the layer D of a Ref sample after a thermal shock
in air at 600°C – Crack free microstructure
A similar but lesser dense crack network can be observed with obviously great difficulty
in the microstructure of the layer C as shown on figure 4.41. Indeed, like for the layer B,
cracks with a characteristic low crack width are propagating throughout the
microstructure and radially to the pores. The crack density is lesser important than in the
microstructure of the layer B. And the cracks are stopped when they reach a pore.
The apparent porosity and the bulk density of the small pieces constituting the samples
are measured according to Archimedes’ principle and DIN EN 991-1. The results are
summarized in table 4.1.
199
Results
The examination of the microstructure of the different sample layers is verified through
the measurement of the apparent porosity and the bulk density of each layer. Indeed, a
porosity gradient is observed in the casting direction from the bottom of the sample to the
top with an apparent porosity of 18,0 vol-% and a bulk density of 2,95 g.cm-3 for layer D
up to an apparent porosity of 19,2 vol-% and a bulk density of 2,95 g.cm-3 for layer A.
Table 4.1: Apparent porosity and bulk density of the layers A, B, C and D of Ref
samples after sintering at 1500°C for 6 hours according to Archimedes’
principle (DIN EN 991-1)
Apparent porosity [vol-%] Bulk density [g.cm-3]
Layer A 19,2 0,2 2,91 0,01
Layer B 19,0 0,2 2,92 0,01
Layer C 18,7 0,1 2,94 0,01
Layer D 18,0 0,1 2,95 0,01
As shown on figure 4.43, the layer A of Ref sample exhibits a primary crack network,
which is characterized by a considerable crack width up to around 50 µm. Widespread
ramifications of large cracks are noticeable on this figure. Such cracks reveal a crack
width of around 20 µm. In addition to those large cracks, some microcracks are visible in
the surrounding areas of the large cracks. Those last mentioned cracks show a low crack
width below 10 µm with an order of magnitude of 1 µm. Independently of the crack
width, all those cracks are propagating straight on throughout the matrix with a constant
distance between the crack flanks. To conclude, a dense crack network is observable in
the sample matrix explaining the low stiffness of the layer A. If the large cracks with a
considerable crack width may not be subjected to crack flank friction phenomena, the
secondary crack network composed of microcracks may cause the enhancement of crack
flank friction phenomena.
On figure 4.44, a considerable crack network in terms of crack density is observable.
Contrary to the layer A, the layer B exhibits a primary crack network with a characteristic
crack width of around 10 µm. If a lower decrease of the crack density can be noticed in
comparison with layer A, the widening of the cracks is not so considerable. Such cracks
are propagating straight on throughout the matrix with a constant gap between the crack
200
Results
Fig. 4.43: Magnification (x 500) of the layer A of a Ref sample after a thermal shock in
air at 600°C – Crack branching
201
Results
Fig. 4.44: Magnification (x 500) of the layer B of a Ref sample after a thermal shock in
air at 600°C – Straightforward propagation of the crack with constant crack width
Fig. 4.45: Magnification (x 500) of the layer C of a Ref sample after a thermal shock in
air at 600°C – Crack propagation through a macro-pore
Fig. 4.46: Magnification (x 500) of the layer D of a Ref sample after a thermal shock in
air at 600°C – Low crack density surrounding porous area
202
Results
In a first part, the evolution of the apparent porosity of the castables Ref-sph, Ref-cyl and
Ref-fib determined according to DIN EN 993-1 in function of the volume addition of the
pore forming agent is examined to verify if such an addition does not cause any
non-linear evolution of the apparent porosity. Indeed, unsuitable rheological properties of
the mixtures containing a high quantity of pore forming agent may cause a non-linear
increase of the apparent porosity and a enhancement of the material anisotropy. On
figure 4.47, the evolution of the apparent porosity is illustrated for each pore forming
agent geometry.
Figure 4.47 reveals a linear increase of the apparent porosity with increasing volume
addition of pore forming agent. Starting from an apparent porosity of 18,4 vol-% for the
reference formulation, the apparent porosity increases to reach 32,6 vol-% for an addition
of 5 vol-% of fibre-like pore forming agent, 32,4 vol-% for an addition of 10 vol-% of
spherical pore forming agent and 30,5 vol-% for an addition of 10 vol-% of cylindrical
pore forming agent. The increase of the apparent porosity is almost similar for the
spherical and cylindrical pore forming agents, while the increase of the apparent porosity
is about twice as much as considerable as far as the fibre-like pore forming agent is
concerned. The evolution of the apparent porosity does not reveal any experimental
problem, which could impede the interpretation of the resulting elastic properties.
203
Results
Fig. 4.47: Evolution of the apparent porosity of Ref-sph, Ref-cyl and Ref-fib according to
the volume addition of pore forming agent
To examine the anisotropic behaviour of the studied castables after pore forming agent
addition, ultrasonic measurements are performed in the length, the width and the
thickness of the samples according to the measurement principle described in
chapter 3.2.10. In case of isotropic microstructure, the samples should reveal a similar
increase of the ultrasound travel time with the addition of the pore forming agent in each
direction, i.e. length, width and thickness.
In case of the composition Ref-sph, a two-step increase of the travel time is observed in
each direction of the samples as shown on figure 4.48. In an apparent porosity range from
18,4 vol-% to 28,2 vol-%, a first low increase is noticeable and this increase is more
considerable up to the formulation Ref-sph-10 exhibiting an apparent porosity of
32,4 vol-%. Even if the discrepancy is significant for the porosity value of 32,4 vol-%
with a relative travel time of 154 % in the width, 143 % in the length and 140 % in the
thickness, it is worth mentioning that the trend of evolution for each direction is identical.
This statement is in agreement with the spherical geometry of the pores. For high
porosity values, the higher relative travel time in the width than that in the thickness and
the length may lead to the conclusion of the presence of prolate spheroid pores slightly
stretched in the width direction.
204
Results
In case of the composition Ref-cyl, a three-step increase of the travel time is observed in
each direction of the samples as shown on figure 4.49. In an apparent porosity range from
18,4 vol-% to 19,2 vol-%, a first considerable increase is noticeable. This increase is
lower up to an apparent porosity of 28,7 vol- %. Then, the travel time slightly decreases
up to an apparent porosity of 29,5 vol-% and finally increases to reach its maximum
values for the formulation Ref-cyl-10 exhibiting an apparent porosity of 30,5 vol-%.
Contrary to Ref-sph, formulation Ref-cyl exhibits a significant discrepancy of the travel
time in function of the sample dimension with a relative value of 170 % for the thickness
vs 146 % and 139 % for the width and the length respectively as far as Ref-cyl-10 is
concerned. For high porosity values, the higher relative travel time in the thickness than
that in the width and the length may lead to the conclusion of the presence of prolate
spheroid pores slightly stretched in the thickness direction. Ultrasonic measurements tend
to show a preferential direction of the cylindrical pores with a height parallel to the
casting direction.
Fig. 4.48: Evolution of the relative ultrasound travel time in the length, the width and the
thickness of Ref-sph
205
Results
Fig. 4.49: Evolution of the relative ultrasound travel time in the length, the width and the
thickness of Ref-cyl
In case of the composition Ref-fib, a chaotic increase of the travel time is observed in
each direction of the samples as shown on figure 4.50. In an apparent porosity range from
18,4 vol-% to 25,3 vol-%, a first considerable increase is noticeable followed by a similar
increase up to an apparent porosity of 29,6 vol-%. Then, the travel time slightly increases
to reach its maximum values for the formulation Ref-cyl-10 exhibiting an apparent
porosity of 32,6 vol-%. By analogy with Ref-cyl, formulation Ref-fib exhibits a
significant discrepancy of the travel time in function of the sample dimension with a
relative value of 181 % for the thickness vs 155 % and 144 % for the width and the
length respectively as far as Ref-cyl-10 is concerned. Like previously, for high porosity
values, the higher relative travel time in the thickness than that in the width and the
length may lead to the conclusion of the presence of prolate spheroid pores slightly
stretched in the thickness direction of the sample. Ultrasonic measurements tend to show
a preferential direction of the fibre-like pores with a height parallel to the casting
direction.
206
Results
Fig. 4.50: Evolution of the relative ultrasound travel time in the length, the width and the
thickness of Ref-fib
After examining the anisotropy of the castables through ultrasonic measurements, the
evolution of the Young’s modulus with increasing apparent porosity is studied in this
chapter. RFDA measurements are performed in flexural mode on Ref-sph, Ref-cyl and
Ref-fib samples with geometry 160 mm x 40 mm x 40 mm (L x b x t) according to
ASTM C 1548-02 as detailed in chapter 3.2.8.3 The resulting modulus of elasticity after
addition of pore forming agent is illustrated on figure 4.52.
A clear pore configuration dependency is noticeable on figure 4.51. Indeed, if each tested
formulation exhibits a similar decreasing trend with increasing porosity, a significant
discrepancy regarding the materials stiffness is visible.
Concerning Ref-sph, a two-step decrease of the Young’s modulus is observed with a first
decrease in an apparent porosity range comprised in a range from 18,4 vol-% to
28,2 vol-% while the stiffness decreases from 134 GPa to 79 GPa. A considerable drop of
the stiffness occurs to reach a Young’s modulus value of 63 GPa for an apparent porosity
of 29,0 vol-%. The modulus of elasticity continues to weaken up to reach a minimum
value of 48 GPa for an apparent porosity of 32,4 vol-%. In the whole porosity range from
18,4 vol-% to 32,4 vol-%, Ref-sph exhibits higher stiffness values than Ref-cyl and
Ref-fib.
207
Results
Fig. 4.51: Evolution of the Young’s modulus of Ref-sph, Ref-cyl and Ref-fib in function
of the apparent porosity
As far as Ref-cyl is concerned, the Young’s modulus first drastically decreases from the
initial value of 134 GPa for an apparent porosity of 18,4 vol-% to reach 106 GPa for an
apparent porosity of 19,2 vol-%. The material stiffness remains almost constant up to an
apparent porosity value of 22,2 vol-% with a Young’s modulus of 102 GPa. A linear
decrease of the Young’s modulus is observed in the porosity range from 22,2 vol-% to
30,5 vol-% to reach a minimum value of 50 GPa. In this porosity range, the Young’s
modulus values of Ref-cyl are above those of Ref-fib.
The Young’s modulus of Ref-fib considerably decreases form the initial value of
134 GPa for an apparent porosity of 18,4 vol-% to 95 GPa for an apparent porosity of
21,3 vol-%. The Young’s modulus continuously and quite linearly decreases to reach a
minimum value of 37 GPa for an apparent porosity of 32,5 vol-%. In the porosity range
from 21,3 vol-% to 32,5 vol-%, Ref-fib exhibits the lowest Young’s modulus values
among the different tested formulations.
If the stiffness difference is not considerable for high porosity values above 29,0 vol-%,
this discrepancy is significant in a porosity range from around 24 vol-% to 28 vol-%. For
instance, Ref-sph exhibits a Young’s modulus of 79 GPa for an apparent porosity of
28,2 vol-% while Ref-cyl shows a Young’s modulus of 60 GPa for an apparent porosity
of 28,7 vol-% and Ref-fib reveals a Young’s modulus of 57 GPa for an apparent porosity
of 27,7 vol-%, i.e. the stiffness difference can reach around 20 GPa for a similar apparent
porosity because of the pore configuration.
208
Results
Among all the models related to the porosity dependency of the Young’s modulus
presented in chapter 2.4.3, Boccaccini proposes a model taking into account the volume
fraction of the pores and the pore shape as shown on equation 2.52. As the pore shape
significantly influences the stiffness evolution of the tested formulations, Boccaccini’s
model is applied on Ref-sph, Ref-cyl and Ref-fib. The Young’s modulus value of the
crack free material is assumed to be 390 GPa as the castable is mainly composed of
alumina. The results are illustrated on figures 4.52, 4.53 and 4.54 for Ref-sph, Ref-cyl
and Ref-fib respectively.
Fig. 4.52: Evolution of the stiffness of Ref-sph with increasing porosity and Boccaccini’s
model applied on Ref-sph formulation
Figure 4.52 shows the evolution of the Young’s modulus of Ref-sph with increasing
porosity. Boccaccini’s model is applied on the experimental values with a Young’s
modulus of the crack free material assumed to be close to the Young’s modulus of
alumina (390 GPa). The model consists in defining a suitable shape factor z/x to
minimize the sum of squared deviations between the modelled stiffness and the
experimental values of the Young’s modulus. Three different possibilities can be
distinguished in function of the shape factor z/x:
If z/x is below 1 [-], the pore is assumed to be an oblate spheroid, i.e. the pore is
stretched perpendicularly to the stress direction
If z/x is equal to 1 [-], the pore is assumed to be a sphere
If z/x is above 1 [-], the pore is assumed to be a prolate spheroid, i.e. the pore is
stretched in the stress direction
209
Results
The model applied in the whole porosity range from 18,4 vol-% to 32,4 vol-% fits as
good as possible the evolution of the experimental points for a z/x ratio of 2,89 [-]. In
spite of the addition of spherical pore forming agents, the pores are according to this
model assumed to be prolate, i.e. stretched in the casting direction. This observation is in
disagreement with the observation following the ultrasonic measurement leading to the
presence of pores slightly stretched in the width direction of the sample.
Fig. 4.53: Evolution of the stiffness of Ref-cyl with increasing porosity and Boccaccini’s
model applied on Ref-cyl formulation
Figure 4.54 shows the evolution of the Young’s modulus of Ref-cyl with increasing
porosity. Boccaccini’s model is applied on the experimental values with a Young’s
modulus of the crack free material assumed to be close to the Young’s modulus of
alumina (390 GPa).
The model applied in the whole porosity range from 18,4 vol-% to 30,5 vol-% fits as
good as possible the evolution of the experimental points for a z/x ratio of 3,43 [-]. In
accordance with the addition of cylindrical pore forming agents, the pores are according
to this model assumed to be prolate, i.e. stretched in the casting direction. Moreover,
those pores are more stretched than the assumed pore geometry obtaining after addition
of spherical pore forming agent. This observation is in agreement with the observation
following the ultrasonic measurement leading to the presence of pores stretched in the
casting direction of the sample.
210
Results
Fig. 4.54: Evolution of the stiffness of Ref-fib with increasing porosity and Boccaccini’s
model applied on Ref-fib formulation
Figure 4.54 shows the evolution of the Young’s modulus of Ref-fib with increasing
porosity. Boccaccini’s model is applied on the experimental values with a Young’s
modulus of the crack free material assumed to be close to the Young’s modulus of
alumina (390 GPa).
The model applied in the whole porosity range from 18,4 vol-% to 32,5 vol-% fits as
good as possible the evolution of the experimental points for a z/x ratio of 3,79 [-]. In
accordance with the addition of fibre-like pore forming agents, the pores are according to
this model assumed to be prolate, i.e. stretched in the casting direction. Moreover, those
pores are more stretched than the assumed pore geometry obtaining after addition of
spherical pore forming agent or cylindrical pore forming agent. This observation is in
agreement with the observation following the ultrasonic measurement leading to the
presence of pores stretched in the casting direction of the sample.
By analogy with the ultrasonic measurements, Boccaccini’s model shows a preferential
distribution of the pores in the samples microstructure. Indeed the pores are assumed to
be stretched in the casting direction to form prolate spheroid. The stretch of the spheroid
pore is gradually more significant from the formulation Ref-sph up to the formulation
Ref-fib.
211
Results
After examining the influence of the apparent porosity and the pore shape on the material
stiffness, the evolution of the damping behaviour of formulations Ref-sph, Ref-cyl and
Ref-fib is shown on figure 4.55. An enhancement of the damping values is expected to
occur with increasing porosity as porous areas contribute to the damping of the acoustic
signal. Another focus consists in investigating the effect of the pore shape and the
preferential distribution of the porous areas within the samples microstructure on the
damping of the flexural frequency.
According to the theory, a material exhibiting a high stiffness should also exhibit a low
damping and conversely. The correlation of the figure 4.51 considering the evolution of
the material stiffness with increasing porosity and the figure 4.55 below clearly reveals
the importance of the pore shape on the material damping behaviour.
Indeed, Ref-fib that shows a lower stiffness than the other studied formulations in the
whole porosity range exhibits very low damping values in comparison with the other
formulations. A two-step evolution is thus observed: the damping values slightly increase
from the initial value of 3,33.10-4 [-] for an apparent porosity of 18,4 vol-% to a value of
4,42.10-4 [-] for an apparent porosity of 27,7 vol-%. The damping increases more
significantly in the high porosity range to reach a value of 8,75.10-4 [-] for an apparent
porosity of 32,5 vol-%.
Fig. 4.55: Evolution of the Young’s modulus of Ref-sph, Ref-cyl and Ref-fib in function
of the apparent porosity
212
Results
Ref-sph that shows a higher stiffness than the other studied formulations in the whole
porosity range exhibits intermediary damping values in comparison with the other
formulations. Like previously for Ref-fib, a two-step evolution is observed: the damping
values slightly increase from the initial value of 3,33.10-4 [-] for an apparent porosity of
18,4 vol-% to a value of 6,75.10-4 [-] for an apparent porosity of 29,0 vol-%. The
damping increases more significantly in the high porosity range to reach a value of
11,54.10-4 [-] for an apparent porosity of 32,5 vol-%.
Ref-cyl that shows an intermediate stiffness in comparison with the other studied
formulations in the whole porosity range exhibits higher damping values than those
revealed by the other studied formulations. A two-step evolution is also observed,
however the tendency is not similar as the one of Ref-sph and Ref-fib. Indeed, the
damping values of Ref-cyl continuously and quite linearly increase from the initial value
of 3,33.10-4 [-] for an apparent porosity of 18,4 vol-% to a value of 8,43.10-4 [-] for an
apparent porosity of 28,7 vol-%. The damping slightly decreases in the high porosity
range to reach a value of 7,66.10-4 [-] for an apparent porosity of 30,5 vol-%.
In order to give pertinent explanations related to the damping behaviour of the tested
formulations that is not correlated with the material stiffness with increasing porosity,
SEM analyses are performed on polished samples of each formulation after sintering at
1500°C for 6 hours. Overviews of the microstructure of the sample are shown on
figures 4.56, 4.58 and 4.60. The aim of the microstructure examination is to reveal the
density of the porous areas, to show if the obtained porous area has the expected
geometry according to the addition of the pore forming agent and to show if a distribution
of the pore forming agent during the casting is noticeable. A magnification x 250 or
x 500 is next shown on figures 4.57, 4.59 and 4.61 to examine the interface between the
porous areas and the microstructure.
213
Results
entrapped in the viscous mixture during the manufacturing that release the microstructure
of the sample by evaporating.
Fig. 4.56: Overview of the formulation Fig. 4.57: Magnification x 500 of the
Ref-sph-4 exhibiting an apparent porosity formulation Ref-sph-4 exhibiting an
of 24,7 vol-% after sintering at 1500°C for apparent porosity of 24,7 vol-% after
6 hours sintering at 1500°C for 6 hours
214
Results
Fig. 4.58: Overview of the formulation Fig. 4.59: Magnification x 250 of the
Ref-cyl-4 exhibiting an apparent porosity formulation Ref-cyl-4 exhibiting an
of 24,7 vol-% after sintering at 1500°C for apparent porosity of 24,7 vol-% after
6 hours sintering at 1500°C for 6 hours
215
Results
Fig. 4.60: Overview of the formulation Fig. 4.61: Magnification x 500 of the
Ref-fib-1,5 exhibiting an apparent porosity formulation Ref-fib-1,5 exhibiting an
of 24,2 vol-% after sintering at 1500°C for apparent porosity of 24,2 vol-% after
6 hours sintering at 1500°C for 6 hours
Thermal shocks in air according to DIN EN 993-11 at 950°C are performed on the
formulations Ref-sph-4, Ref-cyl-4 and Ref-fib-1,5 exhibiting an apparent porosity of
24,7 vol-%, 24,7 vol-% and 24,2 vol-% respectively. To focus on the influence of the
pore shape on the thermal shock resistance of refractory castables, those formulations
with a similar apparent porosity are tested. In order to examine the influence of the
apparent porosity, the evolution of the stiffness of Ref exhibiting an apparent porosity of
18,4 vol-% is also studied.
216
Results
Figure 4.62 shows the evolution of the Young’s modulus of the tested formulations after
five progressive thermal shocks in air.
Fig. 4.62: Evolution of the stiffness of the formulations Ref, Ref-sph, Ref-cyl and Ref-fib
after thermal shock tests in air at 950°C according to DIN EN 993-11
The reference formulation shows an exponential decrease of the stiffness with a quick
decrease of the Young’s modulus from 134 GPa after sintering to 98 GPa after the second
thermal shock followed by a slight decrease up to 86 GPa after five thermal shocks. This
formulation exhibits a retained stiffness of 64 % after five thermal shocks.
Formulations Ref-cyl-4 and Ref-fib-1,5 show a similar decrease of the Young’s modulus
and a similar trend as the reference formulation. The stiffness decreases first from 81 GPa
and 79 GPa respectively after the sintering to 47 GPa and 50 GPa respectively after the
second thermal shock followed by a slight decrease up to 37 GPa and 42 GPa
respectively after the fifth thermal shock. Ref-cyl-4 and Ref-fib-1,5 exhibit a retained
stiffness of 46 % and 53 % respectively after five thermal shocks.
Ref-sph-4 differs from the formulations Ref, Ref-cyl-4 and Ref-fib-1,5 with a continuous
exponential decrease of the Young’s modulus. Contrary to the other formulations, no
two-step decrease of the material stiffness is observed in Ref-sph-4. This formulation
exhibits an initial stiffness of 96 GPa after the sintering and a stiffness of 45 GPa after the
fifth thermal shock corresponding to 47 % of the initial stiffness.
By focussing on the retained Young’s modulus of the material, it can be concluded that
the further addition of pore forming agent, independently of the pore shape, in order to
reach an apparent porosity of around 24-25 vol-% does not enhance the thermal shock
resistance of the refractory castable. The pore shape influences nevertheless the thermal
217
Results
shock resistance of the castable. The castable showing fibre-like porous areas exhibits a
higher retained stiffness with 53 % than the formulation constituted of cylindrical pores
with a retained Young’s modulus of 47 %, which is also higher than the stiffness of the
formulation composed of spherical pores with a retained Young’s modulus of 46 %.
Figure 4.63 shows the evolution of the damping behaviour of the tested formulations
after five progressive thermal shocks in air. A clear influence of the pore shape is
noticeable on the damping behaviour of the porous castables.
The reference formulation shows a low increase of the damping in agreement with high
stiffness values exhibited by the formulation after each thermal shock cycle. With an
initial value of 2,57.10-4 [-],Ref shows the lowest damping value after sintering at 1500°C
for 6 hours, which is in accordance with the low apparent porosity of the formulation
(18,4 vol-%). The damping increases first up to the third thermal shock cycle to reach a
value of 3,36.10-3 [-] and then remains almost constant up to the fifth thermal shock cycle
to show a final value of 3,34.10-3 [-] corresponding to 1300 % of the initial damping
value after sintering.
Ref-sph-4 exhibiting intermediary Young’s modulus values after each thermal shock
cycle in comparison with Ref characterized by a higher retained stiffness and
formulations Ref-cyl and Ref-fib that show a lower retained stiffness, reveals the highest
damping values from the second thermal shock cycle up to the fifth thermal shock cycle.
Starting from a value of 5,10.10-4 [-] after the sintering, the damping first drastically
increases up to the third thermal shock to reach a value of 9,53.10-3 [-] and then slightly
increases up to the fifth thermal shock to reach a value of 12,52.10-3 [-] corresponding to
2455 % of the initial damping value after sintering.
Formulations Ref-cyl and Ref-fib, which follow the same trend related to the Young’s
modulus after thermal shocks, exhibit thoroughly different damping behaviours. Like Ref
and Ref-sph, the damping of Ref-fib first linearly increases from the initial value of
4,10.10-4 [-] after sintering to 7,02.10-3 [-] after the third thermal shock cycle and then a
low decrease of the damping is observed up to the fifth thermal shock cycle with a value
of 5,82.10-3 [-] corresponding to 1419 % of the initial damping value after sintering.
Concerning Ref-cyl, the damping first logarithmic increases from the initial value of
7,44.10-4 [-] after sintering to 3,56.10-3 [-] after the third thermal shock cycle and then a
218
Results
low decrease of the damping is observed up to the fifth thermal shock cycle with a value
of 2,29.10-3 [-] corresponding to 308 % of the initial damping value after sintering.
Fig. 4.63: Evolution of the damping behaviour of the formulations Ref, Ref-sph, Ref-cyl
and Ref-fib after thermal shock tests in air at 950°C according to DIN EN 993-11
In order to give further explanations related to the disagreement between the damping
behaviour of the tested formulations and the material stiffness with increasing damage,
SEM analyses are performed on polished samples of the formulations after five thermal
shock cycles in air. Overviews of the microstructure of the sample are presented on
figures 4.64, 4.66 and 4.68. The aim of the microstructure examination is to reveal first
the density of the cracks in order to correlate the microstructural damage with the
resulting Young’s modulus values, and to examine the crack path within the castable
microstructure and more precisely to investigate the crack propagation through the
artificially porous areas. A magnification x 2000, x 2500 or x 5000 is next shown on
219
Results
figures 4.65, 4.67 and 4.69 to examine the interface between the porous areas and the
cracks to reveal if such interfaces can enhance friction phenomena.
Figure 4.64 reveals the overview of the microstructure of Ref-sph-4 exhibiting an
apparent porosity of 24,7 vol-% after five thermal shocks in air at 950°C. A large crack is
clearly propagating throughout the sample matrix and more precisely at the boundary of
the tabular alumina aggregates. Those aggregates are lesser embedded into the castable
matrix than in case of the reference formulation because of the porous areas surrounding
those particles. Therefore, large cracks can easily propagate at the grain boundaries of the
tabular alumina aggregates. On the right side of the picture, the crack path throughout
two spherical pores is noticeable. The crack is propagating radially to the spherical pores:
a perpendicular crack path at the interface of the spherical pore is observed.
A magnification x 2000 of the crack path at the level of a spherical pore is shown on
figure 4.65. The radial propagation of the crack to the surface of the spherical pore is
noticeable. The crack width at the level of the interface between the porous area and the
matrix is around 2 µm. In this region of the microstructure rich in CA6-facets randomly
distributed in the matrix, the enhancement of the crack flank friction phenomena is
expected to take place in case of crack propagation.
Fig. 4.64: Overview of the formulation Fig. 4.65: Magnification x 2000 of the
Ref-sph-4 after five thermal shocks in air formulation Ref-sph-4 after five thermal
at 950°C shocks in air at 950°C
220
Results
A magnification x 5000 of the crack path in the surrounding areas of a cylindrical pore is
shown on figure 4.67. The cracks are propagating tangentially to the interfaces of the
height of the cylindrical pore. The crack width can be locally significant and reach
around 20 µm. Few friction phenomena are expected to take place in such local damaged
areas with a characteristic considerable distance between the crack flanks.
Fig. 4.66: Overview of the formulation Fig. 4.67: Magnification x 5000 of the
Ref-cyl-4 after five thermal shocks in air at formulation Ref-cyl-4 after five thermal
950°C shocks in air at 950°C
Fig. 4.68: Overview of the formulation Fig. 4.69: Magnification x 2500 of the
Ref-fib-1,5 after five thermal shocks in air formulation Ref-fib-1,5 after five thermal
at 950°C shocks in air at 950°C
221
Results
222
Results
To illustrate the importance of the amplitude of the impulse excitation on the accuracy of
the assessment of the elastic properties, figure 4.70 shows the evolution of the Young’s
modulus of Ref and Ca-PSZ after progressive thermal shocks in air at 950°C for different
impulse excitations, namely 30 %, 50 %, 70 % and 100 %. It is easily noticeable that
according to the experimental procedure and more precisely to the mechanical excitation
of the sample, a significant discrepancy of the Young’s modulus can be achieved. This
undermines the inter-laboratory reproducibility of RFDA measurements on damaged
samples.
In a further part of the study, it will be shown that the ensuing behavior with regard to
thermal shock resistance of Ref and Y-FSZ is almost similar and differs from the
behavior of Ca-PSZ and Mg-PSZ also characterized by a similar behavior. This explains
the choice of the formulations examined on figure 4.70. In the following examinations,
the elastic behavior of Ref and Y-FSZ is separately treated from the elastic behavior of
Ca-PSZ and Mg-PSZ in order to facilitate the interpretation of the data.
As far as Ref is concerned, a two-step decrease of the stiffness is observed for each
impulse excitation. A first drastic degradation of the Young’s modulus is observed after
the first thermal shock from 147 GPa after sintering to 131 GPa for 30 % excitation until
129 GPa for 100 % excitation. The second decrease step is lesser significant up to reach a
final stiffness after ten thermal shocks from 107 GPa for 30 % excitation until 98 GPa for
100 % excitation. The stiffness difference tends to increase with increasing thermal shock
cycles and the measured Young’s modulus values for a defined thermal shock cycle
decrease with increasing amplitude of the impulse excitation. If the difference of stiffness
is only 2 GPa after the first thermal shock cycle, this difference reaches 9 GPa after the
whole series of thermal shocks until 10 GPa after the fourth thermal shock cycle with a
stiffness of 124 GPa for 30 % excitation until 114 GPa for 100 % excitation.
Concerning Ca-PSZ, an excitation dependency is also observed even if the trend is not
similar to the one previously observed for Ref. It is worth mentioning that the initial
material stiffness varies from 65 GPa for 100 % excitation to 67 GPa for 30 % excitation
after sintering. The supposed higher crack density after the sintering in Ca-PSZ may
explain the stiffness difference of 2 GPa. A first drastic degradation of the Young’s
modulus is observed after the first thermal shock cycle to reach 55 GPa for 30 %
excitation until 53 GPa for 100 % excitation. From the first thermal shock cycle to the
sixth thermal shock cycle, a further and lower decrease of the Young’s modulus is
observed to reach a stiffness of 37 GPa for 100 % excitation until 39 GPa for 30 %
223
Results
excitation. In this range of thermal shock cycles, the measured Young’s modulus strongly
depends on the amplitude of the impulse excitation as the stiffness varies after the fifth
thermal shock cycle from 40 GPa for 100 % excitation to 47 GPa for 30 % excitation, i.e.
a stiffness difference of 7 GPa is noticeable. A third step is observed in the stiffness
evolution of Ca-PSZ from the seventh thermal shock cycle to the tenth, while
independently of the amplitude of the excitation the Young’s modulus remains constant
to reach 37 GPa after the whole series of thermal shocks.
Fig. 4.70: Evolution of the Young’s modulus of Ref and Ca-PSZ after progressive
thermal shock cycles in air at 950°C with 30 %, 50 %, 70 % and 100 % of impulse of
excitation
In order to mathematically represent the stiffness dependency of the tested materials with
the amplitude of the impulse excitation, the evolution of the Young’s modulus with
increasing voltage delivered to the electromagnet that ejects the hammer providing the
excitation on the lower face of the sample is examined for each thermal shock cycle and
for each tested formulation. Figure 4.71 reveals the evolution of the Young’s modulus of
formulation Ref with increasing voltage delivered to the electromagnet for the ten
thermal shock cycles. Thus, an impulse excitation amplitude of 100 % corresponds to a
voltage of 42 V, while an impulse excitation amplitude of 1 % corresponds to a voltage
of 5 V. In the present part of the study, the abbreviation TS refers to thermal shock while
AS refers to as sintered.
224
Results
Fig. 4.71: Evolution of the Young’s modulus of Ref in function of the voltage delivered
to the electromagnet for excitation generation according to the number of thermal shock
cycles in air at 950°C
It clearly appears on figure 4.71 that the measured Young’s modulus slightly decreases
with increasing voltage. This decrease is more and more considerable with progressive
thermal shock cycles. A linear trend line can be added to the experimental Young’s
modulus values and the examination of the amplitude of impulse excitation dependency
of the Young’s modulus after progressive thermal shocks can be summarized by the
examination of the slope of the Young’s modulus evolution with increasing voltage.
A non-linear coefficient related to the slope of the trend line of the Young’s modulus
evolution with increasing voltage can be defined as follows:
∆𝐸
𝜒𝐸 =
∆𝑉 Equation 4.1
With:
χE: Non-linear coefficient related to the stiffness with increasing voltage for
impulse generation [GPa/V]
ΔE: Stiffness difference [GPa]
ΔV: Difference of voltage delivered to the electromagnet for impulse generation
[V]
225
Results
Figures 4.72 and 4.73 show the evolution of the non-linear coefficient χE with the number
of thermal shocks for formulations Ref and Y-FSZ, and Ca-PSZ and Mg-PSZ
respectively.
Even if the trend of evolution does not look like similar, a two-step evolution of the
non-linear coefficient χE with increasing thermal shocks can be observed. Such an
evolution shows a two-stage microstructural damage of Ref and Y-FSZ with a first period
characterized by the nucleation and the propagation of cracks within the microstructure
and a second period characterized by the unification and the widening of cracks.
Fig. 4.72: Evolution of the non-linear coefficient χE related to the voltage for excitation
generation dependency of the Young’s modulus of Ref and Y-FSZ in function of the
number of thermal shock cycles in air at 950°C
As far Ref is concerned, the coefficient χE first decreases from an initial value of
-0,10 GPa/V after sintering to -3,46 GPa/V after the fourth thermal shock cycle.
Afterwards, an oscillating behavior is observed with χE values fluctuating between
-1,25 GPa/V after the ninth thermal shock cycle and -4,09 GPa/V after the tenth thermal
shock cycle.
Concerning Y-FSZ, the first evolution step constituting of a decrease of the coefficient χE
occurs during the first six thermal shock cycles form an initial value of -0,15 GPa/V to
reach a value of -2,88 GPa/V after the sixth thermal shock. An oscillating behavior is
then observed with fluctuating values of the coefficient χE between -1,42 GPa/V after the
eighth thermal shock and -1,99 GPa/V after the ninth thermal shock cycle.
226
Results
Fig. 4.73: Evolution of the non-linear coefficient χE related to the voltage for excitation
generation dependency of the Young’s modulus of Ca-PSZ and Mg-PSZ in function of
the number of thermal shock cycles in air at 950°C
As far as Ca-PSZ is concerned, the coefficient χE first slightly oscillates around its initial
value of -0,95 GPa/V after sintering up to the third thermal shock. An abrupt decrease of
the coefficient χE is then observed from the third thermal shock cycle with a value of
-0,92 GPa/V to the fifth thermal shock cycle with a coefficient value of -2,47 GPa/V.
Afterwards, the coefficient χE increases up to fluctuate around a zero value between
-0,37 GPa/V after the sixth thermal shock cycle and 0,07 GPa/V after the eighth thermal
shock cycle.
Concerning Mg-PSZ, the coefficient χE first slightly oscillates around its initial value of
-0,69 GPa/V after sintering up to the third thermal shock. An abrupt decrease of the
coefficient χE is then observed from the third thermal shock cycle with a value of
-0,68 GPa/V to the fourth thermal shock cycle with a value of -2,87 GPa/V. Afterwards,
the coefficient χE increases up to fluctuate around a value close to zero between
-0,02 GPa/V after the fifth thermal shock cycle and 0,08 GPa/V after the ninth thermal
shock cycle.
227
Results
Besides the Young’s modulus, the following part aims to show the influence of the
amplitude of impulse excitation on the damping behavior of the flexural resonant
frequency of the sample after progressive thermal shock cycles.
Figure 4.74 shows the evolution of the flexural damping of Ref after progressive thermal
shock cycles in air at 950°C for different impulse excitations, namely 30 %, 50 %, 70 %
and 100 %. Like previously with the Young’s modulus, it is easily noticeable that
according to the experimental procedure and more precisely to the mechanical excitation
of the sample, a significant discrepancy of the damping of the flexural resonant frequency
can be achieved. This also contribute to call into question the inter-laboratory
reproducibility of RFDA measurements on damaged samples.
228
Results
the fourth thermal shock cycle to reach a value of 17,86.10-3 [-], which remains constant
up to the fifth thermal shock cycle (17,74.10-3 [-]). The damping decreases after the sixth
thermal shock cycle until 9,05.10-3 [-] and progressively increases up to the eighth
thermal shock to reach a value of 15,94.10-3 [-]. This value remains almost constant up to
the tenth thermal shock cycle (15,17.10-3 [-]).
Fig. 4.74: Evolution of the flexural damping of Ref after progressive thermal shock
cycles in air at 950°C with 30 %, 50 %, 70 % and 100 % of impulse of excitation
Like previously with the Young’s modulus, in order to mathematically represent the
damping dependency of the tested materials with the amplitude of the impulse excitation,
the evolution of the flexural damping with increasing amplitude of excitation is examined
for each thermal shock cycle and for each tested formulation. Figure 4.75 reveals the
evolution of the damping of the flexural resonant frequency of Ref with increasing
amplitude of excitation for the ten thermal shock cycles. In order to improve the
readability of the diagram, the evolution of the flexural damping after 6, 7, 8 and 9
thermal shocks are not represented on the diagram.
It clearly appears on figure 4.75 that the damping of the flexural resonant frequency
considerably increases with increasing amplitude of excitation. This increase is more and
more considerable with progressive thermal shock cycles. If after sintering, the flexural
damping is 1,27.10-3 [-] for an amplitude of excitation of 30 % and 1,95.10-3 [-] for an
amplitude of excitation of 100 %, i.e. an increase of 154 %, the flexural damping is
4,76.10-3 [-] for an amplitude of excitation of 30 % and 17,86.10-3 [-] for an amplitude of
excitation of 100 %, i.e. an increase of 375 % after four thermal shock cycles. However,
contrary to the Young’s modulus, the interpretation of the damping increase with
229
Results
increasing amplitude of excitation through a linear trend line is not relevant as the
flexural damping exhibits an oscillating like behavior.
Therefore, to illustrate the non-linear behavior of the flexural damping with increasing
amplitude of excitation, a parameter ς is defined for each thermal shock cycle as follows:
With:
ς: Non-linear coefficient related to the flexural damping with increasing
amplitude of impulse excitation [-]
MAX 𝜁𝑓𝑓 : Maximum damping value of the flexural resonant frequency for a defined
thermal shock cycle in a range of amplitude of impulse excitation from
30 % to 100 % [-]
MIN 𝜁𝑓𝑓 : Minimum damping value of the flexural resonant frequency for a defined
thermal shock cycle in a range of amplitude of impulse excitation from
30 % to 100 % [-]
Fig. 4.75: Evolution of the damping of the flexural resonant frequency 𝜁𝑓𝑓 of Ref in
function of the amplitude of the impulse excitation according to the number of thermal
shock cycles in air at 950°C
230
Results
Figures 4.76 and 4.77 show the evolution of the non-linear coefficient ς with the number
of thermal shocks for formulations Ref and Y-FSZ, and Ca-PSZ and Mg-PSZ
respectively.
As far as Ref and Y-FSZ are concerned, a two-step evolution of the non-linear coefficient
ς with increasing thermal shocks can be observed as shown on figure 4.76. Such an
evolution that can be correlated with the evolution of the non-linear coefficient χE related
to the Young’s modulus shows a two-stage microstructural damage of Ref and Y-FSZ
with a first period characterized by the nucleation and the propagation of cracks within
the microstructure and a second period characterized by the unification and the widening
of cracks.
Concerning Ref, a first increase is observed from the initial value 0,7 [-] up to the fifth
thermal shock cycle to reach a value of 12,4 [-] with a maximum value of 13,1 [-]
exhibited after the fourth thermal shock. The non-linear coefficient ς decreases to reach
values that fluctuate around 4,0 [-] from the sixth thermal shock cycle to the tenth, and
more precisely between 3,6 [-] and 5,5 [-] after the seventh and the sixth thermal shock
cycle respectively.
Fig. 4.76: Evolution of the non-linear coefficient ς related to the amplitude of impulse
excitation dependency of the flexural damping of Ref and Y-FSZ in function of the
number of thermal shock cycles in air at 950°C
Concerning Y-FSZ, a first increase is observed from the initial value 0,4 [-] up to the
sixth thermal shock cycle to reach a value of 8,6 [-] corresponding to the maximum value
exhibited by this formulation. The non-linear coefficient ς then decreases to reach a value
231
Results
of 3,5 [-] after the seventh thermal shock cycle. The non-linear coefficient ς finally
slightly decreases to reach a value of 1,2 [-] after the tenth thermal shock cycle.
Contrary to Ref and Y-FSZ, formulations Ca-PSZ and Mg-PSZ exhibit no particular
evolution of the non-linear coefficient ς. The chaotic and oscillating evolution of ς shown
on figure 4.77 reveals first the effects of the pre-existent cracks after sintering on the
damping behavior and the complex friction phenomena occurring between the zirconia
particles and the castable matrix.
As far as Ca-PSZ is concerned, the non-linear coefficient ς starts from a value of 3,4 [-]
after the sintering to oscillate around 15 [-] between 8,8 [-] after the fifth thermal shock
and 21,4 [-] after the seventh thermal shock cycle. The final value of the non-linear
coefficient ς is 20,6 [-].
Concerning Mg-PSZ, the non-linear coefficient ς starts from a value of 12,8 [-] after the
sintering, showing the high microstructural damage of this formulation after sintering, to
oscillate around 20 [-] between 10,8 [-] after the first thermal shock and 38,0 [-] after the
third thermal shock cycle. The final value of the non-linear coefficient ς is 19,4 [-].
Fig. 4.77: Evolution of the non-linear coefficient ς related to the amplitude of impulse
excitation dependency of the flexural damping of Ca-PSZ and Mg-PSZ in function of the
number of thermal shock cycles in air at 950°C
232
Results
After having examined the influence of parameters, namely the apparent porosity and the
pore shape, the heterogeneity of the castable microstructure after manufacturing as well
as the experimental parameters such as the amplitude of the impulse excitation, on the
resulting elastic properties assessment, the following part of the study focusses on the
determination of the thermal shock resistance of the tested formulations Ref, Ca-PSZ,
Mg-PSZ and Y-FSZ. First the thermal shock resistance is determined through thermal
shock tests in air at 950°C according to DIN EN 993-11. A second approach consists in
investigating the effect of the temperature difference of thermal shock on the resulting
elastic properties. In a third and last part, the focus is broadened to the influence of the
heat capacity of the annealing medium on the severity of the thermal shock. Therefore,
quenching tests in molten aluminum at 800°C are performed. Furthermore, such
quenching tests in molten aluminum provide heating thermal shock conditions, while
quenching tests in air provide cooling thermal shock conditions. Different microstructural
damage mechanisms are expected to take place according whether the refractory
materials are submitted to heating or cooling thermal shocks.
In order to give a preliminary estimation of the thermal shock resistance of the tested
formulations Ref, Ca-PSZ, Mg-PSZ and Y-FSZ, the fundamental properties of the
castables with regard to their structure (porosity and density), their mechanical strength
(CCS, MOR and HMOR) as well as their elastic properties (Young’s modulus and
damping) are determined according to their respective standards. In a second part, the
RuL and the CiC of the castables are determined to better understand the
thermo-mechanical behavior of the castables at elevated temperature. In a third part,
dilatometric measurements are performed to assess the coefficient of expansion of the
castable and to interpret in a further part of the study the stiffness behavior of the
castables at elevated temperature through HT-RFDA measurements. All those precious
information are in a last part implemented in the calculation of the Hasselman parameters
R and R’’’’ related to the crack nucleation resistance and crack propagation resistance
respectively.
233
Results
Table 4.2: Structural, mechanical and elastic properties of formulations Ref, Ca-PSZ,
Mg-PSZ and Y-FSZ after sintering at 1500°C for 6 hours
Ref Ca-PSZ Mg-PSZ Y-FSZ
πa [vol-%] 16,6 ± 0,2 19,4 ± 0,2 17,7 ± 0,1 17,7 ± 0,1
πf [vol-%] 3,8 ± 0,1 2,3 ± 0,1 1,7 ± 0,1 1,7 ± 0,1
πt [vol-%] 20,4 ± 0,2 21,7 ± 0,2 19,4 ± 0,2 19,4 ± 0,2
ρb [g.cm-3] 3,1 ± 0,01 3,2 ± 0,01 3,2 ± 0,01 3,3 ± 0,01
CCS [MPa] 169 ± 6 128 ± 7 147 ± 11 198 ± 3
MOR [MPa] 42 ± 2 20 ± 1 11 ± 1 34 ± 2
HMOR at 1300°C [MPa] 23 ± 1 14 ± 2 6±1 18 ± 1
Young’s modulus E [GPa] 145 ± 2 66 ± 2 36 ± 4 125 ± 2
Flexural damping 𝜁𝑓𝑓 [*10-3] 1,92 ± 0,41 12,71 ± 4,32 25,78 ± 8,60 2,24 ± 0,35
234
Results
In terms of porosity and bulk density, formulation Ref is characterized by a low apparent
porosity with 16,6 vol-% and a typical high closed porosity with 3,8 vol-% resulting from
the Bayer process and thus with a total porosity of 20,4 vol-%. The replacement of
tabular alumina aggregates by dense electrofused zirconia aggregates induces a reduction
of the total closed porosity of the formulations based on zirconia aggregates with
2,3 vol-%, 1,7 vol-% and 1,7 vol-% for Ca-PSZ, Mg-PSZ and Y-FSZ respectively.
However, the nature of the doped zirconia has clearly an influence on the closed porosity
of the formulations. This can be explained by the different closed porosity exhibiting by
the electrofused zirconia aggregates in spite of the extreme manufacturing conditions.
Figure 4.78 shows in detail the structure of Ca-ZrO2, Mg-ZrO2 and Y-ZrO2 aggregates
embedded in the calcium aluminate matrix Ca-PSZ, Mg-PSZ and Y-FSZ respectively.
a b c
Fig. 4.78: Closed porosity in a) Ca-ZrO2 aggregate of Ca-PSZ
b) Mg-ZrO2 aggregate of Mg-PSZ and c) Y-ZrO2 aggregate of Y-PSZ
A significant discrepancy is noticeable with regard to the closed porosity and the
distribution of this closed porous areas within the zirconia aggregate. In case of Ca-ZrO2,
two representative aggregates are shown on figure 4.78 a). Those grains exhibit a
considerable density of porous areas that are heterogeneously distributed in the grains in
the form of spherical pores with a mean diameter around 1 µm. The zirconia grains
shown on figure 4.78 b) and c) constitute representative zirconia particles of Mg-ZrO2
and Y-ZrO2 respectively. Both particles exhibit a similar closed porosity, which is
obviously lower than that of Ca-ZrO2. The distribution of those porous areas is more
homogeneous and the mean size of the pores is lower than the mean pore size in Ca-ZrO2
with a mean diameter significantly below 1 µm. The observation of such a closed
porosity in each zirconia raw material, even in the fine grain fractions as revealed by the
SEM pictures on figure 4.78 with a typical grain size of around 10 – 40 µm of the
zirconia particles, calls into question the accuracy of the assessment of the true density of
Ca-ZrO2, Mg-ZrO2 and Y-ZrO2 summed up in table 3.12. Indeed, the true density of the
different zirconia varieties is determined with the assumption of the absence of closed
porosity on the grain fraction sieved under 63 µm. Therefore, the true density values
listed in table 3.12 can be considered as lowered true density values and the correct true
235
Results
density of the different zirconia may be higher than the values listed in table 3.12. A
particle size distribution of Ca-ZrO2, Mg-ZrO2 and Y-ZrO2 is performed according to
Fraunhoffer before the assessment of the true density, as the boundary conditions defined
by Fraunhoffer are thoroughly fulfilled, namely the opacity of the particles and a mean
particle diameter, which is significantly higher than the wave length of the laser beam. A
D50 value of 29,835 µm for Ca-ZrO2, 30,495 µm for Y-ZrO2 and 31,931 µm for Mg-ZrO2
has been respectively measured. Such a high D50 value for each tested zirconia reveals
the considerable size of the zirconia particles used for the assessment of the true density.
Such a value of the mean diameter of the zirconia particle size may explain an expected
discrepancy between the true density values listed in table 3.12 and the theoretical values
by assuming a zero closed porosity. Furthermore, the zirconia addition has also an impact
on the rheological properties of the castable, as the apparent porosity of Ca-PSZ, Mg-PSZ
and Y-FSZ are much higher than that of Ref with 19,4 vol-%, 17,7 vol-% and 17,7 vol-%
respectively. The microstructure examination after sintering shown in chapter 4.3.1.6
reveals a more important density of large spherical pores within the matrix of Ca-PSZ
than in Mg-PSZ and Y-FSZ. Therefore, the addition of zirconia raw materials does not
induce a considerable increase of the castable bulk density as the replacement of tabular
alumina aggregates with a true density of 3,91 g.cm-3 by Ca-ZrO2 aggregates with a true
density of 5,72 g.cm-3 or by Mg-ZrO2 aggregates with a true density of 5,65 g.cm-3 or
even by Y-ZrO2 aggregates with a true density of 5,99 g.cm-3, is responsible for the
increase of the apparent porosity due to detrimental rheological properties. Formulation
Ref exhibits a bulk density of 3,1 g.cm-3 while formulations Ca-PSZ, Mg-PSZ and
Y-FSZ show a bulk density of 3,2 g.cm-3, 3,2 g.cm-3 and 3,3 g.cm-3 respectively. Ca-PSZ
reveals the lowest bulk density of the zirconia based castables, as this material reveals the
highest apparent porosity as well as the highest closed porosity. As far as Mg-PSZ and
Y-FSZ are concerned, both formulations exhibit similar apparent porosity, closed
porosity and bulk density, while a low discrepancy could be expected as the
MgO-particles are slightly coarser as the other zirconia raw materials. The compactness
enhancement can be counterbalanced by other effects such as the detrimental rheological
properties of the mixture or the closed porosity of the raw materials.
In terms of mechanical properties, the enhancement of the CCS is only verified in case of
Y-FSZ. The compressive strength of refractory castables depends in a first instance on
the nature of the aggregates, in a second instance on the apparent porosity and in a third
instance on the crack density within the matrix. In case of Y-FSZ, the microstructure
examination presented in chapter 4.3.1.6.4 shows a crack free microstructure also
characterized by a low apparent porosity, even if this apparent porosity is 1,2 vol-%
higher than that of Ref. This can explain the considerable CCS of Y-FSZ with 198 MPa
in comparison with Ref exhibiting a CCS of 169 MPa. As far as Ca-PSZ and Mg-PSZ are
concerned, a lower compressive strength are observed with 128 MPa and 147 MPa
236
Results
237
Results
A MOR – Young’s modulus – flexural damping map is presented in this part of the study
in order to better distinguish the four studied refractory castables Ref, Ca-PSZ, Mg-PSZ
and Y-FSZ. Three point bending tests are performed on ten prismatic samples of each
formulation according to DIN EN 993-6 after measurement of the Young’s modulus and
the damping of the flexural resonant frequency through Resonant Frequency Damping
Analysis according to ASTM C 1548-02. The damping values presented in this part of
the study are much lower than those previously detailed in table 4.2, as the elastic
properties are determined with the help of the RT-RFDA testing device providing in
general lower damping values than the HT-RFDA testing device due to a lower damping
capacity of the acoustic signal by the sample holder used for RT-RFDA measurements.
Fig. 4.79: MOR – Young’s modulus – Damping of the flexural resonant frequency
𝜁𝑓𝑓 map of formulations Ref, Ca-PSZ, Mg-PSZ and Y-FSZ after sintering at 1500°C for 6
hours
A MOR – Young’s modulus – Damping of the flexural resonant frequency 𝜁𝑓𝑓 map is
shown on figure 4.79. This map should be regarded as a helpful tool in order to have a
more accurate assessment of the materials properties characterized in terms of thermal
shock resistance. Thus the observations detailed previously in table 4.2 are schematically
238
Results
illustrated. This schematic representation refers to the material properties after sintering
at 1500°C for 6 hours according to sintering profile 1 as far as Ref is concerned and
according to sintering profile 2 as far as the zirconia based formulations are concerned.
Ref and Y-FSZ are characterized by a considerable bending strength (42 MPa and
34 MPa respectively), a considerable stiffness (147 GPa and 125 GPa
respectively) as well as low damping values (1,92.10-3 [-] and 2,24.10-3 [-]
respectively)
Ca-PSZ is characterized by a medium bending strength (20 MPa), a medium
stiffness (66 GPa) and medium damping values (12,71.10-3 [-])
Mg-FSZ is characterized by a low bending strength (11 MPa), a low stiffness
(36 GPa) and high damping values (25,78.10-3 [-])
As shown on figures 4.80 and 4.81, Ref and Y-FSZ, and Ca-PSZ and Mg-PSZ are
respectively characterized by a similar thermal expansion behavior.
Indeed, as shown on figure 4.80, Ref and Y-FSZ exhibit a linear behavior in the whole
temperature range up to 1500°C. Those formulations reveal a higher maximum expansion
at 1500°C with 1,33 % and 1,26 % respectively than those of Ca-PSZ and Mg-PSZ with
1,05 % and 1,16 % respectively. The mean value of the thermal expansion coefficient α
between room temperature and 1000°C of Ref and Y-FSZ (8,27.10-6 K-1 and 8,34.10-6 K-1
respectively) is moreover more considerable than that of the Ca-PSZ and Mg-PSZ with
7,97.10-6 K-1 and 8,01.10-6 K-1 respectively. The tested formulations exhibit however a
similar coefficient of thermal expansion as pure alumina. The low monoclinic content in
Y-FSZ has a negligible impact on the thermal behavior of the castable as no contraction
is noticeable during the heating phase and a very low expansion (< 0,001 %) takes place
during the cooling phase at 933°C.
239
Results
240
Results
On the other side, Ca-PSZ and Mg-PSZ reveal a typical hysteretic thermal expansion of a
material containing a non-negligible monoclinic content as shown on figure 2.10. A
contraction is visible during the heating phase at 1058°C and 1075°C respectively and an
expansion occurs during the cooling phase at 905°C and 942°C respectively. However it
is worth mentioning that Mg-PSZ characterized by a high monoclinic content exhibits a
negligible contraction during the heating phase and a negligible expansion during the
cooling phase, both less than 0,001 %. Ca-PSZ characterized by a similar monoclinic
content as Y-FSZ according to XRD analyses detailed in chapter 3.1.2.4 exhibits the
highest contraction during the martensitic transformation from monoclinic to tetragonal
(0,033 %) and the highest expansion during the martensitic transformation from
tetragonal to monoclinic (0,044 %). The higher thermal expansion during the cooling
phase than the contraction during the heating phase may cause a further microstructural
damage and be responsible for the material fatigue in case of cycling high temperature
application.
Table 4.3: Some calculated data relative to the coefficient of thermal expansion,
transformation temperature of zirconia in the tested castables and the
maximum expansion through dilatometric measurement
Ref Ca-PSZ Mg-PSZ Y-FSZ
Mean value of α [10-6.K-1] 8,27 7,97 8,01 8,34
(between RT and 1000°C)
Tmon-tet during the heating phase [°C] - 1058 1075 n.a.
Ttet-mon during the cooling phase [°C] - 905 942 n.a.
Contraction at Tmon-tet [%] - 0,033 < 0,001 n.a.
Expansion at Ttet-mon [%] - 0,044 < 0,001 n.a.
Expansion at T=1500°C [%] 1,33 1,05 1,16 1,26
241
Results
0,20 MPa. Beyond the impact of zirconia, creep phenomena are expected to take place in
function of the nature of the doping agent, which is also likely to react with the calcium
aluminate phases of the matrix at the level of the grain boundaries of the zirconia
aggregates. Another focus consists in investigating the effect of the pre-existent cracks
generated by the sintering process on the creep behavior. Both RuL and CiC
measurements are performed on sintered cylindrical samples.
RuL measurements are carried out on Ref, Ca-PSZ, Mg-PSZ and Y-FSZ after sintering at
1500°C for 6 hours up to 1700°C with a heating rate of 5 K.min-1 and an appropriate
compressive stress for dense shape refractory materials of 0,20 MPa. Figure 4.82 shows
the evolution of the creep behavior of the tested castables with increasing temperature.
242
Results
Fig. 4.82: Refractoriness under Load (RuL) measurement performed on Ref, Ca-PSZ,
Mg-PSZ and Y-FSZ up to 1700°C under 0,20 MPa compressive stress
according to DIN EN ISO 1893
Table 4.4: Main thermo-mechanical properties of Ref, Ca-PSZ, Mg-PSZ and Y-FSZ
derivated from Refractoriness under Load (RuL) measurements according to
DIN EN ISO 1893
Ref Ca-PSZ Mg-PSZ Y-FSZ
εmax [%] 1,05 1,00 0,95 1,29
Tεmax [°C] 1457 1435 1383 1473
T0,5 [°C] 1610 1587 1615 1588
T1,0 [°C] 1655 1613 1651 1618
T2,0 [°C] > 1700 1645 1694 1655
CiC measurements are carried out on Ref, Ca-PSZ, Mg-PSZ and Y-FSZ after sintering at
1500°C for 6 hours at the sintering temperature with a heating rate of 5 K.min-1 up to
1500°C and an appropriate compressive stress for dense shape refractory materials of
0,20 MPa. The temperature and the mechanical load are maintained for 25 h. The aim of
such measurements is to reveal the creep velocity or creep rate in extreme severe
thermo-mechanical conditions to speed up the microstructural phenomena at the origin of
the creep. Figure 4.83 shows the evolution of the creep behavior of the tested castables
with increasing temperature up to 1500°C in a first stage of the experiment and for 25 h
at 1500°C in a second stage of the experiment.
243
Results
By analogy with RuL measurements, two groups of castables can be distinguished with
regard to the creep behavior at constant temperature. Ref and Y-FSZ exhibit a higher
maximum expansion at elevated temperature and more particularly a higher initial
expansion when the temperature of 1500°C is achieved with 1,13 % and 1,12 %
respectively while the initial expansion ε0 of Ca-PSZ and Mg-PSZ is 1,01 % and 0,99 %
respectively. Besides, Ref and Y-FSZ show a similar creep behavior and more
particularly a similar creep rate as soon as the temperature of 1500°C is achieved. Indeed,
Ref and Y-FSZ are characterized by a considerable creep rate during the first five hours
of the experiment at constant temperature with v1,0, v2,0 and v5,0 values of 0,299 %/h,
0,164 %/h and 0,102 %/h for Ref and 0,294 %/h, 0,177 %/h and 0,116 %/h for Y-FSZ
respectively. Those v1,0, v2,0 and v5,0 values exhibited by Ca-PSZ and Mg-PSZ are much
lower with 0,248 %/h, 0,138 %/h and 0,085 %/h for Ca-PSZ and 0,194 %/h, 0,113 %/h
and 0,076 %/h for Mg-PSZ respectively. This significant creep velocity of formulations
Ref and Y-FSZ is also highlighted by the maximum contraction of the sample after 25 h.
Indeed, Ref and Y-FSZ exhibit a contraction of -0,536 % and -0,644 % respectively after
25 h, while the final contraction of Ca-PSZ and Mg-PSZ is -0,499 % and -0,238 %
respectively. Furthermore, if formulations Ref, Y-FSZ and Ca-PSZ exhibit a similar
creep rate during the last 20 h of the experiment with v5h-25h values of 0,045 %/h,
0,047%/h and 0,043 %/h respectively, Mg-PSZ reveals a low creep velocity with a v5h-25h
value of 0,034 %/h in addition to a low creep rate during the first 5 h of the experiment at
constant temperature.
244
Results
It is also worth mentioning that Ref and Y-FSZ show a quite linear expansion during the
heating phase, a low contraction is observed in case of Ca-PSZ and Mg-PSZ during the
transformation of zirconia from monoclinic to tetragonal crystalline phase. By Ca-PSZ,
the contraction resulting from the martensitic transformation is 0,032 % and starts at
1070°C. This contraction is also noticeable in the same order of magnitude during RuL
measurement but at lower temperature, i.e. 1037°C. Concerning Mg-PSZ, if a significant
contraction is not visible during the zirconia transformation, a lower expansion rate is
noticeable in a temperature range from 1050°C to 1150°C. This phenomenon is not
observed in the RuL measurement.
The creep properties of the tested castables Ref, Ca-PSZ, Mg-PSZ and Y-FSZ assessed
from Creep in Compression (CiC) measurements are summarized in table 4.5.
Table 4.5: Main thermo-mechanical properties of Ref, Ca-PSZ, Mg-PSZ and Y-FSZ
derivated from Creep in Compression (CiC) measurements according to
DIN EN 993-9
Ref Ca-PSZ Mg-PSZ Y-FSZ
ε0 [%] 1,13 1,01 0,99 1,12
ε1 [%] 0,83 0,76 0,79 0,82
ε2 [%] 0,67 0,62 0,68 0,64
ε5 [%] 0,36 0,37 0,45 0,30
v1,0 [%/h] 0,299 0,248 0,194 0,294
v2,0 [%/h] 0,164 0,138 0,113 0,177
v5,0 [%/h] 0,102 0,085 0,076 0,116
v5h-25h [%/h] 0,045 0,043 0,034 0,047
The thermal shock resistance behavior of Ref, Ca-PSZ, Mg-PSZ and Y-FSZ with regard
to the crack initiation resistance and crack propagation resistance can be assessed by the
estimation of the Hasselman parameters R and R’’’. By taking into account the stiffness
and the bending strength detailed in table 4.2 as well as the coefficient of thermal
expansion previously estimated in chapter 4.3.1.3, both R and R’’’ parameters can be
calculated. Hasselman parameter R’’’ of each tested formulation is calculated according
245
Results
to the equation 2.27. Hasselman parameter R is defined according to the equation 2.24.
However this parameter can also be assessed according to the following equation:
𝜎 .(1−𝜇)
R= [𝑐𝑚] Equation 4.3
𝛼. 𝐸
The Hasselman parameters R and R’’’ as well as the stored elastic energy after sintering
of each tested formulations are summarized in table 4.6. The stored elastic energy is
calculated according to equation 2.31.
Table 4.6: Estimation of R and R’’’ Hasselman parameters of Ref, Ca-PSZ, Mg-PSZ
and Y-FSZ
Ref Ca-PSZ Mg-PSZ Y-FSZ
R [K] 28 ± 1 35 ± 0,5 35,5 ± 0,1 26 ± 1
1,0.10-4 1,8.10-4 3,3.10-4 1,3.10-4
R’’’ [Pa-1]
± 0,4.10-4 ± 0,2.10-4 ± 0,6.10-4 ± 0,2.10-4
Eel [kPa] 5,9 ± 0,6 3,2 ± 0,5 1,6 ± 0,4 4,8 ± 0,8
As far as the crack initiation resistance is concerned, Ref and Y-FSZ reveal similar R
values with 28 [K] and 26 [K] respectively. Those values are lower than those of Ca-PSZ
and Mg-PSZ with 35 [K] and 35,5 [K] respectively. Such values should be nevertheless
called into question as Ca-PSZ and Mg-PSZ already present some cracks after sintering,
while the calculation of the Hasselman parameter R assumes a crack-free material.
Ca-PSZ and Mg-PSZ are furthermore characterized by a higher crack propagation
resistance with a R’’’ value of 1,8.10-4 [Pa-1] and 3,3.10-4 [Pa-1] respectively. Those R’’’
values are more considerable than those of Ref and Y-FSZ that exhibit R’’’ values equal
to 1,0.10-4 [Pa-1] and 1,3.10-4 [Pa-1] respectively.
Ca-PSZ and Mg-PSZ that reveal a high resistance to crack initiation and crack
propagation according to the calculation of the R and R’’’ parameters respectively are
also characterized by a low stored elastic energy with 3,2 [kPa] and 1,6 [kPa]
246
Results
respectively. On the other hand, Ref and Y-FSZ that reveal a low resistance to crack
initiation and crack propagation according to the calculation of the R and R’’’ parameters
respectively are characterized by a considerable stored elastic energy with 5,9 [kPa] and
4,8 [kPa] respectively.
247
Results
the tabular alumina aggregates (see fig. 4.86 and 4.87). Those pores have a mean
diameter comprised between 1 and 10 µm as shown on fig. 4.87.
Fig. 4.86: Stable interfaces between TA Fig. 4.87: High crystallized matrix
grain and the matrix in formulation Ref surrounding porous TA grains of
after sintering at 1500°C/6h formulation Ref after sintering at
(magnification 500) 1500°C/6h (magnification 2000)
248
Results
sintering process already generate the nucleation of pre-existent cracks. This means that
the further thermal shock tests performed within the framework of this study are carried
out on samples that exhibit a non-negligible crack density before tests.
Fig. 4.88: Microstructure overview of Fig. 4.89: Low crack density in the
formulation Ca-PSZ after sintering at microstructure of Ca-PSZ after sintering at
1500°C/6h (magnification 60) 1500°C/6h (magnification 150)
Fig. 4.90: Intergranular cracks propagating Fig. 4.91: Stable interfaces between
through the matrix of Ca-PSZ after zirconia particles and the matrix in
sintering at 1500°C/6h (magnification 500) Ca-PSZ after sintering at 1500°C/6h
(magnification 2000)
Some spherical pores resulting from the manufacturing process are parsimoniously and
heterogeneously present in the microstructure. A stronger magnification shown on
figures 4.89 and 4.90 reveals that those pre-existent cracks are propagating throughout
the matrix and at the grain boundaries of tabular alumina and zirconia middle-sized
aggregates. A detailed examination of the crack configuration is illustrated on
figure 4.90. By propagating throughout the matrix, the crack is deviated by the zirconia
particle. In general, the interfaces between the zirconia particle and the matrix reveal a
249
Results
strong connection between the formed calcium aluminate phases and the zirconia
particles.
Such pictures of Ca-PSZ show that microstructural damage can take place in this
formulation even if the principal underlying observation consists in showing a dense
microstructure constituted of tabular alumina and zirconia grains of different fractions,
which are well embedded and homogeneously distributed in the calcium aluminate
matrix. Figure 4.89 also reveals the local high concentration of CA 6 facets surrounding
the small and middle-sized tabular alumina and zirconia aggregates resulting from an
optimum dispersion of the particles during the manufacturing. Zirconia particles exhibit
different shape namely spherical and splintered. Most of the Ca-ZrO2 grains are
middle-sized with a mean diameter between 50 µm and 200 µm. They correspond to
splitter of zirconia crystals. Those zirconia particles exhibit furthermore a high density of
closed pores as shown on fig. 4.91.
250
Results
Mg-ZrO2 particles, while cracks are propagating at the grain boundary of the zirconia
particles in case of Ca-PSZ.
By analogy with the Ca-ZrO2 particles, Mg-ZrO2 particles exhibit different shape namely
spherical and splintered. Most of the Mg-ZrO2 grains are middle-sized up to big-sized
with a mean diameter comprised between 500 µm and 1 mm. Figure 4.93 shows local
high concentration of big Mg-ZrO2 particles, while other microstructural areas of the
samples is only constituted of tabular alumina rich parts. Those Mg-ZrO2 particles appear
to be dense with a low concentration in closed pores (see fig. 4.95).
Fig. 4.94: Intergranular crack propagation Fig. 4.95: Crack propagation at grain
between zirconia particles in Mg-PSZ after boundary of zirconia particle in Mg-PSZ
sintering at 1500°C/6h (magnification 500) after sintering at 1500°C/6h
(magnification 2000)
251
Results
spherical pores resulting from the manufacturing process are by analogy with Ref
parsimoniously and heterogeneously present in the microstructure. Those pores appear to
have a similar size as Ref, i.e. comprised between 50 µm and 300 µm, however the pore
density seems to be more important. A stronger magnification shown on fig. 4.97 and
4.98 reveals the undamaged interfaces between the zirconia aggregates and the calcium
aluminate matrix. That reveals besides the high thermal stability of the interfaces between
tabular alumina aggregates and the matrix, a high thermal stability between the zirconia
particles and the matrix during sintering.
Fig. 4.98: Stable interfaces between Fig. 4.99: High CA6-facets concentration
Y-ZrO2 grain and the matrix in formulation surrounding ZrO2 grain in formulation
Y-FSZ after sintering at 1500°C/6h Y-FSZ after sintering at 1500°C/6h
(magnification 500) (magnification 2000)
On fig. 4.99, and as observed in zirconia based formulations Ca-PSZ and Mg-PSZ, a high
density of CA6-facets around the zirconia aggregates are noticeable. That reveals the
more challenging defloculation of the castable formulation in presence of zirconia during
252
Results
4.3.2 Thermal shock resistance of Ref, Ca-PSZ, Mg-PSZ and Y-FSZ in air
according to DIN EN 993-11
A first approach to examine the thermal shock behavior of the castable formulations Ref,
Ca-PSZ, Mg-PSZ and Y-FSZ consists in performing thermal shock tests in air at 950°C
according to DIN EN 993-11. The experimental procedure for such thermal shock tests is
stressed in chapter 3.2.7.1.1. The aim of this experiment is to provide first information
with regard to the stiffness and damping response of the tested formulations after abrupt
quenching from 950°C to room temperature. As those formulations exhibit specific
mechanical, elastic and damping properties after manufacturing, different evolution of
stiffness and damping are expected to be examined. The thermal shock resistance is
assessed through the measurement of the retained Young’s modulus and the damping
increase through Resonant Frequency Damping Analysis as well as the retained MOR
through three-point bending strength measurements.
253
Results
Young’s modulus after progressive thermal shocks in air at 950°C of the tested
formulations.
Fig. 4.100: Evolution of the Young’s modulus of Ref, Ca-PSZ, Mg-PSZ and Y-FSZ after
progressive thermal shocks in air at 950°C according to DIN EN 993-11
On the one hand, Ref and Y-FSZ exhibit a low two-step decrease of the Young’s
modulus with increasing thermal shocks. Starting from close Young’s modulus values
after sintering, 147 GPa for Ref and 125 GPa for Y-FSZ, the stiffness undergoes a
significant decrease after the first thermal shock cycle to reach 130 GPa and 113 GPa
respectively. The second decrease of the Young’s modulus is characterized by a lower
rate to reach 102 GPa and 98 GPa respectively after the whole series of thermal shocks. It
is worth mentioning that Ref shows a decrease of 45 GPa while the stiffness decrease of
Y-FSZ is only 26 GPa.
On the other hand, Ca-PSZ and Mg-PSZ follow a three-step decrease with increasing
thermal shock cycles. Starting from Young’s modulus values after sintering of 66 GPa
for Ca-PSZ and 36 GPa for Mg-PSZ, the stiffness undergoes a significant decrease after
the first thermal shock cycle to reach 53 GPa and 28 GPa respectively. The second
decrease of the Young’s modulus is characterized by a lower rate from the first thermal
shock cycle to the sixth thermal shock cycle as far as Ca-PSZ is concerned and to the
seventh thermal shock cycle as for Mg-PSZ to reach 37 GPa and 16 GPa respectively.
254
Results
The third evolution step corresponds to constant values of Young’s modulus. The final
stiffness of Ca-PSZ and Mg-PSZ is 37 GPa and 16 GPa respectively after the whole
series of thermal shocks. The stiffness decrease after ten thermal shock cycles is 28 GPa
for Ca-PSZ and 20 GPa for Mg-PSZ.
Fig. 4.101: Evolution of the relative Young’s modulus of formulations Ref, Ca-PSZ,
Mg-PSZ and Y-FSZ after progressive thermal shocks in air at 950°C
according to DIN EN 993-11
The examination of the relative Young’s modulus evolution after progressive thermal
shocks shown on figure 4.101 leads to similar observations, namely the distinction
between the formulations Ref and Y-FSZ, and Ca-PSZ and Mg-PSZ that exhibit a similar
behavior in terms of retained stiffness.
On the one hand, Ref and Y-FSZ exhibit a low two-step decrease of the Young’s
modulus with increasing thermal shocks as well as the highest relative Young’s modulus
values after the whole series of thermal shocks with 69 % and 78 % of their initial
Young’s modulus values respectively. The stiffness undergoes a significant decrease after
the first thermal shock cycle to reach 88 % and 90 % respectively. The second decrease
of the Young’s modulus is characterized by a lower rate to reach the final stiffness values
detailed above. It is worth mentioning that Ref shows a higher relative decrease of
stiffness than that of the Y-FSZ with 30 % and 21 % respectively.
On the other hand, Ca-PSZ and Mg-PSZ follow a three-step decrease with increasing
thermal shock cycles to reach final relative Young’s modulus values that are considerably
lower than those of formulations Ref and Y-FSZ with 58 % and 45 % respectively. The
255
Results
stiffness undergoes a significant decrease after the first thermal shock cycle to reach 80 %
and 76 % respectively. The second decrease of the Young’s modulus is characterized by
a lower rate from the first thermal shock cycle to the sixth thermal shock cycle as far as
Ca-PSZ is concerned and to the seventh thermal shock cycle as for Mg-PSZ to reach
56 % and 45 % respectively. The third evolution step corresponds to constant values of
Young’s modulus until that Ca-PSZ and Mg-PSZ exhibit the final stiffness values
previously detailed. The stiffness decrease after ten thermal shock cycles is 41 % for
Ca-PSZ and 54 % for Mg-PSZ. The kinetic approach is in agreement with the global
examination of the material behavior in terms of elasticity, as formulations Ca-PSZ and
Mg-PSZ exhibiting considerable decrease of the absolute Young’s modulus in
comparison with their initial stiffness after sintering also show a significant damage
kinetic.
The examination of the evolution of the damping of the flexural resonant frequency after
progressive thermal shocks aims to provide further information with regard to
microstructural damage processes. By analogy with the Young’s modulus, two different
approaches can be adopted: either the resulting absolute damping is studied or the relative
damping corresponding to the damping after a defined thermal cycle number divided by
the initial damping value of the material. The first approach constitutes a global
examination of the material behavior in terms of crack flank friction enhancement as
microcracking progressively takes place within the microstructure of the material, while
the second approach refers to a kinetic aspect of the crack flank friction enhancement of
the material. Both approaches are presented in the part of the study hereby.
Figure 4.102 shows the evolution of the damping of the flexural resonant frequency after
progressive thermal shocks in air at 950°C of the tested formulations, while figure 4.103
shows the evolution of the relative damping of the flexural resonant frequency after
progressive thermal shocks in air at 950°C of the tested formulations.
256
Results
Fig. 4.102: Evolution of the damping of the flexural resonant frequency 𝜁𝑓𝑓
of Ref, Ca-PSZ, Mg-PSZ and Y-FSZ after progressive thermal shocks in air at 950°C
according to DIN EN 993-11
On the one hand, Ref and Y-FSZ exhibit lower damping values than that of Ca-PSZ and
Mg-PSZ after each thermal shock cycle. The damping increase of Ref and Y-FSZ is quite
linear from an initial value of 1,92.10-3 [-] for Ref and 2,24.10-3 [-] for Y-FSZ to achieve
15,70.10-3 [-] for Ref and 10,11.10-3 [-] for Y-FSZ.
On the other hand, Ca-PSZ and Mg-PSZ are characterized by a high initial damping
value, 12,71.10-3 [-] for Ca-PSZ and 25,78.10-3 [-] for Mg-PSZ. Concerning Ca-PSZ, the
damping follows a drastic increase after the first thermal shock cycle to reach
25,11.10-3 [-] followed by a decrease after the second thermal shock cycle up to
16,71.10-3 [-]. The damping of Ca-PSZ continuously increases afterwards up to the fifth
thermal shock cycle to reach a damping value of 33,43.10-3 [-]. The damping remains
constant up to the ninth thermal shock cycle with a value of 29,92.10-3 [-] and finally
increases after the last and tenth thermal shock cycle up to 36,12.10-3 [-]. As far as
Mg-PSZ is concerned, the damping progressively increases up to the third thermal shock
cycle exhibiting the highest value (37,46.10-3 [-]). The damping slightly decreases up to
the ninth thermal shock cycle to reach a value of 28,66.10-3 [-] and considerably
decreases after the tenth thermal shock cycle up to 19,07.10-3 [-].
257
Results
Fig. 4.103: Evolution of the relative damping of the flexural resonant frequency 𝜁𝑓𝑓 of
Ref, Ca-PSZ, Mg-PSZ and Y-FSZ after progressive thermal shocks in air at 950°C
according to DIN EN 993-11
The examination of the evolution of the relative damping of the flexural resonant
frequency partly leads to similar observations. On figure 4.103, two groups of materials
can be distinguished.
On the one hand, Ca-PSZ and Mg-PSZ exhibit lower relative damping values than that of
Ref and Y-FSZ after each thermal shock cycle except for the first thermal shock where
Ca-PSZ reveals a high relative damping value of 202 %. Concerning Ca-PSZ, the relative
damping, except for the first thermal shock cycle, continuously and regularly increases to
achieve 284 % after the whole series of thermal shocks. As far as Mg-PSZ is concerned,
the relative damping is slightly superior to the initial relative damping value and reaches
a maximum value of 145 % after the third thermal shock cycle and slightly decreases
after this third thermal shock cycle up to the tenth cycle to reach a value inferior to the
initial relative damping with 74 %.
On the other hand, Ref and Y-FSZ, in spite of low absolute damping values, exhibit a
significant increase of relative damping. This is partly explained by the low absolute
damping values after sintering with 1,92.10-3 [-] for Ref and 2,24.10-3 [-] for Y-FSZ and
the considerable damping increase in comparison with these initial values. Formulation
Y-FSZ is thus characterized by a logarithmic increase of the relative damping to reach
values fluctuating around 500 % after the sixth thermal shock up to the end of the
experiments. Y-FSZ exhibits a final relative damping value of 451 %. Formulation Ref
shows an increase of the relative damping up to the fourth thermal shock cycle with a
258
Results
value of 678 %. The relative damping of Ref decreases up to the seventh thermal shock to
achieve a value of 444 % and progressively increases to achieve 819 % after the tenth
thermal shock, corresponding to the maximum relative damping of the four tested
formulations.
If the evolution of the retained Young’s modulus after progressive thermal shocks aims to
assess the evolution of the microstructural damage and more specifically the evolution of
the crack density, the retained bending strength of castables after progressive thermal
shocks also depend on the crack configuration and crack localization within the sample
microstructure and the possible toughening mechanisms. Therefore, it appears that the
study of the retained modulus of rupture in function of the retained Young’s modulus
provide further information in terms of influence of the crack network and damage
mechanisms on the mechanical and elastic properties of the castable formulations.
Figure 4.104 shows the evolution of the Young’s modulus of formulations Ref, Ca-PSZ,
Y-FSZ and Mg-PSZ after progressive thermal shocks in air at 950°C according to
DIN EN 993-11 in function of the retained modulus of rupture.
Fig. 4.104: Evolution of the modulus of rupture of Ref, Ca-PSZ, Mg-PSZ and Y-FSZ
after progressive thermal shocks in air at 950°C according to DIN EN 993-11
259
Results
The examination of the evolution of the retained Young’s modulus and the retained
modulus of rupture leads to similar observations in terms of similarities with regard to the
castable behavior after progressive thermal shocks. On figure 4.104, two groups of
materials can be distinguished.
On the one hand, Ref and Y-FSZ exhibit high stiffness and bending strength values after
progressive thermal shocks. The slope of the degradation of the modulus of elasticity in
function of the decrease in modulus of rupture is furthermore low for both formulations.
Indeed, Ref reveals a stiffness of 147 GPa and a modulus of rupture of 42 MPa after
sintering while formulation Y-FSZ reveals a stiffness of 125 GPa and a modulus of
rupture of 34 MPa after sintering. After ten thermal shocks, those properties drop to reach
102 GPa and 18 MPa for Ref respectively and 98 GPa and 13 MPa for Y-FSZ
respectively. This corresponds to a stiffness decrease of 2 GPa for a bending strength
decrease of 1 MPa for Ref, while a stiffness decrease of 1 GPa for a bending strength
decrease of 1 MPa for Y-FSZ. It is however worth mentioning that if the degradation of
the stiffness with decreasing bending strength is linear in the whole series of thermal
shocks for Y-FSZ, this decrease is only linear after the first five thermal shocks for Ref.
Afterwards, formulation Ref reveals a drastic decrease of the stiffness while the bending
strength remains constant. During the first five thermal shock cycles, Ref shows a
stiffness decrease of 1 GPa for a bending strength decrease of 1 MPa.
On the other hand, Ca-PSZ and Mg-PSZ exhibit low stiffness and bending strength
values after progressive thermal shocks. The slope of the degradation of the modulus of
elasticity in function of the decrease in modulus of rupture is furthermore high for both
formulations. Indeed, Ca-PSZ reveals a stiffness of 66 GPa and a modulus of rupture of
20 MPa after sintering while Mg-PSZ reveals a stiffness of 36 GPa and a modulus of
rupture of 11 MPa after sintering. After ten thermal shocks, those properties drop to reach
37 GPa and 11 MPa for Ca-PSZ respectively and 16 GPa and 7 MPa for Mg-PSZ
respectively. This corresponds to a stiffness decrease of 3 GPa for a bending strength
decrease of 1 MPa for Ca-PSZ, while a stiffness decrease of 5 GPa for a bending strength
decrease of 1 MPa for Mg-PSZ.
260
Results
Figures 4.105 – 4.108 show the evolution of the microstructure overview of Ref samples
after one, three, five and ten thermal shocks. The aim of those SEM pictures is to reveal
the increasing crack density with increasing thermal shock experiments in air at 950°C.
As shown on figure 4.105, after one thermal shock cycle, only few cracks are nucleating
parsimoniously within the matrix. On the top left corner of figure 4.105, a crack with a
considerable distance between the crack flanks is propagating throughout some pores
while others are nucleating in a perpendicular direction to a big pore induced by the
manufacturing process. Nevertheless, it has to be noticed that the crack density is low and
the ensuing cracks are generated in areas of the matrix exhibiting some manufacturing
defects.
261
Results
Figure 4.106 shows the microstructure of Ref after three quenching tests. The cracks
density is slightly higher than the crack density after one thermal shock. Most of the
cracks are still propagating throughout the matrix and through the porous areas of the
matrix more particularly.
On figure 4.107, the crack network is dense and the cracks are propagating at the grain
boundaries of tabular alumina aggregates.
Finally, figure 4.108 shows the considerable crack density in the microstructure of Ref
after ten quenching tests. Big cracks are propagating throughout the sample at the grain
boundaries of tabular alumina aggregates causing a partial debonding of those aggregates
as shown on the bottom part of figure 4.108. Besides the crack density increase, a
widening of the cracks is also observed.
Fig. 4.109: Crack nucleation within the Fig. 4.110: Crack propagation around
matrix of formulation Ref after 1 TS in air tabular alumina grains within the matrix of
at 950°C formulation Ref after 3 TS in air at 950°C
Fig. 4.111: Crack widening within the Fig. 4.112: Crack widening at the grain
matrix of formulation Ref after 5 TS in air boundary of tabular alumina grain of
at 950°C formulation Ref after 10 TS in air at 950°C
262
Results
Figures 4.109 – 4.112 show the path of the cracks in the matrix of Ref. The aim of those
pictures is to examine the nucleation of the cracks and the enhancement of the friction
mechanisms occurring after progressive thermal shock tests.
On figure 4.109, a crack with width of around 5 µm is propagating throughout the
CA6-rich matrix and at the grain boundary of a tabular alumina grain. Such a crack path
is deviated by small tabular alumina grains and defines interfaces between the CA6-rich
matrix with random oriented calcium aluminate particles and tabular alumina grains.
As shown on figure 4.110, after three quenching tests, the increase of thermal stresses
causes the expansion of the crack within the matrix and also a relative widening of the
crack. The mean distance between the crack flanks is around 5 µm but can reach 10 µm
in case of CA6 / CA6 interfaces. The crack path stops at the surface of a pore on the top
right corner of figure 4.110 while a crack deviation followed by a crack bridging occurs
on the bottom of figure 4.110.
After five quenching tests, the crack width increases up to reach a mean value of 10 µm
along the whole crack path as shown on figure 4.111. It could be assumed that the crack
widening first takes place within the CA6-rich matrix that provides a lower resistance to
damage mechanisms than the tabular alumina grains. The big crack propagating on the
left side of figure 4.111 seems to be split into two smaller cracks by hitting the tabular
alumina grain of the middle of figure 4.111.
After 10 quenching tests, the crack width reaches its maximum value, above 15 µm and
locally up to 20 µm, and this phenomenon of crack widening takes place at the level of
the grain boundary of tabular alumina aggregate as shown on figure 4.112.
Figures 4.113 – 4.116 show the evolution of the microstructure overview of Ca-PSZ
samples after one, three, five and ten thermal shock(s). The aim of those SEM pictures is
to reveal more considerable damage mechanisms with increasing thermal shock cycles in
air at 950°C than previously as far as Ref was concerned. Independently of the number of
quenching tests, the crack density within the microstructure of Ca-PSZ is much higher
than that of Ref inducing local debonding of aggregates from the matrix.
As shown on figure 4.113, the microstructure of Ca-PSZ after one quenching test is
similar to the microstructure of the samples after sintering at 1500°C during 6 hours.
Indeed, if no important damage is noticed within the microstructure, some cracks are
nucleated at the grain boundaries of tabular alumina grains and more particularly the
aggregates of big size.
263
Results
After three quenching tests, some cracks are propagating throughout the matrix as shown
in the middle of the figure 4.114 and joining different tabular alumina grain boundaries as
shown on the bottom left corner of the figure. Those cracks are nevertheless short.
After five quenching tests, more cracks are visible within the matrix and those cracks as
shown on fig. 4.115. A widening of the cracks is noticeable at the grain boundary of the
tabular alumina aggregates while in the meantime a transgranular crack is propagating
throughout the tabular alumina aggregate on the top right corner of the figure.
After ten quenching tests, as shown on figure 4.116, the crack density is considerable. In
the middle of the figure, a crack is propagating at the grain boundary of a tabular alumina
grain and throughout a zirconia particle. A transgranular crack is also visible within the
tabular alumina grain of the bottom of the figure.
264
Results
Figures 4.117 – 4.120 show the path of the cracks in the matrix of Ca-PSZ. The aim of
those pictures is to examine the nucleation of the cracks and the enhancement of the
friction mechanisms occurring after progressive thermal shock tests at the interfaces
between zirconia particles and the matrix.
Fig. 4.117: Crack nucleation within the Fig. 4.118: Crack propagation around a
matrix of formulation Ca-PSZ after 1 TS in zirconia grain within the matrix of
air at 950°C formulation Ca-PSZ after 3 TS in air at
950°C
Fig. 4.119: Increasing crack network Fig. 4.120: Crack widening at the grain
around zirconia particles within the matrix boundary of tabular alumina grains and
of formulation Ca-PSZ after 5 TS in air at zirconia grains of formulation Ca-PSZ
950°C after 10 TS in air at 950°C
As shown on figure 4.117, a crack is propagating throughout the matrix and is deviated
by zirconia particles present on the crack path. The mean distance between the crack
flanks is irregular and varies between 1 µm and 5 µm. In non-damaged area of the
microstructure, both tabular alumina and zirconia grains are well embedded in the
microstructure.
After three quenching tests, cracks are present at the level of the grain boundaries of
zirconia grains by joining the grain boundaries of several neighboring zirconia particles
265
Results
as shown on figure 4.118. Contrary to the microstructure observed previously after one
thermal shock, each zirconia exhibits a local debonding from the matrix.
After five quenching tests, the debonding of the zirconia particles is more considerable as
shown on figure 4.119. Indeed, the gap between the zirconia particle and the matrix can
reach locally 2-3 µm. Some fine transgranular cracks are also visible within the porous
structure of the zirconia particle on the bottom left corner of figure 4.119.
After ten quenching tests, cracks are propagating throughout the matrix, by being locally
deviated by small tabular alumina particles. Such cracks further propagate through
zirconia particles. A crack is visible in the middle of figure 4.120 hitting first the
elongated zirconia particle perpendicularly. It has to be noticed, that no deviation of the
crack path is observed. Then the crack is propagating through the matrix by being locally
deviated by the CA6 facets and the small tabular alumina aggregates before continuing its
propagation through the zirconia particle of the left side of the picture. A decrease of the
crack width is observed along the crack path. By analogy with the previous picture, the
crack is propagating through the porous structure of the zirconia particle.
Figures 4.121 – 4.124 show the evolution of the microstructure overview of Mg-PSZ
samples after one, three, five and ten thermal shocks. The aim of those SEM pictures is to
reveal the higher degree of microstructural damage of Mg-PSZ in comparison with Ref,
Ca-PSZ and Y-FSZ after each thermal shock cycle and to show the kinetics of the
damage mechanisms that occur early with regard to this composition Mg-PSZ.
As shown on figure 4.121, the microstructure of Mg-PSZ after one thermal shock test is
similar to the microstructure of the samples after sintering at 1500°C for 6 hours. A
considerable crack network is already present in the microstructure of Mg-PSZ. Those
big cracks are mostly propagating at the grain boundaries of big tabular alumina
aggregates.
After three quenching tests, some cracks are propagating throughout the matrix as shown
in the top left corner of the figure 4.122 and some transgranular cracks are also visible
within the tabular alumina aggregate of the top left corner.
After five quenching tests, more cracks are visible within the matrix and those cracks and
exhibit a similar crack width as previously. Tabular alumina aggregates seem to be
particularly damaged, either by exhibiting transgranular cracks like in the bottom left
corner of figure 4.123 or by revealing a local debonding due to the considerable interface
between the tabular alumina aggregate and the matrix damaged by the intergranular crack
expansion.
266
Results
After ten quenching tests, as shown on figure 4.124, the same damage mechanisms are
noticeable like previously in terms of transgranular cracks propagating throughout tabular
alumina grains and intergranular cracks propagating throughout the matrix and at the
grain boundary. Furthermore, cracks that propagate perpendicularly to the surface of
zirconia particles are also visible.
Figures 4.125 – 4.128 show the path of the cracks in the matrix of Mg-PSZ. The aim of
those pictures is to examine the evolution of the interfaces between the zirconia particles
and the matrix.
As shown on figure 4.125, a small crack is propagating throughout the matrix and hits
perpendicularly the grain boundary of the zirconia particle. A further propagation of the
crack is noticeable at the level of the grain boundary of the zirconia particle.
267
Results
After three quenching tests, similar damage mechanisms are observed as previously as
shown on figure 4.126. Indeed, a crack is propagating throughout the matrix and hits
perpendicularly the zirconia particle and continues propagating at the grain boundary of
the zirconia particle. A crack widening is furthermore observed.
After five quenching tests, cracks are still propagating at the level of the grain boundary
of the zirconia particles as shown on figure 4.127. Furthermore, some damaged areas are
also visible within the zirconia particle exhibiting a local high porosity.
After ten quenching tests, a widening of a crack propagating throughout the matrix before
hitting perpendicularly the zirconia particle is observed (see figure 4.128). The crack
width can reach locally 10 µm. The distance between the zirconia particle and the matrix
remains nevertheless almost the same (around 2-3 µm).
Fig. 4.125: Crack nucleation within the Fig. 4.126: Crack propagation around a
matrix of formulation Mg-PSZ after 1 TS zirconia grain within the matrix of
in air at 950°C formulation Mg-PSZ after 3 TS in air at
950°C
Fig. 4.127: Increasing crack network Fig. 4.128: Crack widening at the grain
around zirconia particles within the matrix boundary of tabular alumina grains and
of formulation Mg-PSZ after 5 TS in air at zirconia grains of formulation Mg-PSZ
950°C after 10 TS in air at 950°C
268
Results
Figures 4.129 – 4.132 show the evolution of the microstructure overview of Y-PSZ
samples after one, three, five and ten thermal shocks. The aim of those SEM pictures is to
reveal the low damaging level of this composition during the whole experimental
procedure in comparison with the other tested formulations.
As shown on figures 4.129, 4.130 and 4.131 the microstructure of Y-FSZ after one, 3 and
five thermal shock(s) does not exhibit any particular damage mechanisms as both tabular
alumina and zirconia particles are well embedded in the calcium aluminate matrix, which
also does not reveal any crack.
269
Results
After ten thermal shocks, the first cracks are observed and the crack density is
non-negligible. First, a crack is propagating at the grain boundary of the tabular alumina
aggregate of the middle of the picture inducing a local debonding o the aggregate by the
bottom. And a small transgranular crack is propagating throughout the tabular alumina
aggregate of the right side of the picture.
Figures 4.133 – 4.136 show the evolution of the interface between zirconia particles and
the matrix of Y-FSZ after progressive thermal shock experiments.
Fig. 4.133: Crack nucleation at the grain Fig. 4.134: Crack propagation around
boundary of a zirconia particle of zirconia grains throughout the matrix and
formulation Y-FSZ after 1 TS in air at at the grain boundary of a tabular alumina
950°C grain of formulation Y-FSZ after 3 TS in
air at 950°C
270
Results
After the first quenching test, the zirconia particle is well embedded in the matrix as
shown on figure 4.133. It is worth mentioning that the CA6-density of the area
surrounding the zirconia particle is quite considerable.
After three quenching tests, a similar microstructure is noticeable as before even if small
cracks are visible in the matrix as shown on figure 4.134. Those cracks have no real
contact with the interface of the zirconia particle.
After five quenching tests, a crack which is first propagating throughout the matrix is
then deviated by the small zirconia particle as shown on figure 4.135. The distance
between the grain boundary of the zirconia particle and the matrix is low (around 1 µm).
After ten quenching tests, crack density in the surrounding area of the zirconia particle is
considerable. A large crack is propagating throughout the matrix by being locally
deviated by small tabular alumina particles and is further deviated by a first small
zirconia particle and finally a second zirconia particle. The distance between the crack
flanks is irregular and varies between 2 and 20 µm.
271
Results
Scanning Electron Microscopy analyses are performed on Ca-PSZ samples after sintering
at 1300°C for 6 hours according to sintering profile 3 in order to examine the crack
density within the microstructure, when the thermal stresses resulting from the sintering
process are reduced and to investigate the nature of the crack flank friction phenomena at
the interface of the zirconia or tabular alumina aggregates and the matrix.
Figures 4.137 – 4.140 reveal the microstructure of a sample Ca-PSZ after sintering at
1300°C according to sintering profile 3.
272
Results
both zirconia and tabular alumina grains are well embedded in the calcium aluminate
matrix.
Some spherical pores resulting from the manufacturing process are parsimoniously and
heterogeneously present in the microstructure. However, the pores noticed on
figure 4.137 exhibit a more considerable mean diameter than that of the sample Ca-PSZ
after sintering at 1500°C, with a mean diameter above 500 µm. A stronger magnification
shown on figures 4.138 and 4.139 reveals that the interfaces between the tabular alumina
and zirconia aggregates and the matrix do not show any crack, even if sometimes some
pores resulting from the manufacturing process are visible at those interfaces.
The stronger magnification shown on figure 4.139 reveals a matrix, which is composed
of some CA6 facets embedded in CA2 crystals that look like not so crystallized that the
CA6 facets. A microcrack-free matrix composed essentially of CA2 constitutes the
preponderant discrepancy between the Ca-PSZ sample sintered at 1300°C as shown on
figure 4.140 in comparison with the sample Ca-PSZ sintered at 1500°C as shown on
figure 4.91.
273
Results
identical, the same closed porosity is measured with 2,3 vol-% for a sintering temperature
of 1300°C vs 2,3 vol-% for a sintering temperature of 1500°C. Therefore, a difference of
0,5 vol-% is also observed with regard to the total porosity.
Table 4.7: Structural, mechanical and elastic properties of Ca-PSZ after sintering at
1300°C for 6 hours according to sintering profile 3 and at 1500°C for 6 hours
according to sintering profile 2
Ca-PSZ (1300°C) Ca-PSZ (1500°C)
πa [vol-%] 18,9 ± 0,4 19,4 ± 0,2
πf [vol-%] 2,3 ± 0,0 2,3 ± 0,0
πt [vol-%] 21,2 ± 0,4 21,7 ± 0,2
ρb [g.cm-3] 3,2 ± 0,0 3,2 ± 0,0
CCS [MPa] 156 ± 11 128 ± 7
MOR [MPa] 27 ± 1 20 ± 1
HMOR at 1300°C [MPa] 14 ± 1 14 ± 2
Young’s modulus E [GPa] 120 ± 2 66 ± 2
Flexural damping 𝜁𝑓𝑓 [*10-3] 0,26 ± 0,06 1,61 ± 0,19
274
Results
elastic properties. Thus, Ca-PSZ sintered at 1300°C exhibits a stiffness 54 GPa higher
than the same formulation sintered at 1500°C, with 120 GPa and 66 GPa respectively.
Like the CCS, this increase of the Young’s modulus is partly explained by the lower
apparent porosity and mostly caused by the crack-free microstructure of the castable as
shown previously in chapter 4.4.1, even if the transformation of CA2 phases into CA6
phases should lead to a stiffener material. The damping behavior is in agreement with the
material stiffness. Formulation Ca-PSZ sintered at 1300°C exhibiting a crack-free
microstructure reveals a low damping value with 0,26.10-3 [-], while the same
formulation sintered at 1500°C exhibiting a non-negligible crack density after sintering
reveals a considerable damping value with 1,61.10-3 [-]. It is worth mentioning that the
damping values listed in table 4.7 are obtained through RT-RFDA measurements with the
help of the room temperature sample support. Therefore, the damping value of Ca-PSZ
sintered at 1500°C (1,61.10-3 [-]) is much lower than the damping value listed in table 4.2,
namely 12,71.10-3 [-], as those values are obtained RT-RFDA measurements with the
help of the sample support for high temperature measurements exhibiting higher damping
properties.
Thermal shock experiments in air at 950°C according to DIN EN 993-11 are performed
on samples of Ca-PSZ after sintering either at 1500°C or 1300°C. Resonant Frequency
Damping Analysis is performed on each tested prismatic sample after each thermal shock
cycle in flexural/torsion mode in order to measure the samples Young’s modulus, shear
modulus and Poisson’s ratio as well as the damping of the flexural and torsion resonant
frequencies. The focus of this chapter is on the evolution of the Young’s modulus as well
as of damping of the flexural resonant frequency. By analogy with the thermal shock
resistance of Ref, Ca-PSZ, Mg-PSZ and Y-FSZ, both approaches with regard to the
evolution of the absolute and relative stiffness behaviours are presented in this part of the
study.
Figure 4.141 shows the evolution of the Young’s modulus after progressive thermal
shocks in air at 950°C of the Ca-PSZ according to the sintering temperature, while
figure 4.142 shows the evolution of the relative Young’s modulus after progressive
thermal shocks in air at 950°C of the tested formulations according to the sintering
temperature.
275
Results
Fig. 4.141: Stiffness evolution of Ca-PSZ after progressive thermal shocks in air at
950°C according to DIN EN 993-11 in function of the sintering temperature
Fig. 4.142: Evolution of the relative stiffness of Ca-PSZ after progressive thermal shocks
in air at 950°C according to DIN EN 993-11 in function of the sintering temperature
The evolution of the Young’s modulus of Ca-PSZ sintered at 1500°C has already been
previously examined in chapter 4.3.2.1. The stiffness of Ca-PSZ follows a two-step
decrease with increasing thermal shock cycles. Starting from Young’s modulus values
276
Results
after sintering of 66 GPa, the stiffness undergoes a significant decrease to the sixth
thermal shock cycle to reach 37 GPa. The second evolution step corresponds to constant
values of Young’s modulus. The final stiffness of Ca-PSZ is 37 GPa after the whole
series of thermal shocks. The stiffness decrease after ten thermal shock cycles is 28 GPa.
As far as formulation Ca-PSZ sintered at 1300°C is concerned, a traditional two-step
decrease of the Young’s modulus is observed. Starting from an initial Young’s modulus
of 120 GPa, the stiffness strongly decreases up to the third thermal shock cycle to achieve
a value of 81 GPa and exhibits a low decrease from this third thermal shock cycle to the
tenth cycle to reach a final value of 69 GPa.
Even if Ca-PSZ sintered at 1300°C exhibits a Young’s modulus almost two times higher
than that of Ca-PSZ sintered at 1500°C after the sintering, it is remarkable to observe that
both materials reveal a similar evolution of the relative stiffness as shown on
figure 4.142.
On the one hand, Ca-PSZ sintered at 1500°C follows a three-step decrease with
increasing thermal shock cycles to reach final relative Young’s modulus values that are
identical to those of Ca-PSZ sintered at 1300°C (57 %). The stiffness undergoes a
significant decrease after the first thermal shock cycle to reach 81 %. The second
decrease of the Young’s modulus is characterized by a lower rate from the first thermal
shock cycle to the sixth thermal shock cycle to reach 57 %. The third evolution step
corresponds to constant values of Young’s modulus until that the formulations Ca-PSZ
exhibits the final stiffness values previously detailed. The stiffness decrease after ten
thermal shock cycles is 41 % for Ca-PSZ sintered at 1500°C.
On the other side, Ca-PSZ sintered at 1300°C follows a traditional two-step decrease of
the relative Young’s modulus. The stiffness strongly decreases up to the third thermal
shock cycle to achieve a value of 67 % and exhibits a low decrease from this third
thermal shock cycle to the tenth cycle to reach a final value of 57 %.
The damping behavior of Ca-PSZ sintered at 1300°C and 1500°C after progressive
thermal shocks in air at 950°C is analyzed through Resonant Frequency Damping
Analysis. The focus is on the damping behavior of the flexural resonant frequency. By
analogy with the thermal shock resistance of Ref, Ca-PSZ, Mg-PSZ and Y-FSZ
previously examined, both approaches with regard to the evolution of the absolute and
relative damping behaviours are presented in this part of the study. Figure 4.143 shows
the evolution of the absolute damping behaviour after progressive thermal shocks in air at
950°C of the Ca-PSZ according to the sintering temperature, while figure 4.144 shows
277
Results
the evolution of the relative damping behaviour after progressive thermal shocks in air at
950°C of the tested formulations according to the sintering temperature.
Fig. 4.143: Evolution of the flexural damping of Ca-PSZ after progressive thermal
shocks in air at 950°C according to DIN EN 993-11 in function of the sintering
temperature
The evolution of the absolute flexural damping of Ca-PSZ sintered at 1500°C has already
been previously examined in chapter 4.3.2.2 in case of damping measurements performed
with the help of the sample support for high temperature measurements.
On the one hand, Ca-PSZ sintered at 1500°C is characterized by a high initial damping
value of 1,61.10-3 [-]. The damping follows a drastic increase after the first thermal shock
cycle to reach 3,25.10-3 [-] followed by a decrease after the second thermal shock cycle
up to 2,12.10-3 [-]. The damping continuously increases afterwards up to the fifth thermal
shock cycle to reach a damping value of 4,12.10-3 [-]. The damping remains constant up
to the ninth thermal shock cycle with a value of 3,79.10-3 [-] and finally increases after
the last and tenth thermal shock cycle up to 4,58.10-3 [-].
On the other hand, Ca-PSZ sintered at 1300°C is characterized by a low initial damping
value of 0,26.10-3 [-], which is more than 6 times lower than the initial flexural damping
value of Ca-PSZ sintered at 1500°C. The damping follows a first continuous and regular
increase up to the third thermal shock cycle to reach a value of 3,36.10-3 [-] and a second
low increase up to the tenth thermal shock cycle to reach a final absolute value of
3,90.10-3 [-]. Even though the formulation Ca-PSZ sintered at 1300°C exhibits much
lower damping values than the same formulation sintered at 1500°C after sintering, both
278
Results
formulations reveal similar absolute damping values from the second thermal shock cycle
to the tenth.
Fig. 4.144: Evolution of the relative flexural damping of Ca-PSZ after progressive
thermal shocks in air at 950°C according to DIN EN 993-11 in function of the sintering
temperature
Contrary to the relative Young’s modulus of the formulation Ca-PSZ, which seems to be
independent of the sintering temperature as shown previously on figure 4.141, the
damping behavior of the flexural resonant frequency strongly depends on the sintering
temperature as shown on figure 4.144.
On the one hand, Ca-PSZ sintered at 1500°C exhibits lower relative damping values than
that of formulation Ca-PSZ sintered at 1300°C after each thermal shock cycle. The
relative damping of Ca-PSZ sintered at 1500°C continuously and regularly increases up
to the fifth thermal shock cycle to achieve 256 % and follows a second and lower
increase up to the tenth thermal shock cycle to reach 284 %.
On the other hand, Ca-PSZ sintered at 1300°C exhibits high relative damping values after
progressive thermal shocks. The relative damping drastically increases up to the third
thermal shock cycle to reach 1306 %, then slowly decreases up to the eighth thermal
shock cycle to reach 1537 %, afterwards strongly increases after the ninth thermal shock
cycle to reach 1853 % corresponding to the maximum relative damping value during the
whole series of the thermal shocks, and finally decreases after the ninth thermal shock to
reach 1518 %, which is above five times higher than the final relative flexural damping
of Ca-PSZ sintered at 1500°C.
279
Results
280
Results
281
Results
After 10 thermal shock cycles, some big cracks are propagating throughout the matrix
and more particularly at the grain boundaries of tabular alumina and zirconia grains. The
main crack visible on figure 4.149 also causes a local fragmentation of the tabular
alumina aggregate in the top right corner of the figure 4.149. It may be assumed that the
big crack propagates throughout the tabular alumina grain as transgranular crack and then
continues propagating throughout the matrix. The analysis of the interface between the
zirconia grain and the matrix in presence of crack shows that the crack is deviated by the
zirconia particle. The crack also causes local damage of zirconia particle as shown on
figure 4.150, so that small zirconia crystals are localized in the crack path between the
main zirconia particle and the calcium aluminate matrix. A crack widening is observed
after the tenth thermal shock in comparison with the crack observed after five thermal
shocks with a crack width of around 5 µm up to 10 µm.
The influence of the temperature difference during thermal shock experiment is examined
on Ca-PSZ samples sintered at 1300°C for 6 hours according to sintering profile 3. The
thermal shock experiments are performed in air according to DIN EN 993-11 at four
different testing temperatures: 750°C, 850°C, 950°C and 1050°C. An increasing
microcracking of the microstructure of the castable is expected with increasing
temperature difference as well as the enhancement of friction phenomena. The aim of this
part of the study is to quantify the microstructural damage in function of the external
thermal stresses with the help of Resonant Frequency Damping Analysis. It is worth
mentioning that the testing thermal shock temperatures are all below the martensitic
transformation of zirconia, so as the effects of the martensitic transformation are not
included in the microstructural damage phenomena at the origin of the degradation of the
elastic properties.
282
Results
Table 4.8: Young’s modulus and damping of the flexural resonant frequency of Ca-PSZ
samples required for thermal shock experiments at 750°C, 850°C, 950°C and
1050°C
TTS = 750°C TTS = 850°C TTS = 950°C TTS = 1050°C
Young’s modulus E [GPa] 113 ± 1 114 ± 2 120 ± 4 114 ± 3
Flexural damping 𝜁𝑓𝑓 [*10-3] 0,33 ± 0,03 0,33 ± 0,04 0,26 ± 0,06 0,31 ± 0,03
As shown on table 4.8, the Ca-PSZ samples required for thermal shocks at 750°C, 850°C,
950°C and 1050°C exhibit same elastic properties with Young’s modulus ranging
between 112 GPa and 119 GPa and damping values of the flexural resonant frequency
comprised between 0,26.10-3 [-] and 0,33.10-3 [-]. The set of samples for thermal shocks
at 950°C distinguishes itself slightly by exhibiting a higher Young’s modulus with
120 GPa and a lower damping of the flexural resonant frequency with 0,26.10-3 [-], while
the Young’s modulus of the other sets of samples is around 113 GPa and the damping of
the flexural resonant frequency around 0,32.10-3 [-].
Resonant Frequency Damping Analysis is performed on each sample after each thermal
shock for each testing temperature difference in flexural/torsion mode to examine the
evolution of the Young’s modulus, the shear modulus, the Poisson’s ratio and the
damping of the flexural and torsion frequencies after progressive thermal shocks. The
focus of this part of the study is on the evolution of the Young’s modulus and the
damping of the flexural resonant frequency after progressive thermal shocks according to
the temperature difference. As shown on table 4.8, as the samples exhibit similar
Young’s modulus values after the sintering at 1300°C for 6 hours, it can be assumed that
the examination of the resulting Young’s modulus is similar to the examination of the
relative Young’s modulus. Figure 4.151 shows the evolution of the relative Young’s
modulus after progressive thermal shocks in air according to the temperature difference
of thermal shock.
As shown on figure 4.151, the degradation of the Young’s modulus directly depends on
the severity of the thermal shock: the higher the temperature difference of thermal shock,
the higher the decrease of the stiffness. For each testing temperature, a traditional two-
step exponential evolution is noticeable with a strong decrease of the Young’s modulus
during the first three thermal shocks followed by a low decrease up to the tenth thermal
shock.
283
Results
Concerning the thermal shock tests at 750°C, the strong decrease of the Young’s modulus
takes place during the first thermal shock to reach a relative value of 91 %. After this first
thermal shock cycle, the Young’s modulus slightly decreases up to the tenth thermal
shock to reach an absolute value of 82 %.
Concerning the thermal shock tests at 850°C, the two-step decrease of the Young’s
modulus is noticeable with a residual Young’s modulus after three thermal shock cycles
of 79 % followed by a low stiffness decrease up to the tenth thermal shock to reach 74 %.
In case of thermal shock test at 950°C, the degradation of the Young’s modulus is clearly
more considerable with a residual Young’s modulus after three thermal shock cycles of
67 % followed by a low stiffness decrease up to the tenth thermal shock to reach 57 %.
Finally, as for thermal shocks at 1050°C, the highest decrease of stiffness is observed
with a residual Young’s modulus after three thermal shock cycles of 53 % followed by a
low stiffness decrease up to the ninth thermal shock to reach 43 %. The tenth thermal
shock results in sample fracture, and more particularly to a partial sample fracture or
questionable residual Young’s modulus values.
Fig. 4.151: Evolution of the relative Young’s modulus of Ca-PSZ after progressive
thermal shocks in air at 750°C, 850°C, 950°C and 1050°C
284
Results
The focus of this part of the study is henceforth shifted on the evolution of the damping
of the flexural resonant frequency after progressive thermal shocks according to the
temperature difference. As shown on table 4.8 and by analogy with the Young’s
modulus, as the samples exhibit similar flexural damping values after the sintering at
1300°C for 6 hours, it can be assumed that the examination of the resulting flexural
damping is similar to the examination of the relative flexural damping. Figure 4.152
shows the evolution of the relative damping of the flexural resonant frequency after
progressive thermal shocks in air according to the temperature difference of thermal
shock.
4
3
Fig. 4.152: Evolution of the relative damping of the flexural resonant frequency of
Ca-PSZ after progressive thermal shocks in air at 750°C, 850°C, 950°C and 1050°C
As shown on figure 4.152 and in accordance with the evolution of the material stiffness,
the damping increase directly depends on the severity of the thermal shock: the higher the
temperature difference of thermal shock, the higher the ensuing crack flank friction.
Contrary to the evolution of the Young’s modulus, the evolution trend is different
according to the temperature difference.
For low temperature differences, namely 750°C and 850°C, the relative damping of the
flexural resonant frequency increases quite linearly with increasing thermal shocks to
reach 357 % and 797 % respectively after the whole series of thermal shocks.
285
Results
In order to examine the possible correlation between the degradation of the stiffness and
the increase of the crack flank friction after progressive thermal shocks, the logarithm of
the resulting damping of the flexural resonant frequency is plotted in function of the
logarithm of the retained Young’s modulus according to the temperature difference of
thermal shock.
The Young’s modulus – flexural damping relationship is examined on figure 4.153. The
aim of such a diagram is to correlate the two-step decrease of the resulting Young’s
modulus and the complex enhancement of the crack flank friction phenomena after
progressive thermal shocks.
The observations mentioned previously relative to the strong degradation of the stiffness
and the considerable increase of the crack flank friction with increasing temperature
difference are noticeable on figure 4.153. Indeed, the red curve corresponding to the
stiffness-damping response after thermal shocks at 1050°C occupies the top left corner of
the diagram referring to lower Young’s modulus and higher damping values while the
286
Results
This proves that it is challenging to foresee the stiffness and damping behaviors of
refractory castables after progressive thermal shocks in air beyond a temperature
difference threshold, which is between 850°C and 950°C.
287
Results
Fig. 4.154: Crack nucleation at grain Fig. 4.155: Crack propagation within the
boundaries of PSZ castable after one matrix and throughout the Al2O3
thermal shock cycle in air at 1050°C aggregates of PSZ castable after three
(magnification: 2000) thermal shock cycles in air at 1050°C
(magnification: 60)
Fig. 4.156: Crack deviation along zirconia Fig. 4.157: Cracks’ unification and crack
grains of PSZ castable after five thermal propagation throughout ZrO2 aggregates of
shock cycles in air at 1050°C PSZ castable after nine thermal shock
(magnification: 500) cycles in air at 1050°C
(magnification: 500)
288
Results
As shown on figure 4.154, the first microstructural damage stage after one thermal shock
cycle corresponds to the nucleation of cracks at the grain boundaries of tabular alumina
grains. Such small cracks with a width below 5 µm are successively deviated by the small
tabular alumina grains within the matrix.
The next damage mechanism illustrated on figure 4.155 occurring after three thermal
shocks is the nucleation of transgranular cracks throughout tabular alumina aggregates.
Those grains are propagating at the phase boundaries of the tabular alumina polycrystal.
The mean distance between the crack flanks is similar to the intergranular cracks
described previously and does not exceed 5 µm.
After five thermal shock cycles, the intergranular crack is deflected or deviated by the
partially stabilized zirconia grains that apply redial stresses in the neighbouring area of
the grain as shown on figure 4.156. The crack first perpendicularly hits the partially
stabilized zirconia grain and propagates at the interface of the grain with the matrix
causing a partial debonding of the grain and finally continues its propagation
perpendicularly to the grain surface in the exact direction of the initial crack path. After
five thermal shocks, some cracks are locally widening with a width above 10 µm.
After the whole series of thermal shocks, the damage of zirconia grains is noticeable as
shown on figure 4.157. When the crack hits the surface of a zirconia grain, the crack is
not deviated like previously mentioned but is propagating throughout the zirconia
monocrystalline grain causing a partial fragmentation of the grain.
After examining the thermal shock resistance under heating and cooling thermal shock
conditions provided by thermal shock in air according to DIN EN 993-11, the elastic
behavior of Ref, Ca-PSZ and Y-FSZ within the framework of heating thermal shock is
examined by performing quenching tests in molten aluminum at 800°C as explained in
chapter 3.2.7.3. As zirconia exhibits a higher coefficient of thermal expansion than
alumina, by neglecting the martensitic transformation in first instance, zirconia particles
contract to a higher extent than the matrix during the cooling phase inducing
circumferential compressive stresses and radial tensile stresses. Since in ceramics cracks
occur most of the time in tension, this special case will induce debonding around the
zirconia particles. On the other side, zirconia particles expand to a higher extent than the
matrix during the heating phase inducing circumferential tensile stresses and radial
compressive stresses causing a radial matrix microcracking around the zirconia particles.
The aim of the this part of the study is to examine the damage mechanisms occurring at
the level of the grain boundary of zirconia particle and to reveal the discrepancy in terms
289
Results
of thermal shock resistance of formulations Ref, Ca-PSZ and Y-FSZ under heating
thermal shock conditions.
Table 4.9: Young’s modulus and damping of the flexural resonant frequency of
formulations Ref, Y-FSZ and Ca-PSZ samples required for thermal shock
experiments at 950°C according to DIN EN 993-11 and quenching test in
molten aluminum at 800°C
Type of thermal Young’s modulus Flexural damping 𝜁𝑓𝑓
Formulations
shocks [GPa] [*10-3]
DIN EN 993-11 147 ± 2 0,32 ± 0,07
Ref
Molten Aluminum 140 ± 3 0,43 ± 0,04
DIN EN 993-11 120 ± 4 0,53 ± 0,06
Ca-PSZ
Molten Aluminum 114 ± 3 0,62 ± 0,06
DIN EN 993-11 125 ± 2 0,58 ± 0,04
Y-FSZ
Molten Aluminum 124 ± 2 0,58 ± 0,09
As shown on table 4.9, the samples elaborated for quenching tests in molten aluminum at
800°C exhibit similar elastic properties as the samples elaborated for thermal shock tests
in air at 950°C according to DIN EN 993-11. It is worth mentioning that those damping
measurements are performed with the room temperature RFDA testing device. This
explains the low damping values in comparison with those obtained in chapter 4.3.2, as
those last measurements were performed with the high temperature RFDA testing device.
Despite the significant similarities in terms of elastic properties between the samples
quenched in air or in molten aluminum, it has to be noted that the samples elaborated for
quenching test in molten aluminum exhibit a lower stiffness than those for thermal shock
in air for each formulation. Indeed, Ref samples for quenching tests in molten aluminum
reveal a stiffness 7 GPa lower than the samples for thermal shocks in air with 140 GPa
and 147 GPa respectively. Ca-PSZ samples for quenching tests in molten aluminum
reveal a stiffness 5 GPa lower than the samples for thermal shocks in air with 114 GPa
and 120 GPa respectively. Y-FSZ samples for quenching tests in molten aluminum reveal
290
Results
a stiffness 1 GPa lower than the samples for thermal shocks in air with 124 GPa and
125 GPa respectively. The lower stiffness of the samples elaborated for quenching test in
molten aluminum is in agreement with the higher damping values shown by those
samples. Indeed, Ref samples for quenching tests in molten aluminum reveal a flexural
damping 0,11.10-3 [-] higher than the samples for thermal shocks in air with 0,43.10-3 [-]
and 0,32.10-3 [-] respectively. Ca-PSZ samples for quenching tests in molten aluminum
reveal a flexural damping 0,09.10-3 [-] higher than the samples for thermal shocks in air
with 0,62.10-3 [-] and 0,53.10-3 [-] respectively. Y-FSZ samples for quenching tests in
molten aluminum reveal a similar flexural damping as the samples for thermal shocks in
air with 0,58.10-3 [-]. Therefore, owing to the low discrepancy with regard to the elastic
properties, it is meaningful to examine the evolution of the relative Young’s modulus and
relative damping of the flexural resonant frequency after progressive quenching tests in
molten aluminum.
4.6.1 Evolution of the relative Young’s modulus of Ref, Ca-PSZ and Y-FSZ
after progressive quenching tests in molten aluminum at 800°C
Resonant Frequency Damping Analysis is performed on each sample after quenching test
in molten aluminum after cooling up to room temperature. The measurements are
performed in flexural/torsion mode for the assessment of the Young’s modulus, the shear
modulus and the Poisson’s ratio as well as the damping of the flexural and torsion
resonant frequencies. In the following part of the study, the focus is on the evolution of
the relative Young’s modulus and relative damping of the flexural resonant frequency
after progressive quenching test. Figure 4.158 shows the evolution of the relative
Young’s modulus of Ref, Ca-PSZ, and Y-FSZ after quenching tests in molten aluminum
at 800°C. During such quenching tests, castables are subjected to heating thermal shock
when soaked into molten aluminum and also cooling thermal shock during cooling up to
room temperature. Nevertheless, the damage resulting from the cooling are considered as
being negligible compared to the damage resulting from the quenching into the molten
aluminum, as the sample cooling rate is low. Therefore, the resulting elastic and
mechanical properties measured in the present part are discussed by supposing that the
damage are created during the heating thermal shock.
It is noticeable that the tested formulations do not exhibit a similar stiffness evolution
after quenching tests in molten aluminum at 800°C as after thermal shocks tests in air
according to DIN EN 993-11. The nature of thermal shock, either heating or cooling, has
a considerable impact on the microstructural damage of the refractory castable samples.
Indeed, while Ca-PSZ exhibits a strong decrease of the relative Young’s modulus after
thermal shocks in air, this formulation reveal a more suitable thermal shock resistance in
291
Results
case of quenching tests in molten aluminum. Ca-PSZ shows a significant decrease of the
relative Young’s modulus, but this decrease is not as significant as the decrease of the
relative Young’s modulus of Ref and Y-FSZ. The relative Young’s modulus of Ca-PSZ
reaches 93 % after the first thermal shock and then follows a quite constant and low
decrease up to the tenth thermal shock to achieve a final relative value of 83 %.
Concerning Ref, the relative Young’s modulus strongly decreases after the first thermal
shock cycle to reach a relative value of 90 % and then decreases quite linearly up to the
tenth thermal shock cycle to reach a final value of 75 % corresponding to the lowest
retained Young’s modulus among the three tested formulations.
Y-FSZ is characterized by a similar considerable decrease after one thermal shock with a
relative Young’s modulus value of 88 %. However, the second decrease of stiffness of
the two-step decrease evolution is not as considerable as the decrease of stiffness of Ref.
The linear decrease of the relative Young’s modulus is around 1,1 % per thermal shock
cycle to reach a final value of 79 %. It is noticeable that the decrease of the relative
Young’s modulus in this second step of the two-step evolution of Y-FSZ is the lowest in
comparison with Ca-PSZ and Ref with a value of 1,2 % per thermal shock cycle and
1,6 % per thermal shock cycle respectively.
Fig. 4.158: Evolution of the relative Young’s modulus of Ref, Ca-PSZ and Y-FSZ after
quenching tests in molten aluminum at 800°C
292
Results
The damping behavior of the flexural resonant frequency of Ref, Ca-PSZ and Y-FSZ
after progressive quenching tests in molten aluminum is examined through Resonant
Frequency Damping Analysis and is plotted on figure 4.159. A direct correlation between
the damping of the flexural resonant frequency and the Young’s modulus is noticeable as
formulation Ca-PSZ exhibiting a low decrease of the Young’s modulus also reveals a low
damping increase, and conversely, formulations Ref and Y-FSZ exhibiting a high
decrease of the Young’s modulus also reveals a significant damping increase.
Fig. 4.159: Evolution of the relative damping of the flexural resonant frequency of Ref,
Ca-PSZ and Y-FSZ after quenching tests in molten aluminum at 800°C
As far as Ca-PSZ is concerned, the relative damping of the flexural resonant frequency
first increases up to the third quenching test to reach a value of 386 %. Then the damping
remains constant up to the fifth thermal shock cycle with a value of 402 %. The damping
increases again after the sixth quenching test to reach a value of 496 % and finally
remains constant up to the last quenching test with a value of 493 % corresponding to the
lowest resulting damping value after the whole series of quenching tests among the three
tested formulations.
In case of Ref, the relative damping considerably increases up to the fifth thermal shock
to reach a value of 1226 % and remains constant up to the tenth thermal shock cycle with
293
Results
a final relative damping value of 1192 %, except for the seventh thermal shock cycle
characterized by a local high relative damping with 1525 % corresponding to the highest
relative damping exhibited by the formulation. The final relative damping value of Ref is
2,4 times higher than the final relative damping value of Ca-PSZ.
Concerning Y-FSZ, the relative damping of the flexural resonant frequency continuously
and progressively increases after the quenching tests in molten aluminum with the
exception of two local minima after two and nine thermal shocks with a relative damping
value of 290 % and 991 % respectively. Y-FSZ exhibits after ten quenching test a relative
damping value of 1386 %, which is 2,8 times higher than the final relative damping value
of Ca-PSZ.
4.6.3 Comparison of the crack distribution within the samples of Ca-PSZ after
thermal shock tests in air according to DIN EN 993-11 and after
quenching test in molten aluminum at 800°C
The heterogeneous distribution of cracks may have an influence on the assessment of the
resulting elastic properties after thermal shock experiments. In case of the thermal shock
tests in air, and as examined in chapter 4.2.1, the density of cracks generated by the
abrupt cooling of the upper surface of the samples follows a gradient in the sample
thickness and is more and more considerable when the distance to the upper surface of
the sample is lower. In case of quenching test in molten aluminum, all surfaces of the
sample are in contact with the annealing medium and a more homogeneous distribution
of cracks within the sample microstructure is expected.
To verify this expectation, the resulting Poisson’s ratio after progressive thermal shocks
in air according to DIN EN 993-11 and after progressive quenching tests in molten
aluminum is assessed through Resonant Frequency Damping Analysis by determining the
Young’s modulus and the shear modulus in flexural/torsion mode.
The comparison of the resulting Poisson’s ratio according to the nature of thermal shock
experiment is shown on figure 4.160 in case of Ca-PSZ sintered at 1300°C for 6 hours.
Measured negative Poisson’s ratio values do not have any scientific significance in the
presence study. They only show the heterogeneous crack distribution within the sample
microstructure.
Even if the degradation of the Young’s modulus after thermal shock tests in air of
Ca-PSZ is more considerable than the decrease of the Young’s modulus after quenching
tests in molten aluminum, i.e. the crack density is not comparable between both thermal
shock procedures, it is noticeable that the Poisson’s ratio shows a more considerable
decrease after thermal shock tests in air than after quenching tests in molten aluminum.
Indeed, the Poisson’s ratio of Ca-PSZ shows a first significant decrease after one
294
Results
quenching test in molten aluminum to reach a value of 0,10 [-], while the initial value
after sintering is 0,16 [-]. Then the decrease of the Poisson’s ratio is lower up to the last
quenching test. The Ca-PSZ samples exhibit a final mean Poisson’s ratio of 0,04 [-].
As far as the thermal shock experiments in air are concerned, the Poisson’s ratio
considerably decreases up to the third thermal shock to reach a negative value (-0,01 [-]).
The Poisson’s ratio slowly decreases up to the sixth thermal shock cycle to reach a value
of -0,03 [-]. After the seventh thermal shock cycle, the Poisson’s ratio strongly decreases
to reach a value of -0,07 [-] and this value remains constant up to the last thermal shock
cycle. A difference of 0,11 [-] is noticeable after the whole series of thermal shock tests
in air and after ten quenching tests in molten aluminum. Such a discrepancy of behavior
with regard to the Poisson’s ratio is in agreement with the expected more homogeneous
distribution of cracks after quenching tests in molten aluminum than after thermal shock
experiments in air.
Fig. 4.160: Evolution of the Poisson’s ratio of Ca-PSZ sintered at 1300°C for 6 hours
after thermal shock tests in air at 950°C and after quenching tests in molten aluminum at
800°C
295
Results
tests in molten aluminum at 800°C and to investigate the nature of the crack flank friction
phenomena at the interface of the zirconia or tabular alumina aggregates and the matrix.
The aim of the SEM analyses is also to examine if the microstructural damage
mechanisms caused by heating thermal shocks are of a different nature than those
resulting from cooling thermal shocks.
Figures 4.161 – 4.164 show the evolution of the microstructure overview of Ref samples
after one, three, five and ten thermal shocks. The aim of those SEM pictures is to reveal
the increasing crack density with increasing quenching tests in molten aluminum at
800°C.
296
Results
As shown on figure 4.161, after one thermal shock cycle, only few cracks are nucleating
parsimoniously within the matrix. On the top right corner of figure 4.161, cracks with a
considerable distance between the crack flanks are visible. Those cracks hit
perpendicularly the surface of a tabular alumina aggregate.
Figure 4.162 shows the microstructure of Ref after three quenching tests. A
homogeneous distribution of the cracks within the microstructure is noticeable.
On figure 4.163, in addition to intergranular cracks, a transgranular crack is visible
throughout the tabular alumina aggregate on the top left corner after five quenching tests.
Finally, figure 4.164 shows the considerable crack density in the microstructure of Ref
after ten quenching tests. Big cracks are propagating throughout the sample at the grain
boundaries of tabular alumina aggregates causing a partial debonding of those
aggregates. Besides the crack density increase, a widening of the cracks is also observed.
Fig. 4.165: Crack nucleation within the Fig. 4.166: Crack propagation around
matrix of formulation Ref after one tabular alumina grains within the matrix of
quenching test in molten aluminum at formulation Ref after three quenching tests
800°C in molten aluminum at 800°C
Fig. 4.167: Crack widening within the Fig. 4.168: Crack widening at the grain
matrix of formulation Ref after five boundary of tabular alumina grain of
quenching tests in molten aluminum at formulation Ref after ten quenching tests in
800°C molten aluminum at 800°C
297
Results
Figures 4.168 – 4.168 show the path of the cracks in the matrix of Ref. The aim of those
pictures is to examine the nucleation of the cracks and the enhancement of the friction
mechanisms occurring after progressive quenching tests.
On figure 4.165, a small crack with a width of around 3 µm is propagating throughout the
CA6-rich matrix and perpendicularly to a tabular alumina grain. A radial matrix
microcracking around the tabular alumina particle takes place because of the
circumferential tensile stresses and radial compressive stresses present at the interfaces of
the tabular alumina grains during the heating phase.
As shown on figure 4.166, after three quenching tests, the increase of those stresses
causes the expansion of the crack within the matrix and also the widening of the crack.
The mean distance between the crack flanks is around 5 µm. The crack is then
propagating at the interface of another tabular alumina grain, so that the crack path is
continuously deviated by the tabular alumina particles.
After five quenching tests, the crack width increases up to 7-8 µm as shown on
figure 4.167. It could be assumed that the crack widening first takes place within the
CA6-rich matrix that provides a lower resistance to damage than the tabular alumina
grains.
After 10 quenching tests, the crack width reaches its maximum value, above 10 µm, and
crack widening takes place at the grain boundary of tabular alumina aggregate as shown
on figure 4.168.
Figures 4.169 – 4.172 show the evolution of the microstructure overview of Ca-PSZ
samples after one, three, five and ten thermal shocks. The aim of those SEM pictures is to
reveal increasing crack density in a lesser extent with increasing quenching tests in
molten aluminum at 800°C than previously as far as Ref was concerned. Independently
of the number of quenching tests, the crack density within the microstructure of Ca-PSZ
is much lower than that of Ref.
As shown on figure 4.169, the microstructure of Ca-PSZ after one quenching test is
similar to the microstructure of the samples after sintering at 1300°C for 6 hours. Tabular
alumina and zirconia grains are well embedded in the calcium aluminate matrix and no
crack is visible.
After three quenching tests, some cracks are propagating throughout the matrix as shown
in the middle of the figure 4.170 and also perpendicularly to the surface of a zirconia
grain as shown on the top of the picture. Those cracks are nevertheless short and reveal
however a non-negligible crack width.
298
Results
After five quenching tests, more cracks are visible within the matrix and those cracks,
even if they exhibit a similar crack width as previously, are much longer than after three
quenching tests. The cracks either propagate throughout the matrix as shown on the top
right corner of figure 4.171 or perpendicularly to the surface of tabular alumina grains as
shown on the top left corner of figure 4.171.
After ten quenching tests, as shown on figure 4.172, the crack density is more
considerable and a crack widening is also noticeable. In the middle of the figure, cracks
are clearly propagating perpendicularly to the surface of a tabular alumina grain and a
zirconia grain. A partial debonding of the zirconia particle may also take place.
Figures 4.173 – 4.176 show the path of the cracks in the matrix of Ca-PSZ. The aim of
those pictures is to examine the nucleation of the cracks and the enhancement of the
friction mechanisms occurring after progressive quenching tests.
299
Results
As shown on figure 4.173, small and short cracks with characteristic length of around
20 µm and width inferior to 1 µm are nucleating within the matrix and perpendicularly to
the surface of tabular alumina and zirconia particles after the first quenching test.
Furthermore, it appears that zirconia particles are slightly debonded from the matrix.
After three quenching tests, a crack expansion as well as a crack widening is observed as
shown on figure 4.174. Indeed, the mean distance between the crack flanks can locally
reach 10 µm. The main crack is emerging perpendicularly at the surface of the zirconia
particle. Then the crack is deviated by a small tabular alumina particle and is finally
widening in the CA2-rich matrix.
Fig. 4.173: Crack nucleation within the Fig. 4.174: Crack propagation around a
matrix of formulation Ca-PSZ after one zirconia grain within the matrix of
quenching test in molten aluminum at formulation Ca-PSZ after three quenching
800°C tests in molten aluminum at 800°C
Fig. 4.175: Increasing crack network Fig. 4.176: Crack widening at the grain
around zirconia particles within the matrix boundary of tabular alumina grains and
of formulation Ca-PSZ after five zirconia grains of formulation Ca-PSZ
quenching tests in molten aluminum at after ten quenching tests in molten
800°C aluminum at 800°C
300
Results
After five quenching tests, several cracks are propagating in the surrounding area of
zirconia particles and all those cracks are emerging perpendicularly at the surface of the
zirconia grain. When those cracks meet a tabular alumina particle during the propagation,
the crack further propagates at the grain boundary of the tabular alumina grain as shown
on figure 4.175.
After ten quenching tests, a widening of a dense crack network in the surrounding area of
zirconia particles can be observed. If crack width after five quenching tests is around
5 µm, it can reach locally a value around 10 µm as shown on figure 4.176. The crack
density is so high that small tabular alumina and zirconia particles are partially debonded
from the matrix. The crack that propagates in the middle of figure 4.176 is perpendicular
to the zirconia particle of the top and also perpendicular to the tabular alumina particle of
the bottom left corner and is slightly deviated by a small tabular alumina particle during
its propagation.
Figures 4.177 – 4.180 show the evolution of the microstructure overview of Y-FSZ
samples after one, three, five and ten thermal shocks. The aim of those SEM pictures is to
reveal the increasing crack density in a lesser extent with increasing quenching tests in
molten aluminum at 800°C as well as the different nature of microstructural damage
within the Y-FSZ sample. If an increase of the crack density is in general noticeable, it is
worth mentioning that instead of intergranular cracks that could be observed in Ref and
Ca-PSZ, the main cracks that propagate in Y-FSZ are transgranular. The crack density of
those transgranular cracks increases with increasing quenching tests.
As shown on figure 4.177, the microstructure of Y-FSZ after one quenching test in
molten aluminum is similar to the microstructure of this formulation after sintering. If the
tabular alumina and zirconia particles are well embedded in the calcium aluminate
matrix, some cracks are nevertheless visible at the grain boundaries of big tabular
alumina aggregates. An intergranular crack is noticeable on the bottom left corner of
figure 4.177. This crack is propagating throughout the matrix and in a parallel way to the
surface of the coarse tabular alumina aggregate.
After three quenching tests, no significant increase of the crack density is visible. If some
cracks are noticeable at the grain boundary of tabular alumina aggregates in the middle of
figure 4.178, a transgranular crack can be observed throughout the tabular alumina
aggregate of the top left corner.
301
Results
After five quenching tests, more transgranular cracks can be observed throughout the
tabular alumina aggregates and more specifically on the top left corner of figure 4.179.
Those cracks can be large and reach a crack length superior to 1 mm.
Finally, after ten quenching tests, a widening of the transgranular cracks is observed on
figure 4.180. On the top right corner of figure 4.180, a large transgranular crack is
propagating from one side to the other and the crack width is locally superior to 10 µm.
Besides the transgranular cracks, some cracks are visible at the interface of the big
tabular alumina aggregate of the bottom right corner.
Even if the main damage mechanisms in Y-FSZ after thermal shocks correspond to
transgranular cracks as revealed by the previous microstructure overviews (see
figures 4.177 – 4.180), the present part focusses on the examination of intergranular
cracks. Figures 4.181 – 4.184 show the path of the cracks in the matrix of Y-FSZ.
302
Results
The aim of those pictures is to examine the nucleation of the cracks and the enhancement
of the friction mechanisms occurring within the matrix after progressive quenching tests.
Fig. 4.181: Crack nucleation at the grain Fig. 4.182: Crack propagation around
boundary of a zirconia particle of zirconia grains throughout the matrix and
formulation Y-FSZ after one quenching at the grain boundary of a tabular alumina
test in molten aluminum at 800°C grain of formulation Y-FSZ after three
quenching tests in molten aluminum at
800°C
After the first quenching test, small cracks are visible in the CA6-rich matrix in the
surrounding are of an yttria stabilized zirconia particle. Those cracks with a length of
around 10 – 20 µm and a width of around 1 µm are propagating perpendicularly to the
surface of the zirconia particle. Those small cracks are continuously deviated by the CA6
facets which are entangled with each other by forming a micro-scale interlocking
303
Results
structure. The zirconia particle in the middle of figure 4.181 is partly debonded from the
matrix in its left-hand portion.
After three quenching tests, the crack density sensibly increases in the surrounding areas
of small zirconia and tabular alumina particles as shown on figure 4.182. A
non-negligible crack widening obviously occurs with increasing quenching tests. The
cracks that propagate perpendicularly to the surface of those small tabular alumina and
zirconia particles exhibit a crack width the crack flanks, which can locally reach 10 µm.
The crack expansion is also noticeable with a crack length of around 50 µm.
After five quenching tests, a dense crack network is visible in the calcium aluminate rich
part of the matrix as shown in the middle part of figure 4.183. Some big cracks are
nucleating from the grain boundary of a coarse tabular alumina aggregate and
perpendicularly to its grain boundary, and are progressively deviated by small tabular
alumina particles during their propagation. A rambling crack path is noticeable and the
cracks have a length around 200 µm and a width of around 10 µm.
After ten quenching tests, the surrounding area of coarse zirconia and tabular alumina
aggregates are damaged, while the damage is localized in the surrounding area of small
particles of zirconia or alumina after the first and the third quenching test. The damage
mechanisms are however identical with a propagation of crack perpendicularly to the
surface of the zirconia particles as shown on the top right corner of figure 4.184 or to the
surface of tabular alumina particles as shown on the left side of figure 4.184. Those big
cracks exhibit a length up to 500 µm and a width of around 10 µm. In the top middle part
of figure 4.184, a crack is nucleating perpendicularly to the surface of a splintered
zirconia grain and is deviated by another zirconia particle around 250 µm on the right
side.
304
Results
application. The present part of the study correlates the elastic properties evolution during
cycling temperature difference through HT-RFDA measurements as explained in
chapter 3.2.8.3.3 and the creep behavior of the four tested castables during cycling
temperature difference, in the identical temperature range from 900°C to 1500°C,
through cycling Refractoriness under Load measurements as explained in chapter 3.2.6.3.
Another alternative consists in performing a HT-RFDA measurement up to 1500°C on
samples after thermal shock experiment and comparing the high temperature elastic
behavior of the sample after sintering and after thermal shock test. Before analyzing the
elastic behavior of the four tested castables at elevated temperature, the non-linear
behavior of the materials stiffness characterized at high temperature with regard to the
amplitude of excitation is similarly treated as in chapter 4.2.2 at room temperature in
order to define an experimental procedure applicable for formulations Ref, Ca-PSZ,
Mg-PSZ and Y-FSZ independently of their damage extent. In the following part, Ref
samples are sintered at 1500°C for 6 hours according to temperature profile 1, while
zirconia based samples, namely Ca-PSZ, Mg-PSZ and Y-FSZ, are sintered according to
temperature profile 2.
The amplitude of the impulse excitation dependency of the elastic properties of castables
after progressive thermal shocks in air according to DIN EN 993-11 and at room
temperature is examined in chapter 4.2.2. It has been shown that the acoustic response of
the materials considerably depend on the extent of mechanical excitation delivered by the
automated tapping device. It has also been shown that this non-linear behavior also tends
to increase with the material damage. In this part of the study, HT-RFDA measurements
are performed on each castable formulation after sintering at 1500°C for 6 hours.
Formulation Ref is sintered according to sintering profile 1 illustrated on figure 3.13 and
formulations Ca-PSZ, Mg-PSZ and Y-FSZ are sintered according to sintering profile 2
illustrated on the same figure. HT-RFDA measurements are performed up to 1500°C for
3 hours according to the heat treatment detailed on figure 3.24. Those measurements are
successively performed on different representative samples of each formulations with an
amplitude of impulse excitation from 30 % to 80 % with a step of 10 %. As Ca-PSZ and
Mg-PSZ can be distinguished from formulations Ref and Y-FSZ by a significant crack
density after sintering as SEM pictures could reveal in chapter 4.3.1.6, the aim of the part
of this study is to define a correlation between the non-linear acoustic response at
elevated temperature with the crack density of the samples as well as with further damage
mechanisms occurring during the heat treatment of HT-RFDA measurement such as the
305
Results
Fig. 4.185: Evolution of the relative Young’s modulus of formulation Ref after sintering
at 1500°C for 6 hours with an amplitude of excitation of 30 % and 70 % (measurement
performed in accordance with the heat treatment detailed on figure 3.24)
306
Results
Because of microstructural healing mechanisms that will be examined later on, the
Young’s modulus slowly increases during the dwell time to reach a relative value of
around 74 % after three hours.
During the cooling phase, the relative Young’s modulus also follows a two-step evolution
with a first increase up to around 1250°C to reach a value of 86 %, i.e. the relative
Young’s modulus progressively increases of around 4,9 % each 100°C and with a second
increase up to room temperature to regain its original value with 100 %, i.e the relative
Young’s modulus progressively increases of around 1 % each 100°C in this second
temperature range of the cooling phase.
It is worth mentioning that the stiffness evolution is perfectly reversible in the second
evolution phase between 1250°C and 1500°C. However, a low hysteretic behavior is
noticeable at low temperature from room temperature to 1250°C. Ref exhibits a
spindle-shaped behavior with relative Young’s modulus values that are higher during the
cooling phase up to around 3 % at 600°C.
Independently of the amplitude of the impulse excitation, the damping of the flexural
resonant frequency follows a two-step evolution with damping values that exponentially
decrease from 1500°C to 1200°C – 1250°C by reaching around 1,0.10-2 [-] while the
initial damping values in the beginning of the cooling phase at 1500°C are quite similar
with 6,344.10-2 [-] for an amplitude of impulse excitation of 30 % and 6,653.10-2 [-] for
an amplitude of impulse excitation of 70 %. The damping decreases then slowly up to
room temperature to reach a final value after complete cooling of 0,379.10-2 [-] for an
amplitude of impulse excitation of 30 % and 0,397.10-2 [-] for an amplitude of impulse
excitation of 70 %.
The damping behavior of Ref is in good correlation with the stiffness behavior in
function of the amplitude of the impulse excitation. While a two-step linear Young’s
modulus behavior is observed for an amplitude of impulse excitation of 70 %, it is worth
mentioning that a two-step linear damping behavior is observed for the same amplitude
of impulse excitation. For lower amplitude of excitation and namely 30 %, abrupt and
sudden damping increase or decrease are observed during the cooling phase such as the
considerable increase of damping from 515°C to 371°C where the damping values
307
Results
increase from 0,106.10-2 [-] to 1,155.10-2 [-] or the strong damping decrease from
0,575.10-2 [-] to 0,117.10-2 [-] between 270°C and 261°C.
Fig. 4.186: Evolution of the damping of the flexural resonant frequency of formulation
Ref during the cooling phase after sintering at 1500°C for 6 hours with an amplitude of
excitation of 30 % and 70 % (measurement performed in accordance with the heat
treatment detailed on figure 3.24)
In order to mathematically represent the stiffness dependency of the tested materials with
the amplitude of the impulse excitation, the evolution of the relative Young’s modulus
with the relative voltage delivered to the electromagnet that ejects the hammer for
providing the mechanical excitation (from 30 % to 80 % with a step of 10 %) is examined
for each tested formulation from room temperature up to 1500°C with a measuring step
of 50°C. Figure 4.187 reveals the evolution of the relative Young’s modulus of
formulation Ref with increasing voltage at different temperature, namely RT (for Room
Temperature), 500°C, 1000°C, 1250°C, 1450°C and 1500°C during the heating phase.
The Young’s modulus measured at temperature T with a relative voltage of 30 % is
considered as reference value for the calculation of the relative Young’s modulus values
at same temperature T for relative voltage values comprised between 40 % and 80 %.
The maximum voltage value delivered to the electromagnet is 42 V.
308
Results
Fig. 4.187: Evolution of the relative Young’s modulus of formulation Ref with increasing
voltage delivered to the electromagnet at different temperature, namely RT, 500°C,
1000°C, 1250°C, 1450°C and 1500°C
It clearly appears on figure 4.187 that the Young’s modulus of Ref slightly decreases
with increasing voltage for each examined temperature except for 500°C characterized by
a low increase. This decrease is more and more considerable with increasing temperature.
A linear trend line can be added to the experimental relative Young’s modulus values and
the examination of the amplitude of impulse excitation dependency of the relative
Young’s modulus in function of the temperature can be summarized by the examination
of the slope of the relative Young’s modulus evolution with the increase of the relative
voltage delivered to the electromagnet for each formulation.
A similar non-linear coefficient as the coefficient χE defined in chapter 4.2.2.1
(equation 4.1) related to the slope of the trend line of the relative Young’s modulus
evolution with increasing voltage can be therefore defined for high temperature analysis
as follows:
∆𝐸
𝜒𝐸,𝑇 =
∆𝑉 Equation 4.4
With:
309
Results
χE,T: Non-linear coefficient related to the relative stiffness with the relative
voltage delivered to the electromagnet for impulse generation at
temperature T [% of Young’s modulus / % voltage] i.e. [-]
ΔE: Relative stiffness difference [% of Young’s modulus]
ΔV: Relative voltage delivered to the electromagnet for impulse generation [%
of voltage, i.e. voltage value divided by the maximum voltage value (42V)]
The evolution of the non-linear coefficient χE,T of formulations Ref, Ca-PSZ and Y-FSZ
during the heating phase and the cooling phase of the heating treatment is shown on
figures 4.188 and 4.189. The microstructural damage of Mg-PSZ after sintering at
1500°C for 6 hours results in a challenging data acquisition during HT-RFDA
measurements with increasing amplitude of impulse excitation. Therefore, the non-linear
behavior relative to the stiffness is exclusively examined on formulations Ref, Ca-PSZ
and Y-FSZ exhibiting a reasonable crack density after sintering.
During the heating phase, Ref and Y-FSZ reveal a similar behavior with regard to the
non-linear response relative to the material stiffness with increasing voltage for impulse
generation as shown on figure 4.188. Both formulations depend to a very small extent on
the voltage as the non-linear coefficient χE,T remains almost constant in the whole
temperature range from room temperature to 1500°C. As far as Ref is concerned, the
non-linear coefficient χE,T ranges between -0,02 [-] observed at 1500°C and 0,01 [-]
observed at 500°C, i.e. a very low amplitude of the non-linear coefficient χE,T is observed
in the whole range of temperature (0,03 [-]).
A similar behavior is observed in Y-FSZ. The non-linear coefficient χE,T is comprised
between -0,02 [-] observed at 1450°C and 0,02 [-] observed at 1100°C, i.e. a very low
amplitude of the non-linear coefficient χE,T is observed in the whole range of temperature
(0,04 [-]).
Concerning Ca-PSZ, the amplitude of impulse excitation dependency of the material
stiffness appears to be insignificant from room temperature to 1050°C during the heating
phase as the non-linear coefficient χE,T is comprised between -0,02 [-] observed at 350°C
and 0,02 [-] observed at 600°C as shown on figure 4.189. However, above 1050°C, the
non-linear coefficient χE,T drastically and almost linearly increases up to 1500°C from
-0,01 [-] to a value at the beginning of the dwell time at 1500°C with 0,15 [-]. A local
maximum of χE,T is noticeable at 1450°C with 0,16 [-]. As soon as the martensitic
transformation of zirconia occurs during the heating phase leading to a debonding of the
zirconia particles from the castable matrix, the acoustic response of Ca-PSZ in terms of
stiffness significantly depends on the voltage delivered to the electromagnet for impulse
generation and subsequently on the amplitude of the impulse excitation.
310
Results
Fig. 4.188: Evolution of the non-linear coefficient χE,T from room temperature up to
1500°C during the heating phase of formulations Ref, Ca-PSZ and Y-FSZ
During the cooling phase, Ref and Y-FSZ reveal a similar behavior as previously during
the heating phase with regard to the non-linear response relative to the material stiffness
with increasing voltage. Both formulations depend to a very small extent on the
amplitude of impulse excitation as the non-linear coefficient χE,T remains constant in the
whole temperature range from room temperature to 1500°C. As far as Ref is concerned,
the non-linear coefficient χE,T is comprised between -0,02 [-] observed at 1500°C and
0,0 [-] observed at room temperature, i.e. Ref only exhibits negative values of non-linear
coefficient χE,T and a very low amplitude of the non-linear coefficient χE,T is observed in
the whole range of temperature (0,02 [-]).
A similar behavior is observed in Y-FSZ. The non-linear coefficient χE,T is comprised
between -0,05 [-] observed at 1450°C and 0,03 [-] observed at 1500°C, i.e. a very low
amplitude of the non-linear coefficient χE,T is observed in the whole range of temperature
(0,07 [-]). Except for the two last values of non-linear coefficient χE,T at 1450°C and
1500°C, the parameter χE,T fluctuates in a same range of values as during the heating
phase in the same temperature range. The low decrease of the non-linear coefficient χE,T
between 1500°C and 1450°C may result from a sudden and slight debonding of zirconia
particles from the matrix.
Concerning Ca-PSZ, the values of the non-linear parameter χE,T remain in the same order
of magnitude during the cooling phase up to 900°C as the values observed at elevated
temperature during the heating phase. The non-linear parameter χE,T fluctuates in a
temperature range from 1500°C and 900°C between 0,18 [-] observed at 1450°C and
311
Results
0,26 [-] observed at 1000°C. From 900°C to 800°C, the non-linear coefficient χE,T
drastically decreases from 0,24 [-] up to 0,04 [-]. After this strong decrease occurring in
the temperature range of the martensitic transformation of zirconia, the non-linear
coefficient χE,T slightly decreases up to room temperature to exhibit a value of 0,0 [-] at
room temperature.
Fig. 4.189: Evolution of the non-linear coefficient χE,T from room temperature up to
1500°C during the cooling phase of formulations Ref, Ca-PSZ and Y-FSZ
It is worth mentioning that formulations Ref, Ca-PSZ and Y-FSZ exhibit a non-linear
coefficient χE,T value of 0,0 [-] at room temperature after HT-RFDA measurement. By
focusing on Ca-PSZ, it can be assumed that the examination of the non-linear coefficient
χE,T shows that microstructural damage such as debonding results in an increase of the
Young’s modulus with increasing voltage and subsequently with increasing amplitude of
excitation, while microcracking results in a decrease of the Young’s modulus with
increasing amplitude of excitation. However, no negative value of the non-linear
coefficient χE,T is observed after the martensitic transformation of zirconia during the
cooling phase showing that microcracking effects cannot lead to an expected decrease of
the material stiffness with increasing amplitude of impulse excitation if debonding takes
place during the material thermal history.
312
Results
In order to mathematically represent the damping dependency of the tested materials with
the amplitude of the impulse excitation, the evolution of the damping of the flexural
resonant frequency with the relative voltage delivered to the electromagnet that ejects the
hammer for providing the mechanical excitation (from 30 % to 80 % with a step of 10 %)
is examined for each tested formulation from room temperature up to 1500°C with a
measuring step of 50°C. Figure 4.190 reveals the evolution of the damping of the flexural
resonant frequency of formulation Ref with the relative voltage at different temperature,
namely RT, 500°C, 1000°C, 1250°C, 1450°C and 1500°C during the heating phase.
Fig. 4.190: Evolution of the damping of the flexural resonant frequency of formulation
Ref with the relative voltage for impulse generation at different temperature, namely RT,
500°C, 1000°C, 1250°C, 1450°C and 1500°C
It clearly appears on figure 4.190 that the damping of the flexural resonant frequency of
Ref is linearly dependent on the voltage and subsequently on the amplitude of excitation
for each examined temperature. A higher scatter of the measured damping values is
nevertheless noticeable at elevated temperature and more particularly at 1500°C. A linear
trend line can be added to the experimental damping values and the examination of the
amplitude of impulse excitation dependency of the damping of the flexural resonant
313
Results
A similar non-linear coefficient as the coefficient χE,T defined previously in equation 4.2
related to the slope of the trend line of the flexural damping evolution with increasing
voltage can therefore be defined for high temperature damping analysis as follows:
∆𝜁𝑓
𝜒𝜁𝑓,𝑇 =
∆𝑉 Equation 4.5
With:
The evolution of the non-linear coefficient 𝜒𝜁𝑓,𝑇 of formulations Ref, Ca-PSZ and Y-FSZ
during the heating phase and the cooling phase of the heating treatment is shown on
figures 4.191 and 4.192. Like previously with the examination of the non-linear response
in function of the amplitude of impulse excitation, the microstructural damage of
formulation Mg-PSZ after sintering at 1500°C for 6 hours results in a challenging data
acquisition during HT-RFDA measurements with increasing amplitude of impulse
excitation. Therefore, the non-linear behavior relative to the flexural damping is
exclusively examined on formulations Ref, Ca-PSZ and Y-FSZ exhibiting a reasonable
crack density after sintering.
314
Results
Fig. 4.191: Evolution of the non-linear coefficient 𝜒𝜁𝑓,𝑇 from room temperature up to
1500°C during the heating phase of formulations Ref, Ca-PSZ and Y-FSZ
During the heating phase, formulations Ref, Ca-PSZ and Y-FSZ show a different
behavior relative to the non-linear response in terms of damping with increasing voltage
and subsequently with increasing amplitude of excitation.
As for Ref, the non-linear coefficient 𝜒𝜁𝑓,𝑇 appears to be unstable at low temperature and
more particularly from room temperature to 500°C by oscillating between 0,4 [-],
observed at 150°C and -4,4 [-], observed at 350°C. After this fluctuating behavior, the
non-linear coefficient 𝜒𝜁𝑓,𝑇 slightly and linearly increases with increasing temperature to
reach -0,1 [-] at 1250°C. Finally, the non-linear coefficient 𝜒𝜁𝑓,𝑇 exponentially increases
with increasing temperature to reach 14,1 [-] at 1500°C.
Concerning Ca-PSZ, the non-linear coefficient 𝜒𝜁𝑓,𝑇 is quite constant in a wide range of
temperature, and more precisely from room temperature up to 1050°C with a value close
to zero and equal to 0,6 [-] at 1050°C. The non-linear coefficient 𝜒𝜁𝑓,𝑇 then decreases up
to -3,3 [-] at 1350°C and finally increases to reach values close to zero at elevated
temperature (-0,2 [-] at 1500°C).
As far as Y-FSZ is concerned, the non-linear coefficient 𝜒𝜁𝑓,𝑇 fluctuates around values
close to zero in a wide range of temperature, and more precisely from room temperature
up to 1300°C by analogy with formulation Ca-PSZ. The values of 𝜒𝜁𝑓,𝑇 oscillate between
-2,3 [-] observed at 1200°C and 0,7 [-] observed at 350°C. The non-linear coefficient
𝜒𝜁𝑓,𝑇 then increases up to 2,4 [-] observed at 1250°C corresponding to its maximum
315
Results
value in the whole temperature range of the heating phase and finally strongly decreases
up to -8,5 [-] observed at 1500°C.
Fig. 4.192: Evolution of the non-linear coefficient 𝜒𝜁𝑓,𝑇 from room temperature up to
1500°C during the cooling phase of formulations Ref, Ca-PSZ and Y-FSZ
Once again, during the cooling phase, formulations Ref, Ca-PSZ and Y-FSZ show a
different behavior relative to the non-linear response in terms of damping with increasing
voltage and subsequently with increasing amplitude of impulse excitation.
As for Ref, the non-linear coefficient 𝜒𝜁𝑓,𝑇 exponentially decreases from 1500°C up to
1250°C with 26,7 [-] and 1,9 [-] respectively. During the second phase of the cooling,
namely from 1250°C to room temperature, the non-linear coefficient 𝜒𝜁𝑓 ,𝑇 values are very
low. Those values are located in a narrow range between -1,1 [-] observed at 350°C and
1,9 [-] observed at 1250°C. The evolution of the non-linear coefficient 𝜒𝜁𝑓,𝑇 with
temperature during the cooling phase is similar to the evolution of the damping of the
flexural resonant frequency during the same cooling phase.
Concerning Ca-PSZ, the evolution of the non-linear coefficient 𝜒𝜁𝑓,𝑇 also begins with a
strong decrease from 1500°C up to 1250°C but in a lesser extent, while the non-linear
coefficient 𝜒𝜁𝑓,𝑇 evolves from 4,0 [-] at 1500°C to 0,6 [-] at 1250°C with a local
maximum at 1400°C with 4,7 [-]. After this decrease, and during the second phase of the
cooling, namely from 1250°C to room temperature, the non-linear coefficient 𝜒𝜁𝑓,𝑇 values
are very low. Those values are comprised in a narrow range from -0,8 [-] observed at
100°C and 1,1 [-] observed at 600°C.
316
Results
317
Results
Figure 4.193 shows the evolution of the Young’s modulus of Ref and Y-FSZ, as those
formulations exhibit a similar elastic behavior at elevated temperature. The evolution of
the Young’s modulus of Ca-PSZ and Mg-PSZ is shown on figure 4.194 for similar
reasons. All those measurements are performed in accordance with the heat treatment
detailed on figure 3.25 with an amplitude of impulse excitation of 30 %.
Ref
Y-FSZ
Fig. 4.193: Evolution of the Young’s modulus of formulations Ref and Y-FSZ after
sintering with increasing temperature up to 1500°C (measurement performed in
accordance with the heat treatment detailed on figure 3.25 with an amplitude of impulse
excitation of 30 %)
Concerning Ref, during the heating phase, a first linear decrease of the Young’s modulus
is observed from room temperature, starting from 149 GPa up to around 1250°C to reach
a Young’s modulus of around 128 GPa, i.e. the Young’s modulus progressively decreases
of around 2 GPa each 100°C. Another decrease of the stiffness is observed from 1250°C
up to 1500°C to reach a value of 109 GPa at the beginning of the 3h-dwell time, i.e the
Young’s modulus progressively decreases of 8 GPa each 100°C in this second
temperature range.
318
Results
Because of microstructural healing mechanisms that will be examined later on, the
Young’s modulus slowly increases during the dwell time to reach a value of 110 GPa
after three hours.
During the cooling phase, the Young’s modulus also follows a two-step evolution with a
first increase up to around 1250°C to reach a value of 128 GPa, i.e. the Young’s modulus
progressively increases of around 7 GPa each 100°C and with a second increase up to
room temperature to regain its original value with 149 GPa, i.e the Young’s modulus
progressively increases of around 2 GPa each 100°C.
It is worth mentioning that the stiffness evolution is perfectly reversible in the second
evolution phase between 1250°C and 1500°C. However, a low hysteretic behavior is
noticeable at low temperature from room temperature to 1250°C. Ref exhibits a
spindle-shaped behavior with Young’s modulus values that are higher during the cooling
phase than during the heating phase up to 5 GPa at 626°C.
319
Results
Contrary to formulations Ref and Y-FSZ, Ca-PSZ and Mg-PSZ reveal a much more
noticeable hysteretic behavior with regard to the Young’s modulus at elevated
temperature. If a decrease of stiffness is observed in Ref and Y-FSZ with increasing
temperature, the Young’s modulus of Ca-PSZ and Mg-PSZ increases with increasing
temperature and continue increasing during the cooling phase up to the martensitic
transformation temperature before drastically decreasing up to room temperature to reach
a room temperature stiffness slightly inferior to the stiffness of the material at the
beginning of the heating phase.
As for Ca-PSZ, the Young’s modulus starts from 68 GPa at room temperature and first
decreases up to reach a minimum value of 59 GPa at 735°C. The Young’s modulus
increases up to the martensitic transformation of zirconia occurring at 1051°C and
reaches a value of 66 GPa. A low stiffness decrease is observed after the zirconia
transformation and the Young’s modulus exhibit a local minimum at 1059°C with
65 GPa. The Young’s modulus logarithmically increases up to 1500°C to reach a value of
81 GPa at the beginning of the 3h-dwell time. A significant microstructure healing occurs
during this 3h-dwell time as the castable exhibits a stiffness of 84 GPa at the beginning of
the cooling phase at 1500°C. During the cooling phase, the Young’s modulus of Ca-PSZ
continuously increases up to the martensitic transformation temperature of zirconia
occurring at 905°C by reaching a maximum value of 116 GPa. The zirconia
transformation inducing crack nucleation in a low temperature range results in a drastic
and exponential decrease of the material stiffness up to room temperature. After this
second heat treatment following the sintering program, a further microstructure damage
is generated as the material reveals a lower Young’s modulus at the end of the cooling
phase at room temperature with 63 GPa vs 68 GPa at the beginning of the heating phase
at room temperature.
In case of Mg-PSZ, a similar elastic behavior is observed even if the evolution appears
not to be as continuous as for Ca-PSZ, because of the high crack density in the
microstructure of Mg-PSZ after sintering that makes the data acquisition challenging.
The Young’s modulus starts from 33 GPa at room temperature and first decreases up to
reach a minimum value of 30 GPa at 777°C. The Young’s modulus increases up to the
martensitic transformation of zirconia characterized by an unstable evolution of the
Young’s modulus values in a temperature range from 943°C to 983°C and reaches a local
minimum value of 36 GPa at 963°C. The Young’s modulus logarithmically increases up
to 1336°C to reach a value of 51 GPa. Up to 1368°C, no trustful resonant frequency
could be detected. This explains the absence of Young’s modulus values in this
temperature range. From 1368°C up to 1500°C, the stiffness of Mg-PSZ remains quite
constant with 43 GPa and 43 GPa respectively. By analogy with Ca-PSZ, a significant
microstructure healing occurs during this 3h-dwell time as the castable exhibits a
stiffness of 47 GPa at the beginning of the cooling phase at 1500°C. During the cooling
320
Results
Ca-PSZ
Mg-PSZ
Fig. 4.194: Evolution of the Young’s modulus of formulations Ca-PSZ and Mg-PSZ after
sintering with increasing temperature up to 1500°C (measurement performed in
accordance with the heat treatment detailed on figure 3.25 with an amplitude of impulse
excitation of 30 %)
The damping behavior of the flexural resonant frequency of Ref, Ca-PSZ, Mg-PSZ and
Y-FSZ is examined at elevated temperature through HT-RFDA analysis in order to
provide further information with regard to microstructural mechanisms occurring during
the heat treatment. This part of the study focusses on the damping behavior of
321
Results
formulations Ref, Ca-PSZ and Y-FSZ during the heating and cooling phase. Indeed, the
damaged microstructure of Mg-PSZ results in considerable friction phenomena with
increasing temperature and a chaotic damping evolution. The examination of the
damping evolution during the heating phase aims to investigate the effect of debonding
on the damping behavior and to correlate the damping behavior with the creep behavior.
The examination of the damping evolution during the cooling phase aims to investigate
the effect of the volume expansion induced by the martensitic transformation of zirconia
on the damping behavior.
Figure 4.195 shows the evolution of the damping of the flexural resonant frequency of
formulations Ref, Ca-PSZ and Y-FSZ during the heating phase up to 1500°C.
Concerning Ref, the flexural damping remains almost constant in a wide temperature
range from room temperature to 1200°C. Starting from a damping value of 0,49.10-2 [-],
the flexural damping reaches a value of 0,66.10-2 [-] at 1200°C. From 1200°C to 1500°C,
the flexural damping exponentially increases to reach a final value of 5,40.10-2 [-] at the
beginning of the 3h-dwell time. In a temperature range from 1380°C to 1440°C, a low
decrease of the increase rate of damping is noticeable.
A similar behavior is observed in Y-FSZ. The flexural damping remains almost constant
in a wide temperature range from room temperature to 1000°C. Starting from a damping
value of 0,23.10-2 [-], the flexural damping reaches 0,45.10-2 [-] at 1000°C. This first step
of evolution is followed by a low increase of damping up to 1250°C with a damping
value of 1,04.10-2 [-]. The flexural damping increases finally exponentially up to 1450°C
to reach a value of 8,13.10-2 [-]. From 1450°C to 1500°C, the damping evolution is
unstable and the values evolve between 7,04.10-2 [-] observed at 1473°C and 8,74.10-2 [-]
observed at 1463°C. The final damping value of Y-FSZ at the beginning of the 3h-dwell
time is 7,50.10-2 [-].
In case of Ca-PSZ, the damping of the flexural resonant frequency remains almost
constant up to 920°C. Starting from a damping value of 0,90.10-2 [-], the flexural
damping reaches a value of 1,13.10-2 [-] at 920°C. At 920°C, the damping strongly
increases to reach a value of 2,02.10-2 [-] at 980°C and remains almost constant up to
1060°C with 1,93.10-2 [-]. From 1060°C to 1105°C the flexural damping considerably
decreases to reach a local minimum of 0,85.10-2 [-], while debonding induced by the
martensitic transformation of zirconia takes place. Finally, the flexural damping
exponentially increases to reach a value of 4,10.10-2 [-] at 1328°C. From 1328°C to
1500°C, the flexural damping evolution of Ca-PSZ is unstable and evolves between
322
Results
4,05.10-2 [-] observed at 1392°C and 5,07.10-2 [-] observed at 1384°C. The final damping
value of Ca-PSZ at the beginning of the 3h-dwell time is 4,95.10-2 [-].
Fig. 4.195: Evolution of the damping of the flexural resonant frequency of Ref, Ca-PSZ
and Y-FSZ from room temperature to 1500°C during the heating phase (measurement
performed in accordance with the heat treatment detailed on figure 3.25 with an
amplitude of impulse excitation of 30 %)
Figure 4.196 shows the evolution of the damping of the flexural resonant frequency of
formulations Ref, Ca-PSZ and Y-FSZ during the cooling phase from 1500°C to room
temperature after the 3h-dwell time.
Concerning Ref, a reversible evolution of the damping is observed as the damping
behavior of this formulation is similar during the heating and cooling phase. Starting
from a damping value of 6,02.10-2 [-] at 1500°C after the 3h-dwell time, the flexural
damping first exponentially decreases up to 1180°C by exhibiting a value of 0,73.10-2 [-].
The flexural damping then remains almost constant up to room temperature and regains
its initial value with 0,50.10-2 [-] at the end of the heat treatment. It is also worth
mentioning that an increase of the flexural damping occurs during the 3h-dwell time as
323
Results
the value of damping is 5,40.10-2 [-] at the beginning of the 3h-dwell time and reaches
6,02.10-2 [-] after 3h at 1500°C.
Fig. 4.196: Evolution of the damping of the flexural resonant frequency of Ref, Ca-PSZ
and Y-FSZ from 1500°C to room temperature during the cooling phase (measurement
performed in accordance with the heat treatment detailed on figure 3.25 with an
amplitude of impulse excitation of 30 %)
In case of Ca-PSZ, and contrary to formulations Ref and Y-FSZ, no reversible damping
evolution is observed. Indeed, starting from a value of 4,56.10-2 [-] after the 3h-dwell
324
Results
time, which is slightly lower than the damping value at the beginning of the dwell time
with 4,95.10-2 [-], the damping first decreases up to 1478°C to reach 4,23.10-2 [-]. The
damping then increases up to 1396°C to reach 6,09.10-2 [-], corresponding to the
maximum value in the whole temperature range. The damping of Ca-PSZ strongly
decreases up to 1150°C by exhibiting a damping value of 0,88.10-2 [-]. The damping
remains almost constant up to 911°C with 0,59.10-2 [-]. While the martensitic
transformation of zirconia takes place with ensuing microcracking, the damping suddenly
increases up to 0,98.10-2 [-] at 901°C. The damping is quite constant up to 336°C
characterized by a value of 0,91.10-2 [-] before decreasing up to 198°C and reaches a
value of 0,52.10-2 [-] and finally linearly increases up to room temperature to reach a final
value of 1,59.10-2 [-], which is considerably higher than the initial value of damping
before heating treatment (0,90.10-2 [-]).
Refractory materials are in most of the case submitted to thermal cycles between two
high temperatures during application. A first approach to determine the effect of those
thermal cycling conditions on the material properties consists in post-mortem
characterizing the material structural, elastic and mechanical properties. A second
approach, constituting the focus of this part of the study, consists in examining in situ the
effect of thermal cycling on the elastic and mechanical properties of the material.
Therefore, two techniques of measurement are used. High Temperature Resonant
Frequency Damping Analysis (HT-RFDA) aims to assess the elastic behavior and more
precisely the evolution of the Young’s modulus of the material between two high
temperatures, namely 900°C and 1500°C. In such experimental conditions, the induced
stresses within the microstructure of the material are generated by the cycling
temperature change with a low heating and cooling rate (2 K.min-1). Refractoriness under
Load (RuL) aims to estimate the thermo-mechanical behavior and more precisely the
evolution of the thermal expansion of the material between two high temperatures,
namely 900°C and 1500°C in presence of compressive load. In such experimental
conditions, the induced stresses within the microstructure of the material are generated by
the cycling temperature change with a low heating and cooling rate (2 K.min-1) and also
by the mechanical compressive load applied on the cylindrical sample, namely 0,20 MPa.
HT-RFDA measurements under cycling thermal conditions are performed between
900°C and 1500°C according to ASTM C 1548-02 and the experimental procedure is
detailed in chapter 3.2.8.3. Cycling RuL measurements are performed between 900°C
and 1500°C according to DIN EN ISO 1893 and the experimental procedure is detailed in
325
Results
chapter 3.2.6.3. It is worth mentioning that the studied temperature range comprises the
martensitic transformation temperature of zirconia in Ca-PSZ and Mg-PSZ during the
heating and the cooling phase. Therefore, the influence of the zirconia transformation on
the castable fatigue is examined. Another focus consists in correlating both cycling RuL
and HT-RFDA measurements to establish a causal relationship between the additional
mechanical load and the degradation of material properties.
HT-RFDA measurements are performed on Ref, Ca-PSZ, Mg-PSZ and Y-FSZ between
900°C and 1500°C. The evolution of the Young’s modulus of Ref and Y-FSZ during
the 15 thermal cycles between 900°C and 1500°C is shown on figure 4.197, while
figure 4.198 shows the evolution of the Young’s modulus of formulations Ca-PSZ and
Mg-PSZ during the 15 thermal cycles between 900°C and 1500°C.
All those measurements are performed in accordance with the heat treatment detailed on
figure 3.15 with an amplitude of impulse excitation of 40 %.
Concerning Ref, starting from a value of 140 GPa at room temperature, the Young’s
modulus continuously decreases to reach a value of 103 GPa, i.e. 74 % of the initial
value, at 1500°C. The Young’s modulus increases during the cooling phase to reach a
value of 129 GPa at the end of the first thermal cycle at 900°C. A sawtooth evolution of
the Young’s modulus is observed during the further thermal cycles with almost linear
decrease of the Young’s modulus during the heating phase from 900°C to 1500°C and a
linear increase of the Young’s modulus during the cooling phase from 1500°C to 900°C.
The Young’s modulus values exhibited by Ref at 1500°C are almost constant and evolve
between a minimum value of 102 GPa after the first cycle and a maximum value of
106 GPa corresponding to 76 % of the initial value after the ninth thermal cycle. After
15 thermal cycles, the Young’s modulus of Ref at 1500°C is exactly 104 GPa, i.e. 75 %
of the initial value. The Young’s modulus values of Ref at the end of the cooling phase at
900°C tends to increase with the number of thermal cycles up to reach a value of
133 GPa after eight thermal cycles, which is 4 GPa higher than the Young’s modulus
exhibited after the first thermal cycle at 900°C. Then the Young’s modulus slightly
decreases to reach a value of 132 GPa after the last thermal cycle. A low increase of the
Young’s modulus is observed after the whole series of thermal cycles as Ref exhibits a
326
Results
value of 142 GPa at room temperature, which is 2 GPa higher than the initial value before
fatigue test.
900°C
900°C
1500°C
1500°C
Fig. 4.197: Evolution of the Young’s modulus of formulations Ref and Y-FSZ during the
15 thermal cycles between 900°C and 1500°C through HT-RFDA measurements
(measurement performed in accordance with the heat treatment detailed on figure 3.15
with an amplitude of impulse excitation of 40 %)
A similar elastic behavior is revealed by Y-FSZ, with the major exception that the
Young’s modulus values exhibited after the cooling cycles at 900°C are in the same order
of magnitude as the Young’s modulus at room temperature, while a difference of around
10 GPa is observed in case of Ref. Indeed, starting from a value of 126 GPa at room
temperature, the Young’s modulus continuously decreases to reach the exact value of
90 GPa, i.e. 72 % of the initial value, at 1500°C. The Young’s modulus increases during
the cooling phase to reach a value of 125 GPa at the end of the first thermal cycle at
900°C, which is only 1 GPa lower than the initial value. By analogy with Ref, a sawtooth
evolution of the Young’s modulus is observed during the further thermal cycles with
almost linear decrease of the Young’s modulus during the heating phase from 900°C to
1500°C and a linear increase of the Young’s modulus during the cooling phase from
1500°C to 900°C. The Young’s modulus values exhibited by Y-FSZ at 1500°C slightly
and progressively increase with the number of thermal cycles to reach a final value of
95 GPa after 15 thermal cycles, corresponding to 75 % of the initial value, and by
reaching a minimum value of 90 GPa after the second cycle, corresponding to 71 % of
the initial value. The Young’s modulus values of Y-FSZ at the end of the cooling phase
327
Results
at 900°C remains almost constant after each thermal cycle and evolve between 125 GPa
after four thermal cycles and 126 GPa after the last thermal cycle, which is only 1 GPa
lower than the Young’s modulus value after the heating phase of the first thermal cycle at
1500°C. After final cooling, Y-FSZ regains its initial value with 126 GPa.
T=Tm
900°C
1500°C
T=Tm
900°C
1500°C
Fig. 4.198: Evolution of the Young’s modulus of formulations Ca-PSZ and Mg-PSZ
during the 15 thermal cycles between 900°C and 1500°C through HT-RFDA
measurements (measurement performed in accordance with the heat treatment detailed on
figure 3.15 with an amplitude of impulse excitation of 40 %)
328
Results
a further part (see figure 4.217). After the first cooling cycle, Ca-PSZ reveals a stiffness
of 111 GPa at 900°C. The values exhibited after each heating cycle at 1500°C remain
almost constant during the whole series of thermal cycles. At 1500°C, the material
stiffness of Ca-PSZ evolves between a minimum value of 78 GPa after the first heating
cycle up to a maximum value of 82 GPa after the fifteenth and last heating cycle. A low
increase of 4 GPa is then observed between the first and the last heating cycle at 1500°C.
At the zirconia transformation temperature during the cooling phase of the thermal cycle,
the stiffness of Ca-PSZ evolves between a minimum value of 111 GPa after the first
cooling cycle up to 113 GPa after the eighth cooling cycle. After the last cooling cycle,
the Young’s modulus of Ca-PSZ is 112 GPa. A detailed examination of each cooling
cycle is performed in a further part (see figure 4.217). This examination shows that the
material stiffness of Ca-PSZ estimated at 900°C continuously decreases after each
cooling cycle and evolves from a maximum value of 111 GPa after the first cooling cycle
to a minimum value observed after the fifteenth cooling cycle, which is equal to 91 GPa.
A considerable stiffness decrease of 20 GPa is then observed between the first and the
last cooling cycle at 900°C. Ca-PSZ exhibits a final Young’s modulus value of 55 GPa at
room temperature after the whole series of thermal cycles. A significant stiffness
decrease of 14 GPa is observed between the room temperature values before and after the
test. During each heating cycle from 900°C to 1500°C, a decrease of the Young’s
modulus is noticeable up to the martensitic transformation of zirconia. This martensitic
transformation causes a stiffness increase in a narrow temperature range explaining the
local peak values during the heating phase. The increase of the Young’s modulus after
the zirconia transformation is more and more considerable with increasing thermal
cycles. The impact of the thermal fatigue on the ensuing effect of the material stiffness
evolution during the martensitic transformation of zirconia is treated in a further part (see
figure 4.216).
Formulation Mg-PSZ reveals a similar elastic behavior as Ca-PSZ. However, the
considerable crack density within this formulation is responsible for the challenging
evaluation and interpretation of the stiffness evolution and more particularly after the
ninth thermal cycle. If the Young’s modulus values at 1500°C after each heating phase
remain constant up to this ninth thermal cycle, with 39 GPa after the first heating phase
and 40 GPa after the ninth heating phase, the material stiffness at low temperature and
namely at the temperature of transformation of zirconia during each cooling phase tends
to decrease, with 79 GPa during the first cooling phase and 63 GPa during the ninth
cooling phase. The material stiffness before the martensitic transformation of zirconia
during the cooling phase corresponds to the maximum values of stiffness on figure 4.198.
The material stiffness decrease continues up to the fifteenth thermal cycle revealing the
continuous material damage with the number of thermal cycles. Not only does the
material stiffness at the temperature of transformation of zirconia decrease after each
329
Results
thermal cycle, but the material stiffness continuously decreases with the number of
thermal cycles and more particularly after the ninth thermal cycle. Indeed, from the ninth
thermal cycle, the material stiffness at the end of the heating at 1500°C starts to decrease,
while this property is almost constant during the first nine thermal cycles. This statement
is proven by the drastic degradation of the material stiffness by comparing the Young’s
modulus value of Mg-PSZ at room temperature before and after the test with 33 GPa and
20 GPa respectively. It is worth mentioning that the stiffness increase induced by the
martensitic transformation of zirconia during the heating phase is also observed in
Mg-PSZ.
Figure 4.199 shows the evolution of the material stiffness of Ca-PSZ and Y-FSZ during
the first heating cycle from room temperature to 1500°C and the last cooling cycle from
1500°C to room temperature. The aim of such a presentation of the results is to compare
this elastic behavior during cycling measurements and HT-RFDA measurements after
sintering illustrated on figures 4.193 and 4.194 for Y-FSZ and Ca-PSZ respectively.
The evolution of the Young’s modulus of Y-FSZ and Ca-PSZ during the first heating
phase and the last cooling phase of the thermal cycling HT-RFDA measurements is
exactly the same as the stiffness evolution of those castables after sintering illustrated on
figures 4.193 and 4.194 respectively with the exception of some discrepancies.
Y-FSZ
Ca-PSZ
Fig. 4.199: Evolution of the Young’s modulus of Ca-PSZ and Y-FSZ during the first
heating phase and the last cooling phase of the thermal cycling HT-RFDA measurements
330
Results
331
Results
Figures 4.200 – 4.203 show the microstructure of Ref after cycling HT-RFDA
measurements between 900°C and 1500°C.
Fig. 4.200: Low crack density in the Fig. 4.201: Presence of cracks at grain
microstructure overview of formulation boundaries of tabular alumina aggregates
Ref after cycling HT-RFDA measurements in formulation Ref after cycling HT-RFDA
between 900°C and 1500°C measurements between 900°C and 1500°C
(magnification 60) (magnification 150)
Fig. 4.202: Detail of cracks propagating at Fig. 4.203: Friction phenomena and crack
grain boundaries of formulation Ref after branching at the level of grain boundaries
cycling HT-RFDA measurements between of tabular alumina grains in formulation
900°C and 1500°C (magnification 500) Ref after cycling HT-RFDA measurements
between 900°C and 1500°C
(magnification 2000)
On the microstructure overview shown on figure 4.200, only few cracks are visible
within the sample matrix and more particularly at the level of the grain boundaries of the
tabular alumina aggregates. The preponderant part of the microstructure is constituted of
332
Results
Figures 4.204 – 4.207 show the microstructure of Ca-PSZ after cycling HT-RFDA
measurements between 900°C and 1500°C. On the microstructure overview shown on
figure 4.204, a non-negligible crack density is visible within the sample matrix and more
particularly at the level of the grain boundaries of the tabular alumina and zirconia
aggregates. Some transgranular cracks are also visible in the tabular alumina aggregate
on the right side of the figure 4.204. The preponderant part of the microstructure is like
previously as far as Ref was concerned constituted of tabular alumina and Ca-ZrO2
aggregates perfectly embedded in the calcium aluminate matrix.
A stronger magnification shown on figures 4.205 and 4.206 reveals that those cracks are
propagating throughout the matrix and at the grain boundaries of tabular alumina and
zirconia middle-sized aggregates. A detailed examination of the crack configuration is
illustrated on figure 4.207. By propagating throughout the matrix, the crack is either
deviated by the zirconia particle or this crack can in some cases propagate throughout the
zirconia particle in function of the nature of the crystalline phase.
Such pictures of Ca-PSZ show that microstructural damage take place in this formulation
during cycling HT-RFDA measurements as the crack density is slightly higher after the
fatigue test in comparison with the sample microstructure after sintering, even if the
principal underlying observation consists in showing a dense microstructure constituted
of tabular alumina and zirconia grains of different fractions, which are well embedded
and homogeneously distributed in the calcium aluminate matrix.
333
Results
Fig. 4.204: Low crack density in the Fig. 4.205: Crack parallel to the grain
microstructure of formulation Ca-PSZ after boundary of a tabular alumina aggregate in
cycling HT-RFDA measurements between formulation Ca-PSZ after cycling
900°C and 1500°C (magnification 60) HT-RFDA measurements between 900°C
and 1500°C (magnification 150)
Fig. 4.206: Crack path along and through Fig. 4.207: Crack path along and through
Ca-ZrO2 particles in formulation Ca-PSZ Ca-ZrO2 particles in formulation Ca-PSZ
after cycling HT-RFDA measurements after cycling HT-RFDA measurements
between 900°C and 1500°C between 900°C and 1500°C
(magnification 500) (magnification 2000)
Figures 4.208 – 4.211 show the microstructure of Mg-PSZ after cycling HT-RFDA
measurement between 900°C and 1500°C. On the microstructure overview shown on
figure 4.208, a considerable crack density is visible within the sample matrix and more
particularly at the level of the grain boundaries of the tabular alumina and zirconia
aggregates. This crack density is much higher than that present in Ca-PSZ and also higher
than the density of the pre-existent cracks after sintering. Some transgranular cracks are
334
Results
also visible within the zirconia aggregates on the bottom of figure 4.208. The extreme
high crack density is in agreement with the high monoclinic content in the Mg-ZrO2 raw
material and the low stiffness of the Mg-PSZ samples after the cycling heat treatment.
Fig. 4.208: High crack density in the Fig. 4.209: High crack density in the
microstructure overview of formulation microstructure of formulation Mg-PSZ
Mg-PSZ after cycling HT-RFDA after cycling HT-RFDA measurements
measurements between 900°C and 1500°C between 900°C and 1500°C
(magnification 60) (magnification 150)
A stronger magnification shown on figures 4.209 and 4.210 reveals that those cracks are
propagating along the grain boundaries of zirconia middle-sized aggregates causing a
light debonding of those particles from the matrix and also to a large extent throughout
the matrix and the grain boundaries of coarse tabular alumina aggregates. Those last
mentioned cracks are in most of the case characterized by a more considerable crack
width than the cracks localized at the interfaces between the zirconia particles and the
335
Results
Figures 4.212 – 4.215 show the microstructure of Y-FSZ after cycling HT-RFDA
measurement between 900°C and 1500°C. On the microstructure overview shown on
figure 4.212, cracks are scarcely visible within the sample. A stronger magnification is
required to reveal an extreme low crack density within the matrix and more particularly
at the level of the grain boundaries of the tabular alumina and zirconia aggregates. The
preponderant part of the microstructure is by analogy with Ca-PSZ constituted of tabular
alumina and Y-ZrO2 aggregates perfectly embedded in the calcium aluminate matrix.
A stronger magnification shown on figures 4.213 and 4.214 reveals that those cracks are
propagating throughout the matrix and at the grain boundaries of tabular alumina and
zirconia middle-sized aggregates. Contrary to Ca-PSZ, by propagating throughout the
matrix, the crack is deviated by the zirconia particle and does not propagate throughout
the zirconia particle revealing the stability of the crystalline phase. The stronger
magnification shown on figure 4.215 reveals a local damage of the Y-ZrO2 particle that
might result from the polishing operation necessary for the SEM analysis and does not
obviously demonstrate a resulting microstructural damage from the cycling heat
treatment.
Such pictures of Y-FSZ show that microstructural damage take place in this formulation
even if the principal underlying observation consists in showing a dense microstructure
constituted of tabular alumina and zirconia grains of different fractions, which are well
embedded and homogeneously distributed in the calcium aluminate matrix. Figure 4.215
also reveals the local high concentration of CA6 facets surrounding the small and
middle-sized tabular alumina and zirconia aggregates resulting from an optimum
dispersion of the particles during the manufacturing.
336
Results
Fig. 4.212: Low crack density in the Fig. 4.213: Low crack density in the
microstructure overview of Y-FSZ after microstructure of formulation Y-FSZ after
cycling HT-RFDA measurements between cycling HT-RFDA measurements between
900°C and 1500°C (magnification 60) 900°C and 1500°C (magnification 150)
Fig. 4.214: Sequence of crack deviation by Fig. 4.215: Some transgranular cracks
Y-ZrO2 particles in the matrix of Y-FSZ through a Y-ZrO2 particle in formulation
after cycling HT-RFDA measurements Y-FSZ after cycling HT-RFDA
between 900°C and 1500°C measurements between 900°C and 1500°C
(magnification 500) (magnification 2000)
As shown on figure 4.198, the Young’s modulus of Ca-PSZ is strongly affected by the
martensitic transformation of zirconia during the heating and cooling phases of cycling
HT-RFDA measurements. A more detailed examination of the ensuing effect of zirconia
transformation is presented in this study. The focus consists in investigating the evolution
of the stiffness increase or decrease due to the zirconia transformation with increasing
337
Results
thermal cycles. Figures 4.216 and 4.217 illustrate the evolution of the Young’s modulus
of Ca-PSZ between 900°C and 1500°C during the heating phase and the cooling phase
respectively.
As shown on figure 4.216, for each heating cycle with the exception of the first cycle, the
Young’s modulus slightly decreases up to temperature comprised between 1000°C and
1040°C before remaining quite constant in a narrow temperature range and then
considerably and logarithmically increasing up to 1200°C. The Young’s modulus finally
decreases linearly from 1200°C to 1500°C. In this part of the study, it is assumed that the
temperature of the martensitic transformation of zirconia during the heating phase
corresponds to the temperature, where the Young’s modulus of Ca-PSZ starts to increase.
Furthermore, the increase of the Young’s modulus is more and more noticeable with
increasing number of thermal cycles.
As shown on figure 4.217, for each cooling cycle with the exception of the first cycle, the
Young’s modulus first increases in a temperature range from 1500°C to 1310°C up to
1280°C. The second step of the evolution corresponds to a further but lower increase of
the Young’s modulus up to 900°C – 940°C according to the thermal cycle. Between
900°C and 940°C, the Young’s modulus of Ca-PSZ considerably decreases up to 900°C.
In this part of the study, it is assumed that the temperature of the martensitic
transformation of zirconia during the cooling phase corresponds to the temperature,
where the Young’s modulus of Ca-PSZ starts to decrease.
Fig. 4.216: Evolution of the Young’s modulus of Ca-PSZ between 900°C and 1500°C
during the successive heating phases of cycling HT-RFDA measurements
338
Results
Fig. 4.217: Evolution of the Young’s modulus of Ca-PSZ between 900°C and 1500°C
during the successive cooling phases of cycling HT-RFDA measurements
Fig. 4.218: Evolution of the martensitic transformation temperature in Ca-PSZ with the
number of cooling cycles of cycling HT-RFDA measurements
339
Results
Both Young’s modulus and damping examinations show that the martensitic
transformation temperature linearly increases with the number of cooling cycles. This
temperature increase is around 2°C per cycle and more precisely 2,14°C per cycle
according to Young’s modulus measurements and 1,98°C per cycle according to damping
measurements. According to Young’s modulus measurements, the martensitic
transformation of zirconia takes place at 905°C during the first cooling cycle while it
takes place at 936°C during the fifteenth cooling cycle. According to damping
measurements, the martensitic transformation of zirconia takes place at 903°C during the
first cooling cycle while it takes place at 932°C during the fifteenth cooling cycle. It is
also worth mentioning that the damping measurements provide in general a temperature
of transformation slightly lower than the temperature assessed through Young’s modulus
determination. Both Young’s modulus and damping evolutions show that the temperature
of the martensitic transformation of zirconia increases with the number of thermal cycles
and this temperature tends to the martensitic temperature of the raw Ca-ZrO2 material
with further increasing number of thermal cycles.
Cycling RuL measurements between 900°C and 1500°C are performed on the
formulations Ref, Ca-PSZ, Mg-PSZ and Y-FSZ according to the measurement principles
detailed in chapter 3.2.6.3. Figure 4.219 shows the evolution of the thermal expansion of
Ref, Ca-PSZ, Mg-PSZ and Y-FSZ during the 15 thermal cycles between 900°C and
340
Results
1500°C. For each formulation, the thermal expansion linearly increases during the
successive heating phases between 900°C and 1500°C and linearly decreases during the
successive cooling phases between 1500°C and 900°C.
Fig. 4.219: Evolution of the thermal expansion of Ref, Ca-PSZ, Mg-PSZ and Y-FSZ
during the 15 thermal cycles between 900°C and 1500°C through cycling RuL
measurements
By analogy with CiC measurements detailed in chapter 4.3.1.4, two groups of castables
can be distinguished.
On the one side, Ref and Y-FSZ exhibit a high expansion at 1500°C. Starting from an
initial expansion of 1,26 % and 1,21 % respectively, the maximum expansion at 1500°C
after each cycle exponentially decreases to reach a value of 0,61 % and 0,35 %
respectively after the fifteenth thermal cycle at 1500°C. Those final values of expansion
correspond to 49 % and 29 % of the initial value respectively.
On the other side, Ca-PSZ and Mg-FSZ exhibit a lower expansion than Ref and Y-FSZ at
1500°C. Starting from an initial expansion of 0,94 % and 0,95 % respectively, the
maximum expansion at 1500°C after each cycle exponentially decreases to reach a value
of 0,28 % and 0,23 % respectively after the tenth thermal cycle as for Ca-PSZ and after
the fifteenth thermal cycle as for Mg-PSZ. Those final values of expansion correspond to
30 % and 24 % of the initial value respectively.
In order to accurately examine the creep rate of the four studied formulations, the
maximum expansion at 1500°C is plotted in function of the time as shown on
341
Results
figure 4.220. As the creep velocity follows an exponential decay function, an exponential
trend line is added. The exponential coefficient corresponds to the creep rate of the
maxima at 1500°C in function of the thermal cycles.
Fig. 4.220: Evolution of the creep rate of Ref, Ca-PSZ, Mg-PSZ and Y-FSZ by taking
into account the maxima at 1500°C in function of the thermal cycles through cycling RuL
measurements
As shown on figure 4.220, the creep rate of Ref differs from the creep rate of Ca-PSZ,
Mg-PSZ and Y-FSZ, since formulation Ref exhibits a very low exponential coefficient
with -0,005 [-]. The zirconia based formulations reveal a similar creep rate with an
exponential coefficient of -0,009 [-] for Y-FSZ followed by -0,010 [-] for Mg-PSZ and
-0,011 [-] for Ca-PSZ.
Even if a linear expansion of Ca-PSZ is observed during each heating phase as well as a
linear contraction during each cooling phase as shown on figure 4.219, a detailed
observation of the expansion behavior in the temperature range from 900°C and 1500°C
shows that this thermal expansion behavior is affected by the martensitic transformation
of zirconia and more precisely by the ensuing volume change of this transformation. A
more detailed examination of the ensuing effect of zirconia transformation is presented in
342
Results
this study. The focus consists in investigating the evolution of the thermal expansion or
contraction due to the zirconia transformation with increasing thermal cycles.
Figures 4.221 and 4.222 illustrate the evolution of the thermal expansion of Ca-PSZ
between 900°C and 1500°C during the successive heating phases and the successive
cooling phases respectively.
Fig. 4.221: Evolution the thermal expansion of Ca-PSZ during the successive heating
phases from 900°C to 1500°C through cycling RuL measurements
As shown on figure 4.221, during each heating phase of the cycling RuL measurement, a
linear expansion of the sample occurs from 900°C to 1050°C – 1080°C. Between 1050°C
and 1080°C, a low contraction of the sample occurs due to the contraction of the zirconia
particles of Ca-PSZ that are able to transform from monoclinic to tetragonal. After this
transformation, the material expands linearly up to 1500°C with a similar expansion rate
as before the martensitic transformation. It is assumed that the temperature where the
sample starts to contract during the heating phase corresponds to the temperature of
transformation of zirconia.
As shown on figure 4.222, during each cooling phase of the cycling RuL measurement, a
linear contraction of the sample occurs from 1500°C to 900°C – 940°C according to the
number of thermal cycles. Between 900°C and 940°C, a low expansion of the sample
occurs due to the expansion of the zirconia particles of Ca-PSZ that are able to transform
from tetragonal to monoclinic. After this transformation, a lower expansion of the sample
343
Results
is observed up to 900°C. It is assumed that the temperature where the sample starts to
expand during the cooling phase corresponds to the temperature of transformation of
zirconia.
Fig. 4.222: Evolution the thermal expansion of Ca-PSZ during the successive cooling
phases from 1500°C to 900°C through cycling RuL measurements
Fig. 4.223: Evolution of the martensitic transformation temperature in Ca-PSZ with the
number of heating and cooling cycles through cycling RuL measurements
344
Results
In the following part of the study, the evolution of the martensitic transformation
temperature during heating and cooling phases is examined. Figure 4.223 shows the
evolution of the martensitic transformation temperature in Ca-PSZ with the number of
heating and cooling cycles through cycling RuL measurements.
During both heating and cooling phases, the martensitic transformation temperature
linearly increases with the number of thermal cycles. The increase rate is exactly the
same for each thermal phase with an increase of 2,93°C per cycle. Therefore, a shift of
2,93°C of the hysteresis to higher temperature with regard to thermal expansion is
observed with increasing thermal cycles. In case of heating phase, the martensitic
transformation of zirconia occurs at 1050°C during the second heating phase and takes
place at 1077°C during the twelfth thermal cycle. In case of cooling phase, the
martensitic transformation of zirconia occurs at 909°C during the first cooling phase and
takes place at 939°C during the eleventh thermal cycle.
In order to better show the influence of the compressive stress on the temperature shift of
the martensitic transformation of zirconia, figures 4.224 and 4.225 reveal the evolution of
the martensitic transformation temperature in Ca-PSZ with the number of heating and
cooling cycles through both cycling HT-RFDA and RuL measurements.
As shown on figure 4.224, if a linear increase of 2,93°C per cycle of the temperature of
zirconia transformation is observed during the successive heating phases through cycling
RuL measurements, no obvious trend of evolution can be noticed through cycling high
temperature RFDA measurements. Indeed, the temperature assessed through Young’s
modulus measurements first strongly decreases from 1047°C during the third thermal
cycle up to 1012°C during the fifth thermal cycle. The temperature than continuously
increases to reach 1027°C during the twelfth heating cycle with a local maximum at
1031°C during the tenth thermal cycle. The temperature finally decreases to reach a final
value of 1004°C during the fifteenth and last heating cycle.
345
Results
compressive stress tends to increase the rate of the temperature shift up to 2,93°C per
cycle, as the cycling RuL measurements reveal.
Fig. 4.224: Evolution of the martensitic transformation temperature of Ca-PSZ with the
number of heating cycles through cycling HT-RFDA and RuL measurements
Fig. 4.225: Evolution of the martensitic transformation temperature of Ca-PSZ with the
number of cooling cycles through cycling HT-RFDA and RuL measurements
346
Results
4.7.4 Thermal shock resistance of Ref, Ca-PSZ, Mg-PSZ and Y-FSZ after
high temperature thermal shocks between 900°C and 1500°C
High Temperature Thermal Shocks (HTTS) are performed on formulations Ref, Ca-PSZ,
Mg-PSZ and Y-FSZ according to the experimental procedure described in
chapter 3.2.7.4. After having examined the elastic behavior as well as the creep behavior
in the temperature range from 900°C to 1500°C, the thermal shock resistance of the
tested formulations is examined in this temperature range. Cycling HT-RFDA have
revealed a stability with regard to the stiffness of Ref, Ca-PSZ and Y-FSZ with the
number of thermal cycles while a strong decrease of the Young’s modulus is observed for
Mg-PSZ. In case of Ca-PSZ, the ensuing effect of the martensitic transformation are
more and more considerable with increasing cycles. Those cycling HT-RFDA
measurements have been performed with a low heating and cooling rate (2 K.min-1).
Within the framework of HTTS experiments, samples are submitted to three progressive
thermal cycles from 900°C to 1500°C and vice versa with a significant heating and
cooling rate. In the following part of the study, the post-mortem elastic behavior of the
four studied castables after three thermal shock cycles at elevated temperature between
900°C and 1500°C is examined by performing a HT-RFDA measurement up to 1500°C
for 3 hours according to the measurement principle detailed in chapter 3.2.8.3.2. Such a
post-mortem HT-RFDA measurement aims to provide further information with regard to
the stiffness decrease in the whole temperature range of measurement, as well as some
microstructural healing effects occurring typically during the 3h-dwell time of the heat
treatment at 1500°C, and the elastic response of the castable during the martensitic
transformation of zirconia when the sample is previously damaged.
As Ref and Y-FSZ on the one side and Ca-PSZ and Mg-PSZ on the other side reveal a
similar elastic behavior during cycling HT-RFDA measurements between 900°C and
1500°C, the elastic behavior of Ref and Ca-PSZ are examined in the following part of the
studied by assuming a similar elastic behavior after HTTS experiments.
Figure 4.226 shows the stiffness evolution of Ref with increasing temperature up to
1500°C after sintering and after HTTS experiment.
If the stiffness of Ref is almost reversible after the sintering with a typical two-step
evolution with temperature during the heating and cooling phases, another trend of
evolution is noticeable after HTTS experiment.
347
Results
Fig. 4.226: Comparison of the stiffness evolution of formulation Ref after the sintering at
1500°C for 6 hours and after three thermal shock cycles between 900°C and 1500°C
After this HTTS experiment, Ref exhibits a retained stiffness of 109 GPa at room
temperature, which represents 73 % of the stiffness value before thermal shock. The
Young’s modulus first linearly decreases up to 1250°C and reaches a value of 89 GPa. In
this first stage of stiffness evolution, the Young’s modulus decreases by 2 GPa each
100°C. A drastic decrease of the Young’s modulus of 12 GPa is then observed between
1250°C and 1350°C to reach a value of 77 GPa. The last stage of stiffness evolution
during the heating phase is a low decrease from 1350°C to 1500°C with a decrease rate of
2 GPa each 100°C. Ref exhibits a stiffness of 74 GPa at the beginning of the 3h-dwell
time at 1500°C. A significant stiffness increase is observed during the 3h-dwell time and
Ref reveals a stiffness of 82 GPa at the beginning of the cooling phase at 1500°C. An
increase of 7 GPa is observed corresponding to a mean microstructural healing of
2,5 GPa per hour. A strong increase of the Young’s modulus is observed during the
cooling from 1500°C to 1360°C up to reach a value of 107 GPa. A considerable increase
rate of Young’s modulus of 18 GPa each 100°C is observed in the first evolution stage of
the cooling phase. The Young’s modulus of Ref slowly increases from 1360°C to 820°C
with a stiffness of 121 GPa. The increase rate of the Young’s modulus is 2,5 GPa each
100°C. The Young’s modulus finally remains quite constant up to room temperature with
a final value after cooling of 119 GPa. An increase of 10 GPa is noticeable after the heat
treatment showing that the microstructural healing takes place to a large extent during the
3h-dwell time at 1500°C corresponding to 76 % of the stiffness increase. However, the
microstructural healing continues during the cooling phase and mainly in a high
temperature range and contributes to 24 % of the Young’s modulus increase.
348
Results
Contrary to Ref, a similar trend of stiffness evolution is observed in Ca-PSZ after the
sintering and after the HTTS experiment with the exception of a more important increase
of the Young’s modulus during the 3h-dwell time at 1500°C and a more considerable
decrease of the Young’s modulus during the cooling phase at the temperature of
martensitic transformation of zirconia as shown on figure 4.227.
After this HTTS experiment, Ca-PSZ exhibits a retained stiffness of 52 GPa at room
temperature, which represents 72 % of the stiffness value before thermal shock. The
Young’s modulus of Ca-PSZ first slightly decreases up to 1054°C to reach a value of
44 GPa. In this first stage of stiffness evolution, the Young’s modulus decreases by 1 GPa
each 100°C.The martensitic transformation of zirconia results in a logarithmic increase of
the Young’s modulus up to 1500°C to reach a value of 68 GPa. A significant stiffness
increase is observed during the 3h-dwell time at 1500°C, such that Ca-PSZ reveals a
stiffness of 79 GPa at the beginning of the cooling phase at 1500°C. An increase of
11 GPa is observed corresponding to a mean microstructural healing of 3,5 GPa per hour,
while an increase of 3 GPa is observed after the sintering during the 3h-dwell time at
1500°C corresponding to a mean microstructural healing of 1 GPa per hour. A strong
exponential increase of the Young’s modulus is noticeable in a first stage of stiffness
evolution during the cooling phase from 1500°C to 1290°C to reach a value of 101 GPa.
The stiffness of Ca-PSZ then slightly increases up to the martensitic transformation
temperature, i.e. 924°C, by reaching a value of 105 GPa. A low increase rate of the
Young’s modulus is thus observed between 1290°C and 924°C with 1 GPa each 100°C.
The microcracking of the microstructure during the martensitic transformation of zirconia
induces a considerable decrease of the Young’s modulus up to 800°C to reach a value of
63 GPa. The Young’s modulus finally linearly decreases up to room temperature to reach
a final value of 48 GPa at room temperature. A decrease of 4 GPa is noticed after the heat
treatment. This decrease corresponds to 8 % of the initial stiffness value before heat
treatment. This stiffness decrease is however lesser important than after the sintering
equal to 6 GPa and corresponding to 13 % of the initial value before heat treatment.
349
Results
Fig. 4.227: Comparison of the stiffness evolution of formulation Ca-PSZ after the
sintering at 1500°C for 6 hours and after three thermal shock cycles between 900°C and
1500°C
In this part of the study, the damping behavior of formulation Ref and Ca-PSZ is
examined in order to verify the microstructural mechanisms responsible for the stiffness
evolution and to investigate the impact of those microstructural mechanisms on the crack
flank friction phenomena. Figures 4.228 and 4.229 show the evolution of the damping of
the flexural resonant frequency of formulations Ref and Ca-PSZ respectively with
temperature. The Young’s modulus evolution of those formulations is also plotted on
those diagrams in order to correlate both stiffness and damping behaviours.
Concerning Ref, only two important damping changes occur during the heating and the
cooling phases as shown on figure 4.228. A first considerable damping increase takes
place between 1188°C and 1304°C with values from 3,49.10-2 [-] to 11,20.10-2 [-]
respectively. This damping increase can be correlated with the Young’s modulus
decrease taking place between 1250°C and 1350°C. The damping then decreases at more
elevated temperatures to reach a minimum value of 5,75.10-2 [-] at 1412°C while a
traditional exponential damping increase is expected to take place at high temperature.
350
Results
Fig. 4.228: Evolution of the stiffness and the damping of the flexural resonant frequency
of formulation Ref after three thermal shock cycles between 900°C and 1500°C through
HT-RFDA measurement
During the cooling phase, the damping first increases from 1500°C to 1442°C with
values of 7,29.10-2 [-] at the beginning of the cooling phase at 1500°C and 15,71.10-2 [-]
at 1442°C. This damping increase can be correlated with the Young’s modulus increase
taking place between 1500°C and 1360°C. The flexural damping finally exponentially
decreases up to around 1200°C to reach a value of 1,77.10-2 [-].
As far as Ca-PSZ is concerned, some drastic damping changes are noticeable at
characteristic temperature during both heating and cooling phases as shown on
figure 4.229. Indeed, during the heating phase, a considerable increase of the damping is
observed from 850°C to 958°C, while the damping values evolve from 1,03.10 -2 [-] to
6,56.10-2 [-]. This strong damping increase can be correlated with the low stiffness
increase occurring in this temperature range (1 GPa), while the Young’s modulus
globally decreases from room temperature to 1054°C. Crack flank friction is triggered by
the considerably higher thermal expansion of zirconia particles that are able to partly
close some microcracks in the surrounding areas of the particles. Those damping values
strongly fluctuate but in general remain constant up to the martensitic transformation
temperature of zirconia. After this transformation, the debonding of the zirconia particles
results in a decrease of the friction phenomena. This explains the strong decrease of the
damping values (5,98.10-2 [-] before transformation and 3,38.10-2 [-] after
transformation). At elevated temperature, the strong increase of the damping is caused by
the increase of the crack flank friction phenomena due to the thermal expansion of
351
Results
zirconia particles that partly close some microcracks in the surrounding areas also
explaining the increase of the Young’s modulus in the temperature range
1054°C – 1500°C.
Fig. 4.229: Evolution of the stiffness and the damping of the flexural resonant frequency
of formulation Ca-PSZ after three thermal shock cycles between 900°C and 1500°C
through HT-RFDA measurement
During the cooling phase, and more precisely from 1500°C to 1290°C a strong increase
of the damping occurs from 5,66.10-2 [-] at the beginning of the cooling to 11,45.10-2 [-]
at 1290°C, while the Young’s modulus considerably increases in this temperature range.
After reaching this temperature, the damping strongly decreases to reach values below
2,0.10-2 [-] from 1200°C. No damping evolution is then observed until the martensitic
transformation of zirconia. Indeed, a considerable damping increase occurs at 904°C to
reach a value of 11,68.10-2 [-], while the Young’s modulus starts to decrease from 924°C.
The sudden volume expansion of the transformable zirconia particles explains those
important friction phenomena taking place in such a narrow temperature range.
As shown on figures 4.228 and 4.229, a similar Young’s modulus evolution is observed
for formulations Ref and Ca-PSZ during the 3h-dwell time. Indeed, a stiffness increases
352
Results
Fig. 4.230: Stiffness increase during the 3h-dwell time at 1500°C of formulations Ref,
Ca-PSZ and Y-FSZ
As shown on figure 4.230, the relative Young’s modulus of Ref, Ca-PSZ and Y-FSZ
logarithmically increases with the dwell time. However, different kinetics can be
observed.
Ref exhibits the lowest increase rate of the relative Young’s modulus during the 3h-dwell
time. Indeed, Ref reveals a Young’s modulus increase of 9 % after 3h-dwell time.
However, even if the evolution trend is logarithmic, a further stiffness increases is
expected to take place beyond the three hours of dwell time.
In case of formulations based on zirconia, a more considerable stiffness increase is
observed. Ca-PSZ exhibits a relative Young’s modulus increase of 12 %, while Y-FSZ
reveals a relative Young’s modulus increase of 13 % after the 3h-dwell time. It is worth
mentioning that the stiffness increase of Ca-PSZ and Y-FSZ during the first two hours of
dwell time is much more important than the stiffness increase of formulation Ref with a
relative stiffness increase of 11 % and 13 % for Ca-PSZ and Y-FSZ respectively vs 7 %
for Ref. Nevertheless, this stiffness increase of Ca-PSZ and Y-FSZ is less important
353
Results
during the last hour of the 3h-dwell time with 1,7 % and 0,3 % for Ca-PSZ and Y-FSZ
respectively vs 2,0 % for Ref. This shows that a longer dwell time favors the
microstructural healing of formulations that are not based on zirconia.
Figures 4.231 – 4.234 show the microstructure of Ref after three thermal shock cycles
between 900°C and 1500°C.
On the microstructure overview shown on figure 4.231, only few cracks are visible
within the sample matrix and more particularly at the level of the grain boundaries of the
tabular alumina aggregates. Those cracks visible on the top left corner of figure 4.231 are
however large with a considerable crack width. The preponderant part of the
microstructure is constituted of tabular alumina aggregates perfectly embedded in the
calcium aluminate matrix. A stronger magnification shown on figures 4.232 and 4.233
reveals that those cracks are mostly propagating at the grain boundaries of the tabular
alumina aggregates. A transgranular crack is also visible on the bottom right corner of
figure 4.232. A detailed examination of the crack configuration is illustrated on
figure 4.234. The crack is continuously deviated by small tabular alumina particles and
propagating throughout the CA6-rich matrix before continuing propagating at the grain
354
Results
boundary of a tabular alumina aggregate. The distance between the crack flanks is locally
around 10 µm, even if large cracks are visible on the microstructure overview. Such
pictures of Ref show that microstructural damage take place in a lesser extent in this
formulation as a dense microstructure constituted of tabular alumina grains of different
fractions, which are well embedded in the calcium aluminate matrix is mostly observed.
A low crack density increase is then observed after the HTTS experiment in comparison
with the microstructure after sintering explaining the low stiffness decrease.
Fig. 4.231: Microstructure overview of Fig. 4.232: Low crack density in the
formulation Ref after HTTS experiment microstructure of formulation Ref after
between 900°C and 1500°C HTTS experiment between 900°C and
(magnification 60) 1500°C (magnification 150)
Fig. 4.233: Intergranular crack propagation Fig. 4.234: Detail of intergranular crack
at grain boundary of tabular alumina propagation at grain boundary of tabular
particles in formulation Ref after HTTS alumina particles in formulation Ref after
experiment between 900°C and 1500°C HTTS experiment between 900°C and
(magnification 500) 1500°C (magnification 1000)
355
Results
Figures 4.235 – 4.238 show the microstructure of Ca-PSZ after three thermal shock
cycles between 900°C and 1500°C.
Fig. 4.235: High crack density in the Fig. 4.236: Debonding of a tabular alumina
microstructure of formulation Ca-PSZ after aggregate from the matrix in formulation
HTTS measurement between 900°C and Ca-PSZ after HTTS measurement between
1500°C (magnification 60) 900°C and 1500°C (magnification 150)
Fig. 4.237: Crack propagation throughout Fig. 4.238: Crack deviation by a Ca-ZrO2
Ca-ZrO2 particles in formulation Ca-PSZ particle exhibiting some transgranular
after HTTS experiment between 900°C and microcracks in formulation Ca-PSZ after
1500°C (magnification 500) HTTS experiment between 900°C and
1500°C (magnification 2000)
On the microstructure overview shown on figure 4.235, a high crack density is visible
within the sample matrix and more particularly at the level of the grain boundaries of the
tabular alumina and zirconia aggregates as well as a big transgranular crack that
propagates throughout the tabular alumina aggregate in the middle of the picture. A large
356
Results
Figures 4.239 – 4.242 show the microstructure of Mg-PSZ after three thermal shock
cycles between 900°C and 1500°C. On the microstructure overview shown on
figure 4.239, a considerable crack density is visible within the sample matrix and more
particularly at the level of the grain boundaries of the tabular alumina and zirconia
aggregates. This crack density is much higher than that present in Ca-PSZ and also higher
than the density of the pre-existent cracks after sintering. Some transgranular cracks are
also visible within the tabular alumina aggregates on the top right corner of figure 4.239.
If the most of the cracks are propagating at the grain boundaries of the tabular alumina
and zirconia aggregates, some of them are propagating perpendicularly to the surface of
the tabular alumina particles like the particle of the bottom of picture 4.240.
A stronger magnification shown on figures 4.240, 4.241 and 4.242 reveals that those
cracks are propagating along the grain boundaries of zirconia middle-sized aggregates
causing a light debonding of those particles from the matrix and also to a large extent
throughout the matrix and the grain boundaries of coarse tabular alumina aggregates.
Those last mentioned cracks are in most of the case characterized by a more considerable
crack width than the cracks localized at the interfaces between the zirconia particles and
the matrix. A detailed examination of the crack configuration surrounding a zirconia
particle is illustrated on figure 4.242. The zirconia particle is almost debonded from the
357
Results
matrix with a mean gap of around 5 µm between the grain boundary and the matrix. Big
cracks are propagating perpendicularly to the surface of the zirconia particle revealing
typical damage mechanisms during the heating phase. Those cracks exhibit a higher
width of around 10 µm. Some thin cracks are propagating throughout the zirconia
particles. Those transgranular cracks exhibit a mean crack width of around 1 µm. As
shown on figure 4.242, a meandering of the transgranular crack propagation through the
closed porous areas of the zirconia particle is observed.
358
Results
Figures 4.243 – 4.246 show the microstructure of Y-FSZ after three thermal shock cycles
between 900°C and 1500°C. On the microstructure overview shown on figure 4.243,
cracks are scarcely visible within the sample. A stronger magnification is required to
reveal an extreme low crack density within the matrix and more particularly at the level
of the grain boundaries of the tabular alumina and zirconia aggregates. Tabular alumina
and Y-ZrO2 aggregates are perfectly embedded in the calcium aluminate matrix.
Fig. 4.243: Low crack density in Fig. 4.244: Crack propagation throughout
formulation Y-FSZ after HTTS experiment the matrix of formulation Y-FSZ after
between 900°C and 1500°C HTTS experiment between 900°C and
(magnification 60) 1500°C (magnification 150)
Fig. 4.245: Crack deviation by a Y-ZrO2 Fig. 4.246: Thin crack at grain boundary of
particle in formulation Y-FSZ after HTTS a Y-ZrO2 particle in formulation Y-FSZ
experiment between 900°C and 1500°C after HTTS experiment between 900°C and
(magnification 500) 1500°C (magnification 2000)
359
Results
A stronger magnification shown on figures 4.244 and 4.245 reveals that those cracks are
propagating throughout the matrix and at the grain boundaries of tabular alumina and
zirconia middle-sized aggregates. Contrary to Ca-PSZ, the crack is deviated by the
zirconia particle revealing the stability of the crystalline phase. However those cracks
exhibit a local mean distance of around 10 µm between the crack flanks as shown on
figure 4.246.
Such pictures of Y-FSZ show that microstructural damage take place in this formulation,
even if the principal underlying observation consists in showing a dense microstructure
constituted of tabular alumina and zirconia grains of different fractions, which are well
embedded and homogeneously distributed in the calcium aluminate matrix.
360
Discussion
5. Discussion
Within the framework of the present study, resulting elastic properties of refractory
castables after progressive thermal shocks are examined through Resonant Frequency
Damping Analysis (RFDA). Beneficial and detrimental influence of stabilized zirconia
addition to tabular based castable can be investigated through RFDA measurements at
room and elevated temperature. Furthermore, the impact of damage and healing
mechanisms on a long-term stiffness evolution of the tested castables constitutes a
substantial research focus to establish whether such an addition of partially stabilized
zirconia can result in a material toughening under near reality industrial conditions. This
study stresses furthermore the importance of the stabilized zirconia phase as well as the
retained monoclinic content on the resulting elastic behaviour of castable under repeated
heating and/or cooling conditions.
Nevertheless, the interpretation of elastic properties measurements of high alumina
castables through RFDA requires first an examination of each parameter influencing such
measurements. Three variables shall be taken into consideration before correlating
induced microstructural damage in castables and subsequent elastic properties.
In the first instance, the studied castables exhibit a heterogeneous structure in terms of
chemistry and porosity. Therefore, it is necessary to study the elastic properties of dense
ceramics based on alumina and zirconia to be able to assess the influence of the interfaces
between zirconia and alumina particles on those elastic properties by assuming a pore
free microstructure.
In the second instance, refractory castables exhibit a heterogeneous microstructure with
heterogeneous distribution of pores and fine and coarse particles due to manufacturing
process. Such a heterogeneity shall also be taken in consideration to call into question the
obtained elastic properties results.
In the third and last instance, RFDA as Impulse Excitation Technique (IET) also
generates non-linear effects on the acoustic behavior of the tested samples because of the
excitation amplitude required for the signal acquisition and treatment.
The results of the present study are thus discussed according to the following structure in
order to proceed to a critical and constructive interpretation of the resulting elastic
behaviour of high alumina castables after progressive thermal shocks. In a first part, as
the present study focuses on the impact of partially stabilized zirconia addition of thermal
shock resistance, the elastic properties of dense Al2O3-ZrO2 ceramic materials are
examined after sintering and after quenching tests in water. From this part, deeper
knowledge about the interfaces between zirconia and alumina particles in terms of crack
nucleation and friction can be gained. Then in a second part, in order to be able to
transfer those knowledges in the dense Al2O3-ZrO2 system to refractory castables, factors
361
Discussion
The evolution of the Young’s modulus with the zirconia content after sintering is in
agreement with the evolution of both bulk density and apparent porosity. After sintering,
the modulus of elasticity steadily decreases from 403 GPa for pure alumina formulations
up to 208 GPa for pure zirconia formulations. Concerning the pure alumina composition,
the measured Young’s modulus value is in agreement with the theoretical stiffness of
alumina [RIC05] [SAL07], while the measured Young’s modulus value of the pure
zirconia composition is below the theoretical value of partially stabilized zirconia
[CAR08]. If no cracks are visible in the microstructure of formulation 0A-100Z after
sintering (fig. 4.19), a cluster of pores is however visible. On fig. 4.3, the value of the
Young’s modulus of formulation 50A-50Z is furthermore higher than the expected
stiffness of this composition by taking into account a linear decrease of the Young’s
modulus with increasing zirconia content. The sensitivity of the Young’s modulus
measurements through RFDA reveals a characterizing considerable stiffness of this
composition associated with a low apparent porosity. Such a local increase of stiffness in
a binary composition is not revealed by a Voigt’s model [VOI10] and would have not
been revealed neither by a HS-Model of Hashin and Shtrikman [HAS63] nor by Reuss’
model [REU29], as those models estimate the Young’s modulus of the heterogeneous
material as the sum of the contribution of the Young’s modulus of each phase inducing
362
Discussion
therefore either a linear evolution of the stiffness like in the illustrated Voigt’s model on
fig. 4.3 or a parabolic evolution like in the HS-Model or Reuss’ model. Furthermore,
Voigt’s model should be considered as the upper limit of stiffness of a multi-phase
material. Such unexpected Young’s modulus values locally removed from the theoretical
rule of mixtures value have already been observed [TUA01], but this concerns alumina
composites comprising monoclinic and tetragonal zirconia in same proportion. Tuan et al.
showed that Al2O3/ZrO2 composites based on tetragonal stabilized zirconia reveal values
close to the rule of mixture.
Furthermore, formulation 50A-50Z exhibits a microcrack-free microstructure composed
of homogeneously distributed alumina and zirconia particles (fig. 4.15). Such a
microstructure can be modelled as interconnected alumina rich clusters and
interconnected zirconia rich clusters. This microstructure cannot explain this abnormal
high Young’s modulus value of composition 50A-50Z.
Damping measurements show a linear increase of the crack flank friction with increasing
zirconia content after sintering from 1,86.10-4 [-] for pure alumina to 7,51.10-4 [-] for pure
zirconia (fig. 4.5). Composition 50A-50Z reveals a damping value slightly inferior to the
expected damping value by considering a linear increase.
Furthermore, the non-linearity coefficients αf (fig. 4.6) and βf (fig. 4.7) after sintering are
locally low for composition 50A-50Z. This is in agreement with the local high Young’s
modulus value in the sense that this composition should have a low crack density and
tougher connections between zirconia particles and alumina particles compared to close
compositions. Such microcracks can however not be revealed by SEM pictures.
After examining the influence of the crack free microstructure of the dense ceramic
samples after sintering as shown on SEM pictures 4.8 – 4.14, the influence of crack
density and crack configuration on elastic properties of those dense ceramic samples is
discussed next. After a first quenching test in water from 400°C, a stiffness decrease is
observed in the whole Al2O3 – ZrO2 binary system. First, the trend of stiffness evolution
is similar to the Young’s modulus behavior after sintering. However, a more considerable
decrease of the Young’s modulus is noticeable for a zirconia content above 87,5 wt-%.
Furthermore, a local maximum in stiffness is still observed for formulation 50A–50Z
after first quenching test.
The microstructure examination of the tested samples shows that the formulations rich in
zirconia namely 50A–50Z, 12,5A–87,5Z and 6,25A-93,75Z reveals cracks that propagate
across the sample. This global damage of the samples may explain a higher relative
decrease of the stiffness of those samples. As for the alumina rich formulations, namely
100A-0Z and 25A-75Z, some cracks are visible in the sample corner and not throughout
the inner parts of the samples. Independently of zirconia content, cracks propagate in a
363
Discussion
straight line throughout the sample with a low curvature radius (fig. 4.20-4.25), except
for formulation 6,25A-93,75Z, characterized by the highest relative stiffness decrease
after quenching test (above 20 %), that exhibits a sharp edged crack unification, as shown
on fig. 4.24.
Damping values show a clear increase of crack flank friction after a first quenching test
in water. Damping values strongly depend on the zirconia content and oscillate between
1,61.10-3 [-] and 2,29.10-3 [-] for a zirconia content between 0 wt-% and 75 wt-% and
drastically increase for zirconia rich formulations to reach 7,87.10-3 [-].
By analogy with the damping behavior, coefficients αf and βf oscillate within a range of
low values for low zirconia content. Then, both coefficients strongly increase for a
zirconia content ranging between 87,5 wt-% and 100 wt-%. The zirconia content clearly
influences the nature of crack configuration (fig. 4.26-4.31). Two factors are competing
to enhance friction phenomena and ensuing non-linearity of the resonant frequency and
of damping. The first factor is the irregular interface separating alumina particles at the
level of the crack walls being separated by a mean distance of 1 µm. The second factor
corresponds to the higher coefficient of friction between zirconia particles. Zirconia
containing samples exhibit a microcracked microstructure with a primary crack network,
characterized by a crack width up to 6-7 µm. These cracks are propagating straight on
throughout the microstructure Furthermore, some microcracks of the secondary crack
network characterized by a very low crack width (around 1 µm) generate more
considerable friction phenomena. Such an observation is in agreement with [PAS05],
who revealed the enhanced friction at the level of sliding interfaces between zirconia and
alumina in comparison with alumina/alumina interfaces. When the zirconia content
increases, the damping and the non-linearity of the acoustic signal remain almost
constant. This can be explained by the synergetic decrease of the friction phenomena
caused by the reduction of irregular interfaces at the level of crack walls separating
alumina particles and the increase of the friction between zirconia interfaces being
characterized by a higher coefficient of friction. After a second thermal shock cycle, a
further stiffness decrease of the tested formulations is only observed for a zirconia
content between 0 wt-% and 25 wt-%. A third thermal shock has no influence on the
retained Young’s modulus of each tested formulation as shown on fig. 4.3 and 4.4. This
reveals that formulations with a zirconia content up to 25 wt-% still exhibit sufficient
stored elastic energy that acts as driving force for crack initiation and crack propagation
after first thermal shock. The effect of this progressive damage of the microstructure of
alumina rich samples can unfortunately not be monitored by damping measurements
because of the competition between the formation of new cracks generating much friction
and much non-linearity and the widening of the cracks resulting in a lowering of the
friction.
364
Discussion
After examining the effect of friction phenomena generated at the interfaces between
zirconia and alumina particles, the focus directs now to the influence of the
heterogeneous microstructure of high alumina refractory castables induced by the
manufacturing process on the elastic properties.
After thermal shock cycle in air at 600°C, the Young’s modulus of each layer decreases
as shown on fig. 4.36. As expected, the stiffness of the upper layer A drastically
decreases as this upper layer is directly in contact with the annealing medium. Layers B
and C reveal a low relative stiffness decrease (10 % for layer B and 9 % for layer C).
Layer D also reveals a high relative decrease of the stiffness (35 %). Those results reveal
the heat transfer that occurs during such a thermal shock in air between first the upper
layer and the atmosphere and then between the lower layer and the steel plate acting as a
heat transfer support. Layers B and C exhibit similar heat transfer properties as
intermediate layers between the upper layer A and the lower layer D. This observation is
strengthened by the damping values shown on fig. 4.37, which are in agreement with the
Young’s modulus values. The examination of the microstructure of the different layers
365
Discussion
reveals a gradient of crack density form the upper layer A to the lower layer D as shown
on fig. 4.39-4.42. As shown on fig. 4.39, large cracks propagate straight on throughout
the matrix and radially to the big pores until continuing its crack path beyond the pores as
far as the upper layer A is concerned. Some cracks with similar crack path are also visible
in the microstructure of the intermediate layer B as shown on fig. 4.40. Such a crack path
can be regarded as a crack meandering by the pores network [RIT88]. In intermediate
layer C (fig. 4.41) and further in the lower layer D (fig. 4.42), crack density and width are
so low that those cracks are hardly visible on the SEM pictures. If a similar crack density
is noticeable for the intermediate layers B and C, which is in correlation with comparable
stiffness values after thermal shocks, a severe microstructural damage is observed in the
microstructure of the upper layer A, which corroborates the considerable decrease in
stiffness of this layer. However, if a drastic decrease of Young’s modulus is also
observed for the lower layer D, no crack is visible in the microstructure thereof.
Concerning the examination of the non-linear behavior of the acoustic signal of the
different layers, a direct correlation between the Young’s modulus, the damping of the
flexural resonant frequency and the parameters αf and βf is observed: the lower the
Young’s modulus, the higher the damping and the higher the parameters αf and βf. It is
worth mentioning that the values of coefficients αf and βf of layer B are slightly higher
than those of layer C. This would imply the occurrence of damage mechanisms in the
layer B that still do not take place in layer C. Such damage mechanisms cannot be sensed
at this stage by Young’s modulus measurements. Therefore, it could be assumed from
this study that the Young’s modulus measurements reveal the current severity of damage
of samples, while damping measurements and examination of the non-linear behavior of
the acoustic signal inform about the further chronology of damage when similar damage
mechanisms occur. Such a prediction of the sequence of damage mechanisms sensed first
by non-linearity assessment, then by damping measurement and finally recorded by the
determination of the Young’s modulus is in agreement with the study of the damage
mechanisms in dense materials based on metal alloy [ABE02].
366
Discussion
Pore forming agents of different shapes have been therefore added to a standard high
alumina castable formulation to obtain spherical (Ref-sph), cylindrical (Ref-cyl) and
fibre-like (Ref-fib) pores. A quite linear increase of the apparent porosity is thus observed
for each pore forming agent in function of the added volume of pore forming agent. This
linear increase shows that the rheological properties of the tested castables were not
influenced by the addition of the pore forming agent.
367
Discussion
high damping behavior. Indeed, the microstructure of Ref-cyl-4 (fig. 4.58-4.59) reveals
some local damaged regions in the surrounding areas of the pores that may be explained
by an eventual more considerable water absorption by this pore forming agent, while a
clear, continuous and spherical boundary of the porous area and the matrix is observed in
castable Ref-sph (fig. 4.56-4.57).
Concerning castable Ref-fib, the measured damping values are not in agreement with the
measured stiffness. Indeed, Ref-fib that shows a lower stiffness than the other studied
formulations in the whole porosity range (fig. 4.51) exhibits very low damping values. A
two-step evolution is thus observed: the damping values slightly increase from an
apparent porosity of 18,4 vol-% 27,7 vol-% and then increase more significantly up to
32,5 vol-%. However, no particular change in stiffness or ultrasound travel time is
observed for an apparent porosity of 27,7 vol-%. The microstructure of Ref-fib-1,5
(fig. 4.60-4.61) reveals no straight-lined interfaces between the porous areas and the
matrix. The extremity of the flake-like pore is constituted of a high concentration of
microscopic scale porous areas with no defined shape with intermediary regions rich in
alumina fillers and calcium aluminate matrix. The local low distance between the regions
composed of fine particles should favour friction phenomena within the microstructure.
The surrounding areas of the pore can be therefore regarded as damaged parts of the
microstructure exhibiting microporous areas that could be considered as microcracking.
The preferential orientation of the pores in the microstructure that is assumed from the
ultrasound measurements could be proved by applying Boccaccini’s model on the
experimental values (fig. 4.51). A z/x ratio of 2,89 [-] is calculated for Ref-sph, while a
z/x ratio of 3,43 [-] and 3,79 [-] is calculated for Ref-cyl and Ref-fib respectively
(fig. 4.52 - 4.54). Those values show that for each formulation, the pores are assumed to
be prolate, i.e. stretched in the casting direction. The stretching of the pores is more
significant for Ref-cyl and much more significant for Ref-fib. This ranking is directly
correlated with the difference of the ultrasound travel time between the thickness of the
samples and the ultrasound travel time in the width and the length. However those results
should be called into question as the tested castables exhibit a plurality of porosities, such
as the open porosity due to the pore forming agent, the open porosity due to casting, and
the closed porosity of tabular alumina aggregates. Even if the Young’s modulus of the
tested formulations is plotted in function of the measured apparent porosity, the closed
porosity may also influence the results. The calculated z/x ratio of 2,89 [-] of Ref-sph
shows furthermore that the influence of the open porosity due to casting is considerable
in the calculation of the z/x ratio. Hence, this high value of z/x ratio far from the expected
1 [-] for spherical pores may be explained by the preferential orientation of the air
bubbles trapped in the mixture during casting that are moving upward during vibration.
Thus, the contribution of the open porosity due to manufacturing can be assumed to
368
Discussion
correspond to the difference between the calculated z/x ratio of Ref-sph and the
theoretical z/x ratio value for spherical pores, namely 1 [-]. This contribution would be
1,89 [-] and the resulting z/x ratio of the porosity due to pore forming agent would be
1,54 [-] and 1,9 [-] for Ref-cyl and Ref-fib respectively, showing a considerable
stretching of the fibre-like pores in the thickness of the samples. This is in agreement
with the ratio of the thickness to length of this fibre-like pore forming agent in
comparison with cylindrical pore forming agent. This is also merely in the same order of
magnitude as the measured values of the shape factor in porous alumina samples
exhibiting cylindrical pores disclosed by Boccaccini et al. [BOC93] [BOC97].
Thermal shocks in air at 950°C are then performed on formulations Ref-sph-4, Ref-cyl-4
and Ref-fib-1,5 exhibiting an apparent porosity of 24,7 vol-%, 24,7 vol-% and 24,2 vol-%
respectively, in order to examine the influence of the pore shape on the thermal shock
resistance of high alumina castables with similar apparent porosity. The evolution of the
stiffness of those castables is compared to that of formulation Ref to examine furthermore
the impact of the apparent porosity on the thermal shock resistance. As Ref exhibits the
most considerable retained Young’s modulus after five thermal shocks, it can be
concluded that a further increase of the apparent porosity of this castable would not result
in a further enhancement of the thermal shock resistance. Moreover, if the apparent
porosity is increased, the pore shape influences the subsequent behavior of the castable.
Indeed, the castable showing fibre-like porous areas exhibits a higher retained stiffness
(53 %) than formulation constituted of cylindrical pores (47 %), which is also higher than
formulation composed of spherical pores (46 %).
The microstructure examination of those tested castables shows crack initiation at the
grain boundaries of tabular alumina aggregates in Ref-sph (fig. 4.64), while cracks are
propagating tangentially to the interfaces of the height of the cylindrical pores in Ref-cyl
(fig. 4. 66). On the other side, no crack is visible in the microstructure of Ref-fib except
for few transgranular cracks propagating throughout tabular alumina aggregates
(fig. 4.68). The high density of micro-porous areas in the microstructure of Ref-fib would
therefore partly absorb the energy of thermal shock resulting in such a higher retained
stiffness.
The high damping values of Ref-sph can be explained by the dense microcrack network
(fig. 4.64) and the radial propagation of the crack to the surface of the spherical pore
(fig. 4.65). The crack width at the level of the interface between the porous area and the
matrix is around 2 µm. In this region of the microstructure rich in CA6-facets randomly
distributed in the matrix, the enhancement of the crack flank friction phenomena takes
place. The damping increase in Ref-fib can only be explained by the nucleation and the
propagation of transgranular cracks through tabular alumina aggregates (fig. 4.68) as no
369
Discussion
damage is noticeable in the surrounding region of the fibre shaped porous areas
(fig. 4.69). Even if Ref-cyl exhibits a high crack density with tangential propagation of
cracks in the height of the cylindrical pores (fig. 4.66), a stronger magnification shown on
fig. 4.67 shows a debonding of the damaged areas from the matrix. A large crack network
encircles the cylindrical porous area, wherein friction phenomena are not favoured to take
place because of the significant crack width that could reach locally 20 µm. It is worth
mentioning that a damping increase is observed for all pore shape configuration, while a
decrease of coefficient of friction should be expected when a crack is propagating
throughout a porous area [BRA11]. However, the accurate examination of the interfaces
between the porous areas and the crack should be performed to reveal if the
microstructure surrounding those porous areas can inhibit (case of crack widening near
cylindrical pores) or on the contrary enhance friction phenomena (case of interlocking
structure near spherical pores inducing crack surface roughness [RIT88]).
370
Discussion
Concerning Ref and Y-FSZ, a two-step evolution of the non-linear coefficient χE with
increasing thermal shocks may be interpreted by a two-stage microstructural damage with
a first period characterized by the nucleation and the propagation of cracks within the
microstructure and a second period characterized by the unification of cracks and the
widening of cracks. If the non-linearity of the sintered castable exhibiting a crack free
microstructure may be explained by the heterogeneous microstructure constituted of hard
phases (aggregates) separated by a soft bond system (matrix) [OST01], the main source
of non-linearity corresponds to the nucleation and the propagation of cracks. From the
examination of the damage mechanisms occurring in the microstructure of Ref and
Y-FSZ, a clear correlation can be made between the nucleation of isolated cracks that
induces a linear decrease of the non-linear coefficient χE and the unification of cracks and
the propagation of large cracks throughout the castable microstructure that induce an
oscillating behavior of the non-linear coefficient χE.
371
Discussion
stabilized zirconia. From the evolution of the microstructure of these formulations after
progressive thermal shocks, it is observed that samples exhibiting a crack network
essentially initiated at the grain boundaries of tabular alumina aggregates after sintering
and after thermal shocks will reveal a constant non-linear coefficient χE until cracks are
propagating through the matrix and through tabular alumina aggregates inducing a
decrease of χE. This coefficient further oscillates when further thermal shocks induce a
widening of the cracks.
The amplitude of impulse excitation also influences the damping behavior of the flexural
resonant frequency (fig. 4.74). Thus, the damping of same Ref samples can vary between
4.92.10-3 [-] and 17,86.10-3 [-] after four thermal shocks according to the amplitude of
excitation. This observation first shows the questionable inter-laboratory comparison of
results in terms of damping if the experimental procedure is not exactly the same as the
amplitude of excitation is not defined in the standard. However, contrary to the Young’s
modulus, wherein a clear linear decrease of stiffness is observed with increasing
amplitude of excitation (see fig. 4.71), the damping increases generally with increasing
amplitude of excitation, but no clear evolution trend is noticeable (fig. 4.75). Therefore,
to illustrate the non-linear behavior of the flexural damping with increasing amplitude of
excitation, a parameter ς is defined according to equation 4.2. The plotting of non-linear
coefficient ς with the number of thermal shocks shown on fig. 4.76 and fig. 4.77 reveals
same trend of evolution for formulations Ref and Y-FSZ, and Ca-PSZ and Mg-PSZ
respectively.
By analogy with the non-linearity of the Young’s modulus, the evolution of the
non-linearity of the damping after progressive thermal shocks may be interpreted by a
372
Discussion
two-stage microstructural damage for Ref and Y-FSZ with a first period characterized by
the nucleation and the propagation of cracks within the microstructure up to 4 thermal
shocks for Ref and 6 thermal shocks for Y-FSZ, wherein the non-linear parameter ς
steadily increases, and a second period characterized by the unification and the widening
of cracks up to the tenth thermal shock, wherein the parameter ς steadily decreases.
In order to accurately study the influence of the amplitude of the impulse excitation on
the elastic properties of high alumina castables at elevated temperature, coefficients χE,T
related to the Young’s modulus and 𝜒𝜁𝑓,𝑇 related to the damping of the flexural resonant
frequency are calculated by assessing the slope of the evolution of the relative Young’s
modulus and the damping of the flexural resonant frequency ζf respectively with
increasing voltage delivered to the electromagnet that ejects the hammer for providing
the mechanical excitation according to equations 4.4 and 4.5 respectively.
The evolution of the non-linear coefficient χE,T of formulations Ref, Ca-PSZ and Y-FSZ
is examined during the heating and cooling phases of the heating treatment (fig. 4.188
and 4.189). First, Ref and Y-FSZ depend to a very small extent on the amplitude of
impulse excitation as the non-linear coefficient χE,T remains almost constant in the whole
373
Discussion
temperature range from room temperature to 1500°C during both heating and cooling
phases.
Concerning Ca-PSZ, the non-linear coefficient χE,T drastically and almost linearly
increases above 1050°C up to 1500°C from -0,005 [-] to 0,154 [-] at 1500°C. This shows
the contribution of the martensitic transformation of zirconia on the non-linear behavior
of the acoustic signal at elevated temperature. During the heating phase, debonding of
zirconia particles takes place from 1050°C. Such a microstructural damage mechanism
influences the acquisition of the Young’s modulus according to the amplitude of impulse
excitation, so that the stiffness increases more and more with increasing impulse intensity
after the martensitic transformation. The increase of the non-linear coefficient χE,T is
furthermore quite logarithmic up to a temperature of 1350°C showing that the further
expansion of debonded zirconia grains partially filling the voids between the zirconia
particles and the matrix tends to limit the non-linear behavior.
During the cooling phase, the considerable volume expansion of zirconia particles
induces microcracking. As the coefficient χE,T is quite high between 1500°C and 900°C
during the cooling phase with values above 0,2 [-], this shows that the Young’s modulus
tends to increase with increasing amplitude of impulse excitation in a microstructure
configuration, wherein zirconia particles are debonded from the matrix. From 900°C
(according to fig. 4.189), the martensitic transformation takes place inducing
microcracking of the microstructure. This microcracking explains that the Young’s
modulus tends to decrease with increasing amplitude of impulse excitation. Both
mechanisms compensate for each other explaining that the χE,T values from 800°C to
room temperature are almost similar to the values observed during the heating phase in
the same temperature range. Thus, from the examination of the non-linear behavior of the
Young’s modulus of Ca-PSZ after one thermal treatment following the sintering, it can
be concluded that no further damage takes place. Moreover, it is worth mentioning that
the decrease of coefficient χE,T from 900°C to 800°C during the cooling phase is more
considerable that the increase of same coefficient after martensitic transformation during
the heating phase. That reveals a higher sensitivity of the non-linear behavior of the
Young’s modulus with regard to microcracking in comparison with debonding. More
significant effect of microcracking over debonding could also be observed during high
temperature acoustic emission measurement [BRI08]. The narrower range of
temperature, wherein coefficient χE,T is drastically decreasing, can lead to an accurate
measurement of the martensitic transformation temperature during the cooling phase.
Figures 4.191 and 4.192 show the evolution of the non-linear coefficient 𝜒𝜁𝑓,𝑇 of
formulations Ref, Ca-PSZ and Y-FSZ during the heating and cooling phases. While an
374
Discussion
375
Discussion
Zirconia is added as substituents of medium grain sized tabular alumina and reveals a
similar particles size distribution (fig. 3.6), in order to achieve comparable Andreassen
packing coefficients close to 0.30 (see tab. 3.8). This results in limiting the impact of the
particle packing on the flow properties of the slurry as well as on the porosity and the
pore size distribution of the sintered product, and finally on the high temperature
mechanical properties [MYH94] [MYH96], more particularly the material stiffness
[WAN84]. However, a monomodal particle size distribution is observed for MgO-ZrO2
grains tending to coarser grains than the multimodal distribution of middle-sized zirconia
grains of CaO-ZrO2 and Y2O3-ZrO2 (fig. 3.6). By working with three zirconia raw
materials with similar grain size distribution, the influence of the grain size on the
generation of a crack network during the cooling phase defined through a fracture
mechanics approach and the principle of brittleness number BN is almost limited
[HAR97].
376
Discussion
The addition of zirconia revealing different retained monoclinic content may furthermore
have an impact on the castables thermal shock behavior.
The calculation of the volume fraction of monoclinic phase according to equation 3.1
[POR79] [MIL81] is performed from XRD analyses carried out on all three zirconia raw
material (fig. 3.10 - 3.12). MgO doped zirconia reveals a considerable volume fraction of
monoclinic content (above 70 vol-%), while Y2O3 and CaO doped zirconia exhibit a
lower monoclinic content (under 5 vol-%) with 4,9 vol-% and 1,5 vol-% respectively (see
tab. 3.11).
Those estimated monoclinic contents from XRD analyses can be completed by the study
of the dilatometric behavior of tested castables shown on fig. 4.80 and 4.81. Indeed,
formulations Ref and Y-FSZ exhibit a linear behavior in the whole temperature range up
to 1500°C. The mean value of the thermal expansion coefficient α between room
temperature and 1000°C of Ref and Y-FSZ (8,27.10-6 K-1 and 8,34.10-6 K-1 respectively)
is around 5% higher than that of formulations Ca-PSZ and Mg-PSZ with 7,97.10-6 K-1
and 8,01.10-6 K-1 respectively. No contraction or expansion is noticeable in Y-FSZ
revealing a low monoclinic content. On the other side, Ca-PSZ and Mg-PSZ reveal a
typical hysteretic thermal expansion of a material containing monoclinic phase as shown
on figure 2.10 [SAL07]. However, Mg-PSZ characterized by a high monoclinic content
exhibits a negligible contraction during the heating phase and a negligible expansion
during the cooling phase, both inferior to 0,001 %. Ca-PSZ characterized by a similar
monoclinic content as Y-FSZ according to XRD analyses exhibits the highest contraction
during the martensitic transformation from monoclinic to tetragonal (0,033 %) and the
highest expansion during the martensitic transformation from tetragonal to monoclinic
(0,044 %). Furthermore, the thermal expansion during the cooling phase is higher than
the contraction during the heating phase. From these dilatometric measurements, it can be
obviously assumed that CaO-ZrO2 has a higher retained monoclinic content than
Y2O3-ZrO2. The low hysteretic thermal expansion of Mg-PSZ, while MgO-ZrO2 exhibits
according to XRD analysis a huge content of retained monoclinic phase, can be explained
by the dense network of large cracks (fig. 4.92 - 4.95), acting as “buffer” when further
expansion of the particles occurs during heat treatment following the sintering.
Finally, DTA-TG measurements shown on tab. 3.12 reveal the high retained monoclinic
content in Mg-ZrO2 through the considerable measured enthalpy during the heating phase
(99,5 µVs.mg-1), while those DTA-TG measurements show a low enthalpy for Ca-ZrO2
and Y-ZrO2 with 7,8 µVs.mg-1 and 7,2 µVs.mg-1 respectively.
From this estimation of the retained monoclinic content, two families of castables can be
distinguished: Ref and Y-FSZ exhibiting no (Ref) or a negligible content of monoclinic
377
Discussion
The estimation of the retained monoclinic phase in the tested zirconia raw materials is in
agreement with the damaged microstructure thereof. On one side, as shown on
fig. 4.96 - 4.99, Y-FSZ shows tabular alumina and zirconia aggregates that are well
embedded in and bonded to the matrix.
On the other side, as shown on fig. 4.88 - 4.91, microcracks are noticeable in the
microstructure of Ca-PSZ after sintering. Those cracks are mainly propagating at the
grain boundaries of large tabular alumina and zirconia middle-sized aggregates. By
propagating throughout the matrix, the crack is either deviated by the zirconia particle or
is propagating throughout the zirconia particle (see fig. 4.90). In general, the interfaces
between the zirconia particle and the matrix reveal a strong connection between the
formed calcium aluminate phases and the zirconia particles.
Concerning Mg-PSZ, a considerable crack density is visible within the sample matrix and
more particularly at the level of the grain boundaries of the tabular alumina aggregates
(fig. 4.92 - 4.95). This crack density is much higher than that in Ca-PSZ. Some
transgranular cracks are also visible within the tabular alumina aggregates (fig. 4.92).
Those pre-existent cracks are also propagating along the grain boundaries of zirconia
middle-sized aggregates causing a light debonding of those aggregates from the matrix
and also to a large extent throughout the matrix and the grain boundaries of coarse tabular
alumina aggregates (see fig. 4.93 and 4.94). As shown on figures 4.94 and 4.95, cracks
are also propagating perpendicularly to the grain boundary of the Mg-ZrO2 particles,
while cracks are only propagating at the grain boundary of the zirconia particles in case
of Ca-PSZ. This shows that during cooling, cracks are initiated by the considerable
volume expansion of zirconia particles transforming into monoclinic. Such cracks are
next merging in the debonding area of zirconia particles.
The doping of the raw material aims to obtain metastable tetragonal zirconia particles
capable of transforming into monoclinic during cooling phase when microcracking
occurring. Therefore, beyond the estimation of the retained monoclinic content, the
estimation of the tetragonal content also constitutes an important criterion to assess the
ability of the castables to resist to thermal shocks through toughening mechanisms. The
estimation of this phase is unfortunately rendered difficult by the superposition of the
peaks related to the tetragonal phase and those related to the cubic phase in case of XRD
analyses.
As the manufacturing process of those stabilized zirconia powders influences the
formation of solid solutions between zirconia and the doping agent and as the
manufacturing steps of production of studied stabilized zirconia are unknown, the
378
Discussion
assessment of zirconia phase in raw material powders based on phase diagram should be
also called into question. Therefore, no reliable estimation of tetragonal phase could be
carried out in the present work.
Fig. 4.79 illustrates the dependency between the Young’s modulus, the MOR, and the
crack flank friction properties of the four tested castables to reveal the distinction
between formulations Ref and Y-FSZ characterized by a considerable stiffness (147 GPa
and 125 GPa respectively), a considerable bending strength (42 MPa and 34 MPa
respectively) and low damping values (1,92.10-3 [-] and 2,24.10-3 [-] respectively) and
formulations Ca-PSZ and Mg-PSZ. Ca-PSZ is characterized by medium bending strength
(20 MPa), a medium stiffness (66 GPa) and medium damping values (12,71.10-3[-]),
while Mg-PSZ is characterized by a low bending strength (11 MPa), a low stiffness
(36 GPa) and high damping values (25,78.10-3 [-]). This diagram shows first the effect of
zirconia addition in a crack free microstructure by comparing Ref and Y-FSZ resulting in
a low decrease of stiffness (15%), a low decrease of bending strength (18 %) and a low
increase of damping (17 %). Then, this diagram also shows the correlation between
damping increase and crack density increase: the damping increases by 567 % for
Ca-PSZ and 1151 % for Mg-PSZ, while the Young’s modulus decreases by 47 % and
71 % for Ca-PSZ and Mg-PSZ respectively. This correlation is allowable as after
sintering the distance between the crack flanks in both formulations Ca-PSZ and Mg-PSZ
is similar with a mean crack width of around 1-2 µm between zirconia particles and the
matrix (see fig. 4.91 and 4.95).
According to Dong et al. the tested castables can be classified independently of the
microstructural damage according to their stiffness – damping behavior [DON13]. The
calculation of the parameter │E*│. tan ψ would give a value of 0,3 GPa for Ref and
Y-FSZ, while this parameter is 0,8 GPa for Ca-PSZ and 0,9 GPa for Mg-PSZ. The
obtained value for Ref and Y-FSZ is characteristic for conventional materials as their
value is below 0,6 GPa, while the value for Ca-PSZ and Mg-PSZ is characteristic for
materials suitable for vibration damping applications as their value is over 0,6 GPa.
The assessment of all those elastic, mechanical and thermal can be implemented into the
calculation of Hasselman parameter in order to predict both crack initiation and crack
propagation resistances. The values of R and R’’’ shown on tab. 4.6 reveal the distinction
between castables Ref and Y-FSZ from castables Ca-PSZ and Mg-PSZ. Indeed, Ref and
Y-FSZ are characterized by relative high R-values, with 28 [K] and 26 [K] respectively
and by low R’’’-values, 1,0.10-4 [Pa-1] and 1,3.10-4 [Pa-1] respectively. This shows the
effect of zirconia addition to high alumina castables on thermal shock behavior in case of
crack free microstructure after sintering. Indeed, such an addition induces a moderate
decrease of the crack initiation resistance (2 [K]) but a considerable increase of crack
379
Discussion
propagation resistance (0,3.10-4 [Pa-1]). On the other side, Ca-PSZ and Mg-PSZ are
characterized by high R-values, with 35 [K] and 35,5 [K] respectively, and by
considerable R’’’-values with 1,8.10-4 [Pa-1] and 3,3.10-4 [Pa-1] respectively.
Nevertheless, no particular attention should be acknowledged on the calculated R-value
of Ca-PSZ and Mg-PSZ as those formulations already exhibit a microcrack network after
sintering [HAS69]. For similar reasons, the comparison between Ca-PSZ and Mg-PSZ in
terms of crack propagation resistance is questionable as Mg-PSZ has already reached a
considerable level of damage after sintering (fig. 4.92 or 4.93). In such a configuration of
damaged structure, further thermo-mechanical stresses induce a widening of the already
present cracks in addition to a moderate further propagation of such cracks.
However, Ref and Y-FSZ are characterized by a considerable stored elastic energy,
5,94 kPa and 4,75 kPa respectively showing that a lot of energy is available to be used as
driving force for crack initiation and propagation. If the addition of Y-ZrO2 particles
results in a decrease of 20 % of the stored elastic energy, adding CaO-ZrO2 and Mg-ZrO2
results in a more considerable decrease of the stored elastic energy, 46 % and 74 %
respectively. As the crack driving energy Gf is proportional to the stored elastic energy
[SCH04], it should not be considered that the addition of zirconia does not result in a
rising R-curve as expected [SCH04], as some cracks are already noticeable in the
microstructure of the tested castables. Furthermore, no impact of the doping agent, as it
could be studied by [HAN00], can be stressed from the calculation of the stored elastic
energy for similar reasons.
First, the thermal shock behavior of those tested castables is studied in terms of
quenching tests in air at 950°C. This distinction between those four tested castables can
be verified, by considering both evolution of the Young’s modulus and relative Young’s
modulus with increasing thermal shocks (fig. 4.100 and 4.102), as well as the evolution
of the bending strength as shown on fig. 4.104.
380
Discussion
Indeed, on the one hand, Ref and Y-FSZ exhibit a low two-step decrease of the Young’s
modulus with increasing thermal shocks. On the other hand, Ca-PSZ and Mg-PSZ follow
a three-step decrease with increasing thermal shock cycles.
The stiffness evolution of the tested castables can be correlated with characteristic
damage mechanisms observed within their microstructure. In case of Ref and Y-FSZ, a
two-step evolution of the microstructure damage can be observed. After the first thermal
shock cycle, few cracks are nucleated principally in the matrix of the castables as shown
on fig. 4.105 for Ref and 4.129 for Y-FSZ. Those formed cracks progressively propagate
throughout the matrix. The toughening mechanism occurring in Ref and Y-FSZ is
therefore the deviation of the cracks by tabular alumina and zirconia aggregates
preventing the crack from propagating straight on through the microstructure as it can be
seen on fig. 4.111 for Ref and on fig. 4.136 for Y-FSZ with a sawtooth crack path. Such a
crack propagation can be regarded as a contact shielding strengthening mechanism
characterized by a considerable crack surface roughness [RIT88].
Furthermore Ref and Y-FSZ exhibit high stiffness and high bending strength values after
progressive thermal shocks. The slope of the degradation of the modulus of elasticity in
function of the decrease in modulus of rupture is low. This corresponds to a stiffness
decrease of 2 GPa for a bending strength decrease of 1 MPa for Ref, while a stiffness
decrease of 1 GPa for a bending strength decrease of 1 MPa for Y-FSZ is observed. For
these formulations, the evolution of stiffness and bending strength can be correlated with
parameter R’’’. The higher the parameter R’’’, the lower the decrease of the material
stiffness in function of the decrease of its bending strength.
In case of Ca-PSZ and Mg-PSZ, more complex damage mechanisms are observed. For
both formulations, the three-step degradation of material stiffness can be correlated with
a three-step damage mechanism. Indeed, the first decrease of the Young’s modulus is
caused by the nucleation of new cracks. Then, the widening of the cracks takes place in a
further stage of damage inducing a more and more significant debonding of large
aggregates from the matrix. After reaching a threshold of crack widening and crack
propagation, the further thermal shocks have no additional impact of the damaged
material microstructure explaining the constant measured Young’s modulus values.
Furthermore, Ca-PSZ and Mg-PSZ exhibit a low stiffness and a low bending strength
values after progressive thermal shocks. The slope of the degradation of the modulus of
elasticity in function of the decrease in modulus of rupture is furthermore high. This
corresponds to a stiffness decrease of 3 GPa for a bending strength decrease of 1 MPa for
Ca-PSZ, while a stiffness decrease of 5 GPa for a bending strength decrease of 1 MPa for
Mg-PSZ is observed. In case of Ca-PSZ and Mg-PSZ, the considerable decrease of the
stiffness compared to the decrease in bending strength can be explained by two factors:
first, the retained MOR of those castables is enhanced by the splintered shape of both
381
Discussion
CaO-ZrO2 and MgO-ZrO2 particles (see fig. 4.89 for Ca-PSZ and fig. 4.93 for Mg-FSZ)
that causes an interlocking structure having a beneficial impact on retained material
bending strength [BAB11]. Such an interlocking structure is more visible in the
microstructure of Ca-PSZ. Then, the results of RFDA measurements strongly depend on
debonding of aggregates, as the measured stiffness drastically decreases with crack
widening (fig. 4.101). Nevertheless, no correlation is noticeable between the decrease of
the material stiffness, the decrease of material bending strength and the thermal shock
parameter R’’’ in case of Ca-PSZ and Mg-PSZ. Therefore it can be concluded that the
correlation between R’’’ parameter and the decrease of the material stiffness and bending
strength is not valid for castables exhibiting pre-existent cracks after sintering.
The distinction between castables Ref and Y-FSZ, and castables Ca-PSZ and Mg-PSZ is
also observed in terms of damping evolution as shown on fig. 4.102 (in case of absolute
damping values) and on fig. 4.103 (in case of relative damping values).
On the one hand, Ref and Y-FSZ exhibit lower damping values than that of Ca-PSZ and
Mg-PSZ after each thermal shock cycle. If no trustful correlation between the evolution
of the retained Young’s modulus and the evolution of the damping for Ref and Y-FSZ
can be performed, it is worth mentioning that the damping evolution of those
formulations can be correlated with the evolution of the non-linear behavior of those
castables with regard to the stiffness (χE) and the damping (ς).
The assessment of the non-linear coefficients χE and ς combined with the measurement of
the damping would conclude to a two-step damage mechanism in those formulations: a
first step characterized by the crack nucleation and the crack propagation during the first
four thermal shocks for Ref and six thermal shocks for Y-FSZ followed by an unification
and a crack widening during the subsequent thermal shock cycles. A comparison of the
crack width in Ref (see fig. 4.109 - 4.112) shows a mean distance between the crack
flanks of around 5 µm after one thermal shock (see fig. 4.109), three thermal shocks (see
fig. 4.110) and five thermal shocks (see fig. 4.111). After 10 cycles, the distance between
the crack flanks reaches its maximum value, above 15 µm and locally up to 20 µm, and
this phenomenon of crack widening takes place at the level of the grain boundary of
tabular alumina aggregate as shown on figure 4.112. In case of Y-FSZ, this
microstructure observation is not as obvious as in case of Ref as few cracks are visible
within the matrix after one (see fig. 4.133), three (see fig. 4.134) and five (see fig. 4.135)
thermal shocks and those cracks exhibit furthermore irregular distance between their
flanks. However a fair estimation of this distance between the flanks would be of around
5 to 10 µm up to the fifth thermal shock, while this distance can reach 20 µm after ten
thermal shocks (see fig. 4.136).
A simplified interpretation of damping evolution of castables Ca-PSZ and Mg-PSZ
would consist in a two-step evolution that can be correlated with the evolution of the
382
Discussion
Young’s modulus. Indeed, in a first stage, further cracks are propagating through the
matrix of the castables Ca-PSZ and Mg-PSZ that already reveal cracks after sintering.
Those cracks are mainly propagating at the grain boundary of tabular alumina and
zirconia particles (see fig. 4.113 - 4.115 for Ca-PSZ and fig. 4.121 and 4.122 for
Mg-PSZ). This induces a decrease of the Young’s modulus (see fig. 4.100 and 4.101) and
an increase of the damping (see fig. 4.102 and 4.103). Then in a second stage, crack
widening occurs with a more and more considerable debonding of the aggregates from
the matrix, while cracks still continue to propagate throughout the matrix generating new
friction surfaces (see fig. 4.116 for Ca-PSZ and fig. 4.123 and 4.124 for Mg-PSZ). The
competition between the creation of new friction interfaces generating damping and the
crack widening reducing the damping explains the constant values of damping observed
in Ca-PSZ from the fifth thermal shock cycle and the low decrease of damping in
Mg-PSZ from the third thermal shock. In case of Mg-PSZ, the low decrease of damping
shows that the crack widening is more important than the propagation of cracks. If single
damage mechanisms can be observed within the material microstructure, namely crack
initiation or crack propagation [KOV15] or debonding in heterogeneous materials
[IDR13], an increase of damping is expected to take place. However, when damage
mechanisms of different nature take place synergistically, those damage mechanisms
compete with each other. Thus, formation of new cracks or propagation of pre-existent
cracks result in a damping increase, while crack widening or an advanced debonding of
particles from the matrix result in a damping decrease.
The main inter granular damage mechanism observed in each castable corresponds to the
propagation of cracks tangentially along the surface of tabular alumina aggregates or
zirconia particles. During such cooling thermal shocks, radial tensile stresses are
383
Discussion
preponderant at the grain boundary of tabular alumina and zirconia aggregates because of
the coefficient mismatch between the calcium aluminate matrix and the zirconia and/or
tabular alumina particles [BEA04]. This explains the propagation of cracks mainly at the
grain boundary of the zirconia and/or tabular alumina particles.
384
Discussion
1300°C therefore influences the thermal shock behavior of the castable in terms of
expected crack initiation and crack propagation resistance. While after sintering at
1500°C, Ca-PSZ exhibits a R parameter of 35 [K] and a R’’’ parameter of 1,8.10-4 [Pa-1],
this castable exhibits after sintering at 1300°C a R parameter of 25,5 [K] and a R’’’
parameter of 1,8.10-4 [Pa-1]. As after sintering at 1300°C, the castable exhibits a crack
free microstructure, the high R parameter value is meaningful and is slightly lower than
that of Y-FSZ (26 [K]). Furthermore, the calculated R’’’ parameter reveals a considerable
crack propagation resistance, slightly greater than after sintering at 1500°C.
However, this prediction of the thermal shock behavior is not in agreement with the
measured Young’s modulus values after progressive thermal shock tests in air at 950°C
compared with the already discussed results of the same formulation sintered at 1500°C.
Indeed, despite the expected enhanced crack formation resistance, the Young’s modulus
of Ca-PSZ sintered at 1300°C follows an exponential decrease characterized by a drastic
decrease after the first two thermal shock cycles to reach a stiffness of 87 GPa (i.e. a
relative value of 73 %) followed by a low rate decrease up to the tenth thermal shock to
reach a final value after ten thermal shocks of 69 GPa (i.e. a relative value of 57 %) as
shown on fig. 4.141 and 4.142. Furthermore, both formulations show almost the same
evolution of the relative Young’s modulus after progressive thermal shocks, exhibiting
the same relative Young’s modulus (57 %) after the whole series of thermal shocks
(fig. 4.142).
The SEM pictures shown on fig. 4.145 – 4.150 revealing the microstructure of Ca-PSZ
sintered at 1300°C after progressive thermal shocks do not however show a dense
network of cracks. If no crack is visible on the castable overview after one (see
fig. 4.145) and five (see fig. 4.147) thermal shocks, large cracks are visible after ten
thermal shocks (see fig. 4.149). By analogy with Ca-PSZ sintered at 1500°C, the main
damage mechanism occurring in Ca-PSZ sintered at 1300°C corresponds to a debonding
of the particles from the matrix. However, contrary to the castable sintered at 1500°C,
wherein the debonding is noticeable at large scale (see fig. 4.113 – 4.115), a stronger
magnification is needed to reveal the debonding of small tabular alumina or small
zirconia particles from the matrix in case of the sintered castable at 1300°C (see
fig. 4.146 and 4.148). This shows that a light debonding of small particles from the
matrix in high alumina castables can be monitored through RFDA measurements, so that
a drastic decrease of the material stiffness after progressive thermal shocks does not
automatically correspond to the formation of cracks.
Debonding and formation of cracks in Ca-PSZ sintered at 1300°C result in increasing the
damping behavior of the castable at a different extent. The first increase of damping can
be correlated by the light debonding of small tabular alumina particles shown on
385
Discussion
fig. 4.146, while the second damping increase can be correlated to the crack formation at
the grain boundaries of both tabular alumina and zirconia aggregates shown on fig. 4.148
and 4.150. This shows different friction phenomena observed between tabular alumina or
zirconia particles and the matrix according whether the matrix is mainly composed of
CA6 facets or CA2 particles. Furthermore, in comparison with formulation Ca-PSZ
sintered at 1300°C, the low relative damping values of Ca-PSZ sintered at 1500°C can be
explained by the local distance between the particles and the matrix of around 5 µm as
shown on fig. 4.117 – 4.119. Such a distance between the crack flanks limits the friction,
while a less significant debonding is observed in case of Ca-PSZ sintered at 1300°C with
very thin cracks propagating at the grain boundary of the aggregates (fig. 4.146 and
4.148). Finally, after ten thermal shocks, if large cracks are visible at the grain boundary
of zirconia grains with a crack width of around 5 up to 10 µm, the superficial damage of
the zirconia particle resulting in forming micro-debris inserted in the cracks generating
friction phenomena also contributes to the damping increase defined as corrosion debris
inducing crack closure and crack surface interference (fig. 4.150) [RIT88].
As expected, the degradation of the Young’s modulus directly depends on the severity of
the thermal shock: the higher the temperature difference of thermal shock, the higher the
decrease of the stiffness (see fig. 4.151) [PER10]. For each testing temperature, a
traditional two-step exponential evolution is noticeable with a strong decrease of the
Young’s modulus during the first three thermal shocks followed by a low decrease up to
the tenth thermal shock. Ca-PSZ exhibits a relative final Young’s modulus value of 82 %,
74 %, 57 % and 43 % for quenching temperatures of 750°C, 850°C, 950°C and 1050°C
386
Discussion
respectively after the whole series of thermal shocks (except for quenching temperature
of 1050°C, wherein the Young’s modulus could not be trustfully assessed after the tenth
thermal shock). In accordance with the evolution of the material stiffness, the
enhancement of friction phenomena directly depends on the severity of the thermal
shock: the higher the temperature difference of thermal shock, the higher the ensuing
crack flank friction (see fig. 4.152) [PER10]. Contrary to the evolution of the Young’s
modulus, the evolution trend is different according to the temperature difference, while a
proportional increase of the damping behavior is expected with increasing temperature
difference [PER10]. However, Pereira et al. studied the damping behavior of high
alumina castables only based on alumina raw materials. Therefore the addition of
zirconia influences the friction phenomena within the castable microstructure.
However, if no direct correlation can be drawn between the evolution of the Young’s
modulus and the evolution of the damping according to the quenching temperature, the
logarithm of the flexural damping is linear in function of the logarithm of the Young’s
modulus for each tested temperature difference as shown on fig. 4.153.
The decrease rate of the logarithm of the flexural damping in function of the logarithm of
the Young’s modulus is the same for low temperature difference (750°C and 850°C). For
higher temperature differences such as 950°C and 1050°C, said decrease rate increases
with the temperature difference showing that prediction of the damping evolution in
accordance with the Young’s modulus evolution is only possible for low temperature
difference (below 850°C). Beyond a temperature difference threshold, which is in the
present case between 850°C and 950°C, such a prediction is challenging.
The considerable thermal stresses generated at quenching temperature above this
temperature difference threshold induce a quick crack widening that limits the friction
phenomena resulting in a limited increase of the damping.
For these reasons, the prediction of the damping behavior of a castable subjected to
thermal shocks in function of its stiffness is possible when progressive damage
mechanisms are restricted to crack initiation and crack propagation without crack
widening after each thermal shock. As soon as crack widening takes place, the ensuing
decrease of friction results in reducing the increase of damping values, while the Young’s
modulus of the castable keeps on decreasing.
Furthermore, the multi-step evolution of the damping can be directly correlated with
typical damage mechanisms. Indeed, a four-step increase of the relative damping is
noticeable in case of thermal shocks at 1050°C: first, the damping drastically increases up
to the second thermal shock corresponding to the nucleation of cracks at the grain
boundary of tabular alumina grains (fig. 4.154). Then, this first increase is followed by a
second significant increase up to the fifth thermal shock cycle corresponding to the
387
Discussion
propagation of cracks within the matrix and throughout the tabular alumina aggregates
(fig. 4.155). Afterwards, the relative damping remains constant up to the eighth thermal
shock cycle corresponding to crack deviation along zirconia particles (fig. 4.156)
inducing a partial debonding of the zirconia particles from the matrix with a relative high
gap limiting friction phenomena. Finally, the damping undergoes a last considerable
increase after the ninth thermal shock before fracture. This last increase corresponds to
the crack propagation throughout zirconia particles resulting in the nucleation of new
crack flanks with high friction phenomena (fig. 4.157) [PAS05].
During such thermal shock tests, damage mechanisms mainly occur during the quenching
into the annealing medium, this annealing medium exhibiting a considerable heat
capacity [BUY70]. While Ref and Y-FSZ exhibit a more suitable thermal shock
resistance in terms of stiffness decrease after progressive thermal shocks in air compared
388
Discussion
to Ca-PSZ (see fig. 4.100 and 4.101), Ca-PSZ exhibits a more suitable thermal shock
resistance than Ref and Y-FSZ after quenching tests in molten aluminum (fig. 4.158).
The high retained Young’s modulus value of Ca-PSZ is in agreement with the low
damaged microstructure of this castable after each thermal shock. Indeed, the
microstructure of Ca-PSZ after the first quenching test shown on fig. 4.169 is similar to
the microstructure of the castable as sintered (see fig. 4.137 - 4.140), i.e. a crack free
microstructure. Zirconia particles are however slightly debonded from the matrix
showing that very few damage mechanism also occur during the cooling phase. Few
cracks mostly propagating perpendicularly to the surface of zirconia particles are visible
after three thermal shocks (see fig. 4.170) and five thermal shocks (see fig. 4.171). A
relative dense crack network is only visible after ten thermal shocks with perpendicularly
propagation of the cracks to the surface of zirconia particles.
As far as Ref is concerned, the nucleation of large cracks is already visible after the first
thermal shock (fig. 4.161). By analogy with Ca-PSZ, the cracks are mostly propagating
perpendicularly to the tabular alumina aggregates. After three quenching tests, the
castable shows a homogeneous distribution of large cracks (see fig. 4.162), while
transgranular cracks are further noticeable after five quenching tests (see fig. 4.163). The
microstructure of Ref after ten thermal shocks reveals a dense network of large cracks,
wherein cracks which are propagating perpendicularly to the surface of large tabular
alumina aggregates are further merging at the grain boundary of other tabular alumina
aggregates (fig. 4.164).
Contrary to Ref and Ca-PSZ, the low retained Young’s modulus values in Y-FSZ are not
induced by the propagation of cracks throughout the matrix but by the propagation of
transgranular cracks through large tabular alumina aggregates (fig. 4.179 after five
thermal shocks and fig. 4.180 after ten thermal shocks). A stronger magnification is
needed to show that after the first and the third thermal shock, small cracks propagating
perpendicularly to small zirconia particles are nucleating in the castable matrix (fig. 4.181
after one thermal shock and 4.182 after three thermal shocks). Those cracks are further
deviated by zirconia particles inducing a debonding of those zirconia particles from the
matrix (see fig. 4.181 - 4.184).
The damping behavior of the tested castables after progressive quenching tests can be
correlated with the damage mechanisms. Indeed, on one hand, Ca-PSZ reveals low
damping values after each progressive thermal shock (fig. 4.159). A stronger
magnification reveals that after the fifth and the tenth quenching tests, the crack network
in the surrounding areas of zirconia particles is more noticeable (fig. 4.175 and 4.176),
inducing a partial debonding of those zirconia particles with a low mean distance value
between the zirconia surface and the matrix, around 1-2 µm.
389
Discussion
On the other hand, Ref and Y-FSZ reveal almost similar damping values after each
quenching test and exhibit quite the same final damping values after the whole series
with a relative value of 1192 % for Ref and 1386 % for Y-FSZ (fig. 4.159). First, the high
damping values of Ref can be explained by the dense crack network (fig. 4.164) and by
the propagation of numerous small cracks throughout the microstructure with a mean
crack width of around 3 µm generating friction between CA6-facets (fig. 4.166 and
4.167). After five thermal shocks, a crack widening is observed, wherein the distance
between the crack flanks keeps on growing to reach 7-8 µm after five thermal shocks and
more than 10 µm after ten quenching tests. The competition of the nucleation and the
propagation of new cracks in the microstructure of Ref and the crack widening of pre-
existent cracks explains the quite constant damping values from the fifth quenching test
to the tenth as shown on fig. 4.159.
In case of Y-FSZ, the increase of damping can be first explained by the nucleation of
transgranular cracks, particularly from the fifth thermal shock (see fig. 4.179 and 4.180)
and the nucleation of small cracks perpendicularly to the surface of small zirconia
particles as shown on fig. 4.181 and 4.182. A partial debonding of the zirconia particles
from the matrix generated by the nucleation of such cracks characterized by a low
distance between the zirconia particles and the matrix, typically 1-2 µm, enhances the
damping behavior of the castable. After ten thermal shocks, large cracks (with a more
considerable width of around 10 µm) that are propagating perpendicularly to the surface
of medium sized zirconia particles are also visible in the matrix (fig. 4.184).
The main inter granular damage mechanism observed in each castable corresponds to the
perpendicular propagation of cracks either to the surface of tabular alumina aggregates or
zirconia particles as illustrated above. This shows the significant radial compressive
390
Discussion
stresses generated by tabular alumina particles and zirconia particles onto the matrix
during such heating thermal shocks as the coefficient of thermal expansion of both
alumina and zirconia is higher than that of calcium aluminate phases [BEA04].
Finally, the comparison of the evolution of the Poisson’s ratio of castable Ca-PSZ after
quenching test in molten aluminum and after standardized thermal shocks in air clearly
shows that thermal shocks in air induce a more anisotropic behavior of the castable after
progressive thermal shocks, as in such experimental procedure, stresses are mainly
localized on the upper part of the samples as it could be observed on the microstructure
of sample Ref (see fig. 4.39 - 4.42). During quenching test in molten aluminum, stresses
are more homogeneously distributed on each face of the sample.
Castables Ref and Y-FSZ can be distinguished from Ca-PSZ and Mg-PSZ by their
Young’s modulus evolution with increasing temperature after sintering.
On the one hand, Ref and Y-FSZ show a low hysteretic behavior between heating and
cooling phases (fig. 4.193). For both formulations, the Young’s modulus first slowly
decreases from their initial value at room temperature up to a temperature around
1250°C. This first stiffness decrease coincides with the theoretical linear behavior of the
Young’s modulus in function of the temperature for mono-phase materials [WEB99].
Afterwards, the Young’s modulus follows a drastic second decrease up to 1500°C. In this
temperature range, the evolution of the Young’s modulus may be more regarded as
obeying an exponential-law [WAC61]. During the cooling phase, the Young’s modulus
of both formulations increases up to 1250°C exactly as the Young’s modulus was
decreasing in this temperature range during the heating phase. Then, in the second part of
the cooling phase between 1250°C and room temperature, a low hysteretic behavior is
noticeable as the Young’s modulus of both formulations is higher during the cooling
phase than during the heating phase. The Young’s modulus at room temperature after
cooling is similar as the Young’s modulus of the castable before HT-RFDA measurement
revealing that no further microstructural damage occurs during HT-RFDA. Such an
elastic behavior observed in Y-FSZ is in agreement with the elastic behavior of the model
material constituted of a vitreous glass matrix and 30 vol-% spherical alumina inclusions
391
Discussion
studied by Joliff et al. [JOL08]. Based on the Hashin and Shtrikmann’s model HS-, a
stabilization of the stiffness should be observed during the cooling phase of Y-FSZ from
800°C and room temperature. However, a low decrease of stiffness is observed within
this temperature range (around 2 GPa). Such a stiffness decrease during the cooling phase
may be explained by the debonding of the inclusion from the matrix as suggested by
Joliff et al. [JOL08], and may also be explained by phase transformation with
characteristic volume expansion inducing crack network generation [PAT12].
Furthermore, no particular decrease of stiffness is observed during the cooling phase of
Ref as shown on fig. 4.193, except for a limited flattening of the curve between 200°C
and room temperature, showing that the evolution of the Young’s modulus of Ref tends
to follow more closely the theoretical stiffness evolution defined by Hashin and
Shtrikman’s model HS-.
Concerning Ca-PSZ and Mg-PSZ, a more significant hysteretic behavior can be observed
(fig. 4.194). First the Young’s modulus is almost constant from room temperature up to
the martensitic transformation temperature of zirconia, namely 1051°C for Ca-PSZ and
963°C for Mg-PSZ. It is noticeable that no linear decrease of the Young’s modulus is
observed in this temperature range contrary to Ref and Y-FSZ. This can be explained by
further microstructural mechanisms preventing the stiffness decrease, such as the higher
thermal expansion of zirconia particles filling the voids and/or cracks in the surrounding
areas of the matrix. Then the Young’s modulus logarithmically increases for both
formulations up to 1500°C. Contrary to Ref and Y-FSZ, a non-reversible thermo-elastic
behavior is observed during the cooling phase. The Young’s modulus increases up to the
martensitic transformation temperature of zirconia, namely 905°C for Ca-PSZ and 959°C
for Mg-PSZ. Then, in a second part of the cooling phase, the Young’s modulus
drastically decreases up to room temperature. This decrease is exponential-like for
Ca-PSZ. As for Mg-PSZ, a two-step decrease is observed from 959°C to 903°C followed
by a second drastic decrease from 815°C to 683°C. Afterwards, the Young’s modulus
slightly decreases up to room temperature. Those HT-RFDA measurements reveal further
damage for formulations Ca-PSZ and Mg-PSZ already exhibiting some cracks within the
microstructure after sintering, as a stiffness decrease at room temperature at the end of
HT-RFDA measurements of 8 % for Ca-PSZ and 17 % for Mg-PSZ is observed.
392
Discussion
Damping measurements performed during the heating phase (see fig. 4.195) and cooling
phase (see fig. 4.196) reveal a two-step evolution in case of Ref and Y-FSZ. This
two-step evolution can be correlated with the two-step decrease of the Young’s modulus.
Thus, the increase of damping between 1200°C and 1500°C corresponding to the strong
decrease in stiffness can be correlated with the formation of liquid phases especially at
the grain boundary of large tabular alumina and zirconia aggregates [SCH00] [PEZ96]
that induces grain reorientation, which generates friction.
By analogy with the creep behavior [SCH00b] [NOW72], the evolution of the damping is
thermo-activated obeying to the Arrhenius-law [VER94] or more generally to an
exponential-law [WEL04] [SIM14] and the increase rate of the damping can be
correlated with the creep rate of the castable during CiC measurements [DRY89]. While
the damping increase is correlated in the literature with diffusion mechanisms in dense
ceramics, in the present studied castables, the damping increase is correlated with
formation of viscous phases at grain boundaries of aggregates.
Thus, Y-FSZ exhibiting a strong exponential behavior of the damping during both
heating and cooling phases (see fig. 4.195 and 4.196) is also characterized by the highest
v5h-25h value (0,047 %/h), while Ref and Ca-PSZ exhibiting similar exponential increase
of damping during both heating and cooling phases (see fig. 4.195 and 4.196) are
characterized by similar v5h-25h values (0,045 %/h and 0,043 %/h respectively).
Furthermore, the examination of the damping evolution of the zirconia based castables
Y-FSZ and Ca-PSZ at low temperature shows a low increase of the crack flank friction at
temperature around 420°C for Y-FSZ and 450°C for Ca-PSZ. Such damping increases
can be correlated with the reverse transformation from monoclinic to tetragonal phase of
the monoclinic particles previously formed at lower temperatures during cooling after
sintering or due to Low Temperature Degradation [BAS99]. As this damping increase is
more significant for Ca-PSZ, this reveals that the retained monoclinic content in
CaO-ZrO2 is higher than in Y2O3-ZrO2.
393
Discussion
cycling HT-RFDA measurements permit to examine the impact of fatigue on the material
stiffness and a fortiori on the material thermo-mechanical properties.
However such stable stiffnesses after fatigue test are not directly correlated with an
unchanged microstructure. Indeed, Ref reveals some cracks as shown on fig. 4.200 and
4.201. Those cracks are mostly localized at the grain boundary of large tabular alumina
aggregates (fig. 4.201 and 4.202). Such a propagation of cracks, tangentially to the
surface of large tabular alumina aggregates, shows that the cracks are mostly nucleating
during the cooling phase, while radial tensile stresses emerge at the grain boundary of
tabular alumina particles. However, those cracks are characterized by a considerable
distance between the crack flanks, locally around 20 µm (fig. 4.203). Such an increased
crack density should have induced a decrease of the material stiffness after the test.
Nevertheless, the numerous thermal cycles between 900°C and 1500°C furthermore
contribute to further sintering mechanisms in terms of complete conversion of CA2
particles into CA6 facets resulting in a stiffening of the matrix, Moreover, cycling thermal
cycles also generate healing mechanisms due to the formation of viscous phases at
elevated temperature. Concerning Y-FSZ, a lower crack density is observed than in Ref
(fig. 4.212 and 4.213). By analogy with Ref, large cracks are mainly localized at the grain
boundary of large tabular alumina aggregates (fig. 4.213 and 4.214). Such cracks reveal
locally a considerable width of around 20 µm (fig. 4.214). Furthermore, superficial
cracks are also observed in some Y2O3-ZrO2 particles.
Concerning Ca-PSZ and Mg-PSZ, no sawtooth stiffness evolution is noticed (fig. 4.198).
As for Ca-PSZ, the Young’s modulus increases from an initial value at room temperature
of 68 GPa to 78 GPa at 1500°C and the stiffness values at 1500°C slightly increase from
78 GPa after the first thermal cycle to 82 GPa after the last thermal cycle. The material
stiffness then increases up to the temperature of transformation of zirconia close to 900°C
before drastically decreasing up to 900°C. At this temperature of 900°C, the Young’s
modulus decreases from 111 GPa after the first thermal cycle to 91 GPa after the fifteenth
thermal cycle revealing further damage mechanisms after progressive thermal cycles due
to the repeated expansion of non-stabilized zirconia particles.
394
Discussion
By analogy with Ref and Y-FSZ, large cracks are noticeable within the microstructure of
Ca-PSZ, mainly at the grain boundary of large tabular alumina aggregates or tangentially
to large tabular alumina aggregates (fig. 4.204 to 4.206). The distance between the crack
flanks is much lower than in case of Ref and Y-FSZ with a mean distance of around 5 µm
(fig. 4.207). This tangential crack propagation shows that the cracks are mostly nucleated
during the cooling phase and this crack initiation is favored by the martensitic
transformation of zirconia. During its propagation, the crack is either crossing over
zirconia particles through the closed pore network or is deviated by other zirconia
particles (fig. 4.206 and 4.207). The multiple crack deviation by the zirconia particles
consuming a considerable energy explains the limited crack widening. Even if the crack
density is more considerable within the microstructure of Ca-PSZ (fig. 4.204) compared
to that of Ref and Y-FSZ, leading to local debonding of large tabular alumina aggregates,
further sintering mechanisms as well as healing mechanisms are sufficient to compete
with such damage mechanisms, explaining the stable measured Young’s modulus values
at 1500°C. As far as Mg-PSZ is concerned, a similar elastic behavior as Ca-PSZ is
observed during the first nine thermal cycles. After this ninth thermal cycle, the crack
density in the microstructure of Mg-PSZ is such that the stiffness monitoring cannot be
trustfully interpreted. This sample damage is essentially due to the repeated
transformation of the numerous and large monoclinic zirconia particles resulting in the
decrease of the stiffness. Even if a dense crack network is already observed after sintering
(fig. 4.92 to 4.94), Mg-PSZ exhibits a critical crack density after the fatigue test as shown
on fig. 4.208 and 4.209. Large cracks are propagating orthogonally to the surface of
zirconia particles (fig. 4.208) revealing that such cracks are mainly initiated during the
cooling phase after the martensitic transformation of zirconia, whereby radial
compressive stresses are applied by the transformed zirconia particles onto the matrix.
Besides those orthogonal cracks, transgranular cracks crossing over zirconia particles are
also observed (fig. 4.208) as well as a considerable debonding of large zirconia particles
with a local gap of around 10 µm (fig. 4.210 and 4.211). These damage mechanisms all
together result in a strong decrease of the material stiffness.
By focusing now on Ca-PSZ, fig. 4.198 has shown that the stiffness of this formulation is
strongly affected during the cooling phase by the successive martensitic transformations.
The examination of the Young’s modulus evolution during heating cycles (see fig. 4.216)
and cooling cycles (see fig. 4.217) shows that the martensitic transformation temperature
can be accurately assessed through HT-RFDA measurements. As most of the damage
mechanisms occur during the cooling phase, the emphasis is laid on the evolution of the
temperature of transformation during the cooling phase. Associated to a strong stiffness
decrease, an increase of damping is furthermore observed as soon as the martensitic
transformation has taken place. The study of the evolution of the transformation
395
Discussion
temperature of zirconia in Ca-PSZ during the cooling phase (fig. 4.218) reveals that both
damping and Young’s modulus measurements agree to conclude to a linear increase of
the martensitic temperature after progressive cooling cycles. The increase rate is 2,14°C
per cycle according to Young’s modulus measurements and 1,98°C per cycle according
to damping measurements.
Ref and Y-FSZ are characterized by high initial expansion (1,26 % and 1,21 %
respectively) as shown on fig. 4.219. This maximum expansion is exponentially
decreasing to reach a final value of 0,61 % and 0,35 % respectively. As far as Ca-PSZ
and Mg-PSZ are concerned, both are characterized by a lower initial expansion at 1500°C
(0,94 % and 0,95 % respectively). This maximum expansion at 1500°C exponentially
decreases to reach a value of 0,28 % and 0,23 % respectively.
On the other side, Ref can be distinguished from the zirconia based formulations by the
low creep rate deduced from the exponential decay of the maximum expansion at 1500°C
as shown on fig. 4.220. Ref exhibits indeed a creep rate with a very low exponential
coefficient (-0,005 [-]), while the zirconia based castables are characterized by an
exponential coefficient of -0,009 [-] for Y-FSZ followed by -0,010 [-] for Mg-PSZ and
-0,011 [-] for Ca-PSZ. First, the discrepancy between the zirconia based materials and
formulation Ref can be explained by the lower temperature of viscous phase formation.
Single RuL measurements have shown a temperature of maximal expansion of 1435°C
for Ca-PSZ and 1383°C for Mg-PSZ, while this temperature of maximal expansion is
1457°C for Ref (tab. 4.4). Furthermore, even if Y-FSZ exhibits a higher temperature of
maximum expansion (1473°C as shown on tab. 4.4), the lower viscosity of the viscous
phase also contributes to accelerate the creep rate. Therefore, Y-FSZ exhibiting the most
considerable v5h-25h value during CiC measurements as shown on tab. 4.5 (0,047 %), may
reveal the viscous phase with the lowest viscosity explaining this significant creep rate.
The v5h-25h values of Ca-PSZ and Mg-PSZ shown on tab. 4.5 (0,043 % and 0,034 %
respectively) can be considered as key indicators showing that the viscosity of the
viscous phases formed in those castables is higher than those formed in Ref exhibiting a
v5h-25h value of 0,045 %.
396
Discussion
Therefore, two criteria are competing in terms of creep enhancement: the temperature of
formation of viscous phases (low for Ca-PSZ and Mg-PSZ) and the viscosity of the said
viscous phase (low for Y-FSZ). This explains the similar high creep rate during cycling
RuL measurements for formulations Ca-PSZ, Mg-PSZ and Y-FSZ, while Ref exhibits a
low creep rate during cycling RuL measurements. Furthermore, the pre-existent cracks in
the microstructure of Ca-PSZ and Mg-PSZ shown fig. 4.88 and 4.92 respectively can
furthermore be partially filled with said viscous phases facilitating a reorientation of the
grains under compressive stresses enhancing therefore creep mechanisms.
397
Discussion
The shift of the martensitic transformation temperature of zirconia in Ca-PSZ with the
number of thermal cycles is illustrated in the following diagram.
constant ∆
750 800 850 900 950 1000 1050 1100 1150 1200 1250
Temperature (°C)
Tm of coarsest zirconia particles
The damage mechanisms observed after HTTS are similar to those observed after cycling
HT-RFDA measurements. First, Ref reveals a low density of cracks as shown on
fig. 4.231 as it could also be observed after cycling HT-RFDA (fig. 4.200 and 4.201).
Those cracks are mostly propagating at grain boundaries of large tabular aggregates
(fig. 4.231 to 4.234) as it could also be observed after cycling HT-RFDA on fig. 4.201
and 4.202. Furthermore, transgranular cracks crossing over tabular alumina grains are
also observed (fig. 4.232). Like after cycling HT-RFDA tests, the cracks are large and
exhibit a mean width that can locally reach 20 µm as shown on fig 4.234.
As far as Y-FSZ is concerned, the crack density is low but higher than that observed in
Ref (fig. 4.243). A large crack is visible through the whole sample (fig. 4.243 and 4.244).
Such a damage mechanism is not observed after cycling HT-RFDA measurements
(fig. 4.212 and 4.213). Those cracks are progressively deviated by zirconia particles (see
398
Discussion
fig. 4.245) and the distance between the crack flanks can locally reach 10 µm at grain
boundaries of zirconia particles (fig. 4.246). At lesser extent, such cracks are also
observed in Y-FSZ after cycling HT-RFDA measurements (fig. 4.214).
In case of Mg-PSZ, a dense unified crack network is observed as shown on fig. 4.239.
Those large cracks exhibiting a mean width of around 10-20 µm (see fig. 4.242) are
propagating straight forward and perpendicularly to the surface of zirconia particles
(fig. 4.239 - 4.242). The cracks are further propagating at the grain boundary of large
tabular alumina aggregates resulting in a partial debonding (fig. 4.239). Such damage
mechanisms are also observed after cycling HT-RFDA measurements but in a lesser
extent as shown on fig. 4.208 - 4.211.
The repeated transformation of zirconia induces radial compressive stresses onto the
matrix forming the dense network of large cracks. However, Mg-PSZ based on zirconia
with a high retained monoclinic content can be distinguished from Ca-PSZ based on
zirconia with a low retained monoclinic content. Indeed, most of the cracks in Mg-PSZ
are propagating orthogonally to the surface of zirconia particles resulting from the radial
compressive stresses generated by the volume change of non-stabilized zirconia particles.
In case of Ca-PSZ and in a lesser extent in case of Y-FSZ, most of the cracks are
propagating at the grain boundary of large tabular alumina aggregates and are either
deviated by small zirconia particles or they cross over other zirconia particles. This
shows that such a crack network is generated during the successive cooling phases when
radial tensile stresses are induced at the grain boundaries of tabular alumina and zirconia
aggregates. This also shows furthermore the influence of the closed porosity of zirconia
particles on crack propagation. Depending on whether the zirconia particle is porous or
not, the crack is either deviated by the zirconia particle or can propagate through the pore
network of the zirconia particle. Furthermore, the crack path is also influenced by the
bonding properties of the interfaces between aggregates and matrix. Indeed, as it is
observed in Ca-PSZ after sintering (see fig. 4.91), Mg-PSZ (see fig. 4.95) and Y-FSZ
(see fig. 4.99), calcium aluminate phases are highly concentrated at the grain boundary of
399
Discussion
zirconia particles, while some microporous areas can be observed at the grain boundary
of tabular alumina particles. This can be explained by the challenging deflocculation of
acidic zirconia particles in a basic calcium aluminate matrix. Therefore, zirconia particles
are strongly embedded within the matrix explaining that the energy required to separate
those zirconia particles from the matrix despite the considerable coefficient of thermal
expansion mismatch should be much more important than the energy required for
separating the tabular alumina particles from the matrix.
By hitting zirconia particles, three crack paths can be observed: either the crack is
deviated by the ensuing volume expansion of the stabilized zirconia particle transforming
into monoclinic resulting in toughening mechanisms [CAR07] through zone shielding
[RIT88], or the crack is propagating at the grain boundary of zirconia particles, wherein
the local concentration in calcium aluminate phases is abnormally relative low (see
fig. 4.238), or the crack is propagating throughout the pore network of the zirconia
particle (see fig. 4.120).
The weak interfaces between tabular alumina particles and the matrix are already
noticeable in formulation Ref (see fig. 4.233 and 4.234). This shows that the difference of
thermal expansion between the tabular alumina particles and the calcium aluminate
phases is sufficient to generate cracks [BEA04].
Example of crack deviation resulting from the volume expansion of zirconia particle in
Ca-PSZ after 3 quenching tests in molten aluminum (left) and schematic illustration of
toughening mechanism (right)
400
Discussion
Example of tangential crack propagation in Ca-PSZ after HTTS experiment (left) and
schematic illustration of weak bonding between zirconia particle and matrix (right)
401
Conclusions
6. Conclusions
To conclude, the objective of the present study was the post-mortem and in situ
examination of microstructural damage in high alumina refractory castables subjected to
thermal shocks through Resonant Frequency Damping Analysis (RFDA). Two major
research axes were examined: a first part of the present study was dedicated to the
measuring principle of RFDA. The results gained in Young’s modulus and damping
should be called into question as heterogeneous materials are considered to be a source of
experimental uncertainties. Secondly, thermal shock resistance enhancement of high
alumina castables by addition of stabilized zirconia was another focus of research.
402
Conclusions
In other words, the stiffness values measured by RFDA should be corrected by taking
into consideration the coefficient χE to limit as much as possible the impact of the
impulse excitation.
The influence of the impulse excitation on both stiffness and damping behavior of
castables has also been sensed at elevated temperature. The coefficient χE,T related to the
non-linearity of the Young’s modulus with temperature follows a sudden drastic increase
when debonding takes place, while this coefficient abruptly decreases in case of
microcracking. The coefficient 𝜒𝜁𝑓,𝑇 related to the non-linearity of the damping with
temperature strongly increases from a characteristic material temperature. This increase is
nevertheless dampened by creep phenomena occurring at elevated temperature that
moderate the increase of the non-linearity.
Anisotropy of microstructure of castables due to manufacturing process
If the assessment of the Young’s modulus and the damping of refractory castables
through RFDA leads to an integrative information of the elastic properties, refractory
castables should be regarded as materials exhibiting an anisotropic porosity in the casting
direction. The upper layer of such materials reveals a denser network of pores and
therefore a lower stiffness, a higher damping as well as a higher non-linearity behavior.
Furthermore, this upper layer is first subjected to a steeper thermal gradient during
thermal shock test in air. Besides pore distribution, the upper layer also reveals an
anisotropy of crack distribution and crack geometry after testing. Young’s modulus
measurements therefore reveal the intensity of damage, while damping measurements
and analysis of the non-linear behavior of the acoustic signal inform about the further
chronology of damage when similar damage cases are repeatedly superimposed.
Pore shape
The elastic properties of castables are influenced by the pore shape. Castables
exhibiting similar apparent porosity values may reveal a different stiffness according to
whether the pores are spherical, cylindrical or fibre-like. Fibre-like pores significantly
decrease the material stiffness, while the decrease of the Young’s modulus is limited in
case of spherical pores. Boccaccini’s model applied on the tested castables shows the
strong correlation between the shape of the pore forming agent and the calculated z/x
ratio. The pore forming agent has a preferential direction within the microstructure. This
preferential direction of the pore forming agent could be proved by ultrasound
measurements. Furthermore, if the influence of the apparent porosity on the thermal
shock behaviour of dense ceramic materials is well known [SAL17], the pore shape
seems to play a further role in the enhancement of thermal shock resistance. Radial
propagation of cracks through spherical porous areas within the porous castable matrix is
detrimental for the thermal shock resistance as this microstructural damage induces a
significant degradation of the material stiffness and a considerable damping increase. A
propagation of cracks tangentially to the surface of cylindrical pores within the porous
403
Conclusions
castable matrix has a limited impact on the degradation of the material stiffness and on
the damping increase. It is worth mentioning that the studied castables, besides the
artificial shaped pores, also reveal an interconnected pore network within the matrix as
well as closed pores within the aggregates that also influence the thermal shock resistance
of said castables. The impact of the pore shape on the thermal shock resistance of
ceramic materials needs to be performed on dense materials to conclude to the influence
of the pore shape. The thermal shock resistance of porous castables can be therefore
enhanced by the control of the pore shape and namely by the control of the z/x ratio, as
the more elongated the pore, the higher the retained Young’s modulus and the lesser the
crack density.
The examination of the elastic behaviour of dense ceramic materials of the Al2O3-ZrO2
system combined with the study of the parameters influencing the monitoring of the
404
Conclusions
First, refractory castables are designed to study the influence of different stabilized
zirconia on the toughening mechanisms in presence of thermo-mechanical stresses.
13,75 wt-% of tabular alumina grains are substituted by zirconia particles. Such an
amount corresponds in the previously discussed dense Al2O3-ZrO2 system to
formulations exhibiting a significant stored elastic energy inducing further damage
mechanisms after progressive thermal shocks. Three types of zirconia are tested:
Mg-ZrO2, Ca-ZrO2, and Y-ZrO2 exhibiting a considerable (> 80 vol-%), moderate
(~ 5 vol-%) and low (< 5 vol-%) retained monoclinic content according to XRD and
DTA-TG analyses performed on the raw zirconia powders and according to dilatometric
and HT-RFDA measurements performed on the tested castables Mg-PSZ, Ca-PSZ and
Y-FSZ.
Those raw material characteristics have a strong impact on the microstructure of the
castable after sintering, and therefore on the mechanical and elastic properties of the
castables and finally on their thermal shock resistance. Thus, Ref and Y-FSZ are
characterized by a microcrack free microstructure after sintering, wherein tabular alumina
and Y-ZrO2 particles are well embedded within the matrix, explaining considerable
stiffness, MOR, HMOR or CCS values. On the other side, Ca-PSZ exhibits few cracks
after sintering, which are more specifically located at the grain boundary of Ca-ZrO2
particles. Such a deterioration of the microstructure induces a moderate decrease of
stiffness, MOR, HMOR or CCS values. Finally, Mg-PSZ exhibits a considerable crack
network after sintering. Such a significant deterioration of the microstructure induces a
drastic decrease of stiffness, MOR, HMOR or CCS values. Therefore, Ref and Y-FSZ
can be distinguished by their high stored elastic energy and their high crack propagation
resistance, represented by parameter R, contrary to Ca-PSZ and Mg-PSZ characterized by
a reduced stored elastic energy and an improved crack propagation resistance represented
by parameter R’’’.
The distinction between Ref and Y-FSZ on one side and Ca-PSZ and Mg-PSZ on the
other side is also noticed in terms of thermal shock behaviour according to standardized
quenching method in air. Indeed, Ref and Y-FSZ are characterized by a low (absolute or
relative) stiffness decrease and by a low decrease of the bending strength after
progressive thermal shocks, while Ca-PSZ and Mg-PSZ are characterized by a higher
(absolute or relative) stiffness decrease as well as a more drastic decrease of MOR after
progressive thermal shocks. Such resulting elastic and mechanical behaviours can be
explained by the low density of induced cracks within the microstructure of Ref and
405
Conclusions
Y-FSZ, while large cracks can be observed in the microstructure of Ca-PSZ and Mg-PSZ.
Concerning Ca-PSZ and Mg-PSZ, the splintered shape of both Ca-ZrO2 and Mg-ZrO2
particles causes an interlocking structure and therefore enhances the mechanical
properties of the castables [BAB11]. Thermal shock experiments induce in these
castables the propagation of pre-existent cracks through the matrix as well as the
widening of these cracks. Moreover, the assessment of the material Young’s modulus
through RFDA measurements strongly depends on the nature of the microstructural
damage in such a way that the measured stiffness drastically decreases with crack
widening when namely debonding of aggregates occurs. Concerning Ref and Y-FSZ, the
two-step evolution of the non-linear coefficients χE and ς correlated with the two-step
evolution of the damping conclude to a two-step damage mechanism. The first step is
characterized by the crack nucleation and the crack propagation during the first four
thermal shocks for Ref and six thermal shocks for Y-FSZ followed by unification and
crack widening during the subsequent thermal shock cycles. As for Ca-PSZ and Mg-PSZ,
the two-step evolution of the damping can be correlated with the two-step evolution of
the Young’s modulus. Indeed, in a first stage, further cracks are propagating through the
matrix of the castables Ca-PSZ and Mg-PSZ that already reveal cracks after sintering.
Those cracks are mainly propagating at the grain boundary of tabular alumina and
zirconia particles. This induces a decrease of the Young’s modulus and an increase of the
damping. Then, in a second stage, crack widening occurs with a more and more
considerable debonding of the aggregates from the matrix, while cracks still continue to
propagate throughout the matrix generating new friction surfaces. The competition
between the creation of new friction interfaces generating damping and the crack
widening reducing the damping explains the constant damping values observed in
Ca-PSZ from the fifth thermal shock cycle and the low damping decrease in Mg-PSZ
from the third thermal shock. In case of Mg-PSZ, the low damping decrease shows that
the crack widening is more important than the propagation of cracks.
The sintering temperature has a theoretical beneficial influence on the thermal shock
behaviour of the Ca-ZrO2 based castable. Lowering the sintering temperature from
1500°C to 1300°C induces the enhancement of both crack initiation and crack
propagation resistances characterized by the increase of both R and R’’’ parameters and
furthermore leads to a microcrack free microstructure after sintering. The Young’s
modulus of the castable sintered at those temperatures follows the same degradation trend
in terms of relative values. Such a stiffness decrease in the castable sintered at 1300°C
should however be tempered by the significant sensitivity of Young’s modulus
monitoring through RFDA measurements when the castable microstructure exhibits a
light debonding of zirconia particles from the matrix. The extreme low gap (below 1 µm)
406
Conclusions
between zirconia particles and the matrix after progressive thermal shocks explains the
damping enhancement of the castable while no particular crack formation is observed.
The thermal shock behaviour of Ca-ZrO2 based castable furthermore depends on the
temperature difference of thermal shock. An increase of the quenching temperature
results in a more considerable degradation of the material stiffness and a more significant
material damping enhancement. However, if no direct correlation can be drawn between
the evolution of the Young’s modulus and the evolution of the damping according to the
quenching temperature, the logarithm of the flexural damping is linear with the logarithm
of the Young’s modulus. The decrease rate of the logarithm of the damping in function of
the logarithm of the Young’s modulus is the same for low temperature difference (750°C
and 850°C) showing that the damping evolution of a castable is predictable when
microstructural damage are restricted to crack initiation and crack propagation without
crack widening. As soon as crack widening occurs, namely for temperature differences
above a temperature threshold being between 850°C and 950°C, crack flanck friction is
reduced limiting the increase of damping values, while the Young’s modulus keeps on
decreasing. Furthermore, the multi-step evolution of the damping can be directly
correlated with typical damage mechanisms. Indeed, a four-step increase of the relative
damping is noticeable in case of thermal shocks at 1050°C: first, the damping drastically
increases up to the second thermal shock corresponding to the nucleation of cracks at the
grain boundary of tabular alumina grains. Then, this first increase is followed by a second
significant increase up to the fifth thermal shock cycle corresponding to the propagation
of cracks within the matrix and throughout the tabular alumina aggregates. Afterwards,
the relative damping remains constant up to the eighth thermal shock cycle corresponding
to crack deviation along zirconia particles inducing a partial debonding of the zirconia
particles from the matrix with a relative high gap between the zirconia particle and the
matrix limiting friction. Finally, the damping undergoes a last considerable increase after
the ninth thermal shock before fracture. This last increase corresponds to the crack
propagation throughout zirconia particles resulting in the nucleation of new crack flanks
with high friction properties [PAS05].
While Ref and Y-FSZ exhibit a more suitable thermal shock resistance in terms of
stiffness decrease after progressive thermal shocks in air compared to Ca-PSZ, the
contrary is observed after progressive quenching tests in molten aluminum. Thus, during
cooling thermal shock in air, radial tensile stresses are generated at the grain boundary of
both tabular alumina and zirconia particles inducing the propagation of cracks at the grain
boundary of those particles. During heating thermal shocks, radial compressive stresses
are generated at the grain boundary of both tabular alumina and zirconia particles
inducing the propagation of cracks perpendicular to the aggregate surface. If the
interfaces between zirconia particles and the matrix are prone to crack initiation and
407
Conclusions
crack propagation in case of cooling thermal shocks, those interfaces are much more
stable in case of heating thermal shock. Furthermore, quenching tests in molten
aluminum generate homogeneous stresses within the castable microstructure in
comparison with standardized thermal shock in air, wherein the same sample face is
subjected to more severe thermal gradient. The examination of the evolution of the
castable Poisson’s ratio after progressive thermal shocks should be therefore taken into
consideration in order to call into question the RFDA signal acquisition. The absence of
cracks propagating tangentially to the grain boundaries of alumina and zirconia
aggregates after quenching test in aluminum reveals that the cooling rate during such
quenching tests is not sufficient enough to damage said castables. Therefore, it can be
assumed that microstructure damage is only induced by thermal stresses during heating
thermal shock in case of quenching test in aluminum.
By analogy with the measurement of material elastic properties at room temperature, Ref
and Y-FSZ can be distinguished from Ca-PSZ and Mg-PSZ according to their elastic
behaviour at elevated temperature. On the one hand, Ref and Y-FSZ show a low
hysteretic behavior between heating and cooling phases with a quasi-reversible stiffness
evolution during the cooling phase that almost follows the Hashin and Shtrikmann’s
model HS-. Concerning Ca-PSZ and Mg-PSZ, a more significant hysteretic behavior as
well as a non-reversible thermo-elastic behavior during the cooling phase can be
observed revealing the significant retained monoclinic content in both Ca-ZrO2 and
Mg-ZrO2. Damping measurements at elevated temperature reveal a thermo-activated
behaviour obeying an Arrhenius-law [VER94] or more generally an exponential-law
[WEL04] [SIM14] and the increase rate of the damping can be correlated with the creep
rate of the castable during CiC measurements [DRY89]. The creep behavior of the
castable is influenced by the competition between the temperature of formation of the
viscous phases and the viscosity of said viscous phase [DRY89]. Thus, the formation of
viscous phases takes place at lower temperature when those viscous phases are rich in
MgO and CaO (below 1400°C and around 1400°C respectively), while first viscous
phases rich in Y2O3 are formed at temperature around 1500°C. However, the enrichment
of the viscous phases in MgO or CaO increases the viscosity, while the enrichment of the
viscous phases in Y2O3 decreases the viscosity. Quantity and viscosity of the viscous
phases at grain boundaries of aggregates influence the reorientation of the grains within
the matrix and therefore influences both creep and damping behaviour at elevated
temperature. Therefore, the lower viscosity of the viscous phases formed in Y-FSZ
compensates the lack of quantity of viscous phases compared to Ca-PSZ and Mg-PSZ.
408
Conclusions
of heating thermal shock, but has a detrimental impact in terms of thermal shock
resistance in case of cooling thermal shock. Refractory castables are in practice subjected
to both heating and cooling thermal shocks. Therefore, the examination of the fatigue
resistance through cycling HT-RFDA and RuL measurements is a key indicator to
conclude whether the addition of zirconia in high alumina refractory castable constitutes
a basis of material development.
According whether the castable exhibits or not a hysteretic behaviour during single
HT-RFDA measurements, this would influence the elastic behaviour of the castable
during cycling HT-RFDA measurements at elevated temperature. On one side, in both
formulations Ref and Y-FSZ, the cycling thermal treatments between 900°C and 1500°C
generate few cracks mainly localized at the grain boundary of large tabular alumina
aggregates and a widening of these cracks occurs during the fatigue test. The sawtooth
stiffness evolution with constant stiffness values observed during the test may be
explained by further sintering mechanisms resulting in a material stiffening and by
healing mechanisms induced by the formation of viscous phases at elevated temperature.
In case of Ca-PSZ and Mg-PSZ, the retained monoclinic content is the driving force of
damage mechanisms during fatigue test. Compared to Ref and Y-FSZ, castable Ca-PSZ
exhibits a more considerable crack density after the test, leading to local debonding of
large tabular alumina aggregates. However, further sintering mechanisms as well as the
healing mechanisms occurring during fatigue tests are sufficient to compete with these
damage mechanisms, explaining the stable measured Young’s modulus values at 1500°C.
Concerning Mg-PSZ, the retained monoclinic content is critical and therefore the induced
compressive stresses at the grain boundary of transformed zirconia particles during the
cooling phase generates a considerable crack network within the microstructure
explaining the significant stiffness decrease during fatigue test.
Both HT-RFDA and HT-RuL measurements can monitor a shift of the martensitic
transformation temperature of zirconia in Ca-PSZ during the cooling phase and at a lesser
extent during the heating phase. It can be concluded that in the only presence of thermal
stresses (case of HT-RFDA), the hysteretic behaviour in terms of stiffness is shifted by
2,14°C per cycle, while in presence of further mechanical stresses (case of HT-RuL), the
shift of the hysteretic behaviour in terms of thermal expansion is more considerable
(increase by around 37%) and reaches 2,93°C per cycle. This shift of the martensitic
transformation temperature can be explained by the wide particle size range of zirconia
raw material. As explained by Becher et al., the tetragonal-to-monoclinic formation
temperature decreases with decrease in tetragonal grain size [BEC92]. After each cycle,
successive sub-range of zirconia particles generates cracks in the microstructure
monitored as stiffness decrease (HT-RFDA) and thermal contraction (HT-RuL). Both
409
Conclusions
HT-RFDA and HT-RuL measurements monitor the start martensitic temperature but
cannot monitor the end martensitic temperature as this one is below 900°C.
During High Temperature Thermal Shocks (HTTS) experiments, similar damage
mechanisms are observed as those during cycling HT-RFDA measurements. The more
considerable heating and cooling rate between two elevated temperatures accelerate the
damage kinetics. Therefore, after three cycles, few cracks mainly localized at the grain
boundary of large tabular alumina aggregates and a widening of these cracks occurs in
Ref and Y-FSZ, while a complete debonding of large tabular alumina aggregates is
observed in Ca-PSZ and a dense crack network mainly propagating orthogonally to the
zirconia particles is noticed in Mg-PSZ due to the induced compressive stresses at the
grain boundary of transformed zirconia particles during the cooling phase.
To conclude, the crack path throughout the matrix of a high alumina refractory castable
containing zirconia mainly depends on the stabilized phase, the closed porosity of the
zirconia particle and on the interface between the zirconia particle and the matrix. By
hitting zirconia particles, three cases can be observed: either the crack is deviated by the
ensuing volume expansion of the stabilized zirconia particle transforming into
monoclinic resulting in toughening mechanisms [CAR07] through zone shielding
[RIT88], or the crack is propagating along the grain boundary of zirconia particles,
wherein the local concentration in calcium aluminate phases is abnormally relative low,
or the crack is propagating throughout the pore network of the zirconia particle.
Furthermore, the weak interfaces between tabular alumina particles and the matrix are
already noticeable in formulation Ref. The difference of thermal expansion between the
tabular alumina particles and the calcium aluminate phases is sufficient to generate
cracks that can easily propagate at the grain boundary of tabular alumina aggregates
exhibiting low connection with the matrix [BEA04].
If the retained monoclinic content in the stabilized zirconia raw material is the main
factor influencing the thermal shock resistance of the high alumina refractory castable,
the nature of the stabilized phase also has an impact on the resulting microstructure after
thermal shock.
First, during heating thermal shock, radial compressive stresses are generated at the grain
boundary of the zirconia particle independently of the nature of the phase leading to the
formation of cracks propagating orthogonally to the zirconia particle.
During cooling thermal shocks, the nature of the zirconia phase has much more influence
on the resulting microstructure. First in case of the monoclinic phase (Mg-PSZ), a severe
radial compressive stress field is generated at the grain boundary of the zirconia particle
during the martensitic transformation leading to a dense crack network propagating
orthogonally to the surface of the zirconia particle. In case of the tetragonal phase
410
Conclusions
(Ca-PSZ), toughening mechanism can occur when a crack is hitting the surface of a
stabilized zirconia particle capable to transform into monoclinic with ensuing volume
expansion that could result in a phase transformation induced crack closure [RIT88] or in
a decrease of the crack propagation kinetics. And finally, in case of the cubic phase
(Y-FSZ), cracks are propagating at the grain boundary of the zirconia particles as radial
tensile stresses are generated at the grain boundary.
If in the present study, damping and Young’s modulus evolution are discussed according
to the ensuing damage observed in the sample microstructure, it is challenging to
correlate the elastic properties with damage quantification: Indeed, SEM micrographs
first reveal microcracking in sample microstructure at two-dimensional scale. It is
erroneous to consider that the crack density shown on a SEM micrograph is
representative of the volume damage of the sample. In the present study, it has been
shown that thermal shocks in air induce a crack density gradient within the sample
thickness. Finally, techniques like soaking the sample in colorant to reveal cracks within
the sample matrix have also the drawback to reveal the apparent porosity rendering the
distinction of cracks as well as their quantification challenging. For all these reasons, a
correlation between crack density and damping evolution is questionable.
411
Outlook
7. Outlook
Two major research axes were proposed in the study hereby: a first part was dedicated to
the measuring principle of RFDA and a second part to the thermal shock resistance
enhancement of high alumina castables through the addition of stabilized zirconia. In
both research axes, further work could be performed.
With regard to the Resonant Frequency Damping Analysis, this technique could reveal
damage mechanisms at room and elevated temperature as well as material characteristics
such as the temperature of transformation of zirconia. It has been shown during fatigue
tests that the martensitic transformation temperature is shifted to higher temperature after
progressive thermal cycles. Further experiments could be performed in order to examine
the effect of the heating and cooling rates on this temperature shift. On one side, the
thermal stresses could be increased by raising both heating and cooling rates and on the
other side, this increase of the thermal stresses could be correlated with the increase of
thermo-mechanical stresses during HT-RuL tests, wherein the load is increased from
0,20 MPa to 0,50 MPa. Furthermore, cycling DTA-TG measurements could be also
performed on raw Ca-ZrO2 material to assess if the shift of the martensitic transformation
temperature is also visible while the only zirconia raw material is subjected to thermal
cycling and not the zirconia material contained in the matrix of the high alumina castable.
If the assessment of the Young’s modulus through dynamic and static measurement
methods results in different results, it could however be of interest to study the elastic
behaviour of the tested high alumina castables at elevated temperature through HMOR
experiments in order to compare the elastic behaviour at elevated temperature obtained
through RFDA measurements and through HMOR measurements. Such HMOR tests
performed at regular intervals at elevated temperature could furthermore enable to study
the evolution of the stored elastic energy with temperature in order to determine
temperature ranges, wherein the stored elastic energy is reduced or critical. The
combination of both HT-RFDA and HMOR experiments could therefore provide
information related to the part of the stored elastic energy released by friction and by
strain development respectively.
The present study mainly focused on the thermo-mechanical and thermo-elastic
properties of high alumina castables. However, such materials are furthermore subjected
to chemical stresses during application. Therefore, HT-RFDA measurements could be
performed on castables, wherein a slag material is infiltrating the castable during the
experience in order to examine the influence of the corrosion kinetics on the stiffness
evolution of the castable.
412
Outlook
In terms of material development, some results have shown the influence of the shape of
the aggregates on the degradation of the mechanical properties after thermal shocks. A
more elongated grain shape seems to have beneficial effects on the enhancement of the
retained mechanical properties. Therefore, the thermal shock resistance could be assessed
on high alumina castables based on the same stabilized zirconia characterized by
different grain shape in order to examine the influence of the interlocking structure on the
mechanical properties improvement. This work could therefore be focussed on the
relative flexible deformation of the castable improving the release of the stored elastic
energy.
Grain shape of zirconia aggregates has an expected technical effect on the enhancement
of the creep behaviour under bending conditions for instance. The influence of the length
of lenticular zirconia aggregates in the castable formulation can be therefore studied by
the examination of failure map according to the temperature, the load and the time until
failure under such conditions of temperature and load [MUN99]. Thus, the shape of the
zirconia aggregates can be optimized in order to achieve improved creep behaviour
without imparting detrimental consequences on other thermo-mechanical properties of
the castable.
If the addition of partially stabilized zirconia as aggregates influences the propagation of
the crack path within the microstructure of the castables, it could be furthermore of
interest to examine the influence of zirconia incorporated in the castable formulation
either as filler or as binding agent as for instance using a calcium zirconate cement or a
combination of calcium zirconate and calcium aluminate cement [KAN14]. Besides
413
Outlook
obtaining a macro-scale deviation of the cracks by the zirconia aggregates, those cracks
could furthermore be deviated by small zirconia particles contained in the matrix. This
would induce toughening mechanisms at macro and micro-scale resulting in the
enhancement of the fracture toughness of the material.
Besides having an influence on the stabilized zirconia phase, the doping agent also has an
impact on the viscosity and the temperature of formation of the viscous phases formed at
elevated temperature at the grain boundary of the zirconia particle influencing therefore
the creep behaviour as well as the thermo-elastic properties of the castable. If CaO as
doping agent enables the formation of tetragonal phases that can transform into
monoclinic within the framework of toughening mechanisms, Y2O3 can also have
beneficial effect in terms of crack healing as the viscosity of the formed viscous phase is
reduced resulting in the infiltration of those viscous phases in the microcrack network
surrounding the zirconia particles. A new stabilized zirconia raw material could be
therefore developed in order to exhibit both properties by studying the doping of zirconia
with both CaO and Y2O3 or by studying the effect of other stabilizer such as CeO on the
creep behaviour.
Furthermore, a more accurate qualitative measurement of the monoclinic, tetragonal and
cubic phases after progressive thermal shocks could be realized through Raman
spectroscopy. Such an analysis could quantify qualitatively the proportion of stabilized
tetragonal zirconia particles that transform into monoclinic within the framework of
toughening mechanisms. This could help to develop stabilized zirconia particles prone to
contribute to the enhancement of the material fracture toughness [ZAN16].
Cycling HT-RFDA and RuL measurements could also be performed on ideal model
materials based on alumina and zirconia with defined content of zirconia and known
tetragonal content in zirconia raw material. This would allow to study the impact and the
evolution of successive martensitic transformation of zirconia on material elastic and
mechanical properties, wherein the fundamental material properties in terms of density,
stiffness, thermal expansion are known. The obtained results could be used to understand
the elastic and mechanical behaviour of porous castables.
414
References
8. References
415
References
[AUT13]: Autier, C., Azema, N., Taulemesse, J.M., Clerc, L.: “Mesostructure
evolution of cement pastes with addition of superplasticizers
highlighted by dispersion indices”, Powder technology, Vol. 249,
2013, pp. 282-289
[BAB11]: Babelot, C., Guignard, A., Huger, M., Gault, C., Chotard, T., Ota, T.,
Adachi, N.: “Preparation and thermomechanical characterization of
aluminium titanate flexible ceramics”, Journal of Materials Science,
Vol. 46, 2011, pp. 1211-1219
[BAK94]: Bakunov, V. S.: “High-Temperature Creep of Refractory Ceramics”,
Refractories, Vol. 35, No.11, 1994, pp. 353-359
[BAR12]: Bargel (Hrsg.), H.-J., Schulze, G.: “Werkstoffkunde”, 11. Auflage,
Springer-Verlag Berlin Heidelberg, 2012
[BAS99]: Basu, B., Donzel, L., Van Humbeeck, J., Vleugels, J., Schaller, R.,
Van Der Biest, O.: “Thermal expansion and damping characteristics of
Y-TZP”, Scripta Materialia, Vol. 40, Issue 7, 1999, pp. 759-765
416
References
417
References
[CHA15]: Djangang, C.N., Tealdi, C., Cattaneo, A.S., Mustarelli, P., Kamseu, E.,
Leonelli, C.: “Cold setting refractory composites from cordierite and
mullite-cordierite design with geopolymer paste as binder: Thermal
behavior and phase evolution”, Materials Chemistry and Physics,
Volume 154, 2015, pp. 66-77
[COP73]: Coppola, J. A., Bradt, R. C.: “Thermal Shock Damage in SiC”,
Journal of the American Ceramic Society, Vol. 56, No. 4, 1973, pp.
214-218
[CRI88]: Criado, E., Pena, P., Caballero, A.: “Influence of processing method
on microstructural and mechanical properties of calcium hexaluminate
compacts”, Science of Ceramics, Vol. 14, 1988, pp. 193-198
[CUT96]: Cutard, T., Viatte, T., Feusier, G., Benoit, W.: “Microstructure and
high temperature mechanical properties of TiC0,7N0,3-Mo2C-Ni
cermets”, Materials Science and Engineering A209, 1996, pp. 218-227
[DAS12]: Das, S., Mitra, N.K., Das, S.: “Sintered properties and sintering
behavior of MgO-ZrO2 composite hydrogel prepared by
coprecipitation technique”, Science of Sintering, Vol. 44, Issue 1,
2012, pp. 33-45
[DIN92]: Dinger, D. R., Funk, J. E.: “Particle packing. III: Discrete versus
continuous particle sizes”, Interceram, Vol. 41, Issue 5, 1992, pp. 332-
334
[DIN CEN/TS Prüfverfahren für dichte feuerfeste Erzeugnisse – Leitlinien zur
15418]: Prüfung von durch Flüssigkeiten hervorgerufene Korrosion an
feuerfesten Erzeugnissen, 2006
418
References
419
References
420
References
[GRO11]: Gross, D., Seelig, Th.: “Bruchmechanik – Mit einer Einführung in die
Mikromechanik”, 5. Auflage, Springer-Verlag Berlin Heidelberg,
2011
[GRUE59]: Grüneisen, E.: “The state of a solid body”, traduction from National
Aeronautics Space Administration, 1959, pp. 1-52
[GUA15]: Guan, Y., Li, C., Lu, Y., Sun, L., Tang, J., Wang, C., Zhao, Y.:
“Refractory Castable with High Thermal Shock Resistance”, Patent,
Publication date: 18.11.2015
[GUI98]: Guirado, F., Gali, S., Chinchon, J.S.: “Thermal Decomposition of
Hydrated Alumina Cement CAH10”, Cement and Concrete Research,
Vol. 28, No. 3, 1998, pp. 381- 390
[HAN00]: Hannink, R.H.J., Kelly, P.M., Muddle, B.C.: “Transformation
Toughening in Zirconia Containing Ceramics”, Journal of the
American Ceramic Society, Vol. 83, No. 3, 2000, pp. 461-487
[HAR14]: Harten, U.: “Physik – Eine Einführung für Ingenieure und
Naturwissenschaftler”, 6. Auflage, Springer-Verlag Berlin Heidelberg,
2014
[HAR95]: Harmuth, H.: “Stability of crack propagation associated with fracture
energy determined by wedge splitting specimen”, Theoretical and
Applied Fracture Mechanics, Volume 23, Issue 1, 1995, pp. 103-108
[HAR96]: Harmuth, H., Rieder, K., Krobath, M., Tschegg, E.: “Investigation of
the non-linear fracture behaviour of ordinary ceramic refractory
materials”, Materials Science and Engineering, Vol. 214, Issues 1-2,
1996, pp. 53-61
[HAR97]: Harmuth, H., Tschegg, E. K.: “A fracture mechanics approach for the
development of refractory materials with reduced brittleness”, Fatigue
& Fracture of Engineering Materials & Structures, Vol. 20, Issue 11,
1997, pp. 1585-1603
[HAS63]: Hashin, Z., Shtrikman, S.: “A variational approach to the theory of the
elastic behaviour of multiphase materials”, Journal of the Mechanics
and Physics of Solids, Vol. 11, 1963, pp. 127-140
[HAS69]: Hasselman, D. P. H.: “Unified Theory of Thermal Shock Fracture
Initiation and Crack Propagation in Brittle Ceramics”, Journal of The
American Ceramic Society, Vol. 52, No. 11, pp. 600-604, 1969
[HAS70]: Hasselman, D. P. H.: “Strength Behavior of Polycrystalline Alumina
Subjected to Thermal Shock”, Journal of The American Ceramic
Society, Vol. 53, No. 9, 1970, pp. 490-495
[HEI10]: Heimann, R. B.: “Classic and Advanced Ceramics: From
421
References
422
References
pp. 826-831
[KAK08]: Kakroudi, M. G., Yeugo-Fogaing, E., Gault, C., Huger, M., Chotard,
T.: “Effect of Thermal Treatment on Damage Mechanical Behaviour
of refractory Castables: Comparison between Bauxite and Andalusite
Aggregates”, Journal of the European Ceramic Society, Vol. 28, Issue
13, 2008, pp. 2471-2478
[KAK09]: Kakroudi, M. G., Huger, M., Gault, C., Chotard, T.: “Anisotropic
behavior of andalusite particles used as aggregates on refractory
castables”, Journal of the European Ceramic Society, Vol. 29, 2009,
pp. 571-579
[KAM97]: Kamitani, K., Grimsditch, M., Nipko, J. C., Loong, C., Okada, M.,
Kimura, I.: “The elastic constants of silicon carbide: A Brillouin-
scattering of 4H and 6H SiC single crystals”, Journal of Applied
Physics, Vol. 82, Issue 6, 1997, pp. 3152-3154
[KAN14]: Kang, E.-H., Yoo, J.-S., Kim, B.-H., Choi, S.-W., Hong, S.-H.:
“Synthesis and hydration behavior of calcium zirconium aluminate
(Ca7ZrAl6O18) cement”, Cement and Concrete Research, Vol. 56,
2014, pp. 106-111
[KAT92]: Katchanov, M.: “Effective elastic properties of cracked solid”, Applied
Mechanics Reviews, 1992, pp. 70-76
[KER06]: R. PDS-US-S71-8/06, “Product Data Sheet Secar71”, KERNEOS,
2006
[KHO11]: Khoeini, M., Rastegar, H., Hafizpour, H.R.: “Preparation of Layer
Nano-Silicate/Alumina Castable Composites”, Journal of American
Science, Vol. 7, Issue 6, 2011, pp. 630-634
[KHA98]: Khan, M. S., Islam, M. S., Bates, D. R.: “Cation doping and oxygen
diffusion in zirconia: a combined atomistic simulation and molecular
dynamics study”, Journal of Materials Chemistry, Vol. 8, Issue 10,
1998, pp. 2299-2307
[KIN55]: Kingery, W. D.: “Factors Affecting Thermal Stress Resistance of
Ceramic Materials”, Journal of the American Ceramic Society, Vol.
38, No. 1, pp. 3-15, 1955
[KOL04]: Kollenberg, W.: “Technische Keramik - Grundlagen, Werkstoffe,
Verfahrenstechnik”, Essen: Vulkan-Verlag, 2004
[KOV15]: Kovarik, O., Hausild, P., Capek, J., Medricky, J., Siegl, J., Musalek,
R., Pala, Z., Curry, N., Bjorklund, S.: “Damping measurement during
resonance fatigue test and its application for crack detection in TBC
samples”, International Journal of Fatigue, Vol. 82, No. 2, 2015, pp.
300-309
423
References
424
References
[LIU07]: Liu, X., Büchel, G., Buhr, A.: “Review of Tabular Alumina as High
Performance Refractory Material”, Interceram Refractories Manual,
2007, pp. 6-12
[LIU09]: Liu, B., Feng, X., Zhang, S.-M.: “The effective Young’s modulus of
composites beyond the Voigt estimation due to the Poisson effect”,
Composites Science and Technology, Vol. 69, 2009, pp. 2198-2204
[LIU12]: Liu, J., Shan, G., Zhu, G., Zhu, J., Zhu, Q.: “Casting Material
Comprising Andalusite, Mullite, Silicon Carbide, Hollow Ball Mullite,
Pure Calcium Aluminate Cement, Alumina Micro Powder, Zirconium
Silicate, Melamine and Polycarboxilic Acid”, Patent, Priority date:
25.10.2011, Publication Date: 20.06.2012
[LTI11]: Ltifi, M., Guefrech, A., Mounanga, P.: “Effects of sodium
tripolyphosphate addition on early-age physico-chemical properties of
cement pastes”, Procedia Engineering, Vol. 10, 2011, pp. 1457-1462
[LUZ15]: Luz, A.P., Gomes, D.T., Pandolfelli, V.C.: “High-alumina phosphate-
bonded refractory castables: Al(OH)3 sources and their effects”,
Ceramics International, Vol. 41, Issue 7, 2015, pp. 9041-9050
[MAN09]: Manhart, C., Harmuth, H.; “Resonant Frequency and Damping
Analysis of Refractories with and without reduced Brittleness”, in
Unified International Technical Conference on Refractories
UNITECR, Salvador, Brasilien, 2009
[MAX73]: Maxwell, J. C.: “A treatise on electricity and magnetism”, Oxford,
UK, Clarendon Press, 1873, p. 365
[MCC94]: McColm, I. J.: “Dictionary of Ceramic Science and Engineering”,
New York: Plenum Press, 1994
[MIL81]: Miller, R.A., Smialek, J.L., Garlick, R.G.: “Phase stability in Plasma-
Sprayed, Partially Stabilized Zirconia-Yttria, Advances in Ceramics,
Vol. 3, Science and Technology of Zirconia”, Edited by Heuer A.H.,
Hobbs, L.W., American Ceramic Society, Columbus, OH, 1981, pp.
241-253
[MIY11]: Miyaji, D.Y., Pereira, A.H.A., Rodrigues, J.de A., “Thermal Shock on
High Alumina Castables Investigated by the Measurements of
Young’s Modulus as a Function of Temperature”, Proceeding 54th
International Colloquium on Refractories, Aachen, 2011, pp. 38- 39
[MIY14]: Miyaji, D.Y., Otofuji, C.Z., Rodrigues, J. de A.: “The load-
displacement curve of steady crack propagation: an interesting source
of information for predicting the thermal shock damage of
refractories”, Unitecr 2013 Proceeding, 2013, pp. 811-816
425
References
426
References
427
References
428
References
429
References
[SAK09]: Sako, E.Y., Braulio, M.A.L., Milanez, D.H., Brant, P.O., Pandolfelli,
V.C.: “Microsilica role in the CA6 formation in cement-bonded
refractory castables”, Journal of Materials Processing Technology,
Vol. 209, 2009, pp. 5552-5557
[SAK85]: Sakuma, T., Yoshizawa, Y. I., Suto, H.: “The microstructure and
mechanical properties of yttria-stabilized zirconia prepared by arc
melting”, Journal of Materials Science, Vol. 20, 1985, pp. 2399-2407
[SAL07]: Salmang, H., Scholze, H.: “Keramik”, 7. Auflage, Telle, R. (Hrsg),
Springer-Verlag Berlin Heidelberg, 2007
[SAL18]: Salmang, H., Scholze, H.: “Keramik”, 8. Auflage, Telle, R. (Hrsg),
Springer-Verlag Berlin Heidelberg, 2018
[SAL12]: Salvini, V. R., Pandolfelli, V. C., Bradt, R. C.: “Extension of
Hasselman’s thermal shock theory for crack/microstructure
interactions in refractories”, Ceramics International, 2012, Vol. 38,
pp. 5369-5375
[SAN00]: Souza Santos, P., Souza Santos, H., Toledo, S.P.: “Standard transition
aluminas. Electron microscopy studies”, Materials Research, Vol. 3,
No. 4, 2000, pp. 104-114
[SAR01]: Sarpoolaky, H., Zhang, S., Argent, B. B., Lee, W .E.: “Influence of
Grain Phase on Slag Corrosion of Low-Cement Castable
Refractories”, Journal of the American Ceramic Society, Vol. 84, Issue
2, pp. 426-434
[SCH00a]: Schaller, R.: “Mechanical spectroscopy of the high-temperature brittle-
to-ductile transition in ceramics and cermets”, Journal of Alloys and
Compounds, Vol. 310, 2000, pp. 7-15
[SCH00b]: Schmitt, N., Hernandez J.F., Lamour, V., Berthaud, Y., Meunier, P.,
Poirier, J.: “Coupling between kinetics of dehydration, physical and
mechanical behaviour for high alumina castable”, Cement and
Concrete Research, Vol. 30, 2000, pp. 1597-1607
[SCH04]: Schacht (Hrsg.), C. A.: “Refractories Handbook”, Marcel Dekker, Inc.
New York, Basel, 2004
[SCH87]: Schmid, H.K.: “Quantitative Analysis of Polymorphic Mixes of
Zirconia by X- Ray Diffraction”, Journal of the American Ceramic
Society, Vol. 70, No. 5, 1987, pp. 367-376
[SER04] Serena, S., Sainz, M.A., de Aza, S., Caballero, A.: “Thermodynamic
assessment of the system ZrO2-CaO-MgO using new experimental
results Calculation of the isoplethal section MgO.CaO-ZrO2”, Journal
of the European Ceramic Society, Vol. 25, 2005, pp. 681-693
430
References
431
References
432
References
[WEG13]: Węglewski, W., Bochenek, K., Basista, M., Schubert, Th., Jehring, U.,
Litniewski, J., Mackiewicz, S.: “Comparative assessment of Young’s
modulus measurements of metal-ceramic composites using mechanical
and non-destructive tests and micro-CT based computational
modeling”, Computational Materials Science, Vol. 77, 2013, pp. 19-
30
[WEL04]: Weller, M., Haneczog, G., Kestler, H., Clemens, H.: “Internal friction
of γ-TiAl-based alloys with different microstructures”, Materials
Science and Engineering A, Vol. 370, 2004, pp. 234-239
[WEL93]: Weller, M., Schubert, H., Kountouros, P.: “Mechanical and dielectric
loss measurements in Y2O3–ZrO2 and TiO2–Y2O3–ZrO2 ceramics”,
Science and Technology of Zirconia V, ed. Badwal, S. P. S., Bannister,
M. J. and Hannink, R. H. J. Technomic Publ. Co., Lancaster and
Basel, 1993, pp. 546–554
[WÖH09]: Wöhrmeyer, C., Fryda, H., Parr, C., Auvray, J.-M., Guillaumin, V.:
“Mineralogy and microstructure evolution along the curing, drying and
firing process of calcium aluminate bonded refractory castables”,
Proceeding 52nd International Colloquium on Refractories, Aachen,
2009, pp. 30-34
[YEP07]: Yeprem, H.A.: “Effect of Iron Oxide Addition on the Hydration
Resistance and Bulk Density of Doloma”, Journal of the European
Ceramic Society, Vol. 27, Issues 2-3, 2007, pp. 1651-1655
[YOU08]: Yousef, S. G., Rödel, J., Fuller Jr., E.R., Zimmermann, A., El-Dasher,
B.S., “Microcrack Evolution in Alumina Ceramics: Experiment and
Simulation”, Journal of the American Ceramic Society, Vol. 88, Issue
10, 2005, pp. 2809-2816
[ZAN16]: Zanocco, M.; “Raman spectroscopic analysis of zirconia toughened
alumina ceramic (ZTA) in presence of different metal stains and ZTA
retrieval femoral heads”, Master’s degree thesis, 2016
[ZEN41]: Zener, C.: “Theory of the Elasticity of Polycrystals with Viscous Grain
Boundaries”, Physical Review, Vol. 60, 1941, pp. 906-908
[ZIV09]: Zivcová, Z., Cerny, M., Pabst, W., Gregorová, E.: “Elastic properties
of porous ceramics prepared using starch as a pore forming agent”,
Journal of the European Ceramic Society, Vol. 29, 2009, pp. 2765-
2771
[ZumG86]: Zum Gahr, K.H. (Hrsg.): “Reibung und Verschleiß bei metallischen
und nichtmetallischen Werkstoffen”, DGM Informationsgesellschaft
Verlag, Oberursel, 1986, p. 383
433
References
434
Resume
Name: Traon
Surname: Nicolas
Professional Since 2016: Patent examiner at the European Patent Office, The
activity: Hague - Netherlands