Complete_MSc_thesis_Mark_Meijerink.

Complete_MSc_thesis_Mark_Meijerink.
Mark Johan Meijerink
Coating of MoSi2 healing particles for
self-healing thermal barrier coatings
i
ii
Coating of MoSi2 healing particles for
self-healing thermal barrier coatings
By
Mark Johan Meijerink
in partial fulfilment of the requirements for the degree of
Master of Science
in Chemical Engineering
and Materials Science and Engineering
at the Delft University of Technology,
to be defended publicly on Friday October 9, 2015 at 16:00.
Supervisor:
Thesis committee:
Dr. ir. W.G. Sloof
TU Delft
Prof. dr. Ir. S van der Zwaag,
Dr. ir. W.G. Sloof,
Dr. ir. J.R. van Ommen,
Dr. E.M. Kelder,
TU Delft
TU Delft
TU Delft
TU Delft
This thesis is confidential and cannot be made public until December 31, 2015.
An electronic version of this thesis is available at http://repository.tudelft.nl/.
iii
Abstract
To increase lifetime of the protective thermal barrier coatings (TBC) in jet engines and other gas
turbines, a self-healing approach based on MoSi2 healing particle addition is considered. However,
due to rapid oxygen transport in yttria-stabilized zirconia (YSZ), a common TBC material, premature
oxidation is a major problem. This thesis investigates the feasibility of coating MoSi2 sacrificial
particles with a protective Al2O3 shell to prevent this oxidation, while still retaining particle
availability upon damage. Two different chemical methods, namely a sol-gel procedure and atomic
layer deposition with residual chemical vapor deposition were successfully utilized to coat MoSi2
healing particles.
The microcapsule composition and integrity has been investigated by means of scanning electron
microscopy coupled with energy dispersive x-ray spectroscopy, x-ray diffraction and x-ray
photoelectron spectroscopy. The results demonstrate that after calcining at 1200 °C for 1h in argon
α-Al2O3 shell can be formed and the shell remains intact. Subsequently the heat treated
encapsulated particles were embedded in YSZ matrix followed by healing tests at 1100 and 1200 °C.
The crack-healing tests proved that the shells produced by both methods remain intact at high
temperatures, but also the coatings have a protective effect compared to uncoated MoSi2.
Moreover, the embedded particles show a crack healing effect, indicating the feasibility of this selfhealing concept.
Samenvatting
Om de levensduur van thermal barrier coatings (TBCs) in gasturbines en vliegtuigmotoren te
verlengen, wordt een zelfherstellende coating overwogen, gebaseerd op de toevoeging van MoSi2
deeltjes aan deze coating. Een uitdaging in dit systeem is echter de zeer snelle oxidatie van deze
deeltjes door de hoge snelheid van zuurstoftransport in yttria stabilized zirconia (YSZ). In dit werk
wordt daarom de haalbaarheid van het aanbrengen van een Al2O3 beschermlaag op de MoSi2
deeltjes onderzocht, die het zelfherstellende mechanisme niet blokkeren. Hiervoor zijn twee
verschillende chemische methoden voor het aanbrengen van deze coating vergeleken, namelijk solgel en Atomic Layer Deposition met residual Chemical Vapor Deposition (ALD/rCVD).
De eigenschappen en microstructuur van de met beide methoden succesvol geproduceerde
microcapsules zijn geanalyseerd met behulp van scanning electron microscopy (SEM) gecombineerd
met energy dispersive x-ray spectroscopy (EDS), x-ray diffraction (XRD) en x-ray photoelectron
spectroscopy (XPS). Deze resultaten geven duidelijk aan dat het mogelijk is om na calcineren in argon
op 1200 ᵒC gedurende 1 uur, een beschermlaag van α-alumina gevormd kan worden en dat deze laag
intact blijft. Deze microcapsules zijn daarna ingebed in YSZ, gevolgd door hersteltesten op 1100 en
1200 ᵒC. Deze testen lieten zien dat de capsules inderdaad in staat zijn de deeltjes te beschermen,
vergeleken met niet beschermde deeltjes en intact blijven op hoge temperaturen. Ook het composiet
blijk van enig zelfherstellend vermogen, wat aangeeft dat dit inderdaad een interessant concept is.
iv
List of Tables
Table 2.1: Important thermal properties, namely melting temperature, coefficient of thermal
expansion and thermal conductivity of the main materials tabulated based on data from the
(Japanese) National Institute of Materials Science (NIMS). .................................................................. 10
Table 3.1: Standard free energy of formation at 1000 ᵒC for each oxide present in the system from its
element, per mole of oxygen consumed, along with the equilibrium partial oxygen pressure. ........... 25
Table 4.1: Overview of conditions for each aluminium oxalate sample................................................ 37
Table 4.2: Overview of ALD sample conditions. .................................................................................... 38
Table 5.1: PSD percentiles and Sauter particle diameter for the measured samples calculated from
laser diffraction data. ............................................................................................................................ 46
Table 5.2: The calculated BET specific surface areas based on isotherm data for each sample. .......... 47
Table 5.3: EDS elemental concentration measurements in atom% of the points shown in Figure 5.9 b.
............................................................................................................................................................... 49
Table 5.4: Combined EDS measurements for each sample and the number of measurements with
significant Al detected. .......................................................................................................................... 51
Table 5.5: Combined EDS results for the aluminium tri-sec-butoxide samples with average atom% Al
detected and the amount of measurements that found less than 0.9 atom% Al. ................................ 54
Table 5.6: EDS elemental concentration measurements in atom% of the points shown in Figure 5.18 b.
............................................................................................................................................................... 60
Table 5.8: Measured hardness and crack length from Vickers HV10 indentation and resulting fracture
toughness for an SPS sample of YSZ and the YSZ-MoSi2B composite. .................................................. 75
List of Figures
Figure 2.1: Jet engine layout (a) and interface between turbine blade and hot gas (b). (Padture, Gell et
al. 2002), (Clarke and Phillpot 2005) ....................................................................................................... 3
Figure 2.2: Causes of spallation in TBCs illustrated with (a) showing the source of compressive stresses
and (b) the coalescence of microcracks and spallation. (Turteltaub 2013)............................................. 4
Figure 2.3: The self-healing thermal barrier coating system with on the left the whole turbine blade
coating system, zooming in on the particles on the right. Upper part is before healing and the lower
part after healing. (Sloof 2014) ............................................................................................................... 5
Figure 2.4: The corundum crystal structure with (a) showing the regular structure with both Al3+ and
O2- (Askeland and Phulé 2003) and (b) showing the locations of the empty alumina sites in the
structure (Chiang, Kingery et al. 1997). ................................................................................................... 7
Figure 2.5: The temperature dependent phase diagram of molybdenum and Silicon, along a three
component phase diagram of Mo, Si and O at 1200 ᵒC. (Fujiwara and Ueda 2007) .............................. 8
Figure 2.6: The crystal structure of the thermodynamically stable tetragonal structure of MoSi2.
(d’Heurle, Petersson et al. 1980) ............................................................................................................. 8
Figure 2.7: The ZrO2-Y2O3 phase diagram in the ZrO2 rich region, showing the different phases of
zirconia and their stability depending on temperature and yttria content. (Subbarao and Gokhale
1968)........................................................................................................................................................ 9
Figure 2.8: The crystal structure of cubic YSZ with Y substituting randomly for Zr (Singhal and Kendall
2003)........................................................................................................................................................ 9
Figure 2.9: The temperature-pressure phase diagram of silica (Koike, Noguchi et al. 2013). ................ 9
v
Figure 2.10: Crystal structure of trigonal α quartz (Lager, Jorgensen et al. 1982).................................. 9
Figure 2.11: Crystal structure of zirconium silicate (Mao 2013). .......................................................... 10
Figure 2.12: A schematic representation of the sol-gel process, with multiple possible microstructures
depending on processing route (Brinker and Scherer 2013). ................................................................ 11
Figure 2.13: The fraction of alumina species present in an aqueous solution as a function of pH at 25
ᵒC (Wang and Muhammed 1999).......................................................................................................... 14
Figure 2.14: Zeta potential for SiC in water and alumina sol as function of pH (Yang and Troczynski
1999)...................................................................................................................................................... 15
Figure 2.15: A schematical overview of one cycle in the ALD process (Kim). ........................................ 16
Figure 2.16: Density, refractive index and growth rate of Al2O3 coatings on PET as function of
temperature (Groner, Fabreguette et al. 2004). ................................................................................... 18
Figure 3.1: The temperature dependent Mo-B phase diagram (Liao). .................................................. 22
Figure 3.2: The temperature dependent Al2O3-ZrO2 phase diagram (Lakiza and Lopato 1997). .......... 23
Figure 3.3: The temperature dependent Al2O3-Y2O3 phase diagram (Fabrichnaya, Seifert et al. 2001).
............................................................................................................................................................... 23
Figure 3.4: The temperature dependent Al2O3-SiO2 phase diagram (Degterov and Pelton 1996). ....... 24
Figure 3.5: The temperature dependent SiO2-ZrO2 phase diagram (Butterman 1967) ......................... 25
Figure 3.6 The calculated temperature dependent Y2O3-SiO2 phase diagram (RouNSow, Grsns et al.
1971)...................................................................................................................................................... 25
Figure 3.7: A schematic representation of the evolution of the coated MoSi2 system at high
temperatures in an oxygen-rich environment. In this system, it is assumed that yttria and zirconia are
not able to diffuse through alumina and that molybdenum will not oxidize. ...................................... 26
Figure 3.8: Example of the thickness of each layer after 24 hours at 1000 ᵒC. .................................... 33
Figure 3.9: The influence of temperature on total oxidation (equivalent silica thickness) of the system
after 24 hours for 900 ᵒC to 1200 ᵒC with 25 ᵒC increments with the highest temperature having the
highest oxidation rate. .......................................................................................................................... 33
Figure 3.10: The effect of partial oxygen pressure on total oxidation after 24 hours at 1000 ᵒC, with a
partial pressure varied from 10-14 to 1 bar in power of 10 increments with the highest partial oxygen
pressure having the highest oxidation rate........................................................................................... 33
Figure 3.11: The influence of alumina and mullite grain size on total oxidation of the system after 24
hours at 1000 ᵒC with grain size varied from 50 to 500 nm with 50 nm increments with the smallest
grain size having the highest oxidation rate. ........................................................................................ 33
Figure 3.12: The influence of the initial alumina layer coating thickness, with coating thickness
ranging from 10 to 1000 nanometers and the highest thickness having the lowest oxidation rate. ... 34
Figure 3.13: The influence of the initial mullite layer coating thickness, with coating thickness ranging
from 5 to 50 nanometers and the highest initial thickness having the lowest oxidation rate. ............ 34
Figure 3.14: The influence of the initial SiO2 layer coating thickness, with coating thickness ranging
from 5 to 50 nanometers and the highest initial thickness having the lowest oxidation rate. ............ 34
Figure 3.15: Example of the thickness of each layer after one year at 1000 ᵒC.................................... 34
Figure 4.1: The Alpine 100 MRZ laboratory zig-zag classifier used for wind sifting and its different
parts....................................................................................................................................................... 35
Figure 4.2: The molecular structures of (a) aluminium oxalate, (b) aluminium tri-isopropoxide and (c)
aluminium tri-sec-butoxide, as provided by Sigma-Aldrich. ................................................................. 36
Figure 4.3: The setup used for sol-gel experiments with heating, stirring and nitrogen supply. .......... 36
vi
Figure 4.4: The ALD setup used in this experiment, showing the whole setup including bubblers and
gas cleaning (left) and the reactor with connections (right). ................................................................ 38
Figure 5.1: Laser diffraction results for MoSi2 and MoSi2B with the volume particle size distribution
(left) and cumulative volume distribution (right). ................................................................................. 42
Figure 5.2: Scanning Electron Microscope (SEM) images of untreated MoSi2 powder at different
magnifications. ...................................................................................................................................... 43
Figure 5.3: X-ray diffractograms of MoSi2 and MoSi2B with phases present. ....................................... 43
Figure 5.4: N2 physisorption adsorption/desorption curve of MoSi2 base material, with reference
pressure P0 = 0.1 MPa............................................................................................................................ 44
Figure 5.5: Laser diffraction PSD results for the coarse fractions after wind sifting compared to the
material before wind sifting. ................................................................................................................. 45
Figure 5.6: Laser diffraction PSD results for the fine fraction of MoSi2 batch 1 after wind sifting
compared to the material before wind sifting. ..................................................................................... 45
Figure 5.7: Nitrogen physisorption isotherms of the two wind sifted MoSi2 coarse batches and the
starting material.................................................................................................................................... 46
Figure 5.8: SEM images of MoSi2 particles (left) and MoSi2B particles (right) after wind sifting.......... 47
Figure 5.9: The two morphologies present for all aluminium oxalate sol-gel coatings except those
resulting from experiment two, high-resolution (left) and low-resolution EDS image with
measurement points indicated (right). .................................................................................................. 48
Figure 5.10: An SEM image of particles resulting from experiment 2, only being coated by small
patches of aluminium oxalate instead of almost complete coverage................................................... 49
Figure 5.11: SEM images of the three final sol-gel experiments, with neither pregelation nor nitrogen
bubbling (a), dispersed phase after pregelation (b), with nitrogen bubbling (c) and the agglomerated
part at the bottom after pregelation (d). .............................................................................................. 50
Figure 5.12: Particles coated with aluminium tri-isopropoxide precursor based gel at different
magnifications. ...................................................................................................................................... 52
Figure 5.13: SEM images of four of the aluminium tri-sec-butoxide experiments, with 10g (a) and 20g
(b) Al(OC4H9)3 separated by evaporation and 10g (c) and 20g (d) Al(OC4H9)3 separated by
centrifugation. ....................................................................................................................................... 53
Figure 5.14: TGA results for an appropriate sample of each type of precursor compared to uncoated
MoSi2 as a blank. ................................................................................................................................... 55
Figure 5.15: Cross-section SEM-BSE images of coated particles for the SG-10g sample (left) and the
SG-20g sample (right). ........................................................................................................................... 57
Figure 5.16: Coating thickness distribution for the SG-10g sample from cross-section analysis. ......... 57
Figure 5.17: Coating thickness distribution for the SG-20g sample from cross-section analysis. ......... 57
Figure 5.18: SEM image of particles coated by ALD with morphology characteristic for all ALD
experiments with a resulting thin (<100 nm) coating (left) and an image with EDS measurement
points indicated (right). ......................................................................................................................... 59
Figure 5.19: TGA results for the 4 minute TMA, 5 minute purge ALD/rCVD sample compared to
uncoated material. ................................................................................................................................ 60
Figure 5.20: The effect of TMA and water dosage time on resulting average coating thickness as
obtained by the EPMA method.............................................................................................................. 61
Figure 5.21: The effect of purge time on resulting average coating thickness for two different TMA
dosage times as obtained by the EPMA method. .................................................................................. 61
vii
Figure 5.22: SEM images of the two samples with thicker coatings, namely ALD-25C (left) and ALD40C (right). ............................................................................................................................................. 63
Figure 5.23: Measured thickness for ALD/rCVD samples with different number of cycles, all with 4
minutes of TMA dosage, 5 minutes water dosage and 5 minutes purge per cycle. .............................. 64
Figure 5.24: Cross-section SEM-BSE images of coated particles for the ALD-25C sample (left) and the
ALD-40C sample (right).......................................................................................................................... 64
Figure 5.25: Coating thickness distribution for the ALD-25C sample from cross-section analysis. ....... 65
Figure 5.26: Coating thickness distribution for the ALD-40C sample from cross-section analysis. ....... 65
Figure 5.27: A linescan of the coating of the ALD-25C sample with the scanned region (left) and the
atomic percentages detected for each element as a function of distance (right)................................. 66
Figure 5.28: SEM images of precalcined SG-10g sample (left) and high temperature annealed sample
(right). .................................................................................................................................................... 66
Figure 5.29: Morphology of sol-gel samples after heat treatment with the precalcined (450 ᵒC, 14h)
only SG-20g sample (left) and the SG-20g sample subsequently annealed at 1200 ᵒC (right). ............ 67
Figure 5.30: XRD diffractograms of the SG-10g sample annealed at different final temperatures and
including the sample before any heat treatment and after precalcination. ......................................... 68
Figure 5.31: XRD diffractograms of the SG-20g sample annealed at different final temperatures and
including the sample after precalcination. ............................................................................................ 68
Figure 5.32: Cross-section SEM images of a heat treated particle, namely SG-20g at 1200 ᵒC (with
precalcination), showing a BSE image (left) and a SEM image (right). ................................................. 69
Figure 5.33: Coating thickness distribution for the SG-20g sample heat treated at 1200 ᵒC................ 70
Figure 5.34: Morphology of ALD samples after heat treatment with the 25 cycle sample (left) and the
40 cycle sample (right), both annealed at 1200 ᵒC................................................................................ 70
Figure 5.35: XRD diffractograms of the ALD-25C sample annealed at different final temperatures and
including the sample after precalcination. ............................................................................................ 71
Figure 5.36: XRD diffractograms of the ALD-40C sample annealed at different final temperatures and
including the sample after precalcination. ............................................................................................ 72
Figure 5.37: Relative weight change as a function of time for two blanks and the MoSi2B 6wt% Al
SG20g coated sample during a TGA test at 1000 ᵒC in synthetic air for 100h. ..................................... 73
Figure 5.38: SEM images at different magnifications of MoSi2B coated with Al2O3 according to the SG20g sol-gel procedure and heat treated at 450 ᵒC and 1200 ᵒC in argon. ............................................ 74
Figure 5.39: XRD diffractograms of the coated MoSi2B particles before and after heat treatment..... 74
Figure 5.40: SEM-BSE images of two different indents at different magnifications with HV10 (left) and
200N force (right). ................................................................................................................................. 76
Figure 5.41: SEM images of an indent before (left) and after (right) heat treatment in air at 1100 ᵒC
for 1 hour (heating and cooling rate 5 ᵒC/min). .................................................................................... 77
Figure 5.42: SEM images of cracks close to an indent with a BSE image (left) and a SEI image (right),
showing the presence of crack filling. ................................................................................................... 78
viii
List of abbreviations
ALD = Atomic Layer Deposition
rCVD = Residual Chemical Vapor deposition
YSZ = Yttria Stabilized Zirconia
TBC = Thermal Barrier Coating
TGO = Thermally Grown Oxide
BC = Bond Coat
PZC = Point of Zero Charge
TMA = Trimethylaluminium
SEM = Scanning Electron Microscopy
EDS = Energy-Dispersive x-ray Spectroscopy
XRD = X-Ray Diffraction
BET = Brunauer-Emmett-Teller physical adsorption model
XPS = X-ray Photo-electron Spectroscopy
ICP-OES = Inductively Coupled Plasma Optical Emission Spectroscopy
XRF = X-Ray Fluorescence
TGA = Thermo-Gravimetric Analysis
DSC = Differential Scanning Calorimetry
EPMA = Electron Probe MicroAnalysis
SPS = Spark Plasma Sintering
SG-10g = wind sifted MoSi2 particles coated with 10g aluminium tri-sec-butoxide per 10g MoSi2
SG-20g = wind sifted MoSi2 particles coated with 20g aluminium tri-sec-butoxide per 10g MoSi2
ALD-25C = wind sifted MoSi2 particles coated with the ALD/rCVD method using 25 cycles
ALD-40C = wind sifted MoSi2 particles coated with the ALD/rCVD method using 40 cycles
MoSi2B = MoSi2 particles containing 2 wt% alloyed boron
PSD = Particle size distribution
BSE = BackScatter Electron image
ix
Table of contents
Abstract ................................................................................................................................................... iv
Samenvatting........................................................................................................................................... iv
List of Tables .............................................................................................................................................v
List of Figures............................................................................................................................................v
List of abbreviations ................................................................................................................................ ix
1-Introduction ......................................................................................................................................... 1
1.1 General .......................................................................................................................................... 1
1.2 Protection of healing particles ...................................................................................................... 1
2-Theory .................................................................................................................................................. 3
2.1 Thermal barrier coatings ............................................................................................................... 3
2.1.1 Regular thermal barrier coatings............................................................................................ 3
2.1.2 Self-healing in thermal barrier coatings ................................................................................. 4
2.1.3 Protective shells for self-healing capsules.............................................................................. 5
2.1.4 Material properties of main components .............................................................................. 6
2.2 Sol-gel .......................................................................................................................................... 11
2.2.1 Sol-gel chemistry .................................................................................................................. 11
2.2.2 Sol-gel coatings ..................................................................................................................... 13
2.2.3 Effect of pH ........................................................................................................................... 13
2.2.4 Effect of temperature ........................................................................................................... 15
2.3 Atomic layer deposition (ALD) ..................................................................................................... 15
2.3.1 Atomic layer deposition chemistry....................................................................................... 15
2.3.2 ALD on particles .................................................................................................................... 16
2.3.3 Atomic layer deposition with residual chemical vapour deposition (ALD/rCVD) ................ 17
2.3.4 Surface activation ................................................................................................................. 18
2.4 Heat treatment ............................................................................................................................ 18
2.4.1 Transformation and kinetics sol-gel coatings ....................................................................... 18
2.4.2 Transformation and kinetics ALD/rCVD coatings ................................................................. 19
2.5 Crack formation and healing in YSZ ............................................................................................. 20
Thermodynamics and Diffusion ............................................................................................................ 21
3.1 Thermodynamics and kinetics of the self-healing TBC system ................................................... 21
3.1.1 Oxygen and oxidation behaviour of MoSi2 ........................................................................... 21
3.1.2 Alumina/YSZ ......................................................................................................................... 22
3.1.3 Alumina/MoSi2 ..................................................................................................................... 23
x
3.1.4 Alumina/Silica ....................................................................................................................... 24
3.1.5 Silica/YSZ............................................................................................................................... 24
3.1.6 Coated particle system evolution ......................................................................................... 25
3.2 Diffusion of species in the coated healing particles .................................................................... 26
3.2.1 Bulk diffusion ........................................................................................................................ 26
3.2.2 Defect diffusion .................................................................................................................... 28
3.2.3 Diffusion model .................................................................................................................... 30
3.2.4 Results model ....................................................................................................................... 32
Materials and Methods ......................................................................................................................... 35
4.1
Wind sifting ........................................................................................................................... 35
4.2
Sol-gel methods ..................................................................................................................... 35
4.2.1
Aluminium oxalate method ........................................................................................... 36
4.2.2
Aluminium tri-isopropoxide method ............................................................................. 37
4.2.3
Aluminium tri-sec-butoxide method ............................................................................. 37
4.3
Atomic Layer Deposition/residual Chemical Vapour Deposition .......................................... 37
4.4
Heat treatment procedures .................................................................................................. 39
4.5
Characterization and performance ....................................................................................... 39
4.5.1
Characterization raw material ....................................................................................... 39
4.5.2
Characterization coated particles.................................................................................. 39
4.5.3
Characterization heat treated particles ........................................................................ 41
4.5.4
Performance testing final particles ............................................................................... 41
Results and Discussion .......................................................................................................................... 42
5.1
Characterization starting materials ....................................................................................... 42
5.1.1
Starting material ............................................................................................................ 42
5.1.2
Wind sifting ................................................................................................................... 45
5.2
Sol-gel coating ....................................................................................................................... 48
5.2.1
Oxalate method ............................................................................................................. 48
5.2.2
Aluminium tri-isopropoxide method ............................................................................. 52
5.2.3
Aluminium tri-sec-butoxide ........................................................................................... 52
5.3
Atomic Layer Deposition/Residual Chemical Vapor Deposition ........................................... 59
5.3.1
Mechanism .................................................................................................................... 59
5.3.2
Increased number of cycles ........................................................................................... 62
5.4
Heat treatment ...................................................................................................................... 66
5.4.1
Effect of atmosphere ..................................................................................................... 66
xi
5.4.2
Heat treatment of sol-gel coatings ................................................................................ 67
5.4.3
Heat treatment of ALD/rCVD coatings .......................................................................... 70
5.5
Performance .......................................................................................................................... 72
5.5.1
Thermogravimetric stability .......................................................................................... 72
5.5.2
Embedded particle stability and healing ....................................................................... 73
Conclusions and Recommendations ..................................................................................................... 79
6.1 Conclusions .................................................................................................................................. 79
6.2 Recommendations....................................................................................................................... 80
Acknowledgements ............................................................................................................................... 82
Bibliography........................................................................................................................................... 83
xii
1-Introduction
1.1 General
In 1972, the club of Rome brought to public attention one of the major problems humanity faces. In
their well-known report [1] the limits to economic growth and increasing human prosperity are
described, focusing particularly on the limited supply of oil. As oil is currently the main source of fuel
for transportation and suitable alternatives are not able to supply enough fuel in a cost-effective
manner, it is vital to use current reserves as efficiently as possible.
This is especially true for gas turbines and other high-temperature turbines. A good example is the
aviation industry, where fuel costs for jet engines can account for as much as 30% of the overall costs
of a flight [2]. According to Carnot's theorem, the best way to increase efficiency would be to
increase the operating temperature [3]. However, current turbines already operate at temperatures
significantly above the creep limit of the used nickel superalloys [4]. To prevent breakdown of the
structural parts, a (~0.5 mm thick) thermal barrier coating (TBC) in combination with internal gas
cooling are applied to prevent overheating. This allows especially the most creep-sensitive parts, the
turbine blades, to endure these extreme environments [5].
However, due to thermal expansion coefficient mismatch between the TBC, usually made of yttriastabilized zirconia (7 wt% Y2O3–ZrO2, YSZ) and the nickel superalloys, application of these coatings is
difficult and significant mismatch stresses arise during heating and cooling of the engine. Even
though a (~250 µm) bond coat (BC) with a (0.6-3.0 µm) thermally grown oxide (TGO) for oxidation
protection of the superalloy is applied, this mismatch together with the growth of the oxide layer
results in unavoidable crack growth and spallation damage in the TBC and frequent replacement of
the coating is therefore required [6]. However, the work of Carabat et al. [7] used a different
approach to repair damage autonomously, based on the inclusion of sacrificial MoSi2 healing
particles that oxidize, expand and fill the crack when it is close to the particle.
1.2 Protection of healing particles
There is however a challenge still to be overcome with this proposed system. This is because YSZ is
very transparent to oxygen at the turbine operating temperatures (1250-1500K depending on engine
and location in the TBC [5]) and therefore significant premature oxidation of MoSi2 is present. To
prevent premature oxidation, a shell has to be applied around these particles that both protects
against oxidation and allows cracks to grow through it to allow for oxidation when damage is
present. Based on a preliminary literature study, which can be found in appendix I, the most suitable
materials for such a coating were found to be α-alumina (α-Al2O3), zircon (ZrSiO4) and mullite
(Al6Si2O13).
However, application of coatings on MoSi2 has rarely been investigated, mainly due to the excellent
oxidation resistance of the bulk material at high temperature, resulting from the formation of a thick
homogeneous SiO2 layer. For small (10-30 µm) particles, this formation of a native oxide layer would
not be feasible though, as this would require most of the particle to be oxidized even before
incorporation into the TBC [7]. Therefore a protective coating has to be applied beforehand and
during the aforementioned preliminary literature study, the most suitable methods were found to be
sol-gel and Atomic Layer Deposition (ALD) routes.
1
The goal of this research is therefore to find an optimal route to produce MoSi2 particles coated with
α-Al2O3 that prevents significant premature oxidation, while at the same time allowing cracks to grow
through the coating. This will be done by optimizing sol-gel and ALD techniques followed by thermal
treatment to create different coatings. First, particles will be coated with both methods and they will
be compared on thickness and morphology. This is followed by heat treatment experiments and
subsequent comparison of resulting microstructure, which includes crystallinity, defect types and
defect concentrations obtained, hardness and grain size. Oxidation tests will also be performed on
the final particles and compared to a developed diffusion model to investigate oxidation resistance
and follow microstructural development during operation. Finally, healing tests will be performed to
show the validity of the self-healing concept and the possibility of crack propagation through the
coating.
The structure of this report is as follows. Chapter 2 introduces the background needed to understand
the TBC system, the self-healing system, the chemical methods to apply coatings and the subsequent
heat treatment. In chapter 3, the thermodynamics of the system are described, followed by the
development of a diffusion model. Chapter 4 then describes the experiments performed for
synthesis and characterization of the coated particles, testing performance and validating the
diffusion model. The results of these experiments are then shown and discussed in chapter 5,
followed by the main conclusions in chapter 6 and recommendations for continuation of the research
in chapter 7. Furthermore, the original research plan is described in appendix II.
2
2-Theory
This chapter introduces the main concepts regarding self-healing thermal barrier coatings and
provide an overview of the most essential literature on coating particles. First, the thermal barrier
coating system will be described in more detail, together with an introduction of the self-healing
system, the particle shell and the materials involved. This is followed by an introduction of the two
main chemical coating methods, sol-gel and Atomic Layer Deposition (ALD/rCVD) and how their
individual chemistries can be optimized for obtaining protective shells. Furthermore, theory of the
required subsequent heat treatment is also described. Finally, an introduction of crack propagation
through the shell is presented.
2.1 Thermal barrier coatings
2.1.1 Regular thermal barrier coatings
A thermal barrier coating (TBC) is any type of coating that is used to limit heat transport across this
coating. Such coatings are present in many different applications, but as mentioned in the
introduction, the focus in this work is on TBCs for gas turbines. A gas turbine is a type of internal
combustion engine, which consists of three stages: the rotating compression area, the combustion
zone and the exhaust, as is shown in Figure 2.1a.
Gas turbines use compressed air and chemical energy contained in the fuel to produce hightemperature, high-pressure gas to power the rotating compressor at the start of the engine, which is
connected to the exhaust area by a shaft. The remaining available work is either used to power any
other devices connected to the shaft (an electricity generator for example) or can exit the exhaust
area at high velocity to produce thrust (such as in a jet engine). Although current gas turbines are
already an efficient way to convert energy, with modern combined cycle (in which waste heat is used
by a regular steam turbine) electricity producing gas turbines reaching a turbine thermal efficiency of
39.5% and a total efficiency of nearly 60% [8], significant improvements are still possible. According
to Carnot's theorem, increasing operating temperature could still result in a significant efficiency gain
[3].
(a)
(b)
Figure 2.1: Jet engine layout (a) and interface between turbine blade and hot gas (b). [9], [10]
To allow for these high operating temperatures, advanced cooling methods and increasingly complex
layers of coatings are necessary to surpass the temperature limits of currently used nickel
superalloys, especially in the most critical component: the turbine blades. An overview of the current
3
coating system is shown in Figure 2.1b [10], in which the superalloy turbine blade is shown on the
left. Cold air is blown through these turbine blades to cool them [11] and allow a temperature
gradient to exist. The blade is coated with a ~100-250 µm bond coat (BC) containing significant
amounts of aluminium. This aluminium is oxidized to produce a continuously growing Al2O3 thermally
grown oxide (TGO) to protect the blade against high temperature oxidation. Compositions of these
superalloys and bond coats are very complex and shown in the preliminary study in appendix I.
On top of this TGO, a 0.1-0.5 mm TBC is present to protect the entire blade against the immense heat
of the combusted gases, which can reach a gas temperature in excess of 1500 ᵒC [8]. Finally a film of
cooling air is also blown along the outside of the blade to limit heat transport from the hot gas to the
surface of the TBC, allowing for a maximum TBC surface temperature of roughly 1200 ᵒC . This
however still requires a thermal gradient of 200 ᵒC over the coating to reach the limit of the nickel
superalloys [4]. For this reason, TBCs are often made from partially yttria stabilized zirconia (YSZ)
containing roughly 7 wt% yttria, although other materials are also under investigation [12].
However, almost all thermal barrier coatings are oxides, which suffer from an important drawback:
their low thermal expansion coefficient compared to nickel superalloys. This results in a significant
thermal expansion mismatch and subsequent compressive stresses in the TBC during cooling of the
gas turbine. Because oxides are relatively brittle, these stresses generate small cracks in the TBC,
especially close to the interface with the TGO [6]. This process is shown in Figure 2.2a. These small
cracks can then coalesce to form larger cracks and cause further delamination. Combined with the
compressive stresses, this can cause buckling and finally complete spallation of the TBC in certain
areas when the cracks start to grow perpendicular to the coating, as can be seen in Figure 2.2b.
Because of the sensitivity of the nickel superalloys to higher temperatures, the final result is
extremely rapid degradation of the turbine blade. To prevent degradation and possible catastrophic
failure, TBCs have to be inspected and replaced regularly [13].
(a)
(b)
Figure 2.2: Causes of spallation in TBCs illustrated with (a) showing the source of compressive stresses and (b) the
coalescence of microcracks and spallation. [14]
2.1.2 Self-healing in thermal barrier coatings
Instead of damage management, which consists of complete replacement of the coating once it is
too damaged to continue functioning, another option is the use of self-healing materials. These
materials are able to repair damage before failure occurs and can thereby prolong the lifespan of
4
materials [15]. A method for applying the self-healing concept to TBCs was suggested by W.G. Sloof
and S. van der Zwaag [16]. This concept introduces MoSi2 particles of 20 to 25 µm as a self-healing
agent, coated with a shell of Al2O3 to prevent oxidation. These particles are then introduced in the
TBC, close to the TGO, where most of the damage forms.
The self-healing mechanism is based on the oxidation of MoSi2 to form SiO2 and gaseous MoO3
according to reaction 2.1. When a crack grows through the shell, the particle is exposed to oxygen
and the reaction is able to proceed. Because the molar volume of 2SiO2 is larger than the molar
volume of MoSi2, the material will expand to 238% of the original volume upon complete oxidation
and is therefore able to fill the crack with SiO2, while the MoO3 sublimates and escapes through the
pores of the YSZ. SiO2 can also react with the matrix of ZrO2 to form ZrSiO4, better known as zircon.
As the toughness of zircon is higher than the toughness of YSZ, complete strength recovery and crack
healing is possible under the right conditions [17]. This self-healing mechanism is also illustrated in
Figure 2.3.
2 MoSi2 (s) + 7 O2 (g) → 2 MoO3 (g) + 4 SiO2 (s)
(2.1)
Figure 2.3: The self-healing thermal barrier coating system with on the left the whole turbine blade coating system,
zooming in on the particles on the right. Upper part is before healing and the lower part after healing. [18]
2.1.3 Protective shells for self-healing capsules
As mentioned before, MoSi2 poses challenges though. The material is supposed to oxidize rapidly at
temperatures between 1000 ᵒC and 1200 ᵒC in an oxygen-containing atmosphere with a partial
oxygen pressure PO2 between 10 and 10000 Pa[19]. Furthermore, YSZ is very transparent to oxygen,
indicated by its common use as solid oxide fuel cell barrier material [20]. The YSZ used in TBCs is also
very porous to accommodate the compressive stresses to a certain extend [6].
Although bulk MoSi2 can form a protective SiO2 coating at temperatures above 800 ᵒC, the thickness
of this coating usually several µm [21], which would consume a significant part of the particle
material. The formation of SiO2 would also lead to a reaction with ZrO2 to form zircon. To prevent this
from happening, a coating is necessary. Appendix I shows the different materials investigated for this
study, recommending Al2O3 as the most optimal shell material, while other interesting choices are
mullite (Al6Si2O13) and zircon (ZrSiO4).
5
The main purposes of this shell are to protect the particle itself from high-temperature oxidation and
to allow a crack to grow through the coating. Therefore the main requirement for this shell is to have
a low diffusion of oxygen and counterions. Other requirements for low diffusion is the absence or
minimization of defects that can act as a fast diffusion pathway. This includes, among others, pores,
cracks, grain boundaries and vacancies. This also requires the final shell to be completely closed and
of homogeneous thickness. The diffusion of oxygen through the coating will be discussed in more
detail in chapter 3 however.
It is also important that defects do not form during manufacture of these shells, the TBC system or
during operation in the coating. According to previous research, especially cracks are likely to occur
[22]. These cracks form due to stress build-up caused by either phase transformations and associated
volume changes or excessive oxidation of the substrate MoSi2 and resulting volume expansion. Stress
build-up from phase transformations can be prevented by ensuring a stable and fully densified phase
is created before operation. In the case of Al2O3, the only stable phase is the α phase or corundum
structure, but many transition aluminas are known and usually form before the α phase [23].
Therefore, proper heat treatment to obtain the α phase and obtain full densification is required to
form the final coating.
The interfaces of the shell with the substrate and the TBC itself could also act as another form of
stress. This is partially due to growth stress from SiO2 oxidation on the substrate side, which is
another important reason to limit diffusion through the shell. Another possibility on both sides of the
shell is the mismatch in thermal expansion. The coefficients of thermal expansion for Al2O3 and MoSi2
are very similar however, resulting in low stresses in the coating on this side. There is however a
thermal expansion mismatch between YSZ and the particles, which would probably result in stress
build-up around the particles. This is studied in more detail by Turteltaub et al. [14]
For crack propagation through the shell and into the particle, Turteltaub [14] found that interface
strength is a critical factor. The preferred interface should be strong at the TBC side and should be
relatively weak at the particle side. Flaws in the particle could help due to coalescence of the two
cracks. This also holds for cracks in the shell that do not reach either of the interfaces, but in this case
the crack would not necessarily grow through the protective layer and reach the particle, which is
undesired. If the crack would reach sufficiently far into the shell, accelerated local oxidation due to a
thinner coating might fracture the shell completely anyway though. However, flaws in the shell
should still be avoided because they could initiate cracks when stresses are present. A more
elaborate discussion on the interfaces with both sides is presented in chapter 3.
2.1.4 Material properties of main components
To understand the self-healing TBC system better, some information on the materials involved is also
necessary. Therefore, some information on Al2O3, MoSi2, YSZ and the healing products SiO2 and
ZrSiO4 is presented here, along with the most important thermal properties in Table 2.1.
Alumina (Al2O3) is one of the most well-studied ceramics due to its many applications which are the
result of its excellent properties and its availability. Alumina in its stable α phase is a very hard and
relatively strong ceramic that has an extremely high melting temperature of 2072 ᵒC [24]. This high
melting temperature in combination with other favourable thermal properties, such as a low
coefficient of thermal expansion, as is shown in Table 2.1 and its low ionic conductivity at high
temperatures, making it a very good candidate for corrosion protection scales [25].
6
The only thermodynamically stable phase is α-Al2O3, which crystallizes in the corundum crystal
structure shown in Figure 2.4. However, many other metastable phases are known and many of them
have important uses, resulting from their often high surface areas. One of the more famous
examples is the extensive use of γ-Al2O3 as both a catalyst and a catalyst support [26]. These
metastable phases are usually divided by packing of the oxygen anion lattice in either an FCC (for γ,
δ, θ and η among others) or HCP (for α, κ and χ) packing [23]. The difference between the separate
phases is the distribution of the Al3+ cations in the anion lattice.
(a)
(b)
3+
2-
Figure 2.4: The corundum crystal structure with (a) showing the regular structure with both Al and O [27] and (b)
showing the locations of the empty alumina sites in the structure [28].
Molybdenum disilicide (MoSi2) is one of the thermodynamically stable intermetallic phases of
molybdenum and silicon, the phase diagram of which is shown in Figure 2.5. Usually considered to be
a ceramic, its high melting point of 2030 ᵒC [29] makes this material an interesting refractory. Due to
its intermetallic nature, its electrical conductivity is high [29]. Bulk MoSi2 is also resistant to high
temperature oxidation due to the formation of a protective SiO2 scale at high temperatures. Due to
the combination of these properties, MoSi2 is often used as a heating element in applications that
require high temperatures.
The crystal structure of MoSi2 is tetragonal (space group I4/mmm), although a metastable hexagonal
structure also exists at low temperatures [30]. The stable tetragonal structure is shown in Figure 2.6.
7
Figure 2.5: The temperature dependent phase diagram of
molybdenum and Silicon, along a three component phase
diagram of Mo, Si and O at 1200 ᵒC. [31]
Figure 2.6: The crystal structure of the thermodynamically
stable tetragonal structure of MoSi2. [30]
Zirconia or ZrO2 is another refractory ceramic that is well-known for its high temperature stability. It
is in fact one of the most refractory oxides currently known, with a melting point of 2715 ᵒC.
However, pure zirconia has three known phase transformations, namely the low temperature
monoclinic phase (<1170 ᵒC), the intermediate tetragonal phase (1170 ᵒC - 2340 ᵒC) and the high
temperature cubic phase (>2340 ᵒC). These phase transformations are associated with significant
volume changes; an 8% increase in the case of the tetragonal to monoclinic transformation. High
temperature phases are also difficult to retain due to the transformation temperatures being
relatively high too.
High temperature phases can however be stabilized by doping with other oxides. The most common
oxide for stabilization is yttria (Y2O3) and can stabilize both the tetragonal and the cubic phase at
room temperature. Adding sufficient yttria (roughly 18 mole%) even results in the cubic phase being
the thermodynamically stable phase at room temperature, as is shown in the ZrO2-rich zirconia-yttria
phase diagram in Figure 2.7[32]. Usually, only 7-8 mole% of yttria is used though, stabilizing the cubic
phase enough to be metastable at room temperature when quenching from the liquid phase,
resulting in cubic yttria stabilized zirconia (YSZ). The crystal structure of this cubic YSZ is shown in
Figure 2.8, with Y3+ substituting randomly for Zr4+, forming oxygen vacancies in the process, according
to reaction 2.2 in the Krüger-Vink notation.
Y2O3 → 2 YZr' + 3 OO + VOᵒ ᵒ
(2.2)
These oxygen vacancies and their mobility at high temperature are the reason that YSZ is very
conductive to oxygen at higher temperatures. Although this can be useful for some applications, such
as fuel cells[33], it is detrimental to any system that is sensitive to high-temperature oxidation [6].
8
Figure 2.7: The ZrO2-Y2O3 phase diagram in the ZrO2 rich
region, showing the different phases of zirconia and their
stability depending on temperature and yttria content.
[32].
Figure 2.8: The crystal structure of cubic YSZ with Y
substituting randomly for Zr [33].
The main healing agent in the TBC system is SiO2, another very important and well-studied ceramic.
Although less temperature-resistant than the other ceramics mentioned before, it is still a refractory
material with a melting temperature of 1713 ᵒC [34]. Silica is one of the main constituents of earth's
crust and besides having very interesting high-temperature properties, is also known for the many
different (usually metastable) crystal structures it can form, especially in combination with other
oxides such as Al2O3, CaO, MgO or iron oxides. These are known as the silicates and many are
important minerals.
Figure 2.9: The temperature-pressure phase diagram of
silica [35].
Figure 2.10: Crystal structure of trigonal α quartz [36].
9
Pure SiO2 can exist in multiple phases, depending on temperature. Trigonal α quartz is the most
stable room temperature crystal structure, but at higher temperatures hexagonal β quarts, tridymite
and cristobalite become more stable, as is shown in the SiO2 phase diagram in Figure 2.9 [35].
Whether tridymite is stable or metastable is sometimes disputed however, mainly because small
amounts of impurities are required for the transformation from quartz to tridymite. In very pure
silica, quartz will directly transform into cristobalite [37]. All crystal structures are however based on
SiO2 tetrahedra, as is illustrated in Figure 2.10, which shows the crystal structure of α-quartz. Along
with these crystal structures, another form of these tetrahedra is amorphous silica, which is also
often formed and remarkably stable. This is illustrated by glass, the most well-known amorphous
form of SiO2 (along with some other components).
When SiO2 reacts with the YSZ according to reaction 2.3, zircon or zirconium silicate (ZrSiO4) gets
formed. This extremely resilient silicate of zirconia is both very tough and strong for a ceramic
material. Because of this and its suitable thermal properties [32], complete strength recovery of the
TBC is possible. The zircon structure is similar to the SiO2 structure in that it also consists of
tetrahedra, but in zircon the ZrO4 and SiO4 tetrahedra alternate. They crystallize in a body-centered
tetragonal crystal structure (space group I 41/amd), which is shown in Figure 2.11.
ZrO2 (s) + SiO2 (s) → ZrSiO4 (s)
(2.3)
Figure 2.11: Crystal structure of zirconium silicate [17].
Table 2.1: Important thermal properties, namely melting temperature, coefficient of thermal expansion and thermal
conductivity of the main materials tabulated based on data from the (Japanese) National Institute of Materials Science
(NIMS).
Material
Al2O3
MoSi2
ZrO2 7mol%Y2O3
SiO2
ZrSiO4
*Decomposes
Melting temperature Coefficient of thermal expansion Thermal conductivity
(ᵒC)
(10-6 K-1)
(W m-1 K-1)
2072
8.0
39
2030
7-10
70
2715
10.3
2
1713
0.59
1.4
1676*
5.0
3.5
10
2.2 Sol-gel
Although it is now understood that the materials involved in the self-healing system are suitable, this
will only hold if produced in a suitable manner. One of the methods selected in appendix I that would
be able to achieve the required final microstructure described in both appendix I and chapter 2.1.3 is
the sol-gel method, which will be introduced here.
2.2.1 Sol-gel chemistry
A sol-gel process is difficult to define and currently many different definitions exist. These definitions
vary in whether the degree of mixing and homogeneity should be on the atomic level or if colloidal
mixing could also be considered sol-gel processing. Another point of discussion is whether sol-gel can
also include the processing of a single oxide or that only multicomponent oxides should be included.
Furthermore, novel sol-gel techniques allow for the production of certain nitrides, sulfides and
carbides, raising the question to whether sol-gel should be restricted to oxides. To avoid any
confusion, in this thesis the broad definition of a sol-gel process from the book of A.C. Pierre [38] will
be used, which reads: "A sol-gel process is a colloidal route used to synthesize ceramics with an
intermediate stage including a sol and/or a gel state."
Figure 2.12: A schematic representation of the sol-gel process, with multiple possible microstructures depending on
processing route [39].
This process is schematically shown in Figure 2.12 and highlights that any sol-gel process consists of
multiple steps and starts with the preparation of a solution or multiple separate solutions of
precursors. This solution should then condensate, either by a chemical reaction or by removal of the
solvent to form a sol. A sol is a colloidal solution, which means it is a stable suspension of colloidal
solid particles. This can consist of either very small particles of bulk material or large macromolecules
(usually consisting of 103 atoms or more) [40]. It should be noted here that stable is of course
relative. A sol usually has a tendency to evolve to a more gel-like state and is only stable for a limited
amount of time. This is usually not a problem, as long as the sol is stable long enough for its
application.
11
When this sol evolves to a gel, a network starts to form. Particles connect or macromolecules bond,
rigidifying the system. This network still contains a significant amount of liquid, but due to its
significant and abrupt increase in viscosity, does not completely behave like a liquid anymore [38].
Due to the small particles and homogeneous distribution of components in the liquid, the resulting
gel is very homogeneous. This results in a very homogeneous end material as well.
After the sol or gel is formed, it is first dried to remove the liquid phase completely. In the case of
gels, this dried gel is called a xerogel if dense or an aerogel if porous. Many drying and forming
processes can be used to obtain a variety of materials and components. Gels with sufficient viscosity
can be formed easily but will retain their shape while drying and solidifying. Sols do not necessarily
need to gelate before being applied. Especially in the case of coatings, several techniques exist to
apply a sol to a substrate and have gelation initiate afterwards, as is also shown in Figure 2.12. This is
especially useful if phase separation is an issue for the sol-gel system [38]. When sol-gel is used to
form ceramic components, a subsequent heat treatment is often required to obtain the desired final
phase(s). This heat treatment for alumina gels will be described in chapter 2.4.
One of the challenges of sol-gel processes is the complexity of the process due to the many phases
and chemical species involved. There are many different parameters in each stage that influence the
final result. Furthermore the chemistry of sol-gel techniques is complex and not fully understood
[38]. Many parameters are important in this system, such as the chemistry of the precursor
solution(s)/sol/gel, the precursor used, the processing route and the possible addition of binders,
fillers and other compounds [41].
Precursors for ceramic materials can be divided in three broad categories: metal salts, alkoxides and
powders obtained from bulk material. Metal salts are usually quite soluble in the solvent and form
colloids by hydrolysis and subsequent condensation of OH bridges, as described in [38]. Alkoxides on
the other hand are metals bonded to the oxygen atom of a deprotonated alcohol and react rapidly
with H2O, combined with condensation, as is shown in reaction 2.4, taking an aluminium alkoxide as
an example. Finally, powders obtained from bulk material only have to be properly dispersed to form
a sol, assuming they are sufficiently small in size.
Al(OR)3 + 2 H2O  AlOOH + 3 R-OH
(2.4)
Chemical environment and concentrations of the species present have a significant influence on the
hydrolysis and condensation reactions and therefore on the sol-gel process. This is mainly due to
electrostatic interactions between the ionic and polar species present. Therefore, pH and polarity of
the solvent are especially important for behavior and evolution of the system.
Another important parameter is the processing route, which is the way the system will evolve from
the precursor solution(s) to the final product. Of critical importance here are the sol aging time and
temperature, as these are critical to control gelation of the sol. This gelation is due to the aggregation
of the colloidal particles. Due to brownian motion of very small particles, collisions are frequent and
if attractive forces between the particles are larger than repulsive forces, they rapidly stick together
and form a gel. Attractive forces between the particles are mainly Van-der-Waals forces [42], while
repulsive forces that can prevent or delay this gelation are usually either electrostatic or steric [41].
12
2.2.2 Sol-gel coatings
Although sol-gel methods are very versatile, coating substrates using a sol-gel method usually
requires the system to be a sol or at least not fully gelated, as is shown in Figure 2.12. And even
though alumina is one of the most investigated materials produced by sol-gel, most of these are
related to bulk alumina, especially for catalytic applications. However, coatings of alumina prepared
by a sol-gel route have been investigated and used frequently for different applications [39] [43] [44].
Some processes use aluminium salts or sometimes combinations of salts and alkoxides [45] in an
aqueous environment. Most of the investigations focus on the Yoldas method [46] though, which
hydrolyzes an aluminium alkoxide in an aqueous environment to obtain alumina gels that can be
used for coatings [44]. This route has been applied to many substrates with many different functions,
of which the corrosion and scratch protection of stainless steel [41] and the coating of carbide
cutting tools to protect against wear and high temperature oxidation [47].
It should be noted though that the stainless steel protective coating was not completely transformed
to α-alumina and it therefore failed after a few or in some cases even one thermal cycle from room
temperature to 900 ᵒC and back to room temperature due to transformation stresses. This highlights
the importance of proper heat treatment again and that sol-gel methods are sensitive to cracking
from volume changes, which is also observed in many other investigations [41]. The coating of MoSi2
with alumina by sol-gel has however not been investigated yet due to its excellent high-temperature
oxidation resistance as a bulk material [29], except for the work of Carabat et al. [22].
An advantage of bulk materials is that application of the sol is relatively straightforward, as the
component to be coated can simply be dipped in the aqueous sol and slowly pulled out, a process
named dip coating [38]. Another possibility is the use of spin coating, which is also relatively
straightforward. However, the proposed self-healing system requires that particles of 20-25 µm are
coated. Therefore, these methods of sol application are not feasible and a different method has to be
used.
Some investigations have been done on the coating of particles by sol-gel, for example on phosphors
[48], magnetic particles [49] and silicon carbide particles [50] [51], in all cases with the goal to
protect the particles from the environment. For the aforementioned articles, the Yoldas method with
some modifications is used in all cases. The main difference is however that in all cases the particles
are added before the alkoxide and water are mixed. Therefore the sol is formed while particles are
already in suspension, aiding greatly in sol and gel formation on the particles instead of phase
separation of a gel, which is also often observed [41]. These particles are however both smaller and
with a lower density (except for the magnetic particles) than the MoSi2 particles proposed in the selfhealing system (0.5 µm instead of 20 µm) and therefore easier to disperse and to keep dispersed.
Although no work on MoSi2 sol-gel coating with Al2O3 exists, the work done on SiC is very useful, due
to the very similar surfaces, both consisting of a layer of native oxide SiO2 [52]. When preparing a
coating for particle shell molar ratios of 1:100 or 1:150 of aluminium alkoxide to water [50]. The
effect of pH and temperature are also important and discussed in the next chapters.
2.2.3 Effect of pH
The effects of pH on the sol-gel coating process of SiC particles has been thoroughly investigated by
Yang and Shih [51] and also by Yang and Troczynski [53]. They found that the effect of pH is twofold.
13
First, the acid changes the nature of the alumina species in the sol [54], which is shown in Figure
2.13. The exact species that can be present or are present in a given sol are disputed however [41].
Nevertheless, the species and therefore the pH have a signficant influence on both the crystallinity
and the type of crystal of the resulting colloidal particles [51]. At high pH the sol forms bayerite
(AlOH3) with high crystallinity, while at lower pH a more amorphous structure of boehmite (AlOOH) is
observed [41]. This is also reinforced by the observation that base-catalyzed sols in general allow for
more growth of particles, resulting in larger and more crystalline colloids, while acid-catalyzed sols
generally promote aggregation of small particles into a more homogeneous, but less crystalline gel
network [38]. This results in a more homogeneous, dense and amorphous coating derived from acidcatalyzed gels and thought to be caused by partial dissolution of the constituents by the acid.
Figure 2.13: The fraction of alumina species present in an aqueous solution as a function of pH at 25 ᵒC [54].
Yang and Shih also observed that bayerite is not able to coat SiC, even when additional acid is added.
This can be explained by surface charges on the SiC particles. Ionic materials such as the SiO2 on the
surface of most silicides will absorb either protons (H+) or hydroxyl ions (OH-) in water depending on
the pH, following the equilibrium described in equation 2.5. In this equation the reaction will move
more towards the right with increasing pH. This results in the buildup of a charged double layer,
measured by the so-called ζ-potential (zeta potential). Aqueous
M-OH2+ + H2O ↔ H3O+ + M-OH + OH- ↔ M-O- + H2O
(2.5)
For every oxide, there is a so-called point of zero charge (PZC), at which the reactions balance each
other out and the surface has no net electrical charge. For a pH>PZC, the surface will absorb more
OH- ions, which results in a negative electrical charge, while for pH<PZC, the surface will absorb more
H+ ions and the surface will be positively charged.
For SiO2, this PZC is pH 2.5-3.0 [55], which means that in all experiments performed by Yang and Shih,
the surface was negatively charged. As can be seen in Figure 2.13, bayerite also has a negative charge
and the particles and sol will therefore repel each other. Boehmite on the other hand has a PZC
14
around 9 [55] and will therefore be positively charged for a pH below this. The significant attractive
forces resulting from these opposite charges can be used easily to ensure complete and
homogeneous coatings on particles. An example of this attraction and the subsequent increase in
zeta potential as function of pH is shown in the research of Yang and Troczynski and shown in Figure
2.14.
Figure 2.14: Zeta potential for SiC in water and alumina sol as function of pH [53].
2.2.4 Effect of temperature
Sol-gel is known as a low-temperature process, especially for ceramics, because it is a liquid phase
process and therefore limited by the solvents freezing and boiling point. For water, this limitation is
roughly between 0 and 100 ᵒC, which is a narrow range for temperature effects. Nevertheless, the
effect of temperature in alumina sols was investigated by Pierre and Uhlmann [56] and found a
significant influence of temperature on sol and gel behaviour.
This research was based on acidic sols with varying ratios of nitric acid (HNO3) to aluminium tri-secbutoxide (Al(OC4H9)). They found that the density of the gel in a sol of 90 ᵒC has a maximum at a ratio
of HNO3:Al of 0.07, while at room temperature, density decreased monotonously with increasing
HNO3 concentration. This indicates it is possible to control and maximize solid loading with pH at high
temperature. Furthermore the structure was also found to be different, as crystallinity of the
boehmite seemed to increase with higher temperature.
2.3 Atomic layer deposition (ALD)
The second method that was selected in appendix I is the Atomic Layer Deposition (ALD) method.
This method, together with proper annealing techniques, should also be able to obtain the required
microstructure and will be described here.
2.3.1 Atomic layer deposition chemistry
Atomic layer deposition is an elegant technique that is related to Chemical Vapour Deposition (CVD)
and is an important technique for the deposition of thin films utilizing gas phase reactants. It is
15
characterized by the use of two self-limiting half-reactions with the surface to deposit conformal
solid films on this surface. Because of these self-limiting reactions, the film thickness can often be
controlled up to single atomic layers, hence the name atomic layer deposition [57].
One of the most common ALD deposition processes is that of alumina (Al2O3) from
trimethylaluminium (TMA) and water (H2O) [58]. This process is illustrated in Figure 2.15 and the two
half-reactions and the final result of the two reactions are schematically shown in reaction 2.6, 2.7
and 2.8 respectively. Reaction 2.8 does never occur in pure ALD, as the precursors are never
introduced in the reactor at the same time. It is the final result of the combinations of 2.6 and 2.7. In
reactions with CVD, reaction 2.8 is likely to occur however.
It should be noted that reaction 2.6 also has another option in which one molecule of TMA reacts
with two surface groups, leaving only one methyl group able to react with water. As can be seen in
these reactions and the figure, TMA first reacts with hydroxyl groups present on the surface,
followed by a purge step to remove excess reactant. Then water is added to the reactor to allow the
oxidation of the other methyl groups present on TMA, forming new hydroxyl groups in the process.
Finally the excess water is also purged and the process can be repeated to deposit another cycle [59].
M-OH + Al(CH3)3 (g)  M-O-Al(CH3)2 + CH4 (g)
(2.6)
M-O-Al(CH3)2 + 2H2O (g)  M-O-Al(OH)2 +2CH4 (g)
(2.7)
2 Al(CH3)3 (g) + 3H2O (g)  Al2O3 (s) + 6CH4 (g)
(2.8)
From this description, it is evident that any starting sample needs to have hydroxyl groups on the
surface, which results in many materials being challenging or unsuitable for ALD. However, due to
the thin native SiO2 layer being present on the surface, MoSi2 can easily be activated to form SiOH
groups that would be able to react with TMA. This activation will be discussed later in this chapter.
Figure 2.15: A schematical overview of one cycle in the ALD process [60].
2.3.2 ALD on particles
The current main application of ALD is for thin coating deposition on wafers and other relatively flat
substrates. However, deposition on particles is possible although some additional challenges are
present. The surface area to be coated is in general significantly larger than the same mass of wafers,
which requires the supply of more reactant. Another problem is the slow mass transfer in a bed of
16
particles, even when porosity is relatively high. This would result in non-homogeneous coatings from
retained TMA during the purge and lack of access of TMA to the least accessible surfaces or
prohibitively long cycle times [61].
A solution to both of these problems is the use of a so-called fluidized bed to coat particles. In a
fluidized bed, a gas is blown through a bed of particles with sufficient velocity to suspend the
particles in this gas flow. This causes both the particles and gas to act as a fluid, increasing gas-solid
contact and mixing enormously [62]. Furthermore, by using a carrier gas to fluidize the system and
evaporation of the reactants, the amount of TMA that can be supplied to the reactor can be
increased significantly. Although fluidization of particles smaller than roughly 20 µm can prove
challenging [62], it is possible provided that some agglomeration is present [63]. These authors also
found that ALD under atmospheric conditions (25 ᵒC, 1 bar) is possible.
2.3.3 Atomic layer deposition with residual chemical vapour deposition (ALD/rCVD)
One of the remaining challenges for manufacturing high-temperature diffusion barriers with ALD
however is the low thickness growth per cycle. This is advantageous for producing nanostructured
materials, but for particles that require thicker coatings, this low growth again results in prohibitively
long processing times. Although cycle times depend on the reactor used and especially the residence
time of the gas in the reactor [63], a single cycle on lab scale usually requires between 10 and 30
minutes and as is evident from the TGO layer in the TBC system, alumina layers need to be at least
several 100 nm in thickness. Utilizing pure ALD, which has a layer growth of 0.1-0.16 nm/cycle, this
would take about 2000 cycles or 20000 minutes/14 days of continuous operation in the best case. In
some cases the addition of catalysts that aid in decomposition of the precursor can be added to
significantly increase growth per cycle, such as the well-known example of SiO2 deposition being
catalyzed by trimethylaluminium [64]. Unfortunately, no such catalyst is currently known for TMA
itself.
However, another possibility has been found by Garcia-Trinanes and Valdesueiro [65] [63]. In this
research, dosing a significant excess of both precursors compared to the amount of reactive groups
present at the surface at ambient conditions resulted in higher growth per cycle rates. The
explanation given in these articles is that operation of the ALD process below the boiling point results
in condensation or physisorption of reactant molecules that can subsequently react with the other
reactant during the next half-cycle, resulting in a CVD-like component of the ALD process.
The mechanism is not completely understood however. It has been shown that temperatures above
the boiling point do indeed result in more ideal ALD with growth per cycle close to that of literature
values [63]. For room temperature ALD, the effect of single precursors is not understood however
and neither is the effect of purge time. One would expect that a shorter purge time would result in
less re-evaporation of reactant and a resulting increase in growth per cycle, which corresponds with
the findings of other authors that at low temperature partial CVD will occur when purge times
become too short [66].
For the effect of single precursors, it is possible that both precursors condense, but also that only one
of the precursors condenses in significant amounts, while the other precursor only reacts with the
condensed precursor. Based on purge times required by Groner et al. [66] during low temperature
ALD, H2O takes significantly longer to remove from the reactor than TMA, hinting at water
condensation being more important than TMA condensation.
17
Temperature also has an effect on the microstructure of the resulting coatings. According to Groner
et al. [66] the density of alumina coatings deposited by ALD on PET substrates decreases significantly
at lower temperatures, from 3.0 to 2.5 g/cm3, as is shown in Figure 2.16. Both of these densities are
however significantly lower than the bulk density of 3.99 g/cm3 of α-alumina, or 3.5-3.7 g/cm3
commonly reported for other amorphous alumina films [67]. This is partially explained by the ALD
process depositing amorphous alumina instead of crystalline alumina [58], but another likely factor is
the increased hydrogen and hydrocarbon content and microporosity resulting from the low
temperature process [63].
Figure 2.16: Density, refractive index and growth rate of Al2O3 coatings on PET as function of temperature [66].
2.3.4 Surface activation
Although it is already possible to use this process to coat SiC with only a native SiO2 layer,
improvements in growth per cycle can be achieved by activation of the surface with ozone [65]. A
possible explanation could be that the reaction with ozone increases the number of reactive groups
present on the surface, as is observed in Si wafer bonding [68] [69]. However, as the mechanism is
expected to be based on condensation, a more likely explanation would be that this pre-treatment
also has an effect on the condensation of reactants and possibly the deposited Al2O3 film, as the
effect is also observed after many cycles with coatings several 100 nm thick. The nature of this effect
is however unknown.
2.4 Heat treatment
Although closed and homogeneous films of Al containing material can be applied to MoSi2 particles,
these films are not yet in the stable α phase, as ALD films are generally amorphous, while the sol-gel
films are usually boehmite (AlOOH) that is partially amorphous and partially in the γ phase [23].
Furthermore, the deposited alumina does not have the density required and some impurities left
from the process are still present. To attain the desired dense α alumina films and remove
combustible impurities and H2O, heat treatment is required, which will be discussed here.
2.4.1 Transformation and kinetics sol-gel coatings
For sol-gel coatings consisting of AlOOH, transformation to the α phase requires following a
transition sequence of multiple metastable aluminas. Before that however, removal of both water
still contained in the gel and crystal water contained in AlOOH is necessary according to equation 2.9.
The transformation of boehmite to γ-alumina usually takes place between 300 and 500 ᵒC [23], which
18
is also the temperature range at which residual organic material can be burned away with sufficient
access to oxygen.
2 γ-AlOOH (s)  γ-Al2O3 (s) + H2O (g)
(2.9)
Further transformation from γ-alumina goes through δ-Al2O3 between approximately 700-800 ᵒC,
followed by θ-Al2O3 between 900-1000 ᵒC and finally to the α-phase from 1000-1100 ᵒC according to
Levin et al. [23]. Because γ, δ and θ are metastable phases, it is possible for them to coexist in the
sample depending on the annealing conditions, which makes distinguishing between them difficult.
Furthermore, all of these metastable phases have an oxygen anion packing in the FCC phase, while αalumina has an HCP anion lattice. Due to this, the transformation from θ to α is the slowest step with
the highest activation energy of 557 kJ/mol [70], which is also visible in the unusually high
temperature required to obtain the thermodynamically stable α phase. This transformation is also
thought to occur through a nucleation-growth based process, which makes the formation of grain
boundaries unavoidable.
Another issue with the transformation of sol-gel coatings is the volume change, which causes stress
buildup. This volume change results from the equilibrium densities of the transition aluminas being
lower than that of α-alumina, with 3.6-3.7 g/cm3 for most transition aluminas and 3.99 for the α
phase [23]. Boehmite has a density of approximately 3.08 g/cm3, with approximately 15% of this
mass consisting of H2O that will evaporate during the transformation. This significant difference in
density results in high tensile stresses in the coating and can result in major cracking and spallation of
alumina sol-gel coatings on bulk samples [41]. Heat treatment procedures should therefore aim to
reduce these stresses with a proper temperature profile.
The evaporation of water, oxidation of hydrocarbons and decomposition of other contaminants such
as nitrates also results in micropore formation during heat treatment [38]. These pores are the result
of gaseous molecules escaping from inside the coating and should be closed during heat treatment
by sintering to obtain a densified coating mostly free of porosity.
However, due to the formation of these pores, the substrate can be exposed to oxygen during
thermal treatment, resulting in unwanted oxidation. Therefore, heat treatment should be performed
in an inert atmosphere.
2.4.2 Transformation and kinetics ALD/rCVD coatings
The transformation sequence of coatings resulting from ALD/rCVD is unfortunately not as well
understood and two possible sequences can be found in literature. Most coatings originating from
CVD or CVD related processes deposit κ-alumina when performed at a temperature above 600 ᵒC,
which transforms directly to α-alumina around 1100 ᵒC [23]. This is however very dependent on the
substrate, as Andersson et al. could produce α-Al2O3 directly on a chromia substrate [71].
In low-temperature (150 ᵒC) CVD experiments with reactive magnetron sputtering, the γ phase was
produced directly however [72], which transformed directly to α-alumina as well. Other researchers
have investigated alumina ALD coating crystallization, but did not report crystal structure [73], but
based on temperature the α structure would be most likely. According to Levin et al. [23], most
amorphous alumina films crystallize in the γ phase however and follow the same transition path as
the sol-gel coatings. Based on this information, it is therefore more likely that γ-alumina is more likely
19
as a transition phase. However, all authors find that transforming ALD or CVD coatings to α-alumina
requires very high temperatures in excess of 1100 ᵒC and therefore a high activation energy, even in
cases with a coating thickness in excess of 1 µm [72]. Because of this, grains that do form will likely
remain smaller than those resulting from sol-gel when subjected to the same heat treatment.
ALD coatings suffer from some of the same problems as sol-gel coatings, in particular the volume
change associated with the transformation of amorphous to crystalline coatings. However, although
some impurities in the form of H2O and hydrocarbons are present, their concentration is far lower
than for sol-gel coatings, even for room temperature coatings. This prevents the formation of pores
during heat treatment and reduces the need for further densification during heat treatment. The
formation of porosity can however act as a stress relieving mechanism and ALD coatings could
therefore have higher stresses present in the coating during heat treatment [38], making it more
likely cracks will form.
2.5 Crack formation and healing in YSZ
Similar to most other ceramics, crack formation and failure in YSZ occurs in a brittle manner. In
essence, brittle fracture occurs when the stress is sufficient to break the bonds between atoms in a
solid material. However, without a stress concentration mechanism, this stress would be extremely
high [74]. For ceramics, this stress concentration is usually at the tip of flaws present in the material.
Due to a lack of stress relief mechanisms such as the formation of dislocations and other plastic
deformation mechanisms, cracks can grow relatively rapidly and with little obstruction [75], which is
why their behaviour is described as brittle. This is especially true for ceramics loaded in tension.
Due to the absence of significant plastic deformation, ceramics can usually be described by linear
elastic fracture mechanics. Therefore, the system can be described with a stress intensity factor,
which is defined in equation 2.10 in which KI is the stress intensity factor, σy the far-field applied
stress, a the crack length and f(ϕ) a dimensionless parameter correcting for crack and loading
geometries and angle of loading. As is evident from this equation, fracture strength is not an intrinsic
property of materials, but is dependent on the critical stress intensity factor or fracture toughness KIC
of the material and the flaw size and geometry of the system [75].
(2.10)
Although this is a useful description for well-understood systems, TBCs are very complex and are
loaded in multiple directions [9]. Furthermore, this description breaks down for flaw sizes that are
too small, making it difficult to describe the onset of fracture well [74]. However, a study from Hille
et al. [13] showed that cracking starts in the TGO as a result of thermal cycling and slow TGO growth
due to further oxidation of the Al reservoir in the bond coat. Stresses and crack growth seem to
increase in severity with a higher roughness of the interface between the TGO and the TBC. Upon
growing sufficiently, they will grow into the TBC and follow the pattern described in chapter 2.1.1 of
growth, coalescence, perpendicular growth and subsequent spallation. Arresting this crack growth by
self-healing mainly aims to reduce the crack size to slow growth and fill the crack to remove it. As the
ZrSiO4 is tougher than the YSZ in the TBC, new cracks will have a tendency to grow around the healed
area instead of through it.
20
Thermodynamics and Diffusion
This chapter will introduce the main thermodynamic and kinetic considerations of the self-healing
TBC system. Because of the high temperatures involved in this system, transport of matter is
relatively rapid and thermodynamic considerations become significantly more important than at
lower temperatures. Therefore, as the original system is not the thermodynamically most stable
system in an environment containing oxygen, the system will evolve towards a more stable system.
This chapter will therefore start with the kinetics of MoSi2 oxidation, followed by a thermodynamic
analysis of the interfaces between the different materials and the most likely evolution of the
system. This will be followed by a description of diffusion at high temperatures through a coating and
the construction of a diffusion model of the coated particle system. This model is then utilized to
predict relevant time scales of particle stability.
3.1 Thermodynamics and kinetics of the self-healing TBC system
3.1.1 Oxygen and oxidation behaviour of MoSi2
Owing to its use as heating elements in many high-temperature furnaces, among other high
temperature applications, numerous studies on the high-temperature oxidation of MoSi2 have been
performed [76], [77], [21]. This oxidation behaviour is rather complex, mainly resulting from the two
oxidizable components present in the system, molybdenum and silicon. Oxidation of MoSi2 starts
between 400 and 500 ᵒC and follows reaction 3.1 at temperatures lower than 800 ᵒC.
2 MoSi2 (s) + 7 O2 (g) → 2 MoO3 (s) + 4 SiO2 (s)
(<800 ᵒC)
(3.1)
However, thermodynamic calculations and experimental evidence by Zhu et al. [76] found that above
approximately 800 ᵒC (the melting point of MoO3), the formation of Mo5Si3 and SiO2 according to
reaction 3.2 is thermodynamically more favourable than the formation of MoO3. If sufficient
additional oxygen is present, Mo5Si3 can oxidize further according to reaction 3.3. However, at this
temperature, the vapour pressure of MoO3 becomes significant, removing most MoO3 that could still
form. Therefore, 800 ᵒC is also approximately the temperature at which a closed scale of SiO2 starts
to form, protecting the underlying MoSi2 and Mo5Si3.
5 MoSi2 (s) + 7 O2 (g) → Mo5Si3 (s) + 7 SiO2 (s)
Mo5Si3 (s) + 10.5 O2 (g) → 5 MoO3 (g) + 3 SiO2 (s)
(>800 ᵒC)
(>800 ᵒC)
(3.2)
(3.3)
Below 800 ᵒC, significant MoO3 formation introduces porosity in the formed scale, a phenomenon
often referred to as MoSi2 pest oxidation, as this porosity prevents the formation of a protective
coating. This allows the oxidation of MoSi2 to continue at high rates if no initial protective coating is
present and is a significant issue in bulk MoSi2 applications. However, as mentioned before in the
theory section, particles do need a protective coating in any case, due to the required thickness of
the SiO2 coatings being several µm. This would necessitate the consumption of a significant part of
the healing particle to form this coating.
Due to its interesting high-temperature properties and intermetallic nature, attempts at alloying
MoSi2 have been performed as well [78]. Two of the more interesting elements are boron and
aluminium. As mentioned before, Mao found that boron is able to stabilize the amorphous phase of
SiO2. However, boron is not very soluble in MoSi2 and tends to form separate phases with
21
molybdenum, as is evident from the phase diagram presented in Figure 3.17. Boron does not seem to
form any borides with silicon in the presence of molybdenum though [79]. This indicates that boron
addition will most likely form a separate phase with molybdenum only and which one will depend
mainly on the processing conditions of the MoSi2 production. Because most molybdenum borides
have a remarkably high hardness, boron is often added to MoSi2 to increase hardness [80].
Figure 3.17: The temperature dependent Mo-B phase diagram [81].
Aluminium on the other hand can easily substitute for silicon in the MoSi2 structure, although it is
known to stabilize the usually metastable hexagonal phase of MoSi2 [78]. This results in either a
single phase of MoSixAly or a two-phase system with both the hexagonal and tetragonal phase of
MoSixAly coexisting, depending on the molar ratios of the elements present. The main effect of the
presence of aluminium is however its effect on oxidation behaviour. Because the ΔG of Al2O3 per
mole of oxygen is significantly lower than that of either Mo or Si, aluminium is preferentially oxidized
and is even able to reduce SiO2 according to reaction 3.4 [82]. This limits pest oxidation and results in
an Al2O3 scale instead of an SiO2 scale, although some SiO2 can still form, depending on the local Si, Al
and O activities.
4 Al (s) + 3 SiO2 (s) → 2 Al2O3 (s) + 3 Si (s)
(3.4)
The oxidation of components in MoSi2 and subsequent reactions to form ternary oxides do result in
significant changes in molar volume though. Although this is desired for the self-healing process, as it
will aid in filling and closing cracks, premature oxidation will also result in accumulation of stress, as
described in the theory section. This could result in coating fracture or damage and is therefore
important to prevent. The presence of all these oxides also results in many possible interfaces, which
will be discussed in the next parts.
3.1.2 Alumina/YSZ
The first interface and one of the more critical ones for particle opening is that of the alumina
particle shell and the YSZ matrix. As the latter consists of both ZrO2 and Y2O3, phase diagrams of Al2O3
with both oxides are presented in Figure 3.18 and Figure 3.19 respectively. As is clear from the first
diagram, ZrO2 and Al2O3 do not form any thermodynamically stable ternary compounds.
22
Furthermore, solubility between the oxides is limited, even being practically nonexistent for ZrO2 in
alumina.
For Al2O3 and Y2O3, solubility is very similar, also being practically non-existent for both compounds.
However, alumina and yttria are able to form some ternary compounds at higher temperature. Of
these, only Y3Al5O12 and YAlO3, the famous yttrium aluminium granate (YAG) and yttrium aluminium
perovskite (YAP) are thermodynamically stable at lower temperatures, but other compounds can
form at approximately 1100 ᵒC, which is still close to TBC temperatures.
Figure 3.18: The temperature dependent Al2O3-ZrO2
phase diagram [83].
Figure 3.19: The temperature dependent Al2O3-Y2O3 phase
diagram [84].
It should be noted though that yttria is present in YSZ in this case, not in its pure form. For YSZ and
Al2O3 Lakiza et al. did show that ternary and quaternary compounds are indeed able to form and
thermodynamically stable though [83]. This could be problematic for thin shells protecting the MoSi2
particles.
However, studies on the YSZ-Al2O3 interface in TGO-TBC coatings by Suenaga et al. [85] found that
this interface is easily detectable on a nanometer scale, with pure alumina and pure YSZ being
separated by an interface of only 5-6 nm. This indicates that YSZ and alumina are relatively stable in
contact with each other and even though it is thermodynamically possible to form ternary and
quaternary compounds, kinetics for this seem to be relatively slow.
3.1.3 Alumina/MoSi2
Another important interface is that between the shell and the healing agent. Tests with MoSi2
including Al as an alloying element, mentioned in the oxidation section, indicate that although
aluminium is soluble in the healing agent, its oxide is not soluble at all [78]. This is mainly evident by
the segregation of Al2O3 in the bulk of their material, even in argon atmosphere.
Furthermore, no thermodynamically stable compounds of Al2O3 and either Si or Mo in reduced form
are known. However, as already described in the theory section, MoSi2 at its interface does not
contain reduced Mo or Si, but most likely a thin native oxide layer of SiO2. As will be described in the
23
next section, Al2O3 and SiO2 do have a stable ternary compound and are able to form a strong
interface. Therefore, it is likely that the interface between the shell and healing agent will consist
mainly of this Al2O3-SiO2 interface.
3.1.4 Alumina/Silica
The expected interface between the healing material and alumina is most likely based on SiO2 and
Al2O3. The phase diagram of these two oxides is shown in Figure 3.20 and shows one stable ternary
compound, namely mullite (Al6Si2O13), which is also thermodynamically stable at room temperature,
even though not shown here. Mullite is an important engineering ceramic with many applications, an
interesting one being in porcelain, as it is the compound that gives this material its strength.
Because of this, it is a well-studied ceramic and according to many studies, very likely to form at any
interface between Al2O3 and SiO2 at sufficient temperatures, usually around 1000 or 1100 ᵒC in bulk
materials [86]. This is also an explanation for the often strong interfaces found between alumina and
silicon containing materials that have an SiO2 native oxide layer. Many other aluminosilicates exist
though and are likely to form, although these other aluminosilicates often require counterions to
balance charges and the only known thermodynamically stable phase is mullite.
Figure 3.20: The temperature dependent Al2O3-SiO2 phase diagram [87].
3.1.5 Silica/YSZ
The final interface that is of significant importance is that between SiO2 and YSZ, as this is the
interface that is most important for the healing behaviour. A ternary oxide phase diagram of SiO 2,
ZrO2 and Y2O3 is not available, but phase diagrams of SiO2 with the individual oxides are known and
shown in Figure 3.21 and Figure 3.22. the ZrO2-SiO2 phase diagram shows that there is only one
ternary compound possible, which is zircon (ZrSiO4). As described in the theory section, there is only
one stable crystal structure for zircon as well. For yttria and silica, there are two different possible
phases however, Y2SiO5 and Y2Si2O7, with the latter having four different known polymorphs.
24
Figure 3.21: The temperature dependent SiO2-ZrO2 phase
diagram [88]
Figure 3.22 The calculated temperature dependent Y2O3SiO2 phase diagram [89].
Which of these phases forms preferentially is not known, but experiments on the healing mechanism
by Mao [17] show that the formation of zircon is dominant in the SiO2-YSZ system however. The
presence of quaternary compounds was not discussed however and whether this system is able to
form these is not known.
3.1.6 Coated particle system evolution
After discussion of the oxidation mechanism of MoSi2 and the most likely possible interfaces, it is
important to note that at temperatures present in the TBC system, which can reach to 1100 ᵒC, the
coated particle system will change over time. As is evident from the TGO, even an Al2O3 coating of
several µm thick cannot prevent oxygen diffusion and oxidation of the substrate, only slow it down.
Therefore it is important to understand how the system will most likely evolve over time, based on
thermodynamics and diffusion. The idealized starting situation is therefore graphically depicted in
Figure 3.23, while the Gibbs free energy of formation (ΔGf) and equilibrium partial oxygen pressure
calculated with eq 3.3 at 1000 ᵒC of each compound present in the system is listed in Table 3.2. In
this equation, PO2, eq is the equilibrium partial pressure of oxygen, or the PO2 at which the oxide and
metal are in equilibrium, while Kp is the equilibrium constant for the oxidation reaction. ΔGf is the
Gibbs free energy of formation in J/mol O2, R is the gas constant and T the temperature in K.
(3.5)
Table 3.2: Standard free energy of formation at 1000 ᵒC for each oxide present in the system from its element, per mole
of oxygen consumed, along with the equilibrium partial oxygen pressure.
Compound
SiO2
MoO3 (g)
Al2O3
ZrO2
Y2O3
B2O3
Standard free energy (ΔGf, kJ/mol O2)
-677
-189
-845
-863
-1023
-641
Equilibrium PO2 (bar)
1.68E-28
1.74E-08
2.13E-35
3.92E-36
1.04E-42
5.08E-27
25
As can be seen from both the Gibbs free energy and the equilibrium PO2, the most stable oxides are
Y2O3 and ZrO2, the latter closely followed by Al2O3. It is therefore clear that MoSi2 both with and
without boron is not able to reduce these oxides. This table also shows that of the compounds
present in the MoSi2, silicon will likely oxidize before molybdenum, which agrees with the discussion
in the first part of this chapter. However, it will most likely also oxidize before boron, although the
difference in free energy is significantly smaller. This does however depend on the energy of
formation of MoSi2 and the molybdenum-boron compound that is present in the MoSi2, the latter of
which is unfortunately not known.
Figure 3.23: A schematic representation of the evolution of the coated MoSi2 system at high temperatures in an oxygenrich environment. In this system, it is assumed that yttria and zirconia are not able to diffuse through alumina and that
molybdenum will not oxidize.
3.2 Diffusion of species in the coated healing particles
Diffusion in ceramics is an often complex process due to the complexity of diffusion in solids in
general and the influence of atomic charges and charge separation. One of the main differences of
diffusion in solids compared to liquids or gases is the presence of all kinds of defects that can act as a
fast pathway for diffusion. Furthermore, another important aspect of diffusion of charged species is
that charge neutrality should be maintained, which influences the speed of diffusion of all species
involved. Therefore, this aspect of bulk diffusion will be described first, followed by defect diffusion.
Afterwards, the diffusion model with its main assumptions will be introduced and the results from
this model will be discussed.
3.2.1 Bulk diffusion
Bulk diffusion in ceramics in general is governed by movement of vacancies, electrons and/or holes.
Sometimes, especially with diffusion of elements that are not part of the matrix and relatively small,
these elements occupy interstitial positions and are able to move along these interstitial positions.
Nevertheless, all these forms of diffusion proceed by the means of point defects, which are always
present in ceramics, because it is thermodynamically favourable to form some defects such as
Schottky, Frenkel and anti-Frenkel defects. As the transport of atoms is governed by both the
concentration and mobility of these point defects and defect concentrations in oxides are also often
dependent on the partial oxygen pressure of the gas in contact with the ceramic, it is important to
understand the defect structure in the oxides present.
However, another important aspect is that this transport is driven by a difference in chemical
potential, as the potential of the oxygen at the MoSi2-SiO2 interface is significantly more negative
26
than the chemical potential at the outer surface of the Al2O3 layer. These systems are different from
systems with only self-diffusion being present in that the chemical potential gradient accelerates
movement of ions in the direction in which the chemical potential is lower. In most systems this
diffusion under the influence of a gradient is described by Fick's first law, but this law does not
account for charged species.
To account for charged species, there are two important models to describe oxidation of metals with
resulting film growth, the Cabrera-Mott model for very thin films and the Wagner model for thicker
films. The former assumes that electrons can pass freely from the metal surface to the oxide-gas
interface. This results in an uniform electrical field in the oxide, but restricts the applicability of this
model to very thin films in which most of the film is charged and electrons can essentially tunnel
from one side to the other. Although extensions have been developed for thicker films, this theory is
usually only applied for films of at most several 10s of nanometers [90].
The Wagner model on the other hand assumes that oxidation is mainly governed by diffusion of
charged particles in an oxide that is considered to be mostly neutral, which is usually the case for
oxides thicker than the Debye length [90]. Therefore space charges are neglected in this model and
any movement of ions must be balanced by either other ions or electrons and holes to retain charge
neutrality. The main equation for the Wagner model is shown in equation 3.6.
(3.6)
In which Ji is the flux of species i in mol/(m2*s), Di the diffusion coefficient in the oxide in m2/s, C the
concentration of the respective defects in mol/m3, x the distance in m, µ the mobility and E the
electric field. As the Wagner model assumes charge neutrality and that the system is close to local
equilibrium everywhere in the oxide layer to use equilibrium thermodynamics, equation 3.6 can be
simplified to equation 3.8 with the aid of equation 3.7.
(3.7)
(3.8)
In which D, J and x refer to the same property, while R is the gas constant in J/(K*mol), T the absolute
temperature in K, Ci is in this case the concentration of the species in the bulk material in mol/m 3
and due to unhelpful use of µ for multiple different properties in literature, here µ refers to the
chemical potential in J/mol. If most of oxide layer growth is dependent on the diffusion of only one
element, equation 3.9 can then be used to calculate layer growth. By assuming chemical potential
varies linearly over the oxide layer according to equation 3.10, equation 3.11 can be derived after
integration of equation 3.9 for a single oxide layer.
(3.9)
(3.10)
(3.11)
27
With dox the oxide thickness in m, fi, MOx the amount of moles of the diffusing species needed to form
one mole of the oxide MOx, MMOx the molar weight of the oxide in g/mol, ρMOx the density of the
oxide in g/m3 and t the time in s. Equation 3.11 is the famous parabolic growth rate law for an oxide
scale on a substrate and describes the thickness of the oxide as a function of time. However, the
behaviour of the three different oxides is slightly different and should be discussed individually,
especially with regard to the diffusion coefficient and chemical potential.
Starting with alumina, which is one of the most studied materials regarding diffusion and therefore
one of the most poorly understood materials. Or to quote the answer of A. Heuer on the question
"Oxygen and aluminium diffusion in α-Al2O3: How much do we really understand?": "Not a great
deal" [91]. The diffusion of most species in α-alumina is not very well understood, even to the extent
that even the main defects responsible for this diffusion are still debated. It is however known that
the self-diffusion coefficient Dox < DAl under practically all conditions and at all temperatures in bulk
diffusion [91]. This indicates that in defect-free materials, transport of Al is mainly responsible for
oxidation at higher temperatures.
For mullite on the other hand, this is quite different, as was found by Fielitz et al. [92]. In their
investigations on the formation of mullite from its constituent oxides Al2O3 and SiO2, they found that
aluminium and oxygen had very similar bulk diffusivities, while the Si diffusivity was several orders of
magnitude lower. According to the authors, this can be attributed to the strong covalent bonds
formed by SiO4 tetrahedra, which limits movement of Si atoms.
This same effect is also observed in SiO2 in which oxygen is again the faster-diffusing species [93], at
least in amorphous silica, which is the most likely species to be present. The author explains this
effect partially with strong covalent bonds between SiO4 tetrahedra and the presence of many O
atoms that are only bonded to a single Si atom (which is common in amorphous SiO 2). Another
important effect however is the movement of gaseous/dissolved O2 in the system that can move
significantly faster than network oxygen.
3.2.2 Defect diffusion
As mentioned before, an important difference between diffusion in solids and fluids is the presence
of defects in the former. These defects can exist in many forms and are usually described based on
the amount of dimensions occupied by these defects. 0D or point defects have already been
discussed in the previous part as they are the defects responsible for lattice diffusion.
1D or line defects can also accommodate diffusion along a line. The only example of this type of
defect is a dislocation, which can accelerate diffusion along the dislocation. Although diffusion along
dislocations can be relevant for metals with high dislocation densities, for brittle ceramics,
dislocations are very complex and difficult to form. Therefore dislocation density is very low and the
influence of diffusion along this type of defect can be neglected in this system.
2D or interface defects on the other hand are significantly more important and are one of two types.
The first type of interface is a free surface, which is a gas-solid interface. This surface can either refer
to an external surface or the surface of pores or cracks present in the material. It is often very rapid
compared to bulk diffusion due to the significantly lower activation energy barrier, as no vacancies
have to be formed in this type of diffusion.
28
Another important type of interface defect is the grain boundary, which is any interface between two
crystallites with usually a different crystal orientation. For low angle grain boundaries (having a
misorientation between crystallites < approximately 15ᵒ), this boundary consists of a row of regularly
spaced dislocations, but for high angle grain boundaries, the boundary is both poorly defined and
poorly understood. However, the atomic density is significantly lower in both grain boundaries,
making the formation of defects and movement of atoms far easier [94]. Therefore, grain boundaries
can and will often act as a fast pathway for diffusion, especially at intermediate temperatures.
Finally, the 3D defects, which are either be cracks, pores, any other type of gas-containing area in the
ceramic or another phase included in the original phase. The first type of 3D defect, which is a
gaseous inclusion in the solid material, can easily act as a fast pathway for O2, as diffusion in gases is
significantly faster than in either liquids or solids. Depending on the flaw size, this diffusion could be
Knudsen diffusion in which the molecule mean free path is larger than the width of the flaw andgases
interact significantly with the flaw walls. If the mean free path is small compared to the flaw size
though, regular gaseous diffusion is the main mechanism. However, due to the high temperatures,
the mean free path is generally in the same order of magnitude as the coating thickness (λ=350 nm
at 1000 ᵒC and 1 bar total pressure) and therefore, any diffusion of this type in the coated particle
system is most likely Knudsen diffusion.
Diffusion in foreign phases is also possible if impurities are present. This type of diffusion will only act
as a fast pathway if the diffusivity of the cation and oxygen are higher than in the matrix though.
However, most impurities will have a higher diffusivity than Al2O3 though, as diffusion in alumina is
relatively slow, as described in appendix I.
Calculation of diffusion in systems with fast diffusion pathways is often carried out with the use of an
effective diffusion coefficient for the whole system, Deff. These are generally sums of the diffusion
coefficients in each of the defects with the volume fraction of that defect in the bulk material [90].
For the materials involved in the self-healing TBC system, Al2O3, SiO2 and Al6Si2O13 are the most
important and will most likely be the most affected by any fast diffusion pathways. Therefore it is
important to know how likely these fast pathways are to be present. As mentioned before, diffusion
along dislocations can be neglected in ceramics. For SiO2, grown from MoSi2, 3D defects are not likely
to be present as long as only SiO2 is oxidizing, as thermally grown oxides generally have low porosity
and crack formation as long as they remain sufficiently thin [25]. Furthermore, grain boundaries are
obviously not present in an amorphous phase and other interfaces will most likely not form in the
direction of diffusion.
For α-Al2O3 on the other hand, the formation of 3D defects is very dependent on the manufacturing
method of the coating and subsequent processing. This also influences the presence of free surfaces
significantly. However, the main defect that is almost unavoidable in alumina is the formation of
grain boundaries that can act as a fast pathway in diffusion.
Although even less understood than bulk diffusion in alumina, many investigations on the grain
boundary diffusion of especially oxygen are available, a good example of which is the work done by
Smialek et al. [95, 96]. His work shows that grain boundary diffusion of both oxygen and aluminium is
orders of magnitude faster than bulk diffusion and surprisingly, that oxygen grain boundary diffusion
is actually significantly faster than aluminium grain boundary diffusion. Therefore, it is very important
29
to consider grain boundary diffusion for materials with small grains, which is likely to be the case in
the coatings investigated here.
For mullite, grain boundary diffusion has not been investigated in great detail, but work performed
by Fielitz et al. [92] found that grain boundary diffusion of Si is still significantly slower than Al and O
grain boundary diffusion. All diffusivities were however found to be at most 4 orders of magnitude
higher (depending mainly on temperature, but also other factors), which is a significantly lower
increase than that of Al2O3 when comparing grain boundary and bulk diffusion. The presence of other
defects is however not very likely, as formation of mullite is mainly due to Al and O diffusion and
therefore the morphology is most likely more similar to that of the SiO2 layer, lacking significant
amounts of 3D effects. The only defect that would be likely to form is a crack, due to stress buildup
from oxidation.
3.2.3 Diffusion model
Because of the complexity of diffusion in solid materials, some assumptions are necessary to create a
diffusion model. In this model, the evolution of the system will be that described in section 3.1.6,
with oxygen diffusing inwards through the Al2O3, mullite and SiO2 layers and oxidizing MoSi2
according to reaction 3.2. The formation of all of these layers is assumed to be completely diffusion
limited with no external mass transfer or reaction limitations and the activity of Si at the MoSi2-SiO2
interface is assumed to remain sufficient to prevent any oxidation of molybdenum. Furthermore, as
described in section 3.1.2, the interface between alumina and YSZ is very sharp, indicating very low
diffusivity of Y and Zr in alumina. Due to this, it is assumed that diffusion of these into the alumina is
negligible and can be ignored.
Due to the thick Al2O3 coating already present, it is likely that parabolic oxide growth rate law kinetics
of MoSi2 as described in the Wagner model will indeed be valid and that diffusion through the oxide
layer is indeed the rate limiting step throughout the whole experiment. Furthermore, investigations
of oxidation of MoSi2 indicate that the diffusion of Si in MoSi2 and Mo5Si3 is indeed significantly faster
than the diffusion of oxygen through Al2O3, mullite or SiO2 [97], supporting the validity of this
assumption.
Furthermore, it is assumed that mullite is only formed by the inward diffusion of aluminium and
oxygen through the mullite layer, as described in the work of Fielitz [92]. To model the diffusion, it is
assumed that in this case the Wagner model with its assumptions of charge neutrality and local
equilibrium can be used to describe the formation of all three layers. Although this is possible
without question for the Al2O3 layer, the thickness of the mullite layer and SiO2 layer might initially be
too small to be described properly by the Wagner model. However, after a certain period of time,
these layers will have grown sufficiently to ignore space charges and keep assumptions of the
Wagner model valid and it would simplify the model significantly. Using equation 3.8 and the
assumption of a linear chemical potential presented in equation 3.10, this would result in equation
3.12 for each chemical species in each layer.
(3.12)
It is assumed that only the oxide layer thickness will change with time and the other variables are
only dependent on temperature or remain constant for a certain species and oxide layer. In this
equation, the diffusion coefficient is an effective ambipolar diffusion coefficient taking into account
30
bulk and grain boundary diffusion. It is assumed the layer is perfect and no other defects are present
that could act as a fast pathway for diffusion. To account for temperature, an Arrhenius dependence
is used, which results in equation 3.13 for the effective diffusion coefficient. In this equation, f is the
grain boundary volume fraction usually given by equation 3.14.
(3.13)
(3.14)
In which D0 is the pre-exponential factor in m2/s and ΔGact the activation energy in J/mol, δ the grain
boundary width in m (usually taken as 1 nm for diffusion purposes [95]) and g the average grain size
in m. As for most temperatures the grain boundary diffusion coefficient is at least 10 4 times higher
than the bulk diffusion coefficient [95], [92], [91], equation 3.14 indicates that grains should be at
least 20 µm to have equal contributions of bulk and grain boundary diffusion. Because of the coating
thickness of less than 1 µm, it is likely that they are in general significantly smaller. It is therefore
reasonable to assume bulk diffusion in alumina and mullite is negligible and that oxidation is
governed solely by grain boundary diffusion of oxygen. As amorphous SiO2 does not have any grain
boundaries, the bulk diffusion should be used here.
For the change in chemical potential, equation 3.15 indicates how chemical potential changes with
reactions and differences in partial oxygen pressure. In this equation, µO is the chemical potential of
oxygen atoms in the system, while µO20 is the standard chemical potential of oxygen molecules, both
in J/mol. R and T have their usual meaning and PO2 is the partial oxygen pressure at that location. For
the formation of mullite, it is assumed that the ΔµAl can be entirely attributed to the Gibbs free
energy of reaction of the formation of mullite from the oxides, which would result in equation 3.16
(3.15)
(3.16)
As the system is diffusion limited, a quasi-steady state regime can be assumed in which the pressure
at the gas-Al2O3 interface is the PO2 pressure of the bulk gas and the PO2 pressure at the SiO2-MoSi2
interface the equilibrium partial oxygen pressure of equation 3.2. Although not entirely correct, it is
assumed that the chemical potential is linear over all three oxide layers without any discontinuities at
the interfaces, because data of the latter is unfortunately not available. To simplify the model and to
allow the evaluation of partial oxygen pressures at the interfaces, it is also assumed that diffusion of
oxygen for mullite formation is independent from diffusion due to oxidation and that the charge
balance of the latter is solely compensated for by electron and hole movement, which is reasonable
for very thin layers. The pressures at the interfaces on both sides of the mullite layer can then be
calculated using the equations 3.18 and 3.19, in which the ki's are given by equation 3.17.
(3.17)
31
(3.18)
(3.19)
In which P0 is the partial oxygen pressure at the gas-alumina interface, P1 is the partial oxygen
pressure at the Al2o3-mullite interface, P2 the partial oxygen pressure at the mullite-SiO2 interface
and P3 the equilibrium partial pressure of equation 3.2, all in bar. These equations are based on a
mass balance of oxygen in which any oxygen that diffuses through alumina also has to diffuse
through mullite and silica. Combining equation 3.9 and 3.12 and using equation 3.15 and 3.16 to
calculate the chemical potential, this results in the following equations for the change of thickness in
each layer as a function of time.
(3.19)
(3.20)
-
(3.21)
In which the diffusion coefficients, the mullite Gibbs free energy of formation and the equilibrium
partial pressure of oxygen are functions of temperature. The values used for all parameters in the
evaluation of this model are presented in the matlab code The evaluation of this model was done
numerically using the Matlab script presented in appendix III and the results are presented in the
next part.
3.2.4 Results model
The main predictions of this model are shown in Figure 3.24, Figure 3.25, Figure 3.26 and Figure 3.27
and are somewhat surprising. From the first figure, it is clear that silica scale growth is very rapid and
follows an approximately parabolic growth rate, but that mullite growth is significantly slower. The
latter is expected though, as the diffusion of both oxygen and aluminium through mullite is generally
considered to be very slow.
When investigating the effect of the main parameters temperature, partial oxygen pressure and
grain size, it is clear that temperature has a significant influence. However, the actual influence of
temperature seems smaller than would be expected, based on an Arrhenius relation. However, the
formation of mullite is significantly increased at higher temperature due to accelerated aluminium
diffusion, decreasing the total oxygen diffusion through the system significantly. This might be a very
important prediction as particle lifetime could be significantly prolonged by promoting mullite
formation during manufacture of the coating.
32
-7
6
-6
x 10
1.4
x 10
1.2
5
1
4
d (m)
d (m)
0.8
3
Alumina thickness
Mullite thickness
Silica thickness
equivalent Silica thickness
2
0.6
0.4
1
0
0.2
0
1
2
3
4
5
6
7
8
t (s)
0
9
Figure 3.24: Example of the thickness of each layer after
24 hours at 1000 ᵒC.
1
2
3
4
5
6
7
8
t (s)
9
4
x 10
Figure 3.25: The influence of temperature on total oxidation
(equivalent silica thickness) of the system after 24 hours for
900 ᵒC to 1200 ᵒC with 25 ᵒC increments with the highest
temperature having the highest oxidation rate.
-7
-7
6
0
4
x 10
x 10
5
x 10
4.5
5
4
3.5
4
d (m)
d (m)
3
3
2.5
2
2
1.5
1
1
0.5
0
0
0
1
2
3
4
5
t (s)
6
7
8
9
4
x 10
Figure 3.26: The effect of partial oxygen pressure on total
oxidation after 24 hours at 1000 ᵒC, with a partial
-14
pressure varied from 10 to 1 bar in power of 10
increments with the highest partial oxygen pressure
having the highest oxidation rate.
0
1
2
3
4
5
t (s)
6
7
8
9
4
x 10
Figure 3.27: The influence of alumina and mullite grain size
on total oxidation of the system after 24 hours at 1000 ᵒC
with grain size varied from 50 to 500 nm with 50 nm
increments with the smallest grain size having the highest
oxidation rate.
The effect of partial oxygen pressure on the other hand is extremely small, as the difference in final
coating thickness between 10-14 and 1 bar of partial oxygen pressure is at most 150 nm, which is still
less than 30% of the final coating thickness at 1 bar. At partial pressures that low, it is however very
likely that external transport from the bulk to the surface might become rate limiting instead,
lowering the final oxide layer thickness that would be found. Finally, grain size also has a minor
influence on the oxidation kinetics, slightly lowering the total oxide thickness layer with larger grain
size, as would be expected. However, the difference is relatively small compared to temperature.
To investigate the sensitivity of this model towards initial coating thickness, initial Al2O3, mullite and
SiO2 coating thickness were calculated, as shown in Figure 3.28, Figure 3.29 and Figure 3.30
respectively. This shows that both Al2O3 and SiO2 initial coating thickness have a surprisingly minor
influence on overall oxidation after 24 hours.
33
-7
-7
x 10
6
x 10
5
4.5
5
4
3.5
4
d (m)
d (m)
3
3
2.5
2
2
1.5
1
1
0.5
0
0
1
2
3
4
5
6
7
8
t (s)
0
9
0
Figure 3.28: The influence of the initial alumina layer
coating thickness, with coating thickness ranging from 10
to 1000 nanometers and the highest thickness having the
lowest oxidation rate.
4
5
6
7
8
9
4
x 10
-5
x 10
1
0.9
4
0.8
3.5
0.7
3
0.6
d (m)
d (m)
3
Figure 3.29: The influence of the initial mullite layer coating
thickness, with coating thickness ranging from 5 to 50
nanometers and the highest initial thickness having the
lowest oxidation rate.
4.5
2.5
0.4
1.5
0.3
1
0.2
0.5
x 10
0.5
2
0
2
t (s)
-7
5
1
4
x 10
0.1
0
1
2
3
4
5
t (s)
6
7
8
9
4
x 10
Figure 3.30: The influence of the initial SiO2 layer coating
thickness, with coating thickness ranging from 5 to 50
nanometers and the highest initial thickness having the
lowest oxidation rate.
0
0
0.5
1
1.5
2
t (s)
2.5
3
3.5
7
x 10
Figure 3.31: Example of the thickness of each layer after
one year at 1000 ᵒC.
A likely explanation for the somewhat surprising lack of influence of Al2O3 coating thickness might be
the fast grain boundary diffusion in combination with small grains. Due to this, the effective oxygen
diffusion coefficient is several orders of magnitude higher than the effective diffusion coefficient in
either mullite or silica. This results in a significantly lower resistance to oxidation of the substrate and
the influence of Al2O3 coating thickness being very small.
Initial mullite coating thickness on the other hand does influence the overall oxidation speed, which
agrees with literature and other calculations showing the low diffusivity of oxygen in mullite. Because
of this, it might be useful to promote the formation of mullite by either alloying MoSi2 with additional
Al or additional/longer heat treatment at higher temperatures with low oxygen partial pressures.
Finally, the effect of increasing time (assuming the oxidation layer does not fracture) is essentially the
same as for most parabolic growth rate laws. At 1000 ᵒC the total oxide thickness would grow close
to 10 micron, indicating that most of the particle would be oxidized at this point.
34
Materials and Methods
In this section, a concise description of the methods used to synthesize, characterize and test coated
molybdenum disilicide (MoSi2) particles is presented. The chapter starts with a brief description of
the wind sifting technique required to obtain the right size range of particles, followed by the various
methods of coating synthesis, namely three different sol-gel procedures and (Atomic) Layer
Deposition with residual Chemical Vapour Deposition (ALD/rCVD).
Subsequently, the methods of thermal treatment and determination of the optimal calcination and
transformation are discussed. To obtain information about the material and coating properties of the
particles, extensive characterization and performance testing was employed. The techniques used for
the particle characterization are also presented.
4.1
Wind sifting
Wind sifting was performed prior to the coating process molybdenum disilicide powders to reduce
polydispersity in the size distribution and decrease total surface area of the particles. This was
necessary to remove particles that were considered too fine for incorporation in the final TBC and to
ensure an acceptable volume ratio of coating material to healing material, which is discussed in more
detail in appendix IV. Wind sifting was performed using an Alpine 100 MRZ laboratory zig-zag
classifier, shown in Figure 4.32. In each experiment between 300 and 500 g of MoSi2 powder
(Chempur, D(V,0.5) = 18 μm) was loaded. Airflow was regulated at 15 m3/h and the classifier
rotational speed was kept at 5000 RPM. More information regarding the working principle of the
wind sifting method is available in appendix V.
Centrifugal
zig-zag
columns
Feed
Coarse fraction
collection
Cyclone
Engine
Fine fraction
collection
Figure 4.32: The Alpine 100 MRZ laboratory zig-zag classifier used for wind sifting and its different parts.
4.2
Sol-gel methods
Coating of particles using sol-gel methods was performed using a modified Yoldas procedure, a
method based on alkoxides as precursor for boehmite sols, developed originally by Yoldas et al. [46].
The procedure was modified by Yang and Shih for silicon carbide particles [51], which was used as a
35
starting point for the sol-gel coating of MoSi2 particles, due to the expected similar SiO2 native oxide
layer on the surface of both particles [52].
First, tests with aluminium oxalate (Al2(C2O4)3) were performed for obtaining critical information on
the behaviour of MoSi2 particles in sol-gel experiments, followed by experiments with aluminium
isopropoxide (Al(OCH(CH3)2)3) and aluminium tri-sec-butoxide (Al(OCH(CH3)C2H5)3), the structures of
which are shown in Figure 4.33 a, b and c respectively. All sol-gel experiments were performed in the
setup shown in Figure 4.34, which consisted of a glass beaker mounted on a heating and stirring
plate with thermocouple to regulate temperature and stirring speed. If required, nitrogen gas was
supplied to the reactor by a custom-made ring-shaped bubbler connected to a nitrogen supply, for a
better dispersion of the substrate particles in the suspension.
Figure 4.33: The molecular structures of (a) aluminium oxalate, (b) aluminium tri-isopropoxide and (c) aluminium
tri-sec-butoxide, as provided by Sigma-Aldrich.
N2 flow meter
Thermocouple
Reactor vessel
Bubbler
Heating and stirring plate
Figure 4.34: The setup used for sol-gel experiments with heating, stirring and nitrogen supply.
4.2.1 Aluminium oxalate method
To obtain information regarding the behaviour of MoSi2 particles and the interactions with the sol
and gel during the coating process, aluminium oxalate was employed first, using the procedure
mentioned above and adjusted based on the work of Wei et al. [98] and Saha et al. [99]. First,
aluminium oxalate tetrahydrate was dissolved in a 1.0 M HNO3 solution (prepared from 65 wt%
HNO3 solution, Sigma-Aldrich and distilled H2O) and distilled water was added to obtain desired
concentrations. Subsequently, the suspension was heated to 60 °C and stirred at 800 RPM for 30
minutes to dissolve the aluminium oxalate. Then the MoSi2 particles were added and the
temperature increased to 80 °C to initiate gelation of the aluminium oxalate and kept at this
temperature for 90 minutes while continuing stirring.
36
Afterwards, solid particles and liquid were separated by vacuum filtration and the solid residue was
washed with distilled water three times. Finally, the sample was dried overnight in an autoclave at
110 °C. The weight and volume of the added chemicals were varied significantly for each experiment
and these values can be found in Table 4.3. Furthermore, pregelation of oxalate for 30 minutes at 80
°C and bubbling 200 mL/min of nitrogen through the suspension were both performed in separate
experiments to increase dispersion of the MoSi2 particles.
Table 4.3: Overview of conditions for each aluminium oxalate sample.
Sample
1
1.0 M HNO3
solution
added (mL)
50
2
100
3
4
5
distilled H2O
added (mL)
Al2(C2O4)3
added (g)
20
4.48
MoSi2
powder
added (g)
5
wind sifted
MoSi2 used
N2
bubbling
no
Pregelation
no
0
8.36
10
no
no
no
50
75
9.62
10
no
no
no
50
100
9.62
10
no
yes
no
50
250
9.62
10
yes
no
yes
no
4.2.2 Aluminium tri-isopropoxide method
For the sample based on aluminium tri-isopropoxide (Al(OC3H7)3), 98%, Sigma-Aldrich), 8.467 g
aluminium tri-isopropoxide was dissolved in 200 mL isopropanol (99.8%, Sigma-Aldrich) and 50 mL
ethanol (99.8%, Sigma-Aldrich). 10 g of MoSi2 powder was dispersed in 72 mL MilliQ and 72 mL of
isopropanol by ultrasonication. Subsequently the dispersed powder was added to the dissolved
Al(OC3H7)3 solution, together with 3 mL of 1.0 M HNO3 solution (prepared from 65 wt% HNO3
solution, Sigma-Aldrich and distilled H2O) and the system was heated to 90 °C to gelate for 35
minutes. The solution was cooled down afterwards, left overnight and filtered by vacuum filtration
the next day. Finally, the powder was dried in an autoclave at 110 °C for 35 hours.
4.2.3 Aluminium tri-sec-butoxide method
The samples based on aluminium tri-sec-butoxide (Al(OC4H9)3, 97 wt%, Sigma-Aldrich) were prepared
by heating 300 or 600 mL of distilled H2O, to obtain a molar ratio of 1:150 of Al(OC4H9)3 to H2O and
15 mL of a 1.0 M HNO3 solution to 80 °C, adding 10 g of wind sifted MoSi2 powder to this solution and
then bubbling 200 mL N2/min (99.999% nitrogen) through this suspension. When the temperature
was stable at 80 °C again, either 5, 10 or 20 g of Al(OC4H9)3 was added with approximately 25 mL of
ethanol (96 wt%, Sigma-Aldrich) to aid in transfer of the viscous liquid. The ratio of aluminium tri-secbutoxide to MoSi2 powder was varied to study the influence of solid loading on thickness and
resulting coating quality.
After addition of the precursor, the system was left to gelate for 60 minutes. Afterwards, the
nitrogen flow and stirring were stopped and either left at 80 °C to evaporate all liquid or centrifuged,
in both cases obtaining a paste. When centrifuged, the samples were also washed with distilled
water three times. The resulting paste was then dried in an autoclave at 110 °C for 15 hours and
ground in a mortar to obtain the coated powder.
4.3
Atomic Layer Deposition/residual Chemical Vapour Deposition
Atomic layer deposition with residual chemical vapour deposition (ALD/rCVD) on MoSi2 particles was
performed using a custom-built setup shown in Figure 4.35. This setup consists of a fluidized bed
reactor in which the particles to be coated are fluidized, bubblers containing the precursors and gas
37
cleaning after the reactor to prevent unreacted precursor in the exhaust gases. More information on
the setup and operation is presented in appendix VI.
Cyclone
Exhaust gas
cleaning
Manometer
Glass reactor
column
TMA supply
H2O supply
Supply lines and
connection
Figure 4.35: The ALD setup used in this experiment, showing the whole setup including bubblers and gas
cleaning (left) and the reactor with connections (right).
In each experiment 11 g of wind sifted MoSi2 powder was first dried for 14 hours at 110 °C to remove
physisorbed water that could react with the aluminium precursor. Afterwards, the powder was
activated in a plasma cleaner for 2 minutes and 20 seconds and loaded in the glass reactor column.
To obtain gaseous precursors, a nitrogen gas flow (99.999% purity) of 0.6 L/min (yielding a superficial
gas velocity of approximately 0.3 m/s) was passed through bubblers containing the liquid precursors
trimethylaluminium (TMA, AKZO Nobel, semiconductor grade) and H2O (MilliQ), to evaporate and
transport the precursors to the reactor. To ensure proper fluidization, vibration at 50 Hz of the entire
reactor was utilized. To prevent undesired reactions of TMA with water or oxygen, the whole reactor
system was kept under pure nitrogen and at room temperature.
By variation of TMA and water dosage time, purge times and total surface area of the particles to be
coated, the layer thickness increase (or layer growth) per cycle can be controlled, allowing for very
precise control of final coating thickness. To find optimal conditions for rapid growth, independent
variation of TMA dosage time, water dosage time, purge time and number of cycles was performed,
summarized in Table 4.4.
Table 4.4: Overview of ALD sample conditions.
Sample
TMA dosage time (min)
H2O dosage time (min)
purge time
(min)
number of cycles (-)
1
2
5
10
9
2
4
5
10
9
3
6
5
10
9
4
8
5
10
9
5
4
0.5
10
9
6
4
5
5
9
7
6
5
5
10
8
4
5
5
25
9
4
5
5
40
38
4.4
Heat treatment procedures
To select the optimal heat treatment procedure, multiple tests were performed on coated samples.
As mentioned in chapter 2.4 transformation of boehmite and amorphous alumina from sol-gel and
ALD respectively to the stable α-alumina phase goes through a sequence of alumina phase transitions
with associated volume changes [23]. Furthermore, carbon and water impurities are still present
[63]. Therefore, after drying, a two-step calcination/annealing procedure was necessary.
For the first step, the sample was heated to 450 °C in air with a heating rate of 5 °C/minute and kept
at this temperature for 30 minutes, followed by flushing with argon and continuing calcination at 450
°C overnight (12 hours). Subsequently the sample was cooled down to room temperature with a
cooling rate of 5 °C/minute.
The second step consisted of heating to 900, 1000, 1100 or 1200 °C with a heating rate of 5
°C/minute, keeping the sample at that temperature for 1 hour and then cooling down to room
temperature, again at a rate of 5 °C/minute.
4.5
Characterization and performance
In this part the characterization of the particles and composite will be discussed. Details on the
working principles of the analysis techniques listed can be found in appendix V.
4.5.1 Characterization raw material
A variety of characterization techniques was performed on the prepared samples. The powders as
delivered and the wind sifted samples were analysed using a JSM 6500F, JEOL Ltd. Scanning Electron
Microscope (SEM) with an electron beam energy of 15 keV and a beam current of 600 pA to
investigate morphology. Elemental analysis was performed with Energy-Dispersive X-ray
Spectroscopy (EDS). Powder samples were deposited on carbon tape and if necessary, coated with a
thin film of carbon to prevent charge buildup. Laser diffraction was performed to observe the particle
size distribution with a Malvern Master Sizer X laser diffraction instrument (Malvern Instruments
Ltd.). BET specific surface area was determined by N2 physisorption, utilizing a TriStar II 3020,
Micromeritics, in which approximately 9 g of powder was weighed, degassed at 300 ᵒC overnight in
an Autosorb Degasser from Quantachrome instruments and subsequently analysed.
4.5.2 Characterization coated particles
SEM and EDS were also used to characterize coated particles, again to see morphology of the
samples and whether successful coating has been achieved and a homogeneous coating was
obtained. The same parameters were used for SEM and EDS. X-ray photoelectron spectroscopy (XPS)
was utilized to confirm successful coating application and to obtain information on the chemical
nature of the surface. XPS was performed using a PHI 5400 ESCA from Physical Electronics Inc,
equipped with a Mg/Al dual anode X-ray source and a hemispherical capacitor analyzer. The aperture
of the input lens was a 3.5x1.0 mm aperture and emitted electrons were observed at a 45ᵒ angle
with respect to the sample surface. In all cases, powders were pressed into a high purity indium foil
for analysis and analyzed with with Mg Kα radiation (1.2538 keV). The x-ray source was operated
with 13 kV acceleration voltage and 200 W power, while the vacuum of the analysis chamber was
below 5 ·10-9 mbar. Spectra were taken from 0 to 1100 eV with a step size of 0.5 eV and 100s
acquisition time per step for a survey. For detailed peak analysis, step size was 0.2 eV with 200s
acquisition time per step, while the range was varied depending on the element.
39
Furthermore, X-ray diffraction (XRD) was used to obtain information on crystalline phases present
and to see whether obtained coatings are crystalline or X-ray amorphous. This was done with a
Bruker D5005 diffractiometer from Bruker Corp. and equipped with a Huber incident-beam
monochromator and a Braun position sensitive detector. Powders were dispersed in ethanol and
deposited on a monocrystalline Si wafer. Cu Kα1 radiation (wavelength of 154 pm) was used to
record diffractograms in a 2θ range of 15ᵒ to 90ᵒ with a step size of 0.039ᵒ.
To obtain information on the amount of alumina present and to obtain an estimate of coating
thickness, Inductively Coupled Plasma-Optical Emission Spectroscopy (ICP-OES), X-ray Fluorescence
(XRF) analyses and Electron Probe MicroAnalysis (EPMA) surface layer thickness determination [100]
were performed.
For ICP-OES, approximately 40 mg of the sample was dissolved in a mixture of 6 mL 1.0M HCl
solution with 2 mL 1.0 M HNO3 and 2 mL 1.0 M HF solution with the aid of 30 min heating in the
microwave at maximum power. After destruction, samples were diluted to 50 mL with milliQ and
analyzed with a PerkinElmer Optima 4300 ICP-OES instrument at three different wavelengths per
element. Concentrations of each element (Si, Mo and Al) were calculated with the aid of calibration
solutions for each element, prepared on the same day.
XRF was performed utilizing an Axios Max WD-XRF spectrometer from Panalytical. To be able to
measure the powder, 2.0 g of coated powder was weighed, mixed with a cellulose binder and
pressed into a tablet. This tablet was then measured in the equipment in a vacuum.
To obtain thickness information based on EPMA analysis, the powder sample was deposited on
carbon tape and coated with a thin layer of carbon to prevent charging. This sample was
subsequently inserted in the SEM and K-ratios for four elements (Mo, Si, Al and O) were measured
with the EDS detector by irradiation of two selected spots for two minutes in a central area on a
surface perpendicular to the electron beam. For each sample this procedure was performed on ten
different particles and afterwards coating thickness was calculated with the method of Goldstein et
al. [100] using a numerical iteration program. Acceleration voltage was varied between 5 and 25 kV
depending on the thickness of the coating to obtain approximately equal EDS count signals for Al and
Si peaks. This was necessary to ensure high accuracy and the optimum acceleration voltage for each
sample was determined by trial and error on a different particle before measurement started.
Cross-sections of the particles were also prepared to obtain better understanding of the interface
and confirm coating thickness. Cross-sections were prepared by embedding the 10g and 20g
aluminium tri-sec-butoxide coated particles and ALD 25 and 40 cycle samples in an epoxy resin (G2
epoxy resin, Gatan) between two Si wafers of approximately 5x5 mm. These wafers with the epoxy
resin and particles were subsequently inserted in a cross-section ion polisher (SM-09010, JEOL Ltd.).
Operating conditions were an acceleration voltage of 3.5 kV and an ion current of 18 µA, using Ar
ions from 6N (99.9999%) Ar gas. Cross-sections were subsequently analyzed by SEM and EDS. Finally,
to obtain information about required annealing conditions, Thermo-Gravimetric Analysis (TGA) was
performed.
40
4.5.3 Characterization heat treated particles
Heat treated samples were characterized by SEM combined with EDS to obtain information on
morphology and coating integrity upon annealing, especially to observe the appearance of cracks in
the coating. Furthermore XRD was used to see whether coatings were crystalline and which phases
were present. XRD was also used to estimate grain size of the α-alumina phase. For all methods
mentioned here, equipment and analysis parameters were kept constant and the same as for
characterization of the coated particles in chapter 4.5.2.
4.5.4 Performance testing final particles
To evaluate long-term shell integrity and stability of the particles, they were subjected to an oxidizing
environment at high temperature using TGA. Weight change of approximately 25 mg of 10g and 20g
aluminium tri-sec-butoxide and 25 and 40 cycle ALD samples that were heat treated at both 450 and
1200 ᵒC, was monitored at isothermal hold for 100h at 1000 ᵒC. Furthermore, wind sifted but
uncoated MoSi2 particles were also monitored to act as a blank. Finally an ALD 40 cycles sample that
has only been heat treated at 450 ᵒC was also monitored to investigate whether this would limit
cracking observed in these samples and improve oxidation resistance due to the more amorphous
structure at the start of the experiment. For investigation of the healing properties, the 20g sol-gel
coating procedure and heat treatment at 450 ᵒC and 1200 ᵒC was repeated for wind sifted MoSi2
particles that contained 2 wt% boron. Previous research by Mao [17] found that the addition of B to
MoSi2 improved oxidation and healing kinetics, allowing for healing at lower temperatures by
inhibiting SiO2 crystallization, which was the reason for the use of MoSi2 with 2 wt% B.
Coated particles were mixed with yttria-stabilized zirconia powder (5.2 wt% Y2O3, basic grade,
D(V,0.5) = 40 nm, Tosoh) by adding 5.58 g of coated particles to 22.0 g YSZ powder in a glass jar and
mixing these powders on a roller mill. The mixed powder was subsequently divided in three equal
portions, which were sintered with Spark Plasma Sintering (SPS). During SPS, the powder was first
compressed with a force of 35 MPa, followed by heating to 1450 ᵒC (heating rate 25 ᵒC/min) and was
kept at this temperature for 30 minutes to attain full densification. Sintering was monitored by piston
movement and was assumed to be finished after the piston ceased to move (on average after 5
minutes pressing at maximum temperature). After isothermal holding at 1450 ᵒC, pressure was
released and heating stopped and the system was allowed to cool by subjecting it to a room
temperature atmosphere. This resulted in three cylindrical tablets with a diameter of 20 mm and a
thickness of 5 mm, containing 20 volume % of healing particles.
To investigate healing properties, damage was introduced in the samples by Vickers indentation with
10 kg force after polishing. Additional and more severe damage was also created in one of these
tablets by using Vickers indentation with 20 kg force. The tablet that was only subjected to damage
from 10 kg force was healed at 1200 ᵒC for 30 minutes, introducing it directly in the furnace at 1200
ᵒC without heating and cooling to prevent excessive oxidation. The tablet that was more severely
damaged was healed at 1100 ᵒC for 1h with heating and cooling rates of 5 ᵒC/min to avoid thermal
shock damage. The third tablet was kept at 1100 ᵒC for 48h without polishing or damage to
investigate thermal stability of the composite. Furthermore, several tablets containing only YSZ
powder were also sintered, polished and indented with the same procedure to compare mechanical
properties of composites and pure YSZ. Damage and healing were investigated with optical
microscopy and SEM combined with EDS.
41
Results and Discussion
In this chapter, the results of the various methods of particle shell manufacture will be presented and
discussed. The chapter will start with a discussion of the uncoated particles and the results of wind
sifting. This is followed by the results of the two particle coating methods, starting with the sol-gel
procedures and followed by the results of atomic layer deposition with residual chemical vapor
deposition (ALD/rCVD). Afterwards, the heat treatment of coated particles resulting from both
methods will be analyzed. The chapter will then continue with an analysis of the thermal stability of
the particles and how this compares to the diffusion model and finish with the results of particles
embedded in the self-healing composite.
The following abbreviations will be used in this chapter to address the main samples. Wind sifted
molybdenum disilicide particles that have been coated with a sol/gel originating from 10 or 20 g
aluminium tri-sec-butoxide per 10 g of particles will be referred to as SG-10g and SG-20g
respectively. The wind sifted particles produced with the final ALD/rCVD method that underwent 25
and 40 cycles will be referred to as ALD-25C and ALD-40C respectively. Furthermore, MoSi2 particles
that contain 2 wt% alloyed boron will be referred to as MoSi2B. This is however neither an accurate
description of the boron content (which would be more closely resembled by Mo 3Si6B), nor of the
phase it is present in.
5.1
Characterization starting materials
5.1.1 Starting material
The MoSi2 and MoSi2B powders from Chempur were first characterized as delivered. Particle size
distribution (PSD) was studied by laser diffraction, shown in Figure 5.36. SEM analysis was also
performed on MoSi2 powder for investigating morphology, shown in Figure 5.37. Phase composition
was studied by XRD, presented in Figure 5.38. BET specific surface was measured for the MoSi2
powder with N2 physisorption, the adsorption-desorption curve of which is shown in Figure 5.39,
along with the measured BET specific surface area.
14
120
MoSi2 PSD
12
MoSi2 PSD
100
MoSi2B PSD
MoSi2B PSD
80
Volume (%)
Volume (%)
10
8
6
60
40
4
20
2
0
0
0.1
1
10
100
Particle diameter (um)
1000
0.1
1
10
100
Particle diameter (um)
1000
Figure 5.36: Laser diffraction results for MoSi2 and MoSi2B with the volume particle size distribution (left) and cumulative
volume distribution (right).
42
Figure 5.37: Scanning Electron Microscope (SEM) images of untreated MoSi2 powder at different magnifications.
When investigating the obtained SEM and laser diffraction data, it is clear that the particle size
distribution of both the MoSi2 and the MoSi2B is very polydisperse and far from the 25 µm that would
be the ideal particle size. This is also illustrated by a volume average diameter D(V,0.5) of 18.6 µm
but a D(V,0.1) of only 2.9 µm. Due to the lens used during laser diffraction, particles smaller than 0.5
µm could not be measured during the MoSi2 experiment, but SEM images indicate that such particles
are very likely to be present as well. This would also explain the high BET surface area of almost 0.6
m2/g found during N2 physisorption, which is more than 10 times the surface area that would be
expected for ideal spherical particles of 18 µm (0.053 m2/g for MoSi2The distributions for the raw
materials are however very similar to each other, with the MoSi2B powder being only slightly finer
than the MoSi2 powder.
The SEM images also show that most particles are far from spherical. Many particles have sharp
edges and characteristic cleavage steps and surfaces. These result from the manufacturing process of
Chempur, which is crushing of sintered material, which could also explain the high polydispersity
observed, as crushing is likely to yield a wider distribution than many other manufacturing methods.
□
□
Intensity (A.U.)
□
□
15
□
•
◊
35
MoSi2B
□
□
•
MoSi2
□ MoSi2
• Si
◊ MoB2
□ □
□
□
□
□
□
◊◊
55
2θ (ᵒ)
75
95
115
Figure 5.38: X-ray diffractograms of MoSi2 and MoSi2B with phases present.
43
The X-ray diffractograms show that in both materials the main constituent is indeed MoSi2 in its
stable tetragonal form, as would be expected. It also shows the presence of some residual silicon,
which is most likely due to the production process. As polycrystalline Si is significantly less expensive
than Mo, a slight excess is often added to ensure complete reaction of molybdenum to MoSi2 during
sintering and prevent the formation of Mo5Si3 [101]. Both diffractograms do indeed not show any
Mo5Si3 peaks.
Furthermore, the diffractogram of MoSi2B also has another phase, namely MoB2. This indicates that
boron segregates from the MoSi2, instead of dissolving in the matrix. No peak shifts of MoSi2 from
lattice strain due to the presence of boron are observed either, which would be expected if boron
would actually dissolve in significant amounts in the MoSi2 matrix. Another observation supporting
this is the presence of Si peaks of higher intensity compared to MoSi2 peaks. As boron in MoB2
replaces Si in certain phases, larger amounts of free Si should be present, which is exactly what is
observed. The segregation of boron is important, as this could affect the healing capabilities of the
capsules. As Mao discussed, the main effect of boron is the stabilization of the amorphous SiO2 phase
during oxidation. This does require the boron to be located close to Si though, which is why it is
important for these segregated areas to not be too large. An attempt was made to obtain B
distribution in the particles by EDS element mapping shown in appendix VII, but proved to be
unsuccessful due to energy overlap with carbon (from carbon tape) and the low accuracy of the
detector for low energy x-rays from B. The distribution nevertheless seemed relatively homogeneous
over the MoSi2B particle, although another experiment is recommended.
0.8
MoSi2 N2 physisorption
Quantity N2 adsorbed (cm3/g STP)
0.7
0.6
0.5
BET specific surface area:
0.585 ± 0.003 m2/g
0.4
0.3
0.2
0.1
0
0
0.2
0.4
0.6
0.8
1
1.2
Relative pressure P/P0
Figure 5.39: N2 physisorption adsorption/desorption curve of MoSi2 base material, with reference pressure P0 = 0.1 MPa.
Finally, the N2 physisorption isotherm allows both calculation of the surface area of the as received
samples and the determination of porosity. The lack of a hysteresis between the adsorption and
desorption curves indicates that no mesopores (2-100 nm in size) are present, while the low starting
adsorption and lack of hysteresis at the start show the lack of micropores (<2 nm). The absence of
larger pores is also clear, as these would be visible in the SEM images and no such pores have been
observed. It is therefore concluded that the MoSi2 particles are nonporous and sufficiently densified
during manufacture.
44
As mentioned before, it is also possible to measure the surface area with N2 physisorption using the
Brunauer-Emmett-Teller (BET) physisorption model. For this MoSi2 sample, the BET specific surface
area was found to be 0.585 m2/g, which is very high when a high thickness Al2O3 coating is necessary.
Because of this and the presence of a significant amount of very small particles, wind sifting to
separate fine and coarse particles was deemed necessary.
5.1.2 Wind sifting
Hence, the characterization of wind sifted particles is discussed here. PSD of both the coarse and fine
fractions was again studied with laser diffraction, as shown in Figure 5.40 and Figure 5.41 and the
PSD percentiles are shown in Table 5.5. N2 physisorption isotherms of the two MoSi2 batches are
shown in Figure 5.42, along with the calculated BET surface areas in Table 5.6. SEM images of windsifted MoSi2 and MoSi2B are shown in Figure 5.43. Finally, the possibility of contamination by the
wind sifting equipment was investigated with XRF on MoSi2B before and after wind sifting, the results
of which are shown in appendix VIII.
14
12
MoSi2 batch 1
12
10
MoSi2B batch 1
8
MoSi2 before wind-sifting
8
Volume (%)
Volume (%)
MoSi2 batch 1 fine
10
MoSi2 batch 2
MoSi2 before
wind-sifting
6
6
4
4
2
2
0
0
0.1
1
10
Particle diameter (um)
100
1000
Figure 5.40: Laser diffraction PSD results for the coarse
fractions after wind sifting compared to the material before
wind sifting.
0.1
1
10
Particle diameter (um)
100
Figure 5.41: Laser diffraction PSD results for the fine
fraction of MoSi2 batch 1 after wind sifting compared to
the material before wind sifting.
Figure Figure 5.40 and Figure 5.41 clearly show the classification abilities of the wind sifting machine.
In both coarse fractions of MoSi2 almost all particles below 7 µm in diameter have been removed
from the as delivered powder, while retaining the larger particles. As is evident from the fine fraction
analysis, this separation is not perfect though, as some of the larger particles are still present in the
fine fraction. Their volume fraction has however been significantly decreased. The imperfect
separation can be explained at least partially by the irregular shape of the particles. As the wind
sifting separates based on density and hydrodynamic properties in an air flow, irregular shapes result
in variation of the hydrodynamic properties with variation of orientation [102]. Therefore particles
that are either too large or too small to be expected in the fine or coarse fraction respectively can
still end up there. Another cause could be imperfect dispersion, causing several smaller particles to
stick together or to a larger particle and end up in the coarse fraction.
The relative lack of the smallest particles of 1 µm or lower in the fine fraction can be explained by the
imperfect cyclone that is supposed to separate all fine particles from the gas phase. This cyclone is
not able to separate most of the very fine particles from the gas flow due to their low weight to
surface area ratio. Due to this, the finest particles usually end up in the dust filter of the wind sifting
machine.
45
As is clear from these figures however, good separation is possible and reproducible, as there is only
a small difference between the two MoSi2 batches. However, separation of fine particles has been
less successful for the MoSi2B fraction even though a significant amount of fine material has been
removed. This might be related to the material, as it was already observed in the XRD diffractogram
that the MoSi2B material consists of multiple phases, the distribution of which is not completely
known. If some particles contain a significantly higher percentage of MoB2, their density is likely
higher as well, which alters separation behavior in wind sifting.
Table 5.5: PSD percentiles and Sauter particle diameter for the measured samples calculated from laser diffraction data.
Sample
Specifications Chempur MoSi2 (µm)
Actual measured MoSi2 (µm)
MoSi2 Wind sifted batch 1 coarse (µm)
MoSi2 Wind sifted batch 2 coarse (µm)
MoSi2 Wind sifted batch 1 fine (µm)
D (V,0.1) D (V,0.5) D (V,0.9)
1.86
4.87
9.01
2.88
18.63
42.89
8.315
28.79
57.42
10.24
29.16
58.23
2.04
11.91
47.59
Sauter diameter D (3,2)
8.16
13.78
17.65
3.53
Table 5.5 reinforces this, showing a relatively large increase in the tenth percentile of the PSD, while
showing a significantly smaller relative increase in the 90th percentile for both wind sifted batches
compared to the starting material. The fine fraction shows a relatively small decrease of the tenth
percentile however, strengthening the suspicion that most very fine particles are not collected in this
fraction. This could also explain the observed loss of material, which was between 10 and 20 percent
of the starting weight and the fact that the fine fraction weight usually consisted of less than 10
percent of the starting weight.
0.8
MoSi2 N2 physisorption
0.7
Quantity N2 adsorbed (cm3/g STP)
Wind sifted batch 1
0.6
Wind sifted batch 2
0.5
0.4
0.3
0.2
0.1
0
0
0.2
0.4
0.6
0.8
1
1.2
Relative pressure P/P0
Figure 5.42: Nitrogen physisorption isotherms of the two wind sifted MoSi2 coarse batches and the starting material.
To confirm that a significant part of the fine material has been removed in the coarse samples and to
measure the necessary decrease in specific surface area, N2 physisorption was also performed. As is
clear from Table 5.6 and Figure 5.42, there is indeed a significant decrease in surface area, as the
total surface area is more than halved by wind sifting the material, which is also indicated by a lower
total adsorption in both wind sifted batches. Surprisingly though, a small hysteresis can be observed
46
in the isotherm of batch 1. This would indicate that either wind sifting could introduce porosity in the
particles, which is unlikely, or that some contamination by a porous material has occurred. A
measurement error would also be possible, as these measurements are performed relatively close to
the limit of the machine capabilities and it is not entirely sure that this hysteresis is significant.
Therefore, XRF was performed on the MoSi2B sample before and after wind sifting, the results of
which are shown in appendix VIII. However, no significant changes in elements other than Mo and Si
were found. This indicates that contamination is unlikely, as the higher Mo and lower Si
concentrations can most likely be attributed to measurement errors. Another option might be that
separate particles of pure Si are present, as the presence of Si in the base material was confirmed by
XRD. The density of Si is also significantly lower than that of MoSi2, which would result in significantly
more and larger Si particles in the fine fraction. If contamination has occurred on the other hand, this
would most likely result in an increase of Si instead, as the most likely contaminant would be fly ash,
which consists mostly of silicates. An increase of concentrations of other elements would also be very
likely upon contamination.
Table 5.6: The calculated BET specific surface areas based on isotherm data for each sample.
Sample
Sample mass measured (g)
MoSi2 Starting material
9.2193
MoSi2 Wind sifted batch 1
10.6214
MoSi2 Wind sifted batch 2
7.4539
BET surface area (m2/g) Error (m2/g)
0.585
0.0026
0.289
0.0007
0.244
0.0009
Figure 5.43: SEM images of MoSi2 particles (left) and MoSi2B particles (right) after wind sifting.
Finally, analysis of the SEM images in Figure 5.43 shows that the morphology of the particles remains
unchanged, but the amount of fine particles has significantly decreased, confirming the results of the
laser diffraction experiments. Some smaller particles, mainly stuck to the larger particles, are still
observed however, indicating that dispersion is indeed not perfect. There is also an interesting
difference between the MoSi2 particles and MoSi2B particles, with the latter apparently containing
more fine particles and having a smaller particle size in general. This does agree with wind sifting
observations in which a larger fraction of fine material is found in the MoSi2B powder. It is also likely
that this increase in the number of smaller particles also results in an increase in total surface area.
This could have implications for resulting coating thickness and coverage.
47
5.2
Sol-gel coating
5.2.1 Oxalate method
Experiments with aluminium oxalate were performed to obtain critical information regarding the
behavior of MoSi2 particles in an acidic sol. Previous investigations by Carabat et al. [22] already
found that dispersion of MoSi2 particles in such a sol is difficult due to the high density of the
particles. Furthermore, as mentioned in the theory part, complete coverage is very important. This
coverage can be affected by dispersion, as settled particles will have less contact with the sol, often
resulting in less material being deposited on these particles.
This was also observed in the aluminium oxalate experiments, as two separate morphologies could
be distinguished for all experiments except experiment 2. An example of these two morphologies is
shown in Figure 5.44, in which the more flaky phase consists mainly of aluminium and oxygen, as
measured by EDS shown in Figure 5.44 and Table 5.7. The flaky phase was however mostly present
on the surface of MoSi2 particles and not separate as loose flakes. Two of the particles, namely at
point four and five, on the other hand did not show any significant concentrations of aluminium and
instead has molybdenum and silicon as its main constituents. Although, in the case of point four, the
absence of molybdenum might indicate a problem with the EDS measurement for this point. Another
option might be that this is a particle of pure silicon however, as XRD did show the presence of a
separate Si phase.
Figure 5.44: The two morphologies present for all aluminium oxalate sol-gel coatings except those resulting from
experiment two, high-resolution (left) and low-resolution EDS image with measurement points indicated (right).
However, in other cases, particles did show aluminium coverage as measured by EDS without a flaky
morphology. An example of this can be seen for point 1, which does not contain significant traces of
the flaky phase, but does seem to be covered with a significant amount of aluminium. Therefore it is
likely that the flaky morphology is actually aluminium oxalate that did not dissolve completely, which
is confirmed by the low solubility of aluminium oxalate [103], especially at a pH close to 7. Although
aluminium oxalate solubility increases drastically in more acidic solutions, concentrations used in
these experiments were sometimes higher than this solubility, as this aided in forming a sol and gel.
Due to this, it is not possible to evaluate coating quality based on SEM observations alone. Therefore,
to evaluate coverage, EDS was utilized as an additional observation.
48
Table 5.7: EDS elemental concentration measurements in atom% of the points shown in Figure 5.44 b.
Point 1
Point 2
Point 3
Point 4
Point 5
O content (atom %)
45
64
54
10
-
Al content (atom %)
4.6
32.9
25.7
0.69
1.2
Si content (atom %)
35
1.8
14
89
65
Mo content (atom %)
16
1.5
6.4
0.60
34
In experiment 2, an attempt to increase dissolution by addition of more nitric acid was performed. As
expected, this lowered the pH to roughly 2.5, measured with pH paper. An example of the resulting
particles is shown in Figure 5.45. Although no flaky morphology was observed in this sample, another
observation was the lack of coating on the particles in general with only small patches of the particle
being coated. This is most likely due to electrostatic interactions between particles and sol, as
discussed in chapter 2.2, resulting in both sol and substrate being positively charged and repelling
each other. Another indication that this is the case is the observation that after filtration, the filtrate
is light green, just like the precursor solution/sol before addition of the MoSi2 particles. In all other
aluminium oxalate experiments the filtrate was colorless.
Figure 5.45: An SEM image of particles resulting from experiment 2, only being coated by small patches of aluminium
oxalate instead of almost complete coverage.
An important conclusion from this would be that the surface is indeed behaving like a native SiO2
layer with an isoelectric point near pH 3 and that electrostatic interactions between sol and substrate
are very important in these systems. Therefore it is important to keep the pH of the sol around 4, as
this allows for favorable electrostatic interactions between sol colloids and MoSi2 surfaces and will
improve coverage.
To further improve coverage of particles with the sol-gel method, three more experiments were
performed, in which the effects of gelation before addition of the particles and nitrogen bubbling on
the resulting coating were investigated. Both of these methods attempted to increase dispersion by a
49
higher viscosity and upward movement pulling the particles into suspension respectively. SEM
images shown in Figure 5.46 show the resulting morphologies for the three different experiments.
The experiment with pregelation also formed a significant agglomeration at the bottom, that was
separately investigated and shown in Figure 5.46 d.
(a)
(b)
(c)
(d)
Figure 5.46: SEM images of the three final sol-gel experiments, with neither pregelation nor nitrogen bubbling (a),
dispersed phase after pregelation (b), with nitrogen bubbling (c) and the agglomerated part at the bottom after
pregelation (d).
The figure shows significant difference in resulting morphology, especially related to the presence of
the flaky phase on the particles. In the case of nitrogen bubbling (Figure 5.46 c), the amount of this
phase seems to have increased significantly compared to the sample in which neither of the methods
was used. Although some increase in this phase on the particles seems to be present when
comparing the pregelated phase that remained in suspension with the original sample, the sediment
of that sample does not show any of this phase. Another observation from these images is the
significantly larger particles being present in the latter phase, indicating that sedimentation of larger
particles in particular is indeed a problem in this method.
However, as was observed before, coating is difficult to observe solely based on morphology, which
is why the samples were also compared with EDS measurements. The individual measurements can
be found in appendix IX, while these results are summarized in Table 5.8. For each sample a certain
amount of points were measured with EDS and were compared based on the amount of points that
50
did not show a significant amount of Al, as defined by the EDS software. This significance is a peak in
the EDS spectrum that is at least three times the value of the background noise with an acceleration
voltage of 15 kV. Another comparison was made that lists the amount of points that did have a
significant aluminium detection, but less than 0.90 atom percent. In these cases, the accuracy of the
EDS detector is still uncertain and such a low Al percentage would result in very thin coatings in any
case and is therefore used as another comparison value.
Table 5.8: Combined EDS measurements for each sample and the number of measurements with significant Al detected.
Sample
No pregelation,
no bubbling
Pregelation,
dispersed
Pregelation,
agglomerated
Bubbling
Total EDS
measurements
No significant Al
detected by EDS
Significant Al detected, but Percentage
<0.90 atom%
of total (%)
15
2
2
27
16
0
6
38
13
53
0
3
5
10
39
25
This table confirms the findings based on SEM images that both pregelation and nitrogen bubbling
result in an increase in coverage of particles when comparing whether the Al signal was significant
according to the detector software. When investigating whether more than 0.9 atom % of Al was
present, the difference is however less pronounced, with the pregelation method actually
performing significantly worse and bubbling only performing slightly better. However, due to the
limited number of points measured for the first sample, this difference is probably not significant and
more measurements should be performed if definite conclusions are to be drawn based solely on
EDS.
The higher number of particles with low amount of coating could be accounted for by the higher
amount of solvent required to retain a bubbling regime, with 300 mL being necessary as opposed to
125 mL for the original sample. With the same amount of precursor, this most likely results in less
gelation due to lower overall concentration.
Another interesting observation is that there is also no significant difference between the
agglomerated and dispersed powders in the pregelation experiment. This indicates that coating also
deposits on the particles present in the sediment. The amount of coating detected by EDS is
somewhat less though, which agrees with SEM observations showing almost complete absence of
the flaky phase.
Based on these measurements, the main conclusion is that coating is possible using aluminium
oxalate sols. The resulting coating morphology consists of both a flaky phase that is most likely nondissolved aluminium oxalate and a thin adherent layer on top of particles that is difficult to
distinguish without EDS. Furthermore, nitrogen bubbling is beneficial to contact between sol and
particles and could aid in coverage improvements, while the effect of pregelation is most likely
detrimental to coverage, as was expected according to previous findings described in section 2.3.
Finally, electrostatic interactions are very important in the alumina-MoSi2 particle system and can be
regulated by pH, which should be carefully regulated around 4 to have attractive interactions
between Al sol and the MoSi2 particles.
51
5.2.2 Aluminium tri-isopropoxide method
Another precursor investigated was aluminium tri-isopropoxide, an alkoxide that forms a boehmite
(AlOOH) sol and gel upon reacting with water. Due to this reaction, dissolution of the precursor in
anhydrous alcohols was necessary and particles and water were added at the same time. The
resulting coated particles were investigated by SEM, the results of which can be found in Figure 5.47.
Figure 5.47: Particles coated with aluminium tri-isopropoxide precursor based gel at different magnifications.
As with the aluminium oxalate based samples, two different morphologies can be observed again.
One that is very similar to the uncoated particles and another that consists of somewhat sharp
crystals with a morphology somewhat different from the base material. However, when performing
EDS analysis, both morphologies contain aluminium in significant amounts, although for the latter
morphology significantly higher concentrations are detected. When performing the same analysis of
the EDS measurements as for the oxalate samples in chapter 5.2.1, only 2 out of 12 points or 17%
contain less than 0.9 atom % aluminium. Individual EDS measurements are tabulated in appendix IX.
These results indicate that the formation of boehmite from aluminium tri-isopropoxide results in a
sol that has a more advantageous interaction with the MoSi2 substrate than sols derived from
aluminium oxalate, as the coating has an improved coverage and is more homogeneous, as is evident
from the SEM images and EDS data. An explanation for this observation could be the formation of
boehmite by a chemical reaction, which forms a sol with significantly different morphology than one
formed from metal salt solutions, as explained in chapter 2.2.1. However, a significant disadvantage
of aluminium tri-isopropoxide is its low solubility in most solvents, even its own parent alcohol,
isopropoxide. It is also very difficult to dissolve, making this route very complex and limiting the
amount of alumina that can be deposited on the particles by this method.
5.2.3 Aluminium tri-sec-butoxide
To circumvent the problem of low solubility that aluminium tri-isopropoxide faces, another very
similar alkoxide, aluminium tri-sec-butoxide was also utilized as a precursor. This alkoxide is liquid at
room temperature and miscible with H2O, although it reacts rapidly. Due to these properties,
dissolution of the precursor in advance was not necessary, resulting in a significantly simpler process.
Different ratios of precursor to MoSi2 particles were used with 5g, 10g or 20g of precursor per 10g
MoSi2 to yield particles with an approximate thickness of 100, 200 or 400 nm Al2O3 after heat
treatment respectively if all formed boehmite deposited on the particles. Due to the formed gel
52
blocking the filter pores, separation of particles and liquid could not be performed by filtration.
Therefore liquid evaporation and centrifugation were used as alternative separation methods. These
samples were again characterized with SEM, the results of which can be found in Figure 5.48 for all
10 and 20 g samples. EDS analysis of coating thickness and presence is summarized in Table 5.9, with
the individual measurements available in appendix IX.
(a)
(b)
(c)
(d)
Figure 5.48: SEM images of four of the aluminium tri-sec-butoxide experiments, with 10g (a) and 20g (b) Al(OC4H9)3
separated by evaporation and 10g (c) and 20g (d) Al(OC4H9)3 separated by centrifugation.
These SEM images show that both the evaporated SG-10g and SG-20g have a very similar
morphology as the aluminium tri-isopropoxide sample, containing a significant amount of coating
material. The 5g sample is not shown here, but did show a very similar morphology. However, for the
centrifuged samples it is very different, bearing more similarities with the uncoated material.
EDS investigations in Table 5.9 confirm this, showing a significantly lower average amount of coating
material and more points with less than 0.9 atom% Al. None of the evaporation samples showed any
points with less than 0.9 atom%, indicating that coverage of these samples was complete. XPS data
shown in appendix XI agrees, showing very minor peaks of Si and Mo for the sol-gel samples, while
aluminium, oxygen and carbon are the main constituents detected for the samples that have not
been heat treated yet. EDS also shows an increased average loading on the particles with increasing
ratio of precursor to particle mass, as would be expected. This trend is visible for both evaporated
53
and centrifuged samples, although the significance of the centrifuged samples is uncertain due to the
low number of total measurements.
Table 5.9: Combined EDS results for the aluminium tri-sec-butoxide samples with average atom% Al detected and the
amount of measurements that found less than 0.9 atom% Al.
Average atom% Al
Measurements with <0.90
detected by EDS with
Sample
Total measurements (-)
atom% Al detected (-)
15kV beam (atom %)
5g evaporation
8
0
8.8
10g evaporation
24
0
14.8
20g evaporation
26
0
23.5
10g centrifugation
2
1
1.0*
20g centrifugation
8
1
4.3
*Average of only two measurement points, one of which did not detect any Al and therefore not
significant.
Based on these results, it can be concluded that the centrifuged samples have a significantly thinner
coating than the liquid evaporation samples. A possible explanation for this observation is that
bonding of the gel coating is relatively weak and the forces required to separate liquid and solid
particles could also be strong enough to remove gel from the particles as well. This agrees well with
other authors that found that boehmite gel derived from alkoxides is generally weakly bonded [38,
41].
Furthermore, this could explain the findings of Dey et al. [50] in which a boehmite coating on SiC
significantly thinner than expected was found, which was nevertheless strongly bound. As they used
centrifugation too, it is very likely that in their case the weakly bonded material was removed by
centrifugation as well, leaving only a thin layer of strongly bound material. This strongly bound
material is most likely material that is bound to the particle by electrostatic forces between the
support and the colloidal boehmite particles. As the only heat treatment performed was drying at
110 ᵒC, the formation of an interfacial phase between the boehmite and the SiO2 native oxide layer is
not likely, because this usually requires significantly higher temperatures.
Therefore it is likely that the coated particles prepared by evaporation of solvent consist of both a
thin, strongly adherent layer of boehmite and a boehmite gel on top of those particles that is more
weakly bound by Van der Waals forces originating from boehmite condensation reactions as
explained in chapter 2.2.1. After centrifugation, the latter layer would be removed, but the former
would still be present, which agrees with EDS data still detecting a measurable amount of Al for most
measurement points, but detecting significantly less overall Al compared to the evaporated sample
that is expected to still contain this layer.
When comparing SG-10g with the N2 bubbling aluminium oxalate sample and the aluminium
isopropoxide sample, which all have an approximately similar amount of aluminium, coverage seems
to have increased significantly, especially compared to the oxalate sample. Furthermore, the loading
of aluminium on the particles has significantly increased. In the case of aluminium isopropoxide, the
lack of nitrogen bubbling could explain a part of this difference, but better dispersion and mixing of
the precursor is most likely also a factor.
54
To evaluate coverage and investigate sample behavior during heat treatment, thermogravimetric
analysis (TGA) in synthetic air (80 vol% N2, 20 vol% O2) was performed for the 10g aluminium tri-secbutoxide sample, the aluminium isopropoxide sample, the aluminium oxalate sample with neither
pregelation nor nitrogen bubbling and wind sifted MoSi2 as a blank. The results of these TGA
measurements are shown in Figure 5.49.
106
sample oxalate A
Sample Isopropoxide
104
Relative mass (% of initial mass)
Sample tri-sec-butoxide
102
blank
100
98
96
94
92
0
100
200
300
400
500
600
Temperature (ᵒC)
700
800
900
1000
Figure 5.49: TGA results for an appropriate sample of each type of precursor compared to uncoated MoSi2 as a blank.
This figure shows that the uncoated and wind sifted MoSi2 mostly follows the expected behavior as
described in chapter 3.2. No change in mass is visible before approximately 450 ᵒC and after that
temperature, mass starts to increase due to oxidation. Around 550 ᵒC, mass gain slows down though,
most likely due to the formation of a protective oxide, as sublimation of MoO3 usually starts to
become noticeable only above 800 ᵒC. This temperature is also the temperature at which a decrease
in mass can be observed, indicating that MoO3 is indeed evaporating at this point.
However, even at the final temperature reached in this TGA, only a small part of the total mass has
oxidized. This is because the highest possible relative mass would be 174% of the total initial mass
upon complete oxidation without any MoO3 evaporation. Upon complete evaporation of MoO3
however, the final mass of the remaining SiO2 would be 79% of the initial mass. Although some
oxidation is observed, the weight of this sample does not rise above 104%, indicating that oxidation
in this experiment is not limited by the amount of sample present.
For all sol-gel samples, an initial drop can be observed, which is mainly caused by two different
processes: the evaporation of water present on the particles and in the gel and the decomposition of
the precursors boehmite and oxalate to aluminium oxide. The former process occurs at lower
temperatures, usually below 200 ᵒC and the latter usually at temperatures between 250 and 450 ᵒC
[38].
For oxalate, a distinct two-step decomposition can be observed in the mass change graph shown in
appendix X, which according to literature corresponds to the loss of crystal water in oxalate and the
decomposition of the oxalate itself [104]. Upon oxidation, a large mass gain is observed in the same
55
temperature range as the bare substrate. What is surprising however, is that this mass gain is
actually larger than the blank. An explanation is that this sample was not prepared with wind sifted
material, resulting in a larger surface area and associated higher oxidation. The aluminium oxalate
did not seem to influence this oxidation in any way significant way though.
Both of the alkoxide precursors have a similar profile, but contrary to the oxalate sample, do have a
significant influence on the mass gain around 550 ᵒC. Especially in the case of aluminium tri-secbutoxide, the mass gain upon oxidation is very small and also seems to be delayed to a higher
temperature around 800 ᵒC. For aluminium tri-isopropoxide, the mass gain is still visible around 550
ᵒC, but is significantly smaller than the mass gain of the blank.
These results indicate that the coating is actually present and can already act as a barrier towards
oxidation, even before annealing. It also agrees with SEM and EDS observations in which aluminium
oxalate does not yield a very good coverage, while aluminium isopropoxide and aluminium tri-secbutoxide do have better coverage. The latter was also found to have the best coverage and thickest
coatings, which would correspond to the best oxidation protection. This can also be observed in the
TGA.
Investigating total mass loss in the first part of the TGA also confirms the SEM observations that
aluminium tri-sec-butoxide contains more aluminium. As all samples have been dried before TGA,
most water that was introduced during the sol-gel procedure should have evaporated and most of
the mass loss is due to decomposition of the precursors. The total mass loss of the tri-sec-butoxide is
significantly larger than the mass loss of the isopropoxide sample and slightly larger than the oxalate
sample. As both alkoxides result in boehmite formation on the particles and aluminium oxalate
decomposition results in significantly higher mass loss per mole of alumina, this indicates that
significantly more precursor was present in the tri-sec-butoxide sample.
Although this mass loss can be used to estimate coating thickness, accuracy is limited due to
evaporation of crystal water that might be present or combustion of residual organic material that is
always present. For example estimating the average thickness of SG-10g by this method would yield
an approximate average thickness of 0.4 µm, while the highest possible average thickness would be
0.2 µm.
Therefore, Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES), the EPMA method
mentioned previously in chapter 4.5 and the preparation of cross-sections were utilized as methods
to estimate thickness. Due to interference between the elements molybdenum and aluminium, ICPOES gave inconclusive results for coating thickness. The EPMA thickness estimate method could not
converge and yield results for a significant part of the measurement points due to the large variation
in thickness and associated difficulty to tune the acceleration voltage.
Therefore thickness of SG-10g and SG-20g was only measured based on cross-sections. One of the
SEM backscatter electron image (BSE) images for each sample is shown in Figure 5.50, while the
thickness distributions measured for all particles imaged by cross-section combined are shown in
Figure 5.51 and Figure 5.52. Thickness was measured in ImageJ with the measurement function. The
cross-section images used were backscatter electron images for better contrast. Measurements were
performed completely around each particle with a regular distance between each measurement.
Images used can be found in appendix XII.
56
<0
0-0.05
0.05-0.1
0.1-0.15
0.15-0.2
0.2-0.25
0.25-0.3
0.3-0.35
0.35-0.4
0.4-0.45
0.45-0.5
0.5-0.55
0.55-0.6
0.6-0.65
0.65-0.7
0.7-0.75
0.75-0.8
0.8-0.85
0.85-0.9
0.9-0.95
0.95-1
1-1.05
1.05-1.1
1.1-1.15
1.15-1.2
1.2-1.25
1.25-1.3
1.3-1.35
1.35-1.4
1.4-1.45
1.45-1.5
1.5-1.55
1.55-1.6
1.6-1.65
1.65-1.7
1.7-1.75
1.75-1.8
1.8-1.85
1.85-1.9
1.9-1.95
1.95-2
>2
Percentage of total measurements (%)
<0
0-0.05
0.05-0.1
0.1-0.15
0.15-0.2
0.2-0.25
0.25-0.3
0.3-0.35
0.35-0.4
0.4-0.45
0.45-0.5
0.5-0.55
0.55-0.6
0.6-0.65
0.65-0.7
0.7-0.75
0.75-0.8
0.8-0.85
0.85-0.9
0.9-0.95
0.95-1
1-1.05
1.05-1.1
1.1-1.15
1.15-1.2
1.2-1.25
1.25-1.3
1.3-1.35
1.35-1.4
1.4-1.45
1.45-1.5
1.5-1.55
1.55-1.6
1.6-1.65
1.65-1.7
1.7-1.75
1.75-1.8
1.8-1.85
1.85-1.9
1.9-1.95
1.95-2
>2
Percentage of total measurements (%)
Figure 5.50: Cross-section SEM-BSE images of coated particles for the SG-10g sample (left) and the SG-20g sample (right).
18%
16%
14%
12%
10%
8%
6%
4%
2%
0%
Coating thickness range (µm)
Figure 5.51: Coating thickness distribution for the SG-10g sample from cross-section analysis.
10%
9%
8%
7%
6%
5%
4%
3%
2%
1%
0%
Coating thickness range (µm)
Figure 5.52: Coating thickness distribution for the SG-20g sample from cross-section analysis.
57
In Figure 5.50, the white areas on the BSE images are the MoSi2 particles, while the light gray areas
represent aluminium containing coatings. The black area surrounding both is the resin used to
embed the particles. This agrees with Mo and to a lesser extend Si being significantly heavier
elements than Al or O, which in turn are heavier than C, which makes up most of the resin.
Compositions have been confirmed with an EDS line scan shown in chapter 5.3.2.
These cross-sections also clearly show that every particle has indeed been coated for both the SG10g and SG-20g sample. Another observation is that coating thickness varies significantly within the
same sample and even on individual particles. This was already expected, as sol-gel coatings on
particles in general are not uniform in thickness and due to the difficulty of thickness measurements
by the EPMA method. Although in some places on the particle the presence of coating might be
disputed, upon further magnification, a visible thin coating was observed. However, as is clear, this
thinner coating is present on part of a significant number of particles, especially for the SG-10g
sample, possibly compromising the oxidation resistance of these particles, as the thinnest part is
likely to experience rapid local oxidation.
Coating thickness could also be difficult to measure due to the maximum depth from which
backscatter electrons originate, if the particle surface below the cutting plane for the cross-section is
not perpendicular to this cross-section plane. That would result in small density differences that can
be detected by the BSE detector and result in hazy edges on some sides of the particles, even though
the whole particle is in focus. This effect introduces an additional error in thickness measurements
and could obstruct the observation thin coatings to a certain extend.
When comparing the SG-10g and SG-20g samples however, it is clear that there is a difference in
coating thickness, with the SG-20g sample having significantly thicker coatings. This is even more
evident in the distributions shown in Figure 5.51 and Figure 5.52, in which the SG-20g sample shows
a significantly higher average thickness, with a maximum around 0.4 µm, while for the SG-10g sample
this maximum is closer to 0.1 µm.
It is also interesting to note that the thickness of the dried boehmite gel for the SG-20g sample is
actually close to, but somewhat higher than the calculated average of 0.4 µm for the fully densified
Al2O3, indicating that most of the precursor did indeed deposit on the particles instead of
segregating. This is of course assuming that the density of the gel is close to its expected maximum
density of 3.0# g/cm3, which is not always the case [50]. For the SG-10g sample, the coating thickness
is surprisingly lower than expected though.
Nevertheless these results show that MoSi2 particles can indeed be coated successfully with
boehmite, using aluminium tri-sec-butoxide as a precursor. The amount of precursor is also directly
related to the thickness of the coating, as would be expected. These sol-gel coatings were also found
to have significant variations in thickness and most likely consist of two layers, one of which is thin,
strongly bound and close to the particle, while the outer layer is thicker, but relatively easily
removed. Furthermore, both proper dispersion of particles and gel and electrostatic interactions
originating from pH were found to be important parameters in successful coating. TGA heating tests
show that the coatings from aluminium alkoxides already have a protective effect compared to the
bare particles.
58
5.3
Atomic Layer Deposition/Residual Chemical Vapor Deposition
5.3.1 Mechanism
To elucidate the mechanism of Atomic Layer Deposition with Residual Chemical vapor deposition
(ALD/rCVD), several ALD experiments have been performed in which precursor dosage time and
purge time between precursor dosage were varied independently. Calculations of precursor addition
based on precursor vapor pressure under atmospheric conditions (25 ᵒC, 1 bar pressure) were
performed to ensure reactant addition was not limiting to coating thickness.
To investigate the resulting microstructure after coating, SEM and EDS was performed, the results of
which are shown in Figure 5.53 and Table 5.10. The resulting morphology for all thin ALD coatings
(<100 nm) was very similar for different samples and was not significantly different from the
morphology of the raw material. EDS results did show however that all particles were coated with a
thin layer of alumina. XPS analysis of the thicker samples shown in appendix XI confirms the coating
of the entire particle, with no Mo being detected and very little Si on the coated particles, which
means coating has been successful and coating might even be more complete than in the case of solgel. This indicates that a very conformal coating was deposited on all particles. It is also an indication
for the thickness being substantially higher than pure ALD samples, as the latter would result in
coatings of approximately 1 nm thick, which would be very difficult to detect with an acceleration
voltage of 15 kV and the EDS detector utilized in this research.
Figure 5.53: SEM image of particles coated by ALD with morphology characteristic for all ALD experiments with a
resulting thin (<100 nm) coating (left) and an image with EDS measurement points indicated (right).
Another observation visible in Figure 5.53 is that compared to the wind sifted material, the amount
of small particles present is significantly lower. Although laser diffraction after ALD/rCVD was not
possible due to the lack of sufficient material, SEM images for all ALD/rCVD samples show this
decrease in small particles. A possible explanation is that during fluidization these particles do not
remain in the fluidized bed and instead remain suspended in the gas flow and end up in the cyclone.
This often happens in fluidized beds, as a gas flow sufficient to fluidize larger particles is often able to
retain smaller particles in polydisperse powders [105]. It also agrees with the observation during the
experiment that a small fraction of the powder sticks to the distributor plate on top of the glass
reactor, approximately 0.5 m above the top of the bed. However, most of the material fluidized
properly.
59
Table 5.10: EDS elemental concentration measurements in atom% of the points shown in Figure 5.53 b.
O content (atom %) Al content (atom %) Si content (atom %) Mo content (atom %)
Point 1
Point 2
Point 3
Point 4
Point 5
Point 6
Point 7
65
43
41
62
42
58
43
5.8
4.8
1.9
3.4
1.7
2.7
1.8
20
35
38
24
38
27
37
10
17
19
11
18
13
18
Although the EDS results in Table 5.10 show that Al is indeed present on all particles measured,
which was also observed for other samples, there is some variation in Al content. Most of this
variation could be explained by the fact that EDS actually measures X-rays originating from a region
at and below the surface, as explained in appendix V combined with the observation that the
particles are not flat, but irregularly shaped. Nevertheless, the amount of Al present has significantly
less variation than the sol-gel samples, indicating that coating thickness might be more
homogeneous. The EDS results also show a surprisingly high concentration of oxygen, the origins of
which are not clear. A possible explanation might be the presence of water, as these particles have
not been dried after the ALD/rCVD process, which always ends with the addition of H2O precursor
and subsequent purging for safety reasons. Furthermore, the plasma activation process might have
induced some additional oxidation of MoSi2.
Coverage and behavior of an ALD/rCVD sample, namely the sample with 4 minutes TMA and 5
minutes purge, was also investigated by TGA, the results of which are shown in Figure 5.54. This
figure clearly shows the protective function of particles coated by this method. Weight gain at higher
temperatures is significantly lower than the uncoated material and significantly delayed as well.
When comparing this sample to the sol-gel samples, the weight changes at high temperature are
slightly lower than those of the SG-10g sample, even though this coating is thinner, as is shown in the
next section. Furthermore, there is almost no weight loss at lower temperatures, indicating that no
or very limited decomposition is necessary and that the coating does not contain a very significant
amount of water.
104.5
ALD 9 Cycles 5
minutes purge
Relative mass (% of initial mass)
104
103.5
blank
103
102.5
102
101.5
101
100.5
100
99.5
99
0
200
400
600
800
1000
Temperature (ᵒC)
Figure 5.54: TGA results for the 4 minute TMA, 5 minute purge ALD/rCVD sample compared to uncoated material.
60
To obtain more information regarding the mechanism, thickness was measured using multiple
different methods. These methods were ICP-OES and XRF, both combined with surface area
measurements by N2 physisorption to calculate coating thickness with equation 5.1 and as a third
method the EPMA method. All of these methods use the density found by [66] for ALD at 25 ᵒC to
calculate coating thickness, which was 2500 kg/m3. In this equation, a homogeneous coating
thickness is assumed. The parameters are as follows: x is the weight fraction Al measured by ICP, M is
the molar weight of the corresponding compound in g/mol, ρ the density in g/m3, S the specific
surface area of the compound as measured by N2 physisorption in m2/g and δ the thickness in m.
(5.1)
When analyzing the results of these methods, XRF was found to give an overestimation of the Al
content due to an inhomogeneous distribution of elements in the sample. ICP-OES results, shown in
appendix XIII on the other hand yield a significant underestimation, most likely due to difficulties
with dissolving the most likely formed aluminosilicates. The EPMA surface layer measurement
method was able to measure thickness, the results of which are shown in Figure 5.55 and Figure
5.56. Its accuracy was verified by cross-section analysis of a thicker coating, as shown in chapter
5.3.2. The cross-section and EPMA method agreed well in thickness estimate, indicating that this
might be the most accurate method.
For the EPMA method, measurements shown in the figures below show some interesting details
regarding the mechanism. As can be observed in the graphs, the error based on the discrepancy in
resulting thickness for the various points is large. However, the origin of these errors is not only in
the measurement error, but also in the small variation in coating thickness that is often observed in
ALD with rCVD and sometimes in pure ALD as well. This variation was also observed in the crosssection. As only 20 points were measured, a distribution could not be made. Due to this, the actual
measurement error is most likely smaller than presented here and these points can be seen as an
average coating thicknesses for each sample.
160
90
TMA dosage time
80
4 minutes TMA
dosage
120
6 minute TMA
dosage time
Water dosage time
Coating thickness (nm)
70
Coating thickness (nm)
140
60
50
40
30
20
100
80
60
40
20
10
0
0
0
2
4
6
Precursor dosage time (min)
8
10
Figure 5.55: The effect of TMA and water dosage time on
resulting average coating thickness as obtained by the EPMA
method.
0
2
4
6
8
Purge time (min)
10
12
Figure 5.56: The effect of purge time on resulting average
coating thickness for two different TMA dosage times as
obtained by the EPMA method.
61
Figure 5.55 shows the effect of precursors and as is clear, a large variation in thickness for the
different experiments is observed. For variation of TMA precursor, no clear trend can be observed.
The experiment with 9 cycles of 4 minutes TMA dose was the first experiment performed, which
could explain the relatively low coating thickness found in this experiment compared to the other
experiments. As ALD experiments are relatively complex, some problems might have occurred during
this first experiment that resulted in a lower coating thickness.
If this would indeed be the case, the other samples indicate that the effect of a higher TMA loading
does not increase coating thickness significantly. This is reinforced by the two separate samples at
lower purge times in Figure 5.56, in which hardly any difference in coating thickness could be
detected. Therefore TMA condensation is not likely to be an important factor in residual CVD, which
agrees with findings of other authors [66].
The increase of water dosage on the other hand seems to have a more significant effect, especially if
the coating thickness of sample with five minutes H2O and four minutes TMA dosage time would be
lower than expected due to the aforementioned challenges with the first ALD experiment. However,
due to this observation being based on only two samples, the preparation of more samples is
necessary to ensure that water condensation is indeed an important factor. Based on previous
research discussed in the literature section, this is likely though.
The most significant effect in increasing coating thickness is however in purge time reduction, as
shown in Figure 5.56. For two different TMA dosage times, a decrease of purge time from ten to five
minutes resulted in a significant increase in coating thickness. Based on these experiments combined
with previous work that found that higher temperatures could eliminate rCVD [63], condensation of
reactants, in particular H2O, is indeed the most likely mechanism for the rCVD part of ALD/rCVD.
Therefore, if high coating thickness is desired, the rCVD component should be maximized by short
purges and low temperatures to limit re-evaporation of precursor and precursor amounts sufficient
to allow for maximum condensation and reaction on particle surfaces. The amount of precursor
added should also be significantly more than necessary for complete surface coverage with a single
monolayer, which would correspond to pure ALD.
5.3.2 Increased number of cycles
These recommendations of short purge times, low temperatures and sufficient addition of water and
TMA were therefore followed along with an increased number of cycles to obtain coatings of
sufficient thickness for high temperature protection. The microstructure and morphology of these
particles was again investigated with SEM and EDS, the results of which are shown in Figure 5.57.
Whether an increase in coating thickness was indeed achieved was investigated with the EPMA
method, the results of which are shown in Figure 5.58. To investigate the accuracy of this method,
cross-sections were made for the ALD-25C and ALD-40C samples, an example SEM-BSE image of
which is shown in Figure 5.59 and the coating thickness distributions, shown in Figure 5.60 and Figure
5.61 were compared to the results of the EPMA method.
62
Figure 5.57: SEM images of the two samples with thicker coatings, namely ALD-25C (left) and ALD-40C (right).
The SEM images show that the final result of the ALD/rCVD samples with an increased number of
cycles is similar to the other ALD/rCVD samples. Again, most smaller particles have been removed
from the powder. However, the morphology is slightly different, having a distinctly smoother surface
with edges that are less sharp, indicating that a coating is present. This is confirmed by EDS, which
detects significant quantities of Al on each particle.
The ALD-40C image also shows some small brighter particles that have a different, more flaky
morphology. Images with a higher magnification and EDS measurements show that these particles
are pure alumina particles. These particles most likely formed on the distributor plate and
contaminated the sample during unloading of the powder. However, they are inert and can be
distinguished easily during analysis and do therefore not interfere significantly with other
measurements.
Analysis of the thickness measurements by both the EPMA method and the average of all crosssection measurements in Figure 5.58 confirms that there is indeed a significant amount of alumina
present in these samples. Coatings were found to be significantly thicker than expected based on
extrapolation of the growth per cycle for the samples with lower cycles, with the 25 cycle sample
having an average thickness of approximately 0.65 µm and the 40 cycle sample having a thickness of
1.9 µm, according to the cross-section data. This does mean that growth in this experiment is
distinctly nonlinear though. The most likely explanation is the continuous removal of smaller particles
during the experiment, especially at the moments the distributor plates have to be changed due to
clogging, significantly lowering the total surface area to be coated.
Unfortunately, calculation of the ALD-40C sample coating thickness was not possible due to
numerical instability of the code at these conditions. Therefore, only the cross-section analysis is
available for that sample and no comparison between the EPMA method and the cross-section can
be made for this sample. However, for the ALD-25C sample, both the cross-section and the EPMA
analysis was successful. As is clear from Figure 5.58, these average measurements correspond very
well with each other, the average result of the EPMA method being 665 nm coating thickness and the
average of the cross-section measurements being 650 nm. This difference is well within the margin of
error and therefore it can be concluded that the EPMA based method is a suitable and accurate
method to estimate coating thickness in these samples.
63
2500
Thickness vs Number of cycles
measured by EPMA
Thickness vs Number of cycles
measured by cross-section
Coating thickness (nm)
2000
1500
1000
500
0
0
5
10
15
20
25
Number of cycles (-)
30
35
40
45
Figure 5.58: Measured thickness for ALD/rCVD samples with different number of cycles, all with 4 minutes of TMA
dosage, 5 minutes water dosage and 5 minutes purge per cycle.
Figure 5.59: Cross-section SEM-BSE images of coated particles for the ALD-25C sample (left) and the ALD-40C sample
(right).
As is clear from Figure 5.59, Figure 5.60 and Figure 5.61, there is indeed clearly a coating present,
which is quite homogeneous in thickness, especially when compared to the sol-gel coatings.
Especially noteworthy is the complete absence of very thin coating layers, smaller than 0.35 µm for
the ALD-25C sample and smaller than 1.2 µm for the ALD-40C sample. This would limit local oxidation
at high temperature significantly. The distribution also shows that coating thickness is significantly
more homogeneous than the sol-gel coatings. There is still a distribution, as the deposition from the
rCVD component can vary like regular CVD. Furthermore, there is a human measurement error
associated with measurements from images and another error from the SEM-BSE images, as
described in the sol-gel chapter, which could introduce some additional distribution.
64
Percentage of total measurements (%)
25%
20%
15%
10%
5%
0%
Coating thickness range (µm)
Figure 5.60: Coating thickness distribution for the ALD-25C sample from cross-section analysis.
20%
Percentage of total measurements (%)
18%
16%
14%
12%
10%
8%
6%
4%
2%
0%
Coating thickness range (µm)
Figure 5.61: Coating thickness distribution for the ALD-40C sample from cross-section analysis.
Finally, to ensure that the light gray coatings were indeed Al containing coatings, an EDS linescan was
performed. This linescan is shown in Figure 5.62 and shows clearly that alumina is only present close
to the middle of the scan, which is also the location of the light gray coating surrounding all particles.
Oxygen is also at a maximum in this region, as the expected compound would be Al2O3. The ratios are
not perfect though, as the polymer also contains significant amounts of oxygen and as explained in
appendix V, the EDS detection is actually in a relatively wide area below the surface. This also
explains the presence of Mo and Si in the part of the coating close to the particle.
65
Figure 5.62: A linescan of the coating of the ALD-25C sample with the scanned region (left) and the atomic percentages
detected for each element as a function of distance (right).
5.4
Heat treatment
The main goal of heat treatment of particles is to obtain a completely densified coating of α-Al2O3
without damaging the particles or the coating. This chapter will however show that this is not very
straightforward and will show the required precautions necessary to prevent damage. First, the
coatings resulting from selecting an oxygen containing annealing atmosphere will be shown, followed
by heat treatment of sol-gel coatings under an atmosphere of pure argon. Afterwards, the heat
treatment of ALD/rCVD coated particles will be discussed.
5.4.1 Effect of atmosphere
The main effect of atmosphere is related to the presence of oxygen or oxygen providing molecules
such as H2O and CO2. Before annealing, coatings are not yet fully densified and might be sensitive to
any oxygen providing molecules. To test this, a small part of the SG-10g sample has been subjected
to heat treatment in air instead of argon. Both precalcination at 450 ᵒC only and high temperature
annealing at 900 ᵒC were tested, the results of which are shown in Figure 5.63.
Figure 5.63: SEM images of precalcined SG-10g sample (left) and high temperature annealed sample (right).
These two images clearly show the problem of annealing in air, which leads to significant cracking,
most likely due to oxidation of the MoSi2 and resulting stress buildup beneath the coating. This effect
is already observed during the precalcination that is only performed at 450 ᵒC, which is apparently
still too high to prevent significant oxidation of the substrate. This agrees with the known oxidation
66
behavior of MoSi2, which usually becomes significant around 400 ᵒC [17] and with the observations
during TGA in which the blank started to oxidize around 450 ᵒC. Based on these observations, it was
decided to perform the entire procedure in argon, except for the first 30 minutes of precalcination to
allow oxygen to burn away residual carbon.
5.4.2 Heat treatment of sol-gel coatings
As the main goal of heat treatment was to obtain an intact and completely densified α-Al2O3 coating,
both morphology and crystal structure were investigated by SEM and XRD respectively. Interestingly,
morphology was found to be very similar for all heat treated sol-gel coatings. An example of these
morphologies is shown in Figure 5.64. Furthermore, XRD diffractograms for the SG-10g and SG-20g
samples annealed at different temperatures are shown in Figure 5.65 and Figure 5.66 respectively.
Figure 5.64: Morphology of sol-gel samples after heat treatment with the precalcined (450 ᵒC, 14h) only SG-20g sample
(left) and the SG-20g sample subsequently annealed at 1200 ᵒC (right).
These images show that the sol-gel coatings are still intact after heat treatment, which according to
literature is quite surprising. Sol-gel coatings often have a tendency to crack on many substrates, due
to tensile stresses resulting from densification to the equilibrium density as described in chapter
2.4.1. Although some very small cracks were observed in some of the samples, most of the particles
did not show any cracks large enough to be visible on the SEM.
The most likely explanation for this is the precalcination in combination with annealing in pure argon.
Previous investigations found that cracking was more likely when the precalcination step was not
performed, due to the significantly faster densification. The stresses caused by this densification
result in fracture of the coating, as there is no time for any stress relaxation with temperatures of
900 ᵒC. However, chapter 5.4.1 clearly shows the necessity of pure argon or another inert gas that
does not contain any oxygen. Therefore it is concluded that both precalcination and argon are
necessary for prevention of crack formation.
The x-ray diffractograms in Figure 5.65 and Figure 5.66 show some interesting results. First of all,
MoSi2 is still the main phase, as would be expected. Furthermore, all diffractograms except for the
sol-gel samples that were annealed at 1200 ᵒC show the presence of a small amount of Si, which
indicates that oxidation was mostly prevented during annealing. The most striking observation
however is the almost complete absence of the α phase for all samples except for the 1200 ᵒC
samples and the SG-20g 1100 ᵒC sample. This is however in agreement with most literature, as the
θ→α transition often occurs at temperatures above 1100 ᵒC [23].
67
□
□
□ MoSi2
• Si
◊ α-Al2O3
□
□
Precalcination 450C
900C
1000C
1100C
□
•
Before calcination
□
□
□
□
1200C
□
Intensity (A.U.)
•
◊
15
25
◊
35
45
55
65
75
85
2θ (ᵒ)
Figure 5.65: XRD diffractograms of the SG-10g sample annealed at different final temperatures and including the sample
before any heat treatment and after precalcination.
□ MoSi2
• Si
◊ α-Al2O3
□
Precalcination 450C
900C
1000C
1100C
1200C
□
□
□
□
Intensity (A.U.)
25
□
□
□
□
•
•
◊
15
□
□
◊
35
◊
◊
◊
45
2θ (ᵒ)
◊
◊
55
65
75
85
Figure 5.66: XRD diffractograms of the SG-20g sample annealed at different final temperatures and including the sample
after precalcination.
Furthermore, the diffractograms show that transformation kinetics are very dependent on coating
thickness, as the SG-20g sample shows the presence of significantly more α-alumina than the SG-10g
68
sample. Furthermore, even though transition alumina's are likely to form, the peaks associated with
them are relatively small. They are present however, although the lattice constants of the various
transition phases are very similar, making it difficult to distinguish these phases. A good example for
this behavior would be the small peak next to the MoSi2 peak at a 2θ of 66ᵒ, where most of the
transition aluminas give a signal. This peak increases in intensity with increasing temperature, but
intensity decreases again at 1100 ᵒC and 1200 ᵒC, which is a good indication that the transition
alumina sequence is indeed a likely transformation path for these coatings. Based on these XRD
diffractograms, it is however not possible to determine the precise transition aluminas formed during
heat treatment.
A possible explanation for the relative lack of transition phase peaks could be the small grain size,
which would cause significant peak broadening in XRD analysis. As these aluminas form from an
amorphous gel, grains are often small and do not grow significantly until the sample reaches higher
temperatures. At these temperatures however, the transformation to other aluminas and the final α
phase is also more likely, reducing the amount of transition alumina phase that could be detected.
This would also explain why the α-Al2O3 peaks are relatively broad as well.
Figure 5.67: Cross-section SEM images of a heat treated particle, namely SG-20g at 1200 ᵒC (with precalcination),
showing a BSE image (left) and a SEM image (right).
To investigate other changes in morphology and changes in thickness during calcination, a crosssection of the SG-20g sample calcined at 1200 ᵒC was also prepared. SEM images of this cross-section
are shown in Figure 5.67 and the coating thickness distribution is shown in Figure 5.68. Comparing
the SEM-BSE images, no significant differences can be observed. When comparing the SEM-SEI
images however, the coating seems to contain significant amounts of white spots that were not
present in the coatings that were not heat treated. These small white spots are most likely α-alumina
grains and their size corresponds surprisingly well with the calculations from the XRD analysis,
measuring approximately 50 nm on the SEM-SEI images.
Analysis of the coating thickness distribution reveals that there is surprisingly little difference
between annealed and fresh coating thickness. Some difference is visible in the range of coating
thickness of more than 1 µm, but for coatings with a thickness lower than 1 µm, there is no
statistically significant difference. This might indicate that the coating has not completely densified
and transformed during heat treatment, especially in the regions of thinner coatings. This is in
agreement with the SEM-SEI observations in which the white spots were mainly found in the thicker
69
parts of the coating. If this is indeed the case, heat treatment at higher temperatures and for longer
periods of time might be necessary to transform the coating completely to α-alumina.
9%
Percentage of total measurements (%)
8%
7%
6%
5%
4%
3%
2%
1%
<0
0-0.05
0.05-0.1
0.1-0.15
0.15-0.2
0.2-0.25
0.25-0.3
0.3-0.35
0.35-0.4
0.4-0.45
0.45-0.5
0.5-0.55
0.55-0.6
0.6-0.65
0.65-0.7
0.7-0.75
0.75-0.8
0.8-0.85
0.85-0.9
0.9-0.95
0.95-1
1-1.05
1.05-1.1
1.1-1.15
1.15-1.2
1.2-1.25
1.25-1.3
1.3-1.35
1.35-1.4
1.4-1.45
1.45-1.5
1.5-1.55
1.55-1.6
1.6-1.65
1.65-1.7
1.7-1.75
1.75-1.8
1.8-1.85
1.85-1.9
1.9-1.95
1.95-2
>2
0%
Coating thickness range (µm)
Figure 5.68: Coating thickness distribution for the SG-20g sample heat treated at 1200 ᵒC.
5.4.3 Heat treatment of ALD/rCVD coatings
The goal for heat treatment of ALD/rCVD coatings was the densification of the coating combined
with the transformation from an amorphous layer to α-alumina, while retaining shell integrity. SEM
image of ALD-25C and ALD-40C are shown in Figure 5.69 and were used to investigate integrity, while
XRD spectra of the samples annealed at different temperatures are shown in Figure 5.70 and Figure
5.71 to observe transformations.
Figure 5.69: Morphology of ALD samples after heat treatment with the 25 cycle sample (left) and the 40 cycle sample
(right), both annealed at 1200 ᵒC.
The most striking observation is the appearance of significant cracks in the coating, especially in the
ALD-40C sample. For the ALD-40C sample, most particles had a very similar morphology to the
particle shown in Figure 5.69. An EDS map of the particle shown in this figure was also prepared to
ensure these were indeed cracks and that they reached through the entire coating. These elemental
maps are available in appendix XIV and do indeed show the presence of only Al and O outside the
cracks, but significant amounts of Mo and Si and hardly any Al inside the cracks.
70
The ALD-25C sample on the other hand did show a few particles with cracks, but most particles did
not contain any. This indicates that thickness is indeed a significant factor, as described in chapter
2.5. Opposed to sol-gel coatings, ALD/rCVD particles do not have any mechanisms to relieve stress
from densification and transformation, such as pore formation or evaporation of water, resulting in
more fracture. However, as is clearly shown by the ALD-25C sample, manufacturing coatings with
appropriate thickness can circumvent this problem almost entirely, as would also be expected.
□ MoSi2
• Si
◊ α-Al2O3
□
Precalcination 450C
900C
1000C
1100C
1200C
□
□
□
□
□
Intensity (A.U.)
•
15
25
□
□
□
□
□
•
35
45
55
2θ (ᵒ) (ALD-25C)
65
75
85
Figure 5.70: XRD diffractograms of the ALD-25C sample annealed at different final temperatures and including the
sample after precalcination.
Analysis of the XRD results reveals some additional interesting observations. Similar to the sol-gel
coatings, MoSi2 is the main component and some small Si peaks are present. Again, annealing at
1200 ᵒC results in the disappearance of the Si peak, most likely due to some minor oxidation.
However, no alumina can be detected in any sample other than the ALD-25C sample that was
annealed at 1200 ᵒC and the ALD-40C samples that were annealed at 1100 and 1200 ᵒC. And even in
those cases, the α-Al2O3 peaks that are detected, are very small, especially compared to the sol-gel
derived coatings.
Based on this, it can be concluded that the transformation of ALD/rCVD based coatings to the stable
α-alumina coatings is significantly more difficult than the transformation of the sol-gel based coatings
and most likely means that this transformation has a significantly higher activation energy. This
agrees with observations in literature, in which this transformation was indeed found to be difficult,
as described in the theory section regarding heat treatment.
Another observation is the complete absence in any sample of peaks of any of the sol-gel transition
aluminas. This might be caused by the transition sequence following the amorphous → κ → α
transition, as some hints of κ-alumina were detected. Furthermore, both κ and α alumina have an
71
HCP anion packing, making differentiation between these phases on XRD more difficult. However,
this might also be caused by the relatively small amount of crystallized alumina detected in general,
combined with the small grain size of the α-alumina. More experiments with higher amounts of
crystallized Al2O3 would be necessary to draw definite conclusions, but the κ → α transformation
seems the most likely explanation based on these XRD analyses.
Precalcination 450C
□ MoSi2
• Si
◊ α-Al2O3
1100C
□
1200C
□
□
Intensity (A.U.)
□
•
25
□
□
□
□
•
◊
15
□
□
□
35
◊
◊
45
◊
55
65
75
85
2θ (ᵒ) (ALD-40C)
Figure 5.71: XRD diffractograms of the ALD-40C sample annealed at different final temperatures and including the
sample after precalcination.
It can definitely be concluded however that not all alumina has transformed to the stable α-Al2O3
phase. The cracks present in the ALD-40C sample do however show that the alumina has densified
significantly, although not completely due to the limited amount of α alumina detected, which is the
densest form of alumina. Therefore, ALD/rCVD coatings would need higher annealing temperatures
and probably longer annealing times to obtain a coating containing mostly the stable α phase.
5.5
Performance
5.5.1 Thermogravimetric stability
To test whether coated particles would be able to resist oxidation for the time required, a small
amount of 30 mg was supposed to be loaded in a TGA and kept at 1000 ᵒC for 100 h. Unfortunately,
due to time limitations and equipment failure, only a MoSi2 and MoSi2B blank were tested, along
with a sample of MoSi2B alloyed with additional 6 wt% of Al and with the sol-gel method, using 20 g
of aluminium tri-sec-butoxide. The results of these tests are shown in Figure 5.72. Although the
MoSi2 sample has a slow mass increase at 1000 ᵒC as would be expected due to the formation of a
slow-growing SiO2 layer, there is surprisingly little weight loss at the start of the experiment, in which
significant MoO3 would be expected to evaporate.
72
For MoSi2B on the other hand, the weight gain is significantly more than expected at the start,
followed by a surprisingly fast weight loss. This trend is however not as sharp as the graph shows, but
is rather continuous during the first hour. It is uncertain whether this is due to machine errors or due
to rapid initial oxidation in which the boron present could limit the formation of a closed scale, as in
the regular MoSi2.
110
MoSi2
MoSi2B
108
MoSi2B 6wtAl SG20g
Relative mass (% of initial mass)
106
104
102
100
98
96
0
10
20
30
40
50
60
70
80
90
100
Time (h)
Figure 5.72: Relative weight change as a function of time for two blanks and the MoSi 2B 6wt% Al SG20g coated sample
during a TGA test at 1000 ᵒC in synthetic air for 100h.
Finally, the coated sample seems to have no weight change at all any more after a rapid initial weight
loss. This weight loss might be caused again by a problem with the machine or mass loss from
evaporation of water or carbon species that were not removed during annealing and could rapidly
oxidize due to contact with oxygen at a high temperature. Nevertheless, after this initial weight loss
the coating seems to protect the sample well, even compared to the blanks. This might however also
be partially due to the formation of mullite instead of SiO2 due to MoSi2 alloying with Al, which would
limit oxidation significantly, as was shown in chapter 3. An attempt to test the model and this
possibility with coated Si wafers was unfortunately not possible due to time limitations.
5.5.2 Embedded particle stability and healing
An important aspect of testing coated particles is investigating their performance in an actual
thermal barrier coating (TBC). To this end, the SG-20g sol-gel procedure was repeated with wind
sifted MoSi2B particles, which were subsequently mixed with yttria-stabilized zirconia (YSZ) to obtain
a composite consisting of 80 V% YSZ and 20 V% healing particles. The results of the repeated coating
procedure are shown in Figure 5.73 (SEM) and Figure 5.74 (XRD). EDS results presented in appendix
73
XV show that also in this case complete coverage with Al2O3 was most likely achieved, as all
measurement points do again show the presence of significant amounts of Al.
Figure 5.73: SEM images at different magnifications of MoSi2B coated with Al2O3 according to the SG-20g sol-gel
procedure and heat treated at 450 ᵒC and 1200 ᵒC in argon.
Both the SEM images and the XRD diffractograms are very similar to the SG-20g sample, indicating
that both morphology and behavior during heat treatment and under high temperature conditions
are similar. The main difference in the MoSi2B sample is the presence of MoB2 in the samples before
heat treatment. Surprisingly, this MoB2 seems to disappear during heat treatment, and minor
amounts of Mo seem to appear.
□
Intensity (A.U.)
□
◊
Before heat treatment
□
◊
Precalcination 450C
□
□
◊
◊
□◊
□
□
1200C
□
◊
□ MoSi2
• Si
◊ α-Al2O3
⁰ MoB2
•
⁰
15
◊
□
25
•
⁰
35
⁰
45
55
2θ (ᵒ) (MoSi2B)
65
75
85
Figure 5.74: XRD diffractograms of the coated MoSi2B particles before and after heat treatment.
An explanation for this behavior could be limited oxidation of the sample during heat treatment, as
boron oxidizes before silicon or molybdenum and in this sample the residual Si peak is still detected.
It is also very difficult to completely prevent the presence of oxygen in a furnace without utilizing a
74
reducing atmosphere. However, this oxidation is not necessarily problematic as long as the coating
does not crack or gets damaged and sufficient boron is still present in the particles. It could even be
beneficial as it could also remove some organic impurities that might still be present. Furthermore,
the main peaks of MoB2 overlap with some of the main peaks of corundum, making detection of
MoB2 more difficult once α-Al2O3 forms. Therefore it is likely that there is still some MoB2 present,
but that it is not detectable anymore because of the presence of corundum.
In any case the main conclusion from this data is that the sol-gel procedure is reproducible and able
to coat MoSi2 particles with and without boron with a closed layer of α-Al2O3 after appropriate heat
treatment.
For embedding particles in YSZ, several samples without MoSi2B powder were first prepared to
investigate spark plasma sintering (SPS) of YSZ. The XRD diffractograms of these pure YSZ samples
and the powder from which they were prepared are shown in Appendix XVI and show the presence
of major carbon and minor ZrC and ZrB2 impurities originating from reactions with graphite mould
and boron nitride lubricant necessary during the SPS process. Furthermore, stress development and
sintering is studied by investigating peak broadening at the peak with the highest intensity, which is
the tetragonal ZrO2 peak at a 2θ of 30ᵒ. These overlapping peaks are also presented in Appendix XVI
and show significant differences in peak broadening compared to the powder. These two
observations indicate that some heat treatment might be necessary to relieve stresses and remove
impurities after SPS.
Damage for self-healing was introduced by indentation and also tested on the blank samples. By
using equation 5.2 [106], fracture toughness of the composite can be calculated and compared to the
regular YSZ samples. In this formula, KIC is the fracture toughness in MPa√m, E is the Young modulus
in GPa, H is the measured hardness in GPa, P is the maximum force achieved during indentation in N,
C0 is the total crack length as measured from the center of the indent in m and § is a correction factor
for the equipment and conversion of Pa to MPa combined, with a value of 1.6*10 -8 [106]. The results
of these calculations are shown in Table 5.11. The Young modulus of 200 GPa for YSZ was based on
the work of Christel et al. [107].
(5.2)
Table 5.11: Measured hardness and crack length from Vickers HV10 indentation and resulting fracture toughness for an
SPS sample of YSZ and the YSZ-MoSi2B composite.
Sample
YSZ
MoSi2B
KIC
0.5
(MPa*m )
11.7 ±0.45
9.3 ±0.78
Youngs Modulus E Hardness measured
(Gpa)
(GPa)
200
1.32 ±0.014
200
1.24 ±0.060
Force used
(N)
98.1
98.1
crack length
micrometer
140
±3.3
167
±9.7
When comparing YSZ toughness measured here with the data of Christel et al. [107] and other
authors, a fracture toughness of 11.7 MPa√m is significantly higher than most values for partially
yttria stabilized zirconia, which are usually between 9 and 10 MPa√m. A possible explanation for this
might be contamination with carbon and boron during SPS, which might be able to produce carbides
and borides of zirconia that could increase toughness. This contamination is observed in XRD
diffractograms of the SPS samples.
75
Another possibility might be that cracks formed by indentation with HV10 are slightly too short for
accurate measurements. This would result in an overestimation of fracture toughness, as crack
length is measured from the center of the indent. The second explanation seems more likely, as the
toughness of both ZrC [108] and ZrB2 [109] are significantly lower than the toughness of YSZ.
When comparing the composite with the YSZ sample, a clear difference in fracture toughness is
observed, with the composite containing 20 V% MoSi2B having a significantly lower fracture
toughness. As sintering parameters were the same, it is likely that this difference is caused by the
particles. Although the manufacture of composites often results in a higher fracture toughness, in
this case a lower toughness could be expected, as MoSi2B has a significantly lower KIC than YSZ [110].
Therefore, addition of 20 V% particles of this more brittle material could easily decrease the overall
fracture toughness of the composite. Furthermore, as explained in chapter 3 it is likely that the
interface between the Al2O3 coating and YSZ is relatively weak, providing another crack path with
lower resistance.
Investigation of SEM images of the composite samples themselves, presented in Figure 5.75, also
shows some interesting results. The first and main result is that particles are indeed properly
dispersed in the YSZ powder and that they are still intact after SPS. EDS maps shown in appendix XVII
also indicate that the particle shell was able to protect the MoSi2B particles against reactions with
YSZ, as no O or Zr was detected in the particles, while Mo and Si were the only elements detected
there in significant amounts. Attempts to manufacture a double layer system of uncoated MoSi2 and
YSZ did show reactions between the two powders under SPS conditions, indicating that the coating is
indeed able to protect particles to a certain extend.
Figure 5.75: SEM-BSE images of two different indents at different magnifications with HV10 (left) and 200N force (right).
SEM images also indicate that indentation is a suitable method to introduce damage. Cracks are
sometimes relatively short compared to the indent itself though, making calculation of fracture
toughness less accurate. Another observation is that cracks in general grow through the healing
particles instead of deflecting. This was not exactly as expected, as computer models predicted that
cracks would often grow around particles, but proves that crack growth through the shell is possible.
It has to be noted though that indentation in this project was performed at room temperature, while
damage usually occurs during engine cooling and therefore at higher temperatures. MoSi2 is
significantly less brittle at temperatures above 1000 ᵒC, which might change crack growth behavior. It
76
is therefore recommended to attempt damage introduction at higher temperatures as well to
observe crack growth behavior under operating conditions.
Healing behavior was also investigated by annealing the sample in air at 1100 or 1200 ᵒC. SEM
images of one of the indents before and after healing is shown in Figure 5.76. One of the clear main
problems is the clearly visible extensive fracture after heat treatment in air. This fracture is visible
everywhere on the composite and not just close to the indents. The suspected cause is premature
oxidation and associated volume expansion of particles without any crack nearby. This is supported
by formation of cracks and oxidation products in the sample that was subjected to 48h of air at 1100
ᵒC without damage introduction. Furthermore, significant swelling was observed after heat
treatment of the composites, but only on the side that was in contact with the gas phase.
Figure 5.76: SEM images of an indent before (left) and after (right) heat treatment in air at 1100 ᵒC for 1 hour (heating
and cooling rate 5 ᵒC/min).
However, as shown in Figure 5.77, both smaller cracks and larger cracks that were already present at
the start, do actually fill with material. Based on the BSE image, this material also contains low
amounts of heavier elements, indicating that this is indeed Si filling and possibly healing the cracks.
Based on these observations, better self-healing might be obtained by adjustment of the healing
parameters, such as creating indents with smaller loads, which would result in smaller cracks to heal
and less significant residual stresses. These smaller cracks could then in turn be healed at lower
temperatures or shorter times. Thicker coatings to limit oxidation of non-damaged particles might
also help, although modelling does not seem to agree with this and a more detailed TGA
investigation is recommended. Furthermore, the addition of a small amount of Al to the MoSi2B
could aid in limiting premature oxidation by acting as a sacrificial element that would protect the
MoSi2B from further oxidation. Lowering the content of healing material could also decrease swelling
and still allow for sufficient healing material to repair any damage. This would also decrease
toughness reduction of the composite, at the cost of having less healing material available.
77
Figure 5.77: SEM images of cracks close to an indent with a BSE image (left) and a SEI image (right), showing the presence
of crack filling.
78
Conclusions and Recommendations
6.1 Conclusions
The goal of this thesis was to find an optimal route to produce Al2O3 coated MoSi2 particles and show
the validity of the self-healing Thermal Barrier Coating (TBC) concept. To this end, two chemical
methods were optimized for coating MoSi2 particles, namely Atomic Layer Deposition and a sol-gel
method.
Initial calculations and oxidation modeling did however show the necessity to reduce surface area to
limit the volume fraction of inert coating material in the healing particles. This reduction in total
surface area, along with an increase of average particle size and a removal of most small particles
was successfully accomplished using wind sifting.
For the sol-gel method, the main challenge was found to be proper coverage of the particles with
coating material. This was successfully solved by ensuring prolonged contact between the alumina
gel and the particles with nitrogen bubbling in a slurry bubble column reactor setup, selecting a
suitable precursor and controlling pH to ensure favorable electrostatic interactions between particle
surfaces and the gel. It was found that liquid aluminium tri-sec-butoxide was the most suitable
precursor due to its miscibility with water and its reactivity, resulting in rapid gel formation.
For ALD the main challenge was reaching reasonable thickness in a feasible number of cycles. It was
found that by utilizing residual chemical vapor deposition (rCVD), growth per cycle could be
increased dramatically, resulting in coatings of several 100s of nanometers up to 1.8 µm in only 2540 cycles. The rCVD component is thought to be governed mainly by H2O condensation and the
subsequent reaction of trimethylaluminium with this condensed water.
To obtain stable α-alumina coatings, heat treatment was also necessary. It was found that heat
treatment in argon was necessary to prevent premature oxidation. This two-step heat treatment
resulted in the formation of detectable α-alumina in the coatings of both the final sol-gel and
ALD/rCVD samples, according to the XRD results. This formation of alpha alumina was without
significant crack formation in all samples except for the thickest ALD coating. Therefore it can be
concluded that both methods are suitable to produce MoSi2 coated with α-Al2O3 as long as coating
thickness is properly controlled.
Comparison of the two different methods shows a difference in both microstructure and coating
morphology. Although comparable average coating thickness can be obtained and was
approximately obtained between the SG20g and ALD25C samples, the thickness distribution was
found to be significantly wider for the sol-gel method. Furthermore, behavior during heat treatment
was very different, with transformation to the stable α phase being significantly easier in coatings
produced with the sol-gel method, compared to the transformation of the coatings produced by ALD.
The most likely explanation for this lower activation energy of the transformation sequence in sol-gel
coatings would be a different transformation pathway, with the γ→δ→θ→α being the most likely
path for sol-gel, while the amorphous→κ→α being the most probable transformation pathway for
the ALD samples. XRD data is however inconclusive on this and further research is recommended.
Due to these differences in transformation kinetics, it is likely that sol-gel coatings will have a larger
alumina grain size.
79
Resistance of the produced Al2O3 coatings to oxidation was modeled using a Wagner model approach
and found by varying the most important parameters that surprisingly Al2O3 initial coating thickness
was not an important parameter, while alumina grain size, temperature and mullite layer thicknes
were found to be very influential parameters in the prevention of oxidation. Unfortunately,
experimental validation of these predictions was not possible due to time constraints. Nevertheless,
this does indicate that it is important to promote the formation of larger α-Al2O3 grains and mullite
formation by higher annealing temperatures and/or aluminium alloying in the MoSi2.
Based on TGA stability tests, the model does however overestimate the oxidation speed, even of the
bare sample, although the order of magnitude seems to be correct. The coating does however aid in
protection for the MoSi2B 6wt% Al sample coated with the sol-gel method, when comparing this
sample with the MoSi2 and MoSi2B blanks, although no MoSi2B 6wt% Al blank was available. Due to
the presence of Al in the base material, mullite formation during oxidation might be the main reason
for this though. Furthermore, there is some uncertainty regarding the validity and accuracy of the
TGA results.
Finally, it was found that embedding coated particles in a matrix of yttria-stabilized zirconia (YSZ) is
possible by SPS and that the alumina coating is able to prevent oxidation during the manufacturing
process. However, oxidation and volume expansion of particles during operation was found to be a
problem which resulted in swelling and damage to the composite, indicating that care must be taken
with designing healing composites. However, evidence of crack filling, SiO2 transport along cracks and
zircon formation has also been found with SEM and EDS, showing the feasibility of this self-healing
concept.
6.2 Recommendations
For the continuation of this research, the composite results indicate that the stability of the particles
might be a major challenge. It is therefore recommended to experimentally validate the oxidation
model and its main conclusion that mullite formation is important in preventing further oxidation of
particles.
Furthermore, it is important to perform TGA analyses on the coated and heat treated powders to
investigate whether they are stable under operation conditions and whether there is a significant
difference between the two methods of coating manufacture.
If the thickness of the alumina is indeed found to be less important, it is recommended to continue
coating particles with the sol-gel method, as it is a less sensitive, less dangerous and more
straightforward method. If there is a significant difference between the methods though, a careful
analysis between the advantages and disadvantages of each method is recommended.
Regarding the production methods, it is recommended to try annealing at higher temperatures to
transform more alumina to the stable alpha phase, especially for the ALD samples. Addition of αalumina nanoparticles might also be an option to increase the transformation kinetics of the stable
phase. Furthermore, annealing at higher temperatures might promote formation of additional
mullite, protecting the particles even more.
However, due to the high grain boundary diffusivity found in alumina, it might be even more
interesting to retain as much of the amorphous phase as possible, as this phase does not have any
80
grain boundaries that could act as a fast pathway for diffusion. For sol-gel samples this is very
difficult, but for ALD/rCVD samples, this might be feasible if the pathway is indeed through κ
alumina. As shown in the results, the formation of alpha alumina is challenging even at 1200 ᵒC,
indicating that the amorphous phase might be stable sufficiently long at regular TBC operating
temperatures. If this pathway is preferred, it is also recommended that the transformation kinetics of
the ALD/rCVD samples are investigated more thoroughly by XRD.
If the ALD/rCVD coating path turns out to be the more suitable coating production method, it might
be necessary to perform more experiments to determine the exact mechanism of the rCVD
component and whether the growth mechanism is indeed nonlinear or whether this is only due to a
reduction in total surface area. More experiments will also help in establishing a number of cyclesthickness relation that will be necessary if the coating thickness is to be controlled.
Finally, for optimization of the self-healing system, the first recommendation is to investigate the
influence of alloying Al in the MoSi2B particles, especially to find out whether the added Al can
decrease swelling by reducing oxidation of particles free of damage and nearby cracks. Varying the
volume percentage of healing particles added might also influence the swelling and damage creation
observed and is therefore also recommended. Furthermore, careful consideration of the testing
procedure for damage healing might change the performance during healing, as it is suspected that
indentation with large loading might damage the composite more than expected. Decreasing the
indentation load or using a different method for healing tests might give more accurate results.
81
Acknowledgements
This project would not have been finished without the contributions of several important people and
I would like to thank you all for your help and advice. First of all, I would like to express my sincere
gratitude to my daily supervisors Alexandra-Lucia Carabat and Wim Sloof for their support and
guidance during this project and for their valuable advice both regarding the project and many other
topics. I would also like to thank Ruud van Ommen and Erik Kelder for supervising me during this
project and for the great and helpful discussions. Furthermore, thanks to professor Sybrand van der
Zwaag for being part of the thesis committee and undoubtedly asking those surprising questions that
nobody else could think of.
I would also like to express my thanks to the people of the groups I had the pleasure to work in,
namely the Virtual Materials and Mechanics group of Materials Science and the battery group and
Product and Process Engineering group of Chemical Engineering. In particular I would like to thank
Mojgan Talebi, Wim van Oordt, Henk Nugteren and David Valdesueiro for their help with atomic
layer deposition and wind sifting, both of which I was not familiar with at all at the start of this
project. I would also like to thank Aris Goulas and Filipe Lopez for not bringing a velociraptor suit to
my defence.
From the Virtual Materials and Mechanics group I would like to thank Hans Brouwer, Kees
Kwakernaak and Ruud Hendrikx in particular for their help with all the beautiful (and expensive)
"toys" present in the materials science department. Without their help, it would have been
impossible to do all those characterization experiments required for this work.
Finally I would like to thank everyone that I did not mention for helping me out when I was stuck,
giving me such a great time during my master thesis and for creating such an enjoyable atmosphere
in all of the groups I worked in. It was a pleasure to work with you all and I hope to see you again in
the future!
82
Bibliography
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
Meadows, D.H., et al., The limits to growth. New York, 1972. 102.
Kitching, C., The true cost of transatlantic travel: Infographic reveals how passengers pay
£152 in tax on a London to New York flight (that's a QUARTER of your ticket price), in The
Daily Mail. 2015: London.
Tipler, P.A. and G. Mosca, Physics for scientists and engineers. 2007: Macmillan.
Duhl, M.G., DN, and A. Giamei, The development of single crystal superalloy turbine blades.
Superalloys 1980, 1980.
Clarke, D.R., M. Oechsner, and N.P. Padture, Thermal-barrier coatings for more efficient gasturbine engines. MRS Bull, 2012. 37(10): p. 891-898.
Sloof, W.G., Self healing in coatings at high temperatures, in Self Healing Materials. 2007,
Springer. p. 309-321.
Carabat, A.L., S. Zwaag, and W.G. Sloof, Creating a Protective Shell for Reactive MoSi2
Particles in High‐Temperature Ceramics. Journal of the American Ceramic Society, 2015.
Langston, L.S., efficiency by the numbers, in From the editors Desk; The official blog of ASME's
Mechanical Engineering magazine, J. Falcioni, Editor. 2012, Mechanical Engineering
Magazine.
Padture, N.P., M. Gell, and E.H. Jordan, Thermal barrier coatings for gas-turbine engine
applications. Science, 2002. 296(5566): p. 280-284.
Clarke, D.R. and S.R. Phillpot, Thermal barrier coating materials. Materials Today, 2005. 8(6):
p. 22-29.
Iacovides, H. and M. Raisee, Recent progress in the computation of flow and heat transfer in
internal cooling passages of turbine blades. International Journal of Heat and Fluid Flow,
1999. 20(3): p. 320-328.
Vaßen, R., et al., Overview on advanced thermal barrier coatings. Surface and Coatings
Technology, 2010. 205(4): p. 938-942.
Hille, T.S., et al., Damage growth triggered by interface irregularities in thermal barrier
coatings. Acta Materialia, 2009. 57(9): p. 2624-2630.
Turteltaub, S., Powerpoint presentation: Self-healing Thermal barrier coatings based on
MoSi2. 2013.
Guimard, N.K., et al., Current Trends in the Field of Self‐Healing Materials. Macromolecular
Chemistry and Physics, 2012. 213(2): p. 131-143.
Van der Zwaag, S., et al., Self-healing behaviour in man-made engineering materials:
bioinspired but taking into account their intrinsic character. Philosophical Transactions of the
Royal Society of London A: Mathematical, Physical and Engineering Sciences, 2009.
367(1894): p. 1689-1704.
Mao, W., Kinetics of self-healing reaction in TBC with MoSi2 based sacrificial particles. 2013,
TU Delft, Delft University of Technology.
Sloof, A.L.C.W.G., Powerpoint presentation: Encapsulation of MoSi2 self-healing particles for
high temperature applications. 2014.
Eaton, H.E., et al. EBC Protection of SiC/SiC Composites in the Gas Turbine Combustion
Environment: Continuing Evaluation and Refurbishment Considerations. in ASME Turbo Expo
2001: Power for Land, Sea, and Air. 2001. American Society of Mechanical Engineers.
Singhal, S., Advances in solid oxide fuel cell technology. Solid state ionics, 2000. 135(1): p.
305-313.
Knittel, S., S. Mathieu, and M. Vilasi, The oxidation behaviour of uniaxial hot pressed MoSi 2
in air from 400 to 1400 C. Intermetallics, 2011. 19(8): p. 1207-1215.
Carabat, A., S. Van der Zwaag, and W. Sloof. Encapsulation of sacrificial silicon containing
particles for SH oxide ceramics via a boehmite precursor route. in ICSHM 2013: Proceedings of
the 4th International Conference on Self-Healing Materials, Ghent, Belgium, June 16-20,
2013. 2013. Ghent University; Delft University of Technology.
83
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
Levin, I. and D. Brandon, Metastable alumina polymorphs: crystal structures and transition
sequences. J. Am. Ceram. Soc., 1998. 81(8): p. 1995-2012.
Patnaik, P., Handbook of inorganic chemicals. Vol. 28. 2003: McGraw-Hill New York.
Strafford, K., 'Protective oxide scales and their breakdown'. Materials Science and
Technology, 1998. 14(11): p. 1200.
Moulijn, J.A., M. Makkee, and A.E. Van Diepen, Chemical process technology. 2013: John
Wiley & Sons.
Askeland, D.R. and P.P. Phulé, The science and engineering of materials. 2003.
Chiang, Y.-M., W.D. Kingery, and D.P. Birnie, Physical ceramics: principles for ceramic science
and engineering. 1997: J. Wiley.
Thomas, O., et al., Molybdenum disilicide: crystal growth, thermal expansion and resistivity.
Solid state communications, 1985. 55(7): p. 629-632.
d’Heurle, F., C. Petersson, and M. Tsai, Observations on the hexagonal form of MoSi2 and
WSi2 films produced by ion implantation and on related snowplow effects. Journal of Applied
Physics, 1980. 51(11): p. 5976-5980.
Fujiwara, H. and Y. Ueda, Thermodynamic properties of molybdenum silicides by molten
electrolyte EMF measurements. Journal of alloys and compounds, 2007. 441(1): p. 168-173.
Subbarao, E. and K. Gokhale, Thermal expansion of zircon. Japanese Journal of Applied
Physics, 1968. 7: p. 1126.
Singhal, S.C. and K. Kendall, High-temperature solid oxide fuel cells: fundamentals, design and
applications: fundamentals, design and applications. 2003: Elsevier.
Lide, D.R., CRC handbook of chemistry and physics. 2004: CRC press.
Koike, C., et al., Infrared Spectra of Silica Polymorphs and the Conditions of Their Formation.
The Astrophysical Journal, 2013. 778(1): p. 60.
Lager, G.A., J. Jorgensen, and F. Rotella, Crystal structure and thermal expansion of α‐quartz
SiO2 at low temperatures. Journal of Applied Physics, 1982. 53(10): p. 6751-6756.
Heaney, P.J., Structure and chemistry of the low-pressure silica polymorphs. Reviews in
Mineralogy and Geochemistry, 1994. 29(1): p. 1-40.
Pierre, A.C., Introduction to sol-gel processing. Vol. 1. 2013: Springer Science & Business
Media.
Brinker, C.J. and G.W. Scherer, Sol-gel science: the physics and chemistry of sol-gel
processing. 2013: Academic press.
Atkins, P. and J. De Paula, Elements of physical chemistry. 2013: Oxford University Press.
Dressler, M., Sol-gel preparation and characterization of corundum based ceramic oxidation
protection coatings. 2006.
Shi, J., Steric stabilization. Croup Inorganic Materials Science–Literature review, The Ohio
State University. Columbus, USA, 2002.
Attia, Y., Sol-gel processing and applications. 2012: Springer Science & Business Media.
Kobayashi, Y., T. Ishizaka, and Y. Kurokawa, Preparation of alumina films by the sol-gel
method. Journal of materials science, 2005. 40(2): p. 263-283.
Wang, D. and G.P. Bierwagen, Sol–gel coatings on metals for corrosion protection. Progress in
Organic Coatings, 2009. 64(4): p. 327-338.
Yoldas, B.E., Alumina gels that form porous transparent Al2O3. Journal of Materials Science,
1975. 10(11): p. 1856-1860.
Chen, Y.-C., et al., Ceramic-coated carbide tools by sol-gel process. Journal of materials
science letters, 2000. 19(16): p. 1469-1472.
Avci, N., et al., Stability improvement of moisture sensitive CaS:Eu2+ micro-particles by
coating with sol–gel alumina. Optical Materials, 2011. 33(7): p. 1032-1035.
Li, Y., et al., Fe 3 O 4@ Al 2 O 3 magnetic core–shell microspheres for rapid and highly specific
capture of phosphopeptides with mass spectrometry analysis. Journal of Chromatography A,
2007. 1172(1): p. 57-71.
84
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
Dey, A., et al., Investigation of thermal oxidation of Al 2 O 3-coated SiC powder.
Thermochimica Acta, 2014. 583: p. 25-31.
Yang, C.Y. and W.H. Shih, Effect of acid on the coating of boehmite onto silicon carbide
particles in aqueous suspensions. Journal of the American Ceramic Society, 1999. 82(2): p.
436-440.
Beyers, R., Thermodynamic considerations in refractory metal‐silicon‐oxygen systems. Journal
of applied physics, 1984. 56(1): p. 147-152.
Yang, Q. and T. Troczynski, Dispersion of alumina and silicon carbide powders in alumina sol.
Journal of the American Ceramic Society, 1999. 82(7): p. 1928-1930.
Wang, M. and M. Muhammed, Novel synthesis of Al 13-cluster based alumina materials.
Nanostructured materials, 1999. 11(8): p. 1219-1229.
Dumont, F., D. Van Tan, and A. Watillon, Study of ferric oxide hydrosols from electrophoresis,
coagulation, and peptization measurements. Journal of colloid and interface science, 1976.
55(3): p. 678-687.
Pierre, A.C. and D.R. Uhlmann, Gelation of aluminum hydroxide sols. Journal of the American
Ceramic Society, 1987. 70(1): p. 28-32.
George, S.M., Atomic layer deposition: an overview. Chemical reviews, 2009. 110(1): p. 111131.
Puurunen, R.L., Surface chemistry of atomic layer deposition: A case study for the
trimethylaluminum/water process. Journal of applied physics, 2005. 97(12): p. 121301.
Leskelä, M. and M. Ritala, Atomic layer deposition chemistry: recent developments and future
challenges. Angewandte Chemie International Edition, 2003. 42(45): p. 5548-5554.
Kim, H., ALD Nanodeposition.
Van Ommen, J. and A. Goulas, ENCAPSULATING PARTICLES WITH NANOSCALE PRE-CISION
USING ATOMIC LAYER DEPOSITION. 2015.
Valverde Millán, J., Introduction. The Classical Geldart’s Diagram and the New Type of GasFluidization Behavior, in Fluidization of Fine Powders. 2013, Springer Netherlands. p. 1-6.
Valdesueiro, D., et al., Gas-Phase Deposition of Ultrathin Aluminium Oxide Films on
Nanoparticles at Ambient Conditions. Materials, 2015. 8(3): p. 1249-1263.
Hausmann, D., et al., Rapid vapor deposition of highly conformal silica nanolaminates.
Science, 2002. 298(5592): p. 402-406.
Garcia-Trinanes, P., et al., Enhancing the activation of silicon carbide particles with gas-phase
coating of aluminum oxide.
Groner, M., et al., Low-temperature Al2O3 atomic layer deposition. Chemistry of Materials,
2004. 16(4): p. 639-645.
Weast, R.C., Handbook of chemistry and physics. The American Journal of the Medical
Sciences, 1969. 257(6): p. 423.
Farrens, S.N., et al., Chemical free room temperature wafer to wafer direct bonding. Journal
of the Electrochemical Society, 1995. 142(11): p. 3949-3955.
Amirfeiz, P., et al., Formation of silicon structures by plasma‐activated wafer bonding. Journal
of the Electrochemical Society, 2000. 147(7): p. 2693-2698.
Yoldas, B.E., Ceramic bulletin, 1975(54): p. 286.
Andersson, J.M., et al., Microstructure of α-alumina thin films deposited at low temperatures
on chromia template layers. Journal of Vacuum Science &amp; Technology A, 2004. 22(1): p.
117-121.
Musil, J., et al., Thermal stability of alumina thin films containing γ-Al2O3 phase prepared by
reactive magnetron sputtering. Applied Surface Science, 2010. 257(3): p. 1058-1062.
Zhang, L., et al., Annealing of Al 2 O 3 thin films prepared by atomic layer deposition. Journal
of Physics D: Applied Physics, 2007. 40(12): p. 3707.
Freiman, S.W. and S. Ward, Fracture mechanics applied to brittle materials. Vol. 678. 1979:
ASTM International.
85
75.
76.
77.
78.
79.
80.
81.
82.
83.
84.
85.
86.
87.
88.
89.
90.
91.
92.
93.
94.
95.
96.
97.
98.
99.
Ghosh, S.K., Self-healing materials: fundamentals, design strategies, and applications. 2009:
John Wiley & Sons.
Zhu, Y.T., et al., Thermal Oxidation Kinetics of MoSi2‐Based Powders. Journal of the American
Ceramic Society, 1999. 82(10): p. 2785-2790.
Liu, Y., G. Shao, and P. Tsakiropoulos, On the oxidation behaviour of MoSi 2. Intermetallics,
2001. 9(2): p. 125-136.
Shan, A., et al., Microstructure and Mechanical Properties of MoSi2-X (X= Al, B, Nb) Alloys
Fabricated by MA-PDS Process. Materials Transactions, 2002. 43(1): p. 5-10.
Kuznetsov, S.A., et al., Synthesis of molybdenum borides and molybdenum silicides in molten
salts and their oxidation behavior in an air–water mixture. Surface and Coatings Technology,
2005. 195(2–3): p. 182-188.
e Silva, A.C. and M. Kaufman, Synthesis of MoSi 2-boride composites through in situ
displacement reactions. Intermetallics, 1997. 5(1): p. 1-15.
Liao, K.E.S.P.K., The B-Mo phase diagram.
e Silva, A.C. and M. Kaufman, Applications of in situ reactions to MoSi 2-based materials.
Materials Science and Engineering: A, 1995. 195: p. 75-88.
Lakiza, S.M. and L.M. Lopato, Stable and metastable phase relations in the system alumina–
zirconia–yttria. Journal of the American Ceramic Society, 1997. 80(4): p. 893-902.
Fabrichnaya, O., et al., The assessment of thermodynamic parameters in the Al2O3‐Y2O3
system and phase relations in the Y‐Al‐O system. Scandinavian journal of metallurgy, 2001.
30(3): p. 175-183.
Suenaga, K., et al., Investigations of alumina/spinel and alumina/ zirconia interfaces by
spatially resolved electron energy loss spectroscopy. Journal of the European Ceramic Society,
1998. 18(10): p. 1453-1459.
Schneider, H., J. Schreuer, and B. Hildmann, Structure and properties of mullite—a review.
Journal of the European Ceramic Society, 2008. 28(2): p. 329-344.
Degterov, S. and A. Pelton, Critical evaluation and optimization of the thermodynamic
properties and phase diagrams of the CrO-Cr2O3-SiO2 and CrO-Cr2O3-SiO2-Al2O3 systems.
Journal of phase equilibria, 1996. 17(6): p. 488-494.
Butterman, W.F., W. , Zircon stability and the ZrO2-SiO2 phase diagram. Am. Mineral., 1967:
p. 880-885.
RouNSow, K., G. Grsns, and P. eNn, The structure of zircon: a comparison with garnet. Am.
Mineral., 1971. 56: p. 782-790.
Springer, T. and R. Lechner, Diffusion in Condensed Matter. Vieweg, Wiesbaden, 1998: p. 59.
Heuer, A., Oxygen and aluminum diffusion in α-Al 2 O 3: How much do we really understand?
Journal of the European Ceramic Society, 2008. 28(7): p. 1495-1507.
Fielitz, P., et al., A diffusion-controlled mullite formation reaction model based on tracer
diffusivity data for aluminium, silicon and oxygen. Philosophical Magazine, 2007. 87(1): p.
111-127.
Lamkin, M., F. Riley, and R. Fordham, Oxygen mobility in silicon dioxide and silicate glasses: a
review. Journal of the European Ceramic Society, 1992. 10(5): p. 347-367.
Lejcek, P., Grain boundary segregation in metals. Vol. 136. 2010: Springer Science & Business
Media.
Smialek, J.L., et al., Oxygen Permeability and Grain-Boundary Diffusion Applied to Alumina
Scales. 2013, NASA/TM—2013-217855, 1–14.
Smialek, J.L., Diffusivity in Alumina Scales Grown on Al-MAX Phases. 2014.
Yoon, J.-K., et al., Microstructure and growth kinetics of the Mo 5 Si 3 and Mo 3 Si layers in
MoSi 2/Mo diffusion couple. Intermetallics, 2003. 11(7): p. 687-696.
Wei, X. and D. Chen, Synthesis and characterization of nanosized zinc aluminate spinel by sol–
gel technique. Materials Letters, 2006. 60(6): p. 823-827.
Saha, S.K. and P. Pramanik, Aqueous sol-gel synthesis of mullite powder by using aluminium
oxalate and tetraethoxysilane. Journal of materials science, 1994. 29(13): p. 3425-3429.
86
100.
101.
102.
103.
104.
105.
106.
107.
108.
109.
110.
Goldstein, J.I., et al., The influence of oxide surface layers on bulk electron probe
microanalysis of oxygen—application to Ti‐Si‐O compounds. Scanning, 1993. 15(3): p. 165170.
Yao, Z., J. Stiglich, and T. Sudarshan, Molybdenum silicide based materials and their
properties. Journal of Materials Engineering and Performance, 1999. 8(3): p. 291-304.
Rosenbrand, G.G., The separation performance and capacity of zigzag air classifiers at high
particle feed rates. 1986, Dissertation TU Eindhoven.
Lumsdon, D.G. and V.C. Farmer, Solubility of a proto-imogolite sol in oxalate solutions.
European Journal of Soil Science, 1997. 48(1): p. 115-120.
Dollimore, D. and D. Griffiths, Differential thermal analysis study of various oxalates in
oxygen and nitrogen. Journal of Thermal Analysis and Calorimetry, 1970. 2(3): p. 229-250.
Rhodes, M.J., Introduction to particle technology. 2008: John Wiley & Sons.
Anstis, G., et al., A Critical Evaluation of Indentation Techniques for Measuring Fracture
Toughness. I.--Direct Crack Measurements. Journal of the American Ceramic Society, 1981.
64(9): p. 533-538.
Christel, P., et al., Mechanical properties and short‐term in vivo evaluation of
yttrium‐oxide‐partially‐stabilized zirconia. Journal of biomedical materials research, 1989.
23(1): p. 45-61.
Sciti, D., S. Guicciardi, and M. Nygren, Spark plasma sintering and mechanical behaviour of
ZrC-based composites. Scripta Materialia, 2008. 59(6): p. 638-641.
Chamberlain, A.L., et al., High‐Strength Zirconium Diboride‐Based Ceramics. Journal of the
American Ceramic Society, 2004. 87(6): p. 1170-1172.
Morris, D.G., M. Leboeuf, and M. Morris, Hardness and toughness of MoSi 2 and MoSi 2–SiC
composite prepared by reactive sintering of powders. Materials Science and Engineering: A,
1998. 251(1): p. 262-268.
87
Table of contents Appendices
Appendix I: Protective coatings for MoSi2 for self-healing Thermal Barrier Coatings ............................ 2
Appendix II: Proposal MSc thesis Mark Meijerink................................................................................. 11
Appendix III: Matlab code used for the diffusion model ...................................................................... 16
Appendix IV: Required coating thickness and amount of coating ........................................................ 19
Appendix V: Principles characterization methods used ........................................................................ 21
Appendix VI: The ALD setup .................................................................................................................. 26
Appendix VII: EDS maps of wind sifted MoSi2B particles ...................................................................... 27
Appendix VIII: XRF results of MoSi2B particles ...................................................................................... 28
Appendix IX: EDS individual data points aluminium alkoxide coatings ................................................. 29
Appendix X: TGA mass change graphs .................................................................................................. 31
Appendix XI: XPS results main samples ................................................................................................. 32
Appendix XII: Cross-section thickness measurement images ............................................................... 34
Appendix XIII: ICP-OES results for ALD samples .................................................................................... 39
Appendix XIV: ALD 40 cycle particle EDS map....................................................................................... 40
Appendix XV: EDS data of coated MoSi2B particles .............................................................................. 41
Appendix XVI: XRD diffractograms of YSZ powder and SPS samples .................................................... 42
Appendix XVII: EDS maps of the YSZ/MoSi2B composite ...................................................................... 45
1
Appendix I: Protective coatings for MoSi2 for self-healing Thermal
Barrier Coatings
1. Introduction
One of the major challenges these days is the increasing scarcity of energy. The realization of fossil
fuels being a finite resource is dawning and emissions of carbon dioxide become increasingly
problematic for earth’s climate. Many solutions are being investigated that focus on using different
sources of energy or increasing the efficient use of energy. However, for many applications,
especially gas turbines, increasing efficiency requires increasing the operating temperature. High
temperatures combined with stresses acting on the components, is an extremely demanding
environment and requires very specialized materials.
Even with specially developed materials such as nickel superalloys, there is a limit to the operating
temperature because of creep damage. To protect the structural components of gas turbines, they
are cooled on the inside and insulated by thermal barrier coatings on the outside. One of the
challenges of this protection is that the thermal barrier coating is almost always a refractory ceramic,
yttria-stabilized zirconia being the most prevalent in modern gas turbines and engines. Because of
the significant mismatch in thermal expansion coefficient of these ceramics and the metallic base
material, stresses arise during heating and cooling, resulting in significant spallation damage and the
necessity to replace the coating often.
A solution to this problem, which has been developed at Delft, University of Technology, is to use a
coating that is able to autonomously repair the damage by including a so-called healing agent. If a
crack is present, the material is able to oxidize and fill the crack, thereby healing it. The problem with
this system is however that due to the thermal barrier coating being transparent to oxygen, the
healing agent oxidizes even when no crack is present. To prevent this, the particle has to be coated
with a material that is impermeable to oxygen, but allows the crack to grow through the coating.
The goal of this study is therefore to find suitable materials to protect the healing agents, compare
them and select the most suitable materials for further investigation. Another important aspect of
this is the method of application of this coating, which is not trivial. Therefore the second goal is to
find suitable routes to obtain the protective barrier and select the most promising combinations of
protective material and application route.
To achieve this, first the system of the thermal barrier coating and important requirements for the
coating of the self-healing agents are described in chapter 2, followed by a thorough analysis of the
possible materials to coat the agent in chapter 3. As the main function of the coating is to prevent
oxygen from reaching the inner part of the oxide, a small analysis on the oxygen diffusion is
performed in chapter 4, along with diffusion calculations and required coating thickness to prevent
oxidation. Using calculated thickness, suitable methods of producing this coating are presented in
chapter 5. Finally, in chapter 6 a selection of the most promising material and application method
combinations is performed.
2
2. The Thermal Barrier Coating system
In this chapter, the system of the thermal barrier coatings (TBCs) is discussed to better understand
the challenges present and extract from this information the main requirements for the protective
coating for the healing agent.
The full system is shown in figure 1 (courtesy of W. Sloof) and shows first the location of the coating
in a jet engine. TBCs are present in the hot zone of the engine, mainly on top of turbine blades made
from nickel superalloys. A bond coating is used for growing a protective layer of Al2O3 on top for
protection against corrosion. Although compositions for both alloy and bond coating can vary
significantly, some typical compositions are shown in table 1. On top of that oxide layer, a layer of
ZrO2 with approximately 6-8 wt% of Y2O3 is deposited by plasma spraying as thermal barrier coating.
Figure 1: Jet engine layout (a) and interface between turbine blade and hot gas (b).
(Source: Introduction presentation W. Sloof and A. Carabat)
Table 1: Elemental composition of some single crystal nickel superalloys and a typical bond coating
(weight %). (Source: Lecture W. Sloof on high temperature material applications)
Alloy
Cr
Co
Mo
Re
W
Al
Ti
Ta
Nb
Hf
Y
Ni
CMSX-2 (first generation)
8
4.6
0.6
-
8
5.6
1
6
-
-
-
66.2
CMSX-6 (first generation)
9.8
5
3
-
-
4.8
5.7
2
-
0.1
-
69.6
Rene N4 (first generation)
9
8
2
-
6
3.7
4.2
4
0.5
(b)-
-
62.6
CMSX-4 (second generation)
6.5
9
0.6
3
6
5.6
1
6.5
-
0.1
Rene N5 (second generation)
7
8
2
3
5
6.2
-
7
-
0.2
61.6
CMSX-10 (third generation)
2
3
0.4
6
5
5.7
0.2
8
0.1
0.03
69.57
19
18
-
-
-
21
-
-
-
-
(a)
Bond coating
61.7
0.2
41.8
As mentioned in the introduction, the TBC is subject to high stresses upon heating and cooling from
the thermal expansion mismatch between the coating and the substrate. This is partially
compensated by structuring the thermal barrier coating with long and thin grains, allowing some of
the stresses to be accommodated. Nevertheless, cracks will still grow, first perpendicular to the
surface of the TBC, but when the crack gets close to the interface with the protective Al2O3 scale, it
can be deflected and start to grow horizontally. When too many cracks have grown in the horizontal
direction, this will result in debonding of the TBC and part of it will break off.
3
To prevent this, a self-healing concept based on MoSi2 particles has been proposed, which is shown
in figure 2. If these particles are added to the TBC close to the alumina interface and a crack grows
through them, they will oxidize according to:
2 MoSi2 (s) + 7 O2 (g)  2 MoO3 (g) + 4 SiO2 (s)
In which the volatile MoO3 escapes and the SiO2 fills the crack due to the volume expansion
compared to the original MoSi2. Furthermore, SiO2 can react with ZrO2 to form ZrSiO4 for good
adhesion. Small cracks have even been shown to be completely filled with ZrSiO4, according to
previous research.
However, Yttria stabilized zirconia is very transparent to oxygen, especially at temperatures used for
these coatings. Therefore MoSi2 will even oxidize without a crack being present, creating additional
stresses close to the TBC-Al2O3 interface due to volume expansion. Therefore, a protective coating
for this healing agent is required.
Figure 2:Proposed self-healing mechanism based on coated MoSi2 particles. (Source:
Introduction presentation W. Sloof and A. Carabat)
The two main functions of this coating would be to protect the MoSi2 against oxidation when closed
and to open when a crack is growing close to the particle, preferably even attracting the crack to
grow into the particle. For the first function, it is very important that the coating material forms a
closed shell that remains closed at operating temperatures and has a low oxygen diffusivity at these
temperatures. For the second function, numerical modeling has shown that interface strength is a
critical parameter. A relatively weak interface between MoSi2 and the coating and a stronger
interface between the coating and the yttria stabilized zirconia would yield the best results for
attracting growing cracks.
Therefore the main requirements for the first function is that the shell is mechanically, thermally and
chemically stable in an oxygen-containing environment (PO2 between 10 Pa and 10000 Pa).
Furthermore, it should have a very low oxygen diffusivity and be thick enough to efficiently block
transport of oxygen and other charged species. For a reasonable adhesion on the particle, it is also
important that thermal expansion coefficients match and that the crystal structure and lattice
parameter of the coating is coherent enough with that of MoSi2.
4
As can be seen, many properties are influenced by the interfaces of the coating. Therefore it is critical
to consider these in more detail. However, these interfaces can be influenced by many factors and
require a more detailed investigation. It is therefore proposed to include the interface chemistry and
behavior as an important part of the thesis work.
3. Materials
In this chapter, various suitable materials are discussed. For a suitable coating, the first requirement
would be that it is stable in the range of room temperature up to 1473K and in the presence of
oxygen with a partial pressure somewhere between 10 and 10000 Pa. Although many materials exist
that would be able to sustain such a temperature, oxidation is a significant problem. Therefore,
common carbides and nitrides (such as SiC, Si3N4, TiN etc.) were found to be unsatisfactory for
oxidation protection. This also holds for metals, as most metals that are still solid at 1473K have a
tendency to oxidize, resulting in most cases in a significant volume expansion. Due to this, only
oxides have been considered for further investigation. An overview of the oxides considered, along
with some important properties, can be found in table 2, along with yttria stabilized zirconia and
MoSi2 data for comparison.
These oxides were selected because they were either common in corrosion protection applications
(such as SiO2, Al2O3, Cr2O3, TiO2 and ZnO), were used in other applications as high temperature
barriers (Ta2O5, HfO2,MgO) or were considered because they were both stable at high temperature
and had a high energy of vacancy formation, indicating a low cation and anion vacancy
concentration, indicating slow diffusion of species. The first important properties are of course
related to stability of the material, as it needs to be both solid and stable. Therefore melting
temperature was presented, along with the Gibbs free energy of formation of the oxide from its
parent metal(s) and oxygen gas at 1473K. This free energy is taken from Ellingham diagrams provided
by an online tool of the University of Cambridge.
Table 2: An overview of various oxides used at high temperature, along with important properties.
Material
Al2O3
Melting
temperatu
re (K)
2345
SiO2
1873-1998
ZnO
2248*
Gibbs free energy
at 1473K (kJ/mol
O2)
-801.71
-617.89
-354.6
-569.11
thermal
conductivity
(W/m/K)
crystal structure
coefficient of
thermal
expansion
(10^-5 K-1)
39
corundum
0.8
1.4
quartz
21
Wurtzite
0.4
Not measured
disputed
0.45
fluorite
0.94
fluorite
(monoclinic)
silicate (tetragonal)
1.03
0.54
1.48
0.059
Ta2O5
2145
HfO2
3031
-806.25
1.1
ZrO2
2988
-821.375
2
ZrSiO4
2820
3.5
Mullite
(Al6O9Si2O4)
MgO
1923
-729.3
-585.5
37
CaO
2886
0.3
rocksalt
Cr2O3
2708
-893.47
-968.87
-491.69
silicate
(orthorhombic)
rocksalt
10.0-33
TiO2
2116
-679.375
11.7
rutile
0.85
MgAl2O4
2403
-857.96
15
spinel
0.745
3125
6
corundum
0.5
1.35
0.54-0.75
5
Ta (metal)
3290
-48.7
54
ZrO2 (Yttria
stabilized)
MoSi2
2988
-806.25
2
2303
-130
70
bcc
0.65
fluorite (cubic)
1.03
tetragonal
0.7-0.8
*decomposes at mentioned temperature
Unfortunately, not all of these oxides were found to be suitable. Cr2O3 and ZnO were found to be not
stable under these conditions for prolonged time due to slow sublimation of ZnO (becoming
noticeable around 1400K and atmospheric oxygen partial pressures) and oxidation of Cr2O3 to the
gaseous species CrO3. Furthermore, CaO is quite reactive and can easily dissolve in the zirconia
matrix (as it is also used to stabilize zirconia), making it not very suitable as a coating material.
Although unstabilized (and therefore undoped) zirconia is in itself a reasonable oxygen barrier, phase
transformations with large volume changes make this material unsuitable to form a protective
coating. TiO2 is very stable, but has a tendency to be reduced at higher temperatures in the presence
of a reductant (MoSi2 not being extremely noble could be able to do this), which is not desirable. SiO2
will also react with the zirconia, as the self-healing mechanism is partially based on this reaction and
is therefore also not suitable as coating.
The zirconia lattice of the surrounding material can also accommodate significant amounts of MgO,
HfO2 and Al2O3 and these are also sometimes used to stabilize zirconia instead of yttria, but the rate
at which these materials are able to diffuse in ZrO2 is not known and important for further
investigation if these materials are to be used. However, MgO has a significantly higher thermal
expansion coefficient, indicating that using this material as coating could lead to severe stresses at
the MoSi2-MgO interface. Ta metal was considered because of the formation of its protective oxide,
but might be problematic due to the significant volume expansion upon oxidation. Its oxide is also
interesting, because the crystal structure is not very well defined and consists of two crystalline
phases and an amorphous region. In the range between 1250K and 1600K there is a very slow phase
transformation to the phase stable at higher temperature, but without a significant volume change.
4. Diffusion and Coating thickness
Another major and important property of the coating is to prevent oxidation. Calculating diffusion in
oxidic ceramics is not the easiest task however and due to time limitations, a simpler model based on
Ficks diffusion law, coupled with diffusion coefficients found in literature and summarized in table 3,
has been used for the ceramics that have not been rejected yet in the previous chapter.
Table 3: Diffusion coefficients at 1473K for the dominant species responsible for oxidation.
Oxide
Al2O3
Ta2O5
HfO2
ZrSiO4
Mullite (Al6O9Si2O4)
MgAl2O4
2
D (m /s)
5.75E-18
5.21E-18
1.30E-13
1.94E-20
1.101E-20
1.30E-12
6
Then using (
being flux, D the diffusion coefficient, c the concentration and x distance):
And assuming the particle to be a perfect sphere with a radius of 10 μm and the problem to be a
steady state problem (as the goal is negligible oxidation), this results in the well-known diffusion
equation for a sphere in steady state:
Here, Δc is the concentration gradient of oxygen over the coating. It is assumed the concentration of
oxygen is exactly 0, as inside, its concentration is so low, it can be neglected. Outside, assuming a
partial pressure of roughly 10 kPa, the concentration at the oxide is roughly 0.81 mol/m3. D is taken
from the table above and R1 is 10 micrometer or 10-5 m. To determine φm, or the maximum allowable
mass transport, it is assumed that all other parameters will remain constant and no more than 0.1
wt% of the MoSi2 particle will oxidize over the period of 1 year at 1473K. For a particle of 10 micron,
this means over the lifetime, a total of approximately 1.21*10-12 mole of O atoms should pass
through the coating and react. This results in a maximum allowable flux of 3.83*10-20 mol O/s. Then
using the equation for a perfect sphere:
Then, calculating the coating thickness: dc = R2-R1. This results in the coating thicknesses presented in
table 4. As can be seen, some error is made in the case of mullite and zircon. Due to their low
diffusion coefficient, a coating thickness of only 0.5 nm should be enough according to the
calculations. This is obviously impossible. However, for the other coatings, an interesting estimate is
presented. It is clear that both hafnia and spinel are not very good high temperature oxygen barriers,
as the coating has to be very thick at these temperatures (the criterion cannot be attained with these
diffusion coefficients, hence the negative value).
Table 4: Calculated required coating thicknesses for the selected oxides.
Oxide
Thickness (nm)
Al2O3
156.7
Ta2O5
141.6
HfO2
-1.003E04
ZrSiO4
0.521
Mullite (Al6O9Si2O4)
0.295
MgAl2O4
-1.0003E04
5. Application methods
Another important aspect of the coating is the method of application. Here, the most interesting
application methods will be listed with the main advantages and disadvantages.
7
Precipitation
The creation of a separate solid phase by exceeding the solubility limit of the solid or liquid it is in. In
the case of deposition, this is often bringing two solutions with soluble salts together and in a
chemical reaction, an insoluble salt is formed, that then deposits on the desired substrate. Often a
subsequent calcination step required. This is a relatively easy method with good mixing and an
homogeneous distribution of reactants. However, control over coating thickness and morphology is
difficult.
Selective oxidation
Alloying the desired material with the metal precursor to the oxide and selectively oxidize that metal
on top of the substrate. Results in a very good control of coating thickness and reasonable control
over morphology, but requires alloying of the component and obtaining required thickness might be
difficult for protective oxides, as growth speed will decrease significantly with coating thickness
(therefore high temperature required). Inhomogeneous coating thickness over the particle might be
obtained too, as well as limited surface chemistry control.
Sol-gel
Conversion of dissolved monomeric metal alkoxides or hydroxides into a stable colloid solution,
followed by evaporation of solvent and polymerization to form a gel on a substrate. Subsequent
firing results in the oxide material that is desired. Advantages are the relatively simple and low
temperature processing and good control over structure, morphology and interface chemistry. It also
generates a very homogeneous coating and gives slightly more control over coating thickness than
precipitation, but it is still limited. Deposition of the gel onto small particles is also something that
might prove slightly more difficult.
Chemical Vapor Deposition
The deposition of a thin solid film by well-controlled chemical reactions on the substrate surface by
volatile precursors. In the case of particles best performed in a fluidized bed reactor. Main
advantages include a good coating with easily controllable thickness and good homogeneity. Many
thicknesses can be attained. However, also here a high temperature is required and quite volatile
and toxic chemicals are required. Furthermore, some issues with sticking of particles have been
mentioned, which is not desired. Control of interface chemistry limited and quite a complex growth
process.
Atomic Layer Deposition
The growth of a layer by depositing alternating monolayers of precursor on the substrate and
subsequent firing. Again for particles best performed in a fluidized bed. The main advantage is that
this method gives a very high quality film with very good control over thickness, composition,
structure and morphology. Its main limitation however is the limited coating thickness, as each cycle
only adds one monolayer of material. Even though cycles can be performed very rapidly (a few
seconds per pulse), a thickness of more than 100-200 nm is probably not very feasible using this
method. Interface chemistry is also limited in control.
8
Thermal Spraying
Spraying liquid material on the particles, either metal or oxide. Main advantage is that it is relatively
simple and yields good closed coatings. Main disadvantage is that to spray oxides, extremely high
temperatures are needed, often higher than the melting point of the MoSi2 and when spraying
precursor metals, selective oxidation is still needed. Poor control of interface chemistry, but
reasonable control of thickness and a quite homogeneous coating.
Dry powder coating techniques
Usually involves the use of temperature and pressure in the form of milling balls to mill smaller
particles of the oxide on a larger substrate particle. This is a relatively simple method, but gives poor
control over homogeneity, interface chemistry, thickness, structure and morphology. Furthermore,
the application method damages the material and the coating and introduces stresses and it is
difficult to obtain a closed shell.
Electroless deposition combined with selective oxidation
The use of a metallic precursor and an organic reductant (for example formaldehyde) to precipitate a
metallic layer on top of the particle and subsequently selectively oxidizing that metal to form an
oxide coating with a small metallic layer beneath. Main advantages are again a relatively simple
process, easy control of thickness, closed shell formation and some control over interface chemistry,
composition and morphology. Another benefit might be an additional interface if some metal is left.
This can both act as a barrier to oxygen diffusion and a reserve of oxide forming material if the metal
is not very noble (which gives some additional oxide growth without oxidation of the MoSi2).
Disadvantages are the same as for selective oxidation and that additional oxidation during lifetime
might increase local volume, introducing stresses. Furthermore, the metal should be solid at 1473K.
6. Selection
Previous information shows that only some oxides are suitable for use as a high temperature barrier
coating. Of these, Al2O3, Ta2O5, zircon and mullite seem to be the most interesting due to their
stability and difficult diffusion at elevated temperatures. For suggested routes, precipitation of
alumina and mullite has already been investigated to a certain extend and is therefore not included.
Dry powder coating and thermal spraying do not seem to be very appropriate methods for this
application due to often poor protective coating properties and difficulty of handling liquid oxides
respectively. CVD was also advised against due to some sticking problems and general complexity.
For zircon and mullite, electroless deposition is probably a process too complex, as these are formed
by two different metallic precursors and therefore both deposition and oxidation will become very
difficult to achieve stoichiometrically. Therefore for these materials, selective oxidation is also
advised against. A sol-gel method would probably work very well for these oxides, as good
stoichiometry and crystal structure could be easily achieved with the right precursors. If a better
diffusion calculation finds that layer thickness can remain very thin, ALD might work in the case of
these combined oxides too.
9
For Al2O3 formation, selective oxidation, ALD (combined with some CVD-like characteristics) and solgel are quite interesting methods. Electroless deposition might be possible, but care must be taken
to oxidize all Al during the subsequent oxidation step, as it becomes liquid around 900K, which is way
below operating temperatures. Therefore selective oxidation with alloying seems to be better. ALD
might be interesting, but the required thickness of alumina seems to be very high, making ALD quite
difficult.
For Ta2O5, electroless deposition will work, but care must be taken that the volume expansion of
oxidized tantalum during operation of the full thermal barrier coating is not too significant. Ta2O5 can
also be prepared by sol-gel methods and deposited by ALD, although again in the latter case,
thickness might be an issue again. Selective oxidation might be possible, but is difficult, as the cation
has relatively low mobility.
As these options are still many, a discussion on the best way to proceed is important. In my opinion it
is worthwhile to look into zircon and mullite further, as literature suggests they seem to be very
suitable corrosion protection oxides at high temperature. Alumina is also a very interesting oxide, as
it is one of the most important oxides for many applications, including high temperature corrosion
protection. Utilizing selective oxidation, ALD (if feasible) and sol-gel can lead to interesting insights
regarding different morphology and interface chemistry of the different coating techniques.
Furthermore, the coefficient of thermal expansion matches very well with the substrate and it is a
relatively cheap material.
For tantalum oxide: according to literature, the difference in protective properties between Ta2O5
and Al2O3 are not very significant and together with the phase transformation in the operating
temperature and being significantly more scarce and expensive, it is recommended to not include
this oxide in further research.
It is also still very important to do some more extensive diffusion calculations and research and
understand the interfaces present in the overall system and how they influence both transport and
fracture behavior.
10
Appendix II: Proposal MSc thesis Mark Meijerink
Title: Synthesis of high-temperature oxidation-resistant coatings on MoSi2 healing particles for use in
self-healing thermal barrier coatings.
Introduction
With rising energy prices and increasing scarcity of inexpensive energy, the demand for increasingly
efficient gas turbines and other high-temperature turbines is increasing. The best option to increase
efficiency is to increase operating temperature (1). However, current turbines are already operating
far above the limit of the used nickel superalloys. To prevent breakdown of the structural parts, (~0.5
mm) thermal barrier coatings (TBC) in combination with internal gas cooling are applied to prevent
overheating and allow the turbine blades to endure these extreme environments (2).
However, due to thermal expansion coefficient mismatch between the TBC, usually made of yttriastabilized zirconia (7 wt% Y2O3–ZrO2, YSZ) and the nickel superalloys, application of these coatings is
difficult and significant mismatch stresses arise during heating and cooling of the engine. Even
though a (~250 µm) bond coat (BC) with a (0.6-3.0 µm) thermally grown oxide (TGO) for oxidation
protection is applied, this mismatch results in unavoidable crack growth and spallation damage in the
TBC and frequent replacement of the coating is therefore required (2). However, the work of Zwaag
et al. (3) used a different approach to repair damage autonomously, based on the inclusion of
sacrifical MoSi2 healing particles that oxidize, expand and fill the crack when it is close to the particle.
However, as YSZ is very transparent to oxygen at operating temperatures (1250-1500K depending on
engine and location in the TBC (2)), significant premature oxidation of MoSi2 is present and a coating
that protects against oxidation is required. The proposed system including coated healing particles is
shown in figure 1.
Although many different protective materials exist, α-alumina (α-Al2O3), zircon (ZrSiO4) and mullite
(Al6Si2O13) were found to be exceptionally suitable for this application (4). However, application of
coatings on MoSi2 has rarely been investigated, mainly due to the excellent oxidation resistance of
the bulk material at high temperature, resulting from the formation of a thick homogeneous SiO2
layer. Some research has for example been performed on mullite coatings produced by alloying the
MoSi2 with some aluminium and subsequent selective oxidation (5), but as most MoSi2 applications
11
are bulk material applications, coating of molybdenum disilicide particles has not been investigated
yet, except for work performed by A. Carabat et al. on sol-gel coating with AlCl3 (not published yet)
(6).
In the case of particles with a size in the range of 5-20 µm, formation of a coating by oxidation of the
MoSi2 would not be feasible, as a significant portion of the material will then be oxidized even before
incorporation in the TBC system. Therefore, a protective coating has to be applied beforehand,
preferably by methods that also add the metallic components during the procedure (or methods that
utilize specially alloyed particles with metallic incorporations that already account for controlled
oxidation to form a coating). Methods that were found to be most promising during a study of the
available literature were using a sol-gel application procedure and atomic layer deposition (ALD).
Therefore, the focus of this project is on the synthesis and characterization of a protective Al2O3
coating on MoSi2 particles by sol-gel methods and atomic layer deposition. As the main aim of these
coatings is prevention of oxidation of the substrate, special attention will be given to the influence of
coating thickness and coating microstructure (especially defects present) on oxidation behaviour.
Furthermore, as crack growth through the healing particle is vital for the self-healing mechanism, the
influence of thickness, microstructure and interface properties on crack growth in the TBC system
will also be investigated. Finally, if time allows, the use zircon as coating material will also be
investigated, because diffusion in zircon is reported to be significantly slower than in alumina (7),
allowing for thinner coatings. Also, literature suggests that amorphous zircon is stable at the desired
operating temperature (8), resulting in far fewer defects that can act as a fast pathway for diffusion.
Research question
During this thesis project, the aim is to answer the following research question:
How can an Al2O3/ZrSiO4 coating be created around MoSi2 healing particles for use in self-healing
thermal barrier coatings that prevents premature high temperature oxidation and attracts nearby
cracks?
To help answer this question, the following sub-questions are to be answered:
-What thickness should such a protective coating have to limit diffusion to acceptable levels?
-How do microstructure and defects influence diffusion behaviour in the coating?
-How do microstructure, defects and thickness affect crack attraction and crack growth into the
particle?
-How can ALD and sol-gel methods influence microstructure, amount and type of defects and
thickness
and
what
are
the
differences
between
these
methods?
-How does thermal treatment affect microstructure, amount and type of defects and thickness?
Approach
To answer these questions, an approach based on literature study, experimental work and limited
diffusion modelling is required. It is important to develop a model for diffusion in these coatings to
be able to quantify diffusion through the coatings by different pathways and to obtain information
on required thickness. As mentioned before, experimental work will first focus on coating particles
with alumina and an investigation of coating properties.
12
As mentioned before, no reports in literature exist for coating of MoSi2 particles, but alumina has
been used as a coating for a variety of other substrates (9). Examples include stainless steel (10),
magnetic particles (11) and SiC particles (12) (13) (14). For this method, it is paramount to first
succeed in coating particles. The work of Kim et al. (14) on coating with dispersed boehmite gels or
that of Yang and Shih (13) using aluminium sec-butoxide produced boehmite coatings on SiC particles
that after annealing transformed into α-Al2O3 coatings. Due to very similar surfaces (both SiC and
MoSi2 have a thin SiO2 layer at their surface) and particle sizes not too dissimilar from the MoSi2
particles to be coated, these methods are expected to yield similar coatings on the healing particles
and will therefore be the start of the sol-gel investigation.
As many parameters play an important role in sol-gel procedures, not all can be investigated and
therefore it is proposed to optimize coating synthesis procedures by:




Precursor concentration/solid loading
Time of reaction
Thermal treatment procedures (drying time and temperature, annealing time and
temperature and the influence of an intermediate chemical water drying step)
Particle surface charge in combination with precursor charge
In the case of ALD, alumina is one of the most common materials for coating particles (15) and the
most frequently used precursor is trimethylaluminium (Al(CH3)3 or TMA) (15). Different oxygen
containing precursors exist, but the most common one is H2O. Many other oxygen precursors exist,
but water is by far the most used (15). The main challenge for ALD however is producing the required
thickness of the coating, as using the TMA/H2O system usually yields between 1 and 1.5 Å al Al2O3
per cycle (16)and due to long residence times in the required fluidized bed reactor and large surface
areas to coat, such a cycle often takes more than 10 minutes to finish.
A way around this problem is the approach used by García-Triñanes et al. (17) to use an ALD/CVD
system to increase growth per cycle, while still working in a linear regime for optimal coating
thickness control. However, this approach was tailored to produce porous coatings, which is
undesired in the case of protective coatings. Another option might be to combine this approach with
using the TMA/Aluminium isopropoxide (Al(O-CH2-(CH3)2)3) used by Ritala et al. (18) to deposit more
alumina per cycle depending on the reaction rate of TMA with the alkoxide compared to water. In
normal ALD growth per cycle of this mechanism is however in the same range as the TMA/H 2O
system.
Characterization of the particles is a very important part of the research and to understand the
various properties of the material, analysis could be done with the following techniques:





Scanning Electron Microscopy (SEM) for morphology and thickness of the coating.
X-ray diffraction (XRD) for analysis of the various crystalline phases present.
N2 physisorption (or another suitable physisorption technique) for particle surface area and
porosity.
X-ray photoelectron spectroscopy (XPS) for coating composition and chemical bonding.
Thermogravimetric analysis (TGA) for high temperature stability investigation and for
investigation of suitable annealing profiles (possibly augmented by differential scanning
calorimetry (DSC)).
13

Fracture tests on particles incorporated in YSZ by spark plasma sintering (SPS) to obtain
information about crack growth behaviour.
Other analysis techniques such as ICP-OES to obtain more information on the overall composition
and the Si/Al ratio (how much healing potential is present for a certain mass), nano-indentation,
transmission electron microscopy (TEM) for interface layers and other techniques might also be
considered.
Planning (starting date: 4 august 2014)
Month
August
September
October
November
December
January
February
March
April
May
June
July
Activities
-Initial literature survey
-Writing thesis proposal
-Literature study + report writing
-Diffusion model development
-Start with sol-gel trial experiments alumina
-Alumina sol-gel coating experiments and characterization (includes fracture tests)
-Literature study + report writing
-Safety training for working with TMA (start of the month)
-First alumina ALD experiments and characterization
-Have a working sol-gel procedure for alumina (end of the month)
-Literature study + report writing
-Synthesis of different thickness coatings of alumina by sol-gel
-Synthesis of high thickness and low porosity ALD alumina coatings
-Characterization of coatings
-Progress report and discussion
-Literature study
-Alumina ALD and sol-gel coating synthesis and characterization
-Literature study
-Alumina ALD and sol-gel coating synthesis and characterization
-Annealing and drying of samples with different procedures and characterization
-Midterm presentation
-Literature study
-Continue on sol-gel and ALD parameter optimization
-Trial experiments with zircon coating synthesis and characterization
-Literature study
-Continue on sol-gel and ALD parameter optimization
-Continue on zircon and have working procedures at the end of the month
-Literature study
-Continue on sol-gel and ALD parameter optimization
-Continue on zircon with amorphous/crystalline and thermal treatment
-Literature study
-Continue on sol-gel and ALD parameter optimization
-Continue on zircon
-Literature study
-Finalize experiments
-Full-time report writing
-Finalize reports
-Final presentation(s)
14
Bibliography
1. Overview on advanced thermal barrier coatings. Robert Vaßen, Maria Ophelia Jarligo,
Tanja Steinke, Daniel Emil Mack, Detlev Stöver. 2010, Surface & Coatings Technology,
pp. 938-942.
2. Development of coatings for protection in specific high temperature environments. M.
Schütze, M. Malessa, V. Rohr, T. Weber. 2006, Surface & Coatings Technology, pp. 3872–
3879.
3. Self-healing behaviour in man-made engineering materials: bioinspired but taking into
account their intrinsic character. S van der Zwaag, N.H van Dijk, H.M Jonkers, S.D
Mookhoek, W.G Sloof. 2009, Philosophical transactions of the Royal Society, pp. 16891704.
4. Meijerink, M.J. Protective coatings for MoSi2 for self-healing Thermal Barrier Coatings:
Materials and application routes. June 2014.
5. The mullite coatings on heaters made of molybdenum disilicide. P.S. Kisly, V.Y. Kodash.
1989, Ceramics International, pp. 189-191.
6. W.G. Sloof, A.L. Carabat. Encapsulation of MoSi2 self-healing particles for high
temperature applications. Delft : s.n., 2014.
7. Oxygen diffusion in zircon. E.B. Watson, D.J. Cherniak. 1997, Earth and Planetary
Science L.etters, pp. 527-544.
8. Effect of matrices on the phase transformation of ZrOz in the ZrO2-MOx (MOx = SiOa,
AIzO3) system. Y. Kanno, T. Suzuki. 1989, JOURNAL OF MATERIALS SCIENCE
LETTERS, pp. 41-43.
9. Preparation of alumina films by the sol-gel method. Y. Kobayashi, T. Ishizaka, Y.
Kurokawa. 2005, Journal of Materials Science, pp. 263-283.
10. Alumina coating of austenitic stainless steel by sol-gel process. K. Miyazawa, T.
Sakuma. 1991, Engineering Fracture Mechanics, pp. 967-973.
11. Fe3O4@Al2O3 magnetic core–shell microspheres for rapid and highly specific capture of
phosphopeptides with mass spectrometry analysis. Li, Y. 2007, Journal of chromatography,
pp. 57-71.
12. Investigation of thermal oxidation of Al2O3-coated SiC powder. A. Dey, N. Kayal, A.R.
Molla, O. Chakrabarti. 2014, Thermochimica Acta, pp. 25-31.
13. Effect of acid on the coating of boehmite onto silicon carbide particles in aqueous
suspensions. C. Yang, W. Shih. 1999, Journal of the American Ceramic Society, pp. 436440.
14. Sol-gel alumina environmental barrier coatings for SiC grit. H. Kim, M. Chen, Q. Yang,
T. Troczynski. 2006, Materials Science and Engineering, pp. 150-154.
15. Crystallinity of inorganic films grown by atomic layer deposition: Overview and general
trends. V. Miikkulainen, M. Leskelä, M. Ritala, R. L. Puurunen. 2013, Journal of Applied
Physics.
16. Atomic Layer Deposition: An Overview. George, S.M. 2010, Chem. Rev., pp. 111-131.
17. García-Triñanes, P. Enhancing the activation of silicon carbide particles with gas-phase
coating of aluminium oxide. 2014.
18. Atomic Layer Deposition of Oxide Thin Films with Metal Alkoxides as oxygen sources. M.
Ritala, K. Kukli,A.i Rahtu,P.i I. Raisanen, M. Leskela,T. Sajavaara, J. Keinonen. 2000,
Science.
15
Appendix III: Matlab code used for the diffusion model
function y = Scaledeveloptest(x);
%Parameters to be adjusted
T=1273;
%Temperature of the system in Kelvin
P0=0.2;
%Partial pressure of oxygen in the gas phase
in bar
t=24*3600;
%Time in seconds the simulation should cover (3600s=1h,
86400s=1 day 31536000s=1 year)
deltt=t;
%number of steps to solve for
d_Al_ini=500;
%Thickness of the initial alumina layer in nanometer
d_Mul_ini=5;
%Thickness of the initial mullite layer in nanometer
d_Si_ini=1;
%Thickness of the initial silica layer in nanometer
G_Al=50;
%Grain size in alumina in nm (for GB diffusion)
%Defined parameters (for diffusion coefficients: first letter(s) is species
%diffusing, second is phase diffusion is happening in.
R=8.3145;
%Gas constant in J/K*mol
M_Al=101.96;
%Molecular weight Al2O3 in g/mol
Rho_Al=3.98*10^6;
%Density Al2O3 in g/m3
M_Mul=426.06;
%Molecular weight mullite in g/mol
Rho_Mul=3.22*10^6; %Density mullite in g/m3
M_Si=60.08;
%Molecular weight amorphous SiO2 in g/mol
Rho_Si=2.196*10^6; %Density amorphous SiO2 in g/m3
D_O_Al=(2/(G_Al*10^-9))*1.8*10^-10*exp(-375000/(R*T));
%Total oxygen diffusion coefficient in alumina (including GB diffusion)
C0_O_Al=3*Rho_Al/M_Al;
%Bulk
concentration O in Al2O3 based on theoretical density and molecular weight
Al2O3 (mol/m3)
D_O_Mul=(2*1*10^-9/(G_Al*10^-9))*1.32*10^-(6-4)*exp(-397000/(R*T));
%Total oxygen diffusion coefficient in mullite (including GB diffusion)
C0_O_Mul=13*Rho_Mul/M_Mul;
%Bulk
concentration O in mullite based on theoretical density and molecular
weight Al6Si2O13 (mol/m3)
D_Al_Mul=(2*1*10^-9/(G_Al*10^-9))*9.2*10^-(3-4)*exp(-517000/(R*T));
%Total aluminium diffusion coefficient in mullite (including GB diffusion)
C0_Al_Mul=6*Rho_Mul/M_Mul;
%Bulk
concentration Al in mullite based on theoretical density and molecular
weight Al6Si2O13 (mol/m3)
D_O_Si=4.4*10^-15*exp(-121000/(R*T));
%Total oxygen diffusion coefficient in amorphous silicon
C0_O_Si=2*Rho_Si/M_Si;
%Bulk
concentration O in SiO2 based on density and molecular weight amorphous
SiO2 (mol/m3)
Delt_Mu_Mulform=((-6888.12+0.26597*T)-3*(-1675.7+0.0592*T)-2*(910.7+0.04146*T))/(6*R*T); %Chemical potential per atom Al over mullite
barrier corrected for RT
P3=exp(1000*(-910.7+0.18*T)/(R*T));
%Equilibrium partial pressure of oxygen at Si-SiO2 interface in bar d(:,3)+
k_Al=D_O_Al*C0_O_Al/(2*d_Al_ini*10^-9);
k_Mul=D_O_Mul*C0_O_Mul/(2*d_Mul_ini*10^-9);
k_Si=D_O_Si*C0_O_Si/(2*d_Si_ini*10^-9);
tspan=linspace(0,t,deltt).';
d_ini1=d_Al_ini*10^-9;
d_ini2=d_Mul_ini*10^-9;
d_ini3=d_Si_ini*10^-9;
16
d0=[d_ini1, d_ini2, d_ini3];
Vdata=[D_O_Al, C0_O_Al, M_Al, Rho_Al, D_O_Mul, C0_O_Mul, M_Mul, Rho_Mul,
D_Al_Mul, C0_Al_Mul, D_O_Si, C0_O_Si, M_Si, Rho_Si, P0, P3,
Delt_Mu_Mulform];
ode = @(t,d) Scalediftest(t,d,Vdata);
[t,d] = ode45(ode, tspan, d0);
deq=d(:,3)+d(:,2)*Rho_Mul/M_Mul*2*M_Si/Rho_Si;
end
function dddt=Scalediftest(t,d,Vdata);
D_O_Al=Vdata(1);
C0_O_Al=Vdata(2);
M_Al=Vdata(3);
Rho_Al=Vdata(4);
D_O_Mul=Vdata(5);
C0_O_Mul=Vdata(6);
M_Mul=Vdata(7);
Rho_Mul=Vdata(8);
D_Al_Mul=Vdata(9);
C0_Al_Mul=Vdata(10);
D_O_Si=Vdata(11);
C0_O_Si=Vdata(12);
M_Si=Vdata(13);
Rho_Si=Vdata(14);
P0=Vdata(15);
P3=Vdata(16);
Delt_Mu_Mulform=Vdata(17);
k_Al=D_O_Al*C0_O_Al/(2*d(1));
%kinetic factor for
Al combining all relevant parameters, for abbreviation purposes
k_Mul=D_O_Mul*C0_O_Mul/(2*d(2));
%kinetic factor for
Mul combining all relevant parameters, for abbreviation purposes
k_Si=D_O_Si*C0_O_Si/(2*d(3));
%kinetic factor for
Si combining all relevant parameters, for abbreviation purposes
P2=(P0^((1/(k_Mul/k_Al+1))/((k_Si/k_Mul)+1((k_Mul/k_Al)/((k_Mul/k_Al)+1))))*P3^((k_Si/k_Mul)/((k_Si/k_Mul)+1((k_Mul/k_Al)/((k_Mul/k_Al)+1)))));
%Calculation O2 partial pressure at
mullite-SiO2 interface with linear profile
P1=(P0*P2^(k_Mul/k_Al))^(1/(k_Mul/k_Al+1));
%Calculation O2 partial pressure at Al2O3-mullite interface with linear
profile
J_O_Al=-k_Al*log(P1/P0);
%Calculation O flux in Al layer
J_Al_Mul=-(D_Al_Mul*C0_Al_Mul/(d(2)))*Delt_Mu_Mulform*1000;
%Calculation Al flux through mullite layer
J_O_Si=-k_Si*log(P3/P2);
%Calculation O flux through Si layer
if ((3/2)*J_O_Al)>=J_Al_Mul;
%To
ensure that SiO2 layer does not obtain a negative thickness if aluminium
diffusion through mullite becomes too large.
dddt_1=-J_Al_Mul*M_Al/(2*Rho_Al);
dddt_2=J_Al_Mul*M_Mul/(6*Rho_Mul);
dddt_3=J_O_Si*M_Si/(2*Rho_Si)-J_Al_Mul*M_Si/(3*Rho_Si);
else
17
dddt_1=-J_O_Al*3*M_Al/(2*Rho_Al);
dddt_2=J_O_Al*M_Mul/(4*Rho_Mul);
dddt_3=0;
end
dddt=[dddt_1; dddt_2; dddt_3];
end
18
Appendix IV: Required coating thickness and amount of coating
One of the more important aspects of coated healing particles is the volume of these particles that
can be utilized for healing. The thinner the coating and the smaller the surface area that has to be
covered, the smaller the volume fraction of coating. This results in a larger volume fraction being
healing material that can be used to restore the material.
Therefore, to minimize the presence of inert coating material, total surface area and coating
thickness should be reduced as much as possible. As coating thickness is determined by lifetime
requirements and material, both of which are fixed, it is important to reduce the particle surface area
as much as possible. This importance is shown in the two figures below, one of which calculates the
volume fraction of alumina present in the system as function of the total surface area to be covered
as if the surface is a completely flat surface for different relevant coating thicknesses. This results in
the first equation to calculate total alumina volume fraction. The second figure calculates the
alumina volume fraction by assuming perfectly spherical particles of uniform size, resulting in the
second equation. Also indicated in both graphs are the theoretical minimum surface area of particles
with a volume average diameter of 18 µm (the starting material) and the surface area as measured
by N2 physisorption.
100
90
0.1 micron
0.2 micron
0.5 micron
1 micron
Theoretical minimum surface area
Measured surface area
Volume percentage alumina (%)
80
70
60
50
40
30
20
10
0
0
0.1
0.2
0.3
0.4
0.5
0.6
Specific surface area particles (m2/g)
0.7
0.8
0.9
1
In which VAl is the volume percentage of alumina present in the system in %, SMoSi2 the specific
surface area of the MoSi2 particles in m2/g MoSi2, δAl the coating thickness of the alumina in m and
ρMoSi2 the density of MoSi2 in g/m3.
19
100
0.1 micron
0.5 micron
Theoretical minimum surface area
90
0.2 micron
1 micron
Measured surface area
80
Volume percentage alumina (%)
70
60
50
40
30
20
10
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Specific surface area particles (m2/g)
In which VAl is the volume percentage of alumina present in the system in %, SMoSi2 the specific
surface area of the MoSi2 particles in m2/g MoSi2, δAl the coating thickness of the alumina in m and
ρMoSi2 the density of MoSi2 in g/m3.
Both of these calculations clearly show the need for appropriate control of total surface area, as even
a coating of 0.1 micron would result in a volume fraction close to 25 volume % alumina with both
calculations, even though the particles have an average size of 18 micron. For thicker coatings, the
volume percentage increases rapidly, which in turn means that the amount of healing particles needs
to be increased significantly to obtain the same healing potential.
It is also shown that these particles have a significantly higher specific surface area than would be
expected based on the average particle size. This indicates either a large polydispersity, particles that
are largely nonspherical or a combination of both. If a high size polydispersity is present, surface area
might be significantly decreased by removal of smaller particles, which contribute most to the total
surface area. As the most ideal particle size for healing purposes was determined to be somewhere
between 25 and 30 micron, this removal of small particles would not be a problem for the selfhealing system.
20
Appendix V: Principles characterization methods used
This appendix will introduce the characterization methods used in this thesis and the main principles
involved.
Wind sifting
Wind sifting, also often called zig-zag air classification, is a process in which particles are separated
based on their falling behavior in an upwards air flow. This process is schematically shown in the
figure below (Rosenbrand 1986). The classifier works by introducing powder in the middle of a zigzag channel (hence the name zig-zag classification) and introducing a gas flow at the bottom. As the
powder falls from one part of the channel into the other, it is mixed with the turbulent stream of gas
and aerodynamically light particles disperse in the gas flow and mostly end up in the light fraction at
the top of the classifier. The heavier particles continue falling to the bottom and end up in the heavy
fraction. This essentially results in a countercurrent flow of gas with lighter particles and heavy
particles, with mixing at each change of direction.
Due to the many channels present, particles are separated very accurately depending on their
hydrodynamic diameter, which is mainly dependent on the surface area and weight of the particle.
Therefore the density, porosity and shape of the particle are very important in determining which
particles end up in the light or heavy fraction. In the system used at TU delft, the setup has been
modified slightly, with the channels mounted on a rotating shaft and the centrifugal force replacing
gravity for the heavier particles. This possibly significantly higher force allows significantly smaller
particles to be separated, with a lower limit of approximately 1-2 µm depending on the material and
other powder properties. By regulating centrifugal force (by changing the rotation speed) and the gas
velocity, particles can be separated in a very narrow size range.
21
Scanning Electron Microscopy (SEM) and Energy Dispersive x-ray Spectroscopy (EDS)
A scanning electron microscope (SEM) produces images of a sample by scanning the surface with
high energy electrons. These electrons interact with the material and produce many different signals,
as illustrated in the image below (centre). The main signals are backscattered electrons (BES),
Secondary Electrons (SEI) and characteristic X-rays.
Backscatter electrons are electrons that are reflected which can subsequently be detected by an
appropriate detector. As scattering is more likely when sample electron density is higher, solids with
heavier elements (higher electron density) generally scatter more electrons. Therefore these
materials often appear brighter on a BES image, due to the higher intensity of electrons detected
from that area. If elements and phases present in the sample are known, this allows for easy
distinction between different phases in a SEM.
Secondary electrons on the other hand are electrons expelled from the sample by the primary
electrons. These electrons have low energy and are therefore have a short mean free path in a solid
material (with a maximum of approximately 10 nm for most materials). Due to this short mean free
path, constructing an image with SEI allows for high spatial resolution of the surface of a sample.
The third important signal, the characteristic x-rays, originate from primary electrons removing
another electron from some of the inner orbitals. One of the electrons from a shell above that will
subsequently fall back into this orbital of lower energy, releasing the excess energy as a characteristic
x-ray. As these x-rays are element specific, these can be used to locally identify elements present in
each phase.
As these characteristic x-rays can be formed at any depth in the sample at which the primary beam
still has sufficient energy to remove an electron from the respective core shell and this penetration
depth is known for most acceleration voltages, this effect can be utilized to calculate coating
thickness. Assuming that the material densities and compositions are known, appropriate iterative
procedures can be used to calculate layer thickness, which is described in the article of Goldstein et
al (Goldstein, Choi et al. 1993).
22
X-ray Diffraction (XRD)
X-ray diffraction is capable of measuring the electron densities of a material and thereby determining
the crystal structure(s) present in the sample under investigation. XRD is the scattering of
monochromatic X-rays by interactions with electrons. As electron density in a crystalline solid is
repetitive in certain directions due to the repetition of the atoms and wavelengths of X-rays are in
the same order of magnitude as the atomic distances in a crystalline lattice, interference between
the photons and the solid occurs and regular scattering can take place.
When these scattered waves are not in phase, there is destructive interference, which means the
photons extinguish each other when they are not in phase and no X-rays can be detected at that
angle. However, when they are in phase, there is constructive interference and the X-rays can be
detected. This can be described with Braggs law:
n*λ = 2*d*sin θ
In which n is an integer, λ is the wavelength of the incident X-rays, d is the distance between the
atoms in the lattice and θ is the angle of incidence and diffraction. A schematic overview can be seen
in the figure below (Moulijn, van Leeuwen et al. 1993).
For small particles, the crystal structure is only regular for relatively short distances, resulting in peak
broadening due to uncertainty. Due to this effect, particle or grain size can be estimated using the
Scherrer equation:
In which D is the average crystal diameter in nanometers, K is a constant depending on the crystalline
shape, λ is the wavelength of the radiation used in nm, b is the diffraction broadening, measured as
the full width of the peak at half the maximum height and θ is again the diffraction angle of the peak
under consideration. However, significant stresses in the material on a small scale can also cause
atomic distances to change slightly, also resulting in peak broadening. Therefore, the presence of
microstresses should be verified and taken into account.
Due to macroscopic stresses in a certain direction, atomic distances can also change, but in a regular
way, resulting in a peak shift. This peak shift is dependent on the magnitude of the stress and can
therefore be used to calculate the macroscopic stresses being present.
23
N2 physisorption
N2 Physisorption is a method to study some of the physical properties of the sample. This
characterization technique uses nitrogen as an adsorbate and is carried out at exactly the boiling
point of liquid nitrogen, which is 77.3 K. After weighing and degassing the sample, it is brought into a
high vacuum and cooled to the required temperature. Then small and known volumes of gas are
introduced, while simultaneously measuring the pressure. After each addition of gas, part of the
adsorbate will be adsorbed on the sample. When equilibrium is reached in between the additions,
indicated by an absence of change in N2 pressure with time, the total amount of nitrogen adsorbed
on the sample can be calculated from the pressure change. This yields the adsorption isotherm.
After the final pressure has been reached, the pressure is slowly reduced by taking away small
volumes of gas, which is the desorption part. Depending on the material, the adsorption and
desorption curve can be the same or very different. The difference in these curves is mostly caused
by mesoporosity, originating from capillary condensation effects inside these pores.
The physical interpretation of the isotherm is that as the nitrogen pressure increases, a layer of
nitrogen molecules gets deposited on the surface, due to weak, physical attractive interactions
between the N2 and the surface of the material. If more nitrogen becomes available, subsequent
layers start to form. Because the attractive forces are stronger in the pores as more surface is
available and also closer, multilayers start to form earlier inside the smallest pores. Therefore, the
micropores will be completely filled with nitrogen first and at higher pressures, also the meso- and
macropores will be filled. Due to this effect, that is called pore condensation, the specific surface
area of the total sample, the pore size and pore size distribution can be calculated from the isotherm.
For calculating the specific surface area of a sample, the BET method is used. This is a modification of
the Langmuir theory developed by Brunauer, Emmett and Teller that takes the simultaneous
formation of monolayers and multilayers into account. For the pore size and volume distribution, the
t-method can be used.
X-ray Photoelectron Spectroscopy (XPS)
In X-ray Photoelectron spectroscopy (XPS), a sample surface is irradiated with X-rays, which by the
photoelectric effect subsequently emits electrons. The different binding energies for each element
causes these photoelectrons to have different energies depending on the x-ray source used, the
element from which the electrons originate and the binding state of the element.
Therefore this technique can be used to analyze the top 10 nm of a surface for elemental
composition and with a detector with appropriate accuracy, also the chemical and electronic
environment of each element, yielding valuable information regarding the surface of the material
under investigation.
Laser diffraction
Laser diffraction is a technique that uses the scattering of a laser by solid particles to determine the
geometric dimensions of these particles. The Fraunhofer diffraction theory states that the angle of
long-range diffraction of light is inversely proportional to the size of the particle diffracting it.
Therefore, particle size can be estimated by measuring the diffraction angle of a laser passing
through a suspension containing these particles.
24
If particles of multiple sizes are present, the light intensity of each angle is an indication for the
amount of particles in that size range being present. This allows for easy measurement of a particle
size distribution of a powder.
ThermoGravimetric Analysis (TGA)
Thermogravimetric analysis (TGA) is a thermal analysis method in which weight changes are
measured in time at high temperatures. Often either temperature is kept constant for a certain
period of time or the temperature is increased/decreased with a constant rate, in both cases
measuring weight change of the sample over this period of time.
This analysis technique provides valuable information regarding thermal stability of a sample and
phenomena that cause the sample to either gain or lose weight. Common processes for weight
change are oxidation or other solid-gas reactions, chemisorption/physisorption, decomposition or
evaporation. If the sample composition and environment are carefully controlled, important data on
the rate at which these processes occur at certain temperatures can be gathered and the extend of
these processes can be measured.
X-Ray Fluorescence (XRF)
X-ray fluorescence (XRF) is a technique that also uses X-rays to excite electrons, like XPS. However,
the main difference is that in XRF, characteristic X-rays originating from the falling back of outer
electrons to a vacant orbital in the core shells are measured instead of photoelectrons. These
characteristic X-rays are also element specific, allowing for qualitative analysis. By measuring the
intensities of each wavelength and proper calibration and correction, quantitative analysis is also
possible. As X-rays do not interact in the same way as electrons with solids, energy loss by scattering
is not possible and therefore the penetration depth is usually sufficient for bulk elemental analysis.
Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES)
The ICP-OES technique works by introducing the completely dissolved sample into an argon plasma,
in which the sample is subsequently ionized. Due to the constant ionization and recombination of the
atoms present in the plasma, radiation with characteristic wavelengths of the elements present is
constantly emitted. This radiation is then detected and the intensity of each wavelength is a good
indication of the quantity of that element present.
centre, U. o. G. I. s. a. a. Scanning Electron Microscopy.
Goldstein, J. I., S. Choi, F. Van Loo, H. Heijligers, G. Bastin and W. Sloof (1993). "The influence of oxide
surface layers on bulk electron probe microanalysis of oxygen—application to Ti‐Si‐O compounds."
Scanning 15(3): 165-170.
Moulijn, J. A., P. W. van Leeuwen and R. A. van Santen (1993). Catalysis: an integrated approach to
homogeneous, heterogeneous and industrial catalysis, Elsevier.
Rosenbrand, G. G. (1986). The separation performance and capacity of zigzag air classifiers at high
particle feed rates, Dissertation TU Eindhoven.
25
Appendix VI: The ALD setup
This appendix shows a flowchart of the ALD setup that was used in this research project.
26
Appendix VII: EDS maps of wind sifted MoSi2B particles
This appendix shows the elemental maps of wind sifted MoSi2B particles, as measured by EDS
(acceleration voltage: 15 kV). In these images, the upper left figure shows the SEM-SEI image of the
mapped area and the other figures are intensity maps of the indicated elements.
27
Appendix VIII: XRF results of MoSi2B particles
Presented in the table below are the elemental concentrations of the MoSi2B before and after wind
sifting, as measured by X-Ray Fluorescence (XRF).
MoSi2B as received
Concentration
Element
(wt%)
Mo
58.8
Si
40.5
Fe
0.23
Zr
0.071
Mg
0.063
W
0.061
Al
0.057
Cr
0.039
Bi
0.035
Ti
0.027
Ca
0.024
Cu
0.023
Ni
0.021
absolute
error (wt %)
0.1
0.1
0.01
0.008
0.008
0.007
0.007
0.006
0.006
0.005
0.005
0.005
0.004
Element
Mo
Si
Fe
Mg
Al
Zr
W
Cr
Ti
Ni
Cu
MoSi2B wind sifted
Concentration
absolute
(wt%)
error (wt%)
68.9
0.2
30.5
0.1
0.193
0.01
0.117
0.01
0.049
0.007
0.047
0.007
0.042
0.006
0.042
0.006
0.033
0.005
0.023
0.005
0.014
0.003
28
Appendix IX: EDS individual data points aluminium alkoxide coatings
This appendix shows the individual EDS measurements for the samples coated with an alkoxide as
precursor.
Alumium isopropoxide sample
Aluminium tri-sec-butoxide 20g
evaporated
Aluminium tri-sec-butoxide 10g
evaporated
Concentration Al
detected (atom %)
3.18
error
(atom %)
+/- 0.40
Concentration Al
detected (atom %)
28.58
error
(atom %)
+/-0.20
Concentration Al
detected (atom %)
18.51
error
(atom %)
+/-0.45
1.27
+/- 0.16
29.17
+/-0.20
11.48
+/-0.43
1.5
+/- 0.20
31.77
+/-0.24
17.25
+/-0.76
0.43
+/- 0.17
21.27
+/-0.21
21.98
+/-1.06
2.85
+/- 0.39
22.23
+/-0.18
7.16
+/-0.42
0.29
+/- 0.18
24.18
+/-0.16
2.83
+/-0.43
18.77
+/- 0.30
7.19
+/-0.19
17.19
+/-0.51
3.12
+/- 0.37
3.7
+/-0.22
6.77
+/-0.63
25.63
+/- 0.35
28.97
+/-0.20
16.56
+/-0.41
46.29
+/- 0.64
26.04
+/-0.17
11.35
+/-1.98
32.79
+/- 0.63
30.38
+/-0.21
11.23
+/-0.45
26.42
+/- 0.34
7
+/-0.25
12.53
+/-0.39
18.3
+/-0.72
11.07
+/-0.40
2.4
+/-0.11
8.03
+/-1.19
Aluminium tri-sec-butoxide 5g
evaporated
Concentration Al
detected (atom %)
12.91
error
(atom %)
+/-0.92
27.25
+/-0.18
19.11
+/-0.68
21.5
+/-0.15
9.01
+/-0.60
11.16
7.42
+/-1.27
+/-0.27
21.87
24.73
+/-0.15
+/-0.16
12.29
26
+/-0.48
+/-0.40
6.4
+/-0.45
31.82
+/-0.24
46.93
+/-1.13
13.06
1.63
8.27
9.56
+/-1.53
+/-0.23
+/-0.35
+/-0.43
36.77
31.94
30.08
34.03
+/-0.35
+/-0.32
+/-0.21
+/-0.29
15.68
15.32
12.29
21.81
+/-0.64
+/-0.47
+/-0.43
+/-0.57
33.6
+/-0.29
2.46
+/-0.45
25.66
+/-0.17
11.07
+/-0.34
29
Aluminium tri-sec-butoxide 10g centrifuged
Concentration Al detected (atom %) error (atom %)
2.04
+/-0.23
0
?
Aluminium tri-sec-butoxide 20g centrifuged
Concentration Al detected (atom %)
error (atom %)
1.65
+/-0.23
1.54
+/-0.27
1.7
+/-0.22
10.61
+/-0.72
2.47
+/-0.18
0.49
+/-0.08
14.11
+/-0.82
2.09
+/-0.14
30
Appendix X: TGA mass change graphs
Derivative weight (A.U.)
This appendix shows the mass change graphs of the sol-gel samples that were tested with
ThermoGravimetric Analysis (TGA) after coating to obtain important information regarding annealing
behaviour. In this graph, the starting point at 25 ᵒC is a mass change of exactly 0 mg/min.
0
100
200
300
400
sample oxalate A
Sample oxalate B
Sample Isopropoxide
Sample tri-sec-butoxide
500
Temperature (ᵒC)
600
700
800
900
1000
31
Appendix XI: XPS results main samples
This appendix presents the results of the XPS analyses performed on the sol-gel samples prepared
with 10 and 20 g of aluminium tri-sec-butoxide (SG10g and SG20g respectively) and the ALD/rCVD
samples that underwent 25 and 40 cycles (ALD25C and ALD40C respectively). The first figure shows
the entire spectrum recorded with peaks indicated. The two figures below that show the results for
the O 1s and the Al 2p peaks in more detail.
SG10g
SG20g
ALD25C
Counts (A.U.)
ALD40C
1000
800
600
400
200
0
545
540
535
530
525
Binding energy (eV)
520
SG10g
SG10g
SG20g
SG20g
ALD25C
ALD25C
ALD40C
ALD40C
515
Counts (A.U.)
Counts (A.U.)
Binding energy (eV)
510
90
85
80
75
Binding energy (eV)
70
65
These results clearly show that the main constituent elements are as expected Al and O on the
surface. The indium was detected because of the indium plates used as support in this analysis.
When investigating the O 1s and Al 2p peaks, an interesting difference is visible between the sol-gel
and ALD samples, but not between the two ALD samples or sol-gel samples themselves. This
indicates a clear difference in chemical environment at the surface of the coatings produced with sol32
gel compared to ALD. This would be expected as the sol-gel samples should be consisting of
boehmite (AlOOH) while the ALD samples should most likely consist of pure Al2O3 with some hydroxyl
groups on the surface.
For both the O and Al binding energies, the shift is 1 eV between sol-gel and ALD (535.0 eV vs 534.0
eV for the O 1s and 77.8 vs 76.8 for the Al 2p energy), with the sol-gel being slightly more #weakly or
strongly?# bound. Deconvolution and attribution of the various peaks is however difficult, as the
literature peak values are slightly lower (between 530 and 532 eV for the O 1s and around 75 eV for
the Al 2p) than measured here. It is however clear that there is a difference in chemical environment
between the two production methods.
Furthermore, hydroxyl groups seem to have a tendency to shift the binding energy to slightly higher
values, which would indicate that the sol-gel samples are likely to contain more hydroxyl groups than
the ALD samples. This would be as expected, due to the expected lower hydroxyl content from the
production process of ALD samples. If this is indeed the case, this is another argument in favor of the
proposed mechanism of water condensation, as TMA condensation would likely result in a
significantly higher content of hydroxyl groups in the bulk of the material due to the excess of water
present during the water addition step.
Finally, XPS also gives an indication of coverage. If the intensity of elements present in the support is
limited, this indicates that coating was successful. The first figure already shows a very low intensity
for Mo and Si in the sol-gel samples and no peaks for the ALD samples. To ensure that this is indeed
the case, a more detailed analysis of the Mo 3d5 peaks was made, which confirms the absence of Mo
at the surface of the ALD particles and the very limited presence of Mo in the sol-gel samples,
indicating that coating was successful with both methods.
Counts (A.U.)
SG10g
SG20g
ALD25C
ALD40C
250
245
240
235
230
225
220
Binding energy (eV)
33
Appendix XII: Cross-section thickness measurement images
This appendix shows the Scanning Electron Microscopy Backscatter Electron Spectrum (SEM-BES)
images used to construct the coating thickness distributions shown in the results chapter.
Sol-gel 10g images
Sol-gel 20g
34
35
Sol-gel 20g after heat treatment at 1200 ᵒC
36
ALD 25 cycles
37
ALD 40 cycles
38
Appendix XIII: ICP-OES results for ALD samples
ICP-OES results for some of the samples are shown in the table below and did give results
significantly lower than those of the EPMA method for those samples. This discrepancy most likely
stems from the difficulty of dissolving aluminosilicates that most likely formed at the interface.
However, due to Al and Mo interference, precision and accuracy were also limited, adding another
factor that could explain this observation. This is evident from the blank, which contained less than
0.01 wt% Al, already containing 0.70 wt% Al according to ICP-OES. This is a significant amount
compared to the 0.8-1.0 wt% detected in the other samples that were coated by alumina. #Should
ICP results be reported? Not sure whether results relevant enough#
Table 5.1: ICP-OES results and calculated thickness of the alumina layer.
Sample
Blank
4 min TMA
8 min TMA
0.5 min H2O
Measured Al weight percentage (wt%) calculated thickness alumina layer (nm)
0.7
0.99
8.7
0.99
8.9
0.83
3.8
39
Appendix XIV: ALD 40 cycle particle EDS map
This appendix shows the elemental maps of a particle coated by ALD with 40 cycles, after heat
treatment at 1200 ᵒC, as measured by EDS (acceleration voltage: 15 kV). In these images, the upper
left figure shows the SEM-SEI image of the mapped area and the other figures are intensity maps of
the indicated elements.
40
Appendix XV: EDS data of coated MoSi2B particles
This appendix shows the EDS data of MoSi2B particles coated with the sol-gel procedure mentioned
in chapter 4 and annealed at 1200 ᵒC. The SEM image indicates the locations of the points, while the
detected atom % concentrations of each element can be found in the table below.
Table with atom % of each element detected.
C-K
O-K
Al-K
Si-K
Mo-L
Point 1
20.72
17.37
4.9
38.38
18.63
Point 2
13.25
32.99
6.5
32.75
14.51
Point 3
9.25
39.52
13.29
26.4
11.53
Point 4
12.76
30.82
4.09
36.51
15.81
Point 5
12.16
51.7
31.58
3.72
0.84
Point 6
17.38
11.62
2.66
44.23
24.1
Point 7
2.96
48.34
37.86
9.01
1.84
Point 8
17.08
9.37
2.06
46.09
25.39
Point 9
9.77
55
25.36
6.52
3.36
Point 10
12.02
16.56
6.69
43.39
21.34
41
Appendix XVI: XRD diffractograms of YSZ powder and SPS samples
This appendix shows the XRD results of the investigation on the YSZ powder from Tosoh and three of
the dummy SPS samples manufactured under different conditions. The powder did not receive any
heat treatment after manufacture, while SPS-253 was processed in the SPS with a maximum
temperature of 1350 ᵒC for 10 minutes and both SPS 266 and 267 were processed at 1450 ᵒC for 5
minutes. Pressure exerted by the machine was 35 MPa in all cases, but sample SPS 266 received an
additional heat treatment in air for 1 hour at 1000 ᵒC (heating and cooling rate 5 ᵒC/min). XRD results
are shown in the two figures below, with the first figure showing the entire diffractogram and the
second figure magnifying the area around the largest tetragonal ZrO2 peak around a 2θ of 30ᵒ.
SPS253
SPS 266
SPS 267
Intensity (A.U.)
YSZ Tosoh Powder
15
25
35
45
55
65
75
85
2θ (ᵒ)
SPS253
SPS 266
SPS 267
Intensity (A.U.)
YSZ Tosoh Powder
27
27.5
28
28.5
29
29.5
30
30.5
31
31.5
32
2θ (ᵒ)
42
Appendix XVII: EDS maps of the YSZ/MoSi2B composite
This appendix shows the elemental maps of two areas of the YSZ/coated MoSi2B composite before
and after heat treatment at 1200 ᵒC for 1h, as measured by EDS (acceleration voltage: 15 kV). In
these images, the upper left figure shows the SEM-SEI image of the mapped area and the other
figures are intensity maps of the indicated elements, starting with the before images.
43
After healing for 1h at 1200 ᵒC (different area).
44
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertising