Page 1
1
On the Degradation Mechanisms of Thermal
Barrier Coatings: Effects of bond coat and
substrate
A thesis submitted to the University of Manchester for the degree of
Doctor of Philosophy
in the Faculty of Engineering and Physical Science
2014
LIBERTY TSE SHU WU
School of Materials
Page 2
2
Table of Contents
List of Figures ................................................................................................................ 4
List of Tables ................................................................................................................ 13
Abstract ........................................................................................................................ 15
Declaration ................................................................................................................... 16
Copyright Statement .................................................................................................... 17
Acknowledgement ....................................................................................................... 18
Vita ............................................................................................................................... 19
CHAPTER 1 Literature review .................................................................................... 20
1.1 Background information ........................................................................................ 20
1.1.1 Two types of top coat .......................................................................................... 22
1.1.2 Two types of bond coat ....................................................................................... 23
1.1.3 Oxidation and diffusion behavior in Pt/Pt-Al diffusion bond coats .................... 24
1.1.4 TGO residual stress ............................................................................................. 26
1.1.5 Stress evolution and failure mechanism near TGO interface ............................. 29
1.1.6 The rumpling behavior of Pt modified bond coats .......................................... 34
CHAPTER 2 Interfacial fracture toughness ................................................................ 43
2.1 Background information ........................................................................................ 43
2.2 Overview of the methods for measuring interfacial fracture toughness ................ 44
2.2.1 Conventional methods: Pull-off, double cantilever beam, and scratch adhesion
test, etc. ........................................................................................................................ 46
2.2.2 Indentation .......................................................................................................... 50
2.2.3 Barb ..................................................................................................................... 62
2.2.4 Four point-bending.............................................................................................. 66
2.3 Shortcomings of current interfacial fracture toughness models for indentation .... 70
2.4 Factors affecting interfacial toughness measurement ............................................ 73
CHAPTER 3 Scope and aims of research .................................................................... 75
3.1 Project introduction ................................................................................................ 75
3.2 Considerations and requirements for choosing an adequate adhesion test ............ 76
3.3 Project objectives ................................................................................................... 78
CHAPTER 4 ................................................................................................................ 80
Microstructure parameters affecting interfacial adhesion of Thermal Barrier Coatings
by the EB-PVD method ............................................................................................... 80
4.1 Introduction ............................................................................................................ 80
4.2 Experimental details............................................................................................... 83
4.3 Results .................................................................................................................... 91
4.3.1 Microscopy observation ...................................................................................... 91
Page 3
3
4.3.2 Lifetime and interfacial adhesion of TBC systems ............................................. 95
4.3.3 TGO thicknesses, residual stresses and rumpling at the TGO/bond coat interface
...................................................................................................................................... 97
4.3.4 Phase transformation and Ni/Ti segregation ..................................................... 100
4.4 Discussion ............................................................................................................ 105
4.4.1 Approximation of interfacial fracture toughness Gc ......................................... 105
4.4.2 Linear fitting of crack rate with degradation factors ........................................ 114
4.4.3 Possible degradation mechanisms by ’ formation rate .................................... 119
4.4.4 Stress relaxation by rumpling ........................................................................... 120
4.4.5 The dynamic feature of interfacial fracture toughness...................................... 123
4.5 Conclusion ........................................................................................................... 125
CHAPTER 5 .............................................................................................................. 127
The degradation mechanisms of TBCs: Driving force for spallation versus interfacial
fracture toughness Gc ................................................................................................. 127
5.1 Introduction .......................................................................................................... 127
5.2 Experimental details............................................................................................. 130
5.3 Results .................................................................................................................. 135
5.3.1 Microscopic observation near the TGO/bond coat interface ............................ 135
5.3.2 EBSD mapping analyses near the TGO/bond coat interface ............................ 137
5.3.3 EPMA mapping analysis near TGO/bond coat regions .................................... 144
5.3.4 Hardness by nano-indentation near the TGO/bond coat interface .................... 146
5.3.5 Simulation of phase transformation by Thermo-Calc ....................................... 147
5.4 Discussion ............................................................................................................ 149
5.4.1 Interfacial evolution during thermal exposure .................................................. 149
5.4.2. Phase transformation due to substrate/bond coat chemistry ............................ 152
5.4.3. Phase transformation induced interfacial rumpling ......................................... 154
5.4.4 Driving force for spallation versus interfacial fracture toughness .................... 159
5.5 Conclusion ........................................................................................................... 163
CHAPTER 6 .............................................................................................................. 164
Summary and suggested future work ......................................................................... 164
6.1 Summary .............................................................................................................. 164
6.2 Suggested future work ......................................................................................... 166
References .................................................................................................................. 168
Page 4
4
List of Figures
Fig. 1.1. A schematic diagram showing the multi-layer structure of thermal barrier
coating systems (TBCs) [6]…………………………………………………………..21
Fig. 1.2. A plot showing the improvement in thermal cyclic lifetime of thermal barrier
coating systems (TBCs) with 7 wt% yttria in ZrO2 [7]………………………………22
Fig. 1.3. A comparison of the coating morphology between APS (left) [6] and
EB-PVD (right), showing splat-like and columnar structures, respectively…………23
Fig. 1.4. A comparison of the microstructure between overlay (left) [13] and Pt-Al
diffusion bond coats (right). The diffusion zone is several times thicker in the overlay
bond coat……………………………………………………………………………..24
Fig. 1.5. The evolution of mean frequency shift measured for specimens that failed
after: (a) 190 (b) 225 (c) 207 (d) 237 cycles [28]…………………………………….28
Fig. 1.6. Low stress evolution map showing gradual coalescence of isolated stresses
[28]…………………………………………………………………………………...29
Fig. 1.7. TGO/bond coat interfacial roughness profile of three bond coats: a) LT Pt-Al
b) HT Pt-Al c) Pt diffusion against thermal cycles [35]. The Pt-diffusion bond coat
maintained similar roughness throughout the thermal cycles after 10
cycles…………………………………………………………………………………32
Fig. 1.8. Ternary Ni-Al-Pt phase diagram for three different bond coats: a) LT Pt-Al b)
HT Pt-Al c) Pt- diffusion at 1100°C and 1150°C showing a shift from to ’phase
with decreasing Pt and Al contents [36]……………………………………………...33
Fig. 1.9. The strain response in the bond coat throughout a thermal cycle. Point 7
denotes the onset of isothermal dwelling. The onset of cooling may start at either
point 1 (stress relaxation) or 1’ (residual stress) depending on the duration of
isothermal exposure [39]……………………………………………………………..35
Page 5
5
Fig. 1.10. The stress computed beneath an undulation in the bond coat s a) during
cooling and b) heating with phase transformation; c) during cooling and d) heating
without phase transformation [41]. Bond coats with phase transformation experience a
sharp rise/decline in the plane stresses during cooling and heating………………….38
Fig. 1.11. The difference between 11-22 computed for different bond coat yield
strengths [41]. In general, bond coats with lower yield strengths experience lower
plane stresses during phase transformation due to creep relaxation…………………39
Fig. 1.12. a) The difference between 11-22 computed for different transformation
temperature b) top coat 22 stress computed for different transformation temperature
[40]. Bond coat plane stresses can be relaxed when the transformation temperature is
raised…………………………………………………………………………………41
Fig. 1.13. The average amplitude growth of bond coat surface rumpling as a function
of martensitic transformation temperature showing an increasing rumpling trend with
higher transformation temperature. Note that the Y-axis indicates the % of total
growth that occurred during the transformation [45]………………………………...42
Fig. 2.1. An illustration showing the three pure modes of loading…………………..45
Fig. 2.2. Figures showing the loading angle in barb (left) and pull-off (right)
[48,55]………………………………………………………………………………..46
Fig. 2.3. Illustration of the tensile based direct pull-off test [59]…………………….47
Fig. 2.4. Illustration of the sample and set-up used in the double-cantilever beam
test……………………………………………………………………………………48
Fig. 2.5. a) Pile-up in front of scratching indenter, and b) buckling failure at the
interface of thin coatings [67]………………………………………………………..50
Fig. 2.6. a) the scratching indenter generates a wedge crack some distance ahead; b)
the coating lifted up the wedge crack to cause an interfacial crack c) a through
Page 6
6
thickness crack occurs right in front of the indenter leading to complete spallation
[67]…………………………………………………………………………………...50
Fig. 2.7. Indentation-induced delamination during loading [58]…………………….51
Fig. 2.8. Indentation-induced delamination during unloading [58]………………….51
Fig. 2.9. Figures showing the stages of fracture in a brittle coating with a hard
substrate due to a nano-indenter load [58]…………………………………………...52
Fig. 2.10. Illustration showing the half penny shaped later crack centered under the
indent, which is dependent on film and substrate properties, residual stress, etc
[70]…………………………………………………………………………………...53
Fig. 2.11. SEM image showing a delaminated black diamond film with a
straight-sided buckling behaviour [72]……………………………………………….55
Fig. 2.12. Schematic representation of the relationship between crack length and
diagonal length [79]………………………………………………………………….57
Fig. 2.13. Illustration showing the cross-sectional indentation right on the interface
[51]…………………………………………………………………………………...58
Fig. 2.14. Schematic representation of the crack/indent length – loading relation
[79]…………………………………………………………………………………...60
Fig. 2.15. SEM image showing an example of the crack generated at the interface due
to indentation below [82]. The indentation caused an upward displacement in the
entire coating with the highest point roughly above the central line of the
indent…………………………………………………………………………………61
Fig. 2.16. Illustration of the barb specimen showing a partial removal of the TBC
[55]…………………………………………………………………………………...63
Fig. 2.17. Schematic representation of the barb loading setup (shear mode)
[55]……………...........................................................................................................63
Page 7
7
Fig. 2.18. A plot of the typical loading curve obtained from the barb test on a EB-PVD
TBC system [55]……………………………………………………………………..64
Fig. 2.19. a) Fracture surface at low magnification b) a high-magnification SEM
image showing embedded TGO in the bond coat and the TBC remained on the TGO
[55]…………………………………………………………………………………...65
Fig. 2.20. Illustration of the modified four-point bending specimen and loading setup
with the newly applied stiffener layer for preventing segmentation in the ceramic
coating [54]…………………………………………………………………………..67
Fig. 2.21. A typical force-displacement plot showing the onset of delamination as
indicated by the sudden drop in load P [54]………………………………………….70
Fig. 2.22. Illustration of the two instrinsic failure modes of TBCs
[93]…………………………………………………………………………………...72
Fig. 2.23. Plots showing the dynamic change in the stress, loading angle, and energy
release rate of TBC interfacial degradation [89]……………………………………..73
Fig. 4.1. A comparison of the relative spallation lifetime of commercial TBCs
subjected to thermal cycling, where the X-axis indicates the types of superalloy while
legends indicate bond coats (LCBC = Pt-diffusion, RT22LT = LT Pt-Al, CN91PA =
HT Pt-Al) [26]……………..........................................................................................85
Fig. 4.2. SEM image showing the placement of the Vickers indenter at the TGO/bond
coat interface…………………………………………………………………………88
Fig. 4.3. SEM image showing the crack propagation to the left of the indent……….88
Fig. 4.4. SEM image showing the end of crack propagation to the left of the
indent…………………………………………………………………………………89
Fig. 4.5. A comparison of typical crack lengths between those generated by 9.8 and
4.5N indentation loads, showing differences of roughly two-folds in
Page 8
8
length…………………………………………………………………………………90
Fig. 4.6. Representative interfacial microstructure between the three different bond
coat systems (LT Pt-Al, HT Pt-Al, Pt-diffusion) paired with TMS-82+ superalloy. The
HT Pt-Al bond coat had the least rumpling, yet failed to survive up to
200hrs…………………………...................................................................................93
Fig. 4.7. Representative interfacial microstructure between the three different
superalloys (CMSX-4, TMS-82+, SRR99) paired with LT Pt-Al bond coat (SRR99
failed after 30hrs, and no photos are available for 100 and 200 hrs time trials). The
SRR99 specimen had the most ’phase formation and rumpling, and could not survive
up to 50hrs…………………………............................................................................94
Fig. 4.8. The variation of crack lengths vs. isothermal oxidation (1135°C) hours for the
five different TBC systems (i.e. the prefix TMS-82+ is substrate, while Pt-diffusion is
bond coat) using a load of 4900mN………………………………………………….96
Fig. 4.9. The TGO growth kinetics vs. isothermal oxidation (1135ºC) hours for the five
different TBC systems……………………………......................................................98
Fig. 4.10. The TGO compressive stress vs. isothermal oxidation (1135°C) hours for
the five different TBC systems…………………………………………………….....99
Fig. 4.11. The evolution of interfacial rumpling vs. isothermal oxidation (1135°C)
hours for the five different TBC systems…………………………………………...100
Fig. 4.12. The amount of ’ formation attached to the TGO/bond coat interface vs.
isothermal oxidation (1135°C) hours……………………………………………….101
Fig. 4.13. EDX mapping showing Ti distribution near the TGO/bond coat interface of
the coating after 30hrs of isothermal exposure. An obvious Ti enrichment can be seen
in the bond coat of the SRR99 specimen…………………………………………...102
Fig. 4.14. EPMA mapping showing the elemental mapping of Ni, Ti, Al, and O for
Page 9
9
SRR99 LT Pt-Al after 30hrs of isothermal exposure (the color bars are values in wt
%). The regions of Al and Cr depletion correspond exactly to the enrichment in Ni
and Ti contents…………...........................................................................................103
Fig. 4.15. a) The mole-fraction of different phases existing as a function of Al content
in the SRR99 LT Pt-Al sample obtained from calculation using Thermo-calc, and b)
The comparison of Ti concentration in and ’ phases of the three
alloys………………………………………………………………………………..105
Fig. 4.16. The crack radius (i.e. sum of the length of half indent and crack lengths)
values used in the calculation of Gc in this work…………………………………...109
Fig. 4.17. Micrograph showing the location of the u measurements (6 per
indent)……………………………………………………………………………….111
Fig. 4.18. The upward displacements (i.e. taken as the crack openings at the
TGO/bond coat interface near the corner of indents) used in the calculation of Gc in
this work…………………………………………………………………………….111
Fig. 4.19. The approximated values of Gc for the five TBCs based on the semi-cirular
plate model………………………………………………………………………….112
Fig. 4.20. Micrographs showing the extent of YSZ damage between as-coated TBC
and TBC subjected to 30hrs of isothermal exposure………………………………..113
Fig. 4.21. A quantitative comparison of crack rates amongst the five TBC
specimens……….......................................................................................................116
Fig. 4.22. A quantitative comparison of TGO growth rates amongst the five TBC
specimens…………………………………………………………………………...116
Fig. 4.23. A quantitative comparison of average TGO stresses amongst the five TBC
specimens…………………………………………………………………………...117
Fig. 4.24. A quantitative comparison of TGO rumpling rates amongst the five TBC
Page 10
10
specimens…………………………………………………………………………...117
Fig. 4.25. A quantitative comparison of ’phase formation rates amongst the five TBC
specimens…………………………………………………………………………...118
Fig. 4.26. Plots showing the linear dependency of crack rate on a) ’phase formation
rate, b) TGO thickness growth rate, c) TGO length growth rate (rumpling rate) and, d)
TGO average stress. The crack rate only increases linearly with the rate of ’phase
formation……………………………………………………………………………119
Fig. 4.27. The dependency of stress on the rumpling increase rate (highly linear
dependent)…………………………………………………………………………..122
Fig. 4.28. A semi-quantitative plot illustrating the variation of driving force G (steady
state) and fracture toughness Gc between 30 and 100 hrs of isothermal exposure. △G
gives an indication of the remaining spallation life in the TBC…………………….124
Fig. 5.1. EPMA mapping showing the formation of Ti concentrated ’ in the bond coat
region near the oxide-bond interface of the SRR99 LT Pt-Al (containing traces of
TiO2 in the TGO) and TMS-82+ HT Pt-Al after 30 and 100hrs of isothermal exposure,
respectively………………………………………………………………………….129
Fig. 5.2. SEM micrographs showing a comparison of bond coat regions between 30
and 50hrs isothermally exposed SRR99 LT Pt-Al and CMSX-4 LT Pt-Al specimens,
with a unique ’ formation at the grain boundaries of the SRR99 specimen……….136
Fig. 5.3. SEM micrographs showing a comparison of bond coat regions between 30
and 50 isothermally exposed TMS-82+ HT Pt-Al and TMS-82+ LT Pt-Al specimens,
with a complete martensitic transformation in the HT-Pt-Al bond coat……………137
Fig. 5.4. EBSD mapping showing a comparison of grain morphology of the bond coat
region near the oxide-bond interface of the TMS-82+ HT Pt-Al and TMS-82+ LT
Pt-Al prior to isothermal exposure, with the HT Pt-Al bond coat having much larger
Page 11
11
grains………………………………………………………………………………..138
Fig. 5.5. EBSD mapping showing the lath martensitic structure of the TMS-82+ HT
Pt-Al specimen, in contrast to the phase of the TMS-82+ LT Pt-Al after 30hrs of
isothermal exposure…………………………………………………………………139
Fig. 5.6. EBSD mapping showing the unique ’ precipitation at the grain boundaries
of the SRR LT Pt-Al, in contrast to the clean and uniform phase of the CMSX-4 LT
Pt-Al specimen after 30hrs of isothermal exposure………………………………...139
Fig. 5.7. EBSD mapping showing the degradation in pattern quality of martensitic
region relative to the amount of ’ formation in the bond coat of 100hrs isothermally
exposed CMSX-4 LT Pt-Al…………………………………………………………140
Fig. 5.8. EBSD mapping showing more severe degradation in pattern quality of
martensitic region with increasing amount of ’ formation in the bond coat of 100hrs
isothermally exposed TMS-82+ LT Pt-Al 100hrs…………………………………..141
Fig. 5.9. EBSD mapping showing very severe degradation in pattern quality
associated with even more ’ formation in the bond coat of 100hrs isothermally
exposed TMS-82+ HT Pt-Al………………………………………………………..141
Fig. 5.10. EBSD mapping highlighting the inherent difference between the ’
microstrucuture of the Pt-diffusion bond coat and the microstructure of the LT Pt-Al
bond coat……………………………………………………………………………142
Fig. 5.11. A comparison of the misorientation profile near the oxide-bond coat region
between the TMS-82+ Pt diffusion (top) and SRR99 LT Pt-Al (bottom) after 100hrs
and 30hrs of isothermal exposure, with much sharper peaks in the SRR99 specimen
indicating regions of ’ precipitation………………………………………………..143
Fig. 5.12. EPMA mapping showing the chemistry near the oxide-bond coat regions of
the TMS-82+ Pt-diffusion specimens after 30 and 100hrs of isothermal exposure, with
Page 12
12
uniformly depleted layer of Al and Ti while enriched with Cr……………………..145
Fig. 5.13. EPMA mapping showing a comparison of Cr content near oxide-bond coat
interface between TMS-82+ LT Pt-Al and CMSX-4 LT Pt-Al after 100hrs of
isothermal exposure, with CMSX-4 showing no obvious Ti presence……………..146
Fig. 5.14. A plot showing the hardness values by nano-indentation of various bond
coats near the oxide-bond coat interface, with increasing trend in both hardness values
and standard deviations……………………………………………………………..147
Fig. 5.15. The mole fraction of different phases existing near the TGO/bond coat
interface of the bond coats in a) CMSX-4 LT Pt-Al 100hrs, and b) TMS-82+ LT Pt-Al
50hrs specimens as a function of varying Cr and Ta contents, respectively. These
elements generally help stabilizing phase………………………………………...149
Fig. 5.16. A qualitative illustration of the TGO stresses evolution and interfacial
degradation during isothermal heating and subsequent cooling of the 30hrs
isothermally exposed SRR99LT Pt-Al and the 100hrs isothermally exposed TMS-82+
HT Pt-Al. The SRR99 had more rumpling due to ’ formation at the grain boundary,
while the TMS-82+ had a relatively higher residual stress in the TGO due to
martensitic transformation…………………………..................................................160
Page 13
13
List of Tables
Table 4.1
Nominal composition of the three superalloys used in this work
(wt%)…………………………………………………………………………………84
Table 4.2
Visual representation of the five TBC systems used in this
work……………………………..................................................................................84
Table 4.3
The nominal composition in wt% of the bond coat of the SRR99 specimen as
specified in the calculation of Thermo-Calc relative to its original bulk
composition…………………………………………………………………………104
Table 4.4
The effective in-plane Young’s modulus of the bi-layer coating (YSZ-TGO) after
different isothermal exposures (These values were estimated from Fig. 6 in
Ref.[106])………………...........................................................................................107
Table 4.5
The coating thicknesses (YSZ+TGO thickness) values used for the calculation of Gc
in this work………………………………………………………………………….107
Table 4.6.
A comparison between the Gc reported in Ref. [82] and the TMS-82+ Pt-diffusion
here by approximating u as half-diagonal of the indent mark………………………108
Table 4.7
The comparison between the rankings of performance and contribution factors studied,
showing an exact match between performance and the ’ formation factor………...115
Page 14
14
Table 5.1
Nominal compositions of the three superalloys used in this work (wt%)…………..131
Table 5.2
Visual representation of the five TBC systems used in this work…………………..131
Table 5.3
A comparison between the estimated and actual (EDX scanned) bond coat
compositions in as-coated CMSX-4 specimen……………………………………...134
Table 5.4
The nominal composition in wt% of the CMSX-4 LT Pt-Al and TMS-82+ LT Pt-Al
bond coats as specified in the Thermo-Calc calculation……………………………148
Table 5.5
A summary of the phase evolution, hardness, and TGO stress of all the bond coats
studied between 30 and 100hrs of isothermal exposure…………………………….151
Page 15
15
Abstract
The operating efficiency and reliability of modern jet engines have undergone
significant improvement largely owing to the advances of the materials science over
the past 60 years. The use of both superalloys and TBCs in engine components such
as turbine blades has made it possible for jet engines to operate at higher temperatures,
allowing an optimal balance of fuel economy and thrust power.
Despite the vast improvement in high temperature capability of superalloys, the
utilization of TBCs has brought the concern of coating adhesion during their usage.
TBCs are prone to spallation failure due to interfacial rumpling, which is driven
primarily by thermal coefficient mismatch of the multi-layered structure. Although
interfacial degradation of TBCs has been widely studied by detailed numerical and
analytical models, the predicted results (i.e. stress state and rumpling amplitude) often
deviate from that obtained by experiments. This is largely due to the lack of
consideration of the influence of bond coat and substrate chemistry on the interfacial
evolution of TBC systems. It is only in recent year that more and more study has been
focused on studying the role of chemistry on the interfacial degradation of TBCs.
The purpose of this PhD project is to clarify how the bond coat and substrate chemical
compositions dictate the mechanisms of interfacial degradation, leading to the final
spallation. A cross-sectional indentation technique was utilized to quantitatively
characterize the adhesion of oxide-bond coat interface among 5 systematically
prepared TBC systems. The adhesion of isothermally exposed oxide-bond coat
interface was then correlated with different microstructure parameters, in an attempt
to identify the key parameters controlling the TBC spallation lifetime. EBSD and
EPMA analyses were conducted on the bond coat near the oxide-bond coat interface,
in order to understand the relationship between the key parameters and specific
alloying elements. The results clearly demonstrated that the phase transformation of
bond coat near the oxide-bond coat interface plays the dominant role in the
degradation of interfacial adhesion. Particularly, the co-existence of ’ and martensitic
phases, each having very different thermomechanical response under thermal
exposure, can generate a misfit stress in the TGO layer, and ultimately causes early
TBC spallation. In addition, the phase transformation behavior has been closely
associated with the inherent chemistry of the bond coat and substrate.
Page 16
16
Declaration
No portion of the work referred to in this thesis has been submitted in support of an
application for another degree or qualification of this or any other university or institute
of learning.
Page 17
17
Copyright Statement
(1) The author of this thesis (including any appendices and/or schedules to this thesis)
owns certain copyright or related rights in it (the “Copyright”) and he has given The
University of Manchester certain rights to use such Copyright, including for
administrative purposes.
(2) Copies of this thesis, either in full or in extracts and whether in hard or electronic copy,
may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as
amended) and regulations issued under it or, where appropriate, in accordance with
licensing agreements which the University has from time to time. This page must form
part of any such copies made.
(3) The ownership of certain Copyright, patents, designs, trademarks and other
intellectual property (the “Intellectual Property”) and any reproductions of copyright
works in the thesis, for example graphs and tables (“Reproductions”), which may be
described in this thesis, may not be owned by the author and may be owned by third
parties. Such Intellectual Property and Reproductions cannot and must not be made
available for use without the prior written permission of the owner(s) of the relevant
Intellectual Property and/or Reproductions.
(4) Further information on the conditions under which disclosure, publication and
commercialization of this thesis, the Copyright and any Intellectual Property and/or
Reproductions described in it may take place is available in the University IP Policy (see
http://documents.manchester.ac.uk/DocuInfo.aspx?DocID=487), in any relevant Thesis
restriction declarations deposited in the University Library, The University Library’s
regulations (see http://www.manchester.ac.uk/library/aboutus/regulations) and in The
University’s policy on Presentation of Theses.
Page 18
18
Acknowledgement
First and foremost, I would like to express my sincere gratitude to my doctoral
supervisor, Prof. Ping Xiao, for his immeasurable amount of guidance and support
throughout the course of my PhD study. Without him, this PhD project could not have
been possible in the first place.
I would also like to extend this gratitude to my brother, Dr. Rudder Wu, for his
invaluable contribution to this PhD project. For all these years, he has been my
mentor in academic work and many aspects of life. In addition, I would like to thank
Dr. Xiaofeng Zhao, now working at Shanghai Jiao Tong University, for guiding a
significant portion of this research.
My appreciation also goes to Ms. Hong Gao and Dr. Tadaharu Yokokawa of the
National Institute for Materials Science (NIMS, Japan) for facilitating the analysis of
EBSD and EPMA, respectively. They offered great support on the technical aspects of
the experimental work. I would like to thank the technical staffs in Materials Science
Centre (University of Manchester, UK), especially Mr. Kenneth Gyves and Mr. Andrij
Zadoroshnyj for offering technical support on the cross-sectional indentation work
and raman spectroscopy, respectively. My gratitude extends to Dr. Hiroshi Harada and
Dr. Kyoko Kawagishi of NIMS, Japan, and Mr. Rodney Wing of Chromalloy - UK for
the provision of research facilities and specimens.
Special thanks to my colleagues and friends from the faculty, including Mingwen Bai,
Ying Chen, Kuan-I Lee, Justyna Kulczyk-Malecka, Prince Wang, Geng Xing, Fan
Yang, and Erica Yu. If it wasn’t for their presence, my PhD life would have been a lot
more difficult.
I am especially grateful to my parents for their endless love and support. I thank them
for raising me up and sharing their life philosophy with me.
Page 19
19
Vita
Mr. Liberty Wu was born in Taipei (Taiwan), and spent most of his childhood time there.
At the age of 11, his parents made an important decision to immigrate to Canada. He had
been living in Vancouver (Canada) ever since until receiving a bachelor’s degree in
applied science from the University of British Columbia in 2009. Afterwards, he spent
one year in Tsukuba (Japan), working as a research assistant at the National Institute for
Materials Science (NIMS, Japan). In 2010, he moved onto pursuing a PhD degree in
Materials Science and Engineering at the University of Manchester (UK). He has been
working on the research of thermal barrier coatings under the supervision of Prof. Ping
Xiao. As part of the academic curriculum, he also spent 3 months every year at NIMS
(Japan), carrying out some of his research activities.
-by Prince Wang
Page 20
20
CHAPTER 1
Literature review
1.1 Background information
Ni-based superalloys have been the material choice for the turbine blade in the hot
section of jet engines and land based power generator due to the superior mechanical
properties at high temperatures, particularly the ability to resist creep and fatigue.
Since the invention of jet engine, there has been ongoing effort in raising the turbine
inlet gas temperature, thus, enhancing the fuel efficiency. Inevitably, the alloy
materials used for turbine blades experienced numerous thermal breakdown issues as
the hot gas stream exiting the combustor exceeded the melting temperature of the
metal parts. To address this issue, scientists from NASA first proposed the use of
ceramic coatings on turbine blades in the early 1940, and began adopting this
technology to new prototype planes and rocket applications between 60s and the early
70s [1-3]. From the 1970s to the early 80s, the newly improved plasma sprayed and
EB-PVD thermal barrier coating systems (TBCs) found their utilization in civil jet
engines [4,5]. The TBCs system consists of a yttria-stablized zirconia (YSZ) ceramic
top coating layer deposited onto an underlying substrate, joined with an intermediate
bond coat material, which serves as a reservoir for the preferential formation of an
oxidation resistant alumina thermally grown oxide (TGO) layer (see Fig. 1.1).
Page 21
21
Fig. 1.1. A schematic diagram showing the multi-layer structure of thermal barrier
coating systems (TBCs) [6].
The top coat, having a tailored composition of approximately 7 wt% yttria, was
determined to maintain the most stabilized tetragonal phase during cyclic thermal
loading, leading to desirable mechanical properties [7,8]. This optimal phase was
reported to have significantly improved TBC lifetime by scientists from NASA in the
70s (see Fig. 1.2). The surface temperature of underlying superalloy of turbines could
decrease by as much as 167°C by using the YSZ coat; thereby, increase the blade
lifetime more than threefold [9].
Page 22
22
Fig. 1.2. A plot showing the improvement in thermal cyclic lifetime of thermal barrier
coating systems (TBCs) with 7 wt% yttria in ZrO2 [7].
1.1.1 Two types of top coat
The top coat, depending on the types of deposition technique applied, can have
different structural appearances. Usually, the coating layer is deposited using either air
plasma spray method (APS) or electron beam physical vapour deposition (EB-PVD).
The APS coating has a splat-like structure with inter-lamella gaps in-plane with the
substrate; while EB-PVD type gives a characteristic columnar structure with
thru-thickness inter-column gaps (Fig. 1.3). For the APS coating, the lateral strain due
to thermal mismatch between the different layers of coating is accommodated by the
porous structure between the successive layers of splats. The EB-PVD coating, on the
hand, accommodates this lateral strain by its columnar structures with gaps in
between. After exposing to high temperature oxidation, the topcoat-scale interface is
much rougher in the case of APS coating than that of the EB-PVD. The columnar
structure of the EB-PVD coating allows much higher strain compliance between the
different layers; hence, giving a much smoother and adhesive topcoat-scale interface
[9]. Nonetheless, the EB-PVD is known to be more costly while being restricted by its
Page 23
23
line-of-sight deposition characteristics.
Fig. 1.3. A comparison of the coating morphology between APS (left) [6] and
EB-PVD (right), showing splat-like and columnar structures, respectively.
1.1.2 Two types of bond coat
Similar to the top coat deposition methods, bond coats can also be deposited using
two common techniques: overlay coating or Pt/Pt-Al diffusion bond coats. The former
type generally consists of thermally sprayed metallic alloy in the form of MCrAlX,
where M is usually Ni or a combination of Ni and Co, and the X indicates the minor
element addition, such as silicon, zirconium, hafnium or yttrium. As a result of the
overlaid structure, desirable surface properties could be obtained for specific
applications in coatings of this type. Nonetheless, due to the independent nature of
coating composition, severe inter-diffusion of chemical activities is inevitable
between the bond coat and substrate during thermal cycling (Fig. 1.4).
Unlike the overlay bond coats, the diffusion type utilizes electrodeposition method,
where the bond coat materials are diffused into the superalloy substrate rather than
Page 24
24
forming a discrete layer. These coating materials have tailored compositions to ensure
a stable formation of alumina scale upon oxidation (Fig. 1.4). For some diffusion
coatings, electrodeposition of thin layers of Pt is done, following a vapour-phase
aluminization. Where as in the so-called “low cost” bond coats patented by
Rolls-Royce and Chromalloy UK [10], the subsequent aluminization is omitted in the
process. It is demonstrated that Pt diffusion process alone could enhance the outward
migration of aluminum, as long as the Pt content is maintained above 15 at % [11].
The major drawback of diffusion coating is that a thicker coating is difficult to
achieve in comparison with the plasma spraying technique of the MCrAlY type [12]
Fig. 1.4. A comparison of the microstructure between overlay (left) [13] and Pt-Al
diffusion bond coats (right). The diffusion zone is several times thicker in the overlay
bond coat.
1.1.3 Oxidation and diffusion behavior in Pt/Pt-Al diffusion bond coats
Due to the porous columnar structure and high ionic diffusivity of YSZ in the ceramic
top coat at high temperature, the ingress of oxygen can readily oxidize the underlying
substrate, leading to catastrophic failure of the turbine structure [14-16]. The
Page 25
25
oxidation is controlled by the rate of the metal ions migration through the bond coat
or the oxygen diffusion into the sample. The preferred formation of oxide in TBCs is
an adherent and slow growing layer characterized by the parabolic growth, without
internal oxidation [17,18].
During the initial stage of oxidation, a transient form of alumina, θ-Al2O3 is formed
on the surface of Pt/Pt-Al diffusion bond coats. This alumina specie is known to
induce a large compressive stress on the unconstrained sample. Soon, the desirable
-Al2O3 began to transform from the original θ phase, which imposes a tensile stress
against the θ phase. With continuous alumina growth, the overall stress eventually
becomes a steady state of low compression [19].
The interdiffusion of bond coat elements during oxidation will cause a significant
change to the microstructure of bond coat. Interdiffusion zone (IDZ) rich in Ni and
various refractory elements, and voids will appear in the bond coat, which causes a
decrease in load bearing strength of the TBCs [20]. The void formations are thought
to have occurred due to the interdiffusion of Al and outward migration of Ni from
substrate to bond coat. The diffusivities of these two elements differ significantly, and
hence, leading to a flux region of vacancy upon interdiffusion. The vacancy sites then
coalesce to form voids, which is known as Kirkendall effect [21,22].
The primary failure mechanism of turbine components lies in the in-service TBC
spallation due to the coefficient of thermal expansion (CTE) mismatch between the
ceramic top coat of yttria-stabilized zirconia (YSZ) and the bond coat near the
thermally grown oxide (TGO) interface [23,24]. While serving as a reservoir of Al to
Page 26
26
promote the preferential formation ofα-Al2O3 TGO protective layer, bond coat will
inevitably react with superalloy substrate through outward chemical diffusion from
the substrate at high temperature, which eventually affects the TGO formation
dynamics, microstructure, and interfacial stability [25]. Traditionally, it was thought
that faster TGO growth rate and higher degree of rumpling at the TBC/bond coat
interface imply a shorter spallation lifetime. However, in more recent studies, certain
bond coat/substrate systems showed a relatively longer lifetime despite having either
a thicker TGO and or more rumpling than other TBCs [26]. This phenomenon raises
the incentive to investigate the resisting mechanisms to spallation near the TGO
interface by focusing on the chemical effect of the bond coat and substrate.
Currently, there have been very few systematic attempts at studying the effect of bond
coat and substrate chemistry on TGO spallation behavior. In order to better
understand the interfacial evolution during oxidation, there needs to be a reliable
method of quantifying the interfacial toughness of TBCs subjected to oxidation prior
to complete spallation of the coating. Chapter 2 will briefly review some of the most
commonly used measurement techniques of coating-substrate adhesion, with a
particular focus on the practical aspects of each technique.
1.1.4 TGO residual stress
The evolution of residual stress in the TGO layer of the TBC system has been widely
studied in recent years. One particular method known as photo-luminescence
piezo-spectroscopy (PLPS) has been used to measure the residual stress in the TGO.
Christensen et al [27,28] showed that the stress could be obtained from the
Page 27
27
stress-induced frequency shift of Cr3+
impurities in solid solution of TGO. This
technique excited much interest due to its non-destructive nature.
Stress is measured by the peak shift relative to an unstrained sapphire reference by
assuming a state of equi-biaxial plane stress, and a random distribution of the
crystallographic texture in the TGO. Based on the relationship obtained by He and
Clarke [29], the frequency shift can be expressed in terms of biaxial stress by the
following equation (Eq. 1-1). Ar laser is used to excite the Cr3+
luminescence spectra,
and their frequency shifts were monitored against cyclic hours (See Fig. 1.5). From
the results of the work by A. Selcuk and A. Atkinson [28], a general trend across
different specimens can be observed, in which the mean compressive stress rises over
the first two thirds of life due to gradual stiffening of the non-planar TGO, and later
decrease toward the end of life by the relaxation of increasingly numerous local
damage events.
∆=5.07 (cm-1GPa-1)(Eq. 1-1)
Page 28
28
Fig. 1.5. The evolution of mean frequency shift measured for specimens that failed
after: (a) 190 (b) 225 (c) 207 (d) 237 cycles [28].
Low stress mapping indicates isolated stress regions at first, then they gradually
become more numerous and tend to coalesce into larger regions of damage near the
end of life (Fig. 1.6). Early works of PLPS mapping analysis indicate that the
luminescence line-width is much wider than expected, and cannot be explained by
surface roughness alone. Therefore, a constrained deconvolution approach was
developed by several researchers to split up the spectra into two stress contribution; a
high stress range (2.0-4.5GPa) and a low stress range (1- 1.5GPa) [30].
Page 29
29
Fig. 1.6. Low stress evolution map showing gradual coalescence of isolated stresses
[28].
1.1.5 Stress evolution and failure mechanism near TGO interface
In the PLPS works of Tolpygo and Clarke [31], specimens with smoother surface
(polished) were found to have higher TGO stress (shift) that either maintained
throughout thermal-cycling, or showed no obvious decline compared to those with
rougher bond coat surface. Selcuk et al [30] also indicated that the compressive stress
in a planar TGO due to thermal expansion difference between the alumina and
CMSX-4 substrate is significantly higher than that of a non-planar TGO morphology.
Page 30
30
Despite the different initial surface morphology, all specimens were found to become
more or less rumpled after thermal-cycling.
Various studies have reported that the TGO stress can fall below expected values by
reasons such as: non-planar TGO morphology; plastic deformation (yield and creep in
the bond coat) during cooling; or damage in the TGO (fracture or detachment from
the bond coat). The failure mechanism of typical Pt-Al bond coats has been widely
studied and various results indicate that the plastic deformation in the bond coat plays
an important role in the type of cracking/stress around the TBC interface [32].
The final fracture path in TBC is believed to be closely related to the surface
roughness of bond coat. Experimental results indicate that on specimens with rough
coating interface, rumpling induced damage near the YSZ interface took the dominant
role in the final spallation. Those with smoother coating interface displayed more
damage near the bond coat/TGO interface. In the latter case, the stress accumulated in
the growing TGO thickness would have contributed as the driving energy for
spallation [33]. A recent study suggests that the failure mechanism is strongly
dependent on the differences in high temperature mechanical properties of the bond
coats [34]. In their experiment, TBCs with Pt-Al bond coats underwent phase
transformation from the as-deposited single phase structure into a two phase (Pt,
Ni)Al and '-(Pt, Ni)3Al during the course of Al depletion in high temperature
oxidation.
It has been proposed that the non-uniform volume change associated with this
transformation could lead to rumpling. However, in the recent study by R.T. Wu et al
Page 31
31
[35], oxidation result disproves this transformation phenomenon as a major role in
inducing bond coat rumpling. According to their studies, the specimen with LT-Pt-Al
bond coat exhibited the most rumpling, yet its oxidation rate was determined to be the
slowest among the three systems investigated. While on their HT Pt-Al bond coat
specimen with relatively lower Al content, results indicate that its rate of rumpling
was considerably lower than that of the LT Pt-Al system. These findings showed the
opposite of what was expected from phase transformation induced rumpling.
Clearly, the phase transformation accompanied with Al depletion cannot be the sole
mechanism for inducing rumpling. Several studies have pointed to the possible
control of rumpling by the resistance to high temperature plastic deformation of bond
coat structures. In the thermal cycling studies by Wu et al [35], the Pt-diffusion bond
coat, which contained a two phase and 'microstructure similar to that of the base
superalloy, had the least noticeable change in the interfacial waviness against thermal
cycling time (Fig. 1.7). Whereas in the Pt-Al bond coat systems, which underwent a
phase transformation from the as-deposited single phase structure into a two phase
(Pt, Ni)Al and '-(Pt, Ni)3Al, had a very drastic increase in roughness with thermal
cycling [34].
Page 32
32
Fig. 1.7. TGO/bond coat interfacial roughness profile of three bond coats: a) LT Pt-Al
b) HT Pt-Al c) Pt diffusion against thermal cycles [35]. The Pt-diffusion bond coat
maintained similar roughness throughout the thermal cycles after 10 cycles.
A previous study [35] also suggested that a finely dispersed 'within the matrix can
Page 33
33
contribute to a high temperature strengthening effect in the Pt-Al systems. The extent
of the 'precipitation is not only dependent on the rate at which Al depletion takes
place by oxidation or interdiffusion, but most importantly how close the composition
is to the /'phase boundary [36]. The HT Pt-Al bond coat, having slightly lower
Pt and Al content near its TGO/bond coat interface after thermal exposure, is
relatively closer to the /'phase boundary than the LT Pt-Al systems (Fig. 1.8). As
a result of a higher volume fraction of precipitated ', HT Pt-Al should have better
creep resistance than that of the LT Pt-Al bond coats at elevated temperature. This
hypothesis is consistent with the experimental results by R.T. Wu et al, in which the
leaner Al containing (HT PT-Al) bond coat exhibited a significantly slower rumpling
rate than the Al-rich LT Pt-Al systems [35]. The authors suggested that the role of
high-temperature mechanical properties in different bond coats, in particular high
temperature plasticity, requires further clarification in order to better understand their
rumpling behavior.
Fig. 1.8. Ternary Ni-Al-Pt phase diagram for three different bond coats: a) LT Pt-Al b)
HT Pt-Al c) Pt- diffusion at 1100°C and 1150°C showing a shift from to ’phase
Page 34
34
with decreasing Pt and Al contents [36].
1.1.6 The rumpling behavior of Pt modified bond coats
A series of papers by Tolpygo and Clarke was devoted to the understanding of
rumpling mechanisms in Pt modified bond coats [37-39]. Tolpygo et al [39]
explained the dependence of rumpling on the cycle length using a semi-quantitative
plot as shown in Fig. 1.9. Plastic yielding can occur during both cooling and heating
when the bond coat creep strength is low at elevated temperatures (i.e. segments 2-3
an 6-7). Due to an intrinsic asymmetry in the plastic response of the bond coat to
tension and compression, the bond coat is in a state of residual compression at the end
of heating cycle (See point 7). The amount of creep relaxation, therefore, undulation
of the bond coat, will then be dependent on the duration of isothermal dwelling.
Further, Tolpygo et al [39] calculated the creep strain (i.e. surface elongation) of the
bond coat due to creep relaxation of the CTE mismatch from previous thermal cycling
at dwelling temperature, and compared that with the actual surface roughness of the
bond coat observed from experiments. The result suggests that CTE mismatch alone
cannot generate the amount of surface undulation observed. Hence, there must be
other mechanisms of inducing surface undulation, which presumably take place at
elevated temperature ranges of thermal cycling (i.e. isothermal dwelling). This will be
the focus of discussions on rumpling mechanisms in later chapters, as the study only
involves isothermal exposures of TBCs.
Page 35
35
Fig. 1.9. The strain response in the bond coat throughout a thermal cycle. Point 7
denotes the onset of isothermal dwelling. The onset of cooling may start at either
point 1 (stress relaxation) or 1’ (residual stress) depending on the duration of
isothermal exposure [39].
Several rumpling inducing mechanisms during high temperature dwelling were
proposed by Tolpygo et al [40]. The stress associated with the lateral growth strain of
oxide, which is constrained by the underlying metal substrate, can lead to in-plane
compression of the oxide. By taking into account of all the strains acting on the oxide
layer (i.e. metal-oxide bi-layer interaction) throughout a complete thermal cycle, the
growth strain of the oxide was found to cause rumpling of the underlying bond coat
due to creep relaxation of the strain [40]. The magnitude of the growth strain is
determined by the diffusion rate of the oxide forming ions within the oxide scale [40].
Tolpygo et al [38] also suggested that the bond coat tends to swell during high
Page 36
36
temperature oxidation, as a result of the intrinsic difference in diffusivities between Al
and Ni (i.e. Kirkendall effect). They found that the wavelength of the undulation
stayed similar throughout thermal cycling, while amplitude continued to grow. This
wavelength is surprisingly similar to the grain size of the bond coat near the surface,
suggesting that rumpling may be induced by an uneven response to swelling between
inner parts of the grain and the regions close to the grain boundary.
Further, the volumetric changes associated with to martensitic and 'phase
transformation during thermal cycling were also investigated for possible role in
inducing rumpling. The 'phase transformation was not a requirement for inducing
rumpling, as bond coat had already form undulation in early thermal cyclic times prior
to any 'phase transformation. However, the authors suggest that 'phase
transformation can promote a further growth in rumpling. For the verification of
martensitic transformation, samples were subjected to thermal cycling in a two-zone
furnace at 1150°C and roughly 750-800°C with holding times of 1hr and 10 minutes,
respectively, for a total of 100 cycles. Since the onset temperature of martensitic
transformation for their specimens was well below 750°C, martensitic transformation
would not occur across those temperature ranges during thermal cycling except for the
final cooling to room temperature. The two-zone thermally exposed samples were
then compared to those that had undergone regular 1hr thermal cycling between
1150°C and room temperature for 100 cycles. The results suggested that both set of
samples achieved the same rumpling magnitude, hence, suggesting that martensitic
transformation has little or no effect on the rumpling accumulation during thermal
cycling [39]. This is, however, contradictory to the modeling work by Glynn et al
described below [41].
Page 37
37
It has been shown from previous research that with modest quenching from
temperatures above 1000°C, a Ni-rich (61 to 68 at. %) NiAl bond coat can transform
into a tetragonally distorted L10 martensite, instead of the ordered -(Ni,Pt)Al phase
at room temperature. [42,43]. Numerous modeling work by Glynn et al [41,44] was
devoted in understanding the general stress response of the TBC system due to the
phase transformation in the bond coat. Figure 1.10 below depicts the difference in
stress response of the bond coat during cooling and heating with/without the
occurrence of martensitic phase transformation. Fig 1.10a (the case with
transformation) indicated a sharp increase in 11, 33 (in-plane bond coat stress
directly below an undulation) and 22 (normal stress in the bond coat directly below
the undulation) due to the transformation from to martensite during cooling at
around 500°C.
Page 38
38
Fig. 1.10. The stress computed beneath an undulation in the bond coat s a) during
cooling and b) heating with phase transformation; c) during cooling and d) heating
without phase transformation [41]. Bond coats with phase transformation experience a
sharp rise/decline in the plane stresses during cooling and heating.
The effect of the martensitic transformation on the overall TBC stress is clearly
demonstrated in Figure 1.10, as the stress level can be more than two folds at times
when martensite transformation took place.
Page 39
39
Glynn et al suggested that overall TBC stress behavior is closely related to the
low-temperature bond coat yielding strength. The stress state in the bond coat directly
below the undulation had been reported to be confined such that y = abs(11-22),
and was plotted against temperature during a typical cyclic loading for four different
yield scenarios (see Figure 1.11). It can be demonstrated from the plot that all bond
coats except the one with the highest yield strength (2000MPa) underwent yielding
during the to martensite transformation upon cooling.
Fig. 1.11. The difference between 11-22 computed for different bond coat yield
strengths [41]. In general, bond coats with lower yield strengths experience lower
plane stresses during phase transformation due to creep relaxation.
Page 40
40
Among all the scenarios studied, only the 500MPa case underwent a substantial
amount of yielding (see flat part of the curve) upon heating during the reverse
transformation from martensite to at around 620°C. Due to this plastic deformation
of the bond coat, both the 22 tensile stress of bond coat and top coat generated during
previous cooling cycle was much relieved. This was however, not the case for the
1000MPa, as the high tensile stress generated during cooling cannot be relaxed due to
the absence of yielding upon heating.
The importance of phase transformation temperature on the overall TBC stress was
also investigated by Glynn et al using three different scenarios: 600°C, 800°C, and
1000°C by assuming that the to martensite and the reverse transformation to be the
same temperature. As expected, when the transformation took place at around 600°C,
much of the strain was accommodated elastically due to the fact that creep was
minimal at this temperature range; thus, receiving the highest accumulation in bond
coat stress and having the least amount of rumpling (Fig. 1.12 and 1.13).
Page 41
41
Fig. 1.12. a) The difference between 11-22 computed for different transformation
temperature b) top coat 22 stress computed for different transformation temperature
[40]. Bond coat plane stresses can be relaxed when the transformation temperature is
raised.
It can summed up from the above analysis that a desirable bond coat would have two
attributes: 1) a high/strong bond coat yield strength and 2) a martensite
transformation temperature of no more than 500°C, so that deformation occurs below
the creep range. Having these two attributes could largely alleviate the rumpling
Page 42
42
growth, consequently, prevent early spallation of TBCs (see Figure 1.13).
Fig. 1.13. The average amplitude growth of bond coat surface rumpling as a function
of martensitic transformation temperature showing an increasing rumpling trend with
higher transformation temperature. Note that the Y-axis indicates the % of total
growth that occurred during the transformation [45].
Page 43
43
CHAPTER 2
Interfacial fracture toughness
2.1 Background information
For functional coating materials, the structural integrity of the coating layer is of
primary concern during operation. Even though considerable efforts have been
dedicated towards improving the reliability of TBCs, the failure mechanisms are not
fully understood. This is in part, due to the complex nature of the coating structure,
which is further complicated by a failure process that is highly sensitive to processing
technique and pattern of use. During in-service thermal and mechanical loading, a
series of events that include crack nucleation, propagation and coalescence will occur
in most cases between either the top coat and TGO, or the bond coat and TGO, and
sometimes be accompanied by a cohesive failure within one of the layers. The
interfacial stability is thus an overall balance between the crack driving force and the
resistance to crack propagation along the relevant interfaces [26,46,47,48].
It has been proposed that a universal damage parameter on the basis of (apparent)
interfacial fracture toughness is necessary to model the damage accumulation as a
function of loading history. The stress field and mechanical response for an interfacial
crack have been characterized with mathematical solutions. Usually, the interfacial
fracture toughness is being evaluated based on one of the following parameters:
adhesive load, force, stress, intensity factor, or energy release rate [49-51].
Consequently, more than a dozen different testing methods have been developed for
estimating the interfacial fracture toughness of TBC-systems (Demarecaux [52],
Page 44
44
Vasinota [53], Hofinger [54], Kagawa [55], etc). In these testing techniques, a
controlled delamination of the interfaces is being induced through mechanical loading.
Each approach is more or less bounded to limitations, both experimentally and
material wise.
2.2 Overview of the methods for measuring interfacial fracture
toughness
To critically examine the practicality of various measurement techniques, it is
worthwhile considering the necessary quality for an ideal adhesion test method. As
Rickerby [48] pointed out, the requirement for an ideal test method should bear the
following: (i) non-destructive, (ii) easily adaptable to routine testing of geometrically
complex shapes, (iii) relatively simple to perform and interpret, (iv) amenable to
standardization and automation, (v) reproducible and, if possible, quantitative and (vi)
directly related to coating reliability in specific applications. Back in the time when
this review paper was being written, there were very few if any practical methods that
could fulfill a number of these criteria. Through years of effort, a number of test
approaches have been developed to satisfy some of these requirements, though many
challenges and limitations still exist in using these techniques. Some of the most
common approaches include: (1) traditional methods that rely on direct application of
force to induce coating delamination (2) four point-bending tests of notched
multilayer beams (3) indentation techniques (4) barb test. The main issue with the
current techniques lies in the limitation of loading angle that can be used for testing.
Many structural materials, such as TBC, undergo complicated mixed phase loading
angle at the interface instead of a fixed or pure modes (i.e. mode I opening and mode
II in-plane shear) that are associated with the individual measuring techniques (see
Page 45
45
Figure 2.1). For instance, the barb and pull-off tests shown in Figure 2.2 are limited to
probing coatings under pure mode II and I loading, respectively. [56,57]. In addition,
other challenges and limitations such as difficulty in sample preparations and
experimental setup also hamper the effectiveness of evaluating the interfacial
mechanical properties.
With the above considerations in mind, the next sections will closely examine the
advantages and disadvantages of some of most commonly used approaches in the
field of interfacial toughness measurement. It is intended, through this review, that
one would become familiar with the applicability of the available techniques, and
develop the key knowledge necessary in adopting them to specific material systems.
Fig. 2.1. An illustration showing the three pure modes of loading.
Page 46
46
Fig. 2.2. Figures showing the loading angle in barb (left) and pull-off (right) [48,55].
2.2.1 Conventional methods: Pull-off, double cantilever beam, and scratch
adhesion test, etc.
About three decades ago, researchers began to seek and develop methods, which
could be used to measure the interfacial adhesion of modern coating-substrate systems.
Some preliminary methods for determining adhesion have thus been proposed. All
these methods are based on mechanical approach of loading as seen in practical
applications of the pull-off, double cantilever beam, and scratch adhesion test
methods, etc. These tests, though destructive in nature, have set the trend for the
development of interfacial adhesion measurement techniques in the last three decades
[48].
Page 47
47
Pull-off methods. The term is a generalization of several methods that utilize the
application of a force normal to the coating-substrate interface to quantitatively
determine the practical adhesion. The tape test, being one of the popular methods,
applies a pressure sensitive tape to the coating surface to pull the coating off and
thereafter determines the peel force per unit tape width. Although the test is easy to
carry out, it is only limited to thick and rather weakly adhering coatings due to the
fact that the strength of the bonding tape is limited to about 20MN m-2
. An alternative
to the tape test would be to glue two rods to the coating and substrate ends from
which a pure tensile force can be imposed on the coating to detach it from the
substrate (Fig. 2.3). This method is capable of measuring materials with higher
interfacial fracture toughness; however, alignment must be perfect to ensure uniform
loading across. In practice, pure tensile loading condition is hard to achieve, and
instead, these tests often involve a complex mixture of tensile and shear forces, which
renders the interpretation of the results difficult. Further, there is the possibility of
adhesive glue diffusion through thinner coatings during the test, affecting the
film-substrate interfaces. [58]
Fig. 2.3. Illustration of the tensile based direct pull-off test [59].
Page 48
48
Double cantilever beam test. Two rectangular arms, one of which is spray coated with the
specimen of interest, are joined together by adhesive epoxy and subjected to
cantilever type loading at the drilled pins using an actuator controlled displacement
(Fig. 2.4).The loading mechanism in the test involves almost pure mode I loading to
the coating bonded between two rigid arms. This unique setup requires sawing a
pre-crack with a crack tip to reaching a depth in excess of the penetration layer of the
coating by adhesive epoxy. Post-test examination of the crack surface is necessary to
confirm whether the specimen failed by adhesive or cohesive, in order to verify if the
experiment is successful or not [60]. This setup works quite well for thick coatings;
however, is not suitable when applied to thin coatings due to the difficulty in
generating the pre-crack and the bonding of the rigid plates may damage the
interfaces to be tested.
Fig. 2.4. Illustration of the sample and set-up used in the double-cantilever beam test.
The energy release rate can be approximated as follow
Page 49
49
(Eq. 2-1)
where E and t are the Young’s Modulus and thickness of the glass beam; respectively,
and δ is the crack opening, L is the crack length [61].
Scratch adhesion test methods. The scratch adhesion test involves drawing a stylus over the
coating surface with a stepwise or continuously increasing normal force until the
coating delaminates. A critical load, Lc, is defined as the load at which the coating is
removed in a regular way along the whole channel length. The testing result could be
influenced by intrinsic factors (i.e. loading rate, scratching speed, indenter tip radius,
instrument-specific designs, etc.), as well as extrinsic parameters such as substrate and
coating modulus, friction coefficient, surface roughness, etc. As such, those factors,
along with the residual stress in the coating must be accounted when applying the
model used for calculating interfacial fracture toughness. The critical load and the
resulting toughness value could only be used to compare different materials given that
the same failure mode occurs. Buckling failure mode is commonly observed in thin
coatings subjected to the scratch test, whereas in thick coatings, a through thickness
crack is more likely to occur leading to a wedge spallation failure (Fig 2.5,2.6). In the
case of buckling failure, it is difficult to make a reliable comparison between coatings
on different substrates, since significant amount of plastic deformation occurred in the
substrate depending on the material hardness. It is only the case of wedge spallation
that such comparisons between different coatings are meaningful. In general, scratch
test can provide a qualitative assessment of hard coatings on softer substrates;
however, for thin and soft coatings, significant amount of plastic deformation may be
associated with the delamination, making it difficult for an adequate comparison
Page 50
50
between different materials [62-67].
Fig. 2.5. a) Pile-up in front of scratching indenter, and b) buckling failure at the
interface of thin coatings [67].
Fig. 2.6. a) the scratching indenter generates a wedge crack some distance ahead; b)
the coating lifted up the wedge crack to cause an interfacial crack c) a through
thickness crack occurs right in front of the indenter leading to complete spallation
[67].
2.2.2 Indentation
The majority of the work related to the adhesion assessment by indentation techniques
is based on the notion that the crack initiates at and propagates along the interface
using an indenter loaded from the top. Despite this assumption, there are various types
of models, each of which is associated with a different failure mechanism. Thus, care
Page 51
51
must be taken to identify the delamination mechanism before choosing an appropriate
model to apply. In general, two types of delamination mechanism can occur during
indentation for soft coatings on hard substrates: (a) compressive stress-induced
delamination during loading, (b) tensile stress triggered delamination during
unloading (Fig. 2.7, 2.8). For specimens with relatively weak interfaces compared to
the coating and substrate, failure tends to occur during loading. In the case where the
interface is relatively stronger, or if the whole coating system tends to dissipate the
induced stress in the form of plastic deformation, then interfacial failure usually
occurs as a single buckling during unloading [58].
Fig. 2.7. Indentation-induced delamination during loading [58].
Fig. 2.8. Indentation-induced delamination during unloading [58].
For indentation of a brittle coating on a hard substrate, the delamination mechanism
can become much more complicated than in the case of soft coatings. Considering
that the coating is more likely to fail by buckling instead of shear delamination, a
slightly modified description of the failure mechanism originally proposed by Li et al
[68] was made in Chen’s work [58]. First, through thickness cracks are generated in
Page 52
52
the radial direction close to the indenter tip. With further penetration, the crack
openings expand with buckling delamination of the coating to their sides. Lastly,
secondary radial cracks form further away from the indenter, leading to partial or full
spallation of the buckled coating (see Fig. 2.9).
Fig. 2.9. Figures showing the stages of fracture in a brittle coating with a hard
substrate due to a nano-indenter load [58].
The earliest work on the modeling of interfacial toughness of indentation began with
the He et al’s analysis of the crack formation initiating from and out of two dissimilar
elastic solids. The study focuses on determining the crack propagation behavior
through the interface by comparing the relative energy release rate between cracks
that either passes through or gets deflected by the interface. Their work became the
modeling basis for other’s analytical work [69].
Marshall and his co-workers first developed a model that accounts for the energy
Page 53
53
release rate during delamination by comparing the cases of unbuckled and buckled
coatings. The equation for the strain energy release rate is derived as follows [70]:
(Eq. 2-2)
Where the term α is derived from the slope of the plot of buckling load versus edge
displacement, which is 0.38 for materials with a Poisson’s ration of 1/3. For
non-buckling fracture, α is equivalent to 1, and the residual stress term thus does
not contribute to the energy needed to drive delamination.
(Eq. 2-3)
and the indentation stress is given by
(Eq. 2-4),
where Vi is the indentation volume (which can be estimated from either the
load-displacement curve and indenter geometry or the profile of the indentation
impression) and a is the crack length. The subscript f, i and B denote the film,
indentation, and buckling properties respectively (see Fig. 2.10 below).
Fig. 2.10. Illustration showing the half penny shaped later crack centered under the
Page 54
54
indent, which is dependent on film and substrate properties, residual stress, etc [70].
Hutchingson and Suo made further development to this model as shown by the
modified equations below. [71]
(Eq. 2-5,2-6)
where σr is the residual stress. This modified model assumes that no strain energy
is dissipated through possible cracks found in the substrate.
In a more recent work, Yeap et al modified this model into two separate equations that
are used to describe both non-buckling and buckling cases [72].
(Eq. 2-7 non-buckling)
(Eq. 2-8 buckling)
where
(Eq. 2-9 critical buckling stress)
Page 55
55
in which the term Y is a dimensionless constant, as determined to be 1.488 and 1 for
circular buckles and straight-sided buckling respectively. The term t and r are the
coating thickness and half radius of the circular crack front normal to the wedge
indent respectively (Fig. 2.11).
Fig. 2.11. SEM image showing a delaminated black diamond film with a
straight-sided buckling behaviour [72].
The indentation induced stress is given by
(Eq. 2-10)
, and Vi is the volume of plastic indentation as follow
(Eq. 2-11)
Page 56
56
, while the volume of delaminated material is denoted by the term Vc as follow
(Eq. 2-12)
Ritter J et al developed a new approach for determining the interfacial shear strength
with respect to coating hardness based on contact radius at the onset of delamination
[73].
This method is based on the Matthewson’s stress analysis work of a linearly elastic
coating on a rigid substrate using an axisymmetric rigid indenter [74]. The
experimental apparatus is designed such that a direct in-situ observation of the
debonding process can be recorded along with the measurement of indenter contact
dimension .
Mencik et al estimated the fracture toughness of coating based on the cases of radial
cracking, chipping, and delamination (see Fig. 2.9 before), each having a different
method of calculation. Consequently, they derived the interfacial toughness by
treating the interfacial failure as mode I and cint as the appropriate flaw size at the
interface [75]. From this approach, they obtained an estimated value of 0.18MPa m0.5
for the average interfacial toughness of their hybrid coatings-glass specimens, which
lies within the range of results found by previous work. Despite having a good
accuracy, the difficulty in determining the cint could potentially lead to wide scatter of
results [76,77].
Other miscellaneous indentation techniques
Page 57
57
In addition to improved models as discussed previously, several special indentation
techniques have also been developed. For instance, it has been proposed to use a
wedge indenter to measure adhesion of thin metal coatings instead of the
axisymmetric indenter such as conical or pyramids type, as the wedge indentation is
relatively easier to induce interfacial failure due to its geometry. This special
technique was then applied to a brittle coating on a ductile substrate using a
stress-analysis approach based on the expanding cavity model [78]. In this model,
radial crack formation is not a problem since there is no formation of tensile hoop
stress in the coating. One disadvantage, however, is that the bending effect in the film
and substrate are completely ignored during the indenting cycle, leading to significant
errors for ductile substrates.
To eliminate the effect from plastic deformation, Choulier [79] and Dal Maschio [80]
proposed using Vickers indentation at the cross section of the coating system close to
the interface, while Chicot et al [51] further modified the technique by introducing a
critical load to crack initiation relationship (Fig. 2.12)
Fig. 2.12. Schematic representation of the relationship between crack length and
Page 58
58
diagonal length [79].
Based on this linear loading vs. crack relationship, a model was established by
considering the mean geometrical characteristics of the substrate and coating couple,
and that the interface behaves locally like an homogeneous material of which elastic
and plastic properties would result from the contribution of both coating and the
substrate [51].
Mathematical expression of the critical load P
(Eq. 2-13 represents the cracking ability of the interface itself, independent of material
geometry)
The terms in Eq. 1-20 are C=1.854E5, P= load in N, a=crack length generated in m,
and HR and HS=hardness of coating and substrate respectively (see Fig. 2.13). The
exponential n in the above equation is a function of the coating thickness as described
in the linear equation below:
Fig. 2.13. Illustration showing the cross-sectional indentation right on the interface
[51].
Page 59
59
(Eq. 2-14)
By taking the ln of this relationship as a function of loading values P (N), n can then
be obtained simply as the slope of the plot. Hence, in order to determine the
exponential n, it is necessary to conduct cross-sectional indentation on the same
specimen using several different loads.
Mathematical expression of the apparent interface toughness
(Eq. 2-15)
Where
(Eq. 2-16)
One issue affecting the experimental outcome of the measured interfacial fracture
toughness is the existence of residual stress resulted from the differential contraction
between coating and substrate during cooling. Chicot found that if an annealing
treatment is performed before indentation, the straight lines relating to samples with
different coating thicknesses on bilogarithmic curve of crack length vs. load will
converge onto a restricted area where they all intersect at a point (Fig 2.14). In other
words, there exists a load, independent of the coating thickness that corresponds to the
onset of cracking, also known as cracking ability. This point may be considered as a
characteristic property of the coating-substrate adhesiveness [51].
Page 60
60
Fig. 2.14. Schematic representation of the crack/indent length – loading relation [79].
Once a crack is formed, the propagation will first go through the zone I, where the
cracks remain in the interface plane. Later moving into zone II as indicated by the
sudden change in the curves of the bilogarithmic diagram (Fig. 2.14). From that point
onward, the crack fronts move into the coating and thus no longer correspond to
delamination of the interface. As shown by Figure 2.12, the crack initiated at a critical
load of 5 N corresponding to a critical crack length of 29µm. Prior to this point, the
curve is simply the impression size of the indenter [51].
Although the cross-sectional indentation method as detailed above is simple, the
difficulty of placing the indent tip accurately on the interface has been known to
hamper the reliability of this technique. This is especially the case for complicated
material systems such as the TBC, where the layers of materials have very different
hardness properties. To address this issue, Wang et al [82] recently came up with a
technique based on a clamped circular plate model, which could measure the
interfacial fracture toughness of TBCs by simply placing the indent in the substrate
close to the interface (Fig. 2.15). The energy release rate G can be obtained with the
Page 61
61
following equation.
(Eq. 2-17)
where Ec, hc, a, and u are the effective in-plane modulus of the coating, the thickness
of the coating, the crack radius, and the central upward displacement of the coating
due to the indentation (see Fig. 2.15), respectively.
Fig. 2.15. SEM image showing an example of the crack generated at the interface due
to indentation below [82]. The indentation caused an upward displacement in the
entire coating with the highest point roughly above the central line of the indent.
Luminescence mapping, where area of TGO delamination (i.e. stress relaxed region)
could be distinguished, was utilized to account for the crack radius, in addition to
SEM observation. FEA method was utilized to accurately account for the upward
Page 62
62
displacement term u. The interfacial fracture toughness of TBC was determined to be
approximately 29±9 Jm-2
between 35 and 100 thermal cycles, which is similar to the
results obtained using barb test.
2.2.3 Barb
Kagawa et al [55] has developed a new testing technique to facilitate the evaluation of
interfacial fracture toughness, which is specifically aimed for TBC systems under
predominantly mode II loading. They described a test that obviates many of the
challenges and limitations with the testing techniques previously mentioned (i.e.
limitation in loading angles, sample geometries, coating thickness, etc). The testing
routine, based on barb geometry, is taken from a common approach used to evaluate
the fiber reinforced ceramic and/or metal matrix composites. It has been demonstrated
that the barb test can be useful toward investigating the crack growth behaviour in
mode II loading, leading to coating delamination [55].
The barb test specimen is fabricated from an EB-PVD processed TBC specimen. The
TBC layer is carefully ground away using WC polishing tool, leaving only a portion
of the original ceramic coating at the top. The remaining TBC layers were notched at
distance of 3mm from the end of the specimens for the purpose of defining the length
of crack growth. Two identical specimens prepared this way, were glued together
using adhesive epoxy and alignment tool to form the final testing piece (Fig. 2.16).
Page 63
63
Fig. 2.16. Illustration of the barb specimen showing a partial removal of the TBC
[55].
Figure 2.17 illustrates the loading fixture equipment used for the barb pullout test.
Sapphire plates were carefully positioned to provide optimal load transfer to the
coating, and to avoid any frictional load transfer to the metallic substrate. A
continuous record of force-displacement response during the loading was kept using a
digital memory scope (ORM 1200, Yokogawa Electric, Tokyo, Japan) [55].
Fig. 2.17. Schematic representation of the barb loading setup (shear mode) [55].
Page 64
64
From the force-displacement plot, the curve increases linearly until a maximum load,
which corresponds to the first crack initiation as observed by a long-standoff distance
optical microscope. Following the peak load, three to four instantaneous force drops
were typically observed, suggesting that crack propagation occurs incrementally until
the final detachment (Fig. 2.18).
Fig. 2.18. A plot of the typical loading curve obtained from the barb test on a EB-PVD
TBC system [55].
The fracture surface is observed with high-resolution SEM micrographs as shown in
Figure 2.19. Based on the microstructural feature of the fracture surface, the
delamination is shown to proceed essentially at the TGO layer, with the crack
propagation occurring in either the TGO/TBC or the TGO/bond coat interface. A
quantitative analysis using this method is done on the EB-PVD TBC specimen using
the model described by Hutchinson et al [71].
(Eq. 2-18)
Page 65
65
Fig. 2.19. a) Fracture surface at low magnification b) a high-magnification SEM
image showing embedded TGO in the bond coat and the TBC remained on the TGO
[55].
Where σsub and σtbc are the uniaxial stress in the substrate and coating layer
resepectively. Esub and Etbc are the Young’s modulus of the substrate and the coating,
which are taken as 200GPa and 44Gpa respectively. The stress values are calculated
as follow:
(Eq. 2-19)
Substituting the above into Eq. 2-18 yields.
Page 66
66
(Eq. 2-20)
An average toughness value of 70J/m2
was obtained using Eq. 2-20. This is slightly
higher than those reported elsewhere by indentation and wedge impression testing
[53,56]. Yukawa et al gave several possible explanations for this phenomena, such as
the effect of residual stress at the interface and the Young’s moduli used in the
analysis, as well as possible friction effect resulting from longer contact zone and
larger normal compressive stress acting on the TBC layer due to the barb testing setup.
Clearly, the effect of TBC properties and delamination mechanisms need to be better
understood in order to yield an accurate analysis in Barb testing. The testing protocol
described by Yukawa et al; however, seems applicable to evaluate the interfacial
toughness of the TBC systems. In a more recent work, Kagawa et al [83] re-analyzed
the barb test measurement based on analytical solutions of a beam theory. It was
determined that the crack actually attains a steady-state at mixed loading angle
between 60-70°, due to a bending moment generated from the support. With this
refined work, an average toughness value of 36J/m2 at approximately 65° loading
angle was obtained, which is more consistent to the value obtained in another
literature work [84].
2.2.4 Four point-bending
Charalambides et al [85] first formulated the critical energy release rate of bi-material
interface through the use of a four-point bending test. This method offers many
advantages such as well-established testing routine and simple sample geometry;
Page 67
67
however, it is mostly restricted to material composites with relatively high fracture
toughness in the debonding layers due to the difficulty in accounting for vertical
cracking, segmentation. In addition, a critical thickness exists for which the energy
necessary for crack propagation can be stored at the interface [54].
In the work by Hofinger et al [54], a modified version of Charalambides’s method is
presented, which enables one to determine the critical energy release rate for
interfacial delamination of a thin and brittle coating. The issues with segmentation
and critical thickness are avoided by attaching the test specimens with a stiffener. The
specimen of the modified test is shown in Figure 2.20. The specimen is notched in the
center with a symmetric interfacial precrack introduced between the inner loading
span.
Fig. 2.20. Illustration of the modified four-point bending specimen and loading setup
with the newly applied stiffener layer for preventing segmentation in the ceramic
coating [54].
An analytical solution is provided to describe the loading condition as shown by the
setup above.
Mb is the constant bending moment along the cross section.
Page 68
68
(Eq. 2-21)
where P, b, and l are the applied force, the sample width and half of the difference
between outer and inner span, respectively.
A steady-state energy release rate can be achieved, provided that the crack length is
large compared to the total thickness of coating and stiffener [86]. Base on the Hook’s
law, a modified equation of Charalambides’ steady-state energy release rate can be
formulated as follow:
(Eq. 2-22)
Where the second moments of inertia, I2 is given as
(Eq. 2-23)
and
Page 69
69
(Eq.2-24)
with
(Eq. 2-25)
Where h, E, and v denote the layer thicknesses, Young’s modulus, and Poisson’s ratio,
respectively, and the subscripts 1, 2, and d refer to the ceramic layer, substrate, and
stiffener. In the analysis, Possison ratio is assumed to be the same for all layers, and
that the stiffener is assumed to be equivalent to the substrate material. A plot of the
normalized energy release rate vs. stiffener to substrate thickness ratio shows that with
increasing thickness ratio hd/h2, the effect of Young’s moduli ratio E2 to E1 on the
energy release rate is decreased accordingly. This implies that by stiffening the top
coat, the energy supplied to the crack growth at the interface is increased under
constant substrate loading conditions.
During the mechanical testing, a force vs. displacement curve was recorded until the
onset of the delamination as indicated by a sudden drop in stiffness from the curve’s
straight line (Fig. 2.21). By knowing the critical load Pc and the corresponding crack
length, the critical energy release rate can be obtained by the analytical equation in Eq.
2-22.
Page 70
70
Fig. 2.21. A typical force-displacement plot showing the onset of delamination as
indicated by the sudden drop in load P [54].
In the four-point bending work by Thery et al. [87], the interfacial fracture toughness
of an as-coated EB-PVD TBC on a β-NiAl bond coat was measured to be 110 J/m− 2
,
which decreased to a range between 55 and 23 J/m− 2
after 50 and 150 1h thermal
cycles at 1100C°. The as-coated toughness in this case seems to be much lower than
the values reported in the work of the cross-sectional indentation and barb test, while
the value after prolonged thermal cycling is just slightly higher. The possible reasons
for the discrepancy in the as-coated value between the different tests will be briefly
discussed in the following sections.
2.3 Shortcomings of current interfacial fracture toughness models for
indentation
Despite the on-going effort in the development and improvement of new and existing
models, none of the current interfacial fracture toughness models could account for
the complex nature of delamination mechanism, crack morphology, and stress
dissipation. For instance, in all of the stress based models, the mathematical models
could only account for either buckling or non-buckling load alone, without possibly
Page 71
71
taking into account of mixed modes. In addition, many of these models assume the
crack to be circularly in nature with certain measured radius, where as in most real
situations, crack openings have far more complex shapes, and are therefore difficult to
define.
As a result of the simplicity and assumptions in these models, one tends to either over
or underestimate the toughness values, leading to large discrepancy when comparing
the calculated values to those obtained by others using both similar or different
techniques. Further, there has not been a model yet that could be adequately applied to
a coating of multi-layered structures with non-homogeneous interfacial morphology
such as in the case thermally cycled TBC. So far, most of the existing models have
been developed based on the delamination of a single flat interface between two
dissimilar materials.
One major issue with many of the current interfacial toughness models is that the
amount of alloy creep or plastic deformation is often overlooked [88]. This is
problematic particularly for the indentation on ductile materials. To minimize the
effect of plastic deformation, the stress-analysis model based on Vickers indentation
from the cross-section has been proposed.
Another problem with the current techniques of measuring interfacial fracture
toughness lies in the difficulty in representing the true interfacial toughness values of
interfaces with complicated geometries, such as TBCs. The rumpling of the
TGO/bond coat interface due to thermal cycling makes it difficult to account for the
exact loading phase angle Ψ of the intrinsic TBC failure mode. According to Choi et
Page 72
72
al [89], the TBC interface undergoes a dynamic loading phase angle Ψ by a
competition between edge and buckle delamination (see Fig. 2.22). The energy
release rate along the straight side and front of the blister crack of a mixed
edge-buckle delamination can be denoted as Gside and Gfront, respectively. The dynamic
energy release rate of intrinsic TBC failure, G, as a result of the mixed mode loading
can be elucidated using Figure 2.23 below by the work of Choi et al [93], where Go
and c are the steady energy release rate parallel to the crack front and the critical
stress for buckling, respectively (see Eq. 2-26,2-27).
Fig. 2.22. Illustration of the two instrinsic failure modes of TBCs [93].
(Eq. 2-26)
(Eq. 2-27)
When o (stress generated at the interface) reaches the c (defined in Eq. 2-27 above),
the G values rises accordingly, especially fast on the side of the blister. Soon after,
Gside reaches the value of Go with a loading angle equivalent of 45°. As o continues
to increase, the side will reach closer to 90 ° (mode II), and the front will be close to
60 °.
Page 73
73
Fig. 2.23. Plots showing the dynamic change in the stress, loading angle, and energy
release rate of TBC interfacial degradation [89].
2.4 Factors affecting interfacial toughness measurement
As described in the previous section, the interfacial toughness values obtained from
different test methods will obviously be different due to the phase angle applied in the
test setup. For instance, the values obtained using the barb test is slightly higher than
those from the cross-sectional indentation due to the fact that the loading angle in the
barb test is known to be relatively higher (65°- 90°). The calculated fracture toughness
will be higher as a result of the contact and friction associated with shear loading (i.e.
mode II). Although the exact phase is unknown in the cross-sectional indentation test
[82], it is reasonable to assume that the angle is much smaller than the barb test due to
a significant portion of out-of-plane displacements. Hence, the results obtained from
the cross-sectional indentation method can be considered fairly similar to that of the
barb test.
Page 74
74
The residual stresses in the TGO and YSZ are of great importance when measuring
the interfacial fracture toughness of TBCs. As described in the work by Wang et al
[82], the stored strain energy in the TGO and YSZ would be partially released when
buckling of the TBC occurs, contributing to a greater driving force for coating
spallation. This additional driving force must be accounted by the second term in Eq.
2-28 (see below); otherwise, the calculated interfacial fracture toughness results may
be significantly underestimated.
In addition, the interfacial fracture toughness is likely to be affected by the sintering
and anisotropic microstructure of the YSZ top coat. For instance, in the four point
bending method, inter-columnar fracture and shear displacement were often observed
on specimens subjected to short thermal cycling/isothermal oxidation [87]. These
types of damages, resulted from low sintering level of the YSZ, may significantly
decrease the critical load, Pc, necessary for crack propagation under bending, thus,
yielding a much smaller interfacial fracture toughness value than those obtained from
barb or cross-sectional indentation methods. This issue is largely resolved after
prolonged thermal exposure, as the sintering of the YSZ leads to higher stiffness in
the overall coating.
(Eq.2-28)
Beside the loading phase angle of the interface and residual stresses of coating, other
factors such as sample differences due to coating deposition methods, coating
thicknesses, bond coat structural change, and phase transformation would also give
rise to a difference in the value of interfacial fracture toughness.
Page 75
75
CHAPTER 3
Scope and aims of research
3.1 Project introduction
There is a wide consensus that the leading degradation mechanism in TBC systems is
associated with the interfacial rumpling or ratcheting of the bond coat during the
course of high temperature oxidation and cooling. This act of distortion resulting from
biaxial compression induces tensile stresses perpendicular to the interface [90-94].
Eventually, crack nucleation and growth can take place at TGO/bond coat and
TGO/Top coat interfaces, as well as within the TGO oxide layer.
Many literatures have identified the TGO growth rate as the key influence in the
lifetime of TBC systems, and that spallation is mainly attributed to the strain energy
lying within the TGO with negligible contribution from the top coat [95,96]. However,
it is evident from recent studies that the bond coat and substrate have clear influence
on the degradation mechanism [97,98]. As pointed out in many previous works,
despite having faster TGO growth rate and more rumpling, certain bond coat and
substrate pairs still outperformed others in lifetime experiments.
To this day, there is still a lack of fundamental understanding of the exact degradation
mechanism leading to rumpling and final spallation. It is only in recent years that
more works have dedicated to the study of the effect of substrate and bond coat
compositions on the lifetime of TBC system. The influence of superalloy composition
Page 76
76
on TBC lifetime had received very little attention in the past while the life-limiting
factors are yet to be clarified.
3.2 Considerations and requirements for choosing an adequate
adhesion test
Chen et al [58] summarized several key considerations for the selection of adhesion
tests in their review paper on interfacial toughness. To produce accurate results,
plastic deformation in the coating or substrate should be avoided, and the stress
responsible for delamination should be maximized. For stress based analysis, the
geometric parameters, which are raised to high powers, should be accurately
accounted for.
In many instances, an accurate measurement of crack geometry can become tricky
due to the difficulty in defining the crack geometry or propagation path. Also, one
should not overemphasize the contribution of residual stress in overall delamination,
since the driving force depends on the combination of residual stress and the applied
loading [58].
To choose an appropriate testing routine, a number of questions should be taken into
considerations as summarized below:
1. What are the bulk properties of the coating and substrate?
- In most techniques, interfacial delamination relies on the transfer of stress
through the coating or the substrate layers. Hence, these layers should
possess a higher cohesive strength than that of the interface. Bending test
Page 77
77
is usually not suitable for brittle materials; whereas, the double cantilever
tests could be used due to the lower bending stresses inflicted.
2. How thick is the coating?
- Depending on the coating thickness, there often exists a practical limit at
which certain testing techniques could be used reliably. In pull-off tests,
for example, coatings of less than 10 m cannot be examined since the
defects are too small to reach the critical stress level prior to the failure of
the adhesive glue. For thin coatings, methods that generate buckling
failures such as bulge tests and indentation are the most suitable choices.
On the other hand, thick coatings with high bending stiffness may undergo
through thickness shear cracking upon indentation from the top; however,
the cracks may not divert into interface if the coatings are either brittle or
have high interfacial toughness.
3. How high is the adhesive strength?
- In the case of poor adhesion (Gint < 5Jm-2
), nearly every test could be
utilized to determine the interfacial toughness. As soon as the adhesion
strength increases, the number of testing choices becomes limited,
especially for thin and brittle coatings. To achieve the large stress required
to drive delamination, methods that generate additional stresses through
frictional contact, such as scratch and indentation tests, are the best
choices.
4. Is the residual stress controllable?
Page 78
78
- Though the stressed overlayer method finds many applications in interfacial
toughness investigation, the addition adhesive layer on top must be
deposited such that the residual stress is large to induce spallation in the
coating of interest, yet the overlayer must have superior adhesion with the
coating than that of the tested interface. This is not always easy to achieve
in many coating/material systems.
5. What type of in-service adhesion failure is involved?
- Ideally, it is best to choose the testing techniques that best mimic the
in-service loading conditions at which the coating fails. For example, the
stressed overlayer method finds useful application in coatings subjected to
frequent thermal cycles and multiple depositions during processing.
3.3 Project objectives
Many recent studies have shown that the rate of TGO growth and interfacial rumpling
are only part of the contributing factors for degradation. There is sufficient evidence
from previous research works suggesting that the alloying contents in the bond coat
and substrate can have significant influence on the interfacial evolution; hence, the
spallation lifetime of TBCs.
In this project, a systematic study of the effect of alloying compositions on high
temperature oxidation behavior is done for TBCs with different industrial Ni-based
superalloys and Pt/Pt-Al diffusion bond coats. The attempt here is to identify the
variation in TGO interfacial evolution between different systems of substrates and
Page 79
79
bond coats, and determine the key degradation factors in controlling TBC lifetime.
Ultimately, this work aims to clarify and establish the missing link between the
degradation mechanisms and chemistry of the bond coat and substrate.
The proposed work can be summarized by the step-by-step objectives listed below:
1. To measure the interfacial adhesion (semi-quantitative) of the TBCs based
on the cross-sectional indentation technique.
2. To compare the interfacial evolution of the different TBC systems in terms
of microstructure parameters such as, TGO growth kinetic, TGO residual
stress, phase distribution of the bond coat near the TGO/bond coat
interface, etc.
3. To establish a relationship between those microstructure parameters and the
interfacial adhesion of the TBCs, and identify the parameter(s) that has the
most influence on interfacial adhesion.
4. To carry out in-depth analysis of the chemistry and phase transformation of
the bond coat near the TGO/bond coat interface using EDX, EPMA, and
EBSD, etc.
5. To clarify how the key degradation mechanism is associated with the
alloying chemistry of the substrate and bond coat.
Page 80
80
CHAPTER 4
Microstructure parameters affecting interfacial adhesion of
Thermal Barrier Coatings by the EB-PVD method
4.1 Introduction
A.G. Evans and R.A. Miller et. al [90,99] first suggested that the main intrinsic cause
of TBC spallation is due primarily to thermal coefficient mismatch between the
ceramic-metal layers and bond coat oxidation. They also paved the way toward the
development of TBC life models by expressing the critical failure criterion as a
function of the strain energy accumulation near the TBC interface. There is now a
wide consensus that the leading degradation mechanism in thermal barrier coating
systems (TBCs) is associated with the interfacial rumpling of the bond coat during the
course of high temperature oxidation and subsequent cooling, which arises mainly
from mismatch in the coefficient of thermal expansion between the multilayered
structures of the TBCs [23,24]. This act of distortion, resulting from biaxial
compression of the thermally grown oxide (TGO) layer, induces tensile stresses
perpendicular to the interface and eventually leads to spallation [90,91-94].
EB-PVD has been the choice for depositing the state-of-the art ceramic top coat for
many years; while thermal sprayed (i.e. air plasma spraying) coating methods are still
widely used in non-moving components of the jet engine and land-based turbines. It
should be noted that the phenomena observed and mechanisms proposed in this work
are specifically for TBC systems with EB-PVD top coat and diffusional bond coat,
Page 81
81
and thus, may not be fully relevant to coatings made by thermal spraying and other
methods due to fundamental differences in coating morphology and chemistry.
As more industrial grade TBCs are being developed nowadays, the focus of the latest
TBC research has been shifted toward the comparison of different coating systems. It
is only in recent years that the influence of bond coats and substrates on the modes of
degradation is being studied and compared between different TBCs, in an effort to
understand the key factors affecting the TBC spallation lifetime [26,35,100-102].
Despite the significant progress made in the study of TGO residual stresses and
modelling of TBC failure little work has taken into account of how these stresses and
failure modes could be influenced by the combination of bond coat and substrate alloy.
While the modelling community [23,24,34,103] made detailed numerical description
of the interfacial undulation and corresponding stress state during cyclic oxidation,
their models tend to overlook the dependence of these values on the complicated
nature of bond coat chemical diffusion/phase transformation, etc. Hence, their
prediction of TBC lifetime expressed in terms of a failure criterion (i.e. TGO
thickness) could deviate largely from that obtained in real experiment. So far, there
has not been a successful attempt to come up with a model that could accurately
predict the lifetime of a state-of-the art TBC.
In order to gain a better understanding of the influence of substrate and bond coat
compositions on the degradation mechanisms of TBCs, it is necessary to adopt one of
the aforementioned techniques in Chapter 2 for measuring the TGO/bond coat
interfacial fracture toughness prior to spallation. Conventional methods (i.e. pull-off,
double cantilever beam, etc.) cannot be used to quantify the interfacial fracture
Page 82
82
toughness due to limitations such as coating thickness and inherent coating adhesive
strength of TBCs. There have been few successful measurements of the interfacial
fracture toughness of TBCs so far, using techniques such as four-point bending,
cross-sectional indentation, and barb test, etc. However, as discussed in section 2.2.3,
2.2.4 and much of the last chapter, these techniques either require tedious
experimental setup (i.e. four-point bending) or extensive geometric/material property
measurement routines.
Considering that a large number of TBC systems will be studied in this project, it is
necessary to adopt a simple yet effective approach to measuring the interfacial
toughness of TBCs. Of all the techniques reviewed in Chapter 2, the cross-sectional
indentation right on the interface developed by Choulier and Dal Maschio [79,80]
seems to satisfy this demand; hence, it will be utilized in this project to provide a
semi-quantitative measurement of interfacial adhesion of TBCs. The semi-circular
plate model used in the indentation work by Wang et al [82] will then be applied to
approximate the interfacial fracture toughness values.
This study intends to clarify the effects of bond coat and substrate on the degradation
of TBCs, by examining the evolution of TBC interfacial adhesion, TGO growth and
rumpling, and the ’ formation in bond coats against isothermal oxidation. A
systematic analysis was carried out on five different industrial grade TBCs with the
yttria-stabilized zirconia (YSZ) top coat produced by the electron beam physical
vapour deposition (EB-PVD) method on top of 3 different bond coats and substrates,
which then allowed us to establish a relationship between TBC adhesion, TGO growth
and rumpling, bond coat phase distribution. The results have clearly indicated the
Page 83
83
dynamics (i.e. dependence) of oxide-bond coat interfacial adhesion with respect to the
phase distribution of the bond coats and TGO growth rate. Moreover, it has been
clearly demonstrated that accurate prediction of TBC spallation lifetime could only be
achieved by taking into account the influence of substrate and bond coat chemistry on
the bond coat-TGO interfacial adhesion.
4.2 Experimental details
Three different industrial standard single-crystal nickel based superalloys were used
for the following work: CMSX-4, SRR99, and TMS-82+. The nominal chemical
compositions for these three alloys are given in Table 4.1. These fully heat treated
(solution and primary aged) alloys were processed by conventional investment casting
methods into cylindrical rods (10mm diameter), having the long axis aligned closely
to the {0 0 1} direction. Each rod is further sliced into disk shape (10mm diameter
and 4mm thickness) and spot welded onto Nimonic sticks. On the other hand, three
different bond coats, namely HT Pt-Al (where HT implies high temperature and low
activity aluminium environment), LT Pt-Al; where LT implies low temperature and
high activity aluminium environment), and Pt-diffusion, were subsequently deposited
onto the surfaces of the disk shape alloys. The CMSX-4 and SRR99 specimens were
only paired with LT Pt-Al bond coat for the purpose of studying the influence from
superalloys alone. Table 4.2 below provides a summary of the five different TBC
systems used for this work.
Page 84
84
Table 4.1
Nominal composition of the three superalloys used in this work (wt%)
Substrat
e
Co Cr Mo W Al Ti Ta Hf Re Ru C Ni
SRR99 5.0 8.0 - 9.5 5.5 2.2 2.8 - - - - Bal.
TMS-82
+
7.8 4.9 1.9 8.7 5.3 0.5 6.0 0.1 2.4 - - Bal.
CMSX-
4
9.6 6.5 0.6 6.4 5.6 1.0 6.5 0.1 3 - - Bal.
Table 4.2
Visual representation of the five TBC systems used in this work
SUPERALLOY SUBSTRATE
BOND COAT CMSX-4 TMS-82+ SRR99
HT Pt-Al ✔
LT-Pt-Al ✔ ✔ ✔
Pt diffusion ✔
These commercially available TBCs, fabricated by Rolls-Royce Plc and High
Temperature Materials Unit of NIMS Japan, were specially selected to cover different
generations of superalloys (SRR99: 1st generation; CMSX-4: 2
nd generation;
TMS-82+: 2nd
generation). The first generation superalloys typically contain more Ti,
but less Re and Ru, while the latest superalloys tend to be Ti-free with higher Re and
Ru. These alloys were then paired with three common bond coat systems: Pt-diffusion,
LT Pt-Al, and HT Pt-Al. The three bond coat systems tend to form very different
concentration profiles in the bond coat region (i.e. Al and Pt contents), as a result of
the different high temperature diffusion processes used during coating fabrication.
The diversity in substrates and bond coats (i.e. difference in compositions) helps to
identify different failure/degradation characteristics. Figure 4.1 below indicates the
Page 85
85
spallation lifetime of various TBC systems [26], including the substrates and bond
coats used in this work. It can be seen that the CMSX-4 and TMS-82+ have relatively
better spallation lifetime performance than the SRR99 alloy when paired with any of
the three bond coats. On the other hand, the Pt-diffusion (LCBC) bond coat tends to
have the best spallation lifetime performance when paired with any alloy, followed by
the LT Pt-Al (RT22LT), while the HT Pt-Al (CN91PA) comes last in terms of
spallation lifetime.
Fig. 4.1. A comparison of the relative spallation lifetime of commercial TBCs
subjected to thermal cycling, where the x-axis indicates the types of superalloy while
legends indicate bond coat (LCBC = Pt-diffusion, RT22LT = LT Pt-Al, CN91PA = HT
Pt-Al) [26].
The HT Pt-Al and LT Pt-Al bond coat specimens were processed by first
electrodepositing a thin layer of platinum of 5 and 7 m, respectively, followed by a
vacuum heat treatment at 1100°C for 1hr. Next, 5 hrs vapour phase, and 20 hrs pack,
Page 86
86
aluminization processes were applied at 1080°C and 870°C, respectively. The
Pt-diffusion bond coat was processed similarly by first electroplating a 10 m layer of
platinum, and subsequently treated to a vacuum treatment process at 1150°C for 1hr
without the aluminization process. A further heat treatment in argon atmosphere was
applied on all specimens for 1hr at 1100°C, prior to the deposition with a ZrO2/7wt.%
Y2O3 (YSZ) top coat of 175m by EB-PVD. Lastly, a vacuum heat treatment (1100°C,
1hr) and ageing (870°C, 16hrs) were applied to all specimens prior to testing for
quality assurance purposes.
Five specimens for each TBC system in a total of 25 sliced pieces were subjected to
isothermal exposure at a furnace temperature of 1135°C for 0, 30, 50, 100, and 200hrs.
According to previous studies [37,39], thermal cycling of TBCs leads to significant
interfacial rumpling upon early oxidation times, making it difficult to achieve a
smooth planar interface along the indentation orientation at the interface. Therefore,
selection of isothermal exposure instead of cycling exposure was to be able to study
the influence of the inherent bond coat and substrate chemistry on interfacial adhesion
of the TBCs after thermal exposures.
The exposed specimens were then mounted in a Struer Citovac vacuum impregnation
chamber using a low-shrinkage epoxy resin in order to preserve the specimen edges,
and polished to mirror finish for microstructural characterization. A high resolution
KEYENCE VHX-1000 optical microscope was used to analyze the microstructural
changes on the entire cross-section of all the specimens. Micrographs of the entire
TGO/bond coat interfacial area were recorded for every specimen.
Page 87
87
Cross-sectional interfacial adhesion test was carried on the thermally exposed
specimens by Vickers micro-indentation based on similar techniques proposed by
previous researchers [51,79,80], using an AMT series MMT-7 Micro-Vickers hardness
machine. The adhesion results were then used to elucidate the mechanisms underlying
the TGO/bond coat interfacial degradation. The symmetrical Vickers indenter was
placed in a way such that one of the diagonal of the indent is as close to the
TGO/bond coat interface as possible (see Fig. 4.2). For an indent to be considered
accurately placed, the entire diagonal needs to lie within a distance of approximately
3m from the TGO/bond coat interface. Indents that did not meet this standard were
discarded from the averaging of crack lengths. Although no systematic study was
carried out regarding the sensitivity of the test results to accuracy of hitting the
interface, it can be seem later in section 4.3.2 that the standard deviations of the crack
length results were generally low enough for a reasonable comparison of the
specimens. Cracks were deliberately introduced at the interface, and the crack length
was calculated by measuring the crack propagation from both corners of the indent.
The length is defined by the total distance of the cracks initiated, starting from either
of the two corners of the indent to the points where the cracks terminate by a distance
of at least 10m (See Fig 4.3 and 4.4). For every specimen, a minimum of 6
indentations was conducted along the interface in order to obtain an overall average
crack length to represent the adhesion of that particular specimen.
Page 88
88
Fig. 4.2. SEM image showing the placement of the Vickers indenter at the TGO/bond
coat interface.
Fig. 4.3. SEM image showing the crack propagation to the left of the indent.
Page 89
89
Fig. 4.4. SEM image showing the end of crack propagation to the left of the indent.
Many test trials using different indentation loads were carried out on as coated TBC
samples (i.e. without oxidation). It was found that a load of at least 9800mN (peak
holding time of 10s) was necessary to generate cracks that were visible under SEM.
For the oxidized samples, particularly those treated to longer times, such a load
tended to generate very long cracks, making it difficult to make at least 6 indents per
specimen over the sample cross-section. Hence, all of the isothermally exposed
specimens (i.e. 30-200hrs) were indented with a lower load of 4900mN and a peak
holding time of 10s. Figure 4.5 shows a comparison of the crack lengths of the same
TBC system generated with the two different loads. The crack lengths of the
specimens differed by approximately two folds when the loading doubled. It should
be noted; however, that not all specimens or instances of isothermal exposure follow
this doubling trend.
Page 90
90
Fig. 4.5. A comparison of typical crack lengths between those generated by 9.8 and
4.5N indentation loads, showing differences of roughly two-folds in length.
The entire coating cross-section of every indented specimens was observed and
photographed using a FEG-SEM at an accelerating voltage of 20kV and a working
distance of 10mm. In addition, compositional mapping for all the major alloying
elements was carried out on the 30hrs specimens near the TGO/bond coat interfacial
region at 10,000X magnification. An additional chemical mapping analysis was done
using EPMA (Electron Probe Microscope Analysis) for the SRR99 LT Pt-Al specimen
under an accelerating voltage of 15kV and a beam current of 20nA.
Residual stress analysis was conducted on all five TBC systems with a set of newly
sliced specimens, which were isothermally exposed at 1135°C for 30, 50, 80 and
100hrs. The residual stress was measured using the photo-luminescence
piezo-spectroscopy method [29,30,104,105], by acquiring Cr3+
luminescence spectra
from -Al2O3. This experiment was carried out using a Renishaw Raman spectrometer
fitted with an optical microscope (Renishaw 1000 Ramanscope system UK). A green
Page 91
91
argon laser with a wavelength of 514.5 nm was used as the laser source. The spectra
were taken from at least 5 different locations away from the sample edge, each of
which had an acquisition time of 30s. During the course of experiment, the Raman
machine was calibrated once prior to testing a specimen by taking a spectrum from a
strain-free single crystal sapphire sample. The residual stress in the TGO was
estimated from the R2 peak shift given off from the TGO relative to the unstrained
sapphire, assuming a planar equibiaxial stress: (GPa) = Δv (cm-1
) /5.07 [25].
Measurement of the interfacial rumpling was carried out by tracking the TGO/bond
coat interface with three cross-sectional SEM images over a total distance of 573 m.
Each position along the interface was represented by a pair of x (in-plane with the
specimen’s cross-sectional surface) and y (out-of-plane direction parallel to EB-PVD
coating growth direction) coordinates, and that the displacement between two points
was approximated by a straight line.
4.3 Results
4.3.1 Microscopy observation
Fig. 4.6 shows some representative micrographs of three different bond coats (LT
Pt-Al, HT Pt-Al and Pt diffusion) paired with TMS-82+ superalloys. According to
these micrographs, TGO thickness, rumpling and precipitates can be observed and
compared. For example, the TMS-82+ LT Pt-Al specimen seemed to have slightly
thicker TGO growth and noticeably more rumpling than that of the TMS-82+ HT
Pt-Al at 30hrs. At 50hrs, both specimens began to form large pieces of precipitation
Page 92
92
near the TGO/bond coat interfacial region. Such precipitates have identified as ’
phase formed after thermal treatments. Fig. 4.7 shows representative images of LT
Pt-Al bond coat with CMSX-4, SRR99, TMS-82+ superalloys as substrate, where the
SRR99 specimen has a much thicker TGO growth than that of the TMS-82+ and
CMSX-4. At 30hrs, large pieces of ’ phase could already be seen in the SRR99 LT
Pt-Al specimen, yet both the CMSX-4 and TMS-82LT Pt-Al specimens had only trace
amount of those near the grit line further away from the interfacial region.
Page 93
93
Fig. 4.6. Representative interfacial microstructure between the three different bond
coat systems (LT Pt-Al, HT Pt-Al, Pt-diffusion) paired with TMS-82+ superalloy. The
HT Pt-Al bond coat had the least rumpling, yet failed to survive up to 200hrs.
Page 94
94
Fig. 4.7. Representative interfacial microstructure between the three different
superalloys (CMSX-4, TMS-82+, SRR99) paired with LT Pt-Al bond coat (SRR99
failed after 30hrs, and no photos are available for 100 and 200 hrs time trials). The
SRR99 specimen had the most ’phase formation and rumpling, and could not survive
Page 95
95
up to 50hrs.
4.3.2 Lifetime and interfacial adhesion of TBC systems
A qualitative analysis of the TGO/bond coat microstructure is given in this section for
the three superalloys and bond coats. First, the thermally treated specimens were
examined by optical microscopy for whether or not spallation has occurred (defined
as approximately 20% detachment of the top coat by cross-sectional length). It was
found that the SRR99 LT Pt-Al specimen was amongst the first to fail between 30 and
50 hrs, as its coating remained mostly intact at 30hrs, but major spallation was soon
observed at 50hrs. The TMS-82+ HT Pt-Al also had reached its spallation life
between 100 and 200 hrs, as indicated by over 20% of coating detachment at 200hrs.
All remaining specimens survived well into 200hrs exposure without noticeable
coating detachment at all.
A quantitative analysis is done using the cross-sectional indentation technique to rank
the interfacial adhesion for the three different substrates and bond coats. Fig. 4.8
shows the crack lengths against thermal exposure time for the five different TBC
systems. CMSX-4 with LT Pt-Al bond coat with lifetime longer than 200hrs exhibited
the shortest crack lengths of all specimens, having lengths no greater than 50m
throughout the entire thermal exposure time. SRR99 LT Pt-Al with lifetime between
30-50hrs, on the other hand, had the longest cracks after only 30hrs thermal exposure
with an average length of 128m. The three TMS-82+ specimens had a somewhat
intermediate crack resistance, with the HT Pt-Al bond coat having the longest average
Page 96
96
crack length of 128 m after 100hrs thermal exposure.
Fig. 4.8. The variation of crack lengths vs. isothermal oxidation (1135°C) hours for the
five different TBC systems (i.e. the prefix TMS-82+ is substrate, while Pt-diffusion is
bond coat) using a load of 4900mN.
It was noticed that the CMSX-4 LT Pt-Al and TMS-82+ HT Pt-Al specimens did not
follow a typical increasing crack length trend with the number of exposure hours. The
CMSX-4 specimen, for instance, had a declining length trend going from 44 m at
30hrs to 17 m at 100hrs, before rising to 38 m at 200hrs. The HT Pt-Al specimen
also had similar behavior, for which its crack length went from 111 m at 30hrs to 36
m at 50hrs, before rising to 128 m at 100hrs (failure after 100hrs). This
phenomenon suggests the need for further investigation on these two specimens, by
Page 97
97
studying their interfacial evolution, and by comparing them with other specimens.
4.3.3 TGO thicknesses, residual stresses and rumpling at the TGO/bond coat
interface
Fig. 4.9 shows the oxidation kinetics (average TGO thickness vs. time1/2
) curve. First,
it can be seen that the TGO growth rate started out quite similar for all (ranging
approximately from 0.0806 to 0.0830 m/hr), except the SRR99 LT Pt-Al specimen
which had a higher growth rate of over 0.100 m/hr. This specimen was not able to
survive up to 50hours. As the exposure time increased, the variation in thicknesses
between different specimens became more apparent, particularly for the comparison
between CMSX-4 LT Pt-Al and other specimens. The CMSX-4 LT Pt-Al specimen
had maintained a very thin TGO layer, and that its growth rate became relatively slow
between 30 and 100hrs at only 0.00572m/hr, whereas the 2nd
slowest specimen has
a growth rate of 0.0137m/hr. As shown earlier, CMSX-4 LT Pt-Al had the best
interfacial adhesion, and a lifetime longer than 200hrs.
Page 98
98
Fig. 4.9. The TGO growth kinetics vs. isothermal oxidation (1135ºC) hours for the five
different TBC systems.
Fig. 4.10 shows the residual stress in the TGO of every specimen from 30 hr to 100hrs,
obtained using luminescence measurements. SRR99 LT Pt-Al specimen shows the
lowest stresses in the TGO which spalled after 30hrs. Among the other 4 samples, the
stress level in TGO of both TMS-82+ HT Pt-Al and CMSX-4 LT Pt-Al shows
decreasing trend with increase in the thermal treatment time, while the other 2
samples show slight increase of the stresses with increase in the thermal exposure
time.
Page 99
99
Fig. 4.10. The TGO compressive stress vs. isothermal oxidation (1135°C) hours for
the five different TBC systems.
Fig. 4.11 shows the evolution of TGO length against thermal exposure time for up to
100hrs. It can be seen that the initial degree of rumpling was very similar for different
samples, but a noticeable difference was soon observed after 30 hrs of thermal
exposure. TMS-82+ LT Pt-Al had the most rumpling through the course of
experiment up to 100hrs, while the TMS-82+ HT Pt-Al specimen maintained the least
rumpling throughout the entire exposure time. On the other hand, the TMS-82+
Pt-diffusion specimen did not show much increase in rumpling throughout the entire
exposure time.
Page 100
100
Fig. 4.11. The evolution of interfacial rumpling vs. isothermal oxidation (1135°C)
hours for the five different TBC systems.
4.3.4 Phase transformation and Ni/Ti segregation
The precipitation mentioned previously is identified as to ’ phase transformation,
which according to the Ni-Al-Pt ternary phase diagram in the work of Gleeson et al
[36], occurs when Al is undergoing depletion as a result of oxidation. Quantification
of this phase was carried out by measuring the total amount by area (m2) of the
precipitates that are attached to the TGO/bond interface over a distance of
approximately m. This quantification was done using the aid of an image
analysis program. Fig. 4.12 shows the content of the ’ phase as function of thermal
treatment time. The SRR99 LT Pt-Al specimen had the fastest ’ formation in the
Page 101
101
early 30hrs with the shortest lifetime. The 2nd
most ’formation can be seen on the
TMS-82+ HT Pt-Al specimen, followed by TMS-82+ LT Pt-Al and CMSX-4 LT
Pt-Al (see Fig 4.12).
Fig. 4.12. The amount of ’ formation attached to the TGO/bond coat interface vs.
isothermal oxidation (1135°C ) hours.
EDX analysis was done on all specimens near the TGO/bond coat interfacial area. Fig.
4.13 shows the Ti maps while other elemental maps either had no obvious variation
between them, or were slightly influenced by the dilution effect from the
aluminization process. A noticeable level of Ti concentration was present near the
bond coat of the SRR99 LT Pt-Al system, including a patch of highly concentrated
area near the interface. This was however, much less obvious in the case of the
Page 102
102
TMS-82+ and CMSX-4 LT Pt-Al systems. The observed Ti segregation phenomenon
on SRR99 LT Pt-Al was further examined using EPMA, and the results are shown in
Figure 4.14. A clear segregation between Al/Cr and Ni/Ti could be seen near the
TGO/bond coat interfacial area.
Fig. 4.13. EDX mapping showing Ti distribution near the TGO/bond coat interface of
the coating after 30hrs of isothermal exposure. An obvious Ti enrichment can be seen
in the bond coat of the SRR99 specimen.
Page 103
103
Fig. 4.14. EPMA mapping showing the elemental mapping of Ni, Ti, Al, and O for
SRR99 LT Pt-Al after 30hrs of isothermal exposure (the color bars are values in wt
%). The regions of Al and Cr depletion correspond exactly to the enrichment in Ni
and Ti contents.
As shown in the previous sections, the phenomena of early to ’phase
transformation and Ti segregation were both found near the TGO/bond coat region of
the SRR99 LT Pt-Al specimen. Hence, there is an interest to investigate whether or
not there is any link between the two phenomena. A thermodynamic software, named
Thermo-Calc was used to identify Ti distribution in the and ’ phases of the Pt-Al
bond coat system. A set of conditions including the isothermal temperature used
(1409K), the oxidizing atmospheric pressure (1 atm), the Al concentration of
as-coated Pt-Al bond coat at the interface (19 wt%), and the overall composition of
the major alloying elements (Ni balanced) were specified as the input for the
Page 104
104
calculation. At any given Al concentration, we assume that the system is in a stable
equilibrium state after certain amount of time. Note that the 19 wt% Al was chosen
based on the EDX analysis result of the SRR99 LT Pt-Al sample studied by Wu et al
[26]. With increase of the Al content, all other remaining alloying elements were
proportionally reduced based on their original wt%, so as to reflect the dilution effect
from the aluminization process (see Table 4.3).
Table 4.3
The nominal composition in wt% of the bond coat of the SRR99 specimen as
specified in the calculation of Thermo-Calc relative to its original bulk composition
Substrate Co Cr Mo W Al Ti Ta Hf Re Ru C Ni
SRR99 5.0 8.0 - 9.5 5.5 2.2 2.8 - - - - Bal.
SRR99 (bond
coat)
4.3 6.9 - 8.2 19.
0
1.9 2.4 - - - - Bal.
Fig. 4.15a gives the mole-fraction of different phases in SRR99 LT Pt-Al obtained
based on the calculation in Thermo-Calc, starting from 19 wt% Al and ending at 10
wt%. This range is set in order to simulate Al depletion due to isothermal oxidation of
the experimental specimen. The Ti distribution results in the and ’ phases, were
calculated with 12% Al concentration, ensuring that both phases are present on the
three alloys compared (note that the relative Ti distribution trend does not vary with
Al%). Fig. 4.15b shows the mole fraction of Ti in each of the two phases for the three
different alloy specimens. The calculation results suggest that the depletion of Al
leads to formation of the ’ phase while the Ti enriched in the ’ phase. It should be
noted that the database that came with this version of Thermo-calc here does not
contain the Pt element. Hence, the thermodynamic results were calculated under the
assumption that the Pt content is substituted entirely by Ni element. Since Pt and Ni
Page 105
105
belongs to same platinum group metals with similar chemical and physical properties,
these elements tend to mix together in solid solution (i.e.,(Pt,Ni)Al of the -NiAl
bond coat phase). Hence, it is reasonable to assume that substitution of Pt by Ni in the
calculation does not influence the phase formation results obtained above.
Fig. 4.15. a) The mole-fraction of different phases existing as a function of Al content
in the SRR99 LT Pt-Al sample obtained from calculation using Thermo-calc, and b)
The comparison of Ti concentration in and ’ phases of the three alloys.
4.4 Discussion
4.4.1 Approximation of interfacial fracture toughness Gc
As mentioned in Chapter 2, it is desirable to obtain a quantitative comparison of the
adhesion between any two coating systems based on experimental results, in
particular, microstructure features generated by the cross-section indentation. Out of
all the techniques introduced in Chapter 2, the semi-circular clamped model using
Page 106
106
cross-sectional indentation as the mean to generate fracture appears to be applicable to
produce reliable fracture toughness values of TBCs without the hassle of complicated
sample preparation and loading setup [77]. Hence, it is utilized here to quantify the
interfacial fracture toughness of the five TBC systems, which can be represented by
the energy release rate Gc with respect to crack radius, a, as shown in Eq.4-1 below:
(Eq. 4-1)
It should be noted that the indentation in this work generated mostly out-of-plane
stresses (i.e. opening tensile force at the interface) without buckling in the
delaminated coating. Therefore, it is reasonable to neglect the influence of in-plane
stresses (i.e. stored residual stress in the coatings), which can contribute to additional
energy release. It can be noted that the crack length generated by indentation is much
less than the critical spallation size by buckling, eg. a critical buckling crack radius of
886m for the 130m TBC coating in Wang’s work [86]. The values of Ec are
approximated from the data plot in Fig. 6 of the work on YSZ stiffness using
miniature 3 point-bending by Wang et al [106]. Since the bi-layer coatings in that
work consisted of 8 wt% YSZ and a TGO scale formed on a Pt-diffusion bond coat,
the effective modulus obtained should bear close resemblance to that of the coatings
used in this work as similar conditions were used to thermal treat the TBCs (eg.
EB-PVD was applied at 1000°C, while furnace temperature of 1150°C was used for
thermal cycling). The coating thickness hc is taken as the sum of YSZ thickness
(175m) and TGO thickness (given in Figure 4.9). The Poisson’s ratio is taken as
0.3 is this case. The values of Ec and hc are summarized below for 0, 30, 50, and
Page 107
107
100hrs isothermal exposures in Table 4.4 and 4.5, respectively. Figures 4.16 shows the
crack radius of the five TBCs isothermally exposed to 0, 30, 50, and 100hrs. Note that
the crack radius is calculated as the sum of the crack lengths (given in Figure 4.8) and
half-diagonal of the indent marks. The central out-of-plane displacement, u, cannot be
measured accurately, since the displacement consists of both elastic and plastic
deformation. Hence, it was first approximated as the half-diagonal of the indentation
mark. The calculated Gc values, however, were many orders of magnitude higher than
the values given in previous literatures [54,82] (Table 4.6), thus suggesting that
half-diagonal of the indent mark is an overestimation for u.
Table 4.4
The effective in-plane Young’s modulus of the bi-layer coating (YSZ-TGO) after
different isothermal exposures (These values were estimated from Fig. 6 in Ref.[106])
Isothermal hrs 0 30 50 100
Ec (GPa) 7.5 13 18.5 13.5
Table 4.5
The coating thicknesses (YSZ+TGO thickness) values used for the calculation of Gc
in this work
hc (m)
Isothermal hrs 0 30 50 100
CMSX-4
LT Pt-Al 175 177.42 177.73 177.82
TMS-82+
LT Pt-Al 175 177.49 177.81 179.34
Page 108
108
TMS-82+
HT Pt-Al 175 177.44 177.74 178.92
TMS-82+
Pt-diffusion 175 177.45 177.54 178.40
SRR99
LT Pt-Al 175 178.15 -- --
Table 4.6.
A comparison between the Gc reported in Ref. [82] and the TMS-82+ Pt-diffusion
here by approximating u as half-diagonal of the indent mark.
Gc (J/m-2
)
Isothermal hrs 0 30 50 100
TMS-82+
Pt-diffusion 2.5x10
8 2.7x10
6 1.2x10
6 1.2x10
6
Thermal cycles
(1-hr cycle) 2 30 50 100
CMSX-4
Pt-diffusion 2300 80 42 42
Page 109
109
Fig. 4.16. The crack radius (i.e. sum of the length of half indent and crack lengths)
values used in the calculation of Gc in this work.
In order to get a more reasonable value of Gc, u value should be much smaller than
the half-diagonal of the indent. From SEM observation, the diamond shape indent
caused mostly inward displacement (i.e. plastic deformation) of the YSZ with
minimal central out-of plane (i.e. parallel to YSZ columns) deformation. By applying
an opening tensile force at the TGO/bond coat interface using indentation, there is
very little upward movement in the YSZ top coat due to its porous/soft structure.
However, the bond coat likely displaced downward, for which the displacement u can
be roughly approximated as the crack openings at the TGO/bond coat interface near
the two corners of the indent mark (Fig. 4.17). 6 measurements (i.e. 3 from each of
the corner parallel to interface) were taken from a single indent, and used for the
Page 110
110
averaging of upward displacements. Figure 4.18 shows the u values of the five TBCs
isothermally exposed to 0, 30, 50, and 100hrs. Finally, the interfacial fracture
toughness is measured by using all the above attributes and summarized in Figure
4.19. There seems to be a wide scattering of Gc values between the five TBCs, which
follow a similar trend to that of the crack length results shown earlier. The Gc values
given here for the as-coated TBCs (i.e. 0hr) seem to be much lower than those
reported in the work by Wang et al [82]. In that work, the Gc for their CMSX-4
Pt-diffusion specimens exposed to only 2 thermal cycles had a much larger value
(2300Jm-2
). It is possible that the u values (taken as the crack openings due to
downward displacement of bond coat) for the as-coated TBCs were significantly
underestimated for the calculation of Gc in the case here. Despite using a load two
times greater (9.8N), the u values (i.e. crack openings) are roughly the same if not
lower than those of later isothermal hours (i.e. 30-200hrs), which were indented with
4.5N. As shown in Figure 4.20, there seems to be significantly more damage in the
as-coated YSZ in the form of inter-columnar fracture and shear displacement than the
same coating after 30 hrs of isothermal exposure. This can be explained by the
sintering of the YSZ top coat after prolonged thermal exposure, which prevented the
stress generated by the indentation from dissipating in the form of YSZ damage in
isothermally exposed TBCs.
Page 111
111
Fig. 4.17. Micrograph showing the location of the u measurements (6 per indent).
Fig. 4.18. The upward displacements (i.e. taken as the crack openings at the
TGO/bond coat interface near the corner of indents) used in the calculation of Gc in
this work.
Page 112
112
Fig. 4.19. The approximated values of Gc for the five TBCs based on the semi-cirular
plate model.
Page 113
113
Fig. 4.20. Micrographs showing the extent of YSZ damage between as-coated TBC
and TBC subjected to 30hrs of isothermal exposure.
Other uncertainties may also contribute to the discrepancy between the Gc values
obtained here and those given by other literatures. For instance, the Ec values taken
from Wang’s work [106] were measured on specimens that underwent thermal cycling
for the same amount of time instead of isothermal exposure as in this work. There
would certainly be a difference in the Ec values between isothermal exposed and
thermal cycled TBCs of the same duration, due to different amount of top coat
sintering and residual stresses present in the top coat and TGO. In addition,
differences in coating deposition methods, coating thicknesses, bond coat structural
change, and phase transformation may also be sensible reasons for the discrepancy in
the Gc values, as these differences can change any of the inputs in Eq. 4-1.
Nevertheless, the interfacial fracture toughness values calculated here for the
TMS-82+ Pt-diffusion are fairly similar to that of the CMSX-4 Pt-diffusion specimen
Page 114
114
reported in Wang’s work [82]. The Gc values obtained for the 30, 50, and 100hrs
isothermally exposed TMS-82+ Pt-diffusion specimens are 62, 31, and 26 J/m-2
,
respectively, which are fairly close to the values obtained by Wang as shown in Table
4.6.
4.4.2 Linear fitting of crack rate with degradation factors
In order to compare interfacial crack resistance under indentation for different
samples, linear fitting has been done on crack length versus thermal exposure time
(hrs) and then the slope of each fitting has been defined as the crack rate. A higher
slope means a higher crack tendency. Therefore, it should be noted that the crack rate
(m/hr) is not a measure of the crack propagation speed under indentation, instead, is
used for the comparison of crack tendency among different samples under indentation.
In addition, TGO thickness growth rate, TGO length growth rate (rumpling) and
’phase formation rate are obtained based on plots in Fig. 4.9, 4.11 and 4.12 where
’phase formation rate was obtained based on initial treatment time only. As the TGO
stresses do not vary significantly during thermal treatments (Fig. 4.10), therefore
average stress value for each sample during the thermal treatment period is taken for
comparison of TGO stresses in different samples. It should be noted again here that
the purpose to define these parameters is not to obtain absolute value of each
parameter, instead, it is to compare properties of different samples, i.e. to identify
effect of different factors on crack resistance and lifetime of TBCs. Table 4.7 gives
ranking of crack rate, TGO thickness growth rate, TGO length growth rate (rumpling),
TGO stresses and ’phase formation rate for the 5 different samples. In addition,
quantitative values for these parameters are also given in histograms for the 5
Page 115
115
different samples (Fig. 4.21-4.25).
Table 4.7
The comparison between the rankings of performance and contribution factors studied,
showing an exact match between performance and the ’ formation factor
Performance Contributing Factors
Best(Factors:
low)
Worst
(Factors:
high)
Crack rate TGO growth
rate
TGO Stress TGO
rumpling rate
’ formation
rate
CMSX-4 LT
Pt-Al
CMSX-4 LT
Pt-Al
SRR99 LT
Pt-Al
TMS-82+
HT Pt-Al
CMSX-4 LT
Pt-Al
TMS-82+ LT
Pt-Al
TMS-82+ Pt
diffusion
TMS-82+ LT
Pt-Al
TMS-82+ Pt
diffusion
TMS-82+ LT
Pt-Al
TMS-82+
Pt-diffusion
TMS-82+
HT Pt-Al
CMSX-4 LT
Pt-Al
CMSX-4 LT
Pt-Al
N/A
TMS-82+
HT Pt-Al
TMS-82+ LT
Pt-Al
TMS-82+
HT Pt-Al
TMS-82+ LT
Pt-Al
TMS-82+ HT
Pt-Al
SRR99 LT
Pt-Al
SRR99 LT
Pt-Al
TMS-82+ Pt
diffusion
SRR99 LT
Pt-Al
SRR99 LT
Pt-Al
Page 116
116
Fig. 4.21. A quantitative comparison of crack rates amongst the five TBC specimens.
Fig. 4.22. A quantitative comparison of TGO growth rates amongst the five TBC
specimens.
Page 117
117
Fig. 4.23. A quantitative comparison of average TGO stresses amongst the five TBC
specimens.
Fig. 4.24. A quantitative comparison of TGO rumpling rates amongst the five TBC
Page 118
118
specimens.
Fig. 4.25. A quantitative comparison of ’phase formation rates amongst the five TBC
specimens.
Fig. 4.26 gives the crack rate as function of ’phase formation rate, TGO thickness
growth rate, TGO length growth rate (rumpling rate), and TGO average stress. Both
the ’phase formation rate and TGO thickness growth rate appear to correlate with the
crack rate (i.e failure of TBCs) in a linear relationship. In contrast, the TGO stress
shows no obvious correlation with the crack rate. Finally, the TGO rumpling
measured based calculating the growth rate of TGO length (i.e. rumpling rate) also
has no apparent correlation to TBC failure, which appears to contradict previous
studies [107,108].
Page 119
119
Fig. 4.26. Plots showing the linear dependency of crack rate on a) ’phase formation
rate, b) TGO thickness growth rate, c) TGO length growth rate (rumpling rate) and, d)
TGO average stress. The crack rate only increases linearly with the rate of ’phase
formation.
4.4.3 Possible degradation mechanisms by ’ formation rate
Although an obvious correlation can be seen between the ’ formation rate and the
crack rate, there is no clear explanation regarding how the early ’ formation may
Page 120
120
have contributed to the loss of interfacial adhesion at this point. It is speculated that
the presence of ’ phase (rich in Ni and Ti) near the interface could either hinder the
formation of alumina, or may alternatively form titanium oxide. These observations
suggest that the two weakest samples: SRR99 LT Pt-Al and TMS-82+ HT Pt-Al
specimens had undergone a chemical diffusion that was quite different from other
specimens, and that the formation of Ni and Ti enriched ’ phase either directly or
indirectly reduced the TGO adhesion. Further chemical diffusion analysis is necessary
to fully understand the effect of Ni and Ti-containing ’ phase on the interfacial
adhesion.
It has been known that fast growing Co and Ni-rich oxides could form as TGO aside
from the alumina growth, and are more pronounced in specimens with lower
aluminum concentration profile in their bond coats. This failure mechanism may have
been closely linked to the early ’ formation, which is reflected by the adjusted
r-square values in the linear fit of Fig. 4.26.
4.4.4 Stress relaxation by rumpling
A surprising phenomenon was observed from the crack length measurements for
CMSX-4 LT Pt-Al and TMS-82+ HT Pt-Al, of which fluctuation in crack length was
found. The residual stress shown earlier indicated that both the CMSX-4 LT Pt-Al and
TMS-82+ HT Pt-Al specimens underwent a stress decline from 30 to 100 hrs and 30
to 80 hrs respectively (Fig. 4.10). Interestingly, declining trends in crack length were
also found at the same time intervals on those two specimens. This phenomenon was
not observed in the other three specimens, for which their corresponding crack length
Page 121
121
all increased with thermal exposure. Further, quantification of the bond coat/TGO
interfacial rumpling showed that CMSX-4 LT Pt-Al and TMS-82+ HT Pt-Al are the
only specimens undergoing further rumpling between 30 and 50 hrs. By taking into
account this increasing rumpling behaviour with the corresponding stress relaxation
shown earlier, it is believed that the observed fluctuation in crack length may have
been due to the relaxation of TGO residual stress by rumpling [28,30,31].
Fig. 4.27 shows that average TGO stress reduces with increase in rumpling rate,
which suggests that TGO rumpling can relax TGO residual stress. It has been shown
by geometric factors that a greater out-of plane tensile stress tends to accumulate at
the protuberance of the TGO/bond-coat interface, and at the YSZ side of the interface
along the flanks of the protuberance for specimens with rougher interface [109,34].
This claim is supported by the fact that specimens with increased interfacial
roughness tend to fail not only at TGO/bond coat, but also at TGO/YSZ interfaces
[83]. While such an introduction of stresses along the interfaces is undoubtedly
detrimental, our experiment results have shown several cases of a specimen with
higher degree of rumpling, yet its interfacial adhesion and lifetime were much better
relative to those with less rumpling.
Page 122
122
Fig. 4.27. The dependency of stress on the rumpling increase rate (highly linear
dependent).
Traditionally, both residual stress and rumpling at the TBC interface were considered
as important factors attributing to the TBC failure. However, in this study, a direct
correlation could not be established between these two factors and the TBC spallation
life (Fig. 4.26). The CMSX-4 LT Pt-Al specimen has been shown to possess the best
crack resistance amongst all the specimens, despite having a moderate TGO stress and
a moderate rumpling rate (Table 4.7). The reason behind this excellent performance
may be largely due to a slow ’ formation. Up to 100hrs, the CMSX-4 specimen had a
marginally less ’ formation than both the TMS-82+ LT Pt-Al and HT Pt-Al
specimens. Beside the kinetics of ’ formation, the TGO growth rate also seemed to
bear some influence on the crack resistance. Throughout the entire thermal exposure,
CMSX-4 specimens demonstrated the slowest TGO growth kinetics relative to
specimens of other substrates, although the difference was much less obvious than in
Page 123
123
the case of ’ formation.
4.4.5 The dynamic feature of interfacial fracture toughness
One key implication of this work is that the TBC spallation lifetime is dependent on
both the variation of driving force G for spallation and interfacial fracture toughness
Gc. This concept had been initially proposed by Wu et al [35], arguing that the focus
of TBC modeling had been largely placed on the origin of the driving force for
spallation, without thoroughly considering the dynamic nature (i.e. variation) of the
interfacial fracture toughness with respect to the substrate chemistry. In that original
work, this dynamic nature was indirectly demonstrated based on preliminary
observation of the TBC spallation lifetime. The current work, in contrast, has clearly
demonstrated the sensitivity of the interfacial fracture toughness to substrates and
bond coat systems. This can be illustrated in Fig 4.28, a semi-quantitative plot of the
variation of the driving force G (steady state) for spallation and interfacial fracture
toughness Gc of TBC specimens between 30 and 100hrs isothermal exposure. The
driving force G, in this case, is calculated by the equation for thin film of thickness h
on a thick substrate and the case of edge delamination [110,111],
(Eq. 4-2)
where E and are the Young’s modulus and Poisson’s ratio of the film, taken as
380GPa and 0.25, respectively. The value is taken as the TGO stress measured in
section 4.3.3. One can see that the TMS-82+ HT Pt-Al having a drastic decline in
interfacial fracture toughness Gc along with its rapid increasing driving force G after
Page 124
124
80hrs of isothermal exposure, resulted in the early failure of the specimen. On the
other hand, TMS-82+ LT Pt-Al still had vast remaining life as shown by the difference
in energy △G2. Note that the Gc values are roughly approximated by taking the ratios
of crack lengths between the TMS-82+ HT Pt-Al and other specimens at different
times, and multiplying them by the interfacial adhesion value Gc of TMS-82+ HT
Pt-Al, while assuming that the Gc value is 20% higher than its steady state energy
release rate G at 100hrs just prior to failure. Hence, the plot does not necessary
represent the true values obtained using proper testing procedures. Note that the rising
and dropping trend seen in this plot for TMS-82+ HT Pt-Al is due to the fluctuating
crack lengths obtained previously.
Fig. 4.28. A semi-quantitative plot illustrating the variation of driving force G (steady
state) and fracture toughness Gc between 30 and 100 hrs of isothermal exposure. △G
gives an indication of the remaining spallation life in the TBC.
Page 125
125
4.5 Conclusion
The results of this work have led to the following conclusions:
The cross-sectional indentation technique has been shown as an effective way to
give a semi-quantitative indication of the TBC interfacial adhesion. The crack
propagated mostly along the TGO/bond coat interface, and the results
correspond well with the known spallation lifetimes of commercial TBCs.
Among the 5 TBC samples, the growth rate, rumpling, and residual stress of
TGO as well as the phase distribution of the bond coat have been compared in
relation to the TBC adhesion.
An approximation of the interfacial fracture toughness, Gc, was carried out based
on the semi-circular plate model by Sanchesz et al [112] and the later work by
Wang et al [82]. Some of the calculated results, particularly for the as-coated
TBCs, were many orders of magnitude lower than those reported in previous
work [55,75,82]. Nonetheless, the Gc values for the TMS-82+ Pt-diffusion
specimens in this work (i.e. 30-100hrs isothermally exposed) were determined
to be similar to that of the CMSX-4 Pt-diffusion in Wang’s work [82].
The bond coat phase distribution, i.e. to ’ transformation due to depletion of
Al, appears to be the dominant factor in degrading the interfacial adhesion of
the specimens studied while Ni and Ti concentrate in the ’ phase. In addition,
the TGO growth rate also contributes to the degradation of the TBC adhesion.
It is speculated that the highly concentrated Ni and Ti-containing ’ formation
attached to the interface can either hinder the stable formation of alumina or
alternatively form titanium oxide, which would weaken the TBC interface.
Page 126
126
The experimental results have shown that interfacial adhesion not only varies
with different materials, but also depends dynamically with the thermal
exposure history of oxidation. However, there is no standard procedure in the
measurement of Gc so far, resulting in large disparity across the values
obtained by different research groups.
Page 127
127
CHAPTER 5
The degradation mechanisms of TBCs: Driving force for
spallation versus interfacial fracture toughness Gc
5.1 Introduction
For achieving a balance between fuel economy and optimal thrust power in modern
jet engines, there is an ongoing pursue for a higher turbine inlet gas temperature [6].
In order to cope with the ever-demanding operating conditions, state-of-the art
thermal barrier coatings (TBCs) based on yttria-stabilized zirconia (YSZ) have been
widely applied on top of the bulk Ni-based superalloy, to protect the underlying alloy
from approaching its melting temperature. Due to the porous columnar structure and
high ionic diffusivity of YSZ in the ceramic top coat at high temperature, the ingress
of oxygen can readily oxidize the underlying substrate, leading to catastrophic failure
of the turbine structure. To address this issue, an intermediate bond coat layer,
commonly made by either a thermally sprayed MCrAlX (M=Ni or combination of Ni
and Co, and X indicated the minor element addition) or electrodeposition of platinum
aluminide, is applied between the ceramic top coat and the bulk superalloy. The bond
coat contains a relatively higher amount of Al content than the bulk alloy, and hence,
serves as a reservoir of Al to promote the preferential formation of a protective
alumina as the thermally grown oxide (TGO) layer [14,15,16].
Despite these protective measures, the TBCs are still prone to spallation failure
primarily driven by the interfacial rumpling of the bond coat during the course of high
Page 128
128
temperature oxidation and thermal cycling [90,99,23,24]. The general consensus for
the cause of interfacial rumpling is thought to be the mismatch in the coefficient of
thermal expansion between the multilayered structures of TBCs [23,24]. Despite
considerable progress that has been made on modeling the interfacial evolution of
TBCs in terms of the stress state and the degree of rumpling, there has been a lack of
understanding on how these values could be influenced by the chemical diffusion and
phase transformation of bond coat and substrate during thermal cycling
[23,24,34,103]. It is only in recent years that more and more studies have focused on
comparing and understanding the influence of bond coat and substrate compositions
on the degradation of TBCs [26,35,95,101,102].
As identified in the previous chapter, the rate of to ’ transformation driven by Al
depletion was determined as a dominant factor in degrading the interfacial adhesion of
the TBC samples studied. It was speculated that the ’ formation containing highly
concentrated Ti near the bond coat/TGO interface may weaken the interface by
forming TiO2. However, preliminary analysis by electron probe microscope analysis
(EPMA) indicated that while traces of TiO2 were present in the TGO layer of the
SRR99 LT Pt-Al specimen, the same phenomenon could not be observed on the
TMS-82+ HT Pt-Al specimen, which also had similar Ti enrichment, after the ’
formation near the bond coat/TGO interface (see Figure 5.1). This implies that the
formation of less adhesive oxide cannot be the only mechanism causing the interfacial
degradation of TBCs. Both SRR99 LT Pt-Al and TMS-82+ HT Pt-Al showed very
poor interfacial adhesion according to the results of the previous chapter.
Page 129
129
Fig. 5.1. EPMA mapping showing the formation of Ti concentrated ’ in the bond coat
region near the oxide-bond interface of the SRR99 LT Pt-Al (containing traces of
TiO2 in the TGO) and TMS-82+ HT Pt-Al after 30 and 100hrs of isothermal exposure,
respectively.
The current study is therefore, devoted to clarifying the difference in the degradation
mechanisms between the SRR99 LT Pt-Al and TMS-82+ HT Pt-Al specimens. A
systematic approach similar to the previous chapter, was carried out by examining
five different industrial grade TBCs based on 3 different bond coats and substrates.
Electron backscatter diffraction (EBSD) pattern maps near the interface were used in
conjunction with high resolution optical micrographs, EPMA chemical mappings, and
nano-indentation hardness values to elucidate the underlying mechanisms behind the
early spallation of TBCs. The results have clearly indicated that the TBC interfaces
may degrade not only by the formation of less adhesive oxides (i.e. TiO2), but also
due to the misfit stress induced in the TGO layer resulting from the mechanical
mismatch between the different phases during cooling. Moreover, the results here
have clearly demonstrated the previously proposed concept that the TBC spallation
Page 130
130
life is dependent on the dynamic variation of both the driving force for spallation G
and interfacial fracture toughness Gc .
5.2 Experimental details
The same set of TBCs, consisting of three industrial standard single-crystal nickel
based superalloys: CMSX-4, SRR99, and TMS-82+, used in the previous chapter, was
utilized in this current work. Table 5.1 shows the nominal chemical composition for
these alloys. These fully heat treated alloys (solution and primary aged) were
processed using conventional investment casting methods into cylindrical rods (10
mm diameter), having the long axis aligned closely to the {001} direction. These rods
were further sliced into disk shape (10 mm diameter and 4 mm thickness) and spot
welded onto Nimonic sticks. On the other hand, three different bond coats, namely
HT Pt-Al (where HT implies a high temperature and low activity aluminization
process), LT Pt-Al (where LT implies a low temperature and high activity
aluminization process), and Pt-diffusion (without aluminization), were subsequently
deposited onto the flat surfaces of the button-shaped discs. The TMS-82+ alloy was
paired with all three bond coats for the purpose of studying the effect of bond coat,
while CMSX-4 and SRR99 were only paired with LT Pt-Al bond coat to study the
influence from superalloy substrate, when the type of bond coat is fixed. Table 4.2
below provides a summary of the five different TBC systems used in this work.
Page 131
131
Table 5.1
Nominal compositions of the three superalloys used in this work (wt%)
Substrat
e
Co Cr Mo W Al Ti Ta Hf Re Ru C Ni
SRR99 5.0 8.0 - 9.5 5.5 2.2 2.8 - - - - Bal.
TMS-82
+
7.8 4.9 1.9 8.7 5.3 0.5 6.0 0.1 2.4 - - Bal.
CMSX-
4
9.6 6.5 0.6 6.4 5.6 1.0 6.5 0.1 3.0 - - Bal.
Table 5.2
Visual representation of the five TBC systems used in this work
SUPERALLOY SUBSTRATE
BOND COAT CMSX-4 TMS-82+ SRR99
HT Pt-Al ✔
LT-Pt-Al ✔ ✔ ✔
Pt diffusion ✔
The HT Pt-Al and the LT Pt-Al bond coat specimens were prepared by first
electrodepositing a thin layer of platinum of 5 and 7m, respectively followed by a
vacuum heat treatment at 1100°C for 1 h. Next, 5hrs vapour phase, and 20hrs pack,
aluminization processes were applied at 1080°C and 870°C, respectively. The Pt
diffusion bond coat specimen was first electroplated with a 10m layer of platinum
and then vacuum heat treated at 1150°C for 1hr without the aluminization process. A
further heat treatment in argon atmosphere was applied on all specimens for 1hr at
1100C, prior to the deposition of a ZrO2/7wt% Y2O3 (YSZ) top coat of approximately
175m in thickness by electron beam physical vapor deposition (EB-PVD). Lastly, a
vacuum heat treatment (1100 °C, 1hr) and ageing process (870°C, 16hrs) were applied
Page 132
132
to all specimens for quality assurance purposes.
A total of 25 specimens (5 for each TBC system) were subjected to isothermal
exposure at a furnace temperature of 1135°C for 0, 30, 50, 100 hrs. These specimens
were then mounted in a StruerCitovac™ vacuum imprenation chamber using a
low-shrinkage epoxy resin in order to preserve the specimen edges, and polished to
mirror finish for microstructural characterization. A high resolution KEYENCE
VHX-1000 optical microscope was used to observe the microstructural features on the
entire cross-section of all the specimens. Micrographs across the entire TGO/bond
coat interfacial area were taken at 5000x magnification for every specimen. High
resolution quantitative elemental analysis was carried out at the TGO/bond coat
interfacial area of all the thermally exposed specimens using EPMA (Electron Probe
Microscope Analysis). A beam current of 20nA was applied with a beam size of about
1m, at an accelerating voltage of 15kV. The scans were done at a step size of 0.1m
in both X and Y directions for an area of 30.00m x 25.00m.
In order to conduct Orientation Imaging Microscopy (OIM) analysis, an additional
polishing step was applied to the specimens using the 0.06 m Buehler Mastermet
colloidal silica suspension to ensure the removal of surface deformation, which is
essential for EBSD analysis. A JEOL FE-SEM equipped with EBSD detector and
OIM data collection software was used to carry out the orientation analysis near the
bond coat/TGO interfacial region of the 0, 30, and 100hrs specimens. The scans were
done at a step size between 0.4 to 0.6 m in both X and Y directions, for an area of
28.000 m x 25.000m. For specimens that had inherently large grains, an additional
scan was conducted using the same step size and area settings as before.
Page 133
133
Nano-indentation (MTS nano-indentation XP) was carried out along the
cross-sectional surface of bond coat for every specimen. A total of 10 indentions were
conducted at locations approximately 15m below the TGO/bond coat interface on
every specimen, and were separated by a distance of at least 20m from each other.
The indenter load was controlled using a depth limit of 1000nm and a peak holding
times of 10s. The hardness results were used to elucidate the mechanical response of
the bond coat after the cooling stage.
For the purpose of carrying out a more realistic Thermo-Calc analysis of the phase
distribution in the bond coat at the TGO/bond coat interface, energy-dispersive X-ray
spectroscopy (EDX) was utilized to determine the major alloying compositions near
the oxide-bond coat interface (unlike the approximating approach used in the previous
chapter) of the 100hrs isothermally exposed CMSX-4 LT Pt-Al and the 50hrs
isothermally exposed TMS-82+ LT Pt-Al specimens. A total of 23 point scans were
conducted at locations approximately 1-3m below the TGO/bond coat interface on
both specimens in order to avoid collecting secondary electron signals from the TGO
layer (i.e. oxygen and aluminum). Average values of the alloying elements in wt%
were obtained and used in the Thermo-Calc simulation of section 5.3.5. Table 5.3
shows a comparison between the bond coat composition near the TGO/bond coat
interface of the as-coated CMSX-4 LT Pt-Al specimen using the estimation approach
(i.e. Chapter 4) and actual EDX scan (i.e. this chapter).
It can be clearly seen that the estimating approach by proportionally decreasing other
elements in the previous chapter (i.e. dilution effect of aluminization) is flawed, as the
diffusivities tend to vary significantly between different elements. For instance, the W
Page 134
134
and Ta contents were significantly reduced/diluted in the actual bond coat as
compared to the bulk CMSX-4, while Mo and Ti maintained similar contents to that
of the bulk CMSX-4. In addition, it was assumed that only Al and Ni content varied in
the bond coat due to oxidation, while the rest of the elemental compositions
maintained the same as in their as-coated state. This is definitely not realistic, as
inter-diffusion of certain elements is known to take place during oxidation. It should
be noted; however, that the previous chapter was mainly looking at the relative
amount of Ti solubility in and ’ phases of bond coat. Hence, the discrepancy in the
bond coat compositions should not have a major influence in the general distribution
trend.
Table 5.3
A comparison between the estimated and actual (EDX scanned) bond coat
compositions in as-coated CMSX-4 specimen
Bond coat
composition
Co Cr Mo W Al Ti Ta Re Pt Ni
CMSX-4 LT
Pt-Al
(estimated)
8.31 5.63 0.52 5.54 19.00 0.09 5.69 2.60 N/A Bal.
CMSX-4 LT
Pt-Al
(scanned)
4.96 2.57 0.50 0.93 20.0 0.14 0.76 0.77 30.7 36.1
Substrate
composition
Co Cr Mo W Al Ti Ta Re Pt Ni
CMSX-4 9.6 6.5 0.6 6.4 5.6 1.0 6.5 3 - Bal.
Page 135
135
5.3 Results
5.3.1 Microscopic observation near the TGO/bond coat interface
Figure 5.2 shows a side by side comparison of the bond coat regions between two 30
and 50hrs isothermally exposed superalloys (SRR99 and CMSX-4) paired with the LT
Pt-Al bond coats. It can be seen that the phase morphology differs significantly
between the two alloys. The SRR99 LT Pt-Al specimen showed short circuit diffusion
along the grain boundaries of grains, and formed a secondary phase of ’
precipitates after 30hrs. In addition, large ’precipitates were also observed near the
TGO/bond coat interfacial region. The corresponding 30hrs isothermally exposed
CMSX-4 LT Pt-Al specimen, on the other hand, maintained a uniform layer of
phase without the presence of ’ precipitates at. After 50hrs of isothermal exposure,
spallation of the TGO had already occurred on the SRR99 LT Pt-Al specimen, while
the TGO layer of the CMSX-4 LT Pt-Al specimen remained intact. Moreover,
CMSX-4 LT Pt-Al was found to contain the martensitic phase (lath structures) along
with phase.
Page 136
136
Fig. 5.2. SEM micrographs showing a comparison of bond coat regions between 30
and 50hrs isothermally exposed SRR99 LT Pt-Al and CMSX-4 LT Pt-Al specimens,
with a unique ’ formation at the grain boundaries of the SRR99 specimen.
Similarly, Figure 5.3 shows a side by side comparison of the bond coat regions
between 30 and 50hrs isothermally exposed HT Pt-Al and LT Pt-Al bond coats paired
with the same TMS-82+ superalloy. There was a significant difference in the phase
distribution of these two bond coats. The TMS-82+ HT Pt-Al specimen was found to
contain a uniform distribution of martensitic phase (coarse lath structures); while the
TMS-82+ LT Pt-Al specimen maintained mostly phase with minimal formation of
martensites and ’precipitates after 30hrs of isothermal exposure. For the 50hrs
isothermal exposed specimens, large pieces of ’ precipitates began to form on the
TMS-82+ HT Pt-Al specimen beside the martensitic phase. The TMS-82+ LT-Pt-Al,
on the other hand, began to show more distribution of martensites and some ’
precipitates along with the phase.
Page 137
137
Fig. 5.3. SEM micrographs showing a comparison of bond coat regions between 30
and 50 isothermally exposed TMS-82+ HT Pt-Al and TMS-82+ LT Pt-Al specimens,
with a complete martensitic transformation in the HT-Pt-Al bond coat.
5.3.2 EBSD mapping analyses near the TGO/bond coat interface
EBSD pattern maps are utilized in this section to further highlight the unique features
in bond coat morphology and phase distribution of the TMS-82+ HT Pt-Al and
SRR99 LT Pt-Al specimens, as the EBSD map is able to indicate the grain boundaries
by distinguishing the orientation of individual crystals. It should be noted here that the
EBSD pattern of the TGO layer could not be accurately indexed (most likely due to
its large residual stress induced crystal distortion), and hence, was deliberately
removed from all maps for the ease of viewing. Figure 5.4 shows a side by side
comparison of the grain morphology between the HT Pt-Al and LT Pt-Al specimens
Page 138
138
paired with the same TMS-82+ superalloy, in the as coated condition. Before
isothermal exposure, the TMS-82+ HT Pt-Al specimen was found to contain much
larger grains than the TMS-82+ LT Pt-Al bond coat. The larger grains can be
explained by the fact that the aluminization processes of the TMS-82+ HT Pt-Al were
carried out at a relatively higher temperature of 1080°C in a low activity aluminum
environment. Based on the fundamental physics of crystallography, grain growth
tends to be much more rapid due to higher diffusivity of atoms at high temperatures.
Fig. 5.4. EBSD mapping showing a comparison of grain morphology of the bond coat
region near the oxide-bond interface of the TMS-82+ HT Pt-Al and TMS-82+ LT
Pt-Al prior to isothermal exposure, with the HT Pt-Al bond coat having much larger
grains.
A similar comparison (Figure 5.5) of 30hrs isothermally exposed specimens shows
that the TMS-82+ HT Pt-Al bond coat transformed completely from the as coated
phase to an acicular martensitic phase, while the TMS-82+ LT Pt-Al specimen
possessed mostly the phase having enlarged grain size. A significant degradation
can be seen in the pattern quality of the 30hrs isothermally exposed TMS-82+ HT
Pt-Al most likely due to crystal distortion due to associated with the significant
Page 139
139
martensitic phase transformation. Figure 5.6 provides a side by side comparison of the
phase distribution between the SRR99 LT Pt-Al and CMSX-4 LT Pt-Al specimens
after 30hrs isothermal exposure. The SRR99 LT Pt-Al specimen was found to exhibit
the formation of ’ phase at the grain boundary and within the matrix of phase as
secondary precipitate. The CMSX-4 LT Pt-Al map, on the other hand, showed very
uniform grains of clean phase.
Fig. 5.5. EBSD mapping showing the lath martensitic structure of the TMS-82+ HT
Pt-Al specimen, in contrast to the phase of the TMS-82+ LT Pt-Al after 30hrs of
isothermal exposure.
Fig. 5.6. EBSD mapping showing the unique ’ precipitation at the grain boundaries
of the SRR LT Pt-Al, in contrast to the clean and uniform phase of the CMSX-4 LT
Page 140
140
Pt-Al specimen after 30hrs of isothermal exposure.
Figure 5.7, 5.8 and 5.9 show a comparison between the phase distribution of CMSX-4
LT Pt-Al, TMS-82+ LT Pt-Al and TMS-82+ HT Pt-Al after 100hrs of isothermal
exposure. Note that an electron microscopic image is provided to accompany each
individual EBSD map, in order to help identifying the different phases present. The
CMSX-4 LT Pt-Al specimen experienced a transformation from phase to a mostly
coarse martensitic structure along with minor amount of ’phase. The TMS-82+ LT
Pt-Al specimen, on the other hand, showed the presence of more ’phase along with a
coarse martensitic structure. The pattern quality of the martensitic region is
significantly lower relative to the neighboring grains of ’phase. Lastly, the formation
of much larger ’grains could be found in the TMS-82+ HT Pt-Al specimen with
patterns of martensitic regions below recognition.
Fig. 5.7. EBSD mapping showing the degradation in pattern quality of martensitic
region relative to the amount of ’ formation in the bond coat of 100hrs isothermally
exposed CMSX-4 LT Pt-Al.
Page 141
141
Fig. 5.8. EBSD mapping showing more severe degradation in pattern quality of
martensitic region with increasing amount of ’ formation in the bond coat of 100hrs
isothermally exposed TMS-82+ LT Pt-Al 100hrs.
Fig. 5.9. EBSD mapping showing very severe degradation in pattern quality
associated with even more ’ formation in the bond coat of 100hrs isothermally
exposed TMS-82+ HT Pt-Al.
The EBSD map of the 100hrs isothermally exposed TMS-82+ Pt-diffusion specimen
is compared to that of the 30hrs isothermally exposed SRR99 LT Pt-Al specimen to
highlight the inherent difference between the and ’ structure of the Pt-diffusion
bond coat and the structure of the LT Pt-Al bond coat (see Figure 5.10). It can be
seem that the TMS-82+ Pt-diffusion specimen consisted of small grains of more or
Page 142
142
less uniform sizes, whereas, the SRR99 LT Pt-Al specimen contained much larger
grains with ’ precipitates of various sizes. Figure 5.11 shows a comparison of the
misorientation profile of the bond coat grains near the TGO/bond coat interface
between the TMS-82+ Pt-diffusion and SRR99 LT Pt-Al specimens. The arrow lines
in Figure 5.10 indicate the location where the line plot was taken from, starting from
the left arrow toward the right one. The flat regions of the plot indicate the location
and width of each individual grain, as the crystal orientation of should be consistent
throughout a grain. The TMS-82+ Pt-diffusion specimen consists of grains having
consistent size and orientations, while sudden fluctuation in the orientation can be
seen within grain filled with many precipitates fine grains in the SRR99 LT Pt-Al
specimen.
Fig. 5.10. EBSD mapping highlighting the inherent difference between the ’
microstrucuture of the Pt-diffusion bond coat and the microstructure of the LT Pt-Al
bond coat.
Page 143
143
Fig. 5.11. A comparison of the misorientation profile near the oxide-bond coat region
between the TMS-82+ Pt diffusion (top) and SRR99 LT Pt-Al (bottom) after 100hrs
and 30hrs of isothermal exposure, with much sharper peaks in the SRR99 specimen
Page 144
144
indicating regions of ’ precipitation.
5.3.3 EPMA mapping analysis near TGO/bond coat regions
EPMA analysis was carried out to study the 30hrs and 100hrs isothermally exposed
TMS-82+ Pt-diffusion, in order to study the chemical diffusion and segregation
behavior of its bond coat (see Figure 5.12). After 30hrs of isothermal exposure, the Al
content near the TGO/bond coat region of the TMS-82+ Pt-diffusion specimen was
uniformly depleted with corresponding enrichment in the Cr content. The Ti content
was found to segregate below the Cr enriched layer at the exact location where Cr
depletion occurred. After 100hrs of isothermal exposure, the Ti and Cr chemistry
remained very similar to that of the 30hrs isothermally exposed while further Al
depletion took place. It should be noted that traces of chromium oxide can be found in
the TGO layer for both exposure times.
Page 145
145
Fig. 5.12. EPMA mapping showing the chemistry near the oxide-bond coat regions of
the TMS-82+ Pt-diffusion specimens after 30 and 100hrs of isothermal exposure, with
uniformly depleted layer of Al and Ti while enriched with Cr.
Figure 5.13 provides a direct comparison of the Ti and Cr distribution between the
TMS-82+ LT Pt-Al and the CMSX-4 LT Pt-Al specimens having isothermally
exposed for 100hrs. Similarly, Ti-Cr segregation was found in the TMS-82+ LT Pt-Al
specimen, except that the segregation occurred in small patches along bond coat
region near the TGO/bond coat interface rather than layers as in the previous case. On
the other hand, the CMSX-4 LT Pt-Al specimen showed a rather consistent Cr
distribution with much less Ti content near the interface. Chromium oxide was
detected in the TGO layers of both specimens.
Page 146
146
Fig. 5.13. EPMA mapping showing a comparison of Cr content near oxide-bond coat
interface between TMS-82+ LT Pt-Al and CMSX-4 LT Pt-Al after 100hrs of
isothermal exposure, with CMSX-4 showing no obvious Ti presence.
5.3.4 Hardness by nano-indentation near the TGO/bond coat interface
To study the mechanical property of the bond coat region near the TGO/bond coat
interface, hardness results measured by nano-indentation is plotted in Figure 5.14 here
for the 30, 50 and 100hrs isothermally exposed specimens. Note the error bar here
represents the standard deviation of 10 measurements for each specimen. After 30hrs
of isothermal exposure, the SRR99 LT Pt-Al specimen had the highest hardness and
standard deviation. In general, all specimens except the TMS-82+ Pt-diffusion,
underwent an increasing trend in both average hardness and standard deviation
Page 147
147
between 50 and 100hrs. The TMS-82+ HT Pt-Al had the highest hardness and
standard deviation among all specimens listed here after 100hrs of isothermal
exposure. The TMS-82+ Pt-diffusion specimen was found to maintain similar
hardness values and standard deviation going from 30 to 100hrs.
Fig. 5.14. A plot showing the hardness values by nano-indentation of various bond
coats near the oxide-bond coat interface, with increasing trend in both hardness values
and standard deviations.
5.3.5 Simulation of phase transformation by Thermo-Calc
To clarify the effect of Cr content on the phase distribution of bond coat, a
thermodynamic software, named Thermo-calc was used to identify the phase
Page 148
148
distribution of the LT Pt-Al bond coat system when prepared on the CMSX-4 alloy. A
set of parameters including the isothermal temperature used (1409K), and the bond
coat compositions near the TGO/bond coat interface of the 100hrs isothermally
exposed CMSX-4 LT Pt-Al specimen were specified in the calculation. We assume
that the system is in a stable equilibrium state after certain amount of time. Note that
the bond coat compositions were taken from the EDX analysis result as mentioned in
the experimental section (see Table 5.4).
Table 5.4
The nominal composition in wt% of the CMSX-4 LT Pt-Al and TMS-82+ LT Pt-Al
bond coats as specified in the Thermo-Calc calculation
Bond coat
composition
Co Cr Mo W Al Ti Ta Re Pt Ni
CMSX-4 LT
Pt-Al
6.691 3.158 0.502 1.998 11.117 0.0104 5.345 0 20.429 50.211
TMS-82+
LT Pt-Al
4.898 2.887 0.476 1.863 11.787 0.316 3.704 0 22.531 50.902
Figure 5.15a demonstrates the mole-fraction of different phases near the TGO/bond
coat interfaces of the 100hrs isothermally exposed CMSX-4 LT Pt-Al alloy based on
the calculation in Thermo-Calc as a function of Cr content. The results here showed
an increase in the mole fraction of phase corresponding to a decrease in the ’ with
increasing Cr content. The dotted line shows the relative distribution of the bond coat
phases in the 100hrs isothermally exposed CMSX-4 LT Pt-Al specimen. A similar
analysis was carried out to study the effect of Ta content on the phase distribution of
the 50hrs isothermally exposed TMS-82+ LT Pt-Al. The purpose here is to see if a
low Ta content, such as in the case of SRR99 LT Pt-Al specimen, could have an
Page 149
149
influence on the ’ transformation behavior. Again, the bond coat compositions used
for Thermo-Calc analysis were taken from EDX line scans near the oxide-bond coat
interface (see Table 5.4). Thermo-calc results given in Fig. 5.15b also indicated an
increase in the mole fraction of phase corresponding to a decrease in the ’ with
increasing Ta content. However, the extent of the phase variation seemed much less in
this case.
Fig. 5.15. The mole fraction of different phases existing near the TGO/bond coat
interface of the bond coats in a) CMSX-4 LT Pt-Al 100hrs, and b) TMS-82+ LT Pt-Al
50hrs specimens as a function of varying Cr and Ta contents, respectively. These
elements generally help stabilizing phase.
5.4 Discussion
5.4.1 Interfacial evolution during thermal exposure
Table 5.5 is a summary of the phase evolution of the five TBC systems with
corresponding hardness of bond coats between 30 and 100hrs isothermal exposures,
Page 150
150
and their TGO stress obtained from the previous chapter. The phase constituents of
the bond coat seem to have a huge influence on the TGO stress level. The TMS-82+
Pt-diffusion specimen, having pure phase between 30 and 100hrs isothermal
exposures, was measured to have the highest average hardness values, while having
the highest TGO stress throughout thermal exposure. The CMSX-4 LT Pt-Al
specimen with pure phase after 30hrs of isothermal exposure had a slightly lower
TGO stress than that of the TMS-82+ HT Pt-Al with all martensitic structure at 30hrs
(~2.29GPa vs. 2.45GPa).
The previous chapter indicated that the CMSX-4 LT Pt-Al and TMS-82+ HT Pt-Al
were the only specimens to have experienced a decline in TGO stress between 30 and
80hrs of isothermal exposure, and in the case of CMSX-4 LT Pt-Al, the decline
continued all the way to 100hrs. The TMS-82 LT Pt-Al, on the other hand, started out
with relatively lower TGO stress having multiple phases after 30hrs isothermal
exposure, and continued to show an increasing trend in TGO stress until 100hrs of
isothermal exposure. These phenomena suggests that a transformation from pure to
multi-phases can largely alleviate the TGO stress in the systems, despite having
high stress inducing phases such as martensites and ’.
Page 151
151
Table 5.5
A summary of the phase evolution, hardness, and TGO stress of all the bond coats
studied between 30 and 100hrs of isothermal exposure
Phase evolution, hardness, and TGO stress during thermal
exposure
30hrs 50hrs 100hrs
SRR99 LT Pt-Al ’ precipitates (g.b.
and near interface)
+
Stress : 0.237GPa
Hardness:
7.197GPa
-N/A (Spalled)
-N/A
TMS-82+ HT
Pt-Al
All lath martensitic
structure
Stress: 2.446GPa
Hardness:
4.555GPa
Martensitic + ’
(significant)
Stress: 2.327GPa
Hardness:
4.825GPa
Martensitic + ’
(large amount)
Stress: 2.485GPa
Hardness:
8.290GPa
TMS-82+ LT
Pt-Al
Mostly +
martensites + ’
(trace)
Stress: 1.183GPa
Hardness:
4.520GPa
martensites
(significant) + ’
(more trace)
Stress: 1.775GPa
Hardness:
4.679GPa
Martensitic +’
(significant)
Stress: 1.854GPa
Hardness:
5.322GPa
CMSX-4 LT Pt-Al All
Stress: GPa
Hardness:
5.167GPa
martensites
(significant)
Stress: 2.170GPa
Hardness:
4.564GPa
Martensitic + ’
(significant)
Stress: 2.091GPa
Hardness:
5.733GPa
TMS-82+
Pt-diffusion
Uniform layer
near the interface
Stress: 2.682GPa
Hardness:
6.553GPa
Uniform layer
near the interface
Stress: 2.840GPa
Hardness:
6.315GPa
Uniform layer
near the interface
Stress: 2.880GPa
Hardness:
6.41GPa
Page 152
152
5.4.2. Phase transformation due to substrate/bond coat chemistry
In TBCs, Al diffuses faster through the grain boundaries of the bond coat upon
forming Al2O3 during thermal exposure. However, the question still remains as to why
the formation of the ’ phase along the grain boundaries was exclusively observed on
the SRR99 LT Pt-Al in early isothermal times. As previously confirmed by EDX,
EPMA, and Thermo-Calc simulation, Ti and Ni contents have the tendency to
segregate to Al depleted region, corresponding to the location of ’ phase. The SRR99
LT Pt-Al specimen, having a relatively higher Ti and Ni content (see Table 5.1), was
previously measured to have the fastest TGO growth kinetic and the ’ phase
formation rate. Both Ni and Ti sped up the Al depletion process through the grain
boundaries of the SRR99 bond coat, thereby, migrated to the grain boundary to form
the ’ phase.
The unique phase transformation phenomena observed on the SRR99 LT Pt-Al
specimen is believed to be the cause of its relatively poor interfacial adhesion and
spallation lifetime. EPMA results in Figures 5.1 and 5.13 of the SRR99 LT Pt-Al and
TMS-82+ LT Pt-Al specimens, respectively, confirmed traces of Ti in the TGO layer,
while Ti rich ’ precipitates were present near the TGO/bond coat interface. The
detrimental effect of TiO2 formation near the bond coat surface, and subsequently,
degrading the adherence of the alumina scale is well documented [18,113,114]. The
same degradation mechanism, however, does not seem to apply in the case of the
TMS-82+ HT Pt-Al specimen. Despite having Ti containing ’ precipitates near the
TGO/bond coat interface, no obvious Ti was detected (see Figure 5.1) in the TGO
scale as confirmed at multiple locations along the interface. The amount of Ti in the ’
precipitates of the TMS-82+ HT Pt-Al was considerably lower than that in the SRR99
Page 153
153
LT Pt-Al specimen, which may be the reason behind the absence of TiO2.
Thermo-Calc simulation in section 5.3.5 indicated that Cr and Ta tend to stabilize the
phase, preventing the early formation of ’. However, the effect of Cr seemed to be
much greater than that of Ta. A comparison of the EPMA compositional mapping
between the TMS-82+ LT Pt-Al and CMSX-4 LT Pt-Al specimens (Figure 5.13)
revealed that Cr, if having sufficient concentration, can largely inhibit the diffusion of
Ti content outward to the bond coat surface. In fact, the CMSX-4 LT Pt-Al, having
higher Cr content in the bulk alloy (see Table 5.1), was shown in the previous chapter
to have relatively slower ’ formation despite having more bulk Ti content than the
TMS-82+ LT Pt-Al specimen. It should be noted that the stability of a certain phase
may as well be influenced by other alloying elements in the bulk superalloy.
Nonetheless, the results suggested that Cr plays a critical role in the prevention of
early ’ formation.
The complete phase transformation to the coarse martensitic structure in the HT Pt-Al
bond coat is believed to be associated with its martensitic transformation temperature,
Ms. A previous work by Smialek et al [43] indicated that Ms in -NiAl alloy increases
when Al depletion occurs as a result of oxidation and alloy/bond coat inter-diffusion.
As pointed out by Hangen et al [115], the martensitic transformation does not undergo
complete (i.e. 100%) transformation right below the Ms, but instead continues to
increase in volume fraction with decreasing temperature (i.e. increasing cooling). In
the previous EDX line profile study of the Pt-Al bond coats by Wu et al [35], the Al
content in the HT Pt-Al bond coat within a distance of 20m from the TGO/bond coat
interface was determined to be slightly lower (~7 at % lower on average) than that in
Page 154
154
the LT Pt-Al. These findings suggest that the complete transformation to coarse
martensite in the HT Pt-Al bond coat type is most likely due to its inherently low Al
content, hence, having a higher Ms.
5.4.3. Phase transformation induced interfacial rumpling
Qualitative and semi-quantitative assessments of local plastic strain by utilizing both
the changes in the EBSD pattern quality and grain misorientations have long been
conducted in many previous studies [116-118]. It is believed that the progressive
degradation in the pattern quality of the martensitic regions corresponds to an
increasing severity of plastic strain. The HT XRD results in the work by Chen et al
[44] indicated that the molar volume of the L10 martensite is approximately 2%
smaller than that of the , causing a transformation strain of about -0.7% during
cooling. They, along with several other researchers [119], believe that this volume
shrinkage due to martensitic transformation can generate additional stress during
thermal cycling in the bond coat, and thus, can significantly enhance rumpling
growth.
The role of martensitic transformation in causing rumpling should be questioned in
the case here with isothermal single cool-down tests. As pointed out in previous
literatures [39], rumpling amplitude tends to accumulate due to an intrinsic
asymmetry in the plastic response to tension and compression of the bond coat (i.e.
during cooling and heating). Rumpling tends to accumulate slowly, as the amount of
undulation developed per cycle is rather small. It should be noted; however, that the
Page 155
155
martensitic transformation may contribute to lowering interfacial fracture energy by
creating out-of-plane stresses at the bond coat/TGO interface (i.e. crack opening).
Although, the volume shrinkage associated with to ’ phase transformation is
significantly larger than that of to martensitic phase (i.e. 8 to 38% change depending
on whether Al depletion occurred predominantly by oxidation or inter-diffusion) [39],
it should be noted; however, that the process takes place at a much higher temperature
range, in which the creep rate is high. As a result, much of the TGO growth stress is
relaxed in the form of rumpling, as evident from the low stress level and high
rumpling rate of TMS-82+ LT Pt-Al specimen shown in the last chapter. In contrast,
the to martensitic transformation occurred in a significantly shorter time span (i.e.
cooling) at a much lower temperature range (i.e. less creep relaxation), in which the
strain generated by the martensitic transformation together with the CTE mismatch
between the coating layers contributes as additional elastic strain at the bond
coat/TGO interface without inducing rumpling. Hence, it is believed that the effect of
to martensitic transformation is more significant in triggering crack nucleation at the
bond coat/TGO interface.
EBSD results of 100hrs isothermally exposed specimens revealed that the degradation
of the martensites patterns tend to be the most severe when large ’ precipitates were
formed as neighboring grains (see Figure 5.7-5.9). Based on this observation, it is
proposed that the adjacent ’grains, along with the TGO scale, may impede the
contraction of martensites during cooling by imposing a tensile strain. This force
exerted on the martensites is likely to be proportional to the number of ’ precipitates
formed in the surroundings. Hence, TMS-82+ HT Pt-Al, having the fastest ’
formation at 100hrs into isothermal exposure, was shown to have the most severe
Page 156
156
degraded pattern in the martensites. On the other hand, the CMSX-4 LT Pt-Al
specimen with significantly less ’ formation than the TMS-82+ LT PT-Al after
100hrs of isothermal exposure, had noticeably less degradation in its EBSD map.
Note that the amount of ’ precipitates formation near the TGO/bond coat interface
was thoroughly quantified on these TBC systems in the previous chapter, and used as
reference for comparison here.
Tolpygo et al argued in their previous work [39] that reverse martensitic
transformation during cyclic thermal exposure has no discernible effect on the
rumpling of bond coat. This argument, however, does not conform to our results
shown in this work. As discussed earlier, when martensites were found as the
dominant phase (volume %) in the bond coat, such as in the case of the 30hrs
isothermally exposed TMS-82+ HT Pt-Al and the 100hrs isothermally exposed
CMSX-4 LT Pt-Al 100hrs specimens, there was much less pattern degradation (i.e.
severity of plastic strain) in their corresponding EBSD mappings as oppose to the
cases where ’ phases co-existed with martensites (i.e. TMS-82+ HT Pt-Al and
TMS-82+ LT Pt-Al at 100hrs). Moreover, their assumption that no martensitic
transformation occurred in their two-zone cycling temperature range lacks the
consideration that the Ms temperature increases with Al depletion and interdiffusion
of various alloying additions during oxidation, hence, should be considered as a
dynamic parameter [115,116,120]. Clearly, the additional strain generated by the
martensitic transformation may cause rumpling with increasing thermal cycles when
the Ms temperature is high enough for creep relaxation to take place in the bond coat.
The TMS-82+ Pt-diffusion specimen was shown to undergo a uniform Al depleted
Page 157
157
layer near the TGO/bond coat interface, with corresponding enrichment in Cr. The
segregation of Ti was also observed right beneath this Al-depleted and Cr-enriched
layer. This compositional distribution profile seems to be maintained between 30hrs
and 100hrs of isothermal exposure, while the as-coated ’ + phase gradually
transformed to a single phase microstructure. In a previous work [35], the TMS-82+
Pt-diffusion specimen was found to exhibit no rumpling behavior throughout its entire
thermal cycling test, in comparison with other TBC systems. In addition, the
TMS-82+ Pt diffusion specimen had some of the best interfacial adhesion among the
TBCs as indicated in the last chapter, despite having the largest TGO stress of the
entire specimen set during thermal exposure. The reason behind these phenomena can
be related to the uniform layer of phase microstructure near the bond coat/TGO
interface. Unlike the to ’/martensitic transformation of Pt-Al bond coat systems,
there is no significant volume change associated with ’ to transformation, and
hence, does not produce a rumpling growth. Moreover, the Pt-diffusion systems do
not have Ti segregation near the bond coat/TGO interface, and thus would not form
TiO2 in the TGO layer. Contrary to the Pt-diffusion system, the TMS-82+ LT Pt-Al
and TMS-82+ HT Pt-Al systems had relatively poor interfacial adhesion/spallation
lifetime and continuous growing rumpling amplitude during thermal exposure (as
shown in last chapter).
It is proposed here that the co-existence of martensite, and ’ phases with inherently
very different thermomechanical properties may induce a misfit stress in the TGO
layer during thermal cycling, leading to crack formation at the TGO/bond coat
interface and thus coating spallation. This idea is supported by the increasing trend of
the hardness and standard deviation values measured from various bond coat systems
Page 158
158
shown in section 5.3.4. Though not provided in the plot before, the average hardness
of as-coated bond coat systems (i.e. as coated CMSX-4 LT Pt-Al, TMS-82+ HT
Pt-Al and LT Pt-Al) was measured to be 5.42GPa. When the bond coat transformed to
mostly martensitic phase (i.e. TMS-82+ HT Pt-Al after 30hrs of isothermal exposure,
TMS-82+ LT Pt-Al and CMSX-4 LT Pt-Al after 50hrs of isothermal exposure), the
average hardness became 4.60GPa. As more and more ’ phase began to precipitate
out alongside the martensitic structure after 100hrs of isothermal exposure, a drastic
increase in both hardness and standard deviation values from those of the 50hrs
thermally exposed samples can be seen for all Pt-Al systems, especially for the
TMS-82+ HT Pt-Al specimen (see Figure 5.14). These variations in the hardness
value due to different phases are likely to induce a misfit stress in the TGO layer
across the TGO/bond coat interface, as the resistance to plastic deformation during
thermal cycling differs locally in the bond coat.
Specimens that partially underwent to ’ phase transformation were found to have
the highest degrees of rumpling. For instance, the SRR99 LT Pt-Al, TMS-82+ LT
Pt-Al and CMSX-4 LT Pt-Al specimens, all of which underwent partial to ’ phase
transformation, had the highest rumpling growth rate among all specimens between
as-coated state and 100hrs isothermal exposure. In contrast, the TMS-82+ Pt-diffusion
and TMS-82+ HT Pt-Al, both of which underwent a complete phase transformation at
early isothermal hours, had relatively lower rumpling growth rate (See previous
chapter).
Page 159
159
5.4.4 Driving force for spallation versus interfacial fracture toughness
The results of this work have clarified the difference in the degradation behavior
between the SRR99 LT Pt-Al and TMS-82+ HT Pt-Al specimens after 30 and 100hrs
of isothermal exposure, respectively. Figure 5.16 provides a schematic explanation of
the TGO stress evolution and the interfacial degradation process during isothermal
heating and subsequent cooling of the two specimens. During isothermal holding,
TGO began to thicken on both specimens; however, the SRR99 LT Pt-Al had a much
faster oxidation kinetic with traces of TiO2 present at the TGO/bond coat interface
while the TMS-82+ HT Pt-Al specimen maintained mostly pure alumina. The lateral
TGO growth due to the formation of new oxide within the scale interior was
constrained by the underlying metal substrate, which led to an in-plane compression
of the oxide scale. This existence of lateral growth strain at isothermal temperature is
known to cause creep in both the scale and bond coat, and thus induces surface
undulation [40]. In addition, the volume shrinkage associated with ’ formation led to
a local stress concentrated zone in the TGO above the ’ grain. This uneven strain
response between the and ’ regions would further enhance rumpling growth.
Page 160
160
Fig. 5.16. A qualitative illustration of the TGO stresses evolution and interfacial
degradation during isothermal heating and subsequent cooling of the 30hrs
isothermally exposed SRR99LT Pt-Al and the 100hrs isothermally exposed TMS-82+
HT Pt-Al. The SRR99 had more rumpling due to ’ formation at the grain boundary,
while the TMS-82+ had a relatively higher residual stress in the TGO due to
martensitic transformation.
It is interesting to note that despite having a thicker TGO and more ’ formation, the
Page 161
161
TMS-82+ HT Pt-Al after 100hrs isothermal oxidation had significantly less rumpling
compared to the SRR99 LT Pt-Al after 30hrs isothermal oxidation (See last chapter).
As pointed out in a previous work by Tolpygo [38], the bond coat was observed to
swell during high temperature oxidation due to Kirkendal effect, leading to surface
undulation of the bond coat. There seemed to be uneven response to bond coat swell,
where the inner grains of bond coat swelled more (i.e. undulation peaks) than the
grain boundaries (i.e. undulation valleys). Based on this phenomenon, it is believed
that unique ’ formation along the grain boundaries of the SRR99 specimen further
contributed to additional surface undulation due to the volume shrinkage (i.e. uneven
strain) of to ’ transformation, making the specimen more rumpled.
During cooling, TGO stress began to increase due to the CTE mismatch between the
different phases of the bond coat and the TGO layer. The bond coat contracted more
and thus applied a compressive stress on the TGO, while the TGO induced a tensile
stress on the bond coat. The TMS-82+ HT Pt-Al was different from the SRR99 LT
Pt-Al in that its phase began to transform into a coarse martensitic phase,
accompanied by molar volume shrinkage during cooling. This transformation likely
led to a higher stress-build up in the TGO layer above the martensitic phases relative
to the regions above ’ in the HT Pt-Al specimen at room temperature, which may
cause crack opening at the TGO/bond coat interface. This is evident from the
relatively higher TGO stress of the 30hrs isothermally exposed TMS-82+ HT-Pt-Al,
having a poor interfacial adhesion as determined previously. The SRR99 LT Pt-Al, on
the other hand, had much less stress gradient across the TGO, where regions above
the had more stress relaxation than those above the ’, as was more susceptible to
creep at high temperature. The main reason for the early spallation of the SRR99 LT
Page 162
162
Pt-Al can be associated with its faster rumpling growth, while the formation of the
less adhesive TiO2 at the TGO/bond coat interface likely contributed to the spallation.
As shown in the previous chapter, while SRR99 LT Pt-Al attained the highest
rumpling magnitude and ’ formation rate prior to spallation, other specimens (i.e.
TMS-82+ LT Pt-Al and CMSX-4 LT Pt-Al after 100 and 200hrs of isothermal
exposure) were found to gain even higher rumpling magnitude and more ’ formation,
yet continued to survive well into 200hrs of isothermal exposure.
In the previous chapter, a semi-quantitative plot of the variation of the driving force G
for spallation and the interfacial fracture toughness Gc with isothermal exposure was
given to show the remaining lifetime of the TBCs used in this work between 30 and
100hrs. The intention of that plot was to highlight the concept that the interfacial
adhesion of TBCs can only be determined by taking into account the dynamic change
of G and Gc with time and temperature. This concept is again demonstrated by the
degradation behavior of the SRR99 LT Pt-Al and TMS-82+ HT Pt-Al specimens
shown in this work. For instance, the formation of TiO2 at the TGO/bond coat
interface of the SRR99 LT Pt-Al, resulting in the loss of interfacial adhesion, can be
treated as a factor for driving the Gc value lower during isothermal exposure. On the
other hand, the martensitic transformation of the TMS-82+ HT Pt-Al during cooling,
leading to a stress build-up in the TGO, may be considered as a factor behind an
increasing G value.
Page 163
163
5.5 Conclusion
The results of this work have led to the following conclusions:
The interfacial degradation behavior has been associated with the phase
transformation of bond coat during thermal cycling. This is reflected by the
variation of interfacial rumpling and TGO stress with bond coat hardness
values, which are dependent on the phases present.
The progressive degradation in the EBSD pattern quality of the martensitic
regions due to increasing ’ phases nearby has suggested that a misfit stress
may be generated in the TGO layer due to the inherently different
thermomechanical properties between different phases.
The experimental results have clarified the difference in the degradation
mechanism between the SRR99 LT Pt-Al and TMS-82+ HT Pt-Al specimens.
Early spallation of the SRR99 LT Pt-Al was most likely caused by its fast
rumpling rate due to ’ formation at the grain boundary, while formation of
weakly adherent TiO2 scale due to Ti-rich ’ precipitates near the bond
coat/TGO interface likely contributed to the interfacial degradation. The early
spallation of the TMS-82+ HT Pt-Al was mainly driven by the stress build-up
in the TGO as a result of the volume shrinkage (i.e. out-of-plane stress)
associated with the martensitic transformation.
Thermo-Calc and EPMA analysis indicated that Cr and Ta, if given high enough
amount, has the effect of preventing diffusion of Ti toward to the TGO/bond
coat interface by stabilizing the and/or phases.
Page 164
164
CHAPTER 6
Summary and suggested future work
6.1 Summary
Current numerical approaches in modeling the intrinsic failure of TBC relies largely
on the notion that spallation occurs when the accumulating strain energy stored in the
coating exceeds a fixed critical value resembling the interfacial adhesion. If this is to
be entirely correct, one would expect this critical value of interfacial adhesion varies
with different materials, but stays independent of their thermal exposure history. In
this study, a unique cross-sectional indentation technique was developed to
quantitatively characterize the adhesion of oxide-bond coat interface among 5
systematically prepared material systems. The results not only re-confirmed that
interfacial adhesion is a material specific property in general, but more importantly,
conclusively demonstrated that the adhesion is dynamic, in particular with time and
temperature. With an aim of further understanding the dynamics (i.e. establishing
correlations between time and temperature dependent microstructure effects),
parameters such as the oxide growth rate, rumpling of the oxide-bond coat interface,
and phase transformation of bond coat were studied as a function of thermal exposure.
It has been conclusively demonstrated here that the oxide-bond coat interfacial
adhesion bears strong dependency on the phase distribution of the bond coats and
TGO growth rate, while receiving little influence from TGO rumpling and residual
stress.
Page 165
165
The influence of substrate and bond coat chemistry on the degradation mechanisms
leading to the early spallation of thermal barrier coatings (TBCs) has not been well
understood despite years of research effort. This is largely due to the sheer number of
factors (i.e. interfacial rumpling, thermally grown oxide (TGO) growth kinetics, and
TGO residual stress, etc.) that all seem to contribute to the degradation of TBCs. To
clarify the chemical effect, a complete investigation, utilizing the EBSD pattern
variation of bond coat near the oxide-bond coat interface, was carried out along with
various other characterizations on the isothermally exposed TBC specimens.
Specimens that underwent partial phase transformation from to’ near the interface
were determined to have larger rumpling magnitude with lower TGO stress, as
compared to those that transformed completely to a single phase (i.e. Pt-diffusion
type). It is evident that the formation of ’ along the grain boundaries can significantly
enhance rumpling, while martensitic transformation during cooling creates
out-of-plane stresses at the oxide-bond coat interface. These two degradation
mechanisms are likely to be the main reason behind early spallation of TBCs. To
alleviate these two mechanisms, it is necessary to minimize the formation of ’ at the
oxide-bond coat interface. Cr and Ta contents in as-fabricated substrate were
determined to stabilize and phases (i.e. inhibiting fast γ’ formation). Having less ’
formation near the interface can also prevent the formation of weakly adherent
titanium oxides, since Ti preferentially segregates to the γ’ phase.
Page 166
166
6.2 Suggested future work
By examining the results obtained from the works of Chapter 4 and 5, it can be
summarized that the spallation lifetime of TBCs is dictated by a work of many
different degradation mechanisms (i.e. TGO growth kinetics, type of TGO species,
and martensitic transformation, etc.). The key to understand these factors lies in the
backward trace of key elemental influence from bond coat and substrate chemistry.
Despite many significant findings on the degradation mechanism of TBCs in this
thesis, many aspects of the results can be further refined by carrying out the following
list of work.
- Fabricate Ti, Cr, and Ta modified nickel aluminide coatings (i.e. ternary Ni-Al-X
or quarternary Ni-Al-Pt-X alloys, where X indicates minor elements such as Ti, Cr
and Ta). Similar microstructural studies could be carried out on these types of
specimens after subjecting them to thermal cycling for different length of times.
XRD (X-ray diffraction) analysis can be utilized to identify the overall oxide
compositions formed on the surface. In addition, EPMA can be used to determine
the differences in local chemistry between the different layers and regions of
oxide formation. It will be particularly interesting to identify and compare the
oxide species grown above different regions of bond coats, where the existing
phases are different from each other (i.e. regions of , ’, and martensites, etc.).
EBSD analysis can be applied to identify the different phases of the bond coat
near the oxide-bond coat regions.
- Try to characterize the oxide-bond coat interface of TBCs. with the high
resolution capability of TEM. It would be interesting to study the oxide phases
Page 167
167
and crystal orientations above different regions of bond coats having different
phases (i.e. regions of , ’, and martensites, etc.). Specimens containing
oxide-bond coat interfaces of different regions can be prepared using FIB-INLO
technique (Focused Ion Beam In-Situ Lift-Out).
- Despite the success of a semi-quantitative approach to ranking the interfacial
adhesion of TBCs in this work, the results are limited in that a cross comparison
between coatings of other material systems is not possible. As mentioned in the
literature review chapter, a universal damage parameter on the basis of interfacial
fracture toughness is necessary to quantify the interfacial adhesion between any
two materials. It would be worthwhile to further refine the approximation of
interfacial fracture toughness, Gc, in section 4.4.1, so that a more accurate
toughness results can be obtained.
Page 168
168
References
[1] W.N. Harrison, D.G. Moore, J.C. Richmond NACA, 1947, TN-1186
[2] L.N. Hjelm, B.R. Bornhorst, NASA Tech Memo, X-57072 1961, 227–253
[3] H.J. Jr. Price, R.L. Schacht, R.J. Quentmeyer NASA 1973, TN D-7392
[4] S. Stecura, NASA, 1976, TM X-3425
[5] C.H. Liebert, F.S. Stepka, NASA, 1976, TM X-3352
[6] N.P. Padture, M. Gell, E.H. Jordan. Science, 296 (2002), pp. 280-284
[7] S. Stecura, NASA, 1978, TM-78976
[8] R.A. Miller, R.G. Garlick, J.L. Smialek. Ceram. Soc. Bull, 62 (1983), pp.
1355-1358
[9] J.A. Haynes (Ph.D thesis). Oxidation and Degeneration of Thermal Barrier
Coating System, University of Alabama at Birmingham, 1997
[10] D.S. Rickerby, R.G. Wing. Article including thermal barrier coating substrate,
US patent, 5 981 091 (1999)
[11] B. Gleeson, W.Wang, S Hayashi, D. Sordelet. Mater. Sci. Forum, 461–464
(2004), pp. 213-222
[12] D.R. Clarke, C.G. Levi. Annu. Rev. Mater. Des, 33 (2003), pp. 383-417
[13] K. Kawagishi, H. Harada, A. Sato, K. Matsumoto. Superalloys (2008), pp.
761-768
[14] W.J. Bridely, R.A. Miller. Surf. Coat Technol, 43-44 (1990), pp. 446-457
[15] S. Stecura. Thin Solid Films, 73 (1980), pp. 481-489
[16] S. Stecura. Thin Solid Films, 182 (1989), pp. 121-139
[17] H.E. Evans. Int. Mat. Rev, 40 (1995), pp. 1-40
[18] J.L. Smialek, G.H. Meier. Superalloys II (1987), pp. 293-326
Page 169
169
[19] P. Y. Hou, A. P. Paulikas, B.W. Veal. High Temp. Mat, (2005), pp. 373-380
[20] O. Lavigne, C. Ramusat, S. Drawin, P. Caron, D. Boivin, J.L. Pouchou.
Superalloys (2004), pp.667-675
[21] J. Angenete, K. Stiller, E. Bakchinova. Surf. Coat Technol, 176 (2004), pp.
272-283
[22] S. Shanker, L.L. Seigle. Metall. Trans. A, 9A (1978), pp.1467-1476
[23] A.G. Evans, M.Y. He, J.W. Hutchinson. Prog. Mater. Sci, 46 (2001), pp. 249-271
[24] M.Y. He, J.W. Hutchinson, A.G. Evans. Acta Mater, 50 (2002), pp. 1063-1073
[25] L. He, Z. Xu, J. Li, R. Mu, S. He, G. Huang. J. Mater. Sci Technol, 25 (2009),
No. 6, pp. 799-802
[26] R.T. Wu, K. Kawagishi, H. Harada, R.C. Reed. Acta Mater, 56 (2008), pp.
3622-3629
[27] R. J. Christensen, D. M. Lipkin, D. R. Clarke, K. Murphy. Appl.Phys. Lett, 69
(1996), pp.3754-3756
[28] A. Selcuk, A. Atkinson. Acta Mater, 51 (2003), pp. 535-549
[29] J. He, D.R. Clarke. J. Am. Ceram. Soc, 78 (1995), pp. 1347-1353.
[30] A. Selcuk, A. Atkinson. Mater. Sci. Eng, A335 (2002), pp.147-156
[31] V.K. Tolpygo, D.R. Clarke. Surf. Coat Technol,163-164 (2003), pp.81-86
[32] E.P. Busso, J. Lin, S. Sakurai, Acta Mater. 49 (2001), pp. 1529-1536
[33] G. Lee, A. Atkinson, A. Selcuk. Surf. Coat Technol, 201 (2006), pp.3931-3936.
[34] Busso E.P., Qian Z.Q., Taylor M.P. and Evans H.E. Acta Mater, 57 (2009), pp.
2349-2361
[35] R.T. Wu, X. Wang, A. Atkinson. Acta Mater, 58 (2010), pp. 5578-5585
[36] B. Gleeson, W. Wang, S. Hayashi, D. Sordelet. Mater Sci Forum, 213 (2004), pp.
461-464
Page 170
170
[37] V.K. Tolpygo, D.R. Clarke. Acta Mater, 48 (2000), pp. 3283-3293
[38] V.K.Tolpygo, D.R. Clarke. Acta Mater, 52 (2004), pp. 5129-5141
[39] V.K. Tolpygo, D.R. Clarke. Acta Mater, 52 (2004), pp. 5115-5127
[40] V.K. Tolpygo, J.R. Dryden, D.R. Clarke. Acta Mater, 46 (1997), pp. 927-937
[41] M.L. Glynn, M.W. Chen, K.T. Ramesh, K.J. Hemker. Mater. Trans. A, 35A
(2004), pp. 2279-2286
[42] S. Rosen, J.A. Goebel: Trans. TMS-AIME, 1968, 242, pp. 722-24.
[43] J.L. Smialek, R.F. Hehemann. Metall. Mater. Trans. A, 4 (1973), pp. 1571-1575
[44] M.W. Chen, M.L. Glynn, R.T. Ott, T.C. Hufnagel, K.J. Hemker. Acta Mater, 51
(2003), pp. 4279-4294
[45] A.W. Davis, A.G. Evans. Mater. Trans. A, 37A (2006), pp. 2085-2095
[46] J. Yan, T. Leist, M. Bartsch, A.M. Karlsson. Acta Mater, 56 (2008), pp.
4080-4090
[47]S.Q. Guo, D.R. Mumm, A.M. Karlsson, Y. Kagawa. Scripta Mater, 53 (2003) pp.
1043-1048
[48] D.S. Rickerby. Surf. Coat Technol, 36 (1988), pp. 541-557
[49] M.J. Stiger, R. Handoko, J.L. Beuth, F.S. Pettit, G.H. Meier. TMS-Proceeding
ISBN 0-87339-489-5 ( 2001) pp. 1
[50] M. Bartsch, B. Baufeld. Proc. Europ. Conf. Fract. –ECF 14, 1 (2002), pp. 209
[51] M.D. Drory, J.W. Hutchinson. Proc. R. Soc. Lond. A, 452 (1996), pp. 2319-2341
[52] D. Chicot, P. Demarecaux, J. Lesage. Thin Solid Films, 283 (1996), pp. 151-157
[53] Aditad Vasinonta, Jack L. Beuth. Eng. Frac. Mech, 68 (2001), pp. 843-860
[54] I. Hofinger, M. Oechsner, H.A. Bahr, M.V. Swain. Int. J. Fract, 92 (1998), pp.
213-220
[55] Y.F. Liu, Y. Kagawa, A.G. Evans. Acta Mater, 56 (2008), pp. 43-49
Page 171
171
[56] D.R. Mumm, A.G. Evans. Acta Mater, 48 (2000), pp. 1815-1827
[57] M.R. Begley, D.R. Mumm, A.G. Evans, J.W. Hutchinson. Acta Mater, 48 (2000),
pp. 3211-3220
[58] J. Chen, S.J. Bull. J. Phys. D: Appl. Phys, 44 (2011), 034001
[59] J.W. Beams. Science, 120 (1954), pp. 619-625
[60] P. Ostojic, R. McPherson. J. Am. Ceram. Soc, 71 (1988), pp. 891–899
[61] P.G. Charalambides, J. Lund, A.G. Evans, R.M. Mcmeeking. J. Appl. Mech, 111
(1989), p. 77-81
[62] S.J. Bull. Advanced Techniques for Surface Engineering ed, (1992), p. 31-64
[63] J. Malzbender , G. de With. Surf. Coat. Technol, 135 (2001), pp. 202-207
[64] J. Malzbender, G. de With. Thin Solid Films, 368 (2001), pp. 68-78
[65] J. Malzbender J, G. de With. Wear, 239 (2000), pp. 21-26
[66] J. Malzbender, G. de With. Wear, 236 (1999), pp. 355-359
[67] S.J. Bull, E.G. Berasetegui. Tribol. Int, 39 (2006), pp. 99-114
[68] X.D. Li, D.F. Diao, B. Bhushan. Acta Mater, 45 (1997), pp. 4453-4461
[69] M. Y. He, A. G. Evans, J. W. Hutchinson. Int. J. Solids Struct, 31(1994), pp.
3443–3455
[70] D.B. Marshall, A.G. Evans. J. Appl. Phys, 56 (1984), pp. 2632-2638
[71] J.W. Hutchinson, Z. Zuo. Adv. Appl. Mech, 29 (1992), pp.63-191
[72] K.B. Yeap, K.Y. Zeng, D.Z. Chi. Acta Mater, 59 (2008), pp. 977-984
[73] J.E. Ritter J, T. Lardner, L. Rosenfeld, M. Lin. J. Appl. Phys, 66 (1989), pp.
3626-3634
[74]M.J.Matthewson. J.Mech.Phys. Solids, 29 (1981), pp. 89–113
[75] J. Mencik. Mechanics of Components with Treated Coated Surface,” (Dordrecht:
Kluwer), (1996)
Page 172
172
[76] J. Malzbender, G. de With, J.M.J Toonder. Thin Solid Films, 366 (2000), pp.
139-149
[77] J. Chen, S.J. Bull. Thin Solid Films, 517 (2009), pp.3704-3711
[78] J.J. Vlassak, M.D. Droy, W.D. Nix. J. Mater. Res, 12 (1997), pp. 1900-1910
[79] D. Choulier. PhD-Thesis, Universite de Technologie Compiegne, (1989)
[80] R. Dal Maschio, V.G. Sglavo, F. Rigoni, L. Bertamini, E. Galvanetto. Proc. Int.
Thermal Spray Conf, (1992), pp. 949-951
[81] M. Bartsch, L. Mircea, J. Suffner, B. Baufeld. Key Eng. Mater, 290 (2005),
pp.183-190
[82] X. Wang, C. Wang, A. Atkinson. Acta mater, 60 (2012), pp. 6152-6163
[83] Y.F. Liu, Y. Kagawa, A.G. Evans. Acta. Mater, 56 (2008), pp. 43-49
[84] C. Mercer, J.R. Williams, D.R. Clarke, A.G. Evans. Proc. Royal. Soc. Ser. A, 463
(2007), pp. 1393-1408
[85] P.G. Charalambides, H.C. Cao, J. Lund, A.G. Evans. Mech. Mater, 8 (1990),
pp.269-283
[86] A.G. Evans, M.D. Drory, M.S. Hu. J. Mater. Res. Soc, 3 (1988), pp. 1043-1049
[87] P.Y. Thery, M. Poulain, M. Dupeux, M. Braccini. J. Mater. Sci, 44(2009), pp.
1726-1733
[88] H.E. Evans. Surf. Coat Technol, 206 (2011), pp. 1512-1521
[89] S.R. Choi, J.W Hutchinson, A.G. Evans. Mech. Mater, 31 (1999), pp. 431-477
[90] A.G. Evans, G.B. Crumley, R.E. Demaray. Oxid of Met, 20 (1983), pp. 193-216
[91] X.Y. Gong, D.R. Clarke. Oxid of Met, 50 (1998), pp. 355-376
[92] A.G. Evans, M.Y. He, J.W. Hutchinson. Acta Mater, 45 (1997), pp. 3543-3554
[93] Z. Suo. J. Mechan. Phys. Solids, 43 (1995), pp. 829-846
[94] M.Y. He, A.G. Evans, J.W. Hutchinson, Acta Mater, 48 (2000), pp. 2593-2601
Page 173
173
[95] H.E. Evans, R.C. Lobb, Coor. Sci, 24 (1984), pp. 209-222
[96] H.E. Evans, G.P. Mitchell, R.C. Lobb, D.R.J. Owen. Proc. Roy. Soc, 440A (1993),
pp. 1-22
[97] P. Deb, D.H. Boone, T.F. Manley. J. Vac. Sci. Technol A, 5 (1987), pp. 3366-3372
[98] R.C. Pennefather, D.H. Boone. Int. J. Press. Vess. Piping, 66 (1996), pp. 351-358
[99] G. Chang, W. Phucharoen, R.A. Miller. Surf. Coat. Technol, 30 (1987), pp. 13-28
[100] U. Schulz, M. Menzebach, C. Leyens, Y.G. Yang. Surf. Coat. Technol, 145-147
(2001), pp. 117-123
[101] B.A. Pint, J.A. Hayne, Y. Zhang. Surf. Coat Technol, 149 (2002), pp. 236-244
[102] X. Zhao, J. Liu, D.S. Rickerby, R.J. Jones, P. Xiao. Acta Mater, 59 (2011), pp.
6401-6411
[103] A.M. Karlsson. Key Eng. Mater, 333 (2007), pp.155-166
[104] Q. Ma, D.R. Clarke. J. Am. Ceram Soc, 76 (1993), pp.1433-40
[105] D.M. Lipkin, D.R. Clarke. Oxid. Met, 45 (1996), pp. 267-280
[106] X. Wang, S. Tint, M. Chiu, A. Atkinson. Acta Mater, 60 (2012), pp. 3247-3258
[107] K. Vaidynathan, H. Gell, E.H. Jordan. Surf. Coat. Technol, 133-134 (2000), pp.
28-34
[108] N.M. Yanar, F.S. Pettit, G.H. Meier. Metall. Mater Trans, 37A (2006), pp.
1563-1580
[109] E.P. Busso, L. Wright, H.E. Evans, L.N. McCartney, S.R.J. Saunders, S.
Osgerby, J. Nunn. Acta Mater, 55 (2007), pp. 1491-1503
[110] S.R. Choi, J.W. Hutchinson, A.G. Evans. Mech Mater, 31 (1999), pp.437-447
[111] M.Y. He, D.R. Mumm, A.G. Evans. Surf. Coat Technol, 185 (2004), pp. 184-93
[112] J.M. Sanchez, S. El-Mansy, B. Sun, T. Scherban, N. Fang, D. Pantuso, W. Ford,
M.R. Elizalde. J.M. Martinez-Esnaola, A. Martin-Meizoso, J. Gil-Sevillano, M.
Page 174
174
Fuentes, J. Maiz. Acta Mater, 47 (1999), pp. 4405-4413
[113] M. Levy, P. Farrell, F. Pettit. Corrosion-NACE, 42 (1986), pp. 708-717
[114] H.M. Tawancy, L.M. Al-Hadhrami. J. Eng. Gas. Turb. Power, 133 (2011), pp.
042101-6
[115] U.D. Hangen, G. Sauthoff. Intermetallics, 7 (1999), pp. 501-510
[116] A.J. Wilkinson, D.J. Dingley. Acta Metal. Mater, 39 (1991), pp. 3047-3055
[117] E.M. Lehockey, Y. Lin, O.E. Lepik. EBSD in Materials Science (2000), pp.
247-264
[118] M. Kamaya, A.J. Wilkinson, J.M. Titchmarsh. Nucl. Eng. Des, 235 (2005), pp.
713-725
[119] D.S. Balint, J. W. Hutchinson. J. Mechan. Phys. Solids, 53 (2005), pp. 949-973
[120] R. Kainuma, H. Ohtani, K. Ishida. Metall. Mater. Trans. A, 27 (1996), pp.
2445-2453