SPRINGER BRIEFS IN MATERIALS

Mohit Gupta

Design of
Thermal Barrier
Coatings
A Modelling
Approach

123
SpringerBriefs in Materials

More information about this series at http://www.springer.com/series/10111

Mohit Gupta

Design of Thermal Barrier

Coatings
A Modelling Approach
Mohit Gupta
University West
Trollhättan, Sweden

ISSN 2192-1091 ISSN 2192-1105 (electronic)

SpringerBriefs in Materials
ISBN 978-3-319-17253-8 ISBN 978-3-319-17254-5 (eBook)
DOI 10.1007/978-3-319-17254-5

Library of Congress Control Number: 2015937967

Springer Cham Heidelberg New York Dordrecht London

© The Author(s) 2015
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the
editors give a warranty, express or implied, with respect to the material contained herein or for any errors
or omissions that may have been made.

Printed on acid-free paper

Springer International Publishing AG Switzerland is part of Springer Science+Business Media

(www.springer.com)
Dedicated to my beloved parents
Smt. (Late) Raj Kumari Gupta and Shri Ram
Niwas Gupta
Preface

Thermal barrier coatings (TBCs) are now key elements in the design of advanced
gas turbines. TBCs are used in a wide variety of modern gas turbine applications
such as power generation, marine and aero engines. Applications of TBCs result in
higher gas turbine efficiency as well as enhanced lifetime of components due to the
thermal protection provided by TBCs. An extensive research is being carried out in
this field to continue the implementation of new and advanced coating systems to
achieve even lower emissions as well as fuel costs.
A high performance TBC would exhibit low thermal conductivity, high strain
tolerance and long lifetime. Thus, these three properties need to be optimised by
controlling the microstructure defects which control the thermal–mechanical prop-
erties of TBCs as well as the topcoat–bondcoat interface roughness which controls
the lifetime of TBCs. The purpose of this book is to describe a design methodology
which could be implemented to obtain an optimised TBC to be used for next gen-
eration gas turbines.
Chapter 1 introduces the topic in detail and describes the scope of this book.
A general background knowledge about the processing technology and materials in
TBCs is given in Chap. 2. The characteristics of TBCs in terms of microstructure,
properties and failure mechanisms are discussed in Chap. 3. The experimental
methods commonly used to characterise TBCs are described briefly in Chap. 4.
Chapter 5 describes a modelling approach to design the thermal–mechanical
properties of TBCs which consists of real microstructure images as well as artifi-
cially created images, and the results are obtained by using that approach. Chapter
6 describes a modelling approach to design the topcoat–bondcoat interface rough-
ness in TBCs which consists of real interface roughness profiles, and the results are
obtained by implementing that approach. Chapter 7 describes a diffusion-based
modelling approach to study oxide formation in TBCs during service conditions,
and the results are obtained by implementing that approach. The conclusions from
the modelling approach described in this book are given in the last chapter as a
methodology to design an optimised TBC.
This work was performed at the Production Technology Centre (PTC), Trollhättan,
as a part of the Thermal Spray research group at University West. The major portion

vii
viii Preface

of this work was done during the doctoral studies of the author. The author would like
to express his gratitude to his main supervisor during his doctoral studies, Prof. Per
Nylén, for his guidance, great support and valuable suggestions during this work.

Trollhättan, Sweden Mohit Gupta

25th February 2015
Contents

1 Introduction ................................................................................................. 1
1.1 Scope and Limitations.......................................................................... 4
References....................................................................................................... 5
2 Background ................................................................................................. 7
2.1 Thermal Spraying................................................................................. 7
2.1.1 Atmospheric Plasma Spraying ................................................. 7
2.1.2 High Velocity Oxy-Fuel Spraying ........................................... 9
2.1.3 Liquid Feedstock Plasma Spraying.......................................... 9
2.2 Thermal Barrier Coatings .................................................................... 9
2.3 Coating Formation ............................................................................... 11
2.4 Process Parameters............................................................................... 11
2.5 Coating Materials for TBCs ................................................................. 12
2.5.1 Topcoat ..................................................................................... 12
2.5.2 Bondcoat .................................................................................. 14
2.5.3 Thermally Grown Oxides......................................................... 14
References....................................................................................................... 15
3 Characteristics of TBCs.............................................................................. 17
3.1 Microstructure ...................................................................................... 17
3.2 Heat Transfer Mechanism .................................................................... 20
3.2.1 General Theory ........................................................................ 20
3.2.2 Application to TBCs ................................................................ 22
3.3 Mechanical Behaviour ......................................................................... 23
3.3.1 Stress Formation ...................................................................... 23
3.3.2 Young’s Modulus ..................................................................... 24
3.3.3 Non-linear Properties ............................................................... 25
3.4 Interface Roughness ............................................................................. 26
3.4.1 Roughness Relationship with Lifetime .................................... 26
3.4.2 Stress Inversion Theory............................................................ 27

ix
x Contents

3.5 Oxide Formation .................................................................................. 28

3.6 Failure Mechanisms ............................................................................. 29
References ...................................................................................................... 30
4 Experimental Methods ............................................................................... 33
4.1 Microstructure Characterisation........................................................... 33
4.2 Thermal Conductivity Measurements .................................................. 33
4.3 Young’s Modulus Measurements ......................................................... 35
4.4 Roughness Measurements.................................................................... 36
4.5 Lifetime Testing ................................................................................... 37
References ...................................................................................................... 40
5 Modelling of Properties of TBCs ............................................................... 43
5.1 Thermal Conductivity .......................................................................... 43
5.1.1 Analytical Models .................................................................... 43
5.1.2 Numerical Models.................................................................... 45
5.2 Young’s Modulus ................................................................................. 47
5.2.1 Analytical Models .................................................................... 47
5.2.2 Numerical Models.................................................................... 47
5.3 Finite Element Modelling .................................................................... 48
5.3.1 Basics of FEM and FDM ......................................................... 48
5.3.2 Image Based Finite Element Model......................................... 49
5.4 Artificial Coating Morphology Generator ........................................... 54
5.5 Recent Work......................................................................................... 56
References ...................................................................................................... 61
6 Modelling of Interface Roughness in TBCs .............................................. 65
6.1 Simplified Interface Roughness Modelling ......................................... 65
6.2 Real Interface Roughness Modelling ................................................... 66
6.2.1 Two-Dimensional Approach .................................................... 66
6.2.2 Three-Dimensional Approach .................................................. 68
6.3 Results ................................................................................................... 69
References ...................................................................................................... 71
7 Modelling of Oxide Growth in TBCs ........................................................ 73
7.1 Diffusion-Based Modelling.................................................................. 74
7.1.1 Previous Work .......................................................................... 74
7.1.2 Real Interface Roughness Modelling ....................................... 74
7.1.3 Recent Work............................................................................. 76
7.2 Mixed Oxide Growth ........................................................................... 78
References ...................................................................................................... 80
8 Conclusions: How to Design TBCs? .......................................................... 81
8.1 Future Work ......................................................................................... 82
About the Author

Mohit Gupta comes from Lucknow in India and received his bachelor’s degree in
mechanical engineering in 2009 from Indian Institute of Technology Kanpur, India.
He received his master’s degree in mechanical engineering in 2010 and doctoral
degree in production technology in 2015 from University West, Sweden. He is
currently employed as a university lecturer at University West. His research inter-
ests are finite element modelling, plasma spraying and solid oxide fuel cells.

xi
Abbreviations

2D Two-dimensional
3D Three-dimensional
APS Atmospheric plasma-sprayed
CAD Computer-aided design
CFD Computational fluid dynamics
CMAS Calcium–magnesium–aluminosilicate
CSN Chromia, spinel and nickel oxide
CTE Coefficient of thermal expansion
DoE Design of experiments
DSC Differential scanning calorimetry
EB-PVD Electron beam-physical vapour deposition
ECP Ex situ coating property
FDM Finite difference method
FEM Finite element method
FOD Foreign object damage
HVOF High velocity oxy-fuel
LFA Laser flash analysis
LOM Light optical microscopy
LPPS Low pressure plasma spraying
Micro-CT Micro-computed tomography
MIP Mercury intrusion porosimetry
OOF Object-oriented finite element analysis
PS-PVD Plasma spray-physical vapour deposition
SANS Small-angle neutron scattering
SEM Scanning electron microscope
SPPS Solution precursor plasma spraying
SPS Suspension plasma spraying
TBC Thermal barrier coating
TCF Thermal cyclic fatigue
TGO Thermally grown oxide

xiii
xiv Abbreviations

VPS Vacuum plasma spraying

XMT X-ray microtomography
XRD X-ray diffraction
YSZ Yttria-stabilised zirconia
Chapter 1
Introduction

Thermal barrier coating systems (TBCs) are widely used in modern gas turbine
engines in power generation, marine and aero engine applications to lower the metal
surface temperature in combustor and turbine section hardware. The application of
TBCs can provide increased engine performance/thrust by allowing higher gas tem-
peratures or reduced cooling air flow, and/or increased lifetime of turbine blades by
decreasing metal temperatures. TBC is a duplex material system consisting of an
insulating ceramic topcoat layer and an inter metallic bondcoat layer. It is designed
to serve the purpose of protecting gas turbine components from the severe thermal
environment, thus improving the efficiency and at the same time decreasing
unwanted emissions. Turbine entry gas temperatures can be higher than 1,500 K
with TBCs providing a temperature drop of even higher than 200 K across them [1].
TBCs were first successfully tested in the turbine section of a research gas turbine
engine in the mid-1970s [2].
Improvement in the performance of TBCs remains a key objective for further
development of gas turbine applications. A key objective for such applications is to
maximise the temperature drop across the topcoat, thus allowing higher turbine
entry temperatures and thus higher engine efficiencies. This comes with the require-
ment that the thermal conductivity of the ceramic topcoat should be minimised and
also that the value should remain low during prolonged exposure to service condi-
tions. In addition to this, longer lifetime of TBCs compared to the state-of the art is
of huge demand in the industry. In case of land-based gas turbines, a lifetime of
around 40,000 h is desired. Therefore, substantial research efforts are made in these
areas as TBCs have become an integral component of most gas turbines and are a
major factor affecting engine efficiency and durability.
The coating microstructures in TBC applications are highly heterogeneous, con-
sisting of defects such as pores and cracks of different sizes. The density, size and
morphology of these defects determine the coating’s final thermal and mechanical
properties and the service lives of the coatings [3–6]. To achieve a low thermal con-
ductivity and high strain-tolerant TBC with a sufficiently long lifetime, an optimisation

© The Author(s) 2015 1

M. Gupta, Design of Thermal Barrier Coatings, SpringerBriefs in Materials,
DOI 10.1007/978-3-319-17254-5_1
2 1 Introduction

between the distribution of pores and cracks is required, thus making it essential to
have a fundamental understanding of microstructure–property relationships in TBCs
to produce a desired coating.
Failure in atmospheric plasma-sprayed (APS) TBCs during thermal cyclic
loading is often within the topcoat near the interface, which is a result of thermo-
mechanical stresses developing due to thermally grown oxide (TGO) layer growth
and thermal expansion mismatch during thermal cycling. These stresses induce the
propagation of pre-existing cracks in as-sprayed state near the interface finally lead-
ing to cracks long enough to cause spallation of the topcoat [7–14]. The interface
roughness, although essential in plasma-sprayed TBCs for effective bonding
between topcoat and bondcoat, creates locations of high stress concentration.
Therefore, the understanding of fundamental relationships between interface rough-
ness and induced stresses, as well as their influence on lifetime of TBCs, is of high
relevance.
The traditional methodology to optimise the coating microstructure is by under-
taking an experimental approach. In this approach, certain set spray parameters are
chosen based on prior experience with the equipment and process, or as is the case
quite often, spray parameters from a factorial design experiments are chosen.
Thereafter, coatings are deposited using these parameters and evaluated by different
testing methods. This procedure could be iterated until the specified performance
parameters are obtained. As it might be apparent, this experimental approach is
extremely time consuming and expensive, apart from the drawback that it does not
enhance our knowledge of quantitative microstructure–property relationships.
Therefore, the derivation of microstructure–property relationships by simulation
would be an advantage. Simulation approach, apart from being time saving and cost
effective, is highly useful for the establishment of quantitative microstructure–property
relationships. New coating designs can be developed and analysed with the help of
simulation in a much easier manner compared to the experimental approach.
Another advantage of using simulation is that the individual effects of microstruc-
tural features (such as defects and roughness profiles) could be artificially separated
and analysed to study their effect on properties of TBCs independently which is not
possible by experimental methods. However, it must be noted here that the relation-
ships between process parameters and microstructure have to be established via an
experimental approach [4, 6]. A simulation approach is schematically described in
Fig. 1.1, showing the steps required to achieve a high performance TBC requiring
low thermal conductivity and long lifetime. It should be noted that the microstruc-
tural parameters for modelling are obtained through the link with experimental
spray parameters during the optimisation routine via process maps as described
further in Sect. 2.4.
A two-step approach has been described in this book to design an optimised TBC
by modelling:
1. Design microstructure defects in topcoat: As the thermal–mechanical properties
of TBCs are highly dependent on coating microstructure, the microstructure
defects can be designed to optimise the topcoat so as to achieve low thermal
conductivity and high strain tolerance which would imply long lifetime.
1 Introduction 3

Fig. 1.1 Block diagram showing the steps required to obtain a high performance TBC

2. Design topcoat–bondcoat interface: As the crack propagation in TBCs near

topcoat–bondcoat interface is highly dependent on the interface topography, the
interface can be designed to optimise the topography so as to reduce thermo-
mechanical stresses leading to crack propagation and thus enhance lifetime.
4 1 Introduction

1.1 Scope and Limitations

The present study is general in itself as it is based on analysis of microstructure

images and roughness profiles and is not dependent on material or equipment used
to fabricate them. However, other factors might have to be considered if this study
is applied to other materials or coating applications.
The limitations in this work can be broadly divided into the following categories:
1. Process
The study was limited to APS for topcoats and mainly high velocity oxy-fuel
(HVOF) spraying for bondcoats using powder feedstock. The effect of varying
bondcoat spray parameters on bondcoat microstructure and surface roughness
was not considered.
2. Material
Only one topcoat material, namely zirconia, was considered with different stabi-
lisers such as yttria and dysprosia. NiCoCrAlY was the only bondcoat material
considered in this study.
3. Experimental evaluation techniques
The experimental analysis performed in this study was limited to the scope of the
used technique. Microstructure was evaluated with light optical microscopy
(LOM) and scanning electron microscope (SEM) which could be incapable of
detecting all fine pores and cracks present in the microstructure. Lifetime testing
was limited to thermal cyclic fatigue (TCF) testing.
4. Modelling
The effect of changing chemistry was not considered in this study. The microstruc-
ture considered in this study consisted of only the defect morphology. The ther-
mal–mechanical properties modelled in this study were thermal conductivity and
Young’s modulus. Lifetime assessment considered in the modelling work was
based on TCF testing.
Most of the material properties used in the modelling work were considered to be
temperature independent even though the properties could change significantly over
the wide range of temperatures considered. As the focus was set on qualitative analy-
sis rather than on predicting exact values, this assumption was considered to be valid
in this case. The effect of radiation was not considered when predicting thermal
conductivity as it can be assumed to be scattered due to the porous microstructure.
Two-dimensional (2D) domain was considered in several models used to deter-
mine coating properties which could affect the final values, though it can be used
effectively for comparative purposes. Virtually designed microstructures were limited
by the scope of the software used. Focus was placed on the individual influence of
microstructural features on the final properties of the coating.
The effect of coating thickness was not considered in the stress analysis model.
The failure mechanism considered in the modelling work was limited to spallation
of coating due to thermo-mechanical stresses; other failure mechanisms such as
erosion were not considered.
References 5

References

1. Golosnoy IO, Cipitria A, Clyne TW (2009) Heat transfer through plasma-sprayed thermal
barrier coatings in gas turbines: a review of recent work. J Therm Spray Technol 18(5–6):
809–821
2. Miller RA (1997) Thermal barrier coatings for aircraft engines: history and directions. J Therm
Spray Technol 6(1):35–42
3. McPherson R (1981) The relationship between the mechanism of formation, microstructure
and properties of plasma-sprayed coatings. Thin Solid Films 83(3):297–310
4. Friis M (2002) A methodology to control the microstructure of plasma sprayed coatings. Ph.D.
thesis, Lund University, Sweden
5. Vassen R, Träeger F, Stöver D (2004) Correlation between spraying conditions and microcrack
density and their influence on thermal cycling life of thermal barrier coatings. J Therm Spray
Technol 13(3):396–404
6. Sampath S, Srinivasan V, Valarezo A, Vaidya A, Streibl T (2009) Sensing, control, and in situ
measurement of coating properties: an integrated approach toward establishing process-
property correlations. J Therm Spray Technol 18(2):243–255
7. Evans HE (2011) Oxidation failure of TBC systems: an assessment of mechanisms. Surf Coat
Technol 206:1512–1521
8. Hille TS, Turteltaub S, Suiker ASJ (2011) Oxide growth and damage evolution in thermal
barrier coatings. Eng Fract Mech 78:2139–2152
9. Kim D-J, Shin I-H, Koo J-M, Seok C-S, Lee T-W (2010) Failure mechanisms of coin-type
plasma-sprayed thermal barrier coatings with thermal fatigue. Surf Coat Technol 205:
S451–S458
10. Vaßen R, Giesen S, Stöver D (2009) Lifetime of plasma-sprayed thermal barrier coatings:
comparison of numerical and experimental results. J Therm Spray Technol 18(5–6):835–845
11. Trunova O, Beck T, Herzog R, Steinbrech RW, Singheiser L (2008) Damage mechanisms and
lifetime behavior of plasma sprayed thermal barrier coating systems for gas turbines—part I:
experiments. Surf Coat Technol 202:5027–5032
12. Chen WR, Wu X, Marple BR, Patnaik PC (2006) The growth and influence of thermally grown
oxide in a thermal barrier coating. Surf Coat Technol 201:1074–1079
13. Zhu D, Choi SR, Miller RA (2004) Development and thermal fatigue testing of ceramic thermal
barrier coatings. Surf Coat Technol 188–189:146–152
14. Schlichting KW, Padture NP, Jordan EH, Gell M (2003) Failure modes in plasma-sprayed
thermal barrier coatings. Mater Sci Eng A 342:120–130
Chapter 2
Background

2.1 Thermal Spraying

Thermal spraying is a branch of surface engineering processes in which metallic or

non-metallic coating material (in powder, wire or rod form) is heated to a molten or
semi-molten state and then propelled towards a prepared surface by either process gas
or atomisation jets. These particles adhere to the surface and build up to form a coating.
The workpiece on which the coating is deposited, or the substrate, remains unmelted.
The credit for inventing thermal spraying process belongs to M. U. Schoop
(Zurich, Switzerland) [1] who, along with his associates, developed equipment and
techniques for producing coatings using molten and powder metals in the early
1900s. The process was initially called “Metallization.” In 1908, Schoop patented the
electric arc spray process. In 1939, Reinecke introduced the first plasma spraying
process. Advancements in thermal spray equipment technology saw much higher
pace after the 1950s. Different variations of thermal spray technique exist today, such
as flame spraying, HVOF, APS and vacuum plasma spraying (VPS).
A major advantage of thermal spraying is that it can be used to deposit a wide
variety of materials without a significant heat input. In theory, any material that melts
without decomposing can be used for spraying without any undue distortion of the
part. A major disadvantage is that thermal spraying is a “line of sight” process,
although new processes such as plasma spray-physical vapour deposition (PS-PVD)
allow even shadowed areas to be coated [2].

2.1.1 Atmospheric Plasma Spraying

Plasma is an electrically conductive gas containing charged particles. When a gas is

excited to high energy levels, atoms lose hold of some of their electrons and become
ionised producing plasma containing electrically charged particles (ions and electrons)

© The Author(s) 2015 7

M. Gupta, Design of Thermal Barrier Coatings, SpringerBriefs in Materials,
DOI 10.1007/978-3-319-17254-5_2
8 2 Background

with temperatures ranging up to 20,000 K. The plasma generated for plasma spray
process usually incorporates one or a mixture of argon, helium, nitrogen and hydrogen.
The advantage of plasma flame is that it supplies large amounts of energy through
dissociation of molecular gases to atomic gases and ionisation.
A typical plasma spray process can be described in the following steps:
• First, a gas flow mixture (H2, N2, Ar) is introduced between a water-cooled copper
anode and a tungsten cathode.
• A high intensity DC electric arc passes between cathode and anode and is ionised
to form a plasma to reach extreme temperatures.
• The coating material in the form a fine powder conveyed by carrier gas (usually
argon) is introduced into the plasma plume formed due to the flow gases via an
external powder port and is heated to the molten state.
• The compressed gas propels the molten particles towards the substrate with
particle velocities ranging from 200 to 800 m/s.
Figure 2.1 shows a photograph taken during the plasma spray process indicating
the different components of the spraying unit.
Materials suitable for plasma spraying include zinc, aluminium, copper alloys,
tin, molybdenum, some steels, and numerous ceramic materials [1]. The advantage
of using plasma spray process compared to combustion processes is that it can spray
materials with very high melting points (refractory metals like tungsten and ceram-
ics like zirconia). On the other hand, it requires higher cost and increases the pro-
cess complexity.

Fig. 2.1 A photograph taken during atmospheric plasma spraying process

2.2 Thermal Barrier Coatings 9

In some cases, it could be advantageous to perform plasma spray process under

controlled environment under low pressure or vacuum, correspondingly termed as
low pressure plasma spraying (LPPS) or VPS. The use of controlled environment
could improve the coating quality due to reduced oxidation while spraying. However,
these processes increase the costs for the set-up equipment as well as processing
times, thus leading to significantly increased overall costs.

2.1.2 High Velocity Oxy-Fuel Spraying

In high velocity oxy-fuel (HVOF) spraying, a mixture of a fuel gas (such as hydrogen,
propane and propylene) and oxygen is ignited in a combustion chamber at high pres-
sures and the combustion gases are accelerated through a long de Laval (convergent–
divergent) nozzle to generate a supersonic jet with very high particle speeds. This spray
process generates extremely dense and well-bonded coatings.
The coatings usually sprayed by HVOF process are hard cermets like WC/Co or
Cr2C3/NiCr or MCrAlY (M = Ni and/or Co) applied as bondcoats to aircraft turbine
blades [1]. The advantage of using HVOF, apart from being suitable for making
dense coatings, is that the coatings contain few oxides due to the low process
temperature making it very attractive for TBC bondcoat applications.

2.1.3 Liquid Feedstock Plasma Spraying

The limitation of minimum particle size which could be used for APS led to the
development of plasma spray technology based on using liquid feedstock, mainly in
the form of suspension or solution, correspondingly known as suspension plasma
spraying (SPS) or solution precursor plasma spraying (SPPS). Since powders below
5 μm in size are difficult to feed and inject into the plasma torch, nanostructured
coatings are obtained by dispersing/dissolving nano- or sub-micrometric powder in
a liquid media to create a suspension/solution, respectively, and using it as a feed-
stock [3]. Suspensions are based on either water or an organic solvent in the form of
alcohol which is typically ethanol. Solutions are typically made of nitrates or chlo-
rides which are oxidised during the spray process forming an oxide particle that
forms the coating [3].

2.2 Thermal Barrier Coatings

A typical TBC (schematically shown in Fig. 2.2) consists of an intermediate metallic

bondcoat and a ceramic topcoat that provides the temperature drop across the coating.
In addition to a low thermal conductivity, topcoats should also have phase stability
during long-term high-temperature exposure and thermal cycling. The present
10 2 Background

Fig. 2.2 A typical thermal

barrier coating system

Fig. 2.3 A simplified comparison of APS (left), EB-PVD (centre) and SPS (right) TBC properties

state-of-the-art topcoat material is a 6–8 wt% yttria-stabilised zirconia (YSZ)

ceramic applied usually with plasma spray on a bondcoat of NiCoCrAlY [1]. YSZ
has a low thermal conductivity compared to other ceramics such as alumina. Also,
it is durable and chemically stable with a high melting point which makes it a good
choice. Further benefits of this material are discussed in detail in Sect. 2.5.1.
The ceramic top layer is typically applied by APS, electron beam-physical
vapour deposition (EB-PVD) or SPS. Figure 2.3 summarises a simplified compari-
son of APS, EB-PVD and SPS TBCs. The main purpose of the topcoat is to provide
thermal protection. In addition, it should be susceptible to the thermo-mechanical
stresses arising during the operating conditions.
The bondcoat provides improved bonding strength between the substrate and the
topcoat. It also reduces the interface stresses arising due to the difference in coeffi-
cients of thermal expansion (CTEs) of ceramic topcoats and metallic substrates.
As at high temperatures the porous ceramic topcoat is transparent to the flow of
oxygen and the exhaust gases, an aluminium-enriched bondcoat composition is
used to provide a slow growing, adherent aluminium oxide film known as TGO.
TGO layer provides oxidation protection to the substrate as alumina has very low
diffusion coefficients for both oxygen and metal ions.
2.4 Process Parameters 11

2.3 Coating Formation

Plasma-sprayed coatings are built up particle by particle. Each individual molten or

semi-molten particle impacts the substrate surface and flattens, adheres and solidi-
fies to form a lamellae structure called splat. Unmelted particles are bounced back
from the substrate reducing the deposition efficiency of the coating. When a spheri-
cal liquid droplet strikes a flat surface with a high impact velocity, it tries to flatten
to a disc but the radially flowing thin sheet of liquid becomes unstable and disinte-
grates at the edges to form small droplets. This process is interrupted by rapid solid-
ification in the case of plasma spray as the substrate is well below the melting point
of the droplet, with a cooling rate as high as 106 K s−1 [4]. Major part of the heat
energy from the particle is transferred to the substrate by conduction, and solidifica-
tion starts at the interface between the particle and the substrate. Heat transfer due
to convection and radiation makes a small contribution at the conditions of plasma
spray process [5].
Small voids exist between the splats, which can be formed due to incomplete
bonding between splats owing to lack of adhesion, relaxation of the residual stresses
mainly during the rapid cooling of the splat or trapped gases. The real area of con-
tact between splats is around 20 % [4], due to which the properties of the coatings,
such as mechanical, thermal and electrical properties, are very different compared
to the sprayed bulk material. The major factors influencing the structure of a coating
are the temperature, velocity and size distribution of the incident particles, apart
from substrate temperature and roughness [6].

2.4 Process Parameters

Plasma spray process is influenced by a number of process parameters. The depen-

dence of microstructure on so many parameters has also been an enigma in terms
of process control. A vast number of parameters need to be monitored and con-
trolled, and several others are very difficult to control, such as electrode wear,
humidity in surrounding air and fluctuations in controllable parameters. Several
studies have been performed in the past emphasising on examination of process–
structure–property relationships based on process maps to overcome this drawback
to some extent [5, 7–9]. These process maps, however, are restricted to the specific
spray gun, powder and all other parameters, except the ones varied to conduct the
study, which makes implementation of the process maps from one gun to another
rather difficult. Some of the process parameters in plasma spraying are presented
in Fig. 2.4.
12 2 Background

Fig. 2.4 Plasma spray process parameters

2.5 Coating Materials for TBCs

2.5.1 Topcoat

The ceramic topcoat provides thermal insulation to the substrate underneath. Some
of the basic properties to be exhibited by the material used as topcoat are (1) low
thermal conductivity, (2) high melting point and (3) phase stability. The material
most widely used as topcoat is 6–8 wt% YSZ due to its low thermal conductivity,
relatively high CTE and adequate toughness [10]. Apart from YSZ, other ceramics
which are used as TBC materials are mullite, Al2O3, TiO2, CeO2 + YSZ, La2Zr2O7,
pyrochlores and perovskites [11].
Zirconia ceramics are one of the very few non-metallic materials which have
good mechanical properties as well as electrical properties, apart from having low
thermal conductivity. These properties are exhibited due to the particular crystal
structure of ZrO2, which is principally a fluorite-type lattice. Pure ZrO2 has a mono-
clinic crystal structure at room temperature and undergoes phase transformations to
tetragonal (at 1,197 °C) and cubic (at 2,300 °C) at increasing temperatures with a
melting point of about 2,700 °C [11, 12]. Figure 2.5 shows the phase diagram for the
zirconia–yttria system. The Zr4+ ions in the cubic ZrO2 have very low coefficients of
diffusion as they are very immobile which gives ZrO2 a very high melting point and
good resistance to both acids and alkalis. On the other hand, the O2− ions are mobile
2.5 Coating Materials for TBCs 13

3000 Liquid (L)

2500 L+C

Cubic (C)
2000
Temperature [°C]

Tetra
1500 gonal
(T) T+C

1000
Monoclinic (C)
M+T

500
M+C
non transformable
M tetragonal T°
C
0
0 2.5 5 7.5 10
Mol % Y2 O3

Fig. 2.5 Phase diagram for the zirconia–yttria system [13]

due to the presence of vacancies, which makes the cubic phase ion conducting and
thus providing it good electrical properties. The main sources of ZrO2 in nature are
zircon (ZrSiO4) and baddeleyite. Apart from TBCs, ZrO2 has widespread applica-
tions in fuel cells, jewellery (cubic ZrO2), oxygen sensors, electronics, etc.
The volume expansion which occurs due to the phase transformation from cubic
(c) to tetragonal (t) to monoclinic (m) phases causes large stresses which will pro-
duce cracks in pure ZrO2 upon cooling from high temperature. Thus, several oxides
are added to stabilise the tetragonal and/or cubic ZrO2 phases like yttria (Y2O3),
ceria (Ce2O3), magnesium oxide (MgO) and calcium oxide (CaO). Specific addition
of cations like Y3+, Ca2+ into the ZrO2 lattice causes them to occupy the Zr4+ positions
in the lattice which results in the formation of anion vacancies to maintain charge
equilibrium.
YSZ is preferred to CaO- or MgO-stabilised zirconia for TBCs as YSZ coatings
have been proved to be more resistant against corrosion [11]. Also, YSZ coatings
14 2 Background

exhibit highest degree of resistance to coating failure due to spallation and an

excellent thermal stability [14]. Its major disadvantage is the limited operating
temperature (<1,473 K) for long-term application [11]. At higher temperatures, the
metastable t’-tetragonal phase, which is the main phase present at the room tem-
perature, transforms first to tetragonal and cubic (t + c) phase and then to monoclinic
(m) phase during cooling, which results in volume contraction and, consequently,
the formation of cracks in the coating [11]. Also, as the YSZ coatings possess high
concentration of vacancies, they facilitate oxygen transport at high temperatures
which results in the oxidation of the bondcoat. This leads to the failure of TBCs due
to spallation of the ceramic. The latter problem is solved by inducing the formation
of oxidation-resistant TGO layer between the topcoat and bondcoat by using alloys
based on NiAl with various additions such as Cr, Co, Pt, Y and Hf as bondcoat
materials [10].

2.5.2 Bondcoat

The bondcoat provides oxidation protection of the substrate and improved adhesion
of the topcoat. Some of the basic requirements from a bondcoat material are (1)
resistance to interdiffusion with the substrate and (2) high creep strength with suitable
ductility.
The typically used thermal sprayed bondcoat materials consist of a variety of
MCrAlX alloys, where M = Ni and/or Co and X = Y, Hf and/or Si [15]. Nickel is
added to enhance oxidation resistance while cobalt for corrosion resistance [16].
Aluminium is added to bondcoat to act as a local aluminium reservoir to provide a
slow growing TGO α-alumina during service conditions to provide oxidation pro-
tection. Alumina is preferred to other oxides due to its low thermal diffusivity and
better adherence [17]. Chromium is added to enhance oxidation and corrosion
resistance [18]. Yttrium is added to provide protection from sulphur diffusion into
the coating by acting as a gettering site and promoting formation of alumina as
well as its adhesion [17].

2.5.3 Thermally Grown Oxides

The oxidation of bondcoat in service conditions results in the formation of a TGO

layer near the topcoat–bondcoat interface. The major oxide formed due to the oxi-
dation of a typically used MCrAlY bondcoat is alumina. Other oxides which are
usually formed are chromia ((Cr,Al)2O3), spinel (Ni(Cr,Al)2O4, nickel oxide (NiO)
and silica [19]. Chromia, spinel and nickel oxide are usually abbreviated as CSN in
literature. The oxidation of bondcoat has been recognised as one of the major causes
for TBC failure [19, 20]; it will be discussed in detail in Sect. 3.5.
References 15

References

1. Davis JR (ed) (2009) Handbook of thermal spray technology. ASM International, Materials Park
2. Niessen K, Gindrat M (2011) Plasma spray-PVD: a new thermal spray process to deposit Out
of the vapor phase. J Therm Spray Technol 20(4):736–743
3. Fauchais P, Etchart-Salas R, Rat V, Coudert JF, Caron N, Wittmann-Teneze K (2008)
Parameters controlling liquid plasma spraying solutions, sols, or suspensions. J Therm Spray
Technol 17(1):31–59
4. McPherson R (1989) A review of microstructure and properties of plasma sprayed ceramic
coatings. Surf Coat Technol 39–40:173–181
5. Friis M (2002) A methodology to control the microstructure of plasma sprayed coatings, Ph.D.
thesis, Lund University, Sweden
6. Bianchi L, Leger AC, Vardelle M, Vardelle A, Fauchais P (1997) Splat formation and cooling
of plasma-sprayed zirconia. Thin Solid Films 305(1–2):35–47
7. Vassen R, Träeger F, Stöver D (2004) Correlation between spraying conditions and microcrack
density and their influence on thermal cycling life of thermal barrier coatings. J Therm Spray
Technol 13(3):396–404
8. Sampath S, Jiang XY, Matejicek J, Leger AC, Vardelle A (1999) Substrate temperature effects
on splat formation, microstructure development and properties of plasma sprayed coatings part
I: case study for partially stabilized zirconia. Mater Sci Eng A 272:181–188
9. Vaidya A, Srinivasan V, Streibl T, Friis M, Chi W, Sampath S (2008) Process maps for plasma
spraying of yttria-stabilized zirconia: an integrated approach to design, optimization and
reliability. Mater Sci Eng A 497:239–253
10. Evans AG, Clarke DR, Levi CG (2008) The influence of oxides on the performance of
advanced gas turbines. J Eur Ceram Soc 28(7):1405–1419
11. Cao XQ, Vassen R, Stoever D (2004) Ceramic materials for thermal barrier coatings. J Eur
Ceram Soc 24(1):1–10
12. Schafer U, Schubert H, Carle VM, Taffner U, Predel F, Petzow G (1991) Ceramography of high
performance ceramics—description of materials, preparation, etching techniques and description
of microstructures—part III: zirconium di oxide (ZrO2). Pract Metallogr 28:468–483
13. Scott HG (1975) Phase relationships in the zirconia-yttria system. J Mater Sci 10(9):1527–1535
14. Alpérine S, Derrien M, Jaslier Y, Mévrel R. Thermal barrier coatings: the Thermal Conductivity
challenge, Proceedings of the 85th Meeting of the AGARD Structures and Materials Panel,
Oct 15–16, 1997 (Aalborg, Denmark), NATO-AGARD-R-823, 1998, p 1-1–1-10
15. Haynes JA, Ferber MK, Porter WD (2000) Thermal cycling behavior of plasma-sprayed thermal
barrier coatings with various MCrAIX bond coats. J Therm Spray Technol 9(1):38–48
16. Curry N (2014) Design of thermal barrier coatings. Ph.D. thesis, University West, Sweden, No. 3
17. Evans AG, Mumm DR, Hutchinson JW, Meier GH, Pettit FS (2001) Mechanisms controlling
the durability of thermal barrier coatings. Prog Mater Sci 46:505–553
18. Yuan K (2014) Oxidation and corrosion of new MCrAlX coatings: modelling and experiments,
Ph.D. thesis, Linköping University, Sweden
19. Chen WR, Wu X, Marple BR, Nagy DR, Patnaik PC (2008) TGO growth behaviour in TBCs
with APS and HVOF bond coats. Surf Coat Technol 202:2677–2683
20. Vaßen R, Giesen S, Stöver D (2009) Lifetime of plasma-sprayed thermal barrier coatings:
comparison of numerical and experimental results. J Therm Spray Technol 18(5–6):835–845
Chapter 3
Characteristics of TBCs

The characteristics of TBCs are very different from that of the bulk material. Apart
from the presence of several defects present in plasma-sprayed coatings, the splat
interfaces present in coatings are rough, resulting in multiple contact points ranging
from micrometre to nanometre scale. These contact points can have very different
bonding characteristics. These features have a significant implication on the mac-
roscale properties of TBCs.

3.1 Microstructure

As discussed earlier, the microstructure of APS ceramic coatings is significantly

influenced by the process parameters. This influence results in a complex micro-
structure with various forms of porous features. Common features present in a
ceramic coating are presented in Fig. 3.1. This image of coating cross section has
been taken by a SEM using backscattered electron detector mode which is a com-
mon technique used for analysing the microstructures in TBCs.
As it can be observed in Fig. 3.1, the microstructure consists mainly of large porous
features, commonly referred to as globular pores, or simply pores, and several long
and narrow porous features, commonly referred to as delaminations and cracks. They
are also called interlamellar (delaminations, horizontally oriented) and intralamellar
(vertical cracks or those non-horizontally oriented). It could be difficult to differenti-
ate between delaminations and cracks, especially at lower magnifications, and so they
are quite often simply referred to as cracks. Some fine scale porosity and partially
melted particles can also be noticed in the microstructure image.
Globular pores are formed due to incomplete packing of the deposited particles or
due to defects in the structure. This effect is magnified in the case of the low energy
spray parameters since in that case the particles are unmelted/partially melted and so
they do not flatten out adequately on impact. Delaminations are formed due to
incomplete bonding between consecutively deposited particles, usually during

© The Author(s) 2015 17

M. Gupta, Design of Thermal Barrier Coatings, SpringerBriefs in Materials,
DOI 10.1007/978-3-319-17254-5_3
18 3 Characteristics of TBCs

Fig. 3.1 SEM micrograph illustrating common features present in APS coating microstructures

successive passes of the spray gun. Cracks are formed due to the stresses developed
within the coating when the sprayed particles cool down. The rapid cooling of
particles results in large shrinkage which induces tensile stresses as the underlying
material tries to prevent the shrinkage. These tensile stresses are released by the for-
mation of cracks in the coating. Under appropriate spray conditions, vertical cracks
in the coating can propagate to form long cracks orthogonal to the surface known as
segmentation cracks. One such microstructure image is shown in Fig. 3.2.
The individual splats have a typical columnar grain structure caused by the
directional solidification of the particles. This structure can be typically observed in
a fracture surface of the coating as shown in Fig. 3.3. It can be observed that the
columnar grains grow as usual along the direction of heat flow from top to bottom [2].
The grain structure and size depends on the processing conditions of the particles
during spraying and can vary significantly during service conditions which could
affect the coating properties [3].
Thermal–mechanical properties and lifetime of TBCs depend mainly on the
various microstructural features present in the topcoat. A thermal sprayed YSZ
coating has a thermal conductivity around 0.5–1 W m−1 K−1, as compared to the
thermal conductivity of 2.5 W m−1 K−1 for bulk material [4]. A large amount of pores
and cracks perpendicular to the heat flow will provide better insulation properties to
the coating [5]. On the other hand, these horizontal cracks might propagate due to
the thermal and mechanical stresses during operating conditions and eventually lead
to failure of the coating by spallation [4]. Segmentation cracks increase the flexibility
3.1 Microstructure 19

Fig. 3.2 Microstructure image showing vertical cracks in the topcoat

Fig. 3.3 A microstructure image showing the fracture surface of the coating in as-sprayed
condition [1]
20 3 Characteristics of TBCs

of the coating as they help in relaxing the residual stresses within the coating [6].
Again, on the other hand, these vertical cracks enhance the thermal conductivity of
the coating as they allow the flow of high temperature gases thus increasing the
heat transfer to the substrate [6]. To achieve a low thermal conductivity and high
strain-­tolerant TBC with a sufficiently long lifetime, an optimisation of microstruc-
tural features is required.

3.2 Heat Transfer Mechanism

3.2.1 General Theory

The theory of heat transfer in crystalline solids is described very well in the literature [7].
It has been briefly reviewed in other sources as well [6, 8–10].
Heat energy can be transferred by three mechanisms in crystalline solids—
electrons, lattice waves (phonons) and electromagnetic waves (photons) [8]. The total
thermal conductivity Κ, which is the sum of the three components, can be expressed
in general form as [8]
1 N
k= å Cp v jl j
3 j =1 j
(3.1)

where Cp is the specific heat at constant pressure, N is the total number of energy
carriers, v is the velocity of a given carrier (group velocity if the carrier is a wave)
and l is the corresponding mean free path.
The electrons are capable of transferring energy only when they are free of inter-
actions with the crystal lattice. This type of electrons is present only in metals and
partly in metal alloys, especially at high temperatures. The electronic thermal con-
ductivity part Κe is proportional to the product of the temperature and the electron
mean free path, with νe being independent of temperature [8]. The electron mean
free path has two parts: residual mean free path, which is related to the scattering of
electrons by defects, and intrinsic mean free path, which is related to the scattering
of electrons by lattice vibrations. The residual mean free path is independent of
temperature while intrinsic mean free path is directly proportional to temperature.
At low temperatures, the electrons are mainly scattered by the defects, while as the
temperature increases, the scattering mechanism by the phonons becomes more and
more dominant. Therefore, the electronic component of thermal conductivity is pro-
portional to temperature at low temperatures, becoming less dependent as tempera-
ture increases, and finally becoming independent of it at high temperatures [8].
Lattice thermal conduction, or heat energy transport by phonons, occurs in all
types of solids, with the phenomenon being dominant in alloys at low temperatures
and in ceramics. At low temperatures (T), the component of thermal conductivity
related to heat transport by phonons, Κph, may be represented by an exponential term
3.2 Heat Transfer Mechanism 21

exp(T*/T), whereas it is inversely proportional to temperature at high temperatures

[8]. Here T* is the characteristic temperature of the material which is generally
proportional to the Debye temperature. Κph may also be expressed as [10]
1
3ò
k ph = Cv r vl p (3.2)

where Cv is the specific heat at constant volume, ρ is the density, and lp is the mean
free path for scattering of phonons.
Heat energy transport by photon conduction (radiation) occurs especially at
high temperatures in materials transparent to infrared radiation such as ceramics
(over 1,200–1,500 K) and glasses (over 900 K) [8]. The radiative component of the
thermal conductivity, Κr, can be expressed as [10]
16
kr = s s n 2 lr T 3 (3.3)
3

where n is the refractive index, lr is the mean free path for photon scattering (defined
as the path length over which the intensity of radiation will reduce by a factor of 1/e)
and σs is the Stefan–Boltzmann constant.
In real crystal structures, scattering of phonons occurs due to their interaction with
lattice imperfections in the ideal lattice, like vacancies, dislocations, grain boundar-
ies, atoms of different masses and other phonons. Phonon scattering may also occur
due to ions and atoms of different ionic radius as they distort the bond length locally,
and thus elastic strain fields might be present in the lattice. The phonon mean free
path lp is defined as [10]
1 1 1 1 1
= + + + (3.4)
lp li lvac lgb lstrain

where li, lvac, lgb and lstrain are mean free paths associated with interstitials, vacancies,
grain boundaries and lattice strain, respectively. The intrinsic lattice structure and
the strain fields mainly affect the phonon mean free path in conventional materials,
with the grain boundary term having the least effect.
Thermal conduction in gases depends on the molecular mean free path, λ, of the
gases. λ is a function of temperature and pressure P and is proportional to T/P for
ideal gas behaviour [6]. The thermal conductivity of a gas, Κg, in a constrained
channel of length dv can be expressed as [6, 11]

k g0
kg = (3.5)
1 + BT / ( dv P )

where κg0 is the normal (unconstrained) conductivity of the gas at the temperature
concerned and B is a constant which depends on the gas type and the properties of
the interacting solid surface [6].
22 3 Characteristics of TBCs

3.2.2 Application to TBCs

In a real engine environment, TBCs protecting the substrate receive radiation which can
be classified into the following two categories—far-field and near-field radiations [9].
Figure 3.4 shows the temperature distribution across a typical TBC system during ser-
vice conditions from the hot gases in the combustor to the substrate. The operating
temperature increase obtained due to the application of TBC is indicated in the figure.
Far-field radiation comes from the combustion gases which are at temperatures
around 2,000 °C, having a spectral distribution same as that of a black body at that
temperature. This far-field radiation makes a small contribution as the hot combus-
tion gases have finite thickness and limited opacity and thus have reduced emissivity,
which results in the far-field radiation being reduced by a geometrical factor.
Near-field radiation comes due to the layer of cooler gas, at around 1,200 °C,
which is adjacent to the topcoat. The topcoat ceramic surface, which is at a similar
temperature as this gas layer, emits radiation which can pass through the partially
transparent ceramic topcoat to the metallic bondcoat and the substrate. This near-­
field radiation contributes to the thermal conductivity of the ceramic and can affect
it significantly at elevated temperatures [9].
The three important factors influencing the heat conduction in typical thermal
sprayed coatings are the dimensions of the grains, the impurities and the porosity [8].

Fig. 3.4 Temperature distribution during typical service conditions across a typical thermal
barrier coating system
3.3 Mechanical Behaviour 23

The grain dimensions depend on the solidification conditions of sprayed liquid

droplets, which depends mainly on spraying technique, cooling of the substrate and
the thickness of the sprayed coating (particles solidifying on previously deposited
layers have lower solidification rates and thus larger grain sizes) [8]. In case of
oxides, however, the grain size might influence the phonon mean free path only at
low temperatures, with the crystal structure being the major factor influencing mean
free path rather than the grain size [8].
The typical impurities present in plasma-sprayed coatings are, besides oxides,
the copper and tungsten particles coming from the electrodes of the torch and the
sand blasted particles at the interface between coating and substrate. The effect of
these impurities on thermal properties of coatings can be ignored if the electrodes
are cleaned systematically and the substrate is cleaned after sand blasting [8].
The thermal conductivity of gas inside a pore is close to that of free gas if the
dimensions of the pore are much larger than the mean free path (L < 10λ), but it can fall
significantly below the free gas value for even moderately fine structures (L < ~1 μm) [8].
Both pore thickness and gas pressure can affect the pore conductivity significantly, as
well as the temperature [11]. Convective heat transfer within the pores can be neglected
for plasma-sprayed TBCs, as it is only likely to be significant if the pores are large
(L > ~10 mm) [6]. Also, convective heat transfer through the porosity network can also
be neglected, even though most porosity in TBCs is normally interconnected [6].
As zirconia-based ceramics are electronic insulators with electrical conductivity
occurring only at high temperatures by oxygen ion diffusion, there is no contribu-
tion to thermal conductivity due to electrons. Thus heat transfer in zirconia takes
place only by lattice vibrations (phonons) or radiation (photons) [10]. Adding yttria
to zirconia modifies the lattice structure locally by introducing ion vacancies and
generating local strain fields due to the incorporation of large dopant atoms, which
results in lower intrinsic mean free path due to enhanced scattering of phonons and
thus reduced thermal conductivity [10]. The radiative heat transfer part in zirconia-­
based plasma-sprayed TBCs becomes significant only at temperatures above
~1,500 K and thus was neglected in the present modelling work [6]. The c­ ontribution
from radiation depends on the radiation scattering length, which increases as the
grain and pore structure coarsens and intersplat contact area increases as well as the
coating thickness increases [6, 12].

3.3 Mechanical Behaviour

3.3.1 Stress Formation

The residual stresses in APS coatings in the as-sprayed condition arise due to two
main factors [13, 14]:
1. Quenching: These stresses arise due to the rapid solidification of single particles
during spraying while their contraction is restricted by the adherence to the sub-
strate. Due to the temperature difference between the substrate and the particles,
tensile stresses are generated in the particles known as quenching stresses.
24 3 Characteristics of TBCs

2. Thermal mismatch: These stresses arise due to mismatch in CTE of the coating
and the substrate during cooling after spraying. As ceramic topcoats have much
lower CTE than the metallic substrates, the CTE mismatch results in compressive
stresses in the coating as the substrate shrinks more during cooling.
The stress state in the coating is due to the combination of the above two factors
leading to stress generation in APS coatings. In addition, several other factors could
also attribute to the final stress state in the coating, such as temperature gradients
during and after deposition, stress relaxation processes (plastic deformation,
cracking, etc.), phase transformations and chemistry changes [14]. In general, topcoats
in TBCs have low residual stresses in as-sprayed condition mainly due to the brittle
nature of the ceramic which results in stress relaxation [15].
The stress state in TBCs during service conditions changes as the TGO layer is
gradually formed; this phenomenon is discussed in detail in Sect. 3.5.

3.3.2 Young’s Modulus

The mechanical properties of thermal sprayed coatings are highly dependent on the
microstructure as it is significantly different from that of conventionally processed
materials. The elastic modulus or Young’s modulus and Poisson’s ratio are the basic
parameters associated with mechanical behaviour of materials from an engineering
context.
Young’s modulus is the most commonly used parameter in industry for TBCs to
describe their mechanical behaviour. It determines the coating’s response under a
state of tension or compression. Young’s modulus is required to evaluate the
­parameters describing the material mechanics in TBCs such as thermal stress and
residual stress.
Most of the studies on mechanical properties of TBCs have been based on evalu-
ating Young’s modulus and elastic anisotropy at low stresses, apart from studies
investigating hardness, creep behaviour, etc. It has been observed that ceramics
show up to three to ten times lower stiffness constants than the corresponding well-­
sintered materials [16]. Different Young’s moduli in different directions parallel and
perpendicular to the surface have also been observed. This anisotropy is attributed
to the preferred orientation of the planar defects present in the topcoat, which affects
the local compliance substantially and results in a lower value of measured Young’s
modulus [17]. Analytical models have also been developed which explain this
behaviour [16, 18]. Similar behaviour in tension and compression was assumed in
these models.
The evaluation of modulus provides indication of the coating integrity, porosity
and bonding quality between splats. It also provides an idea of the thermal stress
developed in the coating during operation since the modulus is roughly proportional
to the induced thermal stresses [19].
3.3 Mechanical Behaviour 25

3.3.3 Non-linear Properties

Recent developments have shown that ceramic coatings exhibit anelastic mechanical
response [20, 21]. Their behaviour both in tension and compression is strongly non-
linear. In general, increasing tensile stress results in a lower value of coating modu-
lus [22]. Generally, anelastic responses arise due to two factors: phase transformation
and geometrical condition [21]. Since YSZ coatings usually exhibit a stable tetrago-
nal structure during the whole range of operating temperature, phase transforma-
tions should not occur. Thus the anelasticity should be due to the geometrical
aspects of the coating.
The non-linearity seems to be driven by unique microstructural features present
in the TBCs, specifically micro-cracks and weak splat interfaces. The opening and
closing of cracks and sliding of sprayed lamellae over each other give rise to a non-­
linear response. The apparent stiffness decreases with increasing tensile stress as the
cracks faces open apart, while it increases with increasing compressive stress as the
cracks faces are closed together. The frictional sliding of unbonded interfaces
between the splats results in dissipated energy during the loading–unloading cycle
thus giving rise to hysteresis.
The anelastic response of a APS YSZ coating during the bilayer curvature mea-
surements using ex situ coating property (ECP) sensor is shown in Fig. 3.5 [2]; the
measurement set-up details are discussed further in Sect. 4.3. The coating, initially
under a state of compression after deposition, is heated from room temperature to
a certain temperature (stress transition from state A to B in Fig. 3.5). Due to the

Fig. 3.5 Anelastic response of APS YSZ coating during bilayer curvature measurements [2]
26 3 Characteristics of TBCs

difference in CTE between substrate and coating materials, thermal mismatch

stresses arise which result in the change of stress state in the coating from compres-
sion to tension during heating and then back to compression as the system is cooled
down to the start temperature. The non-linear behaviour of the coating during both
heating and cooling part of the cycle can be clearly noticed, as well as the hyster-
esis in the stress–strain curve. It should be remarked here that the results exhibit
nearly complete elastic recovery indicating anelastic response.

3.4 Interface Roughness

3.4.1 Roughness Relationship with Lifetime

Surface roughness is an essential requirement for an APS coating to adhere to the

substrate (bondcoat in case of TBCs) on which it is sprayed. Mechanical anchoring is
believed to be one of the key mechanisms responsible for adherence of the coating,
which would be enhanced by higher surface roughness [23]. If the roughness is too
low, the coating adhesion will not be adequate. On the other hand, if the roughness
is too high, large thermo-mechanical stresses would be induced on the topcoat,
especially due to TGO growth, and the coating could spall off. Additionally, local
aluminium depletion could occur near the local protrusions within the bondcoat
close to the topcoat–bondcoat interface. This could result in the formation of the
fast-growing detrimental spinel oxides and thus even higher stresses. Daroonparvar
et al. showed this phenomenon experimentally where Ni infiltrated through the
micro-cracks across the Al2O3 layer leading to the formation of mixed oxides (CSN)
which formed protrusions in the TGO that initiate failure mechanisms of the topcoat
[24]. Thus an optimised surface roughness is required for achieving a TBC with
long lifetime. Fauchais et al. have shown that the height of surface roughness hills
must be about one-third to one half of the mean splat diameter for good adhesion of
APS coatings [25]. The ideal level of bondcoat roughness for APS topcoats for good
adhesion of the coating is believed to be in the range Ra = 6–12 μm [3]. The bond-
coat roughness could be influenced by the spray parameters as well the bondcoat
powder feedstock size and size distribution [26–28].
Topcoat–bondcoat interface roughness is one of the important parameters which
determine the lifetime of a TBC. Eriksson et al. studied this effect by considering
four specimens with same chemistry but with different bondcoat roughness. It was
observed that increasing Ra resulted in a higher TCF lifetime [29]. These results
were in agreement with a detailed modelling work done earlier by Vassen et al.
where it was concluded that lower roughness results in longer cracks near the
topcoat–bondcoat interface, which would imply earlier failure [28]. However, in
another recent work done by Curry et al., two samples with same chemistry as well
as similar Ra were found to have significantly different lifetimes, which was under-
stood to be due to the presence of different topographical features [30]. These results
3.4 Interface Roughness 27

show that the traditionally used Ra is not sufficient to characterise the coatings
and more sophisticated procedures should be used which could characterise the
three-­dimensional (3D) surface profile in a more precise way.

3.4.2 Stress Inversion Theory

A theory for crack propagation mechanism for APS TBCs has been proposed in
earlier works, namely ‘stress inversion’ theory explained by using a simplified sinu-
soidal wave profile to represent the topcoat–bondcoat interface [28, 31, 32].
According to the proposed theory, in the initial as-sprayed state without a TGO
layer, tensile stresses exist in the hills while compressive stresses exist in the valleys
within the topcoat near the topcoat–bondcoat interface as shown in Fig. 3.6a. This
stress state is inverted as the TGO layer grows during thermal cyclic loading as
shown in Fig. 3.6b. Thus, a crack starts from the hill and propagates to the adjacent
valley as the TGO is formed joining the corresponding crack from the other side
eventually leading to the spallation of the topcoat as schematically illustrated by
the dashed line in Fig. 3.6. The thickness of the TGO when the stress inversion
takes place depends on several factors such as the loading conditions, geometry and
material [34, 35].
This effect which occurs due to different CTEs of different layers in the TBC
could be explained by considering the bondcoat hill as small cylinder surrounded by
a concentric cylindrical shell representing the topcoat and topcoat profile as small
cylinder surrounded by a concentric cylindrical shell representing the bondcoat for
the bondcoat valley [28]. In the as-sprayed state, the metallic bondcoat contracts
more due to higher CTE than the ceramic topcoat which leads to tensile stresses in
the topcoat near the hills and compressive stresses near the valleys. After the TGO

Fig. 3.6 A schematic illustration of the stress behaviour in the topcoat near the bondcoat surface
(a) in as-sprayed condition and (b) after TGO growth [33]
28 3 Characteristics of TBCs

layer forms, the stress state can be understood by replacing the bondcoat by TGO in
the concentric cylindrical shell model. Since the TGO is even more stiff (lower CTE)
than topcoat, compressive stresses are now introduced in the topcoat near the hills
and tensile stresses near the valleys. This explanation has been verified with the help
of finite element modelling in several works [32, 36].
The stress inversion theory has been observed to follow the trend in earlier works
when the time to stress inversion was compared with the experimental lifetime of
TBCs. A 2D sinusoidal profile representing the topcoat–bondcoat interface rough-
ness was used in finite element models by Vassen et al. to evaluate the residual
stresses developed in TBCs and it was observed that the time to stress inversion was
shorter for samples which failed earlier in experiments [28].

3.5 Oxide Formation

As soon as the TBC is put into operating conditions, the bondcoat starts to undergo
oxidation due to the exposure to high temperatures. The YSZ topcoat used typically
in a TBC is transparent to oxygen due to two effects: (1) zirconia is transparent to
oxygen flow due to the presence of vacancies and (2) the interconnected porosity
network present within the topcoat allows free flow of oxygen (air). Therefore, the
bondcoat metallic alloy is designed to act as a local aluminium reservoir allowing
the formation of slow growing α-alumina which could provide oxidation resistance
to the substrate. Alumina is the primary and most stable oxide formed for a
NiCoCrAlY bondcoat during operation [37]. This is due to the fact that alumina has
a lower formation energy than the other main element such as Ni, Co and Cr present
in the bondcoat [38]. The formation energy of alumina (ΔG°) is given by the follow-
ing relationship:

(
DG° = RT ln aAl 4 / 3 × PO 2 ) (3.6)

where R is the gas constant, T is the temperature in Kelvin, aAl is the activity of
aluminium and PO2 is the partial pressure of oxygen. The activity of aluminium
depends on the bondcoat composition.
The alumina layer is supposed to suppress the formation of other detrimental
oxides during the extended thermal exposure in service thus improving lifetime of
the TBC. However, other oxides such as CSN are formed at a later stage due to
microstructural defects such as cracks present within the alumina layer and deple-
tion of alumina within the bondcoat depending on several factors such as the com-
position of the bondcoat and the operating conditions as described by (3.6).
The oxide formation process at high temperatures typically during TCF testing is
usually categorised in three stages [38, 39]:
1. Transient stage: In this stage, oxides form at a rapid rate. Almost all of the
alloying element present in the bondcoat can be oxidised during this stage,
the variety mainly dependent on the chemical composition and microstructure of
3.6 Failure Mechanisms 29

the coating. This stage is typically very short, in the range of less than one hour
for Ni–Cr–Al alloys above 1,000 °C [40].
2. Steady stage: In this stage, alumina grows in a continuous and dense form linearly
which acts as a protective layer. The alumina layer could be cracked or spalled
off due to thermo-mechanical stresses and reformed [38].
3. Acceleration stage: When the aluminium content in the bondcoat becomes
lower than a critical level, other mixed oxides such as CSN start to form aggres-
sively leading eventually to failure.
In case of alumina, the oxide growth is believed to be mainly driven due to two
parallel diffusion mechanisms: inward diffusion of oxygen, mostly along grain bound-
aries, towards the TGO–bondcoat interface and outward diffusion of aluminium along
the already formed alumina particles towards the topcoat–TGO interface [37]. However,
it is generally accepted that the inward growth of alumina is the predominant step and
the reaction occurs primarily at the bondcoat–TGO interface [41, 42].
The growth of the TGO layer causes stress in the TBC both due to swelling of TGO
resulting in ‘growth stresses’ as well as stresses generated due to the mismatch in CTEs
between the different layers in the TBC. The growth stresses are usually relaxed quite
quickly due to the high creep rates in the ceramic topcoat layer at elevated temperatures
[34]. However, the thermal mismatch stresses, generated especially near the topcoat–
bondcoat interface when the TBC is cooled, could be detrimental and result in the
formation of cracks and/or elongation of pre-existing cracks near the topcoat–bondcoat
interface which could eventually lead to failure [34].
The growth of mixed oxides is highly disadvantageous for a TBC as these oxides
grow at a rapid rate and have an associated volume expansion which could result in
very high growth stresses leading to spallation and thus failure. It has been shown
that these oxides form protrusions in the TGO which could initiate failure mecha-
nisms of the topcoat [24]. Bulky mixed oxide clusters may form from individual
splats which are cut off from the rest of the bondcoat and quickly oxidise as they run
out of aluminium. It has been observed in previous work that these oxide clusters
are prone to cracking and CSN-nucleated cracks would assist the crack coalescence
process leading to the formation of large cracks that could result in spallation and
thus failure [39]. Eriksson et al. observed a significant reduction in TCF lifetime of
TBCs which formed more CSN [29].

3.6 Failure Mechanisms

Failure mechanisms in TBCs are very complex and are usually a mixture of several
mechanisms. The primary failure method is spallation of the coating due to a frac-
ture in the ceramic layer and/or the TGO layer. These types of failures occur if
either the local stresses increase, the material strength decreases or a combination of
the two. The stress can increase due to mismatch in CTE, temperature gradient and
TGO formation. The changes in stress distribution induce the propagation of
30 3 Characteristics of TBCs

pre-­existing cracks near the interface finally leading to cracks long enough to cause
spallation of the topcoat. The formation of mixed oxides can significantly enhance the
crack propagation and coalescence as discussed in detail in Sect. 3.5. The topcoat
could become more brittle due to sintering at high temperatures and/or propagation
of cracks, while the bondcoat could become more brittle due to depletion of
aluminium which would lead to growth of more brittle oxides.
Some of the other failure mechanisms of a TBC are: (1) damage induced by
particle impact, such as erosion and foreign object damage (FOD), (2) cracks
formed due to the penetration of deposits of calcium–magnesium–aluminosilicate
(CMAS) formed due to ingress of sand and dust present in the atmosphere into the
turbine engine [43, 44].

References

1. Curry N, Donoghue J (2012) Evolution of thermal conductivity of dysprosia stabilised thermal

barrier coating systems during heat treatment. Surf Coat Technol 2019:38–43
2. Dwivedi G (2011) On the anelastic behavior of plasma sprayed ceramic coatings: observations,
characterizations and applications. Ph.D. thesis, Stony Brook University, USA
3. Curry N (2014) Design of thermal barrier coatings. Ph.D. thesis, University West, Sweden,
No. 3.
4. Friis M (2002) A methodology to control the microstructure of plasma sprayed coatings. Ph.D.
thesis, Lund University, Sweden
5. Hasselman DPH (1978) Effect of cracks on thermal conductivity. J Compos Mater
12:403–407
6. Golosnoy IO, Cipitria A, Clyne TW (2009) Heat transfer through plasma-sprayed thermal
barrier coatings in Gas turbines: a review of recent work. J Therm Spray Technol 18(5–6):
809–821
7. Klemens PG (1969) Theory of thermal conductivity in solids. In: Thermal conductivity, vol. 1,
Tye RP (ed.) Academic, London, pp 1–68
8. Pawlowski L, Fauchais P (1992) Thermal transport properties of thermally sprayed coatings.
Int Mater Rev 37(6):271–289
9. Klemens PG, Gell M (1998) Thermal conductivity of thermal barrier coatings. Mater Sci Eng
A 245(2):143–149
10. Nicholls JR, Lawsona KJ, Johnstone A, Rickerby DS (2002) Methods to reduce the thermal
conductivity of EB-PVD TBCs. Surf Coat Technol 151–152:383–391
11. Golosnoy IO, Tripas SA, Clyne TW (2005) An analytical model for simulation of heat flow in
plasma-sprayed thermal barrier coatings. J Therm Spray Technol 14(2):205–214
12. Lim G, Kar A (2009) Modeling of thermal barrier coating temperature due to transmissive
radiative heating. J Mater Sci 44:3589–3599
13. Clyne TW, Gill SC (1996) Residual stresses in thermal spray coatings and their effect on
interfacial adhesion: a review of recent work. J Therm Spray Technol 5(4):401–418
14. Matejicek J, Sampath S, Dubsky J (1998) X-ray residual stress measurement in metallic and
ceramic plasma sprayed coatings. J Therm Spray Technol 7(4):489–496
15. Matejicek J, Sampath S, Brand PC, Prask HJ (1999) Quenching, thermal and residual stress in
plasma sprayed deposits: NiCrAlY and YSZ coatings. Acta Mater 47(2):607–617
16. Kroupa F (2007) Nonlinear behavior in compression and tension of thermally sprayed ceramic
coatings. J Therm Spray Technol 16(1):84–95
17. Guo S, Kagawa Y (2004) Young’s moduli of zirconia top-coat and thermally grown oxide in a
plasma-sprayed thermal barrier coating system. Scr Mater 50:1401–1406
References 31

18. Sevostianov I, Kachanov M (2001) Plasma-sprayed ceramic coatings: anisotropic elastic and
conductive properties in relation to the microstructure; cross-property correlations. Mater Sci
Eng A 297(1–2):235–243
19. Matejicek J, Sampath S (2003) In situ measurement of residual stresses and elastic moduli in
thermal sprayed coatings Part 1: apparatus and analysis. Acta Mater 51:863–872
20. Liu Y, Nakamura T, Srinivasan V, Vaidya A, Gouldstone A, Sampath S (2007) Non-linear
elastic properties of plasma-sprayed zirconia coatings and associated relationships with
processing conditions. Acta Mater 55(14):4667–4678
21. Liu YJ, Nakamura T, Dwivedi G, Valarezo A, Sampath S (2008) Anelastic behavior of plasma-­
sprayed Zirconia coatings. J Am Ceram Soc 91(12):4036–4043
22. Nakamura T, Liu YJ (2007) Determination of nonlinear properties of thermal sprayed ceramic
coatings via inverse analysis. Int J Solids Struct 44(6):1990–2009
23. Nusair Khan A, Lu J, Liao H (2003) Effect of residual stresses on air plasma sprayed thermal
barrier coatings. Surf Coat Technol 168:291–299
24. Daroonparvar M, Hussain MS, Yajid MAM (2012) The role of formation of continues
thermally grown oxide layer on the nanostructured NiCrAlY bond coat during thermal exposure
in air. Appl Surf Sci 261:287–297
25. Fauchais P, Etchart-Salas R, Rat V, Coudert JF, Caron N, Wittmann-Teneze K (2008)
Parameters controlling liquid plasma spraying solutions, sols, or suspensions. J Therm Spray
Technol 17(1):31–59
26. Nowak W, Naumenko D, Mor G, Mor F, Mack DE, Vassen R, Singheiser L, Quadakkers WJ
(2014) Effect of processing parameters on MCrAlY bondcoat roughness and lifetime of
APS–TBC systems. Surf Coat Technol 260:82–89
27. Rajasekaran B, Mauer G, Vaßen R (2011) Enhanced characteristics of HVOF-sprayed MCrAlY
bond coats for TBC applications. J Therm Spray Technol 20(6):1209–1216
28. Vaßen R, Kerkhoff G, Stöver D (2001) Development of a micromechanical life prediction
model for plasma sprayed thermal barrier coatings. Mater Sci Eng A 303:100–109
29. Eriksson R, Sjöström S, Brodin H, Johansson S, Östergren L, Li X-H (2013) TBC bond coat–top
coat interface roughness: Influence on fatigue life and modelling aspects. Surf Coat Technol
236:230–238
30. Curry N, Markocsan N, Östergren L, Li X-H, Dorfman M (2013) Evaluation of the lifetime
and thermal conductivity of dysprosia-stabilized thermal barrier coating systems. J Therm
Spray Technol 22(6):864–872
31. Pindera M-J, Aboudi J, Arnold SM (2000) The effect of interface roughness and oxide film
thickness on the inelastic response of thermal barrier coatings to thermal cycling. Mater Sci
Eng A 284:158–175
32. Jinnestrand M, Sjöström S (2001) Investigation by 3D FE simulations of delamination crack
initiation in TBC caused by alumina growth. Surf Coat Technol 135:188–195
33. Gupta M, Eriksson R, Sand U, Nylén P (2014) A diffusion-based oxide layer growth model
using real interface roughness in thermal barrier coatings for lifetime assessment. Surf Coat
Technol. doi 10.1016/j.surfcoat.2014.12.043
34. Vaßen R, Giesen S, Stöver D (2009) Lifetime of plasma-sprayed thermal barrier coatings:
comparison of numerical and experimental results. J Therm Spray Technol 18(5–6):
835–845
35. Bäker M (2012) Finite element simulation of interface cracks in thermal barrier coatings.
Comput Mater Sci 64:79–93
36. Ahrens M, Vaßen R, Stöver D (2002) Stress distributions in plasma-sprayed thermal barrier
coatings as a function of interface roughness and oxide scale thickness. Surf Coat Technol
161:26–35
37. Busso EP, Lin J, Sakurai S, Nakayama M (2001) A mechanistic study of oxidation-induced
degradation in a plasma-sprayed thermal barrier coating system. Part I: model formulation.
Acta Mater 49:1515–1528
38. Yuan K (2014) Oxidation and corrosion of new MCrAlX coatings: modelling and experiments.
Ph.D. thesis, Linköping University, Sweden
32 3 Characteristics of TBCs

39. Chen WR, Wu X, Marple BR, Patnaik PC (2006) The growth and influence of thermally grown
oxide in a thermal barrier coating. Surf Coat Technol 201:1074–1079
40. Haynes JA, Rigney ED, Ferber MK, Porter WD (1996) Oxidation and degradation of a plasma-­
sprayed thermal barrier coating system. Surf Coat Technol 86–87:102–108
41. Hille TS, Turteltaub S, Suiker ASJ (2011) Oxide growth and damage evolution in thermal
barrier coatings. Eng Fract Mech 78:2139–2152
42. Ma K, Schoenung JM (2011) Isothermal oxidation behavior of cryomilled NiCrAlY bond
coat: homogeneity and growth rate of TGO. Surf Coat Technol 205:5178–5185
43. Vaßen R, Jarligo MO, Steinke T, Mack DE, Stöver D (2010) Overview on advanced thermal
barrier coatings. Surf Coat Technol 205:938–942
44. Drexler JM, Chen C-H, Gledhill AD, Shinoda K, Sampath S, Padture NP (2012) Plasma
sprayed gadolinium zirconate thermal barrier coatings that are resistant to damage by molten
Ca–Mg–Al–silicate glass. Surf Coat Technol 206:3911–3916
Chapter 4
Experimental Methods

4.1 Microstructure Characterisation

Various techniques have been used in the past for qualitative and/or quantitative
analysis of microstructure of TBCs. Mercury intrusion porosimetry (MIP) enables
the measurement of total porosity for open pores and the evaluation of pore size dis-
tribution over a wide range [1, 2]. Small-angle neutron scattering (SANS) can obtain
statistical results for small scale defects as well as orientation information [3–5].
Image analysis is a robust, reliable and inexpensive method to characterise TBC
microstructures from cross-­section images which can be used to obtain information
about porosity, pores and cracks distribution and orientation, etc. [6]. The commonly
used techniques to capture microstructure cross-section images of TBCs are LOM
and SEM.
During the image analysis procedure, several images are taken along the coating
cross section to capture the variation in the microstructure. These images are then con-
verted to binary format by thresholding using an image analysis software such as
Aphelion (ADCIS, France) and Image J (National Institutes of Health, USA). The
images have to be taken carefully at a certain magnification so as to capture the relevant
coating microstructure details. High magnification images result in a loss of global
coating information, while too low magnification images are unable to capture small
scale details present in the microstructure. Other factors which could affect the image
analysis results are threshold level, microscope effects (such as image brightness and
contrast and operating mode) and coating location where the image has been taken [6].

4.2 Thermal Conductivity Measurements

The most widely accepted method for evaluating thermal conductivity of TBCs is
laser flash analysis (LFA) technique [7]. The state-of-the-art measurement set-up is
shown in Fig. 4.1.

© The Author(s) 2015 33

M. Gupta, Design of Thermal Barrier Coatings, SpringerBriefs in Materials,
DOI 10.1007/978-3-319-17254-5_4
34 4 Experimental Methods

Fig. 4.1 The laser flash analysis apparatus used for measuring thermal conductivity (Courtesy of
NETZSCH, Germany)

In this method, a laser pulse is shot at the substrate face of the sample and the
resulting temperature increase is measured at the other face of the sample with an
infrared detector. The signal is then normalised and the thermal diffusivity is calcu-
lated using an equation based on one of the existing models. One such equation is
as follows:

a = ( 0.1388L2 ) / t( 0.5) (4.1)

where α is the thermal diffusivity, L is the thickness of the sample and t(0.5) is
the time taken for the rear face of the sample to reach half of its maximum rise.
Thermal conductivity (K) can then be calculated if the density (ρ) and specific heat
capacity (Cp) of the coating are known using the following relation:
K = a . Cp . r (4.2)

Samples are normally coated with a thin layer of graphite or gold before the
measurements. Since zirconia is translucent to light in the wavelength of the laser,
4.3 Young’s Modulus Measurements 35

the presence of graphite layer is essential to prevent the laser pulse from travelling
through the ceramic layer.
Density can be measured using one of the several existing methods such as
Archimedes displacement, mercury porosimetry and gas absorption methods. These
measurements are based on the assumption that all porous features present in the
coating are interconnected, which is not true in every case. An indirect approach for
evaluating density could be to measure the porosity level from image analysis of
microstructure cross-section images and then calculating the density based on the
density of bulk material. Usually, the density evaluated using image analysis method
is lower than the mercury porosimetry method, which could be attributed to the
effect of unconnected or sealed porosity detected by the image analysis method.
Specific heat capacity data of the coatings can be used from existing databases,
though it is recommended to measure it for each coating material using differential
scanning calorimetry (DSC) for accuracy.
Thermal conductivity is measured either on free-standing ceramic coating, which
can be peeled off from the substrate, or on the complete TBC system. In the latter
case, thermal conductivity of topcoat is calculated afterwards by measuring the
thicknesses of individual layers and thermal conductivity values of only the substrate
and substrate sprayed with bondcoat. Thicknesses are measured from microstructure
cross-section images using an optical microscope. The measurement of thickness
can be difficult due to high roughness of the interfaces in TBC systems and can result
in high errors in final measurements since thermal diffusivity is proportional to the
square of thickness [7].
LFA technique is essentially non-contact and the equipment is available
commercially from several companies. A major drawback is that only small flat
samples can be used for measurement and thus measurement on a real component
is not possible.

4.3 Young’s Modulus Measurements

Traditional techniques used for measuring Young’s modulus of APS topcoats are
micro-indentation, four-point bending, etc. Another technique which has been
developed in recent years is bilayer curvature measurements. The ECP sensor which
consists of non-contact displacement lasers and thermocouples to monitor the
displacement and temperature simultaneously at the back of the sample during a
heating and cooling cycle is used for these measurements [8]. The bilayer curvature
measurement set-up is shown in Fig. 4.2.
The techniques used to measure Young’s modulus of TBCs can be classified into
three categories [9]:
1. Mechanical loading methods: These techniques are based on the direct measurement
of the deformation of the test material upon application of an external force.
Examples include bending tests of coating system or a freestanding coating,
tensile or compression tests and indentation tests.
36 4 Experimental Methods

Fig. 4.2 The bilayer curvature measurement set-up using the ex situ coating property sensor
(Courtesy of ReliaCoat Technologies, NY, USA)

2. Resonance vibration methods: These techniques are based on the principle that the
resonance frequency of a material depends on its elastic modulus of the material.
Examples include acoustic emission techniques [10].
3. Pulse-echo methods: These techniques are based on the principle that the velocity
of sound in a material depends on the elastic modulus of the material. Examples
include laser-induced ultrasonic techniques [11].
It has been observed in earlier work that the modulus value depends highly on the
technique used to evaluate it and it could be different even if the test specimens are
identical between different techniques [12]. This effect, which is related to the local
variations of the microstructure within the coating, could be explained by the scale of
the testing technique which ranges from microscopic (nanoindentation) to mesoscopic
(micro-indentation) and macroscopic (bending tests, ultrasonic testing) dimensions [12].
Therefore, elastic modulus needs to be considered as an engineering value for quantify-
ing and ranking coating systems rather than as a fundamental property.

4.4 Roughness Measurements

The traditional technique to measure surface roughness is to use a surface profilometer.

Surface profilometer is a contact technique which determines the profile of a surface
by dragging a stylus along it. The equipment is able to output the 2D roughness
parameters such as Ra, Rq and Rz. This technique is cheap and very simple to use.
However, recent works have shown the 2D surface roughness parameters might
not be sufficient to characterise a coating [13]. Other techniques such as white light
4.5 Lifetime Testing 37

Table 4.1 Some of the ISO 25178 feature parameters calculated for HVOF and APS bondcoat
samples
Parameter (units) Parameter description
Spd (1/mm2) Density of hills (number of hills per unit area)
Spc (1/mm) Arithmetic mean hill curvature (arithmetic mean of the principal curvatures
of hills within a definition area)
Sha (mm2) Mean hill area (average area of the hills connected to the edge at a particular
height)

interferometry and laser stripe projection could be useful to analyse these coatings.
The major advantages of using these techniques are: (1) A 3D profile could be cap-
tured with these techniques which could provide better visualisation of the surface
features. (2) Very low resolution can be obtained with these techniques; for exam-
ple, the vertical resolution obtained with interferometry is in the range of nanome-
tres. (3) The newly formulated ISO 25178 parameters could be measured and
calculated with the help of these techniques thus enabling to characterise the 3D
profile quantitatively. The disadvantages of using these techniques are that they are
very expensive and the data analysis could be a complex procedure.
The feature parameters are a new family of parameters which has been integrated
into the ISO 25178 standard. Feature parameters are derived by creating segmenta-
tion motifs over a surface which makes it possible to identify specific areas such as
hills and valleys. Some of the ISO feature parameters are given in Table 4.1.
Quantifying the density, size and shape of the hills and valleys of the bondcoat
surface could be highly beneficial to assess the stresses’ characteristics near the
topcoat–bondcoat interface.
Figure 4.3a shows a surface profile of a bondcoat sample captured using white
light interferometry. The figure shows the presence of large hills over the surface
which are suspected to be partially melted/unmelted particles formed during the
spray process. Figure 4.3b shows segmentation motifs captured using stripe projec-
tion technique for the same sample. The presence of the hills over the surface could
be quantitatively analysed with the help of this image.

4.5 Lifetime Testing

Assessing the lifetime of TBCs is a challenging task since testing them in real
environments is very time consuming and expensive. Accelerated and simplified tests
are used to reduce testing times by exposing the TBCs to harsher environments than
reality. Due to the over-simplification of testing methods, there can be large discrepan-
cies between the performance in testing and operating conditions. Nevertheless,
the lifetime tests provide a fair idea of the coating behaviour and can be used to
judge different coatings. Common methods for testing the lifetime of TBCs are TCF
and thermal shock tests. Both these methods are well practised in the industry.
38 4 Experimental Methods

a
50
45
40
35
30
25
20
0 15
50
10 10
0
15
0 5
20
0
25 0

0
0

55
30
-5

0
0

50
35

0
0 -10

45
40

0
0

40
µm 45 -15

µm 50
0

3
50 -20

0
30
0
55 -25

0
0

25
60

0
0 -30

20
65 0
15
0 -35
70 0
0 10 -40
75
0
50

Fig. 4.3 (a) Bondcoat surface profiles captured using white light interferometry technique and
(b) segmentation motifs captured using stripe projection technique [14]

In TCF tests, the samples are typically first heated in a furnace at high temperatures
around 1,100 °C for one hour and then cooled down for ten minutes using com-
pressed air to approximately 100 °C. These two steps are repeated until failure. The
criterion for failure is deemed to be more than 20 % visible spallation of the coating.
4.5 Lifetime Testing 39

Fig. 4.4 Samples during TCF testing showing spallation of a few samples; the indicated areas
show the failed parts of the samples

A visual record of the samples’ surface is made at the end of each cycle using a
webcam which is used later for calculating the number of cycles to failure. Figure 4.4
shows one such photograph taken during the testing where it can be seen that a few
of the samples have the topcoats spalled off while a few are still intact. The failed
areas in the samples are indicated in the figure.
Thermal shock testing simulates the rapid heating and cooling experienced by
coatings during service. As the name suggests, the coatings are heated up to a speci-
fied temperature and then cooled down rapidly with air or by dipping them into
water. Burner rig test is a common way of testing thermal shock behaviour. In this
test, samples are heated on the coating face with a combustion burner flame and then
cooled down with compressed air. Samples are cooled on the back face with water
or air to set up a thermal gradient within the TBC system. A typical cycling time for
burner rig tests is five minutes heating and two minutes cooling.
Figure 4.5 shows a microstructure cross-section image of part of the topcoat–
bondcoat interface after failure during TCF test. The topcoat zirconia layer was
sprayed by APS while the bondcoat NiCoCrAlY layer was sprayed by HVOF tech-
nique. The dark grey layer in the image between the topcoat and bondcoat consists
of TGO formed during the test.
Both TCF and thermal shock testing methods have their own limitations. TCF
tests have the benefit that they provide information about sintering of topcoat, TGO
growth in TBC system and the topcoat’s ability to sustain the induced stresses due
to TGO growth and thermal mismatch. In this way, results obtained from TCF tests
40 4 Experimental Methods

Fig. 4.5 Topcoat–bondcoat interface after failure showing the TGO growth

are highly dependent on the bondcoat material and structure. The major limitation
of TCF tests is the time taken for the tests to be completed, since it can take more
than a month to obtain lifetime results. Another limitation is that since the samples
are normally maintained in an isothermal heating environment, a thermal gradient is
not present in the TBC system which is present in service conditions. Homogeneous
temperature state can lead to higher stresses and thus lower lifetime than reality [15].
Thermal shock tests induce large thermal stresses in the topcoat–bondcoat interface
and they are mainly dependent on the strain tolerance of topcoats. There is little
time in these tests for any significant effect of bondcoat oxidation or sintering of
topcoats on coating lifetime. Though thermal shock tests are much faster as they
normally take less than a week to be completed, they have a disadvantage that they
are much more severe than reality. An alternative testing method which involves a
combination of TCF and thermal shock tests was suggested in an earlier work for
better representation of actual engine conditions [16].

References

1. Portinhaa A, Teixeira V, Carneiro J, Martins J, Costa MF, Vassen R, Stoever D (2005) Surf
Coat Technol 195:245–251
2. Cernuschi F, Golosnoy IO, Bison P, Moscatelli A, Vassen R, Bossmann H-P, Capelli S (2013)
Microstructural characterization of porous thermal barrier coatings by IR gas porosimetry and
sintering forecasts. Acta Mater 61(1):248–262
References 41

3. Kulkarni A, Wang Z, Nakamura T, Sampath S, Goland A, Herman H, Allen J, Ilavsky J, Long G,

Frahm J, Steinbrech RW (2003) Comprehensive microstructural characterization and predic-
tive property modeling of plasma-sprayed zirconia coatings. Acta Mater 51(9):2457–2475
4. Wang Z, Kulkarni A, Deshpande S, Nakamura T, Herman H (2003) Effects of pores and inter-
faces on effective properties of plasma sprayed zirconia coatings. Acta Mater 51(18):
5319–5334
5. Allen AJ, Ilavsky J, Long GG, Wallace JS, Berndt CC, Herman H (2001) Microstructural
characterization of yttria-stabilized zirconia plasma-sprayed deposits using multiple small-­
angle neutron scattering. Acta Mater 49(9):1661–1675
6. Tan Y (2007) Thermal design and microstructure-based property assessment for thermal spray
coating systems, Ph.D. thesis, Stony Brook University, USA
7. Taylor RE (1998) Thermal conductivity determinations of thermal barrier coatings. Mater Sci
Eng A 245(2):160–167
8. Matejicek J, Sampath S, Brand PC, Prask HJ (1999) Quenching, thermal and residual stress in
plasma sprayed deposits: NiCrAlY and YSZ coatings. Acta Mater 47(2):607–617
9. Waki H, Oikawa A, Kato M, Takahashi S, Kojima Y, Ono F (2014) Evaluation of the accuracy
of Young’s moduli of thermal barrier coatings determined on the basis of composite beam
theory. J Therm Spray Technol 23(8):1291–1301
10. Ma XQ, Mizutani Y, Takemoto M (2001) Laser-induced surface acoustic waves for evaluation
of elastic stiffness of plasma sprayed materials. J Mater Sci 36:5633–5641
11. Patsias S, Tassini N, Lambrinou K (2006) Ceramic coatings: effect of deposition method
on damping and modulus of elasticity for yttria-stabilized zirconia. Mater Sci Eng A 442:
504–508
12. Margadant N, Neuenschwander J, Stauss S, Kaps H, Kulkarni A, Matejicek J, Rössler G
(2006) Impact of probing volume from different mechanical measurement methods on elastic
properties of thermally sprayed Ni-based coatings on a mesoscopic scale. Surf Coat Technol
200(8):2805–2820
13. Curry N, Markocsan N, Östergren L, Li X-H, Dorfman M (2013) Evaluation of the lifetime
and thermal conductivity of dysprosia-stabilized thermal barrier coating systems. J Therm
Spray Technol 22(6):864–872
14. Gupta M, Skogsberg K, Nylén P (2014) Influence of topcoat-bondcoat interface roughness on
stresses and lifetime in thermal barrier coatings. J Therm Spray Technol 23(1–2):170–181
15. Ranjbar-Far M, Absi J, Shahidi S, Mariaux G (2011) Impact of the non-homogenous temperature
distribution and the coatings process modeling on the thermal barrier coatings system. Mater
Des 32:728–735
16. Bolcavage A, Feuerstein A, Foster J, Moore P (2004) Thermal shock testing of thermal barrier
coating/bondcoat systems. J Mater Eng Perform 13(4):389–397
Chapter 5
Modelling of Properties of TBCs

Since the macroscopic properties of TBCs are highly dependent on their microscopic
structure, modelling of properties of TBCs usually consists of a representation of the
present porosity dispersed within a solid phase. Several studies have been performed
in the past using both analytical and numerical models for evaluating thermal–
mechanical properties of TBCs; a few of them are discussed in Sects. 5.1 and 5.2. The
approach used by the author in a recent work is described in Sects. 5.3 and 5.4, and the
results using this approach in a recent work are presented in Sect. 5.5 [1–4].

5.1 Thermal Conductivity

5.1.1 Analytical Models

Since the end of nineteenth century, several analytical models have been developed
for determining the thermal conductivity of multiphase solids, especially in porous
materials. Brief reviews of such works have been done earlier [5–10]. Most of these
models consisted of randomly distributed non-interacting pores or ellipsoids, or
periodic structures. The first model developed by Maxwell–Eucken is given as [5]
k tot 1- P
= (5.1)
k d 1 + 0.5P

where Κtot is the thermal conductivity of a porous material, Κd is the thermal conduc-
tivity of a dense material and P is the fraction of porosity, assuming that the thermal
conductivity of pores Κp is small compared to Κd. More recently, another model was
developed by Klemens which is given as [11]
k tot 4
=1 - P (5.2)
kd 3

© The Author(s) 2015 43

M. Gupta, Design of Thermal Barrier Coatings, SpringerBriefs in Materials,
DOI 10.1007/978-3-319-17254-5_5
44 5 Modelling of Properties of TBCs

Hasselman [12] developed a model which calculated the effect of cracks of various
orientations on the thermal conductivity of solid materials. It was concluded that
maximum thermal insulation is obtained with cracks perpendicular to the direction
of heat flow while cracks parallel to the direction of heat flow have no effect on
thermal conductivity.
In this book, focus is placed on models based on plasma-sprayed coatings.
The first analytical model for predicting thermal conductivity of plasma-sprayed
coatings was proposed by McPherson [13] which involved regions of good and poor
contact between lamellae where the regions of poor contact act as thermal resis-
tances. The model can be given as [13]
k tot 2 fd
= (5.3)
kd pa

where f is the fraction of ‘true contact’, δ is the lamellae thickness and a is the radius
of individual contact areas. Li et al. [14] further developed this model and proposed
quantitative structural parameters to characterise the deposit lamellar structure
instead of porosity content, the most important parameter being the bonding ratio at
the interfaces between lamellae. Thus, they incorporated the thermal resistance of
the lamellae into the model. Boire-Lavigne et al. [15] also included the oxide layer
resistance in the contact areas in the model. They used this analogy to determine the
thermal diffusivity of plasma-sprayed tungsten based coatings, where the geometri-
cal parameters were calculated from an image analysis procedure.
Bjorneklett et al. [16] compared experimental values of thermal diffusivity for
ZrO2-7 % Y2O3 and Al2O3-3 % TiO2 plasma-sprayed coatings with analytical models
based on effective medium theories. Models representing three different microstruc-
tures were considered—a continuous ceramic with dispersed pores, a continuous ceramic
matrix with continuously interconnected pores and dispersed ceramic loosely bonded
together. It was shown in this work that last two microstructure models conformed
­better to the theoretical values compared to the first one. This implied that the thermal
conductivity model consisting of an agglomerate of ceramic particles behaves similar
to reality compared to a solid ceramic sheet with pores. This behaviour is also in agree-
ment with the formation process of a plasma-­sprayed coating.
Sevostianov et al. [17] developed a model which gave an explicit relation between
the anisotropic thermal conductivities and the microstructure parameters of plasma-­
sprayed TBCs. The microstructure was assumed to be composed of two families of
penny-shaped cracks—horizontal and vertical. The orientational scatter was also
accounted for in the model by an appropriate orientation distribution function.
The effective conductivity was dependent mainly on the crack densities and their
orientational scatters, and the overall porosity (P) played only a secondary role [17].
The thermal conductivity in the direction perpendicular to the substrate was given
by the expression [17]
k tot 8a v
=1 - (5.4)
kd 3 (1 - P )
5.1 Thermal Conductivity 45

where αv is the component of crack density tensor in the direction perpendicular to

the substrate that incorporates the crack densities of both pore families and their
orientational scatters. An explicit correlation between elastic and conductive prop-
erties, dependent only on the elastic and conductive properties of the bulk material,
was also derived later (Ref [45]). The simplified expression is given as [18]
Ed - Etot k - kd
» 2.14 tot (5.5)
Etot kd

where Ed is Young’s modulus of bulk material and Etot is Young’s modulus in the
direction normal to the substrate. More accurate correlations for the two cases of ide-
ally parallel and randomly oriented cracks were also given later, which indicated
that the actual cross-property factor would be slightly lower than the one in the last
equation [19]. Equations (5.4) and (5.5) were also verified on YSZ coatings by using
the image analysis data from photomicrographs of the coatings and comparing the
predicted values with experimental values which were in good agreement [19].
Lu et al. [20] made a model which included different kinds of pore morpholo-
gies—randomly oriented pores, aligned but spatially random pores, periodic pores
and zigzag pores, with distributed pore shapes. Finite element method (FEM) was
used to compute the conductivity related to zigzag pores. They applied the model on
TBCs made by EB-PVD.
Golosnoy et al. [6] developed an analytical model which comprised solid layers
separated by thin, periodically bridged, gas-filled voids. The model was based on
dividing the material into two independent zones—one characterised by unidirec-
tional serial heat flow occurring through lamellae and pores and the other by chan-
nelled conduction through the bridges between the lamellae. Radiative heat transfer
and the effect of gas pressure inside the pores were also included in the model, and
thermal conductivity was computed over a range of temperatures. Cipitria et al. [21]
used this geometrical model to develop a sintering model based on the application of
the variational principle to diffusional phenomena. It also accounted for the effects of
surface diffusion, grain boundary diffusion and grain growth. Good agreement was
found between these two models and experimental results.

5.1.2 Numerical Models

Hollis [22] developed a model using FEM to compute the thermal conductivity of
VPS and APS tungsten coatings. The domain of the model was a 2D image recorded
using SEM. The image was thresholded based on the grey scale level in the micro-
graphs. The dark phase represented the pores while the light phase represented
tungsten. A finite element mesh was created over the image. It was observed that
VPS coatings were well represented by the 2D model while APS coatings were not
due to their complex microstructure [22]. A method to compensate for 3D pore
structure effects was also proposed by choosing the maximum effective pore length
46 5 Modelling of Properties of TBCs

for APS coatings and sectioning and shifting the model prior to performing the node
temperature calculation [22].
A finite difference method-based approach was given by Dorvaux et al. [7] by
developing software for the computation of the thermal conductivity of porous coat-
ings from binary images of real material cross sections. This approach takes into
account the complex morphology of the ceramic. It was used to determine the con-
tribution of each pore family (globular pores, cracks etc.) to thermal insulating
capabilities with the help of image analysis. It was concluded that major heat insula-
tion is provided by the cracks oriented perpendicular to the heat flow direction [7].
This approach was used later as an alternative to diffusivity measurements for rank-
ing coatings according to their heat insulation capacity with regard to the morphol-
ogy [23]. It was also used to study the relationship between sintering effects and
thermal conductivity increase and to study the effect of pressure on thermal conduc-
tivity of TBCs [24]. This approach was associated with a morphology generator to
develop the software Tbctool (ONERA, France) for building a predictive tool which
can be used to generate plasma-sprayed TBC-like morphology. This tool has been
discussed in detail in Sect. 5.4.
Bartsch et al. [25] compared the results from finite difference (Tbctool) and finite
element simulations for determining the thermal response from binary images of
EB-PVD TBC microstructures. It was observed that the results are more accurate
with decreasing mesh size and when the calculated section area of the microstruc-
ture image is larger. The finite element codes resulted in higher conductivity values
compared to finite difference codes, for the same sections and similar mesh width.
It was concluded that finite difference codes take less computer memory for calcula-
tions for larger models [25].
Kulkarni et al. [26] used SANS to study the effect of material feedstock charac-
teristics (powder morphology) on the properties of plasma-sprayed YSZ coatings.
A 2D finite element model representing the coating microstructure was built from
the volumetrically averaged information available from the SANS data. The voids
were divided into three families—interlamellar pores (assumed to be hexagonal in
shape), intrasplat cracks and globular or irregular pores. Aspect ratios were assumed
for each void family and the volume fractions, mean pore dimensions and orienta-
tions were taken from anisotropic multiple SANS data. The predicted thermal con-
ductivity was higher than the experimentally ­ measured values due to lack of
information on the splat boundaries and pore size distribution [26].
Tan et al. [27] used processed SEM images of coating microstructure to generate
a 2D finite element mesh and predict the effective thermal conductivity using a
commercial finite element code for YSZ, molybdenum and NiAl coatings. It was
observed that the modelling procedure was not sensitive to slight changes in the
threshold level of the images. For YSZ coatings, the predicted thermal conductivity
values were higher than the values measured by LFA, but they still followed the
same trends, for both before and after annealing the coatings [27].
Liu et al. [28] developed a 3D finite element model which included spherical
pores and unmelted particles, and ellipsoidal cracks, each inside a unit cell.
5.2 Young’s Modulus 47

The unmelted particles were assumed to be completely debonded from the substrate.
The unit cell model was used to predict the thermal insulation behaviour of different
features on YSZ plasma-sprayed coatings.
Qunbo et al. [29] used digital image processing to create a finite element mesh
over a real microstructure image to predict the thermo-mechanical properties of
TBCs and concluded that transverse cracks in the coating are most significant for
thermal insulation.

5.2 Young’s Modulus

5.2.1 Analytical Models

Li et al. [30] developed an idealised model for estimating Young’s modulus of thermal
sprayed ceramic coatings consisting of the stacking of few micrometre-thick lamel-
lae using circular plate bending theory. Two components of elastic strain of the
coating under tensile stress were considered in the model, one related to the local-
ised elastic strain at the regions of contact area between lamellae and the other
related to the elastic bending of the lamellae between bonded regions. It was shown
that the bending component becomes significant only when the percentage bonding
ratio between lamellae becomes less than 40 %.
Sevostianov et al. [31] developed a model for Young’s modulus as a function of
microstructural parameters similar to the one described in Sect. 5.1.1 for thermal
conductivity and proposed an equation as given below:

( )
é 8 ( 4 -n 0 ) 1 -n 02 a ù
-1

Etot = Ed ê1 + v
ú (5.6)
ê 3 ( 2 -n 0 ) 1- P ú
ë û

where ν0 is Poisson’s ratio of the material; other parameters are defined in earlier
equations.
Azarmi et al. [32] investigated APS deposited alloy 625 coatings and compared
Young’s modulus measured by uniaxial tension test with several analytical models
available for predicting Young’s modulus of porous materials and a finite element
modelling technique OOF (discussed in Sect. 5.3.2). A significant difference was
observed between the predicted values by analytical models and experimentally
measured values of the elastic modulus.

5.2.2 Numerical Models

Michlik et al. [33] used a FEM approach using a finite element package OOF
pre-­processor to generate the mesh and employ an in-house developed solver using
extended finite element method (XFEM). The XFEM approach accounted for the
48 5 Modelling of Properties of TBCs

existence of cracks in TBCs which enabled to study the effect of sintering of coating
on elastic modulus.
Amsellem et al. [34, 35] used both 2D and 3D analyses of APS microstructures
to determine Young’s modulus of the coatings. Two-dimensional finite element
meshes were created from SEM images of microstructures while 3D microstruc-
tures were obtained using X-ray microtomography (XMT). The limitations of both
2D and 3D modelling approaches were discussed and it was concluded that the 2D
approach was suitable for characterising mechanical properties due to the low reso-
lution of XMT [34]. Three-dimensional approach was assessed to be beneficial for
studying the shape and orientation of pores as well as to simulate the stress concen-
tration in the coating.
Apart from the numerical techniques discussed above, another finite element
technique which has been used extensively to determine the thermal–mechanical
properties of APS coatings is OOF, as described in the following section.

5.3 Finite Element Modelling

The over-simplification of porosity description in the analytical models limits the

predictive capacity and the certainty for the complex interconnected porous struc-
tures. Numerical schemes seem to be more promising in such cases as they involve
a material structure which is close to the real structure [7, 32]. In cases where the
spatial geometry is too complex for the usage of analytical models, numerical mod-
elling such as FEM and finite difference method (FDM) is beneficial.
Two important issues must be considered for numerical modelling of complex
microstructural features present in APS coatings. First, the model sample volume
must be large enough to contain sufficient microstructural details and hence repre-
sent the average property of the sample. Second, fine details of the microstructural
features have to be captured within the model volume so that the actual properties
are estimated precisely.

5.3.1 Basics of FEM and FDM

Both FEM and FDM are numerical approximation techniques to solve a set of equa-
tions which are typically not easily solvable analytically, for example due to com-
plicated irregular geometry, complex material composition, etc. In both methods,
the geometry is fractioned in small basic units, the problem is set up individually in
each basic unit and then solved together to calculate the result. Only lines/squares/
cubes can be used as basic units in FDM, whereas arbitrarily shaped basic units can
be used in FEM thus enabling better and less complex description of the geometry.
The FEM hence requires less computational power and has a better accuracy than FDM.
5.3 Finite Element Modelling 49

In addition, FEM is more suitable as an adaptive method since it easily facilitates for
local refinements of the solution.
FEM is applied to most engineering problems either by writing a custom-made
finite element program using softwares such as Matlab (MathWorks Inc., MA,
USA) or by using an already developed commercial finite element program.
Commercial finite element tools have a variety of built-in features for modelling
different engineering problems but the cost associated with purchasing these tools
and learning the way how to use them is usually high. Custom-made finite element
programs might have a small cost initially but they handle a small variety of prob-
lems at a time and are less flexible. The solution of a finite element problem involves
the construction of a matrix equation of type AX = B where A is the stiffness matrix,
X is the vector containing all the unknowns and B is the vector that depends on the
boundary condition.

5.3.2 Image Based Finite Element Model

The possibility to determine a material’s macroscopic properties based on its

microstructure is of great importance to materials science. The object-oriented
finite element analysis (OOF) code, developed at the Center for Computational and
Theoretical Materials Science (CTCMS) at US National Institute of Standards and
Technology (NIST), enables complex two-dimensional microstructures to be
modelled using images of the actual microstructures. Its current version, OOF2, is
available as public domain software [36]. This technique has been employed exten-
sively in recent years to simulate thermo-mechanical properties of APS coatings.
Wang et al. [37] compared the models based on SANS and OOF for predicting
thermal conductivity of plasma-sprayed YSZ coatings. The effect of splat boundar-
ies on thermal conductivity was also discussed, and it was concluded that splat
interfaces account for 25–60 % of the total reduction in conductivity, with larger
values for higher splat interface density coatings [37].
Jadhav et al. [38] used OOF and an analytical model to predict thermal conduc-
tivity of APS and SPPS coatings and compared them with the experimental values
measured by LFA. It was concluded that OOF model fitted better with experimental
values compared to the analytical model and it captured the effect of real micro-
structures on the thermal conductivities more accurately [38]. The model was also
used later to determine residual stresses in SPPS TBCs [39]. The low value of pre-
dicted residual stresses depicted that the TBC microstructures produced by SPPS
were strain tolerant.
Bolot et al. [40] predicted thermal conductivity of AlSi/polyester coatings with
two different numerical codes—TS2C (based on finite differences) and OOF.
Some discrepancies between the values obtained were observed, why the code
was developed further by changing the discretisation method which gave better
results [41].
50 5 Modelling of Properties of TBCs

5.3.2.1 Modelling Procedure

OOF is very efficient in capturing the actual microscopic images and generating the
required finite element meshes based on the colours present in the image [42, 43].
Though it is limited to 2D images, it can be very useful in determining macroscopic
properties of materials consisting of more than one phase.
The model geometry information for analysis is obtained from the microstructure
image which is used as an input. Different phases/features present in the microstruc-
ture image are then grouped based on their colour and/or contrast. Each group of
pixels can be assigned a specific material property. The microstructure images are
usually represented in greyscale. Even though OOF2 offers several tools for creating
pixel groups of a greyscale image, it is simpler to modify the image before using it in
OOF2. For example, if there are two phases present in the image, the image can be
converted into binary format by thresholding to reduce the number of hues, thus
making it more suitable for an automatic and repetitive method as shown in Fig. 5.1.
Rough skeletons are first generated on the image using an adaptive meshing rou-
tine. The skeleton is a grid that adapts to all pixel group boundaries present in the
image. Initially individual elements may contain pixels corresponding to different
individual phases. During the meshing procedure, the elements are refined and the
nodes are moved so that the material interfaces are well defined. Based on the bound-
aries of the defined pixel groups present in the image, the skeleton is refined to
capture the irregular boundaries marking the different phases of the microstructure.
Finer elements are automatically generated near the interfaces to account for higher
gradients.
There are several skeleton modifying options in OOF2, many dependent on the
two types of ‘energies’ of the elements: shape energy and homogeneity energy [44].
The total energy (E) of an element is calculated with the following equation:

E = a h . E _ hom + (1 - a h ) . E _ shape (5.7)

Fig. 5.1 A small portion of a microstructure image in (a) the original greyscale and (b) binary format
5.3 Finite Element Modelling 51

Fig. 5.2 A skeleton modification process in the FE software OOF based on the image in Fig. 5.1b

where E_hom is the definition of the homogeneity energy, E_shape is the definition
of the shape energy and αh is an adjustable parameter between 0 and 1. A high
value of αh is used when the priority is the homogeneity of elements, i.e. it is more
important that the element nodes match the image boundaries rather than keeping a
good shape.
An example of a skeleton modification process is shown in Fig. 5.2. Figure 5.2a
shows an initial skeleton, without any consideration to the image boundaries. This
skeleton is generated in the first step based on the maximum element size defined by
the user. Figure 5.2b shows the skeleton after a number of Anneal iterations have
been performed. The Anneal routine moves nodes randomly and accepts moves
according to a given criterion [32]. In Fig. 5.2c the Refine method has been used,
which splits elements into smaller pieces to make the skeleton correspond better to
the boundaries [32]. In Fig. 5.2d, e several additional skeleton modifying steps,
e.g., Rationalise, have been used to further optimise the skeleton. The Rationalise
option fixes badly shaped elements by modifying them and adjacent elements or by
removing them completely [43].
During the skeleton creation process, OOF2 displays a homogeneity index
showing the average homogeneity level of all elements. This value should increase
during the process, as the skeleton gets finer and more adapted to the boundaries.
A high value of homogeneity index is desired which would imply that the different
pixel groups are well defined by the created mesh. If the index reaches 1, all
elements are 100 % homogeneous.
When the meshing procedure is completed, the elements are assigned the property
of the material with the dominant pixel group present in each element to generate
52 5 Modelling of Properties of TBCs

the final mesh. Once the mesh is created, the equation system and boundary conditions
are defined over the model.
The main quality concern in OOF is capturing the required microstructural details
present in the images. Since the fine details of phases/features present in the image
depend on the quality of image, SEM images are preferred over optical microscope
images. Lower resolution images also result in higher errors due to inefficient discre-
tisation of the image during the thresholding procedure. Dorvaux et al. and Tan et al.
have discussed the uncertainties arising due to the effects of image threshold level,
image location across the sample cross section and image magnification and size on
the final results [7, 27]. These factors must be kept in mind when using OOF for
predicting macroscopic material properties. In general, an image covering a large
area of cross section with high resolution should be used. Several images acquired
across the sample cross section must be used so as to reduce the errors induced due
to the effect of local features on predicted properties.
The quality of mesh generated over the group of pixels representing the different
phases present in the microstructure image also has a major influence on the model-
ling results. It is found that minimising the maximum scale of element size is crucial
in addition to setting a small minimum element size. Figure 5.3 shows the finite
element mesh generated with coarse and fine element sizes. The red and white areas
represent the two phases (pixel groups) present in the image. As shown in Fig. 5.3,
the mesh with fine element is capable of capturing the small microstructural details
which are missed out in the coarse mesh, since the material property for an element
is assigned on the basis of the dominant pixel group present in that element. Thus, a
finer mesh is required for higher accuracy of the predicted results.
An advantage of using OOF2 is that the meshes created in OOF2 can be saved and
exported in different file formats, which enables the use of OOF2 meshes in other
finite element software applications, such as ANSYS Workbench (ANSYS Inc., USA).

Fig. 5.3 Finite element mesh with different maximum and minimum element sizes. It can be
clearly noticed that the mesh in (b) is able to capture much finer details compared to (a)
5.3 Finite Element Modelling 53

This is highly beneficial when more tedious calculations are necessary and/or addi-
tional simulation options are required. Another advantage is that once the meshing
procedure is optimised for one of the microstructure images for a ­sample, the entire
procedure can be automated. This can be done by saving the input instructions in a
script form and changing only the text referring to input image file for other images
for the sample. Thus, errors induced due to the operator can be reduced.

5.3.2.2 Property Evaluation

Once a finite element mesh has been created over the microstructure image, a simple
set-up could be made to evaluate properties such as thermal conductivity and
Young’s modulus using the mesh as shown in Fig. 5.4.
Thermal conductivity (λ) can be evaluated by setting up a temperature difference
(ΔT) across the top and bottom boundaries while keeping the other boundaries insu-
lated to calculate the heat flux (dQy/dt) and then using the steady-state heat equation
in 2D domain given as

dQy / dt = l × Ay ( DT / L y ) (5.8)

where Ay is the cross-sectional area perpendicular to the y-axis and Ly is the thickness
of the model domain along the y-axis as shown in Fig. 5.4.
Similarly Young’s modulus (E) can be evaluated by setting up a prescribed defor-
mation (δx) along the x-direction to calculate the average stress (σx) across the model
and then using the stress–strain equation given as

s x = E (d x / L x ) (5.9)

where Lx is the thickness of the model domain along the x-axis as shown in Fig. 5.4.
Since the area representing the image is only a small fraction of the total
cross-­sectional area of the coating and the microstructural features might vary
significantly from one part of the coating to the other, a number of images need to
be analysed to achieve a statistical significance.

Fig. 5.4 Simple set-up to evaluate thermal conductivity (left) and Young’s modulus (right)
54 5 Modelling of Properties of TBCs

5.4 Artificial Coating Morphology Generator

Tbctool software was developed by ONERA, France, within the Brite Euram HITS
project (Project BE96-3226 HITS, task 1.4). It is an interactive computer program
which was created especially for TBC development. As mentioned earlier, it has
been used in the past to predict thermal conductivity of TBCs based on FDM.
However, the major advantage of Tbctool is that it can be used as a predictive tool
for generating plasma-sprayed TBC-like morphology. The software generates
randomised microstructure images based on several pre-defined criteria. Figure 5.5
shows an artificially generated microstructure image using Tbctool simulating a
plasma-sprayed coating morphology. The generation process is described in detail
in the reference manual for Tbctool [45] and is described only briefly here.
The generation process is divided into four major sequences which are executed
step by step: spheroids or globular pores generation, linked cracks or cracks connected
to globular pores, free cracks and dust or fine scale pores. Each sequence has several
parameters which can be defined using a statistical distribution function. The function
can be defined using analytical distribution laws such as Uniform, Normal, Log
Normal and Custom or tabulated laws with numerical definition. This statistical or
probabilistic definition of parameters creates the randomness in the generation
process. However, it must be kept in mind that these sequences can end up in an infinite
loop, if too strict constraints are imposed on the generation parameters.
In the spheroids generation process, first the amount of spheroidal porosity needs
to be defined. After that, the software chooses a set of particles from the ‘particles
library’ on the basis of other specifications, such as particle area, elongation, orienta-
tion and compactness, so that the total area of the particles chosen becomes larger than

Fig. 5.5 An artificial image created using Tbctool morphology generator simulating a plasma-­
sprayed coating morphology
5.4 Artificial Coating Morphology Generator 55

Fig. 5.6 Different crack families in Tbctool [45]

the amount of defined spheroidal porosity level. The particle library, which consists
of thousands of particles or globular pores, is incorporated within the software.
This particle library was developed by extracting features from the actual microscopic
images of several APS coatings. Then the software assigns random positions to the set
of chosen particles in the field on the basis of other specifications such as overlapping
rejection probability and minimum distance between particles.
The linked cracks sequence is divided into three sub-sequences: main or primary
cracks, secondary cracks and segment cracks. Figure 5.6 shows the different linked
crack families. The primary cracks start from the spheroids and the secondary cracks
start from the primary cracks, whereas the segment cracks may start from spheroids
and/or other cracks. First, the amount of linked cracks porosity and the fractions of
main linked cracks and secondary linked cracks need to be defined. All the cracks
start from ‘seeds’, which are generated according to various specifications such as
seed birth probability and seed fraction. All cracks are built according to several
parameters related to crack shape (e.g. shape pulsation and shape damping) and
crack size (e.g. crack length and crack thickness).
The free cracks generation sequence is very similar to the segment cracks gen-
eration process described above, the difference being in the nature of the seeds used.
The dust generation sequence is the simplest one with the amount of dust porosity
specified being generated on the basis of dust distance specified so that there is no
overlapping of particles.
The main limitation of Tbctool morphology generator is that it uses geometric
methods which are not directly related to the process physics. This means that the
morphology generation parameters are somewhat arbitrarily related to the actual
process parameters. The generation procedure widely uses random sub-processes
which are difficult to optimise all together and might not have a physical significance.
This limitation can also be used as an advantage since new coating m ­ orphologies
can be fabricated virtually and explored for further analysis. However, in this case,
56 5 Modelling of Properties of TBCs

the input parameters need to be linked to process parameters to ensure that the
generated morphology is possible to achieve experimentally as well. Another limita-
tion is the finite size of the particle library, so all types of particles cannot be generated
using Tbctool, especially if one considers new coating materials. Comparisons
between real and artificially generated microstructures and a methodology to design
an optimised microstructure using Tbctool are presented in the next section.

5.5 Recent Work

In the first part of this work, the aim was to establish the fundamental understanding
of relationships between coating microstructure and thermal conductivity of TBCs.
A range of coating architectures was investigated including high purity YSZ,
dysprosia-­stabilised zirconia and dysprosia-stabilised zirconia with porosity former.
The microstructures were examined both on as-sprayed samples and on heat treated
samples. First, statistical modelling was used as a ‘screening’ method to determine
the important microstructure parameters influencing the thermal conductivity. The most
interesting finding in this analysis was that pores and cracks in contact seem to have
significant influence on thermal conductivity compared to free pores and free cracks.
As this model does not consider the actual physical phenomenon, the results from
this analysis were validated by FEM developed with OOF using microstructure
images obtained by SEM which provided physical verification of the statistical
results. The microstructure exhibiting the best performance in this study is shown in
Fig. 5.7 where the pores and cracks in contact are indicated. The results, although

Fig. 5.7 Microstructure cross-section image indicating pores and cracks in contact [1]
5.5 Recent Work 57

tentative in nature, indicate that the combination of the two modelling approaches
seems to be a feasible approach to understand relationships between coating
microstructure and thermal conductivity of TBCs. The effect of pores and cracks in
contact on TBC properties was investigated further in this work.
To design new morphologies, Tbctool was first verified on real coatings. First,
Tbctool was used to generate artificial microstructure images using the input param-
eters determined by image analysis of real microstructure images obtained using
SEM. These artificially generated images were used as an input to the finite element
model developed using OOF to predict thermal conductivity. The artificially gener-
ated images were verified by two methods. First, by comparing their image analysis
data with SEM images, and second, by comparing the predicted thermal conductivity
values with the predicted thermal conductivity values from SEM images and the
thermal conductivity values obtained from laser flash experiments. It was observed
that the finite element model ranked the coating systems considered the same way
in terms of thermal conductivity as the experimental values. Also, same ranking was
observed for the artificial images generated. These tentative results indicate that the
images generated by the coating morphology generator Tbctool were similar to
the real coatings. Figure 5.8 shows microstructure cross-section images of one such
TBC topcoat analysed in this work taken by SEM and artificially generated by
Tbctool. It must be noted that these images might not be lookalike but they are simi-
lar from a statistical point of view based on the results obtained above. These results
indicate that the combination of OOF and Tbctool can be used as a powerful
approach to design coatings. Hence, this approach was used in further work as
described below to develop a low thermal conductivity and long lifetime TBC.
Once the finite element model developed using OOF and the artificial coating
morphology generator Tbctool were verified, a combined empirical and numerical
approach was utilised to develop a novel thermal barrier coating. The intention was to
design a coating system which could be implemented in industry within the next three
years. Different morphologies of ceramic topcoat were evaluated: using dual layer
systems and polymers to generate porosity. Dysprosia-stabilised zirconia was also
included in this study as a topcoat material along with the state-of-the-art YSZ using

Fig. 5.8 Microstructure cross-section images of a TBC topcoat (a) taken by SEM and (b) artifi-
cially generated by Tbctool [2]
58 5 Modelling of Properties of TBCs

Fig. 5.9 Microstructure image of the porosity former coating indicating the large globular pores
with connected cracks [3]

high purity powders. In the experiments it was shown that a dysprosia-doped high
purity zirconia coating with large globular pores with connected cracks created due to the
inclusion of polymers exhibited the best performance, i.e. lowest thermal conductivity
and longest fatigue lifetime. The coating microstructure is shown in Fig. 5.9 where the
large globular pores with connected cracks are indicated. The finite element model
was shown to be capable of catching the trend in different coating lifetime through
predicted Young’s modulus values thus making it possible to optimise a coating
design using the model.
To achieve further understanding of why this coating exhibited best performance,
Tbctool was utilised to design three particular kinds of microstructure images contain-
ing only pores, only free cracks with pores and connected cracks with pores as shown
in Fig. 5.10. The porosity level in all these three images was chosen to be the same,
even though they comprised different microstructural features; the porosity level
being similar to the images analysed earlier for porosity former coatings. It was
observed that pores and connected cracks result in lowest Young’s modulus and
thermal conductivity, which also suggests an explanation for the high performance of
porosity former coatings. It also shows that low Young’s modulus and thermal con-
ductivity cannot be achieved by only producing large pores, since that image resulted
in a much higher values. The work implies that the combined empirical and numerical
approach is an effective tool for developing low thermal conductivity coatings with
enhanced lifetime. It was shown that a low conductivity coating with twice the
lifetime compared to the industrial reference today could be produced.
5.5 Recent Work 59

Fig. 5.10 Microstructure images generated using Tbctool representing (a) only pores, (b) free
cracks with pores and (c) connected cracks with pores [3]
60 5 Modelling of Properties of TBCs

To further verify these results as well as the modelling approach, a design of

experiments (DoE) was conducted by varying selected spray parameters, and the finite
element model was used to predict thermal conductivity and Young’s modulus. The
aim was to investigate fundamental relationships between microstructure and thermal–
mechanical properties and obtain an optimised coating microstructure. The lowest
thermal conductivity and Young’s modulus were shown by a coating exhibiting large
globular pores with connected cracks in the microstructure as shown in Fig. 5.11a.

Fig. 5.11 Microstructure cross-section image indicating large pores with connected cracks
(a) coating showing best performance based on the design of experiments, and (b) optimised coating
sprayed with another spray gun [4]
References 61

For the purpose of verifying the modelling results, this microstructure was sprayed
along with a reference coating with a new spray gun. Thermo-­mechanical properties
were predicted using the same finite element model while thermal conductivity and
thermo-cyclic fatigue lifetime were measured experimentally. The coating microstruc-
ture with large globular pores with connected cracks shown in Fig. 5.11b performed
better compared to the reference coating sprayed with standard settings before optimi-
sation. In this way, the modelling results obtained in the first part of this work were
verified. This work shows that the modelling approach in combination with experi-
ments can be used as a powerful tool to achieve optimised microstructures. This
approach is both time saving and cost effective.

References

1. Tano I, Gupta M, Curry N, Nylén P, Wigren J (2010) Relationships between coating micro-
structure and thermal conductivity in thermal barrier coatings: a modelling approach.
Proceedings of the International Thermal Spray Conference, May 3–5, 2010 (Singapore), DVS
Media, pp 66–70
2. Gupta M, Nylén P (2011) Design of low thermal conductivity thermal barrier coatings by finite
element modelling. In: Sudarshan TS, Beyer E, Berger L-M (eds) Surface Modifications
Technologies XXIV, Sep 7–9, 2010 (Dresden). pp 353–365
3. Gupta M, Nylén P, Wigren J (2013) A modelling approach to microstructure-property relation-
ships in thermal barrier coatings. J Ceram Sci Technol 4(2):85–92
4. Gupta M, Curry N, Markocsan N, Nylén P, Vaßen R (2013) Design of next generation thermal
barrier coatings- experiments and modelling. Surf Coat Technol 220:20–26
5. Pawlowski L, Fauchais P (1992) Thermal transport properties of thermally sprayed coatings.
Int Mater Rev 37(6):271–289
6. Golosnoy IO, Tripas SA, Clyne TW (2005) An analytical model for simulation of heat flow in
plasma-sprayed thermal barrier coatings. J Therm Spray Technol 14(2):205–214
7. Dorvaux JM, Lavigne O, Mevrel R, Poulain M, Renollet Y, Rio C (1998) Modelling the thermal
conductivity of thermal barrier coatings. Proceedings of the 85th Meeting of the AGARD
Structures and Materials Panel, Oct 15–16, 1997 (Aalborg, Denmark), NATO-AGARD-R-823,
pp 13-1–13-10
8. Wagh AS (1993) Porosity dependence of thermal conductivity of ceramics and sedimentary
rocks. J Mater Sci 28(14):3715–3721
9. Cernuschi F, Ahmaniemi S, Vuoristo P, Mäntylä T (2004) Modelling of thermal conductivity
of porous materials: application to thick thermal barrier coatings. J Eur Ceram Soc 24(9):
2657–2667
10. Schlichting KW, Padture NP, Klemens PG (2001) Thermal conductivity of dense and porous
yttria-stabilized zirconia. J Mater Sci 36(12):3003–3010
11. Klemens PG (1991) Thermal conductivity of inhomogeneous media. High Temp High Press
23(3):241–248
12. Hasselman DPH (1978) Effect of cracks on thermal conductivity. J Compos Mater 12:
403–407
13. McPherson R (1984) A model for the thermal conductivity of plasma-sprayed ceramic coatings.
Thin Solid Films 112(1):89–95
14. Li CJ, Ohmori A (2002) Relationships between the microstructure and properties of thermally
sprayed deposits. J Therm Spray Technol 11(3):365–374
15. Boire-Lavigne S, Moreau C, Saint-Jacques RG (1995) The relationship between the micro-
structure and thermal diffusivity of plasma-sprayed tungsten coatings. J Therm Spray Technol
4(3):261–267
62 5 Modelling of Properties of TBCs

16. Bjorneklett A, Haukeland L, Wigren J, Kristiansen H (1994) Effective medium theory and the
thermal conductivity of plasma-sprayed ceramic coatings. J Mater Sci 29(15):4043–4050
17. Sevostianov I, Kachanov M (2000) Anisotropic thermal conductivities of plasma-sprayed thermal
barrier coatings in relation to the microstructure. J Therm Spray Technol 9(4):478–482
18. Sevostianov I, Kachanov M, Ruud J, Lorraine P, Dubois M (2004) Quantitative characteriza-
tion of microstructures of plasma-sprayed coatings and their conductive and elastic properties.
Mater Sci Eng A 386(1–2):164–174
19. Sevostianov I, Kachanov M (2009) Elastic and conductive properties of plasma-sprayed
ceramic coatings in relation to their microstructure: an overview. J Therm Spray Technol
18(5–6):822–834
20. Lu TJ, Levi CG, Wadley HNG, Evans AG (2001) Distributed porosity as a control parameter
for oxide thermal barriers made by physical vapor deposition. J Am Ceram Soc 84(12):
2937–2946
21. Cipitria A, Golosnoy IO, Clyne TW (2009) A sintering model for plasma-sprayed zirconia
TBCs. Part I: free-standing coatings. Acta Mater 57(4):980–992
22. Hollis KJ (1995) Pore phase mapping and finite-element modeling of plasma sprayed tungsten
coatings. In: Berndt CC, Sampath S (ed) Advances in thermal spray science & technology.
Sept 11–15, 1995 (Houston, TX), ASM International, pp 403–409
23. Saint-Ramond B (2001) HITS: high insulation thermal barrier coating systems. Air Space
Europe 3(3–4):174–177
24. Poulain M, Dorvaux JM, Lavigne O, Mévrel R, Renollet Y, Rio C (2002) Computation of
thermal conductivity of porous materials applications to plasma sprayed TBCs. In: Advanced
thermal barrier coatings and titanium aluminides for gas turbines, June 17–19, 2002 (Bonn,
Germany), Turbomat
25. Bartsch M, Schulz U, Dorvaux JM, Lavigne O, Fuller ER Jr, Langer SA (2003) Simulating
thermal response of EB-PVD thermal barrier coating microstructures. Ceram Eng Sci Proc
24(3):549–554
26. Kulkarni A, Wang Z, Nakamura T, Sampath S, Goland A, Herman H, Allen J, Ilavsky J, Long
G, Frahm J, Steinbrech RW (2003) Comprehensive microstructural characterization and pre-
dictive property modeling of plasma-sprayed zirconia coatings. Acta Mater 51(9):2457–2475
27. Tan Y, Longtin JP, Sampath S (2006) Modeling thermal conductivity of thermal spray coatings:
comparing predictions to experiments. J Therm Spray Technol 15(4):545–552
28. Liu FR, Zeng KL, Wang H, Zhao XD, Ren XJ, Yu YG (2010) Numerical investigation on the
heat insulation behavior of thermal spray coating by unit cell model (#1182). Proceedings of
the International Thermal Spray Conference, May 3–5, 2010 (Singapore), DVS Media, 2010,
pp 794–798
29. Qunbo F, Fuchi W, Lu W, Zhuang M (2009) Microstructure-based prediction of properties for
thermal barrier coatings, thermal spray 2009: expanding thermal spray performance to new
markets and applications. Marple BR, Hyland MM, Lau Y-C, Li C-J, Lima RS, Montavon G
(eds), May 4–7, 2009 (Las Vegas, Nevada), ASM International, pp 46–50
30. Li C, Ohmori A, McPherson R (1997) The relationship between microstructure and Young’s
modulus of thermally sprayed ceramic coatings. J Mater Sci 32:997–1004
31. Sevostianov I, Kachanov M (2001) Plasma-sprayed ceramic coatings: anisotropic elastic and
conductive properties in relation to the microstructure; cross-property correlations. Mater Sci
Eng A 297(1–2):235–243
32. Azarmi F, Coyle T, Mostaghimi J (2009) Young’s modulus measurement and study of the
relationship between mechanical properties and microstructure of air plasma sprayed alloy
625. Surf Coat Technol 203:1045–1054
33. Michlik P, Berndt C (2006) Image-based extended finite element modeling of thermal barrier
coatings. Surf Coat Technol 201:2369–2380
34. Amsellem O, Madi K, Borit F, Jeulin D, Guipont V, Jeandin M, Boller E, Pauchet F (2008)
Two-dimensional (2D) and three-dimensional (3D) analyses of plasma-sprayed alumina
microstructures for finite-element simulation of Young’s modulus. J Mater Sci 43:4091–4098
References 63

35. Amsellem O, Borit F, Jeulin D, Guipont V, Jeandin M, Boller E, Pauchet F (2012) Three-­
dimensional simulation of porosity in plasma-sprayed alumina using microtomography and
electrochemical impedance spectrometry for finite element modeling of properties. J Therm
Spray Technol 21(2):193–201
36. http://www.ctcms.nist.gov/oof/oof2/. Accessed 24 Feb 2015
37. Wang Z, Kulkarni A, Deshpande S, Nakamura T, Herman H (2003) Effects of pores and interfaces
on effective properties of plasma sprayed zirconia coatings. Acta Mater 51(18):5319–5334
38. Jadhav AD, Padture NP, Jordan EH, Gell M, Miranzo P, Fuller ER Jr (2006) Low-thermal-­
conductivity plasma-sprayed thermal barrier coatings with engineered microstructures. Acta
Mater 54(12):3343–3349
39. Jadhav AD, Padture NP (2008) Mechanical properties of solution-precursor plasma-sprayed
thermal barrier coatings. Surf Coat Technol 202:4976–4979
40. Bolot R, Seichepine JL, Vucko F, Coddet C, Sporer D, Fiala P, Bartlett B (2008) Thermal
conductivity of AlSi/Polyester abradable coatings. In: Lugscheider E (ed) Thermal spray
crossing borders, June 2–4, 2008 (Maastricht, The Netherlands), ASM International.
pp 1056–1061
41. Bolot R, Seichepine JL, Qiao JH, Coddet C (2011) Predicting the thermal conductivity of AlSi/
polyester abradable coatings: effects of the numerical method. J Therm Spray Technol
20(1–2):39–47
42. Langer SA, Fuller ER Jr, Carter WC (2001) OOF: an image-based finite-element analysis of
material microstructures. Comput Sci Eng 3(3):15–23
43. Reid ACE, Langer SA, Lua RC, Coffman VR, Haan S–I, García RE (2008) Image-based finite
element mesh construction for material microstructures. Comput Mater Sci 43(4):989–999
44. Busso EP, Lin J, Sakurai S, Nakayama M (2001) A mechanistic study of oxidation-induced
degradation in a plasma-sprayed thermal barrier coating system. Part I: model formulation.
Acta Mater 49:1515–1528
45. Tbctool manual, Tbctool documentation
Chapter 6
Modelling of Interface Roughness in TBCs

6.1 Simplified Interface Roughness Modelling

The influence of topcoat–bondcoat interface roughness on thermo-mechanical

stresses is of high relevance for understanding the failure mechanisms as well as
assessing the lifetime of TBCs. Analysing the effect of interface roughness on
stresses could be a tedious and extremely complicated task to perform experimen-
tally. Therefore, numerical modelling techniques are commonly used to understand
these fundamental relationships. In most of the earlier works, a 2D or 3D sinusoidal
wave profile has been chosen as a simplification to represent the topcoat–bondcoat
interface [1–16] to analyse the stress distribution in TBCs.
A crack propagation model was proposed by Bäker [2], Ranjbar-Far et al. [7] and
Bialas [11] to analyse the stress distribution in TBCs by using a 2D wave profile to
represent the topcoat–bondcoat interface. Ranjbar-Far et al. used uniform as well as
non-uniform amplitudes to represent the wave profile and an inhomogeneous top-
coat layer with an artificial lamellar structure. It was concluded that the cracking
depends mainly on the interface morphology and the thickness of TGO layer.
Failure mechanisms were analysed by Evans [17] with the help of a 2D finite
element model representing the interface as 2D sine wave profile. Lifetime prediction
models were made by Vaßen et al. [18], Wei et al. [6], Brodin et al. [14] and He et al.
[16] based on the growth of delamination cracks using a 2D sine wave profile for
modelling the interface. The lifetime model by Vaßen et al. was based on the stress
distribution analysis model developed by Ahrens et al. [3].
Kyaw et al. [5] used a 2D sine wave profile to represent the interface and incorpo-
rated the temperature dependency of material properties and sintering of topcoat in the
finite element model. Qi et al. [8] used a 2D sine wave profile to represent the interface
and attributed the stress distribution to the interface roughness and thermal mismatch.
Asghari et al. [9] used a 2D semi-circular profile to represent the interface with a non-
linear material model for topcoat implementing the effect of sintering. Hermosilla
et al. [10] and Busso et al. [4] developed a couple microstructural–mechanical model

© The Author(s) 2015 65

M. Gupta, Design of Thermal Barrier Coatings, SpringerBriefs in Materials,
DOI 10.1007/978-3-319-17254-5_6
66 6 Modelling of Interface Roughness in TBCs

using a 2D wave profile to represent the interface. The TGO growth was determined
by a diffusion model in these studies.
Bednarz [12] did an extensive work in this field by using a 2D profile with different
shapes such as sinusoidal, semi-circular and elliptical. The major conclusion of this
work was that the most significant parameters affecting the stress development
during thermal cycling are time-dependent material properties of topcoat and TGO,
TGO growth behaviour and interface roughness. Jinnestrand et al. [1] developed a
model using a 3D sine wave profile to represent the interface to analyse the stress
distribution.
Ghafouri-Azar et al. [19] and Klusemann et al. [20] made a 2D model to analyse
the residual stresses in tungsten carbide–cobalt (WC-Co) coatings sprayed by HVOF
method. Micrograph images were used to define the coating as well as substrate
geometry for the finite element model including the coating–substrate interface.
As it could be observed from the above references, the general research trend in the
modelling approach in this field has been to develop the material models for individ-
ual layers and simulation conditions so as to represent the real simulation conditions
in a more precise way [5]. However, a simplified roughness profile might not lead to
precise predictions as it does not incorporate the complex surface topography created
during thermal spraying.

6.2 Real Interface Roughness Modelling

In a recent work by Nordhorn et al., finite element analyses of realistic 3D models

captured using laser microscopy and multiple 2D models based on different inter-
face approximation functions were performed to predict the thermally induced
stress fields [21]. The objective was to simulate the 3D reference stress fields using
2D models so as to reduce the required computational time to assess the stress
fields. It was concluded in this work that the 3D model was too complex with respect
to the geometric interface features to be replaced by a single 2D approximating
function. However, the 2D models could correctly reflect essential TGO growth-
related features such as stress inversion.
In Sects. 6.2.1 and 6.2.2 below, the methodology undertaken by the author to
model the interface roughness in 2D and 3D has been described in detail.

6.2.1 Two-Dimensional Approach

The 2D simulations were prepared using cross-sectional microstructure SEM images

of isothermal heat treated samples with different time span in the furnace. The vary-
ing exposure time resulted in altered TGO growth of the samples. Three TGO thick-
ness stages were considered in this work, as-sprayed, 100- and 200-h heat treatments
6.2 Real Interface Roughness Modelling 67

so that a significant difference can be observed. Five images per specimen were
analysed for this study for statistical significance.
The SEM images used in 2D modeling were captured carefully so as to include
sections of the TGO, topcoat and bondcoat layers. These images were processed
with image editing software Adobe Photoshop CS6 (Adobe Systems Incorporated,
USA) as follows. Additional topcoat and bondcoat layers were added to the image
to achieve desired thicknesses of individual layers as shown in Fig. 6.1. All porosi-
ties and cracks present in the SEM image were eliminated while the TGO interface
profile was kept intact. Each layer was given an individual colour.
OOF2 was used to generate a mesh over the processed microstructure image and
thereby enable a 2D simulation in the finite element software ANSYS. Before mesh
generation, different pixel groups are created in OOF2 on the basis of colours present
in the image, and material properties were assigned to these colour groups. A fine
mesh is generated near the boundaries between the layers, and coarser elements are
created where the homogeneity of elements was easily attained, as shown in Fig. 6.1.
The mesh generated in OOF2 was then exported to ANSYS. The substrate layer was
added subsequently in ANSYS Workbench as schematically shown in Fig. 6.1 to
achieve the desired thicknesses of individual layers. A coarse mesh was generated
in ANSYS over the substrate layer and merged with the imported mesh to generate
the final mesh.
Once a finite element model is created using the procedure described above, it
can be used for determining the stresses generated due to the interface profile by
setting up simulation conditions.

Fig. 6.1 SEM microstructure image of a TBC cross section (leftmost) is used to create the model
shown in the centre. The substrate has been cropped. A part of the model illustrating the generated
finite element mesh is shown to the right with varying element sizes in the TGO and adjacent
boundaries [22]
68 6 Modelling of Interface Roughness in TBCs

6.2.2 Three-Dimensional Approach

The surface coordinate file obtained from 3D surface scanning techniques mentioned
in Sect. 4.4 was used as an input for the 3D modelling approach. The bondcoat
surfaces were scanned with white light interferometry technique (MicroXAM, ADE
Phase Shift Technology, USA) before spraying the topcoat layer. The scanned data
were stored in the form of a text file with a list of coordinates. The coordinate files
were used to generate surfaces representing the curvature of the TGO and the adja-
cent faces of the bondcoat and the topcoat. Different TGO thicknesses were consid-
ered in the 3D simulations similar to the 2D simulations to analyse the effect of
TGO growth on stresses.
The coordinate files representing the bondcoat topography obtained from white
light interferometry were pre-processed in MATLAB R2010b (MathWorks, USA)
to obtain an area of 50 μm × 50 μm to avoid tedious calculations as well as for com-
patibility with the computer-aided design (CAD) software, NX Unigraphics 7.5
(Siemens PLM Software, USA). The surface generated in the CAD software was
used to represent both the bondcoat–TGO and the TGO–topcoat interfaces obtained
by translating the surface in the direction perpendicular to the coating deposition
direction. Other layers in the TBC system were then extruded to create appropriate
thicknesses as shown in Fig. 6.2.
The model created in the CAD software was then imported to ANSYS where the
finite element was generated. The mesh concentration was kept dense near the TGO
and coarse at the areas with lower stresses, as shown in Fig. 6.2. The topcoat and the

Fig. 6.2 Representation of the 3D model with cropped substrate and topcoat to the left and an
enlargement of the TGO interface illustrating the varying mesh density present in the models to
the right [22]
6.3 Results 69

bondcoat were split into two bodies each to control the mesh density. Once a 3D
finite element model is created using the procedure described, it can be used for
determining the stresses generated due to the interface profile by setting up simulation
conditions.

6.3 Results

By implementing the procedures described in Sects. 6.2.1 and 6.2.2, finite element
modelling was used to study the residual stress profile in the topcoat–bondcoat
interface using real surface topographies [22]. An attempt was made in this work to
overcome the limitation of simplified representation of topcoat–bondcoat interface
used conventionally by using real surface topographies so as to represent the inter-
face in a more accurate way. The influence of topcoat–bondcoat interface on the
induced stresses in topcoat which eventually affect lifetime was studied by using
time to stress inversion as an indicator as per the stress inversion theory described in
Sect. 3.4.2. The differences in functional performance between APS and HVOF
bondcoat samples observed in an earlier work were evaluated with both 2D and 3D
simulations [23]. Both 2D and 3D simulations were shown to verify the previously
formulated stress inversion theory established using simplified sinusoidal curves.
It was observed that the stress inversion from compressive to tensile stresses
occurred earlier in the topcoat–bondcoat interface for the HVOF samples which
could be a reason for an earlier failure of these samples in lifetime testing.
The bondcoat surface profiles captured using white light interferometry for the
HVOF and APS samples are shown in Fig. 6.3. It was remarked based on the simu-
lation results that possible unmelted particles present on the HVOF bondcoat sur-
face could increase the overall stresses in the topcoat which could also contribute to
earlier failure in lifetime testing. This could also be differentiated with the newly
formulated ISO 25178 feature parameters shown in Table 6.1 derived from the seg-
mentation motifs of these surface profiles.
The height parameter Sa which calculates the arithmetic mean height in 3D
(corresponding to Ra in 2D) was 11.2 μm for HVOF and 11.1 μm for APS, which
would suggest similar roughness level. However, from the feature parameters given
in Table 6.1 based on segmentation motifs, it can be easily noticed that the surface
topography is significantly different. The higher Spd value suggests that APS
bondcoat has higher density of peaks compared to the HVOF bondcoat, and the
higher Spc value suggests that the arithmetic mean curvature of these peaks is also
higher. This implies that the APS bondcoat has high density of sharp and steep
peaks while HVOF bondcoat has relatively flat and large surface area hills, which
is also confirmed by the high Sha value for HVOF bondcoat. This conclusion from
the quantitative measurements can also be easily visualised in Fig. 6.3.
These results support the argument that better characterisation of the surface
rather than the traditional Ra value is required for a fundamental understanding. It was
concluded that the modelling approach using real topographies can be a tool to
70 6 Modelling of Interface Roughness in TBCs

a
µm
50

45
40

35
30

25
20
0
50 15
10 10
0
15
0 5
20
0
25 50 0
0 5
30 0
0 50 -5
35 50
0 4 -10
40 00
0 4
µm 45 0 -15
0 35
50 0 -20
0 30
55 0
0 25 -25
60 0 µm
0 20 -30
65 0
0 15 -35
70 0
0 10 -40
75
0 50
b
µm

40

35

30

25

20

15
0
50 10
10
0
15 5
0
20 0
0
25 60 0
0
0
30 55 -5
0 500
35 0
0
40 45 -10
0
0
45 40 -15
µm 0 0
50 35 -20
0 300
55 0
0 25 µm -25
60 0
0
65 20 -30
0 0
70 15
-35
0
100
75
0 50
80
0

Fig. 6.3 Bondcoat surface profiles captured using white light interferometry technique for (a) HVOF
and (b) APS samples [22]
References 71

Table 6.1 Some of the ISO 25178 feature parameters calculated for HVOF and APS bondcoat
samples [22]
Parameter (units) Parameter description HVOF APS
Spd (1/mm2) Density of peaks (number of peaks per unit area) 49.5 135
Spc (1/mm) Arithmetic mean peak curvature (arithmetic mean of the 1,140 2,844
principal curvatures of peaks within a definition area)
Sha (mm2) Mean hill area (average area of the hills connected to 0.0194 0.00699
the edge at a particular height)

attain valuable insight into the stress distribution in topcoat–bondcoat interfaces.

The results indicated that optimisation of bondcoat roughness can be realised
through this technique.

References

1. Jinnestrand M, Sjöström S (2001) Investigation by 3D FE simulations of delamination crack

initiation in TBC caused by alumina growth. Surf Coat Technol 135:188–195
2. Bäker M (2012) Finite element simulation of interface cracks in thermal barrier coatings.
Comput Mater Sci 64:79–93
3. Ahrens M, Vaßen R, Stöver D (2002) Stress distributions in plasma-sprayed thermal barrier
coatings as a function of interface roughness and oxide scale thickness. Surf Coat Technol
161:26–35
4. Busso EP, Lin J, Sakurai S, Nakayama M (2001) A mechanistic study of oxidation-induced
degradation in a plasma-sprayed thermal barrier coating system. Part I: model formulation.
Acta Mater 49:1515–1528
5. Kyaw S, Jones A, Hyde T (2013) Predicting failure within TBC system: Finite element simula-
tion of stress within TBC system as affected by sintering of APS TBC, geometry of substrate
and creep of TGO. Eng Fail Anal 27:150–164
6. Wei S, Fu-chi W, Qun-bo F, Zhuang M (2013) Lifetime prediction of plasma-sprayed thermal
barrier coating systems. Surf Coat Technol 217:39–45
7. Ranjbar-Far M, Absi J, Mariaux G (2012) Finite element modeling of the different failure
mechanisms of a plasma sprayed thermal barrier coatings system. J Therm Spray Technol
21(6):1234–1244
8. Qi H-Y, Yang X-G (2012) Computational analysis for understanding the failure mechanism of
APS–TBC. Comput Mater Sci 57:38–42
9. Asghari S, Salimi M (2010) Finite element simulation of thermal barrier coating performance
under thermal cycling. Surf Coat Technol 205:2042–2050
10. Hermosilla U, Karunaratne MSA, Jones IA, Hyde TH, Thomson RC (2009) Modelling the
high temperature behaviour of TBCs using sequentially coupled microstructural–mechanical
FE analyses. Mater Sci Eng A 513–514:302–310
11. Białas M (2008) Finite element analysis of stress distribution in thermal barrier coatings. Surf
Coat Technol 202:6002–6010
12. Bednarz P (2007) Finite element simulation of stress evolution in thermal barrier coating
systems. Ph.D. thesis, Forschungszentrum Jülich GmbH, Germany
13. Martena M, Botto D, Fino P, Sabbadini S, Gola MM, Badini C (2006) Modelling of TBC
system failure: stress distribution as a function of TGO thickness and thermal expansion
mismatch. Eng Fail Anal 13:409–426
72 6 Modelling of Interface Roughness in TBCs

14. Brodin H, Jinnestrand M, Sjöström S (2004) Modelling and experimental verification of

delamination crack growth in an air-plasma-sprayed thermal barrier coating. In: 15th European
Conference of Fracture (ECF15), Stockholm, Sweden
15. Nair BG, Singh JP, Grimsditch M (2004) Stress analysis in thermal barrier coatings subjected
to long-term exposure in simulated turbine conditions. J Mater Sci 39:2043–2051
16. He MY, Hutchinson JW, Evans AG (2003) Simulation of stresses and delamination in a
plasma-sprayed thermal barrier system upon thermal cycling. Mater Sci Eng A 345:172–178
17. Evans HE (2011) Oxidation failure of TBC systems: an assessment of mechanisms. Surf Coat
Technol 206:1512–1521
18. Vaßen R, Giesen S, Stöver D (2009) Lifetime of plasma-sprayed thermal barrier coatings:
comparison of numerical and experimental results. J Therm Spray Technol 18(5–6):835–845
19. Ghafouri-Azar R, Mostaghimi J, Chandra S (2006) Modeling development of residual stresses
in thermal spray coatings. Comput Mater Sci 35:13–26
20. Klusemann B, Denzer R, Svendsen B (2012) Microstructure-based modeling of residual
stresses in WC-12Co-sprayed coatings. J Therm Spray Technol 21(1):96–107
21. Nordhorn C, Mücke R, Vaßen R (2014) Simulation of the effect of realistic surface textures on
thermally induced topcoat stress fields by two-dimensional interface functions. Surf Coat
Technol 258:181–188, http://dx.doi.org/10.1016/j.surfcoat.2014.09.032
22. Gupta M, Skogsberg K, Nylén P (2014) Influence of topcoat-bondcoat interface roughness on
stresses and lifetime in thermal barrier coatings. J Therm Spray Technol 23(1–2):170–181
23. Curry N, Markocsan N, Östergren L, Li X-H, Dorfman M (2013) Evaluation of the lifetime
and thermal conductivity of dysprosia-stabilized thermal barrier coating systems. J Therm
Spray Technol 22(6):864–872
Chapter 7
Modelling of Oxide Growth in TBCs

Visualisation of TGO growth in situ by experimental methods is highly non-trivial.

One of the destructive ways of capturing the TGO growth by experiments is to treat
several samples of the same coating for different operating times until failure and
then inspecting the cross-sectional microstructure. However, in this case the TGO
growth cannot be analysed at the same location in the TBC sample as the area
inspected will differ between the samples and may not contain exactly the same
surface topography. A non-destructive way could be to perform micro-computed
tomography (Micro-CT) on a sample in several stages over the sample’s lifetime.
This technique, apart from being extremely expensive and time consuming, can be
performed usually only on a very small sample with relatively low resolution.
Thermography has been successfully performed in the past to analyse the crack
growth during lifetime testing, but a low resolution is obtained with this technique
[1]. High temperature X-ray diffraction (XRD) was used by Czech et al. to determine
the thickness of TGO in situ under isothermal conditions by measuring the intensity
curves, but the TGO profile could not be detected with this technique [2].
As apparent from the above discussion, modelling of TGO formation would be
highly beneficial since using experimental methods would be highly challenging.
The conventional way of modelling TGO growth (dTGO) as a function of time (t) and
temperature (T) is described a parabolic growth law given as [3]

dTGO = ATGO · t p · exp ( - ETGO / kBT ) (7.1)

where ATGO is TGO growth coefficient, p is TGO growth exponent, ETGO is TGO
growth activation energy and kB is the Boltzmann constant. However, this parabolic
growth law does not include the effect of bondcoat roughness on TGO formation.
TGO growth models based on diffusion could thus provide a better reflection of
the reality.

© The Author(s) 2015 73

M. Gupta, Design of Thermal Barrier Coatings, SpringerBriefs in Materials,
DOI 10.1007/978-3-319-17254-5_7
74 7 Modelling of Oxide Growth in TBCs

7.1 Diffusion-Based Modelling

7.1.1 Previous Work

A few studies have been performed earlier to model the TGO layer formation based
on diffusion of species. A multi-scale continuum mechanics approach based on a
coupled diffusion-constitutive framework was implemented by Busso et al. to study the
local stresses induced in TBCs using a parametric unit cell finite element model [4].
A numerical model describing the TGO growth by an oxygen diffusion–reaction
model was developed by Hille et al. to perform an analysis of a representative TBC
system subjected to a thermal cycling process and to make a parametric study for
different fracture strengths of the topcoat to determine its influence on the durability
of a TBC system [5]. A one-dimensional oxidation–diffusion model considering
both surface oxidation and coating–substrate interdiffusion as well as aluminium
depletion during operating conditions was developed by Yuan et al. to predict the
lifetime of TBCs [6]. A diffusion–reaction of aluminium and oxygen to form TGO
in TBCs was studied through an analytical model by Osorio et al. and the results
were compared with experiments to evaluate the TGO growth rate [7]. However, all
of the models discussed above did not incorporate the inherent TBC microstructure
or the topcoat–bondcoat interface topography.
Apart from achieving fundamental understanding of the TGO growth phenomena,
the development of TGO in situ during thermal cyclic loading conditions needs to
be implemented in a diffusion-based model to evaluate stresses developed due to
TGO formation. The following general assumptions based on the references above
could be made for developing a simplified TGO growth model:
1. TGO growth occurs only during the dwell phase of the lifetime testing.
2. The bondcoat mainly consists of aluminium and the TGO formed consists of
pure alumina for a NiCoCrAlY bondcoat. Formation of other oxides can be
neglected.
3. The reaction rate of aluminium and oxygen to form alumina is very high compared
to the diffusion rates of the constituents.
4. The diffusion rate of oxygen within the topcoat can be considered to be very high
since zirconia is transparent to oxygen flow as discussed in Sect. 3.5.

7.1.2 Real Interface Roughness Modelling

A 2D diffusion-based TGO growth model developed by the author using a computa-

tional fluid dynamics (CFD) approach consisting of real topcoat–bondcoat interface
topography extracted from cross-sectional micrographs based on the modelling
assumptions stated above is described here [8].
The cross-sectional microstructure images of TBC samples at failure were used to
define the initial geometry for the model. The microstructure images were imported
7.1 Diffusion-Based Modelling 75

in a CAD software and a spline profile was carefully created over the image to
capture the TGO profile. The upper edge of the TGO layer was defined as the initial
as-sprayed stage in the diffusion model since only the inward growth of TGO occur-
ring due to the diffusion of oxygen within the TGO towards the TGO–bondcoat
interface was considered in the model for simplicity, and the lower edge was defined
as the TGO thickness profile at the failure stage. As several cracks form within the
TGO during testing and thus appear in the images due to the intense cracking lead-
ing to failure, the profiles were created where the TGO profile would have been
before failure occurred based on prior experience. One of the imported microstruc-
ture images taken from a sample is shown in Fig. 7.1 along with the spline profiles.
Bondcoat and topcoat layers were added to the spline profile to create a 2D planar
model by extruding the structure shown with solid lines in Fig. 7.1. The porosity
network and oxide inclusions within the topcoat and bondcoat layers were omitted
from the simulation. Several such models need to be analysed for each coating for
statistical significance. In this way the model could capture the real interface profile
extracted from the microstructure images.
The TGO growth model was made in ANSYS Fluent 14.5 (ANSYS, Inc.,
Canonsburg, PA, USA) using a CFD approach. The generic convection–diffusion
equation for additional scalars was used as the modelling concept, where the con-
vective transport was omitted from the system.
A finite element mesh was created over the model. Fine elements were created
near the topcoat–bondcoat interface to capture the TGO growth accurately while
coarser elements were created near the outer edges. Two solid regions were defined in
the system, one being the quiescent gas region containing oxygen and representing
the topcoat and the other being the solid aluminium region representing the bondcoat.

Fig. 7.1 TGO profile drawn over the microstructure image taken from one of the samples; the
solid lines indicate the 2D model used in the TGO growth model [8]
76 7 Modelling of Oxide Growth in TBCs

The system consisted of the two initial solid scalars, oxygen and aluminium,
transported by diffusion in the solid regions and a third solid scalar, alumina, formed
during the simulation. As the oxygen and aluminium scalars congregate by diffu-
sional transport, the alumina layer was produced.
An image showing the formed alumina concentration was saved after each time
step. The TGO profiles at 50 % concentration of alumina were saved as a list of coor-
dinates after several stages. These stages were selected so as to capture a significant
growth of the TGO from one stage to the next stage. However, the number of time
steps between these stages was kept the same for all the analysed models. The concen-
tration of alumina at which these profiles were saved did not affect the final results.
It must be noted here that since the TGO growth rate depends mainly on the diffusion
coefficients, the approximate TGO thickness in different models at different stages
would be in the same range; however, the TGO profile would be different which
would depend on the initial profile of the topcoat–bondcoat interface. The TGO
profiles extracted from the model in the form of coordinate files were later exported to
NX for generating the TGO profile so as to capture the TGO growth. The timescale
in the simulation was not considered to be relevant as the objective was to extract
the TGO growth profile at different stages until failure.
Figure 7.2 shows three images selected from the ones saved after each time step in
the TGO growth model for one of the profiles from the sample shown in Fig. 7.1;
the scale bar illustrates the alumina concentration in fraction, for example, 50 % = 0.5.
Figure 7.2a shows the initial profile at the beginning of the simulation where no
alumina can be observed. Figure 7.2b shows the intermediate stage where the maxi-
mum alumina concentration can be observed to be around 50 % and very thin layer of
TGO could be observed while Fig. 7.2c shows the mature stage of the TGO growth
where a fairly thick TGO could be observed, considering the 50 % concentration level
as indicated in the figure. The inward growth of TGO towards the bondcoat near the
topcoat–bondcoat interface can be clearly observed in the images.

7.1.3 Recent Work

A 2D diffusion-based TGO growth model was developed using the modelling

approach described in Sect. 7.1.2 by the author in a recent work [8]. The objective
of this work was to make a tentative evaluation of the capability of the combination
of the TGO growth and stress analysis models to assess the lifetime of TBCs in a
comparative way.
The TGO growth model was first validated by comparing the TGO profiles created
by the model with thermally cycled specimens with varying interface roughness.
A qualitative analysis was performed by plotting the profiles created by the model for
50 % alumina concentration at different stages over the microstructure image at
failure and comparing them with the initial and final stages at failure.
Figure 7.3 shows the profiles generated by the model plotted over the microstructure
image at failure for the coating sample shown in Fig. 7.1. The initial as-sprayed and
7.1 Diffusion-Based Modelling 77

Fig. 7.2 Images saved from a 1.00e+00

the TGO growth model 9.50e+01
9.00e-01 Initial as-sprayed profile
showing the (a) initial, (b) 8.50e-01
intermediate and (c) mature 8.00e-01
stages of TGO growth [8] 7.50e-01 topcoat
7.00e-01
6.50e-01
6.00e-01
5.50e-01
5.00e-01
4.50e-01
4.00e-01
3.50e-01
3.00e-01 bondcoat
2.50e-01
2.00e-01
1.50e-01 430µm
1.00e-01
5.00e-02
0.00e-00

b 1.00e+00
9.50e+01
9.00e-01
TGO layer
8.50e-01
8.00e-01
7.50e-01 topcoat
7.00e-01
6.50e-01
6.00e-01
5.50e-01
5.00e-01
4.50e-01
4.00e-01
3.50e-01
3.00e-01 bondcoat
2.50e-01
2.00e-01
1.50e-01 430µm
1.00e-01
5.00e-02
0.00e-00

c 1.00e+00
9.50e+01
TGO layer
9.00e-01
8.50e-01
8.00e-01
7.50e-01 topcoat
7.00e-01
6.50e-01
6.00e-01
5.50e-01
5.00e-01
4.50e-01
4.00e-01
3.50e-01
3.00e-01 bondcoat
2.50e-01
2.00e-01
1.50e-01
430µm
1.00e-01
5.00e-02
0.00e-00
78 7 Modelling of Oxide Growth in TBCs

Fig. 7.3 TGO profiles generated from the TGO growth model indicated by solid lines plotted over
the microstructure image at failure shown in Fig. 7.1 [8]

failure stage profile drawn over the image in CAD software is indicated by dashed
lines while the different TGO growth stages extracted from the TGO growth model
are indicated by solid lines. It can be observed that the generated TGO profiles closely
follow the TGO profile in the microstructure image, and the thickness of the generated
TGO seems to be similar to the actual TGO in the microstructure image.
Thereafter, the TGO profiles at different stages of TGO growth were extracted
from the TGO growth model to the finite element model described in Sect. 6.2 for
stress analysis and the time to stress inversion was evaluated. Three experimental
specimens consisting of the same chemistry but with different topcoat–bondcoat
interface roughness were studied by the models and the time to stress inversion was
compared with the lifetimes measured experimentally. It was observed that the
stress analysis model provided the same ranking for lifetime according to the time
to stress inversion to the three samples as provided by the experiments.
Thus, the combination of the two models presented an effective approach to
assess the stress behaviour and lifetime of TBCs in a comparative way. The TGO
growth model consisting of real interface topography can be used as an effective
tool to visualise the TGO growth in situ which could be otherwise a non-trivial task
if performed experimentally.

7.2 Mixed Oxide Growth

As discussed in Sect. 3.5, the growth of mixed oxides in TBCs during operating
conditions plays a major role in the formation of thermo-mechanical stresses. An
attempt was thus made by the author in a recent work to understand the driving force
7.2 Mixed Oxide Growth 79

for cracking of CSN clusters and their role in TBC spallation in a better way [9].
The purpose of the study was to identify during which step of formation the CSN
cluster cracks and the driving mechanisms for cracking.
First, nanoindentation was performed on a CSN cluster in a TBC sample to map
Young’s modulus of the various oxides included in CSN clusters. The values from
nanoindentation experiments were used to produce a Young’s modulus map by assign-
ing appropriate Young’s modulus to each zone as shown in Fig. 7.4 after making some
simplifications, such as removing porosity and very small material zones.
A finite element analysis was then performed using the established Young’s
moduli and the exact geometry of the CSN cluster derived from micrographs using

Fig. 7.4 (a) Light-optic image of an oxide cluster and (b) the corresponding Young’s modulus
map, created from the values based on nanoindentation experiments after the material had been
identified in each part of the image through microscopy [9]
80 7 Modelling of Oxide Growth in TBCs

the model described in Sect. 6.2.1. Several stages of the CSN cluster formation
process, as well as the volumetric increase associated with the oxidation of the last
Ni-rich core, were modelled.
It was concluded from the modelling results that crack formation in the oxide
clusters can occur due to the induced tensile stresses during cooling in the NiO core.
On cooling, the high CTE of NiO induces tensile stresses in the NiO core and crack
formation occurs in the cluster centre. The stresses introduced due to the oxidation
of the Ni core were observed to relax fast enough at high temperature. Silica parti-
cles in the cluster could possibly assist cracking due to the CTE mismatch but is not
the key factor in oxide cluster cracking.

References

1. Schweda M, Beck T, Singheiser L (2012) Thermal cycling damage evolution of a thermal barrier
coating and the influence of substrate creep, interface roughness and pre-oxidation. Int J Mater
Res 103:40–49
2. Czech N, Juez-Lorenzo M, Kolarik V, Stamma W (1998) Influence of the surface roughness
on the oxide scale formation on MCrAlY coatings studied in situ by high temperature X-ray
diffraction. Surf Coat Technol 108–109:36–42
3. Vaßen R, Giesen S, Stöver D (2009) Lifetime of plasma-sprayed thermal barrier coatings:
comparison of numerical and experimental results. J Therm Spray Technol 18(5–6):835–845
4. Busso EP, Lin J, Sakurai S, Nakayama M (2001) A mechanistic study of oxidation-induced
degradation in a plasma-sprayed thermal barrier coating system. Part I: model formulation.
Acta Mater 49:1515–1528
5. Hille TS, Turteltaub S, Suiker ASJ (2011) Oxide growth and damage evolution in thermal barrier
coatings. Eng Fract Mech 78:2139–2152
6. Yuan K, Eriksson R, Peng RL, Li X-H, Johansson S, Wang Y-D (2013) Modeling of microstruc-
tural evolution and lifetime prediction of MCrAlY coatings on nickel based superalloys during
high temperature oxidation. Surf Coat Technol 232:204–215
7. Osorio JD, Giraldo J, Hernández JC, Toro A, Hernández-Ortiz JP (2014) Diffusion–reaction of
aluminum and oxygen in thermally grown Al2O3 oxide layers. Heat Mass Transf 50:483–492
8. Gupta M, Eriksson R, Sand U, Nylén P (2014) A diffusion-based oxide layer growth model
using real interface roughness in thermal barrier coatings for lifetime assessment. Surf Coat
Technol. doi: 10.1016/j.surfcoat.2014.12.043
9. Eriksson R, Gupta M, Broitman E, Jonnalagadda P, Nylén P, Peng RL. Stress and cracking during
Chromia-Spinel-NiO cluster formation in thermal barrier coating systems. J Therm Spray
Technol. Submitted
Chapter 8
Conclusions: How to Design TBCs?

A two-step approach can be undertaken to design an optimised TBC by modelling:

1. Design microstructure defects in topcoat
The microstructure defects in the topcoat can be designed to reduce thermal
conductivity and enhance strain tolerance of the topcoat by optimising the shape
and size of these defects.
An automated procedure was described in this study using finite element modelling
to predict thermal conductivity and Young’s modulus of TBC topcoats. By comparing
the predicted values with the experimental results, it was shown that finite element
modelling approach utilising real microstructure images can be used as a powerful tool
to predict and derive microstructure–property relationships.
It was shown that coating morphologies with similar microstructural features as
the real microstructure images taken using SEM can be artificially generated by
using an artificial coating morphology generator.
The presence of large globular pores with connected cracks in the coating micro-
structure was shown to significantly enhance the performance of TBCs. Low thermal
conductivity, low Young’s modulus and high lifetime were exhibited by these
coatings. This relationship was derived by designing the microstructure using the
combination of finite element modelling and artificial coating morphology generator
and then verified experimentally.
2. Design topcoat–bondcoat interface
The topcoat–bondcoat interface in TBCs can be designed to increase the time
to stress inversion, thus enhancing lifetime due to slower crack propagation by
optimising the interface topography and oxide formation.
The stress inversion theory established using simplified sinusoidal curves was
verified in this study by both 2D and 3D simulations using real topographies. It was
shown that the modelling approach using real topographies can be a useful tool to
attain insight into the stress distribution in topcoat–bondcoat interfaces. It was shown
that the time to stress inversion can be used as an indicator to assess coating lifetime.

© The Author(s) 2015 81

M. Gupta, Design of Thermal Barrier Coatings, SpringerBriefs in Materials,
DOI 10.1007/978-3-319-17254-5_8
82 8 Conclusions: How to Design TBCs?

The results indicated that an optimal bondcoat topography could be designed through
this technique.
The diffusion-based TGO growth model consisting of real interface topography
developed in this work can be used as an effective tool to visualise the TGO growth
in situ which could be otherwise a non-trivial task if performed experimentally.
The combination of the TGO growth and stress analysis models developed in this
work could be used as an effective approach to assess the stress behaviour and life-
time of TBCs in a comparative way. Based on the fundamental knowledge gained
from these models about the relationships between topcoat–bondcoat interface
roughness, TGO growth and lifetime of TBCs, the models need to be developed
further to realise an optimised interface design.
The finite element model based on analysing real microstructure images was
successfully implemented to achieve fundamental understanding of the phenomena
of stress development and cracking during mixed oxide cluster formation in TBCs.
It was concluded that the major reason for cracking was the large CTE between NiO
and other oxides formed during thermal cycling.
Thus, it can be concluded from the results that the modelling approach described
in this study could be used as a powerful tool to design new coatings as well as to
achieve optimised microstructures and topographies which could significantly
enhance the performance of TBCs.

8.1 Future Work

Further work in this field could be to develop the TGO growth model from 2D to 3D
as it would provide a global view of TGO growth as well as enable to perform stress
analysis in 3D. The inclusion of aluminium depletion and mixed oxides in the TGO
growth model in the future would be highly valuable.
In recent years, coatings produced by SPS/SPPS are becoming of major interest.
The analysis of failure mechanisms as well as the application of the models developed
in this work to SPS/SPPS coatings could be another possibility in the future.