AGARD-R-823

ADVISORY GROUP FOR AEROSPACE RESEARCH & DEVELOPMENT

7 RUE ANCELLE, 92200 NEUILLY-SUR-SEINE, FRANCE

AGARD REPORT 823

Thermal Barrier Coatings

(les Revetements anti-mur de chaleur)

U:.j> ^ST^K^NT A

Papers presented at the 85th Meeting of the AGARD Structures and Materials Panel, held in
Aalborg, Denmark 15-16 October 1997.

19980520 102
North Atlantic Treaty Organization
Organisation du Tratte de l'Atlantique Nord

I3TIC QUALITY INSPECTED 8

 The Mission of AGARD*

According to its Charter, the mission of AGARD is to bring together the leading personalities of the NATO nations in the
fields of science and technology relating to aerospace for the following purposes:

— Recommending effective ways for the member nations to use their research and development capabilities for the
common benefit of the NATO community;

— Providing scientific and technical advice and assistance to the Military Committee in the field of aerospace research
and development (with particular regard to its military application);

— Continuously stimulating advances in the aerospace sciences relevant to strengthening the common defence posture;

— Improving the co-operation among member nations in aerospace research and development;

— Exchange of scientific and technical information;

— Providing assistance to member nations for the purpose of increasing their scientific and technical potential;

— Rendering scientific and technical assistance, as requested, to other NATO bodies and to member nations in
connection with research and development problems in the aerospace field.

The highest authority within AGARD is the National Delegates Board consisting of officially appointed senior
representatives from each member nation. The mission of AGARD is carried out through the Panels which are composed of
experts appointed by the National Delegates, the Consultant and Exchange Programme and the Aerospace Applications
Studies Programme. The results of AGARD work are reported to the member nations and the NATO Authorities through the
AGARD series of publications of which this is one.

Participation in AGARD activities is by invitation only and is normally limited to citizens of the NATO nations.

* AGARD merged with the Defence Research Group of NATO (DRG) on 1 January 1998 to form the Research and Technology Organization (RTO)
of NATO. However, both AGARD and DRG will continue to issue publications under their own names in respect of work performed in 1997.

The content of this publication has been reproduced

directly from material supplied by AGARD or the authors.

®
Printed on recycled paper

Published April 1998

Copyright © AGARD 1998

All Rights Reserved

ISBN 92-836-1073-3

Printed by Canada Communication Group Inc.

(A St. Joseph Corporation Company)
45 Sacre-Cceur Blvd., Hull (Quebec), Canada K1A 0S7
 Thermal Barrier Coatings
(AGARD-R-823)

Executive Summary
Thermal barrier coatings are an emerging technology which will allow either increasing the inlet
turbine temperatures or on the other hand decreasing the working temperature of the metal of the blades
and consequently increasing their life-time.

The Workshop allowed a survey of the state of the art, a description of the existing technologies or of
the technologies under development, a review of the present knowledge of damage mechanisms,
including microstructural, mechanical and thermal aspects, and an account of the advantages and
drawbacks of the various families as perceived by engine manufacturers and users. A final discussion
was held to identify the needs for further R&D.
 Les revetements anti-mur de chaleur
(AGARD-R-823)

Synthese
Les barrieres thermiques sont une technologie emergente qui permettra soit d'augmenter les
temperatures d'entree des turbines, soit de reduire la temperature ä laquelle est soumis le metal des
aubes et par consequent d'en accroitre la duree de vie.

L'atelier a eu pour objet de presenter l'etat de l'art, de donner une description des technologies
existantes ou en cours de developpement, de donner une revue des connaissances actuelles sur les
mecanismes d'endommagement includant les aspects microstructuraux, mecaniques et thermiques, de
rendre compte des avantages et des inconvenients des diverses families tels que les percoivent les
fabricants et les utilisateurs de moteurs.
Une discussion finale a permis d'identifier les besoins futurs en R & D.
 Contents

Page

Executive Summary 111

Synthese IV

Preface Vll

Preface Vlll

Structures and Materials Panel IX

Reference

SESSION I: INTRODUCTION

Thermal Barrier Coatings: the Thermal Conductivity Challenge

by S. Alperine, M. Derrien, Y. Jaslier and R. Mevrel

SESSION II: TBC TECHNIQUES

Paper 2 Withdrawn

Recent Developments in Plasma Sprayed Thermal Barrier Coatings

by P. Fauchais, A. Vardelle and M. Vardelle

In-flight Particle Diagnostics for On-line Process Control during Deposition of

Plasma-Sprayed TBCs
by C. Moreau, P. Gougeon, M. Prystay and L. Leblanc

On Thick Thermal Barriers for Combustor Application

by T. Haubold, H. Gans, D. Schwingel and R. Taylor

Advanced Processing of TBCs for Reduced Thermal Conductivity

by J.R. Nicholls, K.J. Lawson, D.S. Rickerby and P. Morrell

TBCs on Free-Standing Multilayer Components

by P.G Tsantrizos, G.E. Kim and T.A. Brzezinski

Electron Beam Physical Vapour Deposition Thermal Barrier Coatings:

A Comparative Evaluation of Competing Deposition Technologies
by Y. Jaslier and S. Alperine

Processing, Characterisation and Testing of EB-PVD Thermal Barrier Coatings

by W.A. Kaysser, M. Peters, K. Fritscher and U. Schulz
Electron Beam, PVD TBCs: Coating Industrial Blades at an Industrial Scale lOf
by G. Marijnissen

Plasma Assisted CVD of Thick Yttria Partially Stabilized Zirconia Coatings 11

by S. Chevillard, S. Drawin and M.H. Vidal-Setif

SESSION III: TBC PROPERTIES AND DESIGN

Thermal Sprayed Nanostructured Thermal Barrier Coatings 12

by L.T. Kabacoff

Modelling the Thermal Conductivity of Thermal Barrier Coatings 13

by J.-M. Dorvaux, O. Lavigne, R. Mevrel, M. Poulain, Y. Renollet, C. Rio

Microstructure and Thermal Conductivity of Layered Thermal Barrier Coatings Processed 14

by Plasma Spray and Physical Vapor Deposition Techniques
by K.S. Ravichandran, K. An, R.E. Dutton and S.L. Semiatin

Evaluation of Thermal Barrier Coatings from Burner Rig Tests 15

by J-P. Immarigeon, V.R. Parameswaran, D. Chow, D.D. Morphy, P. Gougeon, M. Prystay
and C. Moreau

Graded Design of EB-PVD Thermal Barrier Coating Systems 16

by U. Schulz, T. Krell, U. Leushake and M. Peters

Bond Coat Considerations for Thermal Barrier Coatings 17

by A.M. Freborg, B.L. Ferguson, W.J. Brindley and G.J. Petras

The Effect of TBC Utilization in the Design of Robust Aircraft Combustors 18

by C.A. Arana

Paper 19 Withdrawn

Advantages/Disadvantages of Various TBC Systems as Perceived by the Engine 20

Manufacturer
by P. Morrell and D.S. Rickerby

Report on the Final Discussion D

by S. Drawin

jPaper not available at time of printing

 Preface

Thermal barrier coatings are a technology used to protect the metal of combustion chambers, turbines blades or
veins. Since these parts are air-cooled low-thickness elements, it is possible to reduce the actual temperature of
the metal introducing a large temperature gradient though the ceramic coating. This technology has been used for
several decades in the combustion chambers, and more recently on modern highly-cooled blades.

Since it can allow an increase of the temperature difference between the metal part and the gas flow, up to 100°C,
it can be considered as one of the bottle-neck technologies giving access to higher temperatures in aeronautical
engines. Alternatively, it can allow also, for the same turbine inlet temperature, a reduction of the working
temperature of the metal of the blades, leading to Longer lifetimes of the parts, or a reduction of the cooling air
flow, with an increase in engine efficiency.

Up to now zirconia-base ceramics in particular yttried-zirconia have been used; they are compatible with metal
temperatures up to 900°C for combustion chambers and 1050°C for blades, under fairly low thicknesses.

Two deposition processes are presently in use: plasma spray, which gives relatively thick, low cost, low
conductivity coatings, well adapted to combustion chambers, and for blades Physical Vapour Deposition
techniques which give columnar microstructure which lead to more acceptable damage and to a surface of limited
rugosity, which does not perturb the aerodynamic qualities of the blades but with unfortunately lead to lower
deposition rates, and also higher conductivities.

Research is in progress in several directions: improved or new deposition techniques, new composition of the
ceramics.

The Workshop in this context has been dedicated to:

1. The presentation of families and technologies under development with reference to new ideas and limits.

2. The description of the advantages and drawbacks of the various families of thermal coatings as perceived by
engine manufacturers or users.

3. Understanding damage mechanisms in connexion with properties and microstructure.

4. Modeling the behaviour and the lifetime of the coatings with special reference to the underlying bond coat.

P. COSTA
Chairman of the Subcommittee
 Preface

Les barrieres thermiques sont une technologie qui est utilisee pour proteger le metal des chambres de combustion,
des aubes de turbine mobiles et fixes. Comme ces pieces sont des elements de faible epaisseur fortement
refroidies par un courant d'air arriere, il est possible de reduire la temperature effective du m6tal du fait d'une
importante chute de temperature au travers du revetement ceramique. Cette technologie est utilisee depuis
plusieurs decennies pour les chambres de combustion, et depuis plus recemment pour des aubes mobiles
fortement refroidies.

Dans la mesure oü eile permet d'accroitre la difference entre la piece metallique et le flux gazeux de pres de
100°C, eile peut etre considered comme une technologie goulet pour qui veut accroitre les temperatures dans les
moteurs aeronautiques. En parallele, eile permet aussi, pour une meme temperature d'entr^e de turbine, de require
la temperature du metal des pieces travaillantes, et d'en accroitre ainsi la dur6e de vie, ou encore de require les
flux de refroidissement, avec un accroissement du rendement des moteurs.

Jusqu'ici, ce sont des ceramiques ä base de zircone qui ont ete utilisees, en particulier la zircone-yttriee. Elles sont
compatibles avec des temperatures atteignant 900°C pour le metal des chambres de combustion, de 1050°C pour
les aubes mobiles, et ceci pour des epaisseurs de revetement relativement faibles.

Deux precedes de depot sont actuellement utilises, la pulverisation plasma qui permet d'obtenir des revetements
relativement epais, de coüt modere et de faible conductibilite thermique, bien adaptes au probleme des chambres
de combustion, et pour les aubes des depots en phase vapeur, qui donnent des structures colonnaires qui sont
faiblement endommageables, et un 6tat de surface de bonne qualite ne perturbant pas l'aerodynamique de l'aube,
avec toutefois des vitesses de depot, done des coüts plus eleves, et egalement une conductibilite plus eleväe. Les
recherches se poursuivent dans plusieurs directions: methodes de depots ameliorees ou nouvelles, nouvelles
compositions de la c6ramique.

Dans ce contexte, 1'atelier a ete consacre ä:

1. La description des families et des technologies existantes.

2. La presentation des families et des technologies en cours de developpement, avec une reference particuliere
aux idees nouvelles et ä leurs limites.

3. La description des avantages et des inconv6nients des diverses families de barrieres thermiques tels que les
percoivent fabricants de moteurs et utilisateurs.

4. La comprehension des mecanismes d'endommagement en correlation avec les proprietes des barrieres et leur
microstructure.

5. La modelisation du comportement et de la duree de vie des revetements, avec une reference particuliere pour
le revetement interm6diaire.

P. COSTA
President du Sous-Comite
 Structures and Materials Panel

Chairman: Prof. R. Potter Deputy Chairman: Ir. H.H. Ottens

Business Manager Head of Structures Department
Structural Materials Center National Aerospace Laboratory
Griffith Building (A7) Room 1014 (NLR)
DERA Farnborough P.O. Box 153
United Kingdom 8300 AD Emmerloord
Netherlands

SUB-COMMITTEE MEMBERS

Chairman: Dr. Paul Costa

Directeur Scientifique des
Materiaux
ONERA
29, Av. de la Division Leclerc
92322 Chätillon/Bagne ux Cedex
France

Members: T. Akyurek - TU A. Lasalmonie - FR

E. Campo - IT C. Moura Branco - PO
H. Goncalo - PO R. Potter - UK
P. Heuler - GE T. Ronald - US
W. Van der Hoeven - NE R. Servent - SP
J.P. Immarigeon - CA E. Starke - US
R. Kochendörfer - GE D. Viechnicki - US
L. Kompotiatis - GR J. Waldman - us

PANEL EXECUTIVE

Dr. J.M. CARBALLAL, SP

Mail from Europe: Mail from US and Canada:

AGARD-OTAN AGARD-NATO/SMP
92200 Neuilly-sur-Seine PSC 116
France APO AE 09777

Tel: 33 (0) 1 55 61 22 90 & 92

Telefax: 33 (0) 1 55 61 22 99 & 98
Telex: 610175F
 1-1

Thermal Barrier Coatings : the Thermal Conductivity challenge

S. Alperine1, M. Derrien2, Y. Jaslier1, R. Mevrel3

^NECMA Materials and Process Department, Villaroche, France
2
SNECMA Component Design Department Villaroche France
3
ONERA Materials Science Division BP72 92322 Chatffion Cedex France

ABSTRACT. In this paper, the importance of the the well-known t' phase [5] and its toughening
challenge associated with the control of the thermal mechanisms [6]. Important materials studies have also
conductivity of thermal barrier coatings for turbine been devoted to finding an optimal metallic bond
engines hot stages is being reviewed (yttria stabilised coating to support the ceramic topcoat [7,8]. The
zirconia mostly). It is firstly illustrated by the development of TBC has also been closely linked to
description of a practical aeronautic coated and that of the corresponding deposition processes: plasma
uncoated turbine blade design exercise. The various spraying and electron beam physical vapour deposition
contributions to TBC thermal conductivity are then (EBPVD). A number of studies [4, 9, 10] have thus
reviewed. Their low conductivity finds its source not been devoted to optimise the deposition process
only in the nature of the ceramic layer (highly parameters to obtain favourable structures in order to
disordered material), but also in the morphology of achieve longer life cycles. The general idea there, was
the insulating layer, closely linked to its fabrication to try to relate coatings microstructure to their
process parameters. The influence of various factors thermomechanical resistance.
(such as yttria content, temperature, porosity content
and distribution, etc.) on the thermal conductivity is This necessity to address in the very first place the
examined, largely based on a literature review. In this feasibility and reliability problems for such coatings
field of investigation, the modelling tool should allow has somewhat hindered, until recently, the extent of
to predict, to a certain extent, which morphology research activities directly dealing with their thermal
would lead to the lowest values of thermal conductivity (K). The implicit notion there, was that
conductivity. Eventually attempts are made to identify by using stabilised zirconia as TBC would lead to a
research domains where further understanding is value of K for the coating not significantly different
needed, and to formulate several suggestions from one coating to the other, lower in any case than
concerning possible ways to lower the thermal that of the dense material having the same
conductivity. composition. The scope of this paper is to emphasise
the importance of the challenge associated with the
1. INTRODUCTION control of coating thermal conductivity. It is first
Yttria stabilised zirconia based thermal barrier illustrated through a practical aeronautic turbine blade
coatings are viewed today as the new generation of design. The various contributions to thermal barrier
protective coatings that will allow to take the next coating thermal conductivity are then reviewed. The
quantum step forward to increase turbine engine low conductivity of thermal barriers comes on the one
efficiency, without having to pay the price of hand from the nature of the yttria stabilised zirconia
replacing nickel-base superalloys by some other more ceramic layer, a highly disordered material indeed, and
refractory material [1]. They are also considered by on the other, from the morphology of this layer, closely
diesel engine manufacturers to increase the efficiency linked to coating processes: pores, cracks, etc. The
of their engines. These coatings have been studied influence of various factors (such as yttria content,
and developed for more than 20 years in the western temperature, porosity distribution, etc.) on the thermal
countries and even longer in former USSR. conductivity is examined, largely based upon a
literature review. In this field of investigation, the
It can be said that one of the first concern of materials modelling tool should allow to predict, to a certain
scientists in the TBC field was to optimise the extent, which type of morphology would lead to the
coatings from a materials science standpoint. In the lowest values of conductivity. Eventually, attempts are
early days, important studies have been carried out to made to identify research domains where further
determine for instance which was the most adequate understanding is needed and to formulate several
stabilising agent of zirconia for a given application, suggestions concerning possible ways to lower the
and in what amount [1-3]. As far as very high thermal conductivity.
temperatures applications are concerned (mostly
driven by the aero industry), there was a rapid and 2. A REAL CASE DESIGN STUDY
general move towards 6-8 wt. % yttria partially In this case we consider a high pressure turbine vane
stabilised zirconia (Y-PSZ) because these coatings airfoil of an aero-engine. The component is meshed for
showed the highest degree of resistance to spallation 2D finite element calculation ; for one airfoil, about
in thermal fatigue solicitations [3] and an excellent 350 elements and 1200 nodes are being generated. The
thermal stability. Later, explanations of the thermal barrier coating is represented as a single
phenomenon appeared through the identification of element in the foil thickness.

Paper presented at an AGARD SMP Meeting on "Thermal Barrier Coatings",

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
1-2

^1300 N>7BC
A mission profile is defined for the component, as the
variation with time of relevant engine parameters such Ü1200 I.JVVMII.IV

as altitude and thrust. Aerothermal Navier-Stokes = 1100 1017 __ 1.1 W/m.K

codes are used to calculate thermal exchanges in the
internal cooling passages ; thermal exchanges with hot
I 1000
a 966 ^•^
gases are also computed, taking into account film E 900
4)
cooling (cold gas from the cooling circuit reemitted in H
flow path). The result of these computations is an array 800
of boundary conditions at the external and internal 200 ym 1 mm
walls of the airfoil : hot and cold gas temperatures Figure 2 : Temperature profile across the vane wall
within the boundary layer, values of the wall exchange (leading edge).
coefficients. The next step of the calculation consists
in solving the Fourier equation across the airfoil walls The temperature distribution across the airfoil leading
using the above given limit conditions. The calculation edge is represented on figure 2. At the point considered,
ends up with a mesh of computed temperatures for the metal temperature decreases from 1017°C for an
each node, for every point of the mission profile : T = uncoated airfoil down to 966°C with a TBC having a
f(node #, t). thermal conductivity of 1.9 W/m.K and 933°C with a
low thermal conductivity coating. It is interesting to
note that the ceramic coating external surface
temperature increases correlatively (from 1190°C to
1260°C), this being explained in part by the fact that
the heat flux is kept constant. Excessive temperatures
may provoke an evolution of the ceramic
microstructure : transformation t' —> c +
transformable t, sintering of the external layer
associated with an increase of the thermal conductivity.

In terms of component life duration (taking into

account creep and fatigue phenomena), and considering
the leading edge, the application of a TBC with K = 1.1
■No TBC W/m.K permits to gain more than an order of
-200 M"i TBC; 1.9 W/m.K magnitude on the number of cycles. For less stressed
-200 |im TBC; 1.1 W/m.K regions (pressure side, trailing edge), the benefit is of
course less.
Figure 1 : Metal skin temperature distribution
experienced by a vane during typical flight mission
(leading edge profile).
Pressure
In order to visualise the effect of thermal barrier
side
conductivity, such calculations have been made for an
airfoil without thermal barrier and with a coating of
thickness 200um. In the latter case two values of the Succion
thermal conductivity of the ceramic have been taken side
into consideration : 1.1 and 1.9 W/m.K, typical of
existing plasma sprayed and EBPVD coatings Leading
respectively. edge

Examination of the leading edge metal temperature 100 1000 10000 100000
profile (figure 1) shows that the application of a TBC
(1.9 W/m.K) reduces the maximum temperature Component life in cycles ■ 1.1 W/m.K
(1090°C) of "the metal by 50°C. Decreasing the thermal H1.9W/m.K
conductivity down to 1.1 W/m.K permits to gain a D No TBC
further 32°C. Considering the whole leading edge
profile, the maximum gain even reaches 113°C. Figure 3 : Component life duration (creep andfatigue)
If we consider now that the engine is run so that the
components keep the same life duration, but the gain is
concentrated entirely on reducing the cooling rate and
by way of consequence decreasing the fuel
 1-3

consumption, table 1 shows that up to 0.7% specific cooling, as a result of the limited elastoplastic
consumption can be saved in the cases envisaged. deformation capability of the ceramic. The other is
mainly oriented parallel to the surface and corresponds
Thermal conductivity Specific to poor contacts between different splats. The latter,
(W/m.K) Cooling rate consumption being perpendicular to the heat flux are particularly
1.9 -15% - 0.4% effective in reducing the thermal conductivity of the
1.1 - 25% - 0.7% coating. In addition, a dispersed, spheroidal porosity
contributes to lowering the conductivity.
Table 1 : Reduction in cooling rate and specific
consumption owing to the application of a TBC of In the case of EBPVD coatings, the reduction in
thermal conductivity K on a vane airfoil (for a given thermal conductivity comes mainly from the
lifetime). intracolumnar fine porosity, the intercolumnar porosity
being much less effective as it is arranged
These few results clearly demonstrate the importance perpendicular to the surface. Schematically, the
of lowering the thermal conductivity in terms of columnar coating can be considered, in an electrical
economy and performance. Moreover, a decrease in analogy, more as a set of parallel resistances than a
the thermal conductivity permits to lower the series set which gives a closer image of the
thickness of the coating. Given the fact that a TBC (microcracked) sprayed coating.
represents an added stress on a rotating blade of 10%
(for a 125 urn thick ceramic layer), lowering its
thickness (provided the density is kept constant) will System Thermal conductivity Comment
result in a significant' gain on the stress applied on the Reference
rotating parts, blades and disks. 300K 1300 K
Zr02-7%Y2O3 1.7 2 [4]
EBPVD
Of course, a rigorous computation would have to take Zr02-7%Y2O3 0.9 [12]
into account the radiation contribution to the APS
conductivity within the coating. This is an effect Zr02-7-9%Y2O3 0.9 to 1.1 [13]
which depends on the actual environment seen by the APS Various
airfoils within the turbine stage [11] and difficult to porosity levels
assess at present as the optical properties of zirconia- Zr02-20%Y2O3 0.4 0.55 As sprayed,
based plasma-sprayed and EBPVD coatings are APS 7% porosity
poorly known. ri4i
Zr02-20%Y2O3 1.5 1.7 lOOh, 1480°C,
APS 7% porosity
3. THERMAL CONDUCTIVITY OF ZIRCONIA- ri4i
Zr02-6%Y2O3 1.1 [15]
BASED COATINGS APS
Although the thermal conductivity of thermal barrier Zr02-8%Y2O3 1. [15]
coatings has not been a major research and APS
development topic compared to work on Zr02-12%Y2O3 0.6 [15]
thermomechanical behaviour for instance, a number APS
of characterisations have been published, in particular Zr02-7%Y2O3 1.3 1.3 [16]
to provide designers with data for lifetime prediction EBPVD
models. Looking through some of the data available Zr02-7%Y2O3 0.8 0.8 [16]
and reported in table 2, several points are worth APS
noticing: Zr02-7%Y2O3 0.8 0.8 [16]
APS
- Thermal conductivity values for zirconia based Zr02-8%Y2O3 1.5 1.3 [17]
EBPVD as
coatings are relatively low (within the range 0.5 to 2 coated
W/m.K). This is significantly lower than values Zr02-8%Y2O3 1.9 1.5 [17]
determined on bulk zirconia [18, 19], typically in the EBPVD
2 to 4 W/m.K range, depending on the composition Zr02-7%Y2O, 2.9 . 2.7 [16]
and crystallographic structure. These values lie in the bulk
lower range of thermal conductivity for oxides.
Table 2 : Typical values of the thermal conductivity of
- Thermal conductivities of EBPVD coatings are zirconia-based coatings (EBPVD : electron beam
significantly higher than values measured for plasma physical vapour deposition, APS: air plasma spray)
sprayed coatings having the same composition. Composition percentages in wt%.
This can be explained by the structure of the coatings.
Plasma sprayed coatings exhibit a microcrack
network, with basically two crack populations : one
oriented perpendicular to the coating surface and
originating from cracking of individual splats during
1-4

interest to correctly evaluate this effect with coatings,

System Zr02-Y20, ZrOz-MgO especially at very high temperature.
Phase K, W/m.K K, W/m.K
Cubic 2.3 1.8 - A heat treatment affect the thermal conductivity of
Tetragonal 3.5 4.8 the coatings (increase). After a heat treatment, the
Monoclinic 4.2 5.2 thermal conductivity of TBCs increases, both for
EBPVD [14] and plasma sprayed coatings [14, 20].
Table 3 : Thermal conductivity for zirconia This effect is attributed to the evolution, through a
polymorphs calculated from experimental data on sintering effect, of the porosity (cracks included)
partially or fully stabilised zirconia bulk materials during heat treatment. For example, in plasma sprayed
[18]. The different phases correspond to various coatings, the poor contacts between lamellas in the as-
stabiliser contents. sprayed coatings, with possible nanocrystalline
(amorphous) zones corresponding to rapidly solidified
- Thermal conductivity of EBPVD coatings remains regions, are likely to evolve towards equilibrium
approximately constant as a function of temperature structures, assisted by atom diffusion processes at high
up to about 1100°C, as illustrated in figure 4. This is temperature. One consequence is that thermal
in contradiction with a T1 dependence predicted by conductivity should be determined after a heat
the theory (cf. below) for ceramic materials. At higher treatment conducting to a structure corresponding to
temperature, though, an evolution of the structure the one that will be in service if representative values
involving sintering or phase transformation may are to be used, in particular for the coated blade
result in an increase of the thermal conductivity. The design. Another consequence, is that as surface
same trend is observed for plasma sprayed coatings, temperatures are likely to increase (see preceding part)
in general with a transition at lower temperature. as a result of lowering the thermal conductivity, this
effect has to be taken into account if the benefit of a
low K thermal conductivity is to be retained in service.
2.00
180
> - Values reported cover a surprisingly wide range and
« 1.60 * *********
1 1.40 ■
it is important to keep in mind that they may be
5 * 120 • . . ■ difficult to compare, even for coatings having the same
o E 1.00 nominal composition and porosity, as experimental set
^§0.80 ups can be different, as well as interpretation
E 0.60 procedures from diffusivity determinations. Scattering
jj 0.40 inherent to the fact that thermal conductivity is
I- 0.20 indirectly measured (via thermal diffusivity, density,
0.00 specific heat) and that diffusivity itself is a somewhat
(3 200 400 600 800 1000 12
difficult property to determine. Even with high thermal
Temperature, °C conductivity material such as tungsten, data reported
in the literature are spectacularly scattered, as
■ Asc leposited 0
After 2h * After 5h
1100°C 1100°C
illustrated by Taylor [21]. Moreover, thermal
conductivity values most often are derived from
thermal diffusivity measurements (laser flash in
Figure 4 : Thermal conductivity, as a function of general) and this implies that the porosity is accurately
temperature, of an EBPVD Zr02-7%Y203 thermal known. It would be interesting for several laboratories
barrier coating in as-deposited and heat treated to gather their efforts and, on identical specimens
conditions [17]. conduct a Round Robin test on TBCs, and compare
with absolute, but more lengthy and costly methods.
Given the fact that zirconia is translucent in the
infrared, a radiation contribution to the thermal
conductivity measured (which is actually an effective 4. LOWERING THERMAL CONDUCTIVITY
thermal conductivity) would result in a rise at high
temperature. Such an effect has been reported by 4.1. Intrinsic conductivity and chemistry:
Youngblood et al. [19] in the case of single crystal It has been noticed in the preceding part that zirconia
ZrO2-20wt.%Y2O3 (the thermal diffusivity increases is already a low thermal conductivity material. One
rapidly as a function of T above 600°C). However, in can wonder whether other materials would not present
the case of polycrystals, the same authors observe an a lower thermal conductivity and could be envisaged as
opposite trend and it is likely that in the case of TBC. In fact, experimental thermal conductivity of
coatings (for which the dependence in temperature is pure zirconia, i.e. with no stabiliser, does not seem to
much less important than the T3 predicted by theory), have been reported. This is most certainly due to the
as well as in the polycrystals mentioned, the difficulty of obtaining bulky specimens of pure
numerous interfaces present (gas/solid interfaces, zirconia, zirconia presenting three allotropic forms :
grain boundaries) scatter the penetrating photons and monoclinic up to 1170°C, tetragonal in the range
thereby lower this radiation contribution to the heat 1170°C- 2370°C and cubic up to the melting point
flow. Yet, a thorough study would certainly be of
 1-5

(2680°C), and the tetragonal to monoclinic actually easier starting from an expression proposed by
transformation being accompanied by a large volume Slack [25] for non-metallic crystals and for
variation. The only available data seem to have been temperatures higher than the Debye temperature 6D
derived by Hasselman et al. [18] from experimental and derived form the same principles as the one
determination on various bulk zirconia-based system, followed by [22]:
knowing the relative proportion of the different
crystallographic forms and assuming that the thermal K = B.<M>.8.eD3/T.N2/3.y2
conductivity of a mixture can be expressed simply as a
function of the individual thermal conductivity. As where B is a constant, <M> is the average atomic mass
reported in table 3, the monoclinic form would in the crystal, 83 is the average volume occupied by an
present a conductivity in the range 4 to 5 W/m.K. atom and y is the Griineisen's constant (y = oc/Cv.%
with a the thermal expansion coefficient, % the
Another approach, based on purely theoretical
compressibility and Cv the specific heat). Bearing in
considerations, has been proposed by Klemens [22]
for calculating an «intrinsic thermal conductivity » of mind that 6D is inversely proportional to <M>1/2, it is
cubic zirconia, supposed to contain no stabiliser (and clear that low conductivity crystals correspond to :
therefore no associated vacancies). In a most general
way, the thermal conductivity K can be expressed as: - high atomic mass and weak interatomic bonding
(low 9D, which is a dominant term),
K=(l/3)JC'(a>).v.l((D)dffl - complex crystal structure (large N),
- high anharmonicity (large y).
where C'(ra) dm is the contribution of the specific heat
in the frequency range oo, oo+dro, v is the transverse According to these considerations, other ceramic
materials could then be envisaged as a basis for
phonon velocity and l(m) is the phonon mean free thermal barrier coatings, keeping in mind that other
path.
criterias have to be fulfilled (thermal expansion
coefficient, thermal stability, etc.). Among the
In the high temperature limit and in the case of no parameters which can have a large influence on K,
defect present, the mean free path of phonons is <M> is certainly a privileged one, and to a lesser
limited by three phonon interactions, owing to the
extent N. The Griineisen's constant (y) does not vary
cubic anharmonicities of the lattice forces, and this
expression leads to the following expression of the much from one material to another. An interesting
intrinsic thermal conductivity [23] : candidate, for example, is Ce02 (the atomic mass of
Ce02 is 40% higher than that of zirconia) a material
which has been tested recently by Schulz et al. [26]
K = (3/2)y2 . (n VVODD)!'-1 . N"2'3 and Maloney et al.[27]. However, if EBPVD Ce02-
9wt.%Y203 coatings present a low thermal
where y is the Grüneisen constant, |x is the shear conductivity (around half that of standard EBPVD
modulus, v is the transverse sound velocity, coD the Zr02-8wt.%Y203 coating [27]), they seem to exhibit a
Debye frequency, T the temperature and N the poor erosion resistance, a shortcoming which might be
number of atoms per primitive unit cell. Recently, alleviated though, if a multilayer architecture is
Klemens [22] applied such an approach to zirconia adopted, with a more erosion resistant external layer
and derived for the intrinsic thermal conductivity of added on top of the coating.
zirconia (i.e. that of an hypothetical «pure cubic
zirconia »): As far as compositions for the basis are concerned,
other ways are explored. Observing that only the
K(W/m.K) = 1700/T(K) thermal conductivity at high temperature is relevant,
and that for most ceramics, the thermal conductivity is
According to this expression, the intrinsic thermal a decreasing function of temperature (in principle T"1
conductivity of zirconia: law above), Padture and Klemens [28] are considering
- would amount to 5.7 W/m.K at room temperature, a ceramic materials with complex lattice cells. An
value surprisingly close to the values derived by [18] additional idea in this case is to try to develop an
for pure monoclinic zirconia, oxygen-impermeable dense ceramic layer to avoid or to
- decreases with a classical T1 variation, classical for minimise the oxidation of the bondcoat with formation
non-metallic materials [24], down to 1.3 W/m.K at of a thermally grown alumina at the ceramic/bondcoat
1300K. This last value appears excessively low, interface (locus of the final failure in the most
considering that dense PSZ polycrystals have a advanced systems). It is far from certain however, that
thermal conductivity higher than 2 W/m.K [18] and such a ceramic material can be deposited as a reliable
this represents a strict lower limit of pure zirconia. and dense coating and stay there in service,
(thermomechanical cycles). Moreover, the garnet
Even if absolute values cannot be derived from such compositions these authors propose still have a
simplified theoretical approaches, it is possible to relatively high thermal conductivity, even at 1000°C
infer trends concerning the effect of different (typically above 2.5 W/m.K).
parameters on the thermal conductivity. This is
1-6

In conclusion, possibilities exist to change the average atomic mass, and relative difference in ionic
chemistry of the basis material to decrease thermal size between solute and solvent. Diverse expressions
conductivity. A balance has, of course, to be carefully have been proposed to describe their influence on the
evaluated between the expected gains and the thermal conductivity [see for example 31, 32] and a
consequences on other properties (density, thermal critical study would be necessary to clarify the case of
stability, thermomechanical resistance, erosion zirconia-based materials.
resistance, etc.) to evaluate the final interest of these
potential new candidates. In the case of Zr02-Y203, the replacement of
zirconium ions (M = 91.22, r = 0.080 nm) by yttrium
4.2. Defects : ions (M = 88.90, r = 0.093 nm) should have a minor
An efficient way to decrease the thermal conductivity effect, their atomic mass in particular being relatively
of an insulator is to introduce structural defects which close. However, a significant effect may be expected
constitute obstacles to the propagation of phonons. from other substituents, in particular heavy rare earth
Several types of defect may be envisaged : vacancies, elements (at the expense though of the density, an
substitution ions, dislocations, interfaces (grain important drawback for mobile components). Very
boundaries), etc. few elements lighter than zirconium would satisfy
these criteria, with the exception though of (expensive)
The role of vacancies is particularly important in this scandium. As far as divalent ions are concerned, the
respect for zirconia-based materials. In fact, as pure choice is much more limited. Calcium or magnesium
zirconia cannot be of practical use, due to its cannot be envisaged in high temperature applications
undesirable transformation at around 1100°C, it is due to structural stability problems.
stabilised (partially or completely) by adding
heteroelements (most often Y, Mg, Ca) which Grain boundaries and interfaces can also scatter
stabilise the cubic (or the tetragonal phase). Thus the phonons and reduce the thermal conductivity. These
addition of 20wt% Y203 fully stabilises the cubic processes are most effective at low temperature for
form. Addition of 6 to 8wt. % Y203 permits to obtain ceramics in general [33], and only slightly at high
the so-called metastable t' phase (nontransformable temperature. It is to be noted that in TBCs, the grain
up to around 1200°C). The incorporation of size is much larger (typically 0.1 to 1 urn) than the
heteroelements is accompanied by the introduction of mean free path calculated according to :
vacancies to maintain the electrical neutrality of the
ionic lattice. Thus, introducing two yttrium ions is K~(l/3)C.v.l~0.5nm
accompanied by the introduction of one vacancy
according to the following reaction, using the and it is doubtful that, in these conditions,
nomenclature by Kroger and Vink : conventional grain boundaries may have a significant
effect.
Y203 ->2Yir'+Vo°°+3 0ox
As a conclusion, it can be said that the most efficient
where Y^' is an yttrium ion on a zirconium site ways to decrease the thermal conductivity are to
(single negative charge), V0°° is an oxygen vacancy introduce point defects :
twice positively charged and 00x is an oxygen ion on - vacancies associated with stabilising dopants. In
an oxygen site (no charge). zirconia-based coatings, up to 12 mol.% ROi 5 for a
trivalent dopant R, corresponding to the stabilised
Effectively, incorporating heteroelements such as Ca, cubic form, seem possible. However a fully stabilised
Mg or Y in zirconia lowers its thermal conductivity plasma sprayed material may not present a satisfactory
and a comparison of diffusivity measurements on thermomechanical behaviour.
zirconias with different stabilisers indicates that - for a maximum effect, the dopant should have a very
similar defect populations result in comparable different atomic mass than zirconium, provided that
thermal diffusivity [19]. It is to be noted though that elastic factor does not counterbalance the mass one.
this effect is observed only for stabiliser contents less
than a critical value [29]. For high vacancy contents, Finally, one could wonder what the minimal thermal
they cannot be considered as independent defects and conductivity would be with completely disordered
ordering may appear (as demonstrated by [30] in the materials, in other words what the minimal thermal
case of cubic zirconia stabilised with yttria and conductivity for a refractory material could be. Two
magnesia). It is therefore likely that phonons interact attempts have been proposed to calculate this minimal
less strongly with a population of ordered vacancies thermal conductivity.
than with a population of randomly distributed
defects. Slack [25] derives the theoretical minimum
conductivity assuming that the minimum mean free
Another source of efficient point defects acting as path for a phonon cannot be less than one phonon
phonon scatter centres is substitution ions. This acts wavelength. For pure zirconia, the value calculated is
essentially through two factors : AM/<M>, the ratio of 2.1 W/m.K. In a more recent approach, based on an
the mass difference between solute and solvent to the adaptation of the Einstein model for highly disordered
 1-7

materials, Cahill et al. [34], calculate a minimum example) while keeping the computations tractable on
thermal conductivity of 1.3 for dense zirconia-yttria microcomputers.
systems. These values show that a gain may be
obtained by increasing the atomic disorder, but this This type of approach should offer guidelines for
gain will remain limited if the base material (in designers and coaters, provided the relationship
occurrence zirconia) remains the same. between process parameters and coating morphology is
known.
4.3. Optimising coating architecture :
a. Pores, crack network. b. Thin films.
The fact that the thermal conductivity of EBPVD, and It has been reported by several authors [38-42] that
in particular plasma sprayed coatings is much lower ceramic thin films may exhibit a lower thermal
than for bulk materials of the same composition is conductivity than the corresponding bulk materials.
explained by the presence of pores and microcracks Some examples are reported in table 4.
originating from the coating deposition process, as
described above. It is important to remark that up to Thermal Thermal
now, only few studies have been devoted to optimise Material conductivity conductivity of Ref.
the porosity distribution and crack network with of thin films bulk material
respect to thermal conductivity. In fact, most of the (W/m.K) (W/m.K)
PECVD Si02 0.05 (100 nm) [38]
effort in this field has been devoted instead to the 0.4 (1.4 urn)
influence of these features on the thermomechanical Glassy Si02 0.12 1.3 (amorphous) [391
resistance of the systems, with mainly empirical Ti02 1.5-5 7 [40]
approaches. (depending on
deposition
method)
As a first step, it is necessary to understand and Ti02 0.6 (evaporated) 5.5 - 7.6 (rutile) [41]
describe the relationship between thermal 0.5 (sputtered)
conductivity and morphological features (pores, Zr02(evaporated, 0.8 2-5 [42]
annealed In,
cracks,..). McPherson [35] was probably the first to 730°C)
propose a model for the microstructure of plasma A1N 16 70-220 [42]
sprayed coatings involving regions of good and poor (polycrystalline)
contact between lamellae to explain the low thermal
conductivity of coatings with respect to bulk Table 4 : Examples of thermal conductivity of thin
materials. The regions of poor contact act as thermal films compared with bulk values.
resistances. Considering an electrical analogy, and
assuming a very schematic geometry for the coating, Although this effect is clear, the values reported have
the ratio of thermal conductivity of the coating (Kc) to to be taken with care as measurements of thermal
that of solid material (K0) can be expressed as : conductivity of submicronic films are very delicate and
the error margins reported may be important.
Kc/Ko = 2.f.5/7r.a Different explanations have been proposed for this
effect. Most likely, this effect is due to a combination
where f is the fraction of «true contact», 8 is the of grain-boundary scattering (the grains are
lamellae thickness and a is the radius of the particularly small) and of atomic disorder during
individual contact areas. With a value of 0.2 for f and growth of thin films. Non equilibrium structures may
assimilating a with 5/2, this relation gives Kc/K0 be obtained by PVD or PECVD, in particular for
approximately 1/4, a right order of magnitude. It is to relatively low substrate temperatures (the diffusion of
be noted however that the predictive capacity of such species on the surface of the growing coating, and bulk
an approach is extremely limited due to the simplicity diffusion within the coating being too slow to reach an
of the geometry assumed and the experimental equilibrium state). In this respect, it is to be noted that
difficulty to determine the parameters involved. This the diffusion of cations within zirconia is quite slow.
approach has been employed by Moreau et al. [36] to With the diffusion data (D = D0 . exp-Q/RT with D0 =
describe the thermal diffusivity of plasma-sprayed 0.031 cm2/s and Q = 391 kJ/mol) published by [43],
tungsten based coatings, the geometrical quantities one can estimate that it takes about 8 years for a cation
being determined from image analysis. to travel 1 urn at 1100°C by solid state diffusion .

As described in Dorvaux et al. [37], the most To fully exploit this thin film effect, one could imagine
promising approach for developing a tool to describe thermal barrier coatings constituted of multiple
the thermal conductivity in relation with nanolayers of different materials, such as zirconia and
microstructure seems to be via a finite difference alumina as in [44], with the added advantage of
calculation carried out on digitised images of real incorporating boundary thermal resistances. The
coatings. This type of approach presents the materials of each layer and the process parameters
advantage of taking into account the complexity of the should have to be selected so that prolonged exposure
geometry involved (interconnected porosity for at service temperature would not result in structural
transformations towards equilibrium structures
1-8

(crystallisation of amorphous material for example), ■ designing new coating architectures (nanolayers).
activated by solid state diffusion and accompanied by
an undesirable increase in thermal conductivity. In It is felt that, up to now, several points have received
this respect, it is interesting to note that, as reported little attention and would be worth exploring in more
by Lee et al. [40], annealing a Hf02 thin film at details :
900°C enhances its thermal conductivity by only 20%. ■ characterising the optical properties of the real
It is to be noted also that the fabrication of such coatings, in order to evaluate the possible influence
coatings would certainly require serious modifications of radiation to the heat transfer through the
of techniques employed at present in production, or ceramic coating, and develop ways to decrease this
adapting new techniques to this application contribution.
(sputtering with multiple targets for instance).
■ assessing the thermal stability of the new systems;
with TBCs having low thermal conductivity, the
4.4. Concluding remarks
surface temperatures of the ceramic coating are
Several possibilities exist, in principle to lower the
likely to rise, compared to the temperatures reached
thermal conductivity of TBCs : incorporation of point
with present systems, and provoke a
defects (vacancies, substitution cations), optimising
microstructural evolution of the outmost region of
the microcrack network and, to a lesser extent,
this ceramic layer, through sintering, or phase
porosity distribution, and introducing structural
transformations, which in turn will increase the
disorder via the deposition process. It must be kept in
thermal conductivity.
mind though that, as exposed in the introduction,
TBCs must fulfil a set of requirements, and thermal
Finally, it is important to insist on the fact that
conductivity is only one of them. In particular, it
designing TBCs for engine airfoils requires a
would be highly questionable to develop TBCs that
multidisciplinary approach taking into account various
could not be deposited with processes such as plasma
aspects : materials (ceramic behaviour, oxidation of
spraying or EBPVD (or alternative processes
bondcoat, thermal conductivity, etc.), processes
currently under development). And the design of low
(ceramic morphology), life prediction modelling, blade
thermal conductivity coatings must take into account
design, quality control, commercial (patents),..
the requirements associated with the processes. For
Thermal conductivity is only one aspect of the design
instance, depositing materials having complex
and changing the ceramic composition to lower the
compositions, in particular with elements having
intrinsic conductivity, for example, is likely to have
widely differing vapour pressures, may prove to be too
important repercussions on the other properties. The
difficult to control in production with EBPVD
design of the final coating may have to fully exploit
processes ; similarly, a stack of numerous very thin
the flexibility and capabilities offered by present
layers presenting a highly disordered structure may be
coating processes, by developing multilayer coatings,
possible to fabricate by a PVD (or a CVD) process,
each layer (corresponding to a temperature range and a
but not by plasma spraying. Additional work is
particular environment in service) having a specific
needed however to study the thermal (meta)stability of function.
these defective structures.
Acknowledgements : Particularly fruitful discussions
5. CONCLUSION
with Odile Lavigne, during the preparation of this
Lowering the thermal conductivity of the ceramic
paper, are gratefully acknowledged.
coating of a TBC applied on airfoils in the hot stages
of a turbine represents an important challenge for
References
designing improved performance systems. This paper
has modestly indicated and listed a few paths to tackle
1. Miller R.A., «Thermal barrier coatings - History
this topic, some of which have already been explored :
and directions», in «Thermal Barrier Coating
■ changing the chemistry of the ceramic material in Workshop, NASA CP3312, 1995, pp 17-34.
order to incorporate point defects (vacancies, 2. Stecura S., « Optimisation of NiCrAl/Zr02-Y203
substitution cations) by playing both on the base Thermal barrier system, NASA TM 86905, 1985.
element (zirconia at the .moment) and on the 3. Bose S., DeMasi-Marcin J., «Thermal barrier
stabiliser, the main difficulty being to obtain coating experience in the gas turbine engine at
sufficiently stable phase structures and staying Pratt & Whitney», in NASA CP 3312, 1995, pp
within composition ranges such that the disorder 63-78.
is maximum as far as phonon scattering is 4. Meier S.M., Nissley D.M., Sheffler K.D., Cruse
concerned. T.A., « TBC life prediction model development -
■ optimising the coating morphological features phase II» - PWA Final report; NASA CR 189111,
(porosity, crack network) so that the heat transfer 1991.
perpendicular to the external surface is minimised. 5. Miller R.A., Garlick R.G., Smialek J.L., « Phase
This requires the development of modelling tools stability in plasma-sprayed zirconia-yttria», Am.
such as in the approach described in [36] and a Ceram. Soc. Bull., 62, 1983, pp 1355-1358.
good knowledge of the relationship between 6. Lelait L. (1991) : Etude microstructurale fine de
processes parameters and coating structures. revetements ceramiques de type barriere thermique
 1-9

; incidence sur la resistance thermomecanique de stabilised (yttria) zirconia single crystals», J. Am.
ces revetements. These de l'Universite, de Paris Ceram. Soc. 71(4), 1988, 255-260.
Sud (Orsay). 20. Pawlowski L., Lombard D., Mahlia A., Martin C,
7. Alperine S., Lelait L., «Microstructural Fauchais P., « Thermal diffusivity of arc plasma
investigations of plasma-sprayed yttria partially sprayed zirconia coatings», High Temp. High
stabilized zirconia TBC, in relation with Press. 16, 1984, pp 347-359.
thermomechanical resistance and high 21. Taylor R.E., « An overview of the measurements of
temperature oxidation mechanisms», in Proc. thermophysical properties and some results of
37th ASME international Gas Turbine & molten superalloys and semiconductors», in
Aeroengine Congress (Köln) 92-GT-317 (1992). NASACP10121, 1993,ppl29-142.
8. Brindley W.J., «Properties of plasma sprayed 22. Klemens P.G., «Thermal conductivity of
bond coats», in TBC Workshop, March 27-29, zirconia», in «Thermal conductivity 23»,
1995, NASA CP3312, pp 189-202. Technomics, Lancaster PA (1996), pp 209-220.
9. Pint B.A., Wright LG., Lee Y.J., Zhang Y., 23. KlemensP.G., «Theory of the thermal
Pruzner K., Alexander K.B., « Substrate and bond conductivity of solids», in «Thermal
coat compositions : factors affecting alumina scale conductivity», vol.1, Acad. Press, 1969, 1-68.
adhesion », to be published in Materials Science 24. Kittel C, «Introduction to solid state physics »,
and Engineering. 1986, Wiley.
10. Schulz U., Fritscher K., Ratzer-Scheibe H.J., 25. Slack G.A., «The thermal conductivity of
Kaysser W.A., Peters M. , in Proc. 4th Internat. nonmetallic crystals», Solid State Physics 34,
Symp. on High temp. Corrosion & Protection (Les 1979, pp 1-71.
Embiez, France, 20-24 May 1996), to be 26. Schulz U., Fritscher, Peters M., « EBPVD Y203-
published. and Ce02/Y203-stabilized zirconia thermal barrier
11. Siegel R., Spuckler CM., «Analysis of thermal coatings - crystal habit and phase composition»,
radiation effects on temperatures in turbine engine Surf. Coat. Technol. 82, 1996, pp 259-269.
thermal barrier coatings», in Proc. TBC 27. Maloney M.J., Achter H.S., Barkalow B.K.,
Workshop 97, Cincinnati, OH, 19-21 May 1997, « Development of low thermal conductivity thermal
pp 17-26. barrier coatings». Communication presented in
12. DeMasi J.T., Ortiz M., Sheffler K.D., « Thermal Thermal Barrier Coating Workshop (Cincinnati,
barrier coating life prediction model development. may 1997).
Phase 1 », NASA CR 182230, 1989. 28. Padture N.P., Klemens P.G., «Low thermal
13. Miller RA, Leissler G.W., Jobe J.M. (1993) : conductivity in garnets», J. Am. Ceram. Soc.
Characterization and durability testing of plasma- 80(4), 1997, pp 1018-1020.
sprayed zirconia-yttria and hafnia-yttria thermal 29. Fournier D. et al., to be published (1997).
barrier coatings. Part I. NASA TP 3295. 30. Z.R. Dai, Z.L. Wang, Y.R. Chen, H.Z. Wu, W.X.
14. H.E. Eaton, J.R. Linsey, R.B. Dinwiddie (1994) : Lu (1996) : Local ordering of oxygen vacancies in
The effect of thermal aging on the thermal cubic zirconia stabilized with yttria and magnesia.
conductivity of plasma sprayed fully stabilized I. Electron diffuse scattering study. Phil. Mag.
zirconia. Thermal Conductivity 22, 289-300. 14b. 73(2) 415-430.29b. Z.R. Dai, Z.L. Wang, W.X. Lu
Dinwiddie R.B., Beecher S.C., Porter W.D., (1996) : Local ordering of oxygen vacancies in
Nagaraj B.A., « The effect of thermal aging on the cubic zirconia stabilized with yttria and magnesia.
thermal conductivity of plasma-sprayed and II. Determination of local ordering parameters of
EBPVD thermal barrier coatings», ASME 96- oxygen vacancies. Phil. Mag. 73(6) 1685-1698.
GT-982, pp 1-7. 31. Klemens P.G., «Theory of heat conduction in
15. Morrell P. , Taylor R., « Thermal diffusivity of nonstoichiometric oxides and carbides», High
thermal barrier coatings of Zr02 stabilized with Temp. High Press., 17, 1995, pp 41-45.
Y203 », High Temp. High Press. 17, 1985, pp 79- 32. Fukushima S., Ohmichi T., Handa M., « The effect
88. of rare earths on thermal conductivity of uranium,
16. Staniek G., Marci G, in Proc. Werkstoff- plutonium and their mixed oxide fuels», J. Less-
Kolloquium DLR, 10 dec. 1996, eds M. Peters et Common Metals, 121, 1986, pp 631-636.
al., pp 50-53. 33. Klemens P.G, « Phonon scattering and thermal
17. Portal R, Etude de la conductivite thermique de resistance due to grain boundaries», Internat.
couches minces de Zr02-Y203 deposees par Journal of Thermophysics 15(6) , 1994, ppl345-
EBPVD », Rapport de stage SNECMA,1997. 1351.
18. Hasselman D.P.H., Johnson L.F., Bentsen L.D., 34. Cahill D.G., Watson S.K., RO. Pohl R.O.,
Syed R, Lee H.L., Swain M.V., «Thermal «Lower limit to the thermal conductivity of
diffusivity and conductivity of dense disordered crystals », Phys. Rev. B 46(10), 1992,
polycrystalline Zr02 ceramics : a survey», Am. pp 6131-6140.
Ceram. Soc. bull. 66(5), 1987, 799-806. 35. McPherson R, «A model for the thermal
19Youngblood G.E., Rice R.W., Ingel R.P., conductivity of plasma-sprayed ceramic coatings»,
«Thermal diffusivity of partially and fully Thin Solid Films, 112, 1984, 89-95 .
1-10

36. Moreau C, Boire-lavigne S., Saint-Jacques R.G.,

« The relationship between the microstructure and
thermal diffusivity of plasma-sprayed tungsten
coatings », in Proc. 7th National Thermal Spray
Conf., 20-24 June 1994, Boston, ASM, pp 621-
626.
37. Dorvaux J.-M., Lavigne O., Poulain M., Renollet Y.
Rio C, Mevrel R, «Modelling the thermal
conductivity of TBCs », this conference, paper 12.
38. Brotzen F.R., Loos P.J., Brady D.P., « Thermal
conductivity of thin Si02 films », Thin Solid Films
207, 1992, pp 197-201.
39. Henager C.H., W.T. Pawlewicz, «Thermal
conductivities of thin sputtered optical films».
Applied Optics 32(1), 1993, pp 91-101.
40. Lee S.-M., D.G. Cahill, T.H. Allen, «Thermal
conductivity of sputtered oxide films ». Phys. Rev.
B 52(1), 1995, pp 253-257.
41. Lambropoulos J.C., M.R. Jolly, C.A. Amsden,
S.E. Gilman, M.J. Sinicropi, D. Diakomihalis,
S.D. Jacobs, « Thermal conductivity of dielectric
thin films », J. Appl. Phys. 66(9), 1989, pp 4230-
4242.
42. Shaw-Klein L.J., «Microstructural effect on
thermal conductivity of thin films », Ph. D. thesis.
University of Rochester (USA) 1992.
43. Oishi Y., Ken Ando, Sakka Y., « Lattice and
grain-boundary diffusion coefficients of cations in
stabilized zirconias », Adv. Ceramics, 7, 1984, pp
208-219.
44. Mullen R, Allen W.P., Gell M, Barkalow R.H.,
Noetzel A.A., Appleby J.W., Khan A.A.,
« Multiple nanolayer coating system » PCT WO
96/11288, 1996.
 3-1

RECENT DEVELOPMENTS IN PLASMA SPRAYED THERMAL BARRIER COATINGS

Drs. P. FAUCHAIS, A. VARDELLE and M. VARDELLE

L.M.C.T.S. - Equipe Plasma, Laser, Mat6riaux

Universite de Limoges - UPRES A 6015 - Faculte des Sciences
123, Avenue Albert Thomas - 87060 LIMOGES Cedex - FRANCE

Abstract - at last, the parameters which have to be

controlled during spraying to achieve a much better
This paper presents a review of our present reproducibility of coatings. A special emphasize is
knowledge in the formation of plasma sprayed given to torch voltage fluctuations related to
Thermal Barrier Coatings (TBCs). The following electrodes erosion, powder injection conditions and
points are examined for TBCs made of zirconia finally substrate and coating temperature before
partially stabilized with 8 wt % of yttria: (preheating) during and after (cooling) spraying.
- characteristic times for particle flattening, splat
cooling and solidification with the corresponding 1. INTRODUCTION
cooling rates and microstructure.
- critical preheating temperature of substrates or Thermal barrier coatings (TBCs) are now
previously deposited layers to achieve a good contact included as part of the initial engine of turbine-
with splats. An explanation of the effect of this critical powered aircrafts. They make them possible to
temperature is proposed. achieve an average reduction in the metal temperature
- times between two impact events at the same for turbine blades of 50 to 80°C and reduction of
location and two successive passes. The consequences blade hot-spot temperature of up to 139°C. The latter
on the mean temperature within coating in are only a fraction of the capability of TBCs,
conjunction are underlined for various substrate and estimated at 170°C or more [1]. TBCs have also been
coating cooling devices. used in land-based turbine combustors for 10 years,
- the effect of substrate critical temperature and and in diesel engines where graded coatings are
substrate oxidation stage on coating adhesion/ highly effective. In all cases, the purpose is to
cohesion. enhance component durability while running the
- splat layering and temperature conditions at engine operating temperature. In this paper TBCs for
which a columnar structure can grow through the diesel engines will not be considered. The difference
whole coating. The consequences on stresses develop- between nominal requirements for aircraft and power
ment during and after spraying and coating Young's generation turbine applications are summarized in
modulus, are discussed. table 1 [2].

Requirement Commercial aircraft Power generation

Number of cycles 8000 2,400
Total hours 8000 24,000
Hours at peak conditions 300 24,000
Peak surface temperature >1204°C (2200°F) <1204°C (2200°F)
Peak bond coat temperature 1093°C (2000°F) 954°C (1750°F)
Relative size lx 5x
Table 1 : Comparison of nominal TBC requirements for aircraft
and power generation turbine applications [2].

Paper presented at an AGARD SMP Meeting on "Thermal Barrier Coatings",

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
3-2

The main criteria for power generation use of - a columnar structure under certain conditions
TBCs are the time and temperature effects on bond of evaporation,
coat and substrate, coating densification and changes - a fast tapering at the opening of cooling holes
in thermal or mechanical properties of the coatings. while plasma sprayed coatings have a greater
As underlined in reference [3] for advanced tendency to built up coating at that location,
land-based gas turbines, the goals that are to be - possibility to achieve much more smoother
achieved together with issues and challenges are surfaces,
summarized in fig. 1. - high adhesion/cohesion values (~ 70 MPa),
For a long period (more than 20 years) TBCs - low residual stress at room temperature
were essentially plasma-sprayed in air (APS) [4-7] or (~ 70 MPa),
controlled atmosphere [8]. Since a few years Physical - reduction of temperature of 38-66°C in stage of
Vapor Deposition assisted by Electron Beam (EB- High Pressure Turbine Blade,
PVD) TBCs, have been available on the market [9,10] - high number of cycles in engine testing : 180
especially for aircraft engines. The advantages for blades and 20000 for engine nozzle.
claimed relatively to APS TBCs are the following :

Improving TBC performance requires

a systems approach
Goals Issues & Challenges
Higher thermal • Microstructure, Stabilizer
stability • APS, EBPVD.etc.

Reduced spallation
at the interface
V
Understanding damage
mechanisms
Oxidation,Fatigue, Creep, Phase
& Property changes, Hot
Improved corrosion, Erosion, Diffusion, etc
oxidation/corrosion Reliable life prediction
Protection

f
• PtAI, MCrAlY, Desulfurization
• VPS, HVOF, CVD etc.

Fig. 1 : Approach for improving TBC performances [3].

However, it does not mean necessarily that APS spraying to achieve a much better reproducibility of
TBCs have no future. For example they have a the plasma spray process.
significant advantage in being able to accommodate
the large size components found in modern large 2. PLASMA SPRAYED COATING FORMATION
power generation machines. Each component
application of a TBC requires careful study and All the results presented in this paper concerned
review to determine whether APS or EB-PVD has to fused and crushed powders with 8 wt% Y203. They
be used [3]. Moreover, a better understanding of the are called in the following YSZ.
APS coating formation could result in coatings with
similar properties as those obtained by EB-PVD. The 2.1. Characteristic times and cooling rates
aim of this paper is first to review what is our Table 2 summarizes characteristic time values
knowledge in plasma sprayed coating formation. We for particles in the size range 22-70 urn sprayed with
will first were discuss the splat formation, their d.c. torches and impacting at velocities between 80
layering and the resulting coating adhesion/cohesion and 260 m/s on smooth surfaces (Ra < 0.1 urn) [11-
as well as residual stresses. We will second underline 14].
which parameters have to be controlled during
 3-3

Flattening time Starting of nucleation Solidification time Time between two Time between two
if (us) tn(Hs) ts (us) impact events at the successive passes
same location t; (|is) U (Us)
1-3 0.3-2.5 <10 500-5000 106-107
Table 2 : Characteristic times in d.c. plasma spraying.

These various times show clearly that the

impacting particles and resulting splats will K varies between 0.8 and 1.2941 while a is
experience very different temperature histories : either 0.2 or 0.125 or 0.167.
Typical values for a zirconia particle 30-um in
2.1.1. Single splat diameter, impacting at vp = 200 m/s at a temperature
Many theoretical studies have been devoted to Tp = Tm x 1.3, where Tm is zirconia melting
the impact and flattening of the particles on the temperature, are for example \ = 4.8 which
substrate (see their description in [14]). All of them, corresponds to a splat thickness es ~ 1.1 urn and tf ~
except one, concern impacts on smooth surfaces. 1 us. If the impacting particle velocity increases, %
increases while tf and es decrease and it is the reverse
a) Smooth surfaces if vp decreases. The most recent theories have shown
Phenomena at impact can be characterized by that solidification can start before flattening is
the particle Weber's number (We = pp . d .vp2/ap completed [15]. Solidification depends mainly on
where pp, vp, d are the particle specific mass, impact [14,15] the wettability of the molten material, the
velocity and diameter respectively and ap its surface thermal contact resistance Rtf, with the substrate, the
tension), and Reynolds' number (Re = pp . vp . d/Up splat thickness es and the thermal properties of the
where Up is the particle viscosity at impact). When substrate material. The interfacial contact resistance
flattening starts, We can be rather high (up to 10000) characterizes the real contact with the substrate or the
and the flattening process is independent of particle previously deposited layers. It can vary between 106
surface tension which has importance only when and 10-8 m2.K/W, the last value corresponding to a
flattening is almost completed. These theories end up perfect contact. Fig. 2 represents the time-tem-
in relationships giving the particle flattening degree I, perature evolution of a zirconia splat 1-um thick for
and flattening time tf as functions of the Reynolds two values of R^. For values below 107 m2.K/W (see
number. ^ is defined as the ratio of the splat diameter fig. 2.a), the temperature Tb of the lower surface of
D (assumed to have a disk shape) to that of the the splat decreases very fast while the substrate
impacting particle d. surface temperature T; increases very fast. Over R^ =
lO-7 m2.K/W, there is nearly no temperature gradient
S = K.Re« (1) within the splat (see fig. 2.b) while Tj is much lower.

tf = 2.dp.Reo.2/(3.vp) (2)

Anon -n

3500 -
* 3000 - \Splat surface
§ 2500
g 2000 -
Q) Vsplat bottom
a 1500 -
\ "■^SSSte*™-,,,
o 1000 -
Substrate surface 500 -f
500 - Substrate surface
0
n
50 100
() 5 10
Time (ps) Time (|JS)
2
a) R,,, = 10 s K.m /W b) R^ = 5.HM K.m2/W
Fig. 2 : Splat surface and bottom as well as interface temperature evolutions for two values of R^,
(1-um thick Zirconia splat on a smooth stainless steel substrate at 600 K).
3-4

In both cases the increase in splat surface temperature

is due to the recalescence phenomenon. However,
these results, which have been obtained by assuming a
constant nucleation temperature of 2500 K and an
instantaneous full contact of the whole splat, can be
deeply modified if nucleation process is included in
the model and the evolution of the contact surface
during flattening is accounted for.
The corresponding cooling rates during the first
Us are between 10* and 109 K/s. For R^ < 10-7
m2.KAV the heat transfer rate from droplet to
substrate is very high and the liquid can be cooled
well below the melting point in one microsecond or
less. It is therefore quite plausible that in this very Fig. 3 : Impulse reheating of the first splat surface by
short time interval, the liquid does become the layering splats.
undercooled and nucleation of crystals is delayed until
after the droplet has almost completely spread out. The temperature Td is linked to :
As shown in fig. 2, if during flattening and just - substrate preheating temperature Tpn ;
after, the cooling rate is very high, its slows down - substrate dimension, thickness and thermal pro-
drastically for times longer than 10 to 50 us. The splat perties (controlling the heat sink);
temperature tends then slowly to the mean substrate - relative movement torch to substrate. This mo-
and previously deposited layer temperature Td vement determines the pass thickness ep (the heat
imposed by torch heating, previously deposited pass content of layered splats increases with ep which can
thickness and cooling devices. vary from a few layered splats to a few hundreds) and
the time t, for the torch to come back at the same
b) Rough surfaces location (heating by plasma plume flux);
Only one model accounting for the substrate - cooling devices used : air jets blowing at the
roughness has been proposed [16] up to now. The substrate and coating surface, and air barrier blowing
roughness was schemed by cones. This model gives orthogonally to the plasma jet to reduce in a ratio up
for the flattening degree a relationship similar to (1) to six the plasma plume heat flux [18], but also
but multiplied by a function depending on the cone cooling down the droplets traversing it [19].
height and the valley volume. It results, of course, in
thicker splats which shape is not that of a disk. 2.2. Splat formation
Assuming the same value of Rm than for smooth Experimental set-ups were developed to
substrates, the cooling rate will be lower due to the determine the velocity vp, surface temperature Tp and
higher value of es. diameter d of a single particle prior to its impact and
then its temperature and surface evolutions during
2.1.2. Layering splats flattening as well as the surface temperature evolution
As powder loading is rather low in plasma of the resulting splat during its cooling. The interested
spraying [14], statistically a new particle impacts on a reader can find a presentation of these set-ups in [12-
previously deposited splat in times t; of the order of 14].
500-5000 us i.e. on already solidified splats. The Other techniques allow the collection of a few
surface temperature of the first splat due to the impact thousands of splats all over the spray cone on a
of successive layering splats will depend strongly on controlled temperature substrate to determine their
Rth and the mean temperature of substrate and coating diameter and shape factor distributions [13].
Td. This leads, as schemed in fig. 3, in impulse At last, a detailed observation of the structure of
reheatings with the first splat surface peak the splats collected on smooth surfaces can be obtai-
temperature decreasing progressively due to the ned by Atomic Force Microscopy making it possible
insulation by the successively deposited splats. to see the details of the columnar growth [20].
 3-5

The main results are as follows :

2.2.1. On smooth surfaces

Experimental evidences [21] have shown that
there is a substrate critical temperature Tc, below
which the splats are extensively fingered while over
which, they have almost a perfect disk shape as shown
in fig. 4. For ZrC>2 it was found that this transition
temperature is between 450 and 550 K.
This phenomenon seems to be almost
independent of the particle impact velocity. For
example Table 3 presents results obtained with
-45 + 22 urn particles which velocities range between
200 m/s along the jet axis to 110 m/s at the limit of Fig. 4 : Micrographs of zirconia particles sprayed, on
the spray cone. The data of this table correspond to (a) 350 K substrate, (b) 550 K substrate (impact
about 3000 splats processed by image analysis. On the velocity - 200 m/s).

Substrate temperature Tph (K) 350 573

Analyzed area center periphery center periphery
D splat mean diameter (um) 67 ±24 59 ± 20 107 ± 44 101 ± 36
SF shape factor (urn) 0.68 ± 0.20 0.67 ±0.21 0.95 ±0.13 0.94 ±0.12
CR cooling rates (K/us) 100-200 ~ 500-800 -
Table 3 : Splat mean diameter D, shape factor SF and cooling rate CR with substrate
temperature and location within the spray cone.

hot substrate, the mean Shape Factor SF is close to The latter exhibit on their whole surface, a
one, both in the spray cone central part and its homogeneous microstructure with a columnar growth
periphery even if the mean splat diameter D is perpendicular to the surface [20] (see fig. 5.a) except
slightly smaller in the spray cone periphery, as it in the splat rim (see fig. 5.b) where the columnar
could be expected. SF is defined as 4 n S/P2 where structure is parallel to the substrate. Indeed, due to
P is the splat perimeter and S its surface area (SF = 1 splat curling [24] the rim is not in contact with the hot
for a disk). For cold substrates D is by far smaller substrate and cooling occurs through the already
due to material splashing and SF ~ 0.67 against 0.94. solidified columnar structure of the central part of the
The dispersion is also higher. However, if the splat resulting in a columnar structure parallel to the
substrate is oxidized (due to a too long preheating) substrate. As the cooling is much slower, the size of
splats are again extensively fingered. The oxidation the columns is bigger. On cold substrates (Tph < Tc)
stage depends strongly on the preheating time with the only a very few AFM measurements could be done,
plasma jet and the substrate material [20,22]. Splat the splats being pulled off very easily. In this case, the
cooling rates, measured by fast pyrometry [12-14], are central part of the splat consists of a few islands
3-6 times higher with the hot substrate. For the disk where the contact is good (columnar structure),
shape splats obtained at 573 K, the splat cooling however with bigger columns than on hot substrates,
theoretical curve [23] of a 1-D model has been surrounded by rims where the crystal growth is
adjusted to the experimental curve by varying the parallel to the substrate [20]. The "fingers" exhibit the
thermal contact resistance. Values of R,h ~ 10-7 - same morphology as that of the rims. As a general
10"8 m2.K/W were obtained corresponding to an rule, the size of the columns depends very strongly on
almost perfect contact as confirmed by the crack the cooling rate as illustrated in Table 4 for two
network due to quenching stress relaxation observed different substrate materials and spraying equipments.
in fig. 4.b. On cold substrates, the contact resistance With the RF plasma torch, the particle impact velo-
Rth calculated assuming that the splats have a disk- cities are lower than 80 m/s resulting in splat thick-
shape, were found to be of the order of 10"6 m2K/W. nesses es in the range 2-2.5 urn against 0.9-1.2 um
This poor contact resistance is confirmed by the pul- with d.c. plasma torches. These thicker splats cool
ling off of splats by the tip of a perthometer. This does down slower (in a ratio 6-10) than the thinner splats.
not occur when splats are collected on hot substrates.
3-6

Fig, 5 : AFM view of an YSZ plasma sprayed with a d.c. torch (vp = 210 m/s, Tp = 3400 K)
(a) splat central part (note the cracks) (b) splat rim.

Plasma torch d.c. torch with a 7 mm i.d. nozzle r.f. torch with a 50 mm i.d.
substrate material stainless steel 304 YSZ stainless steel 304L YSZ
splat thickness es (|xm) 0.95 2.2
CR cooling rate (K/us) 645 115 105 16
ts solidification time (us) 0.74 1.23 4.54 7.6
column width (nm) 125 250 320 >400
Table 4 : YSZ column width with respect to cooling conditions [19].

To explain the occurence of the critical

temperature, Vardelle et al. [11,14] have proposed to tuw ■
compare the time tft at which flattening is almost
completed (tft ~ 0.3-0.4 x tf) to the time tn LU
\^^Ts
corresponding to the onset of nucleation (see fig. 6). If a. SJ
z>
tn > tf„ nucleation starts only when flattening is almost I-,
2000 -
completed and it covers the whole splat surface
w
1
allowing it to keep its disk shape. This is confirmed Q. I
Ti
by the observation of splats collected on tilted ■
UJ
substrates : these splats exhibit an elliptical shape with !
n - i 1
a larger thickness in the direction of the inclined tf» tn
substrate. If t„ < tft nucleation starts at a few points
TIME (us)
which impede the liquid flow, resulting in splashing,
which seems to be confirmed by the AFM splat study. Fig. 6 : Characteristic times and temperatures Ts, Tb
tn depends strongly on the maximum value of the and Tj at splat surface, splat bottom and interface
interface temperature Tj which, in turn, is linked to respectively ; Tpf, is the substrate preheating
the preheating temperature Tpn. temperature.
results in a drop of the splat cooling rate. On the 304L
2.2.2. On rough surfaces stainless steel substrates CR was found to be about
Fig. 7 shows SEM pictures (back scattered 650 K/us for smooth hot substrates and 120 K/us for
electrons) of zirconia splats collected on rough 304L cold substrates With the rough substrates, the drop in
substrate (Ra ~ 6 urn) [13]. It can be seen that, CR is lower.
compared to smooth substrates, the particles are by far
more extensively fingered but here again the contact 2.3. Coating formation
is much better with the hot substrate as shown by the 2.3.1. Thin passes
microcrack network exhibited by splat surface. The Coatings were sprayed [13,25] on a rotating
splats collected on rough substrates are thicker. This substrate holder 110 mm in diameter in front of which
 3-7

the torch was moved parallel to the holder axis. The

substrates were either buttons (0 = 25 mm, e = 6 mm)
or beams (2 x 100 x 15 mm3) made of cast iron FT25
(only botton type), ordinary steel XC38, stainless steel
304L and Inconel 100.
The adhesion/cohesion (A/C) of coatings is
closely linked to residual stresses [26,27]. Beside
those resulting from grit blasting which play a role in
A/C [28], the two most important are the quenching
stress aq which is rather low for YSZ coatings [29]
(aq < 20 MPa) and the expansion mismatch stress aAa
which depends strongly on the temperature difference
between the constant spraying temperature Tj and
room temperature. In order to reduce as much as
possible residual stresses, YSZ coating thickness was
limited to 300 um and the pass thickness was 3 um
thanks to a fast relative movement torch/ substrate. In
such conditions as measured a,q 8 MPa [25] and aAa
- -15 MPa at 300°C and 36 MPa at 500°C on 304L
stainless steel. Thus the stress difference at the
interface substrate coating is rather low and the
coating adhesion/cohesion A/C, measured using the
test DIN 50150 is mainly pure A/C. Table 5 presents
typical A/C values obtained on 304L stainless steel,
each result corresponding to the mean value of 5
measurements. It can be seen, that A/C values
increase with substrate temperature but decrease with
preheating time. This result is in good agreement with
the previous observations on splat formation.
Fig. 7 : Splats collected on rough substrates
(Ra ~ 5-6 urn), a) Tsub = 350 K, b) Tsub = 573 K.

Preheating temperature (K) 573 K 773 K

Preheating time (s) 0 60 300 900 0 120 300 900
Coating adhesion/cohesion A/C (MPa) 20±2 50±2 40+2 31±4 20+2 64±5 49±2 45±2
Table 5 : Coating adhesion/cohesion (A/C) values on 304 stainless steel substrates preheated
by the plasma jet at two different temperatures for different times.

The decreasing in A/C values when the preheating 2.3.2. Thick passes
time increases is explained by the time-growth and In order to limit oxidation problems 300 urn
time-composition change of the oxide layer at the thick coatings sprayed on In 100 substrates were
304L surface, as shown by recent studies [22] : the preheated to different temperatures and the pass
plasma plume contains atomic oxygen and oxidation thickness was varied from 25 to 300 um/pass. With 25
is somewhat different from that obtained by urn passes for preheating temperature Tph over 873 K,
preheating substrates in a furnace. Similar results are aq is more than doubled : 20 MPa against 8 MPa. For
obtained on FT25, XC38 and In 100 substrates for Tp ~ 973 K, the fractured cross section of coatings
different temperatures and preheating times. The exhibits a columnar structure growing through the
coating Young's modulii obtained by comparing the whole coating (see fig. 8.a). The Vickers hardness
deflections of XC38 beams measured during spraying with a 5N load is increased to 1200 ± 80 MPa against
[25,28] with values calculated using a 1-D code based 800 ± 70 MPa with the 3-um thick pass (with Tp =
on the equations proposed by Tsui and Clyne [30], are 773 K). Unfortunately when spraying on In 100
in the range 20-30 GPa. However, the precision is substrates preheated at 973 K, the coating detached
poor (± 30%) with such low values. upon cooling probably due to oxidation problems.
3-8

a) b)
Fig. 8 : Fractured cross section of an YSZ coating :
a) at x = 0.4 mm from substrate and sprayed on an already deposited YSZ coating (300 urn on In 100)
preheated at 973 K ep = 25 um, b) sprayed on In 100 preheated at 973 K, x = 0.005 mm.

When spraying on In 100 at 873 K with ep = 3.1. Powders

80 urn, the first 25-30 urn exhibit no columnar growth 3.1.1. Characteristics
except for 2-3 layered splats and afterwards a The main parameters are the YSZ particles
columnar growth through the whole coating. When chemistry and distribution of stabilizer within each
measuring continuously the beam deflection, a slope particle, shape, density and size distribution. The first
change occurs after the time corresponding to the first three parameters are closely linked to powder
25-30 um corresponding successively to aq = 8 and manufacturing process [32] and the last one to
22 MPa. The Young's modulus of this last part is 100- sieving. The density of the starting particles is one of
120 GPa against 50 GPa for the first 25-30 urn. the key parameter, the best results for coating
During A/C tests, debonding takes place in the first thermomechanical properties being obtained with
25 urn and the maximum A/C value reaches 40 MPa dense particles such as fused and crushed ones [33] or
which could be explained by the tougher layer after agglomerated and densified ones [34]. The stabilizer
25 urn [27]. However, it must be underlined that with distribution within each particle has also a great
hot surfaces the sticking of the small (< 1 urn) influence which is enhanced by the faster evaporation
droplets resulting from YSZ particles evaporation and of the Y2O3 stabilizer compared to ZrÜ2- That is
recondensation [31] has to be carefully avoided probably why particles made by sol-gel process, and
between successive passes because it reduces the A/C then plasma densified, allow to increase the thermal
values by a factor at least of 2. The columnar growth shock resistance of the corresponding coatings by
through layered splats and successive passes can be about 100°C compared to fused and crushed particles
explained on the basis of the temperature diagram [35]. In the sol-gel technique Y2O3 grains, in the
shown in Fig. 2. Calculations with thick passes (ep > 0.1 um size range, are uniformly distributed within
25 um) and Tp ~ 973 K show that nucleation of the each particle. Moreover, it has been shown that after
impacting splat is delayed almost to the end of particle plasma treatment, fused and crushed powders present
flattening (t„ ~tf) which could promote the columnar a better homogeneity than agglomerated ones,
growth with the columns of the underlying solidified sintered or not [36].
splat as nucleation sites. Another important questions is which particle
size range has to be used : rather narrow such as - 45
3. REPRODUCIBILITY OF THE PLASMA + 22 um or large - 106 + 10 um ? and which lower
SPRAY PROCESS limit 22, 10 or 5 um ? The answers depend on the
desired properties. As a general rule, YSZ particles
In industrial conditions the reproducibility of which diameter is over 50-60 urn are not fully molten
coatings is by far the most important problem and with d.c. plasma torches because of the heat
many parameters have to be controlled either before propagation phenomenon in the particles [37]. Thus,
spraying or on-line during spraying, and assessed by powders with a large particle size range may result in
coating analysis. coatings with higher porosity but lower mechanical
 3-9

resistance. For the lower limit of the particle results in an increase of the jet instability and a
distribution the injection problems are critical. Most decrease in jet length coupled with a decrease in
of small particles, especially those below 10 urn, coating quality. When a tungsten insert is used this
hardly penetrate" within the plasma jet, travel in its erosion problem may become even more critical when
periphery and are sucked down farther downstream tungsten particles are ejected and imbedded within
resulting in defects in the coating when they stick to coatings. Therefore, it is important to monitor the
it, which is especially the case when the coating anode erosion by the statistical analysis of the voltage
surface is kept over 773 K during spraying. signal [43] or its power spectrum [44], and determine
when this wear will induce poor quality coating, long
3.1.2. Powder injection in the plasma-jet before its failure.
Internal injection allows a better penetration of This type of monitoring makes it possible to
the particles within the plasma jet compared to check rapidly a defective gun assembly resulting for
external injection [37]. However with small particles, example in a poor electrode centering ; a non
requesting a higher carrier gas flow rate, the plasma symmetrical arc attachment influencing deeply the
flow perturbation is more important with internal restrike mode frequency and the voltage signal
injection [38]. The position of the injection port and evolution.
its distance z from the torch axis is very critical. A
variation of z of a few tenths of mm modifies deeply 3.2.2. Real time control
the particles mean trajectory. Thus, any change in the The outputs that have to be set and maintained
injector port location and its wear have to be checked are the particles mean temperature, velocity, trajectory
very carefully and regularly. It is also the case of the and the substrate and then coating surface temperature
internal wear of the injector, a rough internal surface before (preheating), during and after (cooling down)
modifying deeply the powder distribution at the spraying. Among these parameters the most
injector exit compared to that obtained with a new important, at least to our opinion, are the two last
injector with a smooth internal surface. More ones.
precisely (see section 3.2.2) the trajectory distribution Substrate and coating temperature control
of particles within the plasma jet has to be followed depends strongly on the size, shape and thickness of
continuously because it controls acceleration and the parts to be coated, the relative movements
heating of the particles. between the torch and the substrate and powder feed
rate, linked to the pass thickness, and cooling means.
3.2. Plasma spray process If the surface temperature can be followed by IR
3.2.1. Wear of the electrodes pyrometry, the close-loop controller has to be
The arc behavior at the anode and cathode is designed according to substrate and coating heating
quite different. Thoriated tungsten cathode tip wear is and cooling, if possible by using simplified 1-D or
rather fast (~ 1 h). It is due to the fast diffusion of 2-D computer codes easily adaptable to different
thoria which increases drastically the cathode tip geometries and pass thicknesses.
temperature (by about 1000 K) and induces a neck The mean trajectory of the particles within the
erosion below the tip which is blown up after a while plasma jet is very sensitive to the powder injection
[39,40]. The transition of a sharp cathode tip to a conditions which have to be adapted to the spraying
rounded one, which extremity becomes slowly and conditions [37]. With a 1-D photodiode array, it is
progressively wider, induces, within the first working possible to follow continuously the hot particles
hour, a reduction of the flow velocity of about 20% distribution and, when coupling it to a laser sheet, to
and a reduction of the arc voltage of 4 to 6% [41]. The have simultaneously the trajectory distribution of the
widening of the cathode tip reduces progressively the cold particles. The hot particles distribution is very
voltage by 5-8% after 20 h, and the jet velocity by sensitive to any variation of the injection parameters
about 10%. [46]. For example with Zr02 particles (- 45 + 22 urn)
The arc root fluctuations (restrike mode) [41,42] a variation of the powder feed rate of 50 g can be seen
which condition the arc root life time and thus the and the signal intensity is multiplied by 1.9 when the
anode erosion, evolve with operation time and powder flow rate is doubled (from 1 kg/h to 2 kg/h).
number of torch ignitions. This erosion is distributed Only 1.9 and not 2 due to the starting of the load
in an area which length is about 1.5 x d (d being the effect. Any change of the optimum carrier gas flow
internal diameter of the anode). The depth of the rate m°cg modifying the trajectory is immediately
eroded area as well as the life time of the arc root detected. For example a change of 10% relatively to
increase with the operation time. Anode erosion
3-10

its optimum value induces a 10% change of the signal control the induced residual stresses, temperature of
amplitude and position. substrate and coating have to be monitored
Of course, it is also possible to control on line continuously by adjusting the pass thickness and the
the surface temperature of the particles prior to their cooling devices to the coated part dimensions with the
impact [47] or together the temperatures, velocities help of simplified 1-D or 2-D computer codes.
and trajectories of the particles [48]. These robust At last, to improve coatings reproducibility,
easy-to-use optical sensors, now commercialized, strongly linked to the molten state and velocity of
allow us, after positioning the torch in front of the impacting particles, a rather simple set-up has been
sensor head for 1 minute, to check in industrial developed giving in real time the trajectory
conditions the optimization of the spray parameters distribution of hot particles within the plasma jet and
and their time evolution. For this last point after a allowing to check any change in powder flow rate,
certain spraying time the torch has to be positioned carrier gas injection flow rate, injector wear...
again in front of the sensor head. Such devices make Compared to commercial set-ups following particle
them possible to determine when the load effect velocity, surface temperature and flux density it is by
becomes important, measure the time necessary far much simpler and seems to be as efficient. The
before reaching the stability of particle spraying erosion of the torch nozzle has also to be followed by
conditions, increase the reproducibility of the analyzing the voltage fluctuations which vary
spraying process. drastically when conditions are reached where coating
quality is affected and allows also the detection of
Conclusion defective gun assembly.

Plasma sprayed TBCs thermomechanical proper-

Nomenclature
ties depend strongly upon the contacts between splats
A/C adhesion/cohesion (MPa)
and substrate and layered splats.
APS Atmospheric Plasma Spraying
A better understanding of yttria partially CR cooling rate (K/us)
stabilized (YSZ) splat formation made it possible to
D splat mean diameter (urn)
find a critical substrate preheating temperature Tc, in
d particle diameter (urn)
the 400-500 K range over which these contacts are
ep pass thickness (urn)
strongly improved whatever may be the metallic
es splat thickness (urn)
substrate or the YSZ previously deposited layers. It
Re Reynold's number (dimensionless)
seems that preheating over Tc increases the splat-
Rth thermal contact resistance (m2.K/W)
substrate interface temperature allowing delay the
SF splat shape factor (dimensionless)
time tn at which nucleation starts to a value higher
Tb temperature of the lower surface of the splat
than the time to achieve an almost completed
(K)
flattening. The adhesion/cohesion (A/C) of coatings
T(j equilibrium temperature reached by coating
increases with preheating temperature Tpn provided upon spraying (K)
the substrate oxidation is limited. With 304 L stainless
T; interfacial contact between splat and substrate
steel or In 100 substrates which oxidation resistance is
(K)
good. A/C values can reach 70 MPa with Tpn = 773 K
Tm particle melting temperature (K)
and a preheating time with the plasma torch below
Tp particle impact temperature (K)
120 s.
Tph substrate preheating temperature (K)
To compete with EB-PVD coatings, conditions tf flattening time (us)
have been searched to have a columnar structure
tft time at which flattening is almost completed
through the whole coating (up to 1 mm thick). This
(us)
structure is obtained when substrate or coating tempe-
t, time between two successive impacts (us)
rature reaches about 973 K. According to oxidation tn time at which nucleation starts (us)
problems, when spraying in air, the substrate
ts time at which solidification is completed (us)
temperature has to be kept below this temperature and tt time between two successive passes (us)
thus the first 20 um of the coating do not exhibit this Vp particle velocity at impact (m/s)
columnar structure but works are in progress to solve
We Weber's number (dimensionless)
this problem. When spraying under such conditions Up molten particle viscosity (kg/m.s)
the quenching stress increases drastically (from 8 to Pp particle specific mass (kg/m3)
25 MPa) with a corresponding increase of the coating o"p molten particle surface tension (N/m)
Young's modulus (from 50 to 150 GPa). Therefore, to £ flattening degree (E, = D/d)
 3-11

References [18] F. Monerie-Moulin, F. Gitzhofer, P. Fauchais,

M. Boulos and A. Vardelle, J. of High Temp.
[I] W.J. Brindley, Journal of Thermal Spray Chem. Processes 1 (3) (1992) 249-257.
Technology 5_ (4) (1996) 379-380. [19] A. Haddadi, F. Nardou, A. Grimaud "and
[2] W.A. Nelson and R.M. Orenstein, Journal of P. Fauchais, in Advances in Thermal Spray
Thermal Spray Technology £ (2) (1997) 176- Science and Technology (eds.) C.C. Berndt and
180. S. Sampath (Pub.) ASM Int. OH, USA (1995)
[3] W.P. Parks, E.E. Hoffman, W.Y. Lee and I.G. 249-254.
Wright, Journal of Thermal Spray Technology 6_ [20] L. Bianchi, A. Denoirjean, F. Blein and
(2) (1997) 187-198. P. Fauchais, Thin Solid Films 222 (1997) 125-
[4] D.W. Wortman, B.A. Nagaraj and 135.
E.L. Duderstadt, Mater. Sei. Eng. AJ21 (1989) [21] M. Fukumato, S. Katoh and I. Okane, Proc. of
443-440. 14,h ITSC (Ed.) A. Ohmori (Pub.) High Temp.
[5] R.A. Miller, Surf.. Coat. Technol. 20 (1987) 1- Soc. of Japan, Osaka, Japan (1995) 353-358.
11. [22] A. Denoirjean, A. Grimaud, P. Fauchais, J. Pech
[6] B.C. Wu and E. Chang, Thin Solid Films 122 and B. Hannoyer, Influence of substrate
(1989) 185-196. oxidation on alumina splats in plasma spraying,
[7] M. Yoshida, K. Abe, T. Aranami and Y. Harada, Proc. of United Thermal Spray Conf.,
Journal of Thermal Spray Technology 5_ (3) Indiannapolis 15-18 Sept. (1997), (Pub.) ASM
(1996) 259-268. Int. OH, USA.
[8] S. Sodevka, M. Sazaki and K. Ueno, Journal of [23] A. Vardelle, M. Vardelle, P. Fauchais and
Thermal Spray Technology 5_ (3) (1996) 277- D. Gobin, NATO Series E : Applied Science 282
282. (1995)95-121.
[9] D.V. Rigney, R. Vignie, D.J. Wortman and D.W. [24] M. Bertagnoli, M. Marchese and G. Jaccuci, J. of
Skelly, Journal of Thermal Spray Technology 6. Thermal Spray Technology 4 (1) (1995) 41-49.
(2) (1997) 167-175. [25] A.C. Löger, PhD Thesis (in French), Univ. of
[10] A. Maricocchi, A. Bartz and D. Wortman, Limoges, France, March 12"" (1997).
Journal of Thermal Spray Technology 6. (2) [26] T.W. Clyne and S.C. Gill, J. of Thermal Spray
(1997) 193-198. Technology 5 (4) (1996) 401-418.
[II] A. Vardelle, N.J. Themelis, B. Dussoubs, [27] Y.C. Tsui and T.W. Clyne, see [16], (1996) 275-
M. Vardelle and P. Fauchais, Transport and 284.
chemical rate phenomena in plasma sprays, J. of [28] M. Mellali, P. Fauchais and A. Grimaud, Surf,
High Temp. Material Processes 4 (1997), to be and Coatings Technol. 81 (1996) 275-286.
published. [29] A.C. Leger, A. Grimaud, P. Fauchais and
[12] P. Fauchais, M. Vardelle, A. Vardelle, C. Catteau, see [16] (1996) 891-896.
L. Bianchi and A.C. Leger, Plasma Chemistry [30] Y.C. Tsui and T.W. Clyne, An analytical model
Plasma Processing 1& (1) (1996) 99S-126S. for predicting residual stresses in progressively
[13] L. Bianchi, A.C. Leger, M. Vardelle, A. Vardelle deposited coatings. Part 1 : Plasma geometry,
and P. Fauchais, Plasma Sprayed zirconia splat submitted to Thin Solid Films Dec. (1996).
formation and cooling, accepted in Thin Solid [31] K.A. Gross, J. Tikkanen, J. Keskinen,
Films (1997). P. Fauchais, M. Vardelle and A. Grimaud,
[14] P. Fauchais, A.C. Leger, M. Vardelle and Vaporization and ultrafine particle generation
A. Vardelle, Formation of plasma-sprayed oxide during the plasma spraying process, to be
coatings, in Proc. of J. Szekely Memorial Symp. published in United Thermal Spray Conference,
on Materials Processing (Pub.) TMS (1997). Indiannapolis Sept. (1997) (Pub.) ASM Int. OH,
[15] M. Pasandideh-Fard, Y.M. Quiav, S. Chander USA.
and J. Mostaghimi, Physics of Fluids £ (1996) [32] H. Eschnauer, Thin Solid Films, 23. (1980) 1-17.
650-659. [33] F. Gitzhofer, A. Vardelle, M. Vardelle and
[16] H. Fukanama, in Thermal Spray : Practical P. Fauchais, Materials Science and Eng. A147
Solutions for Engineering Problems (ed.) C.C. (1991) 107-120.
Berndt and H. Herman, (Pub.) ASM Int. OH, [34] A. Denoirjean, A. Vardelle, C. Martin,
USA (1996) 647-656. P. Fauchais, T. Cosack, E. Lugsheier, I. Rass, P.
[17] S. Kuroda, T. Dendo and S. Kitahara, Journal of Chandler, R. Mclntyre and H.L. Heijen, in
Thermal Spray Technology, 4(1) (1995) 75-84. Thermal Spray : Int. Advances in Coatings
3-12

Technology (Pub.) ASM Int. OH, USA, (1992)

967-982.
[35] D. Bernard, PhD's thesis, Univ. of Limoges, June
25 (1990).
[36] B. Kolman, J. Forman, J. Dubsky and
P. Chraska, Mikrochim. Acta 114/115 (1994)
335-342.
[37] M. Vardelle, A. Vardelle and P. Fauchais, J. of
Thermal Spray Technology 1 (2) (1992) 117-
128.
[38] B. Dussoubs, A. Vardelle, M. Vardelle,
P. Fauchais and N.J. Themelis, 131" Int. Symp.
on Plasma Chem. 5 (1997) 2056-2061 (ed.) Prof.
C.K. WU, Institute of Mechanics, Chinese
Academy of Sciences, Beijing 100080, China.
[39] X. Zhou and J. Heberlein, Plasma Chem. Plasma
Proc. 16. (1) (1996) 299S-244S.
[40] X. Zhou, B. Ding and J. Heberlein, IEEE Trans.
Components Pkg. and Mfg. Tech. 1£ (3) (1996)
320-328.
[41] P. Fauchais, J.F. Coudert and M. Vardelle,
Transient phenomena in plasma torches and for
plasma sprayed coating generation, accepted in
J. de Physique III.
[42] J.F. Coudert, M.P. Planche and P. Fauchais,
Plasma Chem. Plasma Proc. 16. (1) (1996) 211S-
228S.
[43] M.P. Planche, PhD's thesis (in French), Univ. of
Limoges, France (1995).
[44] Z. Duan, L. Beall, M.P. Planche, J. Heberlein
and E. Pfender, Arc voltage fluctuations as an
indication of spray torch anode condition, to be
published in proc. of UTSC 97, (Pub.) ASM Int.
OH, USA.
[45] K.I. Li, M. Vardelle and P. Fauchais, see [19]
(1995)59-66.
[46] M. Vardelle, A.C. Leger, A. Vardelle and
P. Fauchais, Influence of the variation of plasma
torch parameters on particle melting and
solidification, to be published in Proc. of UTSC
97, (Pub.) ASM Int. OH, USA.
[47] W.D. Swank, J.R. Fincke and D.C. Haggard, see
[19] (1995) 111-116.
[48] C. Moreau, P. Gougeon, A. Burgess and
D. Ross, see [19] (1995) 141-147.
 4-1

In-flight Particle Diagnostics for On-line Process Control during Deposition of

Plasma-Sprayed TBCs

C. Moreau, P. Gougeon, M. Prystay and L. Leblanc

National Research Council Canada
Industrial Materials Institute
75 de Mortagne
Boucherville (Quebec)
CANADA, J4B 6Y4

1. SUMMARY plasma shape or in the injection geometry may have a

Plasma-sprayed TBCs are built by the successive addition significant effect on the particle trajectories and thus on
of molten or partially-molten particles spreading upon their thermal histories and velocities. Significant variations
impact on the substrate surface. Consequently, the in the coating properties may result from these uncontrolled
temperature and velocity of the sprayed particles are among perturbations.
the most important parameters influencing the
microstructure and properties of the deposited coatings. On- Plasma-sprayed coatings are built by the successive addition
line measurement of these parameters, as well as the particle of molten or partially-molten particles spreading upon
trajectories, is thus an efficient way to monitor the state of impact on the substrate surface. Thus, the temperature and
the spray process. This approach should permits to velocity of the sprayed particles immediately prior to their
minimize the influence of uncontrolled parameters such as impact are among the most important parameters
the electrode wear or changes in particle injection influencing the microstructure and properties of the
conditions. deposited coatings [3]. On-line measurement of these
parameters, as well as the particle trajectories, is thus an
A rugged and easy-to-use optical sensor system has been efficient diagnostic tool for characterizing the spray process.
developed to perform on-line particle diagnosis during Since this approach permits to minimize the influence of
spraying in a production environment. In this system, the uncontrollable parameters, one should expect a closer
thermal radiation emitted by the in-flight particles is control of the spraying process leading to a better
collected by a compact sensing head and transmitted reproducibility of the coating properties.
through an optical fiber bundle to a detection cabinet
located away from the dusty environment around the plasma Different techniques have been already developed to carry
torch. out both temperature and velocity measurements of plasma-
sprayed particles [4-10]. However, the required equipment
In this paper, a review of some applications of this diagnosis is difficult to transfer to an industrial environment due to its
system used to determine the influence of the spray complexity and the use of large optical components and
parameters on the particle temperature, velocity and fragile laser devices. Other approaches have been also used
trajectory and to control the spray process during deposition to provide information on the plasma spray process by
of TBCs are presented. The system was used in production monitoring only the particle velocity [11,12] or the thermal
at Pratt & Whitney Canada (PWC) for two months giving radiation from the hot particles [13-16].
information about the stability and reproducibility of the
plasma spray process. Over the last 20 years, the aerospace industry has acquired a
large experience in controlling plasma spraying process.
2. INTRODUCTION Furthermore, the quality standards have led to the
Plasma spraying is a powerful technique that has been development of numerous recipes to control the process.
extensively used for the last twenty years to produce These controls have shown their efficiency as the different
protective coatings on a large variety of substrates. During coatings succeed in passing the corresponding standard tests
this period, the technique progressed mostly in an empirical required by the industry. Using particle monitoring systems
manner, attention being focused on the effect of input spray in the aerospace industry can however be useful to evaluate
parameters on the resulting coating properties [1,2]. the process stability and then the efficiency of the existing
Consequently, the spraying process control was process control. Moreover, more efficient control
implemented by monitoring and regulating the spray approaches such as the one based on the measurement of the
parameters to keep them at a predetermined optimum value. temperature and velocity of in-flight particles may be the
Arc current and power, arc gas flow rates, powder feed rate, only way to achieve higher quality coatings required in
powder carrier gas flow rate, net plasma energy are the future applications.
main parameters usually monitored and ■ automatically
controlled during plasma spraying. This control approach This paper first describes the particle monitoring system
has important drawbacks. Indeed, it is complex because a developed at NRC. This system is dedicated to monitoring
large number of input parameters must be monitored and of in-flight particles in the laboratory and in production. In a
controlled. It is also incomplete because some variables, second time, a review of a few applications of the
such as the electrode wear state, cannot be monitored at all. monitoring system is presented. A focus is made on the
Moreover, due to the huge temperature and velocity influence of the temperature and velocity of yttria-stabilized
gradients in the plasma plume, a very small change in the zirconia (YSZ) particles on the coating structure and on the

Paper presented at an AGARD SMP Meeting on "Thermal Barrier Coatings",

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
4-2

effect of the wear of the electrodes during deposition of

TBCs. The presented results include those collected over a Sensor head
period of 2 months at Pratt & Whitney Canada during the
production of plasma-deposited coatings.

3. PARTICLE MONITORING SYSTEM

The monitoring system, the DPV2000, is based on the
detection and analysis of the thermal radiation emitted by
the hot sprayed particles [17,18]. This system,
commercialized by Tecnar Automation Ltee [19], consists
of a sensor head located near the spray gun, a detection box
linked to the sensor head by an optical fiber bundle and a
computer (PC) for analysis and statistical computation (Fig.
1). The dimensions of the sensor head are 9 cm long and 3
cm diameter. When a hot particle passes in the measurement
volume defined by optics, its image is formed on a two-slit
optical mask fixed on the end of an optical fiber inside the
sensor head. This event gives rise to a two-peak signal at the
output of the photodetectors receiving the radiation through
the optical fiber. Fig. 2 shows a schema of the sensor head
and the corresponding fields of view of the detection optics.
The distance between the slits as well as the magnification Figure 2 Schema of the sensor head and fields of
of the detection optics being known, the particle velocity view of the linear array and two-slit mask on the
can be computed from the time of flight of the particle center fiber.
image between the two slits. The particle temperature is
obtained from two-color pyrometry by analyzing the signals particle jet is collected and imaged on a row of optical
from the two photodetectors receiving the radiation in two fibers located in the sensor head near the fiber used for the
separate spectral ranges. Finally the particle diameter is measurement of the individual particle parameters. The
computed from the absolute radiation intensity at one height of the field of view seen by the optical fiber row is 4
wavelength and the temperature of the particle. cm. The optical fibers are coupled to a CCD linear camera
located in the detection box. The parameters of the particle
Simultaneously to the measurements on the individual jet are computed from the analysis of the video signal of
particles, the monitoring system measures also parameters this CCD camera which is a representation of the spatial
of the particle jet considered as a whole. This information is distribution of radiation in the particle jet. Three different
useful to detect .problems with the injection of the particles parameters are obtained from this analysis: the position of
in the plasma jet. For that purpose, the radiation of the the particle jet relative to the axis of the sensor head, the
width of the particle jet and its total radiative intensity. This
particle jet last parameter is very useful to detect variations in the
sensor ^, powder feeding rate which are detected as time variations of
field of view the intensity. Measurements of individual particle
parameters and particle jet parameters are made
powder simultaneously and give complementary information on the
feeding spraying conditions.

4. RESULTS AND DISCUSSION

4.1 Monitoring of particle properties during coating

\
production
plasma gun The objective of this study was to follow, during a relatively
long period of production, the evolution of the particle
characteristics as measured with the DPV2000 monitoring
system [20]. For that purpose, the procedure to produce
coatings has been slightly modified. Usually, at beginning
of spraying, the operator waits until the process is stable
Particle before moving the torch in front of the part to be coated. For
Parameters this study, when the process became stable, the torch stayed
in place near the sensor head to allow the measurement of
particle parameters for one minute with the monitoring
system. In the same way, once the part was coated, the
process was not stopped immediately and the torch came
back in front of the sensor head for another one-minute
measurement. All the measurements were carried out at 63
Figure 1 Schema of the optical system. The thermal mm from the torch which is the usual spray distance.
radiation from the particles is received by the sensor
head and the corresponding signals analyzed by the An example of results obtained in this way is illustrated in
personal computer. Figs. 3a and 3b. These figures show the evolution of the
 4-3

particle temperature just before and after spraying a YSZ

coating on a specific part. In Fig. 3a, the spraying process _ 3100
starts at 50 seconds and, before reaching the stability criteria ü 3000
at 150 seconds, a large increase of temperature is noted. CD
After the one-minute stability period, around 210 seconds, 3 2900
the torch moves in front of the part for coating deposition
and there are no more particles measured by the monitoring
0 2800
system. In the example of Fig. 3, the spraying process lasts 1 2700
about one hour, and then the torch comes back in front of H
the sensor head for one minute during which the process is 2600
stable. After this minute, at around 4300 seconds on Fig. 3b,
the process is stopped according to the normal procedure of
PWC. Results similar to those of Fig. 3 have been obtained Ü
for all the parts sprayed with YSZ powders. These results
0
3000 - (3UN2
CD
show that the temperature of the sprayed particles is stable 3 2900 - ▲
when the PWC criteria of stability are reached. CO
k_
CD 2800 -
Q.
From these results, the average values on the 60 seconds of E 2700 -
stability can be computed for each measured parameter, i.e. CD
1-
the particle temperature, velocity and diameter. The average 2600 - 1 1 III
temperature of the YSZ powder sprayed with 3 guns used in
production is shown in Fig. 4. These results are obtained
O l\J\J
during production of coatings with three sets of identical
electrodes using the same spraying conditions. The abscissa " 3000 - GUN 3
of Fig. 4 is the number of hours of operation for each set.
Measurements before (down triangles) and after (up = 2900 -
triangles) coating deposition are represented at the same cS 2800 -
abscissa if they correspond to deposition on the same part.
I 2700 -
H
JUUU 2600 - i I 1 1 1
(a)
0
Ü 2800 - Istability
Time (hours)

rn
CD
3
Figure 4 Time evolution of YSZ particle temperature
00 2600 - for 3 guns used in production : (T) before and after
(A) deposition on different parts.

t
L_
CO
Q.
E
CD 2400 - The number of points differs for each gun as the guns were
i
used for spraying different powders and only the
i measurements on YSZ particles are shown. It is worth
??nn i
i I i noting that there is a difference of typically 20°C between
temperature measured before and after deposition on each
50 100 150 200 250
part. Even if this difference is small, it is significant,
Time (sec) indicating that the conditions of the particles have slightly
3000 changed during spraying.

The guns 1,2,3 have been changed after 46, 58 and 48 hours
9 2800 of spraying, respectively. For gun 3, the reason of change is
not imposed by consideration of stability of the process but
CD
is due to a problem of waterproofness in the torch. With this
-I—«
öS
CO 2600 - gun, the process was considered stable according to the
L_
CD PWC criteria and also according to the measurements on the
Q. particles (Fig. 4). The situation is different for guns 1 and 2
E as the temperature and also velocity of the particles (not
CD 2400
shown) vary significantly during the use of these guns in
production. It appears that these variations are often related
2200 to variations of voltage between electrodes which is one of
the parameters recorded by the operators. In fact, when the
4200 4250 4300 4350 4400 4450 voltage variations are too large, the usual procedure is to
stop spraying as it is the case for gun 1 after 46 hours of
Time (sec) operation. Consequently, the two triangles at 46 hours
correspond to a beginning of spraying but no deposition was
Figure 3 Example of particle temperature evolution made as the voltage variation was too large. Even if gun 1
during coating production with YSZ particles, (a) before was used for coating deposition during 46 hours, Fig. 4
deposition, (b) after deposition.
4-4

shows that the spraying conditions of the particles have metallic (NiCoCrW) powders, 27 urn average diameter.
constantly and significantly changed after 35 hours of Figure 5 shows the evolution of the particle jet intensity
operation. It is worth noting that the process was controlled before and after deposition. In this case, the process didn't
by keeping the input power constant. So, as the electrodes reach a stable operation point even after 3 minutes of
become worn, the voltage tends to decrease and the spraying when the torch normally moves to spray the part to
controller reacts by increasing the current in the gun. In the be coated. Such a long time to obtain stable spray
case of gun 1, it clearly appears that such a correction was conditions is likely related to the feeding conditions of the
too large leading to a significant increase of the temperature fine powder into the plasma.
of the sprayed particles. A corresponding increase in the
velocity of the particles was also noted.

The gun 2 was changed because it reached 60 hours of

spraying time. In such a case, changing the gun prevents
eventual problems which can appear during production of
coatings. After the 58 hours of operation of gun 2, no
significant change was noted for the particle temperature.
However, an increase of the particle temperature has been
observed between 40 and 46 hours of gun operation but
after 46 hours the conditions become again similar to what
they were before 40 hours. This variation of temperature
between 40 and 46 hours is correlated with a variation of
the voltage between the electrodes but this change was
considered too small to stop the process. It should be noted
that during production with guns 1 and 2 particle
temperatures have varied by 180°C and velocities by 30
m/s. This means that, in spite of the process control used to
produce coatings, the particle spraying conditions can vary
significantly.

The values for the average particle parameters for each gun
are given in Table 1. The values for each parameter are
average values obtained during the first hours of use of each
gun, that means during the period of time they give
relatively stable conditions for the particles. From Table 1, n 1 r
it appears that the different particle parameters are very
similar for guns 1 and 3 except the difference of 50°C in 600 650 700 750 800 850 900
temperature. The differences with gun 2 are larger as the
temperature difference reaches 82°C. From Table 1, it is Time (sec)
clear that in spite of the control established on the process,
there are still significant differences in the particle Figure 5 Time evolution of the particle jet intensity
characteristics. These differences can be explained by a collected by the CCD camera.
variation in the way the particles are injected in the plasma
jet. This is especially important if the plasma characteristics
change as we can expect if the plasma gun is changed.
4.2 Effect of the wear of the electrodes
Consequently, a control of the particle state appears more
In order to better understand the effect of the wear of the
sensitive to variations in the process conditions than the
electrodes during spraying, a series of measurements of the
current approach based on the control of plasma power.
particle characteristics was carried out in the laboratory
[21]. A torch was mounted with a new set of electrodes
It was observed that, in the case of YSZ particles, the
and the state of the particles was monitored as a function of
spraying conditions are relatively stable during the stability
the spraying time. At regular intervals, YSZ coatings about
periods (Fig. 3) even if the temperature can slightly vary
500 urn thick were sprayed.
during deposition. This is not true in the case of finer
4.2.1 Two-Dimensional Profiles
Four 2-D profiles were measured after 1, 8, 22 and 37
Table 1 Particle parameters measured during the first hours of spraying. Mean temperature, velocity and
hours of operation for 3 guns used in production. diameter of the sprayed particles were monitored during the
scans. These profiles were collected by moving the sensor
GUN1 GUN 2 GUN 3 head with a X-Y unit, scanning in this way the
measurement volume of the DPV2000 across the particle
Temperature (°C) 2687 2769 2734
jet at 65 mm from the gun. The number of detected events
Velocity (m/s) 156 161 152 which is related to the particle flux at the measurement
Diameter (urn) 55 67 60 volume location was also counted for a fixed period of time
Flux (part/min.) 1065 1290 1123 (25 ms). This number can't be taken as the true particle flux
Jet position (mm) -0.9 0.0 -0.5 but is qualitatively related to it in the sense that both follow
Jet width (mm) 4.7 4.6 4.7 the same trends.
Jet intensity (a.u.) 1.0 1.3 1.1
 4-5

Figure 6 represents the evolution of the mean particle

temperature with the spraying time. The temperature
distribution as well as the velocity distribution (not
illustrated here) followed the same behavior during the -
experiment starting with their highest values at the
beginning of the experiment (one hour of spraying) and
10 : /^^'
decreasing continually until the end of the experiment (37
^700// ^xA
hours of spraying). A variation of more than 200 °C is o :
observed between the temperature distributions measured E H 3100 3300-^ y j
at the beginning and at the end of the forty-hour E
experiment. During this time, the velocity distribution has > -10 : 31
tcT^-2900—^?
varied by more than 30 m/s. It is worth noting that, in this - ^ 2700
series of measurements, the plasma current was kept :
constant at 800 A. The results show that even though the -20
After 1 hour of spraying
input spray parameters stayed the same, the particle state -TTTTT-r-T II | 1 1 1 1 | r 1 1 1
has significantly changed. Such temperature variations may
-20 -10 0 10 20
lead to important coating microstructure and thermal
property changes [22]. X(mm)

Figure 7 shows the evolution of the distribution of the 10

number of detected events which is related to the particle
flux. At the beginning of the experiment the distribution
was not symmetric, being nearly 50% wider in the 0
horizontal direction. However, the distribution tended to E
E
get more symmetric with increasing spraying time. It is
also observed in Fig. 7 that the maximum number of events > -io:
decreased with increasing spraying time suggesting that the
more the plasma gun was used the more the hot particles :
were lost from the spray jet. This has also been observed -20
After 8 hours of spraying
recently with alumina [23]. i i i i I i i i i I i i i i I i i i i

-20 -10 0 10 20
4.2.2 Coatings and Deposition Efficiency
Coatings were sprayed after 3, 10, 22 and 37 hours of X(mm)
spraying in order to evaluate the effects of the sprayed
particle state on the microstructure of the deposited 10 -
material through the long-term experiment. Image analysis 2700 S 2900 27l 0
>
showed that the total porosity was the same for all
coatings, at about 16%. E ^
After three hours of spraying, the deposition efficiency
was 55% while at the end of the experiment (37 hours of >• -10
spraying) it fell to 41% reflecting important changes in the
state of the sprayed particles. Figure 8 shows the -20
microstructure of the coatings deposited after 3, 10 and 37 I After 22 hours of spraying
i i i i i i i i i i i i i i i ii i i
hours of spraying. As seen in the figure, a clear evolution
of the coating microstructure took place with increasing -20 -10 0 10 20
spraying time. At the beginning, after 3 hours of spraying,
X(mm)
the deposited coating contained a relatively large amount
of cracking. The cracking pattern, though still present, is
not so apparent in the coating sprayed after 10 hours. The 10 ,2700-
mean particle temperature was nearly the same for these
two coatings whereas the particle velocity differed by
nearly 30 m/s. Coatings sprayed after 22 (not shown in E °
Figure 8) and 37 hours were very similar in their E,
microstructure. However, in these coatings cracking
>
patterns like those observed after 3 and 10 hours of -10 "I 27U0
spraying did not form. These microstructure differences
can be attributed to the different particle state before their -20 :
impact on the substrate. Indeed, in a previous After 37 hours of spraying
i i i i l i i i i l i i i i l i i i i
communication [22] it has been demonstrated with a
similar ceramic powder that a mean particle temperature -20 -10 0 10 20
variation of 200°C, such as the one measured in the present X (mm)
experiment between 3 and 37 hours of spraying, can result
in a more than 10% thermal diffusivity variation as well as
significant microstructure changes. It should be reminded Figure 6 Temperature distribution (°C) of the
that between 3 and 37 hours of spraying the gun power sprayed particles after different spraying times. The
gun axis corresponds to X = 0 mm and Y = 0 mm.
4-6

10 - after 3 hours of spraying

>" -10

-20 -
After 1 hour of spraying
i i i i i i i i i | i i i i | i i i i

-20 -10 0 10 20
X (mm)

10

0
E
E after 10 hours of spraying
> -10

-20
After 8 hours of spraying
i i i I i i i i I i i i i I i i i i

-20 -10 0 10 20
X(mm)

10 :

:
E
E
0
'ffi^s >*
> -10

-20 after 37 hours of spraying

After 22 hours of spraying
i i i i | i i i i | i i i i | i i "T r

-20 -10 0 10 20
X(mm)

10

E °
E.
> -10

-20
THIpI I I | I I I ! |I II I

-20 -10 0 10 20 Figure 8 Pictures of coatings deposited after 3, 10

and 37 hours of spraying. The coatings are about
X (mm)
580, 480 and 520 urn thick, respectively. The
number of vertical cracks decreases with the
Figure 7 Number of detected events (thousands) spraying time as electrode wear becomes
after different spraying times. The gun axis significant.
corresponds to X = 0 mm and Y = 0 mm.
 4-7

varied by only 1 kW. It would have been difficult, if not with higher hydrogen proportion and gas flow rate is
impossible, to predict any microstructure changes from this shown in Fig. 9. It can be seen from this figure that the
power variation. Present results show that coatings cracking pattern is rather similar to that of the coating
prepared after different spraying times with the same input sprayed after 3 hours, as seen in Fig. 8.
spray parameters can result in materials having significant
differences in their microstructures and properties. So after a long spraying time (about 40 hours) by changing
some input parameters it was possible to recover the
4.2.3 Attempts to Recover Initial Particle State sprayed particle parameters obtained after three hours of
After the forty-hour experiment, the electrodes were spraying and to recover the same type of coating structure
disassembled from the plasma gun. They were mounted observed before significant electrode wear took place.
again several months later when attempts were made to try
to recover the initial particle parameters. In a first attempt, 4.3 Controlling the particle state and coating structure
the gun current was set to 900 A. However, no significant In order to implement a feedback loop to control the state of
changes in the mean particle temperature and velocity the in-flight particles, it is necessary to know how the input
were observed neither when gun current was increased spray conditions influence the state of the particles. It is also
from 800 A to 900 A and to 1000 A, nor when the powder necessary to know how the temperature and velocity of the
carrier gas flow rate was changed. Nevertheless, when impinging particles affect the coating structure. For TBCs
hydrogen proportion was changed to 5% (as compared to coatings, one key characteristic is the crack network within
3.9% vol. during the forty-hour experiment) the mean the coating. Indeed, it seems advantageous to maximize the
particle temperature increased by about 150°C (from number of cracks perpendicular to the substrate while
3000°C to 3150°C) whereas the mean particle velocity minimizing the number of parallel cracks [24].
increased by more than 15% (from 148 m/s to 175 m/s).
Moreover when the total gas flow rate was increased from An example of how the arc current and total gas flow rate
52.0 1/min to 60.6 1/min, keeping the hydrogen proportion influence the state of YSZ particles is shown in Fig. 10 [22].
to 5% vol., the mean particle temperature increased from The powder used in this study was a fused and crushed YSZ
3150°C to 3200°C while the mean particle velocity powder (-45 +11 urn). After setting the plasma current and
increased from 175 m/s to 190 m/s.). These velocity and total gas flow rate, the powder carrier gas flow rate was
temperature are similar to those measured after two hours adjusted to position the spray jet axis at a constant angle of
of spraying. The measured gun power was 35 kW, which is about 3 degrees below the torch axis. As shown in the
about 5 kW higher than its value after two hours of process control plot of Fig. 10, an increase in plasma current
spraying. increases both the temperature and velocity of the particles
whereas an increase in the gas flow rate increases their
Using these spraying conditions, 2-D profiles of the velocity but decreases their temperature. Subsequently,
particle jet were measured. They are rather similar to those conditions for spraying particles at an exact temperature and
measured after eight hours of spraying (see Figs. 6 and 7). velocity may be obtained by carefully regulating both the
The numbers of detected events are similar to those arc current and gas flow rates. In this way it is possible to
observed after 8 hours of spraying (Fig. 7). A coating as regulate the spray process in order to spray coatings in a
well as a deposition efficiency test were performed using reproducible manner.
the new input parameters. The deposition efficiency was
measured to be 53% which is essentially the same as what After having established the data shown in Fig. 10, it is
had been measured after three hours of spraying (55%) and possible to optimize the coating structure by varying
much higher than the 41% value measured after 37 hours independently the temperature and velocity of the sprayed
of spraying. The microstructure of the coating prepared
3050

3000
current / /
2950 increase /? *^^ /
/
O

v
S- 2900 /gas flow /
/ increase/
f 2850 - / /
| 2800
h-
2750

2700
-Y
2650 i i i i

220 240 260 280 300 320

Velocity (m/s)

Figure 10 Process control plot showing the effect of

Figure 9 Picture of a coating deposited with higher the arc current and arc gas flow rates on the particle
hydrogen proportion and gas flow rate. The coating is temperature and velocity. The arc gas is Ar with 33%
about 450 urn thick. He and the powder is YSZ (-44 +11 urn).
4-8

3050 3000
#2 II
• 2995 °C Vparticle=250 m/s
3000 - #7 • : «#1 2700 " \ Tparticle=2850 °C
2950
250 m/s 2400
ü
2900
2840 °C .
2100 V •/
I 2850 - * #3
CD

V■
#4 *
CD K 1800 - \•
|2800 */ /
200 m/s
1500 -
2750 •/ ■ /

■^ #5 1200
2700 • ■ ■
#6 900 -Tparticle=3014°C
2650 i i i i i i i i i

140 160 180 200 220 240 260 280 300 320 20 40 60 80
Velocity (m/s) Angle (degrees)

Figure 11 Average temperature and velocity of the Figure 12 Angular crack distribution in coatings 4 and
sprayed particles for coatings sprayed in the study of 5 sprayed with particles at 250 m/s and (•)3014°C
Prystay et al. and (■) 2850 °C, respectively.

particles instead of varying the input spray parameters. For the sprayed coatings increase, the process control becomes a
example, in the work of Prystay et al. [22], the coatings key issue in order to produce, day after day, high quality
were sprayed in the 7 conditions shown in Fig. 11. The coatings in a reproducible manner. Monitoring the
velocity and temperature effects were decoupled by temperature, velocity and trajectory of in-flight particles
selecting operating parameters to generate data points that before their impact on the substrate appears as an efficient
fall along a staircase pattern. For example, the influence of way to characterize the process and to perform feedback on
particle velocity at about 2995 °C is seen by comparing the input spray parameters. Moreover, it is also an efficient
coatings 1 and 7. (The average temperature between the tool to optimize the spraying conditions by varying
particle temperatures of samples 1 and 7 is used. Other parameters that have the most crucial effects on the coating
values shown on the graph were determined in the same properties.
manner.) The effect of velocity at particle temperatures of
2840 °C is seen by comparing samples 3 and 4. The results 5. ACKNOWLEDGEMENTS
of changing particle temperature for fixed particle velocities The authors are grateful to many people that contributed in
are obtained by comparing sample 4 with 5 (200 m/s) and this work. In particular, we would like to thank M.
sample 3 with 2 (250 m/s). Lamontagne and S. Belanger from NRC Canada for their
help in spraying coatings and developing the optical
In this work the angular crack distribution was characterized diagnosis system. Moreover, we are grateful to F. Richard
by image analysis using a dedicated algorithm developed in for taking charge of the project at Pratt & Whitney Canada.
our laboratory. This algorithm treats each pore or crack as
an individual feature of equal weighting and no results 6. REFERENCES
regarding their width are included. The orientation of each
feature is determined and the total length of the features (in 1. Smith, R.W., "Plasma Spray Deposition: A Need for
pixels) as a function of their orientation is calculated. Direct Process Control", in "Review of progress in
quantitative nondestructive evaluation," Vol 8B, D. O.
An example of results is given in Fig. 12 where the crack Thomson and D. E. Chimenti (Eds), Plenum Press,
distribution of coatings 4 and 5 are compared. Both coatings New York, 1989, pp 1169-1183.
were sprayed at particle velocity of 250 m/s but with 2. Fincke, J.R., "Diagnostics and Sensor Development fir
different particle temperatures. In this case, the increase in Thermal Spray Technologies", in "Thermal Spray:
temperature has the effect of sharply decreasing the number International Advances in Coatings Technology," Proc.
of horizontal cracks (0 degree) and increasing the vertical Inter. Thermal Spray Conference, Orlando, ASM
cracking (at 90 degrees). In fact, coating sprayed with International, Materials Park, OH (1992) pp 1-9.
particles at 250 m/s and 3014 °C exhibits only a weak 3. McPherson, R., "The Relationship Between the
preferred angular crack orientation. In addition, although Mechanism of Formation, Microstructure and
this coating is not the densest in this study, it is the coating Properties of Plasma-Sprayed Coatings", Thin Solid
with the highest thermal diffusivity [22]. This suggests that Films, 83, 1981, pp 297-310.
the significant reduction in horizontal cracking decreased 4. Mishin, J., Vardelle, M., Lesinski, J. and Fauchais, P.,
the thermal insulating properties of the coating. "Two-colour Pyrometer for the Statistical
Measurement of the Surface Temperature of Particles
5. CONCLUSION under Thermal Plasma Conditions", J. Phys. E:Sci.
Plasma-sprayed coatings are used in a wide range of new Instrum., 20, 1987, pp 620-625.
and demanding applications. As the quality requirements of
 4-9

5. Vardelle, M., Vardelle, A., Fauchais, P., Boulos, M.I., On-line Diagnostics of Particles under Thermal
"Plasma-Particle Momentum and Heat Transfer: Spraying Conditions", in "Advances in Powder
Modelling and Measurements", AIChE Journal, 29, Metallurgy & Particulate Materials", Vol.6, Advanced
1983, pp. 236-243. Processing Techniques, C. Lall and A. J. Neupaver
6. Sakuta, T., Ohtsuchi, T., Sakai, K., and Takashima, T., (Eds.), Proc. Inter. Conf. on Powder Metallurgy and
"Measurement of In-Flight Particle Parameters in Particle Materials, Toronto, Canada, Metal Powder
Thermal Plasma Spaying Process by the Light Pulse Industries Federation, 1994, pp 199-210.
Analysis Method", in "Proc. Jpn. Symp. Plasma 19. Tecnar Automation Ltee, 3502, First Street, St-Hubert,
Chem.", Vol. 4, 1991, pp. 175-180. Quebec, Canada, J3Y 5K2.
7. Fincke, J.R., Swank, W.D., Jeffery, C.L. and Mancuso, 20. Gougeon, P., Moreau, C. and Richard, F., "On-line
CA., "Simultaneous Measurement of Particle Size, Control of Plasma Sprayed Particles in the Aerospace
Velocity and Temperature", Meas. Sei. Technol. 4, Industry", in "Advances in Thermal Spray Science and
1993, pp 559-565. Technology", C. C. Bemdt and S. Sampath (Eds.),
8. Guselnikov, S.M., Zavarzin, A.G., Lyagushkin, V.P., Proc. National Thermal Spray Conference, Houston,
Mikhalchenko, A.A. and Solonenko, O.P., "Laser Texas, ASM International, 1995, pp 149-155.
Optical Diagnostics Complex For Investigation of 21. Leblanc, L., Gougeon, P. and Moreau, C,
High-Temperature Heterogeneous Jets", in "Plasma "Investigation of the Long-Term Stability of Plasma
Jets", Solonenko and Fedorchenko (Eds.), VSP, 1990, Spraying by Monitoring Characteristics of the Sprayed
pp 163-170. Particles", in "Proc. United Thermal Spray Conf,
9. Coulombe, S., Boulos, M.I. and Sakuta, T., Indianapolis, Indiana, ASM Inter., Sept. 1997, to be
"Simultaneous Particle Surface Temperature and published.
Velocity Measurements under Plasma Conditions", 22. Prystay, M., Gougeon, P. and Moreau, C, "Correlation
Meas. Sei, Technol., 6, 1995, pp. 383-390. Between Particle Temperature and Velocity and the
10. Kudora, S., Fujimori, H., Fukushima, T. and Kitahara, Structure of Plasma Sprayed Zirconia Coatings", in
S., "Measurement of Temperature and Velocity of "Thermal Spray: Practical Solutions for Engineering
Thermally Sprayed Particles using Thermal Radiation", Problems", C.C. Berndt (Ed.), Proc. National Thermal
Trans. Jpn. Welding Soc, 22, 2, 1991, pp. 10-17. Spray Conf, Cincinnati, Ohio, ASM Inter., 1996, pp
11. Frind, G., Goody, C.P. and Prescott, L.E., 517-523.
"Measurement of Particle Velocity in Two Low- 23. Lethtinen, T., Knuuttila, J., Vattulainen, J., Mäntylä, T.
Pressure Plasma Jets", in "6th Inter. Symposium on and Hernberg, R., "Correlations Between In-flight
Plasma Chemistry (ISPC-8)", Montreal, 1983, pp 120- Particle Concentrations and Coating Properties in
125. Atmospheric Plasma Spraying on Alumina", in
12. Mayr, W. and Henne, R., "Investigation of a VPS "Thermal Spray: Practical Solutions for Engineering
Burner with Laval Nozzle using an Automated Laser Problems", C.C. Berndt (Ed.), Proc. National Thermal
Doppler Measuring System", in "First Plasma-Technik Spray Conf, Cincinnati, Ohio, ASM Inter., 1996, pp
Symposium", Vol. 1, 1988, pp 87-97. 525-530.
13. Söelter, H.-J., Müller, U. and Lugscheider, E., "High- 24. Bengtsson, P., Jonannesson, T. and Wigren, J., "Crack
Speed Temperature Measurement for On-line Process Structures in Plasma Sprayed Thermal Barrier
Control and Quality Assurance during Plasma Coatings as a Function of Deposition Temperature", in
Spraying" Powder Metal. International, 24, 1992, pp. "Thermal Spraying - Current Status and Future
169-174. Trends", Proc. Inter. Thermal Spray Conf, Kobe,
14. Hencken, K.R., Tichenor, D.A. and Wang, C.F., Japan, A. Ohmori (Ed.), 1995, pp 347-351.
"Optical Double-Slit Particle Measuring System", U.S.
Patent 4,441,816,1984.
15. Swank, W.D., Fincke, J.R. and Haggard, D.C., "A
Particle Temperature Sensor for Monitoring and
Control of the Thermal Spray Process", in "Advances
in Thermal Spray Science and Technology", Proc. 8th
National Thermal Spray Conf, ASM International,
1995, pp. 111-116.
16. Li, K.-L, Vardelle, M. and Fauchais, P., "Comparison
Between Single and Double Flow Injectors in the
Plasma Spraying Process", in "Advances in Thermal
Spray Science and Technology", Proc. 8,h National
Thermal Spray Conf., ASM International, 1995, pp.
45-50. ;
17. Moreau, C, Gougeon, P., Lamontagne, M., Lacasse,
V. Vaudreuil, G. and Cielo, P., "On-line Control of the
Plasma Spraying Process by Monitoring the
Temperature, Velocity, and Trajectory of In-flight
Particles", in "Thermal Spray Industrial Applications",
C. C. Berndt and S. Sampath (Eds.), Proc. National
Thermal Spray Conference, Boston, ASM Intern.,
Boston, 1994, pp 431-437.
18. Gougeon, P., Moreau, C, Lacasse, V., Lamontagne,
M., Powell I., and Bewsher, A., "A New Sensor for
 5-1

ON THICK THERMAL BARRIERS FOR COMBUSTOR APPLICATION

T. Haubold
H. Gans
BMW Rolls-Royce GmbH
Material Technology
Hohemarkstr. 60-70
D-61440 Oberursel, Germany

D. Schwingel,
R. Taylor
University of Manchester
Materials Science Institute(UMIST)
Grosvenor Street
Manchester, M1 7HS, UK

SUMMARY
1. INTRODUCTION
Basic features for future advanced gas turbines
will be low emission rates and increased A basic feature for the development of advanced
efficiency. To fulfil these requirements a reduction gas turbines will be low emission rates of Nitrogen
of cooling air in the combustor chamber Oxides (NOx), Carbon monoxides (CO) and
demanding more efficient cooling technology Hydrocarbons (HC) to suit expected lower limits
and/or an increasing inner wall temperature is for pollution in the future. One way to fulfil these
needed. The development of thicker thermal requirements is a reduction of cooling air in the
barrier coatings for combustor tile applications is combustor chamber demanding more efficient
one concept followed to enable a reduction in cooling technology and/or increasing inner wall
cooling air. temperatures.

Thermal barrier coatings with a thickness up to Beside the development of more efficient cooling
2mm and improved thermal cycling life have been technologies there are several materials
developed as well as mechanical and technology concepts that are followed at BMW
thermophysical data determined. Rolls-Royce to reduce cooling air requirements in
the combustor:
Combustor segment rig testing demonstrated a i) a development of improved superalloys for
potential of 25% cooling air reduction by the use of combustor application with higher temperature
thick thermal barriers for the tile design capability,
investigated. ii) investigation of the potential of ceramic matrix
composite materials for combustor tiles and
iii) development of thicker thermal barrier coatings
(TBCs) for combustor tile applications.

This paper will focus on the latter.

Paper presented at an AGARD SMP Meeting on "Thermal Barrier Coatings'

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
5-2

Typical current applications of TBCs are thin has been performed at the BMW Rolls-Royce
coatings on guide vanes, blades and combustors APS spray facility (Sulzer Metco A3000S, 9MB
of gas turbines often introduced during overhaul in gun) using different spray parameters. During
order to extend lifetime of parts. Current spraying temperature has been measured by
technology is the use of plasma sprayed bond means of a pyrometer at the coating surface and
coat (BC) such as CoNiCrAlY mainly for adhesion by means of thermocouples at the backside of the
and oxidation protection and partially stabilised substrate. Substrates have been 2mm thick flat
ZrC>2 (about 8wt%Y2C>3) with a thickness below plates of Hastalloy X.
0.5mm. The benefit is a component temperature
reduction up to 150°C resulting in reduced cooling Microstructural investigations are performed at
airflow requirements and/or component durability BRR to determine typical coating features such as
improvement (1, 2, 3). Thickness and lifetime of thickness, porosity and crack density.
the barriers are limited by mechanical and thermal
stresses in the coatings due to the mismatch of Thermophysical and mechanical Properties
thermal expansion coefficients of the ceramic and
the metals and/or oxidation of the bond coat. Four point bend testing has been performed at
UMIST at room temperature and elevated
In a current Brite Euram project (BE7287)1 plasma temperature to determine the nominal strength,
sprayed thermal barrier coatings up to 2mm in the failure strain and Young's modulus. For testing
free standing TBCs have been used by dissolving
thickness have been developed to meet the
away the substrate. This has been possible due to
requirements of improved thermal insulation (4).
the thickness of the coating.
Different microstructures of coatings has been
addressed from different partners. Volvo Aero For the measurement at elevated temperatures, a
Corporation has investigated segmented TBCs high temperature rig has been designed in
(4), BMW Rolls-Royce 'conventional' highly porous accordance with European standards. All parts of
TBCs and ANSALDO Richerce has developed a the rig within the hot zone of the furnace are made
process route to spray highly porous TBCs using out of Alsint 99,7 which can theoretically withstand
a mixed ceramic-polymer powder (5). a temperature of 1700 BC. The capacity of the
furnace enables temperatures up to 1500 BC.
This paper will report on the development of the Figure 1 shows the design of the apparatus. The
highly porous coatings, the investigation of sample is supported by two rollers 20 mm apart.
thermophysical and mechanical properties and In order to minimise the torsional loading, one of
thermal cycling experiments as well as combustor these rollers can also rotate about an axis parallel
sector rig tests performed to determine the to the length of the sample. The specimen is
potential for cooling air reduction of thicker TBCs loaded by a second set of rollers which are
applied on combustor tiles. The sector rig test has isocentric to the first ones. They are 10 mm apart
been performed in the course of a national and both of them allow a rotation about their own
research programme. axis as well as about the axis of the sample.

2. EXPERIMENTAL The Young's modulus is strongly affected by the

ratio of the sample length I and its thickness h and
Spraying turns out to be too low in comparison with results
In general TBC development within the above from adequate tensile tests. In order to come to
an understanding of the influence of the l/h ratio of
mentioned Brite Euram Project has had the
the samples on the calculation of the Young's
objective to reduce Youngs modulus as one modulus, a coating was measured by an
method to reduce thermal mismatch stress. ultrasonic pulse method at 'Fraunhofer Institut
Mechanisms chosen have been segmentation Werkstoffphysik und Schichttechnologie' in
(vertical cracking) or production of highly porous Dresden resulting in a compensation factor for the
coatings. l/h ratio of the 4 point bend test specimen.
Spraying of porous 7%Y203 stabilised Zr02 TBCs

'Partners of the Brite Euram Project BE7287 are: BMW Rolls-

Royce GmbH, Volvo Aero Corporation , ANSALDO Richerce,
CTAS Air Liquide, Fraunhofer Gesellschaft Institut für
Betriebsfestigkeit, University of Lund, UMIST, Kvernes
Technology, IceTec, RWTH Aachen, Rolls-Royce plc.
 5-3

Thermal cycling tests have been performed at

the University Karlsruhe (IKM). The cycling time is
about 60 sec for heating up (TTBC = 1250°C, Tsubstrate
= 800°C) and cooling down to 60°C. Temperature
measurements are performed using a pyrometer
at the TBC surface and thermocouples on the
substrate. Specimen geometry is a flat plate of
10x50mm, substrate thickness has been 2mm.
Cycles are controlled by temperatures of the
substrate, i.e. heating is restarted at a substrate
temperature below 60°C. Maximum temperature
at the TBC of 1250°C is settled for different
specimens by varying distance from torch to
specimen. By this the same temperature gradient
Fig. 1: design of the four-point bending rig, the of TBC surface to substrate back side is
hot furnace area is marked by the grey background guaranteed and temperatures at the interface
bond coat to TBC will be comparable. Additionally
bending of the specimen can be measured using
For the measurement of the thermal diffusivity a, an inductive measurement device from the
the laser flash method was used. The samples backside of the specimen. Fig. 2 exhibit an
examined had a dimension of 10 mm * 10 mm example of the temperature cycle obtained and
with a thickness of about 2 mm. In this the bending of the specimen measured.
experiment, a heat flux through the sample is
caused by an energy pulse from a Nd glass laser
which is incident on the front face of the
oT-substrate[°q + T-sirface [°q 4 Bendng[rmi
specimen. The heat flux causes a temperature
rise at the sample rear side, which is monitored as
a function of time using an InSb detector. From 1500
this trace, the thermal diffusivity can be calculated.
For the samples investigated, the measurements 1200
of the thermal diffusivity were performed under
vacuum. The experiments were done in a
temperature range from 80 9C to 1200 SC(6).

Thermal expansion measurements were

performed in a conventional Al203 pushrod
dilatometer, over a temperature range from 30 aC
to 1000 9C. The heating rate was 60° h'1. After Time [sec]
reaching the maximum temperature, the sample
was cooled at -250 SC h"1. Since the push rod and
the guiding tube it is surrounded by also expand, Fig. 2: Thermal cycles performed at IKM
this effect had to be compensated mathematically.
The temperature gradient across the pushrod-
guiding tube arrangement is taken into account by Combustor Sector Rig Test
a compensation measurement. The latter one was
performed following the same heating and cooling In the course of a German national research
programme as for the actual measurement (6). programme combustion liner tiles with 7%Y203
The samples tested were about 15 mm long and 4 stabilised Zr02 were tested in the BR700 High
mm wide. Their thickness was about 2 mm, but Pressure 90°-sector rig. For the investigations on
varied slightly due to the sprayed coating liner cooling a rig combustor was made and
thickness not being constant. segments of outer and inner liner were cut out.
The free space allows mounting of liner panels
Thermal Cycling
with various cooling designs. Hence, it is possible
to test new cooling concepts using relative cheap
Thermal cycling experiments are one typical test
panel specimen in a representative annular
performed to rank TBCs for gas turbine
combustor. The objective of a test run is to
applications. Tests should simulate heating up and
demonstrate the thermal integrity and the
cooling down of the specimen to create failure
temperature distribution of specimen close to
such as spallation by thermal mismatch stresses
engine take off conditions as well as to evaluate
resulting from cycling.
the potential for cooling air reduction. After ignition
5-4

' "V"";
the conditions are raised to take off rating with an
inlet air pressure of 20 bar and inlet temperature '. ■■■>!

of 885K. The facility limits the inlet pressure to 20 '"• ■.■:.-'•-•.■:£ ■„" ,■■■ u **'•', V- ;
. ,#."■*.-> V:-
bar. ■ ■,''•■ , " * •tf:rc-*."i '•■,■," ' ' ■ *';'".'■'.' i» '.'.% ■''•■ ~'T
•., ■ \ ' "Jx1*f .»■*''•- • .'' ■ ■•' •"' ■.-.'■'.-■ ; ,
w;> . •*■•»',■'.■ m-J ,-.v, >'■* ..*L •*.■•-..•*•»'■

3. RESULTS ■ .v ■-- . "I • -*.*---:*-'■#/ - ,-v **■ * „ .■$*«

■ - '• ■■ ?. - ■'■>/ >■■■>. »■■ /?S*' 'c >""" :■: *■'.-/■"<*
■' «^' *'« -■= ' ■*■<''■'" *# • " * . -\ - '•■■ •:

Spraying '■•■■■'•-.>:■ v ■-.-v.•:*"■/?> •*...." ' •"'- ~„J1*-« \

Several parameter settings have been

investigated to spray highly porous coatings.
Typical temperature spectra measured during
spraying are shown in Fig. 3 a and b.

Porosity values obtained for the different specimen Fig. 4: micrograph of TOP 7,15%porosity, x 150
are in the range of 8.5 to 20% (see fig. 4). For a
number of specimen segmentation cracking has
been found , which has been related to porosity Thermophvsical and mechanical properties
values in the range of 8.5 -12.5%.
Since the structure of the coatings is expected to
vary through thickness two different loading types
have been considered; where the top is either in
the state of compression or of tension.
TOP 8 surface temperature

Mechanical properties determined are listed in

table 1. Numbers given there are mean values out
of the two loading types. Values from the
experiments, where the surface of the TBC is
under compression were found to be slightly
higher in most cases. Youngs moduli found are
lower than typical values reported on plasma
sprayed TBCs, which lies in the range of 10 - 50
GPa (see for example ref. 8, 9, 10).
i- CM CM CM m m n
channel number (time)

Specimen j Youngs | failure ] failure

I modulus I strength I strain
bl ![GPa] : [MPa] ;[%]
TOP 8 substrate temperature
Top 7 I 5.7 115.9 I 0.29
Top 8 I 6.9 113.6 I 0.22
200
Top 11 I 3.3 I 7.4 ;0.26
I

k
Top 12 I8.8 j 30.4 iO.37
150
Top 13 I 7.8 125.1 I 0.35

100
Table 1: Mechanical properties of thick TBCs

SO No significant change in these properties was

J ^— observed in the experiments carried out up to
1000°C.
fl T". r°^™^r?v TT^T:

The failure strength and stress is found to be

correlated to the Youngs Modulus determined
Fig. 3: Temperature readings during spraying from (Fig. 5).
the coating surface (a) and substrate backside (b).
 5-5

For all specimen measured the thermal diffusivity

obtained on cooling was higher than that observed
35 Mean values
during heating, showing that some irreversible
out of both the
■£■ 30 loading types •/ change had occurred during first cycle only. The
values obtained ranged from 0.0029 cm2s"1(Top 7)
ID
0.
§. 25 •/ up to 0.0057 cm2s (Top 8). There is a general
(A
(A
S 20 trend for the diffusivity to decrease with increasing
0)
porosity (Fig. 6a), but the effect is more
15
1 / • pronounced when diffusivity is plotted against
n
"■ 10 horizontal crack density (Fig. 6b).
•/
5 From the diffusivity a, the thermal conductivity X
0- 1 1 1 \- 1 can be calculated as:
0 2000 4000 6000 8000 10000 12000
Young's modulus [MPa] A, = a • p • cp

Fig. 5: Failure stress versus Youngs modulus with p and cp being the density and the specific
heat capacity respectively (6). Values ranged from
The expansion coefficient of the samples tested 0.8 W/mK to 1.5 W/mK for the TBCs investigated
seems to have a rather constant value of 10.7*10" here.
6
K'1, which is typically for plasma sprayed 7%Y203
stabilised Zr02 (8, 11). Obviously, the porosity has
rather little influence. Over the whole temperature Thermal cycling
range, the thermal expansion coefficient is almost
constant and differences during heating and Thermalshock test at BRR (perrformed at
cooling, a typical indicator of phase changes, University Karlsruhe, IKM) used a cycling time of
cannot be determined. 60 sec for heating up (TTBC = 1250°C, Tsubstrate =
800°C) and cooling down to room temperature
a) and has been performed for several specimen up
0.006 to 2000 cycles yet in comparison to a cycling life
below 10 cycles for the first tests. Some specimen
tested to 2000 cycles did not fail at all. Fig 7 was
an overview about number of cycles obtained.

....T.!?.?.r.!53)..!?y!*!?.5.

0.001 • 650 9C on heating

O650 eC on cooling
0.000 \
10 15 20
Porosity [%]
TOP1 TOP2 TOP3 TOP4 TOP7 TOP8 TOP11 TOP12 TOP13
b)
0.006

y= -3.64E-05X+ 5.44E-03 Fig 7: mean number of cycles reached for several

0.005 •-
' Ö- -.. o specimen
0.004
The main influence on thermal cycling life is
» 0.003 - determined to be the temperature during spraying.
A correlation from porosity to thermal cycling life
■= 0.002 + could not be found (Fig. 8 an 9). Specimens
0.001 +
• 650 2C on heating exhibiting segmentation cracks instead of high
O650 SC on cooling porosity tend to show higher life times in the tests
0.000 performed.
10. 20 30 40 50
Horizontal crack density [mm1]

Fig. 6: thermal diffusivity versus porosity (a) and

horizontal crack density (b)
5-6

specimen with a thick TBC was exposed to slightly

25 i
more severe conditions.
20 1 1
The coating surface of the specimen with the
& 15 i 1■ ■ 0.4mm TBC thickness shows only small spots with
(A ■ ■
s temperatures above the 1160°C limit of the
g. 10 ■
■ thermal paint, whereas the 1,35mm thick TBC
5 - specimen has approximately 50% of the surface
hotter than 1160°C. It should be noted however
' ' that the amount of cooling air has been reduced
0 500 1000 1500 2000 2500
TC life
for the thick TBC specimen.

The metal temperature on the cold side of the

Fig. 8: porosity versus thermal cycling life 0,4mm coated tile was around 830°C with peaks
at approximately 900°C. The thick insulated tile
had approx. 50°C lower temperatures on the inner
side.
2500 T

2000 A cooling film on the hot side is still necessary for

£ 1500 thick coated tiles to keep the maximum
O temperature of the top coating below 1250°C. With
I- 1000
the current design temperatures of top - and bond
500 coating a thickness of the ceramic layer of approx.
1,3 mm should provide the desired temperature
50 100 150 200 250 300 350 400 drop of 350 to 400°C between the hot surface and
T Spraying the critical layer where the ceramic adheres to the
bond coating.
Fig. 9: Spray temperature versus thermal cycling
life With a thick coated tile approximately 25% of the
cooling air required for a standard coated tile
could be saved.
FEM simulations have been performed using
thermophysical and mechanical data described
above to compare calculated and measured 4. CONCLUSIONS
bending of the specimen during cycling. The
bending behaviour (see Fig. 2) could be described TBCs with a thickness up to 2mm and improved
qualitatively, quantitative deviations are expected thermal cycling life have been developed as well
to be related mainly to the quality of temperature as mechanical and thermophysical data
data recorded during cycling. determined.

Thermal cycling life is mainly influenced by spray

Combustor Sector Rig Test temperature. Porosity has not found to be of high
importance, but segmentation seems to be
Two panels with different coating thicknesses beneficial in the test performed.
(0.40 mm and 1.35 mm TBC thickness) were
tested. The panel have been cooled by an array of Youngs Moduli in the range of 3 - 9 GPA has
impingement holes on the carrier. At the end of been determined. Failure strengths range from 7
each tile the cooling air is used as a film for the hot to 30 MPA, failure strains from 0.2 - 0.4%.
side of the following tile. The tiles has been
painted with thermal paint to investigate Successful simulation of the bending of the
temperatures and temperature distributions. For specimen during cycling indicates that the
the tiles with thick TBCs on the amount of cooling materials data obtained are of reasonable quality.
air used was reduced compared to the thin TBC
tiles. Combustor segment rig testing demonstrated a
potential of 25% cooling air reduction for the tile
The specimens were not tested under identical design investigated.
conditions. The fuel to air ratio of the test with thin
TBCs was approx. 5% lower than for the test with
the thick coated specimen. This means that the
 5-7

ACKNOWLEDGEMENT

The authors thank all partners of the Brite Euram

project BR7287, especially J. Wigren (VOLVO
Aero Corporation) and Dr. C. Gualco (ANSALDO)
for helpful discussions. The support from Dr. M.
Alaya ( University of Karslruhe, IKM) during
thermal cycling experiment is gratefully
acknowledged.

Part of the work has been performed within the

Brite Euram Project BE 7287 funded by the CEC
under the contract BRE2-CT94-0936.

The combustor sector rig testing has been funded

by the German BMBF under HTGT Turbotherm
contract 0326760B.

REFERENCES

1. R. A. Miller, Journal of Thermal Spray

Technology, Vol 6 (1), 35-42 (1997)
2. A. Bennett, Materials Science and Technology
2,257(1986)
3. F.C. Toriz, A.B. Thakker, S.K. Gupta, ASME
88-GT-279, 1988
4. J. Wigren, J. Dahlin, M-.O. Hansson, submitted
to ASME'98
5. C. Gualco, A. Campora, S. Corcoruto, R.
Taylor, D. Schwingel, S. Oswald, Proceedings
oftheUTSC1997.
6. D. Schwingel, C. Persson, R. Taylor, T.
Johannesson, J. Wigren, High Temperature
High Pressure 1995/1966, Vol 27/28, p. 273-
281.
7. M. Alaya, PhD Thesis 1997.
8. R. McPherson, Surface and Coatings.
Technology, 39/40 (1989) 173-181
9. W. Mannsmann, PhD Thesis 1995.
10.T.A. Cruise, B.P. Johnson and A. Nagy,
Journal of Thermal Spray Technology, Vol 6
(1), 57-66(1997).
11.L. Pawlowski and P. Fauchais, Int. Materials
Review Vol. 37 No. 6 (1992) p. 271-289.
 6-1

Advanced Processing of TBC's for Reduced Thermal Conductivity

J. R. Nicholls", .K. J. Lawson", D. S. Rickerby and P.Morrell

x
Cranfield University, Cranfield, Bedford UK
Rolls Royce pic, Derby, UK

ABSTRACT lp is the phonon mean free path.

a is Stefan Boltzmann's constant = 5.67 x 10"8
Zirconia-8wt% yttria thermal barrier coatings (TBC's) W/(m2.K),
provide the potential to increase the operating T is the thermodynamic temperature (K),
temperatures within the modern aeroengine gas turbine. n is the refractive index of the material,
Of particular interest are the columnar, strain tolerant L is the photon mean free path
thermal barriers produced by electron beam physical lj is the intrinsic mean free path of phonons due to the
vapour deposition which can be applied to rotating lattice and differing ion masses,
components and can reduce the metal surface lvac is the mean free path of phonons due to vacancies in
temperature by up to 150°C. Measured thermal the lattice,
conductivities for this columnar microstructure are lgb is the mean free path of phonons due to grain
typically 1.8-2.0 W/m.K. Should it be possible to reduce boundaries,
this value to that observed for plasma sprayed ceramics Strain is the mean path of phonons due to local strain
(0.9-1.0 W/m.K) then further benefits from the use of centres within the lattice.
such a ceramic layer could ensue including either a
further increase in operating temperature or a reduction
in cooling air requirements.

This paper examines the advanced processing of TBC's

with the aim of reducing thermal conductivity. The role 1 .INTRODUCTION
of coating thickness is reviewed, where it is shown that
for thin coatings, circa lOOum thermal conductivities of The drive to improve engine performance and fuel
1.0 W/m.K can be achieved. This is thought to be due efficiency, while reducing emissions, has meant that the
to increased scattering of thermal waves close to the operating temperatures of the turbine section of aero-
early confused growth region. The possibility of further engines has increased significantly over the last 20
reducing thermal conductivity by using layered years. This has been achieved by novel material design,
structures is examined. These structures have been improved cooling technologies and better manufacturing
manufactured by control of rotation, controlled dwell methods [1]. The latest development in this drive for
and through the use of plasma assistance during the improved performance is to apply thermal barrier
deposition process. The benefits of such layers in coatings to turbine aerofoil surfaces. Thermal barrier
scattering thermal waves is discussed. coatings offer the potential of increasing turbine
operating temperatures by up to 150°C, without any
LIST OF SYMBOLS increase in metal operating temperatures (Figure 1), or
alternatively permits a reduction in the mass of cooling
x total coating thickness air required while maintaining the turbines operating
x, thickness of inner "nucleate" zone temperature giving improved specific fuel consumption.
k thermal conductivity for the complete TBC.
k, thermal conductivity of the inner "nucleate" zone Thermal barrier coatings (TBC's) have been used in the
k2 thermal conductivity of the outer columnar structure gas turbine engine since the 1970's [2-4] to lower mean
kp is the phonon thermal conductivity, metal temperatures. These coatings, based on partially
kr is the contribution to thermal conductivity due to stabilised zirconias, are deposited using thermal spray
radiation, processing and have performed well in service,
Cv is the specific heat capacity at constant volume, extending the lives of combustion chambers and annular
p is the density of the material, platforms of high pressure nozzle guide vanes within the
vis the mean velocity of phonons in the material turbine section of the engine [5]. However, the use of
(speed of sound) thermal sprayed TBC systems have not generally been

Paper presented at an AGARD SMP Meeting on "Thermal Barrier Coatings",

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
6-2

Hot Gas systems by alloying with ceria and yttria [9]. This paper
Temperature discusses the theoretical models of heat transfer in solids
with regard to the reduction of the thermal conductivity,
Temperature Profile
without TBC and reviews these against the measured values on TBC
coatings engineered for low conductivity.

Table 1 Properties of TBC's at Room Temperature

Property/Characteristic EB- Plasma

PVD Sprayed
Thermal Conductivity (W/mK) 1.5 0.8
Temperature Profile Surface Roughness (um) 1.0 10.0
with TBC Adhesive Strength (MPa) 400 20-40
Coolant Young's Modulus (GPa) 90 200
Temperature
Erosion Rate (Normalised to 1 7
PVD)
Ceramic Metal Wall
Figure 1 Schematic of thermal barrier coating systems showing
thermal gradient subtended across the ceramic layer
2. THEORY OF CONDUCTION
extended to high-pressure turbine aerofoils, due to their
poor surface finish, high heat transfer coefficient, low The theory of thermal conductivity in solids is fully
erosion resistance and poor mechanical compliance. described elsewhere and will only be briefly reviewed
Success in coating turbine aerofoils has been achieved here [10-12]. In crystalline solids heat is transferred by
by adopting electron beam physical vapour deposition three mechanisms, (i) electrons, (ii) lattice vibrations and
(EB-PVD) technology to coat these parts. Thermal (iii) radiation. As zirconia and its alloys are electronic
barrier coatings deposited by EB-PVD processes have a insulators (electrical conductivity occurring at high-
good surface finish [5], columnar microstructures with temperatures by oxygen ion diffusion), electrons play no
high strain compliance [5,6] and good resistance to part in the total thermal conductivity of the system.
erosion [5,7] and foreign object damage [7]. EB-PVD
thermal barrier coatings have been used in production The contribution to thermal conductivity from lattice
vibrations (the quanta of which are known as phonons)
since 1989 [8].
is given by:
Unfortunately, the microstructure which gives the EB-
PVD TBC system its high strain compliance and good k
P=^\CvPvlp [1]
erosion resistance leads to a relative high thermal
conductivity when compared to thermally sprayed
coatings (Figure 2): Table 1 compares the properties Whilst equation [1] describes the phonon conductivity,
and performance of commercially produced thermally the total thermal conductivity of the material is higher
sprayed and EB-PVD deposited PTSZ thermal barrier than this figure due to the contribution to the thermal
coatings. Previous work in the development of ceramic conductivity by radiation. Zirconia is susceptible to
alloy compositions indicated that there is opportunity to radiation transport (photon transport) at high-
reduce the thermal conductivity of the zirconia-based temperatures (Figure 3). The contribution to the total
thermal conductivity due to radiation can be written;

k.r =—an
3
T L [2]

Therefore the total theoretical thermal conductivity of

zirconia-based systems is merely the sum of the two
terms:

Bulk EB-PVD Plasma ^D ™Y' [3]

Zirconia TBC Sprayed
TBC

Figure 2 Thermal conductivities of bulk, EB-PVD and thermally

sprayed zirconia-yttria materials at room temperature.
 6-3

'—'S.
\
.
1 0.1
«c
S
H

0.01
0.1 1 10 100
5 wt% Y203 20 wt% Y203
Wavelength (micrometers)

Figure 3 Emittance and transmittance of zirconia as a function Figure 4 Theoretical conductivity of EB-PVD zirconia ceramics
of wavelength. as a function of yttria content.

If the mean free path for radiative transport is assumed

to be the thickness of the coating then, for an application 1
where the surface of the TBC is operating at 1473K and +J- + [4]
/„ /.• +
/„. /„
the interface operates at 1323K the contribution to total
thermal conductivity due to radiation could be as high as
1 W/m.K. In this case the heat flux which could arise Of these the grain boundary term has the least effect on
through radiation is of the order of 80-90 kW/m2 out of the phonon mean free path, whilst the intrinsic lattice
typical total values of 800 kW/m2 seen in high heat flux structure and strain fields have the most significant
regions in engines. Clearly the mean free path of effect. For example, the addition of NiO has a marked
photons is not the thickness of the coating and therefore effect on the intrinsic mean free path in MgO, reducing
radiation is attenuated within the TBC structure reducing the mean free path by a factor of 10 [14]. However, the
the contribution to the total conductivity from the value addition of alloying oxides has the largest effect on
listed above. However, as the operating temperature simple systems, the zirconia-yttria crystal system is more
increases the contribution to thermal conductivity from complex than MgO and the effect of alloying additions
radiation becomes more important. is less. For zirconia based systems, workers in the field
have demonstrated that increasing the level of yttria in
Therefore to lower the intrinsic thermal conductivity of a the alloy, decreases the thermal conductivity [15,16]
system, reductions in the specific heat capacity, phonon which is commensurate with shorter intrinsic mean free
velocity and mean free path, density or refractive index paths with increasing yttria content (Figure 4).
are needed. The specific heat capacity at constant
volume for any system is constant at a value of 3.kbN « Further to the reduction in the intrinsic mean free path
25 J/(K.mol) [13] above the Debye temperature (380K due to alloying, local strain fields and vacancies can be
[13] for zirconia). Therefore to engineer a lower thermal generated in the lattice. These strain fields and vacancies
conductivity in zirconia-based ceramics, the only options act to scatter phonons directly increasing phonon
are to lower the mean free paths of the heat carriers, to dispersion in the lattice. With increased dispersion, there
lower the velocity of the heat carriers, or to lower the is a high probability that phonon-phonon interactions
density of the material. will occur whereby the mean free path will be further
reduced. The benefit of a 20-30% lower thermal
In real crystal structures scattering of phonons occurs conductivity by adding divalent transition metal dopants,
when they interact with lattice imperfections in the ideal as reported by Tamarin et al [17] is attributed to this
lattice. Such imperfections include vacancies, effect and in part to the dopants ability to attenuate
dislocations, grain boundaries, atoms of different masses radiation within the ceramic layer.
and other phonons. Ions and atoms of differing ionic
radius also scatter phonons by locally distorting the bond For thermally sprayed TBC's the typical grain size is of
length and thus introducing elastic strain fields into the the order of 1 um and for EB-PVD TBC's the grain size
lattice. The effects such imperfections cause can be varies from c.a. «1 (am at the interface whilst the
quantified through their influence on the phonon mean column length is 100-250 urn in the bulk of the coating.
free path. This approach has been used by several These dimensions have little effect on the mean free path
workers, for which the phonon mean free path (lp) is of phonons. However, these dimensions can have an
defined by:- effect on the radiative transport in the ceramic, where a
reduction in the mean free path for photons from 250
um to 1 um would cause a reduction in the radiation
contribution to total thermal conductivity from 1 to
6-4

100 200 300 400 500

Thickness (wm)
Figure 6 Effect of coating thickness on thermal conductivity of
EB-PVD at room temperature.

PYSZ ceramic and bulk zirconia are essentially

insensitive to coating thickness this is not the case for
EB-PVD ceramics [19]. This behaviour is illustrated in
Figure 6.

The explanation for this increase in thermal conductivity

with thickness relates to the unique columnar
microstructure which characterise PVD ceramics when
grown from the vapour phase. During deposition,
competitive growth occurs, as illustrated in Figures 7
and 8. This results in a decrease in the number of
oblique columnar boundaries in the outer part of the
coating as grains with "less favoured" growth directions
Figure 5 photomicrographs of thermally sprayed (top) and EB- are eliminated from the coating structure. The number
PVD TBC's (bottom), showing the differences in microstructure.

0.005 W/m.K from equation [2].

In real coating systems grain boundaries are associated

with air-gaps (inter-splat boundary porosity and inter-
column porosity in thermally sprayed and EB-PVD
TBC's respectively - Figure 5) and these have a
significant effect on the thermal conductivity. The
alignment of the inter-splat boundaries in the case of the
thermally sprayed coatings has by far the larger effect
reducing the conductivity of zirconia-yttria TBC
materials from theoretical values of 2.2-2.6 W/m.K to
values in the range 0.7-0.9 W/.m.K (refer to Table 1).
This behaviour is well documented and is due to the
reduction of mean free paths of photons (radiative heat
transfer) by scattering at the splat boundaries, but more
importantly, by the reduction in the composite thermal
conductivity due to the air-gaps [18].

3. THERMAL CONDUCTIVITY IN EB-PVD TBC's

Whilst the thermal conductivity of air plasma sprayed Figure 7 Differences in the column density for the nucleate
zone (top) and the steady state region (bottom)
 6-5

Measured values, [W/m.K]

k=1.0W/m.K k=2.2 W/m.K
Figure 9 Comparison between predicted and measured thermal
Figure 8 Schematic of an EB-PVD TBC and the associated conductivities of EB-PVD TBC's of varying thickness.
variation of thermal conductivity for the two-layer model

and distribution of columnar boundaries is responsible process used to produce the thermal barrier coating
for the reduction in thermal conductivity. More strongly affects the coatings resultant thermal
boundaries mean additional centres capable of scattering conductivity. Lamellar interfaces, parallel to the
thermal waves. Thus the thermal conductivity of an EB- coating/substrate interface as produced in plasma
PVD TBC can be modelled successfully using a two- sprayed ceramic, act to strongly scatter thermal waves
layer model, shown schematically in Figure 8 [19]. The reducing the coatings thermal conductivity. This
inner "nucleate" zone is dominated by defect/grain behaviour is observed to a lesser extent in the inner
boundary scattering and this results in a lower growth zone of EB-PVD deposited coatings due to the
conductivity of around 1.0 W/m.K. With increasing inclined nature of many of these inner boundaries.
thickness, the structure is characterised by a dominant Defect scattering must also play a role in this early
crystallographic texture and the thermal conductivity deposition zone of EB-PVD ceramics. However, due to
increases as the outer part of the coating becomes more the competitive nature of coating growth for vapour
perfect. In this outer zone the thermal conductivity deposited coatings a dominant growth texture is
approaches that of bulk zirconia at circa 2.2 W/m.K. developed. This vertically aligned microstructure of the
The overall coating thermal conductivity results from the outer layer of an EB-PVD coating results in thermal
combined effect of these two layers and can be predicted conductivities close to that for the bulk ceramic.
using a rule of mixtures. For coating thicknesses (x) less
than Xj (lOOum for the schematic illustrated in Figure 8) These observations would suggest that to further reduce
then the thermal conductivity of EB-PVD TBC's one should
introduce layers and/or additional phonon scattering
k = k, [5a] centres into the coating.

while for coating thickness (x) greater than x, 4. REDUCTION OF THERMAL CONDUCTIVITY IN
EB PVD COATINGS BY LAYERING
x
k_ \k\ (x~xi) j.
+
[5b] The former section discussing heat transfer into and
through the ceramic has highlighted a number of
Figure 9 provides a comparison between predicted and methods capable of reducing the thermal conductivity of
measured thermal conductivities for EB-PVD zirconia the zirconia ceramic. To reduce radiative transport
thermal barrier coatings, illustrating the good agreement 'colouring' and 'layering' are possibilities, while
to experiment of this two layer thermal conductivity introducing 'atom dimension defects', 'local changes in
model. ceramic density' and 'local variation in ceramic elastic
modulus' may be effective in reducing phonon
It should be expected that the thermal conductivity of conduction.
chemical vapour deposited (CVD) ceramic should show
a similar thickness dependence to EB-PVD zirconia as As indicated earlier the plasma sprayed TBC structure is
both coating morphologies are grown from the vapour such that the contribution to total thermal conductivity
phase. A study on the thermal conductivity of CVD by radiation is minimised by the lamellar splat
deposited synthetic diamond confirms this hypothesis boundaries at a typical spacing of 1 um. If this
[19,20]. attenuation of the radiative transport mechanism could
be applied to EB-PVD it could make a significant
From the foregoing, it is evident that the manufacturing contribution to the reduction of the thermal conductivity,
6-6

especially at high temperatures. Thus layering of the resistance. The concept is the subject of a European
ceramic appears a most promising route to lower the patent [22].
ceramics thermal conductivity, especially if it can be
coupled with charging the local density or elastic This paper focuses on the use of ion bombardment to
modulus of the material. A suitable periodicity of the create layers in the EB-PVD structure with varying
layers should be 0.2-2.0um (between X and Xf4 for an density.
incident radiation of 0.5-5.0 urn), which is able to scatter
photons and thus reduce radiative transport. This 4.1 Layering by Plasma Assisted Deposition
concept of layering within each column of an EB-PVD
coating is illustrated schematically in Figure 10. This method uses a glow discharge plasma to vary the
density of the ceramic during deposition. Figure 11
To reduce phonon conduction layer periods would have illustrates a typical micrograph of the morphological
to be of near atomic dimensions (0.3-2.0nm) and this is changes that can be introduced.
not considered feasible, although it should be possible to
locally change either the elastic modulus or density of The layers were produced by switching the D.C. bias
the ceramic by introducing atomic level defects into the applied to the substrate between high and low levels
lattice. during deposition. This has the effect of periodically
changing the degree of ion bombardment and thus
The PVD process offers a number of ways to tailor altering the density of the layers produced. The
coating microstructures and thereby introduce multi- micrograph illustrated in Figure 11, was produced by
layers, since both the column angle and morphology switching between a D.C. bias with a peak voltage of
may be influenced by variations in the flux density and 2500V and ground. The degree of bombardment that is
the direction this flux subtends to the substrate surface attained depends on the applied bias and local current
[21]. • density, current densities typically greater than
O.lmA/cm2 are required to achieve a significant density
Three alternative EB-PVD processing routes could be change [22].
used to generate layering within the columns.
The measured thermal conductivity for this
• Route 1 varying the speed of rotation of the parts microstructure is significantly lower than that for a
during coating coating produced without ion bombardments (Figure
12). Reductions of the order of 37-45% compared to
• Route 2 rocking the substrate between two fixed tilt state-of-the-art EB-PVD TBC's have been measured for
angles. these layered structures that are approaching the values
for thermally sprayed TBC systems.
• Route 3 varying the degree of ion bombardment to
change the density of the ceramic.

The principle behind each of these routes is to introduce

interfaces into each of the column structures which are
essentially parallel to the ceramic/bond coat interface,
but without disrupting the overall columnar structure
that provides the coating with its good strain tolerance
and erosion resistance. Thus, these layered structures
should offer the best of the properties currently available
from commercial EB-PVD coatings and thermally
sprayed ceramic coatings, namely low thermal
conductivity, good strain tolerance and good erosion

Figure 11 Layered structures introduced by PAPVD processing

of the ceramic. Top general view, bottom detail of a boundary
region.

Figure 1D Schematic of layering within an EB-PVD column to

reduce radiative heat transfer within the coating.
 6-7

1.2
1.00
I3 1
T3
g 0.8
o 0.67
| 0.6 ■ 0.55 ^B 0 53
o
jE 0.4
a
I 0.2
0
EB-PVD Coloured PAPVD PAPVD Plasma
TBC EB-PVD Layered Layered Sprayed
TBC TBCver.1 TBCver.2 TBC

Figure 12 Relative thermal conductivity of modified EB-PVD

TBC systems compared to state-of-the-art EB-PVD and
thermally sprayed TBC's. Figure 14 An example of a tailored EB-PVD layered structure.

Clearly, the combination of layering at micron followed by a region with bias and high ion current
dimensions and the introduction of density change from density to produce maximum density changes between
layer to layer work in combination to significantly layers. This was followed by a region of low current
reduce the thermal conduction of the coating. As density giving low density differences between layers
discussed earlier the layering periodicity has been and finally a region of high current density giving a
selected to significantly reduce photon transport, while denser structure. (Figure 14).
local changes in layer density act to scatter phonons and
thus reduce thermal conduction by lattice vibrations. 4.2 Addition of dopants to reduce thermal conductivity

Using this process route it has been possible to introduce Tamarin and co-workers [17] have examined the
layers into individual columns, without change the addition of divalent transition metal oxides to reduce the
overall vertical columnar microstructure required for thermal conductivity of zirconia-yttria TBC's. As eluded
good strain, tolerance and good erosion resistance. For to earlier this has two effects, firstly to reduce the
example, it can be seen from Figure 13 that layering phonon transport in the material and secondly a
produces no significant deterioration to the erosion reduction in the radiative transport mechanism. The use
performance of an EB-PVD thermal barrier coating. of a divalent transition metal oxide introduces vacancies
as well as strain centres into the lattice both of which
Also it is possible by programming the sequencing of will reduce the phonon mean free path. In addition this
bias application to tailor the position of the layers in the doping changes the colour of the TBC material to dark
coating. This potentially has benefits in allowing the grey, reducing radiation transport in the visible range
properties of the coating to be varied from the surface to and by inference in the near infra-red.
the interface. For example more erosion resistant
structures could be generated at the outer surface of the Following this concept of dopant additions, further
coating with the application of bias in this region, whilst reduction in phonon thermal conductivity could be
the inner zone of the coating is layered to lower the achieved by the selection of low levels of dopant which
thermal conductivity. To demonstrate this a coating was maximise lattice strain and lattice anharmonicity. This is
produced where the early deposition has no bias applied, best achieved by the addition of transition metal ions
with high mass and large ionic radii.
7.00

£ 6 5. SUMMARY OF REMARKS AND CONCLUSIONS

(0
^5
o This paper has examined the factors that control the
S4 thermal conductivity of thermal barrier coatings and has
w
S3 highlighted the significance of three factors capable of
lowering the coating thermal conductivity.
1.00 0.98

E9 IM 1) Colouring of the coating by addition of dopant

materials can be used to increase the opaqueness to
EB-PVD TBC Layered EB-PVD Plasma Sprayed TBC
TBC infra-red radiation and therefore radiative transport
Figure 13 Erosion performance of varuious modified EB-PVD
through the coating.
coatings.
6-8

2) Atomic level defects, such as atoms of differing

mass, vacancies etc. can be introduced by adding Efladi of
•llaytigand
■trudjfe
dopants or by ion bombardment. These are effective in
reducing phonon conduction, i.e. heat transfer by lattice
vibrations (Figure 15). A recent patent by Strangman
[23] describes a TBC system with stabilised porosity
wherein the pores are coated with a nobel metal (Figure
16). This stabilised poriosity along with the nobel metal
coating is claimed to reduce the thermal conductivity 400 600 800 1000
and improve its sintering behaviour at high Temperature (-C)
temperatures.
Figure 16 Summary of the reductions achievable in thermal
conductivity in zirconia thermal barriers as a result of alloying
3) Layering, can be used to introduce interfaces parallel and layering the structure.
to the ceramic bond coat interface and is effective in
reducing both radiative and phonon transport. The layer
following organisations during work on the development
periodicity when set at levels between 0.5 and 2.0um is
of advanced thermal barrier coating systems: Rolls
effective in reducing photon transport, when coupled
Royce pic, DERA, CEC and the EPSRC.
with a significant change in layer density from one layer
to another. These local changes to the ceramic density
are also believed to be effective in reducing phonon
7. REFERENCES
conduction. Thus layering has been shown to be a most
effective method of modifying the thermal conductivity
1. Brandon J. R., Taylor R. and Morrell P.
of EB-PVD thermal barriers (Figure 16).
"Microstructure, composition property,
relationships in plasma-sprayed thermal barrier
By using a switched D.C. bias, applied to the substrate it
coatings", Surface and Coatings Technology, 50,
is possible to introduce these layers during the
(1992) 141.
deposition processes. The degree of ion bombardment
2. Liebert C. H. et al, "Durability of zirconia thermal
controls the density changes achieved, thus allowing
barrier coatings on air cooled turbine blades in
tailoring of the ceramic microstructure to engineer the
cyclic jet engine operation", NASA TMX-3410
most appropriate mechanical and thermal properties in
(1976).
the coating for the operational requirements placed on it.
3. Grisaffe S. J., "Thermal barrier coatings" NASA
TMX-78848 (1978).
4. Driver D., Hall D. W. and Meetham G. W., in "The
development of the gas turbine engines", Applied
Science Publishers (London), 1981.
6. ACKNOWLEDGEMENTS
5. Rickerby D. S. and Morrell P., "Design and
development of EB-PVD thermal barrier coatings
The authors wish to acknowledge the sponsorship of the
for gas turbines" in "High Temperature
Engineering" Institute of Materials, London UK
.» .» .a? ,/»
(1997) to be published.
6. Strangman T. E. "Tailoring zirconia coatings for
performance in a marine gas turbine environment"
Journal of Engineering for Gas Turbines and
Power, Vol. 112.
7. Nicholls J.R., Yaslief Y. and Rickerby D. S.,
"Erosion and foreign object damage of thermal
barrier coatings" in 4 Int. Symp. on High
Temperature Corrosion, Les Embiez, France, May
1996
8. Meier S. M. and Gupta D. K., "The evolution of
thermal barrier coatings in gas turbine engines
applications", Trans. ASME, 116, (1994), 250.
9. Ragaswami S. European Patent, EP0166 097
(1988)
10. Klemens P. G., "Thermal conductivity of solids",
Figure 15 Schematic of EB-PVD TBC systems with stabilised
porosity after Strangman
Ed. R. P. Tye, Volume 1, Publ. Academic Press
London (1969).
11. Parrott J. E. and Stuckes A. D. "Thermal
 6-9

conductivity in solids", Publ. Pion Limited, (1975).

12. Peterson R. B., "Direct simulation of phonon
mediated heat transfer in a Debye crystal", Trans
ASME J. Heat Transfer JJ6, 815, (1994).
13. Slack G.A.,"Thermal conductivity of nonmetallic
solids", in Solid State Physics; Advances in
research Applications, 34, ppl-74, Academic Press
(1979)
14. Kingery W.D., "Introduction to Ceramics", 2nd
Edition, J.Wiley (1976).
15. Morrell P. and Taylor R., "Thermal diffusivity of
thermal barrier coatings or Zr02 with Y203",
High Temperatures-High Pressures, 17, (1985) 79.
16. Klemens P. and Gell M., "Thermal conductivity of
TBC's", in the proceeding of the TBC Workshop,
Cincinnati Ohio, May (1997).
17. Tamarin Y. A., Kachanov E. B. and Zherzdev S.
V., "Thermophysical properties of ceramic layers
in EB-TBC" in 4 Int. Symp. on High Temperature
Corrosion, Les Embiez, France, May 1996
18. Hassleman D.P.H and Singh J.P, "Effects of cracks
on thermal conductivity", J. Compos. Mater., _12,
403, (1978).
19. Lawson K. J., Nicholls J. R. and Rickerby D. S.,
"The effect of coating thickness on the thermal
conductivity of CVD and PVD coatings", 4th Int.
Conf. on "Advanced in Surface Engineering",
Newcastle, UK, 1996.
20. Graebner J. E., Jin S., Kammlott G. W., Bacon B.,
Seibles L. and Banholzer W., J. Appl. Phys. 7_1,
(1992) 5353.
21. Lawson K. J., Nicholls J. R. and Rickerby D. S,
"Thermal conductivity and ceramic
microstructure" in High Temperature Engineering"
Institute of Materials, London UK (1997) to be
published.
22. Rickerby D. S., European Patent EP 0628090 Bl
(1993).
23. Strangman, T.E., US Patent, 5,512,382, (1995).
 7-1

TBCs ON FREE-STANDING MULTILAYER COMPONENTS

P. G. Tsantrizos
G. E. Kim
T. A. Brzezinski
PyroGenesis Inc.
1744 William, Montreal
Quebec, Canada H3J 1R4

ABSTRACT After heat treatment, the mechanical

properties of the IN-738LC compare
PyroGenesis Inc. has developed a unique favourably to cast materials. Finally, the
process for the production of components cost of spray forming multilayer
designed to operate in the hot section of gas components is lower than the cost of
turbines. The new process involves spray conventional fabricating options.
forming multilayer components by Vacuum
Plasma Spraying (VPS) onto molds and
subsequently separating the mold from the INTRODUCTION
near net-shape free-standing component.
Advanced TBCs have been developed and Combustion system components of gas
incorporated into the multilayer structure in turbine engines are constructed from high
an effort to extend the component's high service temperature materials, such as
temperature performance capabilities. The ceramics and superalloys. For complex hot
spray formed components are heat treated gas containment components, such as
to improve the mechanical properties of the combustor liners and transition ducts, the
superalloys. PyroGenesis has used the current fabrication process consists of:
VPS near net-shape forming process to (i) mechanically forming several sections
fabricate closed components with a TBC of the component; (ii) thermal spraying the
inner layer, consisting of calcia silica inner surface of each section to form the
(CajSiOJ and zirconia partially stabilized thermal barrier coating (TBC);'(iii) welding
with yttria (PSZ), a CoNiCrAIY bond coat, the sections; (iv) thermal spraying the
and an IN-738LC outer layer. Preliminary protective TBC coating on the welds,
results indicate that the spray formed whenever possible; and (v) laser drilling
components have excellent mechanical thousands of small holes into the structure
properties, can operate at much higher through which cooling air is passed, to
temperatures than similar conventionally provide for additional film cooling. Several
fabricated components and require less significant problems exist with components
cooling. The TBCs showed uniform which have been fabricated in this fashion.
thickness and microstructure with a smooth One problem is the inhomogeneity at the
surface finish. The bond coat and welds. Weld regions act as weak sites
structural superalloy layers were very dense from which failure may initiate due to poor
with no signs of oxidation at the interface. quality finish of both the ceramic top coat

Paper presented at an ÄGARD SMP Meeting on "Thermal Barrier Coatings",

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
7-2

and metallic bond coat of the TBC. The coat is deposited by VPS and the top coat
rough surface of the TBC, inherent to this by APS. NASA's Marshall Space Flight
approach and particularly of the weld Center [5] has reported a five fold increase
regions, leads to an undesirable change in in thermal cycling over the conventional
flow pattern of the hot gas. Moreover, TBC, when both the bond and top coats are
because the current fabricating process applied via VPS. Furthermore,
consists of mechanically forming sections PyroGenesis has experimented with
of the component, there is a limitation on advanced TBCs which can be much thicker
the choice of applicable superalloys (only than conventional zirconia and offer
superalloys with relatively high elongation significantly improved thermal protection.
can be used). Finally, the cost of drilling
the thousands of small holes required for PyroGenesis has committed itself to
cooling is exorbitant and the use of cooling addressing the problems associated to
air significantly adds to the size of the current combustion system components and
compressor. their fabrication method, by combining the
advantages of spray forming and advanced
There have been numerous studies on spray TBCs with those of the VPS technology.
forming of metals and ceramics via VPS
[1, 2, 3, 4]. However, to the authors'
knowledge, there has not been any PROCESS DESCRIPTION
published work on spray forming complex
shape components with a ceramic inner A suitable mold material was selected and
layer, followed by multiple metallic layers. surface machined to the inner geometry of
A closed component with an inner ceramic the desired component. The mold was then
layer has potential applications for thermal cleaned and placed inside the VPS
barrier or wear protection components. In chamber, ready for service. The following
the absence of film cooling, the use of a unit operations were then used to fabricate
conventional thermal barrier coating does multilayered combustion system
not provide adequate insulation to protect components by VPS spray forming:
the component. The temperature drop (i) mold preconditioning; (ii) preheating
across conventional TBCs is approximately the mold to a suitable temperature and
120°C. In order to protect the combustion depositing the TBC layer; (iii) preheating
liners and transitions ducts of an advanced the mold to a higher temperature and
turbine operating without film cooling, a depositing the bond coat and the structural
temperature drop across the barrier of superalloy; and (iv) removing the mold
approximately 350°C is required. Thus, from the near net-shape component.
both a new fabrication technique and a Figure 1 illustrates the process.
dramatically improved TBC is needed for
the proposed application. The design of the mold is of particular
importance in the VPS spray forming
PyroGenesis' experience in VPS-applied process. The mold is used both to define
zirconia based TBC systems (both top and the shape of the component and to control
bond coats) for gas turbine engine the energy balance of the process, a critical
applications has revealed superior thermal issue in controlling the stresses within the
cycling performance at a lower cost over spray formed structure.
the commonly used TBCs, where the bond
 7-3

The mold surface was preconditioned using

the plasma jet; this was a critical step mold
which ensured proper bonding between the preconditioning
mold and the deposit during spray forming,
yet allowed for proper detachment at the
T
same interface upon cooling. The deposition
deposition onto the mold started with the of layers
TBC layer (up to 1.5 mm thick), followed
by a CoNiCrAlY bond coat layer (150 /xm),
Y
and finally reinforced with a thick IN-
738LC structural layer (5 mm). Between removal
layers the mold was preheated in an effort of mold
to minimize stresses within the deposited
structure. Upon completion of the
deposition step, the coated mold was cooled
near net-shape
to room temperature in an inert component
atmosphere. The difference in the
coefficient of thermal expansion between
the mold and the deposited layers provides
natural debonding to occur at this interface. Figure 1. Schematic of VPS near net-
Upon the removal of the mold, a near net- shape forming process.
shape component was obtained with a
uniform TBC incorporated into the inner
surface. RESULTS AND DISCUSSION

The component was then heat treated to Macroscopic Features. Several

transform the lamellar grain structure which macroscopic features were evident on the
is typically produced by spray forming to a near net-shape component. The inside
fine equiaxed grain structure with surface of the component, or the top coat
significantly improved mechanical surface of the TBC, had a very smooth (Rz
properties. Finally, the heat treated < 20 pm) surface finish which mirrored
component was machined to the desired the surface finish of the mold. The
shape and tolerances. component's inner geometry was slightly
larger (0.6%) than the machined mold due
to the slight thermal expansion experienced
during processing. If required, the mold
could be easily machined under final
tolerance to compensate for the expansion.

Microscopic Features. The microstructure

of the component cross-section is presented
in Figure 2. The porosity level of the TBC
top coat could be controlled, between less
than 1% up to 20%, to maximize its
thermal barrier characteristics and thermal
shock resistance. The porosity level of the
7-4

TBC presented in Figure 2 is approximately

7%. The CoNiCrAlY bond coat and IN-
738LC structural layers were applied in
dense form; porosity levels were below 1 %
for both layers. Typical to VPS-applied
.■<* ■'"■'%&■ \ ''' '■ "* i

metals, both of the metal layers reveal an

absence of oxidation.

Figure 3 is a micrograph of IN-738LC after • • •- -, ■*, M'Wz • - it, it," " iti', ?*"•> J , • I ' :■ * ' *» ---" -"' '■* «F ■

etching. The sprayed microstructure

reveals the typical features found in as-
sprayed coatings: lamellar and dendritic
structures. After heat treatment the
microstructure is more homogeneous, with
fine grains and lower in porosity
(Figure 4). The average grain size, after
heat treatment, is approximately l/*m.

50 microns

Figure 3. Micrograph of as-sprayed

IN-738LC (etched).
BfÖNO Mechanical Properties. The mechanical
properties of the VPS-applied IN-738LC
were measured and compared to cast IN-
738LC (Table 1). The as-sprayed properties
of the superalloy show higher tensile
.IN738LC •'■
strength and lower ductility, characteristic
of thermally sprayed metals. After heat
n.
.*> treatment, however, the elongation
Y.»
increases dramatically while maintaining
high strength. In fact, the mechanical
properties of the sprayed and heat treated
250 microns
IN-738LC are superior to those of cast
IN-738LC. The tensile strength of the
Figure 2. Micrograph of component sprayed and heat treated superalloy remains
cross-section (unetched). relatively high at elevated temperatures.
 7-5

TBC Development. A number of TBC

options were considered for the envisioned
application. The target was to create a
TBC thick enough to offer a AT of 350°C.
The thickness of the TBC depends on the
thermal conductivity, and therefore,
composition and morphology, of the TBC.
Table 2 shows the theoretical thickness
required for three TBC options.

Table 2. TBC Options for 350°C AT

Material k (W/m K) Thickness
PSZ 0.8 0.7 mm
A1A 6 4.9 mm
CajSiO, 1.7 1.4 mm

50 microns

While PSZ offers the lowest thermal

conductivity, a durable 700 fim. monolithic
Figure 4. Micrograph of VPS-applied PSZ layer is unlikely. Our experience
and heat treated IN-738LC suggests that PSZ layers thicker than
(etched). 300 jitm are susceptible to cracking and
delamination either during spraying or upon
thermal cycling. Pure alumina is also not
Table 1. Mechanical properties of VPS- suitable due to its relatively high thermal
applied and Cast IN-738LC. conductivity.
Cast* VPS VPS+H VPS+HT Following a preliminary screening of 12
@RT @RT T@RT @800°C TBC options, two compositions were
YS 896 - 960 ±10 666 identified for further testing: (i) a
(MPa) MCrAlY/PSZ graded and laminated
UTS 1034 1195 + 1 1161±13 752
structure; and (ii) a PSZ/Ca^SK^ graded
(MPa) 2 structure. The two selected TBC options
were sprayed onto flat coupons and tested
Elong 9 0.6±0.2 9.5±0.5 - for thermal shock resistance, thermal
• (%) cycling, and oxidation at 1100°C.
Hard - 338+72 352+12 -
(HV) A microstructure of the as sprayed
* from technica 1 data pub lished by ] NCO MCrAlY/PSZ TBC is shown in Figure 5.
Inc. The MCrAlY used in this study is AMDRY
7-6

995 (CoNiCrAlY). While the thermal

shock test (heating to 1,100°C and rapidly
quenching in water) showed satisfactory
results for this TBC option, the oxidation
tests proved that graded MCrAlY layers
oxidize very rapidly above 900°C. A
microstructure of the TBC after 25 hrs in
air at 1,100°C is shown in Figure 6.
Under these conditions the coating expands
in volume and delaminates. Based on this
catastrophic failure, no thermal cycling
tests were performed for this option.
mSmM h
nji-j'#is»'äj^
f%*i

300 microns

Figure 6. Micrograph of MCrAlY/PSZ

TBC after 25 hrs oxidation
at 1,100°C.

300 microns

Figure 5. Micrograph of as-sprayed Zr02

graded and laminated
MCrAlY/PSZ TBC.

The PSZ/Ca^iC^ TBC consisted of a 200-

300 pm layer of PSZ followed by a 300-
650 j*m PSZ/CajSiCv graded layer and a
300-1,000 /*m layer of Ca^iCv A typical
micrograph of this TBC option is shown in
Figure 7. This option showed excellent 250 microns
resistance to thermal shock and minimal
weight gain due to oxidation at Figure 7. Micrograph of as sprayed
temperatures as high as 1,300°C. PSZ/Ca2Si04 TBC.
 7-7

Thermal cycling was performed on the The results from the thermal conductivity
PSZ/Ca2Si04 TBC by heating Hat coupons study are presented in Figure 8. Figure 8
to 1,100°C with an oxyacetylene torch and shows that the AT between the hot and cold
cooling rapidly with compressed air. faces of a uncoated stainless steel coupon is
Following 300 cycles no delamination was approximately 200°C when the hot phase
observed. reaches 1,000°C. By comparison, a
conventional PSZ TBC structure offers a
A number of PSZ/Ca2Si04 TBCs were AT of 275°C (Sample Z-10). Under the
produced on flat stainless steel substrates same operating condition an advanced 1.8
and tested for their thermal barrier mm thick PSZ/Ca2Si04 TBC offers a AT of
properties. The test was performed by approximately 420°C (Sample ZC-9).
heating the TBC face with two high Thus, the thermal barrier properties of the
intensity infrared lamps and measuring the PSZ/Ca2Si04 developed in this study can be
temperature of both the hot TBC surface almost 3x better than a conventional TBC,
and the cooler stainless steel surface. A which should be sufficient to meet the
list of the samples evaluated in this study is requirements of the proposed application.
shown in Table 3.

Table 3. PSZ/Ca2Si04 Samples Tested

for Thermal Conductivity. 900
Average Thickness (/xm)
Sample
PSZ Graded Ca2Si04
Z-10 245
C-4 559
ZC-7 238 303 317
ZC-3 308 330 650
ZC-10 210 424 455
ZC-6A 225 584 303
ZC-9 248 655 921

300 ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' '
Sample Z-10 represents a conventional PSZ 650 700 750 800 850 900 950 1,000
TBC sprayed by VPS. Sample ZC-9
Temperature onlhe hot surface, (°C)
represents a TBC which theoretically
should provide a AT in excess of 350 °C
and thus meet the performance
requirements for advanced combustion Figure 8. Thermal properties of
liners and transition ducts without film PSZ/Ca2Si04 advanced TBCs.
cooling.
7-8

SUMMARY

The feasibility of VPS near net-shape

forming multilayered free-standing
components has been demonstrated. This
innovative fabrication process was used to
produce a combustion system component of
a gas turbine engine. The improved
qualities of VPS near net-shape formed
combustion system component include:
(i) an advanced TBC capable of offering
AT in the order of 350°C; (ii) a superior
high-temperature structural superalloy;
(iii) smoother inner TBC surface; (iv) no
irregularities (welds) within the component;
(v) excellent mechanical properties; and
(vi) consistent reproducibility. The cost of
fabricating the components using the VPS
net-shape forming technology is
significantly lower than conventional
methods. Furthermore, it is hoped that the
need for film cooling can be eliminated.

REFERENCES

1. M.R. Kim, R.W. Smith, and D.

Kapoor, Thermal Spray: Practical Solutions
for Engineering Problems, 7-11 (1996).

2. G. Montavon, E. Roussel, C. Coddet,

E. Fauvet, J.P. Huchin, D. Cornu, J.M.
De Monicault, J.C. Garcia, and F.
Pellerin, Proceedings of the 7th National
Thermal Spray Conference, 485-489
(1994).

3. T. McKechnie, P. Krotz, and Y. Liaw,

Proceedings of the 7th National Thermal
Spray Conference, 457-461 (1994).

4. H.-D. Steffens, M. Dvorak, and K.

Nassenstein, Proceedings of the
International Thermal Spray Conference &
Exposition, 369-374 (1992).

5. NASA Tech Briefs MFS-30081.

 8-1

ELECTRON BEAM PHYSICAL VAPOUR DEPOSITION THERMAL BARRIER COATINGS:

A COMPARATIVE EVALUATION OF COMPETING DEPOSITION TECHNOLOGIES

Y.Jaslier, S.AIperine
SNECMA - Materials and Processes Department
Site de Villaroche - Bätiment 41
77550 Moissy-Cramayel
France

ABSTRACT there has been a strong drive to extend the application of

The need for improving the performance and maintenance TBC's to protecting the aerofoil of high pressure turbine
costs of gas turbine engines has led to the development of blades and vanes. Unfortunately, the plasma spray process
advanced thermal protection systems for critical components proved inadequate for depositing ceramic coatings onto high
such as high pressure turbine (HPT) blades and vanes. This pressure turbine (HPT) aerofoils due to the poor coating
led to the concomitant development of advanced coating surface finish retention and cooling hole obstruction
deposition techniques. Electron beam physical vapour problems. It has been alleged that ceramic plasma sprayed
deposition (EB-PVD) first made possible the application of coatings had 'insufficient' spallation resistance for this
corrosion resistant overlay coatings on turbine blades. The application2.
EB-PVD technology has since been gaining ever more
interest world-wide as it stands as the best industrial In the late seventies/early eighties, the EB-PVD technique
technique for the deposition of thermal barrier coatings on was found to be a very interesting potential alternative to
first stage HPT aerofoils. The strategic nature of the EB-PVD plasma spraying for depositing TBC's onto aerofoils.
process means that it has been developing somewhat Refractory materials deposited by EB-PVD typically exhibit a
independently on both sides of the iron curtain during the unique columnar morphology, which has been vehicled as a
cold war years. Today, both American and ex-USSR key feature to accommodate the thermal expansion mismatch
technologies are open to the market. This is the object of this strains with the metallic substrate upon thermal cycling and
paper to compare the two state-of-the-arts both from a to resist severe thermal shocks. A second advantage of the
deposition process standpoint and from a laboratory electron beam physical vapour deposition process is its
evaluation on samples. This evaluation covers structural capability to produce coatings with an acceptable surface
studies as well as thermal cycling testing. The relationship finish as-deposited which is retained in service in spite of the
between the deposition process and the coating functional erosive environment. A third and not least advantage is that
behaviour is discussed. the EB-PVD technique makes controllable and reproducible
the obstruction of cooling holes. This is due to the deposition
INTRODUCTION mechanism that takes place through the condensation of
In the Western countries, the electron beam physical vapour coating vapour in the EB-PVD process rather than through
deposition process was originally developed in the late sixties the impingement of large semi-molten particles in the plasma
by Airco Temescal jointly with Pratt & Whitney for the spray process.
application of overlay type oxidation resistant coatings onto
high temperature blades and vanes of jet engines1. The so- Although the potential of EB-PVD TBC's was recognised,
called MCrAlY's, where M is nickel, cobalt, iron or a the know-how to reliably produce TBC's by electron beam
mixture thereof, offered then superior environmental physical vapour deposition proved to be arduous to establish
resistance compared to diffusion aluminides. In parallel with because of 'infant mortality' problems as reported by the
the development of advanced oxidation resistant coatings, Western pioneers themselves. The strategic nature of the
emerged a new class of high temperature coatings designated EB-PVD applications means that the difficulties have been
as thermal barrier coatings. A thermal barrier coating (TBC) tackled and the process technology matured somewhat
is a duplex system consisting of a heat insulating ceramic independently on both sides of the iron curtain during the
layer deposited on a bond coat underlayer providing adhesion cold war years. Companies such as Airco Temescal, Pratt and
to the ceramic as well as corrosion resistance to the Whitney and Chromalloy made essential contributions to
underlying base alloy. The ceramic coating is commonly make the technology of EB-PVD TBC's industrially viable4.
yttria-stabilised-zirconia for its low thermal conductivity and Production coatings have been flying in PWA engines since
high thermal expansion coefficient. 1989 2.

The early application of TBC's was the protection of In the Eastern countries, the EB-PVD technology was bred in
combustor parts. This application involved applying the prestigious institutes such as the Paton Welding Institute in
ceramic coat on top of an MCrAlY coating used as a bond Ukraine and the VIAM in Russia, with industrial recipients
coat, both deposited by air plasma spraying. One advantage of such as Saturn and Nickolai Kuznetzov (NK). The first
the air plasma spaying technique is to provide the MCrAlY Structure Zone Model for physical vapour deposited coatings
coating with a rough surface onto which the ceramic coating was first published in 1963 by Movchan and Demchishin5
can mechanically bond. Another advantage of air plasma and has been universally referred to since. The design
spraying lies in the heavily microcracked ceramic approach of Russian jet engine manufacturers was such that it
microstructure it produces which is beneficial in terms of was not felt there was a need for coating HP vane aerofoils
strain tolerance and thermal shock resistance. with EB-PVD TBC's. As a result Russian experience in
production was bound to coating blades, not vanes6. This
Since the potential benefits of thermal barrier coatings were detail bears some implications on the development of tool
additive to the gains in high temperature strength capability design and masking technology (or lack of) in the former
of superalloys and in internal and film cooling technologies,

Paper presented at an AGARD SMP Meeting on "Thermal Barrier Coatings",

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
8-2

USSR. EB-PVD TBC's applied on single crystal blades have

been flying on NK engines since 1985.
Preheat chambBr

Today, both the American end ex-USSR technologies are

Loadlock
open to the market. This is the object of this paper to shed
some light on how the two state-of-the-arts compare from the ^2^
technology stand-point. Some of the commonalities and
differences in the TBC concept and manufacturing technology
are first highlighted. Results from a laboratory evaluation of
coatings that are representatives of these technologies are
then presented and discussed.
270° electron beam Blades on
Ceramic ingot
COMMONALITIES AND DIFFERENCES rotating wheel

TBCconstituents Figure 1: Top view schematic of a Western type

Whether they originate from the East or from the West, EB- (Temescal) production EB-PVD coater"
PVD TBC's have overall much in common if one considers This deficiency is manifested in the black colour of the
that they all consist of an oxidation resistant bond coat zirconia coating. The lack of stoichiometry of zirconia
overlaid with a ceramic top coat deposited by electron beam deposits is in itself not a problem since it can be restored via
physical vapour deposition. Typical bond coats are from the a simple heat treatment in air at temperatures as low as
MCrAlY type. Only in the West, does it appear that diffusion 700°C. Some people have claimed that the lack of oxygen
aluminides have been used as bond coats beside MCrAlY's. during the coating atomic build up actually affected the
The range of thicknesses for the MCrAlY and ceramic layers coating microstructure in a detrimental manner due to the
are typically 75-125 and 80-250 microns respectively. material swelling as a result of restoring its stoichiometry".
Another invariant between the two technologies is the Another argument for the need of oxygen bleed inside the
standard ceramic composition which consists of Z1O2- coating chamber, is the need to establish, prior to depositing
6to8wt%Y203 (YSZ). It is interesting to note that this the ceramic, a thin film of aluminium oxide at the work piece
composition range was originally selected from the thermal surface onto which the ceramic condensate can chemically
cyclic life optimisation of TBC's deposited by plasma bond12,13. The oxygen rich environment surrounding the parts
spraying8. in transit from the pre-heating chamber to the deposition
chamber is prone to forming a thin film of alumina acting as
The EB-PVD manufacturing process the 'glue' of the ceramic coat to be deposited. To the
In generic terms, the EB-PVD process involves evaporating authors, the in situ bond coat pre-oxidation prior to ceramic
coating material in a vacuum chamber by means of electron deposition is a stronger argument to justify for the need of
beam heating. The coating vapour so generated condenses on oxygen bleed inside the coating chamber rather than
the work piece at high temperature. The coating flux being stoichiometry or microstructure of the deposit. Yet, the
line-of-sight, the parts are manipulated above the evaporation
Eastern technology of EB-PVD TBC's matured without
source to maximise coating coverage.
identifying oxygen bleed as a key deposition parameter.
The design of Western EB-PVD coaters is based on the Figure 2 is a schematic of a Russian production coater. Its
principle that the evaporating conditions must be kept as architecture is based on a central deposition chamber with
steady as possible with no interruptions, with a batch of parts two load locks on each side. There is no separate preheating
being coated in the deposition chamber at any time of the chambers as in the Western coaters.
campaign. This approach implies that the parts must be
loaded and pre-heated externally to the deposition chamber.
The architecture of Western type EB-PVD coaters is thus Evaporating guns
. Pumps
Electron beams
generally based on a central deposition chamber with
separate pre-heating and unloading chambers (see Figure 1). Parts carrousel Loading
Although the main concern in this approach is to maximise chamber

the coater's productivity and reproducibility, the machine

H \a a
design actually has implications on the deposition physics
that govern the coating properties and this must be borne in iL-,
mind when comparing the two technologies. In the deposition
chamber, the parts are maintained at temperature mainly
thanks to the radiative heating from the molten pool. For Coramlc/metal ingots
metallics however, where the melt pool surface temperature Deposition chamber
Heating gun
is not as high as for ceramics, supplementary heating of the
parts is provided by a resistive radiator9'10. Depending on the
coater's design, whether it be Temescal or Leybold, the parts Figure 2: Top view schematic of an Eastern type
are manipulated in the vapour cloud with various complex production EB-PVD coater
motion patterns involving rotation and tilting. One of the Pre-heating of the parts is accomplished inside the deposition
specificity of the Western technology though is the bleed, chamber by the means of an electron beam gun located in a
inside the coating chamber, of an oxygen containing gas when separate chamber above the deposition chamber. This gun is
depositing ceramics9'10. The argument put forward to justify also used as auxiliary heating of the parts during the coating
for the need of oxygen bleeding is to compensate for the process. Four stacks of ceramic or metal ingots are positioned
dissociation of zirconia upon heating which would otherwise in a row and are evaporated by four individual linear electron
lead to some oxygen deficiency. beam guns. The deposition process is carried out at a vacuum
 8-3

level of a few 10"4 Torr. There is no intentional bleed of a although poorly understood, is known to be moisture related.
reactive gas into the chamber. A typical coating cycle consists Other investigators have reported the effect of moisture on
first in mounting the carrousel holding the parts on the the spallation resistance of thermally grown oxides on bare
driving shaft inside the loading chamber. After the loading superalloys15. In these tests, thermal cycling was interrupted
chamber has been evacuated, the locking vane is opened and every 20 cycles with the test pieces left at room temperature
the carrousel with the cold parts is introduced into the for a time up to 4 hours, after what the specimens were
deposition chamber by translation of the shaft. The parts are inspected and the cycling resumed. A dwell time of 4 hours at
pre-heated inside the deposition chamber by the heating room temperature was chosen so as to let enough chance for
electron beam gun while the evaporating guns are kept shut the moisture to have an effect, if any. The failure criterion
down. Evaporation is started after the work pieces have was a ceramic spalled area of at least 20% of the specimen
reached the right temperature regime. Once the deposition is coated surface.
completed the carrousel is driven back into the loadlock
where it is left to cool down before it is unloaded. RESULTS

It should be noted at this point that Western coaters are Characterization of as-deposited TBC's
generally designed to be dedicated to one type of coatings i.e. The overall TBC systems are presented in Figure 3. It may be
metallics or ceramics whereas, in the Eastern art, the coaters seen that each ceramic coating system exhibits a specific
are designed and used to deposit both types of coatings. ceramic morphology. For a same nominal YSZ composition,
Whether the latter approach is better or worse from the the RT33 process produces a coarser, more columnar shape
production standpoint than the dedicated type coater approach morphology compared to the finer, feather-like, KDPi
is not the debate of the present discussion. The point is that ceramic morphology. Moreover, the KDPI feathered columns
the flexibility of the Eastern type coaters permits to evaporate coarsen towards the outer surface. This effect is not as
metallics concurrently with ceramics or to deposit both in dramatic in the RT33 morphology.
sequence in the same coating cycle and this is essential to
their art14.
wi

SCOPE OF LABORATORY EVALUATION

Coating systems
Two thermal barrier coating systems were evaluated in this
study. The Russian coating was manufactured at the Samara
4
wiW
II,
,.
mmm :-^r- -<■— —-i: J.-*>^i-w-i—
plant of NK Engines under the technical supervision of the
VTAM Institute in Moscow. The American thermal barrier
coating is the well known RT31/RT33 system manufactured
-11«*,
by Chromalloy in the USA. The nominal designation of these :*

systems is given in the table below:

Bond coat/ Bond coat Ceramic top coat

Top coat composition composition !:\1
wc.V Spot Magn Det WD ,
Designation 1MW.4J) g5px__BSE 11.3 T24(etatbrut)
RT31/RT33 Co34Ni20Cr8A10.5Y Zr02-7 wt%Y203
SDP2/KDP1 Ni20Crl2A10.5Y Zr02-7wt%Y203
Li !
* ■ ' I■ i. ' ■ If j- 11 , I I >,/ i V
> It, ';.•[ I
For each system, both bond coat and ceramic coat were
applied by electron beam physical vapour deposition on disk
and cylinder shaped test pieces made out of Hastelloy X. The
;.,.,;,.,:]U'\v' .'. ■ .,.1 ■--.;.«
composition of Hastelloy X is given in the table below in h\k[.'- ;:■■.','•'.//.■ ■■■-■
i* ' I * 1 4 ill I IUli!ilÄilBlB-SiWl^B"
wt%: <Y •■ . '■ ■.'.■: ■!. »i'.'■''' j.«

Ni Cr Fe Mo Co W
bal 22 18 9 1 0.2

Characterization
The morphology of the ceramic layers was evaluated using a
field emission gun scanning electron microscope (FEG-
SEM). Energy Dispersive X-Ray analysis coupled with SEM
Acc.V Spot Magn Det WD
was used to perform chemical analysis in both the as- 115.0 kV 4.0 360x BSE 10.0 VTP21(SDP?-Ätathni«V7fl1
deposited and as-tested state. Ceramic phase composition was Figure 3: SEM back-scattered electron image of the as-
determined using CuKa X-Ray diffraction. deposited RT31ART33 (a) and SDP2/KDP1 (b) TBC
systems
TBC performance was evaluated using a thermal cycling test.
Thermal cycling was conducted on 6 and 10mm diameter bars High magnification examination of the ceramic layers at 50
in a furnace with a cycle consisting of 55 minutes at 1100°C micron height from the bond coat gives some insight into the
with a 5 minute heat up and 15 minute cool down to room column sub-structure (Figure 4). Back-scattered electron
temperature using forced air convection cooling. The imaging reveals internal striation consisting of alternate dark
probability for the ceramic to spall off after many cycles and white layers. This effect is a lot less pronounced in the
increases with dwell time at room temperature. This effect,
8-4

case of the KDP1 ceramic where only minor striation may be Overall bond coat composition was determined using semi-
distinguished in the core centre line of the feathered columns. quantitative EDS analysis over an 80x80microns window
centred within the bond coat depth. The analysis results are
shown in the table below:

Co Ni Cr Al Y Fe
RT31 wt% bal 38 18 8.9 <1 <1
at% 34 18 17
SDP2 wt% - bal 14 12 <1 <1
at% 14 22

A band of white dots identified as yttrium rich precipitates

was found at mid-height in the RT31 CoNiCrAlY (see Figure
3). It was observed at higher magnification that this yttrium
rich phase was nucleated at the y/ß phase boundary. No such
yttrium rich precipitates were observed in the SDP2 NiCrAlY
layer. A feature peculiar to the SDP2 MCrAlY layer was the
presence of a population of chromium rich nodules of 1
micron in average diameter (see Figure 6). These
constituents appeared black in back-scattered electron
imaging. Examination in the secondary electron emission
mode revealed a surface relief suggesting that these nodules
were harder than the matrix. These observations point to the
presence of chromium carbides (Cr7C3?,Cr23C6?), indication
that would need to be confirmed using X-Ray diffraction.

Figure 4: SEM back-scattered electron images showing

details of as-deposited RT33 (a) and KDP1 (b) ceramic
morphologies.

X-Ray diffraction patterns of each ceramic layer are

presented in Figure 5. It was found that each ceramic system
was single phased and consisted of the non-transformable
tetragonal phase (f).

1BBBB
[counts]
9BBB
-

KDP1
,
(28B)
5806

4aee

3BBB
(488)
RT33 (311)
. \
e
£V 38
JB_ 4B 5B
SB 68
bö 78
YB 1i'iei
4B I B'B
BB

Figure 5: X-ray diffraction patterns of RT33(a) and

KDP1 (b) ceramic layers.
Figure 6: SEM secondary electron image of ceramic/bond
All coatings exhibited the (200) preferential growth coat interface for RT31/RT33 (a) and SDP2/KDP1 (b)
orientation. This texture is however less pronounced in the
Russian ceramic. No peak broadening was observed A close examination of the ceramic/bond coat interface
indicating that the diffracting domains were larger than 0,1 reveals that the alumina film developed on the SDP2 bond
micron in size. coat is continuous and thinner than that developed on RT31
(see Figure 6). For the latter, alumina seems to only have
formed at the interface with the y phase i.e. no alumina could
 8-5

be identified at the YSZ/ß-(Ni,Co)Al phase interface using

SEM. Cycles-to-failure (55 minutes at 1100°C)

Right underneath the alumina film formed on SDP2, there

exists a thin (1.5 to 2um) and very fine grained MCrAlY sub- n
i,
layer separated from the bulk of the bond coat by a line of = 6mm
alumina inclusions (see Figure 7).
14— "

ilPfPiiW \\l ■o

"55

o
a
W)
rh
/

41
0 - In
SDP2/KDP1
, Ilk
RT31/RT33

Figure 8: Thermal cyclic performance of RT31/RT33 and

SDP2/KDP1 systems

Characterization of as-tested TBC systems

Micrographs of a typical low life (106 cycles) RT31/RT33
Acc.V Spot Magn Det WD I system are shown in Figure 9. The substrate surface was
15.0kV4.0 lOOOOx BSE 10.0 VTP31 :etat brut(V89)
analysed and consisted of MCrAlY with small isolated
Figure 7: SEM back-scattered electron image of as- alumina patches. The remnant alumina most often contained
deposited ceramic/SDP2 interface region yttrium rich oxides that show up as white constituents using
atomic contrast imaging. The inner side of the ceramic spall
Thermal cycling performance was analysed by EDS and consisted of alumina with isolated
Figure 8 is a bar chart showing thermal cycles-to-failure for yttrium rich oxide phases. The alumina grains imprints can
the SDP2/KDP1 and RT31/RT33 coatings on 10mm diameter be recognised on the bond coat surface. These observations
bars (unless otherwise stated). demonstrate that ceramic spallation occurred through the
propagation of a crack running at the MCrAlY/alumina
Coating spallation life is found to be very scattered for both interface breaking off the yttrium rich alumina protrusions.
systems. There appears however that, under the specific
testing conditions used in this study, the SDP2/KDP1 coating The failure morphology of low life SDP2/KDP1 coatings was
overall exhibits a superior spallation resistance compared to similar to that described above, with the exception that no
that of the RT31/RT33 coating. Mean, low and high thermal yttrium rich oxide phases seemed to have formed. In Figure 10
cyclic lives are shown in the table below for comparison. are shown micrographs of a SDP2/KDP1 coating failed after
96 cycles. This specimen was peculiar in that a continuous
RT31/RT33 SDP2/KDP1 layer of (Ni,Cr Al) bearing oxides, presumably spinels, had
Low 19 96 formed between the alumina scale and the ceramic. This
Mean 260 460 mixed oxide layer that appears voided in cross section
High 710 1081 (Figure 10b) behaved as another site for crack propagation in
addition to the MCrAlY/alumina interface.
Ceramic spallation occurred at room temperature. In most
cases, the spallation event was not progressive but abrupt, In Figure 11 are presented micrographs of the RT31/RT33
with typically 80% of the ceramic loss occurring at once from specimen that failed after 710 cycles. EDS analysis of the
a no spall situation. One exception was the SDP2/KDP1 inner side of the ceramic spall indicated that it consisted
specimen that was stopped after 1080 cycles. This specimen predominantly of YSZ ceramic. Inversely, the substrate
failed via the occurrence of discrete spalls that were surface side consisted mainly of aluminium oxide, implying
dispersed and small in size (no bigger than 2mm) but the that cracking had occurred at the ceramic/alumina interface.
number of which increased with test cycles. The respective fractured surfaces reveal a comparable
topography consisting of equiaxed spherical grains less than 1
The short life specimens typically exhibited a shiny metallic micron in size. Some isolated (Ni,Co,Cr,Al) mixed oxides
surface on the substrate surface where the ceramic had were also observed at the interface between the ceramic and
spalled. The longer the lifetime of the test pieces and the alumina scale although this feature remained marginal.
more grey and mat this surface looked. For the high life
specimens, this surface appeared covered with a grey film of The most durable SDP2/KDP1 specimen (1080 cycles) was
dust. This film could be rubbed off with the finger tip. This assessed on a polished cross-section. The progressive failure
pattern was generally followed by both the RT31/RT33 and of this specimen into discrete ceramic spalls did not make
SDP2/KDP1 systems. In two instances with the SDP2/KDP1 possible to look at freshly spalled surfaces. It can be seen in
coating (96 and 114 cycles), the ceramic free substrate Figure 12a that a crack is running predominantly within the
surface and the inner side of the ceramic spalls were outer region of the alumina scale with some excursions into
blue/green in colour, indicating the presence of nickel and the ceramic. It cannot be determined whether this crack is the
chromium rich oxides. These visual observations were result of metallographic preparation or whether it is a
followed with SEM evaluation. genuine defect generated from the cycling.
8-6

Acc.V Spot Magn Det WD I

150kV4.0 4000X BSE 10.0 V7P21/SDP2.96 cycles(Vll)O)
Figure 9: SEM back-scattered electron images of bond Figure 10: SEM back-scattered electron images of bond
coat surface (a) and inner side of ceramic spalls obtained coat surface (a) and cross section of a low life
from a low life RT31/RT33 system (b) SDP2/KDP1 system (b)
Its location however indicates that the ceramic/alumina contributors to the difference in ceramic morphology that is
interface region is the weakest link. Some isolated nodules of observed. This argument must be put in perspective with the
(Ni,Cr,Al) mixed oxides were also found to have formed fact that Western EB-PVD TBC's deposited either from a
between the alumina scale and the zirconia, locally lifting off Leybold or a Temescal machine are very similar in
the ceramic coat (see Figure 12b). The formation and growth morphology19 with yet massive differences in the coating
of these mixed oxide nodules may be responsible for the chamber configuration. The role of chamber pressure and
discrete ceramic spalls that were observed on this specimen. deposition rate (higher in the Western process) on
The alumina scale thickness on this coating after 1080 cycles morphology should not be ignored.
was measured at 5.5 microns which is to be compared to the
7.5 microns alumina scale developed on RT31/RT33 after Another difference in ceramic structure was found in the
710 cycles (see Figure 13). column striation which is significant in RT33 but only
marginal in KDP1. This well-known feature is usually related
DISCUSSION to rotation of the work piece during deposition20. As the parts
are rotated in the vapour cloud, the surface to be coated sees
As-deposited thermal barrier coating systems alternate regimes of high deposition rate at high temperature
when facing the source (brighter lines) to low deposition rate
Ceramic considerations at lower temperature when facing away from it (darker lines).
The difference in ceramic morphologies obtained from the Lelait et al have identified using transmission electron
RT33 and the KDP1 process clearly demonstrates how microscopy that the darker lines corresponded to aligned
sensitive the morphology of EB-PVD deposits can be to porosity in the ceramic columns. Later, a 'dynamic sintering'
specific deposition conditions. Morphology is classically coating growth model has been proposed to account for the
affected by the substrate surface roughness 16,17, substrate presence of intra-columnar porosity in EB-PVD deposits17.
temperature 5'17, deposition rate and part motion pattern in This model was developed on the consideration that a PVD
the vapour cloud such as rotation speed18. No major deposit builds up as a porous atomic network and that this
differences exist in either the deposition temperature range or spontaneous defectiveness is overcome at high temperature
the bond coat surface finish between the two processes that by diffusion processes resulting in an array of micropores
could account for this effect. The part motion pattern and the (coalesced vacancies). It was proposed that not only the
vapour cloud distribution inside the coating chamber, both ceramic surface temperature played a role in the rotation
very dependent on the coater design, are however very induced striation of the ceramic columns but also the kinetic
different. These parameters are therefore potential
energies of the condensing species which were in turn
affected by gas scattering effects.

It may be understood that the ceramic striation is less

pronounced in the case of Russian ceramics if one considers
that:

• the ceramic surface temperature does not fluctuate with

rotation as much in amplitude as in the Western process
since the part temperature relies to some extent on
■ auxiliary electron beam heating.
• when the surface to be coated faces away from the source,
only the thermalised fraction and back scattered
molecules of the vapour cloud can possibly deposit. In the
KDP1 process, where the evaporation takes place in a
vacuum of a few 10'4 Torr, the surface to be coated hardly
receive any coating flux when facing away from the
source.

Unlike morphology, the phase composition was found to be

an invariant between the two kinds of ceramic. The finding of
the (f) phase in 7YSZ EB-PVD deposits was first
established by Lelait et al21 after previous authors had
concluded to the presence of the cubic phase. This phase is
typically obtained from quenching the cubic phase of bulk
Zr02-7wt%Y203 ceramic . It is also found in plasma sprayed
Zr02-7wt%Y203 deposits due to the rapid solidification
nature of this process. The formation of the (f) phase in EB-
PVD deposits can be explained from the fact that the yttrium
and zirconium oxide molecules condense randomly at a
relative temperature T/Tm of 0.4 approximately (where T and
Tm are the absolute surface temperature and melting point of
zirconia respectively). In this temperature regime, cation
diffusion is too sluggish to allow for the yttrium to
redistribute into the yttrium rich cubic phase and the low Figure 11: SEM secondary and back-scattered electron
yttrium content transformable tetragonal phase. Given that images of bond coat surface (a) and inner side of ceramic
the RT33 and KDP1 processes involve a same ceramic spalls (b) obtained from a high life RT31/RT33 system
composition (zirconia alloyed with 7wt%yttria) and a low
deposition temperature relatively to the melting point of The most peculiar feature that distinguishes the SDP2 bond
zirconia, one could expect the formation of (f) phase in both coat from RT31 is the presence of a fine grained MCrAlY
ceramic deposits. sublayer adjacent to the ceramic coat. This thin layer
originates from the evaporation of a flash of MCrAlY on top
The RT33 ceramic deposits exhibited a preferred (200) of the already mechanically prepared and heat treated bond
orientation that was much more pronounced than in the case coat surface. The Russian art is such that, for each ceramic
of KDP1. A zirconia condensate with a (200) texture builds deposition run, the evaporation of MCrAlY for tens of
up with atomic layers in the coating plane that consist seconds precedes the onset of ceramic evaporation. This
exclusively of oxygen atoms. It may therefore be argued that philosophy probably is meant to deposit the ceramic layer on
the richer oxygen environment characteristic of the RT33 a fresh, reactive and contaminant free surface.
process will favour the (200) growth orientation compared to
a deposition process carried out in vacuum. The nature of the bond coat metal surface is of prime
importance as it controls the nucleation and growth of the
Bond coat considerations oxide film that acts as the glue of the ceramic layer.
MCrAlY bond coats are classically described as consisting of Evidence was shown that this native oxide was different (in
ß-(Ni,Co)Al precipitates in a solid solution of Y-(Ni,Co)CrAl. thickness and continuity at least) whether it was formed at
Both RT31 and SDP2 bond coats fit this pattern. Some the RT31/RT33 or SDP2/KDP1 interface.
differences were found relative to the presence of a third
population of constituents identified as yttrium rich The interfacial oxide film may form as a result of one or a
precipitates in the former and presumed chromium carbides combination of mechanisms such as:
in the latter. These differences most probably reflect different
bond coat processing histories involving work hardening and 1. ex-situ bond coat oxidation prior to the ceramic
heat treatment steps. Some deviation from the nominal deposition step. In both the Western and Russian
composition was noticed in the chromium content of the manufacturing processes, the bond coat metal surface is
SDP2 bond coat. This deficiency of a bond coat applied on an mechanically prepared in a way that is prone to removing
otherwise chromium rich substrate, may be partly attributed any oxide scale formed during a prior heat treatment. It is
to chromium evaporation during the vacuum heat treatments therefore unlikely that the interfacial alumina observed
at 1050°C that preceded ceramic deposition. originates from ex-situ oxidation.
2. in-situ bond coat oxidation during the transient pre-heat 1*1}] f.1-tV\!»
stage in vacuum.
3. in-situ bond coat oxidation during the transit from the
pre-heat chamber to the deposition chamber in a partial
pressure of oxygen, this mechanism standing for the
Western type coating process.
4. in-situ bond coat oxidation during ceramic deposition.
Bearing in mind that the ceramic is porous and that
oxygen diffusivity in YSZ ceramics is significant23 above
900°C, oxygen can be conducted to the aluminium
bearing bond coat metal through the ceramic coating and
fuel alumina growth.
5. ex-situ bond coat oxidation post ceramic deposition
during heat treatments at 1050°-1100°C for times up to 4
hours. Beside oxidation from the residual oxygen present
in a vacuum of 10"" Torr, it is suggested that the reservoir Acc.V SpotMagn Det WD , , „ M,„
16.0 kV 4.0 4000X._ BSE 10J) VTP21/SDP2:1080 cycles(V77)
of oxygen contained in the YSZ ceramic participates to
some degree to bond coat oxidation. This suggestion
follows the observation that white zirconia deposits
typically darken out after a 4 hour heat treatment at
1080°C in a partial pressure of argon, implying tire loss of
oxygen from the ceramic.

None of the mechanisms 2 to 5 proposed above appears to

have activated the oxidation of the RT31 bond coat ß phase.
This observation is somewhat in contradiction with a
previous result reported by Leyens et al24 who studied the
effect of various pre-oxidation treatments of NiCoCrAlY
bond coats. It was found that mainly the ß phase was covered
with oxides after a 4 hour vacuum heat treatment at 1080°C. B
To account for the preferential oxidation of the bond coat
AccV SpotMagn Det WD , , «.„ ,,,„
gamma phase in RT31/RT33, it is proposed that the
15.0kV4.0 1000x SE 10 0 VTP21/SDP21080 cycles(V77)
chromium rich y phase readily forms chromia nuclei during
the transient heat up stage. These nuclei would act as Figure 12: SEM back-scattered and secondary electron
nucleation sites26 and subsequent growth of alumina at a cross-section images of SDP2/KDP1 after 1080 cycles
working oxygen partial pressure that is insufficient to
nucleate alumina on the ß phase.

It is interesting to note that the Russian process produces a

continuous film of alumina between' ceramic and bond coat
metal whereas the Western process fails to form a continuous
film, as-deposited, in spite of the oxygen bleed inside the
deposition chamber.

One could argue that, prior to ceramic deposition, the flash of

evaporated MCrAlY consists mainly of the y solid solution
since the time at temperature is too short to allow for
extensive growth of ß precipitates. The low ß phase content
at the time bond coat oxidation is initiated could be one
argument to explain why the alumina film is continuous
between KDP1 and SDP2. Moreover, the fine grain size of
this MCrAlY sub-layer should promote the formation of a
high density of chromia nuclei. This would further help the
alumina film to cover the whole metal surface. Finally, the Figure 13: SEM back-scattered electron cross-section
role of electron beam irradiation on the bond coat oxidation image of RT31/RT33 failed after 710 cycles.
in the KDP1 process should not be overlooked. Electron
bombardment is known to activate chemical reactions which Thermal cycling performance
would otherwise not take place under thermodynamic EB-PVD TBC's have long been described as failing at the
equilibrium conditions. interface between the TGO (thermally grown oxide) and the
bond coat metal25. It was shown in this work that a same
nominal thermal barrier coating system could exhibit
different spallation mechanisms ranging from a lack of
adhesion at the alumina/bond coat metal to failure at the
ceramic/TGO interface. Adhesive failure at the bond coat
metal/TGO interface was synonymous of short cyclic lives for
both the RT31/RT33 and SDP2/KDP1 systems. Whenever
 8-9

the metal/oxide adhesion was good, TBC spallation life

seemed to be controlled by the integrity of the ceramic/TGO
interface. Failure at this interface could be the result of the
formation of unprotective (Ni,Cr,Al) mixed oxides between
the alumina scale and the ceramic coat as seen in the case of
SDP2/KDP1. The growth of spinels at the YSZ/TGO
interface implies the transport of Ni and Cr cations through
the alumina scale. This could be explained by the formation
of local through-cracks of the scale. No such cracks were
however identified. Even in the absence of mixed oxide
formation, the YSZ/alumina bond in the RT31/RT33 system
seemed to deteriorate upon thermal ageing (see Figure 11).
How the integrity of the YSZ/alumina interface evolves with
time at temperature is an issue that may depend both on the
alumina growth mechanism and ceramic sintering
mechanisms.

MCrAlY coatings are known to exhibit parabolic oxidation

kinetics. The scale thickness was measured on micrographs
from the long life RT31/RT33 and SDP2/KDP1 systems such
as those in Figures 12 and 13. Neglecting the transient
oxidation on heating and cooling, estimated parabolic rate
constants at 1100°C (in g2/cm4s) were determined based on
scale thickness considerations:

System Number AI2O3 scale KP

of cycles thickness (urn) (g2/cm4s)
RT31/RT33 710 7,5 8,4 xlO"1J
SDP2/KDP1 1080 5,5 3,9 xlO'13

These values are consistent with published data on parabolic

rate constants of bare alumina forming alloys27.

In the absence of a fine microstructural characterization of the

scales, it is difficult to account for this difference in oxidation Figure 14: SEM back-scattered cross-section images of
kinetics. It is tempting though to try to correlate the higher RT33 deposited on a LPPS MCrAlY bond coat after 1071
oxidation kinetics of the RT31 bond coat to the presence of cycles (a) and KDP1 after 1080 cycles (b)
yttrium rich oxide phases within the alumina scale. The
oxygen diffusivity of yttrium oxides is much higher than that The ceramic morphology however evolves dramatically with
of pure alumina. These yttrium rich oxides thus may act as ageing (see Figure 14 to be compared with Figure 4, both sets
fast oxygen diffusion paths through the scale to the bond of pictures taken at 50 micron height from the bond coat).
coat/oxide interface and create as many fast growing After more than 1000 cycles at 1100°C, sintering mechanisms
oxidation fronts. have resulted in the columns branching up and microwelding
together. The columnar structure is effectively partly lost.
Bond coat oxidation kinetics is one argument to account for Some loss of ceramic compliance could be expected from this
the superior spallation resistance of the SDP2/KDP1 system structural change potentially adding more stress to the TGO.
over RT31/RT33. The compressive stress within the oxide This effect combined with a potential strength reduction of
scale is maximum at room temperature. It results both from the ceramic/TGO interface with ageing may be responsible
the thermal expansion mismatch between oxide and bond for the crack propagation within the region of this interface
coat metal in addition to growth stresses. Both kinds of after long exposures. How the ceramic deposition process
stresses increase with scale thickness. Moreover, under a affects the integrity and stability of this interface is an open
given stress state, the probability for the scale to contain a question.
critical flaw size also increases with scale thickness.
CONCLUSIONS
Whether the difference observed in ceramic morphology will • Western and Eastern type TBC systems exhibit different
also affect the spallation resistance of TBC's is a complicated ceramic morphologies and TGO films.
issue. XRD patterns at room temperature suggest from the • Thermal cyclic lives were found to be very scattered for
absence of peak shifts that EB-PVD ceramic layers contain both SDP2/KDP1 and RT31/RT33.
negligible residual stresses (no more than a few hundred • Short lives were synonymous of lack of adhesion between
MPa). In another study28, the low residual stress level in as- TGO and bond coat metal.
deposited EB-PVD ceramics was verified within the coating • High life specimens failed in the ceramic/TGO interface
depth by polishing the ceramic layer. region.
• The SDP2/KDP1 exhibited overall superior spallation
resistance and this could be correlated to lower bond coat
oxidation kinetics than in the RT31/RT33 system.
 -10

Acknowledgements: The work concerning the study of nickel-base superalloy », Materials science and Engineering
Russian TBC's was supported by the SPAE.The authors A203 (1995) 388-398
gratefully acknowledge the contributions of E.Celerier and
M.Scher for the thermal cycling work as well as 16.D.V.Rigney, R.Viguie, D.J.Wortman, D.W.Skelly, «PVD
J.M.Duchemin and J.M.Thubert for the SEM work at thermal barrier coating applications and process development
Snecma. The X-ray diffraction studies were performed by for aircraft engines», TBC Workshop 1995, NASA
C.Diot at Onera. Fruitful contacts and discussions with Pr Conference publication 3312
Y.Tamarin at the VIAM Institute in Moscow were much
appreciated. Last but not least, the authors would like to 17.Yann Jaslier, «Development of EB-PVD TBC's: the role
thank Remy Mevrel at Onera for returning his useful of deposition temperature and plasma assistance», PhD
comments on the first draft of this paper. thesis, Cranfield University, 1995

18.U.Schulz, K.Fritscher, H.J.Ratzer-Scheibe, W.A.Kaysser,

References M.Peters, «4th Int.Symp.on high Temp. Corrosion &
1. W.K.Hainan, D.Lee, «EB-PVD process for coating gas rotection» Mat. Les Embiez, Fr.20-24/05 1996, To be
turbine airfoils », High temperature protective coatings published
112th AIME Annual meeting, Atlanta, Georgia, March 7-8,
1983 19. Snecma unpublished work

2. S.M.Meier, D.K.Gupta, K.D.Sheffler, Ceramic TBC's for 20. Maricocchi, Rigney, US Patent 5,350,599
commercial gas turbine engines, JOM, March 1992 Sep. 1994

3. K.D.Sheffler, D.K.Gupta, Current status and future trends 21.LLelait, S.Alperine, C.Diot «Microstructural
in turbine applications of TBC's, Gas Turbine Aeroengine investigations of EB-PVD TBC's », Journal de Physique IV,
Congress and Rxp., Amsterdam, Sep. 30 1988 ASME 88-GT- Colloque C9, supplement au journal de physique HI, vol 3,
286 dec 1993

4. Comments from W.Goward, D.Boone, W.Hainan, 22.M.G.Scott, «Phase relationships in the zirconia-yttria
TBC Workshop 1997, May 19-21, Cincinatti, Oh system », J.Mat.Sci, 10 (1975), pp.1527-1535

5. B.A.Movchan, A.V.Demchishin, « Study of the structure

and properties of thickvacuum condensates of nickel, 23.E.C.Subbarao, «Zirconia-an overview», Science and
titanium, tungsten, aluminium oxide and zirconium dioxide » technology of zirconia, Advances in ceramics, vol.3,
Fiz.metal.metalloved.28N°4, 653-660, 1969 A.H.Leuer, L.W.Hobbs, 1981

6. Y.Tamarin, Private communication 24.C.Leyens, K.Fritscher, R.Gehrling, M.Peters.W.A.Kaysser

« Oxide scale formation on an MCrAlY coating in various H2-
7. A.V.Shavkunov, Private communication H2O atmospheres», Surface and coatings technology 82
(1996) 133-144
8.S.Stecura, Optimisation of NiCrAl/Zr02-Y203Thermal
barrier system, NASA TM 86905, 1985 25.S.M.Meier, D.M.Nissley, K.D.Sheffler, T.A.Cruse,
«Thermal barrier coating life prediction model
9A.Feuerstein, W.Dietrich, H.Lammermann, «Advanced development »J.Eng.Gas.Turb.&Power, 114 (1992), 258-63
PVD overlay coating equipment for aircraft gas turbine
engine applications», Tokyo International Gas Turbine 26.P.Lamesle, «Revetements d'aluminiures modifies par le
Congress, 1987 palladiummecanismes de formation et comportement en
oxydation/corrosion ä haute temperature», These de
10.H.Lammermann, A.Feuerstein, «PVD overlay coating for Doctorat, Universite Henri Poincare, Nancy 1, 1995
blades and vanes of advanced aircraft engines», 1992,
Yokohama International Gas Turbine Congress 27. K.L.Luthra, C.L.Bryant, «Mechanisms of adhesion of
alumina on MCrAlY alloys» Oxid.Met. 26 (5/6) 397-416
ll.T.E.Strangman, »Development and performance of (1986)
physical vapor deposition thermal barrier coating systems »,
Proceedings of the 1987 Coatings for advanced heat engine 28. Onera Rapport technique 22/3729 MY
workshop m-63 to EI-71, July 87

12. T.E.Strangman , US Patent 4,321,311

1982

13. N.E.Ulion, D.L.Ruckle, US Patent 4,414, 249

1983

14.B.A.Movchan, « EB-PVD Technology in the gas turbine

industry:present and future », JOM November 1996

15.M.A.Smith, W.E.Frazier, BA.Pregger, «Effect of sulfur

on the cyclic oxidation behavior of a single crystalline,
 9-1

PROCESSING, CHARACTERISATION AND TESTING OF

EB-PVD THERMAL BARRIER COATINGS

W. A. Kaysser
M. Peters
K. Fritscher
U. Schulz

DLR
German Aerospace Research Establishment
Institute of Materials Research
D-51170 Köln
Germany

Summary engine is another design criterion, more precisely the

specific thrust as expressed in the thrust-to-weight
Thermal barrier coatings (TBCs) are increasingly ratio. Both, specific fuel consumption and thrust-to-
applied to hot components in gas turbines. Contrary weight ratio have been significantly improved over
to plasma spraying, the electron-beam physical va- the years accomplished through increased operating
pour deposition (EB-PVD) process offers the oppor- temperatures as well as improved structural effi-
tunity to generate coatings having a unique columnar ciency [2]. It is very obvious that here advanced ma-
microstructure. The main advantage of this structure terials play a major role. In fact, modern aero engines
is its superior tolerance against straining and ther- represent some of the most demanding and sophisti-
moshock, thus giving it a major edge in lifetime. Fur- cated applications for structural materials in any
thermore, cooling hole closure will be prevented and engineering system manufactured today. This has
the aerodynamic design of the blades will be main- been manifested by the steady increase in service
tained. temperature, product reliability and usage of light-
weight materials [2].
This paper will outline the advantages of EB-PVD for
the production of TBCs on rotating gas turbine com- But engines need not to be only reliable, economical
ponents like blades and vanes. The effect of EB- on fuel and profitable, they also have to be environ-
PVD processing parameters on the microstructural mentally acceptable. In terms of emission nitrogen
evolution and respective lifetimes of partially yttria oxides and carbon dioxide are of prime concern. NOx
stabilised zirconia (PYSZ) TBCs will highlight the is particularly a problem since emission increases
potential of the evaporation process. Alternative with increasing pressure ratio and process tempera-
stabilisers like Ce and La are looked at in terms of ture - both are obvious ways to increase engine per-
increase of application temperatures as well as life formance.
extention of the blades. An extended structural zone
diagram for PVD is proposed incorporating the influ- Blades and vanes of the high pressure turbine sec-
ence of substrate rotation on microstructural evolu- tion of aero engines are among the most highly
tion. Finally, the limits of evaporation processing will stressed parts in engineering components. Internally
be stressed. cooled aerofoils of state-of-the-art Ni-base superal-
loys operate at temperatures of about 1000°C with
short-term peaks yielding even 1100°C which is close
1. INTRODUCTION to 90% of the alloys' melting points. These tempera-
tures are maintained in service due to a highly so-
There is no single parameter to verify the competi- phisticated cooling technology by which however
tiveness of an aircraft. However, it is commonly ac- thermal energy is withdrawn from the aerofoils in the
cepted that direct operating cost (DOC) is a useful order of 1 MW/m2 thus reducing the overall fuel effi-
indication of the potential of a given type of aircraft, if ciency of the engine. The necessity of a close control
a new aircraft is introduced it has to show significant of materials temperatures can be expressed by the
DOC benefits of at least 5 to 10% compared to the simple rule that blade life on creep is halved for every
aircraft currently in service to make it an attractive 10 to 15CC increase in temperature [3].
proposition to an airline. For an aero engine manu-
facturer the DOC is primarily reduced by control of But further increases in thrust-to-weight ratio of ad-
fuel costs through weight savings, drag reduction and vanced aero engines will require even higher gas
specific fuel consumption. For a new engine a realis- turbine inlet temperatures. Figure 1 shows how these
tic picture would call for a 2-3% improvement in spe- temperatures have increased over the years. Today
cific fuel consumption and for about 5% in weight turbine inlet temperatures exceed 1400°C while
reductions [1]. 1760°C are aimed at within the next twenty years.
There is no doubt that this ambitious goal can only be
Apart from specific fuel consumption, the thrust of an met by usage of uneconomical^ extensive cooling

Paper presented at an AGARD SMP Meeting on "Thermal Barrier Coatings",

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
9-2

techniques or by advanced high temperature materi- lives by avoiding hot spot formation and subsequent
als [4] and in particular through the introduction of failure by thermal fatigue. These materials were
electron-beam physical vapour deposition (EB-PVD) broadly accepted for 20 years until their replacement
thermal barrier coatings (TBCs) [5]. by PYSZ (partially yttria-stabilised zirconia) coatings
was initiated. The APS (air-plasma sprayed) PYSZ
TBCs offer outstanding mechanical, chemical and
thermal properties. Two generations of high tem-
perature (>1000°C) TBCs were based on these
unique ceramics (generation I is APS MCrAlY / APS
PYSZ and generation II is LPPS (low-pressure
plasma sprayed) MCrAlY / APS PYSZ) which allowed
in the early eighties to introduce TBCs on highly
thermally loaded parts like vane platforms and vane
airfoils. The need for higher operating temperatures
in turbines, however, inspired material scientists for
half a decade to search for TBCs which could also be
applied on blade platforms and airfoils which are
200
1950 1960 1970 1980 1990 2000 2010 2020 subjected to high thermal and additional mechanical
loads.
Service Date of Gas Turbine Aero Engines

Figure 1: Increase of turbine inlet temperature of Here the electron-beam physical vapour deposition
aircraft gas turbine engines [2] (EB-PVD) technology offered the opportunity to gen-
erate TBCs with vastly superior strain tolerance. This
is due to their specific coating structure which grows
2. THERMAL BARRIER COATINGS from the vapour phase in a columnar form with indi-
vidual ceramic columns being weakly bonded to their
TBCs consist of thin ceramic layers of low thermal neighbour columns as e.g. shown in Figures 8a or
conductivity - typically partially stabilised zirconia - 13.
which are applied on aerofoil surfaces that just have
a metallic corrosion resistant coating. The coating Coating all these high performance parts with reliable
imparts good adhesion of the ceramic to the sub- TBCs would allow to alter the initial component de-
strate. Application of the TBCs enables increasing sign and to operate the TBC bearing components
engine performance/thrust by either increasing the with the coatings as an integral part of them. The
gas temperature or reducing the cooling airflow. exploitation of strain-tolerant TBCs initiated a renais-
Alternatively the lifetime of the turbine blades can be sance of industrial EB-PVD technology in 1987 which
extended by decreasing metal temperatures as sche- until then had been utilised for deposition of metallic
matically outlined in Figure 2 [6, 7]. MCrAIY-type coatings for 20 years. The novel EB-
PVD PYSZ ceramic layers with advanced LPPS
MCrAlY bond coat, designated now as generation III
TBC, offered a roughly 3 times improvement in blade
life or a surface temperature increase on the hot
sections of airfoils of approximately 150K.

4. PROCESSING OF TBCS

Plasma-sprayed (PS) TBCs have been widely ap-

plied to hot components like burner cans since the
sixties while in recent applications to more preten-
tious parts like turbine blades EB-PVD technology is
favoured. Contrary to plasma spraying, EB-PVD
processing offers the opportunity to generate coat-
ings having a unique columnar microstructure with 2
to 25 urn in diameter. The main advantage of this
structure is its superior tolerance against straining
Cooling Air Flow
and thermoshock, thus giving it a major edge in life-
Figure 2: TBCs allow increased engine performance time [8]. Furthermore, cooling hole closure will be
and/or life extension prevented and aerodynamic design of the blades is
maintained. In Table 1 the two processing methods
are compared as far as thermal barriers for turbine
3. HISTORIC DEVELOPMENTS IN TBCS blades are concerned.

Since the early sixties PS (plasma sprayed) calcia- During EB-PVD processing a high energy electron
and magnesia-stabilised zirconia TBCs have been beam melts and evaporates a ceramic source ingot in
used extensively as ceramic top coat on combustion a vacuum chamber. Ingots are bottom fed into the
chamber walls and on burner cans to prolong their crucibles during evaporation to ensure continuous
 9-3

TBC growth. To achieve defined stoichiometry of the the coating systems. Failure in TBC systems almost
zirconia a controlled amount of oxygen is bled into always occurs by TBC spallation due to stresses in
the deposition chamber. Preheated substrates are the ceramic or in the bond coat. But failure of EB-
positioned in the vapour cloud above where the va- PVD processed TBCs happens, unlike those for PS
pour is deposited on substrates at deposition rates of based TBCs, at the ceramic - bond coat interface.
100 to 250 nm/s [9]. Typical columnar microstruc- There is a thermally grown oxide (TGO) of a few urn
tures and aerodynamically smooth surfaces are ob- in thickness which plays a key role in the adhesion of
tained without the need for final polishing or condi- EB-PVD TBCs since it is considered to be the weak
tioning of cooling holes. Due to the columnar micro- link in the system. Its strength and adhesion to the
structure the life time of the TBCs is prolonged and bond coat governs the true location of failure whether
the damage tolerance improved. Typical characteris- first cracks will propagate within or along the TGO -
tics and major advantages of thermally sprayed and bond coat interface. Here spallation is initiated by
on-evaporated TBCs are schematically outlined in accommodation of stresses surmounting a critical
Figure 3 [10]. Figure 4 shows a selection of EB-PVD value.
TBCs on aero engine and stationary gas turbine
blades produced at DLR on semi-commercial single-
source 60kW Leybold and dual-source 150kW von
Ardenne EB-PVD coaters.

Table 1: Plasma sprayed (PS) versus on-evaporated

(EB-PVD) TBCs
PS EB-PVD
Surface Finish good but extra polish excellent
Bond Coat Roughness grit blasted smooth
Bonding Mechanism mechanical chemical
Alloy Flexibility high limited
Typical Thickness 0.2-3mm 0.1-0.3mm
Cooling Hole Closure poor excellent
Coating Source movable fixed
Large Parts predicable favourable costly
Parts per Charge 1 1-10
Investment Costs (%) 100 100-400

PS EB-PVD

TBC»-
BC »-
[HIM11 Figure 4: EB-PVD TBCs on turbine blades for use in
stationary gas turbines, civil aero engines and heli-
copter engines (decreasing in size)
Substrate

A way to minimise misfit stresses in TBC systems is

• horizontal splat structure • columnar coating structure to reduce the growth rate of TGOs. The early overlay
• 0.2-3 mm coating thickness • 0.1-0.3 mm coating thickness bond coatings which were EB-PVD MCrAlYs were
• mechanical bonding / • chemical bonding / followed by PS MCrAlY families. Advanced MCrAlYs
rough BC surface smooth BC surface contain additional elements like Si and/or Ta, Hf for
+ flexible coating chemistry + smooth coating surface providing lower scaling rates and better hot corrosion
+ low initial thermal conductivity + low E in coating plane resistance. These coatings with low vapour pressure
in coating thickness direction + cooling holes stay open elements (mostly refractory elements) are more eas-
+ improved life time ily deposited by LPPS techniques. TGOs grow more
+ superior erosion resistance slowly and thus effectively prolong the lifetime of the
TBCs on them.
Figure 3: A simplified comparison of properties for
plasma sprayed (PS) and evaporated (EB-PVD) Another route to minimise misfit stresses and im-
TBCs (schematic) prove spallation resistance is to make the bond
coatings more creep resistant which will bring more
stability to the TBC root area on thermomechanical
5. FAILURE MECHANISMS AND PREVENTION cycling. This change in mechanical characteristics is
successfully done in recent LPPS MCrAlY overlay
Pertinent usage of TBCs as a reliable component or coats by converting the former y/ß phase structure
future TBC applications as integral design elements into a y/y1 structure by appropriate alloying. Due to the
of highly loaded engine parts need a more compre- introduction of a significant portion of refractory ele-
hensive understanding of the failure mechanisms of
9-4

merits the modified high creep resistance phase 7. INFLUENCE OF STABILISERS ON TBC
structure offers further benefits which refer to a low PERFORMANCE
diffusivity and a reduced coefficient of thermal ex-
pansion (CTE). The latter property is of uppermost Partially yttria-stabilised zirconia (PYSZ) is the cur-
importance for the amount of residual stresses ac- rent state-of-the-art material for TBCs. Unfortunately,
cumulating in the TGO layer. They can be reduced if the material shows insufficient phase stability and
the differences between the CTEs of the respective accelerated sintering at temperatures above 1200°C.
ceramic and metallic partners of the whole system Therefore, further increases of gas inlet temperature
are minimised. require alternative stabilisers with improved phase
Finally the involvement of superalloy substrates in stability. Another argument for new stabilisers in
the interplay of the physical properties of the respec- zirconia or for completely new ceramics is the insuffi-
tive material components and their influence on the cient resistance of current TBCs against chemical
compositional behaviour of TBC systems has to be attack by pollutants in the combustion gas. Hot cor-
addressed. Ni-base superalloys have gained a high rosive decay of TBCs by Na2S04 and vanadates are
degree of technical maturity in the early nineties. reported which involves leaching out of stabilisers
They have passed a lot of development stages, e. g. from parent zirconia. The degradation occurs by
hardening by precipitated y phases, extensive solid transformation of high temperature phases to mono-
solution strengthening, directional solidification and clinic zirconia on cooling. This failure mode may be
finally single crystals to trust in more creep-resistant found mainly in heavy duty engines and off-shore
crystal orientations and continuous grain structures service [15].
[11]. The most creep-resistant crystal orientation
(001), however, provides the least Young's modulus One alternative stabiliser for zirconia is Ce02- The
of only 120 MPa. Obviously it is a drawback for the benefits of ceria-stabilized (CeSZ) TBCs are good
adhesion of TBCs as thin-walled airfoils can be bent corrosion resistance [16-17] and excellent phase
more rigorously under same loads in service than the stability at high temperature [18]. Furthermore, the
conventionally cast counterparts. Three generations thermal conductivity is found to be lower than for
of single crystal superalloy development, however, PYSZ and some benefits for lifetime and thermocyclic
resulted in a reduction of the CTE which compares resistance are reported, too. La203 is another candi-
with the recent low expansion bond coats.
date for replacing Y2O3 in zirconia.

6. SURFACE TRAETMENT/ROUGHNESS To get more insight into single-source EB-PVD proc-

essing of new compositions for TBCs a feasibility
Another key issue for utilising the full potential of study was performed on four differently stabilised
TBCs is .their surface finish. Increasing roughness zirconias (Table 2) using identical substrate alloys
results in higher aerodynamic losses as well as in- and EB-PVD NiCoCrAlY bond coat compositions [19].
creased heat transfer to the airfoil [12]. It also has an
impact on TBC adherence. So the roughness of the Table 2: Differently stabilised zirconias investigated
TBCs needs to be closely controlled.
ZrO, +
The processing steps before TBC deposition are PYSZ 6.5 wt% Y2O3
essential for attaining the targeted surface charac- FYSZ 20 wt% Y2O3
teristics of the final surface. For EB-PVD TBCs the
bond coats have to have a finish on mechanical sur- LaSZ 8 wt% La2Ü3
facing better than 1 to 1.5 urn Ra. Identical pilot sam- CeSZ 25 wt% Ce02 - 2.5 wt% Y2O3
ple parts are attached with thermocouples at repre-
sentative locations in order to have a means to cal- A columnar structure of the TBCs of about 250um
culate in analogy the temperature distribution over thickness was found for all four chemistries with
time within real parts during a real deposition process some noticeable differences between the various
as closely as possible. Optimal T-t histories for the ceramics. FYSZ and CeSZ posses a larger column
thermal cycle during preheating and TBC deposition diameter and a higher degree of ordering compared
have to be assured. The TBC deposition must be to PYSZ and LaSZ. On the other hand, LaSZ has the
performed within an operating window which takes most nonuniform shape of the terminal section of
care of a number of parameters, the most important columns. The microstructure of standard PYSZ lies
of which are [13] between these two extremes with more irregularities
than CeSZ and FYSZ but not as many as with LaSZ.
component temperature during coating
pressure in the coating chamber One possible explanation for the variations in mor-
oxygen partial pressure in the coating chamber phology are the differences of the respective homo-
part rotation rate during coating. logous temperatures TdepoSjtion / Tme|{ing which have
a strong relation to diverse microstructural zones
The final coating will then have a 1 to 1.5 urn Ra within common structural zone diagrams [20-21].
roughness [14]. Further reductions of roughness may Although the deposition temperature was nearly the
be advantageous. same for all four zirconia, their melting points differ
widely, thus resulting in different microstructures ac-
cording to the structure zone models mentioned
 9-5

above. Other effects that affect the microstructure Burner rig tests as well as furnace tests revealed that
include variations in phase composition, ion radii, and composition and phase structure of the variously
surface energy aspects during condensation [22]. stabilised zirconias are closely related to the cyclic
lifetime of the coating system. The only phase found
Phase analyses by X-ray diffraction (XRD) revealed in PYSZ was metastable non-transformable t'. The
that the compositions of the three binary TBCs, equilibrium phase diagram [23] predicts a two-phase
PYSZ, FYSZ, and LaSZ, were close to the ingot mixture consisting of t containing 4 % Y2O3 and c
compositions. For PYSZ exclusive tetragonal (f) containing 16 % Y2O3 at deposition temperature.
phase was identified while FYSZ only contained The c phase, however, can only be maintained at RT
cubic (c) phase and LaSZ exhibited a mixture of under the preconditions that a very low critical grain
mainly tetragonal, minor cubic, and substantial size of c is established and/or the c particles are
amounts of monoclinic (m) phase. In the case of the subjected to high compressional stresses by the
ternary composition Zr02-Ce02-Y2C>3, however, surrounding t matrix. Both mechanisms would cause
analysis showed no constant but fluctuating compo- the half widths of the XRD peaks to be broadened
sitions across the TBC thickness. Due to these varia- which, however, is not observed. Therefore, forma-
tions the surface content of ceria varied among test tion of the t' phase is considered to be caused via
pieces of different deposition runs between 13 and rapid quenching on PVD processing as can similarly
38 wt%. For CeSZ a mixture of cubic, tetragonal, and be assumed for solidifying TBCs during PS deposi-
occasionally monoclinic phases was found [22]. tion.

300- PYSZ
FYSZ
OJ250-
E, /V LaSZ
8 200- CeSZ^-
0
£ 150-
cn
Q)
g 100-

50-

01 k—O rO—-rO-—rO- 1— i
10 20 30 40 50 60 70
Time [h]

Figure 5: Weight loss of cyclically burner rig tested

samples versus time

These four differently stabilised TBCs were subjected

to cyclic burner rig testing at Mach 0.35 gas velocity b)
with cyclic heating up to 1150°C for 57 minutes and
forced air cooling to room temperature for 3 minutes.
The results are summarised in Figure 5. Rapid spal-
lation of FYSZ and LaSZ TBCs was observed as
characterised by high weight losses after short test-
ing times. PYSZ proved to be the most stable coating
system over long times, followed by sudden spalla-
tion of TBC.

CeSZ exhibited a different behaviour. Here a quasi-

continuous weight loss was observed after each
inspection cycle. The different failure mode of CeSZ
in comparison to "standard" failure by spallation of
PYSZ is visible in Figure 6: a step-wise degradation
in layers was observed instead of spallation of the
whole TBC in a single event as noticed in all other
cases. A thin layer of ceramic was still present on top
of the bond coat after the tests. Figure 6: SEM micrographs of burner rig tested TBCs
after 65hrs at 1150°C:
Considering the lifetime of the TBCs a similar ranking a) PYSZ (cross section); b) CeSZ (surface)
was also found in cyclic furnace tests between 150°C
and 1100°C. Here, however, for all stabilisers the
main failure location was found between the ther- A major drawback of PYSZ is the lack of high tem-
mally grown oxide (TGO) and the bond coat rather perature phase stability as supported by annealing
than between TGO and TBC. experiments. Whereas EB-PVD TBCs are stable at
9-6

temperatures up to 1150°C, they transfer after 100 the fluctuation of composition across the thickness of
hrs annealing at 1400°C to a mixture of 48% tetrago- the TBC. This may cause alternating layers of "weak"
nal + 48% cubic + 4% monoclinic (mol%) phases phases including monoclinic and "strong" phases like
[24]. The transformation into three phases suggests a t' [24]. Apparently, the coating will first break in
sequential mechanism consisting of yttrium cation weaker regions. Lattice misfits between the various
diffusion out of t' and destabilisation of t' into c and t layers may further contribute to reduced adhesion
with subsequent phase transformation to m during between the layers. Higher erosion rates of CeSZ
cooling. At temperatures below the stability point t' is found for PS TBCs [28, 31] may also account for this
still the most effective phase for durability in thermal failure mode. The partial loss of this TBC may act as
barrier coatings as can be seen in Figure 5. The t' a strain/stress relief mechanism that allows the re-
microstructure is characterised by outstanding bend- mainder part of the coating, which is thinner due to
ing strength, high crack propagation energy, high partial spallation, to adhere longer. CeSZ apparently
fracture toughness and highly tolerant thermoshock offers some potential in thermal barrier coatings es-
behaviour. Lattice distortion due to tetragonality, a pecially if a columnar microstructure can be utilised.
domain structure and the well-described tweed micro- Multiple source evaporation is however required to
structure inside the t' grains [25] are supposed to be bring about reproducible CeSZ TBCs of stable com-
responsible for the excellent performance of PYSZ. position.

The formation of a single cubic phase for FYSZ TBCs

is in accordance with all findings on EB-PVD as well 8. INFLUENCE OF MORPHOLOGY ON TBC
as PS TBCs known from literature. FYSZ suffers from PERFORMANCE
low thermal shock resistance and poor fracture prop-
erties of the equilibrium c phase. Earlier reports on The microstructure of EB-PVD layers is essentially
PS TBCs [26-27] have shown that FYSZ possesses influenced by four basic processes:
poor thermocyclic behaviour. However, some incon-
sistency in the literature on TBCs of this composition shadowing,
must be recognised. The present results support surface diffusion,
findings [28] that the columnar EB-PVD structure is volume diffusion and
not able to overcome the intrinsic problems of FYSZ desorption.
leading to poor integrity of such coatings after cyclic
testing (Figure 5). Their evolution is roughly predicted in structural zone
diagrams [20-21]. Rotation of substrates during
Monoclinic phases that undergo a phase transforma- deposition, however, is not regarded as an essential
tion during temperature variation will cause rapid contribution to the growth process so far. Neverthe-
spallation of EB-PVD TBCs. The volume change less, it causes an additional microstructural feature in
associated with the phase transformation m ' t cre- zirconia-based TBCs. A beaded or "c-shaped" struc-
ates high stresses. Even strain tolerant structures like ture of the columns is formed due to the continuous
columnar EB-PVD ones are unable to accommodate change of vapour impact angle and amount of vapour
these stresses on transformation. Early spallation of particles that adhere on the surface during each
LaSZ gives clear evidence for such a failure mecha- revolution [32-33]. Elevated substrate temperatures
nism. It has to be taken into account that high rate during EB-PVD cause a higher density of the TBCs
condensation from a vapour phase is not necessarily and a higher hardness as well [34]. In addition to
a process that stabilises equilibrium phases. There- stabiliser type and content the degree of ionisation of
fore, phase diagrams do not allow a safe prediction of the vapour cloud, gas pressure, surface roughness,
the phases of real coatings. deposition rate and vapour impact angle are other
parameters that influence the columnar microstruc-
CeSZ showed the most interesting behaviour. The ture of EB-PVD thermal barrier coatings [22].
condensing matter is evaporated from a ceramic
ingot source material at approximately 3500°C [29]. EB-PVD TBC production upon 3-D coated parts like
The presence of ceria considerably lowers the melt- vanes and blades necessitates substrate rotation
ing interval of the ingot. Selective processes are during deposition. Therefore, the influence of rota-
facilitated which cause the preferential evaporation of tional speed on microstructural evolution of the coat-
high vapour pressure components and enrichment of ings has been investigated in more detail. The differ-
low vapour pressure elements in the melting pool. ences in microstructure between TBCs deposited
Vapour pressures are 8-102 Pa for Zr02 and 9 -10s under two separate process parameter sets are illus-
Pa for CeC-2 at 3060°C [30] and thus differ more than trated in Figure 7. SEM pictures of the surface of
three orders of magnitude. This difference is probably TBCs display that microstructures of EB-PVD TBCs
even higher at the evaporation temperature which is strongly depend on process parameters, particularly
indeed too much for viable single-source EB-PVD substrate temperature and rotation.
processing. As a matter of fact the evaporation proc-
ess becomes instationary resulting in the composi- Aiming at customising microstructures of TBCs for
tional fluctuations observed. However, burner rig specific performance it is noteworthy that substrate
results of CeSZ were promising (Figure 5). Quasi- temperature and rotational speed are alternative
continuous weight loss and a degradation of the process parameters that cause the same micro-
CeSZ TBCs in thin layers were found but no spalla- structural features in the TBCs within certain limits. At
tion. One reason for a step-wise loss is, of course, low temperature and low rotational speed columns
 9-7

often vary in diameter from root to top or from one need of appropriate column features in particular
column to the other. Columns at the root section are TBC thickness zones. Several aspects may contrib-
much thinner than at the top and are enlarging dis- ute to the observed effects including surface tem-
continuously from root to top conically. Increasing perature fluctuations, change of diffusion kinetics and
both temperature and rotational speed improves the enlarged shadowing due to rotation. For instance,
regularity and parallelity of the microstructure and measurements of actual surface temperatures on flat
enlarges the column diameter. After competitive samples during deposition gave temperature differ-
selection of favoured crystal orientations in the thin ences of 60 to 80°C for a low rotational speed com-
root area during the first stages of coating growth, pared to 15 to 25°C for a three times higher speed.
columns start immediately to grow in width to their
terminal diameter. Although the coating density
measured is higher for high temperature/high rota-
tional speed TBCs than in low temperature deposited
TBCs, the columns appear to be less densely
packed.
30 min-'

12 min'

\AAA/

0.4 Tm 0.46 Tm 0.55 Tm

Homologous Temperature

Figure 8: Schematic representation of the influence of

substrate temperature and rotational speed on co-
lumnar microstructure evolution of EB-PVD TBCs

These observations suggest that structural zone

diagrams should be modified when substrates are
rotated during deposition. It is therefore proposed to
incorporate a second axis in the models as a further
degree of freedom similar to sputtered coatings
where e.g. the role of argon pressure was consid-
ered. To give a first idea of a potential model exten-
sion, the observed microstructures for the interval of
homologous temperature investigated are summa-
rised in Figure 8 [35].

9. INFLUENCE OF DEPOSITION ANGLE ON TBC

STRUCTURE
Figure 7: Surface of YPSZ TBCs: a)980°C; 12rpm;
b) 1050°C;30rpm Another processing parameter of interest for EB-PVD
component coating is the angle under which the
vapour cloud hits the substrate. This is of particular
The microstructure can be varied without altering the practical importance since usually substates like tur-
deposition temperature by different rotational speeds. bine vanes and blades irregularly shaped.
At low rotational speed the beaded structure is
formed within the columns while it disappears imme- First it was shown that for both single source and
diately after changing to a higher speed. An increase dual source evaporation morphology and texture of
of column diameter was also found. PYSZ TBCs produced under a perpendicular vapour
incline angle were nearly identical for [35]. In all
This processing tool allows to manufacture micro- cases a {100} plane nearly parallel to the surface was
structurally graded TBCs that can be adapted to the found, an example of which is given in Figure 9 [36].
9-8

The good performance of TBCs on highly loaded In hot corrosion testing of EB-PVD TBCs no evidence
parts in severest environments, achieved in less than was found for hot corrosion of the bond coat and for a
a decade of development, is a good indication for the chemical reaction between the TBC and corrosion
versatility of the columnar structure concept of TBCs. accelerators like molten salts and sulphur com-
The unfavourable crystallographic relations between pounds [37, 38]. However, TBC failure can be initi-
the columns and the direction of highest straining in ated by mechanical attack of solidified salt com-
service, hypothetically addressed before, stand for a pounds. Spallation of small areas of TBC in a funnel-
positive view into the future of TBCs: they still bear a shaped manner with some thinner coating still adher-
potential for further improvements. ent on the bondcoat indicate this failure mechanism.
For columns that posses larger voids between the
interior column faces, salts may penetrate deeper
11. STRUCTURE PROPERTY RELATIONSHIPS into the coating than in denser columnar microstruc-
tures. Consequently, the weight loss is higher for a
Some representative microstructures of TBCs pro- coarser microstructure deposited at higher rotational
duced by single source evaporation were chosen for speed compared to a microstructure deposited at
cyclic furnace testing (Figure 12). All samples had lower temperature and rotational speed [37].
the same EB-PVD NiCoCrAlY bond coat and surface
600
treatments. Samples with a microstructure consisting
of tapered and discontinuous columns due to multiple ^- SDallation
500-
branching on growth exhibit the longest cyclic lifetime •■**&■; i»
(12rpm/1020°C). TBCs with more regular columns of
400- -«.-.
larger diameters appear less tolerant against thermo-
cyclic loading. Burner rig tests confirm these results. (/)
| 300
Obviously, microstructure of TBCs and thermocyclic >.
lifetime are closely interlinked. Figure 12 evidences Ü
200 - ws*$$
the superiority of TBCs with a non-regular micro-
structure consisting of columns with variable diame-
ters over TBC with large uniform column diameters
100 SÜ
and a regular microstructure.
0
12/1020 30/920 30/1060
Revolutions [RPM] /Temperature [°C]
Figure 12: Cyclic lifetime in furnace test between
1100°Cand130°C

OlV j |§; 01o' 200pole figure 12. OUTLOOK AND CONCLUSION

To exploit the full potential of TBCs on highly

stressed rotating turbine components further work
has in particular to be directed to efforts which de-
(001} termine critical stresses more accurately based on
detailed understanding of failure mechanisms of
entire TBC systems with more confidence. The prog-
ress in the manufacture of more reliable TBC sys-
tems on single crystal materials will address the de-
sign of conveniently microstructured TBCs with supe-
(100) rior strain tolerance on advanced bond coats with
predictable TGO formation. The tailoring of respec-
tive microstructures to differing substrates with regard
to optimal adaption of their "porosity" and of the
*(100) thermal expansion mismatch has to be taken into
account. Here the disposal of appropriate EB-PVD
Figure 11: Crystallographic orientation of single technologies will enable to manufacture unique mi-
crystal blades and EB-PVD TBC crostructures for service-tolerant TBCs.

Future applications of TBCs aim at surface tempera-

Changes in stress state, elastic moduli, and adhesion tures of 1250°C and above where aero engines as
between contacting columns that possess high num- well as industrial gas turbines will operate. Today's
bers of protrusions may be responsible for this differ- state-of-the-art PYSZ ceramics, however, exhibit
ent behaviour. On the other hand, in the regular mi- destabilisation of the tetragonal t' phase to monoclinic
crostructure, crack propagation paths were provided and cubic on extended exposure at temperatures
along the parallel columns that cover the whole above 1150°C, and sintering phenomena become
thickness of the TBC. Indeed early crack formation predominant. Alternative ceramics will be needed
was found for these microstructures. with reduced sintering rates, improved phase stability
and lower thermal conductivity. The need for ad-
 9-9

In Figure 10 two EB-PVD PYSZ coatings are com- This question in mind the habit morphology and ori-
pared with vapour incline angles of 25-40° (Figure entation of a TBC will be detailed first. EB-PVD TBCs
10a) and < 10° (Figure 10b) deposited under con- form a preferred {100} texture not only in normal but
stant rotating conditions. It is obvious that the colum- also in in-plain direction which will establish along the
nar structure closely follows the incline angle of the main axis of rotation and perpendicular to it (Figure
vapour cloud to the substrate surface. This shows 11). This axis usually coincides with the main axis of
that the angle of the columns is directly related to the a turbine blade which also is most heavily subjected
vapour incline angle, which is, however, different for to mechanical stressing in service. In essence, the in-
the crystallographic orientation. Texture measure- plain {100} texture orientation of TBCs is the most
ments have revealed that the deposition conditions heavily strained one. Any other crystal orientation
clearly influence the type of texture of EB-PVD TBCs should be better. A look over a TBC in lateral direc-
[36]. For rotated substrates a four-fold symmetry with tion (Figure 7b) clearly shows the longitudinal and
{100} planes nearly parallel to the surface evolves often diamond-shaped cross section of crystals which
independent of vapour incline angle, as indicated in spread with their elongated main diagonals in {100}
Figure 10. The technical importance of textures in direction. So less stress-depressant boundaries will
TBCs is related from the fact that the associated accumulate in this direction compared to all others.
anisotropy in elastic moduli as well as thermal ex-
pansion within the t' phase may very well have a
profound influence on the lifetime of the coating sys- (113) (100)
tems.

Figure 9: Characteristic texture of EB-PVD processed

PYSZ TBCs indicated by {111} pole figure [36]

10. SINGLE CRYSTAL SUBSTRATES

(100)
As discussed before, the'columns of EB-PVD TBCs
exhibit a preferred {100} texture in perpendicular
direction to the substrate. How does this reflect on
advanced substrate material which more and more
tends to be of single crystal nature? The technical
advantage may be twofold, namely a slightly higher
•Wrtfrommw
tV V -M • h "f'< wf
-.-= >
thermal expansion (CTE) in in-plane direction for a M.t m'iimM A ml
better fitting to the bond coat and a hypothetical^
'4 4
* i i!
\Y\
lower Young's modulus (note: the Young's modulus is t

highest in {100} orientation for zirconia monocrystals)

which provides more stress relief at the interface on 1 .; j
mechanical loading.

Some controversy may arise on the appropriate defi-

,:
nition of the "right" Young's modulus which becomes
effective as a stress depressant in in-plain direction.
Is the effective Young's modulus the intrinsic physical
J;^'lJlipl«irfvwr
property of an individual crystal column, or is it the -
much lower - overall Young's modulus of the whole
stack of TBC constituents composed of columnar and
wedge shaped crystals, grain boundaries, loosely
adherent interfaces and some free space e.g. at Figure 10: PYSZ EB-PVD coatings deposited under
abandoned crystal positions? vapour incline angles of 25-40° (a) and < 10° (b)
9-10

H. Schurmann and W.-D. Zimmermann.

14. REFERENCES

[I] S.J. Swadling, An Airframe Manufacturer's View

on the Development of Aero Engines, Proc.
European Propulsion Forum, 1-13, Royal Aero-
nautical Soc, London, 1993
[2] J.C. Williams, The Development of Advanced
Gas Turbines: the Technical and Economic En-
vironment, in: D. Coutsouradis et al. (eds.),
Materials for Advanced Power Engineering, II,
1831-1846, Kluwer Acad. Publishers, Dor-
drecht/Boston/London, 1994
[3] G.F. Harrison, Proc. European Propulsion Fo-
rum, (London, UK: The Royal Aeronautical So-
ciety, 1993)3.1-3.16
[4] K. Fritscher, M. Peters, HJ. Rätzer-Scheibe, U.
Schulz, Superalloys and Coatings, in: H. Buhl
(Editor), Advanced Aerospace Materials, 84-
Figure 13: DLR's 150 kW dual-source EB-PVD labo- 107, Springer, Berlin/Heidelberg/New
ratory coater (von Ardenne Anlagentechnik) for the York/London/Paris/Tokyo/Hong Kong, 1992
production of novel TBCs and bond coats on aeroen- [5] J.T. DeMasi-Marcin and D.K. Gupta, Protective
gine and stationary gas turbine components Coatings in the Gas Turbine Engine, Surface
and Coatings Technology, 68/69 (1994) 1-9
[6] M. Peters, K. Fritscher, G. Staniek, W.A.
vanced ceramic materials will force EB-PVD proc- Kaysser, U. Schulz: "Design and Properties of
essing development to overcome pertinent materials Thermal Barrier Coatings for Advanced Turbine
restrictions. Multiple-source high rate coaters will be Engines", Materialwissenschaft und Werkstoff-
a valuable tool to enable the production of TBCs technik 28 (1997) 357-362
which are composed of low and high vapour pressure [7] J.T. DeMasi-Marcin and D.K. Gupta, "Protective
components (Figure 13). Coatings in the Gas Turbine Engine", Surface
and Coatings Technology, 68/69 (1994), pp.1-9
Advanced electron beam gun design allows the use [8] S.M. Meier, D.M. Nissley, K.D. Sheffler, T.A.
of only one gun in combination with jumping electron Cruse: "Thermal barrier coating life prediction
beam technology for dual source evaporation [39]. model development", ASMEJ. Eng. Gas Tur-
Fast to and fro movements of the electron beam from bine Power 114 (1992) 258-263
one pool to another and provides a quasi-continuous [9] B.A. Movchan, EB-PVD Technology in the Gas
evaporation from multiple sources. Such equipment Turbine Industry: Present and Future, JOM 48
also allows to produce advanced bond coat composi- (11/1996)40-45
tions which may attribute their share to the manufac- [10] M. Peters, G. Staniek, K. Fritscher, U. Schulz,
ture of safer TBC systems. W.A. Kaysser, "Effective Thermal Barrier Coa-
tings for Modern Turbine Engine Design", Proc.
New microstructures encompass compositionally 1997 European Propulsion Forum, 179-188, Ber-
graded, density graded and multilayered arrange- lin, 1997
ments where, especially in the last case, a reduced [II] G.L. Erickson, A New Third Generation Single-
heat conduction may allow the application of thinner Crystal Casting Superalloy, JOM 47 (4/1995)
overlays. CVD techniques can attribute their particu- 36-39
lar thin layer virtues. Basic and applied research [12] F. Soechting, NASA Conf. Publ. 3312(1995)1
capabilities have to focus on alternative materials [13] D.V. Rigney, R. Viguie, D.J. Wortman, D.W.
and processing routes while earnestly guarding all Skelly, "PVD Thermal Barrier Coating Applica-
cost requirements. Non-destructive testing and life- tions and Process Development for Aircraft En-
prediction methodologies for TBC systems have to be gines", Proc. Thermal Barrier Coating Work-
furnished. Finally, research and development efforts shop, NASA Conf. Publ. 3312, pp. 135-150,
have to be combined to fully exploit the temperature Cleveland, OH, USA, 1995
potential of TBCs in aeroengine gas turbines and at [14] A. Maricocchi, A. Bartz, D. Wortman, "PVD TBC
the same time make TBCs mature for extensive us- Experience on GE Aircraft Engines, Proc.
age for rotating components also in stationary gas Thermal Barrier Coating Workshop, NASA Conf.
turbines for energy conversion purposes [6]. Publ. 3312, pp.79-89, Cleveland, OH, USA,
1995
[15] K. Fritscher, M. Peters, H.-J. Rätzer-Scheibe,
13. ACKNOWLEDGEMENTS and U. Schulz: "Superalloys and Coatings", Ad-
vanced Aerospace Materials, ed. H. Buhl,
The authors highly acknowledge the experimental (Berlin, Heidelberg, Germany: Springer-Verlag,
assistance given by J. Brien, C. Kröder, H. Mangers, 1992), pp. 84-107
 9-11

[16] B.A. Nagaraj, A.F. Maricocchi, D.J. Wortman, Coatings", Surface and Coatings Technology,
J.S. Patton, and R.L.Clarke, "Hot Corrosion Re- 50(1992)141-149
sistance of Thermal Barrier Coatings", ASME [32] R.E. Demaray, W.K. Hainan, and S. Shen,
paper 92-GT-44 (1992) "Development of Electron Beam Physical Vapor
[17] B.A. Nagaraj and D.J. Wortman, "Burner Rig Deposition of Ceramic Coatings", Proc. 2nd
Evaluation of Ceramic Coatings With Vanadium- Conf. on Advanced Materials for Alternative
Contaminated Fuels", ASMEJ. Eng. Gas Tur- Fuel-Capable Heat Engines, (Palo Alto, CA:
bine Power112 (1990) 536-542 Electric Power Research Institute 1981), pp.
[18] P. Vincenzini, G. Appiano, F. Brossa, and S. 151-184
Meriani, "Stability of Thermal Barrier Coatings", [33] D.V. Rigney, R. Viguie, D.J Wortman, and D.W.
Proc. 3rd Int. Symp. "Ceramic Materials and Skelly, "PVD Thermal Barrier Coating Applica-
Components for Engines", ed. V.J. Tennery tions and Process Development for Aircraft En-
(1989), pp. 201-210 gines", NASA Conference 3312 (1995) pp. 135-
[19] U. Schulz, K. Fritscher, M. Peters, "Thermocyclic 149
Behavior of Variously Stabilized EB-PVD Thermal [34] K. Fritscher and U. Schulz, "Burner-Rig Perfor-
Barrier Coatings", Journal of Engineering for Gas mance of Density-Graded EB-PVD Processed
Turbines and Power, 119 (1997) in print Thermal Barrier Coatings" in Ceramic Coatings,
[20] J.A. Thornton, "Influence of Substrate Tempe- ed. K.Kokini, ASME MD-Vol.44 (1993), New
rature and Deposition Rate on Structure of York, pp. 1 -8
Thick Sputtered Cu Coatings", J. Vac. Sei. [35] U. Schulz, K. Fritscher, C. Leyens, M. Peters,
Technol., 12 (1975) pp. 830-835 W.A. Kaysser, Thermocyclic Behavior of Differ-
[21] B.A. Movchan and A.V. Demchishin, "Study of ently Stabilized and Structured EB-PVD Ther-
the Structure and Properties of Thick Vacuum mal Barrier Coatings, Materialwissenschaft und
Condensates of Nickel, Titanium, Tungsten, Werkstofftechnik 28 (1997) 370-376
Aluminium Oxide and Zirconium Dioxide", Fiz. [36] U. Schulz, H. Oettel, W. Bunk, "Texture of EB-
Metal. Metalloved., 28 (1969), pp. 83-90 PVD thermal barrier coatings under variable
[22] U. Schulz, K. Fritscher, and M. Peters, "EB-PVD deposition conditions", Zeitschrift für Metallkun-
Y2O3 and Ce02/Y2C>3 Stabilized Zirconia de 87 (1996) 488-492
Thermal Barrier Coatings - Crystal Habit and [37] U. Schulz, K. Fritscher, H.-J. Rätzer-Scheibe,
Phase Composition", Surface and Coatings M. Peters, and W.A. Kaysser, "Thermocyclic
Technology, 82 (1996) 259-269 Behaviour of Microstructurally Modified EB-
[23] H.G. Scott, "Phase Relationships in the Zirco- PVD Thermal Barrier Coatings", Materials Sci-
nia- Yttria System. J. Mat. Sei., 10 (1975) 1527- ence Forum, in press
1535 [38] B.A. Nagaraj, A.F. Maricocchi, D.J. Wortman,
[24] U. Schulz, K. Fritscher, and M. Peters, J.S. Patton, R.L Clarke, "Hot corrosion resis-
"Thermocyclic Behavior of Variously Stabilized tance of thermal barrier coatings", ASME paper
EB-PVD Thermal Barrier Coatings", ASME 92-GT-44(1992)
Turbo Expo, 10-13.6.1996 Birmingham, ASME- [39] P. Lenk, J. Senf, B.-D. Wenzel:" Von Ardenne
paper 96-GT-488 guns and their suitability for PVD processes for
[25] L. Lelait, S. Alperine, C. Diot, and R. Mevrel, thermal barrier coating", Proc. conf. on electron
"Thermal Barrier Coatings: Microstructural Inve- beam melting and refining-state of the art, 1996;
stigation after Annealing", Mater. Sei. Eng., ed. R. Bakish, Bakish Materials Corporation,
A121 (1989)475-482 Engelwood, USA, 182-190
[26] R.A. Miller and C.C. Berndt, "Performance of
Thermal Barrier Coatings in High Heat Flux En-
vironments", Thin Solid Films, 119 (1984) 195-
202
[27] S. Stecura, "Effects of Compositional Changes
on the Performance of a Thermal Barrier Coat-
ing System", NASA Technical Memorandum
78976(1979)
[28] N.P. Anderson and K.D. Sheffler, "Development
of Strain Tolerant Thermal Barrier Coating Sys-
tems", NASA Contract NAS3-22548, Report No.
NASA-CR-168251 (1983)
[29] O. Unal, T.E. Mitchell, and A.H. Heuer,
"Microstructures of Y203-Stabilized ZrÜ2 Elec-
tron Beam-Physical Vapor Deposition Coatings
on Ni-Base Superalloys", J. Am. Ceram. Soc,
77(1994)984-992
[30] N.S. Jacobson, "Thermodynamic Properties of
Some Metal Oxide-Zirconia Systems", NASA
TM 102351 (1989), pp. 1-63
[31] R. Taylor, J.R. Brandon, and P. Morrell,
"Microstructure, Composition and Property Re-
lationships of Plasma-Sprayed Thermal Barrier
 11-1

PLASMA ASSISTED CVD OF

THICK YTTRIA PARTIALLY STABILIZED ZIRCONIA COATINGS

S. Chevillard
S. Drawin
M.H. Vidal-Setif

ONERA - Materials Science Department

BP72
92322 CHATILLON CEDEX
FRANCE

SUMMARY PACVD) [5], electrochemical vapor deposition

(EVD) [6], combustion CVD (CCVD) [7],
Plasma-assisted chemical vapor deposition has been
• physical vapor deposition (PVD): sputtering [8],
successfully used to deposit thick (20 - 150 urn) zirconia
laser PVD [9], electron beam PVD (EBPVD) [10],
- (5-20wt%) yttria coatings on metallic and alumina
and
substrates, with high rates (up to 150 urn/h) at
• plasma spraying [11],
temperatures ranging from 400 to 750°C. Tetrachloride
(ZrCl4) and Y(thd)3 ß-diketonate were used as zirconium for numerous applications : TBCs for gas turbine engines
and yttrium source materials. Vaporization conditions [12, 13], stationary gas turbines [14] and diesel engines
have been studied for both precursors. Morphology, [15], solid electrolytes for fuel cells [16], gas sensors [5],
impurity content and crystallographic structure of the insulating layers for semiconductors, buffer layers for
coatings were investigated. At all deposition YBaCuO supraconductors, hard coatings, optical
temperatures, the coatings exhibited columnar coatings, etc. But few techniques are suited for thick (150
morphology and {100} crystallographic texture. At low to 300 urn) coatings, which require high deposition rates,
substrate temperature (400°C), the pore volume fraction as needed for TBCs.
was high (ca. 50%) and the coatings incorporated some Plasma spraying and EBPVD are currently the only
carbon and chlorine impurities, but at high temperature processes used in production for TBCs [17, 18]. Yet
(750°C), the pore fraction reduced to about 25% and little being typically line-of-sight processes, they are limited to
carbon and no chlorine were detected. Metastable the coating of simple-shaped airfoils, and deposition on
tetragonal «t' » phase was the main constituent of shadowed regions of multiple airfoils may be very
coatings with 7 and 10wt% yttria. Annealing for 24 and difficult [19]. A non-directive alternative deposition
120 hours of Zr02 - 7wt% Y203 coating deposited at process would present definite advantages in this respect.
400°C showed that cubic phase with an yttria content of
about llwt% is allowed to form, at the expense of the CVD techniques are in that sense well fitted. EVD allows
«t' » phase whose yttria content diminishes to 4wt%. deposition rates as high as 100 um/h [16] but requires
This relatively fast evolution towards thermodynamic porous substrates, and too low deposition rates (typically
equilibrium may be related to the high porosity level. 1 um/h) exclude thermal CVD. Plasma activation, at
either radio (13.56 MHz) or microwave (2.45 GHz)
1. INTRODUCTION frequencies, has been widely used for the deposition of
The use of thermal barrier coatings (TBCs) cannot be various layers in microelectronics, hard coatings and
currently by-passed for the enhancement of both oxide diffusion barriers [20]. It presents two benefits :
aeronautical and industrial gas turbine performance and lowering of the substrate temperature, in some cases
efficiency. Zirconia-based compositions, and especially down to less than 100°C, and enhancement of the
6-8wt% yttria partially stabilized zirconia, are widely deposition rate by up to two orders of magnitude.
used, because of an outstanding behaviour at high Moreover, PECVD processes allow to vary the coating
temperature (high chemical and structural stability, high morphology ; in the particular case of zirconia-based
coefficient of thermal expansion, low thermal coatings columnar morphologies can be obtained, as
conductivity) and a good resistance to thermal cycling already shown by Cao [5] for zirconia-yttria solid
[1]. solutions and Bertrand [21, 22] and Gavillet [23] for pure
Numerous depositions techniques of zirconia-based films zirconia. Compared to the lamellar morphology of
are reported in the literature, including : plasma-sprayed coatings, the columnar morphology is
• sol-gel processing [2], generally preferred for cycled applications at high
• chemical vapor deposition (CVD): conventional temperature as in gas turbine engines [19]. The improved
thermal CVD [3], metal-organic CVD (MOCVD) [4], strain tolerance is due to the orientation of the
plasma-enhanced (or assisted) CVD (PECVD or intercolumnar porosity perpendicular to the

Paper presented at an AGARD SMP Meeting on "Thermal Barrier Coatings"

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
11-2

substrate/coating interface, thus allowing contraction and thermocouples

expansion without cracking.
This paper presents the work in progress at ONERA to
develop a PECVD process for the deposition of yttria <J=JAr
partially stabilized zirconia thermal barrier coatings. Y(thd)3
Electrodeless microwave plasma generation has been powder
preferred to radio-frequency, because of the potentially
higher electron densities involved (1017 - 1018 m"3) and
thus higher active species densities and higher deposition
rates.
2. EXPERIMENTAL DETAILS
All deposits were performed in a home-built laboratory-
scale reactor based on the apparatus used by Bertrand
[21, 22]. The reactor can schematically be decomposed
into four parts (Figure 1) [24]:
• the gas generation and distribution system allows the
vaporization of the zirconium and yttrium solid
precursors and their transport, together with argon
carrier gas and oxygen, to the deposition chamber;
• the deposition chamber is a 100 mm in diameter, 280
mm long quartz tube. It is surrounded by a
microwave cavity (180 mm in diameter) coupled
through a waveguide to a 1200 W maximum power
2.45 GHz microwave generator ;
• the substrate (20 mm in diameter, 2 mm thick) is
placed on a temperature-controlled holder resistively
heated by a Ni-Cr wire connected to a 600 VA power
supply. The substrate temperature, Ts, can reach a
maximum of 1000°C for short durations, but this has
to be reduced to 750°C for longer durations due to the
rapid oxidation of the Ni-Cr wire at high temperature
and low pressure. The distance from the injection
nozzle to the substrate can be varied. Substrates are
metallic (NiCoCrAlY, preoxidized at 1100°C to allow
the growth of an adherent alumina film) or ceramic
(sintered a-alumina); a hole is drilled on the sample
side so that a thermocouple can be introduced to
Figure 1 - Apparatus for PECVD of Zr02 Y203
monitor the sample temperature ; thermal barrier coatings.
• the gases are evacuated by a Roots blower (Alcatel,
150 m3 h"1) connected to a primary pump (Alcatel, 33
m3rr). Corrosive gases are condensed in a liquid of them [24, 25]. The main selection criteria were :
nitrogen trap. The total pressure in the deposition • a high vapor pressure to ensure high deposition rates
chamber is controlled by a throttle valve and a (this requirement is less stringent for the Y precusor,
Baratron gauge. because the aimed Y:Zr atomic concentration ratio in
Airtightness is obtained using conventional Viton® O- the coatings is always less than 1:5);
ring seals, except for the gas generation part which is • a good chemical and thermal stability ;
heated up to 250°C where Kalrez® seals were used. All • low toxicity, low moisture sensitivity, low cost.
metallic parts are made of AISI 316 L stainless steel. Zirconium tetrachloride (ZrCl4) and Y(02C,,H|g)3
Experimental conditions for the deposition of Zr02-Y203 (tris(2,2,6,6-tetramethyl-3,5-heptanedionate)yttrium or
TBC coatings are given in Table 1. « Y(thd)3 ») have been finally selected.

2.1 Zr and Y source materials ZrCl4 powder is commercially available at high purity
levels (more than 99.5%) and presents high vapor
Since neither mineral nor metal-organic zirconium and pressures (315 to 4000 Pa) in the range 200-250°C. This
yttrium compounds are gaseous at room temperature, allows the design of a vaporization furnace working at
condensed source materials have to be used and relatively low temperature. In practice, a porous silica
vaporized in a specific manner (bubbler, furnace, etc). crucible is half filled with ZrCl4 and placed in the
The precursors for Zr and Y were chosen from literature temperature-controlled furnace (70 cm3 inner volume).
data and specific investigations were conducted on some Argon carrier gas is then forced through the powder and
 11-3

the ZrCl4-enriched gas flows to the mixing chamber. This deposition experiment and is continuously kept under
percolation technique ensures a good solid to gas mass vacuum. No excessive corrosion is noticed after three
transfer. years operation.
The metal-organic compound Y(thd)3 is available The whole gas generation part, from the furnaces
commercially at slightly lower purity levels (more than downwards to the injection nozzle, is heated at 250°C to
98%). It melts at 174°C and presents a high enough avoid condensation of the gaseous precursors.
vapor pressure at temperatures greater than ca. 160°C. A
2.2 Gas injection
silica boat with a small amount of Y(thd)3 is placed in a
second temperature-controlled furnace. Argon circulates The injection of the precursor gases into the deposition
over the powder surface, which is pasty at the selected chamber is of primary importance as coatings with
temperatures. The Y(thd)3 + Ar gas mixture is then homogeneous thickness are required. This is especially
introduced into the mixing chamber. the case if real complex-shaped turbine blades are to be
coated. But the present laboratory-scale deposition
Oxygen and argon are added to the gas flow in the
reactor has been designed for investigations on small
mixing chamber. Argon is used as an inert carrier gas but
samples, so that this requirement is less severe.
also to facilitate the ignition of the plasma, to improve its
Nevertheless, both to get insight in the fluid dynamics
stability and to adjust the partial and total pressures in the
around the sample and to prepare an up-scaling of the
deposition chamber.
reactor, flow simulations in the deposition chamber have
Only a few grams of Zr and Y precursors are vaporized at been performed [24].
each deposition experiment, as given by crucible mass
The small cylindrical nozzle used by Bertrand [22] has
measurements. This corresponds, in the standard
been replaced by a conical convergent-divergent nozzle
conditions (see Table 1), to approximate mass loss rates
as shown on Figure 2. The nozzle has been designed
of 1.4 g/h and 0.25 g/h and gaseous precursor mole
using one-dimensional isentropic flow relations [26] so
fractions in the mixing chamber of about 2 10"3 and
that a shock wave is located in the divergent part. The
1.5 10"4, respectively. After deposition, the crucibles with
flow at the nozzle exit remains thus subsonic in a wide
the remaining powder are removed from the furnaces,
range of experimental conditions. The flow properties at
after having cooled down to room temperature, and
stored until the next utilisation. Vaporization conditions that location were then used as input boundary conditions
for both furnaces were investigated, and charts of for finite element calculations [27] in the deposition
chamber, neglecting the presence of Zr and Y precursors
precursor mass loss vs. furnace temperature and argon
flow rate were drawn. (less than 1 mol%), as illustrated on Figure 3. For some
calculations, the heating of the gas by the microwave
Although both precursors are moisture sensitive, they can plasma was taken into account, but the numerous
be stored for several months in appropriate dessicators, chemical reactions induced by the plasma were up to now
without significant variation of their properties. To neglected. These simulations will be used to optimize the
prevent corrosion of metallic parts by hydrochloric acid deposition process, with respect to both deposition
(produced by the reaction of moisture with chlorides), the efficiency and coating homogeneity on complex-shaped
whole apparatus is flushed with argon after each substrates.
Table 1 - Experimental conditions for the deposition of
Zr02-Y203 TBC coatings.
Standard
Parameter Range
conditions
02 flow rate (seem) 25 - 500 200
Ar total flow rate (seem) 50 - 1400 800
for Zr furnace (seem) 20 - 700 100
for Y furnace (seem) 10-140 50
Zr furnace temperature (°C) 25 - 250 210
Y furnace temperature (°C) 25 - 250 170
Total pressure in deposition
chamber (Pa) 13-1330 133
Microwave input power (W) 0 - 1200 750
Substrate temperature (°C) 300 - 750 400
Nozzle exit - substrate
distance (mm) 25-90 50
Deposition duration (h) 0.25 - 3 1
Figure 2 - Schematic view of the gas injection nozzle.
11-4

XJ
B

§■

Figure 4 - SEM cross-sectional view of a 2rÖ2 -

substrate 5wt% Y2O3 coating deposited in the standard conditions
(deposition duration : 1 hour).
-T--r--|—r-r-r-r-
Figure 3 - Simulation of the gas flow in the deposition 150
♦ Ts= 400°C
chamber. P = 130 Pa, total flow rate : 1000 seem (80%
A Ts= 600°C
Ar - 20% O2), distance from nozzle exit to substrate : 50
mm, Ts = 400°C, input gas temperature: 220°C, no 3 ■ Ts= 750°C
•S, 100 fully dense
plasma.
3. RESULTS AND DISCUSSION
CO
The structure of the Zr02 - (5-20wt%) Y203 coatings has %ro 50
been characterized using scanning electron microscopy oo
(SEM) and X-ray diffraction (XRD ; CuKa radiation).
The yttrium content in the various phases has been
indirectly obtained by XRD, using the following relations
obtained from Scott's measurements [28] between the 100 150 200
coating mass (mg)
Y03/2 mole fraction, X, and the lattice parameters a and c
(in angstroems): Figure 5 - Coating thickness vs. coating mass for various
for 0.03 <X< 0.13 (tetragonal phase): substrate temperatures and for a fully dense coating
(with p = 6020 kg/m3).
at =5.08 + 0.358 X (1)
ct = 5.195 -0.31 A" (2)
for 0.12 <X< 0.33 (cubic phase): The pore volume fraction in the coating has been
evaluated by the simple relation P= 1 - Pexp/pth w'tn
ac = 5.115 + 0.16X. (3)
Pexp the experimental coating specific mass, measured as
The yttria contents will be given in Y203wt%, which is the substrate mass increase over coating volume ratio,
approximately equal to Y03/2mol%. The impurity content and pth, the specific mass of fully dense Zr02 - 8wt%
has been measured using nuclear reaction spectroscopy Y203 (pth K 6020 kg/m3); because of the errors on the
(NRS) and particle-induced X-ray emission spectroscopy coating mass and volume measurements, these results are
(PIXE), for carbon and chlorine detection (detection estimated to be accurate within ±10 rel%. Figure 5 shows
limits : 0.4 and 0.1 at%, respectively). These techniques the variation of the coating thickness vs. coating mass for
were also used to check the overall Y content in some various substrate temperatures. The slope of the straight
coatings. lines is l/(peXp.S), with S the substrate surface area. This
3.1 Morphology allows to calculate peXp and P, as summarized in Table
2.
All deposited coatings showed columnar morphology, as
illustrated on Figure 4. The columns, originating from the It is seen that the porosity level is highly dependent on
substrate surface, exhibit conical shape, leading to domed the substrate temperature during deposition. This is a
top surface. This morphology, frequently obtained in typical coating growth feature for high deposition rate
CVD, results from the nuclei preferential growth in given processes (especially PVD), as a competition between ad-
directions. particle (atoms, molecules) arrival rate and thermally
 11-5

Table 2 - Measured specific mass peXp and pore volume diffusion becomes more important, by ca. one order of
fraction P, for various substrate temperatures. magnitude between 400°C and 750°C if a leV activation
energy is assumed, so that less porous structures are
Substrate temperature Pexp P allowed to grow. At high temperatures, bulk diffusion
(°Q (kg/m3) can also contribute to the elimination of porosity.
400 2760 54% The porosity is not homogeneous throughout the coating,
600 3500 42% the first micrometers appearing to be more dense.
Annealing at 1100°C in air for 24 hours produces the
750 4420 26%
coalescence of the intergranular porosity which becomes
observable by SEM. This sintering takes place
üü^i throughout the coating, as seen on Figure 6 (bulk of
coating and column base).
The deposition of highly porous coatings could have
been interesting with respect to lowered thermal
conductivities, if the obtained morphology had been
thermally stable up to temperatures corresponding to gas
turbine applications. This is clearly not the case here.
3.2 Deposition rate
The deposition rate is about 55 um/h (48 mg/h) in the
standard conditions. Figure 7 shows that the deposited
zirconia amount (here at Ts = 400°C) is proportional to
the vaporized precursor mass and thus to the partial
pressure of zirconium compounds in the gas phase.
Figure 8 shows that the substrate temperature has very
little effect on the deposition rate. This indicates that the
process is not controlled by surface kinetics. Further
experiments are necessary to determine the limiting step
of the deposition process (mass transport, gas phase
reactions, ...) for various deposition conditions. Note that
the analysis of the deposition mechanism is more
complicated than for conventional thermal CVD, as the
influence of the plasma must be taken into account.
The deposition rate depends both on the gaseous
precursors flow rates (at least for ZrCl4) and the overall
deposition efficiency. The former is related to the
vaporization capability of the ZrCl4 and Y(thd)3 furnaces.
The ZrCL, furnace has been tested up to mass flow rates
corresponding to deposition rates of 150 um/h (about 200
mg/h Zr02 - Y203 deposited on the substrate, at
Figure 6 - SEM cross-sectional view of a coating Ts = 750°C); up-scaling of both furnaces should not
deposited in the standard conditions (Ts = 400°C) and present any problems. The latter is related first to the
annealed for 24 hours at 1100°C in air : bulk of coating physical and chemical processes in the gas phase and at
(top) and substrate-coating interface (bottom). the gas/coating interface and secondly to the gas injection
conditions and to the geometrical configuration of the
activated particle surface diffusion [29]. In the present deposition chamber, i.e. nozzle geometry, substrate size
case, the time interval between ad-particles arrival is and shape, nozzle-substrate distance, etc. Hence, the
about 20 ms, being defined as c/2R with c the lattice deposition efficiencies, defined as the number of
parameter (c«5 10"'° m) and R the deposition rate. At deposited Zr moles to the number of vaporized Zr moles
low deposition temperatures, the distance travelled by the ratio, that can be calculated from Figure 8, i.e. 8-10%,
adsorbed particles by surface diffusion during that time apply only to the current non-optimized deposition
interval is too small to allow the construction of a configuration. For instance, the deposition rate for a
« perfect» crystal so that many voids (vacancy clusters nozzle exit - substrate distance of 75 mm is the half of
or pores) are incorporated into the coating. It is seen on that at 50 mm : this is due to the fact that the flow
the SEM coating cross-sections (Figure 4) that undergoes a conical expansion in the nozzle and in the
intercolumnar porosity cannot account for the whole deposition chamber. An efficiency of at least 20% would
measured 54% pore volume fraction, showing that be achievable after process optimization.
intracolumnar porosity exists which cannot be resolved Most coatings were deposited at rates lower than about
by SEM. At higher deposition temperatures, surface 80 um/h, because beyond that value thickness
 11-6

Table 3 - Measured C and Cl atom contents in Zr02 ■

Y2O3 coatings at low and high deposition temperatures.
150 . •
ra Substrate temperature C content Cl content
E • (°C) (at%) (at%)
in 100 .
• 400 6.8 ±0.5 2.0
E •
O) 750 1.5 <0.1
50 . •
o
ü
•
••
• oxidation by atomic or molecular oxygen leading to
0 - J
h 1 gaseous products (CO, C02) has become favoured.
0 12 3
vaporized ZrCI4 mass (g)
At low deposition temperature, the coatings incorporate
some chlorine. The measured content can appear to be
relatively high if corrosion issues are in mind ; but it also
Figure 7 - Deposited zirconia mass vs. vaporized ZrCfy shows how efficient the plasma decomposition and
precursor mass. Ts = 400°C, P = 133 Pa, flow reaction processes are, by reducing the Cl:Zr atom ratio
rates : direct Ar, 500-700 seem, Zr-furnace carrier gas from 4:1 in the mixing chamber to about 2:30 in the
(Ar), 50-300 seem, O2, 200-300 seem, microwave coating, knowing that at 7^ = 400°C thermal CVD cannot
power: 400-900 W, nozzle-substrate distance : 50 mm. take place. For Ts = 750°C, no chlorine can be detected
in the coatings, here also because reactions of chlorine
compounds with oxygen leading to volatile products are
favoured at that high temperature.
3.4 Crystallographic structure
Zirconia coatings, partially stabilized with yttria with
mass fractions in the range 5-20 %, were deposited at 7^
= 400°C. XRD diagrams for coatings with approximately
7, 10 and 16wt% Y203 are shown on Figure 9 (7^ =
400°C). All coatings exhibit crystallographic texture, the
{100} planes being preferentially oriented parallel to the
sample surface. The yttria content was deduced from the
350 400 450 500 550 600 650 700 750
value of the a lattice parameter. In accordance to the
phase diagram, for both Zr02 - 7wt% Y203 and Zr02 -
Substrate temperature (°C) 10wt% Y203 coatings, the main phase is the metastable
non-transformable phase « t'» (with a small amount of
Figure 8 - ZrC>2 mass deposition rate (in g/h), monoclinic phase « m »), while the Zr02 - 16wt%Y203
normalized to the vaporized Zr&4 rate (in mg/h), as a coatings crystallize in the cubic phase « c ».
function of substrate temperature Ts. The Zr02 - 7wt%Y203 has been more specifically
heterogeneity increases (the thickness ratio at substrate studied. After annealing at 1200°C for 24 hours in air, the
center and edge becoming higher than about 2, in this texture (which is here more precisely (100)) is amplified,
configuration). but from there on remains stable, as shown by further
annealing for 96 hours. During this high temperature
3.3 Impurity content treatment, the «r'» phase is depleted in yttria while
Because these elements are present in the chloride and cubic phase is allowed to form. After 24 hour annealing,
organo-metallic gaseous precursors, chlorine and carbon the yttria content of the « t'» phase is reduced to about
may be incorporated into the Zr02 - Y203 coatings. Table 5wt%, to the benefits of cubic phase (or a tetragonal
3 shows the atom contents measured at two deposition phase with a c/a ratio close to 1) with ca. llwt% Y203,
temperatures. At Ts = 400°C, the C content is not which is the lower limit for the existence of this phase at
negligible (even if this corresponds to only 2wt%, 1200°C. After further annealing during 96 hours, the
approximatively) and reflects the fact that the gas phase yttria content of the « t' » phase still decreases, to about
has been enriched in carbonaceous compounds, 4wt%, but it remains constant for the cubic phase. This
originating from the (partial) dissociation of Y(thd)3 in means that in this second step, the yttria transferred from
the plasma, which where able to adsorb on the coating the «t' » to the « c » phase has been consumed only to
surface. It had not been possible to check whether the allow the increase of the cubic phase volume fraction
carbon is incorporated preferentially as free carbon or as within the sample, as is observed on Figure 10. The
a Zr compound. Nevertheless, at high deposition return to thermodynamic equilibrium, by development of
temperature, the C content is drastically reduced (less the cubic phase, is somewhat faster than that observed by
than 0.5wt%) probably because the kinetics of the Lelait [30] for plasma - sprayed Zr02 - 8wt%Y203
 11-7

70 20 f)
75 26 O 76

Figure 9-XRD diagrams of as-deposited PECVD Zr02 - Figure 10 - {400} region ofXRD diagrams for a ZrÖ2 -
XY2O3, with x = 7wt% (top), Uwf/o (middle) and 7wt% Y2O3 coating. Experimental diagrams and
16wt% (bottom). AI2O3 peaks are due to spoiled areas calculated contributions of the « t'» and « c » phases :
(edges) on an alumina substrate. as-deposited (top), annealed at 1200°C in air for 24
hours (middle) and 120 hours (bottom).
coatings, which appears after about 250 hours at 1200°C. substrates, with high rates (up to 150 um/h) at
This different behaviour of the PECVD sample may be temperatures ranging from 400 to 750CC. ZrCl4 and
explained by the high porosity level of the coating Y(thd)3 were used as source materials; vaporization,
deposited at low temperature (7^ = 400°C): the high handling and storage procedures have been established
amount of internal surfaces thus generated may favour, for both precursors.
comparatively to volume diffusion, the surface diffusion
of Y3+ cations, which are responsible for the evolution of Morphology, impurity content and crystallographic
structure of the coatings were investigated. At all
the «t' » metastable system towards thermodynamic
equilibrium. deposition temperatures, the coatings exhibited columnar
morphology and {100} crystallographic texture. At low
4. CONCLUSION substrate temperature (400°C), the pore volume fraction
was high {ca. 50%) and the coatings incorporated some
Plasma-assisted chemical vapor deposition has been carbon and chlorine impurities, but at higher temperature
successfully used to deposit thick (20 - 150 urn) zirconia- (750°C), the pore volume fraction reduced to about 25%
(5-20wt%) yttria coatings on metallic and alumina and little carbon and no chlorine were detected. It thus
11-8

appears that medium to high temperatures are temperatures », Surf. Coat. Technol. 68/69, 81-85 (1994).
unavoidable to synthesize suitable coatings for thermal [9] Kreutz E.W., Lemmer O., Wesner D.A., Alunovic M.,
barrier applications. Erkens G., Leyendecker T., Voss A., « Electron and laser
Metastable tetragonal «t'» phase was the major radiation as sources of zirconia film deposition », Surf.
constituent of coatings with 7 and llwt% yttria, while Coat. Technol. 74/75, 1005-1011 (1995).
16wt%-yttria coatings exhibited only the cubic phase, in [10] Movchan B.A., « EBPVD technology in the gas
accordance to the phase diagram. Annealing for 24 and turbine industry : present and future », JOM, 40-45 (nov.
120 hours of Zr02 - 7wt% Y203 coating deposited at 1996).
400°C showed that cubic phase with an yttria content of
about llwt% is allowed to grow, at the expense of the [11] Grünling H.W., Mannsmann W., « Plasma sprayed
« t' » phase whose yttria content diminishes to 5wt% and thermal barrier coayings for industrial gas turbines :
4wt%. This relatively fast evolution towards morphology, processing and properties », J. Phys. IV,
thermodynamic equilibrium may be related to the high Colloque C7 suppl. J. Phys. Ill, 3, 903-912 (1993).
porosity level. [12] MeVrel R., « Barrieres thermiques pour aubages de
moteurs aeYonautiques. Etat de 1'art et perspectives », La
Work is in progress to further investigate the applicability Recherche Aerospatiale 5-6, 381-392 (1996).
of PECVD to coat complex-shaped substrates and to
study the relationships between deposition parameter and [13] Meier S.M., Gupta D.K., « The evolution of thermal
coating properties. barrier coatings in gas turbine engine applications»,
ASME J. Eng. Gas Turbines Power 116, 250-257 (1994).
ACKNOWLEDGEMENTS [14] Osyka A.S, Rybnikov A.I., Leontiev S.A., Nikitin
N.V., Malashenko I.S., « Experience with metal/ceramic
This study has benefited from a partial financial support
coating in stationary gas turbines », Surf. Coat. Technol.
from the Delegation G6n6rale de l'Armement (DSP and
76/77,86-94(1995).
SPA6). Thorough XRD investigations performed by C.
Diot and T. Ochin, as well as fruitful discussions, have [15] Parker D.W., «Improving thermal efficiency with
been are greatly appreciated. ceramic thermal barrier coatings», Proceed, of the
Internat. Symp. on Developments and Applications of
REFERENCES Ceramics and New Metal Alloys, Quebec City, Canada,
aug. 29 - sept. 2, 1993, 603-613.
[1] Stecura S., «Optimization of NiCrAl/Zr02-Y203 [16] Minh N.Q., «Ceramic fuel cells», J. Am. Ceram.
thermal barrier systems », NASA TM 86905 (1985). Soc. 76(3), 563-588 (1993).
[2] Maleto M.I., Solovjeva L.I., Turesvskaya E.P., [17]«Thermal Barrier Coating Workshop '95»
Vorolitov K.A., Yanovskaya M.I., «Alkoxy-derived (Cleveland, Ohio, USA, mar. 27-29, 1995), NASA-CP-
Y203-stabilized Zr02 thin films », Thin Solid Films 249, 3312(1995).
1-5 (1994). [18] «Thermal Barrier Coating Workshop 1997»
[3] Yamane H., Hirai T., «Yttria stabilized zirconia (Cincinnati, USA, may 19-11, 1997)
transparent films prepared by chemical vapour [19] DeMasi-Marcin J.T., Gupka D.K., «Protective
deposition », J. Crystal Growth 94, 880-884 (1989). coatings in the gas turbine engine », Surf. Coat. Technol.
[4] Kim J.S., Marzouk H.A., Reucroft P.J., « Deposition 68/69, 1-9(1994).
and structural characterization of Zr02 and yttria- [20] Rossnagel S.M., Cuomo J.J., Westwood W.D., Eds.,
stabilized Zr02 films by chemical vapour deposition », « Handbook of plasma processing technology », Noyes
Thin Solid Films 254, 33-38 (1995). Publications, ISBN-0-8155-1220-1 (1990).
[5] Cao C.B., Wang J.T., Yu W.J., Peng D.K., Meng [21] Seiberras G., «Realisation et caracterisation de
G.Y., « Research on YSZ thin films prepared by plasma- revetements de zircone obtenus par döpöt chimique en
CVD process », Thin Solid Films 249, 163-167 (1994). phase vapeur assist^ par un plasma miro-onde », These
[6] Brinkman H.W., Meijerink J., de Vries K.J., de Doctorat, University Paris 11-Orsay, mar. 24th, 1994.
Burggraaf A.J., «Kinetics and morphology of [22] Bertrand G., M6vrel R., « Zirconia coatings realized
electrochemical vapour deposited thin zirconia/yttria by microwave plasma-enhanced chemical vapor
layers on porous substrates », J. Eur. Ceram. Soc. 16, deposition », Thin Solid Films 292, 241-246 (1997).
587-600(1996). [23] Gavillet J., Belmonte T., Hertz D., Michel H, « Low
[7] Carter W.B., Godfrey S., «Combustion vapor temperature zirconia thin film synthesis by a chemical
deposited partially stabilized zirconia coatings», in vapour deposition process involving ZrCl4 and 02-H2-Ar
« Elevated temperature coatings : Science and technology microwave post-discharges. Comparison with a
I », Dahotre N.B., Hampikian J.N., Stiglich J.J. Eds., The conventional CVD hydrolysis process», Thin Solid
Minerals, Metals & Materials Society, 103-111 (1995). Films 301, 35-44 (1997).
[8] Andritschky M., Rebouta L., Teixeira V., « Corrosion [24] Chevillard S., «Etude d'un procedö de döpöt
and adherence of stabilized Zr02 coatings at high chimique en phase vapeur assists par un plasma micro-
 11-9

onde pour la realisation de revetements de zircone

yttriee», These de Doctorat, University Paris 11-Orsay,
feb. 11* 1997.
[25] Chevillard S., Vidal-S&if M.H., Drawin S., « Yttria
partially stabilised zirconia coatings by microwave
plasma enhanced CVD (MPECVD) », 10* Intern. Colloq.
on Plasma Processes, Antibes (France), jun. 11-15, 1995,
Suppl. Revue «Le Vide : science, technique et
applications » 275,431-434 (1995).
[26] Yahya S.M., «Fundamentals of compressible
flows », Halsted Press (John Wiley & Sons, 1982).
[27] Fluxexpert code, DT2I, Chemin des Preles, 38240
Meylan ZIRST (France).
[28] Scott M.G., « Phase relationship in the zirconia-
yttria system », J. Mater. Sei. 10, 1527-1535 (1975).
[29] Yang Y.G., Johnson R.A., Wadlay H.N.G., «A
Monte Carlo simulation of physical vapor deposition of
nickel », Acta Mater. 45(4), 1455-1468 (1997).
[30] Lelait L., «Etude microstructurale fine de
revetements cöramiques de type barriere thermique;
incidence sur la resistance thermomöcanique de ces
revetements» These de Doctorat, University Paris 11-
Orsay, 1994.
 12-1

THERMAL SPRAYED NANOSTRUCTURED THERMAL BARRIER COATINGS

L. T. Kabacoff
Materials Division
Office of Naval Research
800 N. Quincy Street
Arlington, VA 22217-5660

Summary properties is that, as the physical scale becomes

very small, a larger proportion of atoms are
Nanostructured ceramics exhibit a number of found at surfaces or interfaces. For example, in
enhanced mechanical and thermal properties a 3-D nanostructured material with a grain size
which may make them attractive as thermal of 10 nm, the proportion of atoms at grain
barrier coatings. The phonon contribution to boundaries approaches 50% (depending on the
thermal conductivity is greatly reduced at room width of the grain boundary). This has a
temperature, but becomes significant at profound effect, not only on mechanical
operating temperatures only for grain sizes properties, but also diffusion, ionic
below 20 nm. Improvements in mechanical conductivity and permeability.
properties, such as fracture toughness and
resistance to spalling, can be achieved at larger Nanostructured coatings and structures
grain size, in the range of 30 - 70 nm. Suitable can be fabricated by a number of processing
coatings, consisting of a composite of zirconia methods. Nanoscale multilayers are generally
with a second immiscible phase, such as formed by a sputtering or evaporative process
alumina, can be fabricated by thermal spray. performed in controlled atmosphere.
However, development of techniques to Polycrystalline nanostructured materials are
produce the required microstructure are at an formed by consolidation of either very small
early stage and may require advances in the particles or larger particles containing very
state of the art for thermal spray technology. small grains. The consolidation process
involves the application of thermal energy and
Introduction the challenge is to carry out the consolidation
without engendering excessive grain growth.
Nanostructured materials have traditionally Last year, the Office of Naval Research
been defined as any material having some launched a new program with the goal of
physical length scale smaller than 100 nm. producing nanostructured coatings for a variety
This could be a particle size, layer thickness, of applications using thermal spray processing.
fiber diameter, or grain size (0-, 1-, 2-, and 3- These include resistance to wear, erosion, and
dimension respectively). The choice of 100 nm cavitation. The feasibility of fabricating
is somewhat arbitrary, but is based on the fact nanostructured thermal barrier coatings for gas
that the characteristic length of many turbine engines is also being considered.
interesting mechanical, optical, and magnetic Thermal spray is attractive for several reasons.
phenomena are of this order. Thus, as the layer It is a very fast process, making it relatively
thickness or grain size become very small, the easy to control grain growth. It is also an
associated properties begin to diverge radically inexpensive process which is already in wide
from "conventional" values. Another reason spread use. Implementation will not require
why nanostructured materials exhibit unusual any major capitol investment or extensive

Paper presented at an AGARD SMP Meeting on "Thermal Barrier Coatings",

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
12-2

operator training. The program will material of the same composition.

concentrate on several materials including Understanding of these properties is very
alloys, composites and ceramics. Development incomplete, especially those of nanostructured
of TBC s will mostly deal with composites ceramics. This is partly due to the fact that
containing zirconia. until recently, it has been very difficult to
fabricate good quality test specimens
The changes in mechanical properties reproducibly. Some properties important to
and thermal conductivity which can be consideration of TBC's have been well
obtained in an ultrafine materials can, documented. For example, it has been widely
potentially, be exploited to produce thermal observed that nanoscale ceramic particles have
barrier coatings which are, in some way, greatly suppressed melting temperatures and
superior to a coating of the same material, but also sinter remarkably well at relatively low
with conventional microstructure. Whether or temperature (Ref 1). This is due to the very
not such coatings are possible depends on large surface area leading to a very large
resolving several issues: surface tension (producing a high effective
pressure) and extreme reactivity. The high
1) What grain size must be achieved to obtain surface energy also leads to the stabilization of
significant improvement in thermal resistivity phases favored by the higher pressure. For
and mechanical properties such as toughness example, the ambient stable phase of zirconia
and resistance to cracking and spallation. particles smaller than a critical size is rutile.
Coarsening produces a transformation to
2) Can a suitable nanostructured material be anatase. The critical size is temperature and
synthesized at reasonable cost and in adequate purity dependent, about 30 nm at room
quantity? temperature. This is illustrated in Figure 1. A
consequence of this stabilization is that the use
3) Can the synthesized material be processed ofyttriamustbereexamined. Normal
into a coating which retains the nanostructure? compositions of YSZ are greatly over

4) Can a nanostructured coating be fabricated

with sufficient thermal stability to withstand
typical service temperatures experienced in a
gas turbine engine?

As will be discussed below, research to date

indicates that it may be possible to produce
nanostructured thermal barrier coatings for gas
turbine engines which exhibit significantly
enhanced properties and which are relatively
inexpensive.

Properties of Nanostructured Ceramics

Figure 1. Zirconia noparticles above (anatase)
The properties of a nanostructured and below (rutile) critical size. (Courtesy of C.
material can differ radically from a coarser Berndt)
 12-3

stabilized and exhibit no transformation dramatically reduced as the grain size goes
toughening. The toughening mechanism below 100 nm, the traditional definition of a
reappears if the percentage of yttria is reduced nanostructured material. However, the drop in
to below 2 percent (Ref 2). Another important thermal conductivity at operating temperatures
feature of nanostructured ceramics is reduced within gas turbine engines becomes significant
thermal conductivity (Ref 3). This effect only when the grain size goes below about 20
results from the grain boundaries being closer nm. Several unpublished studies have
-together than the typical distance between point demonstrated that this is indeed the case. Not
defects in a conventional ceramic. Thus, the shown in the calculations is the photon
grain boundary contribution to phonon contribution to thermal conductivity. This is of
scattering becomes anomalously high. The major importance since nanostructured zirconia
calculated phonon contribution to thermal is quite transparent to infrared. Therefore, a
conductivity is illustrated for zirconia (figure 2) nanostructured TBC would have to include
and YAG (figure 3). In both cases, the thermal micron size inclusions, either pores or a
conductivity at room temperature is suitable second phase.

>^
•a
u
3
T3
C -
o 2
u
Intrinsic
S-l
D
X!
5 nm
0
0 200 400 600 800 1000 1200
Temperature, T (°C)

Figure 2. Thermal conductivity of 7YSZ versus grain size and temperaure (Ref. 3)
12-4

It was previously observed that it is difficult to composite, moreover one containing

consolidate nanostructured powder without immiscible phases. Some research has been
excessive grain growth. This is illustrated in done on a composite of zirconia and alumina.
figure 4, which shows final grain size as a Coatings were fabricated by atmospheric
function of sintering temperature (Ref 4). Two plasma spray using several different sources of
things are evident. Grain growth becomes nanostructured starting materials. It was found
significant as temperatures approach 800 C. that the percentage of alumina needed to
Also, porosity is not effective in inhibiting stabilize the structure depended considerably
grain growth, as can be seen by the fact that on the morphology of the thermal spray
sintering of green bodies with differing feedstock and, therefore, the morphology of the
porosity result in the same final grain size. coating. In the best coatings, 30 % alumina
This implies that grain growth in service would stabilized the grain size and, therefore,
be a problem. Clearly, a monolithic tetragonal phase up to 1100 C (Ref 5). A very
nanoceramic cannot be used as a TBC in a gas fine grain size (below 20 nm) could not be
turbine engin. The material must be a achieved because there was complete melting

10 j— i i i i

YAG
,__,
- 8
U
E

^ 6 A \
>->
*> \>lQpnmv\0
N
3 4
^ v^ o
N
^ \>
o
S v ^
^ »x "N-
^
^^w o o
U V* ^

"3
i 2 ■* *■
— —
- —.

5 nm
i i i i i
u
0 200 400 600 800 1000 1200
Temperature, T (°C)

Figure 3. Thermal con ductivity of YAG versus grain size and temperaure (Ref. 3)
 12-5

of the sprayed powder during processing. The TBC. In spite of the lack of a large body of
grain size was limited by the quench rate and definitive data, two facts emerge.
the phase separation of zirconia and alumina Nanostructured ceramics deform
during solidification. The critical size for superplastically at moderate temperature (Ref
stabilization of rutile was larger in the 6). This temperature is typically less than half
composite than that found in individual pure the melting temperature and can be even lower
particles. Typical grain size in these studies for very small grain size. It has also been
was 30 to 50 nm. observed that intrinsic stress can be relieved at
low temperature, perhaps even room
The mechanical properties of nanostructured temperature. Room temperature relaxation has
ceramics depend critically on composition and definitely been observed in nanostructured
microstructure. Almost all of the studies have cermets (Ref 7). It was also observed in
been performed on "fully dense" material (in nanostructured titania by Gleiter (Ref 8).
reality, unintentionally porous). Very little Gleiter attributed this to very high room
work has been done on intentionally porous temperature diffusion (more than six orders of
materials, as one would prefer to have in a magnitude higher than that found in microscale

400-
ZrOr 3mol%Y203 +
+ commercial submicron
(30 nm starting grain size;
200 nm agglomerate size) /+
300- ■ G47% nano
6 D G58% nano
(nano: 15 nm starting grain size;
• 15-30 nm agglomerate size)

200- A
Q
■I A
Ö 100- 97.8%
-56.7%
. E 7^82.2%
5 °C/min heating rate
0 2 hr. hold at tempr
, . , —i ■ r -
600 800 1000 1200 1400

Sintering Temperature (°C)

Figure 4. Grain size versus temperature for sintered nanoscale 3YSZ powder.
12-6

materials). It is clear that these coatings ceramic powder does not exist. Several
should not be used in situations where creep companies have the capability to provide
resistance is required. Another property production capacity, but will need to see an
affected by the ultrafine microstructure is adequate market before doing so. Research
fracture toughness. In the case of thermally quantities are easily obtained from numerous
sprayed nanostrucured zirconia, the biggest sources.
effect may depend on the way the coating is
formed. If the feedstock consists of Development of techniques for fabricating
agglomerated small particles, such as colloidal coatings by thermal spray are at a very early
material, and no melting (or only partial stage. It was noted above that improvement of
melting) occurs, then the coating will lack the thermal resistivity will require an extremely
splat boundaries present in conventional small grain size, of the order of 10 - 20 nm,
coatings. These splat boundaries are a major while improvement of mechanical properties
factor in the failure of conventional thermal can be achieved by reducing grain size only to
sprayed coatings. It is possible to form about 30-70 nm. The difficulty in retaining an
coatings without melting because of the very ultrafine microstructure in the coating during
high sinterability of the small particles. processing an subsequent exposure to elevated
Moreover, use of such small particles is temperature increases dramatically with
necessary if one wants to achieve a grain size reduced grain size. One reason for this is that
small enough to effect thermal conductivity. If the driving force for grain coarsening is highly
melting does occur, the coating will still be grain size dependent. Another reason is that,
nanostructured provided the sprayed material is while it is possible to produce a 30 - 70 nm
a composite of immiscible phases. In this case, grain structure through melting and rapid
the thermal conductivity will not be greatly solidification, a 10 - 20 nm grain structure can
reduced, but the mechanical properties will still only be achieved by avoiding melting, relying
be enhanced. Several companies are pursuing instead on rapid sintering of very fine particles.
both strategies, but the information is This may, if fact, be quite feasible, but it has
considered proprietary at this time. yet to be demonstrated. Clearly, improving
mechanical properties will be much easier than
Discussion and Conclusions reducing thermal conductivity. In both cases,
control of coating microstructure will require
The first requirement for successful fabrication very tight control of the thermal history of the
of nanostructured TBC's be to obtain suitable sprayed particles. This, it turn, will require
material for thermal spray. The synthesis of excellent control of both the size and
suitable sprayable material has not been morphology of sprayed particles, and spray
difficult. Nanoscale ceramic powder can be conditions. It may, in fact, require real time
synthesized by gas condensation, by sol gel and process control at a scale not currently
colloidal processes, by various flame and available. Thus, it may be necessary to extend
plasma processes, and mechanical attrition. the state of the art in thermal spray. This much
Several techniques have also been developed more likely to be true for improvement in
for reprocessing raw powder into a suitable thermal conductivity than for improvement of
form for thermal spray. These reprocessing mechanical properties. Finally, a great deal of
techniques are proprietary and cannot be research still needs to be done in order to
disclosed at this time. At present, commercial determine exactly what composition and
manufacturing of large quantities of nanoscale microstructure to attempt to achieve. This will
 12-7

include a thorough study of creep and stress

relief in the coatings.

In conclusion, the use of nanostructured TBC's

with enhanced properties relative to existing
materials is possible. It will require a great
deal of effort both in research on
nanostructured ceramic composites and in
development of thermal spray techniques
capable of achieving desired microstructures.
The kind of effort required to reduce thermal
conductivity will be quite different from that
required to improve mechanical properties,
with very different attendant risks. Ongoing
work under the ONR program should clarify
most of these issues during the next year.

References

1. Birringer, R., "Nanocrystalling Materials",

Mat. Sei. And Eng., Al 17, 1989, pp 33-43.
2. Mayo, M. J., To be published.
3. Klemens, K. G., "Theory of Thermal
Conductivity of Nanophase Ceramics", in
"Chemistry and Physics of Nanostructureed
and Related Non-equilibrium Materials", TMS
Proceedings, 1997, p. 97.
4. Mayo, M. J., "Processing of Nanocrystalline
Ceramics from Umafine Particles",
International Materials Review, 21, 1996, pp
85-115.
5. Provenzano, V. And Ying, J., To be
Published.
6. Mayo, M J., "High and Low Temperature
Superplasticity in Nanocrystalline Materials",
Nanostuctured Materials, 9, 1997, pp 717-726.
7. Mayo, W. Private Communication.
8. Karch, J., Birringer, R. And Gleiter, H.,
"Ceramics Ductile at Low Temperature",
Nature, 330, 1987, pp 556-558.
 13-1

Modelling the Thermal Conductivity of Thermal Barrier Coatings

J.-M. Dorvaux, O. Lavigne, R Mevrel, M. Poulain, Y. Renollet, C. Rio

ONERA - Materials Science Department
29, av. de la Division Leclerc
92322 Chätillon G§dex
France

ABSTRACT Y2O3, this corresponds to a vacancy content of the order of 4.4

The thermal conductivity of ceramic coatings is a function of the mol.%, providing an efficient source of scattering for phonon
thermal conductivity of the individual constituents (nature of phases, propagation. This also explains why the thermal conductivity of
presence of point and planar defects, etc.) and of the morphology of these materials is a decreasing function of the yttrium content (at
the coating (pores, cracks). This latter aspect is particularly least for relatively low stabiliser contents [5, 6]). Another way of
important for TBCs as it explains why their thermal conductivity is decreasing the intrinsic thermal conductivity of ceramic layers
significantly lower than the corresponding values for dense would be to substitute zirconium ions with heavier metallic ions
materials. [7, 8], an approach followed by Maloney et al. [9] with a work
This paper presents a model developed to calculate the thermal on Ce02-Y203 system, at the expense though of an increase in
conductivity of porous and microcraked ceramic coatings based on a density.
finite difference method applied on digitised images of yttria
partially stabilised zirconia (YPSZ) coatings deposited by plasma Another efficient way to decrease the thermal conductivity of a
spraying. The influence of the different types of morphological material is to introduce microstructural defects such as pores,
features on the thermal conductivity is discussed. voids, microcracks which constitute obstacles against the
through-thickness heat transfer propagation. Fortunately, the
presence of such defects, produced during fabrication, seems to
1. INTRODUCTION be indispensable for the thermomechanical resistance of these
Thermal barrier coatings are constituted of a ceramic external systems. It is doubtful however that their distribution in existing
layer, most generally composed of partially stabilised zirconia coatings is optimised regarding the thermal properties and, in
(Zr02-6 to 8wt% Y203) deposited on top of an oxidation order to design more performant TBCs, it now appears
resistant metallic layer (MCrAlY alloy or modified aluminide). necessary to be able to characterise and understand the
Their main function is to thermally isolate the superalloy relationship between thermal conductivity and microstructural
components, blades or vanes, which they protect from the hot features constituting these heterogeneous materials.
gases circulating in the turbine. Typically, the use of a 250 urn
thick ceramic layer can provide about 200°C temperature It must be reckoned that guidelines that can be derived from the
reduction on the metallic part, thus prolonging its lifetime, literature, in particular from models relating morphology
reducing thermal transients, etc. [1]. (porosity) and thermal conductivity of solid materials are of
limited usage in the case of thermal barrier coatings due to the
Up to now, most of the efforts dedicated to thermal barrier complexity of the microstructures.
coatings have been focused on the obtention of performant
reliable systems and a predominantly empirical approach has A variety of models have been developed to estimate the thermal
permitted their introduction in service even on highly stressed conductivity of multiphased solids, in particular porous
components such as first stage blades [2]. materials. Most of them assume the material constituted of a
dispersion of a more or less concentrated and randomly
Only recently, as witnessed for instance by the dedication of a distributed second phases. While early models proposed by
full session during a recent TBC workshop [3] and on-going Maxwell [10], Eucken [11], Rüssel [12] and Bruggeman [13] for
research programmes [4], significant attention has been example, consider only spherical particles, improved versions
increasingly directed towards characterising, understanding and elaborated by Murabayashi [14], Schulz [15], Koh and Fortini
lowering thermal conductivity, an altogether essential property [16], Cunningham [17] and McLachlan [18] consider ellipsoids
for these systems. characterised by shape and orientation factors.

Zirconia-based thermal barrier coatings exhibit already low As a general rule, the thermal conductivity X of a porous solid
thermal conductivities (typically between 1 and 2 W/m.K). derived with these models can be expressed as :
These low values of thermal conductivity arise both from the
intrinsic value of dense partially (and fully) stabilised zirconia X/Xo = f (s, nf)
materials and from the heterogeneous microstructure of the
coatings deposited by plasma spraying or EBPVD (electron where Xo is the thermal conductivity of the pore-free solid, s is
beam physical vapour deposition). the volume fraction of the pores and n( are adjustable
parameters, to be determined empirically, related to the shape
The introduction of a stabiliser, required to avoid the detrimental and/or orientation of pores.
effect of monoclinic to tetragonal phase transformation at typical
service temperature, is accompanied by the incorporation of a To evaluate the validity of these models, several authors have
substantial amount of vacancies. Thus, in the case of yttria examined the analytical results given by some of these models in
stabilised zirconia, the substitution of two tetravalent zirconium light of experimental values on different systems [19, 20, 21, for
ions by trivalent yttrium ions is accompanied by the example]. Examining experimental data on alumina, graphite,
incorporation of one oxygen vacancy. For PSZ containing 8wt% uranium dioxide, sandstone, silica brick and limestone, and with

Paper presented at an AGARD SMP Meeting on "Thermal Barrier Coatings",

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
13-2

porosity up to 75%, Rhee [19] found a satisfactory correlation and measured values. This is attributed to the complex pore
with the equation proposed by Aizanov and Domashnev [22]. structure, which cannot be properly taken into account by the
For El-Fekey et al. [20] who evaluated the Maxwell, Loeb and limited areas on the cross sectional micrographs serving as input
Murabayashi models on thoria compacts with porosities up to information into the calculation.
46%, the Murabayashi model [14] yields the best result among
the three. More recently, Jackson et al. [21] found that In a study on the thermal conductivity of U02 pellets, Bakker
McLachlan's equation [18] comes closest to modelling the [30] used a finite element method to compute the conductivity of
experimentally measured thermal conductivity of A1N-Ln203 a matrix containing a dispersed phase (pores). Starting from a
systems. photograph of a cross section, the microstructure of the material
is described as a triangular mesh used as input data. Arbitrary
Remarking that the experiments considered may not be suited to temperatures are imposed on the upper and lower boundaries of
determine the quality of theoretical calculations, due to various the corresponding area (other boundaries are taken adiabatic)
experimental inaccuracies and simplifications, Bakker et al [23] and the FEM program computes the conductivity from the
have computed the influence of pores of different shapes thermal flux profile. As explained later, it remains doubtful
(cylinders, spheres, ellipsoids) with the FEM technique. The whether this type approach can be extended to large areas,
comparison with the equations describing the effect of keeping memory storage reasonable.
inclusions on the overall thermal conductivity shows that the
analytical equation of Schulz [15] gives a good description of In the work presented here, the objective was to develop and
such materials. evaluate an alternative approach, based on a finite-difference
calculation for computing the thermal conductivity of thermal
A full review of all the models is beyond the scope of this barrier coating, using as input data digitised images of the real
article. However, the puzzling diversity of conclusions derived material. With this method, the morphology of the ceramic
by these authors point out to the fact that all these models with coating, as complex as it can be, is properly taken into account.
simple algebraic expressions assume that the porosity can be
experimentally assessed with a very limited number of 2. CHARACTERISATION OF TBCs POROUS
parameters (the pore volume fraction e and adjustable STRUCTURE
parameters), which singularly restricts their predictive capacity. In order to provide data to the thermal conductivity calculation
software, reliable procedures have been developed to quantify
To overcome these shortcomings, several authors have the real TBCs porous structure. The success of the modelling
developed more sophisticated theoretical approaches as well as strongly depends on the accuracy and reliability of the results of
finite element calculations to estimate the effective thermal these procedures. The complete microscopic characterisation of
conductivity of multiphase materials. the TBC morphology involves numerous steps as sample
preparation, image acquisition, image analysis and validation
Tzou [24] estimated the thermal conductivity starting from a with macroscopic porosity determination.
general field theory (supposing uniformly distributed internal
cavities), deriving tractable analytical expressions for simple Sample preparation
pore geometries, i.e. insulated spherical cavities and penny- A correct and reproducible metallographic preparation of plasma
shaped cracks. sprayed Zr02 coatings is not straightforward but is essential for
obtaining a representation of the real porous structure. Porosity
In a more sophisticated approach, Furmanski [25] developed an determination by quantitative image analysis and numerical
effective macroscopic description for heat conduction in calculation of the thermal conductivity will give reliable results
heterogeneous materials using an averaging technique and provided that no error is introduced at the sample preparation
Green's function method and applied it to describe a model stage.
composite with randomly oriented parallelepiped inclusions
distributed in an epoxy matrix. Materials
All investigations are performed on a 1 mm thick free standing
The expressions obtained with these approaches become readily TBC plasma sprayed by SNECMA, initially on aHastelloyX
complex and apparently they have been used up to now only in substrate without any bondcoat. The powder is a commercial 8
the case of uniform distributions involving rather simple weight% yttria partially stabilised zirconia spray dried powder
geometrical shapes for the pores (ellipsoid, cubes). For materials (HCST Amperit 827.423).
having complex pore microstructures, such as those met in
plasma sprayed coatings, and in order to be closer to the real Sample preparation
structure (interconnected porous structure), however, numerical The following procedure has been set up. The free standing
schemes appear to be the most promising approaches. coating is vacuum impregnated (Struers/Epovac, P-100 mbar)
with a superfiuid epoxy dye (Struers/Epofix 301) before
With the objective of extending the rather simple analytical sectioning. Sample slices are then cut off with a precision saw
approaches (and poorly predictive) proposed by McPherson [26] (Struers/Accutom-2) using a thin diamond wheel and the lowest
and by Moreau et al. [27] on the thermal conductivity of plasma rotation and sample forward speeds to avoid any damage. The
sprayed coatings (and later improved by Bjorneklett et al.[28]), slices are then manually polished according to the following
Hollis [29] developed a numerical scheme in which the actual steps : 1200 then 4000 SiC papers for grinding, polishing using
pore structures of vacuum (VPS) and air (APS) plasma sprayed a 3 urn diamond spray first on a hard cloth (Struers/Pan W, blue
tungsten coatings are used as the basis of finite-element models lubricant) then on a smoother cloth (Struers/DP-Mol, pink
to calculate the effect of pores on the thermal conductivity of lubricant), final polishing with a !4 urn diamond paste.
plasma sprayed coatings. If, for VPS coatings, whose pore Polishing quality and absence of pull-outs are checked with light
distribution and shapes are relatively simple, this approach gives microscopy after each step. For scanning electron microscopy
a reasonable agreement between calculated and experimental observations a -20 nm thick conducting carbon layer is vacuum
values (respectively 70% and 60% of the bulk value), for APS sputtered within a Balzers MED010 unit.
coatings, instead, a large discrepancy exists between calculated
 13-3

Image acquisition
Cross sections of TBC prepared as previously described have JÄÄiBiteMSJ!ÄiÄiJÄ*«l
been observed in a digital scanning electron microscope
(Zeiss/DSM960). Backscattered electron mode is chosen
fill "J 1 T 1 ['
•-jsoa»

preferentially to secondary electron mode due to the higher I I I 1 i

,mm
I I 1 1
1 j 1
contrast between zirconia matrix and porosity (pores and I I 1 j I 1 f™ |
cracks). The adopted acquisition conditions are the following: I\ i 1
I I j 11 j 1
15 kV acceleration voltage, 9 mm working distance, «low»
electron current. Brightness and contrast are adjusted in order to
I I 1 i I 1 1,

■
obtain a reproducible grey level histogram (Fig. 1). Two
I I 1 iJ «ax»
magnification levels (180x180 um2; 45x45 um2) and two image I'll
[ill 1 i,„
1 k^ ■ lOOOO

resolutions (512x512 or 1024x1024 pixels) have been used for .. 70 ?1 72 73 U ?6 ?S 77 ?8 7SJÖ *i S3 S3 &* 85 8g S? S3 83 SO
characterising the TBCs morphology. Their influence will be orgy twes

discussed later on (see § 4).

Signal to noise ratio is increased by pixel averaging so that a

1024x1024 pixels image acquisition takes about 400 seconds.
Image acquisition is fully automated owing to computer assisted
stage displacement. This allows a statistical characterisation of
the samples.

Image analysis
Image analysis procedures have been developed with two Figure 2: Thresholding.
objectives:
• provide binary images of TBC cross sections which will be
used directly as input data for the thermal conductivity Porosity separation
calculation software, This procedure is based on opening and reconstruction image
• quantify the porous structure in such a way as to be able to analysis operations. It creates three new images from the initial
determine the contribution of each morphological feature to thresholded image : one containing the dispersed globular pores,
the thermal conductivity of a thermal barrier coating. one with the crack network and one with the globular pores and
the cracks directly connected to them.
This will be the starting point for modelling thermal
conductivity as a function of the morphology. This fully automated procedure will be widely used to determine
the contribution of each morphological feature to the
Thresholding conductivity (see § 4). A still more precise description can be
The transformation of SEM grey level images into binary images achieved using the following procedure.
is tht key step on which depends the reliability of the subsequent
results. The threshold level is manually fixed on images with Porosity reconstruction
optimised and reproducible grey level histogram (Fig. 1). From an initial parent binary image it is possible to derive a
series of images (about 300 or more according to the
This last one contains two peaks: a very sharp and intensive morphology complexity) with evolving porosity. This is done by
peak at the 0 level (black level) corresponding to the majority of applying a sequence of erosion image analysis operations to the
the globular pores and a wider one centred at middle grey levels parent image which leads to a final image containing only
which includes cracks and zirconia matrix. «ultimate erodeds ». These pixels exactly correspond to the
globular pore centroi'ds. Starting from this new image, a step by
step reconstruction of the porous features creates the derived
images. The operation is ended when the initial total porosity is
«MCi—T i
j completely reconstructed (Fig. 5).
I i 1 1
1 1 1 An application of this procedure in relation with the thermal
j f
conductivity computation will be shown in the last section.
rl
1 zJ1
i

I
40000
Image assembling
30C00 1 This algorithm is developed to provide large size images of the
[ 1 samples which are more representative for the conductivity
20000
11 I calculation. This is done by an automatic search of maximal
tOCOO

a
20 40 60 80
LL _XJ_J
100 120 140 160 1W 203 220 24Q 260 ■
overlap between two binary images from contiguous and slightly
overlapping fields.
Gtey lev& |

Figure 1: Grey level histogram. Analysis ofglobular pores

Each individual globular pore is characterised by a set of
parameters (area, perimeter, Feret's diameters, number of
The adopted threshold level which accounts for the whole neighbours,...). In this way different pore distributions (in
porosity (pores and cracks) corresponds to the beginning of the number, surface,...) can be obtained (Fig. 3).
lightening of isolated pixels in the solid phase (Fig. 2). This
ensures its reproducibility.
13-4

|0*tilavf|j«Jrpx^|

Start: «mtroids from globs

Reconstruction to globs
Figure 3: Globular pore distribution (% of particles
versus area")
• •
Analysis of the crack network
This procedure is based on the «skeletonization» (image
analysis terminology) of the crack network, followed by a
decomposition in juxtaposed segments of minimal fixed length.
A distribution of these segments with regards to their orientation
is then available (Fig. 4).

- *
Reconstruction to globs and linked cracks

i e 5 S Si 5*?'
I i ! 3 3 3 ! ! 3 i

Figure 4: Segment orientation (average of segment

length versus angle)

Reconstruction to total porosity

Porosity determination
The total porosity of a TBC can be estimated from binary
images of cross sections using image analysis provided that the
number of analysed fields is statistically relevant. A total of 105
♦
images (180x180 um2; 1024x1024 pixels) corresponding to two
transverse perpendicular cross sections have been considered.
The mean porosity value obtained for this image set is 15% with
a 2% standard deviation.
Furthermore the volumetric total porosity of the whole free
standing coating (60x45x1 mm3) has been determined to 12+1 %
from water immersion experiments and X-ray diffraction
theoretical density data. Closed porosity is less than 1 %.

The relatively good agreement between image analysis results

and volumetric porosity shows that the sample preparation
procedure as well as the thresholding one do not introduce too
many artefacts. The higher value obtained by image analysis is
attributed to some remaining pull-outs and mostly to the fact that
backscattered electron detection slightly enlarges the narrowest
cracks.
 13-5

3. NUMERICAL COMPUTATION OF THE THERMAL properties of each node (e.g. the local thermal conductivity) and
CONDUCTIVITY are stored as coefficients of the partial derivative equation to be
It has been shown that the morphology of a zirconia plasma solved. The schematic representation of our problem is shown
sprayed coating is relatively complex in so far as it involves on figure 6:
dispersion of pores with various sizes and shapes and connected
crack network. As we are concerned with the influence of
morphology on thermal conductivity and we want all the
morphological information obtained with SEM to be taken into
account, the modelling method must face this structure
complexity without oversimplification.

Analytical or empirical models mentioned in the introduction

only deal with simplified geometrical shape dispersions and are dT/Sn-o
not well suited for connected porosity. The variety of pore
structure families does not allow an efficient use of such models,
which are generally restricted to one simple type of shapes and
cannot handle connectivity effects with a fair accuracy.

Computations on very complex geometry are tractable with

some numerical methods. The finite-element method is well
known for its possibilities in this field. Recent works [29,30]
demonstrate these capabilities by computing the thermal and div(>.grad(T))-0
electrical conductivities of porous metals or metallic oxides.
Figure 6: Problem definition.
The finite-element method, among others, uses a grid
Then we have to build the finite-difference equation to be solved
representation of the geometry. The mesh properties of this grid
for each node. This is done by means of the well known five
are however submitted to some constraints. Then its application
points scheme, that is :
to very complex shapes may lead to some difficulties. As each
shape to be modelled must be decomposed in simple geometric
(A(ij)+B(i,j)+C(i,j)+D(ij))T(ij)
elements (triangles or quadrilaterals), the resulting amount of
nodes and meshes may become tremendous when the geometry
A(i,j)T(ij-l)+B(i,j)T(i-l,j)
presents very small details. For example the minimal shape that
+C(ij)T(i,j+l)+D(i,j)T(i+l,j)
we may encounter is a single square of one pixel size. If we
want to decompose it in triangular or quadrilateral meshes, this
single pixel produces, at least, four nodes. Continuity If we do not want to treat the discontinuities of these A,B,C,D
constraints on the neighbouring meshes may impose some grid coefficients with specific equations occurring when the local
refinements on the vicinity of this isolated pixel, even if the conductivity exhibits a step, an implicit way has to be found. For
closest shapes are very large. As a consequence the resulting example the well known [31] five weighted conductivities
mesh density in the area close to this square may become very scheme widely used in variable (but continuous) properties
high. For thin cracks (their minimal thickness is one pixel) this problems where the coefficients are defined by :
may lead to more dense mesh distribution and untractable
continuity constraints. Therefore, as the solution methods are A(i,j)=l/2(X(i,j-l)+Mi,j));
generally global (e.g. they are often reduced to a minimisation B(i,j)=l/2(X(i-lj)+X(i,j));
problem), the associated algebraic system may reach an C(i,j)=l/2Wi,j+l)+MiJ));
impracticable size. So the finite element method seems to be D(i,j)=l/2(X(i+l,j)+Wi,j)).
difficult to apply in our field of study if no restriction on the
geometry is acceptable. and which is of second order (for continuous properties) is not
applicable in our case because it smoothes the discontinuities.
On the other hand, the standard finite-difference method seems Indeed, one can easily see that for small sized (e.g. one or two
to undergo equivalent restrictions as it needs some grid pixel thickness) objects this scheme leads to erroneous values as
representation, and therefore, some specific processing on the no node is assigned its real conductivity.
internal boundaries (e.g. boundaries between porosity and
matrix). More suited schemes can be constructed using asymmetrical
distributions of conductivity. For example the Upper Left
However, this can be overcome if: scheme we commonly use is defined by :
• the grid can be constructed directly from the input data
image, A(ij)=^(i,j-1);
• the discretisation does not need specific processing on B(i,j)=X(i-lj);
internal boundaries, C(ij)=X(ij);
• the resulting set of algebraic equation remains practicable. D(ij)=Mij).

Specific implementation of the finite difference method Four first order distinct schemes can be constructed on this basis
The grid construction may be reduced to a very simple process if (the four corner implementations). It can be shown that they lead
we use a regular uniform mesh grid where nodes are simply to a space transformation in the neighbourhood of
defined by the pixels (square pixels) of the input image. With discontinuities as they operate a small translation along bisectrix
such a grid no geometric internal boundary information is of ij axes. For random geometries they are not strictly
associated with meshes. These information are related to the equivalent as they produce small variations (about 10"4 or less)
13-6

of the resulting computed conductivities, but as we will see They can be implemented without extra storage so our criterion
below (Fig. 9) this is not the most critical source of uncertainties is satisfied. Provided that we are able to find some well suited
in the present problem. (Fig. 7) convergence acceleration parameter (over-relaxed
Gauss-Seidel method), they can outperform matrix methods
For the main boundaries (e.g. the boundaries where Dirichlet or when applied on scalar or superscalar computer [33], but they
Neumann conditions are applied) we use the standard method, are notably less efficient when deep vectorisation is allowed.
that is the image method based on the same scheme for
Neumann condition (the Dirichlet condition does not need any So we have implemented an over-relaxed Gauss-Seidel solver
computation as T is imposed). which allows the solution of problems with sizes up to about
6000x6000 on our 512Mo RAM workstation. With the
When the temperature is known for each node, the average or commonly used SEM magnification this theoretically allows the
equivalent thermal conductivity of the whole domain is computation of TBC conductivity for thick samples (1.2x1.2
computed from : mm2), or extended areas of relatively thin coatings (for example
0.20x36mm2).
\A—-dx
>L = Gauss-Seidel over-relaxation factor
/ Ti-T0

where e is the height and 1 the width of the domain, the

integration being done on one of the horizontal boundaries. The
space step being constant, that integration can be performed with
standard quadrature formulas.

The solver
The proposed scheme leads to a set of Nx-2*Ny linear equations
to be solved. As we are concerned with large values of Nx and
Ny (computation with Nx~4200 and Ny~5100 has been
performed), the solver has to be as efficient as possible.

The first difficulty is to have the most compact form of the

problem in order to reduce memory occupation. This can be
done efficiently by reducing the coefficient storage. Figure 7: Convergence acceleration parameter.

With the definition of A,B,C,D we have seen before, it can For huge problems, the main limitation is the duration of
easily be shown that, for a simple diphasic problem, we have to computation. The following table shows computation duration in
store only 23 + 22 + 23 different configurations. So the storage for seconds for a standard 1024x1024 problem with 10"4 precision
A,B,C,D can be drastically reduced provided that we can store criterion (e.g. the convergence criterion on heat flux in relative
some case index for each node. form), for three different superscalar computers :

As this case number is in the range of tens, an array of bytes is HP9000/780 DEC Alpha 500 Pentium Pro
convenient. So the minimal storage required is : 260 405 1164
• one Nx*Ny double precision array for T,
• one Nx*Ny byte array for case markers, The computation was achieved with 3000 iterations using a
• 4*20 double precision scalars for coefficients, convergence acceleration parameter of 1.997. The maximal
• one Nx*Ny byte array for input data. residual error on heat flux (relative value) was about 9.4.10"5.

This is about 10 Moctets for a 1024x1024 problem and up to One can consider this case as the simplest one because the input
210 Moctets for a 4200x5100 problem. So if the solver does not data image was of medium complexity. For higher complexity
need intermediate storage, we expect that such problems can be input the maximal computation time may be twice the indicated
solved on common 256Mo RAM computers. The solution time. For the 4200x5100 problem we have solved only on the
method may be chosen with respect to this criterion. The two FTP workstation, the computation time was about 10 hours in the
common approaches for this kind of solver are iterative matrix most difficult case. Notice that the complexity (which can be
methods and iterative by points methods. evaluated as some function of the number of porous objects)
increases roughly as the area of the domain. Despite that, the
The first kind is well represented by the conjugate gradient relation between the size of the domain and the computation
method [32] which, as other matrix methods, does not satisfy time which is a power function of the maximal dimension of the
our criterion. It needs, at least, two or three times the minimal domain exhibits an exponent close to 3, which is the theoretical
storage defined above. So the maximal dimensions of the value for the over-relaxed Gauss-Seidel method.
problems we could solve with such methods would be reduced
by a factor of V2 or V3. Alternatively, we would need some The initial value for T may also strongly influence the speed of
supercomputer.Therefore, matrix methods are very efficient on convergence. Our standard initialisation is the temperature
computers with high vectorisation capabilities (as Cray for distribution for the homogeneous medium. A possibly more
example), but seems to be less attractive on scalar or superscalar efficient initial distribution may be constructed from the
computers (such as common workstations) [33]. Richardson's extrapolation concept [34]. Let Tm be the solution
The second kind is commonly known as the Gauss or Gauss- of the problem for the domain reduced by a factor 2 in each
Seidel method [32]. dimension (e.g. we compute only one node out of two). This
 13-7

solution may be achieved within a time \J%, where tc is the

computation time for the whole problem. Thermal conductivity computations
MP70A square Adds, SOOX magnification.
The values of T are then determined on alternate nodes using
bilinear interpolation. The global cost of these two operations
remains close to tc/8. As we do not need an "exact" solution
for T1/2, we can reduce the precision criterion in order to reduce
this time. So the initial data construction has a cost less than
tc/10 for example. Our major expectation is that such an E '
initialisation would be very close to the whole solution T. Then *
the amount of computation needed to obtain T would be
substantially reduced.

Experiments have shown that the effective global computation

time may be reduced by a factor of about 2 for low complexity 1000 1500
Pixels
input (globular porosity for example). For higher complexity
input the time reduction is often imperceptible as small details
(cracks for example), which are partially lost in the T-> Tm
input transformation, have a major contribution to the whole Figure 9: Thermal conductivity versus resolution.
solution. So this is not a very efficient method for the reduction
of computation duration because of its lack of generality. extra 1024x1024 image (referenced 1-2) is obtained from the
512x512 one by a pixel duplication technique. The computed
Computation output conductivity is a decreasing function of the resolution whereas
The computation program outputs some convergence the toral porosity is nearly a constant. As results for the 512x512
information and the computed equivalent thermal conductivity. and the 1-2 cases are quasi identical we can assert this is not a
Therefore, the resulting array of temperatures is saved in a discretisation effect. The main justification we found is the loss
compacted binary file. This file can be read by another program of small morphological details (loss of crack continuity for
for visualisation purpose. The temperature, its gradient, and the example) which results from the resolution decrease but does
heat flux can be visualised by means of isovalue representation not strongly affect the total porosity. The computed conductivity
as shown on figure 8. seems to be related to the resolution by :

K - Ki + N

where X^ is the value we would obtain for an ideal resolution

and N is the resolution expressed as the number of pixels on one
side of the image. Thus we are able to find a resolution
independent value of the thermal conductivity for each field if
we have two different resolution input images. This is an easy
thing to do as a 512x512 input image can be constructed from a
1024x1024 one by pixel destruction. So all the following results
are presented with three values (the 1024 resolution, the 512
resolution computed values and the ideal resolution extrapolated
value) for the computed conductivities.

The matrix thermal conductivity problem

«H*-».1«Km-Jj0O» Kv-O.OM Kf-IIJrll »u-tt40, i;i.«lMHW.
The second problem concerns the conductivity data of the
different phases. For the computations, the needed data are the
Figure 8: Heat flux representation. morphology and the thermal conductivity of each phase.

For the porous one, we can use the thermal conductivity of air
Such representations performed on images with separated which is well known. The major problem is for the zirconia
porosity give a qualitative useful description of the effect of each matrix thermal conductivity. At present only bibliographical
morphological feature on the heat transfer through a TBC. values are available. They fall between 2 and 3 W.nf'.K"1 [5], so
the related uncertainties on computed thermal conductivities are
4. APPLICATIONS large. Some in-situ evaluations of the matrix conductivity are in
Before applying these tools to various TBCs, three major points progress by means of microscopic thermal diffusivity
have to be considered. measurements on real TBCs. They may lead to more realistic
values of matrix properties in a near future. In the meantime, we
The resolution problem have used an arbitrary value of 2 W.itf'.K"1 for all
The first one concerns the optimal choice for input image computations.
resolution. As figure 9 shows, the computed conductivity
strongly depends on the input image dimensions.
The 2D to 3D effects
Images of the same physical field with various resolutions (e.g. The third point is related to restriction on actual geometry. All
256x256, 512x512, 1024x1024, 2048x2048 pixels) lead to four the data and results we are able to obtain with these methods are
inputs which have been used for conductivity computation. An
13-8

two-dimensional. Three-dimensional effects can heavily change correspond to the thermal conductivity computed from the
the conductivity if inclusions have geometrical properties which global 4234x5148 image resulting from the assembly of the 35
cannot be evaluated with two-dimensional sections. As Bakker fields according to §2.
[29] mentionned, the three-dimensional problem is unreachable The first important result is that the average values of the 35
with current computers. Therefore the construction of the fields computed conductivities are very close to those obtained
relevant three-dimensional input data sets would be for the global field as it can be seen on figure 10.
impracticable with the resolution we use. So the present work is
limited to the two-dimensional approach. Fractographies Thermal conductivity computations
MT70A serf« (cracks) <>,, 2.0 W.m"'.k ')
performed on plasma sprayed TBCs suggest that the cracks,
which are responsible of the main conductivity reduction, have
depth/width ratios far exceeding one and a two-dimensional
section is an acceptable representation for such geometries. But
this is not the case for globular porosity.
Muls> [.
Owing to these various uncertainties the computation results
have to be considered indicative rather than exact.

Results
These tools were first applied to a free standing TBC plasma
sprayed by SNECMA. Extensive analysis and computations
were done on such coating in order to :
Figure 12: Crack contribution.
• test and validate the different procedures,
• study the influence of different classes of porosity on
thermal conductivity. Therefore the decomposition in small (180x180 urn2) domains
with somewhat arbitrary boundary conditions leads to a correct
estimation of the equivalent conductivity of a larger element of
Thermal conductivity computations
MP70A strie (Ai, 2.0 W.m"'.Kr') TBC.

The second important result comes from computations done for

each class of porosity (Fig. 11 and 12).

Thermal conductivity computations

OM-MPTOilOOO H.2.0W.m"'.K"1)
*i("ls) Di8

1.9 •
1.7
»obi

1.3
Figure 10: Plasma sprayed TBC computed thermal
Lilted
conductivity versus porosity. 1.1 Creeks
Cracks \
A set of 35 adjacent (1024x1024 pixels) fields was obtained
0.000 0.023 0.050 0.075 0.100 0.125 0.1:
with the SEM. The thermal conductivity has been computed on £
each field and is represented versus porosity on figure 10
Figure 13: Influence of porosity class.
Thermal conductivity computations
MP70A serfe (gobukr) (*.„ ZO W.nf'.K."')

For almost identical porosity content, fields containing only the

crack network present a thermal conductivity 10 to 30% smaller
than that of fields with only globular pores. Analysis of
computed conductivities reinforce the intuition that the major
conductivity reduction results from the cracks which are mostly
Mills)
oriented in a direction perpendicular to the heat flow (see Fig. 4)
in plasma sprayed TBCs.

This can be demonstrated even more clearly using porosity

reconstruction (see §2) as computation input (Fig. 13). The rate
of variation of the thermal conductivity with the porosity (e.g.
dXJde) strongly depends on the morphological class.
Figure 11: Globular porosity contribution.
As the crack family has the highest rate of variation this class of
(medium line). The lower line corresponds to ideal resolution porosity is the most efficient in reducing thermal conductivity in
images and is derived from calculations with 1024x1024 and the case of plasma sprayed TBCs.
512x512 resolutions as explained above. The open symbols
 13-9

5. CONCLUSION
A software has been developed for the computation of the 12. Rüssel H.W., J. Am. Ceram. Soc. 18, 1939, pi.
thermal conductivity of porous coatings (plasma sprayed TBCs)
from binary images of real material cross sections. This 13. Bruggeman D.A.G., «Berechnung Verschiedener
approach based on a finite difference method takes directly into Physikalischer Konstanten von Heterogenen Substanzen », Ann.
account the actual complex morphology of the ceramic which is Phys. 24(7), 1935, pp. 636-679.
mostly original with regard to existing models. This numerical
model is able to determine the contribution of each 14. Murabayashi M. et al, « Effect of porosity on the thermal
morphological feature to the thermal conductivity and therefore conductivity of Th02», J. Nucl. Sei. Tech. 6,1969, p. 47
compare different microstructures or coating architectures. In a
near future it will be associated to a morphology generator for 15.Schulz B., in High Temperature-High Pressure 13, 1981, pp
building a predictive tool. This one will be used in particular by 649-653.
engine manufacturers as guideline for modifying coating
spraying conditions in order to obtain specific morphologies 16. Koh J.C., Fortini A., «Thermal conductivity and electrical
leading to optimised coating thermal properties. resistivity of porous material », NASA report NAS3-12012, CR
It is important to note that this approach, developed on TBCs is 1200854(1971)
most general and can be applied to a variety of multiphase media
(refractories, composites, etc.). 17. Cunningham M.E., Peddicord K.L., «Heat conduction in
spheres packed in an infinite regular cubical array », Internat. J.
Acknowledgements : Part of the work reported has been carried Heat and Mass Transfer 24, 1991, ppl081-1088.
out within a Brite Euram project [4] and the support of the
European Commission is gratefully acknowledged. Special 18. McLachlan D.S., Blaszkiewicz M., Newnham R.E.,
thanks are due to SNECMA who provided the plasma sprayed « Electrical resistance of composites », J. Am. Ceram. Soc.
coatings. 73(8), 1990, pp 2187-21203.

19. Rhee S.K., « Porosity-thermal conductivity correlations for

References
ceramic materials », Mat. Sei. Eng. 20, 1975, pp 89-93.
1. Sheffler K.D., Gupta D.K., « Current status and future trends
in turbine application of thermal barrier coatings », J. Eng. Gas 20. El-Fekey S.A., El-Mamoon Yehia M.,E1-Hakim M.N.,
Turbines Power (Trans. ASME), 110, 1988, pp. 605-609. «Mathematical analysis of the dependence of thermal
conductivity on porosity », Powder Met. Internat. 10(2), 1978,
2. Bose S. , DeMasi-Marcin J., «Thermal barrier coating pp 90-91.
experience in the gas turbine engine at Pratt & Whitney », in
NASA CP 3312, 1995, pp. 63-78. 21. Jackson T.B., Virkar A.V., More K.L., Dinwiddie R.B.,
Cutler R.A., «High thermal conductivity aluminium nitride
3. Thermal Barrier Coating Workshop. (Cincinnati, 19-21 may ceramics : the effect of thermodynamics, kinetic, and
1997) microstructural factors », J. Amer. Ceram. Soc. 80(6), 1997, pp
1421-1435.
4. High Insulation Thermal Barrier Systems. Brite Euram
Project BE96-3226 (1996). 22. Aizanov M.I., Domashnev I.A., in Poroshkovaya Met.
8,1968, p 51.
5. Youngblood G.E. , Rice R.W. , Ingel R.P., «Thermal
diffusivity of partially and fully stabilised (yttria) zirconia single 23.Bakker K., Kwast H., Cordfunke E.H.P., « The contribution
crystals », J. Am. Ceram. Soc. 71(4), 1988, pp. 255-260. of thermal radiation to the thermal conductivity of porous
U02 », J. Nucl. materials 223, 1995, pp 135-142.
6. Hasselman D.P.H., Johnson L.F., Bentsen L.D., Syed R., Lee
H.L.,Swain M.V., «Thermal diffusivity and conductivity of 24. Da Yu Tzou, « A universal model for the overall thermal
dense polycrystalline Zr02 ceramics : a survey », Am. Ceram. conductivity of porous media», J. Composite Materials 25,
Soc. Bull. 66(5), 1987, pp. 799-806. 1991,ppl064-1084.

7. Kingery W.D., Bowen H.K., Uhlmann D.R., in 25. Furmanski P., « Effective macroscopic description for heat
« Introduction to ceramics ». J. Wiley (NY, 1976). conduction in heterogeneous materials», Int. J. Heat Mass
Transfer 35(11), 1992, pp 3047-3058.
8. Madarasz F.L., Klemens P.G., « Reduction of lattice thermal
conductivity by point defects at intermediate temperatures », 26. McPherson R., « A model for the thermal conductivity of
Internat. J. Thermophysics. 8(2), 1987, pp. 257-262. plasma-sprayed ceramic coatings», Thin Solid Films 112,
1984, pp 89-95.
9. Maloney M.J., Achter H.S.,Barkalow B.K., « Development of
low thermal conductivity thermal barrier coatings». 27. C. Moreau, S. Boire-Lavigne, R.G. Saint-Jacques, «The
Communication presented in Thermal Barrier Coating relationship between the microstructure and thermal diffusivity
Workshop (Cincinnati, may 1997). of plasma-sprayed tungsten coatings», in Proc. of the 7th
National Thermal Spray Conference 20-24 June 1994, Boston
10. Maxwell J.C., «Treatise on Electricity and Magnetism». (MA). Edited by C.C. Berndt and S. Sampath (ASM, 1994), pp
Oxford University Press. 621-626.

11. Eucken A., « Thermal conductivity of ceramic refractory 28. Bjorneklett A., Haukeland L., Wigren J., Kristiansen H,
materials : calculation from thermal conductivity of «Effective medium theory and the thermal conductivity of
constituents », VDI-Forschungsh. 353 (3-4), 1932, pp. 1-16. plasma-sprayed ceramic coatings », J. Mat. Science 29, 1994, pp
4043-4050.
13-10

29. HolHs K.J., «Pore phase mapping and finite-element

modelling of plasma sprayed tungsten coatings », in Advances in
Thermal Spray Science and Technology. Proc. 8th National
Thermal Spray Conference, 11-15 sept. 1995, Houston, Texas
Edit. C.C. Berndt, S. Sampath (ASM), pp 403-408

30. Bakker K., « Using the finite element method to compute the
influence of complex porosity and inclusion structures on the
thermal and electrical conductivity », Int. J. heat Mass Transfer
40(15), 1997, pp 3503-3511.

31. Nogotov E.F., « Application of Numerical Heat Transfer»,

Series in Thermal and Fluids Engineering, McGraw-Hill Book
Company, 1978.

32. Peyret R., Taylor D., « Computational Methods for Fluid

Flow », Springer-Verlag, 1986.

33. Orignac E., « Compared efficiencies of the conjugate

gradient and the SOR methods applied to the solution of a three-
dimensional laplacian », Private communication, ONERA 1997.

34. Anderson D.A., Tannehill J.C., Pletcher R.H.,

« Computational Fluid Mechanics and Heat Transfer »,
Hemisphere Publishing Corporation, 1984.
 14-1

MICROSTRUCTURE AND THERMAL CONDUCTIVITY OF LAYERED

THERMAL BARRIER COATINGS PROCESSED BY PLASMA SPRAY
AND PHYSICAL VAPOR DEPOSITION TECHNIQUES
K. S. Ravichandran, K. An, R. E. Dutton* and S. L. Semiatin*
Department of Metallurgical Engineering
412 WBB, University of Utah, Salt Lake City, UT 84112, USA
•Materials Directorate, WL/MLLM, Wright Laboratory
Wright Patterson Air Force Base, OH 45433, USA

SUMMARY intermediate NiCoCrAlY alloy bond coating to

improve adherence and to reduce oxidation.
Thermal conductivity is an important design The principal TBC material is Zirconia (Zr02)
parameter for thermal barrier coatings. partially stabilized with about 6 to 8 wt.% Y203
Accurate thermal conductivity data is therefore (hereafter referred to as "YSZ"; for example,
required to ensure proper design and reliability of 8YSZ is used to refer to Zr02 stabilized with
gas turbine blades. In the present research, 8wt.% Y203. Hereafter, the composition of
thermal conductivities of A1203 and 8wt.% Y203 Y203 is quoted in wt.%, unless otherwise
stabilized Zr02 (8YSZ) coatings, including indicated) owing to its reasonable toughness (due
monolithic and multilayer configurations, made to transformation toughening), low density, low
by air plasma spray (PS) and electron beam thermal conductivity, high melting point, and
physical vapor deposition (EB-PVD) techniques, good thermal shock resistance.
were determined from the measurements of
thermal diffusivity and specific heat as a Although YSZ has low thermal conductivity, yet
function of temperature. Thermal diffusivity further reductions in thermal conductivity are
was determined by the laser flash technique. caused by porosity and thermal resistance at
Specific heat was determined by a Differential imperfect interfaces in the coating. However,
Scanning Calorimeter (DSC). Detailed analyses there has been limited work [5-9] performed to
of the results indicate that in the case of PS generate an understanding of the coating aspects
coatings, the thermal conductivity is sensitive to that influence the heat transfer characteristics.
coating density (porosity), interfaces between Additionally, the conductivity data reported in
splats as well as the interface between the the literature exhibit significant variations due
coating and the substrate. In the case of EB- to differences in processing parameters and
PVD coatings, it is shown that the multilayer microstructural characteristics. Among the
conductivity is simply a series representation of microstructural variables that are known to have
monolithic coatings, provided comparisons are an effect on thermal conductivity, only the
made under same microstructural condition. effects of the type and the amount of stabilizer
Further, analyses of sensitivity of the laser flash in Zr02 has been studied in a systematic fashion
technique to variations in the coating and the [5, 7, 8]. Further, microstructural differences
substrate parameters, for the coatings evaluated between bulk A1203 and YSZ and the coatings
in this study, were also performed. The results often exacerbates the difficulty in the
are discussed in the context of coating assessment of coating thermal conductivity.
characteristics, reference conductivity data for
dense materials and the sensitivity of the The primary objective of this study was to
measurement method to coating parameters. understand the issues involved in the assessment
of thermal conductivity of A1203 and YSZ made
by PS and EB-PVD techniques. For this
1. INTRODUCTION purpose, thermal conductivity of coatings were
studied from room temperature to 1000°C. The
High temperature materials are often protected results are analyzed in terms of: (i) coating
by the use of thermal barrier coatings (TBCs) porosity, (ii) thermal resistance at the
[1-4]. These coatings are applied by plasma interfaces, (iii) the discrepancy among the
spray (PS) or electron beam physical vapor thermal conductivity data of monolithic
deposition (EB-PVD) techniques with an materials and (iv) the coating microstructure
relative to that of the dense materials. The

Paper presented at an AGARD SMP Meeting on "Thermal Barrier Coatings",

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
14-2

reliability of the conductivity data is assessed in sprayed condition and on the detached coatings
terms of the sensitivity of the laser flash in the heat-treated condition. Measurements
technique to the uncertainties in the coating and were made from room temperature to a
the substrate parameters. temperature of 1000°C. For the high
temperature measurements, the samples were
heated in a vacuum chamber in 100°C steps; a
2. EXPERIMENTAL PROCEDURE thermal diffusivity measurement was made at
each step. Specific heat measurements of A1203
Plasma-sprayed coatings were obtained by and 8YSZ were also made using a standard
spraying 8YSZ and A1203 onto 3 mm thick Rene Perkin-Elmer Model DSC-2 Differential
95 superalloy substrates measuring 62.5 mm X Scanning Calorimeter with sapphire as the
12.5 mm. Prior to deposition, the substrate reference material. Powders, scrapped from the
surfaces were grit blasted to improve coating as-sprayed coatings were used for this purpose.
adherence. Powders of 8YSZ (Metco 204NS; The standard and the sample were subjected to
average particle size: 10 urn) and A1203 the same heat flux as a blank and the differential
(Plasmalloy Al-1010; average particle size: 5 powers required to heat the sample and standard
Jim) were used. A Plasma Technik Spray system at the same rate were determined. From the
with a single spray nozzle and dual powder feeder masses of the sapphire standard and the sample,
was employed. Both monolithic and multilayer the differential power, and the known specific
coatings involving alternating layers of A1203 heat of sapphire, the specific heat of the sample
and 8YSZ were prepared. The coatings was computed. The thermal conductivities of
exhibited residual porosity. Porosity levels were the coatings were determined using the
determined using measurements of coating mass relationship:
and volume as well as by image analysis and k = ocCpp ...(1)
point counting techniques on optical in which k is the thermal conductivity, a is the
micrographs. To determine if residual porosity thermal diffusivity, Cp is the specific heat and p
could be closed by sintering, heat treatment of is the density of the coating. X-ray diffraction
the coatings was performed at 1300°C for 50 hrs analyses of the coatings were performed using
in a furnace under flowing argon. For this Siemens D5000 unit, with CuKa radiation, to
purpose, 10 mm2 size samples, cut using a
diamond wafering blade, were employed. The identify the phases.
coatings detached from the substrate as units
after the heat treatment. The surfaces of the 3. RESULTS:
detached coatings that corresponded to the PLASMA SPRAYED COATINGS
coating-substrate interfaces were
metallographically polished to remove the metal (a) Monolithic Coatings:
oxide layers formed during the heat treatment. Microstructures of the as-plasma sprayed
EB-PVD coatings were made by electron beam coatings are shown in Fig. l(a&b); those after
evaporation of high purity A1203 and 8YSZ heat treatment are illustrated in Fig. 2(a&b).
sputtering targets on to CMSX-4 single crystal The coatings exhibited a porous structure,
superalloy substrates. The substrate temperature typical of air-plasma-sprayed TBCs. A
was about 300-500°C during the deposition comparison of Figures 1 and 2 reveals little
process. Both monolithic and multilayer change in porosity in A1203 and YSZ layers after
coatings involving alternating layers of A1203 heat treatment. Therefore, porosity levels in
and 8YSZ were prepared by this technique. the as-sprayed condition were assumed for the
heat treated coatings.
Thermal diffusivity measurements were made
using the laser flash technique [10]. The laser Table I. Data on Monolithic Coatin es
flash technique involves heating one side of the Property A1,0, 8YSZ
sample with a laser pulse of short duration and Coating thickness (mm) 0.34 0.37
measuring the temperature rise on the other side Porosity (%) 19 12
with an infrared detector. The thermal Density of monolithic (g/cc) 3.9 5.74
diffusivity is determined from the time required Expt. coating density (g/cc) 3.17 5.06
to reach one-half of the peak temperature and a Calc. coating density (g/cc) 3.28 5.13
transient heat conduction analysis of a two-layer
body. From this analysis, the thermal diffusivity
of a single layer coating on a substrate can be The porosity, density and the thickness of each
independently determined. Measurements were coating are presented in Table I. As mentioned
made on the coatings with substrate in the as- above, porosity levels were determined by a
"mass/volume" (direct) measurement, or by
 14-3

procedures were optimized to minimize particle

pull-out. Additionally, the density values
determined using the porosity levels in Table I
and the theoretical densities were nearly in
agreement, suggesting that the density values
determined by the "mass/volume" technique are
reliable.

,Jft
J
y. «•*
200 fxm
p». *'*'$'.'±->'*--. V*.:-'- '.:'..'■* •».;
Fig. 1(a) Microstructure of PS A1203 in the as- ftV!
sprayed condition

•s*y www,. ■■■■»-■ j..\ <

Fig. 2(b) Microstructure of PS 8YSZ in the heat

treated condition

1.4

1.2 <•».
Diff.-Rene95
tt
E
O)
1 e»
—D-Sp. Ht.-Rene 95 5
^ 200 fim -o--Sp.Ht.-AIO
2 3 0,8 4*<
-*--Sp.Ht.-8YSZ X
Q.
Fig. 1(b) Microstructure of PS 8YSZ coating in ,j, &•—t,—*—*
0.6 (0
the as-sprayed condition
0.4
0 200 400 600 800 1000 1200
Temperature (°C)

Fig. 3. Thermal diffusivity and specific heat

data for Rene 95 substrate, A1203 and 8YSZ

Thermal diffusivity and specific heat data for

the Rene 95 substrate, A1203 and 8YSZ are
''■"' »'"' ' * „ " "* "it .'■■ '■>"-•■' '.. '%■ ''■
presented in Fig. 3. Thermal conductivities of
monolithic A1203 and 8YSZ as-sprayed coatings
as a function of temperature, determined
through Eqn. (1) using the data in Fig. 3, are
presented in Figures 4 and 5, respectively. In
these figures, thermal conductivity of as-sprayed
coatings are compared with that of the solid
Fig. 2(a) Microstructure of PS A1203 in the heat materials incorporating the effect of porosity.
treated condition It is to be noted that that the data for solid
A1203 could not be obtained during the period of
image analysis/point counting techniques on this study, due to difficulties in sintering the
polished cross-sections. The various techniques spray powder. Therefore, the thermal
gave reasonable agreement once polishing
14-4

conductivity data reported in literature is (b) Multilayer Coatings

employed in Fig. 4. Microstructures of multilayer coatings: AZ4 and
AZ8 are shown in Figs. 6 (a&b) in the as-
sprayed condition and in Figs. 7 (a&b) after heat
treatment. A refers to A1203 and Z refers to
8YSZ, with the numbers indicating the number
of layers of each material. For example, AZ4
contains 4 layers of A1203 and 4 layers of 8YSZ,
arranged alternatingly. Table II illustrates the
total coating thickness, individual layer size and
the experimentally measured and calculated
porosity levels, for various multilayer coatings
made by the PS technique.

400 600 800 1000 1200

Temperature (°C)

Fig. 4. Thermal conductivity of A1203 coatings

compared with the porosity-corrected data for
various solid A1203 reported in literature
4
—- Dense 8YSZ
3.5 8YSZ with 12%P
(Calc. From Fully Dense Data)
-■—8YSZ Coating (As-sprayed)
3 □ 8YSZ Coating (Heat Treated)

2.5

u 2 200 |nm
3
■o
C 1.5
o Fig. 6(a) Microstructure of PS AZ4 coating in
u
75 1 the as-sprayed condition
€ 0.5
0)
■c
0 _l_
200 400 600 800 1000 1200
Temperature (°C)

Fig. 5. Thermal conductivity of 8YSZ coatings

compared with the porosity-corrected data of
solid 8YSZ

The degree of reduction in thermal conductivity

due to porosity can be calculated from [11]
kp = k[l-P2/3] •(2)
200 \im
in which kp and k are the thermal conductivities
of the porous material (with a porosity of
volume fraction P) and the fully dense material, Fig. 6(b) Microstructure of PS AZ8 coating in
respectively. During the present work, Eqn. (2) the as-sprayed condition
was found to accurately describe the reduction in
room-temperature thermal conductivity due to Thermal conductivity data as a function of
the presence of pores in solid A1203 and Zr02 temperature for multilayer coatings in the as-
ceramics, when evaluated [11] against the data sprayed and sprayed-and-heat treated conditions
on sintered monolithic materials reported in are presented in Figures 8 and 9, respectively.
literature [12-15]. Therefore, Eqn. (2) was used The conductivity data for fully dense materials
to incorporate the effects of porosity on are also included. In addition, estimates of
thermal conductivity in Figures 4 and 5. conductivity of bilayer coatings (assuming A1203
and 8YSZ in series) using either the porosity-
corrected data for fully dense A1203 and 8YSZ or
 14-5

arrangement, and ta and tz are the thickness

fractions of A1203 and 8YSZ layers,
respectively. Because the thickness fractions of
A1203 and YSZ are nearly the same and equal in
all of the multilayer coatings, these calculations
provide a baseline for comparison with the
measured thermal conductivities of the
multilayer coatings.

— Dense Al O
*S^^Ä^» 2
Dense 8% YSZ
3

AI203(19%P) + 8%YSZ(12%F|)
(Calc. From Fully Dense Data)
o AZ1 (As Sprayed)
100 I I | I I'" a AZ2 (As Sprayed)
A AZ4 (As Sprayed)
o AZ8 (As Sprayed)
6 Al 20+8%
3
YSZ

Fig. 7(a) Microstructure of PS AZ4 coating in (Calc. From Expt. Coating Data)

the heat treated condition 10

o
3
•o
c
5 1
75
E

0.1
200 400 600 800 1000 1200
Temperature (°C)

Fig. 8. Comparison of experimental

conductivity data of as-sprayed multilayers with
200 um the data for monolithic dense materials as well
as calculated data for a bilayer coating
Fig. 7(b) Microstructure of PS AZ8 coating in
— Dense Al 2 O 3
the heat treated condition
Dense 8% YSZ
AMD 3 (19%P) + 8% YSZ (12%F])
Table II. Data on Multilayer Coatings 100
Property AZ1 AZ2 AZ4 AZ8
Thickness (mm) 0.36 0.36 0.35 0.35
Each layer size (mm) 0.18 0.09 0.044 0.022
Porosity (%) 15.5 17 12.7 13.7
Density of solid (g/cc) 4.82 4.82 4.82 4.82
Coating density* (g/cc) 4.07 4.0 4.21 4.16
Coating density (g/cc) 4.17 4.12 4.28 4.24
*Experimentally determined
Calculated from porosity and solid density

the experimentally measured data for

400 600 800 1000 1200
monolithic A1203 and 8YSZ coatings, are
presented. These calculations were performed Temperature (°C)
using Equation (3) in the following form:
KoK7
Fig. 9. Comparison of experimental
kc = conductivity data of heat treated multilayer
(katz + kzta) ...(3) coatings with the data for monolithic dense
in which kc is the composite thermal materials as well as calculated data for a bilayer
conductivity with the layers in series coating
14-6

4. DISCUSSION; quantities. In addition, small intensity maxima,

PLASMA SPRAYED COATINGS characteristic of an amorphous phase, such as
Si02, are also present. Amorphous Si02 is
(a) Monolithic Coatings: known [19] to be present as grain boundary
The conductivity levels in the as-sprayed phase in sintered A1203. Due to its low thermal
condition are substantially lower than the conductivity relative to A1203, a significant
porosity-incorporated data of solid materials, change in conductivity, due to small changes in
for both A1203 and 8YSZ coatings (Figs. 4 & 5). Si02 content can be expected. Figure 4 indicates
However, the data of A1203 and 8YSZ free- that the conductivity data of Youngblood et. al.
standing heat treated coatings agreed well with [16] agreed better with that of the heat treated
the porosity-incorporated data of solid material. The lack of agreement with the data
materials. For A1203, the agreement is good of other reference materials, even after
with the data of Youngblood et. al. [16]. The correcting for porosity, could be due to the
rationale behind this choice is discussed in the microstructural differences between the coating
next section. and these materials. Further study is needed to
clarify the role of purity, phases and
The heat treated coatings differed from the as- microstructure on the thermal conductivity of
sprayed in two respects: (i) a high degree of A1203.
elimination of interfaces between the splats in
1000
the coating due to sintering and (ii) the
elimination of the coating/substrate interface,
since the coatings detached after heat treatment. 800
Therefore, the increase in conductivity is
perceived to be due to the elimination of both of 600
these interfaces in the heat treated condition. •e<
Since dense specimens of A1203, for reference 400
5 I
conductivity measurements, could not be made
from the plasma spray powders, conductivity 200 a -
data reported in literature for A1203 were used in
assessing the conductivities of present coatings. V«#«U*»Av»*i«M,
In this context, one of the primary issues is the 80 100
Angle
accuracy of reference data and the equivalency
of the microstructure of the reference material
to that of the coating. For example, the Fig. 10. X-ray diffraction pattern for as-sprayed
literature thermal conductivity data (Fig. 4) for A1203 coating.
solid A1203 differed significantly, possibly due to
variations in microstructure. The data of (b) Multilayer Coatings:
Youngblood et. al. [16] was generated using AD Figure 8 reveals that the thermal conductivities
995 A1203 powder (Coors Ceramics, Inc., calculated from Equation (3), using the data for
Golden, CO.) sintered to >99% theoretical dense materials and accounting for porosity, are
density. The details regarding purity, phases and significantly higher than the experimental
microstructure were not available. The data of measurements for the multilayers. On the other
Touloukian et. al. [17] is for A1203 having hand, the estimates from Equation (3) using the
>98% theoretical density, but the details on experimental conductivity data for the
purity and phases were not available. The data monolithic A1203 and 8YSZ coatings are in
of Santos et. al. [18] was generated using 99.8% reasonable agreement with the multilayer data.
pure A1203 (A-16SG powder from ALCOA) and The latter agreement may be surmised to be due
correcting the measured conductivity data for to the effect of thermal resistance due to splat
residual porosity. Because of the different interfaces and cracks, which is already included
sources of these powders, considerable variability in the conductivities of the monolithic coatings.
in the thermal conductivity data of dense A1203 The conductivities of all of the multilayer
is seen (Fig. 4). Further, since the details on coatings fall in a narrow band, suggesting that
purity, phases and microstructure are not the contribution from the interlayer interfaces
known, a comparison of them with the A1203 in reducing the overall thermal conductivity of
coating is difficult. The X-ray diffraction the coating is relatively small. It should be
pattern of the A1203 coating is shown in Fig. 10. noted that all the coatings had the same overall
The figure indicates that both oc-Al203 and y- thickness with varying number of layers and
A1203 phases are present in significant corresponding layer thicknesses. Therefore, the
 14-7

total number of splat interfaces in each coating interface thermal resistance [5, 11]. This effect
can be expected to be nearly the same. It due to internal interfaces can be appreciated
appears that due to this similarity, the thermal from the data shown in Fig. 9. The predictions
conductivity levels of multilayer coatings for a bilayer, calculated from the porosity-
differed little and thus can be predicted with corrected data of fully dense materials as well as
reasonable accuracy from the experimental data from the experimental data for heat-treated
of monolithic coatings. monolithic coatings using Equation (3), agree
well with the experimental data for multilayer
While the measured conductivity data for all the coatings. This suggests that heat treatment may
as-sprayed multilayer coatings showed a have eliminated the effects due to the imperfect
significant decrease with temperature, the data splat interfaces in the coatings and that the
after heat treatment (Figure 9) were largely primary microstructural factor that influences
independent of temperature. Several thermal conductivity in this condition is the
microstructural factors should be considered in porosity within the layers. However, since the
understanding thermal conductivity changes coatings detached after heat treatment, it was
after heat treatment. Porosity and thermal not possible to determine the relative
resistance at interfaces can significantly contribution of the coating-substrate interface
influence the thermal conductivity in solids versus that of the splat interfaces.
[4, 5, 10, 11, 15]. Since porosity levels changed
only a little, the effect on thermal conductivity 5. RESULTS: EB-PVD COATINGS
due to this change can be considered negligible.
On the other hand, interfaces between the splats Details of EB-PVD coatings, including total
in thermal sprayed coatings have been suggested coating thickness, individual layer thickness and
to reduce the thermal conductivity due to the density are presented in Table III.

Table III. Description of Coatings, Layer Thickness and Density Values for EB-PVD
Coatings with CMSX Single Crystal Substrates.

Calc. Meas. Subst. Total Thick. Thick.

ID Density* Density Thick. Coating of 8YSZ of A1203
Coating Thick. layer layer
Type (gm/cc) (gm/cc) (mm) (mm) (mm) (mm)
Substrate+
PVD2 bond coat** 8.567 8.52 0.915***
PVD4 PVD2 + 1 5.74 5.41 0.865 0.111 0.111
layer 8YSZ
PVD2 + 1
PVD5 layer each of 4.75 4.69 0.749 0.094 0.049 0.045
A1203 and
8YSZ
PVD2 + 4
PVD6 layer each of 4.65 4.71 . 0.609 0.116 0.015 0.014
A1203 and
8YSZ
PVD2 + « 350
PVD7 alt. A1203 and 5.11 4.64 0.614 0.117 0.00065 0.00035
8YSZ layers
PVD2 + «
PVD8 1000 alt. 5.08 4.83 0.485 0.139 0.00064 0.00034
A1203 and
8YSZ layers
* Calculated from theoretical density of A1203 and 8YSZ, using layer thicknesses.
** Substrate+bond coat was treated as a single material in thermal conductivity measurements and
calculations.
*** Refers to the total thickness of bond coat + substrate.
14-8

is the same (Santos et. al. [18]) as the data

reported in Fig. 4. Additionally, single layer
A1203 coating with the same thickness as the
other coatings, could not be deposited without
cracking and spallation. The data calculated on
the basis of series heat transfer model (Eqn. 3)
using the thickness fractions of layers in each
type of coating, is also presented. It can be seen
that the multilayers have higher conductivity
that that of 8YSZ, but lower than the data of bulk
A1203, as expected. It appears that as the
columnar nature of the layers decreased, the
thermal conductivity also decreased. This is
clearly evident by comparing the data of 2 layer
coating and the 8 layer coating.

Fig. 11. Microstructure of single layer EB-PVD

8YSZ coating

wm m

^itfr^iiski frl Vi

L'j!ry. »':-j6i
BSr-i-'lB'.'Ä

H&M

100 |im (a)

Fig. 12. Microstructure of multilayer coating
with 4 alternating layers of each of 8YSZ and
A1203

Microstructures of EB-PVD coatings are

presented in Figs. 11 through 13. Figures 11 and
12 illustrate the microstructures of the single
layer 8YSZ (PVD4) and 8 layer coating involving
4 alternate layers of each of 8YSZ and A1203,
respectively. Figures 12 (a&b) illustrate the
microstructure of the multilayer coating with
about 350 total alternating layers of 8YSZ and
A1203. The microstructure of single layer 8YSZ
coating is an array of fairly closely packed
columns that run perpendicular to the interface
of coating and substrate. As the total number of
layers was increased, the tendency to exhibit the
columnar structure decreased. The measured (b)
thermal conductivity data as a function of
temperature for all the coatings studied are Fig. 13 (a&b). Microstructures of multilayer
presented in Fig. 14. The solid 8YSZ data is for coating with 350 total alternating layers of 8YSZ
the original bulk material that was used to deposit and A1203
the coating. A1203 could not be sintered to
obtain a dense material. Therefore, the data for
solid A1203 was taken from literature. This data
 14-9

Al203 (Monolithic)
8YSZ (Monolithic) 30000
— -N203 (0.54) + 8YSZ (0.46) (series Calc I
AI2Oa (0.59) + 8YSZ (0.41) (series Calc
Al203 (0.35) + 8YSZ (0.65) (series Calc
8YSZ
1 layer each AI203/8YSZ
4 layer each AI203/8YSZ
-350 alt. layers:
*
Al2 0 3 /8YSZ
-1000 alt. layers: Al203 /8YSZ

V — *■---»-
—»
100
_!__!__!__!__,_
400 600 800 1000 1200
Temperature (°C) Fig. 16. X-ray diffraction pattern for the 8YSZ
coating
Fig. 14. A comparison of the experimental
thermal conductivities of EB-PVD coatings with First, it is of interest to see how the relative
the data for solid materials and the calculated data proportion of phases in YSZ, influenced by Y203
for multilayer coatings level, would influence the conductivity of YSZ.
In Fig. 17, the conductivity data of various dense
YSZ specimens reported in literature [16,20,21],
6. DISCUSSION: EB-PVD COATINGS as well as that of solid 8YSZ evaluated in the
present study, are compared. The data differed
First of all, the thermal conductivity data of significantly among them, owing to the variation
8YSZ coating is higher than the bulk 8YSZ data in the amount of Y203 stabilizer. For example,
by a factor of 1.5. In order to understand this increasing Y203 from 5.3wt.% to 9wt.%, a
discrepancy, some microstructural factors are to significant decrease in the conductivity level can
be considered. The solid 8YSZ and the coating be seen, over the entire temperature range.
deposited using it, were found to be not under the However, at llwt.% Y203, the conductivity
same microstructural condition. Whereas the appears to increase, relative to that of 8-9wt.%
solid consisted of a mixture of monoclinic and Y203 stabilized Zr02 specimens. This trend is
tetragonal phases (Fig. 15), the coating consisted consistent with the variation in thermal
almost entirely of tetragonal phase (Fig. 16). The diffusivity as a function of Y203 content, as
type of phase present in the YSZ microstructure observed by Youngblood et. al. [16].
is known [20] to affect the thermal conductivity
and therefore, this factor should be considered in 5
—o— 9YSZ-Hasselman et. al. [20]
the comparison of the monolithic material and E 4.5 --A--Solid 8YSZ-Present study A
the coating. —o—11YSZ-Mg. Elekt.[21]
/
-«-5.3YSZ-Mg. Elekt. [21]
4
-+- -5.3YSZ-Hasselman et. al. [20]
2000 —a--7YSZ-Youngblood et. al.
3.5 (Single Crystal) [16] /'
u
3 3
■o
C
O
Z> O 2.5
« 2
E
1.5

1
200 400 600 800 1000 1200
M. I
& PS
Temperature (°C)
nLyvAv-ju^wji~
60 80 100
Angle
Fig. 17. Thermal conductivity data for YSZ with
various amount of stabilizer content, compiled
Fig. 15. X-ray diffraction pattern for solid 8YSZ from literature
material
14-10

Thermal diffusivities of the coatings with the

Al 203 (Monolithic) substrate, investigated in this study, were
5.3YSZ (Monolithic)-Mg. Elektron
— -Al203 (0.54) + 8YSZ (0.46) (series Calc determined using a two-layer calculation
100 r , AI;Os (0.59) + 8YSZ (0.41) (series Calc procedure which is discussed in detail elsewhere
Al203 (0.35) + 8YSZ (0.65) (series Calc [10]. The input parameters which enter this
• 8YSZ
E . 0 1 layer each Al 0 /8YSZ calculation are the thicknesses, densities and the
\ □ 4 layer each AI,03/8YSZ specific heat values of the coating and the
\ m -360 alt. layers: AI203/8YSZ
substrate, the thermal diffusivity of substrate and
IS -1000 alt. layers: Al 0 /8YSZ
> ' 2 3
the measured half-times. The sensitivity of each
IB
Ü
3
TJ
10 : "'""""■--^
of these parameters also depends on the relative
0
C
O values between these parameters, i.e. the relative
u
n ■ "a" -a- ■■B—y-g-^l | § layer thicknesses, the relative magnitudes of
E
La
•
diffusivities, etc. The situation is further
0)
c
1-
complicated by the fact that the calculation of
the diffusivity (or conductivity) of the coating is
1 1 1 1 u i i i i—i i i i i i i i i i
an iterative procedure. The effect of
200 400 600 800 1000 1200
uncertainties in the input parameters for the PS
Temperature (°C)
and EB-PVD 8YSZ coatings, on their calculated
thermal conductivity values, was determined by
Fig. 18. A comparison of the experimental introducing positive and negative errors in the
thermal conductivities of EB-PVD coatings with parameters. The changes in the coating
the revised data conductivity values are plotted as a function of
positive and negative changes in different input
The study of Hasselman et. al [20] indicates that parameters, in Figs. 19 and 20 for 8YSZ coatings,
in the range of Y203 content varying from 4.5 to deposited by PS and EB-PVD techniques,
5.5 wt.%, Y203, YSZ almost entirely consists of respectively.
tetragonal phase. Above 5.5 wt.% Y203, a
mixture of tetragonal and monoclinic phases was 100
seen. The proportion of monoclinic phase was >
determined to be 25% in the 8.6YSZ specimen. o
This is consistent with the X-ray data of 8YSZ 3
■a
50 -
dense material (Fig. 15) evaluated in the present c
o
study. Therefore, the present 8YSZ coating and U
the 8YSZ dense material are not truly comparable 75
E
u
due to the differences in the amount of tetragonal 0)
and monoclinic phases. The relative proportions
of phases can affect the thermal conductivity of 5 -50 %Ain p orC —-%A in t
s
coatings due to: (i) the second phase-effect on %Ain t
c
%Ain s0
thermal conductivity in composites [19], (ii) — -%Ain p orC ■ %A in a
phonon scattering at the interphase interfaces -100 =c _l_

[22] and (iii) a change in defect structure with -50 -40 -30 -20 -10 0 10 20 30 40 50
composition [23]. In Fig. 18, the thermal % A in Coating / Substrate Parameter
conductivity comparisons employed in Fig. 14
have been revised using the solid 5.3wt% YSZ Fig. 19. Sensitivity of thermal conductivity of
conductivity data, since, this material entirely PS 8YSZ coating to the changes in
consisted of tetragonal phase. As can be seen, coating/substrate parameters
the agreement between the conductivities of solid
8YSZ and 8YSZ coating as well as between the For the PS 8YSZ coating, in the order of
experimental data and the calculated data for decreasing sensitivity, the coating conductivity is
multilayer coatings has improved. It is to be
noted that the variations between the sensitive to: substrate thickness (ts), half-time
conductivity levels of multilayer coatings appear (S0.5), substrate diffusivity (ocs), coating thickness
to be due to differences in the relative layer (tc), coating density (pc) or specific heat (Cpc)
thicknesses of 8YSZ and A1203 layers.
and substrate density (ps) or specific heat (Cps).
7. ANALYSIS OF SENSITIVITY OF For example, for PS 8YSZ coating (Fig. 19), a -
THERMAL CONDUCTIVITY 10/+10 change in the parameter caused changes
MEASUREMENT in the coating conductivity as: -24/+46 for
substrate thickness, +35/-20 for half-time, +22/-
13 for substrate diffusivity, -21/+21 for coating
 14-11

thickness, -11/+11 for coating density or specific substrate diffusivity, coating thickness, coating
heat and +1/-1 for substrate density or specific density or specific heat and substrate density or
heat. specific heat. For example, a -10/+10 change in
the parameter caused changes in the coating
For the present PS coatings, the accuracy of the conductivity as: -44/+°« for substrate thickness,
conductivity data principally depends on the +105/-35 for half-time, +86/-29 for substrate
parameters that cause the most variability in the diffusivity, -22/+27 for coating thickness, -
present coatings. Since substrate thickness could 15/+16 for coating density or specific heat and
be measured to the accuracy of 0.1 mm, the +5/-3 for substrate density or specific heat (Fig.
uncertainty in substrate thickness (0.03%) is too 20). The order of parameters in sensitivity
small to cause any change in coating ranking is similar to that of the PS coatings.
conductivity. The half-times are recorded using
digital oscilloscopes, so the uncertainty in their It is to be noted that for the EB-PVD coatings
measurement can be considered to be better than the variability encountered in coating parameters
1%. The substrate diffusivity was independently are about the same as that of the PS coatings.
measured using the free-standing sample and However, the EB-PVD coating thickness is a
hence the measurements can be considered factor of 4 smaller compared to the PS coatings.
accurate within ±1%. The density of the The substrate thickness was 1 mm. The sample
substrate were determined within ±1% using the used for thermal conductivity measurement was
Archimedes principle. The specific heat values prepared by mechanically grinding the substrate
of the coating and the substrate can also be from its original thickness of 3 mm to 1 mm.
determined to a level of ±1% or better. This After grinding, the average variation in substrate
leaves the coating thickness and the density as thickness was found to be about +/- 0.04 mm.
the most sensitive parameters involved in the This would suggest that the uncertainty in
present study. The coatings thicknesses were of substrate thickness is +A4%. Therefore, the
the order of 400 |J.m, with the coating surface increased sensitivity of conductivity data in the
roughness of the order of 40-50 urn, indicating case of EB-PVD coating, relative to PS coatings,
that the variability in coating thickness is about seems to arise from the reduced thicknesses of
10%. Additionally, the measurements of density the coating and the substrate.
of plasma sprayed coatings could be in error to
some degree, due to the penetration of liquid 8. CONCLUSIONS
medium through the open pores. Therefore, the
present thermal conductivity data for PS coatings Plasma Sprayed Coatings:
can be expected to be subject to these (1) The reductions in the thermal conductivities
uncertainties. of single layer as well as multilayer plasma-
sprayed A1203 and 8YSZ coatings are brought
100 about by porosity and thermal resistance at the
interfaces in the coating. The thermal resistance
appears to arise from the interfaces between the
splats as well as the coating/substrate interface.
The interlayer interfaces in the multilayer
appears to play only a minor role in influencing
the coating thermal conductivity.

(2) The thermal conductivity levels of

multilayered coatings of A1203 and 8YSZ are
comparable to that of the single layer 8YSZ
coating. Therefore, primarily the interface
-50 -40 -30 -20 -10 0 10 20 30 40 50 thermal resistance and secondarily the porosity
% A in Coating / Substrate Parameter appear to be the parameters that can be
manipulated to reduce the thermal conductivity
Fig. 20. Sensitivity of thermal conductivity of of thermal barrier coatings.
EB-PVD 8YSZ coating to the changes in
coating/substrate parameters (3) The uncertainty in the thermal conductivity
values of the A1203 and Zr02 coatings were
For the EB-PVD coating, in the order of determined to be primarily due to the
decreasing sensitivity, the coating conductivity is uncertainties in the coating thickness and density
sensitive to: substrate thickness, half time. values.
14-12

EB-PVD Coatings: 4. Brink, R. C, Trans. ASME: J. Eng. for Gas

(4) Microstructures of EB-PVD coatings Turbines and Power, 111, 570-577 (1989)
consisted of an array of closely packed columnar 5. Pawlowski, L. and Fauchais, P., Int. Metall.
grains. The columnarity decreased as the number Rev., 31, 271-289 (1992)
of layers increased, in the case of multilayer 6. Pawlowski, L., Lombard, D. and Fauchais, P.,
coatings involving A1203 and 8YSZ. J. Vac. Sei. Technol., 3, 2494-2500 (1985)
7. Brandon, J. R. and Taylor, R., Surf. Coat and
(5) The A1203 and 8YSZ layers consisted of Tech., 39/40, 143-151 (1989)
almost entirely Y-A1203 and tetragonal phase, 8. Morrell, P. and Taylor, R., in Advanced
respectively, in the coatings. This is in contrast Ceramics, Vol. 24, Science and Technology
to the observation of a mixture of a/y phases in of Zirconia, 927-943, The American Ceramic
bulk A1203 and monoclinic and tetragonal phases Society, Westerville, OH (1988)
in bulk 8YSZ. 9. Rudajevova, A., Thin Solid Films, 223, 248-
252 (1993)
(6) An accurate evaluation of thermal 10. Taylor, R. E. and Maglic, K. D., Compendium
conductivity of EB-PVD coating requires accurate of Thermophysical Properly Measurement
reference data on monolithic material having the Methods, K. D. Maglic, ed., 305-336, Plenum
same microstructural condition as the coating. Publishing Co., (1984)
For A1203, the relative proportion of a/y phases, 11. Ravichandran, K. S., J. Am. Ceram. Soc,
impurity levels and microstructure may influence Submitted for Publication, (1997).
the reference conductivity data. For YSZ, the 12. Francl, J. and Kingery, W. D, J. Am. Ceram.
nature and the proportion of phases appears to be Soc, 37, 99-101 (1954)
important (tetragonal, cubic or monoclinic). 13. Saegusa, T., Kamata, K., Iida, Y. and Wakao,
N., Heat Transfer-Japanese Research, 3, 47-
(7) Thermal conductivity of multilayer coatings 52 (1974)
can be predicted from the bulk material data, 14. Swain, M. V., Johnson, L. F., Syed, R and
using the series heat transfer model. This Hasselman, ., D. P. H., J. Mater. Sei. Lett., 5,
suggests that there is no apparent effect of 799-802 (1986)
interface thermal resistance due to the interlayer 15. Pei-feng Hsu and Howell, J. R., Expt. Heat
boundaries. Trans., 5, 293-313 (1992)
16. Youngblood, G. E., Rice, R. W. and Ingel, R.
(8) Thermal conductivity data determined by P., J. Am. Ceram. Soc, 71, 255-60 (1988)
laser flash technique can be subject to 17.Touloukian, Y. S., Powell, R. W., Ho, C.Y.
considerable errors due to the variability in and Clemens, P. G., Thermophysical
coating parameters that are used as input in the Properties of Solids, 2, Y. S. Touloukian and
calculations. The errors increase as the coating C. Y. Ho, ed., Plenum Press, New York
and the substrate thicknesses become small. (1970)
18. Santos, W. N. D. and Taylor, R., High
ACKNOWLEDGMENT Temp.-High Press, 25, 89-98 (1993)
19. Kingery, W. D, Introduction to Ceramics,
The research at the University of Utah was John-Wiley & Sons, New York (1967)
supported by a contract (F33615-92-C-5900) 20. Hasselman, D. P. H, Johnson, L. F, Bentsen,
from Wright Laboratory, Wright Patterson AFB, L. D, Syed, R, Lee, H. M. and Swain, M. V.,
OH, through UES Inc., Dayton, OH. The authors Am. Ceram. Soc. Bull, 66, 799-806 (1987)
thank Dr. S. Sampath, State University of New 21. Stevens, R, Zirconia and Zirconia Ceramics,
York, Stony Brook, NY, and Mr. K. Murphy, p. 30, 2nd Edition, Magnesium Elektron Ltd,
Operhall Research Center, Howmet Corporation (1986).
for preparation of coatings and discussions. 22. Buykx, W. J. and Swain, M. W, Advances in
Ceramics, Vol. 12: Science and Technology
REFERENCES of Zirconia II, 518-527, N. Claussen, M.
Ruhle and A. H. Heuer, eds, American
1. Miller, R. A., Surf. Coat, and Tech., 30, 1-11 Ceramic Society, Columbus, OH (1984)
(1987) 23. Gorelev, V. I. and Palguev, S. F, Izv. Akad.
2. Lammermann, H. and Kienel, G., Adv. Mater. Nauk. SSSR Neorg. Mater, 13, 181-182
Processes, 140, 18-23 (1991) (1977)
3. Herman, H., Berndt, C. C. and Wang, H.,
Ceramic Films and Coatings, J. B. Wachtman
and R A. Haber, eds., 131-183, Noyes
Publications, N. J. (1986)
 15-1

EVALUATION OF THERMAL BARRIER COATINGS FROM BURNER RIG TESTS

J-P. Immarigeon, V.R. Parameswaran, D. Chow and D.D. Morphy

Institute for Aerospace Research
National Research Council of Canada
Ottawa, Ontario, K1A 0R6

P. Gougeon, M. Prystay and C. Moreau

Industrial Materials Institute
National Research Council of Canada
Boucherville, Quebec, J4B 6Y4

SUMMARY which in turn undergoes solid state transformation to the

A series of thermal barrier coatings were deposited on monoclinic phase below 1170°C. This latter phase
superalloy pins by various techniques using different change (tetragonal to monoclinic) is accompanied by a
processing parameters, in order to optimize the durability volume increase of about 9%, which will cause cracking
of the coatings for aerospace applications. These coatings and lead to spalling of the ceramic zirconia layer from the
were subjected to simulated high temperature engine bond coat. This phase transformation, however, can be
operating conditions in a high velocity burner rig. The suppressed by the addition of stabilizing agents such as
durability of the coatings at different temperatures was Y203, MgO, Ce02 and CaO, which lower the phase
assessed at regular intervals from changes in pin weight transformation temperature. Of these, yttria has been
due to oxidation and/or loss of coating. The thermal found to be a good stabilizer for the range of turbine
characteristics of some of the coatings were also assessed operating temperatures and partially stabilized zirconia
on-line by monitoring the core temperatures of the pins. containing 8%Y203 (called YPSZ), has been found to
After thermal cycling, the specimens were subjected to perform well in aero engines (1). Other ceramic systems
metallographic examination, to study the change in based on various oxides (e.g. MgO, CeO, and CaO) and
microstructure brought about by the burner rig cyclic oxide mixtures as well as metal silicates (e.g. Ca2Si04,
tests. The test methodology and the results are discussed. ZrSi04) have also been considered for marine
applications where hot corrosion is a concern (2).
1. INTRODUCTION
Thermal barrier coatings (TBC's), as the name implies, TBC's can be applied by one of several methods,
are protective barriers used to isolate turbine components including flame spraying, plasma spraying, detonation
(usually cooled by air flow from the compressor) from gun projection, and electron beam physical vapour
the high temperature of the engine, so that the deposition (EBPVD). Several modifications and
temperature felt by the substrate is considerably lower innovations have been made in these techniques to suit
than that in which the component is operating. Lowering particular applications and to obtain coatings of high
the metal temperature improves the creep life and reduces integrity and durability (3-5). Plasma spraying, for
the thermal stresses induced by transients caused by example, has been carried out under low pressure (partial
accelerations and decelerations in the engine. Thermal vacuum) to produce coatings of higher density. In a
barrier coatings consist of a metallic bond coat, usually recent NASA Tech Brief (October 1996, p 97), it has
MCrAlY (where M stands for Ni, Co, Fe, or a been reported that a vacuum plasma sprayed zirconia
combination of these) of thickness up to 150 urn, over TBC endured five times as many engine stop/start cycles
which a ceramic oxide overlay coating of thickness up to as coatings applied in air (6). Using EBPVD, graded
500 um is applied. The bond coat layer has higher coatings with gradual transition of composition and
oxidation and hot corrosion resistance than the superalloy properties through the thickness have been produced.
substrate material and it also provides a base for This will prevent an abrupt transition zone, where the
mechanical bonding of the ceramic layer to the metal. likelihood of failure can be high due to drastic differences
in the properties of the metallic and ceramic layers (7).
The most widely used ceramic overlay coating is based Nano-structured TBC's (containing nanograins or
on zirconia, which has a thermal expansion coefficient nanolayers) could provide the basis for new generations
close to that of the superalloy substrate. Pure zirconia, of TBC's (8).
however, undergoes phase transformation in the
temperature range of gas turbine engine operation. Above The microstructures of TBC's vary significantly with
2370°C, Zr02 exists in the cubic form and as it cools deposition technique. Typical microstructures of coatings
down below 2370°C, it transforms to the tetragonal form, applied by air plasma spraying (APS), vacuum plasma

Paper presented at an AGARD SMP Meeting on "Thermal Barrier Coatings'

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
15-2

i s pas» RPÄHRiwii. m\
0491S6 20KV 979321 29KV

(a) (b) (c)

Figure 1 Representative microstructures of TBC's applied by (a) plasma spraying in air (b) vacuum plasma
spraying and (c) EBPVD (S: substrate, B: bond coat, C: ceramic overlay layer, N: electroless Ni film used to
protect ceramic for metallography).

spraying (VPS) and electron beam physical vapour known to fail above this oxide layer. The mechanisms of
deposition (EBPVD) are shown in Figure 1. APS the reactions of the coatings with the environment in the
coatings (Fig. la) usually exhibit a layered structure, with engine and the ultimate coating failure are discussed in
a swirl pattern and a large amount of voids and pores. detail, elsewhere (8-10).
The pores help to increase the insulating capacity of the
oxide. VPS coatings (Fig. lb) are also layered but are Evaluating Thermal Barrier Coatings
denser than APS coatings and the interface between the The effectiveness of a TBC coating is assessed from its
bond coat and the ceramic layer is generally more ability to stay in place without spalling. The best way to
uniform than in air sprayed products. EBPVD coatings assess coating durability is from actual engine tests.
(Fig. lc) have a dense columnar structure, which can These, however, are expensive and time consuming.
accommodate the strains arising from differential thermal Alternatively, the performance of a coating can be
expansion between the metallic and ceramic layers. evaluated in a burner rig, where engine operating
conditions can be closely simulated. Engine
Failure Modes of Thermal Barrier Coatings manufacturers have used burner rig tests in the past to
The life of the overlay ceramic is limited by its evaluate and compare the performance of coatings for
susceptibility to thermal fatigue cracking, while the life of screening purposes, prior to actual engine tests. In
the bond coat is limited by its oxidation and corrosion Canada, military operational groups have also relied on
behaviour. Hot corrosion of the ceramic overlay may also burner rig tests to assess the potential of emerging
be life limiting in marine environments. The combined coatings for use in their engines, under various
effects of stress, time at high temperatures and oxidizing component retrofit programs. However, there is no
environment are complex, causing changes in the standard procedure established for burner rig tests.
microstructure of the coatings as well as the substrates. Manufacturers and users adopt different cycling times
These changes eventually will lead to failure of the and temperatures (heating times from a few minutes up to
coatings. Transverse and longitudinal cracks form at the an hour, at temperatures from 1000°C to 1500°C,
surface and inside the ceramic layer, due to thermally followed by cooling for a few minutes) tailored to suit
induced stresses. Oxidizing gases penetrate through the their end use.
transverse cracks and oxidize the bond coat. The layer of
A1203, which forms at the interface between the bond Objectives of the present study
coat and the ceramic (and possibly between the bond coat At NRC, several studies have been performed using
and substrate), thickens during service. In this layer, burner rig testing to compare TBC's produced by
oxidation products such as Cr203, NiO, Hf02 and spinels different techniques and different processing conditions,
(NiO-Cr203-Al203) are formed by loss of ß-NiAl from and to optimize some of the coatings for use in Canadian
the bond coat. In the case of EBPVD coatings, cracks Forces engines. The objective of this paper is to discuss
form in this layer of oxides, causing the top ceramic the testing procedures and methodologies employed in
coating to spall, whereas plasma sprayed coatings are these studies with a view to evaluating the suitability and
 15-3

practicality of burner rig testing as a laboratory ranking of combustor chemistry are readily reproduced and
method for TBC's. The influence of key test parameters controlled. This is particularly important when assessing
on modes and rates of damage accumulation is discussed. the effects of gas/coating reactions on the durability of
Ways to ensure that the modes known to be life limiting coatings (14,15). Fuel (JP-4), along with compressed air,
in engines are reproduced during rig testing are also is introduced through a pressure-atomizing nozzle into
addressed. the primary section of the combustor, within which
combustion occurs in a swirl stabilized recirculation
2. EXPERIMENTAL MATERIALS AND zone. The expanding combustion gases are cooled by
METHODS dilution within a secondary section, and are exhausted
through a nozzle in front of which the test pieces are
2.1 Materials and Test Coupons placed for exposure to high velocity hot gases. By
The TBC's investigated in this work included independently adjusting the primary and secondary air
experimental coatings produced by EBPVD, APS and flows, along with the fuel flow, it is possible to vary the
VPS, under three development programs, as well as gas velocity and gas temperature independently of each
TBC's produced by established techniques by other. With a standard 2-inch axisymmetric circular
commercial vendors. In the first series of tests, the exhaust nozzle, gas velocity can be varied from Mach 0.2
performance of experimental EBPVD TBC's having to Mach 0.8 and gas temperature from 540°C (1000°F) to
compositionally and functionally graded interfaces 1650°C (3000°F). The latter is controlled by a
between the metallic bond coat and the ceramic layer, programmable microprocessor and measured by a
was evaluated. These coatings were applied to nickel base thermocouple located near the throat of the exhaust
superalloy pins, 0.25" diameter and 4" long, made from nozzle. Nozzles of different sizes can be used to vary the
IN625 and a proprietary single crystal (7). In the second cross sectional geometry of the hot gas jet to
series of tests, zirconia and calcia based TBC's were accommodate different test piece geometries. In the
plasma sprayed onto IN625 pins, 0.5" diameter and 5.5" present studies, the standard 2-inch diameter nozzle was
long, using different conditions of particle velocities and used when testing the Vi" diameter pins and a 3-inch
temperatures, and studied to assess the effects of oversized nozzle was used for the Vi" diameter pins. In
composition and processing conditions on product both cases, the pins were positioned 1.5" away from the
performance (11,12). In the third series of tests, the exhaust nozzle.
performances of a non-graded EBPVD TBC, TBC's with
VPS bond coats and APS ceramic overlays and TBC's The rig in the standard 2-inch nozzle configuration is
With VPS bond coats and VPS overlays, were also shown in Figure 2, with a set of uncoated Vi inch
studied. These latter coatings were applied to the same diameter pins positioned in front of the exhaust nozzle.
pins as those used for the first series of tests. The test pins are mounted on a rotating carrousel, used to
translate the pins back and forth between the hot gases
and a jet of room temperature air from a 2-inch diameter
pipe located beside the hot nozzle. In this way, the pins
can be alternately heated and cooled, simulating thermal
cycling in engines. Rotation of the pins, typically at 100
to 200 rpm, is used to promote gas mixing and ensure
uniform exposure of the pins to the hot and the cold
gases. An electrical slip ring assembly mounted on the
drive shaft of the rotating carrousel can be used for
instrumentation of 8 of the 12 test pins.

The rig in its 3-inch nozzle configuration is shown in

Figure 3, with a set of coated Vi inch diameter pins
positioned in front of the cooling nozzle. With the
oversized carousel required to handle the Vi inch diameter
Figure 2 The IAR burner rig equipped with the 2- pins, the larger exhaust nozzle was needed to achieve
inch exhaust nozzle for testing % inch diameter test efficient heating of the pins. However, at equivalent air to
pins. The slip rig assembly accommodates up to fuel mass flow ratios, the gas velocities and the maximum
eight thermocouples for instrumentation of test pins heating rates and temperatures achievable with the 3-inch
exhaust nozzle are not as large as those produced with the
2.2 Testing and Evaluation 2-inch nozzle, which has a bearing on the modes and
The coated pins were subjected to thermal cycling in a rates of damage accumulation in coatings, as discussed
high velocity/atmospheric burner rig. The design of this below.
rig is based on a dynamic combustor similar to that used
in gas turbine engines (13). The mixing and flow in the In the present studies, the temperatures at various
combustor closely approximate conditions that exist locations at the surface and in the interior of the pins were
within conventional combustors and the critical features monitored by thermocouples mounted flush with the
surface and inside an axial blind hole. The changes in pin
15-4

surface temperature distribution, as a function of gas while a close-up view of the spring loaded mechanism is
temperature and gas velocity and the changes in metal shown in Figure 6, with the test pins removed from the
temperature at the core of the pins, as a function of carrousel, but with the thermocouples still in place. The
number of thermal cycles, were established in this spring-loaded device ensured that the thermocouple beads
manner. For surface temperature measurements, holes were kept in contact with the bottom ends of the holes,
drilled radially and along the axes of the pins were used irrespective of the degree of thermal expansion and
as feedthroughs for the thermocouple leads, to avoid contraction experienced by the pins during thermal
disturbing the heat transfer boundary layer around the cycling. The position of the bottom of the hole was
pins. Several pins with one thermocouple each, and with chosen to correspond to the position within the hot gas jet
thermocouple beads regularly spaced along the axes of known to experience the highest temperature, in order to
the pins, were used for these measurements. A schematic record the maximum metal core temperatures.
cross section of the instrumented Vi inch diameter pins,
where all the bead positions have been combined in a
single drawing for simplicity, is shown in Figure 4. The
surface temperature measurements were performed on
bare pins.

Figure 5 Schematic cross section of the

instrumented Vi inch pins showing the axial blind
hole within which thermocouples were held in
positive contact with the bottom end of the holes to
measure metal core temperatures of the pins
reliably.

Figure 3 The IAR burner rig equipped with the 3-

inch exhaust nozzle for testing % inch diameter test
pins.

0.1» W lWfiOUCM '

3/< I 3/1 I J/< I J/4 '- ■■■• -J -

r jr 7 j £-i h _T

Figure 6 Close up view of the instrumented

Figure 4 Schematic cross section of the carrousel showing the spring loaded thermocouple
instrumented Vi inch pins showing locations of retention fixture which ensured the thermocouple
surface thermocouples positioned at regular beads were kept in contact with the test pins (not
intervals along the pin length. shown) during thermal cycling.

Internal temperatures were only measured in the case of Burner rig tests involved up to 2150 cycles of 4 minutes
the Vi inch diameter pins. The intent was to track the of heating in the high velocity hot gas jet followed by 2
degradation in thermal protection efficiency of their minutes of forced air cooling. In some cases, the gas
TBC's as a function of number of thermal cycles, or test temperature was increased in a stepwise manner, from a
time, through measurements of drifts in heating and minimum of 1300°C to a maximum of 1500°C, to
cooling rates during thermal cycling. To measure the increase the severity of the tests and ensure that failure of
internal metal temperatures, thermocouples (one per pin) the coatings occurred before the end of the tests. The
were mounted with their beads pushed against the bottom surface conditions of the pins and their weight to the
of a blind hole drilled along the axes of the pins, using a nearest 0.5 mg were recorded before testing and after
spring-loaded mechanism. A schematic of the every 50 to 150 cycles. The weight changes provided a
instrumented Vi inch diameter pin is shown in Figure 5, measure of the spalling resistance of the various TBC's to
 15-5

the combined effects of thermal loads and hot gas However, the coated pins reach higher steady state core
erosion. Changes in the heating and cooling rates of the temperatures than do the bare pins, as the rates of heat
pins during thermal cycling were also extracted on-line loss from the free ends of the coated pins are reduced
from internal temperature measurements. Finally, the relative to bare pins. Also, the internal temperatures of
microstructures of the TBC's were compared before and both bare and coated pins are higher than their surface
after testing, and the extent of thermally induced temperatures. This difference increases with gas
microstructural changes, including phase transformation temperature and is higher in coated pins, as shown in
as revealed by x-ray diffraction, was assessed. Table 1.

3. RESULTS AND DISCUSSION

3.1 Effects of Pin Size and Testing Conditions on Pin

Temperatures
The variation of pin temperature during thermal cycling
depends on the heat transferred between the pin and the
hot and cold gas jets. This heat transfer is discussed in
detail and modelled elsewhere, for the case of an airfoil-
like test coupon (16). Considerable heat losses arise from
radiation from the surface of the pins and by conduction
towards their free ends and into the carrousel body.
Consequently, the maximum temperatures reached by the
pins are considerably lower than the maximum
temperature of the impinging hot gases. A record of core Time (s)
temperatures over a complete thermal cycle (4 minutes of
heating in Mach 0.22 1500°C gas, followed by 2 minutes
of air cooling), for 12 coated V2 inch diameter pins, is Figure 8. Surface and internal temperature histories
shown in Figure 7. Pin temperatures are seen to rise revealed by instrumented test pins over parts of the
rapidly at first until, after about two minutes, they heating portion of a thermal cycle involving 4
stabilize. The maximum steady state temperature reached minutes of heating in Mach 0.22, 1500°C gas.
by the pins at the end of the heating cycle depends to
some degree on the TBC and, on average, is about 350°C
lower than the gas temperature. 1600

SFT3-150 O 1200
1400
' '"\ r
;
1400 "i" iooo *.$) JWlv' ■■■:^I t

Tgas:1500 C | Hfth°-22 ' 800 r i ■ ■■ ■

\\~>
1200
^ssss=s=- s0) ♦ Exhaust Gas C * T/C 1 C
P 1000 a. 600 \
T/C 7 C T/C 2 C
E
£
3 800 /''' \, ,» 400 -L * T/C 3 C • T/C 4 C
\
2 \
r V
/: Zireonii baud coatings
g. 600
200 • T/C 6 C
E
//" \ \ 0 1 1 ^ ■■"■
I» 400
/ Thick cilcia band coaling
^W v
100 200 300 400
200
Time (s)
/
100 200 300
Time (s) Figure 9 Surface temperature histories revealed by
instrumented % inch diameter pins over parts of the
Figure 7 Typical thermal cycles experienced by Vi heating portion of a thermal cycle involving 4
inch diameter pins as revealed by thermocouples minutes of heating in Mach 0.45 1400°C gas.
inside axial blind holes in pins (see Figure 5).
The heating and cooling rates for V* inch diameter pins
The temperatures of a series of instrumented Vi inch during thermal cycling, and the maximum metal
diameter pins, as recorded by surface and internal temperatures reached under steady state, were
thermocouples over parts of the heating portion of a significantly higher than those observed on the Vi inch
thermal cycle, are shown in Figure 8. These data were diameter pins, when tested at equivalent gas temperatures
obtained from 5 uncoated pins (Tl to T5), each equipped and mass flow rates of hot gas. For instance, at a gas
with a surface thermocouple, and 2 coated (Tc-Ca8 and temperature of 1400°C, with primary and secondary
Tc-Cal 1) and one bare pin (Tc-bare), each equipped with airflows adjusted to give a gas velocity of Mach 0.45, less
one internal thermocouple. The core temperatures of the than 20 seconds were needed for V* inch pins to reach
coated pins rise more slowly than those of the bare pins, their maximum steady state temperatures, as shown in
due to the thermal insulating effect of the TBC's.
15-6

Temperature Distribution Along the Length of 0.25 Inch Dla.

Temperature Distribution along Length of 0.5 Diameter Test _Test Pinslas a Functionof ^ra Tern garature and yejocttjr
Pins as a Function of Gas Temperature
I
1 1 "--
•^—-•l

1100 ü S^ i
1 ! ' I
I
O • 1500 ^V;
2^ 1000 3
S> N |600
<1400 • 1200 C Flame, PI- 800 P2* 50.0 • 1100CFI8TO. P1-47,5 P2-20.0
3 0)
a 900 Q. • 1200 C Flame, PI- 47.5 P2- 20 0 1300 C Flame. P1 - 47.5 P2- 20.0

o 1300 £ 400 / « 1100CFIame,P1-500P2-400 • 1100 C Flame, P1- 00.0 P2- 50.0

a. -"" + 1200 C Flame, PI-45.0 P2-20.0 A 1200C Flame, P1-500P2-400
E 800 1200 -
^ ■ 1300 CF1BTO,P1- 50.0 P2- 40.0 1300 C Flame. P1 - 45.0 P2- 20.0
0)
K> ■ 1100
• 110OCFtame,P1-45.0P2-2O0
* 1400CFtoTO,P1"47.5P2-20.0
Poly. (1400 C Flame, PI- 500 P2* 400)
* 1400CFIame.P1-450P2-200
x 1400 C Rame, PI- 500 P2- 400
Poly, (1100 C Flame, P1- 45.0 P2- 20 0)
• 1000

§JC 6 #>C 5 WC 4 9IC 3 «C 2

4 3 2 1
Positions along pins Distance From End of Pin in 0.25 Inch Increments

(a) (b)

Figure 10 Profiles of steady state pin (bare) surface temperatures for (a) Vz inch diameter pins, and (b) %
inch diameter pins, measured at different gas temperatures and gas velocities.

temperature, owing to decreased heat losses from the free

Tgas Mach I surface I core 'core AT AT end. Conversely, increasing the pin length will result in
°c # (max) {bare) (coated) (bare) (coated) steeper longitudinal thermal gradients. These gradients
1300 0.2 989 996 999 7 10 can be used to advantage, and their shape optimized, to
1400 0.22 1031 1048 1067 17 36 study the combined effects of thermal fatigue, gas erosion
1500 0.22 1097 1122 1146 25 49 and hot corrosion of ceramic overlays and bond coats.
With corrosive contaminants (sulfur in fuel plus atomized
Table 1 Comparison of surface temperatures, axial sea-salt solution in combustor air) added to the hot gas
temperatures and differences between the two for jet, those regions of the pins exposed to surface
bare and coated Vz inch pins exposed to different temperatures between 700°C to 900°C may suffer fluxing
gas temperatures and velocities. of the ceramic oxide overlays leading to breakdown of
the TBC system. The contaminants may also penetrate
Figure 9. This compares with more than 80 seconds for along thermally induced cracks and attack the bond coat.
Vz inch pins, tested at a higher gas temperature of 1500°C, These forms of hot corrosion could affect the regions of
and the same primary and secondary air flows, Figure 7. coated components where temperature drop below
When both types of pins were heated in a 1400°C gas jet, 900°C, in engines operating in marine environment. The
the maximum steady state temperature of V* inch diameter burner rig test is an ideal tool to study the susceptibility
pins was of the order of 1100°C, whereas it was only of different TBC systems to this form of attack, and its
1040°C for Vz inch pins. This is shown in Figure 10 interactions with other forms of thermally driven damage,
which compares the surface temperatures of % inch and should be investigated.
Vz inch diameter bare pins measured at different axial
positions along their lengths, over a broad range of gas For durability assessment of coatings, it may be
temperatures and velocities (i.e. under different ratios of preferable to use V* inch rather than Vz inch diameter pins,
primary and secondary gas pressures). The differences in because in addition to being more severely thermally
temperature at equivalent mass flow rates of hot gas can shocked, the V* inch pins will accumulate more time at
be ascribed to the higher velocities that were achieved maximum temperature. Thus, over the same number of
with the smaller exhaust nozzle used with lA inch pins,
typically Mach 0.45 for V* inch pins vs. Mach 0.22 for Vz
thermal cycles the bond coat on a 'A inch pin will be
oxidized more than on a Vz inch diameter pin.
I
inch pins. The temperature gradients along the length of Accumulation of oxidation damage, can be life limiting in
both types of pin were, however, qualitatively similar. many applications (8-10) and is more easily simulated
with V* inch diameter pins.
These temperature gradients simulate those present on
actual engine parts and can be adjusted to meet specific 3.2 Damage Accumulation
test requirements by varying pin length or pin position The first sign of damage incurred by the TBC's during
within the hot gas jet. At any given gas temperature and thermal cycling was in the form of a network of cracks
velocity, minimizing the length of the free end of a pin developed over regions of both the V* and Vz inch pins
extending outside the gas jet, will increase maximum pin
 15-7

exposed to the hot gases. The cracked surface of an spalling initiation varied with the deposition method, and
EBPVD coating on a Vi inch pin, after some thermal even more significantly, with processing conditions for
cycling had been applied, is shown in Figure 11. each of the deposition methods. Some coatings from all
four forms of TBC's investigated (EBPVD, APS,
VPS/APS and VPS) were found to survive temperatures
up to 1500°C for several hundred cycles, without
spalling. A series of APS TBC's, after rig testing are
shown in Figure 13. The 1.2 mm thick calcia based
coating, (pin (a) in Figure 13), shows no evidence of
spalling, whereas the zirconia based TBC's (all about 300
um thick, pins (b) to (f),) exhibit varying degrees of
delamination.

Figure 11 Network of surface cracks formed on

EBPVD TBC's soon after initiation of thermal cycling
in burner rig.

The cracks in this instance penetrate right down to the

bond coat, as revealed in Figure 12. A thin oxide layer
has formed at the ceramic bond coat interface, where
delamination occurred. Networks of cracks were also
observed in plasma sprayed coatings. The density of
cracks and their morphology depended on processing
conditions and the degree of cracking was usually more
severe in thicker coatings. Cracks were present even in (a) (b) (c) (d) (e) (f)
as-deposited APS coatings, when the ceramic layer was
thick (300-400 urn). Figure 13 Conditions of a series of APS coated 1/4
inch pins at the end of burner rig testing, involving
2150 cycles at temperatures up to 1500°C; (a) 1.2
mm thick calcia-based TBC; (b) to (fj zirconia-based
TBC's applied under different processing conditions.

LwnaHifBB nnwraira nro

079332 29KV X359 86um

Figure 12 Metallographic cross-section through

EBPVD TBC showing transverse crack through Figure 14. Micrographs illustrating onset of failure
coating penetrating all the way down to the bond mechanism in zirconia based APS TBC, by cracking
coat. above the interface with the bond coat.
Ablation from the surface of the TBC's due to erosive A thin layer of ceramic coat remained attached to the
action of the high velocity hot gases was also observed on surface of the hottest sections of the zirconia pins. This is
all the coatings examined. In the latter stages of the tests, due to spalling occurring in these APS coatings by
most of the TBC's experienced spalling, or delamination, propagation and linking of cracks, about 30 um above the
of their ceramic topcoats. The number of cycles to
15-8

surface of the bond coat, as shown in Figure 14. The trends are due to the rate of bond coat oxidation
vertical cracks may relieve the stresses at the interface, increasing with temperature.
thereby reducing the driving force for delamination. One
of the pins, (b) in Figure 13, shows evidence of bond The data shown in Figure 16 also reveal that processing
coat oxidation, where the ceramic topcoat has been conditions during deposition can be adjusted to achieve
entirely eroded away by the hot gases. These observations improved performance with VPS coatings. The optimized
are generally consistent with expected modes of failure VPS coating showed minimum weight changes for more
for APS coatings (8-10). than 1000 cycles at gas temperatures up to 1400°C,
indicative of good resistance to ablation and spalling.
3.3 Damage Monitoring and Quantification However, it spalled off rapidly after the temperature was
raised to 1500°C, as did the APS coatings.
Weight Change Measurements
Changes in the weight of some coated V% inch pins,
measured over 1150 thermal cycles, are presented in
Figure 15. The weight losses observed for all the TBC's
are due to gas-induced ablation and spalling of the
topcoats, during the test. The thick APS calcia-based
coating appeared to have incurred the least damage in this
test, while the performance of APS zirconia-based
coatings varied greatly with processing conditions. The
relationships between processing conditions,
microstructures and properties of APS coatings are being
investigated at NRC and the results will be published
elsewhere (11,12).

Numbw of CyclM

E
O)
\ p~Vs \
Calcla bind
i
Figure 16 Changes in the mass of a series of
0)
01
c ■::,
V
\
TBC
1
-^
EBPVD, APS and VPS coatings applied to % inch
n ZlrconEa b* ml
1
pins as a function of number of rig thermal cycles.
TBC
\ The thermal cycling schedule is shown in the figure
\ "■-•'•

D)

I All materials previously exposed to

\ Only the EBPVD coating survived the entire duration of
the thermal cycling at 1500°C, because of the higher
' --..
any we ig ht change \ being no ted ""--> strain tolerance of its columnar microstructure. The
apparent shortcoming of APS coatings should be viewed
with caution, however, because the test conditions were
Number of Thermal Cycles not entirely representative of service conditions. The
stresses within the coatings, caused by the difference in
Figure 15 Changes in mass of a series of APS thermal expansion between the coatings and substrates
coated Vz inch pins as a function of number of were much larger in pins than those expected to prevail in
thermal cycles over a test 100 cycles at 1350°C actual service parts, which are cooled by internal air flow.
followed by 1050 cycles at 1500°C. The internal metal temperatures of the uncooled pins
were actually greater than the surface temperatures, by up
Changes in the weight of some coated V* inch pins to 50°C at the centre of the pins under steady state
processed by EBPVD, VPS (bond)/APS (ceramic) and conditions, as seen from Table 1. Under these conditions,
VPS are presented in Figure 16. All coatings experienced the ceramic/metallic bond coat interface would be
weight gains at the onset of testing. This is from subjected to very large shear stresses conducive to
oxidation of the bond coats and, possibly absorption of accelerated spalling, as observed experimentally Thus the
oxygen by the substoichiometric zirconia of the VPS coatings on pins may have faced conditions that were too
coatings. Because the V* inch pins spent more time at the severe, possibly representing an extreme limiting case for
maximum gas temperature, as compared to V2 inch pins, plasma sprayed coatings. Using internally cooled pins
(see Figures 7 and 9 for comparison), there was more and ensuring that adequate time is spent at maximum
time available for their metallic bond coats to oxidize, in temperature during thermal cycling, to promote bond coat
the course of the test. This was followed by weight loss oxidation as experienced in service, would provide a
from ablation and spalling of the ceramic topcoats, until more realistic testing environment and possibly different
the temperature was raised from 1200°C to 1300°C and TBC ranking. Work is underway at IAR to address this
higher, beyond which the weights increased again. These issue.
 15-9

Heating Rates Measurements were used in the present experiments tended to oxidize
The maximum heating rates experienced by coated !4 and fail by thermal fatigue. Type S thermocouples should
pins, as measured by thermocouples located inside the perform better in this application and ought to be tested.
pin's axial blind holes, are plotted in Figure 17 as a Also, the measurements should be made on pins that are
function of the number of thermal cycles. Each data point internally cooled to better simulate the thermal gradients
in Figure 17 corresponds to the maximum slope of the and levels of thermally induced stresses experienced by
rising portions of the temperature/time curves obtained TBC's applied to internally cooled components.
for each of the thermal cycles, examples of which are
shown in Figures 7 and 8. There are several features in CONCLUSIONS
Figure 17 that warrant comment. Firstly, the This work has shown that the modes and rates of
discontinuities in the data arise from interruptions in the degradation of TBC's during burner rig testing of coated
test sequence required to weigh the pins, following blocks solid round pins are influenced by many test parameters,
of 5, 50 or 150 thermal cycles. The initial sharp drop in besides being influenced by differences in deposition
heating rates at the onset of each block is due to the pins method and processing conditions.
being at room temperature at the onset of the first cycle in
a block, and at greater than room temperature for Pins, V* inch in diameter, heat up and cool down much
subsequent cycles, because the pins do not cool down faster than Vi inch pins. They also reach a higher
fully to room temperature at the end of a cycle (see temperature and spend more time at maximum
Figure 7). temperature over the duration of a thermal cycle. Thus,
coatings applied to Vt inch pins will be subjected to
TBC Efficency Breakdown higher transient and steady state stresses, and their bond
coat will oxidize more in the course of a test, as
compared to Vi inch pins.
i ♦ ►

/
12 V& Over the steady state portion of a thermal cycle, the
"

sir K
2 1 internal temperature of a solid pin is higher than its
2 11 *l V'
.«. ¥ t
surface temperature. This difference is larger in a TBC
w
coated pin than in a bare pin. High internal temperatures
► in coated pins cause high shear stresses near the interface
Zirco lia between the ceramic and metallic layers of the TBC,
owing to differential thermal expansion between the two
0 200 400 600 800 1000 1200
media, thereby promoting delamination near this
interface. Because of their highly strain tolerant columnar
Cycle Number
microstructure, EBPVD coatings generally fared better
than thermally sprayed coatings under such testing
Figure 17 Changes in heating rates of a V2 pin conditions. Using internally cooled pins, to better
coated with zirconia as function of number of simulate service conditions, would reduce the magnitude
thermal cycles applied to the pin, showing change in of these stresses and provide a more realistic laboratory
heat transfer characteristics of coated system. ranking of TBC coatings.

As the temperature at the start of a cycle increases, the The performance of each type of TBC examined, is a
apparent heating rate inside the pin decreases. It takes strong function of processing conditions. Analysis of
several cycles for heating rates to become stable because damage modes arising from thermal cycling in a rig
of the complex heat transfer transients affecting the pins confirms that plasma sprayed coatings spall by cracking
and their supporting carrousel. Secondly, the general of the ceramic overlay, above the interface with the bond
upward trend in the data, observed within each block and coat, and in the case of EBPVD coatings, by cracking
from block to block across the plot, may be viewed as an within the oxide layer that builds up at the surface of the
indication of the breakdown in thermal protection bond coat. The failure modes differ because of
efficiency of the TBC system. As the coating ablates, or differences in the degree of strain tolerance of the two
spalls off the test pin, the internal heating rates increase, types of coatings.
as expected, because heat is transferred more quickly into
the pins. Thirdly, there appears to be a threshold number ACKNOWLEDGEMENTS
of applied thermal cycles beyond which the increase in This work was performed at IAR and IMI under projects
heating rate accelerates sharply. This threshold is of the involving collaboration with a number of companies
order of 500 cycles, and corresponds to the point beyond (Cametoid Advanced Technologies Inc., Pyrogenesis
which significant weight loss was recorded for this Inc., Orenda Aerospace Corporation, SURFTEC
coating, as shown in Figure 15. Thus, monitoring the Industrial Consortium), with financial support from the
internal temperatures of coated pins could be used as an Chief Research and Development of the Department of
alternative method for the quantification of damage National Defense. Special thanks are due to Mr. Ray
accumulation in TBC's. However, further work is needed Dainty of IAR, for his support in SEM analysis.
to develop this technique. The type K thermocouples that
15-10

REFERENCES 9. J. H. Wood and E.H. Goldman. "Protective

Coatings" in Superalloys-H, Eds. CT. Sims et al,
1. R. A. Miller, "Thermal Barrier Coatings for Aircraft Pub. John Wiley and Sons, 1987, pp. 359-384.
Engines", Proc. Thermal Barrier Coating Workshop,
NASA Lewis Research Center, Cleveland, Ohio, 10. W.J. Brindley and R.A. Miller, "Thermal Barrier
March 1995, NASA Conf. Pub. 3312, 1995, ppl7- Coatings Evaluation Needs", Pres. at Conf. on NDE
34. of Modern Ceramics, Columbus, OH, 9-12 July
1990, Am. Cer. Soc. & Am. Soc. forNDT, 1990,
2. P.E. Hodge, S. Stecura, MA. Gedwill, I. Zaplatynski pp70-76 (NASA Report, NASA-TM-103708, 1990).
and S.R. Levine, "Thermal Barrier Coatings", J.
Materials for Energy Systems, Vol. 1, 1980, pp. 47- U.M. Prystay, P. Gougeon, C. Moreau, J-P.
58. Immarigeon and V.R. Parameswaran,
"Microstructure and Burner Rig Evaluation of Calcia
3. H. Herman and C.C. Berndt, "Plasma Spray Silica Thermal Barrier Coatings", Proceedings ITSC
Processing of TBC's", Proc. Thermal Barrier 98, Nice, France, May 1998.
Coating Workshop, NASA Lewis Research Center,
Cleveland, Ohio, March 1995, NASA Conf. Pub. 12. P. Gougeon, M. Prystay, C. Moreau, J-P.
3312, 1995, pp. 127-134. Immarigeon and V.R. Parameswaran, "Influence of
Plasma Spraying Conditions on In-Flight
4. D.V. Rigney, R. Viguie and D.J. Wortman, "PVD Characteristics of Zirconia Hollow Sphere Particles
Thermal Barrier Coating Applications and Process and Properties of Resulting TBC's", Proceedings
Developments for Aircraft Engines", Proc. Thermal ITSC 98, Nice, France, May 1998.
Barrier Coating Workshop, NASA Lewis Research
Center, Cleveland, Ohio, March 1995, NASA Conf. 13. R.R. Dills and F.S. Follansbee, Corrosion, NACE,
Pub. 3312, 1995, pp. 135-149. 1977, 33(11), 385.

5. S. Durham, S. Manning Maier, D.K. Gupta and K.D. 14. A.K. Gupta, T. Terada, P.C. Patnaik and J-P.
Schefler, "Ceramic Thermal Barrier Coatings", in Immarigeon, "Evaluation of High Temperature
Advances in High Temperature Structural Materials Protective Coatings for Gas Turbine Engines Under
and Protective Coatings. Eds. A.K. Koul et al, Pub. Simulated Service Conditions", Proc. AGARD-
NRC, Canada, Ottawa, 1994, pp.226-236 NATO Conf. on High Temperature Surface
Interactions, Ottawa, May 1989, AGARD-CP-461,
6. Anonymous, "Vacuum plasma spray coating protects 1989, pp. 14-1/14-31.
turbine blades", a note in Advanced Materials and
Processes, March 1997, p. 12. 15. A.K. Gupta, J-P. Immarigeon and P. C. Patnaik, "A
Review of Factors Controlling the Gas Turbine Hot
7. F. Jamarani, M. Korotkin, R.V. Lang, M.F. Ouelette, Section Environment and their Influence on Hot Salt
K.L. Yan, R.W. Bertram and V.R. Parameswaran, Corrosion Test Methods", High Temp. Technology,
"Compositionally Graded TBC for High Vol. 7, No. 4, November 1989, pp. 173-186.
Temperature Aero Gas Turbine Components",
Surface and Coatings Technology, 54/55, 1992, pp. 16. S. Gendron, N.J. Marchand, C. Korn, J-P.
58-63. Immarigeon and J.J. Kacpryzynski, "An
Experimental Investigation of Convective Heat
8. M. Gell, "Application Opportunities for Transfer at the Leading Edge of a Turbine Airfoil",
Nanostructured Materials and Coatings", Materials Int. Gas Turbine and Aeroengine Congress and
Science and Engineering, Vol. A204, 1995, pp. 246- Exposition, Cologne, Germany, June 1992,
251. Publication ASME 92-GT-248, 1992
 16-1

Graded Design of EB-PVD Thermal Barrier Coating Systems

U. Schulz, T. Krell, U. Leushake, M. Peters

DLR - German Aerospace Research Establishment, Institute of Materials Research,

D-51147 Cologne, Germany.

Summary ority of EB-PVD TBCs is due to their columnar micro-

The most durable TBCs (Thermal Barrier Coatings) are structure. Their processing, microstructure, and lifetime
yttria partially stabilized zirconia coatings which are in service is often described in literature /l, 4, 5, 6, 7, 8/.
applied by EB-PVD (Electron Beam Physical Vapor But there is still a lack of detailed understanding of
Deposition) on metallic bond coats. The EB-PVD proc- failure mechanisms and damage accumulation on a
ess offers the advantage of a superior strain and ther- microscopic scale that can be combined with stress
moshock tolerant behavior of the coatings due to their modeling, thus enabling a phenomenological based safe
columnar microstructure. The interface between ceramic life prediction under the wide variety of application
and metal is the weakest link in this system. Spallation conditions.
occurs in that area which is mainly attributed to oxidation
of the bondcoat and to thermomechanical stresses of this State-of-the-art TBCs for turbine blades consist of an
two-layered coating. oxidation protective bond coat (BC) (typically alu-
The concept of graded materials is one approach to minides or MCrAlY, M=Ni, Co), and a thermally insu-
lower both stresses and also oxidation of the bondcoat. lating yttria partially stabilized zirconia (YPSZ) top
This concept is presented in detail with special emphasis layer. This multilayer system is not thermally stable
on chemically graded alumina-zirconia TBCs. Micro- under long time exposure at elevated temperatures. A
structures and phases of alumina, alumina-zirconia, and thermally grown oxide scale (TGO) is formed by oxi-
of graded coatings deposited by EB-PVD are evaluated. dation at the BC/zirconia interface. This scale consists
Mainly alumina undergoes phase transformation in the mainly of alumina formed by selective oxidation of
mixed and graded layers from the amorphous state to aluminum from the bond coat but it can vary from this
the desired oc-Al203. Finite element modeling was per- ideal constitution particularly in the early stages of
formed to study the influence of coating thickness and oxidation /9, 10/. The TGO and the interfaces to BC and
composition on heat flux and stress distribution. TBC are the crucial locations of the TBC system that
determine in most cases the lifetime. In addition, inter-
diffusion between the bond coat and substrate leads to
1. Introduction microstructural changes in the BC. Another phenomena
Thermal barrier coatings (TBCs) are increasingly used is the reduction of thermal shock resistance of the zir-
in gas turbines for aircraft engines and for power gen- conia layer due to sintering. Having these phenomena in
eration. They must withstand for a long time heavy mind, several potential areas for the application of the
thermal cyclic loads under oxidizing conditions. The Functionally Gradient Materials (FGM) concept for a
application range for TBCs in gas turbines includes TBC system can be suggested.
insulation of the combustion chamber and protection of
turbine blades and vanes. Major advantages and key
factors that rule the lifetime of TBC systems have been 2. Graded Thermal Barrier Coating Systems
described elsewhere /l, 21. Graded materials are characterized by a one or more
dimensional continuous or stepwise change of material
Development of gas turbines clearly aims for increased properties. The change is designed to achieve optimum
gas turbine inlet temperatures (TIT) passing well be- materials behavior for a given loading situation. FGMs
yond 1600°C. Future TBC applications as integral de- differ from bulk materials and composites by their
sign elements of highly loaded engine parts need a macro- and microscopic inhomogeneity. To lower re-
higher temperature capability of the TBC, a better sidual stresses, coatings can be optimized for the case of
bonding of the ceramic on the metal and also a lower fabrication and/or application conditions. Both cases
oxidation rate of the bond coat. Thermomechanical may demand different approaches. The design of graded
stresses caused by this high TIT can only be maintained materials is mainly optimized by numerical or analytical
by usage of uneconomically extensive cooling tech- calculations. By applying the FGM concept, changes in
niques or by the introduction of electron-beam physical the magnitude or location of the critical local stresses
vapor deposited TBCs /3/. The latest generation of can be achieved, although not all stress components are
TBCs offers significantly extended lifetimes compared to reduced in every case. To improve performance it is
plasma-sprayed TBCs, smooth aerodynamically attractive often only necessary to modify the stresses which cause
surfaces and only minor cooling hole closure. The superi- failure.

Paper presented at an AGARD SMP Meeting on "Thermal Barrier Coatings",

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
16-2

can not be maintained under thermocyclic loading due

For TBCs a continuously graded layer between bond to the CTE mismatch. The advantage of a continuously
coat and zirconia would lower the stresses at the inter- graded coating compared to a single alumina interlayer
face substantially. Those systems have been described is the avoidance of a sharp discontinuity in properties
very early for plasma spraying /l 1/ and for EB-PVD too and the associated stress concentrations. Starting with
/12/, but more recently investigations have indicated alumina on top of the bondcoat a continuous increase of
that they are limited in their potential application. the PYSZ content is envisaged to benefit from the FGM
Graded bondcoat-zirconia layers showed improved concept. In addition, by controlled oxide/TBC interface
performance under pure thermal cycling conditions but formation, better adhesion and reduced thermal stresses
they suffer from rapid oxidation of BC particles in the may be realized. Among the large variety of alumina
zirconia due to their high surface area under oxidizing polymorphs, a-AI203 is the most desirable phase be-
conditions. This results in a dramatic performance re- cause of its low oxygen diffusivity and high thermal
duction compared to multilayered TBCs /13, 14, 15, stability at the application temperature range. Therefore,
16/. A possible application field for this kind of graded one target is to get a-alumina. All ceramic layers within
metal-ceramic coating would be limited to non- the graded TBC are produced by PVD techniques, pref-
oxidizing conditions or short time exposure, e.g. reen- erentially by high rate EB-PVD.
try. But for turbine engines these graded coatings are Another possibility to apply the FGM concept to TBCs
not useful. would be a graded layer between metallic bondcoat and
an alumina interlayer. This approach would harden the
In 1995 a joint research program on functionally graded bond coat outer region and therefore lower the creep
materials, funded by the Deutsche Forschungsgemein- rate of MCrAlY. Furthermore, improved adhesion of
schaft (DFG), was started in Germany. Part of this pro- the alumina is expected but some processing problems
gram is the investigation of graded TBCs. It is intended will occur too.
to combine the thermal insulation potential of zirconia
with the low oxygen diffusivity of alumina or similar Fig. 1 shows the state-of-the art two-layer coating com-
materials. Some results in literature indicate that the pared to the new proposed graded ceramic design. First
introduction of an alumina interlayer between bondcoat results of the work carried out on this graded TBC sys-
and ceramic offers a potential solution to some major tem have been published previously /20, 21, 22/.
problems of TBCs by incorporating a layer with re-
duced oxygen diffusivity, thereby reducing the TGO 2.1 Processing of graded TBCs
growth rate and prolonging cyclic lifetime 111, 18, 19/. In order to combine the advantages of columnar and
These alumina layers has been fabricated by CVD and graded structures several issues must be incorporated.
sputtering techniques. Furthermore, adhesion of EB- The vaporization during EB-PVD depends on the vapor
PVD PYSZ TBCs is known to be favored on a-alumina pressure of each compound which makes it difficult to
compared to adhesion on MCrAlY /10/. simultaneously vaporize materials with large differences
in vapor pressure (e.g. Zr02 and Ce02). Using a single
The large mismatch in coefficient of thermal expansion source coater only certain chemical gradients can be
(CTE) between discrete layers within a multicomponent produced, although some types of gradients have been
TBC system results in residual stresses that can cause successfully produced, e.g. density graded PYSZ and
coating failure. Obviously, fully dense alumina layers chemically graded bondcoats /23, 24/. Another attempt
is the use of a special mixture of aluminum, alumina,
and zirconia as a starting composition in form of
two-layer graded pressed tablets on top of YPSZ ingots to get a graded
structure /25/. Controlled chemically graded coatings
can be fabricated by vaporizing from multiple sources
utilizing one or more electron guns.

To develop the graded TBC system described above

several problems had to be solved in advance:
- evaporation material must be in a condition that allows
continuous evaporation
- influence of evaporation conditions on microstructure
spall ation of single alumina layers must be evaluated
- dual source evaporation of alumina and zirconia to get
mixtures of both species must be established.

Finally, grading from alumina to YPSZ was carried out.

Fig. 1: Comparison of conventional and graded de-
sign of TBCs
 16-3

2.2 Experimental procedure 3.2. Alumina coatings

In the present study, a single gun dual source jumping Phase content and morphology of alumina coatings,
electron beam coater was used (for details see 151). It produced by PVD methods, were investigated as a
consists of chambers for loading, preheating and depo- function of process parameters. Chamber pressure dur-
sition and facilities for specimens rotation and manipu- ing alumina evaporation was varied by an optional
lation. Ingots of 50 mm and 63 mm diameter were fed additional oxygen gas flow and by adjustment of the
into the crucibles from bottom during evaporation in pumping rate in order to achieve in some experiments
order to ensure continuous deposition. By adjusting reactive conditions.
electron beam parameters different deposition rates can In the case of lower chamber pressures, extremely high
be achieved for each source via independent adjustment ceramic deposition rates up to 60 um/min occurred. In
of melting pool temperatures. Alumina films were de- Fig. 2 an example is given for a coating deposited under
posited by reactive magnetron sputtering and also by such conditions. It can be concluded that higher deposi-
reactive EB-PVD. tion rates and the associated lower chamber pressures
X-ray diffraction (XRD) was carried out with Cu Ka result in relatively dense, less columnar microstructures,
radiation. If not noticed explicit otherwise patterns are probably as a result of changed adatom mobility /22/.
given corrected for background and Ka 2. Concentration
was measured quantitatively by energy dispersive X-ray On the other hand, a 5 to 8 urn thin alumina layer de-
spectroscopy on polished cross sections. posited by EB-PVD under reactive conditions exhibits a
columnar microstructure. After annealing a tendency to
sintering was observed /20/.
3, Results and discussion Thin alumina layers produced by reactive magnetron
3J. Alumina ingots sputtering showed a rough surface which reproduces the
Semi-industrial or industrial size coaters use a continu- surface of the underlying metal as it is characteristic for
ous feeding system for ceramic rod source materials PVD. In Fig. 3 an example is given where the bond coat
(ingots). They must exhibit a high thermal shock resis- morphology is still visible because peening was omit-
tance to withstand the high thermal gradients associated ted.
with EB-PVD processing. Therefore, ingots generally
have a relative density of about 60%. The current proc- Oiwiousiy, the morphology of alumina films produced
essing method for YPSZ ingots involves powder and by PV 2 strongly depends on deposition parameters and
material specific sintering and calcination and cannot for sputtering on the substrate surface too. Deposition
easily be transferred to other powders or materials. An rate, chamber pressure and atmosphere, film thickness,
alternative method was developed that involves the use and substrate temperature are the most important pa-
of bimodal powder mixtures and a sinterless powder rameters that rule the morphology of alumina.
metallurgical processing /21, 22, 26/. This processing
route can be applied to a large variety of different pow- In the temperature range of deposition alumina under-
ders and materials for ceramic rod fabrication, achiev- goes phase transformations, so that not only microstruc-
ing suitable vaporization behavior and high thermal ture of the coatings but also phase formation depends on
shock resistance at low cost. the substrate temperature. It was intended to get the
most stable alpha phase of alumina.
Alumina deposition was carried out in the temperature
range between 750 and 1150°C. In Fig. 4 the influence

1 mm 10 um

Fig. 2: Cross section of a thick alumina coating Fig. 3: Surface of sputtered alumina layer on top
obtained by EB-PVD at low pressure ofNiCoCrAlYbondcoat
16-4

the annealing temperature, coatings deposited at lower

temperature show faster phase transformation.

3.3. Mixed alumina-zirconia layers

Due to the continuous change of composition and prop-
erties within a graded alumina - PYSZ coating a direct
analyses of phases is difficult. Therefore, mixtures of
alumina and PYSZ were investigated. They represent
individual layers of the graded ceramic.

Co-evaporation of alumina and PYSZ was carried out

using the jumping-beam technique. First, vaporization
25 40 55 70 85 parameters such as scanning pattern, e-beam focus,
Diffraction Angle 2® dwell time and vapor cloud geometry were determined
for each material separately. The several mixtures of
Fig. 4: XRD of EB-PVD alumina films for various alumina/PYSZ with various amounts of the oxides were
deposition temperatures (not corrected) produced by dual source evaporation.

of deposition temperature on phase formation of alu- In Fig. 5 chemical composition is plotted as a function
mina is shown. of sample position above the ingots. Experiments
clearly show that dual source evaporation of two ceram-
XRD peaks were broadened for alumina films deposited ics with widely differing vapor pressures and melting
at low temperatures. This can be attributed to the pres- points is possible using the jumping-beam technique. A
ence of areas within the coating that were not fully mixing of both ceramic vapor clouds takes place and
crystallized. Coatings deposited at the highest tempera- both species condense on the substrates thereby forming
ture showed the presence of only the stable oc-Al203 mixed layers. While the portion of A1203 decreases more
phase. Additionally it was found that chamber pressure rapidly with decreasing distance to the ZrOz-ingot, the
and deposition rate had only little influence on phase portion of Zr02 rises slower and shows a flatter gradi-
formation of alumina which is contrary to results found ent. The zirconia content is predominant in all coatings.
for the morphology. The reason for that is the difference of vapor pressures
Phase transformation was also studied after annealing in and evaporation behavior between Zr02 and A1203 that
the temperature range between 1000°C and 1200CC for leads to a different geometry of the vapor clouds. Fur-
1 to 9 hrs. Annealing at 1000°C had little effect on any thermore it can be assumed that the re-evaporation rate
of the coatings; only a sharpening of diffraction peaks of alumina is higher than that of zirconia because of its
was observed. For all coatings, transformation to ot- lower melting and boiling point.
A1203 started upon annealing at 1100°C. Short term (1
h) annealing at 1100CC resulted in partial transforma-
tion to a-Al203. The higher the deposition temperature
was, the lower was the tendency for phase transforma-
tion. This can be explained by the fact that the driving
force for transformation (i.e. difference between coating
and annealing temperature) is lower. After annealing at
1200°C for 9hrs all coatings consisted mainly of oc-
A1203 with only small amounts of other phases /21, 22/.

Alumina layers deposited by reactive magnetron sput-

tering were amorphous but they showed a rapid phase
-100 -50 0
transformation behavior. Observations are in good position of sanies/mm
agreement with literature data on alumina layers ob-
tained by sputtering 121, 28/ and by reactive aluminium Fig. 5: Composition of mixed alumina-PYSZ layers
evaporation /29/. as a function of position in the chamber

As a consequence, alpha alumina coatings can be

achieved either by in-situ deposition at elevated tem- The morphology of these coatings is exemplified in Fig.
peratures or by post coating annealing. Since the phase 6. All coatings have a columnar microstructure, but
transformation driving force is proportional to the tem- contrary to pure PYSZ TBCs the alumina-zirconia
perature difference between the coating temperature and coatings show a much more closely packed structure.
 16-5

Results of XRD are shown in Fig. 7.

Phase analyses only reveals strongly broadened PYSZ-
peaks which are getting sharper with increasing content
of zirconia. No alumina peaks were found in the as
coated condition. This suggests that alumina only exists
as an amorphous phase, but it is not clear where this
phase is located beside the zirconia crystals. Additional
investigations by transmission electron microscopy are
under way in order to clarify this question.

Because of the broadened peaks zirconia must be very

fine grained which is contrary to pure PYSZ TBCs
where each column is nearly single crystalline. A first
calculation of the grain size based on zirconia peak half-
width gave evidence for 5-15 urn grains, depending on
the alumina content. Thermal conductivity and other
properties of this fine-grained two-phase structure may
significantly differ from values that were calculated
using simple rules of mixture. Since lattice parameters
of PYSZ are only slightly changed by the addition of
alumina one can conclude that only small amounts of
alumina were incorporated into the tetragonal lattice of
the t' phase of PYSZ.

3.4. Graded alumina-PYSZ thermal barrier coatings

By continuously changing the evaporation rates of
alumina and zirconia a change in composition over the
Fig. 6: SEM of mixed alumina-PYSZ layers, frac- coating thickness was achieved. Due to the different
tograph (above) and surface (below) deposition conditions which are necessary to maintain
suitable evaporation of alumina, zirconia and various
Moreover, surfaces of the columns are no flat faces with mixtures thereof, an increase of the substrate tempera-
characteristic shapes as it was found for yttria and ceria ture occurs with increasing zirconia content in the vapor
stabilized zirconias /30/, but they are round shaped with cloud.
cauliflower type appearance. Furthermore they posses Initially a pure and thin alumina layer was put on top of
less opened inter-columnar porosity. Even small a NiCoCrAlY bondcoat, followed by compositional
amounts of alumina can cause these microstructural graded mixed alumina - PYSZ layers. At the top side
changes. Since microstructure was nearly the same for pure PYSZ was deposited. A cross-section of such a
all mixtures, one can conclude that not the amount of graded coating is shown in Fig. 8. The backscattered
alumina but the presence itself is sufficient for the electron imaging (right part of picture) indicates a steep
change in morphology. gradient in chemical composition between alumina and
zirconia. In contrast, change of microstructure from
bottom (alumina) to top (zirconia) appears not continu-
ous.

Zirconia

A120, wt%
39
24
%uttr "Wir^-r Wirtr *HinrT"r KM c
H 1 1 1 1 1 1 1
253035 40 455055606570 75 80859095 100
2ttetain°

Fig. 7: XRD of EB-PVD mixed coatings from alu-

mina and PYSZ Fig. 8: Cross section of a graded alumina-PYSZ TBC
16-6

beam deflection system with capability of beam scan-

ning frequencies up to 1 kHz. It was found that adjust-
ing the beam pattern and beam focus to the different
ceramics was the key point for a reproducible evapora-
tion behavior and also to omit spitting mainly from the
alumina ingot. This task is complicated because zirconia
needs approximately two to three times more energy for
evaporation but beam power cannot be changed quickly
during the jumping procedure.

To illustrate the problems of microstructure investiga-

tion of graded layers XRD patterns of a graded, a stan-
dard PYSZ TBC, and a powdered PYSZ TBC are com-
pared in Fig. 10.
The absence of any alumina peaks in the graded layer
must be noticed which is similar to observations on
mixtures. The thin PYSZ layer on top of the graded
zone additionally complicates the investigation of the
FGM because of the low x-ray penetration depth in
zirconia. The highly textured growth of both EB-PVD
coatings can be seen but some differences in the pre-
ferred orientation can be already determined from sim-
ple diffraction patterns. While PYSZ TBCs grow pri-
marily with {100} planes parallel to the surface /5, 31/
the graded layers posses mainly strong {220} and {113}
peaks. Taking into account the results of mixed layers
(Fig. 7) it is obvious that alumina additions to PYSZ
Fig. 9: Graded alumina-PYSZ coating, composition
(above) and fractograph (below) lower the tendency to preferred growth of some crystal-
lographic orientations. The fine grain size within graded
layers instead of single crystalline columns in case of
pure PYSZ may further contribute to the observed tex-
Fig. 9 evidences a continuous chemical gradient. For ture differences.
better visibility the graded layer is shown as a fractured
cross section. In a final step the fully ceramic graded thermal barrier
Although there is still a step in the gray-scale of the coating system was produced. One important aspect to
SEM picture the concentration profile shows only a consider is the change in overall thermal conductivity of
slightly steeper change of concentration in this area. the graded layer due to the higher thermal conductivity
of alumina compared to zirconia. This also depends on
Experiments have revealed that two-source evaporation the compositional gradient itself, and it has to be com-
can produce a distinct graded layer although only one pensated by an additional zirconia layer. Numerical and
electron-beam gun has been used. Application of analytical computations can help to calculate the appro-
jumping beam technology on ceramics requires a fast priate thermal fluxes and thicknesses of the individual
layers. This topic will be discussed in chapter 4.
A standard PYSZ-TBC of approximately 200 urn thick-
ness was applied on top of the graded layer in order to
maintain the desired thermal conductivity of the whole
system. The TBC is shown in Fig. 11.

The zirconia layer on top exhibits the desirable colum-

nar microstructure typical for EB-PVD coatings. The
graded region, however, is characterized by a coarser
microstructure within the columns and a smaller column
A YPSZpoMlered diameter. Here, spacings between columns seem to be
smaller and the FGM looks more compact compared to
the opened columnar PYSZ region. The temperature
increase during evaporation mentioned above may have
further supported occurrence of this morphology. Col-
umns go through from the graded zone to the PYSZ top
Fig. 10: XRD of different EB-PVD TBCs layer which is necessary for a good bonding. Micro-
 16-7

structural features are in good agreement with observa- calculation of the stress state effects like creep, plastic
tions on mixed layers described in chapter 3.3. deformation, oxidation, and sintering have to be taken
into account. It must contain also stresses after fabrica-
tion, in the transient stages, in the temperature field
during service and after cooling down to room tempera-
ture. All these aspects described above have to taken
into account in order to draw the right conclusions from
PYSZ TBC calculations.
In the following one special aspects of the design of
graded TBCs will be explained in detail.

The designing process is based on a parametric varia-

tion of a chosen grading function, which describes the
change in chemical composition along the coating
thickness. Widely used is a simple power law function:

c(x) = = x" (1)

h
where c is the concentration, d the distance to the inter-
face or surface, h the coating thickness, x the normal-
ized height and p the profile parameter. By variation of
the profile parameter p different chemical profiles are
defined and according to the local composition and
microstructure the local material properties can be set.
A widely used approach is that the grading function also
directly describes the variation of material properties.
Variation of the profile parameter p leads to different
stress situations and the minimization of the critical
stress component responsible for failure can be defined
as an optimum. Variation of the grading function also
effects the FGMs overall thermal conductivity and there-
fore the thermal field. In case of a constant coating
Fig. 11: Morphology of a ceramic graded EB-PVD thickness and of boundary conditions that are given by
TBC system fixed surface temperatures the heat flux depends on the
profile parameter. If the boundary conditions are set
different the maximum temperature of the metal may
4. Finite Element Modeling for graded TBCs exceed a certain value. These variations complicate the
Finite Element Method (FEM) and analytical calcula- comparison of calculated stress fields and the design
process has to be extended to take this effect into ac-
tions provide a useful tool for an understanding of count. The variation of coating thickness itself and the
complex thermal and stress fields. The reliability of the corrected thermal boundary conditions affect the stress
calculations depends strongly on the used material distribution in the graded system.
properties and boundary conditions. Assumptions like
pure linear elastic behavior or certain boundary condi- Thermal conductivity of alumina is substantially higher
tions may have a substantial effect on the results. In than that of zirconia and therefore the thickness of the
most cases the microstructure is modeled on a macro- graded coating has to be increased. The increase can be
scopic scale, since the real columnar microstructure of calculated by defining an effective thermal resistivity R:
EB-PVD or the splashy one of plasma sprayed coatings
is to complex for a microscopic model. Calculations are 1 (2)
based on knowledge of material properties. For bulk
«•-j X{x)-dx
materials, most properties are known or can be meas-
ured. For coatings, however, data availability is limited. with x the normalized height, A,(x) the local thermal
This may cause some uncertainties about the accuracy conductivity. In combination with different rules of
of stress calculations and finite element modeling. The mixtures for the thermal conductivity of composites and
properties of composites are often estimated by differ- the grading function (1) the effective thermal resistivity
ent rule of mixtures according to their microstructure. R can be calculated either analytically or numerically.
The designing process of FGMs should include the
calculation of thermal fields and their stresses and also Fig. 12 shows the effect of the profile parameter p on the
residual stresses during fabrication. For a complete effective thermal resistivity. These curves can be used to
estimate coating thicknesses of same heat flux. To meet
16-8

R(ZJQZ)
deviations because of small temperature gradients.
Compared to a non-graded TBC of 200 urn Zr02 (curve
a) the heat flux of a linear graded coating with the same
thickness (curve b) is more than twice as high. In this
case the graded layer shows higher interfacial tensile
stress, which should decrease performance. Based on
(1) and (2) the overall thickness of the graded layer was
increased to same heat flux conditions (curve d), result-
ing in the lowest compressive interface stresses in the
ceramic but in a dramatic higher thickness too. Alterna-
tively, a 150 um thick zirconia layer was applied on top
of a 200 urn thick graded layer (curve c), resulting in
0.1 0.5 1 5 10 50 100 the same heat flux as the conventional system. Stresses
Profile Parameter p in the bondcoat are much lower than for all other sys-
tems, but interfacial stresses are higher compared to a
Fig. 12: Effective thermal resistivity for different
and d. From these examples one can see that the grading
rules of mixtures and profile parameters
profile significantly influences the stress situation. For
optimization, knowledge of critical stresses that are
this requirement, thickness has to be increased by a
factor of 4.6 in case of a linear graded layer compared to responsible for failure is essential. In case of coating
pure PYSZ. Increase of coating thickness is limited to a thickness of same heat flux the FGM concept offers a
certain level for aero engines. Therefore application of potential for improvement of coating performance.
an additional zirconia layer on top of the graded layer is
preferred to get same heat fluxes.
5. Conclusions
For a loading situation similar to a turbine blade stresses The potential of chemically graded thermal barrier
of different coating systems have been calculated (Fig. coatings to improve lifetime and reliability of turbine
13). In a biaxial stress state the radial stresses of an blades was evaluated. It has been demonstrated, that
infinite plate have been calculated assuming linear elas- graded ceramic coatings can be produced by dual
tic behavior /32. The model involves a 1mm thick su- source EB-PVD and jumping beam technology. Various
peralloy substrate that possesses a NiCoCrAlY bondcoat mixtures of alumina-PYSZ as well as pure alumina and
of 150 urn and the TBC on top. All material properties graded coatings have been produced and analyzed.
were assumed to follow the linear rule of mixture and to 1. Morphology and phases of PVD alumina films de-
be not temperature dependend which causes only little pend on deposition parameters, mainly on deposition
rate, chamber pressure and atmosphere, film thickness,
and substrate temperature. Alpha alumina can be ob-
Bond TBC tained either by high substrate temperatures or by post
300 -Coat .-, Tsurface =1200PC
coat annealing.
\ Tbutton =90(rC
2. Morphology of alumina-zirconia mixtures is more
- 'X \ compact than that of pure YPSZ TBC but it depends to
\ \
a less extend on the amount of alumina. Zirconia is very
u 100 - L \\b
fine grained in an amorphous alumina matrix.
\ " x
\ 3. By defining an effective thermal resistivity coating
£ • \ \
\\
\
\
thickness can be adapted to constant heat flux conditions
.2 400 \ \ x With a graded transition zone and with a modified coa-
->d
-oet ting thickness a variation of the stress field can be
**■----*:
\ \c achieved.
P4 -200 a ^
I.I.I

0 0L2 0.4 0.6 a8 1 1.2 Optimization of coating conditions and the design of the
Distance to Substrate [mm] graded zone with respect to chemical composition pro-
file, phases, microstructure, morphology, and optimized
curve Zr02 FGM heat flux lifetime remains to be carried out in future research.
um urn MW/m2
a 200 0 2
b 0 200 4.2 ACKNOWLEDGEMENTS
c 150 200 2 The authors gratefully acknowledge technical support
d 0 920 2 by J. Brien, C, Kröder, H. Mangers and H. Schurmann..
The work was partially funded by the Deutsche For-
Fig. 13: Effect of corrected coating thickness on schungsgemeinschaft.
stress distribution
 16-9

temperature aero gas turbine components. Surface and

References Coatings Technology 54/55(1992) 58-63

1 Kaysser et al. AGARD this issue 13 Ulion, N.E.; Ruckle, D.L.: Columnar grain ceramic
thermal barrier coatings on polished substrates. European
2 M. Peters, K. Fritscher, G. Staniek, WA. Kaysser, U. Patent EP0044329B1
Schulz: Design and properties of thermal barrier coat-
ings for advanced turbine engines. Materialwissenschaft 14 Alaya, M.; Grathwohl, G.; Musil, J.: A comparison of
und Werkstofftechnik 28(1997) 357-362 thermal cycling behaviour of graded and duplex Zr02-
thermal barrier coatings. Proc. FGM'94, eds. B.Ilschner et
3 J.T. DeMasi-Marcin and D.K. Gupta, "Protective al. Lausanne (1994) 405-411
Coatings in the Gas Turbine Engine", Surface and
Coatings Technology 68/69 (1994) 1 -9 15 Lee, W.Y.; Stinton, D.P.; Berndt, C.C.; Erdogan, F.;
Lee, Y-D.; Mutasim, Z.: Concept of functionally graded
4 S.M. Meier, D.M. Nissley, K.D. Sheffler, T.A. Cruse: materials for advanced thermal barrier coating applica-
"Thermal barrier coating life prediction model develop- tion. J. Am. Ceram. Soc. 97 (1996) 12, 3003-3012
ment", ASME J. Eng. Gas Turbine Power 114(1992) 258-
263 16 Vincenzini, P.: Zirconia thermal barrier coatings for
engine applications. Industrial Ceramics 10(1990)3,113-
5 U. Schulz, K. Fritscher, C. Leyens, M. Peters, WA. 126
Kaysser: Thermocyclic Behavior of Differently Stabi-
lized and Structured EB-PVD Thermal Barrier Coat- 17 Schmitt-Thomas, Kh.G; Dietl, U.: Surface and coat-
ings. Material Wissenschaft und Werkstofftechnik ings technology, 68/69 (1994) 113-115
28(1997) 370-376
18 Sun, J.H., Chang, E.; Wu, B.C.; Tsai, C.H.: The prop-
6 Schulz, U.; Fritscher, K.; Rätzer-Scheibe, H.-J.; Peters, erties and performance of (Zr02-8wt%Y203)/(chemically
M.; Kaysser, WA.: Thermocyclic Behaviour of Micro- vapour-deposited Al203)/(Ni-22wt%Cr-10wt% Al-
structurally Modified EB-PVD Thermal Barrier Coatings. lwt%Y) thermal barrier coatings. Surface and Coatings
4th International Symposium on High Temperature Corro- Technology 58(1993) 93-99
sion and Protection of Materials. Les Embiez 20-24.5.1996
19 Strangman, T.E; Solfest, P.A.: Ceramic thermal bar-
7 Sohn, Y.H.; Biederman, R.R.; Sisson Jr, R.D.: Micro- rier coating with alumina interlay er. US Patent 5 015
structural development in physical vapour-deposited 502(14.5.1991)
partially stabilized zirconia thermal barrier coatings. Thin
Solid Films 250(1994)1-7 20 T. Krell, U. Schulz, M. Peters, WA. Kaysser: Influ-
ence of various process parameters on morphology and
8 Maricocchi, A.; Bartz, A.; Wortman, DJ.: PVD TBC phase content of EB-PVD thermal barrier coatings. Poc.
experience on GE Aircraft engines. Proc. TBC workshop EUROMAT 97, eds. L.A.J.L. Sarton and H.B. Zeedijk,
Cleveland 1995, ed. W.J. Brindley, NASA Conf. Publi- Netherlands Soc. for Materials Science, Vol. 3(1997)
cation 3312 (1995) 79-89 3/29-3/32

9 Schmücker, M.; Fritscher, K; Schulz, U.: Haftmecha- 21 U. Leushake, U. Schulz , T. Krell, M. Peters, WA.
nismen in ausgewählten EB-PVD Wärmedämmschicht- Kaysser: A1203 - Zr02 Graded Thermal Barrier Coatings
systemen. Fortschrittsberichte der DKG, Werkstoffe, by EB-PVD - Concept, Microstructure and Phase Sta-
Verfahren, Anwendung 10(1995)4, 379-384 bility. Proc. 4th Int. Symp. on FGM'96.

10 Fritscher, K; Schmücker, M.; Leyens, C; Schulz, U: 22 U. Leushake, T. Krell, U. Schulz, M. Peters, WA.
TEM Investigation on the Adhesion of EB-PVD Thermal Kaysser: Microstructure and Phase Stability of EB-PVD
Barrier Coatings. 4th International Symposium on High Alumina and Alumina/Zirconia for Thermal Barrier
Temperature Corrosion and Protection of Materials. Les Coating Application. ICMCTF '97, Surface and Coat-
Embiez 20-24.5.1996 ings Technology (in press)

11 Goward, G.W.; Grey, DA.; Krutenat, R.C.: US Pat- 23 Fritscher, K; Schulz, U: Burner-rig performance of
ent No. 4 248 940 density-graded EB-PVD processed thermal barrier coat-
ings, in "Ceramic Coatings" ed. K.Kokini, New York,
12 Jamarani, F.; Korotkin, M.; Lang, R.V.; Ouelette, ASME MD-Vol.44 (1993) 1-8.
M.F.; Yan, K.L.; Bertram, R.W.; Parameswaram, V.R.:
Compositionally graded thermal barrier coatings for high
16-10

24 Schulz, U.; Fritscher, K.; Peters, M.; Kaysser, W.A.:

Processing and behavior of chemically graded EB-PVD
MCrAlY bond coats. Proc. FGM 94, eds. B. Ilschner, N.
Cherradi, Presses polytechniques et universitaires ro-
mandes (1995), Lausanne, 441-446

25 Movchan, B.A.: EB-PVD technology in the gas tur-

bine industry: present and future. JOM 11(1996) 40-45

26 U. Leushake, W. Luxem, C.-J. Kröder, W.-D. Zim-

mermann: German patent application #A19623587.1,
European patent application No. 97108972.7-2111

27 J. A. Thornton, J. Chin: Structure and Heat Treat-

ment Characteristics of Sputter-Deposited Alumina,
Ceramic Bulletin, 56 5 (1977) 504-508

28 R. F. Bunshah, R. J. Schramm: Alumina Deposition

by Activated Reactive Evaporation, Thin Solid Films,
40(1977)211-216

29 A. L. Dragoo, J. J. Diamond: Transitions in Vapor-

Deposited Alumina from 300-1200°C, Journal of The
American Ceramic Society, 50, 11 (1967) 568-574

30 Schulz, U.; Fritscher, K.; Peters, M.: EB-PVD Y203

and CeO/Y203 Stabilized Zirconia Thermal Barrier Coat-
ings - Crystal Habit and Phase Composition. Surface and
Coatings Technology 82 (1996) 259-269

31 Schulz, U.; Oettel, H.; Bunk, W.: Texture of EB-PVD

thermal barrier coatings under variable deposition condi-
tions. Zeitschrift für Metallkunde 87(1996)6,488-492

32 M. Finot, S. Suresh, Multitherm, MIT

 17-1

BOND COAT CONSIDERATIONS FOR THERMAL BARRIER COATINGS

A.M. Freborg1, B. L. Ferguson1,

W.J. Brindley2 and G. J. Petrus'
■Deformation Control Technology, Inc.
Cleveland, Ohio USA
2
NASA Lewis Research Center
MS 24-1, 21000 Brookpark Road
Cleveland, Ohio 44135, USA

Abstract isolating variables in a real material system so that

unambiguous results can be obtained. An alternative is
Bond coat oxidation has been clearly established as a to perform modeling studies to determine trends in the
factor contributing to the failure of a TBC. However, effects of various factors as well as determine the
the mechanism by which oxidation effects TBC failure mechanisms that precipitate failure. There have been
has not been determined. Similarly, other bond coat only a few models that have focused in this important
properties have been cited as influencing TBC residual area for TBCs. These models have examined the TBC
stresses, but neither the magnitude of the effects nor the residual stresses resulting from single cooling cycles
exact mechanisms by which these properties act, and from a stress free state [6, 7] and residual stresses
interact, have been elucidated. A finite element model during thermal cycling [8]. but have all been
was developed to evaluate residual stresses induced by incomplete in that some known factors were not
thermal cycling of a typical plasma sprayed TBC included. The purpose of the current work was to
system. It was found that bond coat properties of develop a more inclusive finite element model to
oxidation, creep, coefficient of thermal expansion characterize the effects of TBC system properties, and
(CTE) and roughness, as well as top coat creep, all particularly bond coat properties, on the residual
strongly influence ceramic layer stresses. Roughness stresses in TBCs. This was accomplished by adding
and CTE were found to have the strong effects in a cyclic oxidation and multiple thermal cycle capability
parametric study, acting primarily as stress multipliers to the previous Petrus model [8], which had already
for other active mechanisms. A detailed examination of incorporated the effects of creep, CTE and roughness
the time dependent properties of creep and oxidation on TBC stresses during a single thermal cycle. As will
indicates that these factors account for evolution of the be shown, the results of this inclusive model indicate
residual stresses with thermal cycling. It is the that creep processes dominate the generation of
combination of these effects over the lifetime of the ceramic layer stress early in the life of a TBC,
coating that drives the generation of stresses required to oxidation dominates stress generation later in life, but
initiate and propagate delamination cracks. that there is interaction between all the effects
throughout coating life that cannot be ignored.

Introduction 2. Methodology

It is universally acknowledged that bond coats for This study used NIKE 2-D stress models [9] to describe
plasma sprayed TBCs have a strong effect on the a burner rig test of a 2.54 cm (1.0 in.) diameter solid
thermal fatigue life of the ceramic layer, but the Waspaloy rod coated with 0.013 cm (0.005 in.) of
specific mechanisms by which the bond coat influences NiCrAlY bond coat, which was in turn coated with a
TBC life have not been established [1, 2, 3, 4]. As an "top coat" of 0.025 cm (0.010 in.) Zr02-8wt.% Y203
example, bond coat oxidation has been clearly linked to (Figure 1(a)). It was assumed that the coating was stress
spallation of the ceramic top coat but the mechanism by free after deposition and cooling to room temperature a
which oxidation causes failure of the top coat has not reasonable assumption in view of the low in-plane
been clearly identified [1, 4, 5]. Similarly, bond coat compressive stresses (approximately 3 MPa) measured
coefficient of thermal expansion (CTE), roughness and on TBCs fabricated with appropriate temperature
creep properties have been proposed to influence TBC control [10]. The model geometry was a thin
failure, but the magnitude of the effects and the axisymmetric slice of the substrate coating system, as
physical mechanisms by which these effects act, and was used previously (Figure 1(a)) [8], with a condition
interact, have not been established [1, 2, 6]. of generalized plane strain. Unlike the Petrus model
[8], the current model used refined elements for the
Part of the difficulty in determining which mechanisms bond coat and top coat at the interface region to allow
and interactions are involved is due to the difficulty in for growth and analysis of an oxide layer. The

Paper presented at an AGARD SMP Meeting on "Thermal Barrier Coatings"

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
17-2

geometry of the bond coat/ceramic interface was no oxidation and a high equal to the oxidation rate
modeled as a sine wave to simulate circumferential found empirically for the alumina former NiCoCrAlY
surface roughness conditions typical for a plasma [11]. The oxide scale was assumed to be of uniform
sprayed TBC (Figure 1(b).) [6, 7, 8].
STEADY STATE
HEATING HOLD COOLING
BOND 1200
INTERFACE

900
m
DIMENSIONS E

SUBSTRATE RADIUS: 12.7 cm (0.5 IN.) 600

K
AVERAGE BOND COAT THICKNESS: 0.130 mm (0.0051 in.) HI
0.
AVERAGE CERAMIC THICKNESS: 0.250 mm (0.0098 in.) 2
300
(a)
1 1 1 1 1 1 1 I I i—-i 1
CERAMIC
0 100 200 300 400 500 600
TIME (s)
Figure 2. Measured burner rig test thermal profile used
in the FEM model.

thickness across the entire interface and was further

assumed to be composed entirely of alumina. The
properties used for the CTE and modulus of the scale
were those for a sintered 95% dense alumina [13]. The
scale was assumed to behave elastically under all
conditions.
10|xm The procedure for incorporating a growing oxide on the
bond coat used the element "birth" and "death"
capabilities of NIKE. Bond coat oxidation was thus
(b) modeled through the replacement of bond coat
elements with the same size of oxide elements at set
Figure 1. Finite element mesh geometry for TBC
time intervals. The first oxide birth was in the elements
oxidation model showing (a) overall geometry
located at the ceramic metal interface. Subsequent
of axisymmetric slice and (b) detail view of
growth occurred through oxide elements birthed at the
the interface region.
scale/ metal interface at the start of the high
temperature hold portion of the cycle. The oxide
The thermal cycle profile for the model was based on
elements were allowed to "grow" during steady state to
the actual burner rig profile (Figure 2) [8]. The model
the full volume dictated by the growth rate equation by
included the temperature gradients that naturally occur
assigning an artificially high CTE to the oxide and
during heating and cooling. It should be noted that the
imposing a small temperature increase during steady
temperature gradient was approximately zero at the
state, similar to the method used by Chang et al. [6]. In
hold temperature since the modeled specimens were not
this way the stress due to oxide growth, relaxation of
internally cooled.
growth stresses and thermal cycling was incorporated in
the model. Since the oxide elements were birthed at the
Creep behavior was defined using a temperature
oxide/bond coat interface, the model simulates the
dependent power-law creep model, with creep behavior
inward growth of an oxide scale.
as defined in the literature [2, 11, 12]. Further
properties information required for the model included
Five factors were intentionally varied, in addition to the
the temperature dependent properties of bulk modulus,
main factor of oxidation. The five factors were bond
shear modulus and coefficient of thermal expansion,
coat roughness, bond coat CTE, bond coat creep, top
and were derived from the literature [2, 11, 12, 13].
coat creep and the number of thermal cycles.
Information on stress as a function of location is also a
The two levels for oxidation were defined as a low of
part of the output of the model. After a few brief
 17-3

comments on the effects of roughness and CTE, the The effects of roughness and bond coat CTE will first
remainder of the paper will concentrate on the effects be discussed to demonstrate general effects that are
of oxidation and creep of the constituents. important to the detailed analysis of the effects of
oxidation and creep that are presented later.
The output of the model included stress as a function of
time, location and direction. It was found that the radial Roughness
stresses in the ceramic (perpendicular to the interface)
provided an adequate indication of the stresses most Interfacial roughness has been the subject of study for
likely to promote delamination cracking. Therefore, TBCs [6, 8] and is a well known consideration for
radial stresses will be presented as a function of time stresses in thin oxide films [14]. The case that is usually
and location to depict the active stress generating treated is for cooling of a two layer system that is
mechanisms in a TBC system and to try to define the assumed stress free at some high temperature, generally
likely failure scenario. It was also found (see Figures 8 the maximum temperature of a thermal cycle. Petrus et
and 9, discussed later) that the highest tensile radial al. [8] started with the assumption of a stress free
stresses were located at the bond coat peak, off-peak condition at room temperature and then modeled one
and valley regions. Therefore, the large volume of data entire thermal cycle for a plasma sprayed TBC. They
generated during FEM modeling will be narrowed to demonstrated that the assumption of a stress free
these locations in order to facilitate comparison of the
various cases presented. RADIAL STRESS CONTOURS (MPa)
"*"11 ^^v-i'o
3. Results and Discussion

The first point to note in the discussion is that the intent

89 ^T~'17
of the model at this stage of development is to examine
mechanisms that generate residual stresses (or strains) ^39 ^A
in TBCs. A reasonable assumption is that high residual
stresses are a cause of ceramic failure. Thus, knowing
67 -
P>
fREGIONI u'
)
how the residual stresses are generated provides insight
on how the coating fails. The intent of a "final" model
is to relate the stress generating mechanisms to failure REGION 3
of the TBC. Thus, the important points to follow in this 89
,4-17
paper are the trends in residual stress development with *11 SAK
the selected factors, not the absolute values of the ^ *
REGION 2
stresses. With this in mind, it is informative to compare
the calculated stresses to those measured in real (a)
RADIAL STRESS CONTOURS (MPa)
systems. Coating stress measurements are of the stress
in the plane of the coating, not the stresses normal to -100 ;J.I
the plane. For zirconia, these stresses are quite low for
the as-deposited coating, between 0 and 5 MPa. After
-221^"^^
"i -64

one cycle, the in-plane stresses typically jump to 20 to

30 MPa, and jump to 25 to 35 MPa after 10 one hour \
cycles to 1100°C [10]. In the current study, the average
240-^^
V*'78
^
compressive residual stresses in the plane of the coating
have been calculated to be as high as 200 MPa after 4 114
_^ 7
I
cycles if coating cracking is not allowed, in obvious
disagreement with the measured values. However,
when the coating is allowed to crack in the model, the -29
-221 ^^
average coating stresses after 4 cycles drop to -100 —64
approximately 20 MPa. This comparison indicates: the I
calculated stress is sensitive to the fracture stress of the 10 Hin
coating, the model is capable of predicting realistic
stresses when fracture is included, it is best to use the (b)
stresses reported in the current paper as an indication of Figure 3. Contour plots of the radial stresses at a rough
the likelihood of coating fracture rather than an interface for (a) a low CTE bond coat and (b)
indication of the true stresses in a system. a high CTE bond coat.
17-4

condition at high temperature is valid for a TBC if the bond coat for these conditions. This large effect
bond coat and / or top coat relax sufficiently during a emphasizes the importance of roughness to the stresses
high temperature hold. When this holds true, Petrus et in the near interface region.
al. achieve qualitatively similar results as previous
studies: a rough interface causes an increase in While it is tempting to suggest that a high CTE bond
delamination stresses (tensile stresses normal to the coat favors cracking at bond coat peaks and should
interface) at the peaks of the metallic asperities while therefore shorten TBC life, a high bond coat CTE also
increasing compression normal to the surface in the favors crack arrest over the valleys at this small number
valleys between asperities, as compared to a smooth of cycles. As will be shown later, oxidation at larger
surface. This finding was reexamined for the current numbers of cycles may have a dramatic effect on this
study using four thermal cycles, instead of one, and situation. The combined effect of CTE and higher
including oxidation. It was found that, even when oxidation at large numbers of cycles has not yet been
oxidation is considered, a smooth surface had stresses investigated, but it is known that CTE interacts with
normal to the interface ranging from 5 MPa at the some of the other parameters studied [15].
interface to 0.5 MPa at the surface. A rough surface,
however, exhibited tensile delamination stresses as high The remainder of the discussion will focus on the
as 67 MPa at bond coat peaks, and normal compressive generation of stresses due to the operation of oxidation
stresses as high as 17 MPa in the valleys (Figure 3(a)). and creep. The effects of the various mechanisms are
Tensile stresses were also developed in the valleys due best understood when presented separately, starting
to oxidation, as will be discussed later. Several points with the simple, all elastic case.
of interest arise from this finding: 1) roughness tends to
cause a substantial increase in stresses in the near Elastic Case
interface region as compared to a smooth interface, and
2) the stress is periodic and should tend to limit In the case where all the constituents are elastic and
cracking to the bond coat peak regions. This result is there is no oxidation, the radial stress profiles for the
qualitatively the same as for previous models that three locations are the same for all four cycles (Figure 4
considered only cooling and only one thermal cycle. A (a)). At the bond coat peak, the ceramic is driven into
third point is noted that the properties of the bond coat compression on heating and returns to zero stress on
have nearly no influence on TBC stresses for the case cooling. The stress in the off-peak location also cycles
of a smooth interface. However, bond coat properties between a compressive stress and zero stress. However,
become extremely important when the interface is the stress in the valley cycles between a zero cool-down
rough [7, 8]. Note that roughness profiles, and hence stress and a tensile stress of approximately 70 MPa
the effect of roughness on stress, are extremely during steady state heating. The change in sign of stress
complex and at an early stage of investigation. from the peak to the valley is a result of the change in
sign of the radius of curvature [14].
Bond Coat CTE
Oxidation is introduced incrementally in the model,
For the case of a smooth interface, it has been with the first increment of oxidation starting with the
demonstrated that the properties of the bond coat are start of the hold time in the third cycle. Oxidation
relatively unimportant. Rather, the substrate dominates changes the stresses significantly from the case of no-
stress generation in the coating. The strength of an oxidation (compare cycles 1 and 2 to cycles 3 and 4 in
effect in the presence of a rough bond coat is Figure 4(b)). Oxidation pushes the peak radial stress
exemplified by the effect of bond coat CTE. Figure 3 toward slightly greater compression for all portions of
shows the radial ceramic layer stresses for low CTE the cycle and reduces the amplitude of the stress per
(Figure 3(a)) and high CTE (Figure 3(b)) bond coats cycle. Concurrently, the off peak cool-down stress
and the standard conditions of oxidation and creep of becomes tensile while the off-peak steady state stress
the top coat and bond coat. The tensile stresses at the remains constant. This results in an increase in the
bond coat peak increase from 67 to 240 MPa with the cyclic stress amplitude for the off-peak location, where
increase in CTE from 14 x 10"6 °C_1 to 20 x 10"6 °C1. an increase in the stress amplitude may have an effect
Also, the valley stresses for the high CTE case are on the fatigue life of this material [16]. The most
compressive, up to -221 MPa, while the valley stresses dramatic effect is on the stress in the valley. The valley
for the low CTE case tend to be mildly tensile. The stress increases rapidly with growth of the oxide. The
factor of approximately three change in stress level at linear, nearly vertical increase in stress reflects the
the peak scales well to the approximately factor of three linear growth rate imposed on the oxide. This rapid
change in CTE mismatch between the ceramic and the increase in stress is exaggerated due to the fact that all
 17-5

200
CYCLE 1
HEATINQ
have a significant effect on stress when no other
_ 150
HOLD mechanisms are active. Interestingly, the four cycle
/ COOLING
stress history in Figure 4 b would result in a system that
S 100 -+- could not support the commonly observed peak-to-peak
ÜJ 50
V cracking. This is due to the fact that the radial stresses

w ffB
DC
^ are zero or compressive over the bond coat peaks for
0

-50
-OFF-PEAK all 4 cycles. A real system, however, includes plasticity
-PEAK and creep (or, identically, stress relaxation) behavior
-100
that can modify these stresses substantially.
-150
500 1000 1500 2000 2500
TIME (seconds) Creep Effects
(a)
Before examining the interactions of top coat and bond
coat creep with oxidation, it is informative to revisit the
effect of these mechanisms on stresses in the absence of
oxidation [8] and to examine how these stresses evolve
with thermal cycling. Figures 5 and 6 show the peak,

1000 1500 2500

TIME (seconds)

(b)
Figure 4. Ceramic layer radial stresses at the peak, off-
peak and valley locations for the case of (a)
no oxidation and (b) oxidation starting at cycle
3.

of the oxide growth for the first three cycles is made to 1,000 1,500 2,000 2,500
occur during a short time in the third cycle in the TIME (seconds)

model. As is shown later, this exaggerated increase in Figure 5. Variation in ceramic radial stress with
stress is unimportant when creep is allowed to occur. position and time for the case of a creep in
The valley cool-down stress increased from zero to 120 the ceramic and an elastic bond coat.
MPa and the steady state stress increased from roughly
50 to 170 MPa, causing a substantial oxidation induced off-peak and valley stresses for systems with creeping
increase in the mean tensile stress. Figure 4(b) provides top coat-elastic bond coat and creeping bond coat-
an indication of the stress levels that are expected if elastic top coat, respectively. While the stress always
stresses due to oxide growth are not relieved. The returns to zero in the all-elastic case (without
change in stress amplitude for all locations is a result of oxidation), creep of the top coat (Figure 5) causes a
the change in the system properties with the addition of deviation from a zero cool-down stress at all locations.
a very thin oxide layer with a drastically different At the peak, the cool down stress progressively
modulus and CTE than the coating materials. increases from the zero stress of the all-elastic case to
roughly 70 MPa during the course of four thermal
This simple case illustrates a trend found throughout cycles. This gradually increasing cool-down stress is a
the results: oxidation tends to cause radial cool down result of partial relaxation of the top coat compressive
stresses at the bond coat peak to become more negative, stresses at the peak during each steady state hold
and off-peak and valley stresses to become more (Figure 5). As with any relaxation process, the
positive. Oxidation also causes a decrease in the peak relaxation rate decreases with time (apparent as a
and valley cyclic stress amplitudes and a concurrent decreasing slope during the hold time) and the cool-
increase in the off-peak stress amplitude. down stress tends toward a steady state value.
However, relaxation is not complete by the end of the
The overall view in Figure 4(b) is that oxidation can fourth cycle. The stresses at the off-peak and valley
17-6

locations also change as a function of time. The off- Combined Effects

peak stresses become progressively more positive with
cycling. The valley cool-down stress is actually more The combination of the effects of top coat creep, bond
positive than the elastic case for the first cycle, but then coat creep and oxidation are compared in Figures 7(a) -
becomes progressively more negative, eventually (c) as the radial cool-down stresses for the peak, off-
reversing sign by the fourth cycle. peak and valley locations. First, note that the stresses

The stress amplitude for each cool-down portion of the ONO-OXIDATION

■ OXIDATION
cycle remains constant with increasing cycling for each
a
location, but the mean stress changes. The reason that &
the stress amplitude remains constant is that there has <— ras
been no change in materials properties of the system
during this process and relaxation is relatively inactive u
I
during the transients. The change in mean stress is due
to the relaxation of stress during the steady state hold.
ALL TOPCOAT BOND ALL CREEP
ELASTIC CREEP COAT
CREEP
CYCLE 1 CYCLE 2 CYCLE 3 CYCLE 4
250
HEATING (a)
■ r
D

\ r \U1r.. \U-1r.
200 coo .ING
1—
OFF-PEAK 250 E NO-OXIDATION

1 150
■ s~- U.. * s
200
150
■ OXIDATION

iJI-rii
0) +" • PEAK

U ■J
«2 loo EC £• 100
i U «
= (O
i
afft
J 3 w 50
50 --- 1 r 1 < oc
III </)
0
u

V ^VLA
^V ILA
r\
-50
0 /\_A -100

-50 V
,1
ALL
ELASTIC
TOP
COAT
BOND
COAT
ALL
CREEP
CREEP CREEP
-100 ' ■ i i
500 1,000 1,500 2,000 2,500 (b)
TIME (seconds)

Figure 6. Variation in ceramic radials stress with 250 BNO-OXIDATION

■ OXIDATION
position and time for the case of creep in the 200

bond coat and an elastic top coat. 150

g£
Bond coat creep has a similar effect to that of top coat 2*
o
creep, but the rate of the change is much higher simply <aDC
because the bond coat creeps more readily (Figure 6)
[2]. The magnitude of the change is also greater within TOP BOND ALL
the first four cycles. This is due to the fact that COAT
CREEP
COAT
CREEP
CREEP

relaxation at steady state hold runs to completion for

(c)
each hold cycle, including the first, resulting in Figure 7. Ceramic radial cool-down stress variation for
stabilization of the thermal fatigue cycle after the very the cases of all-elastic, creeping top coat
first cycle. Therefore, stress profiles for all three (elastic bond coat), creeping bond coat (elastic
locations remain constant for the second through fourth top coat) and "all creep" (combined top and
cycles. This does not mean that bond coat creep stops bond coat creep) in oxidizing and non-
after the first cycle. The evidence of continuing bond oxidizing conditions for (a) the peak, (b) off-
coat creep is apparent as a slight decrease in the stress peak and (c) valley locations.
during hold for all 4 cycles. Other work has also shown
that bond coat creep is active for some materials during resulting from the combination of top coat creep and
cool down on every cycle [2]. As with top coat creep, bond coat creep cannot be expressed by simple addition
bond coat creep causes the peak and off-peak stresses of stresses due to the mechanisms acting separately.
to become more positive and the valley stress to
become more negative. The peak stress is lower, the off-peak stress is
intermediate and the valley stress is higher than would
 17-7

be expected of either mechanism separately. This result Figure 3(a)), shows that the area of the peak, off-peak
illustrates the difficulty in looking at only one and especially the valley tensile stress regions are small
mechanism to characterize the stresses in a system that and not connected. Also, a compressive region exists
has more than one mechanism operating. The various between the peaks (note Region 3 in Figure 8) that
creep cases do, however, remain consistent in the sign would prevent planar cracking from peak top to peak
of the stress at each location. top, as is usually observed in failed coatings. Similarly,
the high tensile stress regions are not in the same plane,
Comparing oxidized to non-oxidized stresses shows so that cracks moving from one region to another
there is interaction among mechanisms at the valley would have to follow a path well off of parallel to the
location but that oxidation and creep are roughly interface. Since cracks typically appear planar and
additive for the other locations. In the valley, oxidation parallel to the interface, this configuration does not
reverses the stress from moderately negative in all three agree with observations.
creep cases to highly positive. An interesting point is
that the magnitude of the stress change is for the RADIAL STRESS CONTOURS (MPa)
creeping bond coat and combined creep case is nearly
double that for the ceramic creep case.

The results appear to indicate that the stresses at the

peak and off-peak locations are tensile whether
oxidation occurs or not. Thus, even if no oxidation
occurs, the stress state at these locations is favorable to
delamination cracking. The stress in the valley region
without oxidation is compressive, clearly not conducive
to cracking in this region, in agreement with previous
considerations of wavy interfaces [14]. However,
addition of oxidation changes the sign of the stress to
become highly tensile, and conducive to delamination
cracking, as noted by previous authors [6, 7]. It should
be remembered, however, that the stresses reported in 10 urn
Figures 7(a)-(c) are from single elements quite close to
(a)
RADIAL STRESS CONTOURS (MPa)
RADIAL STRESS CONTOURS (MPa)
1
122 *£fL 39 \
1.1
67^ >

67 -10
"X
~n'
REGION 2 v.
REGION 3 \>
67s. y//
1.1 !

REGION 2
h
10 um
122 -~-~-Srr 39
-\— L!
H 10 jim
Figure 8. Contour plot of the radial stress in the h
ceramic after cool-down following 4 thermal (b)
cycles for the case of combined oxidation Figure 9. Contour plots of the radial ceramic stress at
and creep. cool-down following 51 cycles for the case of
combined creep and (a) without oxidation
the interface. Examination of a contour plot for the and (b) with oxidation.
general creep case with oxidation, Figure 8 (same as
17-8

This difficulty was resolved by running an additional form a uniform tensile region a short distance from the
combined creep case, with and without oxidation, to a bond coat. Cracking in this large, general tensile region
greater number of thermal cycles. Figures 9 compares then becomes possible. Thus, it appears that early
contour plots of the radial stresses after 51 thermal cracking should occur at bond coat peak regions, aided
cycles for non-oxidizing and oxidizing cases. For 51 by ceramic and bond coat creep, and continued
cycles without oxidation, a compressive stress region cracking is driven by the changes in stress over the
extends from the valley well into the ceramic layer valleys driven by oxidation.
(Region 2 in Figure 9 (a)). These compressive stresses
would limit cracking to the region above the bond coat Summary and Conclusions
peaks. With oxidation, it is evident in both the 4 and
51-cycle cases (Figures 8 and 9(b), respectively) that The failure mechanisms of thermal barrier coatings
the tensile stress at the peak has become compressive. have been examined through a FEM model of residual
In the 51 cycle case the tensile off-peak region has stress generation due to oxidation, top coat creep and
grown in area and become more tensile. Additionally, bond coat creep. The results indicate that top coat and
the compressive radial stress regions over the valley bond coat creep act to generate tensile stresses at bond
after 4 cycles with oxidation (Region 3 in Figure 8), coat peak and off-peak locations, while generating
and after 51 cycles without oxidation (Region 2 in compressive stresses in the valley regions. The primary
Figure 9(a)), have completely disappeared after 51 effect of creep on the cool down stress appears to be
cycles with oxidation (Figure 9b, Region 2). Thus, the limited to short times at temperature. Oxidation acts
stresses at 51 cycles with oxidation is much more opposite to creep, causing increasing tensile stresses in
favorable to crack growth over the valleys than is the the valleys while pushing the peak regions into
four cycle case. However, the stresses at 51 cycles are compression. Continuing oxidation results in an
unfavorable for crack growth at the peak of the bond increase in the size of the tensile region associated with
coat. The histograms in Figure 10 provide a direct the valley, growing to the point that the tensile stresses
comparison of the stresses at each position for each in the valley extend beyond the peak regions.
case. Continued oxidation also causes a slight increase in the
maximum tensile stresses at the off-peak and valley
locations.

The combination of oxidation, top coat creep and bond

coat creep yields a stress distribution in time and in
location along the interface that is not achievable by
any one mechanism acting alone. The picture of the
failure mechanism resulting from this evolution of
residual stresses is as follows: 1) Early cracking at bond
4 CYCLE NO 4 CYCLE 61 CYCLE NO 51 CYCLE
OXIDATION OXIDATION OXIDATION OXIDATION coat peaks, driven by tensile peak stresses generated
due to creep processes. 2) Cracks generated early in life
Figure 10. Radial stress in the ceramic at the end of do not propagate due to creep generated compressive
cool-down as a function of location, number stress regions over the valleys. 3) At higher numbers of
of thermal cycles and oxidation. cycles, and therefore higher oxide thicknesses, stresses
over the valleys become increasingly tensile and the
An overall view of the cracking process begins to size of the tensile region increases. 4) The tensile
emerge from these results. At small numbers of thermal region over the valley is then capable of sustaining
cycles, the relatively rapid processes of ceramic and crack growth, resulting in linking of the near-peak
bond coat creep make cracking favorable at the peak cracks generated during early cycling. This description
and off-peak locations. However, cracking over the of failure is supported by the qualitative observations of
valley locations, except right at the interface, is still cracking progression.
difficult due to compressive stresses and/or low tensile
stresses. As the relatively slow process of oxidation This progression of cracking is clearly a result of the
proceeds, cracking over the peaks becomes less combined action of creep, oxidation and thermal
favorable while cracking at the off-peak and valley cycling. An accurate description of the entire process
locations becomes easier. As oxidation proceeds even requires a model including these factors. However,
further, the regions above the valleys grow further until failure of a TBC is very complex and is clearly not
the tensile regions of neighboring valleys converge to completely described by a simple model. Other factors,
such as: sintering; phase changes in the oxide, bond
 17-9

coat and ceramic layer; ceramic layer cracking; and

compositional changes are also thought to be important
aspects that should be included for a complete
description of the failure mechanism.

References

1. R.A. Miller and C.E. Lowell, Thin Solid Films, 95

(1982) 265.
2. W.J. Brindley, J. Thermal Spray Tech., 6 [1] (1997)
85.
3. DJ. Wortman, E.C. Duderstadt and W.A. Nelson,
ASME Paper 89-GT-134, ASME, New York (1989).
4. R.A. Miller, J. Amer. Cer. Soc, 67 (1984) 517.
5. J.T. DeMasi-Marcin, K.D. Sheffler and S. Bose,
ASME Paper 89-GT-132, ASME, New York (1989).
6. G.C. Chang, W.A. Phucharoen and R.A. Miller, Surf.
Coat. Tech., 30 (1987) 13.
7. W.A. Phucharoen, Ph.D Dissertation, Cleveland State
Univ., Cleveland, OH (1990).
8. G.J. Petrus and B.L. Ferguson, J. Thermal Spray
Tech., 6 [1] (1997) 29.
9. B. Engelmann and J.O. Hallquist, NIKE2D, A
Nonlinear, Implicit, Two-Dimensional Finite
Element Code for Solid Mechanics - User Manual,
UCRL-MA-105413, Lawrence Livermore National
Laboratories, Livermore, CA, 1983.
10. R.T.R. McGrann, J.AS. Graves, E.F. Rybicki, J.R.
Shadley and W.J. Brindley, Proc. of the 9th National
Thermal Spray Conference, Cincinnati, Ohio,
October 7-11,1996, ASM International, Materials
Park, Ohio, 1996, p. 885.
11. J.T. DeMasi, M. Ortiz and K.D. Sheffler, Thermal
Barrier Coating Life Prediction Model Development,
Phase I Final Report, NASA CR-182230, NASA
(1989).
12. High Temperature High Strength Nickel Base
Alloys, INCO, Saddlebrook, NJ, 1984.
13. D.W. Richerson, Modern Ceramic Engineering,
Marcel Dekker, Inc. (1992) pp. 166-186.
14. A.G. Evans, G.B. Crumley and R.E. Demaray, Oxid.
Met., 20 [5/6] (1983) 193.
15. A.M. Freborg, B.L. Ferguson, G.J. Petrus and W.J.
Brindley, Proc. of the TBC Workshop 1997,
Cincinnati, OH, May 19-21, 1997, NASA Lewis
Research Center, Cleveland, OH (1997).
16. CA. Andersson, Fracture mechanics of ceramics.
Volume 6 - Measurements, Transformations, and
High-Temperature Fracture, Plenum Press, New
York, 1983, p. 497.
 18-1

THE EFFECT OF TBC UTILIZATION IN THE DESIGN OF ROBUST AIRCRAFT

COMBUSTORS

Carlos A. Arana
Aero Propulsion & Power Directorate
Air Force Wright Laboratory
Wright Patterson AFB OH USA

ABSTRACT h mean beam length of radiation path

m mass flow rate
As performance objectives of new and derivative n thermal effectiveness
military engines require combustion systems to p pressure
operate at higher pressures and temperatures, bal- q" heat flux
ancing conflicting demands of improved durability, T temperature
stability, and operability becomes more difficult
without technology improvements in combustor Subscripts
liner designs. Since combustor thermo-mechanical
fatigue is currently a significant contributor to en- a air
gine life cycle costs, improved structural durability AN annulus
must be achieved, without compromise of other c casing
combustor requirements. This paper addresses the 8 gas
application and verification of liner cooling schemes w wall
and their interaction with thermal barrier coatings WH flame side
(TBCs) for the design of robust aircraft turbine en- WC cold side
gine combustor liners to meet the above mentioned 3 combustor inlet
demands. An analytical investigation was con- 4 combustor exit
ducted to determine the effect of TBCs on the aver-
age metal temperature for a full annular, semi- 1.0 INTRODUCTION
transpiration cooled combustor liner. The perspec-
tive is from a customer's viewpoint, a combustor Air dominance is maintained by fielding affordable
liner designer who is continuously challenged to and durable high performance air platforms capable
increase combustor temperature rise capability and of delivering payload when and where needed in
operability for new products while maintaining the field command. Key to successful air platforms
cooling flow levels. is the propulsion system. The major key to higher
specific thrust is higher engine operating tempera-
LIST OF SYMBOLS ture. Current combustion systems still operate well
below the stoichiometric chemical limits of kero-
A area sene based fuels. In the future, major increases to
a Stefan-Boltzmann constant these values will be required (Hill, 1997) [1]. It is
a absorptivity estimated that overall pressure ratios are expected
density to grow from a present 20 to over 40, and the tur-
P bine entry temperatures from 1800 to 2400°K. As a
8 thickness
result of these higher combustion pressure and gas
M viscosity temperatures, the heat flux levels in these engines
0 cooling effectiveness will also increase making the task of combustor
Cp specific heat liner durability design extremely challenging.
D combustor diameter Therefore, a critical technology need for the devel-
eff effective opment of advanced gas turbine engines will be the
e emissivity control of heat transfer. Since thermal barrier
fc flame chamber coatings are an effective approach to heat transfer
f/a fuel air ratio control, advanced materials will need to be inte-
h heat transfer coefficient grated with robust thermal barrier coatings and in-
k thermal conductivity novative cooling techniques to minimize increases
L luminosity factor in cooling flow requirements. For a combustor

Paper presented at an AGARD SMP Meeting on "Thermal Barrier Coatings",

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
18-2

liner system, TBCs act as thermal insulators reduc- • Combustor Inlet Temperature: As the inlet
ing the amount of heat transmitted to the combustor temperature to the combustor increases, gas tem-
wall from the products of combustion. The heat perature also increases and, therefore, the liner
flux gets dissipated through radiation back into the temperature increases proportionally requiring
hot gas path, conduction across the coating and more air to maintain a given liner temperature.
substrate, and convection through back side cooling TBCs can ameliorate this issue by enabling a re-
air. duction in the amount of radiant heat flux from the
flame to the liner (figure 2).
2.0 THE PROBLEM

Hot section components are significant mainte-

nance items in today's engines. Almost 7 percent
of unscheduled removals of F100-200 engines in
the F-16 due to core components during 1991 were
caused by combustor or turbine problems. In fact,
during this period of time, the mean time between
failures of the combustor in the -200/F16 system
did not meet the Engine Structural Integrity Pro-
gram requirements (Turner, 1996) [2]. TBCs can 500 800 1100
alleviate this problem, however, at a cost. Cur-
Combustor Inlet Temperature °K (T3)
rently the Air Force spends on average $143.3 per
kilogram ($65/lb) for TBC bond materials. Each
combustor consumes around 0.9 -1.7 kilograms (2- Figure 2. Typical TBC payoff for combustor
3 lbs) of bond coat. For a typical top coat the cost liners
is about $28.7 per kilogram ($ 13/lb) and each com- • Combustor Temperature Rise: It is estimated
bustor consumes around 1.7 kilograms (3 lbs). La- that for the high temperature rise combustors which
bor time for the application of these TBC systems will be required to operate at an equivalence ratio
is approximately 1.84 hours per combustor. of 80 percent, over 85% of the entire combustor air
flow will be required to be introduced in the pri-
3.0 CHALLENGES mary combustion zone to completely react all of
the fuel. As a result, the amount of air remaining
In the future, the entire combustor section of a high for liner cooling and combustor exit temperature
performance aircraft engine, as shown in figure 1, control becomes scarce (figure 3). If the cooling
will be subjected to an increasingly severe thermal air is not reduced, there would be no dilution air
environment which will eventually approach stoi- available to reduce hot streaks. This further com-
chiometric conditions. plicates the task of maintaining the liner tempera-
ture at that allowed for conventional metal materi-
als.
Air ports

Dome Film cooing discharges 100

Fuel Injector
O)
3
O
V
\ x^ E
_o
IE
o

3
Figure 1. Internal view of a typical annular m
0
combustor system 0.01 0.03 0.05 0.07
Fuel-Air Ratio
Two major factors will significantly impact the Figure 3. Increased combustor exit
ability to cool the combustor liner: temperatures reduce available
dilution and cooling air
 18-3

From the practical combustor design standpoint, • 99% efficiency at all operating conditions
the objective is not to use improved cooling meth-
ods to lower wall temperatures below the levels Other combustor system design considerations:
achieved with present methods, but rather to
achieve the same wall temperatures despite reduc- • 2000 engine flight hours and 4000 Low Cycle
tions in the amount of coolant flow as shown in Fatigue TACS - hot parts
figure 4. • 4000 engine flight hours and 8000 Low Cycle
Fatigue TACS - cold parts
Impingement/Film
CMC • 10 hour of operation at any steady-state flight
Uncooled CMC o
Hybrid Liner condition
+900
mat 5 • 1 hour of operation at simultaneous T3/T4
■f
Q.
a Quasi-Transpiration
Ü
ODS <* condition
9>
Single Cristal Effusion Cooled
+600 ^- Impingement/Film CMC go« c
• 350 hour of test cell operation
0)
a. '^ou
B 3
■o
a>
Other requirements:
Quasi-Transpiration a
•a
a +300 Double Wall 10' ■a
m £
a
a Film-Cooled 3
• Buckling margin of safety as dictated by com-
Effusion Cooled o-
Double Wall
Single Wall 0)
DC
bustor geometry (shape and wall thickness) shall
exceed 50 percent under maximum pressure and
TIME temperature
Figure 4. Combustor liner material/cooling
progression for the past 15 years. • 5% creep limit without loss of mechanical
alignment, as dictated by structural geometry to
4.0 GENERAL REQUIREMENTS AND DE- mitigate the interaction of thermal and mechanical
SIGN CONSIDERATIONS stresses

The primary technical objective of combustor re- • Component high cycle fatigue life greater than
search and development is to provide the technol- 107 cycles, as determined by both material and
ogy that will enable the designers attain combustor structure
performance, operability, weight and cost objec-
tives for a particular mission and cycle definition. • -3a material properties (average for crack
propagation)
As systems become more efficient and probably
more complex, the next generation of combustor • 500 0-maximum-0 cycles at maximum power
designs will need to balance conflicting demands
such as increased temperature while improving du- 5.0 COMBUSTOR TBC EVOLUTION
rability, maintainability, and affordability. In order
to achieve these demands an integrated prod- Since the birth of thermal barrier coatings in the
uct/process must be implemented. This process early 1960's, combustors and augmentors have
should enable the production of a robust design benefited from a series of incremental growths in
capable of operating efficiently in a wide variety of this technology. This has resulted in an increased
environments dictated by the intended air vehicle. ability to withstand higher temperatures with a de-
To satisfy the requirements of future cycle parame- crease in the severity of oxidation damage, and an
ters, technology improvements are required. Ad- improved cracking resistance to severe thermal
vanced planning allow us to establish combustor cycling for longer periods of time (Harris,
performance objectives that will support overall 1991) [3]. Early combustor coatings consisted of an
engine system improvements. Some combustor outer ceramic insulative layer of air plasma depos-
specific performance objectives are: ited, 22 weight percent MgO stabilized Zr02, and
an inner metallic bond coat of air plasma deposited
• 50% weight reduction Ni-Cr or Ni-Al (Gupta, et al.,1994)[4].
• 100% increase in volumetric heat release rate
• +900° F maximum combustor exit temperature As these temperatures in the combustor have been
• +400° F combustor inlet temperature increased, the durability of this early combustor
• 300% increase in fuel/air turndown ratio coating has become inadequate. The maximum use
18-4

temperature of the plasma sprayed magnesia- duced through the liner walls downstream of the
stabilized zirconia coating is on the order of dilution jets may not be effective in the combustion
1255°K since magnesia-stabilized zirconia crystal- process, therefore, contributing to potential com-
lographically destabilizes above 1227°K. bustor inefficiency.

By replacing the 22 weight percent magnesia with u.

32 . o o
7 weight percent yttria composition, the spallation
life at temperature above 1255°K has been im-
"53
a 24 p-3
LL
proved substantially (Stecura, 1979) [5]. This im- O)
16 — O
m
provement is approximately 4 times that of 22 per- f
cent magnesia fully stabilized zirconia coating at O 8
1367°K (Gupta et al, 1991)[6].
n

Since the operating temperature of the combustor 32

was increased to take advantage of the increased
temperature capability of the yttria stabilized zir- ?
Li. 24
conia, the underlying bond coat oxidation became a O)
16
problem. By adopting a more oxidation-resistance
NiCoCr A1Y bond coat composition from the tur- O 8
Ü
bine section of the engine, the performance of the
combustor has been improved (Gupta et al.,
1994)[4]. Figure 5 illustrates the progression of _ 5
TBC's for hot section applications. <D

£ 4
D)
Electron Beam Physical
Vapor Deposition £
O 3
J
Generation 3 ^ Nano-layers z
for Reduced Thermal O) 2|—
^ Conductivity

Low Pressure Chamber Plasma Current Low emissions

Sprayed for Improved Bond Strength
Production Combustors

Flame Spray-Air Plasma Deposited Figure 6. Efficiency Benefits for Idle Engine
for Improved Component Durability
Operating Condition
Figure 5. TBC Progression for Hot Section
Applications 6.2 Exhaust Smoke Emissions

Exhaust smoke numbers decrease slightly with

6.0 TBC APPLICATION ADVANTAGES
thermal barrier coatings as reported by Butze, et
al., 1976 [8] from which figure 7 is reproduced, for
Let us look at other potential benefits of applying a cruise condition. Soot is the primary source of
TBC's in the combustor section: flame emissivity at high pressures. The intense
radiation from the ceramic coating back to the
6.1 Efficiency and Idle Emissions flame results in a decrease in soot concentrations
due to reduction of the amount of soot formed ini-
As summarized by the results of Mularz et al., 1978 tially and through burnup of the soot formed.
[7], from which figure 6 is reproduced, the appli-
cation of thermal barrier coating along the inside of 6.3 Combustor Liner Life
the combustor liner allows combustion gases near
the wall to be at higher temperatures. This mini- The average cooling temperature that is required in
mizes wall quenching effects of the combustion a typical production combustor today is graphically
chemical kinetics and therefore, reducing unwanted illustrated in figure 8. The gas temperature re-
exhaust emissions such as carbon monoxide and ferred to in this figure represents the maximum
unburned hydrocarbons. Since the design intent of combustor exit temperature. The allowable operat-
cooling air is to protect the liner surface, air intro- ing temperature for the metal liner is about half
 18-5

way between the maximum gas temperature and the was performed in order to estimate the thermal per-
temperature of the cooling flow (T3). Also shown formance of TBCs and compare the capability of
in this figure is the relationship of the allowable TBC improvements against cooling effectiveness
design metal temperature levels to the incipient improvements and T3 reductions. The heat transfer
metal temperatures. As shown, the allowable metal model outlined in this section closely follows the
temperature levels are near the incipient melt point development approach as detailed by Lefevbre, et
of nickel based superalloys. Therefore, any break- al., 1960 [11].
down in the cooling layer would be detrimental.
Experimental Data
40 r —Uncoated Panel
6. 360 —Coated Panel
E Calculated Data
v. 32 CD
<D *-> 320 o Uncoated Panel
n
■ Coated Panel
i_
E co
3
■i—■ ?80
Z 24 ID
JS c
o
E Test Condition Fuel 1 240
0) 16
LU / A Cruise Jet A
Q.
< O Takeoff Jet A E 200
CO / CD
•4—>

Solid Symbols Denote 0)

O) 160
Ceramic Coated Liner CO
l_

_L j 1 1 1 CD
+■ 1100 1180 1260 1340 1420
>
co
120

Average Exhaust-Gas Temperature ~ K i_

0
r— 80
Figure 7. Effect of ceramic coating on SAE
smoke number 40'.01 _L _L J_
.02 .03 .04 .05 .06

Fuel Air Ratio

2000
1921° K
Figure 9. Liner Temperature Prediction vs.
Incipient Melt
1588° Kf Experimental Results
2x Life
o 1500 1491° Kj"
1477° K|T Figure 10 illustrates the schematic of radiant heat
3 flow through a TBC at isothermal conditions. For
g> iooo this analysis the temperature drop through the ce-
E ramic was assumed to be linear since yttria-
810° K
0)
stabilized zirconia thermal conductivity is nearly
500
constant over the temperature range of interest
(1.5±0.1 W/m»K over 1200 to 2260°K). Emittance
and transmittance were assumed invariant with
Gas Wall Air
temperature. The cycle conditions chosen for the
analysis are as follow:
Figure 8. Exhaust, Wall and Air Temperature
Comparison P3 = 3040KPa, T3 = 1144°K, coolant mass flux =
3.416 Kg/s-m2
7.0 HEAT MODEL It was assumed that the gas temperature next to the
wall is equal to the average flame temperature,
According to experimental results from Butze et having a value close to stoichiometric. This as-
al., 1976 [9], as shown in figure 9, the liner metal sumes the possibility of a hot streak occurring next
temperature reductions available through the use of to the wall due to combined effects of a localized
ceramic coatings can be accurately predicted using low annulus velocity on one side of the wall coin-
a one-dimensional heat transfer model. Having that ciding with a breakdown in the cooling layer on the
in mind, a one-dimensional heat transfer analysis other.
18-6

Stabilized Zirconia
Ceramic Coating
Ceramic Filled Bond
Coating Surface

w E Xl
Hot Combustion °° • Metal
Substrate
Gas Radiation Wie
lel
and Convection
1
Ceramic, Tmetal
Tg Surface

TBC

Figure 10. Schematic of radiant heat flow through the TBC at isothermal conditions.

7.1 Internal Radiant Heat Flux 7.2 External Radiant Heat Flux

The radiation heat flux was calculated for the case Radiant heat flow from the combustor liner to the
of a TBC-coated liner using the results obtained by casing is given by:
Liebert, 1978 [12] where emittance of both the ce- £
ramic translucent layer and the material substrate Wec[TWC~T3 )
<1 RAD = ° A {5}
were taken into consideration.
ec+ew(l-ec){ yAc)
<1"RAD = <y(0.5)egT*-5(Tg2-5 -TWH25) {1}
7.3 Internal Convective Heat Flux

The radiation heat flux was calculated for the case Convective heat flow to the combustor liner from
of an uncoated liner using reference [11]: the gases is given by:

_„(1 + «w) „ r1.5/T2.5

<?' RAD = <T ' n"'' CgTg'iTg^ - TWH25) {2} / m \ 0.8
2 02
1"C0NV = °- " 0.2 (Tg-TWHJ
D \Awßj
The flame emissivity, £g, was calculated using fc
{6}
reference [11]: 7.4 External Convective Heat Flow

Eg=\- exp(-290P LV(/ / a)lb Tfh5) {3} Convective heat flow from the combustor liner to
the annulus air is given by:

where the luminosity factor was calculated using / m \ 0.8

reference [13]: «•W = 0.02-fe (TWA ~ T3 )
Aß
DAN \™H-aJ AN
L = 0.0691(C7//-1.82) 2.71 {4} {7}

7.5 Conduction Heat Flux

where C/H is the fuel carbon/hydrogen ratio by
weight. For future high pressure ratio engines, the
The radial heat flux conducted through the liner
practicality of using this correlation could be
wall is given by:
somewhat dubious. More experimental data will
need to be generated.
 18-7

• Figure 12 shows how the heat fluxes vary as

<1"C0ND ~ T {TWH ~ TWc) W the effective conductivity changes. The heat flux
due to conduction diminishes as the thickness of
the TBC system is increased. For the same case,
where the outside convective and radiation heat fluxes
{9}
eff also decrease as expected. The internal radiation
heat flux is maximum for the case of the uncoated
Metal liner since a small fraction of the heat gets reflected
back to the gases.
7.6 Convective Heat Flux Through Liner

The heat flux picked up by the coolant as it passes 1,600,000 ■

through the semi-transpiration cooled wall is ex- 1,400,000 Qk

St
pressed by
<£j 1,200,000 /-\„
^1,000,000 tfa
9"air = ™airCPairn' {TWH ~ T3 ) 10} ■\ -

* 800,000
* r%in
where T) is the thermal effectiveness, which is a i 600,000 1 * •
R-in
function of the internal heat transfer coefficient I 400,000
within the liner. This heat flux is absorbed by the C-out
200,000 R-ouk *
coolant and therefore reduces the heat load that 'K
0 ''— F —
must be removed by q'coNv and q"RAD on the out-
0 0.2 0.4 0.6 0.8 1
side liner. % TBC thickness (3.81x10-4)m

The following figures summarize the results of the Figure 12. The Effect of TBC Thickness on
investigation: heat fluxes

• Figure 11 shows the potential reduction in

metal wall temperature attainable with a refractory • Figure 13 shows the reduction in metal tem-
coating for the semi-transpiration cooled liner sys- peratures without TBC as coolant temperature de-
tem. It is obvious that changes in TBC thickness creases for the same amount of coolant flow rate.
have a pronounced effect on combustor liner tem- Comparing these results with the results from fig-
perature as reflected in the figure. It is also evident ure 14, it is evident that although the TBC surface
that with the application of the TBC to the liner, temperature is higher, the actual metal temperature
the liner wall metal temperature would be capable (substrate) on the gas side is much lower.
of meeting life requirements as mentioned in sec-
tion 4.
1550
1700 1 *1500 ^
l-ID L/ i
1600 £1450
X. 3 J

> .N \ 1S1400
1500 <u
2
3
as 1400
~—-j
!i=»*_.
-i
vT- rneiai nut siue £1350
<u
T- metal cold sic le_
03 x\ =1300
xX
Q.
E to
1300
x> Sv ^s,
51250
5 1200
\^ 1200
700 750 800 850 900 950 1000
1100 1 1 ■ Coolant Air Temperature ~ K
0 0.02 0.1 0.2 0.5 0.75 1
% TBC thickness (3.81 x 10-4)m

Figure 13. The Effect of Coolant Air

Figure 11. The Effect of TBC Thickness on Temperature Variation on Liner
Liner Wall Temperature Wall Temperature (Uncoated)
18-8

1600
tem. TBC reliability must be equal or greater than
the substrate since the integrity of the coating be-
T-TBC
* 1400 comes prime reliable as loss of the coating can re-
§ sult in rapid deterioration of the liner. Also, the
I 1200 ability to apply TBCs to components with complex
(D
geometries should be improved. As combustor in-
11000 — i -meiai noi siae
L~-~- ._-••■•
let air temperature approaches the maximum al-
I 800 —
C5— — lowable wall temperature, there simply will not be
T-metal cold side
enough cooling potential left to cool the wall ade-
600 quately. This leaves the possibility of no metal wall
700 750 800 850 900 950 1000 structure, no matter how effective, will maintain
Coolant Air Temperature - K
reasonable metal temperatures at those conditions.

Figure 14. The Effect of Coolant Air An analytical investigation was conducted to de-
Temperature Variation on TBC termine, through qualitative trends rather than
Liner Wall Temperature quantitative values, the effect of a ceramic coating
on the average metal temperatures of a full annular,
• Figure 15 shows the impact of liner thermal semi-transpiration cooled combustor liner. This
effectiveness variation on wall temperatures. As investigation was conducted at a pressure of 3040
would be expected, the hot side metal temperature KPa, inlet air temperatures of 1144°K and an
is reduced as the ability of the liner to extract heat overall fuel air ratio of 0.67. The insulating effects
becomes greater for the same amount of coolant as well as increased reflectivity of the ceramic
flow rate coating were responsible for the reduction in heat
transfer through the liner walls.
1700
^ 1600 T-tbc
To solve the heat flux equations a non linear algo-
rithm was used. For that purpose a Newton-
2 1500
Raphson non-linear solver was developed to solve
JS
Q-
1400 the equations in an iterative manner. It should be
1 1300 T-me tal he t side
mentioned that the energy balance equations repre-
h, „„.. ^*™~i h-«^z; sent a gross approximation to the heat transfer
=5 1200 process taking place between the liner structure and
T ■meta cold side- —=-—,
1100 1i
i 1
its surroundings. A more rigorous formulation for
0.5 0.55 0.6 0.65 0.7 the heat transfer process may be developed if spe-
Thermal Effectiveness cific details are required.

It appeared that the reduction in metal temperature

Figure 15. The Effect of Variation of Thermal at any given coating thickness is considerably
Effectiveness on TBC Liner Wall Temperature greater than any improvements achieved solely by
cooling configuration changes.
8.0 SUMMARY
It was noted that a relationship between liner
The quest for higher thrust-to-weight ratio will en- cooling effectiveness and effective conductivity
tail increasing combustor inlet and exit tempera- exists. That relationship was considered in the
tures. This will impose very stringent requirements analysis, however, with some limitations. In other
for the development of robust aircraft combustors. words as cooling effectiveness changes there corre-
A large portion of the heat transferred to the liner sponds a change of effective conductivity.
wall from the hot combusting gases and particles
within the combustor will be by radiation. But 9.0 REFERENCES
significant advances can be obtained by incorporat-
ing third generation TBC into the next production 1. Hill, R.J., "The Future of Gas Turbine Engines",
of combustor liners. To maximize the potential, in "Propulsion and Energy Issues for the 21st
there needs to be a focused effort. The thermal ef- Century", AGARD CP 824, March 1997,
fect of the TBC ceramic layer must become an in- Chapter 2.
tegral element of the combustor liner design sys-
 18-9

2. Turner D., "Enhanced Performance Extended 13. Odgers, J., Kretschmer, D., "A Simple Method
Life Combustor", WL-TR-2136, November for the Prediction of Wall Temperatures in a
1996 Gas Turbine", ASME Paper No.78-GT-90,
April 1978.
3. Harris, D., "Composite Thermal Barrier
Coatings", Fourth National Spray
Conference, CF, May 1991.

4. Gupta D.K., Meier, S.M., "The Evolution of

Thermal Barrier Coatings in Gas Turbine
Engine Applications", Transaction of ASME,
Vol. 116, January 1994.

5. Stecura, S., "Effect of Compositional Changes

on the Performance of a Thermal Barrier
Coating System", NASA TM-78976, 1979

6. Gupta D.K., Meier, S.M., "Ceramic Thermal

Barrier Coatings for Commercial Gas Turbine
Engines", JOM, March 1991

7. Mularz, E.J., Cleason C.C., and Dodds, W.J.,

"Combustor Concepts for Aircraft Gas Turbine
Low Power Emissions Reduction", AIAA/SAE
No. 78-999, 14th Joint Propulsion Conference,
1978

8. Butze, H.F., Liebert, C.H., "Effect of Ceramic

Coating of JT8D Combustor Liner on
Maximum Liner Temperatures and Other
Combustor Performance Parameters",
NASA TM X-73581, December 1976

9. Claus, R.W., Wear, J.D., Liebert, C.H.,

"Ceramic Coating Effect on Liner Metal
Temperatures of Film-Cooled Annular
Combustor", NASA Technical Paper No.
1323, 1979

10. Wassel, A.B., "The Design and Development

of High Performance Combustors", Lecture
Series 93, Von Karman Institute of Fluid
Dynamics, Brussels, 1997.

11. Lefevbre, A.H., Herbert, M.V., "Heat Transfer

Processes in Gas Turbine Combustion
Chambers", Proc. Inst. Mech. Eng., 1960, Vol.
174, No.l2,pp 463-473.

12. Liebert, C.H., "Emittance and Absorptance of

NASA Ceramic Thermal Barrier Coating
System", NASA Technical Paper 1190, April
1978.
 20-1

Advantages/Disadvantages of Various TBC Systems as perceived by the Engine

Manufacturer

P.Morrell and D.S.Rickerby

Rolls-Royce pic
POBox31
Derby DE24 8BJ
United Kingdom

SUMMARY (Figure 1), [Ref.l]. Thermal barrier coatings (TBC's)

offer the potential of allowing an increase in operating
This paper discusses the relative advantages and temperature of 70 to 150°C without any increase in metal
disadvantages of Thermal Barrier Coatings systems operating temperatures or the ability to reduce the mass
(TBC's) produced by thermal spraying and electron- flow of coolant whilst maintaining the operating
beam PVD processing technology from a design point of temperature of the turbine. Both possibilities are very
view. This paper reviews the structure/property attractive to the turbine blade designer. In the first case
relationships for electron-beam physical vapour the equivalent of 2-3 generations of turbine alloy
deposition (EB-PVD) TBC's in contrast to those of development can be achieved with the application of the
plasma sprayed TBC's, particularly with respect to coating, in the second case improvements in specific fuel
thermal conductivity, erosion resistance and mechanical consumption of 0.25% can be realised. This saving may
behaviour. Examples are used to show how, through the not appear significant, but can result in an annual saving
development of customised bondcoat systems, the of the order of £6M for an airline operating a fleet of 747
performance of TBC systems can be considerably or 777 aircraft.
enhanced, an important incremental step towards the
ultimate goal of "designed-in TBC's". USE OF ZIRCONIA-BASED TBC'S

INTRODUCTION By virtue of a low thermal conductivity, TBC's reduce

the heat flux into a component allowing a high thermal
To improve fuel efficiency and performance the gradient to be substended across the thickness of the
operating temperatures of the turbine section of the aero- ceramic (Figure 2). In addition, TBC systems are
engine have been increased significantly over the past designed to match as close as is practical the expansion
twenty years. This temperature rise has been met by the characteristics of the nickel-based component to ensure
use of materials with increasing temperature capability, that the thermal mismatch stresses are minimised. During
the introduction of advanced manufacturing technology the early 1970's, due to their relatively high coefficients
producing cast blades with improved cooling of thermal expansion and low thermal conductivities
effectiveness, and the introduction of directionally [Ref. 2],zirconia partially stabilised with magnesia (20%
solidified and single crystal turbine blade technology MgSZ), calcia (20% CaSZ) or yttria (8wt. % PYSZ) were

1350
1300
1250 Thermal Barrier Coatings
1200
o 1150
£3 1100
Directionally Solidified
2at 1050
Alloys
ingle Crystal
a 1000 Alloys
Ea> 950
900 Conventionally Cast
Alloys
850
800 Wrought Alloys
750
1940 1950 1960 1970 1980 1990 2000 2010

Year

Figure 1 Increase in operational temperature of turbine components made possible by alloy, manufacturing technology and thermal
barrier development.

Paper presented at an AGARD SMP Meeting on "Thermal Barrier Coatings'

held in Aalborg, Denmark, 15-16 October 1997, and published in R-823.
20-2

Hot Gas
Temperature

Temperature Profile
without TBC

1400 1600
Temperature (*C)
Temperature Profile
with TBC
Coolant Figure 4 Relative phase compositions of aged 8% PYSZ TBC's.
Temperature Note the initial coatings were all T face-centred tetragonal
phase.

Ceramic Metal Wall

Figure 2 Schematic of the principle of the thermal barrier coating
PLASMA SPRAYED VERSUS PVD TBC'S
used for plasma sprayed thermal barrier applications.
Early TBC's were manufactured from magnesia or calcia Although plasma sprayed TBC's have performed well in
stabilised zirconia, which performed well in service at service on annular surfaces in the engine, their
operating temperatures below c.a. 1000°C. Above this microstructures do not lend themselves to producing
temperature significant diffusion of the magnesium or coatings with the necessary strain compliance, erosion
calcium ions occurs and precipitates of MgO or CaO are resistance or surface finish required for successful
formed [Ref. 3] which results in an increase in thermal application on blade or nozzle guide vane aerofoils. As a
conductivity (Figure 3) and the monoclinic content of the result their introduction into the high-pressure turbine has
coating. This rise in the monoclinic content leads to been limited. An example of the structure of a plasma
mechanical instability of the coating as a result of the sprayed coating is shown in Figure 6, where the
microcracking associated with the martensitic individual platelets formed from the droplets impinging
monoclinic/tetragonal phase change on thermal cycling. on the surface during the spraying process can clearly be
seen. In contrast, more recently electron-beam physical
This limitation of low operating temperature allied to vapour deposition (EB-PVD) has been used to deposit
phase instability was overcome by the introduction of TBC's which, because of their columnar microstructure,
8% PYSZ in the late 1970's, a material which is relatively exhibit very high levels of strain compliance (Figure 7).
stable for elongated periods at temperatures up to 1500°C The columnar microstructure which forms is due to the
with no precipitation of Y203 from solution; (Figure 4) atomistic processes, nucleation and growth of the coating
[Refs. 4,5]. PYSZ TBC's have performed well in service, from the vapour phase. In addition to their good
significantly increasing the life of components. For compliance, EB-PVD TBC systems also offer other
example, the application of 8% PYSZ TBC to benefits over plasma sprayed TBC's in terms of improved
combustion can walls reduces the thermal stresses in the adhesion, surface finish and erosion resistance which has
can and can result in component lives in excess of 20,000 led to their application onto turbine aerofoils. The effects
of these properties on both coating and engine
hours (Figure 5).
a

* R
■2 ^
E Cooling i

Isothermal
ageing
5 Heating
I2 --
\-
n
200 400 600 800 1000 1200 1400
Temperature (°C) Figure 5 An example of a combustion can with plasma sprayed
TBC on the inner can walls.
Figure 3 Thermal diffusivity of CaSZ TBC showing thermal
instability due to precipitation of CaO out of solution.
performance are discussed in the following sections.
 20-3

Figure 7 Photomicrograph of EB-PVD TBC showing the

Figure 6 Photomicrograph of a plasma sprayed TBC coating. columnar microstructure.
The network of microcracks (platelet boundaries) are clearly
visible together with the porosity associated with these
coatings. Polishing of the thermally sprayed TBC results in a
surface finish of 2-3 urn, however as will be discussed
Surface finish later, erosion of the coating in service will increase the
For aerofoil applications, the coefficient of friction (Cf) surface roughness of the TBC back to values approaching
of the surface of the blade is directly related to roughness 10 um. As a result the designer will favour EB-PVD
[Ref. 4] and in turn the Cf value influences the heat TBC systems for use on high pressure turbine blade and
transfer coefficient and efficiency of the turbine [Refs. nozzle guide vane aerofoils for the aerogas turbine
6,7,8]. The primary loss coefficient is plotted in Figure 9 engine.
for Reynolds numbers between 0.5-1.5xl06 which covers
the range for high-pressure turbine aerofoils. As can be
seen in Figure 8, the as deposited plasma sprayed TBC
will have a significant effect on the performance of the
turbine. The plasma sprayed TBC has a typical surface
roughness of c.a 10 um Ra (Figure 9) and a peak to
valley height of 80-100 urn which results in an increase
in the primary loss coefficient in the turbine of the order
of 2% (from 1.5% to 3.5%) [Ref. 7]. In contrast the EB-
PVD TBC surface finish is close to that of the metal
surface, typically 1.0 um Ra (Figure 10) with a peak to
valley height of 10 um.These coatings exhibit a primary
loss coefficient of the order of 1.5%, and do not appear to
increase the loss coefficient over that observed for the Figure 9 Scanning electron micrograph of the surface of a
plasma sprayed TBC showing the high surface roughness and
uncoated metal surface. undulating nature of the surface.

Thermally Sprayed TBC

0.5 1 1.5 2 2.5

Reynolds number RexlO/8

Figure 8 Primary loss coefficient for a high pressure turbine

aerofoil showing the effects of different surface roughness of
TBC.

Figure 10 Scanning electron micrograph of the surface of an

EB-PVD coating showing the pyramidal end form of the
columns and low surface roughness.
 20-4

measurements are made with the laser ultrasonic method

at high temperatures the signal is very highly attenuated
by the coating indicating that the column boundaries are
200 ^^—-_____^_ Plasma Sprayed TBC playing a significant role. It is expected that when the
coating is in tension and the column boundaries open up,
SOLID ZrO2-8wt%Y203^—^^ the composite tensile modulus of the EB-PVD coating
100
EB-PVD TBC
and column boundaries will reduce significantly more
than the compressive values.
0 — ■ ■

200 400 600 800 The relatively high modulus associated with the plasma
Temperature (°C)
sprayed TBC can lead to high thermal strains in the
coating, especially around tight radii of curvature which
Figure 11 Young's moduli of solid sintered 8% PYSZ, plasma
sprayed and EB-PVD 8% PYSZ TBC systems measured by
are found around the leading edges of high-pressure
dynamic techniques. turbine aerofoils. As an example Figure 12 compares the
microstructures of a plasma and an EB-PVD TBC which
have undergone thermal cyclic testing. In the case of the
For TBC applications in large turbine components plasma sprayed coating, there has been significant
(particularly in the power generation industry) the cracking within the ceramic leading to gross failure of the
relative roughness of the plasma sprayed TBC compared TBC system, whereas the EB-PVD system has suffered
to the blade chord dimensions becomes more favourable little mechanical damage within the ceramic layer.
and the effects of increase friction coefficient become
less important. Similarly, the local gas velocities in the
combustion zones are lower whcih can lead to low
coefficient of friction losses and use of plasma sprayed
TBC systems in such circumstances are not so
detrimental to performance.

Strain Compliance

Several papers have been published detailing the

modulus of TBC systems measured by various techniques
including compressive and tensile tests [Ref. 9], dynamic
vibration and ultrasonic velocity measurements tests
[Refs. 10,11] and the values generated vary depending on
the techniques used. More specifically the values vary
dependent on the displacements the test method imposes
during measurement.

In this paper the moduli of bulk sintered 8% PYSZ

material and EB-PVD 8% PYSZ TBC were measured by
a laser ultrasonic method [Ref. 11] and compared to those
values measured by a resonant technique for plasma
sprayed 8% PYSZ [Ref. 9]. In both methods the
calculation of modulus is based on the measurement of a
compression surface and shear elastic waves in the
coating material. Figure 11 summaries the Young's
moduli of solid sintered 8% PYSZ, plasma sprayed and
EB-PVD 8% PYSZ TBC's. The modulus of the plasma
sprayed TBC systems is similar to that of the sintered
material, with a room temperature value of 200 GPa
reducing to 160 GPa at 1000°C. The EB-PVD TBC in
comparison exhibits a Young's modulus approximately
half of that of the bulk material, ranging from 100 GPa at
room temperature down to 40 GPa at 1100°C. Care has to
be taken to remember that this is the modulus of the Figure 12 Photomicrographs of (a) plasma sprayed and
(b) EB-PVD, TBC's showing the thermal strain damage in the
coating when the columns of the PVD coating are in plasma sprayed TBC following burner rig testing.
contact and they are capable of transferring a
compressive elastic wave through the structure. When
 20-5

0.16
Si02 ERODANT AT ROOM
TEMPERATURE '
„0.12

J Plasma sprayed TBC

DC 0.08 /T

£ \ EB-PBD TBC
l

20 40 60 80
Impingement Angle (")
Plasma Sprayed EB-PVD Substrate Figure 14 Room temperature erosion rate of plasma and EB-PVD
TBC systems as a function of impingement angle.
Figure 13 Erosion resistance of plasma sprayed and EB-PVD
TBC's, compared to that of a metallic substrate.

mechanisms of erosion are well understood [Ref. 13]. At

Again, like surface finish, the EB-PVD TBC systems are low and intermediate temperatures, the erosion of plasma
more attractive to the designer where the coating is sprayed TBC systems involves the removal of coating
subjected to high thermal and mechanical loads during material in the form of the mechanically bonded platelets.
engine operation. The conditions encountered around the This requires relatively low amounts of energy as the
aerofoil surface in the high-pressure turbine represent coating already contains a high number of microcracks,
such an environment and it is not surprising that EB-PVD and particle boundaries, therefore the erodant has to do
is the principal choice in this region. little work on the systems to generate small amounts of
new crack surface to remove the platelet. As relatively
Erosion resistance large masses are removed with little mass of erodant, the
mass erosion rate is high. In contrast, the erosion of EB-
The erosion resistance of TBC coatings has been PVD TBC's requires the generation of a new crack
measured using high-temperature erosion facilities. Early surface across individual columns which needs a
studies [Ref. 12] showed that the erosion resistance of relatively high amount of energy compared with the
EB-PVD TBC's was approximately 8-10 times that of the plasma sprayed coating (Figure 15). When this fracture
plasma sprayed TBC system. Further erosion studies at has occurred, the amount of material which is released is
temperatures up to 910°C [Ref. 11,13] using A1203 relatively small in comparison to the platelets in a
erodant at velocities up to 230 m/s, showed that the plasma sprayed deposit and, as a consequence, the
erosion resistance of EB-PVD at both 20 and 910°C were erosion rate is low. In addition, because the cracking
approximately 7 times that of plasma sprayed TBC's at damage in the EB-PVD TBC due to small erodant
90° impingement (Figure 13). Erosion tests at room particles is limited to the near surface regions, the surface
temperature also showed that the angle of incidence of finish of the coating is not adversely affected by the
the particle also had a notable effect on the erosion rate. erosion process. This is not the case for plasma sprayed
For the plasma sprayed TBC systems the erosion rate TBC's which have been polished, where erosion of the
decrease with the angle of incidence with the rate at 30° coating causes the surface finish to revert quite quickly to
impingement being one third ofthat at 90° (Figure 14) the as deposited values, leading to significant
[Ref. 13]. performance penalties as discussed earlier (Figure 16)
[Ref. 12].

Plasma Sprayed TBC

500 1000 1500

Time (hours)

Figure 16 Effects of erosion on the surface finish of EB-PVD and

Figure 15 Generation of cracks in the near surface region of the
plasma sprayed TBC's.
EB-PVD TBC systems during erosion testing.
The reasons for this erosion behaviour and the
20-6

Indenlor (Rigid elastic)

Coating (Elastic)

Substrate (Elastic - plastic with

strain hardening)

Figure 18 Schematic of the transverse scratch-adhesion test

EB-PVD MCrALY
v bondcoat

Figure 17 Example of large particle damage at high

temperature showing the degree of plasticity shown by the 8%
PYSZ coating.

As 1050'C 1100'C 1150'C 1200'C

Received
At high temperatures there is certainly evidence within Ageing Temperature (100Hours)
EB-PVD TBC's that significant plasticity can occur Figure 19 Scratch test adhesion values for EB-PVD TBC's
especially when large particles, foreign object damage systems on directionally solidified MarM002 alloy.
(FOD), impinge onto the surface (Figure 17). This FOD
mechanism of coating damage has been seen in
development engines where the impact damage matches -DYNA PRED1TICTION OF STRESS
that produced in the high-temperature erosion rig used U 800
. SCRATCH TEST CRmCAL LOAD
Q.
for laboratory assessment. s 40 3-
I
a
In summary, for low angle impingement, such as exists s
on the annular surfaces of the combustion can and nozzle -P-+- 20 =

guide vane platforms, or where gas velocities (and

subsequently particle velocities ) are low, plasma sprayed
0 5 10 15 20
TBC systems can perform as adequately as PVD. DIAMETER (mm)
However where the gas velocities are high and the Figure 20 The effect of radius of curvature on scratch test value
impingement angle moves closer to 90°, EB-PVD TBC and the resolved interfacial strength.
systems offer significant advantages over plasma sprayed
TBC's. Briefly this test involves indenting the substrate of a
coated sample which is metallographically prepared in a
Coating adhesion mount, and moving the indentor from the substrate
though the interface region and the coating to the mount
Traditionally the adhesion of plasma sprayed coatings has material (Figure 18). This process is repeated several
been assessed by the use of bend testing, lap shear testing times at various indentation loads, the crack length at the
and pull-off testing amongst other techniques [Ref 14]. interface of the bondcoat and ceramic layers being
Fr plasma sprayed TBC's of 0.25 mm thickness, as plotted as a function of load and extrapolated to a no-
deposited adhesion strengths are in the range 20-40 MPa cracking critical load. This load is then used directly to
with the strength decreasing as the coating thickness rank coating systems or can be converted via finite
increases. As deposited EB-PVD TBC systems exhibit element modelling to an interface strength in MPa [Ref.
adhesive strengths higher than the glue and generally no 15]. Figure 19 shows the scratch-test data for two EB-
useful data is generated from pull-off testing these PVD TBC systems, one with a EB-PVD MCrAlY
systems.To aid in the development of improved bondcoat bondcoat the other with a platinum-aluminide bondcoat.
systems, a method of assessment of adhesion had to be As can be seen from the figure the test has enabled
developed which could differentiate between the various differences in the ageing performance of the interface
forms of aluminide and MCrAlY EB-PVD TBC systems. region to be studied, even at adhesive strengths far in
This requirement led to the development of the excess of those measurable by the pull-off test (typically
transverse scratch adhesion test [Ref. 15]. equivalent to 5N load on the scratch test). The modelled
adhesive strength for the as deposited coating systems
was 400 MPa (Figure 20).
 20-7

Platinum- EB-PVD VPS

aluminide Deposited Deposited
(datum) MCrAIY MCrAlY

Figure 22 Relative bond strengths of PVD TBC systems on

MarM002 substrates (datum = platinum -aluminide bondcoat)

BONDCOAT DEVELOPMENT

The first bondcoat systems for EB-PVD TBC's were

based on the environmental protection coatings that were
in use, for example MCrAIY coatings produced by EB-
PVD. During the early development phase of bondcoat
technology, it became apparent that the substrate played
an important role in the adhesion of the ceramic,
especially with regard to the diffusion of alloying
elements from the substrate through the bondcoat to the
near interface region with the oxide [Ref. 18]. This led to
a development programme where the rules for the design
of a bondcoat were formulated from a detailed
understanding of the mechanisms of interface
Figure 21 Transport of Hf through grain boundaries in the degradation and ceramic spallation. For example, in the
platinum-aluminide (top) to form Hf02 pegs (bottom) at the
interface with the Al203 thermally grown oxide scale.
case of a platinum-aluminide bondcoat applied to
MarM002 the degradation mode was identified as:

• Short cicuit diffusion of matrix elements along prior

Thermal Conductivity ß-phase ((NiPt)Al) grain boundaries forming hafnia
pegs.
Where as the EB-PVD TBC system offers significant • With continued exposure growth of these hafnia pegs
benefits over plasma sprayed TBC's for aerofoil occurs, these being surrounded by a shell of a-
application in the areas of compliance, surface finish, alumina, and finally,
erosion and adhesion, the thermal conductivity of EB- • Void formation and coalesence which causes a plane
PVD TBC's are higher than their plasma sprayed of weakness at the thermally grown oxide (TGO)
counterparts. A typical plasma sprayed TBC will have a bondcoat interface which reults in ceramic spallation
thermal conductivity of 0.7-0.9 W/m.K [Ref. 4] whilst a (see Figure 21) [Ref. 18].
0.25 mm thick PVD coating exhibits a thermal
conductivity of 1.6 W/m.K [Ref. 16]. Again, these The progressive reduction in the adhesive strength of the
property differences are related to the microstructure of platinum-aluminide EB-PVD TBC systems being
the systems, in the case of the plasma sprayed TBC the previously shown in Figure 19.
platelet boundaries (air gaps) offer a significant resistance
to heat flow whilst the EB-PVD TBC column boundaries Further research showed that even within the MCrAIY
lay parallel to the principal heat flux direction and have family of bondcoats of the same composition, significant
little effect on the conductivity [Ref 17]. Further more, varistion in the adhesive strengths of the ceramic could
due to the structure of the EB-PVD coating, the thermal result from differences in the processing techniques
conductivity varies as a function of the thickness of the (Figure 22) [Ref. 19]. The differences in the adhesion
coating. This variation is related to the high thermal behaviour of the EB-PVD and vacuum plasma sprayed
resistance of the coating near the interface region where (VPS) MCrAIY bondcoats are due to the role of yttrium
the coating has a high density of column boundaries due and the cleanliness of the bondcoat manufacturing
to the nucleation and competitive growth nature of the processes. In the case of the EB-PVD bondcoat, the
PVD process [Ref. 17] yttrium is available as Y in an uncombined form (yttrium
nickel intermetallic) which is mobile and can be taken up
20-8

combustion can and nozzle guide vane platforms, plasma

■ AsrKcFvad
sprayed TBC performs adequately. These applications for
a
c
1.6 ■ 25 hr* ©1150*0
D25hriei170,C
plasma sprayed TBC's will continue and because of their
£ □ 25hrsei190'C relatively low cost compared to EB-PVD TBC's, their
•> 1.2
application to these types of components in the future
0.8
seems to be assured.
0.4

0 L^E. 1 Where EB-PVD technology undoubtedly wins over

Platinum- plasma sprayed TBC's is on the aerofoils of the modern
atuminide
aero-gas turbine, where a good surface finish is
paramount and excellent strain compliance and adhesion
Figure 23 Relative bond strengths of PVD TBC systems on are required. In these areas EB-PVD offers significant
CMSX-4 substrates (datum = platinum -aluminide bondcoat on
MarM002)
benefits over plasma spraying which more than justifies
the higher manufacturing costs.

by the growing oxide. In the case of the VPS coating, the

yttrium is present as Y203 and although dispersed
throughout the coating, cannot migrate and is not REFERENCES
available to be incorporated into the coating or to tie-up
tramp elements coming through from the substrate. 1. Driver D., Hall D.W. and Meetham G.W., in "The
development of the gas turbine engine", Applied
When single crystal substrates were evaluated the cyclic Science Publishers (London), 1981.
lives were lower than the same coating systems on 2. Morrell P. and Taylor R., "Thermal diffusivity of
MarM002 and showed lower adhesion values (Figure thermal barrier coatings of Zr02 with Y203" High
23) [Ref. 19]. One important contributing factor to these Temperatures-High Pressures, F7, .(1985) 79
differences in TBC adhesion relates to the lower carbon 3. Brandt R. "Thermal diffusivity measurements on
levels in the single crystal alloys. Carbon, present in plasma sprayed calcia stabilised zirconia", High-
directionally solidified and conventionally cast alloys as a Temps. High-Press., .13 (1979).
grain boundary strengthener, combines with alloying 4. Morrell P. and Taylor R. "Thermal diffusivity and
elements like titanium and tantalum to form grain phase structure of zirconia-based thermal barrier
boundary precipitates. With the low levels of carbon in coatings", in "Advances in Ceramics" Volume 24:
single crystal alloys there is little to combine with the Science and Technology of Zirconia III, (1988), The
titanium and other alloying additions to stop them American Ceramics Society.
diffusing through to the interface with the thermally 5. Brandon J.R., Taylor R. and Morrell P.
grown oxide. Such elements disrupt the growth of cc- "Microstructure, composition property relationships
alumina and markedly reduce TBC adherence. in plasma-sprayed thermal barrier coatings", Surface
and Coatings Technology, 50 (1992) 141.
To counteract this problem with single crystal alloys a 6. Moody L.F., Trans. ASME, 77, 1265 (1943).
bondcoat system was developed to stop the diffusion of 7. Watt R.M., Allen J.L., Bains, N.C. Simons J.P.,and
these damaging elements from the substrate [Ref. 20]. George M., "A study of the effects of TBC surface
This comprises of an MCrAlY bondcoat with a thin layer roughness on the boundary layer characteristics of gas
of platinum diffused into it prior to the deposition of the turbine aerofoils", ASME87-GT-202 (1987).
ceramic layer. This layer is effective in blocking the 8. Ferreira G., Coney M.H. and Harrogate I.,
diffusion of the damaging elements from the substrate "Observations regarding the influence of surface
and as a result has increased not only the adhesion but finish on the performance of a TBC prepared by
also has improved significantly the temperature plasma arc", presented at ISABE Conference,
capability of the system (Figures 22 and 23). Nottingham 1990.
9. Sheffler K.D., Nissley D.M. and Cruise T.A.,
"Thermal barrier coating life prediction model
CONCLUDING REMARKS development", Trans. ASME, H4, 258, (1992).
lO.Hillery R.V., Pilsner B.H.,.McKnight RL, Cook
This paper has reviewed the advantages and T.S.and Hartle M.S.,"Thermal barrier coating life
disadvantages of both plasma sprayed and EB-PVD TBC prediction model development - Final Report", NASA
systems and their applications in the modern gas turbine CR 180807, (1988).
engine. Although in terms of erosion resistance and strain 11. Morrell P. and Nicholls J.R, ""The development of
compliance the EB-PVD TBC systems has significant environmentally tolerant PVD thermal barrier
advantages over plasma sprayed TBC's, in many coatings - Final Synthesis Report", BRITE/EuRAM
applications in the gas turbine engine, such as the programme BREU-0141, Commission of the
 20-9

European Communities (1995).

12.Toriz F.C., Thakker A.B. and Gupta S.K., "Thermal
barrier coatings for jet engines", Surface and Coatings
Technology, 39/40 (1989) 161.
13.Deakin M.J., Nicholls J.R and Rickerby D.S.,
"Erosion of thermal barrier coatings", In High-
Temperature Surface Engineering Conference, 23-25
September Edinburgh (1997).
14. Chalker P.R., Bull S.J and Rickerby D.S., Materials
Science and Engineering A140 (1991) 583.
15. Johnstone A., Rickerby D.S. and Morrell P., "Scratch
adhesion testing of TBC systems", In High-
Temperature Surface Engineering Conference, 23-25
September Edinburgh (1997).
16.Lawson K.L, Nicholls J.R and Rickerby D.S.,
"Thermal conductivity and ceramic microstructure",
In High-Temperature Surface Engineering
Conference, 23-25 September Edinburgh (1997).
17. Nichols J.R., Rickerby D.S. and Lawson K.L,
"Advanced processing for reduced thermal
conductivity", Paper 6 at 85th Meeting of Structures
and Materials Panel, Aalborg, Denmark 15/16
October 1997.
18. Tawancy H.M., Sridhar N., Abbas N.M. and Rickerby
D.S., "Comparative thermal stability characteristics
and isothermal oxidation behaviour of an aluminised
and platinum-aluminised Ni-based superalloy",
Scripta Met. et Materiala., 33,1431 (1995).
19. Bell S.R Wing R.G. and Woods J.L., "Electron beam
physical vapour deposition (EBPVD) thermal barrier
coatings for industrial gas turbine engines", Presented
at AMSE Turbo '96 Congress, Birminham (1996).
20. "Thermal barrier coatings for a super alloy -article
and method of application", European Patent
EP0718419A2,(1995).
 D-l

RESEARCH and TECHNOLOGY ORGANIZATION

th
85 MEETING of the STUCTURES and MATERIALS PANEL
WORKSHOP 3 « THERMAL BARRIER COATINGS »
AALBORG (DENMARK) - 15-16 OCTOBER 1997

REPORT on the FINAL DISCUSSION

Thursday, October 16th 1997
15hl0 - 16h25

Chairman : Alain LASALMONIE (SNECMA)

Recorder : Stefan DRAWIN (ONERA)

Table of contents

1. « THIN » TBC (AIRFOIL APPLICATIONS)

1.1 THERMAL CONDUCTIVITY

1.1.1 Effect of microstructure

1.1.2 Effect of composition
1.1.3 Modelling
1.1.4 Measurements
1.2 PROCESSING OF TBCs
1.2.1 Plasma spraying (PS)
1.2.2 EBPVD
1.2.3 Layered structures
1.2.4 Plasma assisted CVD
1.3 LIFETIME OF TBCs
1.3.1 Role of bond coat, TGO and oxidation
1.3.2 Testing methodology
1.3.3 Modelling
1.3.4 Graded structures
1.4 OTHER PROPERTIES
1.4.1 Erosion
1.4.2 TBC repair

2. « THICK » TBC (COMBUSTOR APPLICATIONS)

2.1 TBC ON FREE-STANDING COMPONENTS

2.2 TBC FOR COMBUSTORS
D-2

A. Lasalmonie (SNECMA) opened the round table by 1.1.3 Modelling

rapidly summarizing the themes presented throughout the
A. Lasalmonie (SNECMA) - Modelling is important for
workshop (see table of contents above).
optimizing the structure and the properties of TBCs. The
trends need to be quantified and experimental validation
are required.
1. « THIN » TBC (AIRFOIL APPLICATIONS) To study by numerical simulation the influence of the
morphology on thermal conductivity to produce
1.1 Thermal conductivity guidelines for coating manufacturers is an interesting
idea (see paper N°13). Can existing image analysis
1.1.1 Effect of micrestructure techniques, such as those developed e.g. for grain
A. Lasalmonie (SNECMA) - Effects of grain size and boundary studies on metallic alloys, be applied to TBCs?
shape, (macro-)defects (pores and cracks) and layering R. Mevrel (ONERA) - The approaches seem to be
were presented in papers N°l, 6, 8, 9, 14, 20. The needs different.
are more quantitative measurements and studies on the
stability of the microstructure (vs. time and temperature). 1.1.4 Measurements
U. Schulz (DLR) - The quantitative values needed are A. Lasalmonie (SNECMA) - The thermal conductivity
values for thermal conductivity after aging, not on as- measurement techniques on TBCs, especially at high
deposited TBCs. temperatures, should become more reliable and could be
W. Brindley (NASA) - Yes, but informations on initial assessed by a Round Robin on standard specimens.
(as-deposited) and final microstructure, as well as on P. Morrell (ROLLS-ROYCE) - Such a Round Robin was
intermediate states, are needed too, to get an idea of performed five years ago in the United Kingdom.
the important parameters that govern microstructure
evolution during aging. Y. Jaslier (SNECMA) - SNECMA already performed an
intercomparison campaign (two laboratories in
1.1.2 Effect of composition France, one in the United-Kingdom, one in the
United States).
A. Lasalmonie (SNECMA) - Is 8YSZ (8 wt%-yttria
partially stabilized zirconia), the composition that is the W. Brindley (NASA) - The contributions of the radiative
most studied and the most used industrially, really the and conductive components of the thermal
most appropriate composition, since higher yttria conductivity should be assessed.
contents seem to exhibit lower thermal conductivities? Zimmermann (?XXX) - Experiences have been
The role of dopant and the use of other oxides has also to performed on uncoated polished substrates and
be studied (needs for basic research). 100 urn zirconia-coated substrates and showed that
P.Morrell (ROLLS-ROYCE) - Yes, as shown by Y. radiation should not be neglected.
Jaslier in paper N°l, and in other studies, thermal
conductivity decreases with increasing yttria content, 1.2 Processing of TBCs
but if the yttria content is higher than ca. 20 wt%,
ordering phenomena appear and thermal conductivity 1.2.1 Plasma spraying (PS)
increases.
A. Lasalmonie (SNECMA) - The presented communi-
The 8YSZ composition has been developed and cations (papers N°3 and 4) have shown good
optimized by Chromalloy by durability tests on improvements in the process control, a better
plasma-sprayed coatings. This composition has, from understanding of the influence of deposition parameters
then on, be «conservatively» adopted by the and modelling of the coating formation.
aeroengine industry.
Can better structures and properties be achieved by
W. Brindley (NASA) - Temescal 20YSZ EBPVD including columnar inner structure to PS splats, as those
coatings exhibited nasty erosion problems. shown by P. Fauchais (paper N°3)?
A. Lasalmonie (SNECMA) It is time now to investigate Y. Jaslier (SNECMA) - The columns shown by
other compositions. P. Fauchais are generated by a solidification front
starting from the cold substrate and proceeding
V.R. Parameswaran (NRC) - Magnesia stabilized
towards the liquid part of the splat, whereas for the
zirconia has been tested, but gave bad results. Using
columnar structure in EBPVD coatings you have
ceria (at high content : 25%), good results were solidification (condensation) from the gas phase to
obtained up to 1600°C, but ceria was- (partially)
the solid substrate : the properties may not be the
reduced to Ce203 by chromium and aluminum from same.
the substrate.
XXX (?) - A recent study by Boeing showed that LPPS
P. Costa (ONERA) - If ceria and other rare earth oxides
TBCs exhibit a fivefold longer lifetime than APS
are to be studied, the oxides of the elements at the
coatings.
end of the series should give the best thermodynamic
stability. W. Brindley (NASA) - It has to be noticed that a great
variability exists in the properties of PS TBCs from
 D-3

one coating manufacturer to the other. So, great care 1.2.4 Plasma assisted CVD
should be taken when comparisons are made! A. Lasalmonie (SNECMA) - This new deposition
There is place for both processes (PS and EBPVD) in technique seems to present good potentials ; its
aeroengines and other parameters than durability application to coat real parts has not still to be
have to be taken into account. demonstrated and the costs have to be evaluated.
XXX (??) - Have other studies on reproducibility of PS V.R. Parameswaran (NRC) - The use of chlorides (ZrCl4)
coatings been performed (c/paper N°4)? may lead to environmental issues.
A. Lasalmonie (SNECMA) - Equipments are now XXX (DLR) - The ability to coat compex-shaped parts
availavable and installed, for a good control of the remains questionable.
reproducibility.
S. Drawin (ONERA) - The problems with chlorides can
C. Moreau (NRCC) - A good control for a better be solved even for industrial plants ; other zirconium
reproducibility can be achieved by controlling the precursors are currently studied. A larger deposition
particle velocity and temperature and the substrate reactor will be built next year, that will be able to
temperature. The latter depends of the type of thermal coat larger parts, as well coupons as blade and vanes.
contact between splats.
J.R. Nicholls (CRANFIELD Univ.) - The use of a plasma
XXX (??) - How much variability can be tolerated? will make it possible to deposit layered systems,
similarly to what has been achieved in our laboratory.
C. Moreau (NRCC) - For variations of 100°C for the
substrate temperature, different coating micro-
structures were observed. 1.3 Lifetime of TBCs

P. Morrell (ROLLS-ROYCE) - From the point of view of 1.3.1 Role of bond coat, TGO and oxidation
the manufacturer, the shape of the component to be
coated is of prior importance. The now available A. Lasalmonie (SNECMA) - Differences in the coating
processes are a compromise of what the engineers properties are observed when different bond coats are
want and what they can get. used : how can they be explained? The role of the bond
coat surface preparation (roughness) has to be assessed.
1.2.2 EBPVD Various failure modes are encountered in or at the TGO
(Thermal Grown Oxide): the failure mechanisms have to
A. Lasalmonie (SNECMA) - Different coating structures be understood more precisely.
and properties can be obtained by EBPVD. If we refer to
the presentation by G. Marijnissen (INTERTURBINE, M.R. Winstone (DERA) - Are the failure mechanisms the
paper N°10), this is a process that has to be further same for TBCs having experienced 1000 one-minute
controlled : the influence of a lot of parameters has to be cycles and 1000 one-hour cycles? TBC
studied. manufacturers have large databases on the possible
failure modes...
V.R. Parameswaran (NRC) - Grading should have a good
influence on the properties of the EBPVD coatings. W. Brindley (NASA) - Yes, TBCs cycled at maximum
temperature for six or sixty minutes present different
failure modes. It is thus important to design the right
1.2.3 Layered structures
durability tests : maximum temperature, coolant
A. Lasalmonie (SNECMA) - Interesting results have temperature, time at maximum temperature,...
been presented (cf. paper N°6), but no thermal stability
and lifetime values are available for these TBCs. For highly oxidized TBCs, failure occurs at locations
where the experienced temperatures were the highest.
Furthermore, are these structures industrially feasable?
J.R. Nicholls (CRANFIELD Univ.) - These are still too About the surface preparation, the presented paper
new systems : they have only been tested at 1100°C (N°17) adressed PS coatings with a simplified (sine-
wave) roughness. Can the conclusions be generalized
for 50 hours.
to all PS coatings?
A. Lasalmonie (SNECMA) - The layer thickness appears
to be a fundamental parameter. 1.3.2 Testing methodology
J.R. Nicholls (CRANFIELD Univ.) - Yes, the layers A. Lasalmonie (SNECMA) - As shown by paper N°15,
should have a thickness comprised between A/4 and X burner rig tests may be very difficult to interpret: there is
(k: wavelength of the incident radiation) to a need to establish a testing methodology for the
efficiently reduce the radiative contribution to the durability tests.
thermal conductivity. Note that the layers deposited
by K.S. Ravichandran (UTAH Univ., paper N°14) 1.3.3 Modelling
were either too thick (15-49 urn) or too thin (0.06 - A. Lasalmonie (SNECMA) - Thermomechanical
0.15 um) to induce a noticeable reduction of the modelling of the TBC system should allow a better
thermal conductivity. understanding of the failure mechanisms.
D-4

1.3.4 Graded structures 2. « THICK » TBC (COMBUSTOR

APPLICATIONS)
A. Lasalmonie (SNECMA) - The feasability of such
structures has been demonstrated, but extensive lifetime
2.1 TBC on free-standing components
tests have still to be performed.
XXX (??) - This is user-oriented for specific parts.

A. Lasalmonie (SNECMA) - Generally, what is missing 2.2 TBC for combustors

is the users experience on failure mechanisms, on
reliability and aging of the TBCs. A. Lasalmonie (SNECMA) - Thick'TBCs have to be
compared to other solutions (CMC, cooled metals,...).
Another issue is the influence of the substrate
(directionnally solidified alloys vs. single crystals, sulfur
and titanium content and diffusion,...).
P. Morrell (ROLLS-ROYCE) - Physical phenomena are
now understood and coatings have been specifically
designed by ROLLS-ROYCE.
J.R.Nicholls (CRANFIELD Univ.) - Different bond
coats are needed for different single crystals : in
general, now bond coats are designed for specific
substrates.
Y. Jaslier (SNECMA) - In the single crystal substrates,
the grain boundaries having been removed, sulfur
cannot segregate at these locations and has to go
somewhere else.
XXX (KLM) - Experience exists, but is not spread
worldwide (proprietary data!).

1.4 Other properties

1.4.1 Erosion
A. Lasalmonie (SNECMA) - Can the erosion behaviour
of PS coatings be improved?
C. Moreau (NRCC) - Yes, by using for instance powders
with smaller particle size ... but several problems will
then occur (feeding of the powder into the plasma
torch, moisture sensitivity, cost,...).
W. Brindley (NASA) - Another way to solve this
problem would be to find an erosion-resistant
material to spray on the leading edge (and having a
lower thermal conductivity than the superalloy
substrate).
C. Moreau (NRCC) - Here too, feedback from the users
is needed.
P. Morrell (ROLLS-ROYCE) - Currently, TBCs behave
well on combustor cans and vane platforms.
W. Brindley (NASA) - The problem is also to get the
right informations from the TBC coated parts : this
needs periodic inspections, inspections means, and
inspection procedures.
P. Morrell (ROLLS-ROYCE) - And how will the
metallic substrates be inspected if there is a TBC on
it.

1.4.2 TBC repair

A. Lasalmonie (SNECMA) - How is the repair of TBCs
mastered?
 REPORT DOCUMENTATION PAGE
1. Recipient's Reference 2. Originator's Reference 3. Further Reference 4. Security Classification
of Document
AGARD-R-823 ISBN 92-836-1073-3 UNCLASSIFIED/
UNLIMITED

5. Originator Advisory Group for Aerospace Research & Development

North Atlantic Treaty Organization
7 rue Ancelle, 92200 Neuilly-sur-Seine, France
6. Title
Thermal Barrier Coatings
7. Presented at/sponsored by
The 85th Meeting of the AGARD Structures and Materials Panel,
held in Aalborg, Denmark, 15-16 October 1997.

8. Author(s)/Editor(s) 9. Date
Multiple April 1998
10. Author's/Editor's Address 11. Pages
Multiple 188
12. Distribution Statement There are no restrictions on the distribution of this document.
Information about the availability of this and other RTO
unclassified publications is given on the back cover.
13. Keywords/Descriptors
Protective coatings Plasma spraying
Heat resistant materials Coating processes
Barrier coatings Vapor deposition
Combustion chambers Research projects
Turbine blades Electron beams
Ceramic coatings Thermal conductivity
Temperature gradients Bonded coatings
High temperature Reviews
Aircraft engines
14. Abstract
Thermal barrier coatings are an emerging technology which will enable either an increase in
turbine inlet temperatures, or a reduction in the duty temperature of the blades and as a result,
an extension of their life-cycle.
This Workshop presented the state of the art, gave a description of the existing or emerging
technologies, summarised our present knowledge of damage mechanisms, including
microstructural, mechanical and thermal aspects, and reported on the benefits and drawbacks of
the various families as perceived by engine manufacturers and users.
A final discussion attempted to identify future R&D requirements.
 NATO -0- OTAN
7 RUE ANCELLE • 92200 NEUILLY-SUR-SEINE DIFFUSION DES PUBLICATIONS
FRANCE AGARD NON CLASSIFIEES
Telecopie 0(1)55.61.22.99 • Telex 610 176
L'AGARD detient un stock limite de certaines de ses publications recentes. Celles-ci pourront eventuellement gtre obtenus sous forme de
copie papier. Pour de plus amples renseignements concernant l'achat de ces ouvrages, adressez-vous ä l'AGARD par lettre ou par t61ecopie
ä l'adresse indiquee ci-dessus. Veuillez ne pas UUphoner.
Des exemplaires supplementaires peuvent parfois etre obtenus aupres des centres de diffusion nationaux indiques ci-dessous. Si vous
souhaitez recevoir toutes les publications de l'AGARD, ou simplement celles qui concernent certains Panels, vous pouvez demander d'etre
inclus sur la liste d'envoi de l'un de ces centres.
Les publications de l'AGARD sont en vente aupres des agences de vente indiquees ci-dessous, sous forme de photocopie ou de microfiche.
Certains originaux peuvent egalement etre obtenus aupres de CASI.

CENTRES DE DIFFUSION NATIONAUX

ALLEMAGNE ISLANDE
Fachinformationszentrum Karlsruhe Director of Aviation
D-76344 Eggenstein-Leopoldshafen 2 c/o Flugrad
BELGIQUE Reykjavik
Coordonnateur AGARD - VSL ITALffi
Etat-major de la Force aerienne Aeronautica Militare
Quartier Reine Elisabeth Ufficio Stralcio AGARD
Rue d'Evere, B-1140 Bruxelles Aeroporto Pratica di Mare
CANADA 00040 Pomezia (Roma)
Directeur - Gestion de 1'information LUXEMBOURG
(Recherche et developpement) - DRDGI 3 Voir Belgique
Ministere de la Defense nationale NORVEGE
Ottawa, Ontario K1A 0K2 Norwegian Defence Research Establishment
DANEMARK Attn: Biblioteket
Danish Defence Research Establishment P.O. Box 25
Ryvangs Alle 1 N-2007 Kjeller
P.O. Box 2715 PAYS-BAS
DK-2100 Copenhagen 0 Netherlands Delegation to AGARD
ESPAGNE National Aerospace Laboratory NLR
INTA (AGARD Publications) P.O. Box 90502
Carretera de Torrejön a Ajalvir, Pk.4 1006 BM Amsterdam
28850 Torrejön de Ardoz - Madrid PORTUGAL
ETATS-UNIS Estado Maior da Forca Aerea
NASA Center for AeroSpace Information (CASI) SDFA - Centra de Documenta9äo
Parkway Center, 7121 Standard Drive Alfragide
Hanover, MD 21076 P-2720 Amadora
FRANCE ROYAUME-UNI
O.N.E.R.A. (Direction) Defence Research Information Centre
29, Avenue de la Division Leclerc Kentigern House
92322 Chätillon Cedex 65 Brown Street
Glasgow G2 8EX
GRECE TURQUIE
Hellenic Air Force
Air War College Mill! Savunma Baskanligi (MSB)
Scientific and Technical Library ARGE Dairesi Baskanligi (MSB)
Dekelia Air Force Base 06650 Bakanliklar - Ankara
Dekelia, Athens TGA 1010
AGENCES DE VENTE
NASA Center for AeroSpace Information (CASI) The British Library Document Supply Division
Parkway Center, 7121 Standard Drive Boston Spa, Wetherby
Hanover, MD 21076 West Yorkshire LS23 7BQ
Etats-Unis Royaume-Uni
Les demandes de microfiches ou de photocopies de documents AGARD (y compris les demandes faites aupres du CASI) doivent
comporter la denomination AGARD, ainsi que le numero de serie d'AGARD (par exemple AGARD-AG-315). Des informations
analogues, telles que le titre et la date de publication sont souhaitables. Veuiller noter qu'il y a lieu de specifier AGARD-R-nnn et
AGARD-AR-nnn lors de la commande des rapports AGARD et des rapports consultatifs AGARD respectivement. Des references
bibliographiques completes ainsi que des resumes des publications AGARD figurent dans les journaux suivants:
Scientific and Technical Aerospace Reports (STAR) Government Reports Announcements & Index (GRA&I)
STAR peut etre consulte en ligne au localisateur de publie par le National Technical Information Service
ressources uniformes (URL) suivant: Springfield
http://www.sti.nasa.gov/Pubs/star/Star.html Virginia 2216
STAR est edite par CASI dans le cadre du programme Etats-Unis
NASA d'information scientifique et technique (STI) (accessible egalement en mode interactif dans la base de
STI Program Office, MS 157A donnees bibliographiques en ligne du NTIS, et sur CD-ROM)
NASA Langley Research Center
Hampton, Virginia 23681-0001
Etats-Unis

m
Imprime par le Groupe Communication Canada Inc.
(membre de la Corporation St-Joseph)
45, boul. Sacre-Cceur, Hull (Quebec), Canada K1A 0S7
 jMBs^m
NATO -®- OTAN
7 RUE ANCELLE • 92200 NEUILLY-SUR-SEINE DISTRIBUTION OF UNCLASSIFIED

FRANCE AGARD PUBLICATIONS

Telefax 0(1)55.61.22.99 • Telex 610 176
AGARD holds limited quantities of some of its recent publications, and these may be available for purchase in hard copy form. For more
information, write or send a telefax to the address given above. Please do not telephone.
Further copies are sometimes available from the National Distribution Centres listed below. If you wish to receive all AGARD publications,
or just those relating to one or more specific AGARD Panels, they may be willing to include you (or your organisation) in their distribution.
AGARD publications may be purchased from the Sales Agencies listed below, in photocopy or microfiche form. Original copies of some
publications may be available from CASI.

NATIONAL DISTRD3UTION CENTRES

BELGIUM LUXEMBOURG
Coordonnateur AGARD - VSL See Belgium
Etat-major de la Force aerienne
Quartier Reine Elisabeth NETHERLANDS
Rue d'Evere, B-1140 Bruxelles Netherlands Delegation to AGARD
National Aerospace Laboratory, NLR
CANADA P.O. Box 90502
Director Research & Development 1006 BM Amsterdam
Information Management - DRDIM 3
Dept of National Defence NORWAY
Ottawa, Ontario K1A 0K2 Norwegian Defence Research Establishment
Attn: Biblioteket
DENMARK P.O. Box 25
Danish Defence Research Establishment N-2007 Kjeller
Ryvangs Alle 1
P.O. Box 2715 PORTUGAL
DK-2100 Copenhagen 0 Estado Maior da Forga Aerea
FRANCE SDFA - Centro de Documentacäo
O.N.E.R.A. (Direction) Alfragide
29 Avenue de la Division Leclerc P-2720 Amadora
92322 Chätillon Cedex SPAIN
GERMANY INTA (AGARD Publications)
Fachinformationszentrum Karlsruhe Carretera de Torrejön a Ajalvir, Pk.4
D-76344 Eggenstein-Leopoldshafen 2 28850 Torrejön de Ardoz - Madrid
GREECE TURKEY
Hellenic Air Force Mill! Savunma Baskanligi (MSB)
Air War College ARGE Dairesi Baskanligi (MSB)
Scientific and Technical Library 06650 Bakanliklar - Ankara
Dekelia Air Force Base
Dekelia, Athens TGA 1010 UNITED KINGDOM
ICELAND Defence Research Information Centre
Director of Aviation Kentigern House
c/o Flugrad 65 Brown Street
Reykjavik Glasgow G2 8EX
ITALY UNITED STATES
Aeronautica Militare NASA Center for AeroSpace Information (CASI)
Ufficio Stralcio AGARD Parkway Center, 7121 Standard Drive
Aeroporto Pratica di Mare Hanover, MD 21076
00040 Pomezia (Roma)
SALES AGENCIES
NASA Center for AeroSpace Information (CASI) The British Library Document Supply Centre
Parkway Center, 7121 Standard Drive Boston Spa, Wetherby
Hanover, MD 21076 West Yorkshire LS23 7BQ
United States United Kingdom
Requests for microfiches or photocopies of AGARD documents (including requests to CASI) should include the word 'AGARD'
and the AGARD serial number (for example AGARD-AG-315). Collateral information such as title and publication date is
desirable. Note that AGARD Reports and Advisory Reports should be specified as AGARD-R-nnn and AGARD-AR-nnn,
respectively. Full bibliographical references and abstracts of AGARD publications are given in the following journals:
Scientific and Technical Aerospace Reports (STAR) Government Reports Announcements & Index (GRA&I)
STAR is available on-line at the following uniform published by the National Technical Information Service
resource locator: Springfield
http://www.sti.nasa.gov/Pubs/star/Star.html Virginia 22161
STAR is published by CASI for the NASA Scientific United States
and Technical Information (STI) Program (also available online in the NTIS Bibliographic
STI Program Office, MS 157A Database or on CD-ROM)
NASA Langley Research Center
Hampton, Virginia 23681-0001
United States

Printed by Canada Communication Group Inc.

(A St. Joseph Corporation Company)
45 Sacri-Cceur Blvd., Hull (Quebec), Canada K1A 0S7

ISBN 92-836-1073-3