Surface & Coatings Technology 327 (2017) 126–138

Contents lists available at ScienceDirect

Surface & Coatings Technology

j ournal homepage : www . el sevier . com/l oc ate/ surfcoat

CMAS behavior of yttrium aluminum garnet (YAG) and yttria-stabilized

zirconia (YSZ) thermal barrier coatings
a, a a a b b a
Rishi Kumar , Eric Jordan , Maurice Gell , Jeffrey Roth , Chen Jiang , Jiwen Wang , Sarshad Rommel
a
Institute of Materials Science, 97 North Eagleville Road, University of Connecticut, Storrs, CT 06269-3136, USA
b
HiFunda LLC, 270 Middle Turnpike, Storrs, Mansfield, CT 06269, USA

article info abstract

Article history: Calcium magnesium aluminosilicate (CMAS) that is formed from the ingested deposits in gas turbines degrades thermal
Received 10 March 2017 barrier coatings (TBCs), especially for the most widely used material; yttria-stabilized zirconia (YSZ). In the present work, we
Revised 20 July 2017 examine the behavior of yttrium aluminum garnet (YAG) as an alternative material for TBCs. CMAS interaction studies were
Accepted in revised form 8 August 2017
conducted by making composite pellets of YAG-CMAS and YSZ-CMAS pow-ders. These pellets, after being subjected to heat
Available online 9 August 2017
treatment between 1100 °C and 1500 °C were characterized by X-ray Diffraction (XRD), scanning electron microscopy
Keywords:
(SEM) and Energy-dispersive X-ray spectroscopy (EDS), which showed YAG to be almost inert to CMAS whereas YSZ
Thermal barrier coatings exhibited significant phase changes. To test the be-havior of TBCs with YAG and 8YSZ as the topcoat material in a CMAS
Yttrium aluminum garnet environment, cyclic furnace tests were con-ducted in which a controlled amount of CMAS was applied and then the samples
CMAS resistance were cycled to failure. In addition, to simulate the continuous accumulation of CMAS expected in service, a cyclic furnace test
Optical basicity was devised in which a small dose of aqueous solution of CMAS was applied on TBC specimens at the start of every cycle
Solution precursor plasma spray until the samples were cycled to failure. In all these tests YAG TBCs outperformed YSZ in terms of durability. The mech-
anisms of CMAS attack are described and the relative resistance of YAG and YSZ is shown to be consistent with the Optical
Basicity (OB) theory.

© 2017 Elsevier B.V. All rights reserved.

1. Introduction monoclinic phase, accompanied by a detrimental volumetric expansion and

Thermal barrier coatings (TBCs) are widely employed for thermal
protection of gas turbine engines to improve their efficiency, durability and
performance in their high temperature operation [1–3]. Generally, a TBC
comprises four layers - a ceramic top coat, a thermally grown oxide (TGO), a
bond coat and finally a superalloy substrate. TBCs have proven to be a
critical component in hot sections of the gas turbines in order to increase the
turbine inlet temperature (TIT) and attain improved engine efficiency. This
has inevitably led to an increasing demand for yet higher TIT, and hence, for
TBCs that can effectively protect turbine components at these higher
temperatures. The state of art material used for TBCs is 6–8 wt% yttria
stabilized zirconia (YSZ) because of its favorable proper-ties such as high
thermal expansion coefficient, high fracture toughness [4] and low thermal
conductivity [1,5]. However, use of YSZ is limited to a maximum temperature
of 1200 °C because of a) transformation of metastable tetragonal phase into
tetragonal, and further into
Corresponding author.
E-mail addresses: rishi.kumar@uconn.edu (R. Kumar), eric.jordan@uconn.edu
(E. Jordan), maurice.gell@uconn.edu (M. Gell), jeffrey.roth@uconn.edu (J. Roth),
cjiang@hifundallc.com (C. Jiang), jwang@hifundallc.com (J. Wang),
sarshad.rommel@uconn.edu (S. Rommel).
spallation [6], b) high sintering rates of YSZ which causes coating
densification and loss of strain tolerance resulting in spallation [7] and c)
attacks by molten calcium magnesium-aluminosilicate in the form of dust,
sand and runway debris. This infiltrated CMAS degrades the du-rability and
thermal characteristics of the coating through thermomechanical [8,9] as
well as thermochemical effects [10,11]. Infil-trated CMAS causes high
internal stresses in YSZ TBC microstructures due to loss of strain tolerance
and a decrease in the effective coefficient of expansion, and consequent
thermal cycling eventually leads to spall-ation. Molten CMAS is also known
to partially dissolve YSZ coatings, while selectively removing yttrium ions
which destabilizes the metasta-ble tetragonal phase and results in phase
transformation to monoclinic of YSZ accompanied by a 2–5% volumetric
expansion causing spallation in the coating [12,13].
A plethora of alternate material systems have been proposed and ex-
amined as prospective TBC candidates including pyrochlores, fluorites,
garnets and zirconates [14–17]. One such material has been gadolinium
zirconate which has shown excellent properties such as high melting point
(N2000 °C) [18], high temperature phase stability (≤1550 °C) [19], lower
thermal conductivity than 8YSZ [5] and excellent CMAS re-sistance by
forming reaction products capable of blocking further CMAS infiltration in
the top coat of TBC [20]. Despite having these

http://dx.doi.org/10.1016/j.surfcoat.2017.08.023 0257-
8972/© 2017 Elsevier B.V. All rights reserved.
 R. Kumar et al. / Surface & Coatings Technology 327 (2017) 126–138
exceptional properties, gadolinium zirconate has some weaknesses as well.
Not only is it commercially expensive because of high stoichiomet-ric
amounts of gadolinium, but also it lacks in some other critical mate-rial
properties. These include poor thermodynamic compatibility with Al 2 O 3
from the TGO layer [21], worse erosion resistance than air plasma sprayed
(APS) 8YSZ coatings and lower fracture toughness than YSZ [22].
In our previous work [23,24], we demonstrated that yttrium alumi-num
garnet (YAG) TBCs deposited by Solution Precursor Plasma Spray (SPPS)
have superior properties compared to Air Plasma Sprayed (APS) 8YSZ. These
desirable properties include phase stability up to 1500 °C, low thermal
conductivity of 0.95 W/m·k at 1300 °C, improved erosion resistance, higher
thermal cycling lives and increased sintering resistance than APS YSZ [23].
In the current work, we discuss the reac-tivity of YAG with CMAS, while 8YSZ
is used as the reference material. In this paper, we have also compared the
performance of SPPS YAG TBCs with APS 8YSZ coatings that are the current
industrial standard. Ideally, APS YAG coating would provide a head on
comparison however, previous work show for APS YAG the formation of
amorphous YAG [25] phase occurs which will be prone to crystallization
during service and thus cracking. Also, an APS YAG is devoid of stress
relieving vertical cracks that are found in SPPS coatings thus will have a
much larger ther-mal expansion mismatch with the bottom TBC layers. A
head to head comparison with the APS industrial standard is not done here.
The aim of the study is to look at the application of a new TBC material
processed by a relatively young SPPS process and show that it is a candidate
to re-place the current industry standard APS YSZ in the presence of CMAS.
At the same time, we are looking for the first time at the nature of YAG in-
teraction with CMAS. We have also explored the theory of Optical Basic-ity
which has been recently used to predict the reactivity between the top coat
materials with CMAS [26,27]. This theory will be shown to be consistent with
the current results. In the following, the CMAS behavior of YAG TBCs will be
studied. In each case YSZ samples are tested in the same experiment to
provide a comparison basis using identical conditions.
2. Materials and methods
2.1. CMAS powder synthesis and characterization
Four and nine components CMAS (4-CMAS, 9-CMAS) powders were
synthesized using stoichiometric amounts of metal salts along with col-loidal
silica and a titanium sulfate solution. 9-CMAS composition was provided to
us by our industrial partner while 4-CMAS composition was adopted from the
study conducted by Drexler et al. [28]. Synthesis of 4-CMAS aqueous
solution was done by adding stoichiometric amounts of hydrated nitrates salts
of calcium, aluminum and magne-sium (Ca(NO 3 ) 2 ·4H 2 O, Al(NO 3 ) 3 ·9H 2 O,
Mg(NO 3 ) 2 ·6H 2 O, Alfa Aesar, Ward Hill, MA), and Silica (LUDOX® TMA
colloidal silica, Sigma-Aldrich Co. LLC. Spruce Street, MO) in de-ionized
(DI) water. For 9-CMAS, ni-trates of calcium, aluminum, magnesium, iron,
potassium and sodium were dissolved in DI water. L-Cystine was dissolved
separately in a
Table 1
Chemical compositions of 4 and 9 component CMAS used in paste and spritz testing.
Constituents 4-CMAS 4-CMAS 9-CMAS
(wt%) (mol%) (wt%)
CaSO 4 – – 42.00
SiO 2 52.30 51.5 31.57
Al 2 O 3 7.10 4.1 16.50
MgO 3.50 5.2 2.66
CaO 37.10 39.2 1.54
Fe 2 O 3 – – 1.54
K2O – – 1.54
TiO 2 – – 1.33
Na 2 O – – 1.32
127
pre-made solution, and was then mixed with the nitrate solution as a source
of sulfur. After this step titanium sulfate solution and colloidal sil-ica was
added. pH of the solution was constantly measured and nitric acid was used
to keep the pH between 4 and 7 to ensure stability of col-loidal silica as per
manufacturer's recommendation. Both CMAS solu-tions were dried overnight
at 120 °C to convert the solution to gels, which were later heat treated at 600
°C in a tube furnace (CM 1600 Tube Furnace, Bloomfield N.J.). The chemical
compositions of the two CMAS are given in Table 1. Melting points of the
two CMAS were deter-mined using simultaneous differential scanning
calorimetry (DSC) and thermal gravimetric analysis (TGA), (Q600 SDT, 10
°C/min, 100 ml/min N 2 flow, TA Instruments, New Castle DE). Viscosity of
the two CMAS was calculated by Giordano [29] and Fluegel [30] model
and has been shown in Table 2.
2.2. Powder pellet testing
YAG powder (particle size b 50 μm) was obtained by mechanical grinding
of YAG agglomerate (Alfa Aesar, Ward Hill, MA) and then siev-ing using a
50 μm wire mesh. YAG and 8YSZ (Metco 204 NS) powders were then mixed
with 9-CMAS powder (composition listed in Table 1) in a 1:1 ratio by weight.
5 wt% of PVA (low molecular weight, Alfa Aesar, Ward Hill, MA) was added
to the mixture to provide strength to the green body. The mixture was ball
milled, dried and compressed to form powder pellets. The pellets were heat
treated in a box furnace (CM Furnaces Inc., Bloomfield, N.J.) for 24 h at
different temperatures ranging from 1100 °C–1500 °C. Then the pellets were
analyzed for reac-tion using X-ray Diffractometer (D2 Phaser, Bruker AXS,
Madison, WI). For SEM, a small piece of the heat-treated pellets was
embedded in epoxy resin, polished and finally sputtered with Au/Pd coatings.
Field emission scanning electron microscopes (JSM-6350/5F, JEOL USA,
Pea-body, MA) and energy dispersive X-ray spectrometer (Noran system six
EDS, Thermo, Waltham, MA) were used to further analyze the reac-tions
between ceramics and 9-CMAS.
2.3. Thermal cycling
Thermal cycling experiments were simultaneously conducted on SPPS
YAG and APS 8YSZ baseline samples alongside CMAS paste test that is
described in Section 2.4. The temperature profile is shown in Fig. 1. These
tests will help evaluate the performance of both the coat-ings with and
without CMAS in a test related to expected turbine engine performance. For
the thermal cycling experiment and also the forth-coming CMAS experiments,
three samples each of SPPS YAG and APS YSZ baselines were tested. This
was done to account for the dispersion in the cyclic lives of the samples which
is generally observed in furnace testing of TBCs.
Only SPPS YAG coatings (~225 μm thickness) were deposited at our spray
booth on superalloy substrates provided to us by one of our indus-trial
partners which already had a 100 μm MCrAlY bond coat and ~25 μm APS
YSZ as the inner layer. The spray conditions have been adopted from previous
study [24] and are presented in Table 3. The APS 8YSZ
9-CMAS 9-CMAS after sulfate decomposition 9-CMAS after sulfate decomposition
(mol%) (wt%) (mol%)
26.8 0.0 0.0
45.6 41.9 45.6
14.0 21.9 14.0
5.7 3.5 5.7
2.4 25.0 29.1
0.8 2.0 0.8
1.4 2.0 1.4
1.4 1.8 1.4
1.8 1.8 1.8
128 R. Kumar et al. / Surface & Coatings Technology 327 (2017) 126–138
Table 2
CMAS viscosity calculated using Giordano and Fluegel model.
Viscosity (Pa ∗ s) 4-CMAS 9-CMAS
Temperature (°C) Giordano Fluegel Giordano
1180 376.6 33.2 11.9
1200 252.9 21.5 8.4
baseline samples were provided to us by the same industrial partner and were
tested in ‘as received’ condition. These samples had a 250 μm top coat of APS
8YSZ and same bondcoat and substrate as the SPPS YAG, allowing a fair
comparison of coatings. For all the tests half of a sample was used which
initially had a diameter of 25.4 mm.
2.4. CMAS paste test
To evaluate the performance of TBCs, burner rig tests are widely
employed [31] to create a temperature gradient, thereby simulating ser-vice
conditions. In the current study, a simpler test was conceived to test the
interaction of different CMAS compositions ( Table 1) with SPPS YAG and
APS YSZ baseline samples described in Section 2.3. CMAS paste with a
2
concentration of 10 mg/cm was applied on the TBCs that were moist-ened
with DI water to aid paste homogeneity and spreading. This partic-ular
CMAS concentration was chosen based on the reported minimum level
needed to initiate TBC damage by Wellman et al. [32]. These sam-ples were
isothermally cycled in a bottom-loading isothermal furnace (CM Furnaces
Inc., Bloomfield, NJ) to 1180 °C (above the CMAS melting temperature,
Section 3.1) using 1-hour cycles (Refer to Fig. 1), with 10 min of ramp up
time, 40 min of dwell time and 10 min of expedited cooling time using an
external fan, until their failure. Failure was de-fined by 50% of coating
delamination from the substrate. Once the coat-ings failed, XRD was
conducted on the surface of the coatings to study the reaction between the
CMAS and TBCs; SEM and EDS was performed on the coating cross-sections
to investigate the infiltration of CMAS in the TBCs.
2.5. CMAS spritz test
Another CMAS test was carried out where 0.1 ml of 1 wt% aqueous
solutions of CMAS precursor chemicals (4 and 9 Component) were sprayed
on the TBC samples before the start of every cycle. The temper-ature profile
of the experiment was kept similar to the paste test and the samples were
cycled to failure. SEM and EDS were used to study the coating cross sections
and investigate the infiltration of CMAS into the TBCs. Synthesis of the two
CMAS solutions was similar to as described in Section 2.1 while keeping the
equivalent solid loading to 1 wt% in DI water.
Fig. 1. Temperature profile of elevator furnace during CMAS paste and spritz testing.
Table 3
Spraying parameters for SPPS YAG samples.
Parameters
Fluegel Plasma gun Metco 9 MB, GP gun nozzle
Gun power 45.5 kW
4.4
Primary gas/secondary Ar/H 2
1.9
gas
Gas flow rate Ar: 52–57 l/min, H 2 :8–10 l/min
Precursor injection mode Atomization using BETE FC4 spray nozzles, 15 psi
pressure
Precursor feed rate 18–20 ml/min
Standoff distances 35–37 mm
Gun scan speed 550 mm/s
Raster scan step size 2 mm
3. Results
3.1. CMAS powder characterization
Fig. 2 shows the simultaneous DSC and TGA results of the two CMAS
compositions. The 4-CMAS exhibited a complete melting at 1203 °C but the
melting started much earlier from around ~1165 °C. 9-CMAS melted
completely at 1174 °C·in order to run tests with both CMAS composi-tions
simultaneously, a temperature of 1180 °C was chosen for both paste and
spritz testing. We note that a weight change in both 4 and 9-CMAS was
observed. In case of 9-CMAS this can be attributed to the loss of sulfur
however a weight change in 4-CMAS was not expected. It must be the case
that volatile species (nitrates) survived the heat treatment during synthesis of
4-CMAS. Table 1 shows the expected composition of 9-CMAS assuming
sulfur volatilizes at and above 1100 °C.
Temperature-dependent CMAS viscosity is an important parameter that
affects infiltration of coatings thus an attempt has been made to cal-culate
CMAS viscosity using existing models and has been shown in Table 2.
Previous study by Wiesner et. Al [33]. shows that Fluegel model predicted
CMAS viscosity which was in reasonable agreement with the experimental
data, however it has compositional restrictions which may provide inaccurate
results for the 4-CMAS. Thus Giordano model was used for the 4 component
CMAS despite the fact that it pre-dicted more than an order of magnitude
higher than the experimental values for a CMAS like composition. Because of
this we also used the Fluegel model to calculate CMAS viscosity. The
temperature for the vis-cosity calculation (1180 °C) was chosen to match the
temperature of CMAS-TBC interaction in paste and spritz tests. Both the
models predict that 4-CMAS has a higher viscosity than 9-CMAS. The
viscosity for 4-CMAS was higher by a factor of 31 and 8 as predicted by
Giordano [29] and Fluegel [30] models respectively.
3.2. Powder pellet testing
Pellet testing was done using 9 component CMAS only. The cylindri-cal
shaped pellets after a 24-h exposure changed their shape depending upon the
temperature and the material. This change, captured in Fig. 3 was
exacerbated in the case of 8YSZ + CMAS pellets which melted completely
whereas the YAG + CMAS pellet only became rounded. Such a macroscopic
behavior indicates reaction between CMAS and 8YSZ which led to dissolution
and re-precipitation of 8YSZ particles thus forming a ‘puddle’ of reaction
products. The minimal change in the shape of YAG based pellets establish the
lack of interaction between YAG and CMAS and the deformation resulted
from preferential or local-ized melting of CMAS in the pellets but held intact
due to YAG matrix. This theory can be further confirmed by Fig. 4 which
shows the SEM im-ages and EDS of the reaction zone in both 8YSZ and YAG
during the heat treatment at 1500 °C. While edge dissolution of ceramic
particles can be observed in both cases, the YAG particles are still rounded
and discrete. 8YSZ particles have changed their shape from rounded to
irregular,
 R. Kumar et al. / Surface & Coatings Technology 327 (2017) 126–138 129

Fig. 2. Simultaneous DSC and TGA data of a) 4-CMAS and b) 9-CMAS.

primarily due to dissolution in CMAS and re-precipitation. Some CMAS is
also trapped within the particles in the process. X-ray diffraction re-sults for
both YAG and YSZ based pellets are shown side by side for com-parison in
Fig. 5. YSZ based pellets showed primarily tetragonal phase at 1100 °C and
1200 °C with small peaks of monoclinic phase.
Interestingly, at 1100 °C–1300 °C some anorthite (CaAl 2 Si 2 O 8 ) phase can
also be seen with maximum intensity at 1300 °C. At higher temper-atures,
anorthite phase is absent and the same anorthite observation holds true in
YAG + CMAS pellets. With an increase in temperature for YSZ, the
tetragonal to monoclinic phase transformation increases, as indicated by an
increase in monoclinic peak intensity. At 1400 °C and 1500 °C most of the
tetragonal phase has converted to a combina-tion of cubic and monoclinic.
YAG based pellets show almost no reactiv-ity with CMAS at all temperatures.
Crystallization of anorthite is observed between 1100 °C–1300 °C which
disappears at higher temper-atures. At 1400 °C peaks of Ca 4 Y 6 O(SiO 4 ) 6 is
observed which does indi-cate some reactions between CMAS and YAG but
the phase seems to disappear at 1500 °C.

3.3. Microstructure of SPPS YAG and APS 8YSZ baseline samples

Fig. 3. Pictures of powder pellets (a) Before heat treatment (b) YAG + 9-CMAS (1:1 by wt)
after heat treatment for 24 h at 1500 °C, and (c) 8YSZ + 9-CMAS (1:1 by wt) after heat The microstructure of “as received” APS 8YSZ sample is shown in Fig. 6a
treatment for 24 h at 1500 °C. which shows three different layers of the coatings. The top
130 R. Kumar et al. / Surface & Coatings Technology 327 (2017) 126–138

Fig. 4. SEM images of mixed powder pellets post sintering at 1500 °C for 24 h with chemical analysis data acquired at the ceramic-CMAS interface. (a) 8YSZ + 9-CMAS (1:1 by wt), (b) EDS
spectrum at 8YSZ − 9-CMAS interface, (c) YAG + 9-CMAS (1:1 by wt) and (d) EDS spectrum at YAG − 9-CMAS interface.

layer, deposited via APS process, is of 8YSZ and has a thickness of ~250 μm, precursor is pyrolyzed on the sample which causes shrinkage and for-mation
of horizontal pores [24] while the over spray which results in isolated splats
followed by a 100 μm layer of bondcoat and finally the su-peralloy substrate.
adds additional porosity in that layer.
Fig. 6b shows the magnified image of APS 8YSZ coating which has a typical
microstructure obtained through APS process with uniformly distributed splat
boundaries and porosity (~15%, calculated using ImageJ software). 3.4. Thermal cycling without CMAS and failure modes

Fig. 7a is a cross section SEM image of “in house” deposited SPPS YAG
with a thickness of ~225 μm on superalloy substrates pre-coated with 25 μm of
APS 8YSZ and a 100 μm bondcoat layer. The SPPS YAG coating has
approximately uniform vertical cracks which extend through the entire
coating thickness and offers stress relieving during thermal cy-cling
experiments. Fig. 7b shows a magnified SEM image of YAG coat-ings which
depicts the aforementioned vertical cracks and also reveals horizontal pores
that form between the subsequent gun passes by trap-ping of semi-pyrolyzed
precursor an overspray. The under pyrolyzed
Fig. 8 shows the average thermal cycling lives of SPPS YAG and APS
8YSZ baseline samples with standard deviation indicated through bars. SPPS
YAG samples lasted 22% longer than the baseline 8YSZ sam-ples. To analyze
the failure modes, BSE images of the failed coatings were taken. Fig. 9a
shows the failed baseline sample with spallation at YSZ-TGO interface. Fig.
9b shows a magnified image of the failure inter-face with elemental mapping.
The failure is a mixture of failure in the top coat and through the TGO. A
distinct TGO layer (10–15 μm) can be observed. The relatively large TGO
thickness enables a failure

Fig. 5. X-ray diffraction of mixed powder pellets after heat treatments at temperatures from 1100 °C to 1500 °C. (a) 8YSZ + 9-CMAS (1:1 by wt) (b) YAG + 9-CMAS (1:1 by wt)
 R. Kumar et al. / Surface & Coatings Technology 327 (2017) 126–138
Fig. 6. (a) SEM image of as received APS 8YSZ TBC showing topcoat, bondcoat and substrate
layers. (b) Magnified image of the 8YSZ topcoat showing typical microstructure generated by
an APS process with horizontal splats and pores.
mechanism than includes shape change imposed on the ceramic associ-ated
with TGO growth [7,34].
In the case of SPPS YAG the failure occurs in YSZ interlayer-TGO in-
terface and failure at YAG-YSZ interface is also observed as shown in Fig.
10a. A magnified image of the failed region is shown in Fig. 10b with the
elemental analysis. Failure at the YSZ TGO interface probably has a
contribution to stress in the coating due to shape change due to
Fig. 7. SEM image of SPPS YAG TBC showing “in house” deposited YAG topcoat with vertical
cracks, a thin APS 8YSZ inner layer, bondcoat and substrate layers. (b) Magnified image of the
SPPS YAG topcoat showing typical microstructure generated by SPPS process with vertical
cracks for strain tolerance and horizontal pores called inter pass boundaries.
131
Fig. 8. Thermal cyclic performance of SPPS YAG and APS 8YSZ TBCs.
TGO growth while separation at YSZ-YAG interface is most likely due to
differential sintering [24].
3.5. CMAS paste test
CMAS paste tests were done with both 4 component and 9 compo-nent
CMAS and the general failure modes were the same for the two CMAs
compositions however the chemical interactions were different. The lives of
both YAG and 8YSZ coatings decreased drastically with CMAS application
but the reduction in life of 8YSZ sample was more dramatic. 8 YSZ samples
showed a 147 times reduction due to CMAS ex-posure where SPPS YAG
samples showed ~20 times reduction. Fig. 11 shows the cyclic lives of the two
coatings. SPPS YAG coatings performed much better than APS 8YSZ
coatings, with a cycling life greater by at least a factor of 8. The mode of
failure was quite different in the two coatings. APS 8YSZ coatings failed in
the form of flakes, where layer after layer of the coating came off until no
coating remained on the sub-strate. On the other hand, SPPS YAG coatings
failed in one piece at the SPPS YAG to APS YSZ interface (shown in Fig. 12).
On further examina-tion, as shown in Fig. 13, it appears that delamination of
flakes happens in relatively larger horizontal pores in the APS 8YSZ coatings.
This is
Fig. 9. (a) Cross section of a failed APS 8YSZ coating in thermal cycling test showing
spallation at YSZ-TGO interface. (b) Magnified cross section of the spallation region with
elemental mapping showing growth of TGO and separation at YSZ-TGO interface.