Surface & Coatings Technology 205 (2011) 4355–4365
Contents lists available at ScienceDirect
Surface & Coatings Technology
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / s u r f c o a t
Reaction, transformation and delamination of samarium zirconate thermal
barrier coatings
Hengbei Zhao a,⁎, Matthew R. Begley b, Arthur Heuer c, Reza Sharghi-Moshtaghin c, Haydn N.G. Wadley a
a
b
c
Materials Science and Engineering, University of Virginia, Charlottesville, VA 22904, United States
Mechanical Engineering, University of California, Santa Barbara, CA 93105, United States
Materials Science and Engineering, Case Western Reserve University, Cleveland, OH 44106, United States
a r t i c l e
i n f o
Article history:
Received 16 December 2010
Accepted in revised form 9 March 2011
Available online 24 March 2011
Keywords:
Samarium zirconate
Interface reaction
Bilayers
Delamination
a b s t r a c t
The rare earth zirconates have attracted interest for thermal barrier coatings (TBCs) because they have very low
intrinsic thermal conductivities, are stable above 1200 °C and are more resistant to sintering than yttria-stabilized
zirconia (YSZ). Samarium zirconate (SZO) has the lowest thermal conductivity of the rare earth zirconates and its
pyrochore structure is stable to 2200 °C but little is known about its response to thermal cycling. Here, columnar
morphology SZO coatings have been deposited on bond coated superalloy substrates using a directed vapor
deposition method that facilitated the incorporation of pore volume fractions of 25 to 45%. The as-deposited
coatings had a fluorite structure which transformed to the pyrochlore phase upon thermal cycling between 100
and 1100 °C. This cycling eventually led to delamination of the coatings, with failure occurring at the interface
between the TGO and a “mixed zone” that formed between the thermally grown alumina oxide (TGO) and the
SZO. While the delamination lifetime increased with coating porosity (reduction in coating modulus), it was
significantly less than that of similar YSZ coatings applied to the same substrates. The reduced life resulted from a
reaction between the rare earth zirconate and the alumina-rich bond coat TGO, leading to the formation of a
mixed zone consisting of SZO and SmAlO3. Thermal strain energy calculations show that the delamination driving
force increases with TGO and mixed layer thicknesses and with coating modulus. The placement of a 10 μm thick
YSZ layer between the TGO and SZO layers eliminated the mixed zone and restored the thermal cyclic life to that
of YSZ structures.
© 2011 Elsevier B.V. All rights reserved.
1. Introduction
Thermal barrier coating (TBC) systems are an enabling materials
technology for the gas turbine engines used in modern aerospace
applications [1]. Current TBC systems consist of a 100–300 μm thick
thermally protective ceramic layer, a much thinner (up to 5–10 μm
thick) thermally grown oxide (TGO) that protects superalloy turbine
airfoils against rapid oxidation, and a relatively thick (50–100 μm)
metallic “bond coat” applied to the turbine airfoil alloy to slow the
kinetics of oxidation and promote TGO adherence. The bond coat has a
high aluminum concentration to promote slow (parabolic) growth of a
protective alumina-rich TGO layer [2,3]. The thermal protection part of
the coating system traditionally consisted of 7 wt.% yttria stabilized
zirconia (7YSZ) which has a desirable combination of low thermal
conductivity (~2.1 W m− 1 K− 1 at 1000 ≤ T ≤ 1200 °C [4,5]), excellent
thermochemical compatibility with the TGO layer [6,7], and good
durability to thermal cycling [8] because of its high and well-matched
thermal expansion coefficient and low in-plane elastic stiffness
⁎ Corresponding author.
E-mail address: [email protected] (H. Zhao).
0257-8972/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.surfcoat.2011.03.028
(especially when through thickness voids are incorporated in the
coating). This material is relatively easily applied by a variety of thermal
spray [1] and physical vapor deposition methods [9].
As the temperature within gas turbine engines increases, the
continued use of 7YSZ has become problematic. The primary concerns
arise from: (i) accelerated sintering kinetics of porous 7YSZ (which
decreases the pore content of the TBC and thus its in-plane compliance
and thermal resistance) [9–12]; (ii) de-stabilization of the desirable
non-transformable t′ phase [6,13–16] that is responsible for the intrinsic
toughness of 7YSZ [17,18]; and (iii) an increased propensity for
penetration of the coating by molten deposits of calcium-magnesiumalumino-silicates (CMAS) which dissolve the coating and accelerate YSZ
destabilization. Upon solidification, these deposits degrade the strain
tolerance of the coating and lead to premature failure [19–21]. These
concerns, coupled with the potential design benefits of further
reductions of the thermal conductivity relative to 7YSZ [22–24], have
motivated a far-reaching search for alternate thermal barrier oxides
[6,25–27].
The equiatomic rare earth zirconates (REZ's) M2Zr2O7 (M=La→Gd),
with an ordered pyrochlore structure, have attracted significant attention
because of their lower thermal conductivity (~1.2–1.6 W m− 1 K− 1 at
1000≤T≤1200 °C) [23,28–31]. In addition, these materials are phase-
4356
H. Zhao et al. / Surface & Coatings Technology 205 (2011) 4355–4365
stable at temperatures relevant to future gas turbine operation (turbine
inlet temperatures, T≤1500 °C) [32]. They also appear more resistant to
morphological evolution of the pore structure (sintering) than 7YSZ
[31,33,34]; furthermore, one of the rare earth zirconates (Gd2Zr2O7) has
shown potential for inhibiting CMAS penetration by inducing crystallization of the melt [35]. However, the REZ's suffer two fundamental
disadvantages relative to 7YSZ. Firstly, Gd2Zr2O7 is susceptible to reaction
with the alumina TGO [6,36], resulting in the formation of aluminate
interphases by diffusional reactions [37,38]. The other REZ's are also
predicted to be susceptible to similar reactions with alumina TGOs.
Secondly, as with the cubic ZrO2-based phases, the REZ's exhibit a
substantially lower toughness than the tetragonal t′ phase of 7YSZ
[2,17,39,40]. This has deleterious consequences for both their erosion
resistance and other failure mechanisms that involve crack propagation
through the TBC [2,41]. To alleviate the aluminate reaction problem,
gadolinium zirconate TBCs are typically deposited on a TGO-compatible
7YSZ underlayer [6,23,37,42]. However, little is known about the factors
governing the delamination resistance of these bi-layer coatings upon
thermal cycling at temperatures of relevance in gas turbine engines.
Samarium zirconate, Sm2Zr2O7 (SZO), coatings have recently been
deposited on MCrAlY bond coated Hastelloy-X substrates using a
directed vapor deposition (DVD) method [34]. By varying the deposition
temperature and rotation conditions, it was possible to vary the pore
volume fraction in the SZO coating from 20 to 45%. The lowest density
coatings had a very low thermal conductivity (0.5 W/mK), resulting
from the low intrinsic solid material's conductivity and the high void
volume fraction. Inasmuch as the elastic modulus decreases with
increasing pore fraction [43], these columnar coatings also have low
in-plane elastic stiffness and might be anticipated to have good thermal
delamination resistance.
Here, we deposit a series of SZO coatings with widely varying pore
volume fractions upon the same bond coat substrate system used for an
earlier study of thermocyclic delamination of 7YSZ coatings [43]. We
then investigate its response to thermal cycling, and identify the
mechanism of failure and its relationship to the thermochemical
stability of the SZO-TGO interface. An aluminate phase is indeed formed
at this interface, and an elastic thermal strain analysis of the multilayer
system shows a significant delamination driving force at both the TGOmixed zone and mixed zone-SZO interfaces, which results in a
shortened thermocyclic lifetime. We then show that a 7YSZ/SZO
bi-layer TBC system can restore the cyclic life (and return the failure
mode) to that of the 7YSZ system, and examine the factors that govern
the stored strain energy that drives eventual failure at the bond
coat/TGO interface.
2. Experimental methods
The SZO coatings were grown by the DVD method using a deposition
temperature of 1000 °C at various substrate rotation rates [44,45].
A detailed description of the fabrication is presented elsewhere [45].
Bi-layer (7YSZ/SZO) coatings were also deposited using the same
technique. In the bi-layer case a two-source-rod crucible that held
12.7 mm diameter 7YSZ and SZO rods was used, Fig. 1. The substrates
were heated to 1000 °C, and a 10 μm thick 7YSZ coating was deposited.
This was then followed by deposition of an 80 μm thick SZO layer by
translating the electron beam to the second source rod without cooling
or breaking the vacuum system. The substrates were rotated at 6 rpm
during both of these depositions.
A 10:1 helium/oxygen mixture was used to form a transonic gas jet
which entrained and directed the vapor to the substrates. The jet was
created by gas expansion through a nozzle using an upstream pressure
of 140 Pa and a downstream (chamber) pressure of 16 Pa. These process
conditions resulted in a deposition rate of 4 ± 1 μm/min for both 7YSZ
and SZO. The SZO and 7YSZ/SZO coated samples were then thermally
cycled between ambient and 1100 °C under identical conditions to those
used for a study of the thermocyclic response of 7YSZ coatings grown by
a similar method on the identical bond coat/substrate system [43]. The
coatings were characterized by X-ray diffraction (XRD), scanning
electron microscopy (SEM), energy dispersive X-ray analysis and
transmission electron microscopy (TEM).
Fig. 1. The vapor deposition approach used to deposit yttria stabilized zirconia and samarium zirconate coatings. The coatings were deposited on 25.4 mm diameter, 3.2 mm thick
Hastelloy-X substrate with a smooth (polished) NiCoCrAlY bond coat at a temperature of 1000 °C.
H. Zhao et al. / Surface & Coatings Technology 205 (2011) 4355–4365
The TEM cross section samples were prepared by a “lift-out”
technique using a FEI Quanta 3D dual-beam SEM/FIB microscope [46].
A 20 μm × 10 μm, 1.5 μm thick slice was prepared that included the bond
coat and SZO layer using the focused ion beam (FIB). This was lifted out
and placed on a 3 mm copper grid. Further thinning of the sample was
then performed with the sample on the grid (using a modest ion beam
current to minimize FIB damage) until the specimen was electron
transparent. This corresponded to a final foil thickness of 110–130 nm.
The TEM samples were examined in a Zeiss Libra 200 keV FEG
transmission electron microscope with an in-column “omega” filter.
The microscope is equipped with a STEM detector and a Thermo
Noran XEDS system running Noran System Seven (NSS) software. The
analyzer on this XEDS system is capable of handling quite high X-ray
counts, making the instrument suitable for elemental mapping in
reasonable times. The mapping was performed using a 37 μm
condenser aperture with a spot size of 10 nm for 15 min. The
instruments drift compensator was activated to compensate for
possible drift during X-ray data acquisition.
4357
3. Thermally cycled SZO coatings
The as-deposited SZO coatings had a columnar structure with
faceted column tips, Fig. 2(a). The tips had a well-defined 4-sided
pyramidal shape typical of vapor-deposited fluorite-derivative
structures. A detailed analysis of the columnar structure has been
described in an earlier paper [45]. The coatings exhibited a compositional variation across the thickness of up to ±12% SmO1.5, resulting
from the difference in vapor pressure of the constituent oxides and their
scattering rates in the vapor phase. Sm2O3 was most highly enriched
near the first deposited SZO due to its much higher vapor pressure at
typical melt temperatures. A ~0.4 μm thick TGO layer, Fig. 2(d), was
formed as the samples were heated prior to and during deposition.
The SZO coated test coupons were thermally cycled to 1100 °C in air
(with a 1 h hold at peak temperature), and their microstructural
changes examined. Representative micrographs of their outer surfaces
and cross-sections are shown in Fig. 2. After 50 thermal cycles, the sharp
edges of the column tips and growth steps on the facets had become
Fig. 2. The surface (a) and cross-section (d) of the as deposited SZO coating. (b) and (e), (c) and (f) show the coating after 50 and 300 cycles, respectively. The delamination occurs at
the SZO/TGO interface for the spalled samples. The coatings were thermally cycled between 1100 °C and ambient temperature.
4358
H. Zhao et al. / Surface & Coatings Technology 205 (2011) 4355–4365
rounded, Fig. 2(b), indicative of appreciable surface diffusion. Increasing
the number of thermal cycles to 300, Fig. 2(c), caused more substantial
surface reconstruction.
Examination of cross-section views Fig. 2(d)–(f) shows a rapid
increase in TGO thickness with thermal exposure, as expected. The thin
(~0.4 μm thick) TGO layer formed during deposition of the SZO layer
increased with high temperature exposure time and Y-Al oxide “pegs”
developed in the TGO and protruded into the bond coat, Fig. 2(e) and (f).
At ~300 cycles, the TGO thickness had increased to ~4 μm and the
coating spalled at the SZO/TGO interface. Close examination of Fig. 2(e)
and (f), reveals the presence of a ~0.4 μm thick mixed zone above the
TGO layer in both thermally cycled samples. In Fig. 2(f) major failure at
the mixed zone-TG interface can be observed. It appears that the factors
governing delamination (strain energy release rates and interface
toughness's) are similar for the two interfaces and the eventual failure
path jumps between them.
A sample that had been exposed to ~300 thermal cycles with the SZO
coating partially attached was examined by XRD. The result in Fig. 3(a)
corresponds to diffraction from the top surface of the SZO layer.
Compared to the equivalent XRD pattern of the as-deposited coating,
which corresponded to diffraction from a “disordered” cubic fluorite
structure, small additional peaks are present in the thermally cycled
coating. These peaks have been indexed and arise from the ordered
pyrochlore structure. This indicates that the fluorite structure of
as-deposited SZO coating had transformed from fluorite to the ordered
pyrochlore structure after exposure to high temperature for 300 h.
Fig. 3. X-ray diffraction data for the partially spalled sample subjected to ~300 cycles
(shown in Fig. 2). (a) from a region where the SZO coating was still attached to the bond
coat, and (b) from the TGO surface (a region with no SZO coating). Note that the small
pyrochlore peaks occur in the spectrum.
The diffraction pattern from the TGO side of the substrate exposed
by the delamination for this sample is shown in Fig. 3(b). Fig. 2(f)
suggests this pattern is from the “mixed” zone. Diffraction peaks from
the SZO coating are present in the pattern, which indicates that in
some regions the SZO coating remained attached to the TGO after
coating delamination. The distinct peaks (the solid circles) for the
pyrochlore SZO phase are evident in the pattern. In addition to the
peaks from the SZO phase, new peaks have appeared in the XRD data.
Most of them are consistent with SmAlO3 reflections, which imply
that SZO and Al2O3 have reacted to form this new phase.
STEM images of the delamination region of a thermally cycled SZO
sample are shown in Fig. 4(a). It can be seen that in this region of the
sample, the crack propagated at the interface of the mixed zone and
the alumina rich TGO. Elemental mapping was performed on the
corresponding region, Fig. 4(b) and (c). The TGO layer beneath the crack
is Al-rich, while the mixed zone contains Al, Sm and Zr. Fig. 4(b) shows
that Sm had diffused into the TGO layer and reacted with alumina to
form samarium aluminate.
A TEM image of a thermally cycled SZO sample is shown in Fig. 5(a).
The polycrystalline electron diffraction pattern was obtained in the
mixed zone, Fig. 5(b), and was indexed as shown in Table 1, indicating
Fig. 4. (a) The STEM image of thermally cycled SZO sample, (b) the corresponding XEDS
elemental mapping for Al, Cr and Sm and (c) the elemental mapping for Zr. Note that
the debonding occurred at the interface of mixed layer/TGO.
H. Zhao et al. / Surface & Coatings Technology 205 (2011) 4355–4365
4359
Fig. 5. (a) TEM bright field image of cycled SZO sample and (b) the diffraction pattern of the mixed zone.
that these rings originate from the SZO coating and the reaction product,
SmAlO3.
The dependence of the cyclic lifetime upon rotation rate for the SZO
coatings is shown in Fig. 6(a). Like 7YSZ coatings applied to the identical
bond coat, the SZO coating delamination life decreases with the rotation
rate used to deposit the coating (from ~350 to ~150 cycles). The growth
in TGO thickness with thermal exposure time (number of 1-h holds at
1100 °C) is shown in Fig. 6(b). The data have been fitted to a parabolic
TGO growth law; (x− xo)2 = kpt where xo is the ‘as-deposited’ TGO
thickness, x the thickness after accumulating a time t at 1100 °C. The best
fit to the parabolic rate constant, kp, was 0.035 μm2/h.
In order to compare the response of the SZO coatings with that of
7YSZ coatings deposited on the same bond coat/substrate system under
the same deposition conditions, the TGO thickness versus number of
thermal cycles has been plotted for both coatings in Fig. 7(a). The SZO
coatings have a higher TGO growth rate than YSZ whose parabolic
constant was 0.014 μm2/h; less than a half that of the SZO system. YSZ
coatings also had much longer thermocyclic lifetimes than SZO coatings,
Fig. 7(b). Since increases in rotation rate increase the density of the SZO
and YSZ coatings, the lifetime of both coatings decreased as the coating
density was increased.
Several experiments have shown that rare earth zirconates react
with alumina in the TGO layer. Leckie et al. found that Gd2O3 reacts with
Al2O3 when exposed at high temperature [37]. In their case, GdAlO3
appeared to greatly lower the interface toughness of the TBC/TGO. It was
thus anticipated that SmO1.5 would react with Al2O3 to form SmAlO3;
4. SZO coating failure mechanism
The SZO coatings on NiCoCrAlY bond coats clearly have very short
cyclic lifetimes, as much as a factor of four less than identically deposited
YSZ coatings on the same bond coats [43]. The bond coat surface
condition initially governs the spallation failure mechanism. Rough
(grit blasted) bond coats of the composition used here are susceptible to
a “rumpling” failure mechanism with delamination occurring at the
TGO-YSZ interface [43]. For the smooth bond coats used here, rumpling
was suppressed and failure occurred at the bond coat-TGO interface
underneath a YSZ coating. Failure was initiated at small cracks at local
TGO thickness heterogeneities. Complete delamination finally occurred
at the bond coat/TGO interface when the strain energy release rate was
greater than the toughness of bond coat/TGO interface.
Table 1
The d spacings and diffraction planes from the electron diffraction pattern of the mixed zone.
Ring
d(Å)
Diffraction planes
1
2
3
4
5
6
3.35
3.05
2.69
2.17
2.05
1.87
(111) SmAlO3
(222) SZO
(200)SmAlO3/(400)SZO
(202) SmAlO3
(511) SZO
(440) SZO
Fig. 6. Thermal cycling response for SZO coatings deposited on smooth (polished)
NiCoCrAlY bond coat: (a) Cyclic delamination lifetime vs. rotation rate used during SZO
coating deposition; (b) the TGO growth rate vs. number of heating cycles. The samples
were heated from ambient to 1100 °C and held for 1 h prior to forced cooling back to
ambient.
4360
H. Zhao et al. / Surface & Coatings Technology 205 (2011) 4355–4365
Fig. 7. (a) A comparison of TGO growth rate for SZO and YSZ coatings on polished
substrates, (b) The relation between lifetime and density for YSZ and SZO coatings. For the
same density, the SZO coating has more pore fraction than 7YSZ because the fully dense
SZO and 7YSZ materials have the density of 6.7 and 6.0 g/cm3, respectively.
and this has now been verified by XRD. The STEM analysis indicated the
mixed zone to be made up of Al, Sm and Zr, consistent with a two-phase
SZO/SmAlO3 layer.
To examine the mechanism of this premature failure mode we note
that delamination during thermo-mechanical cycling of a coating occurs
when the stored elastic strain energy in the coating system that can be
released during debonding exceeds that needed to extend a crack in the
system. The experiments above indicate that the preferred fracture path
is at the TGO/mixed oxide layer interface, with occasional jumping to
the mixed layer/SZO interface, Fig. 8(a). The steady-state energy release
rate for debonding can be computed from the change in stored elastic
energy associated with debonding of a semi-infinite crack at a given
interface [47]. In the analysis, it is assumed that the multilayer is formed
at elevated temperature with initially stress-free layers. As the
multilayer is cooled, biaxial stresses develop in the layers due to
thermal expansion mismatch between the layers of the coating and with
the substrate. These stresses are used to compute the stored elastic
strain energy in the multilayer prior to debonding. The analysis is
repeated for debonded sections – i.e. the stresses are re-computed in the
multilayer's present after debonding – and the energy release rate is
calculated from the difference between the intact and debonded states
[48]. In the present cases, the substrate is sufficiently thick that the
deformation of the substrate is not affected by the presence of the
additional layers. This implies that debonding does not alter the state of
stress for layers underneath the debonded interface; as such, the energy
release rate is impacted only by a change in the layers above the
debonded interface. It is important to recognize that only the released
layer can contribute to the crack driving force. The material that remains
attached to a thick substrate does not relax and therefore does not
contribute to the debond energy release.
For a SZO coating where failure occurs at the TGO/SZO interface
(above the TGO), the energy release rate is independent of the TGO
thickness, since debonding at this interface does not release the strain
energy stored in the TGO (because of constraint by the bond
coat/substrate). For debonding at the bond coat/TGO interface, the
energy release rate is strongly dependent upon TGO thickness
(and properties), because debonding creates a TGO/SZO bi-layer with
much lower stored elastic energy than when it is adhered. In the
experimental system evaluated here, a mixed layer had formed at the
alumina-rich TGO. The TGO in contact with the bond coat is
predominantly alumina and is similar to that seen in other TBC systems.
However, the mixed layer in contact with the SZO coating is rich in both
aluminum and samarium and appears to consist of SmAlO3 and SZO.
The thermo-mechanical properties of the mixed layer can be
estimated by taking an average of the values of these two compounds.
These and the properties of the other layers are summarized in Table 2,
and an energy release rate calculation performed for the system,
Fig. 8(b). The mixed layer has a very high modulus compared to the SZO
and significant misfit strains with the substrate, which implies
significant stored elastic energy in this layer when it is adhered to the
substrate. When debonding occurs at the TGO/mixed layer interface, the
significant stored elastic energy in the mixed layer is released. This is
completely analogous to the role of the TGO in YSZ systems.
Fig. 8(b) shows the energy release rates for various interfaces in the
system as a function of either TGO or mixed layer thickness. Results are
shown for three debonding scenarios and three values of SZO modulus,
spanning the range of the experiments. The dashed lines correspond to
debonding beneath the TGO, without a mixed layer present; hence the
abscissa in this case is the TGO thickness. The solid lines correspond to
debonding at the TGO/mixed layer interface; the abscissa in this case is
the mixed layer thickness. The dot-dashed lines represent the energy
release rate for just the SZO layer. The third scenario neglects the
presence of the mixed layer and predicts the driving force for a
delamination crack between the TGO and SZO. Since strain energy is
only released in the SZO, this driving force is equivalent to that of SZO
debonding from a mixed layer that stays attached to the substrate
(provided the mixed zone has not significantly altered the total
thickness of the SZO).
In all debonding scenarios, the energy release rate increases with
increasing SZO modulus. For example, the SZO layers with lowest
modulus (highest pore volume fractions) have the lowest elastic strain
energy when released by a crack at the bond coat/TGO interface (brown
dotted lines). Lower values of SZO modulus imply less strain energy in
the adhered state, and consequently less strain released by cracking, and
lower energy release rates.
Broadly speaking, the energy release rate will scale with the
thickness of any highly stressed layer that is released by cracking. For
example, when debonding occurs at the bond coat/TGO interface
(dashed lines in Fig. 8b), the energy release rate scales with the TGO
thickness, as debonding relieves the high TGO stresses present in the
adhered state. When debonding occurs just beneath the SZO layer
(dot-dashed lines in Fig. 8b), the energy release rate for debonding is
constant with respect to oxide layer thickness (or mixed layer
thickness) because the highly stressed layers remain adhered during
debonding, and hence, do not release additional strain energy. In
Fig. 2(f), the debonding occurred at the interface of the TGO with the
mixed layer, the high stresses in this layer in the adhered state being
relieved by debonding. The energy release rate for debonding at the
TGO/mixed layer interface is roughly equal to the average of the other
two cases (underneath the TGO and above the mixed layer), because the
properties of the mixed layer are roughly equal to the average of the TGO
and SZO. That is, the energy release rate is highest for the interface
H. Zhao et al. / Surface & Coatings Technology 205 (2011) 4355–4365
4361
Fig. 8. (a) Illustration of the delamination of an SZO coating at the TGO/mixed zone interface. (b) Calculated energy release rate for the steady-state debonding of an SZO TBC for the various
interfaces as a function of either the alumina TGO thickness (dashed lines and dot-dashed lines), or the mixed layer thickness (solid lines), for three different SZO layer moduli values.
The energy release rate for debonding at the TGO/SZO interface (dot-dashed), at the bond coat/TGO interface (dashed), and at the interface between the TGO and a mixed reaction layer
(solid lines) are shown. When debonding occurred above a layer, the layer's thickness has a minimal effect upon the crack driving force since it remains highly stressed after delamination by
the constraint of the bond coat and substrate.
underneath the TGO because the misfit strains are greatest in the TGO;
the energy release rate for the interface under the mixed layer is smaller,
because the misfit strains for the mixed layer are smaller than those of
the TGO.
The experimental results show that as the TGO and mixed zone
thicknesses increase, debond failure occurred at a combined thickness of
just under 5 μm, with a mixed layer thickness of about 0.8 μm. The
micrographs in Figs. 2, 4 and 5 suggest that the crack path alternates
between the TGO/mixed layer interface, and the mixed layer/SZO
interface. This observation, when coupled with the predictions in Fig. 8,
implies that the interface toughness of these two interfaces is similar
and much lower than that of the bond coat/TGO. For example, for a
mixed layer thickness of 1 μm, the driving force at the mixed layer/SZO
Table 2
Thermo-mechanical properties of the layers used for the multilayer debonding analysis.
Thermo-mechanical properties for debonding analysis: ΔT = −1000 °C
Substrate
Bond coat
TGO
Mixed layer
SZO
Thickness (um)
Modulus (GPa)
Poisson's ratio
CTE (ppm/°C)
5000
175
Variable
Variable
90-mixed thick.
200
90
380
200
30, 36, 42
0.3
0.3
0.2
0.2
0.2
16.6
15.2
8.5
9.0
10.8
interface is inferred to be ~65–90 J/m2 (noting that the TGO/SZO
prediction depends only on the SZO thickness, and hence, is the same as
that of the mixed layer/SZO interface), while that at the TGO/mixed
layer interface is ~70–100 J/m2. The fact that the driving forces are close
to one another (and possibly overlap, depending on local values of
modulus and mixed layer thickness) supports the observation that the
crack is likely to oscillate between these two interfaces.
5. Thermally cycled 7YSZ/SZO bi-layer
5.1. As-deposited bi-layer
The microstructure of an as-deposited bi-layer coating is shown in
Fig. 9. The individual SZO and 7YSZ layers are apparent. The surface
morphology was typical of SZO coatings deposited on polished
NiCoCrAlY bond coats [45]. The thickness of the samarium zirconate
was ~80 μm, while the 7YSZ was ~10 μm thick. A thin, ~0.3 μm, layer of
TGO had formed at the 7YSZ/bond coat interface during deposition.
A polished cross section of the 7YSZ/SZO bi-layer is shown in Fig. 10.
The transition from 7YSZ to SZO occurred with little disruption of the
feathery columnar structure. The continuity of the pre-existing 7YSZ
columnar growth pattern in the subsequent SZO coating is apparent at all
relevant length scales, from the intercolumnar gaps to the intracolumnar
feathers shown in Fig. 10, indicating that the as-deposited 7YSZ/SZO
4362
H. Zhao et al. / Surface & Coatings Technology 205 (2011) 4355–4365
intercolumnar gaps. However, Sm signals are not detected close to the
7YSZ/TGO interface, indicating that this 7YSZ layer acted as a diffusion
barrier.
5.2. 50 cycle response
After the bi-layer was thermally cycled between 1100 °C and
ambient temperature for 50 cycles, the TGO layer behaved in a similar
manner to a single 7YSZ coating on a NiCoCrAlY bond coat. The TGO
layer thickened to ~1.5 μm and oxide precipitates appeared in the BEI
image of the TGO layer, in Fig. 11(a). EDS analysis indicates that the
precipitates contain comparable amounts of Y and Al, which are
probably yttrium aluminum garnet, Y3Al5O12 (YAG). There is no
evidence of localized interface separations between the 7YSZ and SZO
layers. After 50 cycles, the feathery structure on the 7YSZ columnar
surfaces had disappeared. The Sm2Zr2O7 feathery structure appears to
have suffered a slower evolution, similar to that observed in a
7YSZ/Gd2Zr2O7 bi-layer coating [40]. In addition, a mixed zone including
YSZ and alumina is also observed at the 7YSZ/TGO interface.
The formation of a mixed zone is thought to result from dissolution of
the first deposited YSZ in metastable (gamma and theta) alumina oxides
and this then precipitates as the metastable alumina transforms to alpha
alumina [2,51].
The 7YSZ/SZO interface is of particular interest in Fig. 11(d).
Although the top SZO coating is intact and attached to the 7YSZ, there
is a dense region at the interface within the SZO coating. A similar region
has been observed in thermal cycled single layers of SZO and is
consistent with Sm enrichment of the SZO coating [45]. A line scan
analysis (Fig. 12a) was conducted to investigate the diffusion of Sm in
the YSZ layer and indicated that while Sm had diffused into the YSZ
coating, it had not yet reached the YSZ/TGO interface.
Fig. 9. (a) Cross-section view of the as-grown SZO/YSZ bi-layer coating. (b) A high
magnification image of the BC/TGO/7YSZ interface.
bi-layer coating retained a strain-tolerant inter-columnar void structure.
The relatively small lattice misfit between t′-YSZ (a=0.5111 nm and
c = 0.5169 nm) [49] and the as-deposited cubic fluorite SZO
(a=0.5295 nm) [50] may have contributed to this continuous growth
morphology into the SZO layer.
We have measured the chemical composition at five positions along
lines parallel to the SZO/7YSZ interface. The line in the 7YSZ layer
was ~0.5 μm from the interface while that in the SZO was 0.7 μm from
the interface. The average of the five composition measurements along
each line is shown in Table 3. It is evident that Sm is present in the YSZ
coating but Y is not detectable in the SZO coating, indicating Sm
diffusion into the YSZ coating during deposition, probably along the
5.3. 350 cycle response
At 350 cycles, small voids occur locally at the bond coat/TGO
interface, Fig. 11(b). However, the bi-layer coating remained attached to
the TGO. The TGO thickness had also increased to ~3.5 μm. Once again,
no detachment of SZO occurred at the 7YSZ/SZO interface, Fig. 11(e),
even though some spheroidal porosity had formed at this interface. The
more prolonged 350-hour thermal exposure also increased the
penetration of Sm into the YSZ coating. EDS measurements detected a
low concentration of Sm (b1 at.%) near the TGO. A line scan analysis is
also shown in Fig. 12(b). The Sm signal is detectable throughout the YSZ
coating, consistent with the EDS analysis of the YSZ/TGO interface.
Although the diffusion coefficient of Sm in 7YSZ at 1100 °C is
not
pffiffiffiffiffiffi
available, it can be estimated from the observed diffusion distance (2 Dτ )
of ~15 μm after a diffusion time τ=350 h, Fig. 12(b). This estimated
diffusivity (D) is ~10− 17–10− 16 m2/s and is comparable to the reported
diffusivity of ~10− 16 m2 s− 1 for Y in YSZ [52]. We note that the effective
diffusivity may have been accelerated by the unusual microstructure of
these coatings. They contain many inter- and intra-columnar pores which
increase the opportunities for Sm to migrate along void surface paths with
potentially lower barriers to thermally activated diffusion.
5.4. Spallation
Upon further cycling, the bi-layer samples eventually undergo
delamination between 1000 and 1150 cycles with the delamination
Table 3
EDS analysis for lines 1 and 2.
Fig. 10. The polished cross-section of SZO coating on 7YSZ coating. Small colored circles
represent locations where EDS measurements were recorded.
Position
Sm (at.%)
Y (at.%)
Zr (at.%)
Line 1
Line 2
60.2
4.7
0
6.9
39.8
88.4
H. Zhao et al. / Surface & Coatings Technology 205 (2011) 4355–4365
4363
Fig. 11. Evolution of the bi-layer SZO/YSZ coatings deposited at 1000 °C on polished substrates. The bilayer coatings were thermally cycled between 1100 °C and ambient
temperature. Note that the delamination occurs at the TGO/bond coat interface.
fracture path located at the bond coat/TGO interface (Fig. 11c). The
TGO thickness at delamination was ~5.9 μm and wide inter-columnar
gaps, probably due to sintering, had deeply penetrated from the
surface into the SZO coating, Fig. 11(c). A mixed zone was still evident
at the TGO/YSZ interface (Fig. 11c). The two ceramic materials in the
bi-layer coating have very similar thermo-mechanical properties and
appear to have remained compatible with each other. Fig. 11(f)
provides a qualitative comparison of the sintering rate differences
between the SZO and YSZ components of the coating. The SZO still
retained a columnar structure while the columns of YSZ coating had
almost disappeared consistent with the view that SZO has a better
resistance to sintering than YSZ.
The effect of thermal cycling time upon the TGO layer thickness is
summarized in Fig. 13. The parabolic rate constant that best fits the data
has a value of ~0.016 μm2/h, very slightly higher than that of a single
7YSZ coating. A comparison of the delamination lifetimes for 7YSZ, SZO
and 7YSZ/SZO bi-layer coatings is shown in Table 4. All the samples
experienced the same cycling history, i.e. they were heated at 1100 °C
for an hour and followed by 10-min forced air cooling to about 100 °C.
Both 7YSZ and the bi-layer coating have comparable failure lifetimes,
much longer than that of SZO. The use of the bi-layer coating
architecture has therefore enabled exploitation of the lower SZO
thermal conductivity without incurring the deleterious reaction
between SZO and the alumina rich TGO.
6. Bi-layer failure mechanism
The high temperature cycling results indicate that 7YSZ is a
successful diffusion barrier to samarium from SZO even in the presence
of the wide inter-column void gaps needed for strain tolerance. The YSZ/
SZO bi-layer TBC spalled at the bond coat/TGO interface after a similar
number of thermal cycles and by a similar mode to that of a single layer
YSZ on the same polished NiCoCrAlY bond coat. This arises because SZO
and YSZ have similar thermo-mechanical properties, and so the strain
energy release rate calculations summarized in Fig. 8 can be used to
deduce the YSZ, SZO and bi-layer driving forces for fracture.
Delamination at the bond coat/TGO interface releases the stored
strain energy of both the TGO and thermal barrier coating layers and
therefore results in a very large strain energy release which increases
with both TGO thickness and thermal barrier coating modulus, Fig. 8. As
4364
H. Zhao et al. / Surface & Coatings Technology 205 (2011) 4355–4365
Table 4
Comparison of thermal cyclic lifetime of three different coatings deposited on the
polished bond coats at the rotation rate of 6 rpm at 1000 °C.
Coatings
Cycles to failure
Critical TGO thickness at spallation (μm)
7YSZ
SZO
Bi-layer 7YSZ/SZO
1140 ± 110
300 ± 30
1075 ± 75
5.5
3.8
5.9
reported [54] and this may also have contributed to a time dependent
reduction in toughness. Macroscopic delamination would then have
occurred at the time (and TGO thickness) where the drop in toughness
crosses over the increase in stored energy with TGO thickness.
These insights indicate that concepts which delay the rate of TGO
growth are likely to significantly extend the life of a thermal barrier
coating. The use (and maintenance during high temperature exposure)
of a low in-plane modulus coating material/coating pore morphology is
also advantageous. Bi-layer and even tri-layer coating concepts may
open up a wider range of material opportunities. Bond coats or thermal
barrier coatings that provide opportunities for plastic dissipation
(as opposed to elastic storage) of the thermal expansion mismatch
strains are also likely to extend the life of TBC systems. The emergence of
flexible coating deposition technologies may enable expanded
explorations of these possibilities.
7. Conclusions
Fig. 12. Line scan analyses of Sm composition for the bi-layer coating after (a) 50 and (b)
350 thermal cycles, respectively. The coating was grown at 1000 °C using a rotation rate of
6 rpm.
the TGO thickness approached a thickness of 6 μm (the thickness at
which delamination fracture was observed to occur), the predicted
strain energy release rate reached 160–180 J m− 2(depending upon
thermal barrier coating modulus). This can be compared to recent
experimental estimates of an interfacial toughness of 110 J m− 2 for
β-NiAl related systems [53]. We note that the delamination crack
extension at the two-phase bond coat/TGO interface of the system
investigated here might exhibit a higher initial toughness than that of
the intermetallic NiAl system. However, the growth of the TGO was also
accompanied by the gradual appearance of short voids at the bond
coat/TGO interface which may have gradually reduced the toughness of
the interface over time. Sulfur diffusion to this interface has also been
Both SZO and SZO/YSZ bi-layer coatings have been deposited on
polished NiCoCrAlY bond-coated Hastelloy-X substrates using an
electron beam, directed vapor deposition method and their cycling
response has been investigated. We find that:
1. The lifetime of SZO coatings is increased by increasing the coatings
compliance (by incorporating voids) but is significantly less than that
of similar compliance YSZ coatings deposited on the same bond coat.
The reduced life resulted from a chemical reaction between the rare
earth zirconate and the alumina-rich TGO leading to the formation of
SmAlO3 and a reduced toughness interface.
2. The placement of a 10 μm thick YSZ layer between the TGO and SZO
layers eliminated the TGO/mixed zone failure mechanisms and
significantly improved the coating lifetime to that of YSZ coatings.
3. The YSZ layer successfully acted as a diffusion barrier layer to the
migration of samaria to the TGO, inhibiting the chemical reaction
between SZO and TGO and slowing the TGO growth rate, and thus
increasing the delamination lifetime of bi-layer coatings.
4. Eventual delamination of the 7YSZ/SZO bi-layer occurred at the bond
coat/TGO interface. Delamination of this interface releases an
estimated 160–180 J m− 2 of stored strain energy in the coating
system. This strain energy increased with TGO layer thickness and
coating modulus. For the coatings studied here, final failure occurred
by the linking of an array of thin voids at the bond coat/TGO interface
and may have been associated with segregation of sulfur to the
failure interface.
Acknowledgements
We are very grateful to Carlos Levi for many helpful discussions of
the interpretation of this work. We also thank David Wortman, formerly
of GE Aviation for kindly providing the bond coated substrates. We are
grateful to the Office of Naval Research who supported the research
(Contract # N00014-03-1-0297) under the program direction of
Dr. David Shifler.
References
Fig. 13. The evolution of the TGO thickness with number of thermal cycles for the bi-layer
and 7YSZ coated samples. Both sets of coatings were grown on polished NiCoCrAlY bond
coats at 1000 °C.
[1] R.A. Miller, J. Therm. Spray Technol. 6 (1997) 35.
[2] A.G. Evans, D.R. Clarke, C.G. Levi, J. Eur. Ceram. Soc. 28 (2008) 1405.
H. Zhao et al. / Surface & Coatings Technology 205 (2011) 4355–4365
[3] N.P. Padture, M. Gell, E.H. Jordan, Science 296 (2002) 280.
[4] M. Fevre, A. Finel, R. Caudron, R. Mevrel, Phys. Rev., B: Condens. Matter Mater.
Phys. 72 (2005) 104118.
[5] R.A. Miller, Surf. Coat. Technol. 30 (1987) 1.
[6] C.G. Levi, Curr. Opin. Solid State Mater. Sci. 8 (2004) 77.
[7] O. Fabrichnaya, F. Aldinger, Z. Metallkunde 95 (2004) 27.
[8] S. Stecura, in: NASA, Cleveland, OH, 1985.
[9] U. Schulz, J. Am. Ceram. Soc. 83 (2000) 904.
[10] V. Lughi, V.K. Tolpygo, D.R. Clarke, Mater. Sci. Eng., A 368 (2004) 212.
[11] R.G. Hutchinson, N.A. Fleck, A.C.F. Cocks, Acta Mater. 54 (2006) 1297.
[12] D. Zhu, R.A. Miller, J. Therm. Spray Technol. 9 (2000) 175.
[13] J.M. Cairney, N.R. Rebollo, M. Rüler, C.G. Levi, Int. J. Mater. Res. 12 (2007) 1177.
[14] R.A. Miller, J.L. Smialek, R.G. Garlick, Phase Stability in Plasma-sprayed, Partially
Stabilized Zirconia-Yttria, in Science and Technology of Zirconia, The American
Ceramic Society, Inc., Columbus, OH, 1981.
[15] J.R. Brandon, R. Taylor, Surf. Coat. Technol. 46 (1991) 75.
[16] J.R. VanValzah, H.E. Eaton, Surf. Coat. Technol. 46 (1991) 289.
[17] C. Mercer, J.R. Williams, D.R. Clarke, A.G. Evans, Proceedings of the Royal Society
A-Mathematical Physical and Engineering Sciences, 2007, p. 1393.
[18] T.A. Schaedler, R.M. Leckie, S. Kramer, A.G. Evans, C.G. Levi, J. Am. Ceram. Soc. 90
(2007) 3896.
[19] M.P. Borom, C.A. Johnson, L.A. Peluso, Surf. Coat. Technol. 86–87 (1996) 116.
[20] C. Mercer, S. Faulhaber, A.G. Evans, R. Darolia, Acta Mater. 53 (2005) 1029.
[21] S. Krämer, J. Yang, C.G. Levi, C.A. Johnson, J. Am. Ceram. Soc. 89 (2006) 3167.
[22] D.R. Clarke, Surf. Coat. Technol. 163–164 (2003) 67.
[23] M.J. Maloney, in, United Technologies Corporation (Hartford, CT), United States
Patent 6284323, 2001.
[24] J.R. Nicholls, K.J. Lawson, A. Johnstone, D.S. Rickerby, Surf. Coat. Technol. 151–152
(2002) 383.
[25] X.Q. Cao, R. Vassen, D. Stoever, J. Eur. Ceram. Soc. 24 (2004) 1.
[26] D.R. Clarke, C.G. Levi, Annu. Rev. Mater. Res. 33 (2003) 383.
[27] D.R. Clarke, S.R. Phillpot, Mater. Today 8 (2005) 22.
[28] G. Suresh, G. Seenivasan, M.V. Krishnaiah, P.S. Murti, J. Alloy. Comp. 269 (1998) L9.
[29] G. Suresh, G. Seenivasan, M.V. Krishnaiah, P. Srirama Murti, J. Nucl. Mater. 249
(1997) 259.
4365
[30] J. Wu, X. Wei, N.P. Padture, P.G. Klemens, M. Gell, E. Garcia, P. Miranzo, M.I. Osendi,
J. Am. Ceram. Soc. 85 (2002) 3031.
[31] R. Vassen, X. Cao, F. Tietz, D. Basu, D. Stover, J. Am. Ceram. Soc. 83 (2000) 2023.
[32] C. Wang, in: Max-Planck-Institut für Metallforschung, Stuttgart, 2006.
[33] X.Q. Cao, R. Vassen, W. Jungen, S. Schwartz, F. Tietz, D. Stover, J. Am. Ceram. Soc. 84
(2001) 2086.
[34] R. Subramanian, in, Siemens Westinghouse Power Corporation (Orlando, FL), US
Patent 6258467, 2001.
[35] S. Krämer, J. Yang, C.G. Levi, J. Am. Ceram. Soc. 91 (2008) 576.
[36] S. Lakiza, O. Fabrichnaya, C. Wang, M. Zinkevich, F. Aldinger, J. Eur. Ceram. Soc. 26
(2006) 233.
[37] R.M. Leckie, S. Kramer, M. Ruhle, C.G. Levi, Acta Mater. 53 (2005) 3281.
[38] J. Wu, N.P. Padture, M. Gell, Scr. Mater. 50 (2004) 1315.
[39] U. Bast, E. Schumann, Ceramic Engineering and Science Proceedings, 2002, p. 525.
[40] R.M. Leckie, Doctoral Dissertation in Materials, University of California, Santa
Barbara, Santa Barbara, CA, 2006.
[41] A.G. Evans, D.R. Mumm, J.W. Hutchinson, G.H. Meier, F.S. Pettit, Prog. Mater. Sci. 46
(2001) 505.
[42] R. Vassen, F. Traeger, D. Stöver, Int. J. Appl. Ceram. Technol. 1 (2004) 351.
[43] H. Zhao, Z. Yu, H.N.G. Wadley, Surf. Coat. Technol. 204 (2010) 2432.
[44] H. Zhao, F. Yu, T.D. Bennett, H.N.G. Wadley, Acta Mater. 54 (2006) 5195.
[45] H. Zhao, C.G. Levi, H.N.G. Wadley, Surf. Coat. Technol. 203 (2009) 3157.
[46] L.A. Giannuzzi, J.L. Drown, S.R. Brown, R.B. Irwin, F.A. Stevie, Microsc. Res. Tech. 41
(1998) 285.
[47] J.W. Hutchinson, Z. Suo, in: W.H. John, Y.W. Theodore (Eds.), Advances in Applied
Mechanics, Elsevier, 1991, p. 63.
[48] M.R. Begley, D.R. Mumm, A.G. Evans, J.W. Hutchinson, Acta Mater. 48 (2000) 3211.
[49] V. Lughi, D.R. Clarke, J. Am. Ceram. Soc. 88 (2005) 2552.
[50] M.A. Subramanian, G. Aravamudan, G.V. Subba Rao, Prog. Solid State Chem. 15
(1983) 55.
[51] C.G. Levi, E. Sommer, S.G. Terry, A. Catanoiu, M. Rühle, J. Am. Ceram. Soc. 86 (2003) 676.
[52] Y. Oishi, K. Ando, Y. Sakka, Advances in Ceramics: Additives and Interfaces in
Electronic Ceramics, 7, The American Ceramic Society, Columbus (OH), 1983.
[53] P.Y. Théry, M. Poulain, M. Dupeux, M. Braccini, J. Mater. Sci. 44 (2009) 1726.
[54] L. Rivoaland, V. Maurice, P. Josso, M.-P. Bacos, P. Marcus, Oxid. Met. 60 (2003) 137.