Engineering Fracture Mechanics 72 (2005) 2144–2158

www.elsevier.com/locate/engfracmech

Fracture behavior under mixed-mode loading of

ceramic plasma-sprayed thermal barrier coatings
at ambient and elevated temperatures
S.R. Choi *, D. Zhu, R.A. Miller
National Aeronautics & Space Administration, Glenn Research Center, Cleveland, OH 44135, USA

Received 6 January 2004; received in revised form 3 January 2005; accepted 17 January 2005
Available online 1 April 2005

Abstract

The fracture behavior under modes I and II loading of ceramic plasma-sprayed thermal barrier coatings was deter-
mined in air at 25 and 1316 °C in asymmetric
pffiffiffiffi four-point flexure. The mode I fracture toughness was found to be
KIc = 1.15 ± 0.07 and 0.98 ± 0.13 MPa m, respectively, at 25 p ffiffiffiffi 1316 °C. The respective ÔnominalÕ mode II fracture
and
toughness values were KIIc = 0.73 ± 0.10 and 0.65 ± 0.04 MPa m. The empirical mixed-mode fracture criterion best
described the coatingsÕ fracture behavior under mixed-mode loading. The angle of crack propagation was in reasonable
agreement with the minimum strain energy density criterion.
Ó 2005 Elsevier Ltd. All rights reserved.

Keywords: Ceramic thermal barrier coatings; Mixed modes I and II loading; Asymmetric four-point flexure; Fracture toughness
testing; Plasma-sprayed ZrO2–8wt%Y2O3

1. Introduction

Thermal barrier coatings (TBCs) have attracted increasing attention for advanced gas turbine and diesel
engine applications due to their ability to provide thermal insulation to engine components [1–3]. The mer-
its of using the ceramic TBCs are well recognized and include the potential increase in engine operating
temperature with reduced cooling requirements, resulting in significant improvement in thermal efficiency,
performance, and reliability. Plasma-sprayed zirconia-based ceramics are one of the most important coat-
ing materials in light of their low thermal conductivity, relatively high thermal expansivity, and unique
microstructure as a result of the plasma spraying process.

*
Corresponding author. Tel.: +1 216 433 8366; fax: +1 216 433 8300.
E-mail address: sung.r.choi@grc.nasa.gov (S.R. Choi).

0013-7944/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfracmech.2005.01.010
 S.R. Choi et al. / Engineering Fracture Mechanics 72 (2005) 2144–2158 2145

It has been suggested that the important limiting factor encountered in thin plasma-sprayed TBCs is the
relatively low fracture energy of the coatings in planes close to and parallel to the interface. Various efforts
have been made to determine mode I ÔinterfacialÕ fracture toughness of the coatings in the vicinity of the
interface using techniques such as the indentation method, debonding technique, three- or four-point flex-
ure delamination technique, and compact tension test method [4–6]. Failure of the thick TBCs has been
observed within the bulk of the coating material, independent of any delamination that typically occurs
at the interface of a thin coating and a substrate [7].
The majority of data on fracture toughness of both thin and thick coatings have been determined under
mode I loading. However, rarely are structural components or coatings subject to pure mode I loading.
This is particularly true for TBCs that encounter complex thermal and mechanical loading in engine oper-
ations. Recently, Callus and Berndt [8] used a pure-shear technique to determine the interfacial critical
mode II strain energy release rate of some thin coatings. Their data, however, were only for ambient tem-
perature. Mixed-mode data on either thin or thick coatings are rarely available in the literature at elevated
temperatures, despite an important fact that mechanical behavior of coatings should not be based solely on
ambient-temperature properties.
The objective of this work was to determine mixed-mode fracture behavior of free-standing TBCs of
ceramic plasma-sprayed ZrO2–8wt%Y2O3 at an elevated temperature of 1316 °C in air. The choice of this
temperature was based on a typical target temperature of aerospace gas turbine applications. These same
properties were also determined at ambient temperature (25 °C). An asymmetric four-point flexure test
technique was used at both temperatures in conjunction with single-edge-v-notched beam (SEVNB) test
specimens, which yielded simplicity in both specimen and crack preparation and in test procedure.

2. Experimental procedures

2.1. Material and test specimens

The ZrO2–8wt%Y2O3 powder with an average particle size of 60 lm, was first plasma-sprayed on a
graphite substrate measuring 150 by 100 by 6.5 mm to a thickness of about 6 mm, using a Sulzer-Metco
ATC-1 plasma coating system with an industrial robot. The plasma-spray conditions can be found else-
where [9]. A free standing, plasma-sprayed ceramic billet was then obtained by burning away the graphite
substrate at 680 °C in air for 24 h. The billet was machined into the final, rectangular flexure test specimen
with nominal dimensions of 3 by 4 by 50 mm, respectively, in width, depth, and length. The 3-mm-wide face
of flexure test specimens corresponded to the plane perpendicular to the plasma spraying direction. Major
physical and mechanical properties of as-sprayed coating material at ambient temperature [10,11] are
shown in Table 1. Fig. 1 shows a typical fracture surface and a polished surface showing the microstructure
of as-processed coatings, in which large amounts of microcracks and pores are characterized in conjunction
with a platelet (or splat) structure.

2.2. Mixed-mode testing

2.2.1. Preparation of sharp precracks

Sharp v notches were introduced in flexure test specimens, using the single-edge-v-notched-beam method
[12]. This method utilizes a razor blade with diamond paste to introduce a sharp root radius by tapering a
saw cut. Sharp v-notch radii ranging from 4 to 6 lm have been successfully obtained for alumina, glass
ceramic, silicon nitride, zirconia, and silicon carbide ceramics [12]. A starter straight-through notch
0.6 mm deep and 0.026 mm wide was made on the 3-mm-wide face of the test specimens. A steel razor blade
was put into the starter notch sprinkled with diamond paste with a particle size of 9 lm. Typically, a load of
2146 S.R. Choi et al. / Engineering Fracture Mechanics 72 (2005) 2144–2158

Table 1
Typical physical and mechanical properties of plasma-sprayed ZrO2–8wt%Y2O3 thermal barrier coatings at ambient temperature
[10,11]
Density (g/cm3) Elastic modulusa (GPa) PoissonÕs ratio Strength
Type of tests Average strength
(MPa)
5.22 10–35 0.1–0.2 Tension 11–15 (2)b
Compression 300 (77)
Four-point flexure 33 (7)
Biaxial flexure 40 (4)
a
The range refers to from tension to compression as measured by strain gaging.
b
The number in parenthesis represents ±1.0 standard deviation.

Fig. 1. Typical microstructures of plasma-sprayed ZrO2–8wt%Y2O3 thermal barrier coatings: (a) polished surface (a plane normal to
plasma spraying); (b) fracture surface with an arrow indicating plasma-spray direction.

about 10 N was applied through the razor blade with a stroke rate (back-and-forth motion) of 1 Hz and a
stroke distance of about 13 mm in a specially designed polishing machine [10]. The final notch depth and
root radius were 2.0 mm and 20–50 lm, respectively, resulting in the crack size-to-specimen depth ratio of
0.5 (= a/W). The coating material resulted in less sharpness in root radius due to its porous and micro-
cracked nature, compared to typical dense ceramics. However, it has observed that the sharpness ranging
from 20 to 50 lm was sufficient to give a consistent and accurate value of fracture toughness of the coating
material [10]. Note that the through-the-thickness sharp notches thus prepared were aligned parallel with
respect to the plasma spraying direction.
Another method to generate sharp precracks, the single-edge-precracked-beam (SEPB) technique [13],
was not feasible for the coating material: The response of the coatings to Vickers indentation was poor be-
cause of the materialÕs significant porosity (about 10%), microcracks, so that starter indent cracks that were
used to generate precracks were not achieved. Furthermore, the SEPB technique was not appropriate in
elevated-temperature testing because of a phenomenon of crack healing due to oxidation combined with
a very small crack opening displacements, which can result in a significant overestimation of fracture tough-
ness value.

2.2.2. Asymmetric four-point flexure testing

The sharp v-notched flexure test specimen was loaded in asymmetric four-point flexure as shown in Fig.
2. The ratio of mode I to mode II stress intensity factor (SIF), KI/KII was varied by varying the distance of
 S.R. Choi et al. / Engineering Fracture Mechanics 72 (2005) 2144–2158 2147

B A
b

W A S B

Fig. 2. Schematic of asymmetric four-point flexure geometry with accompanying shear force (V) and bending moment (M) diagrams.

the precrack from the center plane, s, as shown in the figure. When the precrack was centered with respect
to the loading point (s = 0), the precrack was subjected to pure mode II loading. As s was increased, KI/KII
also increased. The stress intensity factors, KI and KII are expressed as
a
1=2
K I ¼ rðpaÞ F I ð1Þ
W
a
K II ¼ sðpaÞ1=2 F II ð2Þ
W
where r is the applied (remote) normal stress, s is the applied (remote) shear stress, a is the crack size, and
W is the specimen depth. FI and FII are crack geometry factors in modes I and II, respectively. The normal
and shear stresses are given from the elementary beam theory by
A
B 6sP
r¼ ð3Þ
A þ B bW 2
A
B P
s¼ ð4Þ
A þ B bW
where A and B are distances from a load point as defined in Fig. 2, b is the specimen width, and P is the
applied force. The values of A = 10 mm and B = 5 mm were typically used, but in some cases the values of
A = 12 and B = 6 mm were used for comparison and verification. The test specimenÕs cross section was
W = 4 mm and b = 3 mm.
Several different expressions of FI and FII were suggested by Suresh et al. [14], Wang et al. [15], and He
and Hutchinson [16], and their respective results of FI and FII are illustrated in Fig. 3. As seen from the
figure, no significant difference in FI or FII between the solutions was found, particularly at a/W 6 0.5.
The solution by He and Hutchinson [16] that provided a convenient polynomial expression was used,
and is given as follows:
a a  a 2  a 3  a 4  a 5
FI ¼ 1:122
1:121 þ 3:740 þ 3:873
19:05 þ 22:55
W W W W W W
a
for 6 0:7 ð5Þ
W
2148 S.R. Choi et al. / Engineering Fracture Mechanics 72 (2005) 2144–2158

4
FI ; Suresh
Asymmetric four-point flexure
FII ; Suresh KI = σ(πa)1/2FI
FI ; Wang KII = τ(πa)1/2FII
3 FII ; Wang
FI ; He
FII ; He
FI ; Srawley

FI, FII
2

FI
1
F
II

0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
a/W

Fig. 3. Various solutions [14–16] of crack geometry factors FI and FII applied in the asymmetric four-point flexure configuration.

a  a  a 2  a 3  a 4 
a=W
F II ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 7:264
9:37 þ 2:74 þ 1:87
1:04
W pð1
a=W Þ W W W W
a
for 0 6 61 ð6Þ
W
Note that Eq. (5) was quoted from MurakamiÕs [17] and was almost identical to Srawly and GrossÕ solution
[18] up to a/W = 0.7, as seen in the figure. Also note that the difference in FI and FII between solutions in the
range of a/W = 0.4 to 0.5 was negligible, resulting in
F I  F II ð7Þ
Hence, in the case of a/W = 0.4 to 0.5, the KII/KI ratio (or called ÔmixityÕ) can be simplified from Eqs. (1)–(4)
and (7) as follows:
K II WF II W
¼  ð8Þ
KI 6sF I 6s
   
K II W
b ¼ tan
1  tan
1 ð9Þ
KI 6s
where b is the mixity angle. The angle b was also called the Ôequivalent crack angleÕ by Maccagno and Knott
[19] for situations which do not actually use the inclined crack configuration. The ratio of crack size to spec-
imen width used in this work was chosen to be a/W = 0.5 so that Eq. (8) or (9) could be conveniently used.
The values of s = 0 to 3.6 mm were selected to give uniformly distributed b between pure mode I (b = 0°)
and pure mode II (b = 90°) based on Eq. (9). Pure mode I fracture toughness KIc was determined using a
symmetric four-point flexure fixture with 10-mm inner and 20-mm outer distances.
Each test specimen was placed in a SiC test fixture to ensure that each specified distance s be within a
tolerance of ±0.05 mm by using a micrometer attachment. No experimental attempt was made to see the
effect of variation of s on test results. All testing was conducted in displacement control with an actuator
speed of 0.5 mm/min of an electromechanical test frame (Model 8562, Instron, Canton, MA). Testing was
conducted at 25 and 1316 °C in air. Each test specimen in elevated-temperature testing was heated at a rate
of 20 °C/h and held for thermal equilibration for about 15 min prior to testing. At room temperature, four
and nine specimens were tested for modes I and II, respectively, whereas at 1316 °C, four specimens were
 S.R. Choi et al. / Engineering Fracture Mechanics 72 (2005) 2144–2158 2149

1.4
TBCs/Asymmetric
1.2
RT; A/B=10/5 mm
RT; A/B=12/6 mm
1.0 RT; A/B=5/2.5 mm
1316oC; A/B=10/5 mm

KII [MPa m1/2]

0.8

0.6

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

KI [MPa m1/2]

Fig. 4. A summary of KII versus KI for mode I, mode II, and mixed mode fracture of plasma-sprayed ZrO2–8wt%Y2O3 thermal barrier
coatings at both 25 and 1316 °C in air.

tested for each mode. There were nine mixity values associated with the specimens tested at 25 °C, and four
mixity values associated with those tested at 1316 °C. Typically, two or three specimens of each mixity value
were tested. After testing, the crack size of each tested specimen was determined optically from its fracture
surface based on the three-point measurements in accordance with test standard ASTM C1421 [13]. The
crack propagation angles were also determined.

3. Results and discussion

3.1. Test results

The results of the mode I, mode II, and mixed-mode testing for the coating material at both 25 and
1316 °C are presented as KII versus KI in Fig. 4. Each point represents a single datum. The effect of test
span pdistance
ffiffiffiffi on ÔnominalÕ KIIc1 at 25 °C was found to be insignificant: KIIc = 0.74 ± 0.1 and 0.69 ± 0.09
MPa m, respectively, for A/B values of 10/5 and 12/6 mm (see Fig. 2). Hence, the 10/5-mm distances were
exclusively used throughout
pffiffiffiffi the mixed-mode test program. At 25 °C, KIc = 1.15 pffiffiffiffi± 0.07 and
KIIc = 0.73 ± 0.10 MPa m, and at 1316 °C, KIc = 0.98 ± 0.13 and KIIc = 0.65 ± 0.04 MPa m. Note that
the values of nominal KIIc were 37% and 34% lower than those of KIc at 25 and 1316 °C, respectively.
The coefficient of variation in KIc or nominal KIIc ranged from 6% to 14%, somewhat greater than that ob-
served from dense ceramics [23,24] which is commonly less than 10%. As typical of many ceramics subjected
to ÔfastÕ-fracture testing at 0.5 mm/min, the force versus time trace for most test specimens was linear,
regardless of mixity at either 25 or 1316 °C.
As shown in Fig. 4, because KIIc < KIc the overall KII versus KI relation is represented by an ellipse at
either 25 or 1316 °C with KIc being a major axis. For a given mixity angle ranging from b = 0 to 90°,

1
There is a notion that a crack would not propagate under the influence of KII to reach the critical value of KIIc for homogeneous
materials, making a definition of KIIc ineffective. However, there is another notion that it would be sensible to define a KIIc as the value
at which growth initiates under mode II loading, as appeared in the literature (e.g., [14,20–29]). In this paper, KIIc is termed purposely
ÔnominalÕ mode II fracture toughness (or nominal KIIc) to imply in part both of the views.
2150 S.R. Choi et al. / Engineering Fracture Mechanics 72 (2005) 2144–2158

Fracture toughness, KIc, KIIc [MPa m1/2]

3

KIc: prevoius study [10]

KIc: current study
KIIc: current study
2

Average KIc [=1.0]

Average KIIc [=0.7]

0
0 200 400 600 800 1000 1200 1400 1600
Temperature [oC]

Fig. 5. Fracture toughnesses of KIc and nominal KIIc as a function of temperature for plasma-sprayed ZrO2–8wt%Y2O3 thermal barrier
coatings. The previous KIc data [10] were included. The error bars represent ±1.0 standard deviation.

KEQ [¼ ðK 2I þ K 2II Þ1=2 ] was lower at 1316 °C than at 25 °C. The values of KIc and nominal KIIc were decreased
by 15% and 11% , respectively, when temperature was increased from 25 to 1316 °C. However, this decrease
was marginal, as can be seen in a summary of KIc and nominal KIIc as a function of temperature in Fig. 5, in
which the KIc data previously determined [10] at 25 and 800 °C by the SEVNB method werepincluded ffiffiffiffi with
the current data. KIc remains statistically invariable with an average of KIc = 1.0 ± 0.1 MPapffiffiffiffim, regardless
of test temperature up to 1316 °C. The same would be true for nominal KIIc (= 0.69 MPa m) as depicted
in the figure, notwithstanding the lack of intermediate-temperature data. The insensitivity of KIc to temper-
ature was also observed in a similar but dense (hot-pressed) 10 mol% yttria-stabilized zirconia (10-YSZ) in
which fracture toughness of the pffiffiffiffimaterial—evaluated in air by the same SEVNB method—was almost
unchanged (KIc = 1.6–1.8 MPa m) from 25 to 1000 °C [30].
It should be mentioned that the effect of sintering on fracture toughness during elevated-temperature
testing was found to be negligible based on a sintering study showing that fracture toughness increase of
pffiffiffiffi
the coating material was about 0.12 MPa m/h for the first 5-hour annealing at 1316 °C in air [31]. This
indicates that fracture toughness increase during testing (<20 min) at 1316 °C would be only about 3%.
However, care must be exercised in interpreting fracture toughness when higher test temperature and longer
test duration are used because of sintering effect. This caution applies to any of mechanical testing for plas-
ma-sprayed TBCs at elevated temperatures.
Fig. 6 shows typical examples of crack propagation with different mixities ranging from pure mode II
(KI/KII = 0) to pure mode I (KI/KII = 1). The angle of crack propagation hc was defined as an angle of
the crack plane propagated with respect to the original precrack plane. The direction of crack plane prop-
agated was in most cases straight with some exceptions such as tortuous or not-straight paths. The angle of
crack propagation was greatest in pure mode II, and decreased with increasing KI/KII reaching to hc = 0 in
pure mode I. A summary of experimental data on the angle of crack propagation as a function of KI/KII
determined at both 25 and 1316 °C is depicted in Fig. 7. The overall trend of the angle of crack propagation
was that hc initially decreased quickly with increasing KI/KII and then decreased monotonically at
KI/KII > 3. No significant difference in hc between 25 and 1316 °C was observed. It is noted that the scatter
in hc, for example, hc = 45° to 70° at KI/KII = 0 (pure mode II), seemed more appreciable in the coating
material than in dense advanced ceramics [23], which exhibited a typical scatter of about 15°.
Analysis of fracture surfaces revealed no explicit difference in the fracture morphology of the pure mode
I, mixed-mode, or pure mode II specimens. However, it should be mentioned that unlike dense ceramics the
 S.R. Choi et al. / Engineering Fracture Mechanics 72 (2005) 2144–2158 2151

Fig. 6. Typical examples of specimens showing the angles of crack propagated with respect to KI/KII of plasma-sprayed ZrO2–
8wt%Y2O3 thermal barrier coatings: (a) pure mode II; KI/KII = 0, (b) KI/KII = 2.7, (c) KI/KII = 5.5, and (d) pure mode I; KI/KII = 1.

90
Crack propagation angle, θc [deg]

80
25ºC
70 1316ºC

60

50
40

30
20
10

0
0 1 2 3 4 5 6 7 8
KI/KII

Fig. 7. Crack propagation angles as a function of KI/KII of plasma-sprayed ZrO2–8wt%Y2O3 thermal barrier coatings determined in
asymmetric four-point flexure at both 25 and 1316 °C in air.

coating material has provided a challenging subject on fractography because of its open and microcracked
platelet microstructure. The origin or nature of strength-controlling flaws and the mode of fracture in the
coating material has rarely been identified with any traditional means of fractography (optical and/or SEM)
that would be otherwise sufficient for the case of dense ceramics [11]. An improved means of fractography
is required to better distinguish fracture morphologies in response to different modes of fracture of the
coating material.

3.2. Consideration of mixed-mode fracture criteria

3.2.1. Considered criteria

The noncoplanar crack propagation, as shown in Fig. 7, requires different fracture criteria, typically a
combination of the driving forces KI and KII together with KIc and nominal KIIc. The mixed-mode fracture
criteria for an isotropic, homogeneous material can be broadly categorized in five areas and will be briefly
described then compared with experimental results.
2152 S.R. Choi et al. / Engineering Fracture Mechanics 72 (2005) 2144–2158

3.2.1.1. Maximum principal stress criterion [32]. This criterion assumes that a crack at instability propa-
gates in the direction normal to the maximum principal stress direction. The instability condition is given
by the following relationship:
hc hc hc
K Ic
K I cos3 þ 3K II cos2 sin ¼ 0 ð10Þ
2 2 2
The angle of crack propagation, hc, is a function of KI/KII and is given by
2 ffi3
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2
KI 1 KI
hc ¼ 2tan
1 4  þ 85 ð11Þ
4K II 4 K II

In case of pure mode II, the following relations can be obtained from Eqs. (10) and (11):
K IIc
hc ¼ 70:5 ; ¼ 0:866 ð12Þ
K Ic

3.2.1.2. Minimum strain energy density criterion [33]. This criterion assumes that crack propagation would
take place in the direction along which the strain energy density is a minimum. This assumption gives rise
to the following mixed-mode fracture criterion:
1
K 2Ic
½a11 K 2I þ 2a12 K I K II þ a22 K 2II  ¼ 0 ð13Þ
2ðj
1Þ

where
a11 ¼ ð1 þ cos hc Þðj
cos hc Þ
a12 ¼ sin hc ð2 cos hc
j þ 1Þ
a22 ¼ ðj þ 1Þð1
cos hc Þ þ ð1 þ cos hc Þð3 cos hc
1Þ
with
j ¼ 3
4m for plain strain

j ¼ ð3
mÞ=ð1 þ mÞ for plain stress

where m is PoissonÕs ratio. The angle of crack propagation, hc, is given by

    
KI KI
ðj
1Þ sin hc
2tan
1
2 sin 2 hc
tan
1
sin 2hc ¼ 0 ð14Þ
K II K II
In case of pure mode II, the following relations can be obtained from Eqs. (13) and (14):
   1=2

1 j
1 K IIc 2ðj
1Þ
hc ¼ cos ; ¼ ð15Þ
6 K Ic a22

3.2.1.3. Maximum energy release rate criterion. This criterion assumes that a crack with an infinitesimally
small kink at an arbitrary direction propagates in the direction along the maximum strain energy release
rate [34–37]. Many investigators keep that the strain energy release rate, G, can be expressed in terms of
KI and KII as follows:
G ¼ c11 K 2I þ c12 K I K II þ c22 K 2II ð16Þ
 S.R. Choi et al. / Engineering Fracture Mechanics 72 (2005) 2144–2158 2153

where c11, c12, and c22 are coefficients. Hyashi and Nemeat-Nasser [36] determined numerically the coeffi-
cients and the direction of maximum G as a function of KI/KII. Palaniswamy and Knauss [35] proposed an
approximated mixed-mode fracture criterion on the maximum energy release rate as follows (that was also
used by Suresh et al. [14]):
 2
K I 3 K II
þ ¼1 ð17Þ
K Ic 2 K Ic

where for pure mode II the approximation leads to

K IIc
hc ¼ 77:4 ; ¼ 0:817 ð18Þ
K Ic
The fracture criterion and the crack propagation angle on the maximum G were used based on Eqs. (16)
and (17) in this work.

3.2.1.4. Empirical criterion [21,26,38]. The empirical fracture criterion can be expressed by the following
generalized, simplified form:
 p  q
KI K II
þ ¼1 ð19Þ
K Ic cK Ic

where p and q are parameters (typically p, q = 1–2) to be determined from experimental data through a
curve fitting, and
K IIc
c¼ ð20Þ
K Ic
with 0 < c < 2, an experimentally determined value. Despite the lack of its theoretical frame work in some
cases, the empirical fracture criterion can still provide a much simplified representation (as database, too) of
mixed-mode behavior of a material particularly when other (preceding) fracture criteria do not describe rea-
sonably well the actual materialÕs behavior under mixed mode. The maximum G criterion (Eq. (16)) or the
coplanar crack propagation criterion [39], in fact, is one particular type of the empirical fracture criterion in
its form.

3.2.1.5. Stress intensity factors approach with infinitesimal kinks [40]. This approach is to regard an infin-
itesimal kink emanating from the tip of a straight main crack under mixed mode loading. The resulting
stress intensity factors kI and kII at the tip of the kink are expressed as a function of the stress intensity
factors KI and KII of the main crack as

k I ¼ R11 ðaÞK I þ R12 ðaÞK II

ð21Þ
k II ¼ R21 ðaÞK I þ R22 ðaÞK II

where a is the angle between the main crack and the kink, and Rij are values determined numerically. In the
case of pure mode II loading, the maximum value of kI occurs at a = 76.6°, in which the following condi-
tions are reached provided that the crack grows under kI:

K IIc
hc ¼ 76:6 ; ¼ 0:81 ð22Þ
K Ic
which is similar to the case of the maximum energy release rate criterion.
2154 S.R. Choi et al. / Engineering Fracture Mechanics 72 (2005) 2144–2158

3.2.2. Comparisons with experiment

Fracture locus. Fig. 8 compares the afore-reviewed four different mixed-mode fracture criteria (except
Section 3.2.1.5) with the experimental data, where KII normalized with respect to KIc, KII/KIc, was plotted
as a function of KI/KIc for both 25 and 1316 °C data. The prediction for the minimum strain energy density
criterion was made using a value of m = 0.2. The maximum principal stress criterion seemed to describe rea-
sonably the ambient-temperature experimental data. However, only the empirical criterion was in good
agreement with both ambient and elevated-temperature data, resulting in the following parameters:
p ¼ 1; q ¼ 2; c ¼ 0:65 for 25  C
ð23Þ
p ¼ 1; q ¼ 1:3; c ¼ 0:65 for 1316  C
Previous studies on dense ceramics such as silicon nitrides and alumina [23] and ceria-doped zirconia [24]
showed that the minimum strain energy density criterion resulted in best agreement among the mixed-mode
criteria mentioned above, as that was the case in another study of alumina by Suresh et al. as well [14]. There
exists a remarkable contrast in mixed-mode behavior between the dense and the porous TBC ceramics.
KIIc/KIc ratio. The three mixed-mode fracture criteria (the maximum principal stress, minimum strain
energy density, and maximum energy release rate criteria) as seen from Fig. 8 and the stress intensity factor
approach with infinitesimal kinks all predict the values of KIIc/KIc = 0.8 to 1.1. By contrast, a lower value of
KIIc/KIc  0.65 resulted in the coating material at both temperatures. The value of KIIc/KIc would be the
parameter with the largest influence on the degree of agreement with the prediction of the fracture locus
pattern. In case of KIIc/KIc  1, statistically reasonable agreement would be found in any of the major
mixed-mode fracture criteria mentioned. Metals and PMMA (polymethylmethacrylate) exhibited KIIc/KIc
ratios in the range of 1.0–1.2 with various test techniques [19,27–29]. It is postulated that the lower value
of KIIc/KIc for the TBCs may be attributed to several factors such as different grain/particle orientation of
the splats, weak interface bonding between splats, and impurities or undesirable microstructure along inter-
faces between splats which are formed by plasma-spraying process, resulting in microscopic directionality
and inhomogeneity of the material.
A summary of data on the ratio of nominal KIIc to KIc for various brittle materials by Munz and Fett [21]
as well as data on dense ceramics from previous studies [23,24] was used and compared with the TBC data
obtained in this work, as shown in Fig. 9. The values of KIIc/KIc for the dense ceramics two silicon nitrides

1.4
Experiment
25ºC
1.2 1316ºC

Min. energy density (υ = 0.2)

1.0
Max. principal stress
KII /KIc

0.8
Max. energy release rate

0.6

0.4
HT
0.2 RT
Empirical

0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
KI /KIc

Fig. 8. Plots of KI/KIc as a function of KI/KIc for mixed-mode fracture criteria and experimental data determined for plasma-sprayed
ZrO2–8wt%Y2O3 thermal barrier coatings.
 S.R. Choi et al. / Engineering Fracture Mechanics 72 (2005) 2144–2158 2155

2.0

1.5

KIIc/KIc
1.0

0.5
(This study)

0.0
TBC Al2O3 Si3N4 ZrO2 Glass Graphite
(ZrO2-Y2O3)

Fig. 9. Comparisons of KIIc/KIc of plasma-sprayed ZrO2–8wt%Y2O3 thermal barrier coatings with other brittle materials. Data from
previous studies [23,24] and from Munz and Fett [21].

(one with fine and one, elongated grains) [23], coarse grained alumina [23], and ceria-doped tetragonal zir-
conia polycrystal (Ce-TZP) [24], all tested in asymmetric four-point flexure using naturally sharp precracks,
were invariably around KIIc/KIc  1 and are much greater than those of the TBCs (KIIc/KIc  0.65). How-
ever, as seen from the figure, a large variation of KIIc/KIc varying from 0.6 to 2.0 is noted depending on
materials, test specimens, notch or precrack preparations, and even investigators, etc. As a result of this
significant variation, a reasonably unified trend on the value of KIIc/KIc cannot be made for the brittle mate-
rials provided. However, it should be noted that the value of KIIc/KIc for the TBCs evaluated in this work is
still at the lower end of the data pool.
Crack propagation angle. The data on crack propagation angle shown in Fig. 7 were compared with the pre-
dictions made with the three mixed-mode fracture criteria, and the results are presented in Fig. 10. The min-
imum strain energy density criterion overall seemed to yield a better prediction compared with the other two
criteria. However, because of somewhat significant scatter in hc as well as little difference in prediction between
the criteria, it is difficult to state which criterion gives the best agreement with the experimental data. Note that

90 Experiment
Crack propagation angle, θc [deg]

25ºC
80
1316ºC
70
Max. strain energy release rate
60
Max. principal stress
50

40

30
20
Min. strain energy density
10

0
0 1 2 3 4 5 6 7 8
KI/KII

Fig. 10. Plots of crack propagation angle as a function of KI/KII for mixed-mode fracture criteria and experimental data determined
for plasma-sprayed ZrO2–8wt%Y2O3 thermal barrier coatings.
2156 S.R. Choi et al. / Engineering Fracture Mechanics 72 (2005) 2144–2158

a significant discrepancy was found particularly in the region close to pure mode II, e.g., KI/KII < 1. This dis-
crepancy predominant in the region of KI/KII < 1 was also observed in the previous studies with silicon nitrides
and alumina [23] and other grade alumina by Suresh et al. [14]. By contrast, this discrepancy seemingly unique
in ceramics was not found in amorphous glassy polymer PMMA mixed-mode tested in either asymmetric four-
point flexure [19] or uniaxial tension [33] configuration. A good agreement of propagation angle with the frac-
ture criteria (as well as little scatter in hc) in a wide range of KI/KII close to pure mode II was also found in steel
tested in asymmetric four-point flexure at
196 °C to induce brittle fracture [22].

3.3. Source of discrepancy/error

Discrepancy encountered in this work between theories and experimental data on fracture locus, KIIc/KIc
ratio, and crack propagation angle may stem from two plausible sources: the major one would be from
the materialÕs microscopic inhomogeneity; the second may be from the experimental errors. Although
the material exhibited little directionality at least in response to KIc [31], the material is still considered
microscopically inhomogeneous due to directionally oriented splat, lamellar-like open structure by the
plasma-spraying process (see Fig. 1). Microscopic inhomogeneity of the coating material has been identified
through its strength, constitutive relations, and microhardness in which significantly greater variations in
those properties were typified, as compared with dense, or hot-pressed brittle counterparts [31].
The second source of discrepancy may come from the experimental errors such as force and crack size
measurements and the crack positioning relative to the specific location. The total absolute error, Ea, affect-
ing the mixity, can be analyzed from Eq. (8) as
       
 of   of   of  of 

Ea ¼   
DW  þ   
DF I  þ  DF II  þ  Ds
  ð24Þ
oW oF I oF os II

where f(W, FI, FII, s) = KII/KI, and DW, DFI, DFII and Ds are errors associated with W, FI, FII and s, respec-
tively. Since the error in W, FI or FII is typically less than 1%, the most contributing error is due to s so that
Eq. (24) can be approximated as follows:
 
of 
Ea   Ds
 ð25Þ
os
Fig. 11 shows the results of Ea as a function of variation in s (= Ds/s) for different values of s. The mixity
error is negligible for larger s values. However, the error is greatly augmented when s becomes smaller. In

0.8
W = 4 mm
0.7 a/W = 0.5 s = 0.1
0.6 s = 0.2
Error in KII/KI

0.5

0.4 s = 0.3

0.3
s = 0.5
0.2
s = 1.0
0.1
s = 2.0
s = 4.0
0.0
0.00 0.05 0.10 0.15 0.20 0.25
Variation in s, ∆s/s

Fig. 11. Error in mixity as a function of variation in s for different values of s (in mm), determined from Eq. (25).
 S.R. Choi et al. / Engineering Fracture Mechanics 72 (2005) 2144–2158 2157

other words, the mixity error increases significantly as loading approaches pure mode II. Even small var-
iation or inaccurate measurement in s will result in a significant error in mixity particularly when
s 6 0.2 mm. In this case, an experimental error of about <30% for s = 0.1 mm may be inevitable although
s is maintained even within a 5% tolerance, which would be the case encountered in this work. Hence, spe-
cial care must be exercised in positioning the crack accurately at a specified position within a less-than 1%
deviation—although it is extremely difficult to achieve with such tiny ceramic specimens used—when pure
mode II loading becomes predominant (s 6 0.1). However, it is noted that the mixity is not greatly affected
by KII but by KI as seen from Eqs. (1)–(4), thus influencing little in nominal KIIc.

4. Conclusions

A full range of mixed-mode fracture behavior covering mode I and mode II for plasma-sprayed ZrO2–
8wt%Y2O3 thermal barrier coating was determined in air at 25 and 1316 °C in asymmetric four-point flexure
in conjunction with the single edge v-notched beam (SEVNB) method. The following conclusions were made:
pffiffiffiffi
1. The mode I fracture toughness was found to be KIc = 1.15 ± 0.07 and 0.98 ± 0.13 MPa m, respectively,
at 25 and 1316 °C. pffiffiffiffi The respective nominal mode II fracture toughness was KIIc = 0.73 ± 0.10 and
0.65 ± 0.04 MPa m. Hence, the coating material exhibited an insignificant difference in either KIc or
(nominal) KIIc between 25 and 1316 °C, whereas it exhibited noticeable difference between KIc and (nom-
inal) KIIc resulting in KIIc/KIc = 0.65 at both temperatures. The lower value of KIIc/KIc would be due to
the materialÕs microscopic inhomogeneity, typified through plasma-spraying process.
2. The empirical mixed-mode fracture criterion was in best agreement with the coatingsÕ mixed-mode frac-
ture behavior among the four mixed-mode fracture theories considered. The angle of crack propagation
was in reasonable agreement with the minimum strain energy density criterion.

References

[1] Miller RA. Current status of thermal barrier coatings—an overview. Surf Coat Tech 1987;30:1–11.
[2] Miller RA. Thermal barrier coatings for aircraft engines—history and direction. In: Brindley WJ, editor. NASA CP-3312,
National Aeronautics and Space Administration, Glenn Research Center, Cleveland, OH, 1995. p. 17–34.
[3] Yonushonis TM. Thick thermal barrier coatings for diesel components. NASA CR-187111, National Aeronautics and Space
Administration, Glenn Research Center, Cleveland, OH, 1991.
[4] Tsui YC, Clyne TW. Adhesion of thermal barrier coating systems and incorporation of an oxidation barrier layer. In: Berndt CC,
editor. Thermal spray: practical solutions for engineering problems. Proc 9th Natl Thermal Spray Conf, ASM International,
Cincinnati, Ohio, 1996. p. 275–84.
[5] Shaw LL, Barber B, Jordan EH, Gell M. Measurement of the interfacial fracture energy of thermal barrier coatings. Scr Mater
1998;39:1427–34.
[6] Thurn G, Schneider GA, Bahr HA, Aldinger F. Toughness anisotropy and behavior of plasma sprayed ZrO2 thermal barrier
coatings. Surf Coat Tech 2000;123:147–58.
[7] Wesling KF, Socie DF, Beardsley B. Fatigue of thick thermal barrier coatings. J Am Ceram Soc 1994;77(7):1863–8.
[8] Callus PJ, Berndt CC. Relationship between the mode II fracture toughness and microstructure of thermal spray coatings. Surf
Coat Tech 1999;114:114–28.
[9] Zhu D, Miller RA. Influence of high cycle thermal loads on thermal fatigue behavior of thick thermal barrier coatings. NASA
TP3676, National Aeronautics and Space Administration, Glenn Research Center, Cleveland, OH, 1997.
[10] Choi SR, Zhu D, Miller RA. High-temperature slow crack growth, fracture toughness and room-temperature deformation
behavior of plasma-sprayed ZrO2–8wt%Y2O3. Ceram Engng Sci Proc 1998;19(4):293–301.
[11] Choi SR, Zhu D, Miller RA. Deformation and strength behavior of plasma-sprayed ZrO2–8 wt%Y2O3 thermal barrier coatings in
biaxial flexure and trans-thickness tension. Ceram Engng Sci Proc 2000;21(4):653–61.
2158 S.R. Choi et al. / Engineering Fracture Mechanics 72 (2005) 2144–2158

[12] (a) Kübler J. Fracture toughness of ceramics using the SEVNB method: preliminary results. Ceram Engng Sci Proc
1997;18(4):155–62; (b) Kübler J. Fracture toughness of ceramics using the SEVNB method; round robin. VAMAS Report No. 37,
EMPA, Swiss Federal Laboratories for Materials Testing & Research, Dübendorf, Switzerland, 1999.
[13] ASTM C 1421. Test method for determination of fracture toughness of advanced ceramics at ambient temperature. In: Annual
book of ASTM standards, vol. 15.01. West Conshohocken, PA: American Society for Testing and Materials; 2003.
[14] Suresh S, Shih CF, Morrone A, OÕDowd NP. Mixed-mode fracture toughness of ceramic materials. J Am Ceram Soc
1990;73(5):1257–67.
[15] Wang KJ, Lin HC, Hua K. Calculation of stress intensity factors for combined mode bend specimens. In: Taplin MDR, editor.
Advances in research on the strength and fracture of materials, vol. 4. Waterloo, Canada: ICF4; 1977. p. 123–33.
[16] He MY, Hutchinson JW. Asymmetric four-point crack specimen. J Appl Mech 2000;67:207–9.
[17] Murakami Y, editor. Stress intensity factors handbook, vol. 1. New York: Pergamon Press; 1987.
[18] Srawley JE, Gross B. Side-cracked plates subjected to combined direct and bending forces. In: Cracks and fracture. ASTM STP
601. Philadelphia: American Society for Testing and Materials; 1976. p. 559–79.
[19] Maccagno TM, Knott JF. The fracture behavior of PMMA in mixed modes I and II. Engng Fract Mech 1989;34(1):65–86.
[20] Broek D. Elementary engineering fracture mechanics. Boston: Martinus Nijhofff Publisher; 1983. p. 359–65.
[21] Munz D, Fett T. Ceramics. Berlin, Germany: Springer-Verlag; 1999. p. 45–9.
[22] Maccagno TM, Knott JF. The low temperature brittle fracture behavior of steel in mixed modes I and II. Engng Fract Mech
1991;38(2-3):111–28.
[23] Tikare V, Choi SR. Combined mode I and mode II fracture of monolithic ceramics. J Am Ceram Soc 1993;76(9):2265–72.
[24] Tikare V, Choi SR. Combined mode I–mode II fracture of 12-mol%-ceria-doped tetragonal zirconia polycrystalline ceramic. J Am
Ceram Soc 1997;80(6):1624–6.
[25] Petrovic JJ. Mixed-mode fracture of hot-pressed Si3N4. J Am Ceram Soc 1985;68(6):348–55.
[26] Awaji H, Sato S. Combined-mode fracture toughness measurements by the disk test. J Engng Mater Technol 1978;100:175–82.
[27] Pook LP. The effect of crack angle on fracture toughness. Engng Fract Mech 1971;3:205–18.
[28] Jones DL, Chisholm DB. An investigation of the edge-sliding mode in fracture mechanics. Engng Fract Mech 1975;7:261–70.
[29] Shah RC. Fracture under combined modes in 4340 steel. In: Fracture analysis. ASTM STP 560. Philadelphia, PA: American
Society for Testing and Materials; 1974. p. 29–52.
[30] Choi SR, Bansal NP. Flexure strength, fracture toughness, and slow crack growth of YSZ/alumina composites at elevated
temperature. J Am Ceram Soc, in print 2005.
[31] Choi SR, Zhu D, Miller RA. Effect of sintering on mechanical and physical properties of plasma-sprayed thermal barrier coatings.
NASA/TM-2004-212625, National Aeronautics & Space Administration, Glenn Research Center, Cleveland, OH, 2004.
[32] Erdogan F, Sih GC. On the crack extension in plates under plane loading and transverse shear. J Basic Engng 1963;85:519–27.
[33] Sih GC. Strain-energy-density factor applied to mixed mode crack problems. Int J Fract 1974;10(3):305–21.
[34] Niusmer RJ. An energy release rate criterion for mixed mode fracture. Int J Fract 1974;10(3):305–21.
[35] Palaniswamy K, Knauss WG. On the problem of crack extension in brittle solids under general loading. In: Nemat-Nasser S.
editor, Mechanics today, vol. 4. New York: Pergamon Press; 1978. p. 87–148.
[36] Hyashi K, Nemat-Nasser S. Energy-release rate and crack kinking under combined loading. J Appl Mech 1981;48:520–4.
[37] Ueda Y, Ikeda K, Yoa T, Aoki M. Characteristics of brittle fracture under general combined modes including those under biaxial
tensile loads. Engng Fract Mech 1983;18(6):1131–58.
[38] Shetty DK. Mixed-mode fracture criteria for reliability analysis and design with structural ceramics. J Engng Gas Turbine Power
1987;109(7):282–9.
[39] Paris PC, Sih GC. Stress analysis of cracks. In: ASTM ATP 381. Philadelphia, PA: American Society for Testing and Materials;
1965. p. 30–81.
[40] Melin S. Accurate data for stress intensity factors at infinitesimal kinks. J Appl Mech 1994;61:467–70.