Surface & Coatings Technology 382 (2020) 125025

Contents lists available at ScienceDirect

Surface & Coatings Technology

journal homepage: www.elsevier.com/locate/surfcoat

Transverse ridge cracking in tensile fragmentation tests of thin brittle T

coatings on polymer substrates
Marcus Vinícius Tavares da Costaa, Jessica Bolinssonb, Pierre Fayetc, E. Kristofer Gamstedta,∗
a
Uppsala University, Division of Applied Mechanics, Department of Engineering Sciences, Box 534, SE-751 21, Uppsala, Sweden
b
Tetra Pak AB, DSO Packaging Materials, Ruben Rausings gata, SE-22186, Lund, Sweden
c
Adhemon Sarl, Thin Technology, Avenue Edouard-Dapples 20, CH-1006, Lausanne, Switzerland

ARTICLE INFO ABSTRACT

Keywords: Thin brittle coatings on polymers can be used to improve barrier properties. The interfacial adhesion is known to
Interfacial energy release rate affect cracking of the coating under deformation. It is therefore desirable to experimentally quantify the adhesive
Ridge cracking properties. For thin coating systems, this can be done e.g. by tensile fragmentation tests or nanoindentation
Fragmentation test techniques. For crack opening, an alternative method presents itself in a secondary damage mechanism typically
Metal oxide coatings
found in tensile fragmentation tests. At large strains, the transverse compressive strains can give rise to ridge
Atomic layer deposition
cracks, which may be observed in high-resolution scanning electron microscopy. It is shown that such observations
can be used to estimate the critical energy release rate in interfacial crack growth. The method is exemplified for
20 nm thick coatings of TiO2 and coatings of mixed oxide TiO2 and Al2O3 on polyethylene terephthalate films. The
experiments gave values of 2.5 J/m2 and 1.6 J/m2 for these interfaces, respectively. Energy release rates from a
micromechanical model based on beam theory were found to be comparable with non-linear finite element pre-
dictions for the present crack configurations. Direct measurement of the transverse strains from digital image
correlation resulted in predictions closer to the more exact numerical results as compared with predictions based
on an assumed linear elastic Poisson contraction. The effects of the non-linear stress-strain behavior of the polymer
film could thus be dealt with by measuring the compressive strains giving rise to the ridge cracks. Overall, this
work shows how the ridge cracking can provide insight in interfacial crack growth in predominantly opening
mode, and possibly be used to roughly estimate fracture toughness even for ultrathin coatings in this investigation.

1. Introduction The coating-substrate adhesion, can be characterized by the energy

Ultrathin coatings, with thickness values on the nanometer scale,
are being increasingly used in a wide variety of applications. In de-
velopment of such coatings, it is of interest to characterize the adhesive
properties of the coating-substrate interface. Different characterization
methods have their specific merits and shortcomings. The purpose of
the present study is to show how to utilize the frequently occurring
ridge cracks that form during fragmentation tests to obtain a measure of
fracture toughness in primarily crack opening, which is otherwise not
easily obtained for thin coatings. Here, we consider in particular metal
oxide coating with a nanometer scale thickness on flexible polymer
substrates which is an interesting combination for food packaging ap-
plications. This combination provides an enhancement of the barrier
performance in the carton package [1]. A concern is the cracking of the
brittle coating when subjected to tension and bending in the manu-
facturing process. Such cracks can deteriorate the permeability.
released during crack growth at the interface. The critical energy re-
lease rate (or fracture toughness) G functions as a material property
under certain conditions [2]. In the coating mechanics community,
there is considerable effort to obtain both the cohesive Gc , when there is
a crack propagation across the coating, e.g. channelling cracking [3,4],
and the interfacial Gi , when the cracks cause delamination between
coating and substrate [5]. To measure the cohesive fracture toughness
Gc experimentally for ultrathin coatings, the fragmentation test [6] has
been used, examining channelling of parallel coating cracks, as well as
the nanoindentation test analysing a formation of ring-like cracks [7]
for Gc . For the interfacial fracture toughness , nanoindentation can be
used where detached chips can be observed by scanning electron mi-
croscopy (SEM) [8]. Microscratch tests have also been used to estimate
Gi [9]. Overall, nanoindentation is a widely used method for char-
acterizing the mechanical behavior of brittle coatings. Although the
principle of indentation tests is simple, i.e. load-displacement response

∗
Corresponding author.
E-mail address: kristofer.gamstedt@angstrom.uu.se (E.K. Gamstedt).

https://doi.org/10.1016/j.surfcoat.2019.125025
Received 1 July 2019; Received in revised form 25 September 2019; Accepted 26 September 2019
Available online 14 October 2019
0257-8972/ © 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/BY/4.0/).
M.V. Tavares da Costa, et al. Surface & Coatings Technology 382 (2020) 125025

under normal contact, analyzing the fracture processes can be chal-
lenging for concentrated contact loading, especially for measurements
with soft substrates, such as polymers [10,11].
In the previous work [12], in situ fragmentation tests were carried
out on atomic layer deposited (ALD) coatings of TiO2 (20 nm thick) and
mixed metal oxides of TiO2 and Al2O3 (20 nm thick) on a biaxially-
oriented polyethylene terephthalate (BoPET) film Dupont Teijin Me-
linex ST504 [13] to examine multiple cracking and subsequently to
calculate cohesive and adhesive properties, viz. the coating strength
distribution and the interfacial shear strength. During these tests, ridge
cracking was observed [11] in a direction transverse to the tensile
loading. It occurs through local interfacial delamination of the coating
from the substrate caused mainly by compressive deformation due to
the transverse Poisson contraction [14–16]. This damage mechanism,
termed ridge cracking or buckling delamination in the literature (e.g.
Hutchinson and Suo [17], Thouless [18] and Jensen [5] and etc.), can
be used to estimate the interfacial energy release rate Gi by measuring
the shape, crack opening and crack length of the detached coating.
Despite ridge cracking and buckling delamination occur in same
manner, they are inherently different from one another [18]. Ridged
cracking generally form for brittle coatings with low tensile strains to
failure, and come with, as the term indicates, cracks on top of the ridge.
For more ductile coatings with higher tensile strains to failure, buckling
delamination can occur, where are no cracks in the delaminated
coating. Buckling delaminations tend to be less straight than ridged
cracks, and can propagate in different morphological shapes, e.g. the
so-called telephone cord shape [19]. For utltrathin coatings [13], the
main challenge is that the ridge cracking occurs at the nanoscale, re-
quiring high resolution SEM for observations and measurements.
In this work, we continue to study the TiO2 and mixed oxide sam-
processing temperature closer to ambient conditions for ALD as com-
pared with e.g. chemical vapor deposition (CVD) [12]. In the tensile
fragmentation experiments [11], we observed ridge crack onset which
occurred at the saturation stage of the fragmentation process. The
opening of the fragmentation cracks can emanate from plastic shear
deformation (Kelly-Tyson case) of the substrate or debonding of the
coating. The latter case has been analyzed by Tarasovs et al. [25] to
estimate Gi in mode II. Here, focus is on the delamination primarily in
mode I of the perpendicular ridge cracks. This crack type was first
analyzed by Andersons et al. [20] to estimate Gi from experimental
observations in coatings. Through the statistical fragmentation model
by Hui et al. [26] used previously for the present material system [12],
the coating average stress at saturation can be determined as
c
y( )=µ , (2)
where and are Weibull shape and scale parameters, respectively,
for the tensile strength distribution of the coating. These can be esti-
mated from the fragmentation data given by crack density vs. applied
strain [12]. The parameter µ is the normalized average coating stress,
which can be determined from the shape parameter ρ and are tabulated
in Ref. [26]. Evaluation of Eqs. (1) and (2) for this case gives the
compressive coating stress at saturation
c s
x = Ec x + c µ . (3)
Thouless [18] analyzed ridge cracking based on Euler buckling in-
stability, regarding the delaminated coating as a beam fixed at the crack
tip and a subjected to a compressive force on the ridge end. The critical
compressive stress is
2E
c h 2

ples [13] but focus on ridge cracking in each coating. To quantify Gi , we crit =
48(1 c
2) l (4)
adapt the approach by Thouless [18], based on Euler buckling together
with certain assumptions by Andersons et al. [20] for fragmentation and the general energy release rate is [17].
tests. The model is applied to experimental observations using high- 2)
6 (1 Px 2
resolution SEM and digital image correlation (DIC). Gi = c
M 2 + h2
Furthermore, the energy release rates for relevant crack configura- Ec h3 12 (5)
tions from the analytical expressions were compared with predictions where
using finite element method (FEM) to determine the effect of some
simplifying assumptions in the analytical model. The FEM simulations Px = h ( crit
c
x) (6)
include the inelastic deformation of the polymer substrate observed at
is the compressive load and
ridge cracking onset at large tensile strains [11].
Normally, the tensile fragmentation test is limited to estimating the M= z crit h. (7)
interfacial shear strength and the cohesive strength distribution
[21–23]. Essentially, this work shows that ridge cracking, normally is the bending moment.
regarded as an artefact in tensile fragmentation tests, can be used to The Young's modulus in the beam model Ec is the coating modulus
estimate adhesive fracture properties between very thin coatings and obtained from atomic force microscopy [11]. The crack length l was
non-linear (elastic-plastic) substrates. Some advantages and dis- taken an average for several ridge cracks, as outlined in the results from
advantages of the method will be raised. strain-induced elastic buckling instability for mechanical measurement
of reference [11]. The remaining quantities needed to calculate Gi are
the substrate strain xs and the crack opening displacement z , which
2. Methods were determined experimentally using DIC in macroscopic tensile
testing and observations in high-resolution SEM, respectively.
2.1. Micromechanical model of ridge cracking The propagation of the ridge cracks along the interface under the
loading conditions shown in Fig. 1 occurs in mixed modes I and II. Once
The micromechanical model is based on a small coating fragment the fracture toughness has been determined, it can be partitioned into
with ridge cracking. The fragment length is b and corresponds to the its mode I and mode II components. The mode mixity in terms of the
width of the ridge crack. The length of the ridge crack is l. Fig. 1 shows stress intensity factors of the interfacial delamination can be calculated
the final configuration of a cross section of the ridge crack. The dif- as [17].
ference of lateral Poisson contraction in the substrate and the coating
gives rise to a compressive transverse stress component KII 12 M cos + h Px sin
tan = = ,
KI 12 M sin + h Px cos (8)
c c s c
x = r + Ec ( x + c y) (1)
where the phase factor ( D, D, ) depends on Dundurs' parameters
in the coating [20,24]. Here, Ec is the Young's modulus and c is the of elastic mismatch D and D , [27] as well as the thickness ratio
Poisson ratio of coating, xs is the transverse strain in the substrate and yc between coating and substrate, all defined in Ref. [17]. These para-
is the tensile strain in the coating. The residual stress rc in the coating meters can be determined explicitly from the measurements given in
was neglected due relatively slow process (allowing relaxation) and a the present method.

2
M.V. Tavares da Costa, et al.
Fig. 1. Illustration of the cross section of ridge crack for the micromechanical model.
2.2. Measuring strain field of coated substrate
Polymer substrates generally show high ductility and a non-linear
stress-strain response beyond the initial few percent of tensile strain.
Estimating the transverse coating stress directly from the mismatch of the
Poisson contraction between the coating and the substrate is suitable for
linear elastic deformations [20], but in the case of non-linear deforma-
tion it is advisable to measure the transverse compressive strain xs
needed directly on the sample. To this end, four samples were prepared
with a 4.0 mm × 0.1 mm cross section and a 10 mm of gage length for
tensile testing, corresponding to the same dimensions as in the tensile
fragmentation tests [11,12]. The mechanical tests were performed in a
Shimadzu Autograph AGS-X tensile machine with a crosshead speed of
1 mm/min. The tests were carried out up to 30% of tensile strain.
The strain field was measured in situ with a DIC GOM Aramis stereo
system 5 M. The measurement distance was 270 mm and the images
were captured by 1 frame per second synchronized with the tensile test.
All samples were spray painted with black speckles using an airbrush to
obtain a high-quality contrast [28]. The strain field was calculated as
the average strain components over a virtual gage area of 2 mm2 in the
center of the sample [29]. The Young's modulus was estimated as the
initial slope of the stress-strain curve, and the Poisson ratio as the ratio
of the lateral compressive strain to the longitudinal tensile strain in the
elastic regime. The influence of the coating stiffness on the elastic
properties of the samples was negligible since the substrate was more
than four orders of magnitude thicker than the coatings.
2.3. High-resolution scanning electron microscopy
We first performed ex situ tensile loading, stretching the specimens
up to the ridge cracking onset strain. This strain level was determined
previously in situ using a Deben micro-tensile stage mounted in a
Hitachi TM-1000 table-top SEM operating at 15 kV. The stretched
samples were locked in a portable fixture to maintain the deformed
state, and brought without delay to a Zeiss 1530 field emission gun
SEM. The specimens were coated with an Au–Pd conductive coating
with a Quorum SC7640 polaron sputter coater to minimize the charging
effect. This additional layer of Au–Pd was kept as thin as possible and
has been found not to cause any significant mechanical disturbance
3
Surface & Coatings Technology 382 (2020) 125025
[11]. The thickness of the conductive coating was neglected in the SEM
analysis to measure the crack dimensions, since even small nanoscale
features like edges of the cracked barrier coating could be observed. To
observe the ridge cracks closely with the high-resolution SEM, an ac-
celeration voltage of 5 kV and an in-lens secondary electron detector,
suited for detailed topographic analysis in high magnification, were
used while the specimens were tilted 52° on the sample holder.
2.4. Finite element formulation
We used FEM to validate the micromechanical model for the present
material configuration and geometry. This was done in two steps as
shown in Fig. 2. The first step is to obtain the stress-strain response of the
substrate, which is needed since the ridge crack onset occurs at high
strains in the non-linear regime. When the material model has been
identified from applied stress vs. DIC strain results, the second step is to
use this material behavior in the local model of the ridge crack to cal-
culate the energy release rate at crack tip. Both FEM models were im-
plemented in the solid mechanics module of COMSOL Multiphysics 5.4.
2.4.1. Material model of the substrate
Due to the orthotropic nature of the biaxially oriented PET [30–32],
the elastic constants were calibrated based on tensile testing as de-
scribed above. Bergström and Bischoff [33] identified three stages in
the non-linear stress-strain relationship of polyethylene, each corre-
sponding to a specific deformation mechanism. A similar stress-strain
response was found in our experiments, with an initial linear relation-
ship followed by two exponential curves.
For the initial linear regime, the material was defined as ortho-
tropic. The Young's modulus and Poisson ratio of the substrate in the
longitudinal loading direction (Eys and yx s
) were measured experimen-
tally. For the transverse direction, the Young's modulus and the Poisson
ratio of the substrate (Exs and xy s
) were taken from Ref. [34] where
similar PET films were characterized using the same experimental ap-
paratus. The remaining elastic parameters were determined based on
symmetry properties of the orthotropic stiffness matrix and the ratio
between the elastic parameters for another PET film [31].
For the plastic deformation, a simple isotropic albeit non-linear
model was chosen. The model is integrated in the COMSOL
M.V. Tavares da Costa, et al. Surface & Coatings Technology 382 (2020) 125025

Fig. 2. Scheme of the FEM simulation from (a) tensile testing of the coated PET film to obtain the material model used in (b) the ridge cracking modelling.

Multiphysics software, and only the non-linear material model needs to conditions are assumed, since the thickness of the film is much smaller
be implemented. The total strain than all other dimensions of the sample.
total = e + pl . (9)
2.4.2. Energy release rate
can be decomposed in elastic strain and plastic strain pl .
e
The use of linear elastic fracture mechanics should be questioned in
The yield function
any microscale crack growth. The size of the process zone cannot
F= Tresca + ys (10) generally be neglected in comparison with other crack dimensions, thus
invalidating the general assumptions of linearity (e.g. Ref. [35]). Al-
was defined as sum of the Tresca effective stress and the average of
though the coating thickness, crack lengths and openings are in sub-
yield stress governed by isotropic hardening
micrometer scale, we still opt for linear elastic fracture mechanics, since
ys = ys0 + h ( pe ), (11) the observed cracks were well-defined and showed no obvious cohesive
zones. Inclusion of length scale parameters or cohesive laws would
where ys0 is the initial yield stress and h is the nonlinear hardening require additional parameters not readily quantifiable. Bearing this
function with respect to the effective plastic strain pe . A user-defined compromise in mind, the obtained energy release rates should be in-
hardening function terpreted with care, and not regarded as precise material properties.
h ( pe ) = ys 0 [ exp(a1 pe ) exp(a2 pe ) ] (12) There are various methods to calculate the energy release rate at the
crack tip as reviewed by Matos et al. [36]. Varying the potential energy
was implemented by fitting to experimental results. The fitting para- has been found to be advantageous to determine the energy release rate
meters a1 and a2 were identified by comparison of the experimental for cracks grown at bimaterial interfaces [37]. The crack extension
stress-strain curves. method was adopted in the present fracture simulations, where a high
Finally, the plastic potential mesh density was used at the crack tip to obtain an accurate and con-
Qp ( ) = Qp (I1 ( ), J2 ( ), J3 ( )) = Mises (13) vergent solution. The mesh and geometry are shown in Fig. 2(b). This
simulation was carried out in plane strain [5,17], since the crack front
was expressed in terms of the stress invariants and is equal to the von length was large compared to other dimensions.
Mises criteria. It was used to calculate the increment in the plastic strain The energy release rate
tensor
1
Gi =
Qp Qp I1 Q p J2 Qp J3 b l (15)
p = = + + ,
I1 J2 J3 (14)
was calculated from the variation of the potential energy of the
where is the plastic multiplier. It should be noted that the adopted delaminated coating due to a crack length increment l . A step size of
elastoplastic model has been developed for ductile deformation of l = 2 nm showed convergence for the present configurations. The
metals. However, it was found to fit the experimental curves well and potential energy
could reasonably well capture the material response without too many
1
material parameters. The time-dependent viscous behavior of the = : dV
2 (16)
polymer has been neglected, and the predictions should therefore be
limited to strain rates and temperatures close to the ones used in ex- was determined for prescribed displacements. An initial deformation
periments. was set to account the transverse strain obtained experimentally from
The boundary conditions for the tensile test simulations are illu- PET film. A symmetric boundary condition was chosen along the right
strated in Fig. 2 (a), where the bottom edge is fixed while prescribed edge, roller boundary conditions along the upper edge of the substrate,
displacements y are imposed along the upper edge. Plane stress continuity boundary conditions along the yellow edge and free

4
M.V. Tavares da Costa, et al.
Fig. 3. Stress-strain curves from the FE model 1 (triangles) and DIC experiments
(red lines) of the coated films. (For interpretation of the references to color in
this figure legend, the reader is referred to the Web version of this article.)
boundary along the pink edge (initial crack length) at the interface, as
shown in Fig. 2(b). For these ultrathin and well-bonded coatings, we
did not observe any interfacial slipping ahead of the crack tip. The
coating and substrate elements are perfectly bonded ahead of the crack
tip, and free to separate in its wake. The initial transverse strains are
applied in the substrate, and locally shear deformations are accounted
for in FE model 2, which is not the case for the beam model.
3. Results and discussion
First, the transverse Young's modulus of the coated BoPET film was
determined to be 5.2 ± 0.2 GPa and the corresponding Poisson ratio to
be 0.24 at initial linear strains from the red curves in Figs. 3 and 4. The
results are comparable to similar coated films tested by Casal [34] using
the same test method: 5.3 ± 0.2 GPa for Young's modulus and
0.30 ± 0.02 for Poisson ratio in the transverse direction. Zhang and Li
[32] also studied the elastic properties of biaxially oriented PET films
and found a longitudinal Young's modulus of 5.37 ± 0.04 GPa. The
yield stress at 0.2% offset was found to be 60 ± 4 MPa. The scatter is
presented in terms of the standard deviation. The transverse direction is
the cross direction in the roll of biaxially oriented PET, where the
tensile testing was performed.
Fig. 4. Relation between tensile and compressive strains of coated substrate: FE
model 1 (purple triangles) and DIC experiments (red lines) of the coated film.
The inserts show the measured transverse strain field according to the color
scale legend. (For interpretation of the references to color in this figure legend,
the reader is referred to the Web version of this article.)
5
Surface & Coatings Technology 382 (2020) 125025
Fig. 4 shows DIC images of the transverse strain xs . The strain field
was found to be higher and relatively uniform in the center of the
samples compared with that in areas close to the edges. It is also in the
center where ridge cracking first appears and shows a higher con-
centration [11]. There was only limited scatter between the experi-
mental tests, as can be seen from the red lines in Figs. 3 and 4.
In FE model 1 shown in Fig. 2(a), the measured values of Eys and yx s
above were used in the longitudinal loading direction, Exs = 4.6 GPa
and xy s
= 0.16 [34] in the transverse direction and Ezs = 4 GPa and
s
xz = 0.1 in the out-of-plane direction (based on the obtained values
and Ref. [31]) for the elastic deformation. In the elastic regime, it is
assumed that the material is orthotropic, although the moduli along the
material axis only differ about 13% from each other.
The inelastic (plastic) part of the deformation was then identified by
curve fitting in Fig. 3 as outlined in Section 2.4.1. The purple triangles in
Figs. 3 and 4 are the calculated values from the numerical simulations
using Eq. (12), the fitting parameters a1 and a2 of which were found as
0.75 and −205, respectively. There is a general good agreement with
experimental results in the stress-strain fit in Fig. 3 and the predictions of
the transverse strains in Fig. 4. This indicates that the elastoplastic model
is applicable for our polymer film, although the model has been developed
for non-linear ductile deformation of metals. It can be noticed in Fig. 4 that
there is a considerable difference between the elastic predictions
( xs = y ) and those from the model accounting for the non-linear de-
s
formation, especially for larger strain where the ridge cracking occurs. It is
therefore not advisable to use a linear-elastic model to predict the trans-
verse elastic strains needed in the analysis of the ridge cracks.
As expected, stress concentrations occur at the clamping ends close
to the edges, which is shown in the FEM result inserts for ys in Fig. 3. In
the center of the specimens, however, where the ridge cracking was
observed, the stress state was found to be uniform.
Fig. 5 shows high-magnification SEM images from the thicker TiO2,
(a) and (c), and mixed TiO2/Al2O3 coatings, (b) and (d). The ridge cracks
could not be seen along edges carefully cut by a focused ion beam due to
melting of the polymer substrate. The only way to measure the opening
of the ridge cracks was by observation along the tensile fragment cracks.
The ridge cracking appeared uniformly throughout the coatings as shown
in (a) and (b). It can be seen that ridge cracks may have only one crack
along the top of the ridge, as schematically drawn in Fig. 1. These types
of cracks are analyzed in this work. Other observed cracks show also
secondary cracks in one or both flanks of the ridge. Finally, some un-
cracked ridges, appearing as elongated buckles, were also observed. They
have been analyzed experimentally in e.g. Refs. [11,20].
Through the high-resolution SEM images, such as those in Figs. 5(c)
and (d), a simple approximation can be made for the crack opening. The
20 nm coating thickness was set by the number of ALD cycles [13] and
largely confirmed by the SEM observations. The average crack opening
found to be related to the thickness as z 2.5h for the TiO2 coating
and z 2h for the mixed oxide coating. The debonded coatings had a
beam-like shape viewed from the edge, without any significant bridging
ligaments close to the crack tip. We therefore used linear elastic fracture
mechanics despite the small scales. The average crack length for a large
number of ridge cracks were found to be l = 120 ± 10 nm for the TiO2
coating and l = 112 ± 10 nm for the mixed oxide coating. For this
range of crack lengths, the fracture toughness Gi was calculated using
the non-linear FE model and the micromechanical model (based on
predicted elastic deformation and measured deformation) while
keeping the crack opening and coating thickness constant. These results
are presented in Fig. 6 where a stable crack growth with decreasing Gi
with increasing crack length was confirmed for the investigated crack
configurations and materials. The FEM calculations resulted in
Gi = 2.5 J/m2 for TiO2 interface and Gi = 1.6 J/m2 for mixed oxide
interface. Incidentally, a stronger interfacial adhesion for the TiO2 in
terms of the interfacial shear strength determined from the fragment
density at saturation [12]. The analytical estimates in Fig. 6 are about
10% lower than the numerical predictions, although there is a converge
M.V. Tavares da Costa, et al.
Fig. 5. High-magnification SEM images of the ridge cracking in the 20 nm coatings.
tendency for longer cracks, as is expected since the beam model be-
comes increasingly accurate for longer cracks. The fracture toughness
from the micromechanical model based on the measured transverse
strains are closer to the more accurate FE results than that based on
linear elastic predictions of the transverse strains. The mode mixity
phase angle using Eq. (8) was found to be 45° for the TiO2 coating and
42° for the mixed oxide coating. This ridge crack test can be considered
to measure more opening (mode I) than shear (mode II) cracking for
configurations such as those presented. Crack growth criteria normally
require the energy release rates or stress intensity factors to be parti-
tioned in mode components [38,39].
The magnitude of fracture toughness can be compared with other
results of small scale interfacial cracking, as compiled in Table 1. Al-
though these were determined with different test methods for a variety
of material combinations and with different modelling assumptions, the
orders of magnitude are in concert. The values are even close to limit
values of surface energy. The theoretical surface energy required to
form a new surface by breaking of chemical bonds is about 1.5 J/m2 for
Fig. 6. Fracture toughness at interface Gi evaluated in range of l = 120 ± 10 nm for TiO2 and l = 112 ± 9 nm for mixed oxide coatings. The black line indicate the
average of the measured crack length values.
6
Surface & Coatings Technology 382 (2020) 125025
standard polymers [40]. Overall, the microscale fracture toughness
values are all several orders of magnitude smaller than values typically
observed in macroscale fracture toughness testing (e.g. Ref. [41]). The
volume available for dissipation during crack growth becomes smaller
for smaller cracks and thinner coatings. Since scaling down from macro
to nanoscale crack is generally not possible, a practical approach is to
conduct the fracture characterization on the same length scale as for the
intended application. If opening (mode I) crack growth on the sub-
micron level is of interest, the present method may provide insight and
quantitative estimates of the fracture toughness. It also has the ad-
vantage to provide data for coating fragmentation in tension (quanti-
fication of interfacial shear strength and coating strength distribution),
before the action ridge cracks are formed.
4. Conclusions
Ridge cracking has been found to occur in the transverse direction
in tensile fragmentation tests of brittle coatings on polymer substrates.
M.V. Tavares da Costa, et al. Surface & Coatings Technology 382 (2020) 125025

Table 1
Review values for fracture toughness from different methods and materials.
Method Illustration Coating/substrate materials Coating thickness [nm] Gi [J/m2] Ref.

Fragmentation test under uniaxial load SiNx/Polyester 350 3 [42]

SiNx/Polyimide 600–800 30 [25]
Buckling of coatings SiOx/BoPET 103 15 [20]
Chromium/Polyimide 100 4–8 [43]
Chipping from nanoindentation Sol-gel/glass 1500–1800 1–4 [8]
CNx/SiC 1000 5.5 [44]

Nanosectioning Neat PMMA 60 6.3a [41]

Ridge cracks TiO2/BoPET 20 2.5 Present work

TiO2 and Al2O3/BoPET 20 1.6

a
Ultramicrotomy of nanoscale sections of monolithic material.

These ridge cracks can be addressed by fracture mechanics, thus pro-
viding a test method to estimate the fracture toughness of interfacial
cracking between the coating and substrate mainly in crack opening
mode. This study demonstrates feasibility of the test method down to
ultrathin coatings with a thickness of 20 nm.
A micromechanical model can be used to estimate the interfacial
fracture toughness. The energy release rates from the analytical model
are comparable to results from FEM simulations, with a difference of
about 10% for the ridge cracks observed in this work. The accuracy is
improved if the transverse strains are determined experimentally rather
than predicted by linear elastic Poisson contraction. For polymer films,
which show a significant non-linear stress-strain behavior at ridge crack
onset, it can therefore be recommended to measure the transverse
strains.
The resulting interfacial fracture toughness was found to be ∼2.5 J/
m2 for TiO2/PET interface and ∼1.6 J/m2 for mixed oxide (TiO2 and
Al2O3)/PET interface, which are comparable in magnitude with other
interfaces in thin brittle coatings. An advantage with the ridge crack
test method is that crack growth in mainly opening mode can be ana-
lyzed for very thin coatings. The ridge cracks are widespread in frag-
mentation test samples, and data from many cracks can be collected
from SEM observations. If this approach should be considered as a
characterization method, a limiting factor is that the method requires
high-resolution SEM analysis which was found to be time consuming
and sensitive for the ultrathin coatings analyzed in this work.
Acknowledgement
Financial support and supply of the test samples for this study by
Tetra Pak Packaging Solutions AB are much appreciated. MVTdC is
grateful to the European Commission within the framework of the
Erasmus Mundus Programme, Action 2-STRAND 1, Lot 9, Brazil for
supporting his PhD studies.
References
[1] P. Fayet, C. Neagu, K. Gamstedt, Mechanics-driven Material Design for Optimized
Barrier Films, Naples, USA, 2015.
[2] A.A. Griffith, J.J. Gilman, The phenomena of rupture and flow in solids, Trans.
ASM. 61 (1968) 855–906 1200.
[3] J.L. Beuth, Cracking of thin bonded films in residual tension, Int. J. Solids Struct. 29
(1992) 1657–1675, https://doi.org/10.1016/0020-7683(92)90015-L.
[4] Z.C. Xia, J.W. Hutchinson, Crack patterns in thin films, J. Mech. Phys. Solids 48
(2000) 1107–1131, https://doi.org/10.1016/S0022-5096(99)00081-2.
[5] H.M. Jensen, The blister test for interface toughness measurement, Eng. Fract.
Mech. 40 (1991) 475–486, https://doi.org/10.1016/0013-7944(91)90144-P.
[6] J. Andersons, J. Modniks, Y. Leterrier, G. Tornare, P. Dumont, J.-A.E. Månson,
Evaluation of toughness by finite fracture mechanics from crack onset strain of
brittle coatings on polymers, Theor. Appl. Fract. Mech. 49 (2008) 151–157, https://
doi.org/10.1016/j.tafmec.2007.11.002.
[7] K. Fu, Y. Yin, L. Chang, D. Shou, B. Zheng, L. Ye, Analysis sson multiple ring-like
cracks in thin amorphous carbon film on soft substrate under nanoindentation, J.
Phys. Appl. Phys. 46 (2013) 505314, https://doi.org/10.1088/0022-3727/46/50/
505314.
[8] J.D. Toonder, J. Malzbender, G. de With, R. Balkenende, Fracture toughness and
adhesion energy of sol-gel coatings on glass, J. Mater. Res. 17 (2002) 224–233,
https://doi.org/10.1557/JMR.2002.0032.
[9] M.D. Thouless, An analysis of spalling in the microscratch test, Eng. Fract. Mech. 61
(1998) 75–81, https://doi.org/10.1016/S0013-7944(98)00049-6.
[10] R. Saha, W.D. Nix, Effects of the substrate on the determination of thin film me-
chanical properties by nanoindentation, Acta Mater. 50 (2002) 23–38, https://doi.
org/10.1016/S1359-6454(01)00328-7.
[11] M.V. Tavares da Costa, C. Neagu, P. Fayet, U. Wiklund, H. Li, K. Leifer,
E.K. Gamstedt, Comparison of test methods estimating the stiffness of ultrathin
coatings, J. Coat. Technol. Res. 15 (2018) 743–752, https://doi.org/10.1007/
s11998-018-0085-0.
[12] M.V. Tavares da Costa, J. Bolinsson, R.C. Neagu, P. Fayet, E.K. Gamstedt,
Experimental assessment of micromechanical models for fragmentation analysis of
thin metal oxide coatings on polymer films under uniaxial tensile deformation, Surf.
Coat. Technol. 370 (2019) 374–383, https://doi.org/10.1016/j.surfcoat.2019.03.
035.
[13] E. Dickey, W.A. Barrow, High rate roll to roll atomic layer deposition, and its ap-
plication to moisture barriers on polymer films, J. Vac. Sci. Technol. A 30 (2012)
021502, https://doi.org/10.1116/1.3678486.
[14] A.L. Volynskii, S. Bazhenov, O.V. Lebedeva, N.F. Bakeev, Mechanical buckling in-
stability of thin coatings deposited on soft polymer substrates, J. Mater. Sci. 35
(2000) 547–554, https://doi.org/10.1023/A:1004707906821.
[15] C.M. Stafford, C. Harrison, K.L. Beers, A. Karim, E.J. Amis, M.R. VanLandingham,
H.-C. Kim, W. Volksen, R.D. Miller, E.E. Simonyi, A buckling-based metrology for
measuring the elastic moduli of polymeric thin films, Nat. Mater. 3 (2004) 545–550,
https://doi.org/10.1038/nmat1175.
[16] H. Jiang, D.-Y. Khang, J. Song, Y. Sun, Y. Huang, J.A. Rogers, Finite deformation
mechanics in buckled thin films on compliant supports, Proc. Natl. Acad. Sci. 104
(2007) 15607–15612, https://doi.org/10.1073/pnas.0702927104.
[17] J.W. Hutchinson, Z. Suo, Mixed Mode Cracking in Layered Materials, Adv. Appl.
Mech. Elsevier, 1991, pp. 63–191, https://doi.org/10.1016/S0065-2156(08)
70164-9.
[18] M.D. Thouless, Combined buckling and cracking of films, J. Am. Ceram. Soc. 76
(1993) 2936–2938, https://doi.org/10.1111/j.1151-2916.1993.tb04044.x.
[19] Y. Ni, S. Yu, H. Jiang, L. He, The shape of telephone cord blisters, Nat. Commun. 8
(2017) 14138, https://doi.org/10.1038/ncomms14138.
[20] J. Andersons, S. Tarasovs, Y. Leterrier, Evaluation of thin film adhesion to a com-
pliant substrate by the analysis of progressive buckling in the fragmentation test,
Thin Solid Films 517 (2009) 2007–2011, https://doi.org/10.1016/j.tsf.2008.10.
072.
[21] J. Aveston, G.A. Cooper, A. Kelly, Single and multiple fracture, the properties of
fiber composites, Single Mult. Fract. Prop. Fiber Compos. (1971) 15–26.
[22] Y. Leterrier, Y. Wyser, J.A.E. Månson, J. Hilborn, A method to measure the adhesion
of thin glass coatings on polymer films, J. Adhes. 44 (1994) 213–227, https://doi.
org/10.1080/00218469408027078.
[23] J. Andersons, Y. Leterrier, G. Tornare, P. Dumont, J.-A.E. Månson, Evaluation of
interfacial stress transfer efficiency by coating fragmentation test, Mech. Mater. 39
(2007) 834–844, https://doi.org/10.1016/j.mechmat.2007.02.003.
[24] M. Yanaka, T. Miyamoto, Y. Tsukahara, N. Takeda, In situ observation and analysis
of multiple cracking phenomena in thin glass layers deposited on polymer films,
Compos. Interfac. 6 (1998) 409–424, https://doi.org/10.1163/156855499X00116.
[25] S. Tarasovs, J. Andersons, Y. Leterrier, Estimation of interfacial fracture toughness
based on progressive edge delamination of a thin transparent coating on a polymer
substrate, Acta Mater. 58 (2010) 2948–2956, https://doi.org/10.1016/j.actamat.
2010.01.023.
[26] C.-Y. Hui, S.L. Phoenix, D. Shia, The single-filament-composite test: a new statistical
theory for estimating the interfacial shear strength and Weibull parameters for fiber
strength, Compos. Sci. Technol. 57 (1998) 1707–1725, https://doi.org/10.1016/

7
M.V. Tavares da Costa, et al.
S0266-3538(97)00103-6.
[27] J. Dundurs, Discussion: “Edge-bonded dissimilar orthogonal elastic wedges under
normal and shear loading” (Bogy, D. B., 1968, ASME J. Appl. Mech 36 (1969)
650–652, https://doi.org/10.1115/1.3564739 35, pp. 460–466), J. Appl. Mech..
[28] Z. Wu, S. Zhang, A. Vorobyev, K. Gamstedt, K. Wu, C. Guo, S.H. Jeong, Seamless
modulus gradient structures for highly resilient, stretchable system integration,
Mater. Today Phys. 4 (2018) 28–35, https://doi.org/10.1016/j.mtphys.2018.02.
002.
[29] H. Koerber, J. Xavier, P.P. Camanho, High strain rate characterisation of uni-
directional carbon-epoxy IM7-8552 in transverse compression and in-plane shear
using digital image correlation, Mech. Mater. 42 (2010) 1004–1019, https://doi.
org/10.1016/j.mechmat.2010.09.003.
[30] Y. Leterrier, A. Pinyol, L. Rougier, J.H. Waller, J.-A. Månson, P.J.J. Dumont,
J. Andersons, J. Modniks, M. Campo, P. Sauer, J. Schwenzel, Influences of roll-to-
roll process and polymer substrate anisotropies on the tensile failure of thin oxide
films, Thin Solid Films 518 (2010) 6984–6992, https://doi.org/10.1016/j.tsf.2010.
07.033.
[31] Y. Okabe, N. Takeda, M. Yanaka, Y. Tsukahara, Determination of the orthotropic
elastic constants of thin PET film by an ultrasonic micro-spectrometer, IEEE Trans.
Ultrason. Ferroelectr. Freq. Control 46 (1999) 1269–1275, https://doi.org/10.
1109/58.796131.
[32] S.L. Zhang, J.C.M. Li, Anisotropic elastic moduli and Poisson's ratios of a poly
(ethylene terephthalate) film, J. Polym. Sci., Part B: Polym. Phys. 42 (2004)
260–266, https://doi.org/10.1002/polb.10542.
[33] J.S. Bergström, J.E. Bischoff, An advanced thermomechanical constitutive model
for UHMWPE, Int. J. Struct. Chang. Solids. 2 (2010) 31–39.
[34] J.A. Casal Ramos, Estudio de la fractura de un recubrimiento frágil nanométrico
sobre una pulícula polimérica bajo tension uniaxial, Universidad Simón Bolívar,
2015.
Surface & Coatings Technology 382 (2020) 125025
[35] J.W. Hutchinson, A.G. Evans, Mechanics of materials: top-down approaches to
fracture, Acta Mater. 48 (2000) 125–135, https://doi.org/10.1016/S1359-
6454(99)00291-8.
[36] P.P.L. Matos, R.M. McMeeking, P.G. Charalambides, M.D. Drory, A method for
calculating stress intensities in bimaterial fracture, Int. J. Fract. 40 (1989) 235–254,
https://doi.org/10.1007/BF00963659.
[37] M.R. Begley, J.W. Hutchinson, The Mechanics and Reliability of Films, Multilayers
and Coatings, Cambridge University Press, 2017, https://doi.org/10.1017/
9781316443606.
[38] H.M. Jensen, Mixed mode interface fracture criteria, Acta Metall. Mater. 38 (1990)
2637–2644, https://doi.org/10.1016/0956-7151(90)90276-M.
[39] T. Ikeda, N. Miyazaki, T. Soda, Mixed mode fracture criterion of interface crack
between dissimilar materials, Eng. Fract. Mech. 59 (1998) 725–735, https://doi.
org/10.1016/S0013-7944(97)00177-X.
[40] I.M. Ward, J. Sweeney, Mechanical Properties of Solid Polymers, third, Willey,
2013.
[41] F. Sun, H. Li, H. Lindberg, K. Leifer, E.K. Gamstedt, Polymer fracture and de-
formation during nanosectioning in an ultramicrotome, Eng. Fract. Mech. 182
(2017) 595–606, https://doi.org/10.1016/j.engfracmech.2017.05.044.
[42] N.E. Jansson, Y. Leterrier, L. Medico, J.-A.E. Månson, Calculation of adhesive and
cohesive fracture toughness of a thin brittle coating on a polymer substrate, Thin
Solid Films 515 (2006) 2097–2105, https://doi.org/10.1016/j.tsf.2006.07.012.
[43] F. Toth, F.G. Rammerstorfer, M.J. Cordill, F.D. Fischer, Detailed modelling of de-
lamination buckling of thin films under global tension, Acta Mater. 61 (2013)
2425–2433, https://doi.org/10.1016/j.actamat.2013.01.014.
[44] J. Fernandez-Palacio, I. Arce-Garcia, S.J. Bull, Indentation response of fullerene-like
CNx, Tribol. Int. 37 (2004) 929–940, https://doi.org/10.1016/j.triboint.2004.07.
010.