Wear 254 (2003) 278291

A model for stresses, crack generation and fracture toughness

calculation in scratched TiN-coated steel surfaces
Kenneth Holmberg

, Anssi Laukkanen, Helena Ronkainen, Kim Wallin, Simo Varjus

VTT Industrial Systems, PL 1704, FIN-02044 VTT, Finland
Received 10 April 2002; received in revised form 7 October 2002; accepted 1 November 2002
Abstract
The contact situation in a scratch tester, when a spherical rigid diamond tip is sliding with an increasing load over an elasticplastic steel
plate deposited with a 2 m thick hard ceramic TiN coating is analysed. A three-dimensional nite element model (FEM) for describing
the elastic and plastic behaviour and for calculating the stresses and strains has been developed. It shows that the maximum rst principal
tensile stress is generated in the tail part of the contact area. With increasing load a tetra-armed star shaped stress-eld is generated around
the contact. After about 1 mm of sliding a peak area of maximum rst principal stress is formed in the back-tail region at the border of the
scratch groove, creating the rst visible angular cracks in the coating. This is in agreement with empirical observations. Once substantial
plastic deformation of the substrate has occurred, the maximum tensile stresses are located behind the contact at a distance of 0.51 times
the contact length from the back edge of the contact. These stresses have a horseshoe shaped ridge of maximum values with an opening
in the sliding direction. The change of the state of deformation from sliding over the coating (sliding mode) to deforming the substrate
plastically (ploughing mode) characterises the loss of load carrying capacity of the coated surface system. The model is used for calculating
the fracture toughness of the coating. The critical fracture toughness is equal to the tensile stress times the square root of half of the crack
spacing (K
c
=

b/2) when the crack spacing is smaller than the crack length. For determining the fracture toughness of a 2 m thick
TiN coating on steel substrate a suitable crack eld turned out to be the transversal tensile cracks in the scratched groove. For the studied
case, the fracture toughness of the TiN coating was measured to be K
c
= 7 MPa m
0.5
.
2002 Elsevier Science B.V. All rights reserved.
Keywords: Surface engineering; Coatings; Stress modelling; Fracture; Scratch test; Fracture toughness
1. Introduction
There has been increased interest in the use of coatings
on mechanical components, on tools in the production in-
dustry, on disc drives in the computer industry, on preci-
sion instruments, and on human replacement organs. New
coating deposition techniques developed over the last two
decades offer a wide variety of possibilities to tailor surfaces
with many different materials and structures. In particular,
chemical vapour deposition (CVD) and physical vapour de-
position (PVD) techniques have made it possible to deposit
thin coatings only about 1 m thick in a temperature range
from very high temperatures (about 1000

C) down to room
temperature.
Coating materials such as TiN, TiC, Al
2
O
3
and more re-
cently diamond, diamond-like carbon (DLC) and MoS
2
and
their combinations as multilayers and dopants have been
used with great success. In the best cases, these very thin

Corresponding author. Tel.: +358-9-456-5370; fax: +358-9-456-7002.

E-mail address: kenneth.holmberg@vtt. (K. Holmberg).
coatings have decreased the coefcient of friction and the
wear rate by one or two orders of magnitude. Super-low
friction coefcients down to 0.001 in dry sliding have been
measured [1,2]. The deposition techniques of thin coatings
and their tribological behaviour and applications have been
described by Holmberg and Matthews [3,4] and Holmberg
et al. [5].
A tribological contact with two loaded surfaces in rela-
tive motion is a very complex system that is not easy to un-
derstand nor simulate or predict. The system becomes even
more complex when coatings are introduced on the surfaces.
Studies have been carried out at macrolevel, i.e. the compo-
nent level, at microlevel, i.e. the surface asperity level, and
at nanolevel, i.e. the molecular level [4,6,7].
One problem is that even the parameters used to describe
friction and wear behaviour in coated tribological contacts
are not clear. The parameters related to the macrogeometry
of the contact and the topography are better dened, but the
parameters dening the wear debris and surface layers are
not well dened. Parameters related to load and speed are
well controlled, as are the environmental parameters such as
0043-1648/02/$ see front matter 2002 Elsevier Science B.V. All rights reserved.
PII: S0043- 1648( 02) 00297- 1
K. Holmberg et al. / Wear 254 (2003) 278291 279
temperature, humidity, environmental gases, etc. The mate-
rial parameters are crucial to tribological performance, but
here there is considerable diversity. The importance of ex-
pressing the material response of a coated surface in its basic
material parameterselasticity as elastic modulus, plastic-
ity as hardness and fracture as fracture toughnesshas been
emphasised [810]. Hardness is a very important material
parameter that is much used and well dened by several
in-depth studies [1114]. Youngs modulus is another much
used and well dened material parameter [12,13]. However,
other parameters that attempt to describe the failure or frac-
ture performance, such as the critical load in scratch testing
and load-carrying capacity, are less well dened and their
role is more uncertain.
Because of our limited understanding of contact phenom-
ena, naturally the development of new coatings and their
applications has been very much a matter of trial and error.
This approach, however, has by no means been unsuc-
cessful. Nonetheless, the full advantage of the possibilities
offered by these new techniques requires more systematic
approaches. Excellent work on the simulation of a coated
tribological contact has been reported [1518]. Because of
the complexity of the problem they are still restricted to
necessary simplications of the complete problem, such as
only considering elastic and plastic material behaviour.
It has been observed that hard-coated surfaces very often
fail due to fracture. The basic mechanisms of fracture and
crack initiation in bulk materials is found in textbooks by
Lawn [19], and for a surface under a sliding line contact
by Bower and Fleck [20]. Cracking of thin lms in residual
tension has been analysed by Beyth [21] and the interfacial
cracking between a thin lm and the substrate by Thouless
[22]. Tribological analysis has been made of the role of
a coating fracture on friction and wear based on observed
failure modes and cracking patterns [4,23,24].
Some attempts to nd a suitable method for measuring
the fracture toughness of thin hard coatings have been re-
ported. The indentation test was used by Diao et al. [8] to
estimate the fracture toughness of TiN and Al
2
O
3
coatings
on a cemented carbide substrate, and by Li et al. [25], Li and
Bhushan [26] and Nastasi et al. [27] to estimate the fracture
toughness of diamond-like hard carbon coatings (DLC) de-
posited on silicon substrate. A tensile test method was de-
veloped by Ollivier et al. [28] for coating interface fracture
assessment of DLC lms on PET, and a four-point bending
test by Wiklund et al. [29] for cracking resistance assess-
ment of TiN and CrN coatings on high speed steel.
Oliveira and Bower [30] have published an analysis of
fracture and delamination in elastic thin coatings on a rigid
half-space subject to a sliding contact loaded by an elas-
tic cylindrical indenter. Malzbender and de With [3134]
performed an excellent analysis of the indentation, friction,
stress generation, residual stresses and fracture initiation as-
pects of a diamond stylus sliding on a solgel coated glass
surface. They determined the fracture toughness for the sys-
tem. However, these studies are limited to a rigid substrate.
The models for presenting stress and strain conditions in
a typical sliding contact with thin hard coatings, as well as
the test methods for measuring the fracture toughness of the
surface, have so far both proven limited.
The aim of this paper is to develop a computer model
for simulating the stress distribution, and for calculating the
stress and strain levels and fracture toughness of a coated
surface in scratch test contact conditions.
2. The scratch test
The scratch test was rst suggested for coating adhesion
measurements by Perry [35,36], Steinmann and Hintermann
[37] and Valli [38]. The method is today widely used by
the coating industry and coating development laboratories,
as well as in research for evaluating the tribological prop-
erties of coatings. Its usefulness as an adhesion and quality
assessment method has been discussed [23,3943]. The
scratch test is generally accepted as a good and efcient
method for the quality assessment of coated surfaces, but its
use for coating-to-substrate adhesion assessment has been
criticised by several authors (e.g. [43,44]).
The scratch test consists of pulling a diamond stylus over
the surface of a sample under a normal force, which is in-
creased either stepwise or continuously until failure is ob-
served, as shown in Fig. 1. The normal load at which this
happens is called the critical normal load L
c
(N). It is gener-
ally accepted that the test is suitable for coatings of thickness
ranging from 0.1 to 20 m and this covers a large number
of applications. The diamond stylus has a Rockwell C ge-
ometry with a 120

cone and a 200 m radius spherical tip

dened in the European Standard EN 10109-2. The recom-
mended loading rates are 10 and 100 N/min and the indenter
transverse speed is 10 mm/min. The scratch test procedure is
described in the European Standard suggestion prEN 1071-3
[45].
A schematic presentation of the stylus sliding on a coated
sample is shown in Fig. 2. The material loading and response
Fig. 1. Schematic representation of the scratch test apparatus.
280 K. Holmberg et al. / Wear 254 (2003) 278291
Fig. 2. Schematic illustration of the stylus drawn along the coated sample. The material loading and response can be divided into three phases: ploughing,
interface sliding and pulling a free-standing coating.
conditions have been divided into three independent phases
by Holmberg [46] to illustrate the involved contact and de-
formation mechanisms, as shown in the gure.
Phase one represents the ploughing of a stylus in the sub-
strate material. The substrate material is deformed by plastic
or elastic deformation and a groove is formed. If we assume
that the stylus front surface sliding against the coating is
frictionless, then only the work carried out for plastic defor-
mation is considered.
Phase two represents the bending and drawing of a free-
standing coating like a sheet between a frictionless roller
and a stationary sphere-tipped cone, as shown in Fig. 2. The
sphere-tipped cone has an equal geometry to the sliding sty-
lus. The upper surface of the coating rubs against the stylus
front surface and the force required for pulling the coating is
equal to the frictional force on the coating against the stylus
front surface. The bending movements cause stresses and
stress release in the coating when drawn between the sur-
faces. In this phase the work done for overcoming friction
is considered.
Fig. 3. The surface cracks generated in a scratch test track can be classi-
ed as (a) angular cracks, (b) parallel cracks, (c) transverse semi-circular
cracks, (d) coating chipping, (e) coating spalling, and (f) coating break-
through (modied after Larsson et al. [48]).
Phase three represents pulling the coating from one
point on the surface when its other end is xed just like a
free-standing thin plate, as shown in Fig. 2. The increasing
pulling force results in cracks at the place of maximum
tensile stress. In the real coated sample, the coating is of
course not free-standing but xed to a solid substrate.
When the diamond stylus is drawn over the surface with
an increasing normal load, it typically results in a coating
fracture and spalling pattern as shown in Fig. 3. The for-
mation of cracks in the groove of a scratch tester has been
shown by e.g. [23,43,45,4749]. They can typically be de-
scribed as (a) angular cracks, (b) parallel cracks, (c) trans-
verse semi-circular cracks, (d) coating chipping, (e) coating
spalling, and (f) coating breakthrough.
3. Three-dimensional nite element model for stress
and strain presentation
A three-dimensional nite element model was developed
for calculating the stresses and strains in the coated sur-
face and for identifying the stress concentrations where the
rst cracks of the coated surface are expected to occur.
The scratch test experiment was discretised using the inher-
ent symmetry of the geometry and introducing a nite ele-
ment mesh where mesh sizing is of the order of the coating
thickness. After analysing the convergence behaviour of the
boundary value problem and assessing whether suitable ac-
curacy of contact-related eld variables can be attained, a
suitable mesh density around the contact area was found to
be roughly half the coating thickness. Bilinear hybrid ele-
ments were used in Abaqus 5.8-14 and 6.2-1 nite element
software. The volume of the nite element slit taken to de-
scribe the scratch test conguration was 12 mm 1 mm
K. Holmberg et al. / Wear 254 (2003) 278291 281
0.1 mm(length, width, thickness). The substrate deformation
behaviour was characterised as elasticplastic with isotropic
hardening, while the coating was modelled to behave in a
linear-elastic manner. The sliding spherical tip was mod-
elled as linear-elastic, but Youngs modulus for the diamond
made it behave in a rigid manner.
The kinetic formulation was presented applying the nite
strain deformation description. The contact event was mod-
elled by describing two contact zones, commonly referred
to as the master and slave surfaces, where the master surface
is the one having greater rigidity. During the progress of the
experiment the relative positions of the master and slave sur-
faces dene the contact event. The contact formulation was
of the nite sliding type due to large local deformations and
the distance the tip travels during the experiment. The con-
tact was presented as a hard contact between smooth sur-
faces, i.e. tractions are transferred at the instant of contact but
not before. Computational initial oscillations were treated
with viscous damping. A velocity-independent Coulombian
friction model was used to characterise related surface in-
teractions, and the surface-hardening effect due to plastic
deformation of the substrate is thus included in the model.
The results are inferred and analysed primarily with re-
spect to the rst principal stress. Generally, studies in the
current eld apply the von Mises equivalent stress to char-
acterise deformation and failure events. However, the von
Mises stress has its greatest potential in understanding and
modelling deformation related events, such as the plasticity
of metals, and is not usually associated with the brittle type
of failure. It can be argued, and has often been presented [50]
that local unstable cracking is more dependent on the pre-
vailing tensile stress state than the practically non-existent
state of deformation, and that deformation-related parame-
ters and invariants do not comply with the physical appear-
ance of fracture. Most local approach models for such failure
micromechanisms rely on the rst principal stress. Since the
current work deals with cracking of a brittle layer incapable
of exhibiting much more than elastic deformation all the way
to nal rupture, it can be expected that use of the rst princi-
pal stress as a fundamental stress component to explain the
physical fracture patterns will bring about the most success.
Compared to real scratch tester contact conditions, the
following limitations have been set for the sake of simplicity.
Compressive residual stresses normally occurring in ce-
ramic thin coatings are not included in the simulation
cases reported here. They can be introduced into the used
model, and it is the intention of the authors to report such
results in a separate article.
The stress relaxation effect of previously generated cracks
on the stress distribution is not included in the present
simulations. It is the intention of the authors to develop
the model in order also to take that into account, and to
report such simulation results in a separate article.
The surfaces in the model are ideally smooth, which
means that surface roughness effects are not considered.
The materials are considered to be fully homogenous and
free from contaminants, pinholes and such defects often
occurring in thin ceramic coatings.
4. Stress distribution in the coated surface
4.1. Simulated contact conditions
The above described contact conditions and sliding pro-
cess were simulated by the computer model. The following
parameters were used in the calculations of the stress and
strain distributions.
Scratch test parameters
Sliding distance: 10 mm.
Load: increases linearly from 5 N pre-load at 0 mm dis-
tance to 50 N at 10 mm distance.
Sliding velocity: not included in the model, i.e. the model
is time independent.
Sliding stylus(Rockwell C)
Radius of spherical tip: 200 m.
Material: diamond.
Youngs modulus: 1140 GPa.
Hardness: 80 GPa.
Poissons ratio: 0.07.
Roughness: ideally smooth.
Coating
Thickness: 2 m.
Material: titanium nitride (TiN), deposited by PVD.
Youngs modulus: 300 GPa.
Hardness: 25 GPa.
Poissons ratio: 0.22.
Roughness: ideally smooth.
Substrate
Geometry: ideally smooth plate.
Material: high speed steel.
Youngs modulus: 200 GPa.
Hardness: 7.5 GPa.
Poissons ratio: 0.29.
Yield strength estimated from ultimate bending strength:
4100 MPa.
Strain hardening coefcient: 20.
Friction
The values for the coefcient of friction were measured
from samples corresponding to the above material com-
bination. The value was 0.06 after 0.1 mm of sliding
and increased linearly to 0.13 at 10 mm of sliding.
In the simulations a constant value 0.08 was used for
the coefcient of friction due to friction from interfa-
cial shear which excludes the ploughing component of
friction.
4.2. First principal stresses in the scratched surface
The simulated stress distributions are shown in Figs. 46
at 0, 0.3, 0.5, 0.7, 1.14, 1.6, 2.33 and 4.8 mmfromthe starting
282 K. Holmberg et al. / Wear 254 (2003) 278291
Fig. 4. Static mode topographical stress-eld map showing maximum rst principal stresses on the coating and at the symmetry plane intersection of the
steel sample coated with a 2 m thick TiN coating and loaded by a spherical diamond tip with 5 N load. The loading tip is invisible in the gure. The
contact zone is shown by the white curve.
point. The gures are topographical stress-eld maps where
each colour corresponds to a certain stress level range at
the surface and at the intersection shown in the gure. The
watch angel is similar to that in Fig. 2 but the spherical tip is
invisible in order to display the stresses. The corresponding
contact area of the coated plate and the sphere is shown by
the white curve.
5. Corresponding empirical observations
A typical crack pattern observed from scratch tests per-
formed in the authors laboratory with a test procedure and
samples according to the parameters listed above is shown in
Fig. 7. The rst cracks to occur were angular cracks, found
about 1 mm from the starting point. A photograph of these
rst observable cracks is shown in Fig. 8. The increasing
number of generated angular cracks after 1.5 mm of sliding
is shown in Fig. 9. The angular cracks approaching each
other from both sides after 2.3 mm of sliding and parallel
cracks becoming visible along the track borders are shown
in Fig. 10.
6. Discussion of surface loading mechanisms and
points of crack initiation
During the process of the spherical stylus moving with
linearly increased load over the coated surface, several
stress-generating mechanisms simultaneously inuence the
stress and strain conditions at each place in the coated sur-
face, and a continuously changing stress-eld is experienced.
Three main contact modes can be identied. First the
static mode, in which there is no sliding movement but only
a normal load applied by the indenter (Fig. 4). Second the
sliding mode, in which the indenter slides on the top layer
of the coating, the coating system carries the load and no
remarkable plastic deformation is observed below the coat-
ing in the substrate (Fig. 5). Third the ploughing mode, in
which the coating system can no longer carry the indenter
load and marked plastic deformation occurs in the substrate,
resulting in ploughing (Fig. 6). The maximum normal load
in the sliding mode before changing to the ploughing mode
represents the load carrying capacity of the coating system.
In the static mode at pre-loading, the rst principal stress
values are extremely low, less than 5 Pa, which can be con-
sidered primarily as numerical noise and initialisation of the
computation (Fig. 4). The stress distribution has fairly typ-
ical characteristics of an indentation problem: immediately
under the contact region the stresses are compressive and
outside the stresses are of tensile type. However, the distribu-
tion is not as symmetric as one might expect. For a single ma-
terial domain and system exhibiting innitesimal strains as
well as limited plasticity, such conditions would be met, but
for a system comprised of the coating and an elasticplastic
substrate the situation differs. This results from mismatch-
ing of the coating and substrate and primarily from the plas-
ticity experienced by the steel during different stages of the
loading process. Since the mismatch of mechanical proper-
ties is present even in the elastic loading regime, and since
the steel is forced to endure plasticity very early on in the
loading process, the minute asymmetricities in stresses are
present even during the pre-loading stage of the experiment.
The deformation pattern and its interrelation to relaxation of
K. Holmberg et al. / Wear 254 (2003) 278291 283
stresses can be inferred to be analogous to the shear lip and
plastic hinge formation commonly observed in the plasticity
of metallic materials [51]. Overall, the absolute stress val-
ues at pre-loading are so low in the simulated test conditions
that they are not of practical relevance.
In the sliding mode after 300 m of sliding, the stress
values have increased considerably to a level of 20 MPa
(Fig. 5a). Now the eld of the rst principal stress is rel-
atively symmetric, while the introduction of tension in the
direction of movement is seen to extend the tail end of the
contours. The deformations of the domain are practically
non-existent and in any case do not introduce coupling be-
tween the stress-elds and the deformed shape, which would
affect the stress-elds and their distributions. It is seen that
the maximum peak of the rst principal stress does not lie in
the plane of symmetry. This results from an increase in the
Fig. 5. (a) Sliding mode topographical stress-eld map showing maximum rst principal stresses on the coating and at the symmetry plane intersection
of the steel sample coated with a 2 m thick TiN coating and loaded by a sliding spherical diamond tip with 6 N load after a sliding distance of 0.3 mm.
Sliding direction is from left to right. The sliding tip is invisible in the gure. The contact zone is shown by the white curve. (b) Stress-eld map with
7 N load after a sliding distance of 0.5 mm. (c) Stress-eld map with 8 N load after a sliding distance of 0.7 mm. (d) Stress-eld map with 10 N load
after a sliding distance of 1.14 mm. (e) Stress-eld map with 12 N load after a sliding distance of 1.6 mm.
contact area, which in the numerical simulations moves the
region of maximum contact pressure away from the sym-
metry plane as such, affecting the stress distributions. Since
the indenter does not conform to the deformed shape of the
coating, this phenomenon arises and moves the point of max-
imum contact from the plane of symmetry. The maximum
stresses lie near the edge of the contact.
After 700 m of sliding the maximum stress values have
increased to 800 MPa and the effects of increased protru-
sion are visible (Fig. 5e). Stresses have increased within
the contact area and deformations of the substrate affect
the circular distribution of stresses. The tail end behind the
contact area extends further than in front of the indentation,
being the only visible illustration of applied tension. Finite
deformations are still small enough not to interfere with the
stress-elds. In other words, even though the plasticity of the
284 K. Holmberg et al. / Wear 254 (2003) 278291
Fig. 5. (Continued ).
K. Holmberg et al. / Wear 254 (2003) 278291 285
Fig. 6. (a) Ploughing mode topographical stress-eld map showing maximum rst principal stresses on the coating and at the symmetry plane intersection
of the steel sample coated with a 2 m thick TiN coating and loaded by a sliding spherical diamond tip with 15 N load after a sliding distance of
2.33 mm. Sliding direction is from left to right. The sliding tip is invisible in the gure. The contact zone is shown by the white curve. (b) Stress-eld
map with 26 N load after a sliding distance of 4.8 mm.
substrate interferes with the symmetric development of the
contact stress-elds, protrusion of the tip does not introduce
any elements to the stress-elds that could be interpreted
to arise from nite deformations. As such, the evolution of
stress-elds can be considered to be related primarily to the
tension and protrusion applied by the moving tip rather than
local deformations of the system. Noticeably the eld of the
rst principal stress has nearly symmetric arms extending
from the core region of contact. These result from the inter-
action of elastic stresses within the coating and the elasto-
plasticity of the substrate. Since the coating by denition
cannot exhibit plasticity, it is extremely rigid in compari-
son to the steel substrate. This results in a state of plastic
deformation within the substrate, since it is forced by com-
patibility to carry the high stresses of the elastic coating at
the interface. In order to be able to do so, the substrate de-
forms quite violently, exhibiting large plastic strains. The
peak strains form a typical shape, e.g. a point-like contact
or strains near a free surface, where the plastic strains ori-
ent and form a band of intense straining. The shape and
angular variation of stresses result from the prevailing state
of deformation, i.e. at locations of tensile stress minima the
286 K. Holmberg et al. / Wear 254 (2003) 278291
Fig. 7. Illustration of the cracks observed in scratch tests carried out in the authors laboratory. The experimental parameters are the same as those used
in the computer simulations. The cracks observed are angular cracks at 1 mm, parallel cracks at 2.3 mm and angular cracks joining at 3 mm.
deformations attain their maximum values along with the
von Mises stress. The stress-eld that results can be charac-
terised to have a tetra-armed star shape form. The maxi-
mum stresses at the locations of the tetra-armed stress-eld
occur equally throughout the coating, while the maximum
stresses at the points of the coating bending are located just
at the top of the coating.
After 1140 m of sliding the maximum stresses have
slightly decreased to 200 MPa due to volumetric enlarge-
ment of the stress-eld in the material and changes associ-
ated with the deformed shape, which affects the locations at
which peak stress values are attained (Fig. 5d). This is a rst
illustration that tension from movement of the indentation
tip as well as deformations near the edge of the contact start
Fig. 8. The scratch track after 1.1 mm of sliding with arrow showing the rst visible crack at the surface, which is an angular crack. Tip moving from
left to right.
to interact with the stress-elds. The stress peak previously
located within the contact area is seen to stretch in the di-
rection of movement outside the contact. The peak located
at the tail end at an approximate angle of 45

from the plane

of symmetry is seen to increase in value as well as volumet-
ric range. The fact that the stress peak forms changes in the
tail area demonstrates that the movement-induced tension of
the spherical tip couples with its protrusion, amplifying the
stress peak behind the contact area. Also, the stress peak is
located at the side of the groove, which implies that the lo-
cal deformations may contribute by increasing stresses due
to concentration effects.
After 1600 m of sliding the maximum stresses have
again increased to a level of 800 MPa (Fig. 5e). Since the
K. Holmberg et al. / Wear 254 (2003) 278291 287
Fig. 9. The border of the scratch track after 1.5 mm of sliding, showing an increasing number of angular cracks in the coating. Tip moving from left to right.
Fig. 10. The scratch track after 2.3 mm of sliding, showing angular cracks approaching each other from both sides and parallel cracks being generated
along the track borders. Tip moving from left to right.
288 K. Holmberg et al. / Wear 254 (2003) 278291
stress peak located outside the contact is in the plane of sym-
metry, the deformations at the edge of the groove are sub-
dued by the tension resulting from movement of the tip. The
stresses reach comparably high values within a band starting
from the region of contact all the way to the peak behind the
actual contact. This can be viewed as transition from com-
pressive and local plasticity dominated stress distribution to
a state where tension resulting from movement of the tip
begins to dominate.
In the early stage of the ploughing mode, once the tip
begins to penetrate the system to such extent that the defor-
mations of the hard ceramic layer become noticeable, the
peak values of tensile stresses begin to increase by several
orders of magnitude. This is a consequence of three factors:
(i) the linear-elastic behaviour of the coating necessitating
very high stresses in order for any deformation to occur, in
a fundamentally displacement controlled problem; (ii) the
uncoupled nature of the numerical model, i.e. damage accu-
mulation such as cracking of different types and associated
stress relaxation are not accounted for; and (iii) near plastic
collapse of the substrate due to (i) and (ii), further amplied
once the experiment progresses. Resulting from this, after
2330 m of sliding the maximum stress values in the model
have now considerably increased to an extremely high level
of 30 GPa (Fig. 6a). These stress values do not correspond
to the real situation, as the model does not take into account
the stress relaxation effect of the cracks produced earlier in
the groove which at this stage becomes remarkable. Due to
the ploughing action a marked increase in the contact area
can be observed.
The transition to a tension dominated stress state has
already occurred, and additionally the deformations of the
system have increased tremendously. This has also resulted
in an increase of coating stresses by roughly two orders of
magnitude. A strong tensile peak is seen behind the contact
region right before the part of the substrate that has experi-
enced elastic unloading. The fact that elastic unloading oc-
curs behind the contact at a comparably large distance from
the actual contact region, depending on relative properties
of the coating and the substrate rigidity-wise, introduces a
shape where the tension response is amplied by the de-
formed and unloading geometry. This can be interpreted as
a stress concentration effect in the direction of movement.
Additionally, since the groove begins to grow in depth, the
edge of the groove is responsible for a stress concentration
effect increasing tensile stresses away from the plane of
symmetry and resulting in a horseshoe type of stress dis-
tribution (Fig. 6). The region ahead of the contact is seen to
be under compression resulting from protrusion of the tip.
7. Computation of the fracture toughness of a coated
surface
On the basis of nite element analysis results and experi-
mental measurements for cracking in the scratch groove, the
Fig. 11. Crack eld under far-eld rst principal stress resulting from
superimposed membrane (N) and exural (M) loadings.
critical stress intensity factor for the coating was evaluated.
The problem setting is described in Fig. 11. The coating is
discretised to be under applied far-eld tension in an innite
domain, where this tension results from the combined effect
of applied membrane loading (N) and exural moment (M).
Within the domain of uniform rst principal stress, a crack
eld is modelled and its interaction effects accounted for in
the crack driving force solution. Only the behaviour of the
coating is considered, i.e. the substrate is not present in the
actual computations even though in nite element analysis
it naturally affects the produced values of the rst principal
stress.
Cracking is modelled to occur within the applied rst prin-
cipal stress-eld such that a eld of specic length through
coating cracks is formed with a given cracking density in
the direction of the stylus movement. Crack dimensions,
particularly of density, are inputted into to the K evaluation
throughout the experiment. The analyses are carried out for
two different types of cracks. First, fracture toughness is
computed for angular cracks (type 1: (a) in Fig. 3) located
at the edge of the groove and oriented at about a 45

an-
gle to the direction of the stylus movement. Second, cracks
forming at the bottom of the groove (type 2: (c) in Fig. 3)
and growing from the groove boundary towards the centre
of the scratch are analysed.
The stress intensity factor is computed following the solu-
tions of Paris [52], Isida [53] and Fichter [54]. The solution
of [52] is the limit of the other ones for crack elds where
the density of the crack eld is high enough to overcome
the effects of crack length. This occurs for the current prob-
lem when crack spacing and crack length are of the same
order. The solutions are given as K =
1

b f (a, b), where

1
is the rst principal stress, b crack spacing and f(a, b) a
non-dimensional function dependent on crack length, a, and
crack spacing/density.
The fracture toughness computation results incorporating
crack density and length interaction effects are presented
K. Holmberg et al. / Wear 254 (2003) 278291 289
Fig. 12. Coating fracture toughness as a function of stylus position in the
groove for angular cracks (type 1) and transversal tensile cracks (type 2)
related to the relation between crack length and crack spacing.
in Fig. 12 as a function of test position for both angular
(type 1) and transverse (type 2) cracks. First of all, it is noted
that different crack types are seen to produce different lev-
els of fracture toughness: angular cracks give values around
4 MPa m
0.5
, while transverse crack fracture toughness is ap-
proximately 7 MPa m
0.5
. The fracture toughness for a similar
TiN coating on steel bent to fracture in a four-point bending
test was calculated based on the experimental data given by
Wiklund et al. [29] and the result was an average K-value
of 7.5 MPa m
0.5
with variation of 1.5 MPa m
0.5
.
The interaction between crack spacing and crack length
is seen to be negligible in comparison to fracture tough-
Fig. 13. Coating fracture toughness as a function of stylus position in the groove for angular cracks (type 1) and transversal tensile cracks (type 2). They
are labelled as mean, lower and upper bounds, which means with respect to rst principal stress values. The stress intensity factor (SIF) is calculated
based on crack density.
ness dependency on stylus position and related scatter. Even
though a wide scale of a/b values is considered, the impact
on fracture toughness is only some percentages and can be
neglected. This means that the analyses can focus solely on
crack density as a characteristic crack eld measure.
By incorporating an estimate of accuracy of nite element
based stress results on characteristic crack scale, the results
of Fig. 13 are attained. They are labelled as lower and upper
bounds, which means with respect to rst principal stress
values. For both crack sizes, especially for transverse type 2
cracks, two effects are noted with respect to the toughness
distribution. Initially, at the beginning of the test and in
particular after about 2.25 mm, the fracture toughness values
begin to rise. For the latter parts this is understood to result
from decoupling of damage, cracking and stress-elds in the
nite element simulations.
At the beginning of the test, the effects of smaller crack
size and its relation to dimensions characteristic of nite el-
ement results are interpreted to increase scatter and uncer-
tainty. In part, this is related to the use of stress values, since
current values are surface values, and initially the coating
through thickness stresses has not stabilised to a pure mem-
brane type of loading. The fact that type 1 angular cracks
produce lower values than type 2 transverse cracks can be
attributed to the analysis procedure, i.e. the applied K solu-
tions are best suited for transversely oriented cracks and do
not account for the orientation exhibited by type 1 cracks.
The solution neglects the associated mixed-mode compo-
nent for the crack driving force and as such produces lower
values of fracture toughness.
This study shows that the scratch test can be used as an
instrument for determining the fracture toughness value of
a coated surface by combining numerical stress simulations
290 K. Holmberg et al. / Wear 254 (2003) 278291
of the coated surface and empirical measurements of the
location and density of the rst cracks to occur. The method
has been protected by patent application [55].
8. Conclusions
A FEM model for the scratch test contact system was
created and applied for the conditions of a 2 m TiN coating
on steel substrate. Based on this the following conclusions
can be drawn.
1. The created model describes the stresses and strains in-
corporating elastic and plastic behaviour in a contact ge-
ometry of the scratch test.
2. In the sliding mode a maximum tensile stress area is
generated in the tail part of the contact area.
3. With increasing load a new stress-eld with a tetra-armed
star shape grows around the contact area. At the tail arms,
stress concentrations are amplied at a distance of 12
times the contact length from the edge of the contact at
the border of the scratch groove. The magnitude of these
stress concentrations is of the same level as within the
contact area.
4. The stress concentrations at the tail arms travel with in-
creasing load to the plane of symmetry and increase in
magnitude considerably faster than within the contact
area, and become the dominating tensile stress as the
contact conditions change to ploughing mode.
5. After about 1 mm of sliding in the scratch test, a peak
area of maximum rst principal stress is generated in
the back-tail region at the border of the scratch groove,
creating the rst visible angular cracks in the coating with
this material-coating combination.
6. In the ploughing mode, the maximum tensile stresses are
located behind the contact at a distance of 0.51 times
the contact length from the back edge of the contact, and
they have a horseshoe shaped ridge of maximum values
with an opening in the sliding direction.
7. The change of the state of deformation from sliding mode
to ploughing mode characterises a loss of load carrying
capacity of the coated surface system.
8. The maximum stresses at the locations of the tetra-armed
stress-eld occur equally throughout the coating, while
the maximum stresses at the points of coating bending
are located just at the top of the coating.
9. The model determines the location of the rst cracks to be
initiated and this information can be used for calculating
the fracture toughness of the coating.
10. Critical fracture toughness is equal to the tensile stress
times the square root of half of the crack spacing (K
c
=

b/2), where the crack spacing is smaller than the

crack length, b is the perpendicular distance between
cracks, the tensile coating stress.
11. For determining the fracture toughness of 2 m thick TiN
coatings on a steel substrate, a suitable crack eld turned
out to be the transversal tensile cracks in the scratched
groove. Computations provided the best estimates for K
when the cracks had not grown through the whole groove.
For the studied case the fracture toughness of the TiN
coating was measured to be K
c
= 7 MPa m
0.5
.
References
[1] C. Donnet, Problem-solving methods in tribology with surface-
specic techniques, in: J.C. Riviere, S. Myhra (Eds.), Handbook of
Surface and Interface Analysis, Marcel Decker, New York, 1998,
pp. 697745.
[2] A. Erdemir, O.L. Eryilmaz, G. Fenske, Synthesis of diamond-like
carbon lms with superlow friction and wear properties, in:
Proceedings of the 46th International Symposium on American
Vacuum Society, 2529 October 1999, Seattle, USA, 2000.
[3] K. Holmberg, A. Matthews, Coatings TribologyProperties,
Techniques and Applications in Surface Engineering, Elsevier
Tribology Series 28, Elsevier, Amsterdam, The Netherlands, 1994,
p. 442.
[4] K. Holmberg, A. Matthews, Tribological properties of metallic and
ceramic coatings, in: B. Bhushan (Ed.), Handbook on Modern
Tribology, CRC Press, Boca Raton, USA, 2001, pp. 827870.
[5] K. Holmberg, M. Matthews, H. Ronkainen, Coatings tribology
contact mechanics and surface design, Tribol. Int. 31 (13) (1998)
107120.
[6] I. Singer, H. Pollock (Eds.), Fundamentals of Friction: Macroscopic
and Microscopic Processes, Kluwer Academic Publishers, Dordrecht,
NATO ASI Series, E220, 1991, p. 621.
[7] B. Bhushan (Ed.), Handbook of Micro/Nanotribology, CRC Press,
Boca Raton, USA, 1995.
[8] D.F. Diao, K. Kato, K. Hayashi, The maximum tensile stress on a
hard coating under sliding friction, Tribol. Int. 27 (4) (1994) 267
272.
[9] G.M. Parr, Measurement of mechanical properties by ultra-low load
indentation, Mater. Sci. Eng. A253 (1998) 151159.
[10] K. Holmberg, The basic material parameters that control friction and
wear of coated surfaces under sliding, TribologiaFinn. J. Tribol.
19 (3) (2000) 318.
[11] P. Burnett, D. Rickerby, The mechanical properties of wear-resistant
coatings. I. Modelling of hardness behaviour, Thin Solid Films 148
(1987) 4150.
[12] D. Rickerby, A. Matthews (Eds.), Advanced Surface Coatings: A
Handbook of Surface Engineering, Blackie, Glasgow, UK, 1991,
p. 364.
[13] W. Oliver, G. Pharr, An improved technique for determining hardness
and elastic modulus using load and displacement sensing indentation
experiments, J. Mater. Res. 7 (6) (1992) 15641583.
[14] A. Korsunsky, M. McGurk, S. Bull, T. Page, On the hardness of
coated systems, Surf. Coat. Technol. 99 (1998) 171183.
[15] H. Djabella, R.D. Arnell, Finite element analysis of the contact
stresses in an elastic coating on an elastic substrate, Thin Solid Films
213 (1992) 205219.
[16] H. Djabella, R.D. Arnell, Finite element analysis of contact stresses
in elastic double-layer systems under normal load, Thin Solid Films
223 (1993) 98108.
[17] K. Mao, Y. Sun, T. Bell, A numerical model for the dry sliding
contact of layered elastic bodies with rough surfaces, Tribol. Trans.
39 (2) (1995) 416424.
[18] L. Zheng, S. Ramalingam, Stresses in a coated solid due to shear
and normal boundary tractions, J. Vac. Sci. Technol. A13 (5) (1995)
23902398.
[19] B. Lawn, Fracture of Brittle Solids, Cambridge University Press,
Cambridge, UK, 1993, p. 375.
K. Holmberg et al. / Wear 254 (2003) 278291 291
[20] A. Bower, N. Fleck, Brittle fracture under a sliding line contact, J.
Mech. Phys. Solids 42 (9) (1994) 13751396.
[21] J. Beyth, Cracking of thin bonded lms in residual tension, Int. J.
Solid Struct. 29 (13) (1992) 16571675.
[22] M. Thouless, The role of fracture mechanics in adhesion, in:
Proceedings of the MRS Spring Meeting, 1988, Reno, Nevada, USA,
p. 12.
[23] S.J. Bull, Failure modes in scratch adhesion testing, Surf. Coat.
Technol. 50 (1991) 2532.
[24] S. Bull, D. Rickerby, Evaluation of coatings, in: D. Rickerby,
A. Matthews (Eds.), Advanced Surface Coatings: A Handbook of
Surface Engineering, Blackie, Glasgow, UK, 1991, pp. 315342.
[25] X. Li, D. Diao, B. Bhushan, Fracture mechanics of thin amorphous
carbon lms in nanoindentation, Acta Mater. 45 (11) (1997) 4453
4461.
[26] X. Li, B. Bhushan, Measurement of fracture toughness of ultra-thin
amorphous carbon lms, Thin Solid Films 315 (1998) 214221.
[27] M. Nastasi, P. Kodali, K. Walter, J. Embury, R. Raj, Y. Nakamura,
Fracture toughness of diamond-like carbon coatings, J. Mater. Res.
14 (5) (1999) 21732180.
[28] B. Ollivier, S. Dowey, S. Young, A. Matthews, Adhesion assessment
of DLC lms on PET using a simple tensile tester: comparison of
different theories, J. Adhesion Sci. Technol. 9 (6) (1995) 769784.
[29] U. Wiklund, M. Bromark, M. Larsson, P. Hedenquist, S. Hogmark,
Cracking resistance of thin hard coatings estimated by four-point
bending, Surf. Coat. Technol. 91 (1997) 5763.
[30] S. Oliveira, A. Bower, An analysis of fracture and delamination in
thin coatings subjected to contact loading, Wear 198 (1996) 1532.
[31] J. Malzbender, G. de With, Sliding indentation, friction and fracture
of a hybrid coating on glass, Wear 236 (1999) 355359.
[32] J. Malzbender, G. de With, Cracking and residual stress in hybrid
coatings on oat glass, Thin Solid Films 359 (2000) 210214.
[33] J. Malzbender, G. de With, Friction under elastic contacts, Surf.
Coat. Technol. 124 (2000) 6669.
[34] J. Malzbender, G. de With, Modeling of the fracture of a coating
under sliding indentation, Wear 239 (2000) 2126.
[35] A. Perry, The adhesion of chemically vapour-deposited hard coatings
to steelthe scratch test, Thin Solid Films 78 (1981) 7793.
[36] A. Perry, Scratch adhesion testing of hard coatings, Thin Solid Films
107 (1983) 167180.
[37] P. Steinmann, H. Hintermann, Adhesion of TiC and Ti(C, N) coatings
on steel, J. Vac. Sci. Technol. A3 (1985) 23942400.
[38] J. Valli, TiN coating adhesion studies using the scratch test method,
J. Vac. Sci. Technol. A3 (6) (1985) 24112414.
[39] J. Valli, A review of adhesion test methods for thin hard coatings,
J. Vac. Sci. Technol. A4 (6) (1986) 30073014.
[40] J. Valli, U. Mkel, A. Matthews, Assessment of coating adhesion,
Surf. Eng. 2 (1) (1986) 4953.
[41] A. Perry, Scratch adhesion testing: a critique, Surf. Eng. 2 (3) (1986)
183190.
[42] S. Bull, D. Rickerby, A. Matthews, A. Leyland, A. Pace, J. Valli, The
use of scratch adhesion testing for the determination of interfacial
adhesion: the importance of frictional drag, Surf. Coat. Technol. 36
(1988) 503517.
[43] J. von Stebut, R. Rezakhanlou, K. Anoun, H. Michel, M. Gantois,
Major damage mechanisms during scratch and wear testing of hard
coatings on hard substrates, Thin Solid Films 181 (1989) 555
564.
[44] M. Bromark, M. Larsson, P. Hedenquist, M. Olsson, S. Hogmark,
Inuence of substrate surface topography on the critical normal force
in scratch adhesion testing of TiN-coated steels, Surf. Coat. Technol.
52 (1992) 195203.
[45] European Standard, 1999, Advanced technical ceramicsmethods
of tests for ceramic coatings. Part 3. Determination of adhesion and
other mechanical failure modes by scratch test, prEN1071-3, p. 42.
[46] K. Holmberg, Surface fracture toughness measurement by the scratch
test method, TribologiaFinn. J. Tribol. 19 (3) (2000) 2432.
[47] P. Hedenquist, M. Olsson, S. Jacobson, S. Hogmark, Failure mode
analysis of TiN-coated high speed steel: in situ scratch test adhesion
test in scanning electron microscope, Surf. Coat. Technol. 41 (1990)
3149.
[48] M. Larsson, M. Olsson, P. Hedenquist, S. Hogmark, On the mecha-
nism of coating failure as demonstrated by scratch and indentation
testing of TiN and HSS, Surf. Eng., 1996, submitted for publication
(included in Uppsala Dr. Thesis by M. Larsson, No. 191/1996).
[49] Y. Wang, S. Hsu, P. Jones, Evaluation of thermally-sprayed ceramic
coatings using a novel ball-on-inclined plane scratch method, Wear
218 (1998) 96102.
[50] F.M. Beremin, A local criterion for cleavage fracture of a nuclear
pressure vessel steel, Metall. Trans. 14A (1983) 22772287.
[51] C.C. Chu, A. Needleman, Void nucleation effects in biaxially
stretched sheets, J. Eng. Mater. Technol. 102 (1980) 249256.
[52] P. Paris, Concepts in Fracture Mechanics, Course Notes, Paris
Productions, 1995.
[53] M. Isida, Effects of width and length on stress intensity factors of
internally cracked plates under various boundary conditions, Int. J.
Fract. Mech. 7 (1971) 301.
[54] W.B. Fichter, Stresses at the tip of a longitudinal crack in a plate
strip, NASA Technical Report TR-R-265, NASA Langley, 1967.
[55] PCT International Patent Application WO 02/18905 A1, Method for
measurement of the fracture toughness of a coating or a surface,
International publication, 7 March 2002.