Wear 350-351 (2016) 74–81

Contents lists available at ScienceDirect

Wear
journal homepage: www.elsevier.com/locate/wear

The measurement of wear using AFM and wear interpretation using

a contact mechanics coupled wear model
J. Furustig a,n, I. Dobryden b,c, A. Almqvist a, N. Almqvist b, R. Larsson a
a
Division of Machine Elements, Luleå University of Technology, 971 81 Luleå, Sweden
b
Division of Material Science, Luleå University of Technology, 971 81 Luleå, Sweden
c
Division of Surface and Corrosion Science, KTH Royal Institute of Technology, SE-100 44 Stockholm, Sweden

art ic l e i nf o a b s t r a c t

Article history: Detailed understanding of wear processes is required to improve the wear resistance and lifetime of
Received 31 July 2015 machine components. Atomic force microscopy (AFM) is used to measure surface height profiles with
Received in revised form high precision, before and after a wear experiment. The distribution and depth of wear on steel surfaces
21 December 2015
is then calculated using a relocation method. A numerical investigation of wear based on Archard's
Accepted 1 January 2016
equation is conducted on the same measured surfaces. A good correlation was found between the model
Available online 22 January 2016
and experiment for wear larger than a hundred nm. The wear mechanisms considered in the numerical
Keywords: simulation was thus found to be the cause of the majority of the wear. On the scale of tens of nm the
AFM correlation was limited, but the measured wear was still analysed in detail.
Wear
& 2016 Elsevier B.V. All rights reserved.
Topography
Numerical analysis

1. Introduction optical methods [26,22]. Optical 3D surface measurement techni-

ques such as vertical scanning (white-light) interferometry (VSI)
Wear is quantified by comparing measured quantities of a can have a lateral resolution of 0:55 μm. Furthermore, optical
sample, such as volumetric material loss or mass loss, before and techniques can introduce optical artefacts. In a recent publication
after a wear experiment [4]. It is well established that wear has a by Spencer et al. [28] the difference between AFM and VSI, was
large impact on performance and lifetime of machines [19]. The investigated. The difference between measurements using AFM
financial losses due to wear is a significant part of an industrialised and VSI on an engineering surface was within 7%, in terms of a
nations gross national product. This requires improved wear roughness parameter. In many cases, errors of this magnitude can
resistance, achievable through detailed understanding of wear be acceptable, while in other investigations, such as this work,
processes. One good approach to increase the understanding of the more accurate measurements of the surfaces are desired. Tradi-
wear process, is by means of numerical simulations. An advantage tional stylus profilometry, a contact method to measure line pro-
compared with experimental investigations is access to detailed files of surfaces, has much lower resolution than AFM. AFM is
information of the physical conditions in the contact during the considered herein as the most suitable technique with a highest
wear process. The exact influence of contact conditions on wear, possible lateral resolution of 0.2 nm and a vertical resolution up to
however, is still not well established. As the conditions can be 0.01 nm.
calculated, a deciding factor for improved wear models are accu- Relocation of the surface topography after the wear process is
rate wear measurements. The construction of reliable wear models an important part of the wear analysis. Cabanettes and Rosen [6]
requires exact measurements of wear with high accuracy and recently measured contacting surfaces in a valve train test rig
precise localization. before and after the wear process and relocalised the wear on the
There are several commonly used techniques to measure sur- surface. Many surface roughness parameters and their changes
face topography, such as stylus profilometry, atomic force micro- were investigated and a clear image of the wear distribution was
scopy (AFM) and 3D surface measurement techniques based on provided. As the physical conditions vary and no contact
mechanical analysis was conducted in [6], the study cannot be
n
directly used to formulate wear equations. Similarly, Gåhlin and
Corresponding author. Tel.: þ 46 70 7104144; fax: þ46 920 49 1047.
Jacobson [13] introduced relocalisation of topographies before and
E-mail addresses: [email protected] (J. Furustig),
[email protected] (I. Dobryden), [email protected] (A. Almqvist), after the wear process on the nanoscale, using AFM. No contact
[email protected] (N. Almqvist), [email protected] (R. Larsson). mechanical analysis was included in their study.

http://dx.doi.org/10.1016/j.wear.2016.01.002
0043-1648/& 2016 Elsevier B.V. All rights reserved.
 J. Furustig et al. / Wear 350-351 (2016) 74–81 75

Relocation methods give a detailed map of the wear distribu- pressure is then used to evaluate the wear depths applying
tion on a surface. Tribologically induced contact pressures and Archard's equation. An iterative procedure where the geometry is
local temperatures, used to determine the wear in a continuum updated and a new pressure distribution is repeatedly calculated is
model [2], can be estimated by means of numerical simulations. applied. The results from the measured wear depths from
Relations between the contact conditions and the wear are then experiments are compared with the wear depths predicted by
needed. One commonly used relation is Archard's wear equation. numerical simulation.
In general, applying an equation such as Archard's leads to a non-
linear initial value problem:
2. Method
dV
¼ f ðp; T; …Þ: ð1Þ
dt
A tribological wear test is designed to achieve controlled con-
The wear rate dV dt
depends on for example the pressure p and ditions in the interface between two disks in relative motion, on
temperature T. The model system further considers wear occurring the topography scale. One flat, smoother, and one rough steel
on top of a surface; sub-surface effects, e.g. crack formation, surface are used in a disk on disk test rig. Before and after the
require a different approach. experiment the surfaces are measured with AFM. By using three
The contact stress and temperature inside the interface indentation marks, a worn area on the rough surface is relocated
between rough surfaces in relative motion, can be modelled and after the wear test. The indentation marks were deep relative to
computed by means of numerical simulation methods. A mini- the wear depth. The measurements of the rough surface before
mum energy approach was used by Kalker [18] to formulate a wear is used as input to a numerical wear simulation. The simu-
relation between deformation and pressure. Similar relations lation calculates the gradual change of the surface topography due
between generated heat and temperature rise have been pre- to wear, including iterative updates of the pressure distribution.
sented by Carslaw and Jaeger [7]. Fast Fourier transform (FFT)
methods have been used to efficiently solve the system of equa- 2.1. Experimental details
tions for the temperature [11] and stresses [20] in rough surface
contacts. Multi level multi integration (MLMI) has also been The wear experiment is conducted under dry conditions in a
applied [5]. The MLMI method takes advantage of the fact that the disk on disk (face to face) clutch test machine. The test rig is fur-
local temperature rise or strain on a surface is mainly depending ther described in [9]. Two steel disks with an outer diameter of
on the nearby friction heat or contact pressure. Both methods (FFT 12 cm, as shown in [9], are used as test samples. The test rig is
and MLMI) are similar in terms of numerical efficiency and have chosen due to the conformal contact between the disks.
both been found to agree well with analytically solved special The steel disks were both ground with abrasive paper. In order
cases and with finite element method results. to achieve as close to steady state pressure as possible, one of the
Jamari [16] applied an elasto plastic model of the aforemen- surfaces was further polished in order to achieve a significantly
tioned type to model contact between rough measured surfaces. smoother surface finish compared with its counterpart. Gradually
Jamari also measured the surface topography of the same surfaces finer particles were used, with a final particles radius of 15:3 μm
after a wear process. The results from numerical simulation were for the rougher surface and 1 μm for the flat surface. The average
compared with the results from experiments. Jamari found a roughness (Sa ) after polishing of the flat surface was 6.1 nm,
reasonable agreement between measured wear depths and pre- measured over an area of 20 μm to 20 μm. The idea is that the
dicted plastic deformation. The matching and stitching method same asperities on the rougher of the samples will be in contact
which was used in [16] to determine wear depths, is only suitable with the smoother surface throughout the test. Due to wear on
for non-conformal contacts, where there is a large area of unworn these asperities, the contact pressure will slowly change. The effect
surface surrounding the worn area. A disadvantage with non- of temperature is limited through low sliding speeds and short
conformal contact is that the physical conditions, for instance run-times in the experiment.
contact pressure, vary due to both the roughness and the non- The relocation of the same surface area on the rough steel disk
conformity of the test sample. Relocation and subtraction methods after wearing, is of primary importance in order to study the
such as those used in [6,13,17], constitute a tool to measure wear topography changes due to wear. This is done by employing a
depth locally and this can be used for improving and validating Vickers hardness test to create indentation marks on the surface. A
wear models. The wear depth reported in the aforementioned large square area on the steel surface is marked with 4 indents by
work [6], was measured in the order of micrometers, while in [16] 300 g load. Then a triangular area inside the large area is similarly
and [13] sub micrometer wear depths were found. The wear marked using 3 small indents applying a load of 25 g. Three dif-
depths measured by Gåhlin et al. [13] was determined with a ferent areas were marked with this pattern. The depth of indents
precision in the order of 30 nm. is approximately 1:24 μm. The 4 indents in a square formation are
In this work, a relocation and subtraction method for measur- sizeable enough to relocate the reference points by visual
ing wear on the nanoscale using AFM is applied. In order to enable inspection. The smaller indents are visible by microscope and can
the study of the change in surface topography, the wear experi- be found as they are located in a known position relative to the
ment is designed to produce small wear volumes on the summits 4 deeper indents. The pattern of the indentations was used for
of asperities. Furthermore, a goal with the experimental design is relocating the surfaces after the wear process. The hardness was
to achieve predictable conditions in the interface between the also recorded for the indentations with 300 g and 25 g. The micro
contacting bodies. Conditions that vary mainly due to wear rather indentation using loads of 300 g and 25 g, gave an average Vickers
than due to variations in the relative positioning of the contacting hardness of 311 and 301, respectively. The hardness translates into
surfaces is strived for. The experiment design reaches this goal in an ideal plastic contact pressure limit of 3 GPa, a value acquired by
terms of pressure but not in terms of temperature. Wear grooves numerically reproducing the indentation marks. The flatter
with depths from a few tens to a few hundreds of nm are mea- smoother counter surface had a Vickers hardness of 218. The
sured. The contact pressure between a surface exhibiting the indentations were made after the grinding procedure.
topography measured with the AFM and a perfectly flat and The test is run at 200 N but some fluctuations occurred. The
smooth counter surface is calculated by means of numerical force is kept between 100 and 260 N during the whole test, with
simulations. The numerical simulation utilises FFT. The contact an average load of 197 N. The test is run for 30 s, with a rotational
76 J. Furustig et al. / Wear 350-351 (2016) 74–81

speed of 4 rpm. The moment curve also fluctuated, with a pattern Table 1
similar to that of the measured force curve. The resulting friction Material and system parameters, used to
numerically calculate the contact conditions in
coefficient varied between 0.06 and 0.26. The arithmetic average
terms of pressure and wear, between the two
value of the friction coefficient was 0.17. The variation in load, surfaces.
moment and friction were possibly due to stick-slip, imperfect
surface alignment of the samples or unwanted waviness of the Elastic modulus 210 GPa
samples. Plastic deformation limit 3 GPa
Wear coefficient 2  10  15 m3 /Nm
Nominal pressure 37 kPa
2.2. Surface measurements, filtering and relocation Poisson's ratio 0.29

The AFM measurements were performed with a Solver AFM

(NT-MDT) equipped with a 100 μm head scanner. The surface suitable for numerical evaluation of material removal. The simu-
areas of 70  70 μm were scanned with 1024  1024 data points lations, simulating 30 s of wear, are divided into 100 time incre-
in contact mode using a silicon probe CSC-21 (NT-MDT) with a tip ments. Between each time increment material is removed, due to
height of 7 μm, tip radius less than 20 nm and force constant of wear and plastic deformation. A new pressure distribution is then
2 N/m. The scan velocity was 80 μm/s. calculated. The process is then repeated 100 times. The nominal
The acquired AFM height images were subjected to standard pressure, calculated from the average load over the area of the disk
background removal procedures prior to use in simulation and was held constant at 37 kPa.
relocation. The first data transformation procedure was a line fit- The parameters used for the material pair are summarized in
ting method. A second order regression curve is computed and Table 1. The wear coefficient, estimated from the total wear
subtracted from every recorded horizontal line (fast-scan direc-
volume measured in the experiment, is 2  10  15 Pa  1 . The coef-
tion). The fitting procedure did not significantly influence the real
ficient was calculated from Eq. (2), with a node average wear
surface waviness of the measured surfaces used further in simu-
height of 1.4 nm. The total sliding distance in the measured area
lations. Finally, a simple linear 2D plane was removed from each
image to remove eventual sample tilt and/or drift in the slow scan was estimated to 630 mm. Values on the dimensional wear coef-
direction. ficient in the range of 5  10  16  3:7  10  11 Pa  1 for different
The measured, filtered and planarized surfaces have separate steel–steel contacts can be found in the literature [3,10,23].
coordinate values for each x, y and z, before and after wear. One of The area surrounding the indentation marks can be plastically
the centre points in the indentation marks is therefore used as a deformed during the indentation procedure. Large asperities will
relocation reference for a common origin of the measurements of then protrude on the surface around the indentation marks. This
the surfaces before and after wear. The coordinate systems of the causes the area containing indentation marks to carry a greater
two samples are adjusted thereafter. Following this, the coordinate part of the load compared to the nominal pressure on the disks.
systems are rotated around the z-axis, so that one more centre of The pressure in an area containing indentation marks can be
an indentation mark is aligned. Since the tilt has already been estimated by means of numerical simulations, and is needed for
removed, the surfaces are considered aligned at this point, with wear simulations. This can be achieved by calculating the contact
the indentation marks at the same coordinate values before and pressure on an artificial surface, containing measured features.
after wear data. In order to analyse wear depth, the z-coordinates Numerical simulations are used in this work to estimate wear
of the surface before wear, are subtracted from the surface coor- distribution under the assumption of a flat smooth counter sur-
dinates after wear. In order to conduct this subtraction the sur-
face. Protruding indentation marks in the measured zone carry
faces are first interpolated on the same x–y grid. More precisely,
relatively large load, thus relieving pressure from nearby areas of
the linear interpolation routine implemented in the MATLAB
similar size. By extending the measured area with an artificial
function ‘griddata’ is employed.
surrounding area, having similar roughness but no protruding
Surface patches with and without indentation marks are gen-
erated this way. The area including indentation marks, on which indentation marks, the ratio of pressure carried by the measured
numerical simulations were conducted, was of dimensions area and the nominal pressure can be estimated.
50 μm  50 μm. One smaller area, containing no indentation The size of an area containing indentation marks corresponds
marks, was of dimensions 20 μm  20 μm, and the other smaller to the size of the measured area A, depicted in Fig. 2. The nominal
area was of dimensions 12 μm  12 μm. pressure of the patch containing indentation marks was found to
be 1.1 MPa, about 30 times larger than the pressure compared with
2.3. Simulation details an area of the same size far away from the indentation marks.
If the counter surface is completely flat, the wear on the ground
The pressure distribution on the rough surfaces is calculated by surface is expected to occur only on the large asperities in the
applying a contact mechanics solver as described by Sahlin et al. numerical simulation. The shape of a real surface, however, may
[27]. The model is as the one presented by Tian and Bhushan [30] still cause contact in areas near the indentation marks. As the
and the numerical solution procedure is a development of the counter surface (assumed smooth in the numerical simulations)
procedure presented by Stanley and Kato [29], in order to also corresponds to a very large area in the experiment, it is not fea-
account for plastic deformation of the contacting surfaces. This sible to measure its exact shape. Therefore, the pressure is
linear elastic – perfectly plastic contact mechanics solver has also
expected to be zero from counter surface contact. However, the
been applied by other researchers, e.g., by Almqvist et al. [1] or
nominal pressure of the experiment is used as input to numerical
more recently, Pérez-Ràfols et al. [25]. Wear was introduced in the
simulations of wear in areas B and C (depicted in Fig. 2), in order to
system by removing material according to the relation
investigate whether wear, by the mechanisms that are accounted
h ¼ kps: ð2Þ for in the numerical simulation, is a possible cause of the mea-
were h is the height of material removed on a node, p is the nodal sured nanowear.
pressure, and s is the distance of sliding during one time incre- The numerical simulation was independently performed on the
ment. According to [24], this form of Archards wear equation is three different areas depicted in Fig. 2.
 J. Furustig et al. / Wear 350-351 (2016) 74–81 77

3. Results and discussion volumes in detail. The specific areas B and C contained very few
measurement artefacts. The numerical simulations are performed
Fig. 1 shows the height surface profile with three indents before independently on the different areas. In areas B and C, numerical
(a) and after (b) the wear process, in an area measured using AFM. simulations are utilized as a qualitative tool, to aid the analysis.
The same three indentation marks are clearly observable on the The wear simulation for areas B and C also makes it possible to
images allowing us to compare surface areas before and after the compare the wear mechanisms on the larger (area A) scale with
wear test. These indents are surrounded by the highest asperities those on the smaller (areas B and C) scale.
in the measured area, and are the result of plastic deformation
during the indentation process. It can be seen in Fig. 1 that the 3.1. Wear depth and simulation results
height of asperities around indents is lower and the asperity tips
have been flattened due to the wear process. As an example, the The wear depth in area A, in terms of simulation results and
average heights of asperities around one indent, caused by plastic measurements, are both shown in Fig. 3. It is clear that most of the
deformation during the indentation before the wear test, were wear occurs around the indentation marks. The wear surrounding
294 7 34 nm, as measured with AFM. The average height of indentation marks with a depth of 100–250 nm, is largely located
asperities around indents had decreased to 209 7 10 nm when in the same spots in simulation and measurements. This wear on
measured after the wear test. Any measurement artefacts were the asperities surrounding indentation marks, is due to contact
excluded from further analysis by low pass filtering. In order to pressure between the measured area and the relatively flat
investigate wear depths at different length scales, three sub-areas counter surface. The wear manifests in the simulation partially due
of the measured area were investigated further. to plastic deformation and partially due to the implementation of
The measured area is investigated using the numerical model. Archard's equation. The plastic deformation occurs in the very first
Three parts of the measured area is analysed separately as depic- few time steps, rapidly decreasing with time. In the type of
ted in Fig. 2. Area A, with measured wear depths in the order of numerical simulation that was conducted, such a behaviour is
hundreds of nm, is marked with an A in Fig. 2. Area A, contains expected, as the peak pressures decrease with the smoothing
two indentation marks and is the largest of the investigated areas. effect of wear. In the following time steps there is a state of gradual
Area B and C contain no indentation marks. Areas A, B and C are material removal from the apex of asperities, following Archard's
aligned so that the sliding direction is vertical. If before wear and equation.
after wear data are to be compared, an overlapping region In area A, alongside with the large wear depth measured on the
between the measurements is required. This, together with the asperities surrounding indentation marks, it was also possible to
desired sliding direction limits the size and location of area A. Area measure wear depths smaller than 100 nm in regions of area A
B and area C are smaller areas, partially sub-areas of Area A. Areas located at a distance from the indentation marks. These smaller
B and C are studied in order to investigate nano sized wear wear depths are clearly observable in Fig. 3. However, all of the
nanoscale wear is not captured by the simulation. One possible
explanation for this is the assumption applied in the model, that
the counter surface is flat. The assumption leads to a model where
the highest points on the rough surface are the only points that
can come into contact. In the experiment, the flat surface rough-
ness is controlled by polishing with finer particles compared with
the rough surface. However, this polishing procedure cannot be
expected to completely eliminate waviness. Furthermore, the flat
surface is polished to a finer texture before the experiment, but
wear during the experiment would still cause a change in shape of
the counter surface and wear particles could be produced. The

Fig. 2. The measured surface before wear as seen from above. The measured sur-
Fig. 1. The AFM height images showing the surfaces, before (a) and after (b) wear. face is divided into three areas: A, B and C.
78 J. Furustig et al. / Wear 350-351 (2016) 74–81

Fig. 3. To the left, the AFM height image showing the unworn surface in area A is shown. In the centre is the wear depth measured experimentally. To the right, the
corresponding wear depth predicted by numerical simulations is depicted. Positive wear depths have been excluded from the figure.

Fig. 4. To the left is the AFM height image showing the unworn surface in area B. In the centre, the wear depth of the area illustrated in area B is shown. Straight vertical wear
scars are due to hard asperities on the counter surface that have ploughed the illustrated surface. To the right the corresponding numerical result is seen. Positive wear
depths have been excluded from the figure. Wear depths of 20 nm or more are shown with the same colour. (For interpretation of the references to colour in this figure
caption, the reader is referred to the web version of this paper.)

discrepancies between simulation and experiment can originate the assumptions and limitations of the numerical model. For
from these sources. instance, the smooth surface is polished with particles of a smaller
The wear depths in areas B and C, together with simulation diameter which may lead to sharper asperities on the nanoscale.
results, are illustrated in Figs. 4 and 5. The figures of area B (of Therefore, the smooth counter surface assumption of the model
dimension 20 by 20 μm) and area C (of dimension 12 by 12 μm) loses its validity. Especially long, straight, valley like wear features
demonstrate depth and distribution of nanowear. seen in the observed wear in Fig. 4 at e.g. x ¼ 0 nm and x ¼ 9 nm,
A part of the measured wear marks in area B is also predicted indicate that hard asperities have been ploughing the surface in
by means of numerical simulations, as can be seen in Fig. 4. sliding direction. The plastic deformation is considered in a sim-
However, the simulation does not predict a majority of the wear plified manner in the numerical simulation, as the applied method
observed from the experiment and the deepest observed wear relies on the half-space assumption and consequently disregards
scars are not predicted by the numerical simulation. A disagree- some sub-surface effects which could occur in the experiment.
ment between the measured wear and the wear predicted by Also, it is well known that wear in the valleys of the surfaces could
means of numerical simulation is also clear in area C, see Fig. 5. occur due to third bodies, an effect not considered in the model.
The numerical simulation predicts wear on the highest surface Perhaps the most likely reason for topography changes at this
features. The measured wear, seen in the centre of Fig. 5, reveals scale could be wear by third bodies. The contact conditions can be
significant amounts of wear on other locations of the surface. dramatically affected if a particle gets trapped between the two
The disagreement between simulation and experiment can be surfaces. Wear scars in Fig. 4 that are not vertical (the direction of
due to several possible reasons. The main reason is thought to be sliding), are probable candidates for effects from third body wear.
 J. Furustig et al. / Wear 350-351 (2016) 74–81 79

Fig. 5. To the left is the AFM height image showing the unworn surface in area C. In the centre is the measured wear depth of the area. To the right the corresponding
numerical result is seen. Positive wear depths have been excluded from the figure. Wear depths of 20 nm or more are shown with the same colour. The hatched lines marked
with α, β, γ and δ show the location of the surface profiles depicted in Fig. 6. (For interpretation of the references to colour in this figure caption, the reader is referred to the
web version of this paper.)

Fig. 6. Measured profiles in Area C, before and after wear.

Most of the wear seen in the centre of Fig. 5, is at an angle with the Figs. 4 and 5. It is possible that material (wear particles), down to
sliding direction. This indicates that a large particle passed over the size of atoms could have been detached and reattached on
this area in a single event, causing a majority of the wear in Area C. other locations during the wear process. A wear mechanism is
A further issue with applying the wear model on the nano- inferred, such as shown by molecular dynamics simulations by
metre scale, is that the mechanisms of wear are governed by non- Vargonen et al. [31]. In the wear mechanism, particles (atoms) are
continuum effects, such as adhesive forces between molecules. shovelled along the contacting bodies and attach on new cites.
Molecules or atoms detach and can possibly be reattached else- The study of nanowear with AFM can also be affected by other
where. This type of wear, usually referred to as ‘atomic attrition’, is factors, such as instrumental imperfections and data changes
not inconsistent with the presence of third bodies or sharp aspe- caused by filtration and alignment procedures. In order to mini-
rities on the counter surface. On the contrary, a high probability of mize effects from hysteresis and creep of the piezoceramic tube
breaking and formation of inter atomic bonds requires high energy scanner, a scanner with capacitive scanner displacement sensors
densities, in terms of stress and/or temperature [12,15,14]. Stress was used for height measurements and closed-loop lateral scan-
and temperature in a sliding interface is certainly affected by the ning. However, the sensors can introduce a small jitter i.e. a small
counter surface asperities and the presence of third bodies. shift in the feedback circuit for the lateral x- and y- scanning
Thermal distortion of the material, plastic deformation or a directions. This requires a correction of the measured raw data.
deformation of the reference points, are other possible explana- Also, the very end of the cantilever tip can be worn slightly during
tions for the changes in topography seen in the left side of the scanning and thus lower spatial resolution and increased tip
80 J. Furustig et al. / Wear 350-351 (2016) 74–81

convolution with the surface. The most significant effect on the simulation was successfully utilized on the small scale to interpret
measured z-position, is likely caused by the procedures of filtra- the nanowear measured with AFM, by excluding wear mechan-
tion of the data before and after wear test. It is possible that an isms. The results from the applied method show that the wear
error as large as a few nm is introduced by the filtering process. mechanism differs for large scale (Area A) and small scale (Areas B
The possible offset in x- and z- direction could be estimated and C) wear.
from the surface profiles. By analysing the surface profiles, the The deterministic calculation of wear on rough surfaces con-
possible mechanisms of nanowear discussed above, can be dis- sequently includes accurate consideration of the actual topo-
tinguished. The cross-sections of surface measurements before graphy over the whole nominal contact area. It is therefore
and after the wear test are shown in Fig. 6. The profiles are marked recommended that the validation of deterministic wear models
with Greek letters α, β , γ and δ and are recorded perpendicular to intended to consider topographical effects are performed on small
the sliding direction. The location of the illustrated profiles are nominal contact areas. Optimally, the simulation is sufficiently
shown by the dashed lines in the image to the right in Fig. 5.
effective for the whole contact area to be resolved to the desired
Wear in the order of 2 nm to tens of nm can be observed from
resolution.
the surface profile plots. Some parts of the surface remain almost
The threshold between when a continuum analysis is applic-
unchanged as can be seen in the change of the cross-section plots
able and when molecular effects need to be considered has been
before and after the wear test. This indicates that the surfaces
studied numerically by Luan and Robbins [21], but the threshold
were not in contact at these locations and no third body wear
has not been exactly located, according to Colao¸o [8]. It is thus
occurred. Also, this indicates that the observed changes in height
is mainly due to wear and not the result of a constant offset in z- interesting to note: the agreement, in this work, between simu-
direction caused by applied filtration and alignment procedures. lations applying a continuum mechanics approach and experi-
The possible z-offset was estimated by comparing the depth of two ments, falls of between hundreds and tens of nm.
uninfluenced and unchanged deep valleys in area C. The estimated
discrepancy between the depths of both valleys before and after
the test was found to be 1:4 nm. From comparing the two points, Acknowledgements
the value of 1:4 nm can be seen as a rough indication of the wear
depth measurement accuracy in this study. Sincere gratitude goes to Nowshir Fatima and Kim Berglund,
As seen from the cross-section plots depicted in Fig. 6, the Luleå University of Technology, for their aid with the wear test
surface profile changes significantly at some locations. For equipment. This work was funded by the Swedish foundation for
instance, some surface features appearing as peaks in the before strategic research (Proviking). The Kempe Foundations SMK-2546
wear test profile, are getting split into a few peaks, for example is acknowledged for funding the AFM.
around x ¼ 4 μm in the profiles labelled α and δ. Another inter-
esting feature is that much wear occurs in the valleys, for example
at x ¼  5 μm in the profile labelled β and at x ¼  2 in the profile
References
labelled δ. Similarly, much wear occurred on the side of the surface
peak at x ¼ 5 in the profile labelled α. This can be due to third
[1] A. Almqvist, F. Sahlin, R. Larsson, S. Glavatskih, On the dry elastoplastic contact
bodies but also due to interaction with rough asperities from the of nominally flat surfaces, Tribol. Int. 40 (4) (2007) 574–579.
counter surfaces. Also, it is expectable that the observed nanowear [2] J. Andersson, A. Almqvist, R. Larsson, Numerical simulation of a wear experi-
contains a contribution from ‘atomic attrition’ wear at this scale ment, Wear 271 (11–12) (2011) 2947–2952.
[3] J. Archard, W. Hirst, The wear of metals under unlubricated conditions, Proc. R.
but this cannot be easily determined from the presented results. Soc. Part A 236 (1956) 397–410.
[4] P. Blau, Needs and challenges in precision wear measurements, J. Test. Eval. 25
(1997) 126–225.
[5] A. Brandt, A.A. Lubrecht, Multilevel matrix multiplication and fast solution of
4. Conclusions integral equations, J. Comput. Phys. 90 (1990) 348–370.
[6] F. Cabanettes, B.-G. Rosén, Topography changes observation during running-in
The contact between a flat, smooth surface and a rougher of rolling contact, Wear 315 (2014) 78–86.
[7] H. Carslaw, J. Jaeger, Conduction of Heat in Solids, Clarendon Press, Oxford,
surface was investigated in a relative sliding, disk on disk, wear
1959.
test machine. AFM was used to quantify wear on a surface by [8] R. Colaço, An afm study of single-contact abrasive wear: the Rabinowicz wear
comparing the surface topographies before and after wear. The equation revisited, Wear 267 (2009) 1772–1776.
measurement showed that most of the wear occurred on the [9] N. Fatima, P. Marklund, R. Larsson, Water contamination effect in wet clutch
system, Inst. Mech. Eng. Proc. Part D: J. Automob. Eng. 227 (2013) 376–389.
highest asperities surrounding indentation marks. Also, the [10] A. Flodin, S. Andersson, Simulation of mild wear in spur gears, Wear 207
nanowear of tens nm and more on areas away from the indenta- (1997) 16–23.
tion marks was measured. [11] J. Gao, S. Lee, X. Ai, H. Nixon, An fft-based transient flash temperature model
for general three-dimensional rough surface contacts, J. Tribol. 122 (2000) 519.
A numerical model simulating the contact in the wear test [12] B. Gotsmann, M.A. Lantz, Atomistic wear in a single asperity sliding contact,
resulted in wear coinciding location wise with the measured wear. Phys. Rev. Lett. 101, 16 September 2008, 125501.
The pressure distribution on an area larger than the measured area [13] R. Gåhlin, S. Jacobsson, A novel method to map and quantify wear on a micro-
scale, Wear 222 (1998) 93–102.
was estimated by means of numerical simulations. The asperities [14] T.D.B. Jacobs, R.W. Carpick, Nanoscale wear as a stress-assisted chemical
formed during indentation were large enough to carry a sig- reaction, Nat. Nanotechnol. 8 (2013) 108–112.
nificantly larger load compared with the surrounding area. This [15] T.D.B. Jacobs, B. Gotsmann, M.A. Lantz, R.W. Carpick, On the application of
transition state theory to atomic-scale wear, Tribol. Lett. 39 (2010) 257–271.
suggests that the pressure distribution and thus wear, depends on [16] J. Jamari, Running-in of rolling contacts (Ph.D. dissertation), Universtiteit
the exact surface topography over a large area. A good agreement Twente, 2006.
was found between the simulation and experiment for the wear [17] J. Jamari, D. Schipper, An elastic-plastic contact model of ellipsoid bodies,
Tribol. Lett. 21 (3) (2006) 262–271 〈http://www.scopus.com/inward/record.
larger than 100 nm. However, the simulation of the wear on a scale
url?eid¼2-s2.0-33744944571partnerID ¼40〉.
of tens of nm only correlated on specific locations with the mea- [18] J.J. Kalker, Variational principles of contact elastostatics, IMA J. Appl. Math. 20
sured wear and did not show a good agreement. This behaviour (2) (1977) 199–219.
was expected and indicates that the nanowear occurs due to a [19] H.J. Kim, S.S. Yoo, D.E. Kim, Nanoscale wear: a review, Int. J. Precis. Eng. Manuf.
13 (2012) 1709–1718.
presence of third bodies, whose influence becomes more sig- [20] S. Liu, Q. Wang, G. Liu, A versatile method of discrete convolution and fft (dc–
nificant on the smaller scale. Nevertheless, the numerical fft) for contact analyses, Wear 243 (1–2) (2000) 101–111.
 J. Furustig et al. / Wear 350-351 (2016) 74–81 81

[21] B. Luan, M.O. Robbins, The breakdown of continuum models for mechanical [28] A. Spencer, I. Dobryden, N. Almqvist, A. Almqvist, R. Larsson, The influence of
contacts, Nature 435 (7044) (2005) 929–932. afm and vsi techniques on the accurate calculation of tribological surface
[22] R. Ohlsson, A. Wihlborg, H. Westberg, The accuracy of fast 3d topography roughness parameters, Tribol. Int. 57 (2013) 242–250.
measurements, Int. J. Mach. Tools Manuf. 41 (13–14) (2001) 1899–1907. [29] H.M. Stanley, T. Kato, An fft-based method for rough surface contact, J. Tribol.
[23] U. Olofsson, S. Andersson, S. Björklund, Simulation of mild wear in boundary 119 (3) (1997) 481–485 〈http://www.scopus.com/inward/record.url?eid ¼ 2-
lubricated spherical roller thrust bearings, Wear 241 (2000) 180–185. s2.0-0031190183partnerID ¼40〉.
[24] P. Põdra, S. Andersson, Simulating sliding wear with finite element method, [30] X. Tian, B. Bhushan, A numerical three-dimensional model for the contact of
Tribol. Int. 32 (2) (1999) 71–81. rough surfaces by variational principle, J. Tribol. 118 (1) (1996) 33–42 〈http://
[25] F. Pérez-Ràfols, R. Larsson, A. Almqivst, Modelling of leakage on metal-to- link.aip.org/link/?JTQ/118/33/1〉.
metal seals, Tribol. Int. 94 (2016) 421–427. [31] M. Vargonen, Y. Yang, L. Huang, Y. Shi, Molecular simulation of tip wear in a
[26] C. Poon, B. Bhushan, Comparison of surface roughness measurements by single asperity sliding contact, Wear 307 (1–2) (2013) 150–154.
stylus profiler, afm and non-contact optical profiler, Wear 190 (1) (1995)
76–88.
[27] F. Sahlin, R. Larsson, A. Almqivst, P. M. Lugt, P. Marklund, A mixed lubrication
model incorporating measured surface topography. Part 1: theory of flow
factors, Proc. Inst. Mech. Eng. Part J: J. Eng. Tribol. 224 (4) (2010) 335–351.