Links between energy dissipation and wear mechanismsin solid epoxy/epoxy sliding contact

Olga Smerdova n, Denis Mazuyer, Juliette Cayer-BarriozLaboratoire de Tribologie et Dynamique des Systèmes, CNRS UMR 5513, Ecole Centrale de Lyon, 36 Avenue Guy de Collongue, 69134 Ecully Cedex, France

a r t i c l e i n f o

Article history:Received 14 February 2014Received in revised form9 April 2014Accepted 12 April 2014Available online 2 May 2014

Keywords:WearPolymerSliding frictionEnergy dissipation

a b s t r a c t

This paper covers wear and energy dissipation of solid epoxy induced by the alternative rubbingbetween two samples of identical thermosetting polymer. Varying normal load, sliding velocity andsliding distance, the authors were able to define and discuss wear and friction laws and associatedenergy dissipation. Moreover, traces of several wear mechanisms were distinguished on the wornsurfaces and associated with applied conditions. Observed under higher velocity, polymer softening andlocal state transition were explained by surface temperature estimate and confirmed by infra-redspectroscopy measurements. To conclude this study, all observed phenomena are classified into twowear scenarios according to sliding velocity.

& 2014 Elsevier Ltd. All rights reserved.

1. Introduction

For many years, tribologists have attempted to predict wear ofdifferent tribological systems. In spite of a large number ofpublished experimental and theoretical investigations of thisproblem, a universal wear law, taking into account all parametersaffecting wear, has not yet been proposed. The most famous andone of the oldest wear laws is the law of Archard [1]. Archardconsidered the hypothesis of plasticized contact, where the realcontact area is the ratio of the normal force over the materialhardness, and defined the volume of wear Vwear as follows:

Vwear ¼ kwFNLk=H ð1Þwhere kw is the wear coefficient varying from 10�7 to 10�2 fordifferent materials, H is the hardness of the softest surface, Lk is therelative sliding distance between the materials or kinematic lengthand FN is the normal force. The low value of kw indicates that wear isonly caused by a very small proportion of asperity contacts. Further-more, it does not depend on normal load or sliding velocity.

Many investigators observe similar tendencies in differentfrictional systems. For example, the use of an energetic approachis able to provide similar answers to various wear problems [2–6].By using this approach, the dissipated energy (resp. power) iscalculated from experimentally measured frictional force andsliding distance (resp. sliding velocity). It is then related to themeasured wear volume in order to define the mechanisms ofdissipation and wear.

For instance, Ramalho et al. [3] measured a linear wear volume–energy dissipation dependence for a couple of metals using a cross-cylinder tribometer. A similar linear dependency was observed byFouvry et al. [4,5] for fretting contacts of metals and hard coatings.In their later work on fretting wear [6], the same authors were ableto associate the wear mechanisms, such as plastic deformations,formation of tribologically transformed structure and stable wearregime, with the total amount of dissipated energy. Similar resultswere obtained by Huq et al. [7] in fretting wear during experimentsin humid and dry atmosphere.

Aghdam et al. [8,9] recently proposed a method to predictfriction power loss and wear rate from measurements of contacttemperature. Moreover, in sliding reciprocal friction experimentson steel alloy/brass pin-on-plate couple, they were able to corre-late the average power dissipation to average contact temperaturerise. As with the other works above-mentioned, these authors alsoobserved a linear correlation between average wear rate andaverage power dissipation.

The linear dissipated energy–wear rate relationship was alsoshown to be correct for polymer/metal contacts. In the work ofColaço et al. [10], the worn volume UHMWPE increases linearlywith the dissipated energy, independently of the lubricant, materialused as a counterbody, and of the surface finishing of both thepolymer and the counterbody. To explain their results, theseauthors assumed a negligible contribution to the wear process ofcontact temperature rise. According to them, the major mechanismsresponsible for energy dissipation of the polymer are viscous orplastic deformations Edef and generation of worn particles Ewear, i.e.

Ed ¼ Edef þEwear ð2Þ

Contents lists available at ScienceDirect

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

Tribology International

http://dx.doi.org/10.1016/j.triboint.2014.04.0090301-679X/& 2014 Elsevier Ltd. All rights reserved.

n Corresponding author. Present address: Department of Engineering, Universityof Cambridge, Cambridge CB2 1PZ, United Kingdom. Tel.: þ44 1223 7 48525.

E-mail address: olga.smerdova@eng.cam.ac.uk (O. Smerdova).

Tribology International 77 (2014) 148–159

Many experiments have been carried out to study PMMAfretting wear [11–15], in which in situ and after test observationsrevealed a dependence of wear debris shape and surface damageon loading type and conditions. For instance, Geringer et al. [11]investigated a fretting wear of PMMA against a metal. Theseauthors compared cumulated dissipated energy and C–C bondenergy in order to verify whether the frictional energy wasdissipated in breaking these bonds. Because these values differby a factor of 1000, they concluded that the dissipated energy ismainly used to mix and transform the third body and to expelworn particles from the contact. The good agreement between thetheoretical value of the Fuller-Tabor parameter and that calculatedfrom the experiments encouraged them to consider that onlyadhesion and separation load due to asperities contribute to thedissipated energy, ruling out the energy dissipation contribution:

Ed ¼ EadhþEsep ð3Þ

When frictional loading is cyclic and prolonged, fatigue wearmechanism usually appears. The proportion of fatigue wear con-tribution due to elastic deformations, and abrasive wear contribu-tion due to plastic deformations, depends on elastic modulus andsurface roughness. Dubourg et al. [16] investigated nucleation andpropagation of fatigue cracks into epoxy under fretting wearconditions. In these experiments, epoxy material had a glasstransition above 100 1C and demonstrated brittle behaviour understatic fatigue and wear fretting loading. The material transparencyhelped them to observe that the effects of the material micro-structure on crack propagation mechanisms were predominatedby the mechanical stress-strain state undergone by the material.

The present paper is aimed at introducing some new insightson the wear mechanisms occurring during the friction of solidthermosetting polymer against similar polymer. The evolution ofthese mechanisms during sliding and increase of surface damageunder different tribological conditions is a key to understandingand prediction of failure of these materials. Plane/plane config-uration is chosen as it is more representative of real problems. Inaddition, a wide range of rather severe tribological conditions isapplied in order to let the system develop the maximum of wearregimes under given geometrical and loading configuration. Onecan notice, that the conditions applied in this study are at thesame time close to fretting, because the amplitude of the slidingmotion is just a triple of the contact width, and different from it,because of the macroscopic size of our plane/plane system.

The energetic approach is adopted in order to link the dissipatedenergy to wear. Firstly, a relationship between the wear rate anddissipated energy is examined. Secondly, we relate the dissipatedenergy to the damaged surface area, which seems more appropriatein this case of severe wear and surface deformations. This approachallows us to discuss the energy necessary to develop surface damageby different wear mechanisms which are detected on thoroughlyexamined worn surfaces after different numbers of sliding cycles.An additional surface analysis is performed on the worn specimens.Polymer softening and yielding observed on worn surfaces, as wellas sliding-induced crosslinking, had suggested possible polymerstate transformations, which encouraged us to estimate contacttemperatures during the sliding under different conditions. Finally,all observed wear mechanisms are classified into two qualitatively

different types, and a wear retrospective is proposed to evaluate thescenarios of the genesis of both types.

2. Materials and methods

2.1. Materials

All friction experiments presented in this paper are carried outon cross-linked epoxy resin HexPlys RM10.1. This epoxy resin ismoulded into a large flat plate of 470.5 mm thickness, which iscut into rectangular samples with the approximate dimensions of3074 � 1572 mm2 for the largest specimens and 872 �671 mm2 for the smallest ones. The apparent sliding area of eachslider is carefully measured using image treatment technique andtaken into account in the following calculations. All sample surfacesare successively polished with P600, P1200, P2400 and P4000abrasive silicon papers. The sample surface profiles are measuredwith Surfascan Somicronic tactile profilometer with a stylus of2.5 μm radius tip and a step of 4 μm. The RMS roughness Rq is0.0770.003 μm. The average RMS waviness Wq is 0.0570.001 μm.

The glass transition temperature of this cross-linked epoxymaterialis 68.870.2 1C as measured by several tests of Differential ScanningCalorimetry with heating rate of 10 1C/min. Nanoindentation tests arecarried out on the CSM Ultra Nanoindentation Tester with Berkovitchindenter tip under ambient environmental conditions. The penetrationvelocity of loading and unloading for these static indentation tests is10 mN/min. A pause of 20 s is made after total loading for allindentations. These measurements are treated by the Oliver and Pharrmethod [17] using the value of epoxy Poisson's ratio of 0.4. Values ofelastic modulus of 4.570.1 GPa and of hardness of 256.778MPa ofthe epoxy material are obtained. The physical characteristics given bythe supplier as well as surface mechanical properties measured withnanoindentation technique on polished samples are given in Table 1.

2.2. Experimental setup and tribological conditions

This tribological study is carried out on a linear tribometerdeveloped in LTDS and whose principle is detailed in [18]. In thispaper, the contact configuration on the tribometer is schematizedin Fig. 1. A sinusoidal reciprocating motion between two flatsamples fixed in stationary and moving parts of the tribometeris performed. Hereafter, the largest fixed sample will be called‘track’, while the small sample sliding upon the track will bereferred to as ‘slider’. The normal load, tangential force, positionand sliding velocity are continuously measured and recorded with1 kHz sampling frequency as depicted in Fig. 1.

Four tribological conditions, summarized in Table 2, are applied inorder to point out the effect of normal load and sliding speed onenergy dissipation and wear. The sliding distance for all experiments is10 mm. The applied sliding frequencies of 1 and 6 Hz correspond tothe mean sliding velocity of 20 and 120mm/s, or the maximal slidingvelocity of 30 and 170mm/s, respectively. The constant normal force iseither 20 or 50 N, which corresponds to apparent contact pressure of0.7 and 1.8 MPa, respectively. The test duration under each condition isvaried between 10 and 1000 cycles in order to study the wear andfriction evolutionwith the sliding distance. All the tests are carried outunder ambient humidity and room temperature.

Table 1Physical and mechanical properties of the epoxy resin.

Density ρ

(kg/m3)Thermal conductivityk (W/m K)

Specific heat capacitycp (J/kg K)

Thermal diffusivityχ (m2/s)

Glass temperature(1C)

Elastic modulus(GPa)

Hardness(MPa)

1.1�103 0.19 1�103 0.17�10�6 68.870.2 4.570.1 256.778

O. Smerdova et al. / Tribology International 77 (2014) 148–159 149

2.3. Kinematic conditions

The contact kinematics is also schematized in Fig. 1. Allparameters discussed in this paper will be presented as a functionof the kinematic length of one of the two antagonist samples. Thisvalue is defined as the distance seen by a surface point anddepends on the contact time and sliding velocities of the twocontact bodies as following. Kinematic length LkM of a point Mbelonging to the body A in contact with the body B is calculated as:

LkM ¼ tMðVB�VAÞ ð4Þwhere VA and VB are the velocities of A and B contacting bodies.tM¼b/VA is a time of contact between point M and body B, where bis a contact width.

The kinematic length is similar for all points of the slider and equalto Lks ¼ 1 l Ncycle; where l is the sliding amplitude and Ncycle is thenumber of cycles that varies between 10 and 1000 cycles. However,the points of track surface are under different kinematic conditionsand could be divided into three zones. The surface points of the twoextreme zones, where the contact time depends on the point positionon the surface, have less contact time than those of the central zone,where it is constant for all points. Therefore, the kinematic length isnot similar for these three zones as shown in Fig. 1:

� at the beginning of the track (0rxrb), Lkt¼x, where b is thelength of the contact and x is the position of the point of thetrack for which the kinematic length is calculated,

� in the centre of the track (brxr l), Lkt¼b,� at the end of the track (lrxrbþ l), Lkt¼(bþ l) – x.

Further, we will operate with an average kinematic length Lktfor the track calculated by

Lkt ¼Ncycle

ðlþbÞZ b

0x dxþ

Z l

bb dxþ

Z lþb

lðbþ l�xÞdx

" #

¼ 2Ncycle

ðlþbÞ 2Z b

0x dxþ

Z l

bb dx

" #ð5Þ

Thus Lkt , is given by

Lkt ¼ 2ðblÞ=ðbþ lÞNcycle ð6Þ

3. Experimental results: macroscopic wear and frictioncharacterization

In order to study frictional dissipation and wear of epoxy/epoxysliding system, a multi scale approach is developed. Firstly, amacroscopic approach is used. Several parameters, such as massloss and corresponding wear rate, evolution of friction during eachcycle and throughout the whole test, and dissipated frictional energyare investigated within this framework. Secondly, to identify thewear mechanisms and consecutive surface evolution, a detailedwear mechanism expertise based on the microscopic observations ofworn surfaces after gradually increased test durations is carried out.

3.1. Wear law

In order to measure the mass loss, all samples were weighedbefore and after each test. The mass losses of track and slider areplotted as a function of their respective kinematic lengths underfour sliding conditions in Fig. 2 (a and b). In this plot and in otherplots hereafter, each point represents the result of one experiment.The increase of the mass loss ΔM with the kinematic length Lk isroughly linear for all tested conditions. This linear dependency can

Fig. 1. Tribometer principle, kinematics of motion and measured parameters during one cycle.

Table 2Applied tribological conditions.

Conditions ‘LowVFN’ ‘HighV’ ‘HighFN’ ‘HighVFN’

Normal load (FN), N 20 20 50 50Average apparent contact pressure(p), MPa

0.7 0.7 1.8 1.8

Sliding distance (l), mm 10 10 10 10Frequency (f), Hz 1 6 1 6Average sliding velocity (Vm), mm/s 20 120 20 120Max sliding velocity (Vmax), mm/s 30 170 30 170

O. Smerdova et al. / Tribology International 77 (2014) 148–159150

be defined using one of the following ways:

ΔM ¼ awLk ¼WρLk ¼WsρFNLk ð7Þ

where W is the wear rate, i.e. volume loss per distance unity; Ws isthe specific wear rate, i.e. wear rate per normal load unity, and ρ isthe epoxy density.

The slope value aw, i.e. the rate of wear, differs between thesliders and the tracks issued from the same experiment. While theslope of the former varies from 0.2 to 1.5 mg/m for the less and themost severe conditions respectively, the similar conditions cause avariation of aw for the track from 0.3 to 2.2 mg/m. It is noteworthy,that ‘HighVFN’ conditions produce considerably higher mass loss,than the other three conditions.

The observed linear dependence appears to be in agreementwith Archard's empirical wear law Eq. (1) postulated for metallic

surfaces under plastic deformations. In order to compare mea-sured wear loss with the literature, the slope values, aw, arereduced by epoxy density. Average wear coefficients, wear ratesand specific wear rates of slider and track measured under each offour sliding conditions are reported in Table 3. Findings shows thatall wear characteristics tend to increase with the severity oftribological conditions for both samples separately and for theircouple. Nevertheless, it is noteworthy that the wear coefficientdiffers for the slider and the track, even under the same tribolo-gical conditions. This seems to indicate that both parts (slider andtrack) undergo different thermo-mechanical stress and thereforeexhibit a distinct wear response.

An experimental study [19] performed on several frictionalcouples suggests that there is no significant difference betweenwear coefficients and rates of a couple of identical metals and theepoxy/epoxy couple presented here. However, it is important toemphasize, that although the increase of the mass loss with thekinematic length is linear, as predicted by the Archard model, thewear coefficient reported in this study depends on the normal loadand the velocity arguing against the above model. This may be dueto viscoelastic nature of the polymer and contact deformationsthat are not purely plastic as specified by Archard's hypothesis.

In order to verify this, the plasticity index ψ was calculated forunworn epoxy surfaces using the values of hardness, elasticmodulus and root mean square roughness parameter presentedin Section 2.1. This analysis showed that contact deformations areplastic when the asperity radius is less than 21 mm. Mean radius ofsurface asperities of unworn epoxy surface determined frommeasured roughness profiles is equal to 1074 mm. It is expectedthat local asperity radii would grow due to material transfer fromone surface to another. In that case, the deformations wouldbecome elastic, dominating wear mechanism would be fatigueand the wear rate would be correlated with the rate of fatiguecrack growth, instead of material hardness as by Archard's law.

3.2. Friction maps

Friction and normal forces are continuously measured duringthe experiment. Since the normal force remains roughly constant ateither 20 or 50 N, as illustrated in Fig. 1, Coulomb friction coefficientcan be calculated for each point of the track. The relative symmetryof the friction values between back and forth passes, which can alsobe seen from Fig. 1, makes possible to calculate an average value foreach position. An example of Coulomb friction coefficients for arange of tribological conditions during an experiment of 500 cyclesis presented in the form of friction maps in Fig. 3(a–d), wheresliding position and number of cycle are defined on x- and y-axes,respectively, and the friction coefficient is represented by colour.

As shown in Fig. 3(a), friction remains absolutely stable duringthe experiment under low speed and normal force. An increaseof the normal force to 50 N (see Fig. 3(b)) leads to a ratherlow friction coefficient during the first 50 cycles while it subse-quently increases to 0.7 and then remains stable until the end ofexperiment. Unfortunately, an increase of the sliding velocityprovoked the oscillations of the friction, as illustrated in Fig. 3(cand d). This is due to a coupling between stick-slip motions at theinterface and the mechanical response of the tribometer underFig. 2. Mass loss of (a) slider and (b) track and their linear fits.

Table 3Wear volume loss characteristics.

Conditions ‘LowVFN’ ‘HighV’ ‘HighFN’ ‘HighVFN’

Slider/track mass loss aw (mg/m) 0.2/0.3 0.2/0.7 0.3/0.6 1.5/2.2Slider/track Archard's wear coefficient, kw (�10�3) 1.2/4.4 0.9/2.7 2.9/9.5 6.1/8.8Slider/track wear rate, W (�10�10 m3/m) 0.9/3.5 1.8/5.3 2.3/7.4 11.8/17.1Slider/track specific wear rate, Ws (�10–11 m3/N m) 0.5/1.8 0.4/1.1 1.2/3.7 2.4/3.4

O. Smerdova et al. / Tribology International 77 (2014) 148–159 151

high testing frequency. The frequency of these oscillations is about156 Hz, which is close to the first natural frequency of the rig173.3 Hz. The authors believe that these instabilities do notsignificantly affect the wear mechanisms produced by the sliding,because they concern both solids in a similar manner. An averagevalue of friction coefficient calculated from the friction mapsrepresents well these experiments because of its low variationwith the experiment duration. For instance, its value stabilizesnear 0.6–0.7 for ‘HighV’ conditions, as seen from Fig. 3(c). However,under the highest sliding and loading conditions, the mean valueof the friction coefficient falls after 300 cycles from approximately0.7 to 0.6. The drop of the friction at the top of all four maps(corresponding to the maximum amplitude of displacement of theslider over the track) is probably due to parallelism imperfectionsbetween two planes, but can also illustrate formation and evacua-tion of the 3rd body.

The important conclusion from these friction measurements isthat in spite of small variations during each pass, the averagefriction coefficient is roughly independent of the applied normalforce and the velocity in the considered range of experimentalconditions.

An explanation of this observation lies on the hypothesis offriction governed by interfacial shear τ and local contact pressure Pi.e. the polymer hardness H, following [20]:

τ¼ τ0þαP ¼ τ0þαH � αH ð8Þif the polymer hardness is high. The friction value is then constant

and corresponds to the shear of the interfacial layer. Unfortunately,this hypothesis cannot be verified because the interfacial shearstress of the tested material is unknown.

3.3. Dissipated energy

Friction and wear processes generate frictional energy Ed,which dissipates in cracking, deformation, heating or tribochem-ical reactions. It is usually defined as

Ed ¼Z texp

0FðtÞVðtÞdt ð9Þ

where F(t) is the frictional force as a function of time, V(t) is thesliding speed and texp is the total duration of the frictionalexperiment. Taking into account that the friction is rather stableduring the experiment, as seen from the friction maps Fig. 3, thefrictional force can be considered as F(t)av¼μav FN. According, toEq. (9), the dissipated energy is found to be proportional to thekinematic length as follows:

Ed ¼ ∑Ncycle

i ¼ 1

Z 1=f

0FðtÞVðtÞdt ¼ μavFNLks ð10Þ

The experimental results are shown in Fig. 4 as a function of theslider kinematic length. They clearly show the linear dependenceof the dissipated energy on the kinematic length in agreementwith Eq. (10). Pearson's linear correlation coefficients for these

Fig. 3. Friction maps for 500 cycles under (a) ‘LowVFN’; (b) ‘HighFN’, (c) ‘HighV’ and (d) ‘HighVFN’ tribological conditions. (For interpretation of the references to colour in thisfigure caption, the reader is referred to the web version of this paper.).

O. Smerdova et al. / Tribology International 77 (2014) 148–159152

fittings are between 0.96 and 0.995. Their linear fitting is plottedin Fig. 4 by dashed lines with slopes of 11 N and 24 N approxi-mately. These experimental values are close to those given by therelation μav FN, as predicted by Eq. (10).

The energetic approach presented in the introduction isadopted here to characterize the present epoxy/epoxy system.Combining Eqs. (10) and (7) allows us to associate wear withfrictional dissipation using the following relation:

Ed ¼μavFNaw

ΔM ð11Þ

The energy–mass diagram in Fig. 5 confirms this linear depen-dence between frictional dissipated energy and mass loss. Pearson'slinear correlation coefficients for these fittings are 0.55, 0.93, 0.97and 0.93 for four tribological conditions, respectively.

Eq. (11) can be written as

2Ed ¼ V20ΔM ð12Þ

where V0 ¼ ð2μavFN=awÞ1=2:This equation provides an explanation of the apparent evolu-

tion of the wear coefficients with the normal load and slidingvelocity as seen in Table 2. Thus, the dissipated energy can then beinterpreted as a kinetic energy. According to our experimentalresults, V0 varies between 1.6 km/s and 12 km/s. These values,which are in good agreement with the slopes of the curves in Fig. 5can be compared to that of sound propagation velocity in epoxy,Vc, which is in the range of 3 km/s [21]. As Vc corresponds to crackpropagation speed in fragile rupture mode, these high values of V0

suggests that the mass loss is due to different wear mechanismsaccording to tribological conditions, whether V0 is greater than Vc

or not.Apart from providing a relevant wear law, this approach

revealed a linear dependency between two independently mea-sured macroscopic parameters, mass loss and dissipated energy.

4. Experimental results: surface observations

The macroscopic approach links the initial and final polymerstate by a quantitative characterization, without shedding light onevolution of surface state during the sliding process. The processesoccurring within the plane/plane sliding contact are very complexand require very careful observations to be made. In this section,the authors attempt to retrace the history of the sliding contact,based on the micro and macro observations of the surfaces. Forthis purpose, the surfaces of both contact samples after rubbing

10, 100, 500 and 1000 cycles are observed through an opticalmicroscope and analysed. First, the various types of wear mechan-isms observed at microscale are listed in Section 4.1, followed byfurther remarks about the overall shape of worn traces in Section4.2.

4.1. Wear mechanisms

Only few grooves are observed on the surfaces of both epoxysamples after 10 cycles under all applied conditions. Fig. 6(a) illustrates an example of such grooves observed on a tracksurface after a test under ‘HighFN’ conditions. As explained byprevious studies [22], the shape of worn zone borders can beassociated to the ductility or brittleness of the deformed material.Therefore, the notched borders guide the assignation of the grooveto the one or the other type. The transition between ductileand brittle grooves is function of normal and tangential forces,shape and size of the indenter, visco-elasto-plastic polymerproperties and environmental conditions [23]. The differences ingroove profile and top view of ductile and brittle types areschematized in the insets of Fig. 6(a and b). The grooves shownin Fig. 6(a) exhibit side ridges, which indicates the ductility ofthese grooves. The grooves do not contain any trapped wearparticles. Thus, we suppose that they are produced either by hardinclusions or by hard asperities of the counter-body, i.e. by two-body abrasion. The detection of two-body abrasion between twobodies of similar material is surprising at first sight. This can beexplained by a difference of kinematic length between the trackand the slider: the asperities of both solids do not possess similarinstantaneous hardness due to thermo-mechanical softening. Thelarger and deeper grooves, as those produced after 100 cyclesunder ‘HighFN’ conditions and shown in Fig. 6(b), are brittle.

When local asperity radii grow due to plastic deformations, thetwo-body abrasion mechanism is gradually changing to fatiguewear due to elastic deformations. Initially, subsurface cracks areformed due to multiple passage of conterface asperities. Whenthese cracks reach the surface, they induce third-body formationby detaching large bulk areas. These particle then produce three-body abrasion in both contacting bodies.

Fig. 6(c) and (d) presents third body particles, i.e. worn materialdebris, trapped onto the track surface. These particles werecreated and grew during the sliding of the counter surface. Theycan enlarge and deepen the groove, as in Fig. 6(c), or create a newworn zone, as in Fig. 6(d). A smooth bottom and borders areobserved in the groove indicating that the polymer has probablyyielded.

Fig. 4. Dissipated energy evolution with the slider kinematic length. Fig. 5. Dissipated energy as a function of measured total mass loss points.

O. Smerdova et al. / Tribology International 77 (2014) 148–159 153

A qualitatively different set of worn surface phenomenathat are observed on a track surface after the tests under ‘HighV’conditions is shown in Fig. 7(a–d). The location of each observedarea on the track surface is schematized in top left corner ofeach image. Adhesive forces grow under higher normal loadsand produce higher damage to both surfaces. A material pullout, which is believed to be a consequence of high adhesion,is presented in Fig. 7(a). The tears, observed in Fig. 7(b) aretypical for wear of elastomers under repeated sliding friction[24]. The presence of these tears suggests that high dynamicloading has transformed the cross-linked epoxy into the rubberstate.

Moreover, the similar nature of both contact samples and thesevere tribological conditions favour polymer transfer between thesurfaces. This process is illustrated by the adherence of compactedwear debris trapped into the track surface as in Fig. 7(c) or thegeneral raise of deformed worn zone borders, as it is emphasizedin Fig. 7(d).

4.2. Two groups of worn surfaces

After the tests of 100, 500 and 1000 cycles, all worn surfacesare clearly distinguishable and can be classified into two typesof wear. The examples of slider and track couples, worn accord-ingly to these two types are given in Figs. 8 and 9. In bothexamples, the width of the track and slider worn surfaces aresimilar showing that damage occurs simultaneously onboth surfaces. The slider surface is not entirely worn probablydue to initial non-parallelism between the two planes. The lengthof the track worn zone is consistent with the central zone ofmaximal kinematic length, see Fig. 1. The track surface profilealong the blue arrow is shown over the worn zone picture inFigs. 8 and 9.

Type 1: Fig. 8 illustrates a couple of slider and track surfacesafter 500 cycles under ‘LowVFN’ conditions. The track wornzone contains several individual macroscopic grooves forminga bigger one in the middle zone. The small quantity of matterpiles up on several worn zone borders of both slider and track,but the borders are distinct. The worn surface midline is deeperthan the initial surface. The worn zones of both solids arecovered with the third body formed into rolls perpendicular tothe sliding direction. These wear morphological characteristicsare observed at low velocity (here 20 mm/s), independently ofthe normal load and of the number of cycles (above 100 cycles).Type 2: Worn surfaces after the 500 cycles test under ‘HighV’conditions are shown in Fig. 9. The worn zones look funda-mentally different compared to those in Fig. 8. A surface profiletaken along the track worn zone reveals that its surface level israised with respect to the initial surface as illustrated by theworn zone border in Fig. 7(d). A part of matter is collected andtrapped on the left side of the worn zone as in Fig. 7(c), whilethe dark black areas correspond to deep hollows. Small tearsperpendicular to the sliding direction as in Fig. 7(b) areobserved close to the middle of the track worn zone. A closerlook to some track sample surfaces worn by this type revealslocal colour changes to more or less yellowish or brownishcolours. The slider worn surface is similar to that of the trackpresenting the black hollows, perpendicular tears and smoothlower border. Similar to previous type 1, this wear morphologyobtained at high velocity (here 120 mm/s) is independent ofthe normal load and of the number of cycles (above 100).

Due to the experimental difficulties to create absolutely parallelsliding motion between two plane samples under a constantnormal load, the contact area may vary dramatically betweenthe tests. Therefore, it is difficult to identify the conditions oftransition between the several grooves wear morphology and one

Fig. 6. Microscopic images of several locations on an epoxy track surface worn under ‘HighFN’ conditions. (a) Individual ductile grooves formed after 10 cycles (�20);(b) brittle grooves (�2.5), (c) third body particles ploughing track surface (�20); (d) third body particles remained inside a groove (�20). (b–d) images are taken on thesame track sample after 100 cycles.

O. Smerdova et al. / Tribology International 77 (2014) 148–159154

of the discussed wear zone types. Nevertheless, these experimentsreveal that this transition occurs between 10 and 100 cycles inall cases.

5. Discussions

It seems logical to suppose that the range of polymer transfor-mations observed under different tribological conditions is relatedwith the dissipated frictional energy. Therefore, this section startsby linking the amount of dissipated energy with the surface area ofdamaged polymer. The most important mechanism of energydissipation responsible for polymer state change is contact heat-ing. The authors estimate and discuss contact temperatureincrease, which is directly related to applied load and slidingspeed. Furthermore, a possible change in mechanical properties or

chemical structure in the polymer undergoing severe deforma-tions and transformations is analysed and discussed.

5.1. Energy dissipation in increase of worn zone

Firstly, the macroscopic photos of all worn track samples aretreated with ImageJ software in order to measure their worn areas.The damaged surface area Aw is estimated after this imagetreatment, as it is shown in Fig. 10. All visually detected types ofsurface damage and plastic deformation are considered to beinside the worn area.

The results of this analysis are plotted for all experiments underfour tribological conditions in Fig. 11. Two regions are clearlydistinguished on this diagram: the initial fast growth of the wornarea with slow increase of the frictional energy until approxi-mately Aw¼100 mm2, followed by a faster growth of the track

Fig. 7. Microscopic images of several locations on an epoxy track surface worn for 500 cycles under ‘HighV’ conditions. Figures present (a) material pull-out (�10); (b) cracksand craters on the polymer surface (�10); (c) transferred compacted material (�10); (d) material transfer and softening onto the worn zone front border (�2.5). Thelocation of observed area on the whole worn area is shown in the top left corner of each image.

Fig. 8. Epoxy slider (in the left) and track (in the right) couple after 500 cycles under ‘LowVFN’ conditions. Blue arrow traces the line where the surface profile shown nearbyis measured. The visible pattern outside of the worn zone is on the sample back surface. (For interpretation of the references to colour in this figure caption, the reader isreferred to the web version of this paper.)

O. Smerdova et al. / Tribology International 77 (2014) 148–159 155

worn area associated with a severe increase in dissipatedenergy. With little difference in the slope, the initial linearregion is relevant to all tested tribological conditions. The approx-imate rate of dissipated energy in this zone is 0.5 J/mm2. As itwas concluded from the surface observations, this region ischaracterized by the appearance and increase of the numberof individual grooves. This region corresponds essentiallyto abrasion producing the grooves as presented in Fig. 6 (aand b).

From the second region of this diagram and our surfaceobservations, we can conclude that when the worn area becomeslarger, other wear mechanisms interfere. When the maximaldamaged area is reached, the frictional process continues todissipate energy by surface heating and possible two or three-body abrasion, adhesive or fatigue wear and thermal transformation.If the first region is rather independent of the tribological condi-tions, the second region seems to indicate a strong dependence ofdissipated energy rate and associated wear mechanisms onapplied load and sliding speed.

5.2. Energy dissipation in heating

Relative motion between two bodies produces a work againstfrictional forces, which is mostly dissipated in heating of contactzone. Arising from real contact regions, the heat propagates intosubsurface regions and creates temperature gradients. Unfortu-nately, the temperatures within the contact are very difficult tomeasure. However, they are of great importance because they caninitiate chemical reactions or induce polymer state transformationor even degradation [25].

In order to estimate temperature gradients, contact tempera-tures and partition of the generated heat between two bodies, theArchard [26], Blok [27] and Jaeger [28] theory is used. A contact ofa moving and a stationary body is considered. The average value ofthe quantity of generated frictional heat q per surface unity and

time unity is given by

q¼ μavPVAr

ð13Þ

where Ar is the real contact area. The real contact temperature isthe sum of bulk temperature θb, which is ambient in this case, andflash temperature rise Δθ:

θ¼ θbþΔθ ð14Þ

This quantity of heat is distributed between two bodies differ-ently. Indeed, the problem is not symmetrical; the heat source isstationary for the moving slider and moving for the stationary track.Thus, the coefficient of heat partition δ is introduced.

Fig. 9. Epoxy slider (in the left) and track (in the right) couple after 500 cycles under ‘HighV’ conditions. Blue arrow traces the line where the surface profile shown nearby ismeasured. (For interpretation of the references to colour in this figure caption, the reader is referred to the web version of this paper.)

Fig. 10. Illustration of image treatment technique for worn area estimate.

Fig. 11. Energy dissipated in the formation of the track worn area.

O. Smerdova et al. / Tribology International 77 (2014) 148–159156

The heat source is stationary for the slider. Therefore, its tempera-ture gradient is calculated simply by the following equation:

Δθsl ¼δ q bksl

ð15Þ

where b is the characteristic length of the contact and ksl is thethermal conductivity.

For the track, the heat source is moving, and the temperaturegradient depends on the ratio between the contact time for theheat source tc¼b/Vsl and the time of the temperature propagationin the track volume td¼b2/4χtr. In these relations, Vsl is the velocityof the slider, χtr is the thermal diffusivity of the track material. Thisratio is introduced by the Peclet number LP as following:

LP ¼tdtc

¼ Vslb4 χtr

ð16Þ

The value of this non-dimensional speed criterion gives anestimate of the heat propagation profile. The temperature rise forthe stationary track is given by

Δθav ¼ð1�δÞq

ktr

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiχtrb=2Vsl

sð17Þ

Due to the equality of the track and slider surface temperaturein the contact and as the two material are the same, the coefficientof partition is calculated as following:

δ¼ 12þ ffiffiffiffiffiffiffiffi

2Lpp ð18Þ

Considering the average sliding velocity Vm1¼20mm/s andVm2¼120mm/s, half contact width b/2¼3.10�3 m and thermal prop-erties of epoxy given in Table 1, we obtain the Peclet numbers ofLP1¼177 and LP2¼1059. Both these numbers are very high, indicatinga shallow heat propagation profile and high contact temperatures.

Coefficient of heat distribution between two solids is equal toδ1¼0.05 for ‘LowVFN’ and ‘HighV’ conditions and δ2¼0.02 for‘HighFN’ and ‘HighVFN’ conditions. This means that only 5 and 2%of heat is conducted through the slider under sliding velocities V1

and V2, respectively. Therefore, the temperature gradient propa-gated into the track is much higher. On the contrary, the tempera-ture gradient of the slider is very low, and the temperature isambient Vb in the subsurface layers of the slider.

In order to calculate the contact temperature, the knowledge ofreal contact area is mandatory as specified by Eq. (13). Its value isnot only unknown, but varies significantly during the slidingprocess. At the beginning, it can be considered as a multipleasperity contact, whose area quickly grows as wear occurs. Furtherreal area estimate is difficult due to the continuous repetitive thirdbody formation and evacuation from the contact. For the tem-perature estimate, two limit values for the real contact area areconsidered. The upper limit is given by the apparent surface areaof the slider, while for the lower limit, the Bowden and Tabor [29]formula Ar¼FN/H based on the hypothesis of fully plasticizedcontact is used. The real plasticized contact area under givenexperimental conditions is Ar¼8.8 �10�8 m2 for FN¼20 N, andAr¼19.5�10�8 m2 for FN¼50 N.

The summary of relative average temperature rises for all testsunder four tribological conditions as a function of the slider kinematiclength is presented in Fig. 12. For both real area limits, this graphindicates two phases of temperature rise. The first cycles of fasttemperature increase represent a run-in period and correspond to thesliding area increase and a third body formation. This is coherent withthe initial increase of friction coefficient observed on the friction mapsFig. 3. The second region is presented by the stabilized value of thetemperature rise observed under all conditions. It is noteworthy, thatthe temperature rise is generally higher for the faster sliding velocitiesby both estimate processes.

Although the real flash temperature rises are unknown due tothe lack of knowledge of real contact areas, they should besomewhere between the values calculated with two limiting realarea assumptions. The values of temperature rise calculated withAr2 are very high for all conditions. Taking into account that theglass transition temperature of epoxy is below 100 1C, suchtemperature rises along with the dynamical loading certainlychange the polymer state to rubber-like at least locally.

In order to study epoxy thermal degradation, small epoxychunks cut from the edge of a tested sample were heated atvarious temperatures. While a colour change from transparent toyellow was observed in the sample heated up to 250 1C, theheating of the epoxy up to 300 1C transformed it into dark brown.Finally, 350 1C severely degraded the epoxy sample. As it wasmentioned above, slight colour changes were observed locally onthe surfaces worn under high sliding speed conditions. Theseobservations confirm that local heating could probably reach atleast 250 1C, which is considerably higher than epoxy glasstransition, and transform the polymer to rubber-like state.

5.3. Change of chemical or mechanical properties in the worn zone?

The presented analysis suggests local polymer state transfor-mation under high sliding speed. To verify if applied loading haschanged the chemical organization or mechanical properties of the

Fig. 12. Fig. 12. Average surface temperature as a function of slider kinematiclength under four tribological conditions considering (a) contact area Ar and(b) contact area Aapp.

O. Smerdova et al. / Tribology International 77 (2014) 148–159 157

epoxy in the worn zone, several experimental techniques wereapplied by the authors.

Chateauminois and Briscoe [15] used the nanoindentationtechnique to evaluate the mechanical properties of PMMA weardebris formed during fretting wear against steel ball. Theseauthors found almost no difference between hardness and Youngmodulus of the PMMA debris and those of undeformed PMMA. Webelieve that unlike thermoplastic PMMA, thermosetting epoxyprobably modified its mechanical characteristic because of thestate transformations observed above. However, the curved sur-face of the deformed worn zone and highly adhesive debris did notpermit any trustworthy nanoindentation measurements of localsurface mechanical properties.

An X-ray Photoelectron Spectroscopy (XPS) analysis was per-formed at several locations inside and outside the worn zone ofthe track sample presented in Fig. 9. This spectrometry techniquegives information about the chemical environment of atoms at thesurface. In our case, C1s, O1s and N1s were investigated. Thisanalysis performed on worn and virgin epoxy reveals that thereis no significant difference between the two.

An ATR-InfraRed spectroscopic analysis was also carried out on thesample surfaces before and after sliding for various experimentalconditions. The most relevant parts of the spectra of virgin epoxysample and worn zones of track and slider samples are shown inFig. 13. The first observation that can be made is that peak intensity issmaller for worn epoxy than that detected for virgin epoxy. This mightbe interpreted as a loss of motion ability of the chemical bonds due tosliding. Compared to virgin epoxy, the spectrum obtained for the slidersurface under ‘lowVFN‘ conditions after 1000 cycles displays novibration wavelength at 784, 740 and 699 cm�1. The spectrumobtained for the slider surface under ‘HighFN’ condition after 500cycles is similar. In addition, a low intensity peak appears at1647 cm�1. The antagonist track surface only differs from the sliderone by exhibiting a more marked peak at 1647 cm�1. The loss of thevibrationwavelength at 740 cm�1 (symmetric out-of-plane bending ofthe ring hydrogen), at 699 cm�1 (out-of-plane ring bend), and at784 cm�1 (aromatic out-of-plane vibration of C–H bend), might reflectan increase in stiffness of the molecular network. This effect, accom-panied by the appearance of a new peak at 1647 cm�1, probablycorresponding to the stretch mode of a double-C bond, under 50 N,demonstrates the influence of the experimental conditions on thewear modes. Finally, this new peak is more marked for the tracksurface than for the slider surface, confirming the fact that localheating is more intense, and the temperature gradient much higher inthe fixed track. Nevertheless, this analysis would require furtherdedicated investigation.

6. Conclusions

In this work, an analysis of wear mechanisms and associatedenergy dissipation in mass loss, friction and surface damage byviscoelastic, plastic deformations and state transition and chemicaltransformation due to heating, relevant for the sliding contactbetween two solid epoxy plane samples is proposed. To concludethis work, a wear retrospective with two scenarios of surfacedamage evaluation and wear history is summarized in Fig. 14. Bothscenarios start with two-body abrasion and then develop into oneof two branches as a number of sliding cycles increases. Thefollowing deformations are associated with the second region ofdiagram in Fig. 11, where various wear mechanisms interfere.

Fig. 13. IR spectra of virgin and worn for 500 cycles epoxy under ‘HighFN’conditions. The track and slider surface are indicated for the worn epoxy. Thespectra are shifted on the vertical axis.

Fig. 14. Three-body abrasion (1) and adhesion/thermal (2) wear scenarios for epoxy/epoxy sliding contacts.

O. Smerdova et al. / Tribology International 77 (2014) 148–159158

Scenario 1. The Three-body abrasion scenario is characterized bythe presence of micro or macro grooves on the worn surfaces, seeFig. 6(a–d) and the absence of material transfer. The majority ofwear debris is evacuated from the contact, while the remainingpart slides or rolls inside the moving contact and entails thedeepening of wear grooves by the detachment of new wear debrisfrom their bottom and borders. A secondary effect is the polishingof the surfaces outside the grooves by the third body rolls.A typical worn surface of track that underwent the deformationsby this scenario is described as Type 1 and shown in Fig. 8. Thisscenario occurs under low sliding velocity for both normal loads.

Scenario 2. The Adhesion/Thermal scenario is characterized by twoassociated mechanisms: thermal deformation and adhesion. Ther-mal deformation is expressed in an adhesion of the compactedmaterial to the track surface, see Fig. 7 (c), and its transfer from theslider to the track as illustrated in Fig. 9. High adhesion results inthe material pull-out and in small tears perpendicular to thesliding direction, see Fig. 7 (a and b). Under higher sliding velocity,wear debris detached from the slider is not evacuated from thecontact, but softens and adheres to the track surface, elevating thetrack surface midline. In parallel with this transfer process, thecontact temperature increases as illustrated in Fig. 12. Some part ofthe material is pulled out from the track surface, pushed by theslider to the track worn area borders and remains there ascompacted wear debris. A typical example of track surfacedamaged according to this scenario is described as Type 2 andgiven in Fig. 9. This scenario is generally observed for the track andslider samples under high velocity, independently of normalload value.

The authors believe that the key difference distinguishing twoscenarios is the heat-involved deformation in the surface andsubsurface layers of the polymer. This hypothesis is in agreementwith the contact temperature rise and its gradient in the tracksamples. These calculations reveal that higher velocity causeshigher contact temperature, and the largest part of this frictionalheat (95–98%) is dissipated into the track sample. Higher tem-peratures exceed the polymer glass temperature. This, along withthe dynamic loading, causes softening of the thermosetting poly-mer and shifts it into rubber-like state.

Acknowledgements

The authors are very grateful to Dr. Anne Rubin from ICSStrasbourg for the nanoindentation measurements, to Mr. ThierryLe Mogne from LTDS for XPS measurements, to Mr. BernardBeaugiraud from LTDS for IR measurements, to Dr. Alain Le Botfrom LTDS and Prof. Boris Sarbaev from Bauman Moscow StateTechnical University for their participation to the project.

References

[1] Archard JF. Contact and rubbing of flat surfaces. J Appl Phys 1953;24:981–8.[2] Shakhvorostov D, Pöhlmann K, Scherge M. An energetic approach to friction,

wear and temperature. Wear 2004;257:124–30.[3] Ramalho A, Miranda JC. The relationship between wear and dissipated energy

in sliding systems. Wear 2006;260:361–7.[4] Fouvry S, Kapsa P. An energy description of hard coating wear mechanisms.

Surf Coat Technol 2001;138:141–8.[5] Fouvry S, Kapsa P, Zahouani H, Vincent L. Wear analysis in fretting of hard

coatings through a dissipated energy concept. Wear 1997;203-204:393–403.[6] Fouvry S, Liskiewicz T, Kapsa P, Hannel S, Sauger E. An energy description of

wear mechanisms and its applications to oscillating sliding contacts. Wear2003;255:287–98.

[7] Huq MZ, Celis JP. Expressing wear rate in sliding contacts based on dissipatedenergy. Wear 2002;252:375–83.

[8] Aghdam AB, Khonsari MM. On the correlation between wear and entropy indry sliding contact. Wear 2011;270:781–90.

[9] Aghdam AB, Khonsari MM. Prediction of wear in reciprocating dry sliding viadissipated energy and temperature rise. Tribol Lett 2013;50:365–78.

[10] Colaço R, Gispert MP, Serro AP, Saramago B. An energy-based model for thewear of UHMWPE. Tribol Lett 2007;26:119–24.

[11] Geringer J, Forest B, Combrade P. Wear of poly (methyl methacrylate) against ametallic surface in dry conditions. Polym Eng Sci 2007;47:633–48.

[12] Briscoe BJ, Chateauminois A, Lindley TC, Parsonage D. Fretting wear behaviourof polymethylmethacrylate under linear motions and torsional contact con-ditions. Tribol Int 1998;31:701–11.

[13] Krichen A, Bradai C, Chateauminois A, Kharrat M. Surface damage of poly(methylmethacrylate) under fretting loading. Wear 1999;230:146–55.

[14] Teng J, Sato K. in situ observations of fretting wear behavior in PMMA/steelmodel. Mater Des 2004;25:471–8.

[15] Chateauminois A, Briscoe BJ. Nano-rheological properties of polymeric thirdbodies generated within fretting contacts. Surf Coat Technol 2003;163–164:435–43.

[16] Dubourg MC, Chateauminois A, Villechaise B. in situ analysis and modeling ofcrack initiation and propagation within model fretting contacts using polymermaterials. Tribol Int 2003;36:109–19.

[17] Oliver WC, Pharr GM. Improved technique for determining hardness andelastic modulus using load and displacement sensing indentation experi-ments. J Mater Res 1992;7:1564–80.

[18] Guibert, B. Nauleau, P. Kapsa, E. Guibert. Design and manufacturing of areciprocating linear tribometer, Tribol Couplage Multiphys - JIFT 2006, (Lille,France).

[19] Archard JF, Hirst W. The wear of metals under unlubricated conditions. Proc RSoc London Ser A Math Phys Sci 1956;236:397–410.

[20] Briscoe BJ, Tabor D. Friction and wear of polymers: the role of mechanicalproperties. Br Polym J 1978;10:74–8.

[21] Garrett KW, Rosenberg HM. The thermal conductivity of epoxy-resin/powdercomposite materials. J Phys D: Appl Phys 1974;7:1247–58.

[22] Williams JA. Analytical models of scratch hardness. Tribol Int 1996;29:675–94.[23] Bertrand-Lambotte, P. Sur les méchanismes de rayure des vernis de finition

automobile, PhD thesis, Ecole Centrale de Lyon, France, 2001.[24] Briscoe B. Wear of polymers: an essay on fundamental aspects. Tribol Int

1981;14:231–43.[25] Kennedy F.E., Tian X. The effect of interfacial temperature on friction and wear

of thermoplastics in the thermal control regime; 1994. p. 235–44.[26] Archard JF. The temperature of rubbing surfaces. Wear 1959;2:438–55.[27] Blok H. General discussion on lubrication and lubricants. Inst Mech Eng

London 1937;2:222–35.[28] Jaeger JC. Moving sources of heat and the temperature at sliding contacts. Proc

R Soc NSW 1942;76:203–24.[29] Bowden FP, Tabor D. The friction and lubrication of solids. Clarendon Press;

2001.

O. Smerdova et al. / Tribology International 77 (2014) 148–159 159