1 Introduction

Carbon fibers have the highest specific modulus and highest specific strength of all reinforcing fibers and are widely used as reinforcements in composites [1]. Carbon fiber-reinforced polymeric composites are used in a wide range of industrial applications [2]. In particular, a great efficiency of carbon fibers in improving the tribological performance of polymers has been found [3]. The development of carbon fiber-reinforced tribological materials has been result of a series of engineering studies.

Flöck et al. [4] investigated the influence of the type of carbon fiber [polyacrylonitrile (PAN) vs. pitch-based carbon fiber] on the friction and wear performance of PEEK composites and concluded that preference should be given to the pitch-CF/PEEK composite at an equivalent tribological performance because of its much lower price. Besides the carbon fiber type, the orientation of carbon fibers, no matter continuous carbon fibers (CCF) or short carbon fibers (SCF), has been found to have strong influence on the friction and wear properties of polymeric composites. For example, higher wear resistance of a SCF/PTFE/graphite (10 vol% for each)-filled PEEK composite was registered under higher pressures (3–5 MPa) when the fibers were oriented perpendicular rather than parallel or normal [5]. Thinning, cracking, and debonding of fibers have been suggested as relevant wear mechanisms in polymeric composites [6–8]. Success has also been achieved in the further improvement of tribological performance of carbon fiber-reinforced polymer composites by surface modification of reinforcing fibers [9, 10].

Understanding the separate functions of different fillers can reveal the fundamental mechanisms in the tribological response of polymeric composites. Furthermore, it can pave the route for the design of composites with desired properties. Han et al. [11] studied the micrometer-scale friction of SCF-reinforced PEEK/PTFE composite blends and found that the exposed top surfaces of normally oriented carbon fibers had a slightly higher coefficient of friction than PEEK, while carbon fibers oriented in plane had a clearly lower friction coefficient. Nanometer-scale scratching experiments on SCF-reinforced PEEK/PTFE composite blends illustrated the higher scratch resistance of SCF as compared to the matrix [12]. Typically, the friction coefficients measured in these micrometer-scale experiments are much lower than those measured for the different components in macroscopic tests. The difference can be ascribed to the dominance of adhesive friction in the wear-less regime which is realized in the micrometer-scale experiments. Only recently, the friction behavior of different phases in carbon fiber-reinforced PEEK composites was investigated using the scratching method, and friction coefficients comparable to macroscopic testing were registered [13].

The macroscopic friction and wear properties of carbon fiber-reinforced PEEK composites have been widely studied. Attention has also been paid to the friction characteristics of different phases in PEEK composites microscopically. However, in order to connect micro- and nanoscale results to the macroscopic tribological performance of PEEK composites, the following studies are urgently needed: (1) the friction and damage mechanisms of single carbon fibers; (2) the influence of the contact configuration between asperity and composite components on the tribological mechanisms.

The objective of this study is to study friction and wear mechanisms of carbon fibers in a PEEK composite at the micrometer scale. The results are expected to help reveal fundamental mechanisms of friction and wear reduction of PEEK by carbon fiber reinforcements. To achieve this goal, two different indenters were used in the present study to simulate different possible asperity geometries in the surface of a macroscale counterbody. The conical indenter has a diameter that is larger than the width of exposed carbon fiber sections and was therefore only used to study the wear and damage mechanisms of carbon fibers. The Berkovich indenter, giving higher resolution, was used to investigate friction and wear within carbon fibers. Consistent results are achieved regarding the wear mechanism of carbon fibers using both indenters.

2 Experimental

2.1 Materials

The sample is an injection-molded plate consisting of a poly-ether-ether-ketone (PEEK) matrix filled with 10 wt% short PAN-based carbon fibers and with 10 wt% graphite flakes. The detailed manufacturing procedure is described in Ref. [14]. Before testing, the sample surface was finely polished and cleaned by a three-step protocol. First, the sample surface was ground with water sandpaper of no. 800, 1200, 2400, and 4000. Second, the ground sample surface was polished on a polishing disk with 1 and 0.25 µm diamond pastes. Finally, the polished surface was cleaned in acetone in an ultrasonic bath. On the polished composite surface investigated in this study, only sections of carbon fibers are exposed, which have a typical width of 5 µm and a length of ~50 µm.

2.2 Graphite Structure of Carbon Fibers in the Composite

The graphite structure of carbon fibers in the composite was investigated by Raman spectroscopy (Renishaw InVia Raman Microscope, excitation by an Nd:YAG laser at 532 nm, spectral resolution ≈1.2 cm−1). The spot size in the focal plane was circa 2 μm. With the help of an attached optical microscope, the measuring spot was positioned on a carbon fiber surface, and spectra were recorded as a function of depth in steps of 0.5 µm.

2.3 Scratch Tests

In this study, we will refer to the orientation of carbon fibers with respect to the scratch direction, as illustrated schematically in Fig. 1a. Scratch tests were performed using a Berkovich and a conical indenter. The Berkovich indenter with an apex radius of curvature of 150 nm is used in a Hysitron TriboIndenter TI-950 system. The same indenter is also used to image the surface morphology of carbon fibers after scratching. The friction coefficient during scratching is determined as the ratio of the lateral force and normal force. The normal force ranged from 1 to 10 mN, and the indenter moved at a constant velocity of 0.4 µm/s to a distance of 10 µm. The conical indenter has an apex radius of 7.7 µm with an included angle of 90°. This indenter is used to study the damage behavior of carbon fibers induced by long-term asperity action, implemented in an ASMEC scratch tester at an indenter velocity of 5 µm/s with fixed scratch length of 30 µm. The conical indenter was also used to test the damage behavior of carbon fibers upon indentation into the matrix, which was implemented by indenting ~3 µm away from the center of a fiber end or ~3 µm away from the edge of a perpendicularly oriented fiber (Fig. 1b). Depending on the regime to be studied, the normal load applied on the conical indenter varied from 20 to 40 mN.

Fig. 1
figure 1

Schematic illustration of a different fiber orientations with respect to the scratching direction and b indentation points beside one side or end of a carbon fiber

Full size image

2.4 Friction Imaging by AFM

Friction maps of the exposed surfaces of single carbon fiber in a PEEK matrix were recorded with a Nanowizard 3 setup (JPK Instruments, Berlin, Germany). The normal spring constant of the silicon cantilever was calibrated to be 0.2 N/m using the thermal noise method. Friction experiments were performed under a normal load of 15 nN.

2.5 Morphology Variation of Carbon Fibers

To reveal the damaging mechanisms in carbon fibers after scratch experiments, the composite surface was sputter-coated with a gold layer and then imaged by a scanning electron microscope (Quanta 400 FEG ESEM, FEI).

3 Results

3.1 Graphite Structure of Carbon Fibers in the Matrix

In order to identify damage mechanisms, it is necessary to polish the specimen’s surface before scratching measurement. Exposed carbon fibers were studied by Raman spectroscopy. The spectra recorded on a fiber oriented parallel to the surface in different depths are shown in Fig. 2. The characteristic D band and G band of graphitic materials are clearly observed as long as the focal plane is still situated within the carbon fiber. There is no change in the position and shape of these bands throughout the fiber, which means that these fibers do not exhibit any skin–core structure. This homogeneous structure was already verified for several kinds of carbon fibers [15].

Fig. 2
figure 2

Raman spectra of one carbon fiber in the PEEK composite for increasing depth. The D and G band characteristics for graphitic material are clearly identified as long as the focal plane of the exciting light lies within the carbon fiber

Full size image

3.2 Friction Measured in Scratch Tests on Carbon Fibers

Figure 3 shows the friction results for scratching experiments on carbon fibers and on the surrounding polymer matrix for different fiber orientations. When scratching on parallel carbon fibers, the indenter slides on top of the fiber surface throughout the graph (blue line in Fig. 3). The friction coefficient of carbon fibers is found to be 0.12 (the slight variations are due to the roughness in the scratch route). This value is similar to those found for graphite in macroscopic sliding tests [16, 17]. The other three curves are composed of two levels: a low level associated with carbon fibers and a high level associated with scratching the PEEK matrix. For the latter, a friction coefficient of 0.4 is found, in agreement with a previous study [18]. No significant differences in friction are found for different fiber orientations.

Fig. 3
figure 3

Friction coefficient as a function of tip displacement recorded while scratching with the Berkovich indenter at a normal load of 1 mN across carbon fibers of different orientation and across the PEEK matrix

Full size image

In practical applications, asperities on the counterpart surface are in contact with both the matrix and reinforcing fibers. To resemble typical configurations, scratching was performed at carefully selected spots on the composite surface. The friction curves are shown in Fig. 4, in which the corresponding contact geometry is schematically illustrated adjacent to the individual curves. In these experiments, the indenter slid against both the carbon fibers and the matrix along the whole scratch track. The friction coefficient decreases in the following order: scratching on PEEK, scratching PEEK at the side of a carbon fiber, scratching PEEK between two parallel carbon fibers, and sliding on top of a carbon fiber. When the indenter slid against PEEK at the side of a carbon fiber, the friction coefficient increased at the beginning of scratching (black and blue line). The friction then levels off in the remaining part of scratch route. In the case of sliding on the surface of a fiber or between two fibers, the friction coefficient was kept in the same level in the whole scratching process (pink and red lines). No visible scratch tracks are produced in the carbon fiber surfaces, while the scratching leaves grooves in the PEEK surface. These grooves are significantly shallower in the neighborhood of carbon fibers.

Fig. 4
figure 4

Friction coefficients as function of the tip displacement measured under different contact configurations (schematically illustrated adjacent to the individual curves, which are black, blue, pink and red from top to bottom in the graph) (Color figure online)

Full size image

To study the local friction on carbon fibers, measurements were conducted on a carbon fiber surface using both AFM and the nanoscratch tester. Figure 5a shows the friction maps of a carbon fiber in the PEEK matrix. There is no friction contrast between the central and edge parts of the fiber. The PEEK matrix exhibits higher friction compared with that of carbon fiber. This friction contrast was also reported by Mazeran and Loubet [19]. In the scratch tests, only a minor difference is noticed between the initial friction coefficients recorded at different positions (schematically shown in Fig. 5a). All friction coefficients decrease with increasing number of scratching cycles (Fig. 5b).

Fig. 5
figure 5

Local friction behavior of carbon fibers. a Friction map of a carbon fiber embedded in a PEEK matrix and b variation in the scratch friction coefficient of a carbon fiber with increasing number of scratching cycles against a Berkovich indenter under 8 mN normal load measured in three different positions. The lines with symbol in a indicate the routes for scratch tests in b, and the scratching direction is from bottom to top

Full size image

3.3 Wear Behavior of Carbon Fibers Studied by Scratching

Besides the low-friction characteristic of carbon fibers, the wear performance of carbon fibers is also of concern. Figure 6 shows the morphology of as-worn carbon fiber surface after scratching with a Berkovich indenter for different number of cycles. Within 15 cycles and at the given load, no fracture occurs. Careful analysis revealed that the width of exposed carbon fiber section decreases with increasing number of scratching cycles. While after one scratching cycle no obvious narrowing in sectional plane of the carbon fiber is detected (Fig. 6a), a neck is formed with a decreased width of 4.3 µm (Fig. 6b) after five cycles. The width of the neck is further decreased to 4 µm (Fig. 6c) and 3 µm (Fig. 6d) after ten and 15 cycles of scratching. This leads to the conclusion that thinning of carbon fibers takes place under repeated interaction with asperities. Macroscopically, the thinning of carbon fibers was also reported as a result of sliding of carbon fiber-reinforced PEEK composites against steel counterbody [5].

Fig. 6
figure 6

Topography of the surface scratched across a CF under 10 mN normal load to different cycles. a 1 cycle, b 5 cycles, c 10 cycles, and d 15 cycles (in all the four pictures, the carbon fiber is the same one, which lies in the center horizontally (the triangular impressions are due to indentation of the Berkovich indenter, which are surrounded by matrix pileups)

Full size image

Even in polymer composites reinforced by short carbon fibers, some fibers will lie parallel to the relative sliding direction in tribological contacts. This situation is simulated by scratching a conical indenter against the fiber surface along its length. The morphologies of scratched surfaces of parallel oriented carbon fibers are presented in Fig. 7. There is no noticeable track left on the carbon fiber surface after 50 cycles of scratching (Fig. 7a). For comparison, the morphology of a scratched tilted carbon fiber is shown in Fig. 7b, where the very edge of the fiber started to peel off and the interface between the fiber and PEEK matrix failed. If the scratching is repeated for 500 cycles, a scratch track evolves through depression of the fiber surface, on which micro-cracks are also visible (Fig. 7c).

Fig. 7
figure 7

SEM of a CF surface scratched with the conical indenter under 20 mN normal load to different cycles. a 50 cycles parallel, b 50 cycles tilted, and c 500 cycles parallel. The scratching direction is always from the left to the right; the scratch length is 30 µm

Full size image

To reveal the load dependency of scratching in carbon fibers, the normal load was increased; the resulting morphology is shown in Fig. 8. At 30 mN normal load, no significant damage was left on the fiber surface, but the interface between the fiber and matrix started to fail (Fig. 8a). However, a normal load of 40 mN induced dramatic damage to the fiber, including micro-kinking and micro-fracture of the fiber as well as the interface failure (Fig. 8b). The damage of carbon fibers was further studied by indentation close to the side face of carbon fibers. When the side face of the indenter pushes against one side of a carbon fiber, crescent debris is produced accompanied by cracked small pieces (Fig. 9a). Furthermore, debonding appears at the interface between fiber and matrix. The debonding leads even to full delamination when one end of a carbon fiber is pushed repeatedly by the indenter side face (Fig. 9b). No difference is found in the morphology of debris between indentation at the side and at the end of a fiber.

Fig. 8
figure 8

Morphology of carbon fibers after single-pass scratching with the conical indenter under a 30 mN and b 40 mN normal load. The scratching direction is from left to right

Full size image
Fig. 9
figure 9

Morphology of carbon fibers after 500 cycles of indentation against the side face of conical indenter under 40 mN normal load. a One edge of a carbon fiber and b one end of a carbon fiber

Full size image

4 Discussion

Our experimental results confirm that carbon fibers exhibit lower friction compared with the PEEK matrix. According to the widely accepted two-term model of friction, the friction coefficient can be decomposed into a shear component and a plowing component [20]. In the following, this model will be applied in a quantitative discussion of friction of carbon fibers.

When the PEEK matrix is scratched, the indenter will plow into the PEEK surface. The corresponding plowing friction component was discussed before by Stuart and Briscoe [21]. At the same time, the shearing at the interface between indenter and PEEK also contributes to the total friction. On the carbon fiber surface, the indenter slides without noticeable penetration, in which case the shearing dominates the friction process. As a result, carbon fibers show a much lower friction coefficient compared with the PEEK matrix. When the indenter is slid along one side of a carbon fiber (Fig. 4), one part of the indenter is in low-friction contact with the carbon fiber, reducing the friction due to shearing and plowing against PEEK. Friction is further reduced for the case of sliding between two fibers (Fig. 4), with only minor contributions from the PEEK underneath the indenter. An intermediate friction coefficient is registered in these cases of sliding on PEEK and at the same time along a carbon fiber. The friction coefficients lie between those of the PEEK matrix and of the carbon fibers. The friction coefficient decreases with increasing contact with carbon fibers.

Another observation is that the friction coefficient increases at the initiation of sliding in all cases where PEEK is part of the frictional pair (black and blue lines in Fig. 4). The increase in friction coefficient in the initial phase of sliding is associated with the increasing contact between the front part of the indenter and PEEK, while steady-state plowing is established. In contrast, carbon fibers bear the load effectively in cases of sliding on or between two carbon fibers (pink and red lines in Fig. 4). As a result, the friction coefficients are constant from the beginning because of constant contact size in the whole scratching process. The carbon fiber also bore part of the normal load when the scratch led along one side of a carbon fiber, which can be verified by the decrease in penetration depth compared with pure PEEK (Fig. 10). We consider these load-bearing mechanisms of carbon fibers to be important for corresponding observations in the macro-sliding process of polymer composites [22, 23].

Fig. 10
figure 10

Profile of the scratch track. a On a pure PEEK surface, b on a PEEK/CF pair as depicted in Fig. 4

Full size image

In order to correlate the observed mechanisms with the role of carbon fibers in the friction behavior of polymer composites, we list published friction coefficients of carbon fiber-reinforced PEEK composites in Table 1. In all studies, friction coefficients of the PEEK matrix are decreased after the incorporation of carbon fibers. When comparing the friction coefficients from macroscopic experiments on composites with our results from scratching friction, they would rank between the values measured when scratching the PEEK matrix (µ = ~0.4) and the carbon fibers (µ = ~0.12).

Table 1 Friction coefficient of PEEK composites reinforced with carbon fibers from the literature
Full size table

Macroscopic sliding tests have shown that plowing as the dominating friction and wear process of PEEK is largely inhibited after the incorporation of carbon fibers [22]. This change in the tribological mechanism is in perfect agreement with our scratching results and also consistent with the reduction in the friction coefficient. Overall, the high friction contribution of PEEK and the low friction of carbon fibers lead to the intermediate value for PEEK composites [24]. Our experimental results decompose the friction contribution from the matrix and carbon fibers and elucidate the mechanisms by which carbon fibers improve the tribological behavior of PEEK composites. We suggest that the typical friction coefficient value of 0.25 measured on PEEK composites is due to a plowing which is reduced by the load-bearing capacity of carbon fibers.

The chemical homogeneity across the carbon fibers (see Fig. 2) explains the uniform friction on the carbon fiber surface in AFM measurements (Fig. 5a). In contrast, the interaction between PEEK matrix and SiOx coated tip is much higher due to the oxygen-containing molecules, which results in higher friction on PEEK than on the carbon fiber surface. The carbon fibers are composed of stretched microfibrils with diameters of ~200 nm [15]. Apart from fine scratches from polishing, there is no friction contrast that can be related to the boundaries between microfibrils, indicating a high degree of compaction. In the scratching experiments, the tip radius of curvature (~150 nm) is comparable to the radius of the microfibrils (~100 nm). The initial indenter position may be on fibrils of different size in different relative positions, which could explain the small differences in the initial friction coefficient (Fig. 5b, first cycle). Upon repeated scratching, the groove adapts to the Berkovich indenter and, as a result, the friction coefficient decreases and coefficients from different trials merge with each other. This idea is in agreement with the view that shearing is the dominating mechanism in the friction of carbon fibers.

Carbon fibers can fail in different modes depending on the loading conditions [25]. Under tribological loading, cracking, thinning, and debonding of carbon fibers are believed to be the relevant mechanisms [7]. The contribution of different failure mechanisms varies depending on the orientation of carbon fibers as well as the specific loading parameters. Our results reveal that perpendicular carbon fibers are worn in radial direction by the moving asperity. As a result, the fibers become locally thinner and thinner (Fig. 6). When parallel carbon fibers are scratched by the indenter, fatigue occurs with the formation of micro-cracks (Fig. 7c). Beyond thinning and fatigue, the load-bearing function of carbon fibers imparts improved wear resistance to the composite. These micro-mechanisms shed further light on the macro-tribological properties of carbon fiber-reinforced polymer composites, in which a fatigue mechanism of the composite was also reported [8].

Apart from thinning, carbon fiber fracture is a widely accepted wear mechanism of fiber-reinforced polymer composites, depending on the fiber orientation and testing conditions as well as the matrix properties [7]. Among the various parameters, the normal load is the key factor governing the fracture of fibers [26]. Whatever the relative orientation, the normal load-induced bending momentum plays an important role in the failure behavior of a carbon fiber in polymer matrix [26, 27]. This action impairs the interfacial bonding between carbon fibers and the matrix and leads to interfacial failure. When the asperity touches the side surface of a carbon fiber, the matrix outside the stressed zone prevents the interfacial delamination of fibers from the matrix, and only debonding happens (Fig. 9a). In contrast, indenting the end of a fiber results in delamination at the side walls (Fig. 9b). This result is consistent with the finding that longer fibers are more effective than shorter ones in improving the wear resistance of polymer composites [28], because shorter ones can be easily dug out by the asperities. In addition to interfacial failure, the repeated action of asperity induces the formation of debris (Fig. 9), which can be ascribed to the local shear stress.

Carbon fiber reinforcement is the method of choice in many industrial applications, in which the reliability of polymer components depends on friction and wear performance. Our results establish scratching experiments as quantitative method to reveal mechanisms of tribological reinforcement by carbon fibers. By predicting friction coefficients from simple scratch experiments, efficient ways of a systematic exploration of reinforcements become available with a need for expensive macroscopic tests. In particular, different reinforcement configurations such as in gradient materials can be evaluated in a single sample. Furthermore, in the macroscopic sliding process, carbon fibers are fractured and the matrix is stretched, both of which will slide against the transfer films formed on the counterbody surface. We assume that the surface properties of fractured carbon fibers and stretched matrix are modified by the above process. Building on the work presented here, those friction and wear mechanisms of modified carbon fibers and polymer matrix in the worn composites’ surface can be investigated, which are dominant in the long-term sliding of carbon fiber-reinforced polymer composites.

5 Conclusions

  1. 1.

    Scratching experiments reveal plowing and shearing mechanisms in sliding friction on carbon fiber-reinforced polymers, which contribute to their macroscopic tribological characteristics.

  2. 2.

    For well-defined configurations of fibers and sliding asperity, a quantitative agreement of friction coefficients between macroscopic tests and scratching experiments has been achieved and allows for conclusions regarding the dominant friction mechanisms.

  3. 3.

    Carbon fibers bear the load preferentially in tribological loadings, thereby reducing the shearing contribution to friction of the polymer matrix in PEEK composites.

  4. 4.

    The high wear resistance of carbon fibers is limited only by their thinning in repeated perpendicular interaction with asperities and fatigue in repeated parallel sliding.

  5. 5.

    Friction on carbon fiber surfaces is homogeneous; it neither depends on the orientation of the fiber nor does it change across the fiber’s cross section.