Tribology Letters

, Volume 23, Issue 1, pp 65–75

Wear behaviour of carbon nanotube reinforced epoxy resin composites


    • Fachhochschule Lübeck
  • W. Xu
    • East China University of Science and Technology
  • B. Schädel
    • Fachhochschule Lübeck
  • W. Wu
    • East China University of Science and Technology

DOI: 10.1007/s11249-006-9042-7

Cite this article as:
Jacobs, O., Xu, W., Schädel, B. et al. Tribol Lett (2006) 23: 65. doi:10.1007/s11249-006-9042-7


This paper deals with the effect of a multi-walled carbon nanotube (MWCNT) reinforcement on the wear behaviour of Epoxy (EP) composites. Firstly, various dispersion methods were compared. Secondly, the optimum CNT amount was evaluated. In a third step, dry lubricants were added to the optimised EP/CNT composite. Finally, the influence of the steel counterpart (martensitic bearing steel 100Cr6 and austenitic stainless steel X5CrNi18-10) was studied. The preparation method was found to have a decisive effect on the wear behaviour of the composite. A pre-treatment of the CNTs in concentrated nitric acid proved beneficial. Even more important was the mixing method. A dual asymmetric centrifuge delivered so good wear results that the pre-treatment could be skipped. The optimum CNT content was at about 1 wt%, regardless of the preparation method. The lowest wear rates were found after addition of 10 wt% graphite. MoS2 proved to be less effective and Polytetrafluorethylene (PTFE) even increased the wear. The wear rates against the unalloyed martensitic steel were far higher than against austenitic stainless steel.


carbon nanotubes (CNT)epoxysliding weardispersion methods


Polymers are increasingly used for components that are subjected to tribological loadings like bearing bushes, gear wheels or low-friction coatings. Some of their typical benefits are: easy processing, corrosion resistance, low friction and damping of noise and vibrations. However, the load bearing capacity and thermal stability are inferior to metals and ceramics. Suitable fillers are added to extend the application fields of polymers in tribologically loaded systems:
  • The load bearing capacity and thermal resistance is improved by adding hard particles or fibres (e.g., bronze, glass, carbon, ceramics, etc.). These hard phases additionally increase the heat conductivity so that the temperature in the sliding contact is lowered.

  • Dry lubricants like graphite, MoS2, or polytetrafluorethylene (PTFE) are added in order to reduce the friction coefficient. This also lowers the shear stresses in the mating components and, thus, the wear rate.

Nano-fillers are a rather new group of reinforcements for polymer compounds. They provide several advantages over classical micro-reinforcements. They allow the production of micro-mechanical components and thin coatings, and they usually do not cause embrittlement and deterioration of tensile strength as microscopic fillers often do [1,2]. Several researchers have published studies on friction and wear of nano-particle reinforced thermoplastics [35] and thermosets [68]. All of them found a wear reducing effect of the nano-fillers. Own studies [6] revealed that nano-particle reinforced epoxy (EP) composites mostly exhibit a wear minimum at around 1–2 wt% of nano-fillers for most nano-particle types.

Carbon nanotubes (CNT) are a very new class of nano-materials. They resemble bucky balls that were cut along the equator and elongated by inserting cylindrical graphite tubes [9]. Single-walled CNTs (SWCNTs) have walls consisting of only a single graphite layer, multi-walled CNTs (MWCNTs) have walls consisting of piles of many graphite layers. The tubes are almost free of defects and are based on strong covalent bonds in axial and circumferential direction. CNTs have a diameter of a few dozens of nm and a length of some μm. Due to this high-aspect ratio, CNTs provide a good mechanical reinforcement potential, which has been experimentally confirmed by several researchers [912].

However, only few studies on the effect of CNTs on the tribological behaviour of polymers have been published so far. Chen et al. found that MWCNTs can significantly decrease the wear rate and fiction coefficient of copper and PTFE-based composites [13,14]. Lim et al. [15] investigated carbon/carbon composites coated with CNT reinforced carbon and, too, found that the wear rate decreased continuously with increasing CNT content. In the case of PTFE-based composites [13], the friction coefficient decreased continuously with increasing CNT content up to 30 vol% CNTs, whereas the wear rate passed a minimum around 15–20 vol% (corresponding to a far lower weight percentage). The scanning electron microscopy (SEM) images of typical worn surfaces indicated that the CNTs strengthen the PTFE and effectively reduce adhesive and ploughing wear of the composites. Zoo et al. studied the effect of MWCNT addition on the tribological behaviour of ultra high-molecular weight Polyethylene (UHMWPE) [16]. They added up to 0.5 wt% CNTs. The friction coefficient increased continuously with the CNT content, which is in contrast to the previously cited findings for PTFE. Yet, the wear rate again decreased by about one order of magnitude.

The present paper deals with the effect of CNT reinforcement on the sliding wear behaviour of EP composites. The first step addressed the dispersion problem of CNTs in EP. CNTs are known for their tendency to form rather stable agglomerates [2]. Various dispersion and mixing methods were employed and compared with respect to their effect on processing and wear properties. In a second step, the optimum CNT amount was experimentally evaluated. Finally, some dry lubricants (PTFE, graphite and MoS2, respectively) were added to the optimum EP/CNT composite in order to identify potentials for further improvements.

Experimental details

Base materials

An EP resin was chosen as matrix material for a number of reasons: CNTs are expensive and were only available in small quantities. Thermoplastic compounding methods would have required larger amounts of sample materials. Moreover, EPs can be compounded in the cold state. EP is therefore a good model substance to develop dispersion methods. Last but not least, EPs are often used in practical applications like bearing bushes, repair coatings of worn surfaces, thin low-friction coatings and deep drawing tools.

The resin used in this study was an EP resin of the type Neukadur EP 571 [17] based on bisphenol A, hardened with an aminic hardener Neukadur T9. Resin and hardener were produced by Altropol (Stockelsdorf, Germany).

The CNT used in this study were MWCNTs, produced by Namigang Co. (Shenzhen, China). The parameters of these CNTs were: average diameter 10–30 nm, average length 5–15 μm. The purity of CNTs was better than 95 wt%, the amorphous carbon content was below 3 wt%. The specific surface area was between 40 and 300 m2/g [18].

The dry lubricants were: PTFE powder with an average grain size of 6 μm, graphite powder with an average grain size of 4 μm, and MoS2 with an average grain size of 3–4 μm.

Specimen preparation

A variety of preparation methods were employed. Generally, the subsequent production sequence was followed:

1. Pre-treatment (optional). CNTs have a non-polar surface, which barely interacts with the EP matrix. The pure CNTs in the as-delivered-state exemplify the tendency of MWCNTs to form a kind of loosely packed felt (figure 1). The pores inside these non-polar agglomerates are so small (in the range of a few 100 nm) that the polar EP resin cannot easily infiltrate them. Some CNT samples were therefore pre-treated to improve the bonding between CNTs and matrix. For this purpose, the CNTs were mixed into HNO3 (4.0 mol/l) with a CNT to HNO3 weight ratio of 1:3. The mixture was continually boiled in a reflux condenser at 100 °C for about 11 h while being stirred at a rotational frequency of 300 rpm. After that, the mixture was washed in distilled water until the pH value approached 7 in order to eliminate the HNO3.
Figure 1.

SEM micrograph of CNTs in the as-delivered state.

2. Drying of CNTs. The pre-treated and untreated CNTs, respectively, were dried in an oven at 100 °C for 2 h before further processing. The dried CNTs were put into a plastic bag and stored in an excicator.

3. Mixing of CNTs into the resin. The CNTs were manually crushed with a mortar for about 10 min before being mixed into the resin.

Two different mixing methods were employed. One mixing device was a four-blade stirrer (see figure 2(a)) with about 20 mm diameter and working at a rotational frequency of 2500 rpm. The mixture was prepared in a cup of about 25 mm in diameter. This mixing procedure was found to introduce large amounts of air into the resin.
Figure 2.

Four-blade stirrer (a) and SpeedMixer (b).

The other mixing device was a “SpeedMixer DAC 150 FVZ” (figure 2(b)), a dual asymmetric centrifuge. The samples were filled into a small pot (size: 60 ml) of PP. This pot is inserted into the SpeedMixer with a 30° tilt angle. During the mixing process, the pot rotates around its z-axis while the base plate rotates around the vertical axis in opposite direction. The mixing process is based on a kind of high-power shaking. Due to the centrifugal forces, only little air is entrapped in the mixture. This mixing process lasted 2 min at 2000 rpm.

4. Evacuating. After stirring, the mixture was evacuated and aerated by turns until no more bubbles were generated.

5. Ultrasound dispersion. To achieve a better de-agglomeration, some samples were sonicated after mixing with the SpeedMixer. The parameters were:
  • Apparatus: Dr. Hielscher GmbH UP 400S (400 W max. Power),

  • cylindrical titanium sonotrode with 7 mm diameter and

full amplitude.

The process had to be interrupted every 30 s to avoid thermal softening of the PP pot. The total time of this ultrasound process was 10 min, i.e. 20 cycles, each lasting 30 s.

6. Addition of dry lubricants (optional). Since the samples with 1 wt% sonicated CNTs proved best, this mixture was further modified by adding several dry lubricants. After the mixture of 1 wt% CNTs in EP resin had been sonicated and homogenised in the SpeedMixer, 10 wt% of a dry lubricant (PTFE, graphite or MoS2, respectively) were mixed into the resin using the SpeedMixer at 2000 rpm for 4 min. As a reference, mixtures with dry lubricants but without CNTs were produced.

7. After thorough homogenisation, the final mixture was again cyclically evacuated and re-pressurised until no more bubbles formed on the surface.

8. Hardener was stirred into the mixture. The hardener was directly added to the mixture at a resin-to-hardener weight ratio of 4:1. In some cases, the hardener was mixed into the EP/CNT composites by the SpeedMixer at 500 rpm for 30 s. However, these gentle mixing conditions did not lead to homogeneous mixtures. Very high-mixing speeds, on the other hand, introduced too much air into the final mixture. As a compromise, the hardener was mixed into the composites at 1000 rpm for 1 min into the composites. Nevertheless, one test series was conducted with the sub-optimal mixing speed of 500 rpm.

9. Then the ready mixture including the hardener was again evacuated for about 60 min.

10. When no more bubbles appeared, the mixtures were cast into dies to produce 2 mm thick plates.

11. The dies were put into an oven to cure the sample material. The cure cycle was: (1) heating from 20 to 60 °C in 2 h, (2) heating from 60 to 80 °C in 2 h, (3) finally keeping at 80 °C for 4 h.

12. The wear test specimens were cut from these plates with a handsaw into pieces of 10 × 10 × 2 mm3.

Table 1 lists all samples that were prepared and tested.
Table 1.

List of all sample materials prepared and tested.


Preparation method

Pure EP

Hardener mixed via SpeedMixer at 500 rpm for 30 s

Pure EP

Hardener mixed via SpeedMixer at 1000 rpm for 1 min

EP + 0.2 wt% untreated CNTs

Mixed with four-blade stirrer

EP + 0.5 wt% untreated CNTs

Mixed with four-blade stirrer

EP + 1 wt% untreated CNTs

Mixed with four-blade stirrer

EP + 2 wt% untreated CNTs

Mixed with four-blade stirrer

EP + 4 wt% untreated CNTs

Mixed with four-blade stirrer

EP + 0.2 wt% pre-treated CNTs

Mixed with four-blade stirrer

EP + 0.5 wt% pre-treated CNTs

Mixed with four-blade stirrer

EP + 1 wt% pre-treated CNTs

Mixed with four-blade stirrer

EP + 2 wt% pre-treated CNTs

Mixed with four-blade stirrer

EP + 4 wt% pre-treated CNTs

Mixed with four-blade stirrer

EP + 0.5 wt% untreated CNTs

Mixed with SpeedMixer, hardener at 500 rpm

EP + 1 wt% untreated CNTs

Mixed with SpeedMixer, hardener at 500 rpm

EP + 2 wt% untreated CNTs

Mixed with SpeedMixer, hardener at 500 rpm

EP + 0.5 wt% pre-treated CNTs

Mixed with SpeedMixer, hardener at 500 rpm

EP + 1 wt% pre-treated CNTs

Mixed with SpeedMixer, hardener at 500 rpm

EP + 2 wt% pre-treated CNTs

Mixed with SpeedMixer, hardener at 500 rpm

EP + 1 wt% pre-treated CNTs

Mixed with SpeedMixer, hardener at 1000 rpm

EP + 1 wt% pre-treated CNTs

Sonicated, mixed with SpeedMixer, hardener at 1000 rpm

EP + 1 wt% untreated CNTs, + 10 wt% PTFE

Sonicated, mixed with SpeedMixer, hardener at 1000 rpm

EP + 1 wt% untreated CNTs, + 10 wt% graphite

Sonicated, mixed with SpeedMixer, hardener at 1000 rpm

EP + 1 wt% untreated CNTs, + 10 wt% MoS2

Sonicated, mixed with SpeedMixer, hardener at 1000 rpm

EP + 10 wt% graphite

Mixed with SpeedMixer, hardener at 1000 rpm

EP + 10 wt% MoS2

Mixed with SpeedMixer, hardener at 1000 rpm

Testing procedures

Viscosity measurements

Carbon nanotubes have a strong tendency to form networks inside the mixture [2,9]. These networks increase the resistance of the mixture against shearing, i.e. the viscosity. Hence, the viscosity of the mixtures may give information about the quality of the dispersion level. Various mixtures were therefore examined viscosimetricly. The cone-plate measuring system used, which is sketched in figure 3, works according to ISO 3219. A sample of the mixtures was placed between the base plate and the cone. The cone used here had a radius of 12.5 mm, an opening angle of 1°, and a truncation of 50 μm, which is also equivalent to the distance of the flattened cone and the plate. The cone rotated with stepwise increasing speed while the resulting torque was measured. The shear rate ranged from 10 to 4500 1/s. The measuring time was 10 min for each sample. The measurement delivered plots of the viscosity versus shear rate.
Figure 3.

Sketch of the cone-plate rheometer for viscosity measurements.

The following mixtures were examined in that way:
  • Pure EP.

  • EP with 2 wt% untreated CNTs mixed with the four-blade stirrer.

  • EP with 2 wt% pre-treated CNTs mixed with the four-blade stirrer.

  • EP with 2 wt% pre-treated CNTs mixed with the SpeedMixer.

  • EP with 2 wt% untreated CNTs mixed with the SpeedMixer.

Wear tests

All tests were performed in a ball-on-prism tribometer according to ISO 7148-2/5.2 as shown in figure 4. The specimens were subjected to uniform and unidirectional sliding.
Figure 4.

Schematic of the ball-on-prism tribometer. Side view (left) and front view (right).

The polymeric specimens were glued onto the inner surfaces of the prism with 90° opening angle (figure 5(a)). The prism is fixed at one end of a lever and pressed against a ball with a diameter of d = 12.7 mm via a dead weight. The balls, which rotate uniformly around their vertical axis driven by a simple electro motor, can consist of different materials.
Figure 5.

(a) Prism with specimens and ball and (b) loading situation.

All tests were performed with a dead weight of FZ = 30 N, resulting in a normal load of \( F_{\text{N}} = \frac{{F_{\text{Z}} }} {{\sqrt 2 }} = 21.2\,{\text{N}} \) acting on each specimen (see figure 5(b)). The rotational frequency, f, of 1 Hz resulted in a sliding speed of \( v = \pi \cdot \frac{d} {{\sqrt 2 }} \cdot f = 28.2\,{{{\text{mm}}} \mathord{\left/ {\vphantom {{{\text{mm}}} {\text{s}}}} \right. \kern-\nulldelimiterspace} {\text{s}}} \). Each test lasted ca 60 h, which is equivalent to about 6000 m sliding distance. The contact pressure, p, and the p·v-value cannot be quoted because they decreased continuously during the test due to the spherical contact geometry. Yet a rough estimate can be made: the diameter of the wear mark was around 1–3 mm during steady state. The according average contact pressure is 3–27 MPa and the p·v-level below 0.76 MPa m/s.

One set of specimens was loaded statically in the test rig for about 60 h. Under this static load, the balls penetrated into the specimen surface only a few tenths of a micrometer due to plastic deformation and creep of the polymer. This deformation is negligible in comparison to the wear of the materials, which was in the order of some 10 μm to some 100 μm. The creep was therefore not further considered and skipped in the subsequent tests.

An inductive displacement transducer continuously measured the motion of the lever. The wear of the metallic counterparts was insignificant in all cases so that the measured system wear represented the wear of the polymer-based specimens. The displacement of the lever, h, was converted into a wear volume, V, per specimen (see figure 6): \( V = \frac{\pi } {6} \cdot \left( {\frac{h} {{\sqrt 2 }}} \right)^2 \cdot \left( {3d - 2\frac{h} {{\sqrt 2 }}} \right) \). The wear volume was plotted versus the sliding distance, L, which was derived from the rotational frequency, f, the ball diameter, d and the test duration, t, according to: \( L = \frac{1} {{\sqrt 2 }} \cdot d \cdot f \cdot t \). After a running-in phase, all wear curves were found to become linear. A function of the type \( V = V_0 + wL \) was fitted to the linear part of the curves. The slope \( w = \frac{{dV}} {{dL}} \) was divided by the normal load acting onto a single specimen, FN = 21.2 N, to calculate the specific wear rate, ks:
Figure 6.

Calculation of wear volume (ΔV).

$$ k_{\text{s}} = \frac{w} {{21.2\,{\text{N}}}} = \frac{1} {{F_{\text{N}} }}\frac{{dV}} {{dL}}\left[ {\frac{{{\text{mm}}^{\text{3}} }} {{{\text{Nm}}}}} \right] $$

All experiments were conducted in normal laboratory environment (about 50% r.H. and 23 °C).

Each test was repeated at least three times. The data represented in this paper are the arithmetic mean values of the three tests. The standard deviation of the data is represented by error bars in the diagrams.

Unfortunately, it is not possible to measure friction forces with the tribodata test rig used.

Results and discussion


Figure 7 compares the viscosity versus shear rate curves of pure EP with the flow curves of EP/CNT composites mixed with the four-blade mixer. All mixtures exhibited a Newtoneon flow behaviour at low-shear rates, i.e. the viscosity is more or less independent of the shear rate. Above a shear rate of about 1000 s−1, the alignment of the EP molecules caused a transition to structural viscosity, so the viscosity decreased. Regardless of the shear rate, pure EP had the lowest viscosity (5–6 Pa s). As expected, the addition of 2 wt% untreated CNTs almost doubles the viscosity to about 9 Pa s. Due to this effect, it was not possible to add more than 4 wt% untreated CNTs since the mixture became so viscid that it could not be stirred any more. The pre-treatment of the CNTs in nitric acid reduced the viscosity of the composite again, but it remained above the value of pure EP. The effect of the pre-treatment can be traced back to an improved dispersion: the untreated CNTs have a non-polar surface and thus a weak interaction with the polar EP matrix. The pre-treatment introduces polar functional groups onto the CNT surfaces so that the EP matrix better wets the CNTs, penetrates into the agglomerates and individualises the CNTs more or less.
Figure 7.

Flow curves of pure EP and mixtures of EP and 2 wt% treated and untreated CNTs, respectively. Mixtures were prepared with the four-blade stirrer.

Figure 8 displays the effect of the mixing method. The SpeedMixer and the four-blade stirrer produced mixtures with virtually identical flow curves as long as pre-treated CNTs were used. However, mixing untreated CNTs with the SpeedMixer into the resin roughly decupled the viscosity of the EP and significantly exceeded even the viscosity of the mixture of untreated CNTs in EP prepared with the four-blade stirrer. Moreover, the flow curve is non-Newtoneon right from the beginning; instead, the mixture shows some shear thinning.
Figure 8.

Flow curves of pure EP and mixtures of EP and 2 wt% treated and untreated CNTs, respectively. Effect of mixing method.

This can be explained by different microstructures produced by the different preparation methods: The virgin CNTs form tuft like agglomerates due to the non-polar extremely large specific surface. When mixed into an EP resin, these large agglomerates considerably increase the viscosity of the compound. After pre-treatment in nitric acid, the CNT surfaces are partly polarised so that the resin can easier penetrate into the agglomerates and individualise the CNTs. This results in a reduced viscosity compared to the mixture containing untreated CNTs. The SpeedMixer generates very high-shear forces and impacts inside the mixture so that the primary agglomerates are shattered and the CNTs are more or less individualised, even without pre-treatment. Nevertheless, the matrix cannot properly wet the individualised CNTs because of their non-polar surfaces. Therefore, the CNTs re-agglomerate, but with a modified structure. The observed shear thinning behaviour suggests that a sort of extended 3-D percolation network is formed instead of more or less globular agglomerates as they form in virgin CNT powders. This explanation is also in accordance with the subsequently described wear test results. Gojny et al. [19], too, observed the tendency of CNTs to form percolation networks under certain conditions.

Wear tests

The first approach was to produce the samples in the simplest possible way: untreated CNTs were mixed with the four-blade stirrer into the resin. In the range between 0 and 4 wt% CNTs, no significant effect of the reinforcement on the wear rate could be found (see figure 9) during sliding against X5CrNi18-10. All wear rates were in the range of 10-5 mm3/Nm. Higher CNT contents could not be prepared because the viscosity of the mixtures became too high.
Figure 9.

Effect of CNT content on the specific wear rate of EP/CNT composites during sliding against X5CrNi18-10. CNTs used were untreated and mixed with the four-blade stirrer.

A poor dispersion obviously prevented the CNTs from developing a significant reinforcing effect. The large scatter of the wear data, too, referred to an insufficient homogeneity of the mixture. Moreover, the mixtures needed more than 2 h of evacuation to remove the large amount of air that was entrapped inside the agglomerates.

A second attempt with pre-treated CNTs led to somewhat better results (figure 10). The scatter of the data was significantly smaller, which means that the mixtures were more homogeneous. The wear rate decreased continuously as the CNT content was increased to 1 wt%. However, the wear rate again increased slowly beyond 1 wt% CNT content. This wear minimum in the range of a few weight per cent filler content is typical for nano-composites [6,8]. Nano-fillers have an extremely large specific surface where the matrix material is adsorbed. Therefore, only small amounts of nano-fillers can be mixed in a polymer matrix when every particle should be properly wetted. For example, only 8 wt% SWCNTs can be solved in an EP matrix assuming that each CNT is surrounded by a monomolecular EP layer [20]. That is, 1 wt% is equivalent to about 12.5% of this theoretically maximum possible filler content. Exceeding this limit usually causes embrittlement of the composites and promotes susceptibility to surface fatigue.
Figure 10.

Effect of CNT content on the specific wear rate of EP/CNT composites during sliding against X5CrNi18-10. CNTs used were pre-treated in nitric acid and mixed with the four-blade stirrer.

Figure 11 presents the wear rates of composites where both the CNTs (2 min at 2000 rpm) and later on the hardener (30 s at 500 rpm) were mixed into the resin with the SpeedMixer. Strikingly, the pure EP prepared in this way had a far higher specific wear rate (about 30 × 10−6 mm3/Nm) than the EP resin mixed with the four-blade stirrer (9 × 10−6 mm3/Nm). It was suspected that the mild conditions during mixing of the hardener into the resin could not ensure a homogeneous hardener distribution. As a consequence, the cross-linking density may have been too low locally.
Figure 11.

Effect of the CNT content (pre-treated and untreated) on the specific wear rate of EP/CNT composites during sliding against X5CrNi18-10.

The addition of CNTs, though, had a significant effect. There was a pronounced minimum around 1 wt% CNT content. Beyond this value, the wear rate again increased slightly, in case of the untreated CNTs less pronounced than in the case of pre-treated CNTs. Around the minimum, there was no significant difference between untreated an pre-treated CNTs. Untreated CNTs would be favourable from the practical point of view because the complicated pre-treatment with the environmentally unfriendly HNO3 can be omitted and the composite is less sensitive to minor variations in the dosing process. Yet the pre-treatment is still necessary when low-viscosity mixtures are needed, e.g. for castability reasons. The minimum wear rate of the speed mixed composites prepared with untreated CNTs is slightly smaller than the minimum wear rate of the compounds prepared with the four-blade stirrer and pre-treated CNTs. This shows that the SpeedMixer effectively crushes the agglomerates and disperses the CNTs even without a pre-treatment. This good wear performance of the speed mixed composites of untreated CNTs in combination with the extremely high viscosity of this mixture described above suggests the idea of a percolation network of – more or less – individualised CNTs.

Another important feature of the optimised composites with 1 wt% CNTs is the low scatter of the wear data. This is clearly visible in figure 10 as well as in figure 11. At lower contents, microscopic areas may suffer from CNT depletion and missing reinforcement; at higher contents, more and larger agglomerates may form leading microscopically inhomogeneous properties.

Since the dispersion of the CNTs plays such an important role, one sample of 1 wt% untreated CNTs in EP was additionally sonicated before being homogenised in the SpeedMixer. The wear rate was further reduced, but only by about 20% from 3.22 to 2.65 × 10−6 mm3/Nm.

Scanning electron microscopy inspection of the wear marks gave only little information about the wear mechanisms. CNTs are chemically almost identical with EP and could therefore not be distinguished from the matrix. Figure 12 shows some micrographs of the wear mark of EP including 1 wt% untreated CNTs and mixed with the four-blade stirrer. The wear mark is rather smooth. Some loose wear debris but only little and very small scratches can be found. Adhesive wear seems to have been predominant. However, the smooth area is riddled with a lot of transverse cracks (figure 12(b)) originating from surface fatigue, which is typical for brittle materials like EP.
Figure 12.

Wear mark of 1 wt% untreated CNTs in EP mixed with the four-blade stirrer. (a) Overview, (b) and (c) details. Arrows indicate the sliding direction.

The wear mark of the composite consisting of 1 wt% pre-treated CNTs in EP mixed with the four-blade stirrer looks different (figure 13): first of all, the wear mark has of course a far smaller diameter, which visually documents the significantly lower wear rate. However, only about 50% of the wear mark appears smooth, whereas the remaining part looks rougher. Figure 13(b) shows the borderland between two such areas. The smooth part seems to be back-transfer, which protects the original surface from being polished. This back-transfer was missing in case of the untreated CNTs. The back-transferred material contained numerous small transverse cracks and a few larger longitudinal cracks indicating fatigue damage of this layer followed by delamination wear. Once the back-transfer film locally had delaminated the original surface was again exposed to the sliding. There is less wear debris visible on the wear mark than in case of the composite containing untreated CNTs (figure 13(a)). Moreover, the wear debris has a more needle like conformation (figure 13(d)) compared to the irregularly shaped debris particle in figure 12(b). These needles of some 100 nm diameter must contain a high amount of CNTs because they remained very stable under the electron beam at high magnification and current. These needles may have had a kind of needle bearing effect.
Figure 13.

Wear mark of 1 wt% pre-treated CNTs mixed with four-blade stirrer into the EP matrix. (a) Overview, (b) borderland between rough and smooth surface part, (c) magnification of smooth part and (d) magnification of rough part. Arrows indicate the sliding direction.

The specimens prepared with the dual asymmetric centrifuge featured similar wear marks like the specimen shown in figure 13. However, the share of the area covered by back-transfer varied slightly and was biggest for the compound containing pre-treated CNTs and prepared with the SpeedMixer (figure 14). This was exactly the mixture with the lowest wear rate.
Figure 14.

Overview of the wear mark of 1 wt% pre-treated CNTs in EP, mixed with the SpeedMixer. The arrow indicates the sliding direction.

The best EP/CNT composite so far was the mixture including 1 wt% untreated CNTs, sonicated and homogenised by the dual asymmetric centrifuge. This compound was used as a basis for further improvements by adding 10 wt% PTFE, graphite or MoS2, respectively. Figure 15 compares the wear rates of these materials. The addition of 10 wt% PTFE to the EP/CNT compound increased its wear rate slightly. This was unexpected because PTFE normally has an outstandingly beneficial effect on the wear resistance of conventional [21,22] and nano-composites [6]. The reason for this unusual negative synergy between CNTs and PTFE could not be clarified so far. In contrast, MoS2 and especially graphite significantly improved the wear resistance of the EP/CNT composites. EP/CNT/graphite reached a specific wear rate as low as 3.5 × 10−8 mm3/Nm (please note the exponent -8), which was the lowest wear rate among all combinations tested.
Figure 15.

Specific wear rates of multi-component composites. EP/CNT = EP + 1 wt% untreated CNTs (sonicated and speed mixed), counterpart = X5CrNi18-10. Lubricant content was 10 wt% in all cases.

To check whether the good wear performance of the EP/CNT/graphite and the EP/CNT/MoS2 composites can be traced back to the action CNTs or to the dry lubricants, samples of pure EP (without CNTs) were filled with graphite and MoS2, respectively. Figure 15 shows that these samples, too, perform better than the pure EP but worse than the multi-component composites. Conclusively, neither the CNT reinforcement nor the dry lubrication alone can yield minimal wear rates. There is obviously a positive synergetic interaction between CNTs on the one hand and graphite or MoS2, respectively, on the other hand. This mutual backing is particularly pronounced for the graphite, which chemically resembles the CNTs.

Previous studies [21,22] had revealed that composites containing chemically active fillers like carbon fibres, graphite or MoS2, react sensitively on the chemical composition of the counterpart material. These fillers can cause tribo-corrosion of steel counterparts with low-chromium content. After this tribo-corrosion commences, the composite is rapidly abraded by the corrosion products and the roughened counterpart. This is also true for the CNT composites as figure 16 points out. CNTs as well as MoS2 and graphite perform worse against the unalloyed martensitic steel 100Cr6 than against the corrosion resistant austenitic steel X5CrNi18−10. The effect of the counterpart is again most pronounced for the graphite-containing sample.
Figure 16.

Effect of counterpart material on selected nano-composites. “EP/CNT” means EP with 1 wt% untreated CNTs produced with the SpeedMixer. MoS2- and graphite content: 10 wt%.


The addition of CNTs to an EP matrix can significantly enhance its wear resistance. However, the performance of these composites sensitively depends on the pre-treatment of the CNTs and the mixing procedure. Stirring untreated CNTs into the resin did not cause any improvement. Various methods were found to disperse CNTs effectively in the EP matrix:

  • A functionalisation of the CNT surface, which improves the wettability, enables easier penetration of the matrix into the agglomerates with subsequent individualisation of the CNTs. Boiling in concentrated nitric acid proved to be a suitable pre-treatment. The pre-treated CNTs could subsequently be mixed with a common four-blade stirrer.

  • Crushing of agglomerates by high-energy mixing. A dual asymmetric centrifuge turned out to be very useful for this purpose. A pre-treatment was not necessary in this case to get good results.

Both methods resulted in almost identical wear rates. However, viscosity measurements showed that both mixing methods produced completely different microstructures: The use of pre-treated CNTs lead to a low-viscosity mixture containing more or less individualised CNTs, whereas speed mixing of untreated CNTs made for to a more thixotropic material based on a kind of percolation network of CNTs. Combining the pre-treatment in nitric acid with mixing in the dual asymmetric centrifuge generated similar processing and wear properties as the combination of pre-treatment and four-blade stirrer.

Conclusively, the use of a dual asymmetric centrifuge renders the use of the environmentally unfriendly nitric acid unnecessary but leads to a high-viscosity mixture suitable for press moulding but less applicable for cast moulding. A pre-treatment is inevitable when a low viscosity is needed.

Sonication of the mixture further disintegrated the agglomerates and improved the dispersion of the CNTs. This effect was however of minor magnitude.

An optimal wear resistance was found for about 1 wt% CNT content, regardless of the dispersion and mixing method. The wear rate increase beyond this filler content was least pronounced for the composite produced of untreated CNTs by the dual asymmetric centrifuge.

Moreover, attention has to be paid to the mixing of hardener into the compound. The SpeedMixer did not lead to a sufficiently homogeneous distribution of the hardener inside the rather viscous mixture. The four-wheel stirrer, on the other hand, introduced too much air into the compound. A mixing device working like the four-blade stirrer should be used in a vacuum chamber in future studies.

The addition of dry lubricants had various effects on the wear of the CNT/EP composites. The addition of 10 wt% graphite decreased the specific wear rate significantly. It reduced the wear rate of the EP/CNT composite by a factor of nearly 100 compared to the composite without graphite and more than 300 times compared with the pure EP. MoS2 also decrease the wear rate, but to a lesser degree than the graphite. The EP/CNT filled with 10 wt% PTFE had even a slightly higher wear rate than the PTFE-free EP/CNT. CNTs on the one hand and graphite or MoS2, respectively, on the other hand mutually aid one another with respect to the wear resistance of the composite. This synergetic effect is particularly pronounced for graphite.

The selection of an adequate counterpart material is even more important than the composition of the compound itself. Unalloyed martensitic 100Cr6 steel caused far higher wear rates of the composite than austenitic stainless steel X5CrNi18-10. The sample containing 1 wt% CNTs and 10 wt% graphite delivered the by far best wear results (ks = 3.5 × 10−8 mm3/Nm) against the stainless steel but was extremely sensitive against the counterpart type, the wear rate against 100Cr6 was by about a factor of 1000 higher.


The authors would like to thank some colleagues from the Fachhochschule Lübeck for their support and helpful discussions: Prof. B. Voß conducted the SEM investigations, Dipl.-Ing. E.-O. Reimann introduced us to the viscosity measurements and provided the according test rig, Ms A.-Ch. Heidenreich and Dipl.-Ing. A. Frederich made the equipment available needed for the pre-treatment in nitric acid. Altropol, Stockelsdorf, Germany supplied the epoxy resin free of charge.

Copyright information

© Springer Science+Business Media, Inc. 2006