The primary goal of this study was to grow a tribofilm in situ. To achieve this, the steel colloid probe of apex radius 50 µm was subjected to reciprocating sliding at an applied normal load of 2.9 µN across a scan length of 2 µm. With an adhesion force of 735 nN determined via the pull-off force measurement, a Derjaguin-Müller-Toporov (DMT)[40] estimate of the maximum normal contact pressure and contact radius were 120 MPa and 120 nm, respectively. It should be noted that since this is a multi-asperity contact, as will be shown later, there are asperities which will experience higher normal stresses and the real area of contact will be much smaller than the DMT contact radius would imply.
Figure 1 shows the evolution of the friction force, the total normal stiffness, and probe height, acquired simultaneously, during the reciprocating sliding test. The total normal stiffness is shown, which depends on both the cantilever and contact stiffness. Since the cantilever stiffness is constant throughout the test, any changes are due to variations in the normal contact stiffness. From cycle 1 through approximately cycle 200, the friction doubles and the total stiffness increases by approximately 7%. The probe height increases by 1.2 nm. At least a fraction of this change is due to downward creep of the Z-piezo after the initial transient of applying the normal load, [41] which the feedback circuit compensates for. Two plausible explanations for the friction increase are that asperities are being removed from both counterfaces thus increasing the true contact area, and/or the oxide layer is being removed from the steel, thus exposing the more adhesive and reactive bare steel, which exhibits a higher shear strength. The former would lead to an increase in the real contact area, which increases the normal stiffness and friction. The latter would increase friction due to increased shear strength. In addition, more adhesion and greater contact area would increase formation of adhesive junctions[42] with the exposed bare steel. If an oxide were removed, one would expect a decrease in the topographic height. However, the initial piezo creep effect may obscure such a change.
Starting at cycle 210, the friction begins to drop precipitously before the normal stiffness jumps and the topographic height begins to increase at cycle 225, while the friction continues to decrease. This is clear evidence of the nucleation and growth of a tribofilm. This process occurs during approximately 120 cycles, whereupon the friction has dropped by 81%, the normal stiffness increases by 30%, and the topography increases by 25 ± 2 nm (a close up of this transition region marked in blue is provided in Figure S1.). The topography signal change is due to the probe being lifted by the feedback circuit to maintain a constant normal load, consistent with the growth of a tribofilm on the probe. The increase in the normal stiffness indicates that the real contact area grows substantially if the assumption is made that the tribofilm possesses similar mechanical properties as those grown at macroscale, [18] since it was found in that macroscale study that the tribofilms had much lower average elastic moduli than the initial counterfaces. Thus, a change in only the mechanical properties of the material at the contact should reduce the normal stiffness; the increase seen here can thus only be explained if the contact area increases. That friction is primarily controlled by the tribofilm formation is also supported by the friction image (sometimes referred to as a triboscopy map [43]) in Fig. 2, which is constant across the wear track (except at the left and right edges where the effect of static friction produces an artificial apparent low friction force). If friction were primarily controlled by changes to the a-C:H:Si:O, one would expect a friction dependence on position in the wear track (i.e. one would see vertical features in the image shown in Fig. 2), unless such changes occur homogenously across the wear track. It should be noted, however, that 24.5% of the total friction decrease occurs prior to any detectable change in the probe height, suggesting that chemical and structural changes at the counterface surfaces drive a significant fraction of the friction reduction.
For the remainder of the test, the changes are less dramatic. The topography continues to grow before saturating at a value of 28.4 ± 2.3 nm, suggesting the tribofilm grows to this thickness level and then roughly maintains a steady-state thickness. Notably, the friction is approximately constant during this growth, indicating that the friction reduction is primarily controlled by the chemistry and structure of the tribofilm near the a-C:H:Si:O/tribofilm interface, and not by the overall thickness of the tribofilm. Post-sliding imaging did not reveal any transfer of tribofilm back to the a-C:H:Si:O. There is a slow increase in the friction and normal stiffness, 27% and 8% respectively, throughout the remainder of the test. These results can be explained by some additional growth in the real contact area as the tribofilm grows in lateral extent due to plastic deformation and/or material transport/shearing of the tribofilm.
Figure 3 shows pre- and post-test reverse imaging of the colloidal probe using a TGT1 sample. The post-test measurement was done in tapping mode rather than contact mode to minimize wear of the tribofilm. It is clear from the linescan shown in Fig. 3(c) that the thickness of the tribofilm (33.6 ± 8.8 nm) agrees reasonably well with the 28.4 ± 2.3 nm which was measured during the sliding test, with the difference likely reflecting compression of the tribofilm due to the normal load during the wear test. In fact, a very rough estimate of the Young’s modulus can be made by treating the colloid as a cylindrical punch of radius equal to the DMT contact radius of the original a-C:H:Si:O-steel contact, and also making the more reasonable assumption of incompressible DLC and steel, to yield a value of 517 MPa. This value represents an upper bound, given that we can see in Fig. 3b that the tribofilm diameter far exceeds the DMT prediction, thus the contact area used in the modulus calculation is an underestimate. Note that the tribofilms grown in macroscale sliding experiments also showed much smaller elastic moduli relative to the starting a-C:H:Si:O which had an elastic modulus of 120 GPa.[18]
The tribofilm is elongated in the sliding direction. Since torsion of the cantilever during the two sliding directions of the probe changes the real area of contact for each direction, it might explain elongation of the tribofilm in the sliding direction. However, the magnitude of such an expected change can be estimated from the static friction at the track endpoints in, e.g., Fig. 2, to be approximately 200 nm total. This is not large enough to explain the magnitude of the elongation. The fact that the tribofilm appears to have formed at an offset from the colloid apex (with the center of the tribofilm approximately 1.5 µm to the right of the center of the probe) may be due to an experimental artifact. The topography was measured by scanning the colloid over a spiked TGT1 sample. A difference in the slope of TGT1 sample relative to the a-C:H:Si:O sample under the probe of as little as 2° could produce the offset seen here. It is also possible poor adhesion between the tribofilm and steel allowed it to slide. The presence of wear debris above and below the apex in Fig. 3b outside the horizontal band containing the tribofilm suggests the tribofilm may have moved from where it was originally grown. Prior literature has shown that tribofilm adhesion to steel counterfaces can be poor, depending on tribofilm growth conditions. [44–46]
The wide lateral extent of the tribofilm in all directions relative to the nominal contact area requires explanation. In the macroscale work,[18] where a similar effect was seen for sliding in a flowing N2 environment, we attributed this effect to a combination of plastic deformation and viscoelastic flow due to the soft polymeric properties of the tribofilm. Flow out of the contact area may help explain the self-termination of the vertical growth of the tribofilm, with extra tribofilm volume beyond the thickness where the tribofilm is stable being shed out of the contact area and increasing the lateral extent of the tribofilm.
An additional possibility, given that the thickness of these tribofilms are a significant fraction of the nominal contact radius, is that the nominal contact area increases because of a decrease in the reduced Young’s Modulus E* of the contact due to the formation of the soft tribofilm itself. In the macroscale work, it was found that all regions of tribofilms grown in a humid environment had elastic moduli < 40 GPa, at least a 60% reduction relative to the bearing steel on which they were grown, with many areas exhibiting orders of magnitude smaller elastic moduli. Since the nominal contact area in DMT contact mechanics scales with E*−2/3, a 50% reduction in E* for example leads to a 58% increase in the nominal contact area. Thus, a large reduction of E* due to tribofilm growth can produce a substantial increase in the nominal area of contact.
To examine the tribological response of the steel colloid sliding on a-C:H:Si:O, with and without the tribofilm, a series of friction vs. load measurements were performed. The measurements were obtained at a sliding speed of 3 µm/s across a 5x5 µm2 region with a constant increase in the normal load after each of 128 scan lines. The large transverse spatial increment between scan lines (5 µm / 128 lines = 39 nm) was chosen to minimize wear of the a-C:H:Si:O, and therefore also the dependence of friction on wear-induced changes to the a-C:H:Si:O. It also served to minimize any additional growth of the tribofilm. Each measurement was immediately repeated with a decreasing load to verify repeatability and to capture the load range between pull-in and pull-off. In the bare colloid tests, humid and dry environments were examined. For the colloid-with-tribofilm tests, a smaller load range was used to minimize damage to the tribofilm. Tests were performed in multiple environments (see SI Fig S2), but it was apparent that the tribofilm was evolving throughout these tests, so only the first is presented here, where the tribofilm geometry and structure had not yet changed significantly. This test was performed in the same humid environment in which it was grown (50% RH). All tests were performed on pristine (previously unscanned) regions of the a-C:H:Si:O sample. Note that the maximum load reached, 1.35 µN, is significantly less than the 2.9 µN load used previously to generate a tribofilm on a bare colloid probe.
Results are presented in Fig. 4. Focusing on the bare colloid results, it is clear that there is a strong humidity dependence to the friction. The nearly linear dependence of friction on load is consistent with a multi-asperity contact geometry.[47] The magnitude and slope of the linear dependence are higher in the humid case. Such a humidity dependence was absent in the case of the tribofilm + colloid trials. The humidity dependence in the bare colloid case might be due in part to capillary adhesion. While steel can have large water contact angles suggesting a high surface energy, [48] this is influenced by the presence of surface oxides and organic contamination, [49] which can be removed from the contact zone by sliding. The underlying metal will have a high surface energy [50] and tend to form substantial capillary bridges in a humid environment.
The friction vs. load measurement with a tribofilm present on the colloid shows several distinct, important features. First, the adhesion (the pull-off force reached during retraction) is reduced by at least 80%: 29 nN is obtained, vs. 77 5nN and 849 nN for the humid and dry bare colloid cases, respectively. This is remarkable considering the increase in real contact area upon tribofilm growth, to be discussed below. The reduction in adhesion results in reduced friction across the entire load range, relative to the bare colloid results, shifting the friction vs. load curve to the right. The idea that there has been a change from a patchy, multi-asperity contact to a more continuous area of contact is supported by a power law fit to the contact area vs. load (Fig. 4), where the scaling exponent was 0.52. If we assume a constant shear strength,[51] the expected scaling exponent for a perfect, single-asperity DMT contact is 2/3, as opposed to the case of a multi-asperity contact which obeys Amonton’s law, where the scaling exponent would be 1 (as was seen for the bare colloid). This very low scaling exponent indicates the contact is behaving as a single asperity contact. The fact that the exponent is lower than the lower bound of the range between single and multi-asperity contacts is not fully understood, but may have to do with the fact that the tribofilm is not spherical in shape, but rather more like a flat plateau. In such cases, the contact area would grow more slowly with normal load than would be the case for a spherical contact.[52] This can be understood by considering the extreme case of a rigid, flat punch. Friction vs. load curves in such a case (again assuming constant shear strength) would result in a scaling exponent of 0 since the real contact area does not grow with load.