Running-in process of Si-SiOx/SiO2 pair at nanoscale—Sharp drops in friction and wear rate during initial cycles

Using an atomic force microscope, the running-in process of a single crystalline silicon wafer coated with native oxide layer (Si-SiOx) against a SiO2 microsphere was investigated under various normal loads and displacement amplitudes in ambient air. As the number of sliding cycles increased, both the friction force Ft of the Si-SiOx/SiO2 pair and the wear rate of the silicon surface showed sharp drops during the initial 50 cycles and then leveled off in the remaining cycles. The sharp drop in Ft appeared to be induced mainly by the reduction of adhesion-related interfacial force between the Si-SiOx/SiO2 pair. During the running-in process, the contact area of the Si-SiOx/SiO2 pair might become hydrophobic due to removal of the hydrophilic oxide layer on the silicon surface and the surface change of the SiO2 tip, which caused the reduction of friction force and the wear rate of the Si-SiOx/SiO2 pair. A phenomenological model is proposed to explain the running-in process of the Si-SiOx/SiO2 pair in ambient air. The results may help us understand the mechanism of the running-in process of the Si-SiOx/SiO2 pair at nanoscale and reduce wear failure in dynamic microelectromechanical systems (MEMS).


Introduction
With the development of lithographic microfabrication and micromachining techniques, silicon has become a principal construction material in microelectromechanical systems (MEMS) [1,2]. When the dimensions shrink to nanoscale, the ratio of surface area to volume greatly increases so that the interfacial forces become dominant [3]. As a result, the nanotribological problems involving friction, adhesion, and wear, have become an important concern in Si-MEMS [2][3][4].
As a well-known tribological process, running-in is usually defined as the initial operation of a friction pair until certain friction force and wear rate have reached a steady state [5]. In macroscale devices, surface roughness, apparent defects, and surface waviness induced by manufacturing can affect running-in behavior [6][7][8]. Many new machine parts, such as cylinders and gears, often need to be conditioned through running-in before they are placed into regular service. During running-in, because the peaks of asperities on rough contact surfaces are removed by mechanical interactions and valleys are filled by wear debris, the average surface roughness of specimens would decrease to a stable state [7,8]. The generation of wear debris as a three-body layer might lubricate contact interfaces [9,10]. As a result, both the friction force and the wear rate may decrease and then level off during macroscale running-in.
However, when tests are performed at nanoscale, counter pair normally contact with a single asperity mode due to nanoscale roughness and nanoscale contact area. In this case, the traditional running-in mechanism may not be valid during nanowear process. Nevertheless, similar to running-in processes in macroscale devices, running-in processes in nanoscale 82 Friction 1(1): 81-91 (2013) devices are often observed [11]. Therefore, it is essential to understand the running-in process in the friction pair of nanoscale devices and the variation of the friction force and wear during initial sliding cycles.
In this study, the running-in process of single crystalline silicon coated with native oxide (Si-SiO x ) against a SiO 2 microsphere was investigated using an atomic force microscope in ambient air. The mechanism is discussed based on an analysis of the frictioninduced reduction of adhesion of the Si-SiO x /SiO 2 pair. Finally, a phenomenological model based on the results for friction and nanowear is proposed to explain the running-in process of the Si-SiO x /SiO 2 pair in ambient air.

Material and methods
Wafers of p-Si(100) with thicknesses of 0.5 mm were purchased from MEMC Electronic Materials, Inc., USA. The root-mean-square (RMS) roughness of each silicon wafer was about 0.07 nm over a 500  500 nm area. To simulate the real surfaces of dynamic MEMS, the native oxide layers on the silicon surfaces were not removed. Each surface was partially covered with Si-OH groups, which showed relatively hydrophilic with a water contact angle of 39°. Hereafter, this substrate will be called "Si-SiO x ". As a comparison, a hydrophobic silicon sample was obtained by immersing the original silicon in a 40% aqueous solution of hydrofluoric acid for 2 min, followed by rinsing in distilled water and methanol [12]. Due to the termination of Si-H groups on the silicon surface, this sample was relatively hydrophobic with a water contact angle of 83°. Hereafter, "Si-H" is used to denote this hydrophobic silicon sample.
Nanowear tests of native oxide-coated silicon wafers against SiO 2 tips (Si-SiO x /SiO 2 pair) were performed with an atomic force microscope (AFM, SPI3800N, Seiko, Japan) in an environment chamber with a vacuum capability. As shown in Fig. 1, the SiO 2 tip with a radius of 1 μm (Novascan Technologies, USA) moved horizontally on the silicon wafer surface under a normal load F n . The inset pictures show SEM images of the SiO 2 microsphere and its cantilever. Using a calibration probe with a force constant of 2.957 N/m, the normal spring constants of the cantilever of SiO 2 tips were calibrated to be 10.5-13.8 N/m [13]. If not specially mentioned, the applied normal load F n was 5 μN, the sliding speed was 0.8 m/s, the number of sliding cycles was 2000, and the displacement amplitude D was 100 nm. The total sliding distance in each cycle was 4D. All tests were carried out in air with a relative humidity (RH) of 50%-60% and room temperature of 20 °C -25 °C . After tests, friction forces were calibrated using a silicon grating with a wedge angle of 54°44' (TGF11, MikroMasch, Germany) [14]. To characterize the adhesive behavior of Si-SiO x /SiO 2 pair, the average adhesion forces F a were obtained by twenty pull-off tests. After nanowear tests, the topography of the wear area was scanned by a sensitive silicon nitride tip (MLCT, Veeco, USA), which had a curvature radius of 20 nm and a nominal spring constant of 0.1 N/m.

Running-in process of Si/SiO 2 pair
The friction loops and forces for the Si-SiO x /SiO 2 pair are plotted as a function of the number of sliding cycles N in Fig. 2. The friction loops presented four different shapes over various numbers of sliding cycles, as shown in the inset pictures of Fig. 2. During the first several cycles, the friction loop was in a parallelogram shape, which has been observed elsewhere [15,16]. At the fifth cycle, the parallelogram quickly changed to the shape of an hourglass. After 50 cycles, an ellipse shape with force peaks appeared. When the number of sliding cycles increased to about 200, the friction loops finally changed to an oblique parallelogram, which was preserved throughout the remaining cycles.
Because the shape of the friction loop changed as N increased, the friction force F t could not be calculated by simply taking the force difference between the forward and backward directions and dividing by two. Instead, the average F t was calculated by dividing the total energy dissipated in a sliding cycle (E) by the total sliding distance per cycle (4D) [17]: In the first cycle, the friction force F t of the Si-SiO x /SiO 2 pair was 3.9 μN. As the number of sliding cycles N increased, F t exhibited a sharp decrease during the first 50 cycles, gradually decreased between 50 and 300 cycles, and then maintained a stable value in the remaining cycles up to N = 2000. Compared to F t in the first cycle, the stable value of the friction force decreased by about 62% after the running-in process. The change in shape of the friction loops shown in Fig. 2 is related to the wear of the silicon surface [18]. After the nanowear tests, the resulting wear scars were scanned with a sharp Si 3 N 4 tip, as shown in Fig. 3

(a).
Although the contact pressure (1.3 GPa) between the SiO 2 tip and the silicon surface under a normal load of 5 μN was far less than the yield limit of monocrystalline silicon (7 GPa), material loss occurred and, after only five sliding cycles, a groove with a depth of 1.3 nm was created on the silicon surface [19]. As the number of sliding cycles increased, wear on the silicon surface became more severe, and the wear depth increased to 55 nm after 2000 cycles. To quantitatively characterize the wear behavior of the silicon surface during the running-in process, the wear rate γ was calculated by where Φ is the wear volume and l is the total sliding distance (= 4D  N). As shown in Fig. 3(b), the wear rate γ of the silicon surface underwent a sharp drop during the initial 50 sliding cycles, then gradually decreased to a constant value over the remaining cycles.

Effect of load and displacement amplitude on the running-in process of Si-SiO x /SiO 2 pair
To investigate the effects of experimental conditions on the running-in process, tests were performed at various normal loads F n and displacement amplitudes D in humid air. Figure 4(a) shows F t -N curves of the Si-SiO x /SiO 2 pair obtained at normal loads of 0.5 μN, 3 μN, and 5 μN. The inset picture shows the friction coefficient µ as a function of sliding cycles under the three loads. Here, µ was determined by µ = F t /L, where L is the sum of the normal load F n and the adhesion force F a between tip and silicon sample. As the number of sliding cycles increased, both the friction force F t and the friction coefficient µ, at all tested loads, exhibited a sharp drop within the initial 50 cycles, then gradually decreased to stable values during the remaining cycles. Compared to the initial friction force, the stable friction force was reduced by 74%, 69%, and 62% at normal loads of 0.5 μN, 3 μN, and 5 μN, respectively. When the normal load was lower, the degree of friction reduction was larger. Figure 4 However, the friction loops at 2000 cycles presented different behaviors under the three displacement amplitudes D tested, as shown in Fig. 5(b). When D = 100 nm or 250 nm, the vertical section of the wear scars showed a camber shape and the friction loops formed an oblique parallelogram at 2000 cycles. But when D = 500 nm, the vertical section of the wear scars was in a U slot shape. Since the wear scar had enough space for SiO 2 tip sliding, the friction force exhibited stable values within the central 400 nm and revealed peaked forces at the edges of the wear scars. These results indicate that the shape of the friction loop of Si-SiO x /SiO 2 pair was strongly dependent on the topography of wear scars on the silicon surface [18]. Figure 5(a) exhibits F t -N curves of the Si-SiO x /SiO 2 pair at displacement amplitudes D of 100, 250, and 500 nm. Similar to that at D = 100 nm and F n = 5 μN, the friction force decreased by about 62% during running-in processes at D = 250 and 500 nm. After 2000 cycles, grooves with depths of about 55 nm were formed regardless of the displacement amplitude, as shown in Fig. 5(b). Clearly, the influence of D (> 100 nm) on the friction behavior of the Si-SiO x /SiO 2 pair and on wear of the silicon surface was negligible.   Fig. 4(a) shows the friction coefficient µ as the function of sliding cycles at various loads.

Sharp drop of interfacial force during a running-in process
At macroscale, the decrease of friction during runningin process is normally attributed to the reduction in contact pressure after rough asperities on counter surfaces are flattened under mechanical action [7,8].
In this experiment, since the SiO 2 microsphere and silicon surface provided a single asperity contact, the macroscale topographic smoothening mechanism cannot explain the sharp drop in friction force during the nanoscale running-in process of Si-SiO x /SiO 2 pair. At nanoscale, the traditional Amonton's law cannot be used to calculate the friction force. To analyze the running-in behavior of nanoscale friction, we adapt the model proposed by Tambe et al. [20], which assumes that the nanoscale friction force between contact interfaces is a result of three components: F int due to interfacial adhesion, F def due to deformation, and F stick-slip due to stick-slip of contact interfaces.
Since no jump of tip occurred in our tests, the contribution of F stick-slip can be neglected here, and the friction force F t is mainly attributed to F def and F int . Although there could be some tribochemical contributions to wear, we assume that the substrate profile change in the wear track is mainly due to deformation. This would, at least, give an upper limit to the deformation-related contribution to the friction. Due to the nonlinearity of the elastoplastic stress-strain relationship of silicon, it is difficult to precisely calculate the deformation component of the friction force F def .
Here, the deformation-related force F def is estimated as the sum of the elastic deformation-related force F def-e and the plastic deformation-related force F def-p . This may somewhat overestimate the value of F def . F def-e can be determined by [21] Here, R is the radius of the SiO 2 tip and a is the contact radius between tip and sample, which can be estimated by the DMT model. Using Eq. (4), F def-e was calculated to be 0.194 μN in the first sliding cycle under an adhesion force of 1.35 μN. When the number of sliding cycles increased to 2000, F def-e was estimated to be about 0.177 μN, considering the surface damage and the variation in the adhesion force (see Fig. 7). It seemed that F def-e changed only a little during the nanowear process.
If the loss of material on the silicon surface was fully attributed to mechanical interactions, F def-p could be estimated by [22]   def-p y F S (5) where S is the projected area of plastic deformation along the sliding direction, and  y =7 GPa is the yield limit of monocrystalline silicon [19]. In each cycle, S could be estimated as where   is the wear volume on the silicon surface during each cycle. Using Eqs. (5) and (6), F def-p can be determined by Figure 6 shows the variation in F def (= F def-e + F def-p ) and the rest of the friction component F int (= F t  F def ) of the Si-SiO x /SiO 2 pair as the number of sliding cycles increases. Due to the existence of the oxide layer during the initial cycles and the formation of wear debris after several sliding cycles, the actual yield limit of the deformation area on the silicon substrate may be smaller than the yield limit of monocrystalline silicon cited above (7 GPa). As a result, the calculated values of F def in Fig. 6 may be somewhat overestimated. However, F def showed only a marginal decrease during the initial cycles and remained stable thereafter.
These results indicate that substrate deformation cannot be the main cause for the sharp drop in friction force during the running-in process. Thus, the friction change during running-in can be mainly attributed to the reduction in the adhesion-related interfacial force between the SiO 2 tip and the silicon surface.

Friction-induced reduction of adhesion during the running-in process
At nanoscale, variations in the interfacial force F int are normally induced by the transformation of surface properties of the counter pair [23]. To detect changes in surface properties during running-in, the adhesion force F a of the Si-SiO x /SiO 2 pair was measured before and after wear tests.
Since the size of the wear scar was less than 200 nm × 300 nm under the given conditions, the SiO 2 tip was difficult to locate in the wear area during tests of the adhesion force F a . As a result, F a was measured on a fresh silicon surface by the SiO 2 tip before and after wear tests. As shown in Fig. 7(a), the value of F a was 1.35 μN before the test and decreased to 0.24 μN (a reduction of ~82%) after 2000 sliding cycles. The inset pictures in Fig. 7(a) show SEM images of the SiO 2 tip before and after wear tests. The material transfer layer and wear debris can be observed on the contact area of the SiO 2 tip surface after wear tests. The deposition of substrate wear debris particles on the SiO 2 tip surface could reduce the actual contact area between the tip and substrate surface. This could contribute to the reduction in the adhesion force. Also, it is possible that the transferred silicon wear debris particles might make the tip surface more hydrophobic. If this happens, then the adhesion due to capillary effects could also be reduced [12,24]. In any case, these changes in the tip surface topography and chemistry must be responsible for the overall reduction of the adhesion force of the SiO 2 tip after the wear test.
The thickness of the native oxide layer on a Si wafer is typically about 2 nm [25]. Since the wear depth during the initial running-in period is about 2-3 nm (Fig. 3(a)), it is reasonable to assume that the running-in process removes the hydrophilic silicon oxide layer. To verify this, a wear area having a depth of 5 nm was prepared on a silicon surface by scanning-scratch, as shown in the inset picture in Fig. 7(b). Using a new SiO 2 tip, F a was detected on both fresh and worn silicon surfaces. As shown in Fig. 7(b), compared to that on the fresh silicon surface, F a decreased by ~64% on the worn silicon surface. Based on IR spectra, Mizuhara and Hsu [26] indicated that oxygen is far less reactive than water. Thus, a new oxide layer has difficulty forming on a worn silicon surface during a sliding process. Therefore, the hydrophobic property of the worn silicon surface is attributed to removal of the hydrophilic oxide layer on the Si-SiO x surface during running-in.
To further support this interpretation, the running-in process of a SiO 2 microsphere sliding against a Si-H surface was investigated. As shown in Fig. 8(a), unlike the Si-SiO x /SiO 2 pair, no sharp drop in friction force was observed during the running-in process of the hydrophobic Si-H/SiO 2 pair. At a RH of 50%-60%, the thickness of the adsorbed water layer was estimated to be 0.98 nm on the native oxide-coated silicon surface and 0.4 nm on the hydrophobic Si-H surface [27]. Therefore, the initial friction behavior would have a large contribution from the capillarity effect on the native silicon oxide surface, which would be lacking on the Si-H surface [28]. As a result, the initial friction force of the Si-H/SiO 2 pair was much smaller than that of the Si-SiO x /SiO 2 pair. However, when the native oxide layer was removed after 50 sliding cycles ( Fig. 3(a)), the friction force of the Si-SiO x /SiO 2 pair approached the value of the Si-H/SiO 2 pair (Fig. 8(a)). Figures 8(b) and 8(c) show the wear scars after 2000 cycles on silicon and on Si-H surfaces, respectively. The wear rate of the Si-H/SiO2 pair was ~0.2610 -12 m 3 /Nm, which was very similar to the stable wear rate (~0.3  10 -12 m 3 /Nm) observed after the running-in period (~50 cycles) for the Si-SiO x /SiO 2 pair. These results also suggest that the native oxide layer is removed during the running-in process of the Si-SiO x /SiO 2 pair. Such a running-in process has also been observed on macroscale silicon surfaces and DLC coating surfaces, where a sharp drop in friction force was attributed to the wear or removal of the surface oxide layer [10,29]. Figure 9 compares the adhesion and friction forces before and after 2000 cycles of sliding for the Si-SiO x /SiO 2 and Si-H/SiO 2 pairs. In the case of the Si-SiO x /SiO 2 pair, the adhesion force decreased by ~75% and the friction force decreased by ~62% during running-in. For the Si-H/SiO 2 pair, the adhesion force and friction force decreased by only ~7% and ~9%, respectively. Clearly, the sharp drop in friction force corresponds to a sharp drop in adhesion force.
According to the results for the adhesion force (Fig. 7), both the surfaces of silicon substrate and the SiO 2 tip may change their behavior and induce a drop in friction force during running-in of the Si-SiO x /SiO 2 pair. To detect how changes in the tip surface affect friction behavior, a new SiO 2 tip was first modified by repeated rubbing cycles on a clean substrate surface. After that, the friction force between the modified tip and the pre-worn silicon surface was measured in-situ. To have a comparison, the friction force between the modified tip and the original Si-SiO x surface was measured at a new location. As shown in Fig. 10, the friction forces on both the worn and the original Si-SiO x surfaces showed sharp drops as the number of sliding cycles increased. Further, all the friction forces measured on the worn silicon surface were a little lower than those measured on the original silicon surface. After 2000 sliding cycles, F t decreased by about 50% on the original silicon and about 62% on the worn area. Therefore, modification of the SiO 2 tip during running-in also played some role in the sharp drop of friction force for the Si-SiO x /SiO 2 pair. Figure 11 schematically shows the running-in process of the Si-SiO x surface against a SiO 2 microsphere in ambient air. The silicon surface with a native oxide layer was partially covered with Si-OH groups, which allowed adsorption of water molecules from the humid air [12,24]. While the tip slides on the silicon surface, a water bridge will form between the Si-SiO x substrate and the SiO 2 tip surface, as shown in Fig. 11(a). As a result, both the friction force F t and the wear rate exhibited a large value over the initial several cycles. However, when the native oxide layer on the Si-SiO x surface was removed by the SiO 2 tip via water-induced corrosion [30,31], the Si-Si network of the Si(100) substrate will be exposed to air, and the silicon surface becomes more hydrophobic due to the reduction of Si-OH groups [32]. Because of the reciprocating scratch by the SiO 2 tip, the Si-Si network could be broken by shear stresses transmitted from the Si-O-Si bridges between the SiO 2 tip and the silicon substrate, as shown in Fig. 11(b) [26,33]. At the same time, the tip surface could become more hydrophobic because of dehydroxylation reactions during this process [12,24]. Since the transformation of silanol bonds to siloxane bonds is exothermic, dehydroxylation reactions probably occur on the SiO 2 tip surface under shear stress during the initial sliding process [34]. The hydrophobization of the contact area on the SiO 2 tip and the silicon surface would reduce both the friction force and the wear rate, as shown in Fig. 11(c) [35,36].

Mechanism of the running-in process of Si/SiO 2 pair
In summary, the sharp drops in friction force and wear rate during running-in were caused by hydrophobization of the SiO 2 tip and the silicon surface, as well as by lubrication provided by wear debris. Unlike mechanical interactions in macroscale devices, tribochemical reactions play a dominant role during running-in of nanoscale Si-SiO x /SiO 2 pair.

Conclusions
Using an AFM, the running-in process of Si-SiO x /SiO 2 pair in ambient air was investigated under various normal loads and displacement amplitudes. The main conclusions can be summarized as follows.
Nanowear of Si-SiO x /SiO 2 pair exhibited a typical running-in process. During the initial 50 sliding cycles, the friction force of the Si-SiO x /SiO 2 pair rapidly decreased to 40% of its initial value, and the wear rate on the silicon surface sharply decreased to 10% of its initial value. In the remaining sliding cycles up to 2000, both the friction force and the wear rate gradually decreased to constant values.
During running-in, the sharp drop in friction force between the silicon surface and the SiO 2 tip was mainly attributed to wear of the surface oxide layer and to change in the surface of the SiO 2 tip, which reduced adhesion-related interfacial forces.
An analysis indicated that the running-in process of the Si-SiO x /SiO 2 pair was dominated by removal of the native oxide layer; this may be accompanied by hydrophobization of the contact area on the SiO 2 tip and the silicon surface.