Spatial evolution of friction of a textured wafer surface

Mechanical failure of integrated circuits and micro-electro-mechanical systems (MEMS) demands new understanding of friction in small devices. In present research, we demonstrated an in situ approach to measure sliding friction of a patterned surface composing multi-materials and structures. The effects of materials and surface morphology on friction and electrical contact resistance were investigated. The material transfer at the interface of dissimilar materials was found to play dominating roles in friction. The current work provides important insights from the fundamentals of friction that benefit the design of new micro-devices.


Introduction
Since Leonardo Da Vinci's first design in tangential force measurement [1][2], the surface friction has been studied for centuries [3][4]. The sliding friction can be evaluated against contact conditions and environments. Existing formulas used to calculate friction are based on static conditions involving a tangential force and applied load over an average size of contact area. In micro-meter length scale, macroscopic friction law is not always applicable [5][6]. With the development of micro system the surface texturing has become important in frictional control. Some natural creatures excel in controlling friction employing specially designed texture to benefit their behavior. Geckos' toes have special shaped seta system which produces high friction enabling geckos to walk and climb easily [7][8]. The stridulation of crickets is produced by the friction induced vibration owing to the textured wing [9][10]. In industrial applications, friction and wear have become the one of the major causes of failure in small devices such as micro-electromechanical systems (MEMS) [11][12]. For MEMS actuators composed of components like gears and hinges, friction and wear are inevitable and are important factors for better performance and longer lifespan. In order to improve the quality and the service life of microscale devices, friction and stiction control in interfacial structures is highly desirable [13][14].
In order to understand friction and its impact on surface texture, an in situ method was employed to investigate the tribological properties of the surface during sliding. To achieve precise control, we designed a surface that is textured with two different materials, formed in strips parallel to each other, having different tribological and electrical properties. The electrical contact resistance (ECR) and the coefficient of friction (COF) were evaluated to monitor the change on the surface with evolution of friction. The spatial evolution of friction and its corresponding resistance were expected to reveal the transfer of surface materials due to sliding and assist design of textured surfaces. X-ray photoelectron spectroscopy (XPS) imaging was also conducted to validate the materials transfer.

Materials
The sample is a commercial wafer provided by Semitech Inc. It was chemical-mechanically polished a Xiao, Wang, and Fox contributed equally to this work. * Corresponding author: Hong LIANG. E-mail: hliang@tamu.edu and pre-patterned with parallel nickel [15] and silicon dioxide (SiO 2 ) stripes. The width of the Ni strips is in the range of 70-88 μm while that of SiO 2 strips is in the range of 200-300 μm. Figure 1 shows a photo of the Ni-Si wafer prior to testing. The hardness of Si is 13 GPa [16], SiO 2 is 9 GPa [17], and Ni is 600 MPa [18]. The sample was rinsed with acetone for 20 minutes in an ultrasonic cleaner prior to the friction test. To build a conductive circuit for ECR test, the wafer was attached to an aluminum stage using a carbon tape. A droplet of silver paint was added to the wafer edge so that current could get through the conductive Ni strips. A standard bearing ball of AISI 52100 steel with 6 mm diameter was used to slide against the wafer surface.

Tribological testing and characterization
Friction experiments were conducted using a homemade ball-on-flat tribometer [19] with a linear reciprocal motion. The applied load and the sliding speed are 0.3 N and 2 mm/s, respectively, with a track length of 3.5 mm. The coefficient of friction and the resistance between the pin and the disk were recorded simultaneously. The "triboscopical images" revealing the evolution of COF and ECR with increasing number of rubbing cycles was produced [19]. The calculated Hertzian contact pressure is 93.4 MPa on Ni strips and 57.8 MPa on SiO 2 . After tribotests, surface profile was measured using a white light interferometer (Zygo Corp., NewView 600).
Chemical analysis was conducted using an XPS (Kratos Axis Ultra imaging). The Al Kα line with energy of 1486.7 eV was utilized as incident beam and the typical pressure in the spectrometer was around 10 -7 Pa. Normal incident X-ray was introduced to the sample while the angle between the incident direction and the detector direction is 45°. Images were acquired at 853 eV which is the characteristic binding energy (BE) of Ni 2p3/2 .

Results and discussion
The friction measurement is displayed in a triboscopical image as shown in Fig. 2(a). The X-axis is the number of sliding cycles while the Y-axis is the position of the sliding pin along the linear track on the disk. The Y-axis is composed of 512 elements with a length of approximately 7 μm for each element. The intensity of the brightness in Fig. 2(a) represents the value of COF and brighter zone represents higher COF. The patterns of Ni and SiO 2 stripes on the wafer are clearly distinguished especially in later cycles. A visible change in COF begins around cycle 20. It becomes more evident after cycle 25. Subsequently, bright strips appear at the interface between Ni and SiO 2 indicating wear initiated at the interface. Between cycles 15 and 40, the friction continues to rise and varies in a wide area. Soon after the cycle 45, a high friction zone covering entire sliding track can be distinguished.  The in situ ECR (Ohms, Ω) measurement along the wear track was conducted simultaneously. The ECR data were plotted for all cycles on a tricboscopical image, as shown in Fig. 2(b). The intensity of the brightness represents the measured contact resistance. Higher intensity indicates larger contact resistance. The regions with low contact resistance (blue or dark) are Ni stripes, the bright regions are SiO 2 stripes. The difference between pre-patterned Ni and SiO 2 stripes is more evidently demonstrated in the ECR image. In the first 20 cycles, the measured ECR remains stable. The boundaries between Ni and SiO 2 stripes can be clearly distinguished. As shown in Fig. 2(b), the contact resistance begins to drop from the 21 st cycle. A transition region starts to form at the interface between Ni and SiO 2 stripes. After the 25 th cycle, the decrease of contact resistance is observed in the prior high contact resistance positions, the measured ECR decreasing from 10 6 Ω to 10 5 Ω in 10 cycles. Since all other experimental conditions remain the same, this drop could be attributed to the transfer of Ni during sliding. Conductive Ni transferred to the non-conductive SiO 2 strips greatly reduces the contact resistance in the SiO 2 strips. At the 45 th cycle, the contact resistance reaches the lowest value, 10 4 Ω. At the end of the test, from the 50 th cycle to 60 th cycle, the contact resistance is observed to rise. An increase of COF is also identified in Fig. 2(a) at the end of the test suggesting a severe wear at the corresponding positions. Ni transferred to SiO 2 strips could be worn out leading to the increase of ECR.
Furthermore, the average ECR and average COF of each cycle were calculated and plotted in Fig. 3. The results demonstrate the same trend observed in triboscopical image. In the first 20 cycles, both ECR and COF change little with increasing number of cycles. This result suggests that the friction and wear develop slowly at the beginning. This is due to the super finish of the wafer and effects of asperities are limited. After that, the measured ECR starts to decrease fast and reaches the lowest value around the 45 th cycle, while the COF starts to increase gradually. With more rubbing cycles, wear steadily progresses leading to the change of the wafer surface morphology. The surface becomes rougher in time. The temperature is high at the contact between the asperities of the wafer surface and the steel ball. The Ni particles scratched away from the Ni strip adhere to the ball due to the high temperature and then travel along the SiO 2 strip. In the sling on SiO 2 surface, transfer of Ni from steel ball to SiO 2 could be expected. From cycle 20 to cycle 45, the resistance keeps decreasing which indicates the increase of the conductive film (the transferred Ni film) on the SiO 2 surface. In the last 10 cycles of the test, the measured ECR is observed to raise back, which is accompanied by the increase of COF. The rapid increase in COF demonstrates the severe wear in this stage. At this stage the Ni film is gradually worn out and the thickness of Ni film decreases while the resistance increases accordingly.
In order to investigate the impact of rubbing cycles on the tribological properties, a multi-sequential test was conducted. The resulting optical micrograph of the wear track is displayed in Fig. 4(a). In region A, B and C the number of rubbing cycle is separately 20, 45 and 60. A slight wear can be identified from the color difference in region A. The morphology image obtained by white light interferometer confirms this subtle wear as shown in Fig. 4(b). In region B, the sign of wear is evident and surface damage along the wear track can be seen. Figure 4(c) shows the morphology image in this region. As the number of rubbing cycles increases to 60, the wear track gets much wider and severe surface damage is produced. The Ni strip was worn out as shown in Fig. 4(d). The average surface roughness of the wear track in Fig. 4(b) is 2 nm. And  this value reaches to 24 nm and 67 nm in Figs. 4(c) and 4(d). From the morphological characterization, it is evident that the average surface roughness in the wear track region increases significantly with cycles, and more wear damage is introduced on the prepatterned Ni and SiO 2 stripes.
To confirm the material transfer during sliding, XPS imaging was conducted on the wear tracks. Results are shown in Fig. 5. Figure 5(a) shows a worn region with the wear track crossing two pre-patterned Ni stripes. Figure 5(b) is the enlarged image of the inset in Fig. 5(a). The bright zone indicates the existence of Ni. The distribution of Ni along the wear track can be clearly determined. This confirms the transfer of Ni from pre-patterned Ni strips to SiO 2 strips during the sliding process. The long traveling distance of the Ni can be observed from Fig. 5. The brightness in the wear track is between that of the Ni strip and the SiO 2 strip, which means the concentration of Ni in the wear track is between the two pure strips and mechanical mixing of the two elements happens in the wear track. The gradual decrease of ECR from Fig. 3 demonstrates the mechanical mixing of the elements is an accumulative process during sliding. In this stage, Ni transfer dominates friction. The following increase in ECR suggests the decrease of Ni concentration and the dominating process is wear of the material.
Based on experimental results, three stages can be distinguished in the cyclic sliding process for the parallel strip structures. In the first stage, the COF is stable and the wear is negligible. In the second stage, the COF gradually increases and evident wear occurs. More importantly, for the parallel strips composed of different materials, the softer material transfers from its own strip to the harder strip. If the electrical conductivities of the two materials have a huge difference, the material transfer will cause a remarkable influence and may cause the failure of the component. In the third stage as shown in Fig. 3, the COF increases progressively and the wear develops accordingly. For most MEMS applications, the components will lose the precision and reliability under severe wear. Based on the fact that the metal-silicon strip structure is widely used in semiconductor industry, the friction and wear should be controlled so that the component can work under the first stage because even small amount of metal transfer will damage the electrical conductivity of the whole component.

Conclusion
In order to study the kinetics of materials transfer during sliding, friction and electrical contact resistance of a patterned surface were studied. The influence of materials and material transfer on sliding friction was observed in situ. The average COF gradually increased from 0.18 to 0.67 after 60 rubbing cycles. The average ECR reduced from 10 6 Ω to 10 4 Ω firstly and then increased to 10 5 Ω at the end of tests. The XPS results have proven the transfer of materials for a long distance along the wear track and the mechanical mixing is an accumulative process. This study reveals insight into the possible failure mechanisms of sliding of small devices with textured surfaces.