In order to study the load effect on the fabrication of nanolines on silicon, nanoscratches were created on the Si(100) surface under various normal loads F _n. As shown in Figure 3a, when the normal load was 2 mN, the corresponding Hertz contact pressure was estimated as 6.3 GPa. The scratch damage was very weak, and only a small protuberance was produced on the Si(100) surface. However, when the normal load was above 5 mN, the corresponding Hertz contact pressure was larger than 8.5 GPa. The scratch damage became severe, and the grooves were machined into the Si(100) surface. After etching in 20 wt.% KOH + IPA solution for 40 min, both protuberance and groove were evolved to nanolines with similar height, as shown in Figures 3b, c. The tilt angle of the etched sidewalls is about 50°, indicating that the incline sidewall is in the (111) plane 15 .

It was noted that the width of the nanolines was strongly dependent on the width of the damage area on the Si(100) surface. With the increase in the normal load from 2 mN to 60 mN, the width of the fabricated nanolines on the Si(100) surface increases from 352 nm to 2,219 nm, which are well consistent with the width of the corresponding damage area before etching (from 410 nm to 2,470 nm). The results indicated that the width of the nanolines can be effectively controlled by the scratching load. Moreover, although the scratch under a normal load of 2 mN (the corresponding Hertz contact pressure is 6.3 GPa) induced a very weak damage on the Si(100) surface, it produced a nanoline with the similar height as that obtained by 60 mN after KOH etching. Therefore, a contact pressure of 6.3 GPa was enough to fabricate a mask layer on the Si(100) surface, and a low-destructive nanofabrication technique could then be developed based on this method.

Besides the scratching load, the etching time also shows a strong effect on the fabrication of nanolines on the Si(100) surface. After scratching on the Si(100) surface under a normal load of 60 mN, the specimens were etched in 20 wt.% KOH + IPA solution for various periods. Figure 4 shows the AFM images of the nanolines, and Figure 5 plots the variation of the height of these nanolines with etching time. Due to the mask effect of the native oxide layer, the height of the nanolines kept a very low value during the initial 30 min. After the native oxide layer was etched away, the height of the nanolines reveals a sharp increase with the increase of the etching time. Clearly, the scratched area on the Si(100) surface shows a strong mask ability to the etching of KOH solution. However, since the formation of the scratched area was the combined results of both mechanical interaction and tribochemical reaction, the scratched area of the Si(100) surface may contain a thin superficial oxidation layer and a thick tip-disturbed silicon layer in the subsurface 11 17 . Besides the superficial oxidation layer, the mask ability of the tip-disturbed silicon layer needed to be further investigated.

In order to detect the mask ability of structurally disturbed silicon, a groove structure was produced on the Si(100) surface after a line scratch under F _n = 5 mN, as shown in Figure 6a. It was well known that an oxidation layer may be generated on the surface of the scratch area due to the tribochemical reaction. In the previous studies 12 13 14 15 , such oxidation layer was believed to be the only mask structure on the scratch area. In fact, a thick distorted silicon layer including amorphous and deformed silicon may also be produced under the surface oxidation layer during the scratching process 11 17 . To study the mask ability of the distorted silicon layer in KOH solution, the surface oxidation layer of the Si(100) sample was removed by etching in HF solution for 15 min. As shown in Figures 6a, b, the depth of the groove increases from 3 nm to 8 nm after HF etching. Since the thickness of the oxidation layer on the scratch of the silicon surface was less than 5 nm, the oxidation layer formed on the groove during scratching should have been fully removed by HF etching 22 . When the sample was further etched in 20 wt.% KOH + IPA solution for 25 min, a hillock with the height of 710 nm was formed on the scratch area, as shown in Figure 6c. The results strongly suggested that besides the oxidation layer, the distorted silicon layer can also reveal excellent mask ability in KOH solution.

Based on the friction-induced selective etching method, a series of nanopatterns were fabricated on the Si(100) surface, as shown in Figure 7. In order to improve the fabrication efficiency, the native oxidation layer on the Si(100) surface was removed by HF etching before scratching. A protrusive array of nanolines with a height of about 100 nm (Figure 7a) was produced by line scratch under F _n = 2 mN and post-etching in 20 wt.% KOH + IPA solution for 5 min. A protrusive array of nano-mesas with about 70 nm in height (Figure 7b) was fabricated by scanning scratch under F _n = 15 μN and post-etching for 2 min. A nano-ring with a height of about 52 nm (Figure 7c) was produced by scanning scratch under F _n = 20 μN and post-etching for 2 min, and a nanoword 'TRI' (short for 'Tribology Research Institute') with a height of about 31 nm (Figure 7d) was fabricated by scanning scratch under F _n = 20 μN and post-etching for 1 min.

Therefore, through the control of scratch load, scan trace, and etching time, a nanopattern with the required shape, height, and width can be fabricated on the Si(100) surface with the present method. Compared with fabrication by the prevalent photolithography, this method introduces an optional fabrication approach with more flexibility and lower cost, which could provide new opportunities for the mold production in nanoimprint and prototype fabrication in micro-systems 23 24 .

To understand the fabrication mechanism of the friction-induced selective etching, the cross-sectional microstructure of the scratch on the Si(100) surface was studied by TEM, energy dispersive X-ray spectroscopy (EDX), and a scanning Auger nanoprobe. Figure 8a shows the cross-section TEM (XTEM) bright-field microstructures of the scratch sections. The inserted image in Figure 8a was taken from the single crystal area away from the scratched position. Figure 8b reveals that the scratched area is composed of a thin superficial layer and a relatively thick distorted (amorphous and deformed) case in the subsurface. Both EDX and scanning Auger nanoprobe analysis revealed that oxygen (from SiO_x) can be only detected at the scratched surface areas, but cannot be found from the majority of the amorphous case and the severe deformed zone 25 . Therefore, the mechanical interaction through amorphization and matrix deformation have played a more important role than oxidation reaction during the scratching process on the Si(100) surface.

The above results suggested that the residual scratched structure can still resist against KOH etching after removing the superficial oxidation layer by HF etching. To further understand the etching-resistant mechanism of the structurally disturbed silicon layer, a scanning Auger nanoprobe was used to detect the variation of the chemical composition on the scratched area during the etching process. As shown in Figure 9, a mesa of 5 × 5 μm was fabricated on the Si(100) surface under F _n = 50 μN and scratching cycles of 10 by AFM in a scanning scratch mode 25 . The height of the mesa was measured as 2.5 nm, as shown in Figure 9a. The scanning Auger nanoprobe analysis indicated that the intensity of oxygen on the mesa surface is higher than that on the original silicon surface (Figure 10a) and that the thickness of the oxidation layer is about 1.7 nm (Figure 10b). After the oxidation layer on the silicon surface was removed by etching in HF solution for 3 min, the height of the mesa decreased to 1.4 nm (Figure 9b). Due to the native oxidation of the silicon surface in air, the mesa showed a similar oxidation layer as that on the original silicon surface (Figure 10b). Finally, after etching in 20 wt.% KOH + IPA solution for 2 min, the height of the mesa sharply increased to 70 nm (Figure 9c). Therefore, since the native oxidation layer on the mesa is the same as that on the original silicon surface, the mask behavior of the mesa after etching in HF solution should be mainly attributed to the amorphous and deformed silicon induced by scratching.

It was well known that single-crystal silicon shows strong anisotropic etching behavior in alkaline solutions. Based on the electrochemical model proposed by Seidel et al. 21 , the anisotropic etching behavior of silicon was mainly attributed to the differences on the density of the dangling bond as well as the structure and energy level of the backbond on the silicon surface. When the density of the dangling bond and the energy level of the backbond are high, the etching rate of the silicon surface in alkaline solutions is fast. Since {100} surfaces show the highest dangling bond density and a relatively high backbond energy level, the etching rate of {100} surfaces in alkaline solutions are the fastest. As a comparison, with one dangling bond and three backbonds per surface atom, the {111} surfaces show the lowest dangling bond density and the slowest etching rate in alkaline solutions. Finally, even with the similar dangling bond density as {111} surfaces, the {110} surfaces show a relatively higher backbond energy level and higher etching rate than {111} surfaces.

When the monocrystalline structure of silicon atoms on the Si(100) surface was transformed to amorphous or deformed silicon during the scratching process, the atomic arrangement on the surface became out of order. As a result, the average dangling bond density and backbond energy level of the silicon atoms on the distorted surface were lower than those of the original Si(100) surface 26 . Therefore, the amorphous and deformed silicon is relatively difficulty to be etched and can act as an etching mask of the Si(100) surface in KOH solution. The results may not only help us to understand the fabrication mechanism involved in the friction-induced selective etching technique, but also provide a way towards developing a low-destructive, low-cost, and flexible nanofabrication technique.