**3. Experimental achievements**

Up to now, we have developed a process successfully, and realized a He\*-MAB lithography of Si(111)/Si(110)/Si(100) wafer substrates with SAMs, where the SAMs was formed not on metal/oxide coating layers but directly on silicon substrate. The process of atom lithography using He\*-MAB and organosilane SAMs to create the arrays of silicon micro- and nanostructures on a Si substrate were illustrated in Scheme 1. The first step was the formation of very fine organosilanes (ODTS, DDTS and HTS) SAMs as a resist on Si(111)/Si(110)/Si(100) surface by introducing a hydrogenation process into the pretreatment of silicon wafer and a controlled argon gas atmosphere at elevated

Atom Lithography: Fabricating Arrays of Silicon Microstructures

the exposed SAM resist against chemical etching.

Using Self-Assembled Monolayer Resist and Metastable Helium Beam 59

imperfect SAM formation (i.e., the density of SAM molecules anchored on the Si(111) surface through chemical bonds is not perfectly uniform side by side) on the silicon substrate, which induced a local fluctuation in the etching rate of transferring the latent image of the SAM resist onto the underlying silicon substrate and, consequently, produced a rough surface. Flatter surfaces was obtained using a perfect SAM and under optimal etch conditions. Another characteristic was shown in figure 3, i.e., an upward fringe around the edge of the top of the square micromesas. The etching rates for the silicon wafer in the KOH solution depended strongly on its orientation [33]. Thus, the etching parallel to the Si(111) surface proceeded much faster than that perpendicular to the Si(111) surface and undercuts the sidewalls of a micromesa capped with a durable SAM resist that withstands the etching by KOH solution, leaving a fringe around the micromesa. After the sample was dried, the fringe rolls up or down around the edge of the silicon micromesa. The upward fringe was clearly observed in the 3D image as shown in figure 3(b), suggesting the strong durability of

Fig. 3. Tapping-mode AFM images with sectional analysis of square silicon micromesa array created on Si(111) substrate (Negative patterning, exposure time: 40min; etching time:

We also obtained a positive pattern within the exposed region with another TEM grid window near the border of the irradiation area. The border region accepted considerably sparser irradiation of He-MAB than the center of He-MAB. Consequently, this exposed border region only changed from hydrophobic to hydrophilic without cross-linking between molecules, which leaded to a weak resistance to chemical etching. Figure 4 showed AFM images and the sectional analysis of the microwell array formed on the Si(111) substrate, corresponding to the positive patterning. A square silicon microwell array with a size of ~7×7 µm2 and a depth of ~10 nm was formed. The topography and line profile of the microwell array indicated a much rougher surface with an rms roughness of ~7 nm. In comparison with the depth of ~10 nm, this large rms value suggested that it was difficult to achieve a perfect positive pattern transfer using the ODTS SAM resist, possibly because the

60min, SAM: ODTS) (a): 50×50 µm2; (b) 3D image; (c): 21.7×21.7 µm2).

temperature. After the development of micro stencil masks for He\*-MAB exposure to generate latent patterns in SAMs was carried out, the chemical etching process to transfer the latent patterns into the underlying silicon substrate was examined and optimized. By adjusting experimental parameters (i.e., different masks, exposure time, etching time, etc.), various arrays of silicon microstructures were fabricated successfully. The resulting arrays of silicon microstructures on Si substrates were characterized by the AFM/SEM.

Fig. 2. (a) SEM image of negative patterning (exposure time: 60min; etching time: 5min) and (b) positive patterning (exposure time: 20min; etching time: 5min) on Si(110) substrate with the ODTS SAM and DDTS SAM, respectively. Mask: TEM Cu grid.

The SEM images of two samples were displayed in figure 2. The pattern from the stencil mask (as shown in figure 1(a)) was transferred into the Si(110) substrate, and the resulting micromesa/microwell arrays of silicon fabricated on the Si(110) substrate were very intact and clear (as shown in figure 2). The preliminary SEM observations revealed that the direct fabrication of the array on the surfaces of Si(110) substrates by He\*-MAB and SAMs was viable, and our experimental results indicated that the patterning also appeared on either kind of surface of the Si(111) or Si(100) substrate with various organosilane SAM mentioned above. The main experimental results and discussions were given as following.

### **3.1 Arrays of Si(111) microstructures**

The AFM images with the sectional analysis of the square micromesa arrays formed on the Si(111) substrate with ODTS SAM were shown in figure 3, which corresponds to negative patterning with an exposure time of 40 min and an etching time of 60 min. A square silicon micromesa arrays with a size of ~7×7 µm2 and a nanoscale edge resolution of ~100 nm was successfully fabricated. The height of the step was ~30 nm, indicating an etching rate of 0.5 nm/min. The width of the step edge was 1-3 pixels (41~123 nm). The sharpest step edge had a width of 1 pixel (41 nm) and a height of 22 nm, indicating a steep sidewall with an aspect ratio of 2:1, which was considered to be the achievable limit in the present case. In fact, after subtracting the effects of the tip shape on the lateral resolution induced by the AFM tip with a ~10 nm radius of curvature at its end, the real step width should be smaller than the value mentioned above, and the aspect ratio should be larger. The cross sections were not flat, and the rms roughness of the top area of the micromesa and the pedestal base area were 6.2 nm and 5.8 nm, respectively. This relatively rough characteristic could be ascribed to the

temperature. After the development of micro stencil masks for He\*-MAB exposure to generate latent patterns in SAMs was carried out, the chemical etching process to transfer the latent patterns into the underlying silicon substrate was examined and optimized. By adjusting experimental parameters (i.e., different masks, exposure time, etching time, etc.), various arrays of silicon microstructures were fabricated successfully. The resulting arrays

Fig. 2. (a) SEM image of negative patterning (exposure time: 60min; etching time: 5min) and (b) positive patterning (exposure time: 20min; etching time: 5min) on Si(110) substrate with

The SEM images of two samples were displayed in figure 2. The pattern from the stencil mask (as shown in figure 1(a)) was transferred into the Si(110) substrate, and the resulting micromesa/microwell arrays of silicon fabricated on the Si(110) substrate were very intact and clear (as shown in figure 2). The preliminary SEM observations revealed that the direct fabrication of the array on the surfaces of Si(110) substrates by He\*-MAB and SAMs was viable, and our experimental results indicated that the patterning also appeared on either kind of surface of the Si(111) or Si(100) substrate with various organosilane SAM mentioned

The AFM images with the sectional analysis of the square micromesa arrays formed on the Si(111) substrate with ODTS SAM were shown in figure 3, which corresponds to negative patterning with an exposure time of 40 min and an etching time of 60 min. A square silicon micromesa arrays with a size of ~7×7 µm2 and a nanoscale edge resolution of ~100 nm was successfully fabricated. The height of the step was ~30 nm, indicating an etching rate of 0.5 nm/min. The width of the step edge was 1-3 pixels (41~123 nm). The sharpest step edge had a width of 1 pixel (41 nm) and a height of 22 nm, indicating a steep sidewall with an aspect ratio of 2:1, which was considered to be the achievable limit in the present case. In fact, after subtracting the effects of the tip shape on the lateral resolution induced by the AFM tip with a ~10 nm radius of curvature at its end, the real step width should be smaller than the value mentioned above, and the aspect ratio should be larger. The cross sections were not flat, and the rms roughness of the top area of the micromesa and the pedestal base area were 6.2 nm and 5.8 nm, respectively. This relatively rough characteristic could be ascribed to the

above. The main experimental results and discussions were given as following.

the ODTS SAM and DDTS SAM, respectively. Mask: TEM Cu grid.

**3.1 Arrays of Si(111) microstructures** 

of silicon microstructures on Si substrates were characterized by the AFM/SEM.

imperfect SAM formation (i.e., the density of SAM molecules anchored on the Si(111) surface through chemical bonds is not perfectly uniform side by side) on the silicon substrate, which induced a local fluctuation in the etching rate of transferring the latent image of the SAM resist onto the underlying silicon substrate and, consequently, produced a rough surface. Flatter surfaces was obtained using a perfect SAM and under optimal etch conditions. Another characteristic was shown in figure 3, i.e., an upward fringe around the edge of the top of the square micromesas. The etching rates for the silicon wafer in the KOH solution depended strongly on its orientation [33]. Thus, the etching parallel to the Si(111) surface proceeded much faster than that perpendicular to the Si(111) surface and undercuts the sidewalls of a micromesa capped with a durable SAM resist that withstands the etching by KOH solution, leaving a fringe around the micromesa. After the sample was dried, the fringe rolls up or down around the edge of the silicon micromesa. The upward fringe was clearly observed in the 3D image as shown in figure 3(b), suggesting the strong durability of the exposed SAM resist against chemical etching.

Fig. 3. Tapping-mode AFM images with sectional analysis of square silicon micromesa array created on Si(111) substrate (Negative patterning, exposure time: 40min; etching time: 60min, SAM: ODTS) (a): 50×50 µm2; (b) 3D image; (c): 21.7×21.7 µm2).

We also obtained a positive pattern within the exposed region with another TEM grid window near the border of the irradiation area. The border region accepted considerably sparser irradiation of He-MAB than the center of He-MAB. Consequently, this exposed border region only changed from hydrophobic to hydrophilic without cross-linking between molecules, which leaded to a weak resistance to chemical etching. Figure 4 showed AFM images and the sectional analysis of the microwell array formed on the Si(111) substrate, corresponding to the positive patterning. A square silicon microwell array with a size of ~7×7 µm2 and a depth of ~10 nm was formed. The topography and line profile of the microwell array indicated a much rougher surface with an rms roughness of ~7 nm. In comparison with the depth of ~10 nm, this large rms value suggested that it was difficult to achieve a perfect positive pattern transfer using the ODTS SAM resist, possibly because the

Atom Lithography: Fabricating Arrays of Silicon Microstructures

**3.2 Arrays of Si(110) microstructures** 

than that in the case of underetching.

periodicity of 12.5 um.

Using Self-Assembled Monolayer Resist and Metastable Helium Beam 61

The AFM images of the square micromesa arrays formed on the Si(110) substrate with ODTS SAM was shown in figure 5, which corresponded to negative patterning with an exposure time of 20 min and an etching time of 5 min. A square silicon micromesa array with a micromesa size of ~7×7 µm2 (e.g., the size of one mesh of TEM stencil mask) was fabricated successfully on the substrate. The average height of the step was 27 nm, indicating an etching rate at 5 nm/min. The sharpest step edge had a width of 1 pixel (41 nm) and a height of 26 nm, indicating a steep sidewall with an aspect ratio of 1:1.7. The sectional analysis shown in Figure 5(a) indicated that the rms roughness of the top area of the micromesa and the pedestal base area were 2 nm and 15 nm, respectively. The flat top area of the micromesas could be observed clearly, suggesting the strong durability of the exposed SAM resist against chemical etching. The flaw pedestal base area of the micromesa was also very conspicuous, and this rough characteristic could be attributed to the imperfect SAM formation on the silicon substrate, which induced a local fluctuation in the etching rate of transferring the latent image in SAM resist onto the underlying silicon substrate and, consequently, produced a rough surface. As described in scheme 1, the negative-tone sensitivity of the SAM might be mainly due to the cross-linking polymerization of ODTS molecules induced by the irradiation of the He\*-MAB [19,20]. In addition, it was noteworthy that the etching time had a notable influence on the production of high-quality patterning. In our practical runs, the following etching times were chosen: 5, 10, and 20 min with identical exposure times. The sectional analysis of AFM images of these samples indicated that the rms roughness of top area of the micromesa in the case of overetching was bigger

Fig. 5. Tapping-mode AFM images of periodic arrays of square silicon micromesa fabricated on Si(110) substrate with ODTS SAM (exposure time: 20min, etching time: 5min): (a) AFM image with sectional analysis of the arrays of silicon micromesas with a height of about 26 nm ; (b) 3D views of the arrays of silicon micromesa over a 80 um×80 um area and with a

hydrophilic area might not be much weaker in resisting the etchant than the unexposed hydrophobic area.

Fig. 4. Tapping-mode AFM images with sectional analysis of square silicon microwell array created on Si(111) substrate ( Negative patterning, exposure time: 40min; etching time: 60min, SAM: ODTS) (a): 50×50 µm2; (b) 3D image; (c): 21.7×21.7 µm2).

Since the ODTS SAM resist comprised highly oriented 18C-long-chain hydrocarbon molecules chemically bonded to the silicon substrate, its surface became hydrophobic [30-32]. Previous studies revealed that, under irradiation of He-MAB with a 20 eV internal energy, C-H bond scission and C-C bond scission occured and a C=C double bond could be formed at the outermost surface of the SAM [34-37]. At low doses, breaking of the C-H bond caused H loss, leaving a C-rich film that could be oxidized in air. Consequently, the exposure region was changed from hydrophobic to hydrophilic, leading to a positive pattern [38,39]. At high doses, further degradation of the hydrocarbon molecules caused fragmentation of the carbon chain, and cross-linking polymerization between adjacent fragmented SAM molecules might occur, giving rise to the formation of a highly resistant carbonaceous layer and leading to a negative pattern [36]. In addition, the contamination resist, which was induced by hydrocarbon molecules remaining on the inner wall of the vacuum chambers, remained a concern [40-41]. When the metastable atoms impacted the surface, they transferred their internal energy to these physisorbed hydrocarbon molecules and induced a polymerization change within the exposed region. To clarify whether these resist materials could remain on the surface even after immersion in a KOH solution, a clean Si(111) sample without an SAM was used. Only a trace of negative contrast was observed, indicating a negligible contamination effect. In our negative pattern case, the exposure region of the SAM resist became etching-resistant, which might be mainly ascribed to the cross-linking polymerization of the SAM molecules.
