Despite its good theoretical resolution and mass production ability, the Ion Beam Lithography (IBL) is the least studied NGL tool. At the beginning of the IBL development stage membrane masks were used, but they were soon replaced by stencil masks, because the resolution worsening due to ion straggling in the membrane was considered difficult to overcome.
In this paper, however, we want to demonstrate both by simulation and experiment that using membrane mask, sub 100 nm pattern can be generated with practical mask to wafer distances (~10 mm).
1. straggling problem with membrane masks
From the very beginning of IBL, people concentrated on fabricating least scattering membrane mask, because scattering is directly responsible for the achievable line width. Rensch et al. concluded for 100 nm width line 1 mm-thick silicon channeling mask is needed at about 20 mm mask to wafer distance (half scattering angle about 0.6o). But as they have estimated, with this extremely thin mask, the mask runout can already reach 100nm. Moreover, to achieve less straggling, ions with higher energies are needed, which in turn needs thicker absorber, e.g., 1 mm Au absorber. Even with the state of the art e-beam lithography technique, such mask (aspect ratio about 10) is difficult to fabricate.
As discussed above the frustration of membrane mask was obviously due to the bad lateral resolution coming from the angular straggling, and here is the point where stencil mask replaced membrane mask. But we want to point out that straggling is not really a fatal restriction to the lateral resolution. If it were so, sub 100 nm pattern could not be made by conventional e-beam lithography, because straggling is much severer for electron. The resolution depends on the contrast of resist in the developing process as well as the straggling of ions, and can be drastically improved by proper developing solution and condition. The real problem occurring from the straggling is the dose reduction for narrow structures. Fig.1 shows the lateral dose profile of 400 keV protons passing through a 0.1 mm thick Si3N4 membrane with various slit size in mask (TRIM simulation).
The membrane has similar angular straggling with a perfect 3 mm Si channeling mask (critical angle 1.1o). If the developing condition is so selected as until the half irradiated region is developed, the developed pattern has the same size as the slit in the mask for slits wider than 0.6 mm but for 0.1 ~ 0.4 mm slits, narrower pattern will be formed. For slits narrower than 0.1 mm no pattern will be developed. Nevertheless, this problem is not invincible and can be solved in a way very similar to the proximity correction in e-beam lithography.
2. Preparation of membrane masks
Table 1. summarizes the procedure for manufacturing the membrane masks. Both Si3N4 and Si channeling masks were prepared. Fig. 2 shows the SEM image of the mask pattern. Thick channeling mask was preferred for stability. The Au pattern on the channeling mask serves in this case as scattering center rather than absorber. According to TRIM simulation, more than 97 % of 400 keV protons entering the 3000 A thick Au region are scattered in a angle larger than the critical angle (1.12o) of  silicon and therefore, practically all protons incident on the Au pattern are stopped in the membrane.
Si channeling mask [ ]
|LPCVD of 2mm Si3N4||Thermal diffusion of B for 7hrs at 1100oC for 3mm Si|
|20 A Ti and 700 A Au e-beam deposition|
|1[0.4]mm PMMA(950k) spin coating|
|pattern develop with MIBK:IPA = 1:3 for 60s at 21oC|
|Back side opening of Si3N4 [P+ layer]|
|Au or Ni electroplating with forward-reverse pulsing method|
|backside etching with KOH [EDP(etch stop)]|
400 - 450 keV proton beam from the 1.7 MV Tandem Van de Graaff accelerator in KIGAM was used for irradiation. The beam was scanned over the mask with 64 x 517 frequency to obtain uniform dose. The resist was 2000 A PMMA spin coated on silicon wafer. A special aligner was designed for adjusting beam direction and maintaining precise mask to wafer distance with 1 mm precision. The optimum dose was decided to be 4 x 1013 p/cm2 through experiments on the correlation of PMMA resist develop speed on dose and developing temperature. The developer was 20 % morpholine, 5 % etanolamine, 60 % diethylenglykol - monobutylether and 15 % distilled water. Development was performed for 4 minutes at 35oC as this temperature showed the best contrast. Fig. 3 shows the pattern transferred from the mask pattern in Fig.2.
Micro & Nano Conference 2001, 10. 30 - 11. 2, Shimane Matsui, Japan