1. Introduction

Photolithography is a key process in the fabrication of microelectronic and photonic components and integrated circuits (ICs). In order to comply with Moore’s law, a continual improvement of resolution for photolithography systems is necessary. While the resolution of 193nm photolithography systems can be improved by using high index immersion fluids to increase numerical aperture (NA) or shorter illumination wavelength (e.g., 157nm), such techniques are still unlikely to resolve feature sizes below 30nm without using a pitch doubling technique. Several other lithographic techniques, such as extreme ultraviolet (EUV), imprint and interference lithography, have been suggested for technology nodes at or below 30nm [1–3]. However these techniques impose a large number of technical hurdles (materials, cost, throughput, limited pattern printing capability, etc.) which prohibit their immediate utilization in IC manufacturing. Contact lithography, another photolithographic approach, is questionable for the targeted application due to its severe evanescent wave decay away from the mask [4]. Superlens imaging was recently proposed as a potential alternative for photolithography due to its sub-wavelength resolution achievable with a single exposure process. The sub-wavelength resolution of superlens relies on the negative refractive index material (NRIM) [5], which was predicted to exhibit diffraction free optical imaging [6] by collecting and enhancing near-field evanescent waves through surface plasmon excitation. A “perfect” NRIM (ε=-1 and µ=-1), required to achieve the superlens effect with arbitrary polarization/angle of incidence, is difficult to realize experimentally for UV wavelengths. However, as was suggested by Pendry [6], for normal incidence illumination and TM polarization, negative dielectric permittivity of superlens layer is sufficient for imaging with sub-wavelength resolution. The 40nm resolution imaging of an arbitrary two-dimensional object with a 365nm illumination was experimentally demonstrated [7,8]. However, this resolution is on the order of what is already achieved with state of the art 193nm immersion lithography tools with a NA of 1.35. Therefore, more efforts are needed for superlens imaging in order to scale the resolution down below 40nm.

In this paper we propose a 193nm superlens structure to resolve 20nm feature sizes. The main challenge of efficient plasmon excitation (or permittivity matching) is solved by the introduction of an “index matching layer” between the metal (Al) and dielectric material layer. To evaluate the resolution limit of the proposed structure, the full wave simulations were performed and a 193nm superlens structure was optimized. Simulations indicate that utilization of Al as a mask material improves the resolution at a 193nm wavelength compared to using chromium (Cr) mask material. The simulations further demonstrate that a minimal feature size of 20nm (~λ/10) can be resolved with the designed 193nm superlens structure. A DoF of several nanometers is predicted for the periodic pattern with a 20nm half pitch (HP). Simulations also verify that assistant features (AFs) help to suppress side lobes and improve resolution of a 20nm two-slit imaging structure.

2. Material selection

Previously, silver (Ag) was proven to be a material of choice for a superlens at a 365nm wavelength [9]. In the far UV simple calculation of surface plasmon wavevectors ($k_{\mathrm{sp}}=k_0\sqrt{\genfrac{}{}{0.1ex}{}{{\epsilon }_d{\epsilon }_m}{{\epsilon }_d+{\epsilon }_m}}$, where ε_d and ε_m are the permittivities of the dielectric material and metal material respectively [10].) indicated Al is an optimal material for the 193nm superlens structure. Furthermore, we propose to utilize the index matching layer between the Al layer and the dielectric layer to maximize surface plasmon excitation efficiency. Although exact permittivity matching is not obtainable, Magnesium oxide (MgO) with Re(ε_MgO)=4.08 (compared to Re(ε_Al)~-4.43) at 193nm [11] was identified as the most promising material candidate. Fig. 1 right shows schematics of the proposed 193nm superlens structure. An Al superlens layer (3) of thickness d is “sandwiched” between the index matching layer (2) and the spacer layer (4). An Al mask layer serves as an imaging object. The Al mask material is surrounded by a SiO₂ dielectric layer (1). The bottom of the device is a quartz substrate. The spacer layer is used to facilitate the pattern transfer which will be explained in a later section of this paper. The photoresist layer (5) is to record the image pattern transferred from the superlens layer. The transmissivity of the multilayer thin film structure can be solved by using the characteristic matrix method for a stratified medium [12]. The calculated transmissivity as a function of normalized wavevector k_x/k₀ is shown in Fig. 1 for TM polarization (H along Y direction). For the structure without a super lens layer (1-4-5), the transmissivity for the propagating waves is close to 1 at ~k_x/k₀<1.7 (except a sharp drop at k_x/k₀=1.55), and the evanescent waves decay rapidly at ~k_x/k₀>1.7, as expected for diffraction-limited imaging. The other two structures, both with superlens layers (without index matching layer: 1-3-4-5, with index matching layer: 1-2-3-4-5), have very close transmissivities for the propagating waves at ~k_x/k₀<1.55. However, the superlens structure with the index matching layer has enhanced evanescent wave vector transmissivity especially for k_x/k₀~2-3, leading to improved optical transfer as confirmed by our simulation results which will be discussed in the later sections of the paper. It should be noted that, for ~k_x/k₀>3, the superlens structure (1-3-4-5) has stronger evanescent wave vector transmissivity compared with the structure with the index matching layer (1-2-3-4-5). Since the transmissivity at ~k_x/k₀>3 is generally weak, it does not affect the imaging quality significantly. An optimal Al (ε_Al=-4.43+0.42i [13]) thickness was determined to be 13nm through the transmissivity calculation. SiO₂, spacer layer and photo resist were assumed to have refractive index of 1.55, 1.7 and 1.7 respectively. The thicknesses for SiO₂ layer, index matching layer and spacer layer are determined to be 10nm, 10nm and 8nm, respectively.

 

Fig. 1. Calculated transmissivity of the multilayer superlens structure. Circle: w/index matching layer. Star: w/o index matching. Dot: diffraction limited. Right: schematics of a 193nm superlens imaging structure, where material stacked, from bottom to top, respectively: quartz substrate, Al mask layer, dielectrics layer (1), index matching layer (2), Al superlens layer (3), spacer layer (4) and photo resist layer (5).

Download Full Size | PDF

Different mask materials (Cr and Al) were investigated for the designed 193nm superlens imaging structure. A periodic grating structure with 20nm HP was simulated and the thicknesses of both mask materials are 20nm. Figure 2 shows the calculated total energy density distribution immediately after the periodic mask structure. Simulation results indicate that the energy density around Al mask edge is much stronger than that for the Cr (ε_Cr=-0.65+2.24i [13]) mask due to enhanced plasmon excitation. Therefore the Al mask material provides a stronger image contrast and better imaging resolution through the superlen image formation. Also, peak energy density is not concentrated at the opening area of the periodic structures, but close to the edges of the Al mask due to plasmon excitation-related field enhancement in the Al mask.

 

Fig. 2. Total energy density distribution immediately after the Cr and Al mask materials.

Download Full Size | PDF

3. Resolution and depth-of-focus (DoF)

The resolution of a superlens imaging structure is closely related to the illumination wavelength, mask material properties, metal properties, dielectric properties, and the thickness of each layer [14]. To evaluate the imaging resolution of the proposed 193nm superlens structure, 2D finite difference time domain (FDTD) simulations were performed for a TM polarized incident wave. A perfect matched layer (PML) boundary condition (BC) was used in Z direction and a periodic BC was used in X direction. Periodic BC is a natural choice for the periodic structure simulations. We also applied periodic BC in the simulations of a 20nm two-slit structure however with a large period, ~2µm, to minimize simulation noise. The grid size was 1nm in both X and Z directions. First, we simulated a periodic grating structure with 20nm HP which is similar to that in Fig. 2. The calculated power distributions at the image focal plane are shown in Fig. 3(a) for both superlens structures (1-3-4-5 and 1-2-3-4-5). The result clearly demonstrates that the 20nm feature is resolvable even without an index matching layer. However, the superlens structure with an index matching layer has improved peak power while keeping the minimal power value close to that without using the index matching layer. This means a better image transfer is expected by using an index matching layer. Interestingly, the image is reversed for the superlens without an index matching layer compared to that with an index matching layer. A further investigation of this improved image transfer was made by simulating E_x cross-section distribution along the center line of mask opening in Z direction. The corresponding results are shown in Fig. 3(b) where region 1, 2, 3, 4, and 5 represent mask, SiO_2, MgO for top curve/SiO₂ for bottom curve, Al and photo resist, respectively. A stronger E_x is observed in the region 2, 3, 4 and 5 for the superlens structure with an index matching layer. This enhanced E_x field distribution automatically explains the improved image transfer by using an index matching layer as shown in Fig. 3(a). The physical mechanism of the enhanced E_x is attributed to the higher plasmon excitation efficiency by permittivity matching between Al and the index matching material. The evanescent field amplifications [6] were also observed for both structures in region 4. Furthermore, we simulated power distribution at different image location in order to evaluate DoF and the corresponding results are shown in Fig. 4. The legend numbers represent the imaging distances away from the superlens layer. Image contrast, defined as k=(I_max-I_min)/(I_max+I_min), is calculated as 0.71, 0.89, 0.59 and 0.38 for the four imaging positions in a direction of moving away from the superlens layer. For the imaging location closer to the superlens layer, the peak power value is larger however with stronger side-lobe. The imaging contrast is very poor if the imaging location is too far from the superlens layer. Therefore the superlens imaging has a very stringent DoF requirement. For this 20nm HP periodic structure, a useful DoF is only a few nanometers if a k value of 0.4 is set as threshold value. Beside the periodic structure, a two-slit structure was also simulated. The two slits both have an opening size of 20nm and the spacing between them is 40nm. Figure 5(a) is the simulated energy density distribution. The inner two smaller peaks are the images of the two slits. Surprisingly there are two side lobes with very strong energy density lying outside of useful images (~80nm from mask center, indicated by dashed circule). In order to suppress the side lobes, we followed an approach of placing assistant features (AFs) that are normally adopted in modern photomask design [15]. Two AFs of 10nm width each were placed on both sides of slits (inset of Fig. 5(b)). The AFs are 80nm away from the mask center. Figure 5(b) shows the simulated result with the AF placement. The result clearly shows that the side lobes originally seen ~80nm away from mask center were suppressed significantly and the 20nm slit can be clearly resolved. Therefore, the simulations verified that AFs help to suppress side lobes even in the superlens imaging and therefore improve resolution. A decent common process window should be found among complex patterns in modern mask design. However our simulations have not resulted in a common process window for the 20nm HP periodic pattern and the 20nm two-slit structure (or called isolated or semi-isolated patterns). This is mainly due to the different mask transmission (low transmission for periodic patterns vs. high transmission for isolated or semi-isolated patterns) as indicated by our simulations. Therefore some attenuation mechanism must be introduced to isolated or semi-isolated patterns to reduce their transmission to get decent common process window among different patterns. This will increase chances of superlens to find its practical lithography applications. We are still investigating this issue and will report new progress in a future paper.

 

Fig. 3. (a) Calculated power distribution of a periodic grating structure with and without an index matching layer. (b): E_x cross-section distribution in the Z direction where 1, 2, 3, 4 and 5 represent mask, SiO2, MgO for top curve/SiO2 for bottom curve, Al and photo resist layer respectively.

Download Full Size | PDF

 

Fig. 4. Calculated power distribution of a periodic grating structure with 20nm HP at different image location where legend numbers represent imaging distances from the superlens layer.

Download Full Size | PDF

 

Fig. 5. (a). Calculated energy distribution for a 20nm two-slit structure without assistant feature placement. (b). Calculated energy density for a 20nm two-slit structure with assistant feature placement. AFs are 80nm away from the mask center. The slit opening size is 20nm and the spacing between them is 40nm.

Download Full Size | PDF

For practical applications, the inter-wafer pattern transfer or bond-detach lithographic techniques [16] can possibly be combined with a single superlens exposure step to form a complete superlens pattern transfer process. In such a realization, the thin spacer layer and photoresist layer can be spin-coated and then exposed on the superlens layer, but not developed. The “device” wafer can then be “bonded” to the exposed and then debonded from the superlens structure by a combination of wet chemistry, mechanical force and possibly heating. The superlens structure can then be cleaned while the photoresist layer will be developed (i.e. morphological features will be formed). Our team is working on an experimental evaluation of such an approach.

4. Conclusion

In conclusion, a 193nm Al superlens imaging structure with index matching layer was proposed and designed. Simulations demonstrated that utilization of the index matching layer around the Al superlens layer together with the use of Al mask layer results in a 20nm resolution. A DoF of several nanometers is predicted to be possible for the 20nm HP periodic structure. A 20nm two-slit structure is resolvable by placing AFs to suppress side lobes. However, further research work is needed to get a common process window among different patterns. The proposed 193nm superlens structure together with the suggested novel patterning process may provide a viable way to reach the 20nm lithography node with a single exposure step without the reduction of the illumination wavelength.