1. Introduction

A color filter embodies various colors by selecting a specific wavelength in the visible light region. Color filters are used widely as core elements for such diverse applications as CMOS image sensors, liquid-crystal display devices, and light-emitting diodes [1,2]. Color filters are mainly fabricated by spreading a resin, including dye or pigment in three basic colors (RGB), and then patterning each color using photolithography. Pigment density must be increased to enhance the chromatic properties of the color filter; this can result in low transmittance. In addition, manufacturing a color filter by photolithography may consist of approximately 20 steps, including the black matrix process, and must be repeated three times to construct the RGB patterns. Thus, this process represents a large cost and a significant environmental burden.

Recently, studies of color filters that construct RGB colors from sub-wavelength gratings (which have a smaller period than the incident light wavelength) without pigments have been undertaken. A sub-wavelength grating filter consisting of a few layers incorporating a periodically modulated waveguide has been widely investigated both theoretically and experimentally, because it works as a band-pass filter in a resonance condition, in which incident light is reflected or transmitted by coupling with the resonance modes in the waveguide layer. This approach is known as guided-mode resonance (GMR) [3–6]. Kanamori et al. fabricated transmission color filters by incorporating a one-dimensional sub-wavelength grating into single-crystalline silicon via e-beam lithography (EBL) [7]. The color filters are easily tuned depending on pattern period and possess high transmittance but low chromatic properties, as the color spectrum bandwidth is large. In particular, color filter applications for display may be limited, as the transmittance and colors vary with incident angle. Yoon et al. proposed color filters based on a sub-wavelength-patterned grating in poly-silicon created using a laser-interference lithography technique featuring a wide effective area compared to costly e-beam lithography [8]. They introduced an oxide layer on top of the silicon grating layer as a mask to facilitate the silicon etching and to enhance filtering selectivity. However, transmittance remained dependent on incident angle. A high angular tolerance in the visible spectrum range, independent of incident angle, is essential for displays; until now, no procedure has satisfied that need. In addition, no study exists on the material properties of the grating necessary for a high-efficiency color filter.

We fabricated a two-dimensional (2D) photonic crystal color filter using low-cost, high-productivity nanoimprint lithography. We investigated the theoretical properties of the 2D grating using the rigorous coupled-wave analysis (RCWA) method. The high angular tolerance of the proposed color filter was experimentally verified. The physical relationship between the 2D grating’s refractive index properties and its efficiency was closely examined numerically and experimentally. The fabrication process of the color filter via nanoimprint lithography used in this paper has a prominent advantage: three color filters with different colors can be fabricated on the same substrate by a single patterning process, unlike the conventional process. Moreover, since a resin including dye or pigment is not applied, it is possible for the color filter to have high efficiency and chromatic properties, as well as high thermal and chemical stability.

2. Proposed color filter design

The proposed photonic crystal color filter with 2D sub-wavelength gratings which is inspired by complex micro-structure of Morpho butterfly wings [9] is shown in Fig. 1 . The 2D sub-wavelength grating layer, consisting of silicon with a high refractive index n_p, was formed on a glass substrate with refractive index n_s. The period, width, and height of the grating layer are L, d, and h, respectively. Since the ambient medium is air, which has a low refractive index compared to that of the grating or substrate, the grating layer functions as a planar waveguide. When light is launched with an incident angle θ and an azimuth angle φ towards the color filter, as shown in Fig. 1, various diffracted waves are generated by means of the silicon grating. As long as a phase-matching condition between the incident light and the planar waveguide structure is satisfied, it acts as a band-pass filter in which the diffracted fundamental wave is taken as the output [5]. Moreover, the optical reflectivity is significantly strengthened within a certain range of spectral band centered at a resonance wavelength.

 

Fig. 1 (a) Reflective display with photonic crystal color filter and (b) its geometry.

Download Full Size | PDF

The proposed device was designed and analyzed using a self-made simulation tool based on the RCWA method. In this method, the Maxwell’s equations for electric E(z) and magnetic field H(z) components are reduced to coupled equations for spatial harmonics E ^{( m , n )}(z) and H ^{( m , n )}(z) with indexes m and n run in range −N,…,N, where N is the cutoff parameter. We note that for sub-wavelength grating only m = 0, n = 0 component of light field is reflected for incidence close to normal, while the other harmonics are exponentially decayed from surface. Spatial harmonics of electro-magnetic field with higher indexes are excited and propagating inside the grating, however their amplitudes are rapidly decreased with increasing the indexes n and m. Typically in our simulation we use the cutoff parameter as N = 7. We found that further increasing the cutoff parameter does not influence on results. The glass substrate was assumed to possess a constant refractive index of 1.5 with negligible loss. The dispersion characteristics of the silicone are defined in Fig. 2 . The refractive index of the silicon consists of a real and imaginary part. Different dispersion properties appear according to the crystallinity of the silicon; in particular, the refractive index of amorphous silicon has a relatively large contribution from the imaginary part in the short-wavelength region related to the absorption.

 

Fig. 2 Complex refractive indices of single-crystalline (c-Si), polycrystalline (p-Si), and amorphous silicone (a-Si).

Download Full Size | PDF

The reflective response characteristics for each RGB color of the proposed color filter are shown in Fig. 3 . The period, width, and height of the red color filter (a) are L = 350 nm, d = 175 nm, and h = 120 nm, respectively. For the green color filter (b), they are L = 240 nm, d = 120 nm, and h = 120 nm, respectively. For the blue color filter (c), they are L = 210 nm, d = 105 nm, and h = 100 nm, respectively. The grating duty cycle for all color filters is 0.5. Maximum reflectances of 87.5%, 75.2%, and 63.25% are obtained at wavelengths of 640, 540, and 490 nm for the red, green, and blue filters, respectively. Although the incident angle of the s-polarized wave was 45°, almost no change in maximum reflectance and central wavelengths occurred. Compared to a color filter fabricated with 1D sub-wavelength gratings, the proposed 2D photonic crystal color filter has excellent optical properties that are independent of incident angle. This phenomenon can be analytically described for the photonic band diagram of the modes excited in 2D gratings. The band diagram of leaky modes exited by incident light with p- and s-polarization is shown in Figs. 4(a) and 4(b), where k_|| = k sin(θ) is in-plane wave vector on the grating layer, and k is the incident wave vector. Here we choose photonic crystal color filter parameters to be L = 300, d = 150, h = 100, n_p = 4. The upper region of bold line, denoted as ω = k_||c/n ₀, the so-called light line, can only be coupled with the incident light resulting in resonance reflection peaks. Reflectance intensities of the photonic crystal filter are represented with blue levels. As it seen, the photonic crystal color filter with large refractive index n_s produces almost dispersionless modes ω(k_||). However, the reflectance intensity for s and p-polarization has different features. The reflectance of s-wave does not change significantly in the Brillouin zone, while the reflectance of p-wave decreases over the region, where the second band of ω = k_||c/n_s is crossing. Figure 3(d) shows the reflectance response of the green color filter according to complex refractive indices. This result shows that differences in silicon crystallinity create many differences in the reflectance spectrum. For amorphous silicon, since internal absorption is relatively large it is thought that the maximum reflectance is significantly reduced. It should be noted that the 2D photonic crystal color filter may be easily tuned in period, width, and height, but careful effort to improve silicon crystallinity is required to fabricate a high-efficiency color filter.

 

Fig. 3 Theoretical results of the proposed color filter: (a) red; (b) green; (c) blue; and (d) reflectance characteristics of single-crystalline (c-Si), polycrystalline (p-Si), and amorphous silicon (a-Si).

Download Full Size | PDF

 

Fig. 4 Band diagram of leaky modes (along ΓX direction) excited in photonic crystal color filter by s-polarized incident light (a) and p-polarized incident light (b).

Download Full Size | PDF

3. Proposed color filter fabrication

We first applied the EBL process to verify the outstanding performance of the proposed color filter for a 5 mm × 5-mm filter. Prior to patterning the 2D gratings with the EBL process, we required a silicon-on-glass (SOG) wafer of the desired thickness, either 100 nm (blue color) or 120 nm (green or red color). Using conventional solid-phase crystallization (SPC) with an annealing time of about 10 hours at 800°C, we made an SOG wafer that consisted of a 100- or 120-nm-thick polycrystalline silicon layer and a 300-nm-thick SiO₂ layer on a 525-μm-thick glass substrate. The SiO₂ layer was applied to prevent any impurities in the glass substrate from permeating into the amorphous silicon layer in the SPC process, as silicon crystallinity was important. The amorphous silicon layer deposited via the low-temperature plasma-enhanced chemical-vapor-deposition (PECVD) process resulted in a polycrystalline silicon layer with a crystallinity of over 85% compared to that of the single-crystalline silicon wafer from the SPC process. Note that the difference in crystallinity means a difference in the imaginary part of the refractive index in the short-wavelength region. Thus, we deduced that the reflectance of the color filter corresponding to the short-wavelength region will decrease.

In patterning the proposed 2D structure, EB resist (ZEP 520A) and lift-off EB lithography were applied. After EB exposure, Au/Ti deposition and the lift-off process, a 2D-Au pattern was obtained on the polycrystalline silicon (p-SOG) wafer. This pattern was transferred onto the substrate using non-volatile chlorine-based etching gas in the inductively-coupled plasma-reactive ion etching (ICP-RIE) process. Figure 5 shows the fabricated 2D sub-wavelength grating for the green color filter and the measured reflectances for each color filter. Reflectance was measured with a UV visible spectrometer (Model UV-2450, Simadzu). The measured reflectances of each color filter were 86.6% (red), 73.9% (green), and 31.1% (blue). Except for the blue color filter, reflectance was high, close to the RCWA simulation results. The remarkable point is that we see a “blue shift” in peak wavelength and a left side-lobe peak compared to the simulation results.

 

Fig. 5 (a) FE-SEM result; reflectance for each fabricated color filter: (b) red; (c) green; (d) blue.

Download Full Size | PDF

From XPS and TEM analyses of the p-SOG wafer and the additional simulation according to pattern height and width changes of the color filter, we found that the causes of the “blue shift” in peak wavelength and the left side-lobe peak were resulted from decrease in the effective thickness of the p-Si layer. The high temperature and long process time of the SPC process may have caused the layer near the surface to become amorphous by diffusion. This means that the effective thickness of the p-Si layer was reduced, even though the physical thickness of the p-Si layer as shown in Fig. 5(a) was similar to the desired thickness. Since the reflectance drop in the short-wavelength region adds to this effect, as mentioned above, the blue color filter reflectance decreased dramatically.

To overcome the side effects of the SPC process, a newly designed process that differs from the conventional photolithography process in which dye or pigment is applied is introduced: a nanoimprint (NIL) process, as presented in Fig. 6 . The master for the NIL process can be fabricated on a silicon wafer using the conventional EBL process. We created such a master through e-beam lithography using a JBX-9300FS system (JEOL Ltd.) with a negative CAR-type resist of NEB-22S68 (Sumitomo Ltd.) and the RIE process to obtain an RGB color filter with 2D photonic crystal patterns through the NIL process. Si master used in the NIL process was etched by the conditions as follows; first, it was applied the SiO₂ (300 Å) hard mask to etch the Si master with RF1 bias power of 80 W, RF2 source power of 240 W and the operating chamber pressure of 10 mTorr. Gases used in this step were CF₄ (24 sccm), CHF3 (16 sccm) and Ar (10 sccm), respectively. second, each pattern on the hard mask was transferred into the Si substrate with RF1 bias power of 90 W, RF2 source power of 270 W and the operating chamber pressure of 3 mTorr. Gases used in this step were Cl₂ (30 sccm), O₂ (3 sccm) and Ar (10 sccm). Following these processes, we fabricated a soft-polymer stamp using the nanoimprint process to preserve the relatively costly Si master and easily create a number of stamps. Besides these advantages, a soft stamp can apply pressure more uniformly than a hard stamp like the Si master, resulting in uniform patterning and a large area of extensity for display applications. To make the soft mold, we used the bi-functional urethane methacrylate (FLK MD 700, Solvay Solexis) including 4% w/w photoinitiator (Darocur 1173, Ciba) as the polymer resin. This soft mold did not require additional anti-adhesion treatment on its surface. The amorphous silicon SOG (a-SOG) consisted of a 30-nm-thick hard-mask layer (SiO₂), 100- or 120-nm-thick a-Si layer, and 100-nm-thick barrier layer (SiO₂) on a 525-μm-thick glass substrate, similar to the p-SOG wafer. The barrier layer minimized process errors (e.g., under-trenching) by acting as an etch-stop layer, and also prevented impurities from penetrating into the a-Si layer in the crystallization process. The hard-mask layer on the a-Si layer was applied to facilitate etching of the a-Si layer and to enhance filtering selectivity [8].

 

Fig. 6 Proposed fabrication process for 2D photonic crystal color filter.

Download Full Size | PDF

The 2D grating patterns were formed on the a-SOG wafer by the NIL process using the fabricated soft mold and a UV-curable resin. A mixture of fluorinated oligomer, including perfluoroalkyl chains, and acrylate monomer, including carbon double-bonding, was used as a UV-curable resin for imprinting. The NIL was processed by our in-house NIL system. NIL parameters, including vacuum pressure, pressing force, delay time, and UV-exposure time were optimized and high-fidelity NIL conditions were established. The pattern from the soft stamp was well transferred, preserving the square shape of the Si master. The residual thickness was near zero under 5 nm. After patterning of 2D grating patterns using the NIL process, the oxide layer was selectively etched using the ICP-RIE process, with the NIL pattern serving as a mask. Subsequently, the a-Si layer was also selectively etched using the ICP-RIE process with conditions similar to the Si master etching process.

As mentioned above, silicon crystallinity is important to proposed color filter reflectance enhancement. Although the 2D pattern shapes made by the NIL and ICP-RIE processes are similar to the desired parameters, they are composed of an amorphous silicon. The SPC process is not suitable for crystallization of the patterned a-Si because the process time is long and the process temperature is high. We thus used the excimer laser annealing (ELA) process [10]. In the ELA process, the pattern size is important, because the probability of crystal nucleation is proportional to the volume of the material to be crystallized. That is, since single-nucleus crystallization can be realized in a small volume, irradiation at low laser energy, at which material quasi-melting can occur, may be sufficient. Moreover, for low laser energy, it is expected that the proper degree of multi-shot irradiation is effective for enhancing the extent of the crystallization. After two shot-irradiation at 160 mJ cm⁻² using a pulsed XeCl excimer laser (308 nm) with a pulse duration of 180 ns was applied, local shape changes of only around 7% were observed without large-scale change in shape as shown in Fig. 6.

To prove the optical performance of the 2D photonic crystal color filter made by the proposed fabrication process, UV reflectance was measured after ELA, as shown in Fig. 7 . For each color filter, the measured results (solid lines) are much close to the simulation results (dashed lines). As illustrated in Fig. 7(b), the reflectance spectrum of the patterned a-Si (before ELA) deviated from the desired peak wavelength and its reflectance was low, around 30%. After crystallization by the selected ELA process with minimum shape change, the peak wavelength approached the design value and the reflectance was considerably enhanced at 74.2%, close to the simulated result. This extraordinary phenomenon was predicted, as shown in Fig. 3(d). Also, we observed a difference in peak wavelength before and after ELA in the reflective color of the color filter. Note that a close relationship exists between silicon crystallinity and color filter reflectance. The side-lobe peak shown in Fig. 5(c) does not exist in the reflectance spectrum for the color filter created using the proposed fabrication process. Since the p-Si layer effective thickness does not decrease from diffusion in the SPC, the side effect was overcome using the proposed process, in which crystallization by ELA was performed for a short time after etching. Another significant result is presented in Fig. 7(c). We measured the reflectance of the fabricated color filter while changing the incident angle from 0°°to 45° using a bi-directional scattering distribution (BRDF: RT-300, J&C Tech) system. As can be seen from the measurements, the peak wavelength change without a reflectance drop was only 2 nm, similar to the simulated result.

 

Fig. 7 Reflectance results (a) for each color filter; (b) before and after ELA; (c) with varying incident angles; and Raman spectroscopy results.

Download Full Size | PDF

Thus, it was experimentally proved that the proposed 2D photonic crystal color filter has a high angular tolerance while overcoming the angular dependence of previous 1D photonic crystal color filters. Figure 7(d) shows the Raman spectroscopy results for the fabricated green color filter. By referring to the Raman peak of the c-Si, we determined that the area ratio of the a-Si and c-Si of the fabricated color filter was under 0.04 (4%). This indicates that the average crystallinity of the 2D grating patterns of the fabricated color filter was around 96%, close to the c-Si. Thus, the fabricated color filter made using the proposed fabrication process had high reflectance. Since the volume of the 2D grating patterns is sufficiently small after etching, low laser energy is sufficient to induce single-nucleus crystallization, and thus, the extent of crystallinity, which is closely related to the reflectance, can be dramatically enhanced. The proposed fabrication process in which crystallization is performed after isolation into 2D pattern gratings by etching is reasonable.

4. Conclusions

A 2D photonic crystal color filter was fabricated using low-cost nanoimprint lithography and a crystallization process using ELA. In the nanoimprint lithography, a soft mold replicated from a high-fidelity Si master formed by the EBL and ICP-RIE processes based on non-volatile chlorine-based etching gas was applied to create 2D grating patterns on an a-SOG wafer. All of the patterns were well transferred, with a near-zero residual and a pattern uniformity of 98% for a 1-inch-diagonal area. After etching the a-Si layer using the 2D nanoimprint patterns as an etch mask, low-energy multi-shot ELA successfully enabled single-nucleus crystallization. The reflectance of the proposed color filter rose dramatically due to this ELA process. For the first time, the close relationship between silicon crystallinity and photonic crystal color filter reflectance was examined in detail.

The proposed 2D photonic crystal color filter has a high reflectance, over 70%, and high angular tolerance. Peak wavelength variation is 2 nm as the incident angle varies from 0° to 45°. This was proved analytically and experimentally. Therefore, we expect that the 2D photonic crystal color filter fabricated using the proposed process will be a promising candidate for displays.